Handbook of Cellular Metals

Handbook of Cellular Metals: Production, Processing, Applications.
Edited by H.-P. Degischer and B. Kriszt

Copyright c 2002 Wiley-VCH Verlag GmbH & Co. KGaA
ISBNs: 3-527-30339-1 (Hardback); 3-527-60055-8 (Electronic)
Hans-Peter Degischer,
Brigitte Kriszt (Editors)
Handbook of Cellular Metals
Production, Processing,
Applications
Related titles from Wiley-VCH
W. Schneider, K. Ehrke (Eds.)
Continuous Casting
ISBN 3-527-30283-2
G. Santhoff
Intermetallics
ISBN 3-527-29320-5
Handbook of Cellular Metals
Production, Processing, Applications
Edited by
Hans-Peter Degischer, Brigitte Kriszt
Editors This book was carefully produced. Never-

theless, authors and publisher do not
Prof. Dr. Hans-Peter Degischer warrant the information contained therein
Technische UniversitaÈt Wien to be free of errors. Readers are advised to
Institut fuÈr Werkstoffkunde keep in mind that statements, data, illus-
und MaterialpruÈfung trations, procedural details or other items
Karlsplatz 13 may inadvertently be inaccurate.
A-1040 Wien
Austria Library of Congress Card No.: applied for
Dr. Brigitte Kriszt British Library Cataloguing-in-Publication

Technische UniversitaÈt Wien Data:
Institut fuÈr Werkstoffkunde A catalogue record for this book is available
und MaterialpruÈfung
from the British Library.
Karlsplatz 13
A-1040 Wien
Die Deutsche Bibliothek ±
Austria
CIP-Cataloguing-in-Publication-Data
A catalogue record for this book is available
from Die Deutsche Bibliothek
c WILEY-VCH Verlag GmbH

69469 Weinheim, 2002
All rights reserved (including those of

translation in other languages). No part of
this book may be reproduced in any form ±
by photoprinting, microfilm, or any other
means ± nor transmitted or translated into
machine language without written permis-
sion from the publishers. Registered names,
trademarks, etc. used in this book, even
when not specifically marked as such, are
not to be considered unprotected by law.
printed in the Federal Republic of Germany

printed on acid-free paper
Composition Hagedorn Kommunikation,

Viernheim
Printing Strauss Offsetdruck GmbH,
MoÈrlenbach
Bookbinding J. SchaÈffer GmbH & Co. KG,
GruÈnstadt
ISBN 3-527-30339-1
ISBNs: 3-527-30339-1 (Hardback); 3-527-60055-8 (Electronic) V
Preface
B. Kriszt and H. P. Degischer
There have been a few international and national research and development pro-
grams on cellular metals in recent years: the US Multidisciplinary Research Initia-
tive (MURI) on ultralight metal structures since 1996; a few European research
projects funded within the 4th and 5th EU Framework Program; and a focussed
research program funded by the German Research Council, which began in
1998. These cooperative activities add to the original research activities at the
Fraunhofergesellschaft IFAM in Bremen, the University of Cambridge, Ranshofen
and Vienna in Austria, the Slovac Academy of Science in Bratislava, and several
other places in Europe. A considerable quantity of data has been produced and
presented at international conferences. One of these meetings was organized by
the German Society of Materials (DGM) on 28/29 February 2000 at the Vienna
University of Technology. The proceedings are mainly in German [1] and the
authors wanted another opportunity to publish their results in context with present
knowledge. The reader of proceedings might be overwhelmed by very specific re-
search results on selected topics, obscuring a general understanding of this ex-
panding field of research. Besides the classical books dealing with cellular metals
[2,3], an overview of the state of the art in the year 2000 was needed, covering pri-
mary and secondary processing, characterization of cellular metals, properties,
modeling, and exploitation.
As a result, some of the contributors to the Vienna symposium ªMetallschaÈume
2000º have been asked to extend their papers, by referring to related results of
other researchers and giving a review of their particular topic, whilst maintaining
the detailed specialist knowledge of the author. These contributions are introduced
by coordinators who describe the state of the art in that field. Foamed metals are
described more extensively than other cellular metals because of the actual re-
search activities prevalent in Europe. The European development goal is the appli-
cation of cellular metals in components for motor vehicles, a field in which price
limits and consistent high quality are essential. This handbook aims to give a more
detailed overview of the present state of the art in research and development on
cellular metals with specific emphasis on processing and characterization of
foamed metals than the recent survey [4]. The handbook gives a starting point
VI Preface
for researchers new to the field, and references to topics adjacent to their own spe-
cialty for experts already engaged. Engineers and potential users are encouraged to
consider the application of cellular metals, taking into account the specific peculia-
rities of each material to avoid failures due to miscalculation of processing require-
ments and performance. A guide to the multitude of different types of cellular me-
tals is provided with an indication of particular differences in properties. Research
is ongoing and it is to be hoped that experience with applications, that the book
intends to promote and stimulate, is expanding.
The editors thank all contributors and acknowledge their assistance in adjusting
the content of their contributions in cooperation with each other to produce a con-
cise overview of the state of the art in cellular metals from a European viewpoint.
References
1. H. P. Degischer (ed), ªMetallschaÈumeº 3. M. F. Ashby, A. G. Evans, N. A. Fleck, L. J.

Special Issue, Materwiss. Werkstofftechn. 2000, Gibson, J. W. Hutchinson, H. N. G. Wadley,
31(6). Metal Foams: a Design Guide, Butterworth-
2. L. J. Gibson, M. F. Ashby, Cellular Solids: Heinemann, Oxford 2000.
Structure and Properties, 2nd edn., Cambridge 4. J. Banhart, ªManufacture, Characterisation
University Press, Cambridge 1997. and Application of Cellular Metals and Metal
Foamsº, Progr. Mater. Sci. 2001, 46, 559 632.
ISBNs: 3-527-30339-1 (Hardback); 3-527-60055-8 (Electronic) VII
Foreword
T. W. Clyne
Interest in metallic foams dates from the 1940s, when Sosnick filed a patent on a
production method involving vaporization of low melting point constituents of me-
tallic alloys [1]. Other publications and patents followed in the subsequent two de-
cades, covering concepts such as that of injecting metal into the interstices around
ªspace holderº particles, which are subsequently removed, and the dispersion of
particles, which will release gas via chemical reaction or thermal decomposition.
However, in general, research on the production and performance of cellular me-
tals remained at a low level until the 1990s, when a substantial acceleration took
place. It is now a research topic receiving a high level of attention and various in-
dustrial applications are currently being explored. Updated information about
these activities is available from certain websites [2] and a comprehensive review
has recently been published [3].
Further research is certainly necessary into the development of improved pro-
cessing methods, since much of the material produced hitherto has been of rela-
tively poor quality and/or inherently rather expensive. However, in this context it
is very important to understand the processing-microstructure-property inter-rela-
tionships for cellular metals and the relevance of these to the property combina-
tions required for various applications. Making a foam from a metal, as opposed
to a polymer, boosts the stiffness, range of operating temperature and resistance
to many (organic) solvents, while, in comparison with a ceramic foam, important
advantages are expected in terms of toughness, (thermal and electrical) conductiv-
ity and formability. However, when considering the detailed characteristics, a clear
distinction should be drawn between open-cellular and closed-cellular metal, since,
not only are these two materials made by different processing routes, but in gen-
eral rather separate types of application can be identified for them. Of course, all
types of cellular materials tend to be relatively light, and to have a high specific
stiffness, but these features usually depend primarily on the pore content, whereas
many other properties are much more sensitive to microstructure (cell structure
and nature of cell walls).
VIII Foreword
Closed-Cell Foams
Since the most promising methods for cheap production of bulk material tend to
generate closed-cell foams (commonly of aluminum), attention has been concen-
trated on these for many structural applications. These include various compo-
nents designed to absorb energy progressively under relatively low applied (com-
pressive) loads. In principle, the reduced constraint on cell walls, compared with
solid metal, should mean that large plastic strains can arise throughout via concer-
tina-like deformation, with substantial absorption of energy. However, the perfor-
mance of such components has often been a little disappointing hitherto, with pre-
mature failure commonly occurring within shear bands and such poor tensile duc-
tility exhibited that components tend to fracture readily if any tensile stresses arise,
for example under bending moments. It is now becoming clear that these prob-
lems are substantially reduced if the cell size can be kept fine and uniform pre-
ferably to sub-mm levels. There is thus a strong incentive to develop processing
techniques capable of producing such material in bulk. Furthermore, it is com-
monly the case that the cell walls in these foams contain severely embrittling con-
stituents, such as large ceramic particles and thick oxide films. Some such consti-
tuents are often deliberately introduced during processing in order to raise the
melt viscosity and thus inhibit cell coarsening and drainage [4]. Recent work has
confirmed that such constituents can have highly deleterious effects on the me-
chanical characteristics of the foam and has outlined the mechanisms responsible
for this [5]. Further work is required both on understanding these effects in more
detail and on developing processing routes in which these constituents are suitably
modified or eliminated. Closed-cell foams are also of interest for other types of ap-
plication, such as thermal barriers, although ceramic materials would often be pre-
ferred for these.
Open-Cellular Metals
Suggested and actual applications include filters, catalyst supports, heat exchang-
ers, fluid flow damping conduits (including various types of shock wave dissipation
devices), biomedical prostheses, internally-cooled shape memory actuators, air bat-
teries, and protective permeable membranes and sheathes. Such functional compo-
nents tend to incorporate higher added value than those in purely structural appli-
cations, which is appropriate in view of the generally higher costs of producing
open-cellular metal. The scale of the cell structure is often important for the func-
tional characteristics. This will clearly be a basic specification for filters and fluid
flow limiting devices, but a fine cell size would often be preferred for heat exchang-
ers etc. (subject to limitations imposed by any danger of pore clogging), while bone
in-growth into prosthetic implants might require relatively coarse pores. However,
these properties would often be needed in combination with a minimum strength
and ductility requirement, so a relatively fine, uniform cell structure and a defect-
free cell struts' microstructure might be beneficial from that point of view. A wide
range of metals is being investigated for applications of open-cellular metals, bring-
ing a requirement for improved understanding of process optimization issues in
various alloy systems.
Foreword IX
References
1. B. Sosnick, US Patent 2 434 775, 1948. 4. V. Gergely, H. P. Degischer, T. W. Clyne,

2. http://www.metalfoam.net ªRecycling of MMCs and Production of
http://www.npl.co.uk/npl/cmmt/metal- Metallic Foamsº in Comprehensive Composite
foams/index.html Materials, Vol. 3: Metal Matrix Composites,
http://www.msm.cam.ac.uk/mmc T. W. Clyne (ed.), Elsevier, Amsterdam 2000,
3. J. Banhart, ªManufacture, Characterisation p. 797 820.
and Application of Cellular Metals and Metal 5. A. E. Markaki, T. W. Clyne, ªThe effect of cell
Foamsº Progr. Mater. Sci. 2001, 46, 559 632. wall microstructure on the deformation and
fracture of aluminium-based foamsº Acta
Mater. 2001, 49, 1677 1686.
ISBNs: 3-527-30339-1 (Hardback); 3-527-60055-8 (Electronic) XI
Contents
1 Introduction: The Strange World of Cellular Metals 1
2 Material Definitions, Processing, and Recycling 5

2.1 Foaming Processes for Al 8
2.1.1 Gas Injection: the Cymat/Alcan and Norsk Hydro Process 8
2.1.2 In-situ Gas Generation: the Shinko Wire Process and the FORMGRIP
process 10
2.1.2.1 The Shinko Wire Process [2] 10
2.1.2.2 The FORMGRIP Process 12
2.2 Industrialization of Powder-Compact Foaming Technique 14
2.2.1 Principles of Foam Production 14
2.2.2 Practical Aspects of Foam Production 17
2.2.2.1 Powder selection 17
2.2.2.2 Mixing 18
2.2.2.3 Densification 18
2.2.2.4 Further processing of foamable material 19
2.2.2.5 Foaming 19
2.2.3 State of Commercialization 20
2.3 Making Cellular Metals from Metals other than Aluminum 21
2.3.1 Zinc 22
2.3.2 Lead 22
2.3.3 Titanium 22
2.3.4 Steel 25
2.3.4.1 Powder-Compact Foaming Technique 25
2.3.4.2 Steel Foams from Powder Filler Mixtures 27
2.4 Recycling of Cellular Metals 28
2.4.1 The Remelting of Cellular Metals 28
2.4.2 Recycling of Cellular Metal Matrix Composites 29
2.4.3 Conclusions 32
2.5 The Physics of Foaming: Structure Formation and Stability 33
2.5.1 Isolated Gas Bubble in a Melt 34
2.5.2 Agglomeration of Bubbles: Foam 35
XII Contents
2.6 Infiltration and the Replication Process for Producing Metal

Sponges 43
2.6.1 Replication 44
2.6.2 The Replication Process: General Principles 46
2.6.2.1 Pattern Preparation 46
2.6.2.2 Infiltration 48
2.6.2.3 Pattern Removal 49
2.6.3 Physical and Mechanical Properties of Metal Sponge 51
2.6.3.1 Continuous Refractory Patterns 51
2.6.3.2 Discontinuous Refractory Patterns 51
2.6.3.3 Burnable Patterns 53
2.6.3.4 Leachable Patterns 53
2.7 Solid-State and Deposition Methods 56
2.7.1 Formation from Single Cells: Coreless Methods 58
2.7.1.1 Hollow-Sphere Structures made from Gas Atomized Hollow
Powders 58
2.7.1.2 Hollow-Sphere Structures made from Coaxially Sprayed Slurries 59
2.7.2 Formation from Single Cells: Lost Core Methods 60
2.7.2.1 Hollow-Sphere Structures made by Cementation and Sintering 60
2.7.2.2 Hollow-Sphere Structures made from Galvanically Coated Styrofoam
Spheres 61
2.7.2.3 Hollow-Sphere Structures made from Fluidized Bed Coated Styrofoam
Spheres 61
2.7.3 Bulk Formation: Coreless Methods 63
2.7.3.1 Sintered Metal Powders and Fibers 63
2.7.3.2 Methods Utilizing Special Sintering Phenomena 64
2.7.3.3 Foaming of Solids 65
2.7.3.4 Foaming of Slurries 67
2.7.4 Bulk Formation: Lost Core Methods 67
2.7.4.1 Powder Metallurgical Space Holder Method 67
2.7.4.2 Deposition Methods 68
3 Secondary Treatment of Cellular Metals 71

3.1 Forming, Machining, and Coating 75
3.1.1 High-Temperature Forming 75
3.1.1.1 Specific Problems in Foam Forming 75
3.1.1.2 Process Sequence for Manufacturing 3D Composites with Aluminum
Foam Cores 76
3.1.1.3 Material Behavior at the Solidus 77
3.1.1.4 Forming of Cellular Metals at High Temperatures 78
3.1.2 Machining 79
3.1.3 Coating 79
3.1.3.1 Mechanical Properties of Spray Deposits 80
3.1.3.2 Specific Difficulties in Foam Coating 80
Contents XIII
3.1.3.3 Thermal Sprayed Composites from Metal Foams 81

3.2 Joining Technologies for Structures Including Cellular Aluminum 83
3.2.1 Introduction 83
3.2.2 Feasible Joining Technologies 83
3.2.2.1 Mechanical Fastening Elements 83
3.2.2.2 Gluing 84
3.2.2.3 Welding 84
3.2.2.4 Soldering and Brazing 85
3.2.3 Foam Foam Joints 87
3.2.4 Foam Sheet Joints 87
3.2.4.1 Microstructural Investigations 88
3.2.4.2 Mechanical Properties of Foam Sheet Joints 90
3.2.5 Transferability to Structural Parts 98
3.2.5.1 Tubes 98
3.2.5.2 Hat-Profiles 99
3.2.6 Summary 100
3.3 Encasing by Casting 103
3.3.1 Foam Cores for Encasing by Casting 103
3.3.1.1 Core Production 103
3.3.1.2 Core Attachment 104
3.3.1.3 Mechanical Properties 105
3.3.1.4 Coating of the Foam Cores 106
3.3.2 Shell Casting Processes 107
3.3.2.1 High-Pressure Casting Processes 107
3.3.2.2 Bonding Between Shell and Foam Core 112
3.4 Sandwich Panels 113
3.4.1 Sandwich Foaming Process 114
3.4.2 Industrial Application 116
3.4.2.1 Technological Benefits 117
3.4.2.2 Technical Limitations 118
3.4.3 Joining Technology of AFS 119
3.4.3.1 Laser Welding 120
3.4.3.2 TIG/MIG Welding 120
3.4.3.3 Bolt/Pin Welding 121
3.4.3.4 Punch Riveting 122
3.4.3.5 Riveting Nuts and Screws 123
3.4.3.6 Flow Drilling 123
3.4.3.7 Riveting 123
3.4.3.8 Bonding 123
3.4.4 Cutting of AFS 124
3.4.4.1 Laser Beam Cutting 124
3.4.4.2 Water Jet Cutting 124
XIV Contents
4 Characterization of Cellular Metals 127

4.1 Characterization of Cellular and Foamed Metals 130
4.1.1 Definition of Structural Features of a Cellular Metal and Influence
on Property Profile 130
4.1.1.1 Density and Volume Fraction of Pores 131
4.1.1.2 Shape and Size of Pores 133
4.1.2 Characterization Methods and Quantities of Geometric Architecture
of Real Metallic Foams 136
4.1.2.1 Sample Preparation 136
4.1.2.2 Pore Size 137
4.1.2.3 Pore Shape 139
4.1.2.4 Pore Orientation 140
4.1.2.5 Thickness of Cell Edges and Walls 140
4.1.2.6 Topological Features 141
4.1.3 Characterization of Microstructure of Massive Cell Material 141
4.2 Computed X-ray Tomography 145
4.2.1 Principle of the Technique 145
4.2.1.1 X-ray Radiography 145
4.2.1.2 X-ray Tomography 146
4.2.2 Set-ups 147
4.2.2.1 Medium-Resolution Microtomography 147
4.2.2.2 High-Resolution Microtomography 147
4.2.2.3 Resolution Required for the Study of Metallic Foams 148
4.2.2.4 Reconstruction Method 148
4.2.3 Experimental Results 148
4.2.3.1 Initial Cell Structure 148
4.2.3.2 Evolution of the Structure During a Compression Test 150
4.2.4 Micromodeling of a Foam by Finite Elements 151
4.2.4.1 Direct Meshing of the Actual Microstructure 151
4.2.4.2 Results 152
4.3 Considerations on Quality Features 156
4.3.2 Non-Uniformity of Cellular Metals 156
4.3.3 Macroscopic Parameters 159
4.3.3.1 Type of Cellular Metal 159
4.3.3.2 Surface and Dimensions 159
4.3.3.3 Apparent Density 161
4.3.3.4 Properties 161
4.3.4 Microscopic Features 161
4.3.4.1 Microstructure of the Metal 162
4.3.4.2 Geometrical Features 162
4.3.4.3 Microdefects 164
4.3.5 Mesoscopic Features 165
Contents XV
4.3.5.1 Geometry of Cellular Structure 165

4.3.5.2 Density Distribution 166
4.3.6 Systematics of Quality Features 166
4.3.7 Approximation of a Cellular Structure by a Continuum 168
4.3.7.1 Calculation of Density Maps 168
4.3.7.2 Representation of Non-Uniformity of Densities 172
4.3.7.3 Mesoscopic Basis for Material Modeling 174
4.3.8 Proposal of Quality Criteria 174
5 Material Properties 179

5.1 Mechanical Properties and Determination 183
5.1.1 Young's Modulus 183
5.1.1.1 Influence of the Foam Structure 184
5.1.1.2 Influence of the Foam Density 185
5.1.1.3 Influence of Deformation 186
5.1.2 Compression Behavior 187
5.1.3 Energy-Absorbing and Impact Behavior 190
5.1.3.1 Energy Absorbing Capability 190
5.1.3.2 Impact Behavior 191
5.1.4 Tension Behavior 193
5.1.4.1 General Tensile Behavior 193
5.1.4.2 The Influence of Notches 195
5.1.4.3 Foam-Specific Test Problems in Tension 196
5.1.5 Torsion Behavior 196
5.1.6 Fracture Behavior 197
5.1.6.1 Crack Initiation and Crack Propagation 197
5.1.6.2 Fracture Toughness 198
5.1.6.3 Foam-Specific Test Problems 201
5.2 Fatigue Properties and Endurance Limit of Aluminum Foams 203
5.2.1 Literature Survey of Endurance Data 203
5.2.2 High Cycle Fatigue Properties and Endurance Limit 208
5.2.2.1 Material and Procedure 208
5.2.2.2 Results 209
5.2.3 Mechanism of Crack Initiation 210
5.2.4 Summary 214
5.3 Electrical, Thermal, and Acoustic Properties of Cellular Metals 215
5.3.1 Electrical Properties 215
5.3.2 Thermal Properties 221
5.3.3 Acoustic Properties 225
5.3.3.1 Materials for Sound Insulation 226
5.3.3.2 Sound Absorbing Materials 228
5.3.3.3 Structural Damping 236
XVI Contents
6 Modeling and Simulation 243

6.1 Modeling of Cellular Metals 245
6.1.1 Motivation 246
6.1.2 Micromechanical Modeling of Cellular Materials: Basics 247
6.1.2.1 Analytical and Numerical Models 247
6.1.2.2 Classification of Microgeometries 249
6.1.2.3 Information Obtainable from Micromechanics 251
6.1.3 Selected Results of Micromechanical Simulations 252
6.1.3.1 Influence of Material Distribution in the Cell Walls 252
6.1.3.2 Influence of Wavy and Curved Cell Walls 255
6.1.3.3 Influence of Irregular Vertex Positions 257
6.1.3.4 Microgeometries Containing Cells of Different Sizes 258
6.1.3.5 Influence of Holes and Solid-Filled Cells 260
6.1.3.6 Influence of Fractured or Removed Cell Walls 261
6.1.3.7 Yield and Collapse Surfaces 262
6.1.3.8 Fracture Simulations for Metallic Foams 266
6.1.4 Modeling of Mesoscopic Density Inhomogeneities 269
6.1.5 Macroscopic Modeling and Simulation 272
6.1.5.1 Low Energy Impact on Thin Metallic Foam Paddings 273
6.1.5.2 Crushing of Foam-Filled Crash Elements 275
6.1.6 Design Optimization for Cellular Metals 276
6.1.7 Outlook 277
6.2 Mesomodel of Real Cellular Structures 281
6.2.2 3D Mesomodel 284
6.2.2.1 Elastic Regime 285
6.2.2.2 Plastic Regime 285
6.2.3 Modeling of Uniaxial Compression 288
6.2.3.1 Deformation Band 289
6.2.3.2 Mechanical Properties 292
6.2.4 Discussion 294
7 Service Properties and Exploitability 299

7.1 The Range of Applications of Structural Foams Based on Cellular Metals
and Alternative Polymer Solutions 299
7.1.2 Potential Areas of Use 300
7.1.3 Material Properties 300
7.1.4 Main Component Configurations 301
7.1.5 Application and Attachment Techniques 304
7.1.5.1 Casting 304
7.1.5.2 Thermal Joining Processes 305
7.1.5.3 Mechanical Joining Processes 306
7.1.5.4 Three-dimensional sandwich 306
Contents XVII
7.1.5.5 Alternative Cellular Materials Based on Polymers 306

7.1.6 Effectiveness 307
7.1.6.1 Bending and Torsional Stress 307
7.1.6.2 Impact Stresses 308
7.1.6.3 Axial Load 309
7.1.6.4 Acoustics 310
7.1.7 Outlook 311
7.2 Functional Applications 313
7.2.1 General Considerations 313
7.2.2 Biomedical Implants 314
7.2.3 Filtration and Separation 315
7.2.4 Heat Exchangers and Cooling Machines 315
7.2.5 Supports for Catalysts 317
7.2.6 Storage and Transfer of Liquids 317
7.2.7 Fluid Flow Control 317
7.2.8 Silencers 318
7.2.9 Spargers 318
7.2.10 Battery Electrodes 318
7.2.11 Electrochemical Applications 319
7.2.12 Flame Arresters 319
7.2.13 Water Purification 319
7.2.14 Acoustic Control 319
7.3 Machinery Applications 320
7.3.1 Parameters 321
7.3.1.1 Thermal Behavior 321
7.3.1.2 Pull-Out Strength of Detachable Joints 322
7.3.2 Examples of Application 323
7.3.2.1 Foamed Steel Pipes 323
7.3.2.2 Machine Table 327
7.3.2.3 Cross-Slide 329
7.4 Prototypes by Powder Compact Foaming 330
7.4.2 Methods, Machines, and Molds 332
7.4.2.1 Manufacturing Methods for Precursor Material 333
7.4.2.2 Foaming Process 334
7.4.2.3 Foaming Furnaces 334
7.4.2.4 Foaming Molds 336
7.4.3 Prototypes and their Applications 337
7.4.3.1 Automotive Applications 338
7.4.3.2 Construction and Architecture 340
7.4.3.3 Other Technical Applications 340
7.4.3.4 Improbable Applications 343
XVIII Contents
7.5 Applying the Investment Methodology for Materials (IMM) to

Aluminum Foams 346
7.5.1 Introduction: The Investment Methodology for Materials (IMM) 346
7.5.2 Initial Market Scan for Potential Applications for Al Foams 347
7.5.3 Material Assessment 347
7.5.3.1 Technical Performance 347
7.5.3.2 Cost of Production 348
7.5.3.3 Co-Minimizing Volume and Cost in Energy-Absorbing
Applications 349
7.5.4 Market Forecast 350
7.5.4.1 Market Size for Aluminum Foam 350
7.5.4.2 Market Timing for Aluminum Foams 350
7.5.5 Value Capture 351
7.5.5.1 Industry Structure 351
7.5.5.2 Appropriability of Profits 352
7.5.6 Conclusions: Applying IMM to Aluminum Foams 353
8 Strengths, Weaknesses, and Opportunities 355

8.1 Processing 355
8.2 Properties 358
8.3 Design and Application 360
Index 365
ISBNs: 3-527-30339-1 (Hardback); 3-527-60055-8 (Electronic) XIX
List of Contributors
O. Andersen F. BaumgaÈrtner
Frauenhofer-Institut fuÈr Schunk Sinter Metalltechnik GmbH
Fertigungstechnik und Postfach 10 09 51
Materialforschung (IFAM) 35339 Gieûen
Auûenstelle fuÈr Pulvermetallurgie Germany
und Verbundstoffe
Winterberstr. 28 C. Beichelt
01277 Dresden Wilhelm KARMANN GmbH
Germany Karmannstr. 1
Postfach 26 09
M. Arnold 49084 OsnabruÈck
Lehrstuhl fuÈr Werkstoffwissenschaften Germany
UniversitaÈt Erlangen-NuÈrnberg
Martensstr. 5 T. Bernard
91058 Erlangen Neue Materialien Bayreuth GmbH
Germany UniversitaÈtsstr. 30
94447 Bayreuth
M. F. Ashby Germany
Engineering Design Centre
University of Cambridge H. W. Bergmann²
Engineering Department, UniversitaÈt Bayreuth
Trumpington Street Lehrstuhl Metallische Werkstoffe
Cambridge CB2 1PZ Ludwig-Thoma-Str. 36b
UK 94440 Bayreuth
Germany
J. Banhart
Hahn-Meitner-Institute H. J. BoÈhm
Dept. of Materials ± SF3 Institute of Lightweight Structures and
Glienicker Str. 100 Aerospace Engineering
14109 Berlin Vienna University of Technology
Germany Guûhausstr. 27±29
1040 Vienna
Austria
XX List of Contributors
R. Braune C. Haberling
Lehrstuhl fuÈr Fertigungstechnologie Firma Audi AG
UniversitaÈt Erlangen-NuÈrnberg Abteilung
Egerlandstr. 11 Werkstoffe/Verfahren/Recycling I/EG-34
91058 Erlangen 85045 Ingolstadt
Germany Germany
T. W. Clyne F. Heinrich
Department of Materials Science Lehrstuhl fuÈr Werkstoffwissenschaften
University of Cambridge UniversitaÈt Erlangen-NuÈrnberg
Pembroke Street Martensstr. 5
Cambridge CB2 3QZ 91058 Erlangen
UK Germany
T. Daxner Th. Hipke

Institute of Lightweight Structures Frauenhofer Institut fuÈr Werkzeug-
and Aerospace Engineering maschinen und Umformtechnik (IWU)
Vienna University of Technology Reichenhainerstr. 88
Guûhausstr. 27±29 09126 Chemnitz
A-1040 Vienna Germany
Austria
C. KoÈrner
H. P. Degischer Lehrstuhl Werkstoffkunde und
Institute of Materials Science Technologie der Metalle
and Testing UniversitaÈt Erlangen-NuÈrnberg
Vienna University of Technology Martensstr. 5
Karlsplatz 13 91058 Erlangen
1040 Vienna Germany
Austria
J. Kovacik
B. Foroughi Institute of Materials and Machine
Institute of Materials Science Mechanics
and Testing Slovak Academy of Science
Vienna University of Technology Racianska 75
Karlsplatz 13 P.O. Box 95
1040 Vienna 83008 Bratislava
Austria Slovakia
M. C. Hahn A. Kottar
Lehrstuhl fuÈr Fertigungstechnologie Institute of Materials Science
UniversitaÈt Erlangen-NuÈrnberg and Testing
Egerlandstr. 11 Vienna University of Technology
91058 Erlangen Karlsplatz 13
Germany 1040 Vienna
Austria
List of Contributors XXI
B. Kriszt A. Mortensen
Institute of Materials Science Laboratoire de MeÂtallurgie MeÂcanique
and Testing Ecole Polytechnique FeÂdeÂrale
Vienna University of Technology de Lausanne
Karlsplatz 13 1015 Lausanne
1040 Vienna Switzerland
Austria
U. Mosler
R. Kretz Institut fuÈr Metallkunde
ARC Leichtmetallkompetenzzentrum TU Bergakademie Freiberg
Ranshofen GmbH Gustuv-Zeuner-Str. 5
Postfach 26 09596 Freiberg
5282 Ranshofen Germany
Austria
C. Motz
E. M. A. Maine Erich-Schmid-Institute of Material
Centre for Technology Management Science
Engineering Design Centre Austrian Academy of Science
Engineering Department Jahnstr. 12
University of Cambridge 8700 Leoben
Trumpington Street Austria
Cambridge CB2 1PZ
UK R. Neugebauer
Frauenhofer Institut fuÈr Werkzeug-
E. Maire maschinen und Umformtechnik (IWU)
CR1 CNRS Reichenhainerstr. 88
GEMPPM 09126 Chemnitz
Batiment Saint ExupeÂry Germany
23 Avenue Capelle
69621 Villeurbanne cedex A. Otto
France Lehrstuhl fuÈr Fertigungstechnologie
UniversitaÈt Erlangen-NuÈrnberg
U. Martin Egerlandstr. 11
Institute fuÈr Metallkunde 91058 Erlangen
TU Bergakademie Freiberg Germany
Gustuv-Zeuner-Str. 5
09596 Freiberg R. Pippan
Germany Erich-Schmid-Institute of Material
Science
H. Mayer Austrian Academy of Science
Institute of Meteorology and Physics Jahnstr. 12
University of Agricultural Sciences 8700 Leoben
TuÈrkenschanzstr. 18 Austria
1180 Vienna
Austria
XXII List of Contributors
G. Rausch R. F. Singer
Frauenhofer Institut fuÈr angewandte Institut fuÈr Werkstoffwissenschaften
Materialforschung (IFAM) UnivertitaÈt Erlangen-NuÈrnberg
Lesumer Heerstr. 36 Martensstr. 5
28717 Bremen 91058 Erlangen
Germany Germany
C. San Marchi S. Stanzl-Tschegg

Northwestern University Institute of Meteorology and Physics
Department of Materials Science University of Agricultural Sciences
and Engineering TuÈrkenschanzstr. 18
Evanston IL 60202-3108 1180 Vienna
USA Austria
F. G. Rammerstorfer G. Stephani
Institute of Lightweight Structures Frauenhofer-Institut fuÈr
and Aerospace Engineering Fertigungstechnik und
Vienna University of Technology Materialforschung (IFAM)
Guûhausstr. 27±29 Auûenstelle fuÈr Pulvermetallurgie
1040 Vienna und Verbundstoffe
Austria Winterberstr. 28
01277 Dresden
W. Seeliger Germany
Wilhelm KARMANN GmbH
Karmannstr. 1 M. Thies
Postfach 26 09 Lehrstuhl fuÈr Werkstoffwissenschaften
49084 OsnabruÈck UnivertitaÈt Erlangen-NuÈrnberg
Germany Martensstr. 5
91058 Erlangen
M. Seitzberger Germany
Institute of Lightweight Structures
and Aerospace Engineering B. Zettl
Vienna University of Technology Institute of Meteorology and Physics
Guûhausstr. 27±29 University of Agricultural Sciences
1040 Vienna TuÈrkenschanzstr. 18
Austria 1180 Vienna
Austria
F. Simancik
Institute of Materials and Machine
Mechanics
Slovak Academy of Science
Racianska 75
P.O. Box 95
83008 Bratislava
Slovakia
ISBNs: 3-527-30339-1 (Hardback); 3-527-60055-8 (Electronic) XXIII
List of Abbreviations
AFS aluminum foam sandwich

AVS averaging volume size
BET Brunnauer-Emmett-Teller nitrogen gas absorption method
CCD charged coupled device
CIP cold isostatic pressing
COD crack opening displacement
CT computer tension specimen
CTOD crack tip opening dispacement
CVD chemical vapour deposition
DOF degree of freedom
2D or 3D two- or three dimensional
EB-DVD electron beam directed vapor deposition process
EDM electrodischarge machining
ESRF European Synchrotron Radiation Facility, Grenoble
FE finite elements
FORMGRIP foaming of reinforced metals by gas release in precursors
FPZ fracture-process zone
HIP hot isostatic pressing
IFAM Fraunhofer-Institute for Applied Material Research
IMM investment methodology for materials
INSA Institut National des Sciences AppliqueÂes de Lyon
IP intellectual property
IWU Fraunhofer-Institute for Machine Tools and Forming Technology
LBM lattice block materials
LDC low density core material
LEFM linear elastic fracture mechanics
LKR Leichtmetallkompetenzzentrum, Ranshofen, Austria
MIG metal inert gas arc welding
MMC metal matrix composite
MURI USA multidisciplinary research initiative
PCF powder compact foaming
PM powder metallurgy
PSF point spread function
XXIV List of Abbreviations
PVD physical vapour deposition

RV representative volume
RVE representative volume element
SAS Slovak Academy of Science
SEM scanning electron microscopy
TIG tungsten inert gas arc welding
US ultrasonic
XCT X-ray computed tomography
ISBNs: 3-527-30339-1 (Hardback); 3-527-60055-8 (Electronic) 1
1
Introduction: The Strange World of Cellular Metals
F. Simancik
When Nature builds large load-bearing structures,
She generally uses cellular materials: wood, bone, coral.
There must be good reasons for it.
M. F. Ashby
It is well known that porous structures are good for insulation, packaging, or filter-
ing, but few people believe that they can also be very effective in structural applica-
tions. Thousands of scientific publications deal with the minimization of porosity
in load-bearing parts. Engineers work hard to eliminate pores from castings, pow-
der metallurgy parts, weld joints, or coatings, thinking that a defect-free part is a
pore-free one. With this attitude it is difficult for someone to accept that a load-
bearing material can include pores, even quite large ones. However, large natural
structures of porous materials have existed for thousands of years, demonstrating
how evolution has generated cellular structures that optimize mechanical proper-
ties and structural function for minimum weight.
Mankind tries to learn from nature. Understanding the benefits of natural struc-
tures gives us information to help us produce man-made cellular solids. The cell-
wall material has to be chosen very carefully if the structure is expected to carry
loads. Polymers appear to be insufficiently rigid and ceramics are too brittle. Per-
haps metals could be the right choice. Several of the engineering properties of me-
tallic foams are superior to those of polymeric ones: they are stiffer by an order of
magnitude, they are stable at elevated temperatures, they possess superior fire re-
sistance, and they do not produce toxic fumes in a fire. Moreover, these materials
are fully recyclable without any pollution or waste problems. The latter fact can no
longer be ignored, because the production, disposal, and use of stronger and stiffer
materials in new products often have negative environmental impacts over the
product life cycle.
Owing to their pores, cellular metals possess a set of unusual properties
compared with bulk structural materials: they are crushable, they exhibit a plateau
stress if compressed, and they exhibit a change in Poisson ratio on deformation.
The excellent combination of good mechanical properties (mostly strength and
stiffness) and low weight is the prime advantage. In addition, cellular metals
2 1 Introduction: The Strange World of Cellular Metals
absorb high impact energies regardless of the impact direction, and are very
efficient in sound absorption, electromagnetic shielding, and vibration damping.
Most of the mechanical properties of foam materials can be achieved with other
materials, sometimes more effectively, but foams can offer a unique combination
of several (apparently contradictory) properties that cannot be obtained in one con-
ventional material at the same time (e. g., ultra-low density, high stiffness, the capa-
bility to absorb crash energy, low thermal conductivity, low magnetic permeability,
and good vibration damping). Cellular metals are thus promising in applications
where several of these functions can be combined.
These properties depend significantly on the porosity, so that a desired portfolio
of properties can be tailored by changing the foam density. This is one of the most
attractive features of these remarkable materials.
Cellular material properties also depend on the pore structure. This influence,
imperfectly understood at present, is a topic of intense study. Various constitutive
laws have been suggested for the characterization and modeling of this relation-
ship. These laws, originally developed for polymeric foams, are usually based
on the relative density of the foam, and therefore suppose uniform cellular struc-
ture, at least at a macroscopic level. However, metallic foams are dramatically dif-
ferent from polymeric foams: polymeric foams generally have a regular micro-
structure, whereas metallic foams may be highly disordered with a wide dis-
persion of cell size and shape. Moreover, many imperfections exist in a cell
structure, such as cracks or holes in the cell walls, corrugated cells etc. These
effects are inevitable due to manufacturing at significantly higher temperatures
than in the case of polymers. If these features are not taken into account and
the properties of the foam are characterized only in relation to apparent density,
a higher scatter of properties is to be expected. This is why it is still widely
believed that acceptable reproducibility of the properties of metallic foams is
questionable.
The structure of metallic foams is often non-uniform, especially in the case of
complex 3D parts. It should be noted that a uniform structure is not necessary
for obtaining acceptable and reproducible properties. Anisotropic or gradient
pore structures allow the distribution of load bearing material according to load
conditions (simulating the optimum bone-like structure), without a need to in-
crease the overall weight or volume of the component. Therefore, the challenge
for manufacturing is not to produce a uniform structure, but to achieve reproduc-
ible properties with a controlled non-uniform structure.
If the non-uniform structure is optimal, the crucial question ªWhat is the mate-
rial?º should be answered. It is really difficult to distinguish between material and
structure. If a cellular metal is a material, it is very problematic to define geometry-
independent material characteristics (the strength or elasticity modulus); if it is a
structure made of a certain metal it is almost impossible to define its random
geometry.
Cellular metals can be prepared by various processing methods. They may all be
called ªmetallic foamsº, but they are very different materials, depending on the
manufacturing technique. The production method affects the distribution of the
1 Introduction: The Strange World of Cellular Metals 3
cell-wall material in such a way that the properties of differently manufactured ma-
terials are not comparable.
Metallic foams result from the nucleation and subsequent growth of gas bubbles
in a liquid or semi-liquid metal. They usually have a non-uniform pore structure
(variable pore size and sometimes preferred orientation of pores). The pores are
initially closed, but some defects always appear on cooling, owing to shrinkage of
solidifying metal and gas-pressure reduction in pores. Many solidified cell faces
have non-uniform curvature or are corrugated and have occasional broken walls
that still hang in place. These are the main features of this kind of cellular solid.
Other cellular metals may be manufactured by casting or deposition of the metal
onto templates or place holders, which have to be removed from the final product,
thus creating a porous structure. These cellular structures usually have adjustable
distribution of pore size (according to the template) and their cells are always open.
The manufacturing process dictates not only the properties but also the potential
applications of the foam. Thus foams prepared by the powder compact foaming
(PCF) technique (usually with a dense skin) can be effectively used as net-shape
components, stiffening cores in castings, or in complicated hollow profiles,
whereas the foams prepared by the ªmolten metal routeº (typically large blocks
or panels) can be effectively used as voluminous energy absorbers, cores for sand-
wiches, or for blast protection. The open cellular structures made by investment
casting are good for heat exchangers, sound absorbers, or for electrodes in bat-
teries. The properties arising from the cellular structure produced by a certain
manufacturing technique cannot be effectively achieved using another method.
This also means that cellular metals manufactured differently are not necessarily
competitive materials.
The first attempt to foam a metal was performed by B. Sosnik in 1943 [1]. In
order to create pores, he added mercury to molten aluminum. In 1956 J. C. Elliot
replaced mercury by foaming agents generating gas by thermal decomposition [2],
so now modern scientists and engineers can develop metallic foams without
having to deal with the toxicity of mercury.
In 1959 B. C. Allen [3] invented the PCF route for manufacturing metallic foams
and the basic processing techniques were thus completed.
The success in the preparation of the first metallic foams and discovery of their
remarkable properties started a euphoric enthusiasm for these materials. In 1957
J. Bjorksten stated [4]: ªFoamed metals offer great market potential and might con-
ceivably account for 10 % of all metals produced within 20 yearsº. However he also
said: ªA lot of work still remains to bring about production on a large scale, such as
closer control of density and dimensions.º Unfortunately, only the second of his
statements turned out to be realistic.
Although it is many years since the first patents concerning the manufacture of
metallic foams appeared, the material has not been put into large-scale commercial
production yet. This discouraging fact can be attributed to inadequate design of
components, low reproducibility of properties, a lack of testing procedures and cal-
culation approaches, absence of concepts for secondary treatment, as well as the
production technologies being too complicated and relatively expensive.
4 1 Introduction: The Strange World of Cellular Metals
In spite of disappointments and mistrust, interest in cellular metals is growing.

A new era started, at least in Europe, at the end of the 80s thanks to the activities of
the Fraunhofer Institute for Advanced Materials in Bremen (IFAM). When
J. Baumeister showed his floating aluminum on German TV in the ªKnoff-Hoff
showº this remarkable material found a lot of enthusiastic fans. Stabilization of
the melt via viscosity-increasing additions significantly improved the quality of
the foam structure [5,6]. The development of new foaming techniques enabled rea-
sonable manufacturing costs, and so metallic foams became very attractive to the
transport industry, especially for lightweight stiff body structures and crash absorb-
ing elements.
The first industrial companies (Shinko-Wire, Cymat, Alulight, Schunk, Karman,
Neuman-Alufoam) have already established a group of ªmetfoamº producers and
further companies will join them soon.
At present metallic foams are still insufficiently characterized, and understand-
ing of the process is incomplete, leading to inadequate control and, hence, variable
properties. This gives an impetus for large multilateral scientific activities like the
USA Multidisciplinary Research Initiative on ultralight metal structures (MURI) [7]
or the German Reasearch Council's focussed research program (DFG) [8]. Better
understanding leads to better process control and improved properties. The pro-
ducers themselves have aggressive development programs for their materials.
The next generation of metallic foams will certainly be better.
With all these shortcomings, even the present generation has shown excellent
performance in many case studies. The structures, though still apparently imper-
fect, can be acceptable if they are applied properly in a foam-familiar design, be-
cause cellular metals have to be used in non-traditional ways. The problem-solving
approach, instead of the trial-and-error method, will definitely accelerate the imple-
mentation of cellular metals in real products. An intensive and close collaboration
among scientists, engineers, producers, and end users is crucial for success.
Dr. Bjorksten's: ªNow it's up to industry to decide what to do with itº [4] would
have fatal consequences for the future of cellular metals.
References
1. B. Sosnik, US Patent 2 434 775, 1948. 6. J. Iljoon et al., US Patent 5 115 697, 1992.
2. J. C. Elliot, US Patent 2 751 289, 1956. 7. Ultralight Metal Structure Project, MURI
3. B. C. Allen, US Patent 3 087 807, 1963. Grant No. N00014-1-96-1O28, Washington
4. Modern Metals, 1957, October. 1996.
5. S. Akiyama et al., European Patent 0 210 803, 8. Zellulare metallische Werkstoffe, DFG-Priority
1986. Program 1075, Bonn 1999.
2
Material Definitions, Processing, and Recycling
H. P. Degischer
A great variety of cellular metals is produced by research laboratories and industrial

development departments. Established industrial products are Duocel, Incofoam,
Alporas, and some others are on the brink of market introduction. The different
types of products and prototypes are the result of various combinations of process-
ing, architecture, and metal matrix. The architecture of the cellular structure is the
result of the processing technique, which can be classified according to Fig. 2-1,
but each process has special features typical of the producing company or labora-
tory. None of the manufacturing techniques can be applied to any metal; each is
appropriate for one or other base metal. Cellular structures are those with a relative
Figure 2-1. Processing techniques for cellular metals classified according to the state of the
metal, the formation of the cellular architecture, and the pore-forming ingredients.
6 2 Material Definitions, Processing, and Recycling
density of less than 0.3 [1]. Materials with higher relative densities are called po-
rous materials (for instance, powder compact greens), most of which also can be
produced by the processing techniques listed in Fig. 2-1. The most widely devel-
oped and investigated cellular metals are based on aluminum and its alloys.
The general aspects of processing are described elsewhere [1 3] and specific pre-
sentations can be found in the proceedings of the conferences dealing with cellular
metals [4 10]. The most important processing techniques are described in detail in
subsequent chapters.
Cellular metals are heterogeneous materials formed by a 3D metallic matrix with
gas-containing pores occupying more than 70 vol.-%. Cellular metals are classified
according to the following criteria (Fig. 2-1).
x The metal condition during production of porosity: liquid, solution or emulsion,
solid.
x The forming process involved: casting, foaming, deposition, sintering (including
precursor slurry).
x The method of pore formation: incorporating hollow substrates, removable sub-
strates, or gas (either directly, dissolved, or by means of a dissociating agent).
Open porosity structures can be formed by replication (Duocel [6]), deposition (In-
cofoam [6]) or by construction of solid ingredients with space between (for in-
stance, the 3D networks of lattice block materials (LBM) [5,6] or those prepared
by rapid prototyping techniques [11]). Open-cell solid structures may be called
sponges. Closed cells are produced by embedding or cementation of hollow ingre-
dients (ªsyntactic foamsº), or by foaming in the liquid state.
The expression ªmetal foamsº, strictly valid only for the liquid phase, is often
used to describe the solid product. Mixtures of metal powders and blowing agents
are compacted by extrusion or hot pressing providing a precursor material foamed
above the solidus temperature, a method called ªpowder compact foamingº (Alu-
light [5 10], IFAM-Foaminal [3,5 10], Alufoam [5,7]). The Formgrip material
[3,7] is made by remelting a stir-cast foamable precursor metal matrix composite.
Low-density-core material (LDC [3,6]) is produced by pore formation in the solid
state by the high gas pressure of entrapped dissociating blowing agents. A blowing
agent powder is mixed into the melt either in a crucible (Alporas [1±7]) or in the
gate in pressure die casting (Buehler [10]). Foaming of particle-reinforced metals
takes advantage of the stabilization of gas bubbles by the ceramic ingredients prod-
ucing a cellular metal matrix composite from the melt (Cymat [1 7], COMBAL
[10]). The formation of a gas metal eutectic is the principle for the production
of Gasar foams [1,2,6]. A slip reaction foam technique based on foaming of precur-
sor slurry by chemical reaction and a reaction sintering process for aluminides has
been described [10]. The originally closed cells of foamed metals may not be gas
tight after solidification owing to cracks in their walls.
The resulting cellular metal products can be differentiated by their structural fea-
tures. The term ªstructureº is used for the description of cellular materials at dif-
ferent levels of observation (structology): the geometric architecture of the solid
(skeleton) in the individual cells and their 3D arrangement, the variations of that
architecture within a considered sample or part (degree of uniformity), and the

microstructure of the solid itself and its surface. The cells are formed by plateau
borders (edges and nodes) and, in the case of closed cells, by walls connecting
them [12] with a particular microstructure of the metal, eventually containing
the remnants of the foaming additives and other microstructural inhomogeneities.
The multitude of structural features (see Chapter 4) and their spatial distribution
are still subject to investigation in order to correlate them with processing para-
meters and functional properties, aiming for the development of quality-relevant
specifications. Cellular metals are inherently heterogeneous: real samples usually
exhibit local non-uniformity like variations in cell architecture and mass distribu-
tion.
The simplest description of a cellular metal is given by stating the production
process, the composition of the metal, and the apparent density: producer/process
alloy composition apparent density [g/cm3]. The established symbol for the
corresponding heat treatment condition of light metals could be added; for exam-
ple: Alulight AlSiMgCu 0.45 T1.
References
1. L. J. Gibson, M. F. Ashby, Cellular Solids: 7. J. Banhart, M. F. Ashby, N. A. Fleck (eds),

Structure and Properties, 2nd ed., Cambridge Proc. Metal Foams and Porous Metal Structures,
University Press, Cambridge 1997. MIT, Bremen 1999.
2. M. F. Ashby, A. G. Evans, N. A. Fleck, L. J. 8. T. W. Clyne, F. Simancik (eds) Proc. Metal
Gibson, J. W. Hutchinson, H. N. G. Wadley, Matrix Composites and Metallic Foams, Euro-
Metal Foams: a Design Guide, Butterworth- mat 1999, Vol. 5, Wiley-VCH, Weinheim
Heinemann, Woburn 2000. 2000.
3. J. Banhart, N. A. Fleck, M. F. Ashby (eds), 9. H. P. Degischer (ed) Proc. MetallschaÈume,
Metal Foams Special Issue, Adv. Eng. Mater. Special Issue, Mater. Wissenschaft Werkstoff-
2000, 2(4). techn. 2000, 31(6).
4. J. Banhart (ed.), Proc. MetallschaÈume, MIT, 10. J. Banhart, M. F. Ashby, N. A. Fleck (eds),
Bremen 1997. Proc. Cellular Metals and Metal Foaming Tech-
5. J. Banhart, H. Eifert (eds), Proc. Metal Foam nology (MetFoam 2001), MIT, Bremen 2001.
USA Symposium, MIT, Bremen 1998. 11. A. BuÈhrig-Polaczek, in Proc. Materialsweek
6. D. S. Schwartz, D. S. Shih, A. G. Evans, 2000, Symposium H3, http://www.materials-
H. N. G. Wadley (eds), Proc. Porous and week.org/proceedings.
Cellular Materials for Structural Applications, 12. D. Weaire, S. Hutzler, The Physics of Foams,
MRS Symp. Proc. Vol. 521, MRS, Warendale, Clarendon Press, Oxford 1999.
PA 1998.
8 2.1 Foaming Processes for Al
2.1
Foaming Processes for Al
C. KoÈrner and R. F. Singer
The melt route for processing closed-cell metal foams is very attractive since this
approach allows economic handling of large quantities of material. Melt-route pro-
cesses are also well suited to the use of scrap as feedstock. In order to foam the
melt properly, gas must be introduced. This can be either done by gas injection
or by in-situ gas generation by a chemical decomposition of a foaming agent.
For the production of homogenous foams some prerequisites have to be fulfilled.
If foaming is by in-situ gas generation uniform dispersion of the foaming agent in
the melt within a time that is short compared to the decomposition or reaction
time of the additive is required. In addition, the escape of gas during the foaming
process has to be prevented. One important step to meet these requirements is to
increase the viscosity of the melt. There are several approaches to do this: foaming
in the semi-liquid state, incorporation of ceramic particles [1] or oxidation [2]. The
effect of the particles added is always twofold: beside increasing the melt viscosity
they stabilize the cell walls.
Independent of the specific foaming process it is found that porosity, quantified
by density, and mean cell size are intimately related. Generally, for a particular alloy
or composite the mean cell size can not be chosen independently of the porosity.
2.1.1
Gas Injection: the Cymat/Alcan and Norsk Hydro Process
The Cymat/Alcan and Norsk Hydro melt-foaming process is a continuous, gas-in-

jection method developed simultaneously and independently by Alcan [1] and
Norsk Hydro [3] in the late 1980s and 1990s. A sketch of the process developed
by Alcan is given in Fig. 2.1-1. The patent is now licensed and exploited by the
Cymat Aluminum Corporation [www.cymat.com].
The process employed by Hydro Aluminum, Norway is analogous. A metal ma-
trix composite (Al-wrought or Al-casting alloy matrix 10 30 vol.-% SiC or Al2O3
Figure 2.1-1. Principle of the melt-foaming route employed by Cymat. The foam casting process
for producing flat panels consists of melting and holding furnaces, the foaming box and foaming
equipment, and a twin-belt caster [1].
particles) is used as a starting material. The starting material is molten with con-
ventional foundry equipment and transferred to a tundish where gas, typically air,
is injected via small nozzles incorporated into a rotating impeller, thus forming a
dispersion of small gas bubbles. The bubble size can be controlled by adjusting the
gas flow rate, the impeller design (number of nozzles and their size), and the speed
of rotation of the impeller. The gas bubbles rise to the surface where they accumu-
late. The ceramic particles are trapping gas bubbles owing to the favorable interface
energy and serve as stabilizer of the cell walls and delay their coalescence. They
also reduce the velocity of the rising bubbles by increasing the viscosity of the
melt. That is, the particles reduce the kinetic energy of the rising bubbles and
hence the danger of mechanical rupture when they arrive at the surface. The re-
sulting metal foam, which is still liquid, is carried away by means of a conveyor
belt where it solidifies and cools. The relative density is predominantly controlled
by the process parameters, such as rotor speed, gas flow, and the amount of parti-
cles in the melt, and finally the solidification condition.
This process of casting aluminum foam is capable of producing slabs with a re-
lative density in the range 2 20 % (0.05 0.55 g/cm3). The average cell size is in-
versely related to the density (Fig. 2.1-2a) and is in the range 2.5 30 mm [1].
a)
Figure 2.1-2. Average cell size as a

function of the reciprocal density for:
a) Cymat foam [1]; b) FORMGRIP [7]
products. b)
Figure 2.1-3. Optical micrographs of Cymat foam produced by the gas-injection method: left)
cell structure (density about 0.3 g/cm3), inhomogeneous and anisotropic; right) foam surface
(density about 0.05 g/cm3).
The production facility set up by Cymat is capable of casting foam panels in con-
tinuous length at an average rate of 900 kg/h up to 1.5 m wide with a thickness
range of 25 150 mm. This shows that the process is relatively straightforward
and economical. Challenges that may require more work in the future include
the variation in cell size, the density gradient, and the anisotropy of the cell struc-
ture, which results from mechanical forces from the conveyor belt (Fig. 2.1-3).
Especially earlier prototype have been compressed by the conveyor belt producing
flattened pores causing poor stiffness and strength values along the thickness of
the slabs. It should also be noted that the preparation of the composite feedstock
requires relatively tedious long time stirring processes to achieve the proper homo-
genous distribution of the particles in the melt.
In principle, this foam generation technology also allows the casting of nonrec-
tangular, 2D profiles as well as 3D shapes.
2.1.2
In-situ Gas Generation: the Shinko Wire Process and the FORMGRIP process
The gas-injection method suffers from the fact that a relatively small number of
large bubbles is generated, which leads to rather coarse and irregularly shaped
pore distribution. Two techniques are described below, in which the foaming gas
results from a thermal decomposition of solid ingredient. In this way a huge num-
ber of bubble nuclei is created throughout the melt.
2.1.2.1 The Shinko Wire Process [2]

The manufacturing process of the Alporas foam is a batch casting process patented
by Shinko Wire Company Ltd., Japan (see Fig. 2.1-4) [2].
The installed manufacturing plant is capable of making large sized blocks of
foamed aluminum. For adjusting the viscosity of the molten aluminum 1.5 % Ca
is added at 680 hC and stirred for 6 min in an ambient atmosphere. The addition
and subsequent agitation of an element with a high oxygen affinity facilitates an
oxidation process on the surface of the molten metal and leads to an increase of
Figure 2.1-4. Manufacturing process for Alporas foams [2].
the viscosity by the formation of oxides: CaO, Al2O3, CaAl2O4. There is an appro-
priate stirring resistance for optimizing the foaming ratio [4]. The thickened alumi-
num is poured into a casting mold and stirred with an admixture of 1.6 % TiH2 as a
foaming agent. While vigorously stirring, the TiH2 dissociates and H2 -bubbles are
formed causing molten material to expand and to fill the mold. Then, the foamed
material is cooled by fans to solidify in the casting mold. After removal from the
casting mold. An Alporas block 450 mm wide, 2050 mm long, and 650 mm
high is sliced into plates.
Alporas is an ultra-light material with a closed-cell architecture (Fig. 2.1-5). The
density of the product is 0.18 0.24 g/cm3, the mean cell size is about 4.5 mm. Al-
poras is regarded as the best commercially available aluminum foam in terms of
regular cell micro structure. This is to a certain extent because transfer and defor-
mation of the still liquid foam, as in the Cymat/Norsk Hydro process, is avoided.
The cell architecture is the outcome of a growing process where bubble expand and
coalesce due to cell wall rupture.
Figure 2.1-5. Typical cell structure of an Alporas foam. As a result of the growth process most
cells are far from equiaxed. The global homogeneity is superior to other commercially available
aluminum foams.
2.1.2.1 The FORMGRIP Process

The FORMGRIP (Foaming of Reinforced Metals by Gas Release in Precursors)
process integrates some of the advantages of melt- and powder-route approaches
for foam production into one processing technique [5 7]. It comprises ªbakingº
a precursor material with the foaming agent entrapped in a mold to generate
foam in-situ by dissociation, which is very similar to the established powder routes.
However, the precursor is prepared via melt processing [7]. A diagram of the
FORMGRIP process is depicted in Fig. 2.1-6.
The precursor preparation comprises dispersion of a mixture of AlSi12 powder
and the pre-treated gas-generating TiH2 powder in an Al-9Si/SiCp (particle size
12.8 mm) composite melt by conventional mechanical stirring 1200 rpm for
50 70 s. The critical point is that during this processing step only a limited portion
of hydrogen is released from the foaming agent. This is achieved by the following
precautions: a low temperature of the melt at the moment of the hydride introduc-
tion (T 620 hC), a high melt viscosity due to SiC particles and a pre-oxidized, re-
tarded foaming agent [8]. The thermal pre-treatment of the foaming agent consists
of a two-step thermal oxidation sequence (24 h at 400 hC 1 h at 500 hC), which
slightly reduces the hydrogen concentration of the TiH2. The oxide barrier layer
formed on the powder surface slows down the kinetics of gas evolution. The
amount of the incorporated hydride is about 1.5 wt.-% of the melt mass.
The mixed melt is cast into a mold by cooling and pore growth is suppressed by
cooling. The SiCp particles are also important for foam stabilization, which be-
comes plausible from their distribution in Fig. 2.1-7. The resulting precursor ma-
terial already exhibits a porosity of 14 24 %.
The second stage of the FORMGRIP process comprises heating of the precursor
material above the solidus temperature. As a result, hydrogen released from the
TiH2 diffuses to the bubble nuclei already present and expands them further. Typi-
cal cell structures are depicted in Fig. 2.1-8 [7].
Figure 2.1-6. Diagram of the melt-based FORMGRIP process for production of near net-shape
metal foam parts [7].
Figure 2.1-7. Optical micrographs illustrating the distribution of SiC particles in Al-9Si alloy
based FORMGRIP foams, showing sections through: a) cell wall, b) a node. A significant fraction
of particles is located at the gas/melt interface [5].
Figure 2.1-8. Examples of cross sections of aluminum alloy FORMGRIP foams baked under
different conditions. Porosity, P, levels and the mean cell sizes, d, are: a) P 69 %, d 1.1 mm;
b) P 79 %, d 1.9 mm; c) P 88 %, d 3.1 mm [7].
The relation between relative foam density and average cell diameter fits that de-
rived for Cymat foams. Fig. 2.1-2b shows the mean cell diameter as a function of
the reciprocal density (see Section 4.1). The mean cell diameter is inversely propor-
tional to the density indicating that foam expansion is governed by cell coalescence
and the mean cell wall thickness is constant. The dependence of the cell diameter
on the density is the same for both contents of SiC particles. The influence of par-
ticle size has not yet been investigated. Theoretical work of Kaptay [9] and experi-
mental work of Weigand [10] indicate that a reduction of particle size will not lead
to a higher stabilization and therefore to a smaller critical cell wall thickness.
In terms of geometrical complexity and simultaneous microstructural control,
the FORMGRIP process surpasses the Shinko Wire one. There is no need to trans-
fer material from a mold into a die while the foaming process is going on. In ad-
dition, the precursor material can be shaped before the final baking step. The eco-
nomics of the FORMGRIP process, however, are clearly inferior to the other pro-
cesses discussed in this chapter. This is due to the discontinuous nature of the pro-
cess as well as its various number of processing steps.
14 2.2 Industrialization of Powder-Compact Foaming Technique
References
1. J. T. Wood, ªProduction and Application of 6. V. Gergely, T. W. Clyne, ªA Novel Melt-Based

Continuously Cast, Foamed Aluminumº in Route to Aluminium Foam Productionº in
Proc. Fraunhofer USA Metal Foam Symposium, Metal Foams and Porous Metal Structures,
7 8 October 1997, Stanton, Delaware. J. Banhart, M. F. Ashby, N. A. Fleck (eds),
2. T. Miyoshi, M. Itoh, S. Akiyama, A. Kitahara, MIT Verlag, Bremen 1999, p. 83 89.
ªAluminum Foam, Alporas: The Production 7. V. Gergely, ªMelt Route Processing for Pro-
Process, Properties and Applicationsº in duction of Metallic Foamsº, Department of
Metal Foams and Porous Metal Structures, Materials Science and Metallurgy, Cambridge
J. Banhart, M. F. Ashby, N. A. Fleck (eds), 2000.
MIT Verlag, Bremen 1999, p. 125. 8. A. San-Martin, F. D. Manchester, ªThe H Ti
3. P. Asholt, ªAluminium Foam Produced by (hydrogen titanium) systemº Bull. Alloy
the Melt Foaming Route Process, Properties Phase Diagrams 1987, 8(1), 30 43.
and Applicationsº in Metal Foams and Porous 9. G. Kaptay, ªInterfacial Criteria for Ceramic
Metal Structures, J. Banhart, M. F. Ashby, Particle Stabilised Metallic Foamsº in Metal
N. A. Fleck (eds), MIT Verlag, Bremen 1999, Foams and Porous Metal Structures, J. Banhart,
p. 133 140. M. F. Ashby, N. A. Fleck (eds), MIT Verlag,
4. L. Ma, Z. Song, ªCellular structure control of Bremen 1999, p. 141 146.
aluminium foams during foaming process of 10. P. Weigand, ªUntersuchung der Ein-
aluminium meltº Scripta Mater. 1998, 39(11), fluûfaktoren auf die pulvermetallurgische
1523 1528. Herstellung von AluminiumschaÈumenº,
5. V. Gergely, T. W. Clyne, ªThe FORMGRIP FakultaÈt fuÈr Bergbau, HuÈttenwesen und
process: foaming of reinforced metals by gas Geowissenschaften, RWTH, Aachen,
release in precursorsº Adv. Eng. Mater. 2000, MIT Verlag, Bremen 1999.
2(4), 175 178.
2.2
Industrialization of Powder-Compact Foaming Technique
J. Banhart and F. BaumgaÈrtner
There are many different ways to manufacture cellular materials [1]. One of the
available processes has become increasingly popular in the past few years and is
at the stage of industrial implementation now. The method is sometimes loosely
called the ªpowder-metallurgical routeº, but the term ªpowder-compact foaming
techniqueº seems more appropriate.
2.2.1
Principles of Foam Production
The technique consists of mixing aluminum or aluminum alloy powders with ap-
propriate foaming agents, which get entrapped by compacting this mix to a dense
product called ªfoamable precursor materialª. The powder mix can be compacted di-
rectly by hot pressing, conform extrusion, or powder rolling. Alternatively the pow-
Figure 2.2-1. Foam production by powder-compact foaming technique.
der may be cold compacted for better handing in conventional extrusion or rolling
(Fig. 2.2-1). Heating the precursor above its solidus temperature releases the pres-
sure on the foaming agent allowing decomposition and formation of bubbles. After
cooling a low-density foam structure of originally closed cells is obtained [2,3].
The method is not restricted to aluminum and its alloys: tin, zinc, lead, gold, and
some other metals and alloys can also be foamed by choosing appropriate foaming
agents and process parameters (see Section 2.1.2). The most common alloys for
foaming, however, are pure aluminum or wrought alloys such as aluminum
2xxx, 6xxx, or 7xxx alloys, e. g. AA 2014, 6060, 6061, 6082, or 7075. Casting
alloys such as AlSi7Mg (A356) and AlSi12 are also frequently used because of
their low melting point and good foaming properties, although in principle vir-
tually any aluminum alloy can be foamed by carefully adjusting the process para-
meters.
Quite complex-shaped metal foam parts can be manufactured by expanding the
foam inside a mold, thus confining spatial expansion. An example for one such
part is shown in Fig. 2.2-2. The part, developed in the framework of a feasibility
study, is a novel pantograph horn for an electrical locomotive. This light-weight so-
lution based on aluminum foam replaces traditional cast aluminum parts saving
30 % weight.
A nice feature of the technique is that composite structures consisting of an alu-
minum foam and bulk metal parts can be made without using adhesives. Exam-
ples are foam-filled aluminum sections and sandwich panels with an aluminum-
foam core and metallically bonded steel, aluminum, or even titanium face sheets.
For making such composites the foamable precursor material is first bonded to the
solid section or sheet by co-extrusion or roll-cladding, after which the foamable
core layer is expanded by heat treatment [4,5] (see also Section 3.3).
Figure 2.2-2. Aluminum foam

part (Schunk Sintermetalltechnik,
Giessen).
The advantages of the powder-compact route are obvious and are listed in
Table 2.2-1. Beside the first two features already mentioned, the flexibility arising
from the preparation of the precursor from powders is important. Alloys can be
made simply by mixing inexpensive elementary powders. No ceramic additives
are needed to stabilize the foam, in contrast to some of the melt-route foaming pro-
cesses in which up to 15 % silicon carbide has to be added [1,6]. However, if re-
quired, ceramic powders, metal fibers, or ceramic fibers can be added to the pow-
der blend for special applications, such as for reinforcement or to increase wear
resistance. Naturally, there are also some disadvantages that are inherent to the
process. Metal powder is more expensive than bulk metal and requires effort for
compaction. This rules out applications that require very cheap materials. More-
over, the size of aluminum foam parts that can be manufactured is limited by
the size of the baking furnace, and is therefore smaller than for some of the com-
peting melt-foaming processes. The largest sandwich components that have been
manufactured using the powder-compact foaming technique are about 2 m q 1 m
q 1 cm in size (possibly larger in future). At LKR in Ranshofen a part of similar
size was produced without face sheets (Fig. 2.2-3). True 3D-volume parts are
usually not thicker than 30 cm, a limit which is difficult to shift to higher values.
A large aluminum foam column produced at Fraunhofer IFAM was 1 m high and
18 cm in diameter, weighing 13 kg. In contrast, the liquid-metal route allows for
making panels 15 m in length [7] and 100 cm thickness [8]. However, as these pro-
cesses cannot be used for near-net-shape production and only permit very simple
geometries, they are appropriate for different fields of application. Continuous
foaming of long products is under investigation [9].
The middle column of Table 2.2-1 lists some of the problems that are still en-
countered when foaming aluminum with the powder-compact melting method
but which can, in principle, be solved with further research.
Table 2.2-1. Characteristics of powder-compact foaming method: advantages and disadvantages

that are inherent are listed together with points that are presently problematic, but can be solved
in principle.
Advantage Problem Disadvantage
Net-shape foaming possible Uniformity of pore structure Cost of powders

still not satisfactory
Composites can be Process control must be Very large volume parts

manufactured improved difficult to make
Parts are covered by metal skin Permeable (holes) Coating process

requires sealing
Graded porosity can be Difficult to control

achieved
Flexibility in alloy choice
No stabilising particles have to be

added
Ceramics and fibers can be added
Figure 2.2-3. Rear wall of an automobile made of

aluminum foam (LKR Ranshofen and Daimler-
Chrysler AG, see chapter 7.4).
2.2.2
Practical Aspects of Foam Production
2.2.2.1 Powder selection

The appropriate selection of the raw powders with respect to purity, particle size
and distribution, alloying elements, and other powder properties is essential for
successful foaming. Commercial air-atomized aluminum powders were shown to
be of sufficient quality. However, powders from different manufacturers led to no-

table differences in foaming behavior and empirical criteria have been derived to
facilitate the selection of powders. The cost of powders and the ability of a manu-
facturer to provide sufficient quantity with a constant quality are also crucial.
As already pointed out, alloys can be obtained in different ways. The fre-
quently used alloy AlSi7, for example, can be either prepared by atomizing a
AlSi7 melt, or by blending pure aluminum powder with 7 wt.-% silicon powder,
or, in a third way, by mixing 58 % of standard AlSi12 powders with 42 % alumi-
num powder.
2.2.2.2 Mixing
The mixing procedure should yield a homogeneous distribution of alloying ele-
ments and the foaming agent to ensure that high-quality foams with uniform
pore-size distributions are obtained. Powders are mixed in batches of 500 kg at
Schunk-Honsel in commercial large-scale tumbling mixers with parameters deter-
mined in technological tests. Alternatively, powder mixes can be obtained by aero-
dynamic mixing. For example, Alulight International GmbH Austria mixes alumi-
num and titanium hydride in large containers with 50 80 short pulses of pressur-
ized nitrogen gas.
2.2.2.3 Densification
Powder consolidation can be carried out by various techniques. It has to be ensured
that the foaming agent is completely embedded in the metal matrix and no resi-
dual open porosity remains. One way to obtain foamable precursor material with
nearly 100 % theoretical density is the combined use of cold isostatic pressing
(CIP) and ram extrusion. CIP is first applied to consolidate the powder mix to bil-
lets with a relative density of 70 80 % and a mass of typically 50 kg. These billets
are used in the subsequent extrusion step. Although CIPping is not absolutely ne-
cessary (powders have been filled into thin-walled aluminum cartouches and in-
serted into the extrusion machine without prior consolidation) it has additional ad-
vantages such as the prevention of powder contamination and powder de-mixing.
The CIP billets themselves are not foamable because of their large content of resi-
dual porosity, which causes massive hydrogen losses when the material is heated.
To obtain foamable material, the billets are preheated to 350 hC and extruded as
rods or any other profile. For this a horizontal direct extrusion machine is used
(25MN Schunk-Honsel). The extrusion machine is operated in cycles with a new
billet inserted after each extrusion step. This way rather high outputs can be
achieved.
Foamable material has also been manufactured by rotary continuous extrusion
in the so-called CONFORM process by Mepura (Ranshofen) [10]. Here a rotating
wheel is used to drag the powder into the consolidation chamber from which it is
pulled off in radial direction as a compacted wire. Foamable wires of about 8 mm
diameter were manufactured from wrought alloys containing titanium hydride.
2.2.2.4 Further processing of foamable material

The extruded material can be foamed as it is after consolidation or it can be worked
to the required shape. By conventional rolling, foamable sheets with thicknesses
down to about 2 mm are produced. Optionally, the foamable raw material can
be clad to conventional sheets of metal, of steel or aluminum for example, by at-
taching two sheets to either side of the foamable precursor before rolling. This
way a purely metallic sandwich structure is obtained. By deep drawing, the sheets
and the sandwiches may be transformed to 3-D-shaped sheets for special applica-
tions. In all cases it is favorable to start from near-net-shape precursors in order to
minimize foam flow [11].
2.2.2.5 Foaming
Heat treatment at temperatures above the solidus temperature of the foamable ma-
trix is necessary to produce the foam structure. The gas released by the decompos-
ing foaming agent may form pores in the solid state but only above the solidus are
bubbles formed and the matrix expands up to a maximum volume, that is to a
minimum density. The density and density distribution of the growing foam can
be controlled by several parameters. The foaming agent content in the precursor
material is obviously important, but furnace temperatures and heating rates also
have an influence [12]. The mold material, the mold shape, and the type of furnace
naturally influence the heating rate and have therefore also to be considered. A
careful control of the heating conditions during foaming is essential for obtaining
high-quality foams. The difficulty is that the liquid foam is thermodynamically un-
stable and conditions change constantly during foaming. There are various inter-
mediate stages: at first only the mold is heated directly, whereas the foamable ma-
terial receives heat only indirectly via heat conduction through the mold. Initially
there are merely some point contacts between the piece of foamable material in the
mold and the mold walls. However, as the temperature increases, the precursor
softens and assumes the contour of the mold thus increasing the transfer of
heat. Moreover, heat transfer via radiation gains importance with rising tempera-
tures. The reflectivity of the mold and precursor surfaces may change during the
process and add a further variable. Finally, after foaming has started, the thermal
conductivity of the precursor rapidly decreases thus reducing heat flow. As soon as
the mold has been filled with foam it has to be cooled down below its solidus tem-
perature to stabilize the foam structure. The phenomena during cooling are also
quite complex and difficult to describe for reasons similar to those mentioned
for the heating phase.
Typical densities of aluminum foams are in the range 0.4 0.8 g/cm3 including
the closed skin around the foam body. The final density of a foamed part can be
simply predicted if the volume of the hollow mold and the mass of the inserted
precursor material are known. The foaming mold may be loaded with several
small pieces or one single piece of precursor. Choosing the latter method (which
is preferred by LKR and SAS [13]) one has to take into account that each piece
of the expanding precursor material has a dense aluminum oxide layer on its sur-
Figure 2.2-4. Left) top of aluminum foam part made by inserting various pieces of foamable
material into the mold (dark figures indicate original size of the precursor pieces). Right) foam
part made of two pieces of precursor without achieving bonding between the two pieces.
face, which has to be broken up by expansion of the individual foam pieces. Incom-
plete foaming may cause the foamed pieces to remain separated even after the
foaming process (Fig. 2.2-4b). A relative movement of the foam pieces to each
other helps to break up the oxide films. Fig. 2.2-4 shows an example of a successful
formation of a foamed body from various pieces of the precursor and an example
of failure. In the former case the location of the original individual foam pieces can
still be identified from the contrast in gray scales between the various regions:
darker gray identifies oxide layers of extruded surfaces, brighter gray is the new
(expanded) surface. This effect is currently exploited to create foam panels and
other foamed parts for making designer objects.
2.2.3
State of Commercialization
Currently the foaming technique described is still in the stage of industrial imple-
mentation. Nevertheless, a number of companies have already made commitments
for a future production and are building up facilities [14]. The joint effort of Schunk
Sintermetalltechnik (Gieûen) and Honsel GmbH&Co KG (Meschede) is one exam-
ple. Owing to their collaboration with Karmann the activities are preferentially di-
rected towards foam and foam sandwich parts with a complex 3D geometry (see
Section 3.3). Alulight International GmbH is another example. It is a joint venture
of SHW (Germany) and Eckart Austria. The company offers aluminum foam
panels in sizes up to 625 mm q 625 mm, with thickness of 8 25 mm. Neuman
Alufoam, another Austrian company, also offers foamable precursor material
(extrusions) and foamed parts.
References
1. J. Banhart, ªManufacture, characterisation 8. T. Miyoshi, M. Itoh, S. Akiyama, A. Kitahara,

and application of cellular metals and metal ªAluminum Foam, Alporas: The Production
foamsº Prog. Mater. Sci. 2001, 46, 559 632. Process, Properties and Applicationsº in
2. J. Baumeister, German Patent DE 40 18 360, Metal Foams and Porous Metal Structures,
1990. J. Banhart, M. F. Ashby, N. A. Fleck (eds),
3. J. Banhart, ªFoam metal: the recipeº Euro- MIT Verlag, Bremen 1999, p. 125.
physics News 1999, 30, 17. 9. G. Stengele, H. MuÈcke, A. SchoÈne, German
4. J. Baumeister, J. Banhart, M. Weber, German Patent DE 197 34 394 A 1, 1998.
Patent DE 44 266 27, 1994. 10. H. P. Degischer, H. WoÈrz, DE Patent
5. H.-W. Seeliger, ªApplication Strategies for 4206303, 1992.
Aluminium-Foam-Sandwich Parts (AFS)º in 11. F. BaumgaÈrtner, H. Gers, ªBauteile aus
Metal Foams and Porous Metal Structures, AluminiumschaÈumenº Ingenieur Werkstoffe.
J. Banhart, M. F. Ashby, N. A. Fleck (eds), 1998, 3, 42.
MIT Verlag, Bremen 1999, p. 23. 12. I. Duarte, J. Banhart, ªA study of alumi-
6. P. Asholt, ªAluminium Foam Produced by nium foam formation kinetics and micro-
the Melt Foaming Route Process, Proper- structureº Acta. Mater. 2000, 48, 2349.
ties and Applicationsº in Metal Foams and 13. R. Kretz, F. Simancik, private communica-
Porous Metal Structures, J. Banhart, M. F. tion.
Ashby, N. A. Fleck (eds), MIT Verlag, Bremen 14. http://www.schunk-group.com, http://
1999, p. 133. www.alulight.com, http://www.neuman.at
7. Cymat Corp. (Canada), Product information
sheets, http://www.cymat.com 1999.
2.3
Making Cellular Metals from Metals other than Aluminum
G. Rausch and J. Banhart
The previous section was dedicated exclusively to aluminum foams. For many ap-
plications one would like to use cellular materials made from metals or alloys other
than aluminum. There have been some attempts to manufacture metal foams by
simply adapting the powder-compact process originally developed for aluminum to
other metals by adjusting the properties of the foaming agent and the process para-
meters. This procedure was successful in some cases. However, for high-melting
alloys the powder-compact foaming technique is difficult to implement and espe-
cially for titanium no promising results could be obtained. Here alternative routes
based on advanced powder metallurgy yielded better results. Therefore, in the cur-
rent section the topic will be slightly extended from ªfoamedº to ªcellular or porous
metalsº in a more general sense.
22 2.3 Making Cellular Metals from Metals other than Aluminum
2.3.1
Zinc
Zinc can be foamed by a straight-forward modification of the powder-compact tech-

nique. The foaming agent used for aluminum (TiH2) can be used, although ZrH2
seems to yield slightly better results. Powder properties and mixing procedures are
quite similar to aluminum. Only the pressing and foaming temperature has to be
chosen slightly lower than for aluminum due to the melting temperature of zinc at
419 hC. Foamed zinc shows a very uniform pore structure. This can be attributed to
the fact that the decomposition temperature of the foaming agents TiH2 matches
with the melting temperature of the metal. Therefore, melting and pore formation
occur simultaneously and round bubbles are created from the very beginning. In
contrast to aluminum, there is no solid state expansion range with corresponding
crack formation. Fig. 2.3-1a shows an example of a zinc foam.
2.3.2
Lead
Lead and lead alloys such as Pb Sn and Pb Sb can be foamed by another mod-
ification of the process. TiH2 and ZrH2 cannot be used as foaming agents because
of the low melting temperatures of pure lead (327 hC) and even lower solidus tem-
perature of the alloys. Quite good foams have been obtained by using lead(II) car-
bonate as a foaming agent: it decomposes above about 275 hC and releases CO2 and
water, which act as foaming gas. Fig. 2.3-1b shows an example of a lead foam
2.3.3
Titanium
Owing to its high melting temperature (1670 hC) and relatively low density (4.51 g/
cm3), titanium and its alloys are excellent materials for lightweight applications at
elevated temperatures and are widely used in aeronautical applications. Porous ti-
a) b)
Figure 2.3-1. Zn and Pb foams (width of sample is about 5 cm).
tanium structures have an additional potential for weight reduction and could even
be suitable for functional applications if the pore structure were open.
In principle, there are many possible production methods for cellular materials
based on titanium (see Section 2.4), most of them starting from metal powders.
1. Consolidation of slurry-saturated plastic foam.
2. Foaming and sintering of powder slurries.
3. Reaction sintering of elemental powder mixtures.
4. Foaming of powder compacts containing foaming agents (powder-compact
melting process).
5. Hot isostatic pressing and creep expansion of titanium compacts with en-
trapped inert gas.
6. Sintering of hollow spheres.
7. Sintering of compacted or loose powder filler mixtures.
While some of these methods (1 3) have not yet been investigated very intensively,
the feasibility of the foaming agent process (4) for titanium has been demonstrated
[1]. However, owing to the high temperatures during foaming titanium, the reac-
tivity of this metal with practically any non-inert gas and the lack of appropriate
foaming molds, this method is not suitable for producing shaped titanium foam
components. Hot isostatic pressing of titanium powder with gas entrapment (5)
has been successfully developed for some aircraft applications [2]. Metal hollow
spheres (6) can be produced using wet chemical methods for coating Styrofoam
spheres [3]. Shaping and sintering of these hollow structures typically result in ma-
terials with very low porosity.
One of the most promising methods for manufacturing open porous titanium
materials is the sintering of compacted or extruded mixtures of powders and fillers
that contain removable space-holder materials. The materials are mixed and
shaped by conventional PM techniques. After removal of the space holder the
green samples are sintered at temperatures of 1100 1400 hC. Bram and coworkers
use urea and ammonium hydrogen carbonate as space holders [4], which can be
removed by thermal treatment below 200 hC. Depending on the size and shape
Figure 2.3-2. Open porous titanium made by space-holder technique: left) pore size 1 4 mm;
right) pore size about 500 mm, porosity 55 80 %.
Figure 2.3-3. Pore structure of open porous titanium with 67 % porosity.
Figure 2.3-4. Strength and Young's modulus as a function of density obtained from bending and
tension tests [5].
of the space-holder powder, spherical and angular pores in the range 0.1 2.5 mm
can be obtained, resulting in overall porosities of 70 80 %. It was found that the
sintering activity can be increased by partially substituting titanium by titanium hy-
dride, thus yielding an increased compression strength.
At Fraunhofer IFAM, polymer granules were used as the space holder. They
were removed by a chemical process at temperatures around 130 hC, after pressing.
After space-holder removal, samples are sintered in vacuum at temperatures of
1100 1250 hC. Depending on the particle size of the granules, average pore dia-
meters in the range 200 3000 mm can be obtained. Fig. 2.3-2 shows some typical
samples. Fig. 2.3-3 shows the typical pore structure of samples based on spherical
space-holder granules. Beyond the primary pore structure, some microporosity
(secondary pores) inside the sintered network is visible. It was shown that the sec-
ondary porosity has a strong influence on the overall strength of the samples and
can be reduced by either changing the sintering parameters and/or partially replac-
ing titanium powder by titanium hydride [4]. As for all porous materials, the me-
chanical properties of cellular titanium are a function of density. Fig. 2.3-4 shows
the strength and Young's modulus obtained from bending and tension tests as a
function of density.
2.3.4
Steel
2.3.4.1 Powder-Compact Foaming Technique

The long experience in making aluminum foams from powder metallurgy (PM)
precursors encouraged researchers to transfer this process to higher-melting
materials such as iron-based alloys and steels. The major requirements for an
adaptation of the foaming agent process to this group of materials are the follow-
ing.
x Selection of suitable foaming agents.
x Development of alloys qualified for good ªfoamabilityº.
x Evaluation of compaction methods.
x Adaptation of the foaming process.
The basic requirements for foaming agents are: point of gas emission above
1000 1200 hC (depending on the alloy composition), broad temperature range of
gas emission (up to 1550 hC for nearly pure iron), and sufficient volume of gas re-
lease. It was found that especially metal nitrides and certain carbonates show a sig-
nificant gas emission and qualify for being useful as foaming agents. Examples are
manganese nitride, chromium nitride, molybdenum nitride, calcium carbonate,
strontium carbonate, and barium carbonate [5,6]. Theoretical investigations [5,7]
have shown that both the iron carbon and the iron boron system [7] are able
to meet the basic requirements for being foamed to iron-based metallic foams,
namely: a low melting point matching the decomposition temperature of the foam-
ing agent, and a broad two-phase semisolid region in the phase diagram, thus
creating a wide foaming interval. As for the production of Al based foamable pre-
Figure 2.3-5.Iron-based foams

obtained from the foaming agent
process.
cursor material, extrusion has been successfully used for compacting iron powder
mixtures. The resulting samples are shown in Fig. 2.3-5.
Experiments with powder mixtures of iron and carbon have shown that free car-
bon without any additional foaming agent already leads to a certain degree of po-
rosity. Carbon is oxidized during the foaming process and the resulting gaseous
CO and CO2 creates pores. However, pore size distributions are not uniform and
pore shape is usually rather irregular. The porosity mainly results from large, iso-
lated pores (Fig. 2.3-6). Adding 0.25 % SrCO3 leads to an increase of porosity to
55.5 % (Fig. 2.3-7). The pore structure at this composition appears to be more
homogeneous and the average pore size is obviously lower. Increasing the amount
of SrCO3 results in a further increased porosity (64.3 %). From that it can be con-
cluded that SrCO3 has a significant influence on the achievable porosity and the
maximum expansion. The foaming agent technique has therefore been shown to
be feasible for steel. However, foaming of stainless steel or even superalloys has
not yet been successful and the general state-of-the-art of foaming steel with the
foaming agent method is still far behind the aluminum foaming technology.
Figure 2.3-6.Iron-based metal foams made from extrusion-pressed powder mixtures of

Fe 2.5 % C: left) 0.0 % SrCO3, middle) 0.25 % SrCO3, right) 0.50 % SrCO3.
Figure 2.3-7. Average porosity as a function of foaming agent content (SrCO3).
2.3.4.2 Steel Foams from Powder Filler Mixtures

All foamed metals have essentially closed cells. For certain applications (filters,
membranes, biomedical applications) open porosity is required. For this class of
materials the space-holder technique (see also Section 2.3) can be used. The pro-
cess used for steel is very similar to the one described in the titanium section.
The process starts with a mixture of metal powders and the filler powder. The mix-
ture is compacted, usually by axial compression in a conventional powder press. If
necessary, an additional bonding agent is used in order to achieve a better strength
of the green samples. After pressing an additional drying step is optional. After
this the filler/bonding agent phase is removed from the samples, in a chemical
(catalytic) or thermal process. After complete filler removal the samples are sin-
tered in a furnace under hydrogen atmosphere. Either urea [4] or plastic granules
[8] can be used as space holders. In Fig. 2.3-8 examples of porous 316L and Inconel
600 materials are given, developed by Forschungszentrum JuÈlich GmbH. The po-
rosity of these materials is about 70 % with an average pore size of 1.0 1.4 mm.
Figure 2.3-8.SEM images of sintered specimen: left) stainless steel 316L, 1100 hC, 1 h, particle
size I16 mm; right) Inconel 600, 1250 hC, 1 h, particle size 100 200 mm [4].
28 2.4 Recycling of Cellular Metals
References
1. G. Rausch, T. Hartwig, M. Weber, 5. B. Kriszt, A. Falahati, H. P. Degischer,

O. Schultz, ªHerstellung und Eigenschaften ªMachbarkeitsstudie zur Herstellung von
von TitanschaÈumenº Materwiss. Werkstoff- Eisenbasisschaumº in MetallschaÈume, J. Ban-
techn. 2000, 31, 412 414. hart (ed), MIT Verlag, Bremen 1997,
2. R. L. Martin, R. J. Lederich, Metal Powder p. 59 70.
Rep. 1992, Oct, 30. 6. C.-J. Yu, H. Eifert, M. KnuÈwer, M. Weber,
3. O. Andersen, U. Waag, L. Schneider, ªInvestigation for the selection of foaming
G. Stephani, B. Kieback, ªNovel metallic agents to produce steel foamsº Mater. Res.
hollow sphere structuresº Adv. Eng. Mater. Soc. Symp. Proc. 1998, 521, 145 150.
2000, 2, 192 195. 7. M. KnuÈwer, Herstellung von Eisenschaum
4. M. Bram et al., ªPreparation and Character- nach dem pulvermetallurgischen Treibmit-
ization of High-Porosity Titanium, Stainless telverfahren, Dissertation, UniversitaÈt Bre-
Steel and Superalloy partsº in Metal Foams men, Fraunhofer IRB Verlag, Stuttgart 1999.
and Porous Structures, J. Banhart, M. F. 8. G. Rausch, M. Weber, M. KnuÈwer, ªNeue
Ashby, N. A. Fleck (eds), MIT Verlag, Bremen Entwicklungen zur Herstellung von
1999, p. 197 202. StahlschaÈumenº Materwiss. Werkstofftechn.
2000, 31, 424 427.
2.4
Recycling of Cellular Metals
H. P. Degischer
Cellular metals compete with polymers for some applications. The recyclability of
metals is one of their benefits, enabling ecologically sustainable product life cycles
[1]. Compared to bulk metal products, there are two complications to be tackled in
remelting cellular metals:
x The high surface-to-volume ratio of the order of 100/length unit increases the ex-
tent of surface adsorptions and reactions.
x The low average density, due to the high porosity filled with gas, makes the cel-
lular material float on its melt.
2.4.1
The Remelting of Cellular Metals
Heating cellular aluminum in air enhances the growth of the oxide film not
only in open cellular structures, but also in those foamed by formation of
closed pores in the liquid, or semi-solid state, which usually become permeable
to gas after solidification and cooling. When heating the cellular aluminum up
to the melting point in a short time, about 1 h, the thickness of the oxide film
may reach about 10 mm and any cracks formed will be covered quickly by
ongoing oxidation. The growth will slow down after prolonged heating, when
the thickness reaches about 100 mm [2], which is similar to the cell-wall thickness.
The dehydrated oxide is stable up to about 2000 hC, so the cellular aluminum part is
slowly converted to a cellular alumina structure, which may maintain its macro-
scopic shape if it does not break up under its weight, or an external force.
It is difficult to submerge a cellular part in a melt because it tends to float
and contains a lot of air, which would have to be replaced by melt. Therefore
the cellular structures have to be compressed as much as possible, as is done
with packaging foils, or shredder scrap before melting [1,3]. Cellular metals
can be shredded too and treated as normal shredder scrap. Nevertheless, the
high content of surface oxides reduces the efficiency of metal reclamation.
The only experience is from small batch laboratory trials on scrap from pow-
der-compact foamed aluminum [4], in which efficiencies up to 80 % have
been reached. Oxides and other impurities have to be removed by the usual
cleaning process for melts to the extent necessary for the later production pro-
cess [3]. The purity requirements may be very low for the production of cellular
metals. Impurities in precursor material prepared for powder-compact foaming,
as well as for foaming in the melt, might be advantageous, as long as they act
as nuclei for the formation of pores. There might be even an up-grading of low-
quality scrap when used for the production of foamed metals [5]. The Ti remain-
ing from the blowing agent (usually I0.5 wt.-%) is soluble below 0.12 wt.-% at
the peritectic temperature of the binary aluminum melt [1] and does not de-
grade the quality of the alloy, but provides a grain-refining effect during solidi-
fication. ALPORAS foam contains about 2 wt.-% Ca, which is an element to be
restricted below 0.1 wt.-% in all cast alloys. Ca will be partly oxidized and thus
transferred into the dross, but some care has to be taken in the mix of ingre-
dient scrap for not exceeding the specified impurity levels for the secondary
alloy. The impurity levels of secondary aluminum cast alloys are not as stringent
[1,3], so that scrap of the common types of cellular aluminum can be added to
the remelting furnace.
2.4.2
Recycling of Cellular Metal Matrix Composites
Cellular aluminum produced by foaming particulate-reinforced melts, like the

Cymat process [6] (formerly developed by Alcan [7] and Hydro [8]), the FORMGRIP
process [9], and the shape foaming technique COMBAL [10] may be collectively
named MMC foams. It might be of interest to recycle the MMC matrix of the
foam without significant reduction of the particle content, because of the economic
value of reclaimed MMC based on the primary processing cost to overcome the
nonwettability. Any reuse of MMC saves the effort necessary to bond the two com-
ponents together during processing. Originally, this type of foam was a spin-off of
particulate-reinforced aluminum processing, consequently it has been proposed to
reuse particle-reinforced aluminum alloys in the production of aluminum foam
[11]. One advantage of this recycling route is that the specification requirements
for the production of metallic foams, may not be as stringent as placed on MMC
for bulk components, thus higher quantities of oxide skins and other impurities
may be tolerated, or are even advantageous in foamable aluminum. Figure 2.4-1
shows the material flow cycle for discontinuously reinforced metals and cellular
metals made from MMC indicating the possible interactions between these two
cycles:
x Remelted MMC, with, or without cleaning treatments, can be used as a base ma-
terial for the melt foaming processes and for FORMGRIP.
x Any secondary alloy, including that reclaimed from MMC melts, can be used for
production of cellular structures by any of the known production methods.
x Scrap from MMC foams can be introduced into the MMC cycle, if quality
requirements are met.
The same rules have to be obeyed for the recycling of MMC foams as for the MMC
matrix [11] in addition to the pretreatment (compaction and drying) to reduce gas
evolution. Two main problems have to be tackled when remelting MMC.
x The reactivity of the reinforcement with the melt increases with temperature.
x The dewetting tendency of the reinforcement limits conventional melt cleaning
methods.
Figure 2.4 -1. Recycling of discontinuously reinforced metals and of metal foams based on MMC,
indicating the reuse of secondary MMC for the processing of foamed metals either by the powder-
compact method, or by the melt-foaming technique.
The reaction between 6xxx type wrought alloys and alumina reinforcements is
driven by the Mg content forming MgAl2O4, spinel. An original level of approxi-
mately 2 vol.-% spinel seems to stabilize at just above 3 vol.-% after several remelt-
ing cycles; this does not influence the mechanical properties significantly [12], but
may increase wettability and, consequently, reduce foamability. SiC-reinforced
wrought alloys cannot be remelted without severe aluminum carbide formation
at the interfaces. The formation of Al4C3 would be detrimental as it significantly
reduces the corrosion resistance of the component. Furthermore, it affects the
foamability by increasing the viscosity of the melt and above all by providing wet-
tability. Nonwettability of the additives is necessary for the trapping of gas bubbles.
In the case of Al casting alloys containing 7 12 wt.-% Si reinforced by SiC, there is
the chance to conserve the integrity of SiC by keeping the melt temperature below
750 hC as recommended for primary foundry technology [13].
A detailed study of the recycling of SiC-particle reinforced Al Si casting alloys is
given elsewhere [14], where the quality criteria of the melt are given and remelting,
recycling, and holding practices are described.
x Dry, pre-heated scrap can be added to the melt between 700 and 750 hC.
x During a rest period, usually more dross is formed than on primary material. It
contains mainly oxides and SiC-particles without reducing significantly their con-
tent in the melt.
x The dross has to be skimmed.
x An impeller is introduced to produce strong movement under the surface skin.
x Fluxing and degassing with argon (SF6 may also be used) to remove oxides and
reduce the hydrogen content.
x The melt is allowed to sit and then skimmed.
x The melt is mechanically agitated, without forming a vortex, to distribute the SiC
particles homogeneously.
No loss in particle content was reported and the removal of porosity, oxide films,
and hydrogen were efficient. The amount of dross generated is relatively high
and may amount to more than 10 % of the total weight.
If the ceramic particles should be removed from the aluminum melt, conven-
tional salt addition, or fluxing techniques (as are executed to remove oxide films
[14]) can be applied. Gravity settling allows the fluxed ceramic ingredients to
float to the dross at the top of the aluminum melt. Rotary salt furnace technology
is an established reclamation process to recover aluminum from various mixtures,
including particle reinforced metals; however, this requires 20 50 wt.-% salt [13].
Both wrought and foundry alloys and even machining chips can be recovered
using this technique. Recovery of about 80 % of the available aluminum can be ex-
pected [15].
The efficiency of particle removal by fluxing is related to the probability of con-
tact between the ceramic constituent and the flux. Duralcan, a supplier of particu-
late-reinforced aluminum, proposes to incorporate the salt into a melt agitated by
gas injection [16]. Thus dewetting is achieved with much smaller salt additions
(I1 wt.-% for alumina and about 1.5 wt.-% for SiC) and in combination with
the adsorption of gas to the reinforcement also accelerates particle separation by

floating, where it can be skimmed off.
2.4.3
Conclusions
There are the following possibilities for recycling, or reuse of cellular metals (espe-
cially cellular aluminum).
x Recycling of unreinforced cellular metals can be carried out as for packaging ma-
terial (foils), or shredder scrap, by remelting, with an efficiency probably reduced
to 80 % to produce bulk cast products from the secondary metal. In the case of
recycling ALPORAS, the impurity level of Ca has to be controlled.
x Recycling of MMC foams can be done by extracting the matrix metal for conven-
tional secondary products. The reinforcement may be segregated to the dross and
deposited.
x Recycling of MMC foams can alternatively be achieved by remelting according to
the precautions for MMC recycling, but at a reduced efficiency due to the in-
creased oxide content to produce MMC foams again.
x The foam production itself may be based on an up-grading of secondary material
with higher contents of impurities that increase foamability. In particular, the
cycles for MMC and MMC foams can be closely related.
References
1. Grundlagen und Werkstoffe, Aluminium 7. I. Jin, L. D. Kenny, H. Sang, US Patent

Taschenbuch, Vol. 1, Aluminium Verlag, 5 112 697, 1992.
DuÈsseldorf 1995. 8. W. W. Ruch, B. Kirkevag, NO Patent 1989,
2. American Society of Materials, Proc. 4th World Patent wo 91/01387, 1991.
ASM Int. Conf. Recycling of Metals, 17 18 9. V. Gergely, T. W. Clyne, ªA Novel Melt-
June 1999, Vienna, ASM International, Based Route to Aluminium Foam
Metals Park, OH. Productionº in Metal Foams and Porous
3. K. Krone, Aluminium Recycling, Verein Metal Structures, J. Banhart, M. F. Ashby,
Deutscher Schmelzhuetten e. V., DuÈsseldorf N. A. Fleck (eds), MIT Verlag, Bremen 1999,
2000. p. 83 89.
4. M. Strini, Private communication, ARC- 10. D. Leitlmeier, H. Flankl, ªDevelopment of a
Leichtmetall Kompetenzzentrum Ranshofen, New Processing Technique Based on the Melt
Austria 1996. Route to Produce Near Net Shape Foam
5. H. P. Degischer, F. Simancik, ªRecyclable Partsº in Proc. METFOAM 2001, MIT Verlag,
Foamed Aluminium as an Alternative to Bremen 2001, p. 171 174.
Compositesº in Environmetnal Aspects in 11. V. Gergely, H. P. Degischer, T. W. Clyne,
Materials Research, H.Warlimont (ed), DGM, ªRecycling of MMC and Production of
Oberursel 1994, p. 137 140. Metallic Foamsº in Comprehensive Composite
6. Cymat Corporation, Mississauga, Canada Materials, Vol. 3, T. W. Clyne et al. (eds),
2000, http://www.cymat.com. Elsevier, London 2000, p. 797 820.
12. D. M. Schuster, M. D. Skibo, R. S. Bruski, and Fabrication of Light Metals and Metal
R. Provencher, G. Riverin, ªThe recycling and Matrix Composites, Montreal, TMS, 1992,
reclamation of metal-matrix compositesº p. 598 604.
J. Metals 1993, 45(May), 26 30. 15. T. F. Klimowicz, ªThe large scale commer-
13. Duralcan, Duralcan Composite Casting cialization of aluminum natrix compositesº
Guidelines, Duralcan USA, San Diego, CA J. Metals 1994, Nov, 49 53.
1990. 16. Duralcan, Aluminium Recovery from Metal
14. R. Provencher, G. Riverin, C. Celik, Matrix Composite Scrap, Duralcan USA, Novi,
ªRecycling of Duralcan Aluminium Metal Michigan 1996.
Matrix Compositesº in Proc. Adv. Production
2.5
The Physics of Foaming: Structure Formation and Stability
C. KoÈrner, M. Arnold, M. Thies, and R. F. Singer
A foam is a dispersion of gas bubbles in a liquid in which the bubbles are de-
formed due to their mutual interaction. Cellular metals produced via gas bubbles
in liquid metal are commonly named metal foams although they do not strictly
meet the definition above after solidification. Owing to the surface energy neces-
sary to form the metal±gas interface a foam is never in equilibrium and hence per-
manently trying to lower the internal energy by reducing the internal surface. That
is, the cellular structural state of a foam evolves with time and its actual structure is
a function of its history including all thermal and mechanical influences. The foam
structure is normally strongly disordered and evolves by some combination of
three basic mechanisms: bubble coalescence via film rupture; bubble coarsening
via diffusion of gas from smaller to larger bubbles; and drainage downwards
and out from the foam in response to gravity [1]. It is only in recent years that
some progress has been made towards an understanding of the basic mechanisms
governing the temporal evolution of foams [1±5].
Owing to their opacity and the high temperatures, in-situ observation of the
structure evolution of metal foams is difficult [6]. Fortunately, the solidification pro-
cess is in general much faster than the evolution processes at least as far as the
evolution is not caused by expansion. That is, nearly full information of the
foam formation process can be extracted from ex-situ investigation of foamed sam-
ples in different stages of expansion.
34 2.5 The Physics of Foaming: Structure Formation and Stability
2.5.1
Isolated Gas Bubble in a Melt
Since a foam consists of interacting gas bubbles it is helpful to consider in a first

step an isolated gas bubble in an infinitely extended fluid. The dynamics of a sphe-
rical bubble with radius R is described by the Rayleigh equation [7,8]
S 3 _2 R_ 1 2s
RR R S 4v (Pbubble s s PT ) (1)
2 R r R
where Pbubble bubble pressure, PT equilibrium pressure in the liquid, s sur-

face energy, n kinematic viscosity, r density.
The left hand side of Eq. (1) describes the inertia and viscous forces that both
delay bubble growth. Neglecting viscous effects the time for the formation of a bub-
ble in aluminum with radius R 1 mm at a bubble overpressure of DP 10 4 bar
is given by [7]
r
r
t 0:915 R 0:015s (2)
DP
That is, inertia effects delaying bubble expansion can be neglected if foam forma-
tion is on a time scale of seconds. The contribution of viscous forces for liquid me-
tals is normally very small, for example about 4nr(R/RÇ) z 10 6 bar with RÇ 1 mm/
s; R 100 mm; n 1 mm2/s. On the other hand, if foaming takes place in the
semi-liquid state, where the viscosity of the metal is several orders of magnitude
higher than in the liquid state, viscous forces might delay bubble expansion.
Bubble expansion takes generally more than one second for commercially known
foaming methods for metals [9]. In this case, the viscous and inertia forces can be
neglected and the Rayleigh equation reduces to pressure equilibrium at the gas±li-
quid interface
s s
Pbubble PT S 2 P0 S rgh S 2 (3)
R R
where g gravity constant, P0 ambient pressure, h depth.

The pressure contribution resulting from the surface energy (s 0.2 N/m for
Al) is 0.04 bar and 4 bar for bubble diameters of 100 mm and 1 mm, respectively.
For h 100 mm the hydrostatic pressure is about 0.02 bar. Owing to gravity
there is a pressure gradient present in the melt that deforms the bubble and
makes it move. The bubble is accelerated until a stationary velocity, v, is reached
where the resulting viscous forces balance the buoyant forces (Fig. 2.5-1).
For nearly spherical bubbles the rising velocity v can be calculated from Stokes'
law [7]
s
rgR2 Q
v for R II (4)
3h rpg
where h nr dynamic viscosity.

Figure 2.5-1. Velocity field around a rising bubble

in a liquid. The velocity and deformation of the
bubble depends on the viscosity of the melt, the
surface tension, and the bubble size.
For a pure aluminum melt and bubble radii R of 100 mm and 10 mm the rising
velocity is about 1 cm/s and 100 mm/s, respectively. Additives in the fluid like SiC
or Al2O3 particles influence the movement of the bubbles and are able to stabilize
them [10±12]. They have an effect on both the viscosity of the melt and the surface
tension. How these particles actually operate and how their action can be opti-
mized is not yet understood and still a matter of research.
A gas bubble grows or shrinks due to gas exchange with the surrounding melt
[13]. There is a gas flow from the liquid into the bubble if the concentration of dis-
solved gas in the liquid, as a result of the decomposition of a foaming agent for
example, is higher than the equilibrium concentration in the liquid given by
Henry's law for a given gas pressure in the bubble. Since hydrogen, which is pre-
ferentially used as foaming gas, dissociates when dissolved in aluminum the equi-
librium hydrogen concentration, ceq, at the gas±liquid interface is governed by
Sievert's law, a special form of Henry's law [14]
2760
p mol
ceq 1:4 q 10s3 q 10s T Pbubble p (5)
cm3 bar
where T temperature of the melt.
2.5.2
Agglomeration of Bubbles: Foam
A foam is an agglomeration of many gas bubbles taking a polyhedral form due to

their mutual interaction (Fig. 2.5-2).
Usually, the volumes separated by thin walls are named cells. As mentioned be-
fore a foam is not a static structure but always evolving towards a lower internal
energy. That is, already during the formation of a foam there are processes
going on which try to alter the structure. In a pure liquid these processes are so
fast that the development of a foam is not possible. For this reason additives
have to be added to stabilize the foam. Foams made without a suitable additive
boil, that is, the development of long-lasting membranes is not possible and the
Figure 2.5-2. Cell structure of an aluminum foam

produced by the FORMGRIP process foamed under
different conditions: a) porosity, 79 %; mean cell dia-
meter, 1.9 mm; b) porosity, 88 %; mean cell diameter,
3.1 mm. The arrow marks a residual portion of a
ruptured cell wall [30].
Figure 2.5-3. 2D-lattice Boltzmann simulation critical lower value. The melt boils, the number
of the growth of bubble nuclei by in-situ gas of bubbles decreases, and gas is lost to the
generation. The cell walls are not stabilized, so environment. The structure collapses once a
two bubbles coalesce if the cell wall reaches a critical expansion factor is reached [31].
structure is unstable (Fig. 2.5-3). In order to obtain a foaming material additives

have to be added to the liquid, which slow down cell-wall rupture [15]. Conse-
quently, a metal foam never consists of a pure metal. There must always be addi-
tives like SiC or Al2O3 present [11,16]. These additives are either added deliberately
from outside [11], produced during processing by oxidation [17], or are already
present in form of oxides if metal powders are used [18].
There are two different strategies to introduce the cell forming gas into the melt:
by injection through a nozzle [11,19] or by in-situ gas segregation or generation
[9,17]. The latter can be achieved by a chemical decomposition of a foaming
agent or by creation of a supersaturation of gas in the melt. If foaming is realized
by gas injection the gas bubbles are directly created without bubble nucleation and
growth. The bubble size is determined by various parameters such as the nozzle
geometry, the gas flow rate, and the impeller speed. Experience shows that the bub-
bles generated are relatively large. The bubble shape depends on the viscosity of the
melt, the bubble size etc.
Foaming by in-situ gas generation starts with homogeneous or heterogeneous
bubble nucleation followed by bubble growth by gas diffusion into the nuclei
[13] (Fig. 2.5-4). The rate at which bubbles nucleate homogeneously, NÇ0, is given
by [20]
DGhom 16ps 3
N_ 0 c0 f0 es kT with DGhom (6)
3DP2
Figure 2.5-4. Expansion stages of an AlMg1 completely liquid. Already at the very beginning
foam with 0.4 % TiH2 produced by powder of foam expansion coalescence occurs fre-
compaction. By applying a high ambient pres- quently indicating that the initial number of
sure of 100 bar during heating bubble nuclea- nuclei will not have much influence on the
tion was suppressed until the sample was residual foam structure.
where c0 concentration of gas molecules, f0 frequency factor of gas molecules

joining the nucleus, k Boltzmann constant, DGhom activation energy for homo-
geneous nucleation, DP gas pressure resulting from the dissolved gas following
Sievert's law.
Heterogeneous nucleation occurs when a bubble forms at an interface between
two phases, between a ceramic particle and the melt, for example. The rate for het-
erogeneous nucleation, NÇ1, is given by [20]
DGhet (7)
N_ 1 c1 f1 es kT
where c1 concentration of heterogeneous nucleation sites, f1 frequency factor

of gas molecules joining the nucleus, DGhet activation energy for heterogeneous
nucleation.
In the presence of heterogeneous nucleation sites this type of nucleation will be
favored over homogeneous nucleation because of its lower energy barrier. Potential
nucleation sites are the foaming agent particles and the particles present for foam
stabilization and alloying [18]. For polymers it is known that the nucleation rate can
be influenced by the processing parameters [20±22]. A systematic investigation of
the nucleation rate for metals as a function of the processing parameters is yet
missing but it is expected that the mechanisms are very similar to those in poly-
mers. Owing to the presence of additional particles in the melt, heterogeneous nu-
cleation is probably the main nucleation mechanism.
The shape of the starting gas bubbles changes during growth from spherical to
polygonal (Fig. 2.5-4). From a pure energetic point of view one would expect the
formation of Kelvin or Weaire±Phelan cells [3], which minimize the total internal
surface and hence the total internal energy. Actually, most of the cells are far away
from these ideal structures. That is, the complex formation process of the foam
does not necessarily lead to the energetically preferred structure. When two cells
coalesce due to cell-wall rupture the resulting cell is strongly deformed (Fig. 2.5-
2b). It is hampered to take an energetically more suitable form due to the presence
of the other cells. As a result, the cell geometry is in general not equilateral and the
cell walls are curved due to the pressure difference of neighboring cells. One effec-
tive way to characterize the cell structure is by the shape factor, F, calculated from a
2D cut through the foam [23] (see also Section 4.1)
4p X
n
ai
F (8)
n i li2
where n total number of cells, i cell number, li cell-boundary length, ai

cell area.
The shape factor describes the deviation of the cell geometry from a circle with
F 1. It is found that F decreases much more with decreasing foam density
than expected from the transition of spherical to polygonal cells (Fig. 2.5-5) [23].
That is, the deformation of the cells and the occurrence of bent cell walls increases
continuously during foam expansion. Thus, polygonal cell structures for different
densities are not self-similar.
A further mechanism that might lead to a structural change of the foam is coar-
sening analogous to Ostwald ripening. As a result of the pressure difference of
about 10 3 bar of neighboring cells with different size, there is a concentration gra-
dient (see Eq. 5) and therefore a flow of dissolved gas from smaller to larger cells.
Hence, small cells shrink and large cells grow, so the foam structure coarsens [24].
This effect is very pronounced for aqueous foams due to the small cell wall thick-
ness (about 100 nm). Numerical calculations [25] confirm the experimental finding
that this kind of coarsening is an effect of secondary importance on the relevant
time scale for the production of aluminum foams with a typical cell wall thickness
of 50±300 mm. On the other hand, gas loss by diffusion to the environment can be
substantial for long holding times if the hydrogen partial pressure in the ambient
atmosphere is zero (Fig. 2.5-6).
The formation of an aluminum foam by in-situ gas generation is from the very
beginning intimately correlated with cell coalescence (Fig. 2.5-4). If foam expan-
sion would be a mere enlargement of the structure without cell coalescence, so
the number of cells is constant during foaming, one would expect the mean
pore diameter, D, to be proportional to rrel1/3 (rrel relative foam density). The
Figure 2.5-5. Mean shape factor, F,

of the cell structure as a function of
the foam density for an AlSi10Mg0.6
foam produced by powder compac-
tion [23]. Energetically optimized cell
structures would show a form factor
equal to that of a pentagon.
Figure 2.5-6. Cell structure of an Alulight foam

after heating to 670 hC under an ambient pres-
sure of 100 bar and subsequent foaming by
pressure reduction to 11 bar. The thick metal
layer at the surface is the result of gas loss to
the environment during the 15 min in the liquid
state.
last expression is deduced from a simple cubic plate model for the cells with the
assumption that the volume of the cell-forming material per cell is constant. Actu-
ally, the experimental data (see Fig. 2.5-7) is much better fitted by
d 3d
D p for rrel II 1 (9)
1s 3
1 s rrel rrel
which follows from a cubic plate model assuming a constant mean cell-wall thick-
ness, d. That is, the mean cell-wall thickness is constant during expansion. It de-
pends on the alloy and is found between 50 mm and 300 mm for aluminum
foams. Consequently, given a fixed alloy cell size and relative density are intimately
related and can hardly be influenced by processing parameters [23]. In order to
generate aluminum foams with a lower density and smaller cells the mean cell-
wall thickness has to be reduced. How the cell walls of metal foams are stabilized
by the particles and which properties of them (quantity, size, form) determine the
mean cell-wall thickness is not yet understood and a matter of flow research [18].
Coalescence, rupture of a cell wall, is a mechanical process that depends on the
mechanical stability of the membrane. It has statistical character and occurs if a
local fluctuation of the film thickness leads to faster local thinning because it is
not compensated by restoring forces. The different physical mechanisms that are
responsible for the mechanical stability of a membrane have been extensively in-
vestigated for aqueous films [5,15]. Experimental studies of the stability of metal
membranes are not known in literature. Physical forces between the two surfaces
of the film like the van der Waals or electromagnetic interaction are only relevant
Figure 2.5-7. Mean pore diameter

as a function of the reciprocal
foam density for a wrought alloy
(open symbols) and a cast alloy
(full symbols) foam produced by
powder compaction. The samples
were foamed at various ambient
pressures. The full lines give the
results for a cubic plate model
(Eq. 9) with constant mean cell-
wall thickness of 180 mm for the
wrought alloy and 130 mm for the
cast alloy [31].
for very thin films (I1 mm). For thicker films, as it is the case for metal foams,
there are two effects that might lead to a kind of elasticity of the membrane.
The Gibbs' effect [15] takes into account that a local thinning of the membrane
leads to a local thinning of the surfactants. As a result, the surface energy increases
at this point leading to a restoring force. The Marangoni effect [15] describes that a
gradient of the surface energy leads to a flow of surfactants taking with it a thin
film of fluid in the direction of lower concentration. In order to maximize both ef-
fects the surfactant should minimize the surface energy. At the moment, it is not
clear whether the stability of metal membranes is based on one of these two me-
chanisms. In metals, melt surfaces are covered with oxides and other particles that
might have the required effects on the surface energy.
The origin for cell-wall instability is cell-wall thinning, which can be traced back
to different mechanisms. The first one is simply related to foam expansion itself. If
a cell expands the cell-wall thickness decreases due to mass conservation. During
foam expansion the principal mechanism for cell-wall thinning is by growth. The
second mechanism of cell-wall thinning is drainage due to gravity and capillary
forces. Drainage determines the foam stability in the absence of further expansion
due to gas release, that is, the time that a given foam structure is maintained with-
out destruction. Gravity induces a melt flow from the top to the bottom of the foam
and generates a density gradient [4,5]. Neglecting capillary forces, the draining ve-
locity, vD, can be estimated in a first approximation by the balance of gravity and
viscous stress from shear flow within the Plateau borders [1]
grd2edge
vD Z (10)
h
That is, the draining velocity increases with the cell-edge thickness, dedge.
Capillary forces arise due to the different curvature of the interfaces. They force
the melt from the cell membranes to the cell edges (Fig. 2.5-8). The rate of cell-wall
thinning due to this pressure gradient can be modeled using a Reynolds-type
equation for the flow between two circular parallel discs with immobile wall sur-
faces [5]
2sx 3 (11)
x_ s
3hR2 RPB
where x cell-wall thickness, 1/RPB curvature of the Plateau border, R radius

of the disc.
Integration of the last equation shows that cell walls of 100 mm in pure alu-
minum reduce to 1 mm in I1 s. This result is in contrast to experimental obser-
vations where metal foams are held in the liquid state for several minutes without
a structural change. How a metal foam attains this high stability is not yet under-
stood.
From a practical point of view foam collapse is very important. Foam collapse
usually proceeds from the outer surface to the interior and is closely related to
the appearance of a dense metal layer at the bottom (Fig. 2.5-9).
Figure 2.5-8.Cellular automaton simulation of the material transport

from regions with lower curvature to higher curvature due to capillary
forces.
Figure 2.5-9. Collapse of an alumi-

num foam (powder route precursor
AlSi10Mg0.6). Collapse proceeds from
the outer surface to the interior, the
metal of the collapsed cells drains
down to the bottom. Left) the sample
was heated to 665 hC at 1 bar ambient
pressure and further expanded at
constant temperature by pressure
reduction to 500 mbar. Right) the
sample was heated to 665 hC at 1 bar
ambient pressure and held at constant
temperature for 30 min.
Collapse happens if the amount of gas loss to the environment by diffusion or by

cell-wall rupture is substantial and not compensated by in-situ gas generation. One
has to distinguish between collapse by growth and collapse by aging, due to a long
holding time. Foam collapse by growth occurs when a critical expansion factor is
reached. In this case, cell-wall rupture on the foam surface leads to a pronounced
loss of gas to the environment. Collapse by aging results from gas loss to the en-
vironment by diffusion and cell-wall rupture due to drainage by gravity and capil-
larity.
For the production of foam parts foam rheology plays a crucial role. The rheolo-
gical bearing is important when mechanical forces are present. During foam for-
mation there are always mechanical forces acting on the foam. These might result
from a spatial restriction like a mold, from an injection molding process [26], or a
conveyer belt on which the foam is transported before solidification [11]. The me-
chanical response of foams to applied forces is very complex. Foams show wall slip,
compressibility, and non-Newtonian viscoelasticity [27]. They exhibit a nonzero
shear modulus although made out of gas and liquid, which both individually dis-
play a vanishing shear modulus. Consequently, a foam shows a very complex rheo-
logical behavior including bubble deformation, rearrangement, and avalanche pro-
cesses [28,29]. The resulting pore structure is in general not isotropic but orien-
tated in foaming direction. This behavior is even more pronounced if the foam
fills a complicated mold with dimensions of the same order as the mean pore dia-
Figure 2.5-10. AlSi10Mg0.6 foam

produced by powder compaction.
Since heating was not homoge-
nous, the internal structure repre-
sents different stages of the foam-
ing process. The expansion of the
central region leads to a compres-
sion of the exterior cells that were
formed before.
meter. As a result, pores are deformed and also destroyed during mold filling by
mechanical forces or a non-uniform foaming velocity (Fig. 2.5-10). Hence, the
internal foam structure is strongly influenced by the particular production pro-
cedure.
References
1. A. Saint-Jalmes, M. U. Vera, D. J. Durian, 9. I. Duarte, J. Banhart, ªA study of

ªUniform foam production by turbulent aluminium foam formation ± kinetics and
mixing: new results on free drainage vs. li- microstructureº Acta Met. 2000, 48,
quid contentº Eur. Phys. J. B 1999, 12, 67±73. 2349±2362.
2. S. A. Koehler, S. Hilgenfeld, H. A. Stone, 10. V. Gergely, T. W. Clyne, ªThe FORMGRIP
ªLiquid flow through aqueous foams: The process: foaming of reinforced metals by gas
node-dominated foam drainage equationº release in precursorsº Adv. Eng. Mater. 2000,
Phys. Rev. Lett. 1999, 82(21), 4232±4235. 2(4), 175±178.
3. A. M. Kraynik et al. ªFoam Micromechanicsº 11. J. T. Wood, ªProduction and Application of
in Proc. Foams and Emulsions, Cargese, Cor- Continuously Cast, Foamed Aluminumº in
sica; Kluwer, Dordrecht 1999. Proc. Fraunhofer USA Metal Foam Symp. 7±8
4. S. Hutzler, ªThe Physics of Foamsº, De- October 1997, Stanton, DW.
partment of Physics, University of Dublin 12. G. Kaptay, ªInterfacial Criteria for Ceramic
1998. Particle Stabilised Metallic Foamsº in Metal
5. A. E. Bhakta, E. Ruckenstein, ªDecay of Foams and Porous Metal Structures, J. Banhart,
standing foams: drainage, coalescence and M. F. Ashby, N. A. Fleck (eds), MIT Verlag,
collapseº Adv. Colloid Interface Sci. 1997, 70, Bremen 1999, p. 141±146.
1±124. 13. A. Arefmanesh, S. G. Advani, ªDiffusion-
6. J. Banhart, et al. ªMetal foam evolution induced growth of a gas bubble in a visco-
studied by synchrotron radioscopyº Appl. elastic fluidº Rheologica Acta 1991, 30,
Phys. Lett. 2000. 274±283.
7. L. D. Landau, E. M. Lifschitz, Hydrodynamik, 14. P. Lutze, J. Ruge, ªWasserstoff in Alumi-
3rd ed, Akademie-Verlag, Berlin 1974. nium und seinen Legierungenº METALL
8. S. F. Edwards, K. D. Pithia, ªA model for the 1990, 65, 649±652.
formation of foamsº Physica A 1995, 215, 15. H. Lange, SchaÈume und ihre StabilitaÈt,
270±276. VDI-Berichte, 1972, 182, 71±77.
16. S. W. Ip, J. Wang, J. M. Toguri, 24. C. Monnereau, M. Vignes-Adler,

ªAluminium foam stabilization by solid ªDynamics of 3D real foam coarseningº Phys.
particlesº Canad. Metall. Q. 1999, 38, 81±92. Rev. Lett. 1998, 80(23), 5228±5231.
17. T. Miyoshi et al. ªAluminum Foam, AL- 25. C. KoÈrner, R. F. Singer. ªNumerical Simu-
PORAS: The Production Process, Properties lation of Foam Formation and Evolution with
and Applicationsº in Metal Foams and Porous Modified Cellular Automataº in Metal Foams
Metal Structures, J. Banhart, M. F. Ashby, and Porous Metal Structures, J. Banhart, M. F.
N. A. Fleck (eds), MIT Verlag, Bremen 1999. Ashby, N. A. Fleck (eds), MIT Verlag, Bremen
18. P. Weigand, ªUntersuchung der Ein- 1999.
fluûfaktoren auf die pulvermetallurgische 26. F. Schorghuber, F. Simancik, E. Hartl, US
Herstellung von AluminiumschaÈumenº, patent 5 865 237, 1999.
FakultaÈt fuÈr Bergbau, HuÈttenwesen und 27. W. Hanselmann, E. Windhab, ªÛber das
Geowissenschaften, RWTH, Aachen 1999. Flieûen von Schaum in Rohren; Foam flow
19. P. Asholt, ªAluminium Foam Produced by in pipesº Appl. Rheol. 1996, Dezember,
the Melt Foaming Route Process, Properties 253±260.
and Applicationsº in Metal Foams and Porous 28. D. J. Durian, ªFoam mechanics at the
Metal Structures, J. Banhart, M. F. Ashby, bubble scaleº Phys. Rev. Lett. 1995, 75(26),
N. A. Fleck (eds), MIT Verlag, Bremen 1999, 4780±4783.
p. 133±140. 29. D. J. Durian, ªBubble-scale model of foam
20. J. S. Colton, N. P. Suh, ªNucleation of mechanics: Melting, nonlinear behaviour,
microcellular foam: theory and practiceº and avalanchesº Phys. Rev. E 1997, 55(2),
Polym. Eng. Sci. 1987, 27(7), 500±503. 1739±1751.
21. H.-Y. Kwak, Y. W. Kim, ªHomogeneous 30. V. Gergely, ªMelt Route Processing for
nucleation and macroscopic growth of gas Production of Metallic Foamsº, Department
bubble in organic solutionsº Int. J. Heat Mass of Materials Science and Metallurgy, Cam-
Transfer 1998, 41(4±5), 757±767. bridge 2000.
22. C. B. Park, L. K. Cheung, ªA study of cell 31. M. Arnold et al., ªExperimental and Nu-
nucleation in the extrusion of polypropylene merical Investigation of the Formation of
foamsº Polym. Eng. Sci. 1997, 37(1), 1±10. Metal Foamº in Proc. Materials Week 2000
23. C. KoÈrner et al., ªInfluence of processing [http://www.materialsweek.org/proceedings/
conditions on morphology of metal foams index.htm].
produced from metal powderº Mater. Sci.
Technol. 2000, July-August, 781±784.
2.6
Infiltration and the Replication Process for Producing Metal Sponges
C. San Marchi, A. Mortensen
The process of replicating a structure has been known to the metallurgist for cen-
turies. For example, casting can be thought of as a replication procedure in which a
metal is used to reproduce the negative form of the mold. In the study of porous
materials, however, replication refers to a process used to replicate the open-pore
architecture of a porous material.
44 2.6 Infiltration and the Replication Process for Producing Metal Sponges
2.6.1
Replication
More specifically, this process can be defined as a general three-step procedure for
the production of highly porous materials (Fig. 2.6-1). The three steps are:
1. preparation of a removable pattern;
2. infiltration of the pattern followed by solidification (or an alternative process,
such as curing or cross-linking, which rigidifies the infiltrate); and
3. removal of the pattern.
In some cases, a fourth step, such as pyrolysis, may be required to transform the
porous network into the desired phase. In this text, we focus on the use of the re-
plication process for the production of cellular metals with open porosity. Several
Figure 2.6-1. Diagram of the replication process.

authors have noted that metal (or metallic) sponge is perhaps the most appropriate
terminology for these materials as both the term ªfoamº and ªcellularº imply a
closed-cell architecture [1,2], while porous metals produced by replication must
have an open-pore network. We will thus refer to metals with open porosity of
any density I0.8, produced by replication, as metal sponge.
This process is distinct from deposition techniques, such as those used by Inco
Limited (Swansea SA6 5QR, Wales, UK) to produce porous nickel substrates, called
Incofoam [3]. In that process, a temporary pattern is used, but the pore space is not
replicated: instead, the pattern is coated and the pore space only partially filled.
Sintering methods that utilize a space holder are also not considered replication
for two reasons: the micro-architectural features of the pores are determined by
the metal powder size in addition to the space-holding material and infiltration
is not one of the processing steps. The so-called lattice block materials (LBM)
are prepared using polymer patterns of truss structures and investment tech-
niques; thus the lattice block is rather an engineered structure and cannot really
be considered a metal sponge, which can be ªshapedº and ªformedº while main-
taining the statistics of its properties, as long as the cell sizes are of millimeter
or more dimensions. For more details on lattice block materials contact JAMCORP
(17 Jonspin Road, Wilmington, MA 01887-1020 USA). Although we limit this dis-
cussion to metals, we note in passing that the polymeric sponge technique used to
produce porous ceramics also does not replicate the pattern and should not be con-
sidered a replication process.
In a metallurgical context, the replication technique can be used to produce
metal sponge that is difficult or impossible to produce by other processing meth-
odologies. Replicated metal sponge, for example, by necessity has an open-pore
structure, which is a fundamental difference compared to most commercially avail-
able foamed metals. The structure or architecture of metal sponge produced by re-
plication is flexible and determined by the pattern: porosity as high as 98 % [4] and
as low as 55 % [2] for different pattern materials have been reported and pore sizes
as small as 10 mm have been achieved [5]. In addition, metal sponge can be
produced from virtually any alloy that can be cast (provided a suitable pattern ma-
terial exists). The replication process has been used extensively for the production
of various porous materials including carbon [6], silicon carbide [7], aluminum al-
loys [2,4,5,8±13], magnesium alloys [14,15] and alloys based on iron and nickel
[4,16].
In the following sections we provide details about the production of metal
sponges by replication techniques. We first describe the important physical phe-
nomena that govern each of the three basic steps of the replication process (with
some emphasis on sponge produced from salt patterns for which the most infor-
mation exists in the literature). We then provide some physical and mechanical
properties of metal sponge produced in this manner.
2.6.2
The Replication Process: General Principles
2.6.2.1 Pattern Preparation

The primary requirements of the pattern material are that it be sufficiently refrac-
tory at the casting temperature and that it be chemically stable when in contact
with the melt. Additionally, after infiltration an open network must exist between
the metal ligaments such that the pattern material can be removed. The choice of
the pattern material necessarily determines the range of architectures that can be
produced. Four types of pattern materials are described below: continuous refrac-
tory (investment), discontinuous refractory (pelleted casting sands), burnable (poly-
styrene and resin), and leachable (sodium chloride).
Refractory investment casting material (continuous refractory pattern) is emerg-
ing as perhaps the most common space-holder for replication processing of low-
density metal sponges [4,12±15]. A slurry of an appropriate investment is infil-
trated into a preform that has the form of the desired metal sponge and that
can be easily removed. Commercial polyurethane sponges (generally referred to
as a foam) are typically used, as they are available with variegated properties in
terms of porosity, pore size, and strut size (thickness of the ligaments or beams
that make up the sponge) and can be easily burned out in air. Other preform ma-
terials can also be employed, provided that they can be removed without damaging
the architecture of the investment. The main restriction here is that the investment
must have a particulate size that is significantly smaller than the architecture to be
replicated; also, its relative density must be sufficiently low that the solid invest-
ment grains can be removed from the metal±refractory composite. Investment
casting is a well-developed technology that can be applied to almost any castable
metal.
Another conventional type of space holder is pelleted casting sand, or other
granulated mineral (discontinuous refractory pattern) [16]. In this method casting
sand is mixed with an organic binder to form comparatively coarse agglomerates
that can then be filled into a mold to form the pattern. The porosity of the pattern
is limited by the packing efficiency of the pellets, which are typically spherical.
Maximum random packing efficiency of mono-sized spheres is about 0.64 [17]
and can be greater for mixtures of different-sized pellets, although practical limits
are determined by the necessity to remove the pattern after infiltration with the
melt. Moreover, patterns of packed, discrete elements of this sort (as opposed to
the continuous nature of investment pattern) must be infiltrated such that these
space-holders can still be removed, that is they must not be completely embedded
in the infiltrating metal; this is discussed below. In principle, pelleted casting
sands can be used with any sand-castable alloy.
A rather unusual pattern material is polystyrene [11] (burnable pattern), al-
though any burnable space-holder can potentially be used in the same manner.
Polystyrene spheres are coated with a resin and then filled into a mold before
the resin is hardened. The resin acts to form the connected network between
the spheres; this network is necessary for removal of the pattern. The pore size
is determined by the size of the initial spheres (or particles) and the ratio of poly-
styrene to resin. This type of pattern has the advantage that it is not necessary
to physically remove solid grains from the structure after infiltration, as the pattern
is essentially burned away at the end of the process. On the other hand, this pat-
tern must be infiltrated relatively cold to avoid burning of the pattern during infil-
tration (i. e., while the metal is still liquid) and is thus limited to low-melting alloys
such as aluminum, magnesium, and zinc.
The fourth class of pattern material is leachable: these patterns are removed by
dissolution in an appropriate solvent. The process of dissolution (or leaching) is
less restrictive on size and density than insoluble refractory materials. In practice
however, patterns of leachable material are created from granules or powder, thus
the density is related to the packing efficiency of the granules, which itself depends
on the size (and size distribution) of the granules.
Sodium chloride (NaCl) has been employed most extensively as a leachable pat-
tern, because it is inexpensive and easy to handle. Sodium chloride patterns used
for the production of porous aluminum were first reported [8,9] and patented [10]
in the 1960s, and have advanced more recently [5] (there are also reports of their
use for carbon sponge [6] and for SiC sponge [7], both with a pyrolysis step).
Salt patterns are limited by their melting point; NaCl, for example, is limited to
aluminum and lower-melting alloys, while NaF could potentially be used at tem-
peratures greater than 900 hC. Additionally, highly concentrated saline solutions
are generated during dissolution and can cause significant corrosion in some al-
loys.
Salt grains also have the advantage that they can be sintered to enhance the con-
nectivity of the salt, change the structure of the pattern, and create a free standing
preform that can be handled and subsequently infiltrated [5±7]. The sintering step
is not necessary, however, for two reasons: transient bonding of the grains (perhaps
due to humidity in the salt [8]) may occur during preheating the salt prior to and
during infiltration, and incomplete penetration of the melt locally where individual
grains meet, provides the necessary open-network for salt removal.
Sintering processes depend on a number of parameters, including particle size,
temperature, atmosphere, residual stresses, and time. Such numerous process
parameters might explain some of the varying observations in the literature with
regard to the sintering of sodium chloride. Incidentally, these varying observations
point to the fact that sodium chloride may be a good choice for a pattern material
because of flexibility in controlling the sintering mechanisms and thus the micro-
architectural features of the pattern.
There is a consensus in the literature that large sodium chloride particles sinter
predominantly by an evaporation±condensation mechanism, in which the necks
between the particles grow without densification of the salt compact. Small particle
compacts, on the other hand, can densify significantly during sintering, due to the
predominance of other mechanisms where the center-to-center distances between
particles can decrease, such as bulk diffusion [18±21]. The minimum particle size
that can sinter without densification has been estimated to be about 150 mm
[19,20]. Sintering in vacuum may promote the evaporation±condensation mechan-
ism for smaller particles [22] as supported by the pressure dependence that was
noted for the rate of neck growth when non-densifying sintering mechanisms
are dominant [23].
There are advantages to both classes of sintering mechanisms with respect to
producing metal sponge. Non-densifying mechanisms, such as evaporation±con-
densation should develop a structure dictated by capillary equilibrium: these are
well known [24]. The pore architecture in such a case is expected to be similar
to the regular structures of polymeric sponge. Densifying mechanisms, on the
other hand, provide a means to achieve relative densities significantly greater
(and thus significantly lower for the sponge) than those attainable by packing
and pressing of particles. Sintered salt densities as high as 90 % have been reported
with nearly all the pores as open porosity, although for a particle size of 1 mm,
which would pose challenges in the subsequent processing steps [25].
2.6.2.2 Infiltration
Ideally, infiltration under the influence of gravity is preferred as it does not require
special equipment or procedures. In most cases, however, the metal does not wet
the pattern and some additional force must be applied to the melt to promote (or
assist) infiltration and effect a uniform distribution of metal within the pattern.
Vacuum and low-pressure assisted casting procedures are standard foundry tech-
nologies and, in many cases, these are sufficient for infiltration. Indeed, it has
been reported that spontaneous infiltration is possible when infiltrating salt grains
greater than 4 mm in size. Vibration slightly reduces the size of salt grains that can
be infiltrated under gravity, while vacuum applied to the pattern with a slight over-
pressure on the melt has also been used to infiltrate grains as small as a few hun-
dred micrometers in diameter [8].
The specific pressure required to initiate infiltration (the threshold pressure, P0)
depends to a large extent on the size of the pores to be infiltrated and the volume
fraction of these pores. A detailed physical description of infiltration is beyond the
scope of this work; more detailed information is available elsewhere [26±28]. To the
authors' knowledge, the pressure required to initiate infiltration has not been ex-
plored with specific emphasis on patterns used for replication processes; however,
expressions derived for the threshold pressure in the infiltration of packed ceramic
beds with pure aluminum should be directly relevant. Garcia-Cordovilla et al. offer
a simple semi-empirical relationship [27]
f
P0 MPa z 16 s 0:09
(1 s f )D
where f is the volume fraction solid (or in this context the fractional density of the
pattern) and D is a characteristic diameter of the solid in micrometers. This rela-
tionship is based on data for particles in the range 10±100 mm packed to a density
of 50±60 % and infiltrated with pure aluminum, but it should provide reasonable
magnitude estimates of the pressures required for infiltration over a broader range
of conditions since it contains the functional dependencies on f and D expected

from theory [26±30]. This relation predicts that particles with an average diameter
of 100 mm and packed to 75 vol.-% will require an applied pressure of a couple of
atmospheres to initiate infiltration, while D 4 mm and f 0.98 requires the ap-
plication of about one atmosphere of pressure to the melt, such that infiltration can
be achieved by evacuating the pattern and taking advantage of atmospheric pres-
sure.
The degree to which the metal penetrates the pattern, particularly at the local
level, is also an important consideration and obviously depends on the applied
pressure. When the pattern material is porous, a very high applied pressure will
cause the space between individual grains of investment to be infiltrated, such
that these then cannot be removed. In a system with a wetting angle greater
than ninety degrees, a low (yet finite) pressure will cause infiltration of larger
pores within the pattern, yet cannot force the metal into small channels against
adverse capillary forces. Thus, in patterns created from granulated materials
such as pelleted sand, the regions near contacts between individual grains may re-
main uninfiltrated if the pressure is low enough. These open channels between ad-
joining discontinuous grains are necessary for the removal of the pattern. Control
of the infiltration then is important to effect a uniform distribution of metal in the
large channels of the pattern without penetrating the microstructural elements of
the pattern when it is porous.
Few details have been given in the literature about infiltration in the context of re-
plication processing, although generally the techniques appear to be some form of
vacuum or low-pressure assisted casting methods [4,12±16]. A modified die-casting
method has also been used [11] and any other high-pressure casting methods such
as squeeze casting should be appropriate. For example, gas-pressure infiltration at
pressures up to 80 bar has been used to infiltrate patterns with small pores [5].
After infiltration is complete, the metal must, of course, solidify, which is an in-
tegral part of any casting process. Methods utilized to control solidification shrink-
age depend on the specifics of the casting process, but generally consist of direc-
tional solidification in the case of gas-pressure infiltration or some type of riser
or reservoir in general foundry methods.
2.6.2.3 Pattern Removal

The pattern removal step depends on the characteristics of the pattern. Patterns
from fine-grained ceramic powders such as investment casting compound and ag-
glomerated silica pellets are typically removed by spraying water on the pattern±
metal composite [12±16], such that the water physically breaks down the pattern
into the small grains from which it is made. This requires penetration of the po-
rous network by the spray, or at least that the fluid can ªwickº into the grain struc-
ture of the pattern so that it breaks down. For relatively fine (I1 mm) and dense
sponges i10 %), however, it is unclear how efficient this process would be for sec-
tions many times the pore size. Additional methods such as (ultrasonic) vibration
may also prove useful for removal of solid grains.
Polymer pattern materials can be removed by burning in air as is the case for
polystyrene-resin patterns [11]. The cleanliness of this burning is an important
issue in an industrial setting. These patterns, however, have the advantage that
complex parts can be machined before removal of the pattern and without signifi-
cant damage to the structure of the sponge.
In some cases, leachable patterns can also be machined to complex geometries
prior to leaching and without damaging the architecture of the sponge. In the case
of salt, removal of the pattern is by immersion in water, causing dissolution of the
salt. Dissolution is primarily a diffusive process, as the dissolved salt ions diffuse
from their place-holding position into the water bath through the narrow channels
in the sponge where the water is essentially stagnant. Thus, the rate and total time
of dissolution depend strongly on the size of the specimen. Salt is more difficult to
remove from small pores, particularly if gas is entrapped in the pores; this is the
so-called ªgas lockº phenomenon, which can block the penetration of fluid into the
porous network [8].
Corrosion of metal sponge during leaching of the pattern can be a problem due
to the large surface area of the porous network, thus the salt concentration and the
immersion time must be minimized. This is accomplished by simply changing the
water periodically, however, it is important that the sponge is not drained at this
point, as this will cause gas to be entrapped in pores and promote ªgas lock.º Ad-
ditionally, impurities in the water should be controlled to avoid precipitation of
other salts in these concentrated solutions. Aluminum alloys (the metals for
which salt patterns are most applicable) tend to be relatively corrosion-resistant
in salt solutions. The aluminum±copper alloys are among the least corrosion-resis-
tant in salt solutions, but sponge can still be prepared from these alloys without
degradation by corrosion. Electrochemical protection could also offer additional
protection from corrosion.
A large volume of water is not necessary as salt-saturated water is denser than
distilled water, and hence falls to the bottom of the container where it remains
in the absence of convection. As a result, during the dissolution process a large
concentration gradient is maintained between the outer surface of the salt±metal
composite suspended near the surface of the dissolution bath (but completely im-
mersed) and the bottom of the container. For large pores and low-density sponge,
mixing may modestly enhance dissolution if the velocity of the fluid in the porous
network is affected. In most cases, however, this condition is not met and mixing
the water primarily distributes the salt concentration and reduces the diffusion gra-
dient in the water bath, in turn slowing the leaching process.
The final stage of salt removal from the sponge is rinsing. This can be accom-
plished by flowing water through the structure, or simply by changing the water
several times to incrementally remove the saturated fluid; this process is greatly
facilitated by the higher density of salt solutions. Pressurized gas is also effective
for removing the residual fluid in the sponge such that dissolved salts are not
left in the structure when this water evaporates.
Other methods may contribute to more efficient (faster) removal of salt, but
these must enhance the rate of diffusion of the salt in the fluid. For example, ul-
trasonic vibration has been proposed as a useful method as it should increase the
rate of diffusion with the added benefit of freeing gas bubbles entrapped in the po-
rous network [8].
2.6.3
Physical and Mechanical Properties of Metal Sponge
Published reports of metal sponge produced by replication techniques are summar-

ized in Table 2.6-1. Relevant details of the processes and properties of the resulting
metal sponge are briefly described below in the context of the basic pattern tech-
nologies: continuous refractory, discontinuous refractory, burnable, and leachable.
2.6.3.1 Continuous Refractory Patterns

A variety of metal sponges have been produced from commercial polyurethane pre-
forms: aluminum alloys [4,12,13], magnesium alloys [14,15] and several alloys hav-
ing melting points up to 1500 hC [4]. Infiltration has been conducted with vacuum
or pressure-assisted investment casting techniques. The investment is removed by
means of a water spray, although additional specifics have not been supplied in the
literature. Metal sponge with volumes up to 1500 cm3 have been prepared in this
way; in addition, components, such as modules for heat exchangers integrally cast
with plates and tubes, have been demonstrated [4].
The porous metals produced with this methodology have porosity in the range
92±98 % and pores sizes as small as 0.85 mm, Table 2.6-1. Practical limits of repli-
cation with investment patterns, however, have yet to be clearly identified and may
include a wider range of structural parameters than reported to date. Although size
effects (including density) have not been systematically explored, compressive
strength (plateau stress) of aluminum and magnesium sponge with about 97 %
porosity have been reported to be less than 0.25 MPa and in accordance with pre-
diction based on phenomenological relationships [12±15]. Strain rate sensitivity
has also been reported for these materials although a complete interpretation of
these data has not yet been offered [31,32].
2.6.3.2 Discontinuous Refractory Patterns

Cast-iron sponge has been produced using pelleted silica sand [16]. The size of the
pellets can be varied from about 1 mm to over 5 mm. The porosity of the sponges is
reported to be about 60 %, Table 2.6-1. It appears that gravity casting methods were
used, perhaps assisted with vacuum. The pattern is removed with pressurized
water, but it is unclear if the pellets can be removed from a material several centi-
meters in depth. Preliminary mechanical data in compression have been reported
and feature a significant drop in load carrying capacity after about 2 % strain. Per-
meability tests have also been reported though no systematic conclusions have been
drawn. Incidentally, a similar sponge-like material (from aluminum) was produced
as early as the 1930s in Germany, except that the pattern was not removed [33].
52
Table 2.6-1. Summary of replication techniques used for the production of metal sponge.
Pattern material Metal Infiltration technique Pattern Metal sponge characteristics* Reference
removal
Pore size (mm) Porosity (%)
Investment of poly- Mg vacuum assisted water spray 4.5 97 [14, 15]

urethane sponge (AZ91)
Investment of poly- Al vacuum assisted water spray 4.5 93±96 [12, 13]
urethane sponge (SG91A)
Investment of poly- Al (Cu, Zn, Fe, vacuum or pressure (not reported) 0.85±2.5 92±98 [4]
urethane sponge Ni, Co) assisted
Pelleted casting sand pig Fe vacuum assisted pressurized water 1±5 60±63 [15]
Polystyrene and resin Al modified die casting burning 1.2±3.7 73±86 [11]
NaCl Al gravity, vibration assisted, leaching 0.15±5 (60±63) full range [8]
pressure assisted by not reported
vacuum
2.6 Infiltration and the Replication Process for Producing Metal Sponges
Al 0.45±1.45 55±67 [2]
NaCl (presintered) Al alloys gas-pressure infiltration leaching 0.01±3 65±81 [5]
* Represents range of properties as reported.

2.6.3.3 Burnable Patterns

Aluminum sponge has been prepared from resin-bonded polystyrene spheres
(Table 2.6-1) with porosity in the range 73±86 % and pore sizes in the range
1±4 mm [11]. Infiltration was conducted with a modified die-casting technique
and the pattern was removed by burning in air. Properties of these metal sponges
have not been reported.
2.6.3.4 Leachable Patterns

Aluminum sponge was produced from NaCl as early as 1961 [8]. The patterns were
prepared by packing salt grains in a mold and infiltrating by one of several tech-
niques depending on the size of the salt grains: by gravity (i 3.4 mm), on a vibrat-
ing table (i 1.7 mm), or with an overpressure and simultaneous application of
pressure (j 0.15 mm). The compressive strength of an A356 aluminum alloy
sponge was observed to depend on the heat treatment and at 10 % strain varied
in the range 20±30 MPa for porosity in the range 60±63 %. The porosity undoubt-
edly varied depending on the size (and distribution of sizes) of the salt grains. In
an independent study, Banhart reports a range in porosity between 55 and 67 %;
however, he does not provide any details about the processing other than that
the pattern was leachable [2]. Some basic property indicators (mechanical response
to compression, pressure drop and sound adsorption) are given in this latter report.
Detailed analysis of properties is not given in either reference.
Recently, gas-pressure infiltration techniques have been used to infiltrate
sintered NaCl with grain sizes as small as 10 mm [5]. Porosity was in the range
68±81 % depending, in part, on the size of the salt grains. A SEM micrograph
of an aluminum sponge prepared by replication of sintered NaCl is shown in
Fig. 2.6-2. The mechanical response of a metal sponge depends on the microstruc-
ture of the metal (Fig. 2.6-3), as with any fully dense alloy; however, the properties
of the metal sponge are further influenced by the level of porosity. The shape
of the flow curves in Fig. 2.6-3 are somewhat different from what is generally
observed with commercially available closed-cell metal foams. This is due to
the inherently higher ductility of most of the base metals used to produce the
sponges represented in Fig. 2.6-3 in comparison to commercial cellular metals,
as well as the higher regularity of the architecture of these metal sponges, and
the comparatively low porosity. The stiffness was found to be about half that
expected from simple phenomenological relationships [5].
The processing limits for replication from leachable patterns, like the other pat-
tern technologies, have not been clearly evaluated and there is some divergence in
the literature. The range of available pore sizes and porosity, however, is clearly
considerable.
Figure 2.6-2. Scanning electron micrograph of aluminum sponge produced by replication from
sintered NaCl patterns.
Figure 2.6-3. Compressive response of several aluminum-alloy sponges produced by replication

from sintered NaCl patterns. The salt has grains approximately 0.5 mm in diameter and the
fraction solid is noted as f.
2.6.4
Conclusions
Replication processing provides a versatile and attractive method for the produc-
tion of metal sponge (often called open-celled metal foam). Advantages of this pro-
cess comprise considerable control of the pore size, density, and internal architec-
ture of the sponge, the capacity for net or near-net shape component production,
and the potential for production of components that are part dense, part porous.
Also, the open-porosity that is intrinsic to metal sponge can be put to advantage
in several applications that require fluid flow through the porous metal, such as
filters and heat-exchangers.
The replication process exists in a few variants, which differ mostly according to
the nature of the pattern: continuous refractory, discontinuous refractory, burnable,
and leachable patterns are four classes of patterns that have been employed to date.
The range of density and pore size that can be achieved in replicated metal sponges
depends largely on the pattern material and also on the infiltration process. With
leachable patterns, in particular, these ranges have been shown to be comparatively
wide; at present these also seem to be complementary to what can be achieved with
the methods utilizing insoluble patterns. There is, though, significant room for in-
novation in replication technology as related to the production of metal sponge and
there exists perhaps even opportunities for cross-pollination between the methods
described here.
References
1. D. Weaire, M. A. Fortes, Adv. Phys. 1994, 43, 8. L. Polonsky, S. Lipson, H. Markus, Modern
685. Castings 1961, 39, 57.
2. J. Banhart, Adv. Eng. Mater. 2000, 2, 188. 9. H. Seliger, U. Deuther, Freiburger For-
3. J. Babjak, V. A. Ettel, V. Paserin, US Patent schungshefte 1965, 103, 129.
4 957 543, 1990. 10. H. A. Kuchek, US Patent 3 236 706, 1966.
4. I. Wagner, C. Hintz, P. R. Sahm, in Metal 11. L. Ma, Z. Song, D. He, Scripta Mater. 1999,
Matrix Composites and Metallic Foams, Proc. 41, 785.
Euromat 99, Munich, Germany, T. W. Clyne, F. 12. Y. Yamada, K. Shimojima, Y. Sakaguchi, M.
Simancik (eds), DGM/Wiley-VCH, Weinheim Mabuchi, M. Nakamura, T. Asahina, T.
2000, p. 40. Mukai, H. Kanahashi, K. Higashi, Mater. Sci.
5. C. San Marchi, J.-F. Despois, A. Mortensen, Eng. 1999, A272, 455.
in Metal Matrix Composites and Metallic 13. Y. Yamada, K. Shimojima, Y. Sakaguchi, M.
Foams, Proc. Euromat 99, Munich, Germany, Mabuchi, M. Nakamura, T. Asahina, T.
T. W. Clyne, F. Simancik (eds), DGM/Wiley- Mukai, H. Kanahashi, K. Higashi, Mater. Sci.
VCH, Weinheim 2000, p. 34. Eng. 2000, A280, 225.
6. R. W. Pekala, R. W. Hopper, J. Mater. Sci. 14. Y. Yamada, K. Shimojima, Y. Sakaguchi, M.
1987, 22, 1840. Mabuchi, M. Nakamura, T. Asahina, T.
7. T. J. Fitzgerald, V. J. Michaud, A. Mortensen, Mukai, H. Kanahashi, K. Higashi, J. Mater.
J. Mater. Sci. 1995, 30, 1037. Sci. Lett. 1999, 18, 1477.
56 2.7 Solid-State and Deposition Methods
15. Y. Yamada, K. Shimojima, Y. Sakaguchi, M. 24. P. J. Wray, Acta Metall. 1976, 24, 125.
Mabuchi, M. Nakamura, T. Asahina, T. 25. R. W. Davidge, Proc. Br. Ceram. Soc. 1969,
Mukai, H. Kanahashi, K. Higashi, Adv. Eng. 12, 245.
Mater. 2000, 2, 184. 26. A. Mortensen, I. Jin, Int. Mater. Rev. 1992,
16. U. Mohr, W. Bleck, in Metal Matrix Com- 37, 101.
posites and Metallic Foams, Proc. Euromat 99, 27. C. Garcia-Cordovilla, E. Louis, J. Narciso,
Munich, Germany, T. W. Clyne, F. Simancik Acta Mater. 1999, 47, 4461.
(eds), Wiley-VCH, Weinheim 2000, p. 28. 28. A. Mortensen, in Metal Matrix Composites,
17. R. M. German, Particle Packing Character- A. Kelly, C. Zweben (eds), Vol. 3 of Compre-
istics, Metal Powder Industries Federation, hensive Composite Materials, T. W. Clyne (ed.),
Princeton, NJ 1989. Pergamon, Oxford, UK 2000, Ch. 21.
18. C. S. Morgan, L. L. Hall, C. S. Yust, J. Am. 29. A. Mortensen, J. A. Cornie, Metall. Trans.
Ceram. Soc. 1963, 46, 559. 1987, 18A, 1160.
19. A. A. Ammar, D. W. Budworth, Proc. Br. 30. A. Mortensen, T. Wong, Metall. Trans. 1990,
Ceram. Soc. 1965, 3, 185. 21A, 2257.
20. W. J. Tomlinson, G. Astle, J. Mater. Sci. 31. T. Mukai, H. Kanahashi, Y. Yamada, K.
1976, 11, 2162. Shimojima, M. Mabuchi, T. G. Nieh, K.
21. R. J. Thompson, Z. A. Munir, J. Am. Ceram. Higashi, Scripta Mater. 1999, 41, 365.
Soc. 1982, 65, 312. 32. H. Kanahashi, T. Mukai, Y. Yamada, K.
22. W. D. Kingery, M. Berg, J. Appl. Phys. 1955, Shimojima, M. Mabuchi, T. G. Nieh, K.
26, 1205. Higashi, Mater. Sci. Eng. 2000, A280, 349.
23. J. B. Moser, D. H. Whitmore, J. Appl. Phys. 33. W. Thiele, Metal Mater. 1972, 6, 349.
1960, 31, 488.
2.7
Solid-State and Deposition Methods
O. Andersen and G. Stephani
In this chapter, solid-state and deposition methods for the manufacturing of cel-
lular metals are reviewed with regard to their most important processing features.
Discussion will be limited to methods producing predominantly 3D structures.
Typical values of the cell size, cell wall thickness, and obtainable relative densities
are cited wherever available. Earlier overviews including both methods of manufac-
turing are available [1,2].
The term ªsolid-state methodº means that directly before the formation of the
cellular structure, the metal is in the solid state. However, in some cases limited
amounts of liquid phase appear in the process: during liquid-phase sintering or
brazing of single cells. Except for foaming of solids, all solid-state methods require
a sintering step before the final cellular metallic structure is obtained. Naturally,
powder metallurgical methods play a dominant role in this area.
Porosity in cellular structures made from powders often exhibits two different
size scales: a macroporosity of the order of the targeted cell size, and a micropo-
rosity in the order of the primary particle or, in case of reactive sintering, the
grain size.
For all powder metallurgical processes, the cost of the powders is a major con-
cern. Some of the methods, especially those using slurries of metal powders dis-
persed in a dissolved binder, are capable of using cheap oxides and thus have
the potential for a significant cost reduction.
Deposition methods need to mobilize the metal in order to transport it onto a
porous substrate. This is normally done by transforming the metal into the ionic
or vapor state. From the method of manufacture, it can be deduced that the deposi-
tion methods are only capable of generating an open-cell morphology. Usually, they
also imply a heat treatment for the removal of the precursor on which the metal is
deposited.
A practical differentiation between the methods in question can be made with the
help of two manufacturing parameters, as is shown in Fig. 2.7-1. The chosen criteria,
ªformation of cellular structureº and ªformation of cellsº have direct implications for
the processing itself, as well as for the resulting properties of the structure.
According to this classification, the formation of the cellular structure can be ac-
complished by building it up from single cells, which do not necessarily have to be
in the final metallic state. Alternatively, the structure can be built up from bulk.
The formation of the cells themselves can either be carried out with the help of
a lost core, or in a coreless manner. Within this framework the deposition meth-
ods, for example, constitute a small subset of the lost core bulk processes.
Solid state
and deposition methods
from single cell from bulk
coreless methods
structures from hollow sintered metal powders

powders and fibers
structures from hollow special sintering effects
spheres made by coaxial foaming of solids
spraying of slurries foaming of slurries
lost core methods
cementation process p/m space holder methods
galvanic hollow spheres deposition methods

Figure 2.7-1. Classification of solid-state IFAM hollow spheres
and deposition methods from a manufac-
turing point of view.
Generally, formation from single cells has the potential of a very close control of
the cellular properties. However, in most cases the single cell processes require
more processing steps than bulk formation processes.
Processes using lost cores usually are prone to introducing unwanted interstitials
like carbon and oxygen into the base material, since the cores are predominantly hy-
drocarbons, which have to be removed thermally. Another disadvantage compared to
coreless processes is the additional time required for the removal of the core.
Generally speaking, there is a remarkable variety of solid-state and deposition
methods for making cellular metals, which all have their specific strengths and
weaknesses. It therefore depends very much on the final application, which
method may be best suited in a given situation.
2.7.1
Formation from Single Cells: Coreless Methods
2.7.1.1 Hollow-Sphere Structures made from Gas Atomized Hollow Powders

It is known that gas atomization often generates a small amount (typically 1±5 %)
of hollow particles owing to gas entrapment in the liquid ligaments. By appropriate
separation methods, these hollow particles can be separated from the solid ones
a) Surface 1.0 mm
Figure 2.7-2. HIP-bonded hollow-

sphere structure made from alloy 625
(courtesy IPML, University of Virginia,
b) Cross section 1.0 mm Charlottesville).
and consolidated to yield hollow-sphere structures. Typical sphere diameters are

500±1000 mm with wall thicknesses about 100±300 mm [3].
Different methods for consolidation are reported: sintering, transient liquid
phase sintering with the help of a powdered additive, and hot isostatic pressing
(HIP). Best results with regard to specific stiffness were achieved with HIP bond-
ing, Fig. 2.7-2. This was attributed to the predeformation of the cells and enlarging
of the contact zones.
Structures were made from TiAlV 64 as well as from a nickel-based superalloy
(alloy 625). Relative densities of 0.3±0.12 were achieved.
Other methods of producing hollow metal powders without using lost cores have
been reported, such as coaxial atomization of melts [4], atomization of melts con-
taining a dissolved gas [5], and plasma treatment of powder particles containing an
absorbed gas [6]. However, nothing is known about cellular structures made
thereof.
2.7.1.2 Hollow-Sphere Structures made from Coaxially Sprayed Slurries

Hollow spheres can be produced by spraying a powder slurry containing a solvent
and a polymer binder through the outer orifice of a coaxial nozzle [7]. By hydrody-
namic interaction with the gas passing through the inner orifice, single bubbles are
formed. In flight, they are spherodized by surface tension, and dried. The resulting
hollow spheres can have diameters in the range 1±6 mm with close control (e4 %)
of the dimensions. Wall thickness is typically around 100 mm and can be varied be-
tween 40 and 200 mm [8].
Subsequently, the powder shells are heat treated to remove the organic binder
and, at a higher temperature, to sinter the metal powder particles to form a closed
metallic shell.
In order to obtain cellular structures, the single hollow spheres are filled in
molds and either diffusion bonded or sinter bonded with the help of a metal pow-
der slurry. The use of slurries leads to an undesirable reduction in overall void
space of the structure. Nonetheless, values of the relative density of 0.12 were
Figure 2.7-3. FeCr hollow spheres

bonded together with magnetite slurry
and reduced (courtesy Georgia Institute
of Technology).
achieved with stainless steel structures made from reduced iron and chromium
oxide powders [9], Fig. 2.7-3.
2.7.2
Formation from Single Cells: Lost Core Methods
2.7.2.1 Hollow-Sphere Structures made by Cementation and Sintering

If spherical iron particles are immersed in a stirred CuSO4 solution, the so-called
cementation reaction
Cu2 Fe () Cu Fe2
leads to deposition of pure copper on the particles while simultaneously the iron
core is being dissolved. The process can be carried out until the iron is completely
dissolved in the electrolyte and only hollow copper shells remain [10].
Hollow copper powders were prepared using reduced iron oxide powders as the
lost core. The resulting copper particles had diameters of about 500±750 mm. The
walls are very rough and porous, having a thickness of a few hundred microns. The
hollow copper particles need further heat treatment to give them sufficient
strength for handling. Cellular metallic structures are achieved by packing the hol-
low particles in ceramic molds and bonding them by sintering.
Good results were achieved by coating the copper particles with tin, thus en-
abling liquid-phase sintering. The resulting cellular bronze structures exhibit
good interparticle bonding, dense cell walls, and relative densities of about 0.2
[11], Fig. 2.7-4.
Figure 2.7-4. Cross section of a bronze hollow-sphere structure made by cementation and sin-
tering (courtesy Institute for Chemical Technology of Inorganic Materials, Technical University of
Vienna).
Figure 2.7-5. Brazed nickel hollow-sphere structures with sphere diameter of 4.5 mm made from
galvanically manufactured hollow spheres (courtesy ATECA, Montauban, France).
2.7.2.2 Hollow-Sphere Structures made from Galvanically Coated Styrofoam Spheres

Hollow spheres can be made by galvanic coating of Styrofoam spheres, which act
as a lost core [12]. The Styrofoam is thermally removed. Sphere diameters may
range from 2 to 10 mm. The sphere walls show excellent uniformity in thickness,
which is typically in the few micron range. Thick walls are very expensive to man-
ufacture. Material choice is limited to a few metals suitable for galvanic deposition
(typically copper and nickel).
For making purely metallic structures, the single spheres can be joined by braz-
ing [13] or diffusion bonding. Sphere sizes are in the range of a few millimeters
and relative densities can be smaller than 0.05 since the galvanic process is able
to produce very thin and uniform coatings [14]. With suitable sorting methods
for the Styrofoam cores, very uniform spheres can be obtained, which allow for
the manufacturing of ordered structures, Fig. 2.7-5.
2.7.2.3 Hollow-Sphere Structures made from Fluidized Bed Coated Styrofoam

Spheres
According to previously published work [15], commercially available Styrofoam
spheres can be coated in a fluidized bed process with a suspension containing
metal powder and a binder. The resulting green spheres may be debindered and
sintered to obtain single metallic hollow spheres.
Alternatively, the green spheres can be subjected to a suitable forming process to
obtain green parts made up of single cells [16], Fig. 2.7-6. The resulting green parts
Figure 2.7-6. Process scheme for making

hollow-sphere structures (courtesy IFAM,
Dept. of Powder Metallurgy and Composite
Materials, Dresden).
are heat treated in the same way as the single spheres. In principle, this allows for
the production of hollow-sphere structures from arbitrary metals and alloys.
In order to enhance the mechanical properties of the resulting structures, large
contact areas between the spheres are achieved by applying a slight pressure to the
green spheres during the forming step. In this stage, the elastic Styrofoam core
helps to prevent buckling and leads to a more uniform distribution of the cell
pre-deformation. A typical example of the resulting structure is shown in
Fig. 2.7-7.
The attainable range of the sphere diameters is 0.5±10 mm. Wall thicknesses
have been made from 10 to 1000 mm. Relative densities can be as low as 0.03. Re-
cent overviews of the processing and properties of such structures are given else-
where [17,18].
Materials currently under investigation are numerous, with a focus on stainless
steel 316L [19] and low-cost hollow-sphere structures made from iron oxides [20].
Figure 2.7-7. IFAM 316L hollow-sphere

structure. Left hand side: computer tomo-
graphy images of cross sections (courtesy
IFAM, Department of Powder Metallurgy
and Composite Materials, Dresden).
2.7.3
Bulk Formation: Coreless Methods
2.7.3.1 Sintered Metal Powders and Fibers

Porous metals made from sintered particulates are already in widespread use for
predominantly functional applications. A good overview of classical commercial
processes and products relying on plain sintered metal powders is available [21].
The powders are sifted to narrow size fractions, die compacted, and sintered.
Bronze, stainless steels, titanium, and nickel-based alloys are common in commer-
cial production.
Depending on the particle morphology and packing density, the relative density
of the resulting parts can reach values as low as 0.4, while pore sizes are from 0.5
to 200 mm. However, usually the relative density is quite high (0.8±0.6).
Much lower relative densities can easily be achieved by using particles with large
length-to-diameter ratios: metal fibers. Depending on the length-to-diameter ratio
and the processing route, relative densities of 0.01 can be obtained.
Very fine fibers down to diameters of a few microns can be produced by the bun-
dle drawing technique [22]. Fibers are available in many corrosion and heat resis-
tant alloys: 316 and 302-type, Inconel 601, FeCrAlloy, Hastalloy X, Nichrome, and
some titanium alloys. Diameters of bundle drawn fibers can be as small as a few
microns.
Using broken fiber bundles or short fibers, highly porous structures can be man-
ufactured by applying textile or other dressing techniques with subsequent sinter-
ing to get a rigid structure. Depending on the application, the mean pore size of
such structures can be made as small as a few microns. The relative density typi-
cally is in the range 0.35±0.1, but can also be much smaller.
There are other mechanical technologies for making fibers, such as shaving
from either wire, foils, or bars. These fibers have irregular cross sections, either
triangular or rectangular with a certain surface roughness and equivalent dia-
meters of 25±100 mm. These shaving methods have been recently improved by in-
troduction of superimposed vibrations, which results in a more homogeneous
cross section [23].
Figure 2.7-8. Cross-sectional cut of sintered fiber structure made from melt-extracted short
fibers from FeCrAl 23.15. From left to right: relative density 0.3, 0.2, 0.1 (courtesy IFAM,
Department of Powder Metallurgy and Composite Materials, Dresden).
Using the melt extraction process for making short fibers directly from the melt,
the materials limitations of the mechanical fiber manufacturing routes can be over-
come [24,25]. Highly porous sintered structures from intermetallics such as Ni3Al
[26] and FeAl, as well as from many other ferrous and nonferrous alloys have been
made. Equivalent fiber diameters are 50±150 mm, fiber length may vary between
3 and 25 mm. This allows for relative densities between 0.5 and 0.03 in the final
structure, Fig. 2.7-8.
2.7.3.2 Methods Utilizing Special Sintering Phenomena

In common powder metallurgical production, the formation of pores is an un-
wanted by-product of the sintering process. Many origins of pore formation have
been explained theoretically and attributed to a variety of different physical phe-
nomena. Some of these effects are capable of generating a very large amount of
porosity and thus lend themselves to making cellular metals.
An early example is the so-called Kirkendall effect, in which in a diffusion couple
pores are generated in the component with higher diffusivity [27].
Aldinger [28] has evaluated the potential of an extreme Kirkendall effect for mak-
ing porous structures. Cellular structures were obtained by using coated particles
where the core consists of the component with higher diffusivity. The powders
are then pressed and sintered. Experiments with nickel and cobalt coated beryllium
powders showed a maximum volume increase of 262 %, corresponding to an abso-
lute density of 0.47 g/cm3 and a relative density in the order of 0.2.
The structures exhibit a mixture of closed-cell porosity with diameters of 50 to
100 mm and wall thicknesses around 10 mm, depending on the size of the starting
powders and the thickness of the coatings. Theoretical considerations lead to the
conclusion that under ideal conditions, relative densities can be made arbitrarily
small. In reality, this is hindered by a number of different effects, such as incom-
plete or non-uniform coating of the particles.
In reactive sintering, elemental powders are used instead of pre-alloyed powders
in order to enhance sinterability and reduce cost. Especially in intermetallic sys-
tems, an unwanted swelling of the sintered parts is noticed under certain process-
ing conditions. A general treatment of the different mechanisms for swelling in
the system Ti-Al was given elsewhere [29].
KnuÈwer exploited the swelling effect for manufacturing FeAl structures with
open porosity [30]. Relative densities down to 0.5 were reported.
In contrast to the cell morphology obtained via the Kirkendall effect, these struc-
tures exhibit an almost exclusively open porosity. Due to the different effects taking
place during swelling, a macro- and a microporosity are generated, both being of
an open type (Fig. 2.7-9). Macroscopic porosity is in the order of the primary par-
ticle sizes, typically in the range 50±150 mm. The size range of the microscopic po-
rosity is determined by the grain size of the intermetallic phase generated, and is
typically smaller than the macroscopic porosity by one or two orders of magnitude.
Figure 2.7-9.Porous TiAl3 structure prepared from elemental powder mixture. Large pores
correspond to primary Al particles (courtesy IFAM, Department of Powder Metallurgy and
Composites).
2.7.3.3 Foaming of Solids

Kearns et al. first described a solid-state foaming process where porosity in metals
is generated by expanding a hot-isostatically pressed mixture of argon gas and
metal powder at elevated temperatures and ambient pressure [31].
More recent work utilizes this principle for making TiAlV 64 sandwiches with
a porous core [32]. This is achieved by canning TiAlV 64 powder in canisters of
the same material. After HIP, the resulting billet can be isothermally forged or
hot rolled to achieve sheets. Subsequently, an expansion annealing treatment
(typically at temperatures i0.6Tm for 4±48 h) generates the porous core,
Fig. 2.7-10.
This method, which is known as the low-density core (LDC) process, uses creep
expansion due to the internal gas pore pressure to generate the desired porosity.
Queheillalt et al. have carried out a thorough analysis of the process kinetics
and morphological evolution of the pores [33]. The resulting porosity in the core
region is largely unconnected and shows pore diameters of 10±100 mm. The rela-
tive density lies in the range 0.65±0.5. One advantage is the capability of producing
very large sheets (up to 2000 mm q 1200 mm q 4 mm).
A slightly different approach described previously utilizes the transformation
superplasticity found in allotropic materials [34]. Transformation superplasticity oc-
curs at all grain sizes by biasing with a deviatoric stress (from the entrapped gas) of
internal stresses (from the allotropic mismatch during thermal cycling about the
allotropic temperature range). Hence, the rather lengthy expansion annealing
can be replaced by a significantly shorter thermal cycling treatment.
Low Density Core Process Steps

Ti-6Al-4V can
a) Powder / Can
Preparation
Powder
Ti-6Al-4V packing
powder density D0
Evacuate and backfill with Argon gas to pressure p 0
b) HIP Consolidation
(1040oC, 100 - 200MPa, 2h)
Isolated
pressurized
voids
Final relative density

0.85 - 0.95 p(t), T(t)
c) Hot Rolling
(approx. 850oC, 6-20 passes)
Thinning of
facesheet
Internal gas
Changes in
pressure
pore shape,
matrix
microstructure
d) Expansion Heat Treatment

(900oC - 1200˚C, 4 - 48 h)
Vacuum
furnace
Isolated
porosity
Sandwich
(<40%)
panel
Facesheet
Figure 2.7-10. LDC process scheme (courtesy IPML, University of Virginia, Charlottesville).
2.7.3.4 Foaming of Slurries

There are only few reports on the foaming of slurries, and work has not been very
successful up to date owing to low strengths of the resulting structures. According
to Banhart et al., metal powder slurries containing a foaming agent can be poured
into molds and foaming occurs due to decomposition of the foaming agent [2]. Sta-
bilizing the resulting liquid foam is essential and in some cases is accomplished by
adding a stabilizer capable of polymerization on contact [1]. After foaming, the or-
ganic contents have to be removed thermally, followed by a sintering step. Relative
densities as low as 0.05 have been reported [35].
2.7.4
Bulk Formation: Lost Core Methods
2.7.4.1 Powder Metallurgical Space Holder Method

The use of space holders in powder metallurgy is a well-known technique to create
different pore sizes and achieve relative densities lower than 0.4. The space holder
is mixed with the metal powder and, depending on the final shape of the product,
several methods are used for forming of the powder mixture: die compaction, ex-
trusion, and rolling.
A general difficulty of these methods is the need to remove large quantities of
organic or inorganic space holder material from the parts. Good results can be
achieved with carbamide (urea), which can be thermally removed below tem-
peratures of 200 hC. Porous structures from stainless steel 316L, nickel based
superalloys, and titanium (Fig. 2.7-11) have been made this way (see Section 2.3)
[36,37].
Pore sizes are typically in the range of a few hundred microns up to a few
millimeters, and relative densities of 0.4±0.2 are reported.
Figure 2.7-11. Titanium sample with relative density of 0.3 made by space holder method
(courtesy Research Centre Ju
È lich, Institute for Materials and Processes in Energy Systems).
2.7.4.2 Deposition Methods

Deposition techniques always need a parent structure on which to deposit the
metal. Most common is the use of polymeric (polyurethane) foams. In order to en-
sure the penetration of the structure by the mobilized metal, a thermophysical
shock treatment (reticulation) for the removal of residual membranes is needed.
After deposition, the parent structure is removed thermally or chemically.
If the metal is deposited from the ionic state (either galvanically or electroless), a
pretreatment is necessary that assures either electrical conductivity or, in the case
of electroless plating, a sufficiently rough surface of the parent structure.
Materials choice is limited to nickel, nickel±chromium, silver, copper, and tin.
Relative densities are typically less than 0.1 with average cell sizes from 400 mm
up to a few millimeters.
In order to overcome the limitations in materials choice, it is also possible to
deposit the metal from the vapor state. Albeit being a nickel foam, INCOFOAM
made by Inco Ltd., Canada, is an example of a commercially available product
made by chemical vapor deposition. A recent development in this area is the elec-
tron beam directed vapor deposition process (EB-DVD), which was used to evapo-
rate and deposit an Inconel alloy 625 onto an open celled, reticulated structure
[38], Fig. 2.7-12. The EB-DVD technology uses a combination of an electron
beam gun and an inert gas carrier jet of controlled composition to create thick
coatings with high deposition rates (i10 mm/min) and could thus overcome
the economical limitations encountered to date by vapor deposition methods.
Figure 2.7-12. As-deposited Inconel 625

structure made by EB-DVD, relative den-
sity 0.01 (courtesy IPML, University of
Virginia, Charlottesville).
References
1. V. Shapovalov, MRS Bull. 1994, 19(4), 24 16. O. Andersen, L. Schneider, G. Stephani,

2. J. Banhart, J. Baumeister, in Porous and Ingenieur-Werkstoffe 1998, 7(4), 36.
Cellular Materials for Structural Applications, 17. U. Waag, L. Schneider, P. LoÈthman,
D. S. Schwartz, D. S. Shih, A. G. Evans, G. Stephani, Metal Powder Rep. 2000,
H. N. G. Wadley (eds), Materials Research 55(1), 31.
Society Proceedings, Vol. 521, Warrendale, 18. G. Stephani, U. Waag, P. LoÈthman, O.
Pennsylvania, USA, 1998, p. 121. Andersen, L. Schneider, F. Bretschneider,
3. D. J. Sypeck, P. A. Parrish, H. N. G. Wadley H. Schneidereit in Advances in Powder
in Porous and Cellular Materials for Structural Metallurgy & Particulate Materials, Proc. 2000
Applications, D. S. Schwartz, D. S. Shih, A. G. World Congress Powder Metall. Particulate
Evans, H. N. G. Wadley (eds), Materials Re- Mater. MPIF, Princeton 2000, p. 12 119.
search Society Proceedings, Vol. 521, War- 19. O. Andersen, U. Waag, L. Schneider,
rendale, Pennsylvania, USA, 1998, p. 205. G. Stephani, B. Kieback, Adv. Eng. Mater.
4. J. M. Beggs, T. G. Wang, D. D. Elleman, US 2000, 2(4), 192.
Patent 4 344 787, 1982. 20. L. Schneider, U. Waag, P. LoÈthman in Proc.
5. C. P. Ashdown, J. G. Bewley, G. B. Kenney, 2000 Powder Metall World Congress, p. 510
US Patent 5 024 695, 1991. 21. W. Schatt, K.-P. Wieters, Powder Metallurgy:
6. G. D. Cremer, US Patent 4 162 914, 1979. Processing and Materials, EPMA, Shrewsbury
7. L. B. Torobin, US Patent 4 671 909, 1987. 1997, p. 344.
8. K. M. Hurysz, J. L. Clark, A. R. Nagel, C. U. 22. J. A. Roberts in Handbook of Fillers and
Hardwicke, K. J. Lee, J. K. Cochran, T. H. Reinforcements for Plastics, H. S. Katz, J. V.
Sanders in Porous and Cellular Materials for Milewski (eds), Van Nostrand Reinhold, New
Structural Applications, D. S. Schwartz, D. S. York 1978, p. 579.
Shih, A. G. Evans, H. N. G. Wadley (eds), 23. R. De Bruyne, I. Lefever, J. Saelens,
Materials Research Society Proceedings, J. Vandamme in Advances in Powder
Vol. 521, Warrendale, Pennsylvania, USA, Metallurgy & Particulate Materials, Proc. 1996
1998, p. 191. World Congress on Powder Metall. Particulate
9. J. L. Clark, K. M. Hurysz, K. J. Lee, J. K. Mater., MPIF, Princeton 1996, p. 16±99.
Cochran, T. H. Sanders in Metal Foams and 24. G. Lotze, G. Stephani, W. LoÈser, H. Fiedler,
Porous Metal Structures, J. Banhart, M. F. Mater. Sci. Eng. A 1991, 133, 680.
Ashby, N. A. Fleck (eds), MIT Verlag, Bremen 25. O. Andersen, G. Stephani, Metal Powder
1999, p. 171. Rep. 1999, 54(7/8), 30.
10. H. Danninger, S. Strobl, R. GuÈrtenhofer in 26. S. Steigert, Z. Li, O. Andersen, G. Stephani,
Proc. PM `98: Powder Metall. World Congress T. Schrooten in Proc. Materials Week 2000
Exhib., Grenada, V. Arnhold, A. Romero [http://www.materialsweek.org/proceedings/
(eds), EPMA, Shrewsbury, Vol. 1, 1998, p. 185. index.htm].
11. S. Strobl, M. Kupkova, H. Danninger, M. 27. A. D. Smigelskas, E. O. Kirkendall, Trans.
Kupka, E. Dudrova, L. Kovac, R.GuÈrtenhofer AIM 1947, 171, 130.
in Proc. Int. Conf. FRACTOGRAPHY 2000, 28. F. Aldinger, Acta Met. 1974, 22, 923.
Slovak Republic, L. Parilak(ed.), IMR SAS 29. A. BoÈhm, B. Kieback, Z. Metallkd. 1998,
Kosice, 2000, p. 446. 89(2), 90.
12. M. JaÈckel, German Patent 3 210 770, 1982. 30. M. KnuÈwer, in Metal Foams and Porous
13. A. Rousset, J.-P. Bonino, Y. Blottiere, Metal Structures, J. Banhart, M. F. Ashby,
C. Rossignol, French Patent 8 707 440, 1987. N. A. Fleck (eds), MIT Verlag, Bremen 1999,
14. Product data sheet, MECAPROTEC Indus- p. 203.
tries, Muret, France, 1998. 31. M. W. Kearns, P. A. Blenkinsop, A. C.
15. M. Jaeckel, H. Smigilski, German Patent Barber, T. W. Farthing, Int. J. Powder Metall.
3 724 156, 1987. 1988, 24(1), 59.
32. D. S. Schwartz, D. S. Shih, R. J. Lederich, Wadley (eds), Materials Research Society

R. L. Martin, D. A. Deuser in Porous and Proceedings, Vol. 521, Warrendale, Pennsyl-
Cellular Materials for Structural Applications, vania, USA, 1998, p. 231.
D. S. Schwartz, D. S. Shih, A. G. Evans, 35. S. B. Kulkarni, P. Ramakrishnan, Int. J.
H. N. G. Wadley (eds), Materials Research Powder Metall. 1973, 9(1), 41.
Society Proceedings, Vol. 521, Warrendale, 36. M. Bram, C. Stiller, H. P. Buchkremer,
Pennsylvania, USA, 1998, p. 225. D. StoÈver, H. Baur, Adv. Eng. Mater. 2000,
33. D. T. Queheillalt, B. W. Choi, D. S. 2(4), 196.
Schwartz, H. N. G. Wadley, Metall. Mater. 37. G. Rausch, T. Hartwig, M. Weber, O. Schulz,
Trans. A 2000, 31, 261. Materialwiss. Werkstofftechn. 2000, 31(6), 412.
34. D. C. Dunand, J. Teisen in Porous and Cel- 38. D. T. Queheillalt, D. D. Hass, D. J. Sypeck,
lular Materials for Structural Applications, D. S. H. N. G. Wadley, J. Mater. Res. 2001, 16(4),
Schwartz, D. S. Shih, A. G. Evans, H. N. G. 1028.
3
Secondary Treatment of Cellular Metals
F. Simancik
The introduction of new materials into practice requires suitable technologies for
their machining, forming, surface coating, or joining. Many papers in the literature
deal with the preparation of cellular metals and with the evaluation of their me-
chanical and physical properties, but only limited attention is paid to their second-
ary treatment. There is no doubt that the following unusual features of cellular me-
tals require specific approaches for shaping and fabrication.
x Crushability of the structure (difficult forming and machining).
x High sensitivity to tensile loading (difficult forming).
x Defects in cell walls and in surface skin (difficult coating).
x Outer surface that is usually covered with oxides (difficult coating, brazing, weld-
ing).
x Presence of the melt stabilizing ceramic particles (difficult machining).
Components made of metallic foam are usually prepared in near-net-shapes by
foaming in the mold. These porous components are always covered with a compact
surface skin. However, the skin usually contains various defects like small holes or
microcracks, so it cannot be considered fully dense. The skin significantly im-
proves the properties and the appearance of the foamed part, so it is undesirable
to remove it. In addition, removing the skin would be expensive.
There are other cellular metals made using various templates or placeholders.
They have open-cell structures and no outer skin (the template must be removed
from the structure after solidification). It is obvious that the secondary treatment of
these two kinds of structures will be very different.
Machining
Machining of net-shape foams breaks the dense surface skin and reveals the inner
structure of the pores. Machining is best avoided at the design stage. Nevertheless,
even well-designed components will need to be cut and drilled. Cellular metals can,
in principle, be cut by all conventional methods, but achieving a high-quality sur-
face is substantially more costly than for traditional materials. If ceramic particles
are used for stabilizing the liquid foam, the excessive wear on the machine tools
72 3 Secondary Treatment of Cellular Metals
has to be taken into account, especially if SiC particles are present. Conventional
machining causes bending of thin pore walls in ductile alloys and fracture in brittle
alloys. Pore walls yield or fracture at their thinnest point, which is often out of the
cutting plane. This leads to poor quality of the cutting surface and low precision.
The accuracy and quality of the cutting plane are significantly affected by partial
melting of the thin pore walls and their subsequent diffusion bonding to the cut-
ting tool. High porosity and the resultant low thermal conductivity do not allow for
efficient cooling by transport of heat into the surroundings. On the other hand,
melting and bending of pore walls can be utilized for the strengthening of the cut-
ting surface (bent pore walls are pressed into the surface pores, which improves
their apparent density).
If high precision and a good quality machined surface is required, traditional
methods of machining cannot be used. Electric discharge machining was demon-
strated to be most suitable, as it avoids the mechanical effect on the pore walls and
thus their bending and failure. Material in the cutting plane simply burns out. The
result is a clean cutting plane (without material pressed into the surface pores) and
good dimensional accuracy. The speed of electric discharge machining is substan-
tially greater in porous materials than with dense ones. The main disadvantage of
the process is the need to use dielectric fluid, which penetrates into the porous
structure and needs to be removed after machining by burning: this is environ-
mentally harmful. If a simple cut is required, wire electric discharge machining
in water can be used to overcome this disadvantage. A water beam with abrasive
can also be used for the precise cutting of cellular metals. The quality of the cutting
plane is substantially better than in the case of classical band saws (the bending of
pore walls does not take place). As the cutting speed and the thickness of the plate
increases, the quality of the bottom edge of the cutting plane deteriorates owing to
scatter of the beam. The same effect takes place in plasma or laser cutting.
If the component is foamed, its final shape derives from the mold and forming
is usually not necessary. From the practical point of view, forming can be consid-
ered with respect to flat plates. Sometimes bending or roll-bending of these panels
is required. During bending, cracks initiate in the tensile surface and then they ra-
pidly propagate across the whole section. The bending that can be achieved without
damage is limited by the plasticity of the thin surface skin and the thickness of the
plate (as the thickness of the plate grows, the deformation of the tensile loaded part
of the sample for a given bending radius increases). Panel bending is therefore pos-
sible only for foams made of ductile alloys and/or at higher temperatures (see
Section 3.1).
Surface coating
For surface coating the same basic principles can be applied as for the bulk cell-
wall material, but ªsoakingº of the porous structure should be taken into account
as it significantly restricts those surface treatment procedures in which compo-
nents are immersed (acid pickling, anodizing, etc.). Aluminum foam can be ano-
dized like aluminum, but removal of the electrolyte from the structure is required
to avoid corrosion. Owing to the microporosity of the surface skin, painting with
viscous paints is not problematic. Baking of the paint can result in the formation of
bubbles (blisters in the surface paint) caused by the expansion of gases in pores
and their attempts to escape through surface microcracks. For this reason it is re-
commended to leave some bare areas to enable the foam to ªbreatheº on tempera-
ture changes. If the foam has to be sealed completely, the coating process should
be performed in vacuum until the coating solidifies.
Joining
For practical applications of cellular metals, joining techniques are required. As the
structure of cellular metals is similar to that of wood, methods used for joining
wooden parts, especially gluing and bolting, can also be used with cellular metals.
Moreover, methods of joining metal parts, such as welding and soldering, are also
possible.
Joining cellular metal parts with bolts is simple, quick, flexible, and cost effec-
tive: the joint is also detachable. The strength of the joint depends primarily on
the density and pore size of the foam and increases if foam with a surface skin
is used. Higher strength is achieved with foams having finer porosity (finer poros-
ity provides a larger contact surface for the thread of the bolt). The strength of the
joint can be significantly improved if the drilled hole is filled with a suitable adhe-
sive.
Brazing and Soldering

The brazing or soldering of metallic foams is still questionable. The metal surface
is always coated with various oxides, depending on the alloy used. Oxides impede
the formation of brazed or soldered joints and therefore they must be removed be-
fore or during the soldering. They can be removed either mechanically (scraping)
or chemically (using fluxes). Scraping of oxides usually takes place under a layer of
melted solder to avoid subsequent oxidation. However, this is a laborious operation
demanding a skilled worker and it is costly. The strength of the soldered joint is
usually lower than the tensile strength of the foamed sample owing to the negative
effect of solder on the strength of the basic alloy near the joint. Flux can be used
only for foams with a compact surface skin, otherwise there is a danger of it pene-
trating into the porous structure. It cannot be fully removed from the pores, which
increases the danger of subsequent corrosion. As small cracks always appear on
the foam's surface, soldering with premature fluxing can be used for sandwich
structures. Foams without a surface skin can be soldered without fluxing, but
the strength of the joint is substantially lower than for samples with the surface
skin (oxides cannot be fully removed from the surface of open pores by scraping).
Glued Joints
Metallic foams can be glued with all types of adhesive recommended for gluing the
cell-wall metal. These joints can be quickly and easily made and their strength
often exceeds the strength of the basic foam structure. Because of the simplicity
and flexibility, glued joints can be recommended for preparation of permanent
joints.
74 3 Secondary Treatment of Cellular Metals
However, glued joints also have disadvantages: low thermal stability, different
coefficients of thermal expansion for the adhesive and the bonded material, de-
creasing strength of the glued joint due to aging of the adhesive, and formation
of an insulating layer (thermal and electrical) are some of them. Moreover, if
foams without a surface skin are glued together, a greater consumption of adhesive
(weight increase, cost) has to be expected because of the significantly enlarged sur-
face area.
Welding
Metallic foams can be also welded, if the weldability of the cell-wall material is
good. Welded joints are very promising, but the techniques need further develop-
ment and the costs are still relatively high for practical application. Welding prob-
lems arise mainly from the very low and variable thickness of the surface layer.
This obstacle can be partially overcome using controlled heat input into the
joint, by laser-welding for example. Good joints can be obtained if the foamable
material is used as filler. Welding in combination with such a filler can provide
a method for repair of foamed parts.
Another possibility for joining cellular metals is the use of integral foams con-
taining the parts made from bulk material, which are built into the foamed struc-
ture and if possible diffusion bonded to the foam during foaming. These parts pro-
vide the contact counterparts of the foam for a joint. The bulk material can be
either locally placed inside the foam if the position of joint is known before foam-
ing (Fig. 3-1a), or it can create a large portion of the foam's surface (Fig. 3-1b). The
material of the built-in element should be compatible with the cell-wall alloy with
respect to corrosion resistance and thermal expansion. Locally distributed elements
need precise positioning and suitable fastening to avoid displacement due to foam
motion during foaming, which makes the foaming procedure more expensive. The
use of various coversheets that are diffusion bonded to the foam during foaming
can eliminate the higher costs and lower flexibility of the built-in elements, on
the other hand the higher overall weight must be taken into account.
a) b)
Figure 3-1. Joints made using bulk aluminum parts diffusion bonded to the AlSi12 foam as:
a) locally placed element, b) 3 mm thick Al coversheet.
Further Reading
1. H. P. Degsicher, A. Kottar, ªOn the Non- 4. C. Born, H. Kuckert, G. Wagner, D. Eifler,

destructive Testing of Metal Foamsº in Metal MetallschaÈume, Wiley-VCH, Weinheim 2000,
Foams and Porous Metal Structures, J. Banhart, p. 547. [Ultrasound soldering.]
M. F. Ashby, N. A. Fleck (eds), MIT Verlag, 5. R. Neugebauer, T. H. Hipke, in Porous and
Bremen 1999, p. 213 Cellular Materials for Structural Applications,
2. O. B. Olurin, N. A. Fleck, M. F. Ashby, J. D. S. Schwartz, D. S. Shih, A. G. Evans,
Mater. Sci. 2000, 35, 1079. [Joining using H. N. G. Wadley (eds), Materials Research
wood screws.] Society Proceedings, Vol. 521, Warrendale,
3. K.-J. Matthes, H. Lang, MetallschaÈume, Pennsylvania, USA, 1998, p. 265. [Water jet
Wiley-VCH, Weinheim 2000, p. 558. [Solder- cutting, joining.]
ing of foams.]
3.1
Forming, Machining, and Coating
M. C. Hahn, R. Braune, and A. Otto
Cellular metals are a new class of materials with low densities and novel physical,
mechanical, thermal, electrical, and acoustic properties, which offer potential for
lightweight structures, for energy absorption, and for thermal management [1].
In order for their outstanding material properties to be used, they have to be ma-
chined and processed to build structural components. The relevant techniques are
described in this section.
3.1.1
High-Temperature Forming
At present cellular metals produced from the melt, by deposition, and by solid-state
processes are available. They have very valuable properties in terms of density, stiff-
ness, and energy absorption, but their potential usefulness is limited by the diffi-
culty of forming or fabricating finished parts. A forming process is necessary for
the manufacture of 3D structural parts and a joining technique is required to fab-
ricate such parts into composite components.
3.1.1.1 Specific Problems in Foam Forming

Today it is possible to make 3D composites with cellular aluminum cores from roll-
plated foamable semi-finished products, as described in Section 3.3. This method,
however, is based on the powder-compact foaming technique and is quite costly.
76 3.1 Forming, Machining, and Coating
Foams produced from melts, in contrast, are much less expensive but there are
problems with the forming properties of foamed cellular metals. At ambient tem-
perature, forming the foams gives rise to fractures or to the collapse of the pore
structure in the deformed zone. Local stress and strain concentrations can occur
causing failure at the weakest parts of the cellular structure [2].
Forming at elevated temperatures facilitates the shaping of metallic foams with-
out failure and with the pore structure being retained. At high temperatures the
yield stress and the deformation limits can be significantly reduced so that yielding
can occur. Local stress concentrations can be reduced by creep relaxation. The de-
tails of the process are described in the following section.
3.1.1.2 Process Sequence for Manufacturing 3D Composites with Aluminum Foam

Cores
The process sequence for manufacturing 3D composites with core structures of
foamed aluminum is shown in Fig. 3.1-1. The cover sheets of the sandwich
panel are formed in a first process step to serve as forming tools for the cellular
core material in the subsequent integrated forming and joining process. The result
is a complex near-net-shape composite part.
The aluminum foam is formed at temperatures closed to the solidus tempera-
ture. As an example, the working area in the Al Si phase diagram and a workpiece
made from the alloy AlSi12 are shown in Fig. 3.1-2.
Figure 3.1-1. Process sequence [3].
Figure 3.1-2. Left) working area in Al Si phase diagram [3]. Right) workpiece formed in a steel
mold [3].
The workpiece in Fig. 3.1-2 was formed using a steel mold in a chamber furnace
in the temperature range of the solidus. The forming force in this case was gravity.
The mold was reproduced well, including the corrugations, but a disadvantageous
temperature management and locally higher stresses at the beginning caused a col-
lapse in contact areas at the edges.
3.1.1.3 Material Behavior at the Solidus

The behavior of the material in the temperature range of the solidus is important
for the later forming process of the different foamed metals.
Besides the alloying elements, the production process also influences the ther-
mal behavior. Materials produced by the powder-compact foaming technique
have a surface skin, which impedes heat transfer from the furnace atmosphere
into the foam. Materials produced from the melt mostly have open pores outside
and therefore, because of the large surface, heat transfer into the material is accel-
erated. Consequently, the temperature range of the working area for high-tempera-
ture forming is reached earlier.
Another important factor is the foaming agent. Up to the solidus temperature
metallic foams do not show any permanent changes in the cell structure, but on
reaching the solidus temperature foams that are produced with titanium hydride
(TiH2) as the foaming agent show the formation of new pores in the cell walls.
This results from TiH2 remaining from the production process that releases hydro-
gen when the dissociation temperature is surpassed and the alloy starts to melt.
The cell structure changes as the temperature rises, and on exceeding the solidus
a combination of pores is possible (Fig. 3.1-3, left). Foams that are produced by gas
injection into the melt do not show such a behavior; the pore structure remains
stable even above the solidus, as shown in Fig. 3.1-3, right.
It is obvious that a high-temperature forming process should take place below
the solidus of cellular aluminum to avoid weakening the formed parts by addi-
tional coalescence of pores.
Figure 3.1-3. TiH2 (left) and gas injection foamed (right) foams: influence of temperature and
foaming agent on cell structure.
3.1.1.4 Forming of Cellular Metals at High Temperatures

The temperature at which the forming process takes place is important. To form
structural parts of titanium hydride aluminum foams it is necessary to work
below the solidus temperature to prevent the cell structure from collapse because
the main factor of this procedure is the forming force. Above this temperature
threshold value a controlled forming process is not feasible using this material be-
cause even very small forming forces destroy the cellular structure. Thus, tempera-
ture management is quite difficult. On the one hand, the yield stress has to be re-
duced to allow high deformation, and on the other hand the solidus temperature
must not be exceeded.
The foams produced by gas injection, which are often stabilized by particles,
show a different behavior at temperatures above the solidus. The particles, for ex-
ample silicon carbide (SiC), added to raise the viscosity in the production process
[4] keep the structure stable even in this temperature range because the particles
stabilize the aluminum matrix and the cellular structure. In the temperature re-
gion above the solidus the temperature is the main factor because the material
starts to melt into a semi-solid state. The yield force can be reduced again, whereas
the respective solidus temperature depends on the alloy of the foam. Besides the
temperature, time is an important factor in the forming process because the pro-
cedure is diffusion controlled and thus time dependent. The longer the forming
process lasts the larger are, for example, the resulting bending angles. In addition,
variation decreases, which is reflected in improved reproducibility at slow forming
velocities.
Some examples of high-temperature forming of cellular metals are shown in
Fig. 3.1-4. Foam samples made from the melt by gas-injection with about 15 wt.-%
SiC particles, were bent at temperatures of 5 (top), 10 (middle), and 15 hC above
solidus (bottom) with a constant forming force. In this case the force was the
Figure 3.1-4. Samples of aluminum foam

bent by high-temperature forming.
weight of the bending wing. As can be seen the material can be formed without
collapse of the cell structure, and relatively large bending angles are possible.
Owing to the properties mentioned above, the temperature range and with it the
working area and the process window can be expanded. Thus, this material is very
well suited to high-temperature forming process in contrast to the TiH2 foamed
materials, which only show a small process window.
3.1.2
Machining
Conventional cutting and machining techniques like sawing, milling, or drilling all
are suitable for processing metal foams. But using conventional machine tools
causes surface distortion or damage to low-density metal foams [5]. Nevertheless,
the conventional methods are satisfactory for most applications in which metal
foams have to be machined [6, 7].
For accurate cutting electrodischarge machining (EDM), chemical milling, water-
jet cutting, high speed cutting, or diamond sawing are recommended [5]. This is
especially necessary for machining metallic foam specimens for testing, to avoid
cell damage that would influence the properties of the samples.
Owing to the cellular structure of metal foams there are different methods of
machining the material. The appropriate technique should be selected according
to the required quality of the cutting surface. For cutting of metallic foams rein-
forced with hard particles like SiC, a diamond cutting tool might be required.
3.1.3
Coating
Owing to their properties, metal foams have great potential in lightweight con-
struction. At present there are few applications for metallic foams because
foams with open surfaces show relatively low mechanical strength and are difficult
to join to other parts. So metallic foams are not yet well suited to structural appli-
cations [8]. To further enhance the stiffness and strength of the foams they are
often used as composites with roll-plated or adhesive bonded cover sheets. The
cut surface of metal foam displays open cells and a rough texture, and is vulnerable
to local damage [5]. It can be filled with an epoxy or other resin, or, in the case of
syntactic structures that have a natural skin, the surface can be coated by conven-
tional methods. These manufacturing methods are mostly limited to plane sur-
faces and do not allow complex shaped parts. By using thermal spraying tech-
niques almost any desired form of foam-spray-deposit composites can be produced.
Furthermore, there are considerable advantages concerning recycling and usage at
elevated temperatures compared with adhesive-bonded composites [8].
Figure 3.1-5. Spray deposition diagram [8].
3.1.3.1 Mechanical Properties of Spray Deposits

Thermal spraying is a surface technique using powder or wire materials melted by
an energy source. The molten materials are accelerated and set onto the surface to
be coated by a gas jet (Fig. 3.1-5). The energy source can be a wire arc or an inert
gas plasma. Spraying materials can be metals, ceramics, or plastics.
The mechanical properties of spray deposits are not the same as those of the
bulk material, because there are differences in the structure. Spray deposits
show anisotropic behavior determined by their lamella structure. During the
rapid cooling process of the molten spray particles residual stresses develop,
which, together with impurities or brittle phases like oxides inside the layer,
may cause the lamellae to slide under load. Thus, normally the strength of the
spray deposit is less than that of the intrinsic bulk material [8].
There are methods for improving the mechanical properties of the spray depos-
its. By preheating the substrate the cooling velocity, and with it the shrinkage stres-
ses, can be reduced. The use of inert gas instead of compressed air for spraying
decreases the proportion of brittle phases in the spray deposits. Furthermore, spe-
cific thermal treatment leads to significant improvements of the spray deposit
properties. Metallurgical and positive substance joints between deposit and sub-
strate result from diffusion and increase the adhesion of the deposit. In addition,
diffusion causes an increase of cohesion of the lamellae inside the spray deposit [8].
The boundary layer between the foam core and the spray deposit is of essential
importance for the properties of the composites. This layer has to transmit the
maximum shear stress. If the adhesion of the spray deposit is not strong enough,
this will cause delamination due to the shear stress. The deposit peels off by crack
propagation in the boundary layer and so the whole component fails. Conse-
quently, in the selection of the deposit materials both high specific strength and
favorable spray characteristics have to be taken into account to get an optimum
for the construction of structure parts.
3.1.3.2 Specific Difficulties in Foam Coating

To get a stable boundary layer that is able to transmit the shear stress between
foam core and spray deposit it is necessary to allow the sprayed particles to
interlock with the substrate. To achieve this, the surface to be coated has to be pre-
treated to roughen and to activate it. This is generally done by corundum blasting
[8].
Currently available foamed semi-finished products made via the powder-metal-
lurgical route have a thin smooth surface skin. This surface is often damaged dur-
ing the blasting process, which opens the underlying pores. Thus, a surface treat-
ment that is more suitable for foams is required because the spraying process is
not able to fill the open cells and form a new plane surface. A suitable approach
may be microblasting, in which the semi-finished foam products are blasted
with low pressure and with particles less than 100 mm in diameter. Although the
process parameters have not been fully investigated yet, the first results of the mi-
croblasting technique have been very satisfying.
3.1.3.3 Thermal Sprayed Composites from Metal Foams

Currently aluminum foam is coated by wire arc spraying. By bringing the wires
together in a strong electric field inside a spray gun, an arc is formed which causes
the wire to melt. Then the molten particles are sprayed on the substrate by an ac-
celerating gas jet. This is a widespread industrial process because it is economical
and easy to use [8]. But using compressed air as accelerating gas has a negative ef-
fect on the formation of the spray deposit because of the oxides that arise in the
different layers. In addition, the commercial wires are developed for corrosion pro-
tection and not for high strength and outstanding mechanical properties as re-
quired for metallic foam coating. An obvious improvement in the formation of
the spray deposit can be achieved by the use of nitrogen (N2) as the acceleration
gas, thereby causing significantly less oxides than compressed air (Fig. 3.1-6).
The important process parameters are spraying distance, gas pressure, and arc
power.
The usual spray materials are AlSi6 and NiAl5. AlSi6 is normally used to repair
damaged aluminum components and shows good properties regarding specific
strength and hardness. A thermal treatment to improve the adhesion of the layers,
on the other hand, reduces the mechanical properties of the deposit. NiAl5 is a typ-
ical spraying material used for its excellent adhering properties. It is easy to spray
and facilitates the production of deposits with a very low proportion of oxide when
used with an inert gas for spraying. Thermal treatment reduces the residual stres-
Figure 3.1-6. High (left) and low (right) oxide NiAl5 layers [8].
ses and leads to a relatively small reduction of the hardness of the spray deposit.
One disadvantage is the high density of NiAl5 compared with aluminum alloys.
The use of highly tensile solidifiable aluminum alloys as spray materials is the sub-
ject of investigations at the present time [8].
Another technique more complicated than wire arc spraying but producing
much better spray deposits is plasma spraying (Fig. 3.1-7). The spraying material
is injected into an inert gas plasma as a powder and then heated and accelerated in
the direction of the substrate. The material is an AlSi30 powder, which in conjunc-
tion with suitable process parameters leads to relatively hard deposits when
sprayed. A practical thermal treatment of these spray deposits and their further me-
chanical properties are currently being investigated.
Thermal spraying is a suitable and flexible method of producing composites with
metallic foam cores. But investigations are still necessary with regard to the en-
hancement of the mechanical properties of the deposits as well as the pretreatment
of the metallic foams.
Figure 3.1-7. Left) spray deposit on aluminum foam [8]. Right) plasma sprayed deposit [8].
References
1. M. F. Ashby, A. Evans, N. A. Fleck, L. J. 5. M. F. Ashby, A. Evans, N. A. Fleck, L. J.

Gibson, J. W. Hutchinson, H. N. G. Wadley, Gibson, J. W. Hutchinson, H. N. G. Wadley,
Metal Foams: A Design Guide, Butterworth- Metal Foams: A Design Guide, Butterworth-
Heinemann, Boston, MA, USA, 2000, Ch. 1. Heinemann, Boston, MA, USA, 2000, Ch. 15.
2. M. Hahn, A. Otto, in MetallschaÈume, H. P. 6. Shinko Wire. Technical description: AL-
Degischer (ed), Wiley-VCH, Weinheim 2000, PORAS, Shinko Wire Co. Ltd., Amagasaki,
p. 432 433. Japan 1998.
3. M. Hahn, A. Otto, in MetallschaÈume, H. P. 7. MEPURA. Technical description: Alulight,
Degischer (ed), Wiley-VCH, Weinheim 2000, Metallpulvergesellschaft mbH, Ranshofen,
p. 538 540. Austria 1996.
4. M. F. Ashby, A. Evans, N. A. Fleck, L. J. 8. M. Maurer, E. Lugscheider, in MetallschaÈume,
Gibson, J. W. Hutchinson, H. N. G. Wadley, H. P. Degischer (ed), Wiley-VCH, Weinheim
Metal Foams: A Design Guide, Butterworth- 2000, p. 523 526.
Heinemann, Boston, MA, USA, 2000, Ch. 2.
3.2
Joining Technologies for Structures Including Cellular Aluminum
T. Bernard and H. W. Bergmann
Products of cellular aluminum are on the brink of widespread application. Various

research projects concentrate on the characterization and simulation of the me-
chanical properties of the cellular structures, others target the reduction of the
cost of production of cellular metals. However, until feasible joining technologies
are not available, mass production of components is impossible. Though there are
already well-known processes for the production of foam sheet composites like
the roll-cladding process [1], other technologies have to be investigated for the par-
tial stiffening of hollow extruded profiles.
3.2.1
Introduction
The aim of this contribution is to give an overview of feasible conventional joining

technologies. Section 3.2.2 summarizes common techniques that are regarded as
suitable for cellular metals according to the literature, ensuing an overview of fea-
sible joining technologies for foam foam joints. The main focus however, is put
on foam sheet joints and their properties in different loading cases as this is of
greater interest for structural applications. The structure of the joints and their
properties are described in Section 3.2.3 and Section 3.2.4.
3.2.2
Feasible Joining Technologies
This section summarizes conventional joining techniques that are feasible for com-
pound structures utilizing aluminum foams.
3.2.2.1 Mechanical Fastening Elements

There is a huge variety of mechanical fastening elements available. The possibili-
ties range from hollow spheres- to metal-plugs, as well as nails, screws, and rivets.
Experiments show that there are only a few suitable technologies in which the
strength of the joint is approximately the strength of the applied cellular material
[2].
With nails, only a frictional-fitting joining mechanism can be achieved with the
consequence that the strength of the joint is too low for technical applications.
In contrast to this, fasteners using not only a frictional-, but a form-fitting me-
chanism as well, are preferable. Screws, as well as some types of blind rivets,
show this combined fitting mechanism. The application of wood screws is de-
scribed in the literature [3]. Additionally self-drilling screws were investigated,
84 3.2 Joining Technologies for Structures Including Cellular Aluminum
which have the advantage that no additional hole has to be drilled into the sheet
material.
The performance of the mechanical joints can be improved if the fastening ele-
ments are additionally smeared with epoxy adhesives before driving them into the
foam [4,5].
A further possibility for joining foams with mechanical fasteners is the applica-
tion of integral foams [3], where the foam contains parts made from bulk material.
These parts, screw sockets for example, are integrated into the foamed structure
and bonded to it during the foaming process. However, the locally distributed ele-
ments need precise positioning and suitable fastening in order to avoid a displace-
ment caused by the foam motion during foaming. Consequently, not only is the
weight of the cellular structure increased by integrating these parts, but the cost
of production as well.
3.2.2.2 Gluing
Gluing has been described as a feasible joining technology for aluminum foams
[6]. Different gluing systems have to be distinguished. One-component systems
are, for instance, epoxy-based and cure for approximately 30 min at 180 hC. The
process conditions of these adhesives make them very interesting for car manufac-
turers, as the glue can cure during cathodic electrocoating and no further working
steps are necessary. Epoxy-based systems are available as conventional glues or as
expanding glues. The latter ones are very convenient for bridging high tolerances
and for integrating foam bodies into hollow extruded profiles. In case a thermal
curing process is undesired, two-component systems are available that cure at
room temperature. By combining riveting and gluing it is easily possible to fix
the foam body in the structure while the curing of the glue takes place. Addition-
ally, the properties of the composite are improved in certain loading cases.
3.2.2.3 Welding
Welding processes that have been shown suitable for joining aluminum foams are
laser-beam welding [6] and ultrasonic (US) welding [7].
With the latter process, various geometries of the joining area are possible. They
reach from quadratic joints with a maximum area of 100 mm2 to roller seam welds
with a width of 1 5 mm and a maximum length of 500 mm. Finally circular bond-
ing areas with typical diameters of 18 mm are producible. A characteristic of the
US welding process is that it is a joining process in the solid state based on friction
and consequently not a real welding process, where both joining partners are trans-
ferred into the liquid state for joining [7].
Laser-beam welding has been shown as a viable joining technology for alumi-
num foams [6]. Owing to its characteristics such as the keyhole effect and the
minimal heat input, deeper parts of the structure can be integrated into the
joint as well. Fig. 3.2-1 shows the principle mechanism of surface welding and
the key-hole welding effect for bulk and foam material.
Figure 3.2-1. Surface welding and keyhole effect during laser-beam welding.
3.2.2.4 Soldering and Brazing

Soldering and brazing of aluminum foams is impeded by their oxide layer, which
has to be removed before or during the soldering process. There are two ways of
removing the oxide layer: one is mechanical destruction by scratching, brushing,
or ultrasonics, the other is the application of flux. The latter causes serious corro-
sion problems if it is not removed properly. In consequence, the joint requires good
accessibility so that the flux can be washed out after the soldering process. This
makes soldering with flux unsuitable for sheet foam composites, where a connec-
tion along the whole surface is desired.
When removing the oxide layer mechanically, the use of foams with a casting
skin is advantageous, as the oxide layer is easier to remove.
Soldering
Various Sn- and Zn-based solders can be used for fabricating structures using Al
foams [3]. Reliable joints are said to be possible without the use of fluxes, just
by scratching the oxide layer.
Another suitable technology is the S-Bond process, developed by Euromat [8].
This solder has a chemical composition of SnAg4Ti4 (wt.-%) and a melting
range of 220 229 hC. The S-Bond process is a flux-free soldering process that is
characterized by the mechanical ripping of the oxide layer that has formed around
the molten soldering material. The oxides of the substrate layer are partially ripped
during the soldering process, so that a metallurgical interaction of the soldering
material and the substrate can take place. Additionally, the soldering material
creeps below the oxide layer of the substrate and wets its surface totally. S-Bond
is available as wire, rod or foil, which makes this process an interesting alternative
for extensive joints.
Though flux-free soldering technology can be transferred to aluminum foams,
corrosion preventing measures have to be taken as the Sn- and Zn-based alloys
can cause corrosion problems in combination with aluminum.
Brazing
Brazing (Ti450 hC) is a feasible joining process for aluminum foams [9]. Though
noncorrosive fluxes are available for brazing, it is advisable to coat the structure to
protect it from humidity. The brazes are preferably Al-based and the furnace atmo-
sphere should be inert.
An alternative is the use of a porous filler material [9], which expands during the
brazing process (Fig. 3.2-2). Thus a sudden change in porosity from foam to bulk
material is prevented. The working temperature of the filler material has to be low
enough that the porous base material does not deform during the expansion of the
filler material.
As outlined, there is a variety of suitable joining technologies available for fabri-
cating structures utilizing aluminum foams. However, for a practical investigation
the number of technologies has to be limited to one or two processes per category.
Table 3.2-1 gives an overview and some characteristic properties of relevant joining
technologies.
Figure 3.2-2. Joining of Al foam and solid Al by brazing with a filler that expands to a foam [9].
Table 3.2-1. Characterization and classification of the joining technologies
Joining Formation of Joining Joining class Detachability

technology the joint temperature
Riveting point-contact RT mechanical ±
Screwing point-contact RT mechanical I
Gluing area-contact RT to 180 hC adhesive ±
Riveting/Gluing area/point- RT to 180 hC mech./ ±

contact adhesive
Flux-free soldering area-contact 250 hC thermal ±
Laser-beam welding line-contact Melting Point thermal ±

3.2.3
Foam Foam Joints
As the main requirement of industrial appliers are feasible technologies for

foam sheet composites, the joining of foam±foam structures is shown only
schematically. The authors have carried out feasibility studies for all joining tech-
nologies mentioned in Table 3.2-1 except for laser-beam welding. Laser-beam weld-
ing of aluminum foams is described in [10]. Though Al-foam joints can be made by
this technique there are still many problems. Owing to the collapse of the cellular
structures during welding, large amounts of filler material have to be used, which
increases the density of the structure in the area of the seam. Another problem is
the thickness of the foam structures, because they will not be welded through the
whole cross section. Additionally, the number of joining geometries is limited
when welding Al-foams. Only butt-joints have been investigated so far and it
will be difficult to produce other joining geometries like T-joints and overlap joints.
In contrast, most joint geometries can be realized with area contact joints.
Table 3.2-2 shows the relevant joining techniques and the corresponding joining
geometries.
Table 3.2-2. Feasibility of the joining technologies for various foam-foam joining geometries
Butt-joint T-joint Overlap-joint
Blind riveting ± ± I
Screwing ± I I
Gluing I I I
Riveting/gluing ± ± I
Soldering I I I
Welding I with restrictions with restrictions
3.2.4
Foam Sheet Joints
The following text concentrates on processes for producing foam sheet compo-
sites [2,11]. Though there are other processes, such as the roll-bonding process
[1], further technologies have to be investigated, for integrating foam bodies into
hollow extruded profiles or for the partial stiffening of structural parts. In most
cases, foam sheet composites offer only one-sided accessibility. This has to be
taken into consideration when selecting a feasible joining technology. Under
these circumstances, all joining technologies listed in Table 3.2-1 are feasible.
3.2.4.1 Microstructural Investigations

Figure 3.2-3 shows micrographs of the investigated foam sheet joints.
Screwing (Figure 3.2-3a)

A form-fitting joining mechanism can not only be made by the bulge of a rivet, but
also by the thread of a screw. As the foam material is compressed when the screw
is driven in, enough material is present to fill the single parts of the thread, thus
creating a form-fitting joining mechanism. Its strength, however, is highly depen-
dent on the density of the foam and will decrease with it.
Riveting (Figure 3.2-3b)

Owing to the smooth surface of the rivet, there is only a friction-fitting mechanism
in the area of the shank, whose peculiarity is basically defined by the number of
cell edges adjacent to the hole. As already described (Section 3.2.2.1) a friction-fit-
ting mechanism is not sufficient for metal foams and has to be combined with a
form-fitting mechanism. In case of the displayed rivet, this is made by forming a
bulge at the bottom of the structure, whose diameter is significant bigger than the
pre-drilled hole.
Gluing (Figure 3.2-3c)

Gluing belongs to the group of joining technologies generating a material-fitting
joining mechanism. This is caused by adhesive forces between the surface of the
substrate and the glue. Consequently, well-glued samples should not fail at the in-
terface, as the adhesive forces should be higher than the cohesive forces in the glue
Figure 3.2-3. a) Screwing

b) Riveting
c) Gluing
d) Soldering
e) Laser-beam welding
Figure 3.2-3. Micrographs of the investigated foam sheet joints.
or the substrate. The bonding of the glue and the surface might either be caused by
ionic bonding or polar interactions [12]. Gluing is a feasible joining technology for
foamed metals, independent of the presence of a casting skin. Though the area in
direct contact with the sheet material is drastically reduced in the absence of a cast-
ing skin, the viscosity of the glue is low enough to fill the pores. The loss of surface
can be compensated for by an additional form-fitting joining mechanism.
Soldering (Figure 3.2-3d)

Like gluing, soldering can be used as a joining technology generating an area-con-
tact joining mechanism. In spite of the extensive oxide layer on the metal foam,
interdiffusion zones form between solder and substrate materials, when applying
S-bond. For soldered samples, the presence of a casting skin is advisable. As al-
ready described, relative movement between the parts is necessary during joining
with a flux-free solder like S-bond [8]. Consequently, both surfaces have to be as
plane as possible, so that the wetting of the substrate can take place everywhere.
If there is no casting skin, the solder flows into the pores, but in contrast to
glued samples does not form a material-fitting joint with the cell walls, as the
oxide layer of the molten solder in the pores cannot be torn properly during the
relative movement of the parts. Consequently, the remaining joining area would
be limited to the number of the present cell edges, which would result in a reduced
strength of the structure.
Laser-Beam Welding (Figure 3.2-3e)

The connection of the sheet to the foam with a laser is based on two principal join-
ing mechanisms: a material one and a shape-related one. The heat input of the
laser causes the melting of the sheet material and the underlying foam structure.
After resolidification a metallic connection is created. However, this connection
exists only between the molten sheet material and the molten material of several
cell structures located close to the interface. The keyhole effect supports this me-
chanism. Because of the deep penetration of the laser beam, additional cells that
are located in close proximity to the sheet surface can be included in the joining
zone. Additionally, the concentrated heat input of the laser beam causes only mini-
mal melting of foam structures outside the original joining zone. On the other
hand, the molten material from the cover sheet and filler material flows into the
pores of the foam structure during the welding process and fills them partially.
The additional material prevents the formation of a seam groove and is responsible
for the shape related connection mechanism.
3.2.4.2 Mechanical Properties of Foam Sheet Joints

The load-carrying capability of the joints is highly dependent on the manner in
which the joint is loaded. For this reason the joining technologies have to be tested
in different loading cases.
Geometry of the Samples and Testing Parameters

Figure 3.2.4 shows sketches of the samples used for pull-out and tensile-shear
testing.
The geometries of the sandwich structures for bend tests and the corresponding
testing conditions that were selected according to the given standards are listed in
Table 3.2-3.
Figure 3.2-4. Geometry of the samples for tensile-shear and pull-out testing.
Table 3.2-3. Geometry of the samples for bend tests
Dimensions Testing speed Distance of the

(L q W q H) [mm] bending die/stamp
Four-point bend test 240 q 25 q 10 5 mm/min 200 mm / 100 mm

DIN 53 293
Three-point bend test 145 q 25 q 10 5 mm/min free distance: 45 mm

DIN 50 111
Dynamic bend test 145 q 25 q 10 5.7 km/h 105 mm

Pull-Out Tests
Pull-out tests were carried out with structures utilizing mechanical fastening ele-
ments. After preliminary investigations with a selection of conventional fastening
elements, all fasteners were excluded that form the joint according to a frictional-
fitting joining mechanism. Though good results with frictional-fitting fastening
elements could be measured with large volume samples [5], the thickness of
foams for technical applications such as the partial stiffening of structures should
be limited to less than 10 mm. In this case however, there is not enough material
left for the frictional interaction between cell edges and the fastening element, so
that it will be torn out at low loads. Figure 3.2-5 shows some characteristic
load deflection curves of joints utilizing mechanical fasteners with an additional
form-fitting joining mechanism.
Riveted samples show a larger range of deformation as the bulge has to be pulled
through a hole of smaller diameter. This means that a great amount of deforma-
tion work has to be done. In contrast to this, the diameter of the investigated
screws reduces to the tip of the screw. As a consequence, not much deformation
work can be done when pulling out the screw, after the maximum load has
been reached.
Obviously, the pull-out load is dependent on the density of the foam material.
The effect of relative density on the maximum pull-out load, Fr, is nonlinear
with a power of about 1.5 [5]
Fr Ar3=2 (1)
Figure 3.2-5. Qualitative load deflection curves for screwed and riveted samples in tensile-shear
loading.
Tensile-Shear Tests
Tensile-shear tests were carried out for the joining technologies listed in Table 3.2-1.
The maximum shear loads for the joining technologies are given in Table 3.2-4.
There are two different failure mechanisms that apply to point-contact joining
technologies or area/line-contact joining technologies. Laser-beam welded and
Table 3.2-4. Influence of the joining technology on the maximum load in tensile-shear loading
Joining method Maximum tensile shear load
Riveting 997 e 91.12 N
Screwing 1282 e 130 N
Riveting/gluing 1548 e 182 N
Gluing 2043 e 220 N
Flux-free soldering 2112 e 596 N
Laser-beam welding 580 e 116 N
glued samples tend to fail in the foam close to the joining zone. Consequently, the
mechanical properties of the cellular structure influence the behavior under tensile
shear loading. This is observable at a maximum load that represents the mechan-
ical properties of the foam. After the maximum load is passed, the foam tears al-
most brittle. Laser-beam welded samples show a lower maximum load than glued
samples as their joining area is smaller. Consequently the failure load is highly
dependent on the number of welding seams.
In contrast, riveted and screwed samples show ductile failure behavior under
tensile-shear loading. The fastening element compresses parts of the cellular struc-
ture during the deformation. When the remaining cross section of the foam body
is too small, the foam breaks, which is shown by an abrupt decrease of the load.
Influence of the Foam Density on the Joint Strength for Line-Contact Joints
Figure 3.2-6 shows that the strength of welded samples is highly dependent on the
relative density of the foam [2], i. e. the number of the cell walls in the area of the
joint. In these experiments the casting skin was removed before joining so that the
resulting joining mechanism is based on a fusion of the sheet with the cell struc-
tures as well as by form-fitting due to a flow of the melt into the porous structure.
The failure of the joint is observed to start close to the welding seam.
The failure behavior can be described by a correlation between the foam density
and the strength of the joint.
s Metalfoam Cr2 (2)

max
s Metalfoam
max rmin S ar (3)
There are two ways of assessing the dependence of the strength of laser-beam
welded composites on the foam density. For example, the strength of massive
sheet sheet joints of about 210 N/mm2 (Fig. 3.2-6) would result in a relationship
according to a power law (Eq. 2). In this study however, the joining partner is not a
massive material but a powder-metallurgic produced aluminum. As a conse-
quence, the values of the massive material cannot be reached with a density of
a)
b)
Figure 3.2-6. Influence of the density of the Al foam on the strength of laser-beam welded joints.
2.7 g/cm3. The values of composites utilizing a metal foam with a casting skin in-
dicate a linear relationship between the foam density and the strength of the joint.
Here, the casting skin must be assessed as a massive joining partner and the
resulting density of the casting skin is 1.8 g/cm3 (linear regression of Eq. 3).
The solution of the power law equation gives a lower value of 1.1 g/cm3.
Bend Tests
Influence of the Joining Technology on the Properties in Four-Point Bending

With the help of the slopes in the linear section of the load deflection curves in
four-point bending, it is possible to calculate the effective bending stiffness
(EJ)eff of a sandwich element according DIN 53 293. The optimum joint is an
area joint, so glued sandwich structures show the highest effective bending stiff-
ness, which is equal to the theoretical bending stiffness (EJ)theo. Further results
show that a decrease of the joining area by utilizing other joining techniques
leads to a reduction of the stiffness of the structure.
Therefore, riveted samples show the worst properties due to the puncture joint.
The stiffness of the structure is additionally reduced by the holes that are necessary
Figure 3.2-7. Influence of the joining technology on the energy absorbing behavior of a sandwich
structure in four-point bending after a deflection of 10 mm (see Table 3.2-3).
for integrating the rivets. Welded and riveted/glued samples show an average stiff-
ness, like soldered specimens.
As the integral of the load deflection curve (equal to the energy absorbed dur-
ing the deformation) increases with rising stiffness, the stiffness of the composi-
tes is mirrored in the values of energy absorption up to a deflection of 10 mm
(Fig. 3.2-7).
Related to their mass, structures utilizing a steel sheet (applied joining technol-
ogy: gluing) show the worst properties owing to the high weight of the sheet
material. Composites utilizing a tissue on the tensile side of the structure are
the lightest, but they have a low stiffness owing to the tissue. In consequence,
the amount of energy that can be absorbed is comparatively low.
Four-point bend tests with a constant bending moment along the sample serve
for evaluating the stiffness of a sample. In the case of a three-point bend test
the maximum bending moment is in the area of the bending stamp. Thus,
three-point bend tests serve for evaluating the strength of the structure.
Influence of the Joining Technology on the Properties in Three-Point Bending

As in four-point bending, the joining technology has a strong influence on the
properties of the composite. In contrast to the four-point bend tests, glued/riveted
samples show the best properties in this loading case. Structures that were simply
glued failed owing to cracks in the foam core. By combining the processes, the rivet
relieves the foam core from high stresses and makes the structure less dependent
on inhomogeneities of the foam. This is confirmed by the comparatively low var-
iance of the results. Apart from the structures utilizing a tissue, all samples vary
around the same value for the maximum load. For realizing a good energy absorb-
ing behavior a ductile deformation behavior is necessary as well as a high maxi-
mum load. According to Fig. 3.2-8, which shows the amount of energy absorbed
during three-point bending after a deformation of 10 mm against the mass of
the structure, glued, riveted/glued, soldered, and laser-beam welded structures
Figure 3.2-8. Influence of the joining technology on the energy absorbing behavior of a sandwich
structure in three-point bending after a deflection of 10 mm (see Table 3.2-3).
meet this requirement. Though the strength of the tissues is high enough, it fails
after a comparatively low deformation owing to catastrophic fracture.
Displaying the amount of absorbed energy in Fig. 3.2-7 to 3.2-10 shows that all key
figures located on the same straight line through the origin of the diagram have
the same mass specific value. With an increasing slope of the line, the mass-related
properties increase as well. The goal of optimization is therefore to increase the
slope of the line on which the values are located.
This cannot only be achieved by a variation of the joining technology, but also by
the density of the foam core.
Influence of the Density of the Foam Core on the Properties in Three-Point Bending
The results of three-point bend tests with structures utilizing a foam body without
casting skin and an average density from 0.3 to 1.1 g/cm3 show that there is an almost
linear dependence of the maximum load on the density of the foam core in the inves-
tigated density range.
Figure 3.2-9. Influence of the weight of the sandwich structure on the value of the absorbed
energy up to a deflection of 10 mm (see Table 3.2-3)
Comparing the performance of sandwich panels where the casting skin was still
on the core material to those without casting skin reveals that the sandwich panels
with core material without casting skin have a higher specific energy absorption
(Fig. 3.2-9).
As the foam is compressed in the area of the bending stamps during three-point
bending, the loading mode in this area is consequently similar to a compression
test. However, the casting skin is orientated perpendicular to the load and cannot
enhance the energy absorbing properties. Thus the casting skin just decreases the
mass-related energy absorbing properties in bending.
Influence of the Joining Technology on the Properties in Dynamic Three-Point

Bending
Dynamic three-point bending of sandwich structures serves for simulating sudden
impact, like in crash situations. The goal is to absorb an impact energy that is as
high as possible. This can only occur if the sheets stay bonded well to the foam
core during the whole deformation. Figure 3.2-10 shows the influence of the
different joining technologies on the impact energy that can be absorbed
against the mass of the structure. The results of the dynamic three-point
bend tests basically confirm the results of the quasi-static tests. Glued and
glued/riveted samples show the best properties, followed by soldered speci-
mens, owing to the ductile deformation behavior of the joint in bending.
Welded samples show a high variance. The high dependence of this joining
technology on the porosity of the foam increases with the testing speed. If
some parts of the structure are not joined well, owing to an uneven surface
of the foam or because of big pores in the area of the welding seam, and
if these areas are orientated unfavorably during the deformation, the composite
structure fails almost without absorbing any energy. Under optimum conditions
however, most of the kinetic energy can be absorbed. In this case, continuous
seams show better results than step-welded seams. Step-welded seams are
usually used in order to stop cracks propagating through the seam. However, in
Figure 3.2-10. Influence of the joining technology on the energy absorbing behavior of a sand-
wich structure in dynamic three-point bending (see Table 3.2-3).
the case of step-welded foam-sheet composites the share of insufficient joined

areas is increased artificially by not joined areas, which leads to a decrease of
the properties.
3.2.5
Transferability to Structural Parts
The basic fields of application of aluminum foams are energy absorbing structures
and stiffening of structural elements. The following sections show the transferabil-
ity of the described joining technologies to structural elements.
3.2.5.1 Tubes
The integration of foam bodies in tubes for energy absorbing structures is well de-
scribed elsewhere [13,14]. The results show that gluing as well as laser-beam weld-
ing are feasible joining technologies for this application. Figure 3.2-11 shows an
X-ray of a crash-absorbing element partially filled with Al foam and a detail of
the joining area.
Regarding the triggering of the sequential buckling process, laser-beam welding
shows some synergetic effects in axial loading, as shown previously [13,15]. The
initial peak load of the characteristic load deflection curve in axial loading can
be reduced, which leads to an increase in the effectiveness of the composite struc-
ture in axial loading.
Figure 3.2-11. Foam body integrated in an Al tube by laser-beam welding for an application as a
crash element in the front side member of a vehicle [13].
3.2.5.2 Hat-Profiles
Dynamic Three-Point Bend Tests

All joining technologies listed in Table 3.2-1 are feasible for integrating a rectangu-
lar foam tube into a hat-profile. Figure 3.2-12 shows the single parts of the struc-
ture and the resulting composites after testing. The various joining strategies that
were investigated included one-sided joining and three-sided joining, as shown in
Fig. 3.2-12.
Not only the selected joining technology, but also the joining strategy has signif-
icant influence on the properties of the composite structure. The properties were
investigated in torsion tests and dynamic three-point bend tests. In the latter ex-
periments, the impact mass was 600 kg, which hit the structure at a speed of
10 km/h. Figure 3.2-13 shows load deflection curves for three-sided glued and
laser-beam welded structures.
Figure 3.2-12. Integration of a foam body into a hat profile for improving the torsion, bending,
and energy absorbing behavior of the structure.
Figure 3.2-13. Influence of the joining strategy (one- or three-sided) and the joining technology
on the deformation behavior of a hat-profile in dynamic three-point bending.
Owing to the optimum contact area of glued samples, the sheet material stays
in contact with the foam during the whole deformation of the composite struc-
ture. In the welded sample, however, the sheet can deform independently of the
foam in large areas, which reduces the energy-absorbing potential of the struc-
ture in a defined interval of deformation. Independently of the joining technol-
ogy, three-sided joining improves the energy-absorbing properties of the struc-
ture.
Torsion Tests
The torsion tests shown in Fig. 3.2-14 show that glued samples have better proper-
ties than welded ones. Fewer folds form during the deformation process, which
means that more deformation energy can be absorbed by the porous structure.
In laser-beam welded samples, the welding seams that fix the foam core are the
weakest part of the structure and tear during the loading process.
Corresponding to the bend tests, the experiments with foam-filled hat profiles
show that with increasing area of the joint, the properties of the composites in-
crease.
Figure 3.2-14. Influence of the joining strategy (one- or three-sided) and the joining technology
on the deformation behavior of a hat-profile in torsion.
3.2.6
Summary
In a final overview, Table 3.2-5 and Table 3.2-6 show the properties of the investi-
gated joining technologies for different cases of loading, including a technological
assessment of the joining technologies.
Table 3.2-5. Summary of the properties of the investigated joining technologies in the different
loading cases (++ very good; + good; o overage; ± bad;/ not carried out)
Tensile Pull out test Four-point Three-point Dynamic

shear test bend test bend test bend test
Riveting + ++ ± ± ±
Screwing + + ± ± ±
Gluing + / ++ + +
Riveting/ ++ / o ++ ++
Gluing
Laser beam o / + + o
Welding
Table 3.2-6. Final evaluation of the investigated joining technologies (++ very good; + good;
o overage; ± bad)
Accessi- Costs Recycling Function- Freedom Transferability

bility ality of design on structural
parts
Riveting ++ ++ ++ + o +
Screwing ++ ++ ++ + + +
Gluing o o o o ++ ++
Riveting/ ± o o ++ o +
gluing
Soldering o ± + + ++ +
Laser-beam + ± ++ o o ++
welding
References
1. H.-W. Seeliger, in Metal Foams and Porous 8. D. Pickart-Castillo, F. Hillen, I. Rass,

Metal Structures, J. Banhart, M. F. Ashby, Materwiss. Werkstofftechn. 2000, 31, 553.
N. A. Fleck (eds), MIT Verlag, Bremen 1999, 9. K.-J. Matthes, H. Lang, Materwiss. Werkstoff-
p. 29. techn. 2000, 31, 558.
2. T. Bernard, J. Burzer, H. W. Bergmann, in 10. I. Burmester, M. Goede, J. Bunte, Mater-
Sheet Metal 2000, B. Shirvani et al. (eds), Proc. wiss. Werkstofftechn. 2000, 31, 436.
8th Int. Conf. Sheet Materials, 17 19 April 11. T. Bernard, H. W. Bergmann, Materwiss.
2000, University of Central England, Bir- Werkstofftechn. 2000, 31, 440.
mingham 2000, p.561. 12. W. Bergmann, Werkstofftechnik, Teil 2: An-
3. N. SedliakovaÂ, F. SimaÂncik, J. KovaÂcik, in wendung, 2nd ed., Hanser Verlag, Munich
MetallschaÈume, J. Banhart (ed.), MIT Verlag, 1991.
Bremen 1997, p. 177. 13. C. Haberling, H. G. Haldenwanger, T.
4. Information material of IWU Chemnitz, Bernard, J. Burzer, H. W. Bergmann, Metal
Germany. Foams and Porous Metal Structures, J. Banhart,
5. O. B. Olurin, N. A. Fleck, M. F. Ashby, Adv. M. F. Ashby, N. A. Fleck (eds), MIT Verlag,
Eng. Mater. 2000, 2, 521. Bremen 1999, p. 37.
6. J. Burzer, T. Bernard, H. W. Bergmann, in 14. R. Gradinger, M.Seitzberger, F. G. Ram-
Porous and Cellular Materials for Structural merstorfer, H. P. Degischer, M. Blainschein,
Applications, D. S. Schwartz, D. S. Smith, Ch. Walch, Metal Foams and Porous Metal
A. G. Evans, H. N. G. Wadley (eds), MRS Structures, J. Banhart, M. F. Ashby, N. A.
Symp. Proc. Vol. 521, MRS, Warrendale, PA Fleck (eds), MIT Verlag, Bremen 1999, p. 313.
1998, p. 159. 15. J. Burzer, Beitrag zur Einsetzbarkeit von
7. C. Born, H. Kuckert, G. Wagner, D. Eifler, MetallschaÈumen in der Verkehrstechnik, H. W.
Materwiss. Werkstofftechn. 2000, 31, 547. Bergmann, G. Ziegler (eds), Herbert Utz
Verlag Wissenschaft, Munich 2000.
3.3
Encasing by Casting
C. KoÈrner, F. Heinrich, and R. F. Singer
Foam parts prepared by expansion in a mold show a closed surface skin with a
thickness comparable to the cell-wall thickness of about 200 mm. Even in cases
where the surface near cells are compressed yielding a few layers of cell walls
along the outer skin, it is not suitable for carrying high loads, as required for struc-
tural applications. Casting aluminum around a foam can create components where
a low density foam core is completely surrounded by a massive outer shell. The
shell can be designed in such a way that additional functions besides its load carry-
ing ability can be fulfilled. As only one processing step is required to produce such
a foam core component, production is expected to be very economical.
Potential applications for this kind of foam core castings are space frame nodes,
knuckles, control arms, cross-members, and stiffness providing structural compo-
nents. The described method also offers the potential to produce quasi-hollow cast-
ings where conventional salt or sand cores are replaced by permanent foam cores
in complex geometries.
The various casting methods for the massive outer shell differ by the acting pres-
sures, the die-filling time, and the solidification time, which is mainly determined
by the temperature of the die. Generally, casting of the shell is not completely
straightforward due to the limited compressive strength of the foam cores and
the very thin surface skin, which is not without flaws [1]. A further difficulty results
from the fact that methods have to be developed for fixing the cores.
The following text is based on three publications: squeeze-casting, Kretz and
Kaufmann [2]: high-pressure die-casting, Heinrich, KoÈrner, and Singer [3,4]; and
gravity die-casting, Simancik and Schoerghuber [5].
3.3.1
Foam Cores for Encasing by Casting
3.3.1.1 Core Production

A prerequisite for encasing by casting is a dense foam surface. At the moment
there are two production routes resulting in a solid surface skin, the powder com-
pact foaming, like Alulight and the FORMGRIP process. Both processes comprise
baking of a foamable precursor material in a mold. They differ in the way the pre-
cursor material is produced (Section 2.1). Up to now the aluminum foam cores
used for encasing were exclusively produced via the powder compact technique.
There are two different methods: the furnace baking method and the injection-
molding technique [5]. It depends on the core geometry whether the first or the
second method is more suitable. Generally, for very complex cores the injection
molding technique is particularly suitable. Another difference between the two
methods is the fact that the surface skin of the foams produced via the injec-
104 3.3 Encasing by Casting
Figure 3.3-1. Foam core inserts produced

by compaction of aluminum powder
mixed with TiH2: a) for sand casting
(ca. 300 cm3, baking furnace) [2],
b) for squeeze casting (baking furnace) [2],
c) for die casting (baking furnace) [3,4],
d) for gravity casting (injection molding)
[5].
tion-molding technique is thicker than that of the furnace baking method. This is
an advantage for encasing but it also increases the weight of the core. Figure 3.3-1
shows several core geometries used for different encasing techniques.
3.3.1.2 Core Attachment

A suitable core attachment system has to perform two functions: first it has to keep
the core in place in the die when the die is open and during the movement of die
closing. The more important second function is to maintain the desired distance
between the foam core and the die wall during the casting process because this dis-
tance determines the wall thickness of the castings.
One method to realize a suitable attachment of the core is by creating elongated
wedge-shaped spacers during the foaming process of the cores (Fig. 3.3-2).
These spacers have to be long enough to transfer the acting forces to the foam
core without damage. It is very important to place the spacers corresponding to the
melt flow to avoid flow barriers resulting in ªdead zonesº. This is especially true for
die- and squeeze casting, where the melt freezes quickly.
Figure 3.3-2. Al foam core insert

with spacers to prevent movement in
the die during casting of the shell [4].
3.3.1.3 Mechanical Properties

The aluminum alloys used for the foam cores are wrought or casting alloys. Gen-
erally, the porosity of the surface, from the appearance of cracks or open cells, is
higher for the casting alloys. On the other hand, due to their smaller cell size, cast-
ing alloys are more tolerable against penetration of melt. This is especially true for
small samples.
Besides defects already present there are two important properties of the cores
that determine the processing parameters during high-pressure casting: the
foam compression strength (global strength s CG) and the compression strength
of the surface skin (local strength s CL). The global strength s CG (Fig. 3.3-3) defines
the maximum allowed pressure in the process without foam compression [3].
The samples prepared of the wrought alloy (Fig. 3.3-3) show a significantly
higher scatter of the values compared to the casting alloy. This is a result of the
small sample thickness of 10.5 mm and the coarser pore structure of the wrought
alloy foams in the present case.
It is observed that infiltration of foam cores already occurs at pressures that are
much lower than s CG. This is either the result of open pores in the surface or the
product of a local fracture of the surface skin. As a consequence, an indefinite pore
volume is infiltrated. The local compressive strength s CL can be measured by a
compression test where a liquid (water) exerts a homogenous pressure on a
foam surface of 25 mm q 25 mm (Fig. 3.3-4) [4].
From simple mechanical considerations one finds that the maximum bending
stresses acting in the cell walls that make up the surface are proportional to
p(D2/d2) t p(1/d2r2)
where p pressure, D cell diameter, d cell-wall thickness, r foam density.
As a result, the local compressive strength s CL increases with increasing density.
Figure 3.3-3. Global compressive strength s CG versus density of several Al foam samples from
plates produced via the furnace-baking method [4].
Figure 3.3-4. Average values of the local compressive strength versus density of several foam
samples produced via the furnace-baking method [4].
There is a strong scatter of strength values due to the inhomogeneous cell struc-
ture of the foams.
The results in Fig. 3.3-4 demonstrate that the strength of the foam surface skin
has to be improved for squeeze and die casting, which are characterized by a very
high pressure. For gravity die casting the global strength as well as the local
strength are not so important.
3.3.1.4 Coating of the Foam Cores

In order to obtain reproducible and sound castings, foam infiltration has to be
avoided. That is, for high-pressure casting s CL has to be improved. This can be
achieved by reinforcing the foam surface by an additional coating. The coating
method has to meet two requirements: it has to be applicable to complex geome-
tries and in reasonable thickness. This is fulfilled by a plasma-sprayed coating of
Al99.5 or a coating made from a ceramic slurry. By increasing the thickness of
the coating the local compression strength is increased until the global compres-
sion strength is reached. For even higher thicknesses the foam is no more dam-
aged by local breakage of cell walls but by global compression (Fig. 3.3-5). Experi-
mentally it was found, that for coatings with a thickness higher than 350 mm local
damage is completely suppressed and foam damage is by global compression.
Since both, the local and the global strength show the same dependence on the
foam density the critical thickness of the coating is supposed to be independent
of the foam density.
a) b) c)
Figure 3.3-5. Optical micrographs showing breakage of the surface at high pressure levels;
foam inserts after the local compressive test: c) Al plasma sprayed coating, d 200 mm, local
a) no coating, surface damage occurs at low damage and overall collapse by compression
pressure level preferentially at locations with deformation, d 400 mm, compression [4].
large cells; b) ceramic coating with d 300 mm,
3.3.2
Shell Casting Processes
Depending on the used casting process, different requirements have to be fulfilled

by the foam cores and the processing conditions. The difficulty lies in producing
sound casting shells without infiltration or deformation of the foam core. In addi-
tion, local melting of the surface skin has to be avoided. It was found in former
experiments made in sand and permanent mold casting, that melting of the
skin may occur in particular in the gate area where incoming hot metal can
cause damage to the thin foam skin by heat and erosion [2]. This effect can be
avoided by surface coatings or by use of foam cores with a thicker skin. Available
coatings are lubricants or die coatings, which are in use in the foundry industry to
cover molds and sand cores before casting.
In the following two different approaches are described in more detail: pressure-
less casting and high-pressure casting.
3.3.2.1 High-Pressure Casting Processes

The difficulty during high-pressure casting processes is to avoid infiltration and de-
formation of the foam cores. For this reason, controlled solidification and pressure
build-up are necessary to ensure that the melt solidifies on the surface of the alu-
minum foam core and that growth of a cast skin occurs before the maximum pres-
sure is applied so that the cast shell itself is able to bear the load of casting
pressure.
Squeeze Casting [2]

The squeeze-casting process (Fig. 3.3-6) is characterized by slow, high-pressure
bottom-up filling of the die and provides similar die-filling conditions as in low
pressure permanent mold casting.
The clamping unit of a squeeze-casting machine is similar to that of a conven-
tional high-pressure die-casting machine. The shot unit is placed below the ma-
chine and fills the die vertically. The shot speed is very slow compared to the
high-pressure die-casting process. The shot pressures can be varied from 300 to
1200 bar. This controlled filling is ideal for sensitive inserts such as foam cores.
Figure 3.3-7 shows the arrangement of the foam core in the die. The casting trials
were performed with three foam alloys (Al99.5, AlMgSi1, AlSi10) and two different
materials for the casting of the shell: AlSi9Cu3 and AZ91. Figure 3.3-8 shows the
cross section of an AZ91 magnesium casting with an aluminum foam core. The
gate speed was approximately 0.5 m/s. The melt pressure had to be reduced to a
very low level to avoid foam core deformation and core infiltration. In most trials
infiltration occurred due to the small range of possible parameter variation of the
squeeze-casting machine.
By casting with suitable machine parameters, the UBE squeeze-casting machine
allows the production of castings with wall thickness of 3 23 mm in which the
Figure 3.3-6. Schematic representation of the squeeze-casting process [2].
Figure 3.3-7. Arrangement of the foam core

insert in the squeeze-casting die [2].
Figure 3.3-8. Cross section of a

AZ91 squeeze-casting part with
AlMgSi1 aluminum foam core [2]
(150 x 60 x 39 mm3).
cores are not infiltrated and not deformed. However, the processing window for
achieving good samples without core infiltration or core deformation is a very nar-
row one.
Die Casting [3] [4]

Essential for die casting is controlling dwell pressure with the help of a real-time
control system [3,4]. In the first phase of the conventional die-casting process the
die is filled with melt as fast as possible. In the second phase very high dwell pres-
sures are applied to balance the shrinkage and to minimize porosity of the castings
by compressing the entrapped gases. Because of the high velocity of the casting-
piston a high pressure peak appears at the end of the first phase when the die
is completely filled resulting in massive foam infiltration and collapse (Fig. 3.3-9).
In order to avoid this peak, a real-time controlled die-casting machine is indis-
pensable. With real-time control it is possible to control the casting velocity and
the casting pressure at every step of the casting process. That is, the piston can
be decelerated at the end of the filling phase to prevent the pressure peak.
To use a real-time controlled machine an accurate dosage system is required. The
castings were performed with the standard casting alloy AlSi9Cu3. The melt tem-
perature was about 720 740 hC, die temperature about 250 270 hC. Casting inves-
tigations were carried out on two demonstration parts (Fig. 3.3-10).
Figure 3.3-9. Comparison of dwell pres-

sure with and without real-time controlled
die-casting machine [4].
Figure 3.3-10. Two different demonstration parts with foam core inserts encased by die casting:
a) trapezoid, b) plate [4].
As core materials the casting alloy AlSi10 and the wrought alloy AlMg1 were
used. The upper bound of the pressure inside the die is limited by the global com-
pression strength of the foam core. It is about 60 100 bar and can only be raised
by increasing the foam density.
a)
Figure 3.3-11. a) Infiltrated pore vol-

ume versus specific casting pressure
of several casting experiments with
coated and uncoated foam cores. b)
Porosity in the cast outer skin versus
specific casting pressure of several
casting experiments with coated and
b) uncoated foam cores [4].
Casting experiments with coated and uncoated foam cores were performed for a
plate-like geometry with different specific casting pressures pNS. The specific cast-
ing pressure is the operating dwell pressure in the casting chamber. It has to be
taken into account that the actual acting pressure inside the die is much lower
because of friction. Figure 3.3-11 shows the infiltrated pore volume VI and the
porosity in the skin versus pNS.
As expected infiltration increases with increasing pNS whereas porosity in the
skin decreases. By using coated foam cores a significant reduction of core infiltra-
tion is possible. In accordance with the tests to determine the local compression
strength, the plasma-sprayed plates show the best results. More recent results
achieved with optimized die design and process conditions (evacuation of the
die) demonstrate that the porosity of the skin can be reduced even to less than
5 % without infiltration of the core. The minimum skin thickness is about 2 mm
for die geometries evaluated so far.
Gravity Casting [5]

The foam core samples (AlSi12 and AlMg1Si0.6) used had a density of
0.8 0.9 g/cm3 and were prepared by injection of the foam into sand molds
(Fig. 3.3-12) using the injection-molding technique. They were inserted into
an existing gravity casting die (steel) with conventional gating and risering.
The casting alloy AlSi9Cu3 at 740 hC was used to cast the outer shell around
the foam cores.
In contrast to die- or squeeze casting the acting pressures during gravity casting
are insignificant and much smaller than the local compressive strength of the foam
cores. That is, as long as there are no large cracks in the foam surface infiltration or
compression due to high pressures is not observed during gravity casting. No melt-
ing of the foam core could be detected in the experiments performed (Fig. 3.3-13).
Pouring liquid metal around the foam core results in heating up and consequent
expansion of the gas inside the pores. In contrast to the high-pressure casting
methods the gas is not prevented from diffusing from the core into the melt
where it forms bubbles. That is, the expanding gas can cause porosity in the
shell or even destroy the integrity of the casting because of gas bubble formation.
This expansion of the gas and the deterioration of the shell can be minimized by
preheating the cores before insertion.
Figure 3.3-12. Pressure-less casting: Foam core, encased core, and cross section with foam core
(height of the foamed part 25 cm, maximum width of the cross section 12 cm) [5].
Figure 3.3-13. Aluminum casting with different foam cores: left, AlSi12 foam; right, AlMg1Si0.6
foam (diameter of the foam filled tubes 6 cm) [5].
Another advantage of preheating the cores is that cold shuts, which appear in
regions of the castings with the lowest thickness of the shell, can be avoided.
The preheating of the cores leads to a lower heat flow from the melt into the
core, thus enabling reduction of the wall thickness in comparison to usual sand
cores, which can not be preheated. With preheating temperatures of 400 hC the
minimum shell thickness is found to be about 1.5 mm.
3.3.2.2 Bonding Between Shell and Foam Core

Generally, no bonding develops between the core and the shell during casting be-
cause of the continuous aluminum oxide layer that prevents the core surface from
reaction with the molten metal [5]. The same effect occurs using aluminum cores
in magnesium squeeze castings [2]. The contact time appears to be too short for
significant reactions between the aluminum oxide skin and the magnesium
melt. There are two possible ways to improve the bonding [5].
x Mechanical bonding by flow of liquid metal into the outer foam structure sup-
ported by intentional weakening of the surface skin, e. g., by sand-blasting. Dis-
advantage are that the weight of the casting increases and the bonding occurs
only locally and is difficult to control.
x Metallurgical bonding by coating the cores with various agents supporting diffu-
sion through the aluminum oxide layer. With a suitable metallic coating a metal-
lurgical bonding can be achieved. On this point, however, further research is ne-
cessary.
The shrinkage of the cast shell during solidification leads to a pressure fit of the
inserted core. For most applications metallurgical bonding will not be necessary
to achieve the required component properties such as improved energy absorption
behavior or improved damping properties. The absence of bonding may sometimes
even improve structural damping of the part due to additional energy dissipation at
the shell/core interface [5].
References
1. H. P. Degischer, A. Kottar, Metal Foams and Druckgieûverfahrenº in MetallschaÈume 2000,

Porous metal Strutures, J. Banhart, M. F. Vienna, Wiley-VCH, Weinheim 2000.
Ashby, A. Fleck (eds), MIT Verlag, Bremen 4. F. Heinrich, C. KoÈrner, R. F. Singer,
1999, p. 213 ªEncasing of Al Foams by Die Casting as
2. R. Kretz, H. Kaufmann. ªFabrication of Manufacturing Process for Light Weight
Squeeze Castings with Permanent Alumi- Componentsº in Proc. Materials Week 2000.
num Foam Coresº in EUROMAT, Metal Ma- 5. F. Simancik, F. Schoerghuber, ªComplex
trix Composites and Metallic Foams, Munich Foamed Aluminum Parts as Permanent
1999. Wiley-VCH, Weinheim 1999. Cores in Aluminum Castingsº in Porous and
3. F. Heinrich, C. KoÈrner, R. F. Singer. Cellular Materials for Structural Applications,
ªHerstellung flaÈchiger Leichtbauteile durch MRS, San Francisco, CA 1998.
Umgieûen von Aluminium-SchaÈumen im
3.4
Sandwich Panels
J. Banhart, W. Seeliger, and C. Beichelt
For structural applications metal foams are often used in combination with con-
ventional dense metal structures such as sheets, columns, or more complex-shaped
hollow metal structures. This allows for optimized mechanical properties in a
given loading situation [1]. It also facilitates ªhidingº the metal foam inside a closed
and dense structure, which again is advantageous for corrosion protection. Such
composites containing aluminum foam may be manufactured in various ways.
The most obvious and straightforward one is achieved by adhesive bonding of
pre-fabricated aluminum foams and flat face sheets. However, this approach has
certain disadvantages and is not feasible in all cases. An alternative and preferable
method consists in establishing composites during the foaming process. Foam-
filled columns can be produced by inserting foamable precursor material into a col-
umn and then heating. The foam will eventually rise and fill the column. Another
possibility is to start with a foamed part and to coat it with aluminum by thermal
spraying [2] thus establishing a dense outer skin. Yet another way is to use the
foam as a core for die-casting [3,4]. Aluminum foam sandwich (AFS) panels
may be manufactured very elegantly by roll-cladding face sheets to a sheet of foam-
able precursor material, then creating the desired shape in an optional working
step, and by finally foaming the entire composite [5,6] (Fig. 3.4-1). Foaming will
create a highly porous core structure without melting the face sheets if the melting
points of the foam and the face sheets are different and process parameters are
chosen appropriately.
114 3.4 Sandwich Panels
Figure 3.4 -1. Process steps for making sandwich panels with aluminum foam cores.
3.4.1
Sandwich Foaming Process
The formation of a metal foam sandwich may be visualized by making use of an

X-ray technique that has been developed recently [7] and allows for the monitoring
of the internal structure of expanding metal foams. Since the technological imple-
mentation of the production process for aluminum foam sandwiches still suffers
from occasional flaws that can be traced back to inadequate process parameters
and defects in the foamable material, the X-ray investigations may also help to
identify such problems. One example of this is shown in Fig. 3.4-2. The foaming
of this particular sandwich panel was carried out at a furnace temperature of
750 hC. The first frame corresponds to an early stage of foaming. The foamable
core layer already shows a slight absorption contrast to the face sheets, indicating
that some porosity has already formed at this stage. Moreover, a crack may be seen
on the right upper side running right through the foamable layer. The second
frame, showing the situation just 5 s later, reveals that the foaming of the core
layer takes place in a highly non-uniform way. The restricted heat flux through
the face sheets leads to a temperature gradient and triggers the foaming process
near the interface of face sheet and core layer. As can be seen from Fig. 3.4-2,
the crack in the precursor material has deepened after the initiation of the foam-
ing process and still extends over the entire foam layer. After 22 s of ongoing
foaming, however, the core layer is fully expanded and the crack has disappeared.
Therefore, this type of defect does not lead to an obvious defect in the foam sand-
wich.
In order to obtain a complementary view of expanding metal sandwich struc-
tures, metallographic images have been made of samples that were foamed to a
given expansion stage and were then quenched. Three of these images are
shown in Fig. 3.4-3. The unfoamed sample shows a sharp boundary between
the foamable core (characterized by the angular-shaped gray silicon particles em-
bedded in the light aluminum matrix) and the dense face sheets to the right.
The foamed sample in the middle, which is at a stage corresponding approximately
to that of full expansion, shows the typical microstructure of an under-eutectic alu-
minum-silicon alloy. The light aluminum-rich grains surrounded by the eutectic
phase can be easily identified. The dense face sheet is virtually pore-free and
Series of radioscopic images of an

Figure 3.4 -2.
expanding AlMn1/AlSi7-foam/AlMn1 sandwich [8].
Foamed at a furnace temperature of 750 hC.
shows no structure in the low magnification chosen. The interface of foam and
face sheet lies on a straight line and is well defined.
Finally, the foamed sample on the right hand side of Fig. 3.4-3 represents an
even later stage of expansion. It exhibits a notably coarser grain size distribution
in the foam and a slightly diffuse boundary between foam and face sheet. The
eutectic phase has grown into the former face sheet material by diffusion pro-
cesses and has locally amalgamated with the face sheet alloy: one reason for
the excellent bonding between foam cores and face sheets in properly manu-
factured aluminum foam sandwich parts and the explanation of the absence
Figure 3.4 -3. Metallographic images of sand- shortly before maximum expansion; right)
wich structures with an AlSi7 core. Three dif- sandwich at the onset of face sheet melting.
ferent expansion stages are shown: left) un- The width of each image is 0.75 mm.
foamed precursor material; middle) sandwich
of face sheet delaminations in tests of the tensile strength of sandwich

structures.
3.4.2
Industrial Application
On the basis of technological advantages and limitations, application strategies

have to be developed in order to simplify the decision whether or not AFS
components should be used in a certain production process and to avoid unrealis-
tic decisions. The application of light-weight materials often implies higher costs
and, compared to steel, eventually a loss of stiffness and an increase of ma-
nufacturing problems. For a technically and economically successful appli-
cation of AFS components, a new approach to vehicle body architecture is required
[9].
For an application in the ªBody in Whiteº technique (BIW) about 90 % of the
current design concept requires a complete change. The space frame, for in-
stance, should be designed considering the stiff AFS components in a way that
the special properties of AFS may be optimally exploited. A simple replacement
of steel parts by AFS parts will not suffice since the benefit of the stiffened planar
surfaces are not employed efficiently. Therefore, a new BIW architecture must
be developed. Examples for such new concepts are shown in Fig. 3.4-4 and
Fig. 3.4-5.
Whenever replacing conventional materials by AFS, it must be taken into consid-
eration that the range of AFS properties includes some characteristics that have
previously been achieved by additional parts. For example, the use of AFS for
the fire wall could imply the elimination of heat shields and the associated connect-
Figure 3.4 -4. ªDetroit Show Car 1998º developed by Karmann (Osnabru Èck) for demonstrating
potential uses of aluminum foam sandwich panels. The rear bulkhead and the front firewall (not
directly visible) are made of AFS.
Figure 3.4-5. ªEUROC 99º concept racing car: left) CAD model of entire car; right) space-frame
structure with AFS parts.
ing parts. The application of AFS parts may also lead to an elimination of noise
attenuation materials because of the low structure-borne sound characteristics of
AFS.
Another area of interest to be investigated in the future is that of exterior panel
closures, e. g., doors, hoods, and decklids. This application depends on the
achievement of a Class-A surface with stamped AFS panels. Taking a hood
stamped out of AFS as an example, there is no longer any need for an inner
panel due to the inherent stiffness of the AFS outer layers, thus reducing ex-
penses for material, tools, and the complete assembly. In spite of higher material
expenses, an AFS bonnet may (depending on the shape) be more cost-effective
than a steel hood up to a production volume of 100 000 units. This is attributed
to the reduction of manufacturing and tooling expenses. This means that light
weight AFS constructions may be applied economically in low and middle pro-
duction volumes due to the reduced investment costs compared to conventional
steel components.
In general applications, possible additional costs of AFS panels should be com-
pared to the advantages that can be expected. Even if only one of the improved
characteristics of AFS is required for a particular application, the associated in-
crease of expenditure may be of secondary interest. In general, the application of
AFS components will increase overall material costs, so that an enhanced perfor-
mance must be achieved as justification. In the future, this situation may change
as vehicle operating costs gain more significance due to the higher energy costs. As
a consequence, this implies that the amount of money available for the use of light
materials, which is right now about 2.5 5 Euro per kg weight reduction, will
increase.
3.4.2.1 Technological Benefits

Besides the advantages already known (the combination of high torsional stiffness
with low weight) further properties of AFS must be taken into account as they also
may have a major impact on the implementation strategy.
Acoustic Properties
Considering the customers comfort requirements, a new lightweight body mate-
rial is not allowed to show inferior acoustic properties than steel. Therefore, the
good acoustic properties of AFS should be emphasized. Especially cars with alu-
minum bodies show poor damping properties. To improve this situation, a large
amount of damping material must be added, thus sacrificing the mass saving po-
tential.
A joint research by the TU Dresden and Karmann has revealed that AFS offers a
significantly better acoustic behavior especially in the range 50 400 Hz. Additional
insulation measures may be reduced: weight is saved. The acoustic performance of
special insulation materials will of course not be attained, so that AFS cannot ex-
clusively serve as a sound damping material.
Thermal Properties
Thermal conductivity is another important aspect when selecting body materials.
Due to the entrained air bubbles, the heat transport capabilities of foams are
low. Depending on the density, the thermal conductivity of AFS is reduced to
1/12 1/20 of the conductivity of bulk aluminum. Furthermore, AFS satisfies
most of the fire protection regulations. No adhesives are contained and the AFS
components maintain their shape up to the melting point of 600 hC and in
some cases even above. The exceptional welding characteristics together with mini-
mal distortion justify to characterize AFS as ªthermally very stableº.
Robustness
Sandwich components, as known from aerospace technology, are relatively vulner-
able in impact situations. Even small damages of honeycomb panels may lead to a
complete breakdown of the core panel structure. As a result of the metal link be-
tween core and panels, this does not occur in an AFS panel. Cracks may only occur
in the core and their expansion is limited. A delamination of core and panels has
not been detected with parts manufactured and tested to date. This is very impor-
tant since structural body components are not subject to special checks during the
product life cycle
Other Properties
Additional properties of AFS that enhance product performance are: good energy
absorption, recyclability, and low manufacturing time periods for the sandwich compo-
nents. The foaming process, e. g., takes only 30 45 s even for large parts. There-
fore, mass production with a comparably low number of parallel production lines
is possible.
3.4.2.2 Technical Limitations

When selecting applications for AFS components, their formability and geometry
after the foaming process must be taken into consideration. A constant component
thickness may only be achieved with plane sheets. Complex formed structures will
Figure 3.4 -6. limitations of manufacturability of AFS.
have a variable thickness in different areas of the component. However, these

thickness variations are predictable and may be adjusted to component loads by
simulations, since the upper layers maintain their original geometry during the
foaming process. The form tolerance after foaming is e1 mm. Drills and trim
cuts will be performed with the help of a trimming/calibration tool, so that addi-
tional reference surfaces and flanges may be established. U-form shapes should
generally be avoided, as they show an adverse relation between the side and
base surfaces, which leads to differences in the thickness of the foam (Fig. 3.4-
6). This effect is primarily due to the stiff inner layer, which should shorten up
when the thickness is constant.
There are limitations to the determination of the gage relation between outer
layers and core due to current processing techniques. The minimum thickness
of the outer layer is limited to 0.6 mm. A skin thickness less than this may
cause a degradation of the alloy during foaming and is undesirable. Under
optimum process conditions, the foam layer may expand up to the extent of
seven times of its original thickness. The maximum achievable height ranges
from 25 30 mm. Thicker foam cores should be avoided because of the differ-
ent cooling rates within the foam, which may result in a non-uniform core por-
osity.
3.4.3
Joining Technology of AFS
In order to be able to efficiently exploit the advantages of AFS technology, it does

not suffice to exchange highly stressed parts with AFS parts. AFS technology re-
quires special construction as well as joining techniques, adjusted to the character-
istics of sandwich parts. The joining of AFS parts may be attained by a variety of
possible processes. The most important ones are given in Table 3.4-1.
Table 3.4 -1. Overview of joining techniques for AFS.
Process Details
Laser welding CO2 -laser, Nd:YAG
TIG welding by hand, partly mechanized, by robot

MIG welding
Pin/bolt welding by hand, mechanized
Punch riveting AFS/aluminum; AFS/steel
Riveting nuts/screws M4 M8
Flow drilling M4 M8
Riveting blind riveting, splay riveting
Bonding 1-K and 2-K systems
3.4.3.1 Laser Welding

The process of laser welding is suitable for series production. In order to weld face
sheets of 1.2 mm 3 kW power are required. By bluntly welding sandwiches, only
the face sheets are bonded, while the core layers remain unaffected. As shown
in Fig. 3.4-7a the filigree cell structure is not damaged by the local thermal impact.
In welding flat AFS sheets with linear joints, a maximum speed of 10 m/min has
been achieved. If both face sheets have to be bonded, usually the part has to be
turned thus decreasing the welding speed to a rate less than 5 m/min.
3.4.3.2 TIG/MIG Welding

Welding of AFS parts by common welding techniques is also possible. The princi-
ple techniques worth mentioning are TIG- and MIG-welding (Fig. 3.4-7b). Both
techniques are exceptionally well suitable both for joining two AFS parts and
also AFS parts with aluminum parts. The advantages of these techniques include
flexible application possibilities, the amount of experience in this field and the low
investment costs. The high degree of stiffness of AFS sandwich parts, their low
thermal conductivity and the resulting low thermal distortion rate minimize the
necessary efforts for clamping and fixing the parts. As in laser welding, only the
face sheets have to be welded. Welding rates for manual TIG welding reach
0.3 m/min, those for MIG welding reach 0.8 m/min. Partly mechanized welding
with a linear carriage may achieve a rate up to 1.3 m/min.
Figure 3.4 -7. Different

welding techniques for AFS:
a) laser welding, b) TIG
welding, c) bolt/pin weld-
ing.
a)
b)
c)
3.4.3.3 Bolt/Pin Welding

Another joining technology of considerable interest for AFS sandwiches is that of
welding bolts (Fig. 3.4-7c). These bolts do not transfer high forces but are em-
ployed for fixing cable bundles and wires or as mass contacts. The welding process
is completely controlled and monitored with the help of a welding head with a lin-
ear motor. This technology allows for the regulation of variations of the thickness
of the sandwich. As with the other techniques, the core layer remains intact in the
welding area. Even welding a bolt directly onto a flaw, e. g., a void or a large pore
has no impact on the quality of the bonding as the joining zone is obviously lim-
ited to the face sheet.
3.4.3.4 Punch Riveting

Due to the increase of mixed constructions and the problem of thermal impact on
the structure by assembling parts by welding, this joining technology has been
strongly forced in the recent years. Since 1994, Punch Riveting is applied in
construction of the AUDI A8 and has reached its momentary peak in the construc-
tion of the A2 with 1800 die casts per car. Figure 3.4-8a shows the joining area as
well as a cut of the fusion. Studies of the settling properties of the core layer have
shown another positive characteristic of AFS sandwiches. If AFS is highly com-
a)
b)
c)
Figure 3.4-8. Different joining techniques for AFS: a) punch riveting, b) riveting nuts and screws,
c) flow drilling.
pressed, its tensile strength drops to a value of 50 % plastic deformation, but re-
gains the original mechanical values of the nondistorted sandwich at maximum
compression. This property is a result of the increasing mechanical clutching of
the collapsed cell structures and the likewise increasing friction. The mechanical
values of compression strength and shear stress do decrease in this range but
they stabilize again at a low level under static as well as under dynamic load.
3.4.3.5 Riveting Nuts and Screws

Another possible joining technique consists in the placing of riveting nuts and
screws (Fig. 3.4-8b). The joint cannot transfer high forces. It is rather employed
to fix holders and devices. The size of the nuts and screws may range from M4
to M10, depending on the thickness of the AFS parts and the face sheet.
3.4.3.6 Flow Drilling

Flow Drilling (Fig. 3.4-8c) is an alternative to riveting nuts. Frictional heat is gen-
erated by a multi-polygon that is pressed onto the face sheet in axial direction at a
high revolution rate/speed. The material plastifies and becomes easily formable.
One will get a defined drill-hole and the material of the face sheets will flow
into the core. The length of the formed hole wall will be three to five times the
thickness of the face material. The minimum revolution rate for flow drilling
lies at about 2400 rpm at a spindle moment of 1.5 kW. A coated thread cutter is
used at a revolution rate of 500 rpm. The lifetime of the flow former and the thread
cutter amounts to 10 000 drills and threads. But so far only feasibility studies have
been carried out.
3.4.3.7 Riveting
AFS parts may also be riveted. Especially in mixed constructions, riveting is very
well suitable. It is important to choose a relatively large diameter of the rivet
head since too small a diameter leads to a fastening pressure of the rivet that com-
presses the core. Due to the high surface pressure one will find plastic deforma-
tions on the face sheets.
3.4.3.8 Bonding
AFS with face sheets of aluminum may be bonded with the same technology as
conventional aluminum sheets (Fig. 3.4-9). The same parameters have to be re-
spected as there are the creation of a defined surface, a construction adapt for
bonding and especially the choice of an adequate bonding system. The strength
of today's bonding systems partly exceeds the physical values of the aluminum
foam cores. One of the main advantages of bonding lies in the optimum transmis-
sion of the applied forces. This technology allows the AFS parts to be excellently
integrated into the surrounding structure.
Figure 3.4 -9. Bonding of AFS
3.4.4
Cutting of AFS
Cutting of AFS structures, the unfoamed precursor material as well as the foamed
sandwich panels, can be quite a challenge, especially in the latter case. Here, con-
ventional mechanical cutting techniques cannot be applied in a straight-forward
way because of the danger of uncontrollable deformation of the material. There-
fore, two alternative methods have been evaluated.
3.4.4.1 Laser Beam Cutting

Laser beam cutting of the unfoamed material has been successfully evaluated
[10±12]. A high precision of cutting into the desired net-shape without the need
for further processing was achieved (see Section 3.1).
Cutting of the foamed materials required a special adaptation of process para-
meters to take account of the specific nature of AFS. Both face sheets have to be
cut simultaneously. Moreover, the low-density foam tends to melt more than the
face sheets leading to the deposition of metal and dross at the opposite side of
the sandwich panel, which lower the quality of the lower cutting edge [13]. Good
results with cutting a 12 mm AFS could be obtained by adjusting the laser
power to 5 kW and the cutting speed to 0.8 m/min.
3.4.4.2 Water Jet Cutting

Water-jet cutting of unfoamed AFS precursor is possible without any problems.
Cutting of AFS sandwiches, however, imposes the problem, that abrasive particles
remain in the pores after cutting and cannot be entirely removed even by repeated
swilling with water or solvents. Such contaminations are not acceptable in cases in
which the sandwich panels have to be varnished as they would lead to inferior sur-
face qualities [14].
References
1. M. F. Ashby, A. G. Evans, N. A. Fleck, L. J. 9. W. Seeliger, Entwicklung und Program-

Gibson, J. W. Hutchinson, H. N. G. Wadley, mierung eines Materialmodells fuÈr elasto-
Metal foams: a Design Guide, Butterworth- plastische MetallschaÈume, Thesis, University
Heinemann, Oxford 2000. of Bremen, MIT-Verlag, Bremen 2000.
2. M. Maurer, E. Lugscheider, Materwiss. 10. W. Seeliger, ªFertigungs- und Anwen-
Werkstofftechn. 2000, 31, 523. dungsstrategien fuÈr Aluminiumschaum-
3. F. Heinrich, C. KoÈrner, R. F. Singer, Mater- Sandwich Bauteile, AFS (Aluminum Foam
wiss. Werkstofftechn. 2000, 31, 428. Sandwich)º in Proc. IIR Werkstoff-Kongress,
4. T. HoÈpler, F. SchoÈrghuber, F. SimancõÂk, in 22 23 Jan 2001.
Metal Foams and Porous Metal Structures, 11. R. Braune, A. Otto, in Metal Foams
J. Banhart, M. F. Ashby, N. A. Fleck (eds), and Porous Metal Structures, J. Banhart,
MIT-Verlag, Bremen 1999, p. 79. M. F. Ashby, N. A. Fleck (eds), MIT-Verlag,
5. J. Baumeister, J. Banhart, M. Weber, German Bremen, 1999, p. 119.
Patent 44 26 627, 1994. 12. R. Braune, A. Otto, Materwiss. Werkstoff-
6. J. Baumeister, in Metal Foams and Porous techn. 2000, 31, 562.
Metal Structures, J. Banhart, M. F. Ashby, 13. R. Braune, C. Beichelt, in Lasertechnologien
N. A. Fleck (eds), MIT-Verlag, Bremen 1999, im Automobilbau, Stuttgart, 29 Feb 1 Mar
p. 113. 2000.
7. J. Banhart, H. Stanzick, L. Helfen, T. 14. C. Beichelt, in MetallschaÈume Verfahren
Baumbach, Appl. Phys. Lett. 2001, 78, 1152. und Komponenten fu È r den Leichtbau, J. Bau-
8. J. Banhart, H. Stanzick, L. Helfen, T. meister (ed), Workshop, Bremen, 7 8 Mar
Baumbach, Adv. Eng. Mater. 2001, 3, 407±411 2001.
(2001).
4
Characterization of Cellular Metals
B. Kriszt
We are familiar with natural cellular structures such as wood and bone: these nat-
ural materials are structured hierarchically.
In engineering, man-made cellular materials (porous construction materials and
cellular polymers) have been used for a long time because of their remarkable ther-
mal insulation. A new group of cellular materials has been invented more recently:
the cellular metals. When we study different man-made cellular materials we soon
recognize that they also show a hierarchical architecture. As in nature, every pro-
duction process leads to a different cellular structure, so clearly cellular structures
are determined by the process by which they are made. However, when we want to
describe a cellular metal we say Alporas, Alulight, or Inco foam, and do not de-
scribe exact structural quantities.
So before we can characterize cellular metals, we need to understand the signif-
icant parameters necessary to describe the architecture and the hierarchy.
First, we can think of a cellular metal as a heterogeneous composite material of a
metal and a gas. If only the metal matrix is observed, the microstructure consisting
of grains, precipitates, dendrites, or different phases can be seen, but there is no
information on the architecture of the cellular metal. This hierarchical level is
often called the microstructure of the metal.
The density of the cellular metal depends on the volume fraction of metal and
gas. Consequently, cellular metals are characterized by their density compared to
the parent metal: the relative density. Typical metal foams have relative densities
of less than 0.3. Typical porous metals have relative densities greater than 0.3
but less than 1.The density of cellular metals, in absolute units or as relative den-
sity, is one of their main characteristics. From the density of a cellular metal part,
we have an instrument for describing the material at the ªmacroº level.
The second main characteristic is the cellular architecture. Often fluctuations in
cellular architecture lead to inhomogeneity in density, so the local density distribu-
tion can describe a cellular metal more precisely. This way of describing cellular
structures belongs to the ªmesoº level.
128 4 Characterization of Cellular Metals
However, the definition of macro, meso, and micro is not yet standardized: in
Section 4.1 a scheme is discussed that helps to clarify the classification of the hier-
archical structure.
Real cellular metals do not have perfect ideal cells, but the analysis of ideal cells
is nevertheless helpful for describing the architecture of cellular metals. The ideal
cells model describes the whole structure in terms of at least one space-filling
three-dimensional unit cell; in most cases we can think of polygonal cells. For ex-
ample, honeycombs are networks of channels having hexagonal cross section; me-
tallic foams, sponges, or lattice block material are representatives of real 3D poly-
gonal cellular metals.
Cellular structures in which adjacent cells are separated by cell walls or faces are
called closed-cell foams or metals: the other group, open-cell foams or metals, are
permeable. Sometimes this classification is not clear, because in some cellular me-
tals both closed cells and open cells can be observed.
To get a feeling for a 3-D structure, we conceptualize a space-filling unit cell of
an ideal structure (Fig. 4-1). From a geometric point of view it is helpful to think of
a cellular structure formed by a skeleton of cell edges (struts), which merge at
nodes (vertices).
One space-filling unit cell is the Kelvin tetradaidecahedron cell. This cell has a
low surface area per unit volume. Most foams made by expanding a gas in a liquid
tend to form such a cell shape. For this reason, this cell type is often chosen for
modeling cellular metals [1 6]. Knowing the geometry of the unit cell allows us
to calculate the relative density of the cellular metal. From the example of the Kel-
vin cell, we can elaborate the main structural features used for characterization.
Certain parameters are needed to describe the cell shape, the cell size, the topol-
ogy, the geometry of cell struts, nodes, and walls, and their relation to neighboring
cells. Often it is helpful to define also parameters that describe the imperfections
or the defects of a cellular material compared with the ideal unit cell (see
Section 4.1).
A long list of parameter can be defined, but we have to ask critically if these serve
a useful purpose. If we want to understand relations between processing para-
meters and cell structure (Chapter 2, and Section 4.1), as well as between cell struc-
ture and the property profile (Section 4.1 and Chapter 5), we need to investigate
and quantify the cells and their imperfections as precisely as possible.
Another task that requires structural parameters is modeling the mechanical be-
havior of cellular metals (Section 4.1 and Chapter 6). Input data derived from char-
acteristic cell features of real foams are needed for generating virtual cellular net-
works, such as the Voronoi network, for examples.
Figure 4-1. Space-filling ideal tetrakaidecahedral unit

cell (Kelvin cell) showing cell walls, edges, and nodes.
Typical metallographic methods are used to characterize features of metallic

cells. Ordinary methods such as optical microscopy and quantitative image analysis
give information on pore size, cell-wall thickness, orientation, and neighboring cor-
relation (Section 4.1).
Another task for characterization (which might become more important) is the
inspection of the quality of cellular components: the definition of which is not
yet established (Section 4.3). The basis of any specification is the apparent density.
A rough estimate of density can be gained by weighing a sample of known volume.
A more accurate result can be obtained using the principle of Archimedes. Unfor-
tunately, not all components have a closed surface, so the penetration of the test
liquid has to be prevented or it will invalidate the result.
The local density distribution of cellular components can be determined by X-ray
radiography or computed tomography (XCT) (Section 4.2).
A local density can be determined to check if the distribution of the metal is
homogeneous; if not there might be hard or soft regions in the component. Tech-
nical XCT scanners having a spatial resolution of up to ten microns allow us to
study the cell structure and the defects in the cells. Besides XCT, other nondestruc-
tive testing methods such as ultrasonic testing, eddy current measurements, or
lock-in thermography are under consideration for the inspection of density distri-
bution and defects.
From this introduction it can be concluded that structural analysis is the basis of
cellular metal and foam research. We seek relationships between the structure and
the processing technique, the property profile, and the description by modeling.
Chapter 4 will give an overview of typical structures of metallic foams and show
how structure influences the property profile (Section 4.1). In Section 4.2 an out-
line of XCT, a typical nondestructive testing method for cellular material, is given.
Based on XCT technology, quality features can be derived, which are discussed in
Section 4.3.
References
1. L. J. Gibson, M. F. Ashby, Cellular Solids: 4. N. J. Mills, A. Gilchrist, J. Appl. Mech. 2000,

Structure and Properties, 2nd ed, Cambridge 122, 67.
University Press, UK 1997. 5. J. L. Grenstedt, K. Tanaka, Scripta Mater.
2. M. J. Silvia, L. G. Gibson, Int. J. Mech. Sci. 1999, 40(1), 71.
1997, 39(5), 549. 6. T. Daxner, H. J. BoÈhm, F. G. Rammerstorfer,
3. W. E. Warren, A. M. Kraynik, J. Appl. Mech. R. Denzer, M. Maier, Materwiss. Werkstoff-
1997, 64, 787. techn. 2000, 31, 447-450.
130 4.1 Characterization of Cellular and Foamed Metals
4.1
Characterization of Cellular and Foamed Metals
B. Kriszt, U. Martin, and U. Mosler
It is known from modeling of ideal cell structures and demonstrated by experi-

ment, that the real cell structure of a cellular material is the key factor that deter-
mines its properties. This fact leads to the need for characterization of cellular ma-
terials. The objective of characterization is to define the relevant structural para-
meters and to understand their influence on the properties of the cellular material.
In this section the main structural features are described and their influence on
properties is shown.
4.1.1
Definition of Structural Features of a Cellular Metal and Influence on Property Profile
As can be seen in Fig. 4.1-1, we can define a list of more than 30 structural para-
meters for describing the geometrical structure (architecture) and the microstruc-
ture of cellular materials.
As already discussed in the introduction of Chapter 4, cellular metals are hetero-
geneous materials, like composites, because the pores are embedded in the metal
cell skeleton. The shape and arrangement of the cells is called the geometrical
structure or architecture. In some cellular metals the geometrical architecture
Figure 4.1-1.List of parameters for describing structure of metallic foam; depending on the re-
solution, geometrical structure and microstructure can be defined.
can be observed visually, so the length scale of cell size can range from some tenths
of millimeters to some centimeters. For acquiring images of the outside appear-
ance of the geometrical structure either digital video cameras or ordinary scanners
can be used. Typical magnifications that can be reached with these systems are up
to tenfold.
If we zoom in on the cell material, we will see another hierarchical level or sub-
structure, the microstructure of the cellular skeleton. We can observe the structure
of metallic cell material, micropores, and cracks at a resolution of microns by using
light microscopy or scanning electron microscopy. If transmission electron micro-
scopy is used, the lateral resolution is increased to nanometers and we may observe
dislocation structure or small precipitates.
From this, it becomes clear that cellular materials are more than heterogeneous
materials: they are structured hierarchically. When describing cellular materials,
one has to be aware that different structural features that can influence the proper-
ties of the cellular material can be defined for each hierarchical level.
Understanding such a complex system is rather difficult. As a consequence, no
general material rules based on real structures can be developed to explain the in-
fluences of structure and architecture on properties. Furthermore, no guidelines
for the characterization of cellular metals have been established. In recent years
structural features that influence the properties have been identified but their in-
teractions and effects on real metal foams are not known quantitatively.
4.1.1.1 Density and Volume Fraction of Pores

One of the most significant structural characteristics seems to be the density dis-
tribution. A clear relation between relative density, stiffness modulus, and com-
pression plateau stress was derived from the scaling model of Gibson and Ashby
[1]. The scaling laws were verified recently by others [2,3]. Based on the scaling
laws the influence of inhomogeneous density distribution on mechanical proper-
ties is understood quite clearly (see Section 6.2). Originally the scaling law was der-
ived from structural parameters such as length and thickness of edges of a cubic
cell; these structural features were assumed to be constant and can be replaced
by the relative density [1].
Astonishingly, density is not commonly determined in quantitative structural
analysis [4]. The density distribution of real cellular metals is influenced by
many structural features as shown in Fig. 4.1.-2.
The higher the volume fraction of pores in a cellular material, the lower is the
density of the material. So the volume fraction of pores seems to be a significant
quantity. With the exception of CT tomography (see Sections 4.2 and 4.3), investi-
gations of nontransparent 3D structures is restricted to analysis of cross sections.
Assuming certain boundary conditions, the laws of quantitative image analysis give
relations between 2D and 3D structural features [5]. Using these laws, the volume
fraction of pores can be derived from the area fraction of pores. The volume frac-
tion of pores is a global parameter and gives no information about the arrange-
ment, size, or shape of pores. For example, two types of cellular metal can have
Figure 4.1-2. Dependence of density on structural features of pores and cell skeleton.
the same density, but one samples can have one or two big pores that represent
most of the volume fraction of the gaseous phase, and the other sample (for in-
stance made by a replication technique or investment casting) can consist of a net-
work of nearly uniform small pores. Figure 4.1.-3 shows differences in structures
of cellular material.
Figure 4.1-3. a) Closed-cell foam, geo-

metrical structure of Alulight foam pro-
duced by PCF: irregular pore size and
a) shape can be seen.
Figure 4.1-3. b) Open-cell Ni:

regular and uniform pore size
and shape.
b)
4.1.1.2 Shape and Size of Pores

Even if the pore (or cell) size is uniform and the pores are arranged regularly, this
information is not sufficient to describe the cellular metal. Also, pores can take pre-
ferred orientations.
Because of the structural characteristics explained before, two more sets of quan-
titative parameters have to be defined: one to describes the properties of each pore
(object parameters) such as size, shape, or orientation and the second to character-
ize the topological properties of the cellular material, such as arrangement and
neighborship relations.
For each of these sets of features it is possible to define a certain number of
structural quantities. So, for instance, the pore size can be described either by
x an equivalent diameter of the pore, which is defined as the diameter of a sphere
or circle having the same volume or area, or
x as the maximum, minimum, or average length of the pore.
The impact on properties is different; it has been shown that cell size of carbon
foams has an influence on compression strength, but not on stiffness and fracture
toughness [6].
Even if we assume that pore size has an impact on the mechanical behavior, no
experimental work shows how large pores act in real metallic foams, so they have
to be classified as a ªknock-downº factor, or defect, for mechanical properties. It
has been shown that the size of the pores, combined with a certain shape, for ex-
ample an elliptical shape and a certain orientation, can allow deformation bands to
start [7]. In Fig. 4.1-4, orientated and elongated pores in Alulight are marked. A
study of pore characteristics in Alulight foam revealed that pores in the deforma-
tion zone have significantly different orientation from pores in undamaged mate-
rial [8]. Similar behavior was deduced from in-situ experiments using computed
tomography [9]. Unlike structural applications, the pores size becomes an impor-
tant feature if the cellular metals are used in heat exchangers or in filter systems
(see Section 7.2).
Figure 4.1-4. Inhomogeneous distribu-

tion of pores size: misoriented and
large pores.
Cell Skeleton
After characterizing the pore structure, it still remains to define quantities for de-
scribing the cell skeleton. As explained in the introduction of this chapter, the ske-
leton consists of cell nodes interconnected by cell edges. In a closed-cell structure,
the cell edges are connected by cell walls that separate the cells.
The volume fraction of metal in nodes, edges, and walls defines the density of
the cellular material. The features that can be described are number of nodes,
number of cell edges and walls per cell, length, thickness, and curvature of cell
edges, angle between nodes and cell edges, and shape of cell edge, number of
cell walls connected of edges and of edges connected at nodes (connectivily). A
drawback of quantitative image analysis of nontransparent samples is that it is
difficult to define the exact position and extent of the cell walls, edges, and
nodes, as well as the range of these features.
All these parameters have an influence on mechanical properties. As long as the
cellular material is nearly uniform and perfect, these are the most important para-
meters, but in real foams parameters are necessary to characterize agglomeration
of solid material. Klocker proposed the analysis of covariance function for detection
of clustering [10]. In a comparison of undeformed and crushed Alporas samples,
clustering of cells could be shown by a drop in covariance. Another way of charac-
terizing clustering is the pair correlation function [16], but this method has not yet
been applied for describing cell clustering in metallic foams.
Defects in Cellular Metals and Foams

Another important aspect of characterizing the cell material is the description of
imperfections. All structural features that degrade the properties determined by
the scaling law of Gibson and Ashby, are classified as ªknock-downº factors or de-
fects.
First of all there are geometrical imperfections. These are deviations from ideal
polygons or irregularities in edge pattern. Models (see Chapter 6.1) revealed that
imperfections or defects such as wavy distortions of cell walls, variations of cell-
wall thickness, or non-uniform cell shape decrease the stiffness and strength of
cellular material [11 14]. Unfortunately, most metallic foams tend to show many
of these defects.
Defects in cell edges and walls have been identified as reducing the elastic mod-
ulus of metallic foams significantly. Two different typical defects are known: curved
cell edges and corrugated cell walls and edges. Corrugated cell walls are typically
found in metal foams such as Cymat (Alcan) or Hydro foams, which have rather
thin cell walls compared to the cell size (Fig. 4.1-5). Curved or corrugated cell
walls occur during forming or on solidification.
In modeling, cell edges or walls are often described by beam or shell elements
having an ideal shape and constant thickness. In real foams, the metal is thicker
at cell nodes and edges because of the dynamics of metal flow during foaming.
Flow of the melt also causes the rupture of cell walls. During solidification, dendri-
tic growth at a free surface and shrinkage of the melt can induce roughening of cell
walls, edges, or nodes. Figure 4.1-6 shows the surface of a solidified cell wall. The
shrinkage of the melt during solidification and cooling leads to cell-wall wriggling
and also cracks in cell walls. The influence of initial cracks on toughness or fatigue
behavior is still under discussion.
The formation of micropores is also connected to sucking of liquid into the
nodes (Plateau effect). Owing to the increased amount of liquid in nodes, the
solidification starts at the surface of the nodes caused by the heat flow. While the
Figure 4.1-5. Corrugated cell walls

in foam produced by the melt process
(Combal).
Figure 4.1-6. Defects in cell walls

(microcracks, roughness of surface).
nodes solidify the melt is consumed and, as know from casting, small voids occur
in the middle of the node. This can be observed as micropores in cell nodes or
thicker cell edges.
4.1.2
Characterization Methods and Quantities of Geometric Architecture of Real Metallic
Foams
Although XCT is now established for the investigation of metallic foams (see
Section 4.2), the characterization of metallic foams is usually performed using
quantitative 2D image analysis methods. Most papers concentrate on the character-
ization of cell structure derived from analysis of cross sections. For further reading
on the principles of stereology the reader is referred to the basic literature of Exner
[15] and Ohser [16].
In the following section the structure of cellular metals and metallic foams are
described and typical structural features are discussed. Before going into detail on
structural analysis, some information on the preparation techniques for metallic
foams is given.
4.1.2.1 Sample Preparation

The first step in preparing a metallographic specimen is to cut a sample from a
foamed part. In the early days of foam research, great uncertainty led to insistence
that metallic foam should only be cut using spark erosion machines in order to
prevent damage to cells or cell walls. Our results have shown that using a precision
saw at low cutting speed and force gives results just as good as spark erosion, but
with a reduction in time and cost.
In the next step, the specimens are often embedded under vacuum in epoxy, so
that the pores are filled and a conventional metallographic preparation with
grinding and polishing can follow. The procedure of filling epoxy in the pores
has be repeated several times until all the pores are filled. This method gives
good results if the microstructure of the cell-wall material is studied using light
microscopy. If using scanning electron microscopy on epoxy-embedded samples
the surface can become charged: a higher conductivity embedding substance,
such as Cu-enriched epoxy, will help to improve the resolution and quality of
the images.
Another method of preparing the sample for digitizing by scanning was pro-
posed by Kriszt et al. [17]. In this method the foamed samples are precision cut,
followed by coloring the sample and smooth grinding, so that the contours of
the cell structure on top of the sample regain their metallic luster. This is a prac-
tical way of analyzing the geometrical structure of big areas. After the preparation,
the surface can be scanned using a high resolution mode and converted into a
digital image. Figure 4.1-7 depicts prepared cross sections of Alporas foam pre-
pared by this technique.
Figure 4.1-7. Geometrical structure of

Alporas foam, sample preparation for
digitizing by scanning technique.
4.1.2.2 Pore Size

Figures 4.1-3a, 4.1-4, 4.1-5 and 4.1-7 depict the differences in structure between
metal foams produced by the powder compact technique, such as Alulight or Fraun-
hofer foam, and the melt process route, such as Alporas, Cymat, and Combal Foam.
Alulight seems to have more irregular cell shape and size than Alporas. This is a
typical feature of PCF materials. Another structural feature of metallic foams such
a)
Figure 4.1-8. Pore-size

distribution of Alulight:
a) equivalent diameter versus
frequency of pores;
b) equivalent diameter versus
frequency of area. b)
Figure 4.1-9. Pore-size distribution of

Alporas: a) equivalent diameter versus
frequency of pores; b) equivalent dia-
meter versus frequency of area.
a)
b)
as Alulight is the non-uniform distribution of solid metal in the foamed part (see
Fig. 4.1-3a and 4.1-4). Zones having an enrichment of solid material are referred as
hard spots in the literature [18].
Measurements of the pores size for both foam types have shown that they have
the same average equivalent diameter of about 3 mm. The distribution curves of
equivalent diameter versus frequency of pores of both foam types show a bimodal
shape (Fig. 4.1-8a and Fig. 4.1-9a). The large fraction of small pores is caused by
solidification shrinkage. Interestingly, it seems that the Alulight material always
shows a broader distribution curve. Because of the high number of small pores,
the cell-size distribution is dominated by the small pores. The presentation of
area fraction of a pore class over equivalent diameter helps to emphasise the con-
tribution of big pores to geometric structure (Fig. 4.1-8b and Fig. 4.1-9b). The
presence of outliers can be made more visible if this method of presenting the re-
sults is used. Outliers having a pores size bigger than five times the average pore
size can be detected in nearly all samples produced by PCF techniques, such as
Alulight. This is not the case in metallic foams produced by the melt route.
Figure 4.1-10. Density as a function of pore size for various foam types (data for FORMGRIP
process from [20]).
Measurements of cell size published by Miyoshi et al. distinguish between two

types of Alporas foams: one conventional type for sound absorption having a
cell-size range of 1 13 mm and another type having smaller cells of 1 7 mm [19].
While Alulight and Alporas have a similar pore size, metallic foams such as
Hydro or Cymat foams are known for a pore size of more than 1 cm. The main
difference between these two groups of foams is that for Alulight and Alporas
TiH2 particles are used as the blowing agent and in the second group gas blown
directly into the melt leads to the formation of pores.
As already discussed in Section 4.1.1, the density and the pores size are two
quantities that are used to characterize foams. It is recognized that these quantities
are interdependent, but nevertheless many users demand that density and cell size
be tailored independently. Taking data from the FORMGRIP process [20], it can be
shown that density is a function of pore size (Fig. 4.1-10). The same result was con-
firmed by KoÈrner et al., who investigated pore evolution during foaming [21]. They
proposed that the mean cell size is proportional to the reciprocal value of density.
This holds true if Alulight, Alporas, Hydro, or Cymat foams are examined.
4.1.2.3 Pore Shape

Assuming that cells of metal foams have ideal polyhedral shapes, such as the Kel-
vin cell, it should be possible to find a correlation between that number of sides of
each pore in the 2D cross section and 3D shape [15,16]. But Degischer et al.
showed that the pores in metallic foams do not have an ideal polyhedral unit
cell [18]. Hence, in 2D analysis the characterization of shape is limited to the
use of shape factors for comparison of different foam types.
From quantitative structology different methods for defining shape factors are
known. Klocker defined the ratio between the largest and the smallest principle
cell dimension as the shape anisotropy ratio [10]. A similar definition was used
by Simone (ellipse shape ratio) [22]. Another shape factor is circularity, which is
the ratio of pore area and the square of pore perimeter [23]. For all mentioned
shape factors a value of 1 corresponds to a circular shape. The circularity allows
one to interpret the shape of the cell; the circularity of a rectangle is 1.4 and
that of a regular pentagon is 1.1.
Klocker has shown that the shape anisotropy ratio of Alporas and Alulight has an
average ratio of 1.5 [10]. The ellipse shape ratio of Alulight foam is Gaussian dis-
tributed, and has an average value of 0.6 [17]. In Cymat foam an ellipsoidal shape
ratio of a factor of 1.8 was found [22]. Studies of evolution of pore shape during
foaming revealed that foams having a low expansion rate (high density) tend to
have pores with a circular shape. As the foaming process proceeds the shape of
the pores becomes rectangular or ellipsoidal [23].
4.1.2.4 Pore Orientation

The definition of pore orientation is only sensible if a reference system is defined.
Depending on the target of the investigation, reference axes are oriented to some
main directions of processing, such as the foaming direction or the feed direction
of conveyor-belt systems. Another kind of reference system can be defined by
the direction of loading. The only foam types in which the pores have a systematic
and significant orientation in relation to processing are Hydro and Cymat [22].
Generally, Alulight and Alporas foams do not have a pronounced pore orientation
[17,22].
4.1.2.5. Thickness of Cell Edges and Walls

Although the cell-wall thickness has a great influence on mechanical properties,
very little information on real cell-wall thicknesses is published. This might be
due to the difficulty of defining the transition from node to cell wall or edge.
Hence it is difficult to measure this structural feature with quantitative image anal-
ysis.
Berger showed that the distribution of cell-wall thickness has an asymmetric
shape for foams produced by powder metallurgical technique [23]. The smallest
cell-wall thickness that was found is about 70 mm. The maximum cell-wall thick-
ness was about 500 mm. A cell-wall thickness of about 150 mm was the most fre-
quent thickness. Astonishingly, the thickness of cell walls is independent of
foam density.
Klocker characterized the ratio of mean thickness to mean length of cell wall for
Alulight and Alporas [10]. Both foam types show an extremely asymmetric distri-
bution. The maximum frequency of about 40 % at a ratio of 0.1 in Alporas is
more pronounced than the maximum frequency in Alulight with 15 % at a ratio
of 0.18.
The ratio of mean thickness/length in crush-fronts revealed that in both foam

types the distribution is the same. The most frequent ratio is 0.1. Mosler et al. ana-
lyzed the width of cell edges in Alulight foams [24]. The average width does not
exceed 130 mm, so it can be calculated that the average length of cell edges in Alu-
light foams is about 750 mm.
4.1.2.6 Topological Features

The most difficult task in characterizing the foam structure is the description of
topological relations. If no model for the complex geometrical structure of real
foam is known, our understanding of topology of cellular material based on a
2D analysis is limited. First attempts have been made by calculating tessellation
for cell structures.
Klocker measured the average number of neighbor cells for Alporas and Alulight
[10]. The number of neighbor cells varies between 1 and 10. The average number
of neighbor cells in Alporas is 4.5, in Alulight material the average number is 5
cells, slightly higher than in Alporas. In deformation bands caused by compres-
sion, the analysis of neighbor cell zones gave a smaller number [10].
Another parameter that helps to describe the topological relations is the distance
to nearest neighbors. For Alulight foam Mosler et al. found a mean average dis-
tance of about 900 mm [24].
The characterization of the pattern of 2D cell arrangement can be described by
tessellations. Some information on tessellations of cellular structures can be found
elsewhere [10,24]. Klocker's results are optimistic about the description of Alporas
foam by Voronoi tessellations, whereas the results on Alulight, because of its
strong dependence of cell structure on the position in the specimen, are not so pro-
mising. Klocker suggested the Johnson Mehl tessellation as an alternative that
takes into account the curvature of cell walls. In contrast to Klocker's work, the in-
vestigations of Mosler et al. confirmed that Voronoi tessellation give good agree-
ment with the geometric structure of Alulight [24]. The explanation for the contrary
results might already be given by Klocker, who assumed that inhomogeneity in the
structure of Alulight causes this result.
4.1.3
Characterization of Microstructure of Massive Cell Material
Knowledge of the microstructure (grains, grain boundaries, precipitates, disloca-

tions etc.) of metallic foams is essential, because of the effect of microstructure
on mechanical properties. As has been shown, foamed samples having the same
density but made of different alloys reach different plateau stresses [25]. Different
hardnesses support the theory that the microstructure is responsible for different
plateau strengths [26].
More ductile alloys, such as wrought aluminum alloys, show a ductile collapse of
cells under compression; brittle cast alloys deform by crushing of cells. This differ-
ent behavior is why the strength of a foam made of a wrought aluminum alloy in-
creases continuously, while the second foam shows remarkable fluctuations in

strength [25]. Another phenomena where influence of microstructure can be ob-
served clearly is the increase of strength after tempering treatment of metallic
foams due to precipitation hardening [26,27].
In most metal foams alloying elements or particles have to be added so that a
stable liquid foam is gained (Fig. 4.1-12). These additions often modify the alloy
or even change it completely. For instance, the Alporas foam consists of an unusual
aluminum alloy, additions of 2 % calcium change the constitution completely. The
reader is referred elsewhere for further information on microstructure of Alporas,
Cymat, and Alcan foams or FORMGRIP samples [28,29]. FORMGRIP samples are
stabilized by adding particles such as SiC, so they show a microstructure similar to
that of particle-reinforced composites [29].
All microstructures of foamed materials are in a solidified undeformed state. like
the microstructure of castings. Depending on the chemical composition, metallic
dendrites, eutectic cells, precipitates, or particles are found in metal foams
(Fig. 4.1-11). Even if the cell walls are thin, it can be estimated that the cooling rate
is not too high, because dendrites have a size of hundreds of microns. In many
cases the size of dendrites or eutectic cells is equal to the thickness of cell walls.
Most of the dendrites are equiaxed or sometimes even globular (Fig. 4.1-11a)
[30]. This can be explained in two ways: first the superheating of the melt during
foaming is as much as in common melts, because superheating immediately leads
to collapse of foam. This allows homogeneous distributed nuclei to remain in the
melt, which lead to equiaxed dendrites.
Figure 4.1-11. Microstructure of cell walls.

a) Dendritic structure of Alulight made of a
wrought aluminum alloy. b) Interdendritic eutectic
structure of Alulight made of cast AlSi7 alloy.
Figure 4.1-12. Microstructure of particle-

stabilized metallic foam.
Figure 4.1-13. Ti-rich particle in cell wall.
Another interesting result was found by MuÈller et al. when studying the micro-
structure of Fraunhofer foams made of the casting alloy AlSi7 [30]. They found
some Ti-rich particles after foaming (Fig. 4.1-13). It can be assumed that these
particles are residuals of the TiH2 blowing agent, which are stabilized by the
formation of a diffusion barrier. Braune et al. (Section 3.1) reported that
they observed refoaming in some foamed samples when the samples were
heated before hot forming. Both observations indicate that due to the forma-
tion of intermetallic layers surrounding the TiH2 particle, it is likely that the
dissociation decelerated.
4.1.4
Conclusions
Most of the work on structural analysis of cellular structures is concentrated on the

study of cell architecture or local density, little knowledge has been gained on the
special microstructure of foamed metals influencing the property profile. This de-
velopment in foam research was triggered by material models of cellular struc-
tures, from which we can learn that the architecture and the density of cellular
metal have a dominant impact on the behavior of metals. Most of the structural
data available are measured on metallographic cross sections, scarcely any data
on structure analysis is derived from 3D analysis or XCT data.
But with the ongoing research on cellular metals the investigation of microstruc-
ture is important. Often there are typical defects, such as irregularity in cell shape,
cracks in cellular structure, or micropores caused by the intrinsic properties of the
massive metal. So in future, the emphasis in foam research will be put on the
microstructure, because not all material properties can be explained by the
architecture of the cellular metal and knowledge of microstructure is essential
for understanding processing.
References
1. L. J. Gibson, M. F. Ashby, Cellular Solids: 13. J. L. Grenestedt, in Porous and Cellular

Structure and Properties, 2nd edn, Cambridge Materials for Structural Application, D. S.
University Press, UK 1997. Schwartz, D. S. Smith, A. G. Evans, H. N. G.
2. R. Gradinger; F. G. Rammerstorfer, Acta Wadley (eds), MRS, Warrendale, PA 1998,
Mater. 1999, 47, 143 148. p. 3.
3. M. F. Ashby, A. Evans, N. A. Fleck, L. J. 14. A. E. Simone, L. J. Gibson, Acta Mater.
Gibson, J. W. Hutchinson, H. N. G. Wadley, 1998, 46, 3929.
Metal Foams: A Design Guide, Butterworth- 15. H. E. Exner, Int. Metall. Rev. 1972, 17, 25.
Heinemann, Oxford 2000, p. 42. 16. J. Ohser, U.Lorz, ªQuantitative
4. F. N. Rhines, Microstructology, Riederer Ver- GefuÈgeanalyseº in Theoretische Grundlagen
lag, Stuttgart 1986. und Anwendung, Freiberger Forschungshefte
5. H. E. Exner, H. P. Hougardy, Einfu È hrung in Reihe B, 276, Deutscher Verlag fuÈr Grund-
die Quantitative GefuÈgeanalyse, DGM Infor- stoffindustrie, 1994.
mationgeschellschaft, Oberursel 1986. 17. B. Kriszt, A. Kottar, T. Klocker, H. Knoblich,
6. R. Brezny, D. J. Green, Acta Metall. Mater. H. P. Degischer, Fortschritte in der Metallo-
1990, 38, 2517. graphie, G. Petzow (ed), Prakt. Met. Sonder-
7. B. Kriszt, A. Falahati, K. Faure, H. P. De- band, Vol. 30, DGM-Informationsge-
gischer, Bauteilversagen durch Mikrodefekte, sellschaft, Frankfurt 1999, p. 385.
DVM-Bericht 518, 1998, p. 195. 18. H. P. Degischer, A. Kottar, B. Foroughi, in
8. B. Kriszt, B. Foroughi, K. Faure, H. P. De- X-Ray Tomograhy in Material Science, J. Baru-
gischer, in Metal Foams and Porous Metal chel, J.-Y. Buffiere, E. Maire, P. Merle, G. Peix
Structures, J. Bahnhart, M. F. Ashby, N. A. (eds), Hermes, Paris 2000, p. 165.
Fleck (eds), MIT Verlag, Bremen 1999, p. 241. 19. T. Miyoshi, M. Itoh, A. Akiyama, K. Kita-
9. H. Barth-Smith, A.-F. Bastawros, D. R. hara, in Metal Foams and Porous Metal Struc-
Mumm, A. G. Evans, D. J. Sypeck, H. N. G. tures, J. Bahnhart, M. F. Ashby, N. A. Fleck
Wadley, in Porous and Cellular Materials for (eds), MIT Verlag, Bremen 1999, p. 125.
Structural Application, D. S. Schwartz, D. S. 20. V. Gergely, T. W. Clyne, in Metal Foams and
Smith, A. G. Evans, H. N. G. Wadley (eds), Porous Metal Structures, J. Bahnhart, M. F.
MRS, Warrendale, PA 1998, p.71. Ashby, N. A. Fleck (eds), MIT Verlag, Bremen
10. T. Klocker, ªImage Analysis of Metallic 1999, p. 83.
Foamsº, Thesis, Vienna University of Tech- 21. C. KoÈrner, F. Berger, M. Arnold, S. Stadel-
nology, Austria 1999. mann, R. F. Singer, Mater. Sci. Technol. 2000,
11. J. L. Grenestedt, J. Mech. Phys. Solids 1998, July-August, 781 784.
46, 29. 22. A. E. Simone, ªPorous Metals and Metallic
12. J. L. Grenestedt, K. Tanaka, Scripta Mater. Foamsº, Thesis, Massachusetts Institute of
1999, 40, 71. Technology, Boston, MA 1997.
23. F. Berger, ªEinfluss der Prozessparameter Materialforschung und -PruÈfung, Bad Nau-
auf die Schaumtruktur von Al-SchaÈumen auf heim, Germany 1998, p. 167.
pulvermetallurgischer Baisisº, Thesis, Lehr- 27. D. Lehmhus, C. Marschner, J. Banhart, in
stuhl fuÈr Werkstoffkunde und Technologie MetallschaÈume, H. P. Degischer (ed), Wiley
der Metalle, UniversitaÈt Erlangen, Germany VCH, Weinheim 2000, p. 474 477.
2000. 28. B. Kriszt, O. Kraft, H. Clemens, Materwiss.
24. U. Mosler, G. Heinzel, U. Martin, H. Oet- Werkstofftechn. 2000, 31, 432.
tel, Materwiss. Werkstofftechn. 2000, 31, 29. Y. Sigimura, J. Meyer, M. Y. He, H. Barth-
519 522. Smith, J. Grenstedt, A. E. Evans, Acta Mater.
25. B. Kriszt, B. Foroughi, K. Faure, H. P. De- 1999, 45, 5245 5259.
gischer, Mater. Sci. Technol. 2000, 16, 792. 30. A. MuÈller, U. Mosler, Fortschritte in der
26. B. Kriszt, R. Gradinger, B. Zettl, Metallographie, G. Petzow (ed), Prakt. Met.
ªUntersuchung von Aluminiumschaum, im Sonderband, Vol. 32, DGM-Informations-
speziellen der mechanischen Eigenschaftenº gesellschaft, Frankfurt 2001, p. 279 284.
in WerkstoffpruÈ fung, Deutscher Verband fuÈr
4.2
Computed X-ray Tomography
E. Maire
X-ray tomography has recently emerged as a powerful technique capable of giving

a nondestructive picture of the interior of structural materials [1,2] including cellu-
lar metals and foams [3 6]. This section gives an overview of the technique, the
set-ups that can be used, and some examples of application in the field of the com-
pression behavior of foams. In the final section, we describe how to produce finite
element models from the actual 3D images of the microstructure and we show re-
sults of calculations performed with this method.
4.2.1
Principle of the Technique
4.2.1.1 X-ray Radiography

The X-ray radiography technique is based on the simple Beer Lambert law, which
gives the ratio of N1, the number of photons transmitted after a path of length z
through the thickness of a sample exhibiting a coefficient of attenuation m, over
N0, the number of incident photons. If m varies along the path, the integral of m
over z has to be used
" Z #
N1 (1)
exp s m(x; y; z)dz
N0 path
146 4.2 Computed X-ray Tomography
a)
b)
Figure 4.2-1. a) Diagram of the principle of the X-ray radiography technique used in the present
chapter with a diverging beam for a medium resolution. b) Diagram of the X-ray tomography
technique.
m depends on the composition of the local elementary volume of matter situated in

x,y,z. For aluminum the photoelectric effect dominates for X-ray energies below
60 keV according to
Z4
m(x; y; z) Kr (2)
E3
where K is a constant, r and Z are the density and the atomic number of the in-
vestigated material, and E is the energy of the incident photons. Above 60 keV
the Compton effect dominates, where the attenuation becomes nearly proportional
to the density.
This attenuation law explains the contrast observed in the X-ray radiograph of a
bulky material (see Fig. 4.2-1a) because each point of a detector placed behind the
sample is situated in front of a different path: if the material is heterogeneous, the
integral value of m(x,y,z) varies also with x and y.
4.2.1.2 X-ray Tomography

The drawback of radiography is that a large amount of information is projected on
one single plane and the resulting image can be difficult to interpret if the number
of microstructural features along the thickness of the sample is important. Tomogra-
phy overcomes this drawback by combining the information from a large number of
such radiographs, each being taken with a different orientation of the sample in
front of the detector (Fig. 4.2-1b). If the angular step between each radiograph is
small enough, it is possible from the complete set of radiographs to recalculate
the local value of m(x,y,z) at each point of the sample. This reconstruction is per-
formed by appropriate software based on the filtered back-projection algorithm [1].
4.2.2
Set-ups
Different set-ups can be used. They all include a source, a rotation stage on which
the object is fixed, and an X-ray detector. The easiest way of getting digitized
images is to use directly a two-dimensional radioscopic detector composed of a
screen transforming the X-rays into visible light, which is then transferred by ap-
propriate optic lenses to a cooled CCD camera. The common geometric constraint
of the different set-ups is that the axis of rotation of the sample must be parallel to
the plane of the detector and its image must be aligned with one of the columns of
the CCD (preferably the central column). The crucial point in applying tomography
to materials science is the achievable spatial resolution r defined as the size of the
pixel of the detector in reference of the sample. Its limit value is mainly governed
by the available photon flux at the level of the sample and by the set-up, as de-
scribed in the next two sections. For medical tomography (XCT scanner) this
limit of resolution is of the order of 300 mm. Materials scientists wishing to see
and distinguish features with a size of the order of 1 10 mm had to develop appro-
priate tools.
4.2.2.1 Medium-Resolution Microtomography

For a limit of resolution of the order of 10 mm (medium resolution), a cone beam
system can be used with a classical microfocus X-ray tube as the source. Such a
device has been assembled in the CNDRI laboratory at INSA Lyon. With this diver-
ging geometry the magnification can be varied easily by changing the position of
the sample in the space between the source and the detector. The limit of resolu-
tion is due to the size of the microfocus, which introduces blur in the projected
image. This size has a minimum value because if the source size gets too small,
the flux at the level of the sample becomes so low that the acquisition time re-
quired to record a single radiograph is too long for a realistic analysis. In such a
laboratory set-up, a polychromatic source is used in order to keep the acquisition
time acceptable. This may introduce artifacts due to beam-hardening and does
not allow quantitative reconstruction of the absorption coefficient m.
This kind of standard set-up exists now in many different academic and indus-
trial laboratories. Some commercial and nearly portable set-ups with a good reso-
lution (down to 6 mm) are also available.
4.2.2.2 High-Resolution Microtomography

We emphasized in the previous section that the set-ups using X-ray tubes are
limited because of the low flux delivered by this kind of source. The best quality
images in terms of signal-to-noise ratio and spatial resolution, allowing high-re-
solution microtomography, are today obtained from instruments located in syn-
chrotron radiation facilities. Such a source can be found at ESRF in Grenoble.
The X-ray beam produced at this kind of third generation synchrotron facility
is very interesting for microtomography because of the following original fea-

tures [2].
x Very high intensity of the X-ray beam inside the cone.
x Nearly parallel beam: this simplifies the reconstruction of the images but shifts
the resolution dependance into the detector's performance.
x A monochromatic beam can be used: this suppresses for example beam harden-
ing and cupping effects and allows for a quantitative reconstruction of m.
x Availability of high energy photons (beyond 100 keV) that can penetrate heavy
(high Z) materials.
The ID19 beam line team of ESRF has combined these specific properties with the
availability of a suitable CCD-based detector exhibiting at the same time a large
dynamic range (13 bits), a low noise, and a short transfer time (down to 60 ms
for 1024 q 1024 pixels). This unique combination allows a resolution of the
order of 1 mm to be obtained routinely today.
4.2.2.3 Resolution Required for the Study of Metallic Foams

In the case of metallic foams, a medium resolution set-up is usually enough be-
cause the size of the microstructural features to image is of the order of 100 mm.
The different studies presented in this book, and especially in this section, have
been performed using standard laboratory set-ups but very interesting results
were obtained with the high-resolution set-up in the case of the finer structures
that can be found in polymer foams [7].
4.2.2.4 Reconstruction Method

Reconstruction of the images presented here is achieved using a C-language pro-
gram written by the CEA-LETI (Commissariat a l'Energie Atomique, France) and
based on the Feldkamp algorithm [8]. A ramp filter is selected, thus preserving
the spatial resolution. A DEC 500 MHz workstation is used to reconstruct the en-
tire volume, slice by slice. Reconstruction of a typical volume (350 q 340 q 128 vox-
els) takes approximately 1 hour. The final result is a 3D map of the m of elementary
volumes of size r q r q r inside the sample. This map can give a certain image of
the microstructure if m is transformed into a gray level. Voxels located in the
aluminum will exhibit high values of m (high gray levels) whereas voxels located
in the pores will exhibit low values of m (low gray levels).
4.2.3
Experimental Results
4.2.3.1 Initial Cell Structure

Figure 4.2-2 compares reconstructed slices (in gray level) of two different Norsk
Hydro foams with different density and cell sizes. These slices are 2D images
but extracted from a 3D bloc of numerical data (i. e., the reconstructed volume
a) b)
Figure 4.2-2. a) Slice extracted from the tomographic reconstructed volume of a sample of 6 %
relative density Norsk Hydro foam. b) Slice extracted from the tomographic reconstructed volume
of a sample of 11 % relative density Norsk Hydro foam.
of the sample). The internal structure and cell-wall arrangement is clearly imaged
and the difference between the two samples is easy to visualize. The relative den-
sities of the two materials shown in these figures are 6 and 11 %. A 3D rendering
of the 6 % foam is shown in Fig. 4.2-3. This rendering is obtained by removing the
pixels located in the pores, showing then only the material located in the walls of
the foam. This virtual image really looks like an optical picture of foam, as can be
seen in different sections of this book. From such images, 3D morphological char-
acteristics can be measured such as the wall thickness, the wall dimensions, the
cell size and so forth. These characteristics have a strong influence on the mechan-
ical properties.
Note also that these images can be used with a lower resolution to assess the
density fluctuations inside the material (m is directly related to the density). It is
perfectly well established that the value of the global relative density of a sample
of metallic foam has a direct effect on its mechanical properties but the density
fluctuations have also a strong influence and especially on the scatter of these prop-
erties. This aspect is developed in Chapter 6 of this book.
Figure 4.2-3. 3D rendered view of a 6 % relative density

Norsk Hydro metallic foam.
4.2.3.2 Evolution of the Structure During a Compression Test

The deformation modes of a foam can be studied by compressing a sample and
imaging by tomography its internal structure at different stages of deformation.
Such a study has been performed in [4 6]. The experimental details and main
results of the first study [4], performed by the author of the present chapter, are
summarized below.
a)
b)
c)
Figure 4.2-4. a) Two dimensional recon-

structed slice showing the initial micro-
structure of the compressed sample.
b) Sample slice as in (a) for e 0.065.
c) Sample slice as in (a) for e 0.2.
d) Sample slice as in (a) but e 0.6.
d)
The compression set-up used for this study was an Adamel Lomargy standard
mechanical frame. The sample was placed between two plates and the force/dis-
placement curve was recorded. The resulting s(e) conventional curve was found
to be classical for this material with a yield stress of 0.7 MPa and a slight increase
in the stress during the plastic collapse. We made three steps during the compres-
sion loading to allow the characterization of the internal structure by tomography.
The three steps were made at increasing values of the remanant plastic true strain:
0.065, 0.2, and 0.6, measured directly from the change in dimensions of the paral-
lelepiped after unloading.
Figure 4.2-4 shows a set of 2D reconstructed slices of the same zone of the sam-
ple numerically and nondestructively extracted from the volume at the four com-
pression stages from 0 to 0.6 of remanant true strain. These slices are parallel to
the compression axis, which is vertical on the figure. One can clearly see the defor-
mation mechanism of the studied foam in compression by comparing figures (a)
(initial state) and (b) (remnant strain 0.065). The large value of the plastic strain
at this stage is clearly not due to a homogeneous plastic straining of the whole
sample, but to the local buckling of several walls. Between the second and the
third step, we observe that the number of new buckled walls is small. The defor-
mation process leads first to the complete closing of the collapsed cells instead
of the appearance of new ones. From the observation of the entire population of
buckled walls, we have observed that they are situated in a band perpendicular
to the loading axis (see Fig. 4.2-4c). These 3D nondestructive observations confirm
those made using other techniques [9] or with the same technique by other authors
[3,5,6].
4.2.4
Micromodeling of a Foam by Finite Elements
The internal structure of a metal foam is extremely complex: the cell size and the
length of the walls are parameters that are widely spread. The nonperiodic charac-
ter of the structure leads also to a very difficult calculation of the actual force ap-
plied to each wall compared to the case of honeycombs. The complexity of this en-
tanglement suggests analyzing the distribution of the stress and deformation in
each wall using an appropriate tool: finite element analysis. In the present study
we have a 3D picture of the actual microstructure of the foam. In what follows,
we present an easy way to use this picture to generate meshed models of the actual
microstructure. These models are readable by the Abaqus commercial code, which
we have used to perform calculations, firstly in 2D and then in 3D.
4.2.4.1 Direct Meshing of the Actual Microstructure

As already mentioned, the result of a tomographic inspection (after reconstruction)
is a 3D table of gray levels. Given the experimental resolution of the set-up used,
the volume of the element is 150 mm q 150 mm q 150 mm. The method described
bellow is based on a simple idea: each voxel in the 3D image can be represented by
a cubic element in a finite element mesh. The cubic elements located in the voids
are then removed from the model. Given the experimental conditions we used, we
considered that the boundary between the steel plates of the compression rig and
the aluminum foam was perfect. The displacement of the lower nodes were fixed
in the three directions. We generated a steel plate perfectly bonded to the upper
nodes and applied a displacement along the compression direction to the upper
nodes of this steel plate.
4.2.4.2 Results
All the calculations were performed assuming elastic behavior for the walls.
Two-dimensional (2D)
The procedure described above can be very simply applied to any slice by generat-
ing only one layer of four-node 2D solid elements. We applied this technique for
several slices. The elements used were plane-strain four-nodes elements (PE4 Aba-
qus elements).
Figure 4.2-5 shows a contour plot of the s22 (i. e., the normal component in the
direction of loading) component of the stress tensor in the walls calculated accord-
ing to our procedure for a slice with PE elements. The higher compressive stresses
are shown in black on the figure. What we want to reach with this calculation is not
the absolute value of the stress but the determination of the sign of this stress so
the color bar has been intentionally simplified to distinguish tension and compres-
Figure 4.2-5. Stress map (arbitrary units) calculated using the Abaqus FE code by converting the
actual microstructure of a 2D slice into a mesh using the local procedure described in this
chapter. The stress component shown here is the normal component.
Figure 4.2-6.Enlargement of a region of the stress map shown in Fig. 4.2-5, which shows the
bending character of the stress field when calculated in 2D.
sion. Note that the values of the stresses in the walls seem to form paths of high
compression loading between the top and the bottom of the section. These paths
define ªpillarsº where the compression load is transferred while some other re-
gions are much less loaded. This permits us to locate highly loaded regions in
the material.
If we zoom in a single wall situated in one of these pillars and oriented along the
compression axis, we observe a very special stress state composed of zones in com-
pression (black) and in tension (light gray), which indicate a high bending charac-
ter of the stress (see Fig. 4.2-6) with 2D assumptions. Plane-strain elements imply
that the cells that are defined in Fig. 4.2-6 are section of tubes having an infinite
dimension along the third axis perpendicular to the figure. This tube shape pro-
motes a high bending stress state within the plane of the slice only.
Three-dimensional (3D)
A complete 3D calculation of the entire foam with our technique would require a
1 000 000 element mesh. This amount of data is not tractable with our computer
today. A region of 1003 voxels was selected from the initial raw data. In order to
reduce the number of elements in the final mesh, the voxels of the region were
grouped eight by eight (two by two in the three directions), which led to an increase
in the actual resolution in the final mesh leading to a size of 300 mm q 300 mm q
300 mm for each element. This 503 voxels 3D block was transformed into a mesh
according to the previously described procedure and the mesh was slightly com-
pressed (0.02 % of elastic strain). The calculation took around 140 h on a standard
workstation. This allowed us to analyze the stresses and strains in the walls as illus-
trated in Fig. 4.2-7 and Fig. 4.2-8, which show a contour plot of the normal stress
and of the Tresca shear stress in a layer. As for 2D simulation, the color map has
been simplified to emphasize the sign of the stress rather than its absolute value.
Note that Fig. 4.2-7 shows much less region loaded in tension (black) than the
equivalent 2D calculation, shown for instance in Fig. 4.2-6. This indicates that
when calculated in 3D, the stress state in the walls is completely different than
Figure 4.2-7. Stress map (arbitrary units) cal- applied strain was 0.02 %. The amount of ten-
culated using 3D elements on a representative sile stress is much lower than for the 2D cal-
region of the actual compressed block. The culation (compare with Fig. 4.2-6).
normal stress component is shown here. The
Figure 4.2-8. A map of the value of the Tresca's shear stress calculated in the same 3D condi-
tions as for Fig. 4.2-7.
when calculated in 2D with, in particular, a much less pronounced bending

character. Thus a proper calculation with this technique should be performed on
a representative piece of material in 3D with a bigger computation power.
4.2.5
Conclusions
Tomography is a new powerful technique, suitable for the 3D nondestructive anal-

ysis of the mesostructure of metallic foams and its evolution under compression.
We have presented illustrations which show that it is possible to use this technique
to have a better determination of the microstructural features responsible for the
mechanical behavior of such materials. When used at different compression stages
on the same sample it also enables us to understand perfectly the deformation
mechanisms of metallic foams. In the case of the Norsk Hydro foams, we have
shown that deformation is due to local buckling of the edges and walls and that
these events were all located in a band perpendicular to the compression axis.
These images also give directly produced finite-element meshes of the actual sam-
ples. The size of these meshes can be very big and not tractable with a regular
computer if the entire compression samples are meshed in 3D. Preliminary calcu-
lations were then made on entire sections in 2D, which allow us to visualize the
highly loaded regions in some particular sections and the bending character of
the stress state in the walls, but the local carrying function of the cell edges cannot
be taken into account appropriately. A small 3D block was finally calculated and it
was thus shown that the stress state in the walls has a much less pronounced bend-
ing character in the actual conditions.
References
1. J.-Y. BuffieÁre, E. Maire, P. Cloetens, G. Lor- 6. A. H. Benaouli, L. Froyen, M. Wevers, Ch. 10

mand, R. FougeÁres. Acta Mater. 1999, 47, in X-ray Tomograhy in Material Science, J.
1613. Baruchel, J.-Y. Buffiere, E. Maire, P. Merle,
2. V. Kaftandjian, G. Peix, D. Babot, F. Peyrin, G. Peix (eds), Hermes, Paris 2000.
J. X-ray Sci. Technol. 1996, 6, 94. 7. P. Cloetens, W. Ludwig, J. P. Guigay,
3. H. Bart-Smith, A. F. Bastawros, D. R. J. Baruchel, M. Schlenker, D. Van Dyck,
Mumm, A. G. Evans, D. J. Sypeck, H. H. g. ªPhase contrast tomograhyº in X-ray Tomo-
Wadley, Acta Mater. 1998, 46, 3582. grahy in Material Science, J. Baruchel, J.-Y.
4. E. Maire, F. Wattebled, J. Y. Buffiere, G. Peix, Buffiere, E. Maire, P. Merle, G. Peix (eds),
in Proc. Euromat `99 Conf., Vol. 5, Munich, Hermes, Paris 2000, p. 115 125.
T. W. Clyne, F. Simancik (eds), Wiley VCH, 8. L. A. Feldkamp, L. C. Davis, J. W. Kress.
Weinheim 1999, p. 68. J. Opt. Soc. Am. 1984, 1, 612.
5. H. P. Degischer, A. Kottar, B. Foroughi, 9. A. M. Harte, N. A. Fleck, M. F. Ashby, Acta
Ch. 12 in X-ray Tomograhy in Material Science, Mater. 1999, 47, 2511.
J. Baruchel, J.-Y. Buffiere, E. Maire, P. Merle,
G. Peix (eds), Hermes, Paris 2000.
156 4.3 Considerations on Quality Features
4.3
Considerations on Quality Features
H. P. Degischer and A. Kottar
The quality of a material is defined by its properties with respect to the expected
performance. A general demand for engineering materials (not necessarily for ob-
jects of arts) is reproducibility in the sense that the product exhibits the same prop-
erties within specified limits delivered in different samples or batches in order to
be predictable in its performance and an engineer can design a reliable component
suitable for series production. Material science relates properties to microstructural
features of bulk materials, for which quality criteria can be deduced defining limits
for structural variations (for instance grain size) and for inhomogeneities (for in-
stance frequency and size of inclusions). Usually the processing steps producing
the microstructure are specified to guarantee the correlated properties.
4.3.1
Introduction
Cellular metals are a new class of materials on the brink to industrial application
and therefore the producers have to develop specifications for their product with
respect to the application aimed for. In Chapter 7 various applications are pro-
posed, each requiring different property profiles. Applications in architecture ob-
viously will require different quality criteria than heat exchangers or crash absor-
bers. A design guide [1] based on the available properties of different cellular me-
tals and material laws has been offered to the engineers for product development.
The great variety of processes and products presents a considerable challenge for
the selection of the appropriate cellular material for a potential application.
4.3.2
Non-Uniformity of Cellular Metals
Regular cellular structures like those considered for modeling [2] can hardly be pro-
duced. The most regular 3D structures are achieved in lattice block material (LBM)
[3], which can be simulated by a 3D periodic array of volume elements that repre-
sent the main structural features and exhibit the corresponding properties. The be-
havior of any macroscopic part made of LBM can be calculated (see Chapter 6) by
means of known properties of the representative volume (RV), which contains just
as many cells that the volume related properties do not change by adding more
cells, but would change if one cell were taken off. RV is the smallest volume exhi-
biting the properties of the corresponding cellular structure of a given material
with a certain cell architecture yielding an uniform average density. Consequently,
no dimension of a component should be smaller than the RV, if the performance of
the specified cellular structure were required. Any disturbance in the regularity of
ideal cellular microgeometry (see Section 4.1) could be quantified microscopically

to assign a material quality level.
The engineers expect narrow variations and guaranteed lower limits (in some
cases upper limits as well) of properties of the materials considered, which depend
neither on dimension nor on shape and are valid for each segment of the compo-
nent. A RV should contain all the geometrical features that determine the property
profile of intrinsically heterogeneous materials like cellular metals. In the case where
the geometrical features of a cellular structure scatter in a certain statistical range
within the sample, the question arises, which variations (for instance in cell size or
crack probability) are typical for the cell architecture and can be taken into account
for the definition of a RV (examples are given in Chapter 6.1). Features not being
accounted for in such a RV have to be classified as structural defects (for instance
over-sized pores). Anyhow the frequency and the extent of defects and their effects
on properties have to be known in order to be able to qualify a cellular material [4].
Most real cellular metals exhibit a further type of irregularity besides that of geo-
metrical architecture: the mass may not be evenly distributed along the Plateau
borders and cell walls [5], furthermore the relation between pore volume and sur-
rounding material may vary, resulting in a non-uniform density distribution within
the sample [6]. The local density distribution is greatly affected by geometrical de-
fects as well (for instance an over-sized pore surrounded by thin cell walls causes a
local density minimum). Non uniform mass distributions may occur in cellular
metals in combination with variations in the cell architecture [7].
Cellular materials that are produced as shaped parts with or without skin (either
by casting or by foaming) could yield properties characteristic only for the given
shape fabricated by a certain process. Still, such parts should exhibit reproducible
properties for series applications, not only macroscopically over the whole part, but
as well locally where the load is transferred. The range of scattering of these prop-
erties will define the quality with respect to the requirements. Figure 4.3-1 shows a
skeleton of a cross section of an irregular cellular structure, where no common RV
smaller than the whole image could be defined. This problem exists strictly for
castings in a similar way, where material properties depend on the solidification
process and vary across the thickness of the part. If one puts up with the inhomo-
geneities of cellular metals, these have to be quantified in their range of variation
and with respect to their location within the sample. An irregularly structured
sponge with non-uniform density distribution can be considered as a multiphase
material consisting of an arrangement of sub-domains of various cellular materials
exhibiting different properties each (see Chapter 6.2). The non-uniformity can be
described by different volume elements (sub-domains). Each sub-domain consists
of an ensemble of cells of similar microgeometry and density at least as big as the
corresponding RV. If a sub-domain is big enough, it can be subdivided into a num-
ber of equal RVs. In the cross section of the irregular cell structure in Fig. 4.3-1,
some regions representing different cell structures are marked.
Shape-independent quality criteria for cellular metals for engineering application
can only be defined for semi-products like for instance Alporas or Cymat material
delivered as blocks. Net shape cellular parts have to be considered with respect to
Figure 4.3-1. Skeleton of a cross section of an

irregular cellular structure (compact powder
foamed aluminum), where three regions of
different cell structures are marked, each of them
can be regarded as a reasonably uniform cell
architecture that can define a mesoscopic
sub-domain.
their specific loading conditions related to the geometry and to the surface. In a
first approach, the cellular metal or part is defined by its processing route, although
the interaction of the different processing parameters cannot be fully correlated to
the produced cellular structure yet. Each type of product may be characterized by
investigating the statistics of achieved properties and the correlated structural fea-
tures accessible to quality control [7]. The range of scatter of properties acceptable
for the envisaged service performance has to be defined, which can be related to
the scatter of structural features. Beyond specification limits, defects can be de-
fined, which have to be weighted in their influence on the service properties. Ac-
cording to the actual development stage, it can be expected that some defects could
be eliminated or considerably reduced by improved processing techniques.
A hierarchical approach is proposed for defining quality criteria of cellular me-
tals by addressing different description levels according to the characterization le-
vels described in Section 4.1: macro-, meso-, and microscopic. Corresponding fea-
tures are to be defined that can be quantified and validated with respect to material
properties.
The average density of a part or component is an example of a macroscopic level.
The question is: How reproducible is the density of a given part in series produc-
tion? How much variation is tolerable with regard to the scatter of related proper-
ties? Even for a given average density of a part the properties might scatter owing
to the local scatter of mass densities within the part, which may vary on a meso-
scopic scale. Going into further structural details on the microscopic level, the var-
iations in shape factors, in dimensional and microstructural features of the cell
walls and nodes have to be quantified statistically and assessed with respect to tol-
erable variations or irregularities beyond.
4.3.3
Macroscopic Parameters
All those parameters are referred to as macroscopic, which can be assigned to the
whole cellular part without looking into the structure in detail: the general appear-
ance of a cellular architecture, the composition of the ingredients, and the process-
ing method.
4.3.3.1 Type of Cellular Metal

The terms for identification of cellular metals are defined in Chapter 2. The archi-
tecture of the cellular structures can be deduced qualitatively from the processing
route that defines primarily the type of porosity (open or closed). The degree of
open porosity can be determined by measurements of the permeability for liquids
or gas. Cellular structures produced by foaming are of closed cell character in the
liquid state, but, contrary to polymer foams, metal foams suffer from shrinkage
porosity and hot tearing during solidification and further cooling (see Section 4.1),
which make them permeable to gas and wetting low viscosity liquids. The total
surface, inner and outer, could be determined by BET gas adsorption techniques.
The composition of the ingredients (the metal and the foaming additives) can be
quantified. Usually the ingredients are specified and not the resulting composition
of the matrix metal. The relatively large surface may extract some elements from
the alloy, which may change the composition of plateau borders and cell walls
slightly due to eventual chemical reactions with the surrounding media (for in-
stance surface oxides formed predominantly by one or the other constituent).
The sensitivity of the properties of cellular metals on slight changes in composi-
tions has not been identified as a major concern. In the contrary, impurities
have been found to improve foamability promoting the use of secondary alloys
for foam production (see Section 2.1.3). The oxide content (analogous to metal
powder quality specification) could be analyzed to estimate the portion of metallic
components bound by the surface. In the case of foams made from metal matrix
composites (like Cymat or FORMGRIP [8]), usually the particulate content of the
precursor melt is given, which might differ from the local particulate concentration
within the cellular structure. The overall microstructural condition of the solid
metal can be deduced roughly from the manufacturing parameters like cooling
rates, heat treatment and plastic deformation.
4.3.3.2 Surface and Dimensions

The surface of cellular parts has to be considered separately disregarding any addi-
tional surface treatments: either the cellular material has been machined produc-
ing free ends of the cell walls or it was foamed in a mold forming a more or less
closed skin. Useful quality criteria for the surface depend on the service environ-
ment and greatly on the way the cellular part is connected to the surrounding
system.
In Section 7.3 the problem of deforming the cell walls during mechanical cutting
operations is addressed. The degree of such deformations could be of importance
for following joining operations. The enlargement of outer surface in comparison
with a carefully prepared cross section could be employed to quantify the smearing
of cross sections of elements of the cellular structure by machining.
Foamed and cast cellular metals are usually surrounded by a metal skin. The
oxide content will be much higher than on interior cell surfaces, because that
skin was exposed directly to the atmosphere in the liquid state and during solidi-
fication and cooling. The roughness depends greatly on the used mold and its coat-
ing and may be qualified similar to sand casting quality. Attempts to measure the
thickness of the skin by eddy current tests did not succeed so far. A skin region of
about 1.5 mm thickness can be deduced from Fig. 4.3-2 of an Alulight sample,
where the local density is almost twice of that in the center of the plate. Still the
skin cannot be regarded as solid metal, but consists of porous material.
There are as well pores open to the environment in such skins, which can be
detected by dye penetrant tests [6]. Although these pores may be in the range of
tenths of a millimeter in diameter, they form a passage to the cells below, which
might be of a few millimeters in diameter. Such open pores might adsorb humidity
and thus enhance corrosion attack. These pores prohibit chemical treatments like
anodizing, because the electrolyte penetrates and can hardly be removed. On the
other hand, if those pores are sealed by a coating, which has to be baked, it
might form blisters when the temperature rises and the internal gas pressure in-
creases. Ambient temperature coating processes with a strong sealing effect can be
applied successfully [9]. The number of surface pores per surface area and their
distribution of diameters are quantifiable surface quality features. A pore density
of 1-3 pores/cm2 has been measured on Alulight samples [6].
The dimensional accuracy of net shape metal foam parts depends on the thermal
stability of the mold and the uniform shrinkage of the foam during solidification
and cooling. The cell-size distribution, especially along the surface of the part, af-
fects the surface roughness. Big pores (in the range of one centimeter) may cause
dents due to the drastic reduction of pressure of the gas in the closed pores during
cooling. The achievable tolerances for net shape metal foams may be in the range
of percent, but no detailed study has been published so far.
Figure 4.3-2. Local density distribution across a

20 mm thick Alulight plate (r 0.57 g/cm3) deter-
mined by computed X-ray tomography of a resolu-
tion of I0.3 mm.
4.3.3.3 Apparent Density

The apparent density of a cellular metal part still is the most important selection
criterion. It can be measured by the Archimedes principle provided the surface
can be sealed to avoid penetration of the test liquid. Foamed net shape parts
with a skin can be measured in high viscosity liquids preventing penetration
into the surface pores. Generally, a measurement uncertainty of about e2 % in
the apparent density of a part has to be reckoned with. The result includes the sur-
face skin, the density and thickness of which are usually unknown. This may cause
a considerable uncertainty in the apparent density of the interior structure. The ap-
parent density of the whole sample referred to in Fig. 4.3-2 for instance, with a
high density skin layer occupying about 15 % of its volume, is overestimated by
about 15 % (0.57 g/cm3 for the whole sample, without skin about 0.5 g/cm3).
The reproducibility of the apparent density depends on the processing tech-
nique: cellular metals produced on the basis of polymer foam space holders are
highly reproducible, but powder compact foaming achieved so far only about
e15 % reproducibility in the apparent density of similar samples from small series
prototype production. An overall non-uniform mass distribution can be identified
by determining if there is a difference in the position of the center of gravity of the
shape and that of the mass.
4.3.3.4 Properties
Macroscopic properties of test samples are described in Chapter 5, where they are
usually related to the apparent density of the specimens. The observed scatter in
mechanical properties, like the plateau stress for instance, depends very much
on the uniformity of the cellular structure. In general, the properties are closely
related to the quality of the cellular architecture. The scatter in the test results in-
dicate the variations in structure of the individual test specimens. Most of the tests
are executed without knowing the microgeometry of the test sample exactly, there-
fore there is little evidence on the correlation with structural features, especially
when the tests are destructive. A systematic correlation of properties to the
three-dimensional cellular architecture is highly desirable in order to give rise to
process developments aiming for improvements in the reliability of the service
related performance of the foams.
4.3.4
Microscopic Features
In terms of microscopic features the individual cells are to be considered: the mi-
crostructure of the metal forming the cell walls and their topological parameters
described in Section 4.1.
4.3.4.1 Microstructure of the Metal

The microstructural features of the metal, the cellular structure is made of, are
alike any metal: grain size, dendritic sub-structure, phase distributions, segrega-
tion, precipitates and dislocation densities. These are considered in Section 4.1
in detail. Typical features for foamed metals are remnants of foaming agents, dis-
tribution of additives to increase viscosity, the consequences of the large internal
surface on solidification, of grain growth with only partial impingement, and of
temperature gradients during cooling depending significantly on the position in
the sample. The resulting local properties of the metal are presumably not equal
throughout the cellular structure. An important quality criterion is again the uni-
formity of the microstructure throughout the sample, although characteristic dif-
ferences between Plateau borders and cell walls are inherent. Owing to the
small dimensions of cell walls comparable with the grain size, the mechanical,
chemical and physical properties will differ with respect to those of the polycrystal-
line bulk solid. Generally it has to be assumed that the material properties of the
metal in bulk are not identical to those of the same material in the cell structure. In
the case of molten metal casting and foaming processes, there will be a difference
between faster cooled surface near regions and the slowly cooled interior for alloys
sensitive to cooling rates, especially if the cellular aluminum alloys are age harden-
ing. Some indications for PCF aluminum are given elsewhere [10,11].
4.3.4.2 Geometrical Features

The shape of cells and cell walls has been investigated so far by metallographic
methods either by light microscopy or by scanning electron microscopy. Average
cell-size distributions from cross sections are only representative when the pores
are spherical. The presentation of pore-volume distributions is more relevant
than equivalent linear pore diameter statistics. Average pore size figures are of little
significance because the width of the distribution determines the properties ac-
cording to the weakest link principle. The distributions may be Gaussian only
for hollow sphere structures and those produced by replication techniques,
where the narrow width of the distribution provides a quality measure. Generally,
the shape of the pore volume distribution curves could provide a quality criterion
for the cellular structure. The portion of pores bigger than the average is especially
important for the properties of foamed metals. In a first approach the statistics of
ªover-sizedº pores could be used as quality criterion: for instance, the volume frac-
tion occupied by the biggest 10 % or the pores bigger than 20 % of the sample's
thickness. Not only the size of the big pores is relevant, but as well their positions
and the microgeometry of surrounding pores. A centrally placed big pore would
not affect the stiffness of a component and the strength could be compensated
by surrounding small pores (see Chapter 6.1.3). A surface near big pore could
be harmful at the position of load transfer, where the local pore size distributions
would be of interest.
Some attempts have been made to determine the three-dimensional geometry of
pores by computed X-ray microtomography [5,12]. It gives the most exact tool for
the 3D description of cellular structures. The determination of the maximum and

minimum pore diameters and their orientation distribution would be relevant in-
formation needed to identify an anisotropy of the architecture of the cellular struc-
ture. Examples of the potential of X-ray microtomography are shown in Fig. 4.3-3
and Fig. 4.3-4 and elaborated in [13,14] and in Section 4.2. The 2D architecture of
cellular light metal can be represented similarly to metallographic cross sections
but nondestructively by selecting the intensity distribution of an arbitrary plane
of the 3D microtomography. The examples given in Fig. 4.3-3 and Fig. 4.3-4
have been measured at Bundesanstalt fuÈr MaterialpruÈfung, Berlin [6] and represent
the mass distribution for different aluminum foams in Fig. 4.3-3 within a slice of
40 mm and an aluminum metal matrix foamed in the melt [15] in Fig. 4.3-4 (slice
of 0.1 mm). The difference in the architecture (pore size and shape, frequency of
irregularities) can be qualitatively observed. Cross sections of Plateau borders and
cell walls are only representative if they are oriented end on. Cell walls oriented
parallel to the imaged slice appear as mass concentrations like thick nodes, espe-
cially when the cell size is small like in Fig. 4.3-3a and c. The 3D architecture can
be visualized by iso-surfaces like presented in Fig. 4.3-4b. It can be observed on a
screen by scrolling the data set of 3D tomography slice by slice or by means of an
observation point travelling through the cells by techniques of virtual reality [16].
The distribution of mass between nodes and Plateau borders in open-cell struc-
tures, and additionally in cell walls of closed cell foams can be estimated from me-
tallographic sections or determined by computed X-ray microtomography [5,14].
The uniformity of this material distribution parameter is another quality criterion.
The connectivity of pores can only be determined by computed X-ray microtomo-
graphy [5]. This parameter affects the permeability, some thermophysical and
acoustic properties.
The shape of the struts or cell walls or edges can be compared with ideally
straight members or shells. The buckling could be quantified by the relation be-
tween shortest distance between the ends and the real curved length or by the cir-
cularity in 2D and the sphericity in 3D. Most of these geometric microstructural
features are described in detail in Section 4.1. The statistics of these factors and
Figure 4.3-3. Cross sections of mass distribu- Alulight AlMgSiCu of r 0.39 g/cm3, c) Alu-
tions computed from X-ray microtomograms light AlSi10Mg r 0.37 g/cm3, d) same mate-
(voxel size 40 mm q 40 mm q 40 mm) of foamed rial as (c), but specially foamed small cell ar-
aluminum: a) Alporas of r 0.45 g/cm3, b) chitecture r 0.56 g/cm3.
Figure 4.3-4. Mass distributions computed polygonal cell walls (same magnification as
from micro X-ray tomograms of foamed Al- Fig. 4.3-3; note that not all of them are end-on),
Si9Mg/SiC/20p melt of r 0.11 g/cm3 (voxel b) 3D cell reconstruction of a cuboid volume
size 0.1 mm q 0.1 mm q 0.1 mm): a) cross element (22 mm q 18 mm q 16 mm) indi-
section showing different plateau borders and cating curved shapes of the cell walls.
the identification of predominantly buckled orientations can be correlated to the

initial deformation behavior in tension and compression [17,18] (see Chapter
6.1.3).
4.3.4.3 Microdefects
Of course, pores created in cellular metals due to processing imperfections like
shrinkage pores, imperfect sintering, spraying pores etc., may not to be considered
as essential defects. But in general the consistence of the percolating solid should
be of similar metallic properties throughout the cellular structure. Cellular struc-
tures processed in the melt or at elevated temperatures suffer thermal stresses dur-
ing solidification and cooling. Hot tearing and cracking is very probable because of
the multi-axial stress situations created in walls of little strength (see Section 4.1).
Almost 100 % connectivity of the pores have been determined within powder com-
pact foamed aluminum [5], although almost 40 % of the solid is located within the
cell walls. Cell-wall cracks have been observed frequently in powder compact foams
[19].
The material properties of the metal in the cellular structure have to be consid-
ered to be of poorer quality than those of the same metal in bulk. For those pro-
duced from the melt it is expected that the alloy exhibits the properties of a low
quality casting. If the frequency of microdefects depends on the position within
a sample, their effect should be considered as deviation from uniformity.
4.3.5
Mesoscopic Features
The mesoscopic description of real cellular metals assumes that a sample or part
can be described as an arrangement of volume elements, each representing a char-
acteristic microgeometry. The local inhomogeneities in architecture and in mass
distributions can be classified by different mean cellular structures approximating
the properties of each sub-domain of the material. A mesoscopic model for uniaxial
compression has been proposed [20,21], which divides the samples into layers of
cellular materials with different apparent densities. Generalizing this concept,
any sample or part can be divided into a 3D arrangement of sub-domains of differ-
ent cellular phases of known features and properties. Thus the cellular sample is
considered as a multiphase material (see Fig. 4.3-1), where a cellular phase is
defined by its composition, microgeometry, and mass distribution occupying the
volume of a sub-domain.
The mesoscopic behavior of each sub-domain corresponds to that of the asso-
ciated RV, so that each sub-domain behaves like a cellular material to which macro-
scopic material laws of cellular metals can be assigned. In the extreme case, where
the sub-domains were smaller than RVs (i. e., most cells were essentially different
from their neighbors), such a mesoscopic approximation would not make sense
and complex finite element modeling of the real microstructure would be neces-
sary as shown in Section 4.2. In the best case of samples of uniform microgeo-
metry like Duocel, Incofoam, or those cast by replication, the mesoscopic consid-
eration is superfluous and the macroscopic treatment would result satisfactorily.
In general, the range of variations in size, volume fractions and properties of
the mesoscopic sub-domains can be quantified providing a quality measure on uni-
formity of the considered part.
4.3.5.1 Geometry of Cellular Structure

A sub-domain of a sample or part has to be big enough to behave like a cellular
metal the mechanical properties of which can be approximated by continuum me-
chanics. It should measure about five cells in diameter, if their shape is similar to
each other [22]. This criterion would require a statistically representative number
of more than 100 cells per representative volume. If the shape factors of the
cells varied significantly, the sub-domain should contain as many cells as necessary
to represent the statistical scatter of geometrical features.
In general, the processing method of cellular metals defines a certain type of cell
architecture. Most foamed metals exhibit some gradation in their structure depend-
ing on the dimension of the part and/or on an anisotropy in foaming [7,23] and in
solidification [1]. If the variation of the geometrical features of the cells is not
statistical, but can be related to different regions of the samples, arrangements
of uniform sub-domains can be defined to represent a graded structure. For
instance a sub-domain of the thickness of the surface layer representing the skin
as indicated in Fig. 4.3-2. Systematic variation in the cell architecture can be pro-
duced by intentionally graded or anisotropically cast cellular metals [24]. If the

shape factors of the pores or the material properties are preferentially oriented,
the anisotropic behavior can be assigned to the sub-domains.
The cell-size distribution cannot be considered as a mesoscopic parameter. There
is no directly quantitative measure of mesoscopic geometrical features. Quantified
data would have to be determined microscopically by three-dimensional analysis
and evaluated statistically. Anyhow the pre-dominating geometrical features can
be assigned to each sub-domain.
4.3.5.2 Density Distribution

If the microgeometry of the cells were uniform, except the local variations in cell
size, the density distribution would become non-uniform. Powder compact foamed
samples usually contain bigger cells in the center of samples of a few centimeters
in thickness yielding a graded density distribution (similar to that shown in
Fig. 4.3-2). Furthermore it has to be admitted that inhomogeneities in foamed
metals (over-sized pores and local mass concentrations) cannot be fully controlled
so far. The point is to take them into account by simulation. The non-uniformity in
solid distribution of foamed metals can be characterized by inhomogeneous den-
sity distribution [6]. Regions of similar mass densities can be approximated by a
mean density value [25]; thus the sample can be divided into sub-domains of dif-
ferent mass densities. A procedure to produce such sub-domains will be described
in Chapter 4.3.7.
4.3.6
Systematics of Quality Features
The features characterizing parts or samples made of cellular metals are correlated
to the structural hierarchy described above in a matrix presented in Table 4.3-1,
from which quantifiable magnitudes for quality considerations are deduced.
Most of the microscopical parameters are described in Section 4.1 and require con-
siderable effort to be determined statistically, especially in 3D. The macroscopic
magnitudes can be obtained from the manufacturer and can be complemented
by nondestructive tests except the mechanical properties. The mechanical proper-
ties of uniform cellular metal parts can be calculated by simulating the geometrical
condition with input data from test samples of equal materials. The properties of
non-uniform cellular parts (either intentionally graded or inhomogeneous due to
the processing conditions) can be estimated from their mesoscopic structure
when the scaling laws with respect to the relevant microstructural features are
known.
Mesoscopic parameters can be defined on the condition that ensembles of cells
exhibit similar microstructural features, which form more or less uniform sub-do-
mains of a sample of cellular metal. An attempt is made to characterize a cellular
material without determining microscopic features quantitatively. Irregularities in
a certain depth from the surface can be detected by ultrasonic spectroscopy [26] as
Table 4.3-1. Matrix of features of considered quality aspects and correlated quantities within the structural hierarchy of cellular metals.
Feature Microscopic Mesoscopic Macroscopic

Magnitudes Quantifiability Magnitudes Quantifiability Magnitudes Quantifiability
Cell Cell size, shape Cell volume histo- Definition of vol- 3D data set of arrange- Processing method, Processing para-
architecture and orientation gram shape, shape umes of sub-do- ment of sub-domains cell type, permeabil- meters, category of
of pores and factors and orienta- mains of similar in size and shape com- ity architecture, perme-
cell walls, tion distributions, microgeometry posing the cellular ability coefficient,
connectivity connectivity para- component surface/volume ratio
meter
Dimensional Dimensions of Statistics of dimen- Sensitivity of sub- Dependence of sub-do- Achievable toler- Lateral dimensions,
accuracy constituents of sions of cell edges/ domain volumes main arrangement on ances in dimensions radii at corners
cells, irregulari- faces; buckling, on averaging para- the averaging limits for
ties wiggles, circularity, meters microstructural features
sphericity, cracks
Surfaces Composition, Definition of surface Interface between Boundary conditions Surface formation Fabrication method,
structure, thick- region/skin, form of sub-domains or for interfaces of sub- and topology properties, rough-
ness of outer and constituents, chemical sections of outer domains; specific phase ness measure, sur-
inner surface composition, topology surface for outer surface region face porosity index
Metal matrix Grain structure, Grain size and inter- Similar condition Correspondence of sub- Alloy composition, Chemical analysis of
phase distribu- dendritic spacing, in- (microstructure) of domain volumes with additives, metal con- ingredients, thermo-
tions clusions, precipitates the solid uniform relevant mi- dition mechanical manu-
in function of position crostructure facturing para-
in cell/sample meters
Mass Fraction of mass Mass distribution Similar solid dis- Correspondence of sub- Average, apparent Absolute or relative
distribution in cell constitu- coefficient between tribution, uniform domain volumes with mass density; center mass density, differ-
ents faces, Plateau borders apparent density continuous uniform of gravity ence of centers of
and nodes mass distributions gravity of geometry
and of mass
4 Characterization of Cellular Metals
Properties Metal properties Physical and mechan- Properties of an Criteria for definition of Service properties of Experimental data
167
on microscopic ical properties of solid ensemble of simi- properties of sub- component and simulation re-
scale different in function of position lar cells domains or RV, scaling sults with scatter
from bulk in cell and sample laws and their accuracy and validity limits
long as the cell architecture is essentially uniform. In the case of regular Duocel
samples relative density variations could be observed by eddy current measure-
ments [27]. The following paragraphs describe the possibilities to determine meso-
scopic parameters of non-uniform cellular metals and discuss their applicability for
quality criteria that can predict, if a sample's behavior complies with the specifica-
tions of service requirements.
4.3.7
Approximation of a Cellular Structure by a Continuum
For simplicity it is assumed that the microstructure of the metal and the type of
cell architecture are essentially defined by the processing method. The remaining
variables are the local density distribution and the properties of the solid material,
which have to be calibrated. A non-uniform object can be divided into sub-domains
to which the appropriate material laws can be applied to simulate the whole
part's performance. For this purpose, it is convenient to transform the cellular
structure into a continuum, which behaves like a cellular architecture (crushable
material). The experimental basis is provided by computed X-ray tomography
[13] of the 3D metal distribution of a sample, which will be locally averaged to
yield density values in each point of the sample smoothing away the originally
cellular structure.
4.3.7.1 Calculation of Density Maps

The density mapping method is based on determining a local density value at each
point within the cellular structure by averaging the mass located in an appropriate
neighborhood of the points. The cellular structure can be described ideally by a
function r(r) that takes the values r(r) 0 in pores and r(r) rS in cell walls,
edges, and nodes, where rS is the density of the solid material. In general a local
density ra at r can be defined as
Z
ra (r) g(r s r 0 )r(r 0 )d3 r 0 (1)
where
R g is a weight function that has to fulfill the normalization condition
g(u)d3u = 1. Equation (1) is also written in the form ra g*r, which denotes
the convolution of the function r with g. Cellular structures can be investigated
by X-ray radiography or by X-ray computed tomography as described in
Section 4.2. The density distribution is represented by a voxelized data set mXCT(rijk)
that can be used to calculate a density mapping as defined in Eq. (1). Due to the
limited resolution of the XCT equipment mXCT(rijk) is related to the linear mass
attenuation m(r) of the object by
Z
mXCT (rijk ) PSF(rijk s r 0 )m(r 0 )d3 r 0 PSF m (2)
where the point spread function PSF describes the ªblurringº of the object function
m(r) [28].
For example, the micro XCT slices in Fig. 4.3-5a and Fig. 4.3-5d are a very good
representation of the real foam (i. e., the PSF is a sharp peak), whereas the corre-
sponding slices obtained by medical XCT (Fig. 4.3-5b and Fig. 4.3-5e) are of signif-
icantly lower resolution (the PSF is relatively broad). Since the metallic foams in-
vestigated herein consist of a single alloy, m(r) is proportional to r(r) and analo-
gously to Eq.(2), rXCT PSF*r can be defined. Averaging these rXCT values with
the weight function w yields raaccording Eq.(1)
X X Z
w(rijk s rlmn )rXCT (rlmn w(rijk s rlmn ) PSF(rlmn s r 0 )r(r 0 )d3 r 0
l;m;n l;m;n
Z X Z
w(rijk s rlmn )PSF(rlmn s r 0 )r(r 0 )d3 r 0 g(r 0 srijk )r(r 0 )d3 r 0 ra (rijk ) (3)
l;m;n
It should be noted that rXCT is already a function of the form of Eq. (1), i. e., rXCT
can be understood as a density mapping performed by the tomograph itself and a
Figure 4.3-5. Comparison of a micro XCT with c) calculated density map using averaging
pictures obtained from a human tomograph volumes of 6.8 mm q 6.8 mm q 6.8 mm.
and calculated density distributions for Alulight Alporas of r 0.45 g/cm3 d) micro XCT slice of
Si12Mg of r 0.5 g/cm3: a) high-resolution 40 mm, e) human XCT of a 3 mm slice (0.9 mm
micro XCT for a slice thickness of 40 mm (40 mm q 0.9 mm q 3 mm voxel), f) density map
q 40 mm q 40 mm voxel), b) human XCT of a calculated using the same averaging volume
1 mm slice (0.6 mm q 0.6 mm q 1 mm voxel), as in (c).
Figure 4.3-6. Illustration of the principle of 3D

density mapping based on density values for each
XCT voxel: the density in the center of the averaging
cuboid given by l q n q m voxels is the mean
average density of that averaging volume.
second mathematical averaging process reveals the final mapping ra (see Fig. 4.3-
5c and Fig. 4.3-5f). The density mappings in this section are calculated by arith-
metic mean values of the voxels rXCT(rijk) over a cuboid (see Fig. 4.3-6)
1X
ra (rijk ) rXCT (rlmn ) (4)
N l;m;n
where l, m, n run over the voxel indices of the selected cuboidal surrounding of the
location rijk and N is the number of voxels, i. e., the weight function in Eq. (3) is w
1/N within the cuboid and w 0 otherwise.
Figure 4.3-7 shows a one-dimensional model r(x) of a cellular metal with walls of
the solid density rS and embedded empty pores. Convolution of r(x) with the PSF
results in the depicted rXCT(xi) values and further averaging over intervals [xi 34,
xi 34] with N 69 yields the mean local density function ra(xi).
As long as the averaging volume of the density mapping is bigger than the reso-
lution of the tomography, the resulting density map is almost independent of the
XCT resolution [29]. Using averaging cuboids approximately as big as the cells
suppresses microstructural details, but allows to determine local differences in
Figure 4.3-7. One-dimensional model of the Eq.(4) for each voxel with an averaging interval
density mapping procedure: the real mass of xi e 34 voxels yields the smoothed density
distribution r(x) is imaged by the XCT data values ra(xi). rm is the average density; the
rXCT(xi) of the voxels, which are resulting from variations e Dr correspond to e rm/3.
the convolution of r(x) with the PSF; applying
Figure 4.3-8. Example of the effect of the size 10 mm apart; 2nd to 5th column, calculated
of the averaging volume on the presentation of density distributions of the same planes for
density variations in a AlMgSiCu/Al2O3/20p averaging volumes of 4 mm q 4 mm q 4 mm,
melt foamed sample of r 0.25 g/cm3 [15]: 1st 6 mm q 6 mm q 6 mm, 9.5 mm q 9.5 mm q
column, microscopic XCT images (0.1 mm q 9.5 mm, and 15 mm q 15 mm q 15 mm.
0.1 mm q 0.1 mm voxel) of parallel planes
the material distribution in the cells on a mesoscopic scale. Reasonable averaging

volumes should contain at least essential portions of at least one cell so that there
results no point of zero density. Appropriate averaging volumes for Alporas mea-
sure 4 mm q 4 mm q 4 mm, whereas about 6 mm q 6 mm q 6 mm has
been shown to be suitable for all Alulight foams. Figure 4.3-8 shows an example
of AlMgSiCu/Al2O3/20p foamed in the melt for different averaging volumes. It
contains a considerable fraction of pores bigger than 6 mm [15]. The intensity var-
iations due to the cell walls disappear from 9.5 mm q 9.5 mm q 9.5 mm aver-
aging volume upwards, which is an appropriate mesoscopic scale still representing
the scatter in density within the foam. Even the averaging volume as big as 15 mm
q 15 mm q 15 mm shows some non-uniformity of the sample.
Figure 4.3-9 shows examples of 3D density distributions of Alporas and Alulight
samples for the same averaging volume. The cellular structure disappears, but the
non-uniformity of the mesoscopic mass distribution is clearly visible. The local
a)
b)
Figure 4.3-9. Illustration of the 3D local den- r 0.45 g/cm3; b) Alulight AlSi12Mg of
sity distribution obtained from 4 mm q 4 mm r 0.5 g/cm3; the gray value represents the
q 4 mm averaging volumes for a series indicated deviation from the apparent density
of six layers at distances of 3.7 mm of the same rm (white high density).
samples as in Fig. 4.3-5: a) Alporas of
mass densities for the considered averaging volume size vary in these examples be-
tween 40 % and 160 % of the apparent density of the Alporas sample, but between
0 and more than twice of rm of the Alulight sample. The gray scale presentation in
Fig. 4.3-9 divides the samples into density classes of 0.1 rm. It is clear from the
density distribution of the Alulight sample, that no uniform material behavior
can be expected.
4.3.7.2 Representation of Non-Uniformity of Densities

Histograms of density distributions can be calculated for specified averaging
volumes, which are at least as big as the majority of the cells. Alporas shows the
narrowest scatter of local density in Fig. 4.3-10. Both types of Alulight have their
maximum frequency below the average density caused by the extended tail in
the higher density range. Isolated high density regions may not affect the mechan-
ical properties significantly according to the weakest link principle, but would
represent a surplus of material. The shape of density histograms seems to be
characteristic for the processing method of a cellular metal [15].
Density maps can be used to quantify local density variations. The line of average
density rm and arbitrary limits of density variations of eDr are indicated in the
linear model in Fig. 4.3-7. An allowable density variation can be defined and the
volume fraction of material with mass densities below or above the lower and
upper limit, can be calculated. The 3D arrangement of volumes of extreme density
values can be presented by iso-surfaces of densities smaller or bigger than a given
limit as shown in Fig. 4.3-11. The local position, orientation and connectivity of
such hard and soft sections may be important as well for the performance of a
Figure 4.3-10. Histograms of density distri-

butions of different metal foams with similar
apparent density of r 0.4 g/cm3, showing
the mesoscopic differences for averaging
volumes of 6 mm q 6 mm q 6 mm.
Figure 4.3-11. Presentation of hard and soft sity maps with averaging volumes of 6 mm q
regions within an Alulight AlSi12Mg sample (a 6 mm q 6 mm: a) iso-surface of hard sub-do-
b 22 mm, c 30 mm) of apparent rm mains with r i 1.33 rm; b) iso-surface of soft
0.5 g/cm3 by iso-surfaces computed from den- sub-domains with r I 0.67 rm.
part [30]. If there were cells bigger than the chosen averaging volume, so that
points result in the sample where no contrast is found within the surrounding cu-
boid, the density mapping would contain co-ordinates of zero density. The approx-
imate size, volume fraction, and position of pores bigger than the averaging vo-
lume can be determined by means of the calculated 3D density map. An example
for an Alulight specimen with over-sized pores is shown elsewhere [6].
X-ray radiography can be applied to flat samples. The intensity of the transmitted
columns contains the density integrated over the thickness of the sample yielding
the contrast of pixels in the projected plane. This can be transformed into 2D den-
sity maps by applying the analogous procedure of averaging over m q n pixels in
the image plane. Examples given elsewhere [6] illustrate an improved detectability
of inhomogeneities (like ensembles of big pores, drainage effects, or dense skins)
by planar density mapping. Of course, radiography does not provide any informa-
tion on density variations in the direction of the thickness. A compromise between
3D tomography and radiography is provided by X-ray laminography [31], where the
density distribution of parallel layers of a sample can be determined with moderate
resolution.
4.3.7.3 Mesoscopic Basis for Material Modeling

The earlier developed approach to approximate a non-uniform cellular structure by
an arrangement of sub-domains, which themselves are uniform, can be realized on
the basis of density maps. These density maps are a three-dimensional distribution
of density values ra(xi, yj, zk) defined for every voxel (xi, yj, zk) of the XCT data
source. The determined total range of mass densities can be divided into classes
of density ranges. The ensemble of voxels whose density values lie within one den-
sity range form a sub domain. The average of the density values within the sub-
domains is used as their assigned density value. The widths of density classes
have to be big enough, so that the resulting sub-domains comply with the require-
ment of representing a cellular material, i. e., at least as big as the RV.
The cellular sample can be presented by a 3D, space filling arrangement of N
sub-domains defined by N different values of mass densities. The gray scale con-
tour map in Fig. 4.3-9 can be interpreted as presentation of sub-domains of the
samples in steps of 0.1 rm. The corresponding material modeling and the deduced
simulation of properties of non-uniform cellular metals is described in Section 6.2.
4.3.8
Proposal of Quality Criteria
The main concern is reproducibility of properties of interchangeable products

(aiming for series production) comprising the dimensions of the part and its ser-
vice relevant performance. The usual quality assurance refers to the processing,
which defines the material and the architecture of the cellular structure. The prod-
uct itself can be submitted to quality control regarding the dimensions, the appar-
ent density and the surface topology. The general structure, properties and service
performance can be made plausible by results obtained from corresponding test
samples. An overview of these macroscopic quality features is given in Table 4.3-
2. The specifications for each of them are not yet established. Unfortunately the
accomplishment of the macroscopic quality criteria is not sufficient for a defined
performance of cellular metals with non-uniform cell structures. A mesoscopic
analysis of quality criteria is proposed according to Table 4.3-3.
The mesoscopic quality testing is based on non destructive testing. Computed X-
ray tomography is most powerful tool, the resolution of which is sufficient in the
range below the average cell size. Modern human X-ray tomographs are appropri-
ate for light metals. The derived density maps calculated using an appropriate aver-
aging volume for the investigated type of cell architecture provides the 3D data
base for quantitative analysis of the uniformity of density distributions: density his-
Table 4.3-2. Matrix of features of parts made of cellular metals and correlated macroscopic quality criteria proposing their verification.
Features Quality goals Quality criteria Defect evidence Quality testing
Cell architecture Reproducible and similar Reproducible processing Unsatisfactorily controlled Process records, assessment of
throughout the sample parameters parameters relevant parameter variations
Dimensional accuracy Tolerances of outer Specified variations in Dimensions beyond Contour measurements
dimensions achieved lengths and radii tolerances
Surfaces Reproducible surface Properties, roughness, Dents and blisters Roughness measurement,
features open porosity Surface porosity index Dye penetrant tests,
Chemical analysis, hardness
Metal matrix Defined chemical Uniform solid properties Scatter in chemical analysis Chemical analysis of ingredients
composition of different positions and product
Mass distribution Reproducible density Apparent mass density Scatter in average mass Density measurements,
within specifications density; asymmetry of mass Determination of centers of
distribution gravity of shape and mass
Properties Properties unambiguously Property values within Scatter beyond specifications Representative tests,
related to structural quality specifications Correlation to structure,
Service performance
4 Characterization of Cellular Metals
175
Table 4.3-3. Matrix of features of cellular metals and correlated mesoscopic quality criteria proposing their verification.
Feature Quality goals Quality criteria Defect evidence Quality testing

176
Cell architecture Reproducible space filling Master curves of local Histogram shape, Volume Histogram of mass density dis-
sub-domain arrangement density distributions fractions below and above tribution from XCT; Difference
certain limits of rm and histogram's maxi-
mum; Volume fraction of
r I rm Dr and of r i rm Dr
Dimensional accuracy Sub-domain volumes contain Convergence criterion for Averaging volume size too Averaging volume bigger than
statistically representative size of averaging volume big in relation to sample dominant cell sizes, smaller
number of cells deduced from width of size, large sub-domains of than 1/3 of sample's minimum
density histogram only few not representative thickness; Sensibility tests for
over-sized cells variations of averaging volume
size
Surfaces Outer surface boundary con- Free surface except area of Load transfer area smaller Comparison of modeling with
4.3 Considerations on Quality Features
ditions according to loading, load actuation than interacting sub- deformation experiments re-
independent of interface Spatial displacement of domain corded by XCT or digital video
boundaries sub-domain interfaces Limited correspondence images
of loading, sub-domains,
FE-mesh
Metal matrix Properties insensitive to Experience from microstruc- Definition of critical property Estimated differences of solid
microstructural variations tural examinations of same relevant irregularities properties compared with bulk
within averaging limits type of cellular metal (calibration of input properties)
Mass distribution Properties of sub-domains Determination of location Definition of critical dimen- Definition of critical Dr
represented by the local of soft and hard regions sions, connectivity and posi- Shape, connectivity and size of
average mass density of the tions extreme densities (iso-surfaces)
corresponding volume Check of gradients in main
directions
Properties Appropriate scaling laws and Calibration by experimental Establishing confidentiality Testing stiffness in different di-
constitutive equations tests limits for different types of rections, comparison with com-
known materials puter simulations (e DE(x,y,z)
for small strains)
togram, the volume fractions of the sample with densities outside a certain speci-
fication and the local position of those extremely soft or hard regions. Even the
existence of over-sized pores can be indicated.
The following uniformity criteria are proposed for the example of Alulight
foams:
x the local density variations should be within e30 % around the mean apparent
density;
x regions of smaller mass densities than the specified range (including zero for
over-sized pores) should in any direction not be bigger than 1/3 of the thickness
of the sample;
x the sample's regions with mass densities beyond the specified limits should not
be connected in any preferential orientation causing anisotropy, and should oc-
cupy less than 5 vol.-%.
A stiffness test in different directions is proposed as an experimental check of the
mechanical behavior of a part in order to reveal any significant irregularity and
macroscopic anisotropy. The absolute stiffness values can be compared with simu-
lations of the elastic performance of that part (see Section 6.2). Up to now there is
little experience on the applicability of any quality criterion for cellular metals and
specifications will have to be developed on the basis of the experience that will be
gained from the first series products.
Acknowledgements
The authors are grateful to Leichtmetall Kompetenzzentrum Ranshofen, Austria,

Slovak Academy of Science, Bratislava, Slovakia, and Shinko Wire, Japan for the
provision of test samples. The X-ray tomograms have been recorded at Bundesan-
stalt fuÈr MaterialpruÈfung, Berlin and the Division of Osteoradiology, University of
Vienna.
References
1. M. F. Ashby, A. G. Evans, N. A. Fleck, L. J. Schwartz, D. S. Smith, A. G. Evans, H. N. G.

Gibson, J. W. Hutchinson, H. N. G. Wadley, Wadley (eds), MRS Symp. Proc. Vol. 521,
Metal Foams: A Design Guide, Butterworth- MRS, Warrendale, PA 1998, p. 109 117.
Heinemann, Oxford 2000. 4. J. L. Greensted, in Porous and Cellular
2. L. J. Gibson, M. F. Ashby, Cellular Solids: Materials for Structural Application, D. S.
Structure and Properties, 2nd edn, Cambridge Schwartz, D. S. Smith, A. G. Evans, H. N. G.
University Press, UK 1997. Wadley (eds), MRS Symp. Proc. Vol. 521,
3. M. L. Renauld, A. P. Giamei, M. S. MRS, Warrendale, PA 1998, p. 3 13.
Thompson, J. Priluck; in Porous and Cellular 5. A. Elmoutaouokkail, L. Salvo, E. Maire,
Materials for Structural Application, D. S. G. Peix, in Proc. Cellular Metals and Metal
Foaming Technology (MetFoam 2001), D. S. Schwartz, D. S. Smith, A. G. Evans,

J. Banhart, M. F. Ashby, N. A. Fleck (eds), H. N. G. Wadley (eds), MRS Symp. Proc.
MIT, Bremen 2001, p. 245±250. Vol. 521, MRS, Warrendale, PA 1998,
6. H. P. Degischer, A. Kottar, in Metal Foams p. 53 57.
and Porous Metal Structures, J. Banhart, M. F. 18. A. F. Bastawros, A. G. Evans, Adv. Eng.
Ashby, N. A. Fleck (eds), MIT Verlag, Bremen Mater. 2000, 4, 210 214.
1999, p. 213 220. 19. B. Zettl, S. Stanzl-Tschegg, Materwiss.
7. F. Simancik, in Metal Foams and Porous Werkstofftechn. 2000, 31, 484 487.
Metal Structures, J. Banhart, M. F. Ashby, 20. S. Huschka, S. Hicken, F. J. Arendts, in
N. A. Fleck (eds), MIT Verlag, Bremen 1999, Proc. MetallschaÈume, J. Banhart (ed.), MIT,
p. 235 240. Bremen 1997, p. 189 197.
8. V. Gergely, T. W. Clyne, Adv. Eng. Mater. 21. R. Gradinger, F. G. Rammerstorfer, Acta
2000, 4, 175 178. Mater. 1999, 47, 143 148.
9. Th. Schambron, R. Gramlinger, R. Kretz, in 22. E. W. Andrews, P. R. Oneck, L. J. Gibson,
Proc. Cellular Metals and Metal Foaming Tech- Int. J. Mech. Sci. 2001, 43, 681 699.
nology (MetFoam 2001), J. Banhart, M. F. 23. J. Banhart, J. Baumeister, Metallurgy 1997,
Ashby, N. A. Fleck (eds), MIT, Bremen 2001, 51, 19.
p. 461±466. 24. V. I. Shapovalov, in Porous and Cellular
10. H. P. Degischer, U. Galovsky, R. Gradinger, Materials for Structural Application, D. S.
R. Kretz, F. Simancik, in Proc. MetallschaÈume, Schwartz, D. S. Smith, A. G. Evans, H. N. G.
J. Banhart (ed.), MIT, Bremen 1997, Wadley (eds), MRS Symp. Proc. Vol. 521,
p. 79 90. MRS, Warrendale, PA 1998, p. 281 290.
11. D. Lehmhus, C. Marschner, J. Banhart, 25. B. Kriszt, B. Foroughi, K. Faure,
Materwiss. Werkstofftechn. 2000, 31, 474 476. H. P. Degischer, Mater. Sci. Technol. 2000, 16,
12. B. Illerhaus, J. Goebbels, Materwiss. Werk- 792 796.
stofftechn. 2000, 31, 527 528. 26. A. Wanner, B. Kriszt, Materwiss. Werkstoff-
13. J. Baruchel, J.-Y. Buffiere, E. Maire, P. techn. 2000, 31, 481 483.
Merle, G. Peix (eds), X-Ray Tomograhy in 27. K. P. Dharmasena, H. N. G. Wadley, in
Material Science, Hermes, Paris 2000. Porous and Cellular Materials for Structural
14. H. Bart-Smith, A. F. Bastawros, D. R. Application, D. S. Schwartz, D. S. Smith, A. G.
Mumm, A. G. Evans, D. J. Sypeck, H. N. G. Evans, H. N. G. Wadley (eds), MRS Symp.
Wadley, in Porous and Cellular Materials for Proc. Vol. 521, MRS, Warrendale, PA 1998,
Structural Application, D. S. Schwartz, D. S. p. 171 176.
Smith, A. G. Evans, H. N. G. Wadley (eds), 28. W. A. Kalender, Computed Tomography,
MRS Symp. Proc. Vol. 521, MRS, Warrendale, Publicis MCD, Munich 2000.
PA 1998, p. 71 81 29. A. Kottar, H. P. Degischer, B. Kriszt, Ma-
15. D. Leitlmeier, H. Flankl, in Proc. Cellular terwiss. Werkstofftechn. 2000, 31, 465 469.
Metals and Metal Foaming Technology (Met- 30. H. P. Degischer, B. Kriszt, B. Foroughi,
Foam 2001), J. Banhart, M. F. Ashby, N. A. A. Kottar, in Metal Matrix Composites and
Fleck (eds), MIT, Bremen 2001, p. 171±174. Metallic Foams, Euromat 1999, Vol. 5, T. W.
16. A. H. KoÈnig, H. Doleisch, A. Kottar, B. Clyne, F. Simancik (eds), Wiley-VCH, Wein-
Kriszt, E. GroÈller, in Data Visualisation 2000, heim 2000, p. 74 82.
Springer, Vienna 2000, p. 279. 31. S. F. Buchelle, H. Ellinger, Rev. Prog.
17. W. Sanders, L. J. Gibson, in Porous and Quant. Nondestructive Evaluation, 1989, 8A,
Cellular Materials for Structural Application, 449.
5
Material Properties
R. Pippan
Foaming dramatically extends the range of properties of metals, as demonstrated

in Table 5-1 for some commercially available metal foams. These physical,
mechanical, and thermal properties are usually measured by the same methods
as those used for dense solids. We designate the properties of the cell-wall material
with the subscript ªsº and the properties of the cellular structure without the
subscript ªsº (e. g., rs is the density of the solid cell-wall material and r is the
density of the foam). It is clear that not all properties will change in the same way.
There are properties that do not change, for example:

x crystal structure,
x thermal expansion coefficient, and
x melting temperature (solidus liquidus range).
Other physical properties, such as the heat capacity, are typically linear functions of
the density. More exactly it is the sum of the heat capacity of each phase multiplied
by its weight fraction. Even in low density metallic foams the weight fraction of gas
is small, so that the specific heat of the cellular structure is essentially equal to that
of the parent metal.
Finally, there are many properties that depend on the density (or porosity) not
only linearly, but also on the geometrical structure or the microarchitecture of
the cellular structure: the stiffness, the mechanical strength, the thermal and
electrical conductivity as well as acoustic properties, for example. It is the aim of
this chapter to give a short overview of the most important properties for the
application of cellular metals. A much more general and detailed treatment can
be found in the book on cellular solids by Gibson and Ashby [1]. The increasing
interest in the properties of metal foams is clearly visible from the increasing num-
ber of recent publications in leading materials journals [2 13], proceedings
[14 18], and the data supplied by different research groups.
180 5 Material Properties
Table 5-1. Ranges for properties of commercial metallic foams [6].
Property, symbol [units] Cymat Alulight Alporas ERG Inco
Material Al SiC Al Al Al Ni
Relative density, r/r0 0.02 0.2 0.1 0.35 0.08 0.1 0.05 0.1 0.03 0.04
Structure Closed cell Closed cell Closed cell Open cell Open cell
Young's modulus, 0.02 2.0 1.7 12 0.4 1.0 0.06 0.3 0.4 1.0
E [GPa]
Poisson's ratio, n 0.31 0.34 0.31 .34 0.31 0.34 0.31 0.34 0.31 0.34
Compressive strength, 0.04 7.0 1.9 14.0 1.3 1.7 0.9 3.0 0.6 1.1
s pl [MPa]
Tensile elastic limit, 0.04 7.0 2.0 20 1.6 1.8 0.9 2.7 0.6 1.1
s y [MPa]
Tensile strength, 0.05 8.5 2.2 30 1.6 1.9 1.9 3.5 1.0 2.4
s UTS [MPa]
Endurance limit, 0.02 3.6 0.95 13 0.9 1.0 0.45 1.5 0.3 0.6
s ce [MPa]
Densification strain, eD 0.6 0.9 0.4 0.8 0.7 0.82 0.8 0.9 0.9 0.94
Tensile ductility, eUTS 0.01 0.02 0.002 0.04 0.01 0.06 0.1 0.2 0.03 0.1
Fracture toughness, 0.03 0.5 0.3 1.6 0.1 0.9 0.1 0.2 0.6 1.0
KcIC [MPa.m1/2]
Thermal conductivity, 0.3 10 3.0 35 3.5 4.5 6.0 11 0.2 0.3

l [W/mpK]
Resistivity, R [10 8Vpm] 90 3000 20 200 210 250 180 450 300 500
The parameters that influence the structure-sensitive properties of cellular me-

tals are (ordered by their importance):
x intrinsic properties (properties of cell wall material),
x relative density,
x type of cellular structure (open or closed cells),
x in a closed-cell foam, the fraction of the solid contained in the cell nodes, edges
or the cell faces,
x irregularity or gradients in mass distribution,
x the cell size and size distribution (including exceptional sizes),
x shape of the cells and the anisotropy of cells (including exceptional shapes),
x connectivity of cell edges, and
x defects, by which we mean buckled or broken cell walls.
Some scaling relations have been developed for idealized structures [2], between
the properties of the solid, the density, and the type of cellular structure: a few
of them are listed here. These relations clearly point out that the most important
structural characteristic of a cellular material is the relative density r/rs.
The importance of other structural parameters can be easily seen in the density
dependence of the thermal and electrical conductivity. The thermal conductivity of
metal foam is mainly dominated by the conductivity through the solid. The contri-
bution of the gas, radiation across the pores, and convection within the cell play a
minor role. Therefore, the thermal and the electrical conductivity should behave in
the same manner. The conductivity of the cellular metal should be equal to the
conductivity of the dense solid times its volume fraction (r/rs) multiplied by an ef-
ficiency factor, which is given by the geometry and takes into account the tortuous
path of the heat flow. If we assume that a variation of the relative density is caused
by a proportional change of the thickness of the cell wall and struts, or an increase
of the size of the cells maintaining the cell wall thickness the efficiency factor
should be constant. For Al foams it was shown that a variation of the efficiency fac-
tor can be expressed also in a power law of the relative density, it is about propor-
tional (r/rs)0.5, hence the conductivity is about proportional (r/rs)1.5.
In contrast to thermal and electrical properties, the influence of the density and
the architecture of the cellular metal on the mechanical properties is much stron-
ger and more complex. For load bearing structures, packaging, and crash elements
their importance is evident, however, for functional applications, for example, heat-
exchange systems, damping, and filtering the structural stability is also essential.
In general, one can distinguish between elastomeric, elastoplastic, and brittle
foams (this is not a classification of the solid material, it describes the behavior
of the cellular structure).
In compression, these three types of foam show a similar behavior [2] but for
different reasons. Metallic foams, having a relative densities of r/rsI0.2, exhibit
after an elastic loading a more or less clear plateau regime, leading into a final re-
gime of steeply rising stress. The differences are visible in the mechanism of de-
formation or by unloading of the compressed cell structure: in elastomeric
foams the collapse is by elastic buckling of the cell walls and is more or less reco-
verable; in plastic foams it is the formation of plastic hinges; and in brittle foams it
is the brittle fracture of the cell wall (crush) that causes the plateau region: the last
two are not recoverable.
Cellular metals are usually ductile. Figure 5-1 compares schematically the
stress strain curve of a cellular and a solid metal in tension and compression.
The foam shows a linear elastic regime, which is not so clearly established as in
a solid; followed by a ªhardeningº, which is not only hardening as in a solid, it
is induced by a redistribution of deformed regions; in compression one reaches
then a plateau regime and a final densification; in tension the maximum stress
is reached at relative small strains, typically 1 4 %, which is much smaller than
in a solid metal. The characteristic values are the Young's modulus E (strictly
the stiffness of the cellular structure), the elastic limit s y (which may be slightly
different in tension and compression), the tensile strength, s UTS, the plateau stress
182 5 Material Properties
in compression s pl (where the subscript ªplº means plastic), the densification

strain, the ultimate tensile strain, and the fracture strain ef. The shape of the
stress strain curve in Fig. 5-1 is typical of cellular metals. The characteristic para-
meters for a certain relative density depends on the type of foam, open- or closed-
cell, the anisotropy, and other architectural parameters. The standard scaling rela-
tion [1], however, only takes into account the properties of the solid metal, the re-
lative density, and the type of foam. Such simplified scaling relations exist also for
creep [1], fatigue [19], and fatigue crack propagation properties, but they are based
only on a limited number of experimental evidence. Compression behavior has
been relatively extensively investigated [14 18], whereas studies of the behavior
under tensile and shear loading are rather limited, the latter is especially important
for the design of sandwich panels. A further important difference between a solid
metal and a metal foam becomes visible under multiaxial loading. For example, the
yield behavior in a solid metal is not affected by the hydrostatic component of the
stress tensor that is different in cellular metals [1]. In the following, the different
properties and some specific problems with measuring these properties of cellular
metals are discussed.
Finally it should be emphasized that, for the application of a foam, not only one
property has to be considered. For example, in the automotive industry the stiff-
ness, crush behavior, the energy absorbing, acoustic damping, as well as the
cheap production of a complex shape are important. In such multidimensional re-
quirements metal foams can be the best solution. It is furthermore important to
note that redesign (foam specific) of the component is usually necessary. For exam-
ple, the simple filling of a given shell structure with a foam does not necessarily
improve the property of a component.
Figure 5-1. Schematic comparison of the stress strain behavior of a metal and a metal foam.
References
1. L. J. Gibson, M. F. Ashby, Cellular Solids, 14. J. Banhart (ed.), MetallschaÈume, MIT Verlag,
Cambridge University Press, UK 1997. Bremen, Germany: The proceedings of a
2. W. E. Warren, A. M. Kraynik, J. Appl. Mech. conference held in Bremen in March 1997 (in
1988, 55, 341 346. German).
3. T. G. Nieh, K. Higashi, J. Wadsworth, Mater. 15. J. Banhart, M. F. Ashby, N. A. Fleck (eds.),
Sci. Eng. A 1999, 283, 105 110. Metal Foams and Foam Metal Structures, Proc.
4. J. Grenestedt, K. Tanaka, Scripta Mater. 1999, Int. Conf. Metfoam `99, 14 16 June 1999,
40, 71 77. MIT Verlag, Bremen 1999.
5. J. Grenestedt, F. Bassinet, Int. J. Mech. Sci. 16. A. G. Evans (ed.), Ultralight Metal Structure,
2000, 42, 1327 1338. Division of Applied Sciences, Harvard Uni-
6. A.-F. Bastawros, H. Bart-Smith, A. G. Evans, versity, Cambridge, MA, USA: the annual
J. Mech. Phys. Solids 2000, 48, 301 322. report on the MURI programme sponsored
7. H. Bart-Smith, A.-F. Bastawros, D. R. by the Defense Advanced Research Projects
Mumm, A. G. Evans, D. J. Sypeck, H. N. G. Agency and Office of Naval Research, 1998.
Wadley, Acta Mater. 1998, 46, 3583 3592. 17. D. S. Schwartz, D. S. Shih, A. G. Evans,
8. J. Banhart, J. Baumeister, J. Mater. Sci. 1998, H. N. G. Wadley (eds.), Porous and Cellular
33, 1431 1442. Materials for Structural Application, MRS
9. Y. Yamada et al., Mater. Sci. Eng. A 2000, 277, Proceedings Vol. 521, MRS, Warrendale,
213 217. PA, USA, 1998.
10. K. A. Dannemann, J. Lankford Jr., Mater. 18. H. P. Degischer (ed.), MetallschaÈume,
Sci. Eng. A 2000, 293, 157 164. Proceedings of the DGM Symposium
11. A. Paul, U. Ramamurty, Mater. Sci. Eng. A ªMetallschaÈumeº, 28 29 February 2000,
2000, 281, 1 7. Vienna, Austria, MATWER 31, 6, Wiley-VCH,
12. M. MuÈnch, M. Schlimmer, Mater. Sci. Eng. Weinheim 2000 (in German).
Tech. 2000, 6, 544 546. 19. B. Zettl, H. Mayer, S. E. Stanzl-Tschegg,
13. O. B. Olurin, N. A. Fleck, M. F. Ashby, H. P. Degischer, Mater. Sci. Eng. A 2000, 292,
Mater. Sci. Eng. A 2000, 291, 136 146. 1 7.
5.1
Mechanical Properties and Determination
C. Motz, R. Pippan, and B. Kriszt
5.1.1
Young's Modulus
One of the most important properties of structural materials is the linear elastic
behavior, which is described by a set of moduli. For isotropic materials two moduli
are needed: the Young's modulus, E, and the shear modulus, G, to characterize the
linear elastic response. Some foams are not isotropic owing to the manufacturing
processes and therefore more than two moduli are necessary to describe the linear
elastic behavior.
184 5.1 Mechanical Properties and Determination
In the isotropic case, which will be accounted for in this section owing to sim-
plification and because many closed-cell metal foams are more, or less isotropic,
the following well known relations between Young's modulus E, shear modulus
G, bulk modulus K, and elastic Poisson's ratio n exist
E E
G , K (1)
2(1 S v) 3(1 s 2v)
In the situation of cellular material, or foams it is in general not appropriate to use
the term ªmodulusº, because a foam is a structure (like a framework) and one
should use ªstiffnessº instead. However, it is now usual to consider that new
class of materials as a continuum, hence common material properties are used
to characterize them. But it is important to note that the Young's modulus of a
cellular structure is not a material constant and depends therefore mainly on the
architecture of the cellular metal.
5.1.1.1 Influence of the Foam Structure

There is an essential difference between open-cell and closed-cell structures. Open-
cell foams are represented by a network of connected struts. The main deformation
mechanism is bending of the cell edges and, at higher relative foam densities, rr i
0.1 [1], additional extension and compression of the edges. In closed-cell foams the
cell walls between the cell edges stiffen the structure. Besides bending, extension
and compression of the cell edges and the membrane stresses in the cell walls also
play a major role on the deformation mechanisms. Owing to the higher con-
straints, arising form the existence of cell walls, in closed-cell structures the
Young's modulus is theoretically higher by several magnitudes compared with
open-cell structures of the same relative density.
Imperfections in the structure, like corrugations and curvature of cell walls and
cell edges [2] have also a significant influence on the stiffness. There can be a large
impact on the Young's modulus of a foam if these imperfections reach a larger
extent. Other structural parameters, like cell size, cell shape, and their variations
have a minor influence on the Young's modulus [3]. Since the structural para-
meters are not easily controllable during the manufacturing process, finite element
studies of the dependence of the elastic stiffness in foams with respect to different
parameters have been performed [4,5], which support these experimental observa-
tions.
5.1.1.2 Influence of the Foam Density

The relative foam density rr has the largest influence on all mechanical para-
meters. For lower density foams (rr I 0.2) Gibson and Ashby [2] have applied a
simple bending strut model to a cubic foam structure and found the following
relation between Young's modulus and density
!2
E r
z (open-cell)
Es rs
(2)
!2
E r r
z F2 S(1 s F) (closed-cell)
Es rs rs
The subscript ªsº indicates the properties of the solid from which the cell walls
and cell edges are made. F is the fraction of solid that is contained in the cell edges
for closed-cell foams. Although these equations are based on a simple model of a
cubic foam structure, which considerably deviates from a real foam, there is a fair
agreement with experimental results (Fig. 5.1-1). The equations are valid as long as
the main deformation mechanism is bending of cell edges. For higher-density
foams extension and compression of cell edges becomes more important and
therefore a deviation from these relations should be observed.
Figure 5.1-1. Dependence of the Young's modulus on the relative foam density for different types
of closed-cell (closed symbols) and open-cell (open symbols) aluminum foams [6].
5.1.1.3 Influence of Deformation

Deformation changes the structure of a foam by bending, buckling, stretching, and
cracking of the cell edges and cell walls. Therefore, the Young's modulus will
change with strain. In general, there is only a very small linear elastic part in
the stress strain curve of a ductile foam, which makes the measurement of the
initial Young's modulus difficult. For that reason many authors measure an un-
loading modulus at a certain strain and use this one as Young's modulus, E. It
is important at which strain the modulus is measured because it changes with in-
creasing deformation. Figure 5.1-2 shows the stress strain curves in compression
and tension and the unloading compliance of an Alporas foam (rr 0.1).
It is evident that the Young's modulus with increasing strain |e| decreases much
faster in compression than in tension. This is induced by the fact that buckling cell
edges and cell walls reduce the stiffness much stronger whereas stretching will
increase the stiffness before the initial cracking of cell walls during tension load-
ing. Ultrasonic test methods apply a very small deformation on the foam and per-
mits the measurement of the Young's modulus during deformation experiments.
The obtained Young's moduli are significantly higher than those measured in stan-
dard compression or tension tests [7]. Figure 5.1-3 shows the dependence of the
Young's modulus on the deformation obtained with a ultrasonic test method.
For structural applications usually an ªinitialº Young's modulus is needed. Since
the determination of the initial Young's modulus is difficult, it is suggested to
use the unloading modulus E0.2 at 0.2 % total strain in compression, or tension,
which seems to be a good compromise between measurability and the effect of
plastic deformation.
Figure 5.1-2. Stress strain response with measured unloading moduli indicated for a closed-cell
aluminum foam (Alporas, rr 0.1).
Figure 5.1-3. Dependence of the

Young's modulus (stiffness) on
the deformation at which the
modulus is measured [7]. The
change of density is used as
measure at the deformation in
compression.
5.1.2
Compression Behavior
In compression, cellular metals show a unique stress strain response with a pla-
teau region in which the stress is nearly constant over a wide range of strain
(Fig. 5.1-4). This behavior makes cellular metals interesting for energy absorbing
applications where at a relatively low constant stress a large amount of deformation
can be absorbed. More about energy absorbing and crush behavior can be found in
Section 5.1.3. Depending on the material from which the foam is made, different
deformation mechanisms (elastomeric, brittle, and ductile behavior) can be ob-
served. Metallic foams usually show a considerable ductile behavior.
Many investigations have been performed on compression behavior and energy
absorbing performance of metal foams in recent years. The behavior is similar to
many polymer foams. Thus the basic deformation mechanisms and the stress
strain response are well known. Compression loading at small strains of ductile
foams yields to bending and extension/compression of cell edges and cell walls.
If the stresses in the edges and walls exceed the yield stress s y,s of the solid, the
onset of plastification is reached and the deformation is no longer reversible.
Caused by the inhomogeneous structure of real cellular metals the stress concen-
trations exceed the yield stress in some elements at relatively low strains, which
leads to an early ªplastificationº of the foam. Therefore the linear elastic part of
the stress strain curve of a ductile foam is in general hard developed. Increasing
the load on the foam causes to buckling of cell edges and walls in weaker regions
of the foam. A deformation band perpendicular to the loading direction develops,
in which plastic collapse of the cells takes place. This is accompanied by the begin-
ning of the plateau region in the stress strain curve. With increasing strain, addi-
tional deformation bands are formed until most of the cells have collapsed and the
Stress strain response in compression for a ductile aluminum foam and a brittle
Figure 5.1-4.
alumina foam [11].
densification region is reached [8,9]. Depending on the cellular structure and the
properties of the solid, the plateau region of the stress strain curve may not be
flat, but a slight slope, or waviness of the curve can occur. Two extreme cases
are depicted in Fig. 5.1-4, the behavior of a ductile aluminum foam and a brittle
alumina foam, showing a smooth and a jagged plateau, respectively. For the esti-
mation of the plateau stress s pl depending on the foam density r and the yield
stress s y,s of the solid, Gibson and Ashby used a simple cubic-cell model for the
foam. The plateau stress is reached, when the moment exerted to the cell edges
exceeds their fully plastic moment
!3=2
s pl r
z 0:3 (open-cell)
s y; s rs
(3)
!3=2
s pl r r
z 0:3 F S0:4(1 s F) (closed-cell)
s y; s rs rs
Figure 5.1-5 shows the stress strain curves in compression of Alulight alumi-
num foams with different densities.
The increase of the plateau stress with increasing density is clearly obvious. Dif-
ferent methods exist for the measurement of the plateau stress depending on the
course of the stress strain curve. If there is a stress peak at the beginning of the
plateau, this peak or a clear flat section stress will be taken as plateau stress [10]. In
Figure 5.1-5. Stress strain curves of aluminum foams (Alulight) with different densities in
compression.
the absence of a stress peak the plateau region will be extrapolated to lower strains
and the intersection with the initial elastic line, or the ordinate (Rpe) delivers the
plateau stress [10,27]. The plateau stress plays an important role to characterize
the energy absorbing behavior and is a good material property for the compression
performance of a foam. However, for structural design with foams the plateau
stress may not be applicable, because of the relatively high deformation and
damage of the foam that is accompanied with this stress level. The determination
of a yield stress s y is problematic owing to the usually poorly developed linear
elastic regime of the stress strain curve. To compromise, an offset yield stress,
like s 0.2, or Rp02 should be used, depending on the mechanical response of the
cellular metal.
For the experimental determination of a stress strain curve, standard equip-
ment for compression tests can be used. If the measurement of the Young's
modulus E and the 0.2 % offset yield stress s 0.2 is needed, the displacement should
be measured directly at the specimen with a clip gauge, or a videoextensiometer. To
avoid errors arising from inhomogeneous deformation of the foam the displace-
ment should be measured at more than one position on the specimen. The
specimen size must exceed the average cell size by 7 to 10 times in each principal
direction to reduce the influence of the inhomogeneous cell structure.
5.1.3
Energy-Absorbing and Impact Behavior
Good energy absorbers should support large strains at a relatively low constant
stress level. Owing to their compression behavior, cellular metals (especially
foams) and also some polymer foams, are especially suited for energy absorbing
applications. The advantages of metal foams for these applications are their rela-
tively low density, adjustable plateau stress level, good temperature resistance
and non-flammability. Thus their application fields are the transport and packaging
industries.
5.1.3.1 Energy Absorbing Capability

Cellular metals have a good energy absorbing capacity with a good energy absorb-
ing efficiency because of the extensive plateau regime. In the event of an impact,
the forces are controlled by the stress level at which the kinetic energy is absorbed.
To avoid damage of a package, or injury to a person, the impact forces must not
exceed a certain value. Therefore, not only is the energy absorbing capacity im-
portant, but also the stress strain response of the material plays a major role
(Fig. 5.1-6).
Figure 5.1-6. Comparison of stress strain re- the same for both. The peak stress of the ab-
sponse of two different energy absorbers. The sorber that has no plateau region is higher and
amount of absorbed energy per volume in a this will enhance the impact forces.
certain strain interval is indicated as EV and is
Figure 5.1-7. The energy absorbing efficiency, h, is the ratio between absorbed energy of the
material EV and the absorbed energy of an ideal plastic material EV,ideal in a certain strain range.
The absorbed energy per unit volume EV in a certain strain interval [e1, e2] is
equal to the area below the stress strain curve and can be expressed as
Z e2
EV s(e)de (4)
e1
The efficiency of energy absorption, h, is the ratio between the absorbed energy of
the real material and the absorbed energy of an ideal absorber that is an ideal plas-
tic material
R e2
s(e)de
h e1 (5)
s 0 (e2 s e1 )
Most cellular metals have a nearly flat plateau region and therefore a high energy-
absorbing efficiency depending on the used strain interval, and in general, a good
absorbing capacity (Fig. 5.1-7).
5.1.3.2 Impact Behavior

For the application of cellular metals as crash absorbers in transport systems not
only is the energy absorbing capacity but also the impact behavior important. At
high deformation rates, which may occur in crash situations, the mechanical
properties of the foam may change. There are two points that must be taken
into consideration: the inertial mass of the cell edges and walls and the gas pres-
sure and gas flow inside the cells of a closed-cell foam. Both effects can increase
the flow stress and therefore also the plateau stress of the foam with increasing
strain rate [12]. In open-cell structures no significant increase of flow stress is ob-
served at higher strain rates, which suggests that the inertial masses of the struc-
tural elements have no contribution to the flow stress [12]. In closed-cell foams the
situation is different, because some of the metal foams show a strain-rate depen-
dence. Real closed-cell metal foams are, in general, not completely closed-cell struc-
tures because they frequently contain some initial defects in the cell walls arising
from the manufacturing process. Additional defects are generated during the defor-
mation process when cell walls rupture. The gas flow between the cells through
these defects is strain rate dependent. Thus the gas pressure inside the cells in-
creases during high-speed deformation and contributes to the flow stress [13].
Owing to the strain localization in deformation bands, the local strain rate is
much higher than the measured global one, which will increase the strain rate
effect. As a result, the plateau stress and the energy absorbing capacity of closed
cell metals foams ise [12,13]. Investigations on aluminum foams show an increase
in the plateau stress of 20 90 % when the strain rate increases from 10 3 to 103 s 1
(Fig. 5.1-8). The strain rate effect is mainly controlled by the cell morphology and
the alloy from which the foam is made.
Figure 5.1-8. Strain-rate dependence of the plastic strength of two types of aluminum foams [12].
5.1.4
Tension Behavior
Improvements in the manufacturing processes of metal foams have increased their

quality and opened new applications as a real structural material. Therefore me-
chanical properties, like tension and fracture behavior need to be known in addi-
tion to compression and energy absorbing performance. The failure mechanisms
and the stress strain response deviate from those observed in compression.
5.1.4.1 General Tensile Behavior

The initial deformation mode is very similar to compression and consists of bend-
ing of cell edges in open-cell foams and additional extension and compression of
cell walls in closed-cell foams. In the post-yield stage, however, the deformation
mechanism differs significantly from compression. In compression, plastic, or
brittle collapse of cells occurs within deformation bands perpendicular to the load-
ing axis. Owing to the dominance of tensile stresses in the cell edges and cell walls
in tension, buckling of these elements is improbable. Finite element simulations
show that such plastic instabilities may only occur at very large strains [14],
whereas the strain to fracture of real ductile metal foams is of the order of a few
percent.
The deformation is concentrated in weak regions of the cellular metal. No defor-
mation bands are formed except during the final failure. The final failure me-
chanism depends on the ductility of the material from which the foam is made.
Brittle ceramic foams usually show microcracking of highly stressed cell walls
and edges before final failure. If these defects exceed a critical size, a main
crack propagates through the foam and catastrophic failure occurs. The critical
fracture strength s cr of the foam can be obtained by applying linear elastic fracture
mechanics [2]
!3=2
s cr p r
z C l=a (open-cell) (6)
s fs; s rs
where C is a constant, l is the cell size, a is the critical initial defect size, and s fs,s is
the fracture strength of the solid. For closed-cell metals the situation is more com-
plicated.
In ductile cellular metals (such as aluminum foams) cracking of cell walls and
cell edges can also be observed before a main crack starts to propagate. With in-
creasing deformation, a fracture-process zone starts to develop within which the
whole deformation is concentrated. In closed-cell metals, several cell walls rupture
inside this fracture-process zone. Further deformation leads to development and
propagation of a main crack along the weakest path in the cellular structure.
Only the strongest cell edges remain intact and hold the two foam parts together.
This process is accompanied by a decrease in flow stress, which is evident in the
stress strain curve depicted in Fig. 5.1-10. Figure 5.1-9 shows the corresponding
deformation process and the failure mechanism of a ductile closed-cell aluminum

foam.
The estimation of the peak stress, s UTS, or the fracture strain for ductile cellular
metals is complicated, because a relation for the ductile fracture of the cell edges
and cell walls must be established and incorporated into the geometrical models.
The experiments show that s UTS is usually compatible with the plateau stress in
compression.
Figure 5.1-9.Failure mechanism in aluminum foam in tension. Crack initiation in some weak cell
walls and subsequent development of a fracture-process zone (stress strain curve in Fig. 5.1-10).
Figure 5.1-10. Stress strain response in tension of aluminum foam with some unloading moduli
indicated (Alporas, 3r 0.18).
5.1.4.2 The Influence of Notches

Common metals reveal a different behavior in tension if notches are present. Brit-
tle materials show a lower average net section stress at fracture with notched speci-
mens compared with unnotched ones. Owing to the stress concentration at the
notch root, the fracture strength of a brittle material is reached at a lower average
stress and the specimen fractures. Ductile materials behave differently and show a
higher average net section stress at fracture if notches are present. At the notch
root the deformation is constrained, which results in a higher stress triaxiality
and therefore in a higher flow stress. In the case of metal foams, the situation is
more difficult because the plastic deformation do not take place at constant vol-
ume, which results in a plastic Poisson's ratio npl smaller than 0.5. Therefore,
the transverse strains, and consequently also the constraints at the notch root,
are smaller than in common metals. Under compressive loads the plastic Poisson's
ratio is near zero in the plateau region and the foam should show a notch-insen-
sitive behavior. Whereas in tension (and also in compression at strains smaller
than the onset of the plateau region) the plastic Poisson's ratio is similar to the
elastic one. The elastic Poisson's ratio n depends on the cellular structure and
for most commercially available closed-cell foams (Alulight and Alporas) n is in
the range 0.30 0.40 [15]. Thus, these foams should show a notch sensitivity in ten-
sion. This is supported by finite element simulations [16], which indicate that a
notch sensitivity should be observed, if the plastic Poisson's ratio is greater than
0.3. In Fig. 5.1-11 the stress strain curves of notched foam specimens with differ-
ent notch depth ratios are presented. An increase in the net section stresses and a
decrease in the strains to fracture can be observed with increasing notch depth.
Figure 5.1-11. Influence of the notch depth ratio, a/W, on the stress strain response of a ductile
aluminum foam in tension.
5.1.4.3 Foam-Specific Test Problems in Tension

For the testing procedure standard methods for tension tests can be used. Owing
to the relatively low fracture strains of most cellular metals in tension, the dis-
placement should be measured with a clip-gauge attached directly to the speci-
men and not with the cross-head displacement of the testing machine. The speci-
men size should exceed 7 to 10 times the average cell size in all principle direc-
tions to obtain reliable results. The specimen should be mounted carefully into
the testing machine to avoid damage to the foam. A good method is to glue
the foam specimen to two metal plates that are then mounted into the test
machine. Important mechanical properties that can be obtained from tension
tests are Young's modulus, the peak stress, s UTS, the yield stress s 0.2, the strain
at maximum stress eUTS and the elastic and plastic Poisson's ratios n and npl.
The strain to total fracture does not represent a material property but one of
the individual specimen. Therefore eUTS is recommended, where considerable
degradation starts progressing usually rapidly.
5.1.5
Torsion Behavior
In torsion the cell edges are subjected to bending and the cell walls to a combi-
nation of bending and shearing. The shear behavior is important for the develop-
ment of material laws and also for the application in sandwich structures, where
core shear may occur in particular load situations. Alternative methods for testing
the shear strength are, for example, the double-lap shear test or the ASTM C-273
Figure 5.1-12. Stress strain response of aluminum foam in torsion with no axial stress (black
line) and with tensile axial stress (gray line) [17].
test method. Most foams exhibit significantly larger deformation up to failure

in free torsion compared with tension tests. A typical example is depicted in
Fig. 5.1-12. If the specimen ends are axially fixed during the torsion test, addi-
tional tensile stresses are built up in the sample and the strain to fracture is
reduced to the level typical observed in tension tests. Important properties are
the shear modulus G, the shear strength g t and the shear strain to fracture g t.
By applying tensile or compressive loads on the specimen during torsion tests
a variety of stress states in the sample can be achieved and the failure mechan-
isms can be studied.
5.1.6
Fracture Behavior
Fracture behavior and fracture toughness values may play a major role for the ap-
plication of cellular metals foams in load-bearing structural components. The frac-
ture of brittle foams, where linear elastic fracture mechanic (LEFM) can be applied,
were discussed briefly in the section on tensile behavior.
The situation in closed-cell ductile metal foams is more complicated, where
LEFM usually can not be applied, which is considered in this section.
5.1.6.1 Crack Initiation and Crack Propagation

Owing to the inhomogeneous structure of most metal foams, it is expected that
localization effects may appear during fracture tests. Figure 5.1-13 shows a se-
quence of micrographs of a region in front of the notch root at different load
steps for a ductile aluminum foam. Crack initiation and crack propagation are
clearly visible. Loading a notched or pre-cracked specimen results in plastic yield-
ing in the vicinity of the crack tip [18].
The deformation is localized to the thinnest regions of the cell walls surround-
ing the crack tip. With further loading, the strain to fracture is reached in some
cell walls and microcracks appear in the vicinity of the crack tip. Cell walls that
are few cell sizes away from the crack tip can be plastically deformed too, whereas
other regions remain nearly undeformed. A so-called fracture-process zone (FPZ)
with an extent of several cell sizes develops, which contains localized deformation
and microcracking. The definition of a plastic zone in cellular metals is compli-
cated, because at higher loads localized yielding can occur over the whole liga-
ment length. With increasing load a main crack is initiated at the notch root,
or at the pre-crack by a coalescence of microcracks, and starts to propagate
through the cell structure. The crack follows the weakest path through the struc-
ture and builds thereby secondary cracks and crack bridges. Crack bridging length
and microcracking should depend on the cell structure and the solid of which the
foam is made.
Figure 5.1-13. Diagram of the load displacement curve and the corresponding SEM images to
demonstrate crack initiation and crack propagation in a ductile aluminum foam acquired from in-
situ fracture tests.
5.1.6.2 Fracture Toughness

For ductile cellular metal foams, linear elastic fracture mechanics is usually not ap-
plicable, because they show an extensive plastic regime and the specimen sizes are
limited by the manufacturing processes. Therefore elastoplastic fracture mechanic
tests, like J-integral tests, or COD (crack opening displacement) tests, must be used
to determine the fracture toughness of metal foams. Common standards, like
ASTM E813 for J-integral tests, are optimized for ductile solids and can only be
used in a limited manner for foams. Size restrictions for the specimen geometry
to obtain valid fracture toughness values should be thought about because, for
example, the required specimen size also depends on the cell size, l, of the foam.
For the determination of J Da curves according to ASTM E813 or ASTM E1152,
the following parameters must be measured: the force F, the load-line displace-
ment vLLD, and the crack extension Da. Figure 5.1-14 shows a load versus load-
line displacement curve obtained in a fracture mechanics test and the crack exten-
sion as a function of the load-line displacement of a ductile aluminum foam. The
calculated J Da curve is plotted in Fig. 5.1-15. The ASTM standard provides that
the initial J-value, J0.2, is obtained by the intersection of the blunting line at
0.2 mm crack extension with the straight line achieved from a linear regression
of the J Da curve in the Da interval 0.15 1.5 mm.
A crack extension of 0.2 mm is rather low for metal foams, because it is one
magnitude smaller than the usual cell size (the characteristic microstructural
length) and can be caused by other effects (microcracking, localized yielding)
rather than crack extension, depending on the measurement method used. Fleck
and Ashby [19] introduced a steady-state J-value, JSS, which is measured as the pla-
teau J in the J Da curve. They assumed that this JSS is a measure of the fracture
toughness for cracks, where the bridging zone is fully developed. However, not all
cellular metals show a typical plateau region in the J Da curve and the plateau is
reached at rather high crack extensions. The standardization of the evaluation of
initial, or steady state fracture toughness values from J Da curves is a problem
that has not yet been solved satisfactorily.
Gibson and Ashby [6] established the following relation between the fracture
toughness and the foam density for ductile metal foams
Figure 5.1-14. Load and crack extension versus load-line displacement curve for a ductile metal
foam (Alporas, rr 0.1).
Figure 5.1-15. J-integral versus crack extension plot with fracture toughness values JSS and J0.2
indicated (J-values obtained from Fig. 5.1-14).
!3=2
r
JIC z Cs y; s l (7)
rs
This equation correlates the fracture toughness with the cell size (compare with
Eq. 8, it is assumed that COD l), which is demonstrated only for a limited num-
ber of foams and densities. Other fracture toughness values, like KIC, or CODC, can
be derived from this equation using the well-known fracture mechanics relations.
An alternative method for the determination of the fracture toughness of a duc-
tile material is the measurement of the critical crack opening displacement (COD).
Although this is a direct method, it is rarely used owing to the difficulties in
experimental determination. Schwalbe [20] suggested to measure the COD 5 mm
away from the crack tip, COD5, which can be more easily performed. We have
adapted this technique to foams. Figure 5.1-16 shows COD5 Da and CTOD Da
curves of a ductile aluminum foam (CTOD crack tip opening displacement).
The advantage of this method is the similarity of the COD Da curves for different
foams and the easier evaluation of critical fracture toughness values. The relation
between COD and the J-integral is given by the following equation
J k s COD (8)
where s * is the mean value between yield stress s y and ultimate tensile stress s UTS
and k is 1 for plain stress state and 1,3...2,6 for plain strain state for common
metals. In the case of metal foams the value of k for the (macro) plain strain
state depends on the Poisson's ratio of the foam in the plastic regime. It is im-
Figure 5.1-16. Crack opening displacement measured behind the crack tip and reaches a
(COD) in terms of COD5 and crack tip opening constant value after a certain crack extension,
displacement CTOD versus crack extension whereas COD5 is measured at a fixed position
curves for the same foam as shown in 5 mm behind the initial crack tip.
Fig. 5.1-14 and Fig. 5.1-15. The CTOD value is
portant to note that we have to distinguish the macro- and micro- stress states (the
stress state in the cell wall). Since npl is much smaller than in solid metals, k should
be only somewhat larger than 1.
5.1.6.3 Foam-Specific Test Problems

Fracture mechanics tests are usually performed with compact tension specimens.
The criteria for the load and load-line displacement measurements are given in the
corresponding ASTM standards. The determination of the crack extension Da in
cellular metals is more complicated owing to their inhomogeneous structure. Ex-
perimental tests show that potential drop techniques and compliance methods
give reasonable results for Da. The specimen size is not only limited by fracture
mechanics constraints, but also by the cell size of the structure. The specimen
ligament and initial crack length should not be smaller than ten times the average
cell size and the thickness should be larger than five times the average cell size to
obtain specimen size independent results. It should be noted that at this time there
are no standards for determining critical fracture toughness values for metal
foams.
References
1. W. E. Warren, A. M. Kraynik, J. Appl. Mech. 12. K. A. Dannemann, J. Lankford Jr., Mater.

1988, 55, 341 346. Sci. Eng. A 2000, 293, 157 164.
2. L. J. Gibson, M. F. Ashby, Cellular Solids, 13. A. Paul, U. Ramamurty, Mater. Sci. Eng. A
Cambridge University Press, Cambridge 2000, 281, 1 7.
1997. 14. A. Ableidinger, ªMicro-macromechanical
3. T. G. Nieh, K. Higashi, J. Wadsworth, Mater. Investigations of the Strength and Structure
Sci. Eng. A 1999, 283, 105 110. Behaviour of Metallic Foamsº, Thesis, Vienna
4. J. Grenestedt, K. Tanaka, Scripta Mater. 1999, University of Technology, Austria 2000.
40, 71 77. 15. C. Motz, R. Pippan, ªDeformation Behavior
5. J. Grenestedt, F. Bassinet, Int. J. Mech. Sci. of Aluminum Foams in Tensionº, Acta Mater
2000, 42, 1327 1338. 2001, 49, 2463±2470.
6. M. F. Ashby, A. Evans, N. A. Fleck, L. J. 16. P. R. Onck, ªApplication of a continuum
Gibson, J. W. Hutchinson, H. N. G. Wadley, constitutive model to metallic foam DEN-
Metal Foams: A Design Guide, Butterworth- specimens in compression,º International
Heinemann, Oxford 2000. Journal of Mechanical Sciences 2001, 43,
7. A. Wanner, B. Kriszt, Mater. Sci. Eng. Tech. 2947±2959.
2000, 6, 481 483. 17. M. MuÈnch, M. Schlimmer, Mater. Sci. Eng.
8. A.-F. Bastawros, H. Bart-Smith, A. G. Evans, Tech. 2000, 6, 544 546.
J. Mech. Phys. Solids 2000, 48, 301 322. 18. C. Motz, R. Pippan, in Proc. ECF-13 Conf.,
9. H. Bart-Smith, A.-F. Bastawros, D. R. 6 9 Sept 2000, San SebastiaÂn, Spain,
Mumm, A. G. Evans, D. J. Sypeck, H. N. G. CDROM, 2000.
Wadley, Acta Mater. 1998, 46, 3583 3592. 19. O. B. Olurin, N. A. Fleck, M. F. Ashby,
10. J. Banhart, J. Baumeister, J. Mater. Sci. Mater. Sci. Eng. A 2000, 291, 136 146.
1998, 33, 1431 1442. 20. K. H. Schwalbe, A. Cornec, Fatigue Eng.
11. Y. Yamada et al., Mater. Sci. Eng. A 2000, Mater. 1991, 14, 405 412.
277, 213 217.
5.2
Fatigue Properties and Endurance Limit of Aluminum Foams
B. Zettl, H. Mayer, and S. Stanzl-Tschegg
Fatigue properties of aluminum foams are of great interest, if foamed structures

are subjected to repeated mechanical straining or vibrations. Endurance experi-
ments have been performed under cyclic compression [1 6], cyclic tension
[1,2,5,6], and under fully reversed loading condition [7,8]. These experiments
show that the foam structure rapidly looses strength after a certain number of
load cycles depending on the stress amplitude and the mean stress. Cyclic com-
pression loading leads to the formation of deformation bands, whereas fatigue fail-
ure of foams under cyclic tension stresses is caused by the initiation and growth of
fatigue cracks [2 5]. Under fully reversed loading conditions [7,8] fatigue cracks
cause rupture of specimens, similar to the damage mechanism under cyclic ten-
sion stress [8].
Fatigue investigations of metallic foams described in the literature concentrate
on the regime of numbers of cycles to failure below approx. 106 to 107. In practical
application, however, foamed components may be exposed to greater numbers of
load cycles. Automotive components, for example, may be subjected to more
than 108 cycles during lifetime, and the high cycle fatigue properties and the deter-
mination of an eventual fatigue limit is necessary therefore for a comprehensive
description of the cyclic properties of foams. Although most authors assume an en-
durance limit at 107 cycles, an experimental verification is only found in one report
[7] where several specimens have been tested to 109 cycles or more. In this work,
fatigue data available in the literature will be summarized and the high cycle
fatigue properties of three aluminum foams will be evaluated.
5.2.1
Literature Survey of Endurance Data
Results of endurance tests of several aluminum foams are available in literature.

Fatigue data determined by different authors have been analyzed to evaluate the
influence of material properties and loading conditions on the fatigue behavior
of aluminum foams. Published original data have been obtained from articles
and conference proceedings.
Endurance data represent numbers of cycles to failure for a certain foamed ma-
terial (foam type and material density) at a certain cyclic stress and load ratio. How-
ever, the authors have used different ways to define the magnitude of cyclic loading
and the load-ratio R: some authors have used |s max| / |s min| as a definition of the
load-ratio [1,2,4 6], whereas others have used |s min| / |s max| [3]. In the present re-
port, as well as elsewhere [7,8] the load ratio R is defined in accordance to the
major part of the fatigue literature: R s min/s max [9], where positive values repre-
sent tension stresses and negative values represent compression stresses. Conse-
204 5.2 Fatigue Properties and Endurance Limit of Aluminum Foams
quently, the load ratio R is 1 for fully reversed loading, 0 J R I 1 for cyclic ten-
sion loading and 1 J R J T for cyclic compression loading. The amplitude of a
stress cycle Ds/2 (s max s min)/2 is used to describe the cyclic stress range.
A summary of S N curves described in the literature is shown in Fig. 5.2-1.
The S N data are approximated by power-law curve-fits with a 50 % survival
probability.
Ds (1)
Nn C
2
The values of the exponent, n, and the coefficient, C, (in MPa) have been evaluated
and are shown in Table 5.2-1. Different symbols and lines refer to different mate-
rials and loading conditions (cyclic compression, cyclic tension, or fully reversed
loading). Fatigue properties of the different aluminum foams show significant var-
iations. The cyclic stress amplitudes that lead to mean lifetimes of 105 cycles, for
Figure 5.2-1. S N curves of aluminum foams. Hydro. Loading condition: continuous

Literature references and material properties line cyclic compression, broken line cyclic
are shown in Table 5.2-1. Materials: x Alcan, tension, gray line fully reversed cycling.
y Duocel, k Alporas, b IFAM, v Alulight,
Table 5.2-1. List of references for S N curves shown in Fig. 5.2-1. Coefficients C and n refer to a power-law approximation of data (Eq. 1) [1 6,8,11].
Foam Ref. Density Plateau Loading Load Tested Coefficients

[g/cm3] stress [MPa] condition ratio specimens C [MPa] n
#1 Alcan [1] 0.15 0.25 compr. 10 7 0.133 0.064
#2 Duocel [2] 0.19 0.53 compr. 10 7 0.546 0.079
#3 Alporas [4] 0.22 1.4 compr. 10 5 0.626 0.037
#4 Alporas [6] 0.22 1.37 compr. T 30 0.710 0.028
#5 Alporas [6] 0.25 1.71 compr. T 28 1.028 0.038
#6 Alporas [2] 0.30 1.9 compr. 10 8 1.332 0.063
#7 Alporas [6] 0.30 2.34 compr. T 30 1.759 0.053
#8 Alulight [5] 0.45 6.3 compr. 10 7 3.393 0.038
#9 IFAM [3] 0.63 8.8 compr. 10 13 6.232 0.074
#10 Alulight [5] 0.50 19.5 compr. 10 7 12.514 0.031
#1 1 Alcan [1] 0.15 0.25 tension 0.1 8 0.131 0.108
#12 Duocel [2] 0.19 0.53 tension 0.1 4 0.554 0.081
#13 Alporas [6] 0.25 1.47 tension 0 30 1.068 0.072
#14 Alporas [6] 0.25 1.75 tension 0 33 1.467 0.080
#15 Alporas [2] 0.30 1.9 tension 0.1 5 1.355 0.068
#16 Alporas [6] 0.30 2.25 tension 0 26 2.254 0.093
#17 Alulight [5] 0.86 17.0 tension 0.1 14 13.645 0.090

5 Material Properties
#18 Hydro Al [8] 0.29 3.5 rever. 1 19 5.296 0.111

205
#19 IFAM [8] 0.40 4.4 rever. 1 19 15.691 0.189
#20 Alulight [11] 0.56 7.0 rever. 1 51 6.988 0.110

example, may vary by a factor of 200, depending on the investigated foam and the
load ratio.
The main reasons for the large variations of the fatigue properties of aluminum
foams are their different mass densities and structures. Literature data show a sig-
nificant trend, that fatigue properties of foams increase with increasing plateau
stress. As an example, the stress amplitude for 50 % survival probability of 105 cy-
cles and of 107 cycles, respectively, are shown in Fig. 5.2-2. Irrespective of foam
type and loading condition (cyclic tension, cyclic compression, or fully reversed),
the fatigue strength increases with the plateau stress. The linear relation shown
in Fig. 5.2-2 indicates, that the stress amplitude for mean lifetimes of 105 cycles
is approximately 0.35 of the plateau stress, and Ds/2s pl 0.27 leads to mean life-
times of approximately 107 cycles. This means that the fatigue lifetime of an alu-
minum foam is mainly determined by the cyclic stress amplitude, Ds/2 with re-
spect to the plateau stress, s pl, and Ds/2s pl is an appropriate measure of the mag-
nitude of cyclic loading.
To compare different aluminum foams it is reasonable to normalize the cyclic
properties with respect to the plateau stress s pl. Eq. (1) may be rewritten as follows
Ds C
Nn (2)
2s pl s pl
In Fig. 5.2-3, literature data obtained under cyclic compression, under cyclic ten-
sion, and under fully reversed loading conditions are summarized. Additionally,
mean S N curves for these three loading conditions are shown. The algebraic ex-
pression of the mean curves of the literature data is
Figure 5.2-2. Stress amplitudes at

105 cycles to failure (full symbols,
solid line) and 107 cycles to failure
(open symbols, dashed line) for:
x compressive, k tensional,
v reversed loading. Data obtained
from Fig. 5.2-1.
Ds C
N nmean C0 (3)
2s pl s pl
The values of the exponent, nmean, and the coefficient, C0, of the mean curves are
shown in Table 5.2-2. The largest absolute value of the exponent nmean is found for
fully reversed loading and the smallest for cyclic compression loading.
Using the relative stress amplitude, Ds/2s pl to describe the magnitude of load-
ing, Duocel foam shows good fatigue properties under cyclic compression
Table 5.2-2. Coefficients for mean curves (Eq. 3) under different loading conditions.
Coefficient C0 Exponent nmean
Cyclic compression loading 0.628 0.0491
Cyclic tension loading 0.800 0.0825
Fully reversed loading 1.738 0.1351
a) b)
Figure 5.2-3. Normalized load amplitudes

versus number of cycles to failures for: a)
cyclic compression, b) cyclic tension, c) fully
c) reversed loading.
(Fig. 5.2-3a: #2) and cyclic tension (Fig. 5.2-3b: #12). Duocel foam was the only
open-cell foam where fatigue data were found in the literature. Similar relative fa-
tigue strength is visible for Alporas foam (Fig. 5.2-3a: #3 #7, Fig. 5.2-3b:
#13 #16), Alulight foam (Fig. 5.2-3a: #8,#10, Fig. 5.2-3b: #17, Fig. 5.2-3c: #20)
and IFAM foam (Fig. 5.2-3a: #9, Fig. 5.2-3c: #19). Hydro aluminum foam has
been tested under fully reversed loading conditions only (Fig. 5.2-3c: #18) and
shows the highest relative strength at R 1. Alcan foam, which is the material
with the lowest density, shows the lowest relative fatigue strength under cyclic
compression and cyclic tension (Fig. 5.2-3a: #1, Fig. 5.2-3b: #11).
5.2.2
High Cycle Fatigue Properties and Endurance Limit
To evaluate the high cycle fatigue properties and to determine an eventual fatigue
limit for aluminum foams, specimens have to be tested up to very high numbers of
cycles. These investigations are time consuming using conventional fatigue testing
equipment working at low frequencies, since several specimens have to be tested to
obtain statistically relevant data of this rather inhomogeneous material. Therefore,
the ultrasonic fatigue testing method has been used in this study to investigate the
high cycle regime in an efficient and time saving manner.
5.2.2.1 Material and Procedure

The aluminum foams used in the present investigation are produced from powder
metallurgical prepared precursor material with entrapped titanium hydride as
foaming agent. The chemical composition of the investigated foams is
Al 0.6 wt.-% Mg 0.3 wt.-% Si (named AlMg0.6Si0.3), Al 1 wt.-% Mg 0.6
wt.-% Si (AlMg1Si0.6), and Al 12 wt.-% Si (AlSi12). The commercial name of
the materials is Alulight. The material was produced as rods of 160 mm length
and of 17 mm diameter; material density is 0.56 g/cm3. The specimens were
60 mm long and all materials have been tested as fabricated. Additionally, a series
of AlMg1Si0.6 foam specimens were artificially aged before testing (T5, age hard-
ening at 160 hC for 14 h).
Using the ultrasonic fatigue testing method, specimens are subjected to a fully
reversed tension compression resonance loading at a frequency of approximately
20 kHz. The cyclic loading was controlled with an electromagnetic displacement
gauge, which measured the vibration amplitude at one end of the specimen.
The maximum of the cyclic strain amplitude at the half-length of the specimen
was measured with strain gauges on the surface of the rod. To obtain an average
value of the cyclic strain amplitude, strain gauges with a gauge length larger
than the median cell size were used (measuring area 6 mm q 6 mm). Along
the circumference of a specimen the strain amplitudes measured with strain
gauges coincided within 5 10 %. Experiments were performed in pulsed form
(pulse length: 1000 cycles) and were interrupted by periodic pauses of adequate
length to cool the specimen. A fan was used additionally. The specimen tempera-
tures were 20 25 hC during the experiments. Details of the ultrasonic equipment

are described elsewhere [9].
5.2.2.2 Results
Results of endurance tests with AlMg1Si0.6 foam are shown in Fig. 5.2-4a. Num-
bers of cycles to failure are related to the experimentally determined cyclic strain
amplitude, De/2. Assuming a mean Young's modulus of 3.9 GPa, the stress am-
plitude, Ds/2 can be calculated from the measured cyclic strain as shown in
Fig. 5.2-4a (right ordinate). Specimens that did not fail within 109 cycles (runouts)
are marked with an arrow. Fatigue data are subjected to pronounced scatter, which
is typical for inhomogeneous materials. The solid line indicates a mean fracture
probability of 50 % assuming power-law dependence between stress amplitudes
and cycles to failure.
The S N measurements of AlMg1Si0.6 foam indicate a well-defined fatigue
limit: Specimens either fail before about 107 cycles are reached or they survive
109 cycles or more. The median endurance limit (failure probability of 50 %) is in-
a) b)
c) d)
Figure 5.2-4. Results of endurance fatigue tests under fully reversed tension compression cy-
cling: a) AlMg1Si0.6 foam, b) AlMg1Si0.6/T5 foam, c) AlMg0.6Si0.3 foam, d) AlSi12 foam.
dicated with a line parallel to the abscissa in Fig. 5.2-4a at a stress amplitude of
1.3 MPa, which is approximately 23 % of the plateau stress. However, specimens
may fail at significantly lower cyclic stresses due to the statistical distribution of
defects and areas of stress concentrations. The highest stress amplitude, where
no specimen failed within 109 cycles, is 1 MPa.
Endurance tests on AlMg1Si0.6 T5 foam are summarized in Fig. 5.2-4b. The
heat treatment leads to longer mean lifetimes at all stress levels in comparison
with the not heat treated material. Additionally, the endurance limit increased.
The maximum stress amplitude, where no specimen failed, is 1.2 MPa, and the
median fatigue strength at 109 cycles is 1.4 MPa (approximately 20 % of the plateau
stress).
The results of endurance tests with AlMg0.6Si0.3 and AlSi12 foams are shown in
Fig. 5.2-4c and d, respectively. At a stress amplitude of 1 MPa, one (of five tested)
AlMg0.6Si0.3 foam specimens failed. The fatigue limit for 50 % survival is 1.1 MPa
(approximately 22 % of the plateau stress). The median fatigue strength at 109 cy-
cles of AlSi12 foam is 1.3 MPa (approximately 19 % of the plateau stress). Fatigue
data of AlSi12 foam are subjected to the most pronounced scatter among the inves-
tigated Alulight foams. At stress amplitudes of 1.2 MPa three specimens did not
fail within 109 cycles, whereas one specimen failed at 1 MPa.
5.2.3
Mechanism of Crack Initiation
The initiation of fatigue cracks leads to an increase of specimen compliance and

thus to a decrease of resonance frequency of the ultrasonic vibration. This allows
to determine eventual crack initiation during fatigue cycling. After starting a fa-
tigue experiment, the resonance frequency decreases until about 60 80 % of the
lifetime is reached. If fatigue cycling is stopped then and specimens are cut into
pieces, studies with a scanning electron microscope (SEM) show fatigue damage
at several locations in the interior of the foam. Cracks initiate and grow from al-
ready existing material defects, like holes, precracks, and tied up areas with re-
duced cell wall thickness that were introduced during foaming (Fig. 5.2-5a) or cool-
ing of the material (Fig. 5.2-5b).
No significant further variation of resonance frequency is found after approxi-
mately 107 cycles, if a specimen is cycled at the endurance limit. This indicates
that the process of fatigue damage is stopped, and cracks cannot grow until final
failure. Cracks remain arrested in the interior of cell walls (Fig. 5.2-6a and b)
and are trapped when the crack tip reaches cell nodes. The largest cracks in cell
walls of specimens, which did not fail until 109 cycles, were about 3 mm long.
If specimens are cycled above the fatigue limit, a main crack is formed in the
inner section of the specimen during the remaining 20 40 % of the lifetime
(Fig. 5.2-6c). Microscopical surface investigations of several specimens showed
no evidence of fatigue cracks at the skin surface until about 80 % of the lifetimes
have been reached. This means, that fatigue cracks preferentially initiate in the in-
terior of a foamed rod.
Figure 5.2-5. Cell-wall defects in AlMg1Si0.6 foam introduced: a) during foaming, b) during
cooling.
SEM studies at higher magnification show, that the fatigue fracture surface is re-
latively smooth and transcrystalline. The general fracture appearance is ductile,
and striations are visible in some areas. Figure 5.2-7 shows both, a fatigue fracture
surface in a cell wall (Fig. 5.2-7a and b) and a fracture surface that has been formed
by static rupture of the specimen after fatigue testing (Fig. 5.2-7c). Fatigue cracks
show clear differences compared with cracks formed by static loading or during
cooling of the melt due to the far lower plastic deformation of the material around
the crack path. Fatigue crack paths preferentially follow areas where the cell wall is
thin. In Fig. 5.2-7, the thicknesses of cracked cell walls are approximately
30 60 mm, whereas the most frequent wall thickness obtained by image-analysis
is 250 500 mm [10]. SEM studies of the fracture surface do not show deformation
bands or plastic buckling as it could be found for uniaxial compression or for cyclic
compression fatigue loading.
The role of cell nodes, where fatigue cracks may be trapped, helps to understand
the influence of different foam structures on the fatigue properties. Duocel is an
Figure 5.2-6. Typical fatigue cracks: a) fatigue crack inside a cell wall, length about 300 mm,
b) crack starting from a cell-wall defect, length about 150 mm, c) fatigue crack causing rupture
of a specimen.
Figure 5.2-7. Fracture surfaces of AlMg1Si0.6 T5 (Ds/2 1.96 MPa, number of cycles to failure
8.3 q 104): a) fatigue fracture surface, b) striations indicating crack growth direction, c) fracture
surface formed by static rupture.
open-cell foam with r 0.15 g/cm3 and consists of cell nodes only and approxi-
mately the same amount of material is distributed to nodes and cell walls inside
Alcan foam (r 0.19 g/cm3). Fatigue cracks initiated in Alcan foam may find
lower growth resistance since cell nodes contain less material compared with Duo-
cel foam. This may be one reason, why Duocel foam (Fig. 5.2-3b, #11) shows far
better cyclic tension properties than Alcan foam (Fig. 5.2-3b, #12).
5.2.4
Summary
A survey of literature data (20 S N curves determined with six different aluminum
foams under cyclic compression, cyclic tension, and fully reversed loading) shows
that the fatigue strength is related to the plateau stress, s pl.
x The stress amplitude, Ds/2 for mean lifetimes of 105 cycles (107 cycles) is ap-
proximately 35 % (27 %) of the plateau stress.
x Numbers of cycles to failure and relative stress amplitudes, Ds/2s pl may be cor-
related by a power law function. The largest gradient is found for fully reversed
loading and the smallest for cyclic compression loading.
To investigate the high cycle fatigue properties of aluminum foams, three Alulight
foams have been investigated at numbers of cycles between 104 and 109. Cylindrical
rods with a continuous surface layer and a closed cell structure were tested. The
foam density is 0.56 g/cm3.
x The investigated Alulight foams show a fatigue limit, and cycles to failure above
107 are rare. The median endurance limits, Ds/2 are 1.1 1.4 MPa, which is
19 23 % of the plateau stress.
x Fatigue damage under fully reversed tension-compression loading is governed by
the formation of cracks. Preferential areas for crack initiation are initial defects
like precracks or holes in the interior sections of cell walls. No strain localization
or formation of deformation bands (as reported for cyclic compression tests) is
found.
x Fatigue cracks preferentially grow in the inner sections of cell walls where wall
thickness is small. Below the endurance limit eventual cracks are trapped at
the nodes of cells.
Acknowledgements
The authors acknowledge financial support by the ªFonds zur FoÈrderung der wis-
senschaftlichen Forschung (FWF)º, Vienna, (project no. P13230PHY) and the LKR-
Center of Competence on Light Metals, Ranshofen, Austria, for the grant through
the K-plus project and for supplying the Alulight foams.
References
1. O. B. Olurin, N. A. Fleck, M. F. Ashby, in H. P. Degischer (ed), Vienna, Austria, Wiley-

Proc. Metal Foams and Porous Metal Structures, VCH, Weinheim 2000, p. 488 492.
J. Banhart, M. F. Ashby, N. A. Fleck (eds), 7. B. Zettl, H. Mayer, S. E. Stanzl-Tschegg,
MIT Verlag, Bremen 1999, p. 365 371. H. P. Degischer, Mater. Sci. Eng. 2000, A292,
2. A. M. Harte, N. A. Fleck, M. F. Ashby, Acta 1 7.
Mater. 1999, 47, 2511 24. 8. O. Schultz, A. des Ligneris, O. Haider, P.
3. J. Banhart, W. Brinkers, J. Mater. Sci. Lett. Starke, Adv. Eng. Mater. 2000, 2, 215 18.
1999, 18, 617 19. 9. S. Suresh, Fatigue of Materials, 2nd ed,
4. Y. Sugimura, A. Rabiei, A. G. Evans, A. M. Cambridge University Press, Cambridge
Harte, N. A. Fleck, Mater. Sci. Eng. 1999, 1998.
A269, 38 48. 10. H. P. Degischer, B. Kriszt, Structural
5. K. Y. G. McCullough, N. A. Fleck, M. F. Investigations of Al-Foam using Quantitative
Ashby, Fatigue Fract. Eng. Mater. Struct. 2000, Image Analysis (Project-Report), Institute for
23, 199 208. Material Science, University of Technology,
6. T. BoÈllinghaus, H. v. Hagen, W. Bleck, Vienna 1997.
in Proc. DGM Symp. MetallschaÈume,
5.3
Electrical, Thermal, and Acoustic Properties of Cellular Metals
F. Simancik and J. Kovacik
It is evident that the presence of gas bubbles will significantly affect also electrical,
thermal, and acoustic properties of cellular solids. In the case of polymer foams
these properties play an important role: thermal insulation and noise attenuation
are beside packaging the main fields of application for this kind of materials. How-
ever, only a little attention is paid to the mentioned properties when cellular metals
are considered, although the multifunctionality is often the prerequisite for their
successful application. The aim of this chapter is therefore to reveal and critically
discuss the potential of cellular metals in this field with respect to published mod-
els and rarely reported experimental data. For better understanding of considered
properties, the basic physical background will be also briefly introduced.
5.3.1
Electrical Properties
Cellular metals are in distinction to polymer and most ceramic matrices electrically
conductive. This simple fact changes the typical applications drastically: ceramic
and polymeric foams can be used for insulation or for structural enclosures that
216 5.3 Electrical, Thermal, and Acoustic Properties of Cellular Metals
must transmit microwaves, while the metallic counterparts are good for the oppo-
site purpose. The electrical conductivity of foamed metals, although considerably
reduced by the porosity, is still sufficient enough to provide good electrical ground-
ing, low voltage contacts, or capability to absorb electromagnetic waves. In the case
of metallic sponges with open cell structure, the conductivity allows to distribute
the electric potential on large accessible area, which makes them attractive for bat-
tery electrodes. Cellular nickel is extensively used in this application [1].
It is generally accepted, that cellular material inherits electrical characteristics of
the solid of which it is made. The electrical conductivity s of material is defined
according to Ohm's law
l I
s (1)
S U
where U is a potential drop along the distance l when current I is applied. As can
be seen conductivity depends strongly on the cross-sectional area S of the sample.
In a case of cellular metals only a small portion of cross section is electrically con-
ductive. These are the cell walls with roughly the conductivity of the base alloy. The
main part of the cross section is created by pores, which are usually filled with di-
electric air. The cell walls form a complex 3D network of electric conductors of
varying cross section, whereas cell walls oriented normally to the potential gradient
do not contribute to conduction because of no potential difference. Moreover, the
cells are mostly continuously covered with electrically nonconductive metal oxides.
The cellular structure provides therefore much lower electrical conductivity than
bulk material.
To calculate the electrical conductivity according to Eq. (1) the effective length l of
the cell walls conducting along the potential difference and in the cross-sectional
area S effectively contributing to conduction has to be known. Unfortunately, it
is extremely difficult to determine it in real samples, although for example com-
puted tomography offers this possibility. More reasonable and cheaper is to con-
sider the sample as a black box of given outer geometry with macroscopically
homogeneous and continuos distribution of a material within the box and to inves-
tigate the so called effective electrical conductivity.
It is evident that the dependence of effective cross-sectional area on relative
density of the structure is non linear. One of the first models of this dependence
was introduced by Ashby et al. for the case of open-cell sponge [2]. The structure
was idealized with narrow struts of given length and thickness. The struts were
arranged in a perfect cubic lattice, with nodes linked by struts. Only struts oriented
parallel to potential gradient contributed to conduction. For higher relative den-
sity they added also the contribution of the nodes and obtained the following
relation
! !3=2
s r r
f S(1 s f) (2)
s0 r0 r0
where s and r are the effective property and density of the foam respectively, while
s 0 and r0 are the corresponding properties of the cell-wall material. The first term
of this equation describes the contribution of struts and the second one that of
nodes. Therefore, a structural parameter f was introduced, which characterizes
how the material is distributed between cell nodes and struts. However, in real
sponges the ratio between volume of cell nodes and cell struts is not constant
and varies also with the density. Thus f in Eq. (2) must be obtained experimentally
for each material and is valid only for certain density interval.
Another approach for modeling of the relationship between relative density and
selected physical property uses the principles of percolation theory [3 6] and
covers the whole density range. In this case the cellular metal is considered as
an ªinfiniteº cluster of randomly distributed gas pores in the metallic matrix.
According to percolation theory the effective property near the connectivity
threshold (in this case set to zero) behaves like a power-law function of a relative
density
!t
s r
z (3)
s0 r0
where s and r are the effective property and density of the foam respectively, while
s 0 and r0 are the corresponding properties of the cell-wall material. Constant t
(often called as a critical exponent) can be theoretically predicted for the specific
property. The theoretically estimated value for electrical conductivity in 3D is t
2.00 [7]. The constant z ought to be 1, because for bulk material r r0 and the
effective property s s 0.
Also this approach has some difficulties: It does not take into account defects
such as cracks in pore walls, which influence the conductivity but not the relative
density. Moreover, the theoretical predictions of critical exponents in 3D are still
under development and optimization, especially when finite size clusters or aniso-
tropic orientations are considered.
Only little data for electrical conductivity of cellular metals have been reported
[8 12] up to now (see also Table 5.3-1). Therefore, a serious verification of recent
theoretical models is very difficult, if possible at all.
f 0.33 and f 0.05 were obtained for ERG open-cell sponge and for Alulight
closed-cell foam, respectively, applying Eq. (2) to fit the reported data [2]. The con-
siderable difference in a, confirms an assumption that a is strongly material and
structure dependent and for each case has to be obtained experimentally.
The power law seems to be a good characterization of the relationship between
experimental data of electrical conductivity and density measured on Alulight sam-
ples [3 5] (see Fig. 5.3-1 and Table 5.3-2). The characteristic exponents obtained
experimentally have been found slightly lower than the critical exponent value the-
oretically predicted by the percolation theory. It can be explained by the fact that in
distinction to the theory, where an infinite cluster is taken into account, the real
experimental samples have always a finite size and in the referred case possess
a surface skin. The volume fraction of the surface skin increases with decreasing
Table 5.3-1.Overview of the electrical and thermal conductivity of cellular aluminum with various
density from suppliers data sheets [8 12] measured at room temperature.
Suppliers Density [kg.m 3] Thermal conductivity Electrical conductivity

[W/mK] [m/Vmm2]
Alporas [10] 570 ± 9.9
400 ± 5.0
300 ± 3.8
200 ± 2.6
94 ± 1.0
Alulight [9] 500 13 3.1
400 10 2.6
300 8.9 1.9
Cymat [11] 540 0.91 ±
410 0.7 ±
270 0.4 ±
160 0.36 ±
68 0.28 ±
ERG [8] 330 ± 2.1
250 ± 1.6
200 ± 1.2
140 ± 0.9
Norsk Hydro [12] 300 2.1 0.50
250 2.1 0.36
150 1.5 0.27
size of the sample. The high portion of surface skin can diminish the effect of the
porosity on the reduction of electrical conductivity of the foam. If the surface skin
is removed from the sample the reduction of the conductivity due to increasing
porosity is enhanced and characteristic exponent becomes higher. The similar ef-
fect can be observed when the sample size increases. It implies that the conductiv-
ity of the foam is not only a function of the porosity, but it is strongly affected also
by the shape and size of the foamed part.
Table 5.3-2. The obtained fitting parameters for the dependence of normalized electrical
conductivity on the relative density for Alulight Al 99.7. x 2 is a minimization function [7].
z t x2
5
s/0 (mix.) [4] 0.999 e 0.009 1.55 e 0.02 7.31 qx 10
4
s/s0 (Al 99.7) [4] 0.979 e 0.025 1.75 e 0.06 6.54 qx 10
Although the behavior can be qualitatively understood by noting that with de-
creasing density the cross section available for conduction decreases and the tortu-
osity of the current path increases, a quantitative theory is still lacking [13].
Good electrical conductivity of cellular metals minimizes the penetration of elec-
tromagnetic waves into the structure, especially in the case of closed-cell foam
parts covered by a surface skin. This fact can be successfully utilized for the pro-
tection of electronic devices or humans from the effect of electromagnetic noise.
The ability for electromagnetic wave shielding can be defined by shield effective-
ness depending on the wave frequency and the material thickness. The main
advantage of metallic foam is a possibility to achieve a thickness required for
low magnetic permeability at lower weight than in the case of bulk metals. The
experimental measurements on closed cell Alulight and Si-steel sheet of the
same weight are illustrated in Fig. 5.3-2 and Fig. 5.3-3 [9]. The shielding effectivity
of the foam is even slightly better than that of a bulk aluminum plate of the same
thickness, additionally the foam weight was approximately five times lower in this
case (see Fig. 5.3-2). Cellular aluminum foam exhibits much better magnetic field
Figure 5.3-1. The power-law dependence of normalized electrical conductivity on the relative
density for Alulight Al 99.7 foam (cylindrical samples with surface skin, s 0 37.6 q 106 S m 1,
r0 2700 kg m 3) [4], for the fit parameter see Table 5.3-2.
Figure 5.3-2. Magnetic field shielding effectiveness as a function of frequency for cellular
aluminum (Alulight) and Si-steel samples of the same weight and a massive Al sheet of the same
thickness (t 8.5 mm) as the aluminum foam. Sample size 140 q 140 q t, density of foam
500 kg m 3 (KEC-method) [9].
Figure 5.3-3. Electric field shielding effectiveness as a function of frequency for cellular alumi-
num (Alulight) and Si-steel samples of the same weight. Sample size 140 q 140 q t, the density
of foam 500 kg m 3 with t 8.5 mm (KEC-method) [9].
shielding properties than Si-steel sheet of the same weight. The electrical shielding
effectivity of Alulight is comparable with the properties of Si steel (see Fig. 5.3-3)
up to 10 MHz. For higher frequencies the aluminum foam is superior. Similar
results have been reported for Alporas foam [10]
5.3.2
Thermal Properties
The main thermal properties of practical importance with respect to cellular metals
are the melting point, specific heat, coefficient of thermal expansion, thermal con-
ductivity, thermal diffusivity, emissivity of the surface, fire and thermal shock resis-
tance. Some of these properties are almost the same as for bulk materials, other
ones depend strongly on the porous structure.
Melting point of cellular metals is practically the same as for an alloy from which
it is made. However, the surface is often covered by a continuous oxide layer, the
melting point of which is usually considerably higher: for example in the case of
cellular aluminum. The surface area covered with oxide layer increases with in-
creasing porosity and decreasing pore size. When the oxide layer is sufficiently
thick it can support the porous structure even above the melting point of the
alloy, provided that no significant external force is applied on the structure. If
the structure is heated in air or oxidizing atmosphere at the temperature close
to melting point for sufficiently long time, the thickness of oxide layer increases
and material becomes more stable. After a threshold thickness depending on the
mechanical stability of the sample, the metallic cell walls oxidize completely by
subsequent heating converting the cellular metal into a ceramic foam. This behav-
ior is utilized for instance for the preparation of ceramic sponges from open-cell
metallic precursors like Duocell sponges [14].
The energy needed to increase the temperature of cellular structure by a unit
temperature is almost the same as required for cell-wall materials when the unit
mass is considered. The slight difference can be accounted to the presence of
thin surface oxides and the contained air. However, the specific heat per unit volume
Cv of a cellular structures is of cause significantly lower. It makes them attractive
for applications where low thermal capacity is required, for example for rapid heat-
ing and cooling systems.
The coefficient of thermal expansion a of cellular metals is also almost the same as
for the cell-wall material, while the thermal conductivity l is much lower. Higher
thermally induced distortion has to be expected that increases with raising a/l ratio.
To avoid it, the temperature differences in the cellular structure on heating or cool-
ing should be minimized.
On the another hand if the temperature shock resistance is considered, the situa-
tion is not so unambiguous. When a sudden temperature difference occurs be-
tween outer surface and inner material regions the thermal expansion causes ther-
mally induced strains, the related elastic stress depend on the modulus of elasticity
of the material. The stresses depend on the product a E and as the modulus of elas-
ticity of cellular metal decreases with decreasing relative density the same tempera-
ture difference induces smaller thermal stresses the higher the porosity of the cel-
lular structure, assuming the coefficients of thermal expansion is constant. How-
ever, with decreasing density also the collapse or tensile stress of the material de-
creases. If this stress scales with relative density at smaller exponent than the mod-
ulus of elasticity does, the temperature shock resistance of the foam will improve.
If not, the foam may yield or even break earlier than bulk material. In distinction to
bulk materials, cellular metals are able to withstand much higher compressive
strains without fracture because of the existence of a plateau stress. Therefore
their temperature shock resistance is usually considered to be higher than that
of bulk metals. It should be noted, that the same sudden change of ambient tem-
perature causes higher temperature difference within a foam than within solid
metal because of significantly smaller thermal conductivity of the cellular metal.
It is evident that the thermal conductivity, l, of cellular metal will be significantly
lower than the corresponding conductivity of bulk material. In comparison with
electrical conductivity this property is more complex. Apparent thermal conductiv-
ity of porous structure has four contributions [13]: besides the thermal conductivity
of the solid cell walls also the conductivity of enclosed gas, convection and radiation
effects should be taken into account
l lS S lG S lc S lr (4)
where indexes S, G, c, and r denote the contribution to thermal conductivity of the
solid cell wall material, the gas within pores, convection within the cells, and
radiation through the cell walls. The thermal conduction is additionally affected
by the presence of low conductive metal oxides on the surface of pore walls and
cracks.
However, in a case of cellular metals only the first term in Eq. (4) is important,
the last three are practically negligible of moderate temperature levels. The contri-
bution of conduction through gas lG is very low in comparison with the conduction
through metal, because of low thermal conductivity of gas (thermal conductivity of
air is 0.025 W m 1 K 1). Convection within the cells is important only when the
Grashof number (the ratio between force driving convection and opposing viscous
force) is greater than about 1000 [13]. This value is achieved with air filled pores at
room temperature, when the pore diameter exceeds 10 mm. However, the pore size
in real metallic foams usually is smaller, so the convection should be negligible in
this case.
Radiation through the cell walls is not possible in a case of optically nontranspar-
ent metals. Radiation within the cells can be ignored as well, when the conduction
of the cell-wall material is greater than 20 W m 1 K 1 [15].
The contribution of the metal to the thermal conductivity of cellular structure (ls)
depends on relative density r/r0 and similarly as for electrical conductivity it can be
expressed by power law [3 6]
!t
r
lS l0 (5)
r0
where l0 denotes the thermal conductivity of solid metal. Constant t is the critical
exponent for thermal conductivity and according to percolation theory its theoreti-
cally predicted value in 3D is t 2.0 [7]. The characteristic exponents obtained
experimentally are listed in Table 5.3-3. The power-law dependence is illustrated
in Fig. 5.3-4.
Table 5.3-3. The obtained fitting parameters for the dependence of normalized thermal conduc-
tivity on the relative density for Alulight foam at various temperature. x 2 is a minimization function.
z t x2
5
l/l0 (20 hC) [5] 1.000 e 0.006 1.60 e 0.03 2.97 qx 10
5
l/l0 (100 hC) [5] 1.000 e 0.008 1.55 e 0.03 5.66 qx 10
5
l/l0 (300 hC) [5] 1.000 e 0.010 1.48 e 0.04 9.10 qx 10
5
l/l0 (400 hC) [5] 1.000 e 0.005 1.51 e 0.03 4.00 qx 10
As the thermal conductivity of bulk metals increases with increasing tempera-

ture, the same behavior is expected also in a case of cellular metals, in accordance
with Eq. (5). Experimental values determined for Alulight foams made from var-
ious aluminum alloys at higher temperatures are shown in Fig. 5.3-5. As expected,
with increasing temperature the effective thermal conductivity of foam increases.
However, a little drop of this property has been observed at temperatures above
300 hC. This small reduction may be caused by heat losses due to increasing radia-
tion effects, but more probably by extensive oxidation of the foam surface at higher
temperatures. The alloys (such as AlSi12), which are less sensitive to oxidation,
exhibit also less reduction of the effective thermal conductivity at higher tempera-
tures.
Figure 5.3-4. The power-law dependence of normalized thermal conductivity on the relative
density for Alulight foam at 20 hC. The corresponding fit parameters are given in Table 5.3-3
(cylindrical samples with surface skin, l0 232 W m 1 K 1, r0 2700 kg m 3) [3].
Figure 5.3-5. Thermal conductivity of Alulight (ls 200 W m 1 K 1); v, AlMgSi0.6 base alloy,
foams made of various alloys as a function of r 500 kg m 3, V, AlSi12 base alloy,
testing temperature and density r: y, Al 99.6 r 440 kg m 3 (ls 120 W m 1 K 1), x, AlSi12
base alloy, r 320 kg m 3 (ls 230 W m 1 K 1); base alloy, r 330 kg m 3
Y, AlMgSi0.6 base alloy, r 420 kg m 3
Thermal diffusivity a governs heat flow through a material before a steady state
flow by thermal conduction is achieved. It is defined as a ratio l/Cv between ther-
mal conductivity l and volumetric specific heat Cv. Foams have usually higher ther-
mal diffusivity than bulk cell-wall materials and thus they achieve a steady state heat
flow earlier. This property makes the cellular metals attractive for use as heat ex-
changers.
The emissivity of the foam surface is almost always higher than that one of the
metal from which it is made of. As the actual surface area of the foam is also
much larger at equal outer dimensions the foam will emit or absorb more thermal
radiation than bulk material. This property in combination with small specific heat
per unit volume can be utilized in applications where rapid heating or cooling by
radiation is required.
Although polymer and ceramic foams are used more for thermal insulation than
for any other purpose, application of cellular metals in this field is unlike because
their thermal conductivity l is about two orders of magnitude higher than poly-
mers (see Table 5.3-1). However, higher thermal conductivity and diffusivity of cel-
lular metals can be successfully utilized in heat exchangers or heat sinks for power
electronics, air-cooled condenser towers and regenerators. This application is re-
stricted only to open cell structures enabling flow of cooling gas or fluid without
significant pressure drop. The main design principles for heat exchangers made
from cellular metals are [2]: high conductivity of metal (preferably Cu, Al), turbu-
lent flow of fluid to enhance local heat transfer; and minimized low-pressure drop
between fluid inlet and outlet. For any system there is a trade-off between heat flux
and pressure drop. Optimum can be achieved with cell sizes in the mesorange and
with relative densities of order r 0.2. [16]
Relatively small thermal conductivity, noncombustibility and low weight make

the metallic foams attractive also for application as fire resistant panels. Thermal
fire resistance of a structural component can be defined as the time to failure by ex-
cessive heat transmission. In case of panels it can be defined as a time at which the
temperature of the unexposed surface of the panel reaches a critical value TC, for
example the melting temperature. The thermal fire resistance strongly increases
with increasing Biot number calculated for the panel [15]
Bi hL=l (6)
where h is heat transfer coefficient, L is the thickness, and l is the thermal conduc-
tivity of the panel. For given Biot number the fire resistance is infinite if the im-
pose fire temperature is smaller than a threshold value TF
TF TC S Bi(TC s T0 ) (7)
where TC and T0 are critical and initial temperatures of the panel, respectively. The
Biot number for a panel made of metallic foam is always higher than that one for a
corresponding metal sheet of the same weight. For example, it takes 8 min to heat
up the opposite side of a Cymat sheet (density 160 kg m 3, geometry 24 cm q
30 cm q 5.5 cm) from room temperature up to 500 hC using oxy/acetylene torch
(theoretical temperature 3100 hC) rotating at 15 cm distance from the foam front
side. The burn-through time for corresponding aluminium sheet of the same
weight (density 2700 kg m 3, geometry 24 cm q 30 cm q 0.34 cm) is only 23 s
[11]. This illustrates the opportunities of metallic foams for this application.
5.3.3
Acoustic Properties
Sound is a form of the energy transmitted by the wave of collisions of molecules or

atoms one against the next, and so on. It is important to note that there is no trans-
fer of matter in this case. The speed of longitudinal elastic vibrations c in a given
media is usually proportional to the ratio of the square root of the media modulus
of elasticity E to media density r ratio
s
E
ct (8)
r
For this reason the sound travels at velocity of about 5000 m/s in solid steel or alu-
minum, and only 343 m/s in air at the sea level. The wave velocity of sound is
related to the wavelength l and frequency f of sound wave by c lf. The normal
human ear responds to the frequencies from about 20 Hz up to 20 kHz, which cor-
responds to wavelengths of 17 m to 17 mm. For the noise control and reduction it
is often enough to reduce sound pressure mainly in the range of 500 4000 Hz.
The amplitude of sound wave determines the loudness of sound. The amplitude
of sound pressure is measured in Pascals (Pa), but it is more convenient to use
the logarithmic scale in decibels (dB) [2].
In general, the three basic cases can be distinguished with respect to noise at-
tenuation materials:
x materials for sound insulation: they keep airborne sound out of the protected
space,
x materials for sound absorption: they reduce or fully absorb the sound intensity
within the room where the sound is generated,
x materials for damping: they reduce the amplitude of the structure acoustic re-
sponse and hence the sound intensity born by the structure.
In the following the physical principles that determine the potential use of cellular
metals for sound management purposes will be briefly reviewed considering all
three mentioned cases. Guidelines for material selection and procedures ensuring
successful practical applications will be suggested.
5.3.3.1 Materials for Sound Insulation

The performance of sound insulation material is described in terms of sound re-
duction index, which is proportional to the reciprocal value of the sound transmis-
sion coefficient in logarithmic scale. Sound transmission coefficient for a panel is
given by the ratio of the transmitted to incident sound intensity. The sound reduc-
tion index depends on the frequency and, of course, on the material and geometry
of the structure. For a given material the sound reduction index is controlled by
various mechanisms, which become important in the specific frequency range.
The general form of the transmission loss curve is shown in Fig. 5.3-6 [17].
Coincidence
controlled
aw
sl
Resonance controlled
s
Mass controlled
ma
Stiffness controlled
f
iono
te ns
Sound reduction index [dB]
Ex
Coincidence dip
Effect of damping
Critical Frequency Hz
frequency
Figure 5.3-6. Principal dependence of sound reduction index of a panel as a function of
frequency for given material and structure geometry [17].
At very low frequencies the sound reduction is controlled by the stiffness of the
panel. In this case the application of cellular metals can be very attractive because
of their excellent stiffness to weight ratio (see parameter E/r3 in Table 5.3-4).
At somewhat higher frequencies the sound transmission is determined by the
natural resonance of the panel depending on the used material, panel size, and
mounting conditions. Also in this frequency range the cellular metals can be use-
ful, because of their better damping properties in comparison with bulk metals
(see also later in this Section). Moreover, they can shift the first resonant frequency
towards higher values and thus extend the stiffness controlled frequency range (see
Table 5.3-4). This can be important when the panels of smaller size are used.
However, for typical sound insulation, the most important is the frequency
range, where the sound reduction is controlled by mass of the panel according to
TL0 20log10 (mf s 43 (9)
where TL0 in dB is the sound reduction index for normal incidence, m is the super-
ficial density in kg m 2 of the soft impervious panel and f is the frequency in Hz.
This is known as Mass Law [17]: the sound transmission loss increases with in-
creasing mass of the panel. Therefore, the application of cellular metals for
sound insulation in this frequency range is highly disadvantageous.
Table 5.3-4. Design parameters for the materials suitable for sound insulation. Thickness h, 1st
resonant frequency f1 of fully clamped plate and critical frequency fc are calculated for the square
plate of given weight (10 kg) and area of 1 m q 1 m. Parameter E/33 demonstrates the bending
stiffness at minimum weight of the panel. AlSi12 aluminum foam of density 500 kg m 3 and
modulus of elasticity 5 GPa was introduced for comparison.
Material E r E/r3 h f1 fc f1*

[GPa] [g.cm 3] [m8.s 2.kg 2] [mm] [Hz] [Hz] [Hz]
Lead 14.3 11.2 0.01 0.89 1.8 60000 7
Steel 204 8.1 0.4 1.2 11 9770 43
Reinforced 23 2.3 1.9 4.4 24 4400 95

concrete
Glass 38 2.5 2.4 4.0 27 3900 108
Aluminium 69 2.7 3.5 3.7 32 3220 130
Aluminium foam 5 0.5 40 20.0 110 960 434
Plywood 4.2 0.58 22 17.2 81 2241 323
Plasterboard 1.5 0.75 3.6 13.3 33 3200 131
* 1st resonant frequency calculated for the panel size of 0.5 m q 0.5 m q h m in order to demon-
strate the size effect
At an even higher frequency a phenomenon of coincidence between bending

waves within the panel and the incident sound wave can occur accompanied
with a slight reduction of the sound insulation performance. Therefore this fre-
quency should be designed as high as possible. The lowest critical frequency fc
for the coincidence is given by
r
c2 r
fc (10)
1:8h E
where c is the velocity of sound in air, h is the thickness, E is the modulus of elas-
ticity, and r is the density of panel material. The use of cellular metals usually low-
ers the coincidence frequency and thus it cannot be recommended also in this case
(see Table 5.3-4).
5.3.3.2 Sound Absorbing Materials

There are various ways how the sound absorption takes place: by viscous losses as
the sound pressure pumps air in and out of the absorber's cavities, by direct me-
chanical damping in the absorber itself, by thermoelastic damping. The ability of
material to absorb a sound is characterized by sound absorption coefficient, a. The
sound absorption coefficient is the ratio of the unreflected sound intensity at the
surface to the incident sound intensity. Full sound absorption means that incident
sound wave is neither reflected nor transmitted and its energy is absorbed within
the absorbing material. The sound absorption coefficient varies with frequency and
angle of incidence of sound wave and for any given material is a function of ma-
terial thickness, density, and flow resistance of the structure (depends on fiber or
pore size) [17]. The absorption coefficient for normally incident sound waves of
small samples (diameter up to 100 mm) can be easily measured using an impe-
dance tube. If the sound absorption for all angles of incidence is required to be
known, the measurement in a reverberation room is usually applied. Much larger
samples (often up to 10 m2) are needed in this case. Although the theoretical max-
imum value of a is unity, the latter method can result in data slightly in excess of
1.00 because of edge effects.
The sound wave particle velocity immediately adjacent to a rigid wall is zero, and
therefore the absorber will be most effective, if it is situated in a distance of a quar-
ter wavelength from the surface. This leads to a great thickness of absorber if the
absorption at low frequencies is required.
For sound absorption purposes highly permeable materials such as open-cell
polymer foams and glass or mineral wool fibers are generally used. Flammability
and evolution of toxic gases when subjected to excessive heat is the main disadvan-
tage of polymer foams. On the other hand fibrous materials are very sensitive to
erosion by shedding or fraying especially under the effects of air flow or vibration.
Both type of absorbers usually require various facing materials in order to improve
durability or to protect the absorber from contamination. The facing materials are
often also aesthetic (various surface finishing and colors) with added advantages of
thermal insulation, fire safety, humidity and oil resistance.
Beside bulk absorbers two types of resonant absorbers are in general use, espe-
cially for sound absorption at lower frequencies. The panel absorber is a resonant
system formed by the mass spring combination of the facing panel and the stiff-
ness of enclosed air. The resonant frequency of panel absorber at which the max-
imum absorption is reached is given by
600
f p (11)
md
where m is the superficial density of the panel in kg m 2 and d is the depth of air
gap in cm.
The Helmholtz resonator or cavity resonator is essentially a vessel in which the
mass of air in the neck is driven in and out in resonance upon the stiffness of en-
closed air volume. The resonant frequency is given by
r
S
f 55 (12)
LV
where S is the cross-sectional area of the neck, L is the length of the neck, and V is
the enclosed volume of air.
The perforated panel often used for protection of porous absorbers can provide
an additional resonance effect, similarly, as it is in a case of Helmholtz resonator.
Such semiresonant performance is found when the percentage of open area of
front panel is below 20 30 %. Above this value, the system behaves primarily as
a simple porous absorber [17]. The faced material itself can be constructed more
sophisticated containing a plenty of small Helmholtz resonators (see Fig. 5.3-7).
The frequency at which the maximum sound absorption is reached can be
tuned by cavity geometry. Such combination usually shifts the maximum absorp-
tion performance towards low frequencies, preferably in the range 100 1000 Hz.
Figure 5.3-7. Typical construction of sound absorber: thick glass fiber absorber (1), thin pro-
tective and damping foil (2), facing wood panel with resonator cavity (3), and outer design (4)
[18].
Accordingly, the main parameter for good sound absorption is the permeability
of the absorber. Therefore, cellular metals can be used for this purpose only if they
meet this fundamental requirement. The closed-cell metallic foams are too stiff to
convert sound energy into heat by vibration of their cell walls. According to Eq. (11),
they cannot be effectively applied for panel resonators because of their very low
weight.
Unfortunately, the experimental measurements performed in impedance tube
on various types of cellular metals [19,20] are in a good coincidence with these dis-
couraging expectations (see Fig. 5.3-8). The best sound absorbing performance was
obtained using Access-sponge metals having open-cell structure and thus very low
flow resistance [21]. Such structure absorbs the sound energy mainly by viscous
losses as the sound pressure wave pumps air in and out of the interconnected
pores. Alporas closed cell foam (made by melt-route process) [22] is manufactured
in large blocks from which the absorbers are machined. Machining results in
rough open-pore surface that slightly improves the sound absorption. Alulight
closed-cell aluminum foam [9] is prepared in net shape and therefore is covered
with a surface skin. This relatively flat and pore free skin reduces the sound ab-
sorption performance almost to zero: the sound is entirely reflected.
Figure 5.3-8. Sound absorption coefficient of various types of cellular aluminum foams in
comparison with fiberglass sound absorbing material (directly on rigid background).
Sample Access sponge Alporas Alulight Fiberglass
Thickness [mm] 20 10 9 48
Density [kg m 3] 1060 270 580 40

Access material has a good potential to be applied for sound absorption purposes
after optimization of the structure. In a case demonstrated in Fig. 5.3-8 the flow
resistance was too low to provide sufficient viscous losses. However, the manufac-
turing process allows to adjust the optimum pore size or even to obtain gradient
pore structure in relatively easy and inexpensive way [21]. Figure 5.3-9 demon-
strates the effect of the reduction of open area at the surface on the sound absorp-
tion performance of this material.
Although the sound absorption performance of cellular metals is not very im-
pressive, it can be significantly improved by optimal opening of close cell structure.
In this case the foams can serve as multiple cavity resonator (see Eq. 12) with
much longer ªneckº than in the case of bulk materials of the same weight. The
variability of the pore size will provide a wider frequency range of good sound ab-
sorption than typical resonators. Various methods, such as cutting of surface skin,
sand blasting, compressing, rolling, and drilling of holes were examined to open
the foam structure [19,20]. It was shown that sound absorption of as received Al-
poras aluminum foam can be significantly improved by partial compression, roll-
ing, or hole drilling. However, the combination of these method offers only a little
additional benefits [19].
Similar results were obtained for Alulight foams covered with surface skin (see
Fig. 5.3-10). The significant improvement of the sound absorption was reached al-
ready after opening of only 1.4 % of the surface area by drilling of holes. The open-
ing of the surface area can be more economically realized by simple punching of
the foam surface without the need to make the throughgoing holes (Fig. 5.3-11).
Figure 5.3-9. Sound absorption coefficient of Access aluminum sponge without and with
perforated facing Al sheet (air gap of 48 mm behind the absorber, Access sponge: density
1180 kg m 3, thickness 20 mm; facing Al sheet: thickness 1 mm, 61 holes diameter 2 mm).
Figure 5.3-10. Sound absorption coefficient as a function of opened surface area for Alulight
foam (8.9 mm thick AlSi12 foam, density 580 kg m 3, holes drilled through the sample, 20 mm air
gap between sample and rigid wall).
Surface opening 0.07 % 0.19 % 0.61 % 1.40 % 2.49 %
Number of holes 7 19 61 61 61
Diameter of holes [mm] 1 1 1 1.5 2
As expected [23], the increasing thickness of absorber shifts the maximum of the
sound absorption coefficient towards lower frequencies (see Fig. 5.3-12), while the
shape of the curves remains almost unchanged provided, that the density and pore
size do not depend on thickness. At constant sample thickness, the foams with low
density were found to be better sound absorbers than the foams with high densi-
ties in the frequency range of 200 1700 Hz [19], however the obtained results are
affected by an increasing pore size with decreasing density. Another experimental
measurement [24] indicates that there exists an optimal pore size for a sound ab-
sorption with respect to sample thickness at constant density. This experimental
finding agrees with theoretical calculations made by Wang and Lu [25] for 2D cel-
lular solids of Voronoi structure: the sound absorption performance will be im-
proved with increasing pore size for sample, which is infinitely thick. It implies
that the sound absorption coefficient is a complex function of foam thickness, den-
sity, and pore size.
It is not always feasible to increase the thickness of absorber, because of the
weight limitations. The use of a thinner section of absorber spaced by an air gap
gives very similar results to a thick section (see Fig. 5.3-13), however at signifi-
cantly lower weight. Therefore, the frequency range of maximum sound absorption
Figure 5.3-11. Sound absorption coefficient as a function of opened surface area for Alulight
foam (AlSi12 foam 15 mm thick, density 380 kg m 3, diameter of holes 1 mm, holes were punched
into the depth of 8 mm, directly on rigid background).
Figure 5.3-12. Sound absorption coefficient as a function of absorber thickness (Access AlSi
sponge, average pore size 4 5 mm, density 1000 kg m 3, directly on rigid background).
Figure 5.3-13. Comparison of sound absorp- is thicker foam at zero distance from rigid wall
tion for two designs of sound absorbing panels (Access AlSi sponge, density 1000 kg m 3,
at two constant thicknesses: 1st type is the average pore size 4 5 mm).
foam at given distance from rigid wall, 2nd type
can be tailored by creating an optimum air gap between the absorber and rigid wall
[19,20,24] without need to increase weight. The sound absorption can also be en-
hanced for wider frequency range by the combination of several foam plates
with an air gap between them [20].
Although the cellular metals cannot provide a high level of sound absorption
itself, they can be attractively used as a facing materials, especially in combination
with fibrous absorbers, provided, that the structure is sufficiently opened. The fre-
quency corresponding to the maximal sound absorption can be tuned by the thick-
ness of foam panel or by changing of pore size. The sound absorbing performance
is then comparable with typical solutions using facing wood panels (see Fig. 5.3-14).
The cellular metals offer additionally very high stiffness at low weight, are self-
supporting and stable at elevated temperature, does not release toxic gases when
subjected to excessive heat, possess high durability under the effects of air flow
or vibration, exhibit sufficient fire resistance and can be recycled. Moreover, as a
relatively new material with impressive surface design, it can be very attractive
for architects and designers of building interiors (see Table 5.3-5).
The potential applications are expected in sterile (antiseptic, dust free) environ-
ment, in the structures where nonflammability is important (airplanes, hotels,
commercial, and industrial buildings) or in interiors under difficult conditions (ele-
vated temperature, moisture, dust, flowing gas, and vibrations).
Figure 5.3-14. Sound absorption coefficient of fiber glass itself and faced with AlSi12 Alulight
(thickness 13.2 mm, density 330 kg m 3, diameter of holes 2 mm) with Helmholtz resonators
inside (cavity diameter 6.5 mm, length 8 mm). For comparison the sound absorption charac-
teristic of facing sheet of Topperfo acoustic panels TP 16/16/12-2 [18] is presented.
Table 5.3-5. Surfaces of cellular aluminum for improved sound absorption.
Sample Access Alporas Alulight
Thickness [mm] 25 10 8.9
Density [kg.m 3] 1060 480 450
Weight [g] 208 38 31
Surface Cut Cut Foamed skin
Opening As-received Compressed Drilled holes
Structure
The cellular Alporas already has been applied commercially in sound-absorbing

linings on the underside of highway bridge, and in sound-absorbing wall panels at
the entrance of tunnels in Japan [22].
5.3.3.3 Structural Damping

If the structure is subjected to external excitations by sound waves or by mechan-
ical vibrations, the sound can be transmitted or even born by the structure itself.
This is most important in the cases, when the structure oscillates at its resonant
or eigen frequencies. The amplitude of the structure acoustic response and
hence the sound radiation can be dramatically reduced by increasing the damping
of the structure.
Structural damping is owing of internal friction in a material by conversion of
the vibrational energy into heat and subsequent dissipation of heat into the sur-
rounding area. The high level of damping in the structure can be realized either
by utilization of high damping materials or by friction between sliding surfaces
of closely attached structural components interface damping.
The rate of vibrational dissipation or the damping in the structure can be char-
acterized by the loss factor h. The measurement of the loss factor is normally car-
ried out by one of two basic methods [26]:
The ªfrequency responseº method is used when the damping is high (h j 10 3).
The loss factor can be calculated from the bandwidth Df at the amplitude drop of
3 dB points and resonant frequency fn according to
h Df =fn (13)
The second ªdecay-rateº method is used for materials with lower damping. The
loss factor is calculated from the vibration amplitude decay D in dB s 1 at resonant
frequency after instant stopping of exciting force, according to
h 0:0366 D=fn (14)
The typical structural materials with high strength (for example steel, cast iron,
aluminum alloys) have unfortunately very little damping. On the another hand,
the high damping materials, such as lead, rubber, or soft plastic, have little
strength and cannot be regarded as structural materials [17].
The cellular metals exhibit at least one order of magnitude higher values of the
loss factor h [20]. The dissipation of the vibrations mainly results from the friction
between the attaching surfaces of the cracks appearing in the structure and
partially due to the vibration of the thin pore walls. Thus, the damping can be
enhanced by the reduction of the cell-wall thickness or/and by introducing of
imperfections into the structure. Higher loss factor values are therefore obtained
with for example foams made of casting aluminum alloys, which can be prepared
with very thin cell walls and contain a lot of cracks. If the nonsoluble ceramic
particles, for example SiC, Al2O3, or graphite are additionally introduced into the
structure of metallic cell walls, the damping will be further improved. New sliding
surfaces born in the cell walls in this way are responsible for this behavior (see
Figure 5.3-15. Effect of addition of nonsoluble particles on the damping properties of aluminum
foams and steel tubes filled with them (30 mm diameter q 300 mm, foam density 600 kg m 3).
Fig. 5.3-15). Nevertheless, the loss factor of typical cellular metals is still too low
compared with standard damping materials having h in the range 0.01 0.1.
Recently, there is some effort to apply the cellular metals as stiffeners or cores in
metallic hollow parts, castings, or sandwiches. Beside the stiffening and the in-
creased capability to absorb impact energy, also the reduction of noise and vibra-
tions is expected (see Fig. 5.3-16) [27]. In this case the structural damping is sig-
nificantly affected by the quality of contact between cellular metal and bulk metallic
shell. If the attached surfaces can slide on each other under pressure an additional
interface damping will be obtained. The damping effect is produced by dry friction,
which can provide very effective attenuation of excessive vibrations.
Figure 5.3-16. Frequency spectra of reciprocal mass for a hollow profile and a profile foamed
with various types of Alulight foams (steel tube 22 mm diameter q 245 mm filled with 18 mm
diameter q 245 mm foams of density 700 kg m 3).
Thus, though disappointing, in a case of a very good (usually metallurgical)

bonding, which is a precondition for good mechanical properties, the good damp-
ing cannot be expected. On the other hand, the best damping will be obtained
without bonding or using ªslidingº contact provided by an elastic adhesive (see
Fig. 5.3-17).
The high values of loss factor can indicate the absence of bonding or even insuf-
ficient filling of the cavity (see Fig. 5.3-18) with cellular metal. This effect can be
utilized for the nondestructive testing of the quality of foamed profiles for various
industrial applications.
Figure 5.3-17. Vibrational damping in complex epoxy (foam density approximately 700 kg m 3).
aluminum casting and its variations due to the For comparison the damping solution with
presence of Alulight foam insert injected or bitumen layer (3 mm thick) is also plotted.
fastened with either elastic adhesive or cured
Figure 5.3-18. Effect of the foaming on the values of resonant frequency and damping charac-
teristics of hollow steel profile.
Figure 5.3-19. Dependence of the first resonant frequency on the apparent density of Alulight
foams: longitudinal vibrations of AlSi12 foam 25 mm diameter q 300 mm (left axis) transverse
vibrations of Al 99.7 foam 140 mm q 140 mm q 8.5 mm (right axis).
Figure 5.3-20. Property gain obtained by use of material was used (weight 390 kg, self deflec-
aluminium foam for stiffening of hollow mation 4 mm, first resonant frequency 240 Hz,
chine table (800 mm q 720 mm q 250 mm) damping at first resonant frequency 0.15 %)
[28]. As a reference the table made of bulk
Beside mentioned approaches there is also a possibility to avoid the effect of re-
sonant vibrations by stiffening of the structure. The stiffening arrangements do not
damp the resonances, they merely shift them towards higher frequencies. If the
resonances can be shifted to frequencies that will not be excited during normal op-
eration of the equipment this solution to the problem of reducing vibrations may
be very efficient.
The resonant frequencies generally depend on the modulus of elasticity E and
the density r of the material of which the component is made and, of course, on
its geometry. For example, the first resonant frequency f1 of longitudinal vibration
of cantilever is
s
3:52 EI
f1 (15)
2p rSL4
where I is area moment of inertia of beam cross section S and L is the length of the
beam.
To shift a frequency either the geometry or material ought to be altered. The
changes of the part's geometry are often not possible. The altering of the mate-
rial may not be always successful at least in the case of metallic parts, because
E/r ratio of many metals is almost the same. Here the cellular structure can be
very attractive, while both the modulus of elasticity as well as the density can be
varied without changing the outer dimensions of the component (see Fig. 5.3-19).
This feature can be very effectively utilized for shifting the resonant frequencies in
hollow machine components by filling them with cellular metals. Figure 5.3-20
shows the effect of the utilization of aluminum foam as a filler in hollow machine
table. Beside stiffening and improving of the damping a significant shift of the first
resonant frequency towards higher values has been achieved [28].
References
1. J. Babjak, V. A. Ettel, V. Passerin, US Patent Properties of Solids, Military Academy Lip-

4 957 543, 1990. tovsky Mikulas, 1996, p. 57.
2. M. F. Ashby, A. Evans, N. A. Fleck, L. J. 6. J. Kovacik, Acta Mater. 1998, 46, 5413.
Gibson, J. W. Hutchinson, H. N. G. Wadley, 7. D. Stauffer, A. Aharony, Introduction to Per-
Metal Foams: A Design Guide, Butterworth- colation Theory, 2nd edn, Taylor & Francis,
Heinemann, Oxford 2000. London 1992.
3. J. Kovacik, F. Simancik, in Metal Foams and 8. ERG, Data Sheets, ERG Materials and Aero-
Porous Metal Structures, J. Banhart, M. F. space Corp., Oakland, USA 1996.
Ashby, N. A. Fleck (eds), MIT Verlag, Bremen 9. Mepura, Data Sheets, Mepura GmbH.,
1999, p. 303. Ranshofen, Austria 1995.
4. J. Kovacik, F. Simancik, Scripta Mater. 1998, 10. Alporas, Data Sheets, Shinko Wire Com-
39, 239. pany Ltd., Nakahama, Japan 1994.
5. J. Kovacik et al., in Proc. Int. Conf. Effect of 11. Cymat, Data Sheets, Cymat Aluminium
Non-Standard External Factors on Physical Corporation, Mississauga, Canada, 1996.
12. Norsk Hydro, Data Sheets, Hydro Alumi- 21. F. Grote, P. Busse, Giesserai 1999, 10,
nium Ltd., Sunndalsora, Norway, 1995. 75 78.
13. L. J. Gibson, M. F. Ashby, Cellular Solids, 22. T. Miyoshi, M. Itoh, S. Akiyama, A. Kita-
Pergamon Press, Oxford, 1988. hara, in Metal Foams and Porous Metal Struc-
14. E. S. Park, S. D. Poste, US Patent 4 808 558, tures, J. Banhart, M. F. Ashby, N. A. Fleck
1989. (eds), MIT Verlag, Bremen 1999, p. 125.
15. T. J. Lu, C. Chen, in Metal Foams and Porous 23. H. Utsumo, T. Tanaka, T. Fujikawa,
Metal Structures, J. Banhart, M. F. Ashby, J. Acoust. Soc. Am. 1989, 86, 637.
N. A. Fleck (eds), MIT Verlag, Bremen 1999, 24. F. Chen, D. P. He, in Metal Foams and
p. 391. Porous Metal Structures, J. Banhart, M. F.
16. A. G. Evans, J. W. Hutchinson, in Metal Ashby, N. A. Fleck (eds), MIT Verlag, Bremen
Foams and Porous Metal Structures, J. Banhart, 1999, p. 163.
M. F. Ashby, N. A. Fleck (eds), MIT Verlag, 25. X. Wang, T. J. Lu, J. Acoust. Soc. Am. 1999,
Bremen 1999, p. 45. 106, 1.
17. P. D. Wheeler, in Industrial and Machinery 26. Bruel & Kjaer, Mechanical Vibration and
Noise Control Practice, Institute of Sound & Shock Measurements, K. Larsen & Son, So-
Vibration Research, University of South- borg, Denmark 1980.
ampton, UK 1978, p. 11.1. 27. F. Simancik, F. Schoerghuber, in Porous
18. n'H Akustik Design AG, Data Sheet, and Cellular Materials for Structural Applica-
Lungen, Switzerland, 1999. tion, D. S. Schwartz, D. S. Smith, A. G.
19. T. J. Lu, A. Hess, M. F. Ashby, J. Appl. Phys. Evans, H. N. G. Wadley (eds), MRS Symp.
1999, 85, 7528. Proc. Vol. 521, MRS, Warrendale, PA 1998,
20. J. Kovacik, P. Tobolka, F. Simancik, in p. 151.
Metal Foams and Porous Metal Structures, 28. R. Neugebauer, Th. Hipke, in Me-
J. Banhart, M. F. Ashby, N. A. Fleck (eds), tallschaÈume, H. P. Degischer (ed), Wiley
MIT Verlag, Bremen 1999, p. 405. VCH, Weinheim 2000, p. 515.
6
Modeling and Simulation
F. G. Rammerstorfer, T. Daxner, and H. J. BoÈhm
The modeling and simulation of cellular metals is a field of research that has at-
tracted considerable interest from researchers, research groups, foam manufac-
turers, and industrial end-users. Two major directions of research can be distin-
guished: the design of metallic foams on the one hand and the design of compo-
nents made of cellular metals on the other.
Modeling and Simulation for Material Design

From the point of view of the manufacturers, the design of mechanically efficient
metallic foams is the paramount aspect of foam modeling and simulation, the aim
being the development of ªoptimumº cellular metals. The most important design
variables are the production technologies and the choice of the metal making up
the solid phase of the foam. Besides the apparent density and the properties of
the metal, the mechanical behavior of metallic foams is determined by the geome-
trical arrangement of voids and solid regions in the material, which will be referred
to here as the microgeometry. Consequently, a considerable number of studies have
aimed at exploring the connections between the microgeometry and the mechan-
ical properties of cellular materials and thus belong to a research field known as
micromechanics of materials.
At the length scale of the voids, the deformation of cellular materials tends to be
dominated by local mechanisms that must be accounted for in any modeling effort.
Because detailed descriptions of the microgeometries of large structures or compo-
nents made of foam are far in excess of present capabilities, the most fruitful ap-
proaches for simulating the thermomechanical behavior of metallic foams have
aimed at studying representative regions of appropriate model materials in detail.
Most commonly, actual foams are approximated by periodic microgeometries that
can be described via unit cells subjected to appropriate boundary conditions. Alter-
natively, a geometrically detailed microregion may be embedded in a larger region
for which a much simpler description is employed, leading to embedded cell mod-
els. A discussion of such micromechanical approaches is presented in Section 6.1.2.
Microgeometrical models covering a wide range of complexity can be studied
with the above methods to give predictions for the overall mechanical behavior
244 6 Modeling and Simulation
of metallic foams and link it to specific local deformation processes. In addition,

microscale geometrical imperfections can be introduced in a controlled way and
assessed for their influence on the macromechanical material response, making
it possible to isolate the influence of individual parameters. It has been found dif-
ficult, however, to make fully quantitative predictions on the basis of micromechan-
ical simulations, the main reason being the high geometrical complexity of actual
foams and the dearth of reliable data on the in-situ material behavior of the metal-
lic phase. Nevertheless, micromechanical approaches have provided important in-
struments for gaining a better understanding of the thermomechanical behavior of
metallic foams, see Section 6.1.3.
If the mean cell size or the effective density of the foam exhibit spatial variations
or gradients over a sample or component, additional information can be obtained
by explicitly considering length scales that lie between the micro- and macroscales.
Such mesomechanical studies are described in Sections 6.1.4 and 6.2.
Modeling and Simulation for Component Design

Designers of components made of metallic foams are interested in easy-to-use
methods of describing the constitutive behavior of these materials, for example
in the form of constitutive material laws for use with general purpose finite ele-
ment codes. For such purposes it is neither possible nor desirable to account for
details of the foams' microgeometry at each position in the component. Instead,
the material behavior of the foam is described in terms of a (fictitious) equivalent
homogeneous material. Such constitutive models may be derived from microme-
chanical studies by homogenization, or they may take the form of phenomenolo-
gical macroscopic descriptions that employ material parameters, which have to
be obtained from experiments. Because metallic foams typically show a limited
elastic range, such constitutive models have to be nonlinear for general application.
For some purposes full constitutive descriptions are not required and material
characterization can provide the necessary information. This can take the form
of experimentally based relations, for example in the form of stress strain rela-
tions parameterized by the effective density of the foam. Alternatively, microme-
chanical reasoning may be used to derive generic mathematical relationships,
which can then be fitted to experimental results and provide physically based re-
gression formulae.
Knowledge of the effective elastic stiffness on the continuum level is sufficient
for linear stress analysis, where the foam is treated as a homogeneous, linear elas-
tic solid. Depending on the complexity of the problem, such structural analyses can
either be performed analytically or numerically, for example by using the finite ele-
ment method. If only the macroscopic stresses within the component are of inter-
est, the knowledge of the homogenized material properties (in this case the tensor
of elasticity) will be sufficient. These macroscopic stresses have to be assessed with
respect to failure of the structure by yielding, fracture, or buckling, so that addi-
tional information in the form of macroscopic strength data is required.
More sophisticated structural analysis must take into account the nonlinear be-
havior of metallic foams; this is an absolute necessity when large strains are pres-
ent, for example in crushing or crash situations. Here, incremental macroscopic

material laws, that is, relationships between increments in stresses and strains,
must be available to allow the use of typical numerical analysis tools. Several con-
stitutive material laws describing the overall behavior of cellular metals have been
proposed and applied in the simulation of components consisting of or containing
metallic foams. Obviously, the selection of a particular material law is governed by
the required material parameters and by the effort necessary for calibrating them
by experiments or via micromechanical studies. Because they are based on the use
of an equivalent homogeneous continuum, macroscopic material laws should only
be used for studying components or samples that are considerably larger (and
thicker) than the typical cell size of the foam. Macroscopic material laws for
foams are the focus of Section 6.1.5, where the advantageous use of metallic
foams in energy absorbing structures and modeling strategies for assessing crash-
worthiness are discussed.
Because metallic foams show some potential for controlled spatial variation of
their effective density and, as a consequence, of their mechanical properties,
they offer the possibility of designing functionally graded cellular materials.
Some considerations pointing in this direction can be found in Section 6.1.6.
Considering all the facts mentioned above, the structural analysis of components
made of or employing metallic foams is likely to become a standard procedure in
finite element simulations. As new production technologies and new practical ap-
plications emerge, the micro- and macromechanics of metallic foams continue to
be promising fields for future research.
6.1
Modeling of Cellular Metals
T. Daxner, H. J. BoÈhm, M. Seitzberger, and F. G. Rammerstorfer
In the course of the last decade the modeling and simulation of cellular materials
have gained in importance. Well-established representatives of this group, such as
honeycombs or polymer foams, as well as recently developed materials, such as
metallic foams, have been the subjects of various studies. Cellular materials,
which consist of a solid skeleton in the form of struts and/or cell walls and a
high volume fraction of voids, are highly inhomogeneous. This heterogeneity
leads to thermomechanical responses that are markedly different from those of
bulk solids and it gives rise to material properties that have made cellular materials
attractive for many engineering applications. It also presents an obvious target for
modeling studies aimed at gaining an improved understanding of the mechanical
behavior of cellular materials and structures made of them.
246 6.1 Modeling of Cellular Metals
6.1.1
Motivation
Metallic foams, which are the subjects of the present discussion, have entered the
stage of practical application, their most important mechanical properties being ex-
cellent energy absorption capacity as well as high specific strength and stiffness.
Studying their mechanical behavior has attracted considerable research interest,
both from the theoretical and the applications points of view.
Simulations of the mechanical responses of cellular metals may be carried out at
different length scales. In micromechanical approaches the inhomogeneous struc-
ture of these materials is accounted for at the level of individual cells, cell walls,
struts, and vertices either via statistical arguments or by discrete geometrical mod-
els. The corresponding length scale, which is of the order of the size of individual
cells, is called the microscale in the following. Studies at this level provide informa-
tion on the local deformation and load-transfer behavior that can then be correlated
to the mechanical behavior of the material at the structural level. A good under-
standing of the mechanics of cellular metals on the microscale is especially impor-
tant for identifying advantageous microgeometrical parameters for materials de-
sign and development. Microscopic models for metallic foams are the focus of
Section 6.1.2.
For studying samples and components that are, say, two or three orders of mag-
nitude larger than the size of an individual cell, only the overall thermomechanical
behavior is of interest and local details of the metallic foam do not have to be ac-
counted for explicitly. In other words, at the macroscale the foam is treated as a
homogeneous continuum rather than an inhomogeneous medium. Modeling ap-
proaches appropriate for describing metallic foams at the macroscopic level,
which are often rather application specific, will be discussed in Section 6.1.5. In
addition, questions involving the spatial variations or gradients of cell sizes and
shapes within a given sample or structure may be studied at length scales that
are intermediate between microscale and macroscale. Some simple models in-
volving such a mesoscale will be presented in Section 6.1.4, while Section 6.2 is
largely devoted to mesoscopic approaches.
In the following some important aspects of the present state of the modeling and
simulation of metallic cellular materials are discussed. It should be noted, however,
that no attempt is made to give a full overview of this research topic. For compre-
hensive information on the behavior of cellular materials and of its modeling the
reader is referred to the books by Gibson and Ashby [1,2] and to an overview by
Weaire and Fortes [3]. A recent review of the mechanical behavior of metallic
foams was given by Gibson [4].
6.1.2
Micromechanical Modeling of Cellular Materials: Basics
From the point of view of micromechanical modeling, cellular metals fall into three
groups: honeycombs, which can be studied by 2D models, open-cell foams, the
solid scaffold of which is dominated by beam-like members, and closed-cell
foams, in which membrane- or shell-like cell walls are present. All of these materi-
als typically show a limited elastic range and their mechanical behavior tends to be
dominated by local deformation mechanisms such as bending, buckling, plastic
yielding, and fracture of cell walls and struts. These deformation mechanisms,
in turn, are highly sensitive to details of the microgeometry, which, in practice,
may be quite regular for some honeycombs, but tends to be complex and highly
non-uniform in the case of metallic foams.
The mean field and Hashin±Shtrikman-type methods that play an important role
in continuum micromechanics of composites and materials with small volume
fractions of pores [5], have seen only limited use for cellular materials. Hashin±
Shtrikman upper bounds for the elastic moduli can be evaluated for macroscopi-
cally isotropic cellular materials by prescribing vanishing stiffness for the void
phase (the lower bounds, however, vanish trivially). Typically, the elastic moduli
of metallic foams lie considerably below these upper bounds, but for honeycombs
it was possible to identify microgeometries that realize the upper bounds [6].
As a consequence, most micromechanical studies of cellular materials in general
and of metallic foams in particular have been based on discrete microgeometrical
models. Owing to the high geometrical complexity and irregularity of actual metal-
lic foams such model microgeometries tend to be highly idealized. One common
modeling strategy is based on studying periodic ªmodel foamsº, the thermomechan-
ical behavior of which is fully described by appropriate unit cells. Alternatively, cells
or geometrical units may be studied in isolation without requiring them to be
space filling. A third type of approach employs a geometrically fully resolved micro-
region (ªcoreº) that is embedded in a much larger region in which the microgeo-
metry is not resolved and smeared-out material behavior is used (ªembedded cell
modelsº). Discrete microgeometrical modeling approaches facilitate the controlled
variation of selected geometrical parameters in order to assess their effect on the
mechanical response. These modeling strategies, which have well-established
equivalents in continuum micromechanics of composite materials, allow studying
both the local deformation mechanisms and the corresponding overall behavior,
which can be obtained by homogenization.
6.1.2.1 Analytical and Numerical Models

When a sufficiently high degree of abstraction is introduced into discrete micro-
geometry models of cellular materials, the mathematical description of the micro-
mechanical system becomes accessible to analytical methods and closed-form solu-
tions can be obtained. A by now classical example of such an approach is the work
of Gibson and Ashby [1], who based their models on single cells that do not give
rise to space-filling periodic arrangements and developed a large body of results on

the elastic deformation, elastic buckling, plastic collapse, brittle fracture, visco-
elastic deformation, creep, and creep buckling of honeycombs, open-cell, and
closed-cell foams [7,8]. Developments of this class of models have included,
among others, large deformation analyses of the buckling collapse of honeycombs
[9] and studies on the power-law creep of honeycombs [10].
Analytical formulae for the elastic behavior of a number of perfectly regular per-
iodic cellular geometries were given by Grenestedt [11], and scaling relations for
arrangements of this type were published by Christensen [12]. Tetrakaidecahedral
cells were used in a series of papers by Zhu et al. to describe the elastic behavior
[13] and the finite deformation compressive response of periodic open-cell [14] and
closed-cell [15] foams, numerical methods being used to resolve the resulting equa-
tions in some cases. In addition to studying perfectly regular cellular microgeome-
tries, imperfections of the cells, such as waviness of cell walls and struts, can be
accounted for [16,17]. Analytical methods may also be applied to predict the plastic
collapse of cellular structures as proposed by Santosa and Wierzbicki [18] who pres-
ented a kinematic model for the crushing of a column of truncated cubes.
Generally, analytical descriptions can only be used for relatively simple generic
microgeometries. More complex geometrical arrangements typically have to be
analyzed via numerical engineering methods, the finite element (FE) method hav-
ing become the most popular tool for this purpose.
In some FE based studies of honeycombs it was possible to simulate actual ex-
perimental setups [19]. In most cases, however, unit cells were employed, which, in
combination with appropriate periodic boundary conditions, are capable of resolv-
ing any periodic deformation mode and deformation mechanism of the corre-
sponding infinite cellular arrangement. Some care, however, is required in the se-
lection of a cell for a given analysis. In order to provide reliable results the unit cell
ideally should be a proper representative volume element, the geometry of which
contains the full statistical information on the material's microgeometry [20]. For
modeling the inelastic behavior of cellular materials, the unit cell must be designed
to allow for very large local deformations, and self-contact between the cell walls
has to be provided for if the behavior in the densification range is to be explored.
Furthermore, the choice of the unit-cell geometry is often influenced by the neces-
sity of allowing for realistic deformation and buckling patterns, which in most
cases precludes the use of symmetry boundary conditions. In all cases, of course,
there is a requirement to keep the complexity of the unit cell, and thus the com-
putational costs, within acceptable bounds.
When using unit-cell descriptions it is important to keep in mind that the result-
ing models are periodic in all respects. As an example, inhomogeneous densifica-
tion of cellular materials as predicted by unit-cell models will always take place in
periodic patterns, the period of which is strongly influenced by the choice of the
size of the unit cell. Single crack tips (as opposed to periodic patterns of cracks)
cannot be handled by periodic microgeometries, and free surfaces are restricted
to layer-like geometries in which at least one direction is nonperiodic. For these
types of problem, embedded cell approaches are the methods of choice.
6.1.2.2 Classification of Microgeometries

This section represents an attempt at systematically classifying the most important
types of cellular microgeometries that have been or may be used for unit cells or in
the core regions of embedded cell models. The main classification criteria are the
dimensionality (2D versus 3D) and the microscale morphology (regular periodic,
perturbed periodic, random, and ªreal structureº arrangements) of the geometrical
models.
2D discrete microgeometry models are directly applicable to investigating the
mechanical behavior of honeycombs. Owing to their relative simplicity they have
also been used as tools for studying metallic foams in a qualitative way (in the fol-
lowing planar models are generally referred to as ªhoneycombsº). The ªbaselineº
regular morphology are periodic hexagonal honeycombs, which show in-plane elas-
tic isotropy and, in the context of 2D liquid foams, may be viewed as surfaces of
minimal surface energy. Because hexagonal honeycombs (albeit of reduced sym-
metry) are applied routinely as sandwich cores, cellular metals of this type have
been closely studied both experimentally and analytically [21]. Alternative regular
microgeometries for 2D cellular materials were also discussed in the literature
[11,12].
In contrast to the regular microgeometry of idealized models, real cellular metals
are subject to morphological defects that tend to lead to softer and weaker mechan-
ical responses. Periodic 2D models have been an important means for studying the
influence of imperfections such as curved or corrugated cell walls [16,17,22], per-
turbed cell shapes [23], and non-uniform material distribution between cell walls
and cell vertices (Plateau borders) [24] as sketched in Fig. 6.1-1. Within such hon-
eycomb models local defects can be introduced into cell walls [17,25,26], individual
cell walls can be eliminated [26,27], vertices and the adjoining cell walls can be re-
moved to leave large voids in the honeycomb structure [26,28], and cells or groups
of cells can be filled with solid material to assess the influence of nodal inclusions
[26], Fig. 6.1-2. In addition, periodic arrangements of large and small cells [29] may
be used to study influences of relative cell sizes.
Random honeycomb arrangements can be generated on the basis of Voronoi tes-
selations [17,25,27,30], which may be interpreted in terms of geometries that
emerge when the fronts of bubbles growing with the same linear rate from ran-
Figure 6.1-1. Schematic representation of

imperfections at the level of individual cell
walls: left) Plateau borders, center) corru-
gated cell walls, right) curved cell walls.
Figure 6.1-2. Schematic representation of a

honeycomb containing imperfections in the
form of a large hole, cells filled with mate-
rial, fractured cell walls, and geometrical
perturbations (from left to right).
domly distributed nuclei meet. A different type of random microgeometry are

Johnson±Mehl cells [31], a tessellation that corresponds to cases where new cell nu-
clei are added while others are in the process of growing.
Finally, 2D microgeometries may be based on micrographical sections of metallic
foams [32]. Such ªreal structureº unit cells typically require some manipulation in
the boundary regions in order to achieve periodicity.
Planar models obviously can emulate the behavior of 3D microgeometries only
to a limited extent (note, for example, that in the 2D case there are no equivalents
of open-cell and closed-cell topologies). Accordingly, there has been growing re-
search interest in 3D models of cellular materials despite their high demands on
computational resources. In the case of open-cell foams, beam theory or beam fi-
nite elements provide comparatively inexpensive solutions for describing the me-
chanical behavior of the solid skeleton. Micromechanical studies of closed-cell
foams, however, require an appropriate modeling of the cell faces and, conse-
quently, shell theories must be invoked or shell elements must be used, which in-
crease the complexity of the simulations.
Perfectly regular 3D periodic microgeometries can be generated from space fill-
ing regular polyhedra, leading to models based on cubes, rhombic dodecahedra
and regular tetrakaidecahedra (or Kelvin structures). None of these arrangements
is elastically isotropic, but tetrakaidecahedral geometries have the advantage of giv-
ing face and edge counts per cell that are similar to the average values found in
actual polyhedral foams [33]. Accordingly, tetrakaidecahedral microgeometries
have been widely used for studying open-cell [34,35] and closed-cell [15,32]
foams, see Fig. 6.1-3. Somewhat more complex microgeometries were also pro-
posed in the form of regular arrangements of two populations of polyhedral
voids of different size [18].
Most 3D studies of the effects of imperfect periodic cellular arrangements have
been based on closed-cell tetrakaidecahedral microgeometries subjected to various
kinds of perturbations. Unit-cell models of varying levels of complexity were used
to study the effects of plateau borders [24], of curved or corrugated cell walls
[16,22], of perturbed cell shapes [23,32], and of cell walls that have randomly as-
signed thicknesses [36].
Figure 6.1-3. Regular tetrakaidecahedral model for a

closed-cell foam [32].
In analogy to 2D models, Voronoi diagrams may be used to generate irregular

3D microgeometries for open-cell [37,38] and closed-cell foams [23,39,40]. Because
they typically require large numbers of shell elements, investigations of this type
tend to be computationally expensive.
3D real structure models of cellular materials can be generated on the basis of
microtomographical data, each voxel being typically mapped to a finite element.
Analyses of this kind were pioneered by groups studying the mechanical behavior
of cancellous bone [41,42], which may be viewed as an open-cell foam, and were
recently also reported for closed-cell metallic foams [43]. This modeling strategy
typically gives rise to ragged surfaces of the solid skeleton of cellular materials,
which may lead to difficulties in analyses involving large deformations and large
strains.
6.1.2.3 Information Obtainable from Micromechanics

Before presenting specific results from micromechanical simulations in
Section 6.1.3, a short discussion is given concerning the information on the me-
chanical behavior of cellular metals that can be obtained from micromechanical
studies.
The most basic aspect of the mechanical material characteristics of inhomoge-
neous materials is their linear elastic behavior, which can be described in terms
of an overall elasticity tensor or appropriate effective moduli. Linear analysis re-
quires comparably little effort and, therefore, the linear elastic properties of a
wide range of cellular microgeometries are well researched.
Cellular metals typically have small elastic ranges, beyond which nonlinear be-
havior sets in owing to yielding, loss of stability or fracture at the microscale. In
contrast to bulk metals, the elastic range of metallic foams is also limited under
purely hydrostatic loading. Micromechanical analyses of such behavior may be car-
ried out in terms of material characterization by simulating, for example, uniaxial
or multiaxial tests to obtain homogenized stress versus strain curves. In terms of
the overall behavior the onset of nonlinear responses can generally be described via
appropriate surfaces in stress space, of which the classical von Mises yield surface
for plastic yielding of bulk metals is a well-known example. Analogous surfaces for
cellular metals can be obtained from micromechanical analyses by monitoring the
material response while following loading paths that are radial in macroscopic
stress space. Deviations from linear overall behavior are mainly due to local yield-
ing (typically at the transitions between cell walls and cell edges) or to local elastic
buckling of the cell walls or struts. The onset of nonlinear behavior is, in fact,
caused by yielding in the majority of cases, and microscale buckling plays an ap-
preciable role only for highly regular microgeometries, for which the above sur-
faces may be viewed as ªgeneralized yield surfacesº. If only the onset of plastic
yielding is considered in this context the superposition of elastic solutions is pos-
sible, and only one simulation per dimension of the macroscopic stress space is
required for determining the overall yield surface of macroscopically isotropic ma-
terials.
The above concept can also be used to follow the evolution of the yield surface as
a given macroscopic loading path is traced. To do this, the loads are applied incre-
mentally and complete unloading is carried out after each step followed by the eva-
luation of a new yield surface. This may again be done by superimposing elastic
solutions. It should be noted, however, that the application of linear superposition
analyses for such purposes is limited to the small deformation regime. In cellular
metals geometrical nonlinearities (caused, for instance, by marked bending of cell
walls) may give rise to noticeably nonlinear overall responses even before yielding
sets in.
As a foam is subjected to increasing compressive stresses a load will be reached
at which the first collapse of a cell occurs. In metallic foams the initial collapse of a
cell under these conditions tends to be followed by the growth of the collapsed re-
gion, which typically takes place at stress levels that show only limited variation,
giving rise to a so-called plateau region in the overall stress versus strain behavior.
When a considerable percentage of cells have been ªconsumedº the foam densifies,
the plateau region ends, and a much stiffer response sets in. The accumulation of
considerable compressive strains at nearly constant stresses is of particular interest
for cellular metals that are to be employed for impact energy absorption. Collapse
stresses under uniaxial loading have been studied for many cellular morphologies
and for various imperfections. For general load cases the initial collapse stress
states form an envelope surrounding the initial yield surface, which may be called
a collapse surface. For the point-wise evaluation of collapse surfaces fully nonlinear
analyses along individual load paths are required.
Relatively little work has been reported on the modeling of fracture of metallic
foams, which is the dominant failure mechanism under macroscopic tensile load-
ing. Embedded cell models provide a flexible approach for studying cracks in ide-
alized cellular metals, see Section 6.1.3.8.
6.1.3
Selected Results of Micromechanical Simulations
In this section, a number of selected results on the mechanical behavior of metallic

foams obtained by micromechanical modeling and simulation are discussed,
which cover a wide field of subjects, from the influence of microgeometrical imper-
fections to the evaluation of collapse surfaces in macroscopic stress space.
6.1.3.1 Influence of Material Distribution in the Cell Walls

The production of metallic foams typically involves the solidification of a molten
precursor material into an appropriate shape. In the liquid state a redistribution
of material driven by surface tension results in the formation of concave transition
regions between the cell walls, which are known as Plateau borders and give rise to
non-uniform cell wall thicknesses, see Fig. 6.1-4. This phenomenon may influence
the mechanical behavior of the foam on both the micro- and the macroscales and
has, accordingly, been the subject of considerable research interest.
Figure 6.1-4. Section through aluminum

foam produced by the powder metallurgical
route (courtesy Institute of Materials Science
and Testing, Vienna University of Technology).
Simone and Gibson [24] developed unit cells for hexagonal honeycombs and tet-
rakaidecahedral foams in which Plateau borders were modeled by appropriate
curved and parallel regions. Their results indicate that the distribution of material
in the cell walls has little influence on the Young's modulus and only a moderate
effect on the uniaxial yield strength of closed-cell metallic foams. A different ap-
proach was followed by Chen et al. [17], who studied honeycombs with wall thick-
nesses that increase linearly from the middle of the walls to the vertices. This sim-
plifies the mathematics of the problem sufficiently for analytical solutions for the
overall yield surfaces of the models to be obtained.
Investigations of the deformation patterns of honeycombs in the large strain re-
gime, which determine the materials' energy absorption capacity, can also be car-
ried out with unit-cell models. In such simulations the initiation of densification
can be identified from the occurrence of self-contact of the surfaces of the collapsed
voids, the modeling of which requires special provisions in terms of the unit-cell
geometries. Unit cells with all boundaries running within the cell walls have
been found to answer well, see Fig. 6.1-5, which shows symmetric and periodic de-
formation modes before the onset of densification. The main advantage of unit
cells of the above type is that all free surfaces of the voids face inwards, which al-
lows a straightforward use of contact algorithms. Obviously, such models can also
be adapted for studying the dependence of the deformation patterns on prescribed
distributions of mass between cell walls and vertices.
Drainage of the material into Plateau borders reduces the cell wall thickness,
which is the governing parameter for the strength and the stiffness of the arrange-
Figure 6.1-5. Predicted deformation modes of hexagonal honeycombs with Plateau borders: left)
symmetric, right) periodic.
ment. This is especially important in the case of foams of very low density, in
which the cell wall thicknesses approach zero. At higher densities, however, the re-
distribution of material towards the cell vertices and, therefore, into regions sub-
jected to higher bending moments can be beneficial for the overall properties of
foams [17,24].
As more material is accumulated in the vertices, the plastic hinges under bend-
ing tend to form closer to the mid-region of the cell edges, so that the size of the
nearly rigid regions around the vertex is increased. Although such a zone can ro-
tate during compression it still takes up more space if the distance between the
hinges and the vertex centers is increased. Together with changes in the folding
kinematics this causes the cell walls to contact each other at lower strains than
is the case for microgeometries with less pronounced plateau borders. This effect
dominates the energy absorption potential as shown in Fig. 6.1-6, Fig. 6.1-7, and
Fig. 6.1-8. Even in cases where the collapse and pre-contact loads are increased
by the material redistribution, the decrease of the usable deformation length re-
duces the amount of energy that can be dissipated by plastic deformation before
the densification regime is reached.
Microscale fluctuations of the material distribution in cellular materials may, on
the one hand, be due to thickness variations within individual cell walls as dis-
cussed above in connection with Plateau borders or, on the other hand, be caused
by the individual cell walls in a microgeometry having different thicknesses. The
latter problem was addressed by Grenestedt and Bassinet [36], who developed a
3D tetrakaidecahedral model of a closed-cell foam that contains a total of 112
Figure 6.1-6. Predicted nominal strain at first cell wall contact as a function of the material
distribution between walls and cell vertices (Plateau borders) for three different apparent
densities.
Figure 6.1-7. Predicted normalized collapse stress as a function of the material distribution
between cell walls and cell vertices (Plateau borders) for three different apparent densities.
Figure 6.1-8. Predicted normalized absorbed energy (up to first cell wall contact) as a function of
the material distribution between cell walls and cell vertices (Plateau borders) for three different
apparent densities.
cell walls, the thicknesses of which could be assigned individually. The results
show that the stiffness of regular closed-cell arrangements is rather insensitive
to the presence of cell walls of different thickness. This was explained by noting
that such microgeometries deform primarily by cell wall stretching, which is
less sensitive to wall thickness effects than bending modes induced, for example,
by corrugated cell walls.
6.1.3.2 Influence of Wavy and Curved Cell Walls

In addition to inhomogeneous thickness distributions, the geometries of cell walls
may also be perturbed by deviations from simple linear (honeycombs) or planar
(3D foams) connections between the vertices. Such simple strut or wall geome-
tries, in fact, are good approximations for low-density organic foams. In metallic
foams, however, curved and corrugated (ªwavyº) cell walls typically are present, see
Fig. 6.1-4.
Grenestedt [16] studied the influence of wavy imperfections of the walls on the
elastic stiffness of cellular solids and found that the bulk modulus decreases by
some 50 % when corrugations with an amplitude of twice the thickness of the
cell walls are present. Simone and Gibson [22] documented the detrimental effect
of curved and corrugated cell walls on the stiffness and strength of regular honey-
combs and perfectly tetrakaidecahedral unit-cell models. The Young's modulus was
shown to be more adversely affected by these imperfections than the collapse
stress. Chen et al. [17] gave analytical expressions for the yield surface of regular
honeycombs with wavy cell walls. They found that corrugations significantly re-
duce the hydrostatic yield strength of honeycombs, whereas the deviatoric yield
stress is hardly affected. Evidently, cell-wall bending is activated in addition to
membrane deformations under overall hydrostatic loading.
a) b)
c) d)
Figure 6.1-9. Predicted yield surfaces for regular and irregular honeycombs with different types and
degrees of imperfections [29]. Dashed and dash-dotted lines denote elastic buckling stress states.
The influence of corrugated and curved cell walls on the onset of nonlinear be-
havior was studied by Daxner et al. [29] by evaluating initial yield surfaces.
Figure 6.1-9 shows such curves that represent overall stress states that give rise
to the local onset of yielding or to local elastic buckling in regular and perturbed
honeycombs that have curved and wavy cell walls of varying curvature. In perfect
regular honeycombs in-plane hydrostatic compression leads to pure membrane
stresses in the cell walls, which give rise to an elongated overall initial yield surface
in the s x s y stress plane. For low-density, perfect honeycombs, this yield surface is
truncated by the failure surface for elastic cell wall buckling, see Fig. 6.1-9a, which
shows a kink owing to a change of the buckling mode. This type of behavior is
maintained in the presence of small corrugations, but when the amplitudes of wig-
gles exceed about 5 % of the wall thickness yielding alone determines the onset of
nonlinearity.
The influence of wiggles and curved cell walls on the yield surfaces of honey-
comb microgeometries that are irregular from the outset is much less pronounced,
see Fig. 6-1.9b±d. Perturbing the geometry by randomly displacing cell vertices
causes the yield surface to shrink even more than the presence of severe wiggles,
see Section 6.1.3.3. Corrugated or curved cell walls generally have little effect on
the uniaxial yield strength, but severe wiggles can reduce the in-plane hydrostatic
yield stress by some 35 %. Curved cell walls typically give rise to similar effects as
wiggles, see Fig. 6.1-9c, but cell walls incorporating only slight curvature can lead
to minor increases of the in-plane hydrostatic yield stress.
6.1.3.3 Influence of Irregular Vertex Positions

Generic models for cellular materials are commonly based on regular periodic mi-
crogeometries, the simplest 2D geometry being hexagonal honeycombs. A compar-
ison between such a honeycomb structure and a cross section of an actual metallic
foam as shown in Fig. 6.1-4 makes clear that the latter is far from being a regular
structure. A standard assumption in modeling efforts is that different types of geo-
metrical imperfections may be studied in isolation. Following this strategy, the
present section concentrates on the effects of overall cell geometry while neglecting
the cell wall imperfections discussed above.
It can be shown that the elastic stiffness of a honeycomb model is not very sen-
sitive to irregularities and perturbations of the vertex positions. In fact, upon com-
paring the elastic stiffnesses of hexagonal and Voronoi honeycombs Silva et al. [30]
reported that the overall elastic moduli are some 5±10 % higher in the latter case.
Zhu et al. [44] defined a measure for the irregularity of Voronoi honeycombs which
they correlated to the predicted elastic properties. They found the effective Young's
and shear moduli of periodic Voronoi honeycombs to increase with growing irre-
gularity, whereas the effective bulk modulus decreases. They applied the same me-
thodology to 3D open-cell foams and obtained analogous results [38]. Grenestedt
and Tanaka [23] examined the elastic behavior of 3D closed-cell Voronoi unit
cells, finding a decrease in the bulk modulus of 5±10 % compared to regular tetra-
kaidecahedra.
The uniaxial yield stress of honeycomb models can be reduced by some 40 % by

perturbing the positions of the cell vertices, see Fig. 6.1-9. Silva and Gibson [27]
reported a similar decrease in the yield strength of Voronoi honeycombs compared
to hexagonal honeycombs, but a less pronounced reduction for the uniaxial yield
stress was obtained by Chen et al. [17], who attribute this difference to the use
of periodic rather than mixed boundary conditions for their unit cells. In addition
to Voronoi honeycombs, they studied configurations obtained from hexagonal ar-
rangements by randomly shifting the vertices and obtained results on the uniaxial
compressive yield stress that are in good agreement with Fig. 6.1-9. Both planar
Voronoi microgeometries and perturbed hexagonal arrangements give rise to re-
duced yield stresses under in-plane hydrostatic loading as compared to regular
honeycombs because the deformation mode tends to be cell wall bending rather
than membrane compression [17].
As mentioned above, randomly shifting the vertex positions of regular honey-
combs typically leads to more pronounced reductions of the overall yield limits
than does the introduction of corrugations of the cell walls. In addition, the collapse
stress, which is closely related to the macroscopic plateau stress, is decreased. Again,
the most affected loading condition is in-plane hydrostatic loading. Loss of stability
due to elastic buckling plays a significant role only in extreme cases of near-perfect
microgeometries and very thin cell walls. For microgeometries with significant irre-
gularities of the vertex positions elastic buckling would require uniaxial applied
stresses that are about one magnitude higher than the predicted yield stress.
Clearly, irregular microgeometries are required for obtaining realistic results
from simulations of the mechanical behavior of metallic foams. This is especially
true when multiaxial loads or deformations are to be studied. In the presence of
irregular vertex positions other imperfections such as cell wall corrugations typi-
cally are relegated to secondary roles.
6.1.3.4 Microgeometries Containing Cells of Different Sizes

In metallic foams the sizes of the cells are typically far from uniform, see Fig.
6.1-4, and often regions containing small cells as well as clusters of very large cells
can be identified. In order to gain some understanding of the interaction between
larger and smaller cells, simulations of 2D geometries consisting of generic peri-
odic arrangements of cells of two different sizes, see Fig. 6.1-10, were performed
[29]. They comprise a regular hexagonal honeycomb (HC), serving as reference
configuration, as well as clusters of small cells surrounded by large cells (SBS)
and isolated large cells surrounded by small cells (BSB). The main focus was
on large strain deformation owing to its importance in energy absorption by cel-
lular metals.
When observing stress strain relationships predicted for uniaxial compression,
two qualitatively different deformation and collapse mechanisms become apparent.
If shear localization occurs, the microgeometry can deform at nearly constant ap-
plied stress through cell wall bending, and, after collapse of the shear bands, by
sliding within the collapsed layer, giving rise to comparably smooth stress strain
Figure 6.1-10. Honeycomb geometries for studying the interaction of cells of different sizes [29]:
Regular hexagonal arrangement (HC, left), a cluster of small cells surrounded by big cells (SBS,
center), and big cells surrounded by small cells (BSB, right).
curves. If, in contrast, cell wall buckling and extensive layer-wise collapse perpen-
dicular to the loading direction are the dominant mechanisms, the stress strain
relationship shows marked oscillations on account of the sequential collapse of
ªcell rowsº that are incapable of accommodating subsequent deformation by slid-
ing. Figure 6.1-11 shows deformation modes of the above types for the case of an
imperfect hexagonal honeycomb, with the corresponding stress versus strain
curves being given in Fig. 6.1-12.
In addition to these general observations on collapse modes some specific rules
for the interaction of small and big cells can be identified. Comparing the deforma-
tion patterns developed by arrangements SBS and BSB, see Fig. 6.1-13, one can
observe that in the latter the clusters of small cells do not contribute very much
to the overall deformation and may become obstacles in the densification regime.
The framework of small cells surrounding the large ones in model SBS, however,
provides more uniform stiffness and deforms more evenly. In general, clusters of
small cells tend to be detrimental to the deformation and energy absorption poten-
tial of cellular metals as they typically deform less than their bigger neighbors, the
longer struts of which confer lower bending stiffness and strength.
Figure 6.1-11. Undeformed and deformed loads applied in the horizontal and vertical
configurations (as predicted by unit cells with directions. Note the occurrence of shear loca-
periodic boundary conditions) of an imperfect lization (left) and layer-wise collapse (right),
hexagonal honeycomb subjected to uniaxial respectively.
Figure 6.1-12.Predicted overall stress strain relationships of an imperfect hexagonal honey-

comb [29] corresponding to the deformation modes shown in Fig. 6.1-11.
Figure 6.1-13. Deformed configurations predicted for imperfect models describing small cells
surrounding big cells (BSB, left) and big cells surrounding small cells (SBS, right) based on the
configurations presented in Fig. 6.1-10.
These observations lead to the conclusion that whenever smooth stress versus
strain relationships are desired, the cell size distribution should be as uniform
as possible and the cells should be small compared to the sample or structure to
either prevent localization or to limit its detrimental effects.
6.1.3.5 Influence of Holes and Solid-Filled Cells

A further group of imperfections of cellular geometries may be generated by elim-
inating some vertices as well as the cell walls connected to them, which leads to
holes that are larger than typical cell sizes, or by filling selected cells with solid ma-
terial, see Fig. 6.1-2.
Experiments by Prakash et al. [28] have shown that the filling of some cells leads
to local strengthening of honeycombs, increases the elastic modulus and the de-
gree of strain hardening, but reduces the densification strain. The opposite effect
is caused by removing individual cells or whole cell clusters as demonstrated by
Guo and Gibson [45] in a FE study on intact and damaged honeycombs. They re-
ported on correlations between the undamaged cross-sectional area perpendicular
to the loading direction and the elastic buckling load as well as the plastic collapse
strength. The interaction between separate defects of this type was found to have a
range of about ten cell diameters.
Chen et al. [26] investigated the influence of solid inclusions as well as holes on
the stiffness and on the uniaxial and in-plane hydrostatic yield strengths of perfect
honeycombs and of 2D arrangements in which some 5 % of the cell walls were ran-
domly fractured. Solid inclusions were shown to lead to minor increases of the
elastic stiffness and to have a negligible effect on the uniaxial and in-plane hydro-
static yield strength of both the otherwise perfect and the fractured honeycombs.
Because the solid inclusions introduce additional mass, however, the specific prop-
erties were negatively affected. Large holes were found to induce cell wall bending
in otherwise perfect honeycombs, leading to significant reductions in the bulk
modulus and the hydrostatic yield strength. For prefractured honeycombs the de-
crease in overall stiffness due to holes could be estimated from the reduced overall
relative density of the honeycomb.
6.1.3.6 Influence of Fractured or Removed Cell Walls

Cell walls that are damaged by fracture or that are bodily removed may weaken cel-
lular materials to a considerable extent. Having introduced such defects into regu-
lar hexagonal and Voronoi honeycombs, Silva and Gibson [27] reported that the re-
duction of overall mechanical properties due to the removal of cell walls tends to be
2 3 times greater than that caused by an equivalent (in terms of density) uniform
reduction of the cell wall thickness. A typical Voronoi honeycomb was predicted to
be, on average, 30 35 % weaker than a periodic hexagonal honeycomb of the same
density. The same degree of weakening can be obtained by removing 5 % of the cell
walls.
Experiments on honeycombs [28] showed that removal of cell walls triggers loca-
lized deformation because the weakened cells collapse first. Whether or not cells
with defects interact to form a common deformation band depends on their dis-
tance and their position relative to the symmetry axes and the loading direction.
Albuquerque et al. [46] performed similar experiments on Kevlar honeycombs.
In addition to confirming the results of Silva and Gibson [27] and Prakash et al.
[28] they found that the compressive behavior is hardly affected by having the de-
fects uniformly dispersed or concentrated in a region, provided the concentration
of defects is low.
Upon removing 10 % of the cell walls in Voronoi honeycomb models, Silva and
Gibson [27] obtained a reduction of the compressive strength by some 40 % and
found that mechanical stiffness and strength tend to zero when 35 % of the cell
walls are removed. To study fatigue accumulation in cancellous bone Schaffner et

al. [25] removed struts in an open-cell model once the lengths of fatigue micro-
cracks in them (assumed to grow according to a Paris law) exceeded a user-speci-
fied limit. They found that Voronoi honeycombs are more sensitive to fatigue dam-
age than are regular hexagonal geometries. After comparing several types of mor-
phological imperfections in honeycombs Chen et al. [17] identified the removal of
cell walls as the most critical of them in terms of reduction of the yield strength.
6.1.3.7 Yield and Collapse Surfaces

In Section 6.1.2 the evaluation of yield surfaces, ªgeneralized yield surfacesº and
collapse surfaces from unit-cell analyses was discussed. For a given cellular
metal, the availability of such data allows multiaxial stress states to be assessed
for the onset of nonlinear behavior and for the commencement of cell collapse,
which defines the start of the plateau region in the overall stress versus strain be-
havior.
Initial yield surfaces for hexagonal honeycombs with corrugated or curved cell
walls were mentioned in Section 6.1.3.2. Here, some aspects of yield and col-
lapse surfaces will be discussed for more realistic microgeometries. To study the
behavior of irregular honeycombs, a unit cell was generated on the basis of a
Figure 6.1-14.Unit cell for a periodic ªreal structureº honeycomb model [32] adapted from a
micrograph [16]. Note the wide range of cell sizes, cell shapes, and cell wall imperfections.
section through a sample of closed-cell aluminum foam given by Grenestedt [16],

suitable adaptations being introduced to support periodic boundary conditions, see
Fig. 6.1-14.
The predicted collapse surfaces typically surround the initial yield surfaces, and
for irregular honeycombs the shapes of the two surfaces tend to be somewhat
similar, see Fig. 6.1-15. It may be noted that for uniaxial overall stress states,
where the dominant local deformation mechanism is bending, the collapse stress
exceeds the initial yield stress by a factor of about 2.4, which is far in excess of the
value of 1.5 given by simplified beam analysis (assuming ideally plastic beams of
rectangular cross section and neglecting stress redistribution).
The collapse surface presented in Fig. 6.1-15 is not given for all possible loading
paths, only the purely compressive regime (3rd quadrant) and those parts of the
shear regimes (2nd and 4th quadrants), for which the larger principal stress is
compressive, being covered. These restrictions were necessary because the pre-
dicted stress versus strain responses do not allow the extraction of plateau stresses
once the tensile principal stresses exceed the compressive ones. To obtain peak
stresses and a plateau region from such tension dominated loading paths addi-
tional failure criteria, for instance cell wall fracture, would have to be introduced.
By definition an initial yield surface allows the presence of plastic yielding to be
assessed on the basis of linear analysis. This obviously is an important step in the
development of a macroscopic elastoplastic constitutive model. It is not, however,
sufficient for carrying out nonlinear studies such as crushing or crash analyses, for
which detailed information on the evolution of the yield surface under loading as
provided by a hardening law is indispensable and strain rate effects may have to be
accounted for. One of the hardening laws used in connection with metal foams is
Figure 6.1-15. Overall yield and collapse surfaces [32] predicted for the real structure honeycomb
model presented in Fig. 6.1-14. The collapse surface is not evaluated for load cases for which a
distinct maximum could not be found in the overall stress versus strain relationship.
isotropic hardening [47]. It assumes that the origin and the shape of the yield sur-
face stay the same, but its size increases driven by the accumulated volumetric
plastic strains. In the Crushable Foam model [48] mixed hardening is assumed,
where the yield surface keeps the same shape with increasing size (isotropic hard-
ening), but shifts its origin (kinematic hardening) so that the hydrostatic tensile
yield stress remains constant.
Micromechanical analyses based on the real structure geometry shown in
Fig. 6.1-14 paint a somewhat more complex picture. Figure 6.1-16 displays predic-
tions for the evolution of the overall yield surface corresponding to progressive uni-
axial compressive loading in the x-direction. At several points during the loading
sequence the analysis was interrupted, the model was unloaded and then subjected
to linear superposition analyses to estimate the onset of nonlinear behavior. This
way the yield surface corresponding to the actual state of hardening can be gener-
ated. The initial hardening behavior was predicted to be essentially kinematic, the
yield surface being shifted in the direction of the applied stress without major
changes in shape or size. At elevated compressive stresses, however, the evolving
yield surface changes its shape and contracts as well. Because Fig. 6.1-16 is
based on the superposition of solutions from linear analyses that cannot account
for the geometrical nonlinearities, which may be expected to play an increasing
role as compressive collapse is approached, its reliability in the latter regime is
not fully clear. Accordingly, the results can be interpreted as clearly favoring kine-
Figure 6.1-16. Evolution of the overall yield surface under uniaxial compressive loading in
x-direction predicted for the real structure honeycomb model presented in Fig. 6.1-14.
matic hardening models for low inelastic strains and hinting at possible changes in
the hardening behavior in the high strain regime.
As mentioned before, planar models are directly applicable to honeycombs, but
are not necessarily directly transferable to real foams, which require 3D simula-
tions for quantitative predictions. Denzer developed 3D closed-cell microgeome-
tries [32] based on a body-centered cubic arrangement of tetrakaidecahedral cells,
using symmetry boundary conditions to limit the computational costs. Initial over-
all yield surfaces under compressive loading were evaluated for configurations with
perfect cell geometry as well as for models with statistically perturbed vertex posi-
tions, see Fig. 6.1-3 and Fig. 6.1-17. The general shapes of these yield surfaces in
principal stress space are ellipsoid-like and their major axes are aligned with the
hydrostatic axis, see Fig. 6.1-18. This can be explained by the stress states induced
in the cell walls, which are mainly membrane-like for hydrostatic loads (even
though regular tetrakaidecahedra are not elastically isotropic) whereas uniaxial
loads tend to give rise to bending close to the vertices. These effects are less pro-
nounced for the imperfect models, where the ªaspect ratioº of the yield surface
tends to be lower. This behavior is qualitatively similar to predictions obtained
from honeycomb models, but the yield surfaces for the perfect tetrakaidecahedra
are less elongated along the hydrostatic axis, see Fig. 6.1-9.
By projecting selected points of the yield surface shown in Fig. 6.1-18 (left)
onto the equivalent stress versus mean stress plane a standard representation is
obtained, see Fig. 6.1-19, that can be readily compared with other results. The pro-
jected points can be seen to be scattered around an ellipse the major axis of which
coincides with the mean stress axis, the hydrostatic yield stress being about 2.5
Figure 6.1-17. Irregular tetrakaidecahedral model for a closed-cell foam [32]. As can be seen from
the planar section, perturbations were introduced only in the core of the model to allow for the
use of symmetry boundary conditions.
Figure 6.1-18. Yield surfaces predicted for regular (Fig. 6.1-3) and irregular (Fig. 6.1-17)
tetrakaidecahedral closed-cell foams [32].
Figure 6.1-19. Projection of the yield surface shown in Fig. 6.1-18 (left) onto the von Mises
equivalent stress versus mean stress surface.
times larger than the uniaxial yield stress. It may be noted that many material laws
for foams use yield surfaces that give rise to ellipses when projected to the equiva-
lent stress versus mean stress plane, among them the model of Deshpande and
Fleck [47].
6.1.3.8 Fracture Simulations for Metallic Foams

Unit-cell based micromechanical methods give rise to periodic deformation, stress
and strain fields and, as a consequence, periodic patterns of damage and cracks.
Although this may be acceptable for distributed damage, a useful alternative for
studying small samples consists in employing microgeometries that essentially

model the whole specimen. For example, Schaffner et al. [25] used a Voronoi hon-
eycomb to study fatigue damage of cancellous bone.
Crack tips in discrete cellular microstructures were studied by Gibson and Ashby
[1], who gave expressions for the fracture toughness of brittle honeycombs by con-
sidering failure of the first unbroken cell wall in the path of an advancing crack in
simple planar and 3D models. An approach in which hexagonal honeycombs are
homogenized as a micropolar elastic material, for which the asymptotic crack tip
fields are computed and then used to estimate the displacements and rotations
of the cell walls surrounding the crack tip, was developed by Chen et al. [49]. Prob-
ably the most flexible approach available at present for studying crack tips in cel-
lular materials, however, is the use of embedding techniques. Such a strategy
was followed by Ableidinger [35] for studying the influence of a number of micro-
geometrical and material parameters on the macroscopic fracture mechanical be-
havior of a compact tension (CT) specimen made of an open-cell aluminum
foam subjected to monotonic loading [50].
The region surrounding the crack tip was modeled with a fully resolved, but
highly generic, 3D tetrakaidecahedral open-cell microgeometry discretized with
beam elements. This kernel was embedded into the remainder of the CT speci-
men, see Fig. 6.1-20 (left), which was treated as a homogeneous structure with
an effective material behavior obtained from unit-cell analyses of the same tetrakai-
decahedral microgeometry. In the out-of-plane direction the core consisted of a sin-
gle layer of tetrakaidecahedra, the embedding region having the same thickness,
and symmetry boundary conditions were applied at the top and bottom planes,
so that the model actually corresponds to an inner layer (plane strain conditions)
of the specimen. Tensile loads were applied at the positions of the loading fixtures
of the CT specimen, so that the crack proceeded in fracture mode I on the macro-
scale. Crack propagation at the micro level as resolved in the kernel region was as-
sumed to be caused by sequential ductile failure of the most highly loaded struts,
see Fig. 6.1-20 (right).
From the predicted force versus displacement diagrams of the CT specimens, see
Fig. 6.1-21, effective KI versus Da (crack resistance) curves as shown in Fig. 6.1-22
Embedded cell model of an open-cell foam using a tetrakaidecahedral core and a

Figure 6.1-20.
homogenized outer region (left) and detail of the core around the crack tip (right), from [35].
Note that symmetry with respect to the crack plane is assumed.
Figure 6.1-21. Macroscopic force versus displacement curves of CT specimens made of open-cell
aluminum foams predicted by embedded cell models, see Fig. 6.1-20, for different effective
densities and different aluminum alloys [35].
Figure 6.1-22. Macroscopic crack resistance curves of CT specimens made of open-cell alumi-
num foams predicted by embedded cell models, see Fig. 6.1-20, for different effective densities
and different aluminum alloys [35].
were obtained. For this purpose, the macroscopic stress intensity factor KI was eval-
uated at each maximum of the force versus displacement curves according to stan-
dard expressions for CT specimens, the crack length increments Da being dictated
by the selected microgeometry. Both Fig. 6.1-21 and Fig. 6.1-22 show results for
three values of the apparent density of the foam and for two aluminum alloys dif-
fering in ductility.
Within such a modeling approach parameters such as the mass density of the
foam (which determines the void volume fraction), the size of the cells (which in-
fluences the fracture behavior in terms of an internal length scale), the yield
strength, and the strain to fracture of the bulk material (which govern the ductile
damage and failure of the struts) can be varied easily in order to study their respec-
tive influences on the fracture behavior of open-cell metallic foams.
6.1.4
Modeling of Mesoscopic Density Inhomogeneities
Foams are not only heterogeneous in terms of consisting of solid regions and
voids, but in many cases also show an inhomogeneous distribution of their appar-
ent density, mean cell size and other geometrical parameters. These variations cor-
respond to length scales larger than that of the voids (the microscale) and smaller
than that of the sample or component (the macroscale) and are, accordingly,
termed mesoscopic inhomogeneities in the following. Fig. 6.1-23 (left) shows
such density variations in a cross section of a foam sample.
Shim et al. [51] introduced uniaxial lumped mass±spring mesoscopic models for
the simulation of the dynamic uniaxial crushing of foam samples with uniform ap-
parent density. Systems of elastic-plastic springs were employed by Gradinger and
Rammerstorfer [52] for studying the compressive force displacement behavior of
crush specimens with deterministically distributed meso-inhomogeneities. These
two approaches were combined by Daxner et al. [53] to assess the influence of in-
homogeneous density distributions on the impact response of foam plates with ge-
neric variations in their apparent density, the interaction of density inhomogene-
Figure 6.1-23. Micrographical cross-section of an aluminum foam with three regions of differing
apparent density (left) and sketch of the corresponding mesoscopic simulation model (right) [53].
ities in directions normal to the orientation of the load being also accounted for,
see Fig. 6.1-23 (right).
The influence of mesoscale inhomogeneities in the apparent density is of partic-
ular interest when the metal foam is applied as protective padding. Strong mesoin-
homogeneities reduce the energy absorption efficiency, because they lead to strain
localization resulting in a decrease of the initial plateau stress and a less pro-
nounced plateau region [52,53], as can be seen in Fig. 6.1-24. This holds true for
density variations occurring in the loading direction and perpendicularly to it
[53]. Accordingly, for such applications it is desirable to tune foam production tech-
nologies for achieving uniform density distributions.
When a metallic foam is treated as a homogenized material, inertia effects [54]
lead to wave propagation phenomena that are controlled by the interaction of foam
density, layer thickness, impact energy and impact velocity [53]. At moderate im-
pact speeds those inertia effects give rise to an impact resistance that only slightly
exceeds the static yield strength, see Fig. 6.1-25. Nevertheless, the dynamic effects
of wave reflection and superposition are noticeable under these conditions, as is
evident in Fig. 6.1-26, where the stacking order of foam layers with different den-
sity is shown to significantly influence the time history of impact stresses. The re-
sults obtained with simulations that treat cellular metals as homogenized materi-
als, however, will have to be compared to dynamic tests to assure that no artificial
inertia effects are introduced by the homogenization approach. Possibly, the mass
density used for the homogenized material may have to be adjusted to closely fit
the experimental results for dynamic events.
Figure 6.1-24. Stress strain diagrams pre- a meso-homogeneous plate, the open circles
dicted for foam plates that consist of layers of denote a two-layer structure and all other cases
different density while maintaining the same are three-layered arrangements [53].
overall density. The solid line corresponds to
Figure 6.1-25. Predicted influence of impact velocity and foam density on the initial stress
response s dyn normalized by the static collapse stress s 0,stat [53].
Figure 6.1-26. Predicted contact stress between an impacting mass and a foam cube composed
of two equally thick layers of different densities [53]. The two possible stacking orders of the layers
were simulated.
Mesoscopic mass±spring models as described above are restricted to specific load

cases and, in general, cannot handle arbitrary multiaxial loading paths. More gen-
eralized mesoscopic simulations can be carried out by using the FE method with
standard volume elements, see Section 6.2.
6.1.5
Macroscopic Modeling and Simulation
From the perspective of modeling samples and structural components made of cel-
lular metals, discretization down to the level of individual cells is not only infeasi-
ble with current computer and software technology, but also appears to be imprac-
ticable with regard to the required modeling effort. As a consequence, macroscopic
constitutive descriptions of metallic foams have been developed. The main condi-
tion for their use is that the overall dimensions of the sample or structure must be
significantly larger than the dimensions of single foam cells.
Structural analysis of cellular metals in the elastic range is relatively straightfor-
ward in that only the homogenized elasticity tensor must be known. For isotropic
metallic foams, accordingly, two elastic moduli have to be calibrated; more material
parameters are required for foams of lower symmetry. Experiments, however, have
shown that metallic foams may not display a distinct elastic regime under com-
pressive loading, but rather begin to deform irreversibly at stresses that are low
compared to their collapse or plateau stress. Thus, it is not fully clear to what extent
cellular metals can be treated as linear elastic materials in structural analyses.
Practically relevant constitutive models for metallic foams, accordingly, must be
capable of accounting for nonlinear macroscopic material responses. If, for exam-
ple, the FE method is to be used in the engineering analysis of components made
of cellular metals, constitutive descriptions are required at the integration point
level, which may have to be coded specifically [55].
Macroscopic yield surfaces appropriate for metallic foams typically can be repre-
sented by ellipses in the mean-effective stress plane. The yield surface proposed by
Deshpande and Fleck [47] is symmetrical with respect to compressive and tensile
stress states. Furthermore, it evolves by uniform scaling in all directions (isotropic
hardening) and uses an associated flow rule, which implies that the plastic Poisson
ratio is generally not equal to zero. In contrast, the Crushable Foam material model
[48] is characterized by a constant tensile hydrostatic yield stress, which causes the
origin of the yield surface to move along the hydrostatic axis as the yield surface
expands during hardening. The non-associated flow rule governing this material
model is formulated to always prevent a plastic Poisson effect. The accumulated
plastic volumetric strain is the internal variable driving the hardening process.
For additional macroscopic constitutive models proposed for metallic foams see
[56 58].
For the above material models additional parameters determining the shape of
the yield surface and the hardening behavior must be provided by the user, requir-
ing information on the mechanical behavior of the material under multiaxial load-
ing conditions. Since multiaxial experimental data [47,59] is scarce, the user very
often must rely on appropriate assumptions, which can be derived either from
microstructural FE simulations or via parameter identification techniques (that
is, by minimizing the discrepancy between simulations and experimental results).
Fortunately, applications that are dominated by hydrostatic loading conditions
are rare. To illustrate this point, in the next section a case study will be pre-
sented in which the uniaxial compressive stress strain relationship is the most
important material characteristic: the impact of a large sphere on an aluminum
foam pad.
6.1.5.1 Low Energy Impact on Thin Metallic Foam Paddings

At present paddings made of organic foams, which exhibit relatively low strength,
are widely used for impact protection. In some situations, however, the available
design space is too small to allow an organic foam layer of a depth sufficient for
providing adequate energy absorption. Metallic foams are candidate materials for
use in such cases due to their higher collapse strength.
In the literature impact studies may be found in which the performance of foam
specific material laws is compared to experimental data [60]. Such investigations
provide results in terms of overall quantities such as contact forces or accelerations.
The following discussion refers to simulations of a large rigid sphere impacting a
layer of aluminum foam simulated by the FE code ABAQUS/Explicit (Hibbitt,
Karlsson & Sorensen, Pawtucket, RI, 1998) and using the Crushable Foam
model [48]. Figure 6.1-27 shows an acceleration history predicted for this setup;
the severity of the impact studied can be judged by the impression left in the
foam pad by the sphere, see Fig. 6.1-28.
If the predicted evolution of the stress states in selected material points of the
foam target is followed during the impact event, this data can be used to assess
which level of complexity of the flow rule and the hardening law is required for
performing satisfactory simulations. Essentially, radial paths in stress space indi-
cate that simple isotropic constitutive laws suffice for describing the local material
state, whereas distinct changes in the direction of the stress paths signal that
phenomena such as anisotropic hardening will have to be considered. Soften-
Figure 6.1-27. Predicted acceleration versus time history of a typical impact event.
Figure 6.1-28. Schematic view of an impact event (left) and predicted deformed configuration
after the impact (right).
ing due to fracture of cell walls may also occur when stress states enter tensile re-
gimes.
The evolution of local stress states can be visualized in the mean versus equiva-
lent stress diagram used commonly for describing pressure dependent material
laws. Figure 6.1-29 shows such stress paths for three different material points.
These stress paths lie mostly between the axis of uniaxial compression (inclination
3:1) and the vertical axis that represents pure shear, indicating that these two types
of loading will dominate the local deformation of the foam. The `Axis, top' line in
Fig. 6.1-29 is associated with a point under the center of the impact mass. It is sur-
rounded by an ellipse, which is the projection of the Crushable Foam yield surface
corresponding to the highest value of the hardening parameter, namely the accu-
mulated volumetric plastic strain. Uniaxial stress states can be seen to prevail in
the impact event considered here, indicating that the simulation should not be
very sensitive to changes in the hydrostatic parameters of the material. This is com-
forting, since only limited information is available on the behavior of foams under
hydrostatic pressure. Furthermore, the principal stress axes maintain roughly the
same orientation throughout the impact event, so that the selected hardening
model does not play a major role. These findings narrow down the requirements
on improved material laws for crash simulations involving a comparatively thin
padding and a large impactor.
Evolution of Some Local Stress States During Impact
2.5 111111111111
000000000000
000000000000
111111111111
Initial yield surface 000000000000
111111111111
000000000000
111111111111
Axis, bottom
000000000000
111111111111
2.0
Axis, top
Top, Border
000000000000
111111111111
000000000000
111111111111
000000000000
111111111111
000000000000
111111111111
Effective Stress [MPa]
Uniaxial Compression
Hardened yield surface 000000000000
111111111111
000000000000
111111111111
1.5 000000000000
111111111111
000000000000
111111111111
000000000000
111111111111
000000000000
111111111111
000000000000
111111111111
000000000000
111111111111
1.0 000000000000
111111111111
000000000000
111111111111
000000000000
111111111111
000000000000
111111111111
000000000000
111111111111
00
11
000000000000
111111111111
00
11
000000000000
111111111111
0.5
00
11
0.0
-1.5 -1.0 -0.5 0.0 0.5 1.0 1.5
Negative Mean Stress; Pressure [MPa]

Figure 6.1-29. Stress states of selected material points visualized as paths in the von Mises
equivalent versus mean stress plane during a typical impact event.
6.1.5.2 Crushing of Foam-Filled Crash Elements

The energy absorption characteristics of tubular metal structures during axial
crushing may be improved by filling them with appropriate lightweight materials
such as honeycombs or foams. The main mechanisms providing improvements
(compared to an empty tube) are the compression of the filler material itself and
the activation of interaction effects between filler and tube. These interactions
lead to higher energy dissipation of the tubular members owing to changed buck-
ling modes and also increase the energy dissipation of the filler material owing to
multiaxial compression. Experiments in which aluminum foams were applied as
filler material reveal that particularly efficient crush elements may be designed
this way [61 63].
Concerning the numerical analysis of such energy absorption structures the FE
method may be applied, see [61,64] and Fig. 6.1-30. For preliminary design and for
parameter studies, however, simplified methods, which only consider the essential
variables, can also be utilized. Experiments show that the collapse process of thin-
walled structures is, in general, accompanied by the development of localized plas-
tic mechanisms, that is, the formation of a more or less complicated pattern of
folds and wrinkles. Based on these observations simplified kinematic mechanisms
were proposed to predict the crushing response of thin-walled members of simple
geometry under different loading conditions, see Fig. 6.1-31. Such models typically
lead to small computer codes or even to closed-form solutions and, therefore, can
be used efficiently in the preliminary design of energy absorbing systems [65 68].
Using simple design formulae it is also possible to apply advanced optimization
strategies for predefining mass efficient foam-filled crash elements [68,69].
100
FE analysis / filled
FE analysis / empty
3
No. 1396 / 0.52 g/cm
80 No. 1404 / empty
60
Load [kN]
40
20
0
0 20 40 60 80 100
Displacement [mm]
Figure 6.1-30. FE models of the crushing of filled and empty square tubes: Comparison of
computed and measured load±compression curves [61] (left) and stages of compressive axial
deformation (right).
Figure 6.1-31. Folding mechanisms for empty prismatic profile (left) and simplified deformation
kinematics of foam core (right).
6.1.6
Design Optimization for Cellular Metals
With growing knowledge about the mechanical performance of cellular metals,

and in particular of metallic foams, methods for the design of foam components
for applications such as packaging, energy absorption and sandwich structures
have evolved significantly [1,7,54]. The ability of cellular metals to compete with
other materials can be underlined by emphasizing their multifunctionality [70].
Furthermore, it was shown that in certain structures, such as cylindrical sandwich
shells, metal foam cores are more weight efficient than other stiffening concepts
[71]. In studies dealing with the design and optimization of metal foam compo-
nents usually a foam of uniform density has been assumed.
Metallic foams show some potential for being produced with controlled spatial
variations of their density, introducing inhomogeneities on a mesoscale as defined
in Section 6.1.4. In the case of foams produced via powder metallurgical routes the
foaming process, and, as a consequence, the final density distribution may be con-
trolled by suitable choices of the distribution of the blowing agent and of process
parameters such as temperature and pressure. This suggests employing such
foams as graded materials in space filling lightweight structures designed in anal-
ogy to cancellous bone, a natural functionally graded cellular material that displays
increased density in regions of high loading, see Fig. 6.1-32. Reiter et al. [72,73]
presented algorithms for the optimization of composite materials that are based
on procedures for simulating the natural adaptation of bone to applied loads.
Daxner et al. [74] implemented the algorithm of Reiter [73] for a self-adapting
material, providing options for converging towards a uniform distribution of the
local strain energy density or towards the satisfaction of a local yield criterion by
iteratively adapting the local density. The former choice gives rise to configurations
of increased stiffness and the latter to density distributions in which the material
shows uniform safety against plastic failure.
It was shown that strength optimization with material parameters that are typi-
cal for metallic foams leads to continuous solutions showing density gradients, see
Fig. 6.1-33, which may indeed be realizable by suitable processing technologies. At-
tempts to optimize the specific stiffness of foam structures resulted in the forma-
tion of discrete structures such as frameworks of struts. This indicates that cellular
metals may not be ideal for stiffness-optimized structures, at least in the presence
of concentrated loads. For sandwich beams with foam cores as studied by Vonach
et al. [75] smooth mesoscopic density distributions were also obtained for stiffness
optimized configurations under distributed loads.
6.1.7
Outlook
A number of different strategies for modeling the mechanical behavior of cellular

metals were presented. Their present use in studying these materials and their
contributions to improving the understanding of metallic foams were discussed.
The future of micromechanical modeling can be expected to bring an increased
interest in modeling 3D cellular arrangements and in transferring methods that
have become successful in the analysis of planar morphologies into the 3D do-
main. This process is facilitated by the steadily increasing capabilities of computer
hardware and software. Therefore, the advent of micromechanical models that bet-
ter capture the behavior of 3D microgeometries (especially for closed-cell foams)
may be forecast. Larger and more sophisticated 3D micromechanical models will
provide more precise information on the behavior of cellular metals under multi-
axial loading states. This, together with additional information from corresponding
multiaxial experiments, can be expected to lead to new insight into phenomena re-
lated to the macromechanics of cellular models and may give rise to the develop-
ment of more refined constitutive theories.
Figure 6.1-32. Section through a typical

human femur.
Figure 6.1-33. Strength optimized symmetric beam with fixed supports under a central load.
Highly stressed regions such as the supports and the lower and upper regions in the middle of
the beam accumulate the highest local foam density [74]
In view of the engineering potential of cellular metals and of the research chal-
lenges they pose, it appears safe to say that studies of these materials on all length
scales discussed above will remain a fertile field of study for many years to come.
Acknowledgements
Many of the results presented in this contribution were obtained in connection

with the Brite/EURAM project `EAMLIFe' (BE96-3605). We gratefully acknowledge
the funding of this project by the Commission of the European Union. We also
would like to thank the Institut fuÈr Verbundwerkstoffe (IVW) in Kaiserslautern,
Germany, in particular Mr. Ralf Denzer, for the excellent collaboration, which in-
cluded the presentation of a joint paper [32] at the `Symposium MetallschaÈume'
in Vienna, February 2000, excerpts of which were used here. Last but not least,
we want to thank the Institute of Materials Science and Testing of Vienna Univer-
sity of Technology for providing information and participating in many valuable
discussions.
References
1. L. J. Gibson, M. F. Ashby, Cellular Solids: 23. J. L. Grenestedt, K. Tanaka, Scr. Mater.

Structure and Properties, Pergamon Press, 1999, 40, 71 77.
Oxford, 1988. 24. A. E. Simone, L. J. Gibson, Acta Mater.
2. L. J. Gibson, M. F. Ashby, Cellular Solids: 1998, 46, 2139 2150.
Structure and Properties, 2nd edn, Cambridge 25. G. Schaffner, X. D. E. Guo, M. J. Silva, L. J.
University Press, UK 1997. Gibson, Int. J. Mech. Sci. 2000, 42, 645 656.
3. D. Weaire, M. A. Fortes, Adv. Phys. 1994, 43, 26. C. Chen, T. J. Lu, N. A. Fleck, Int. J. Mech.
685±738. Sci. 2001, 43, 487 504.
4. L. J. Gibson, Ann. Rev. Mater. Sci. 2000, 30, 27. M. J. Silva, L. J. Gibson, Int. J. Mech. Sci.
191 227. 1997, 39, 549 563.
5. P. Ponte Castan Ä eda, P. Suquet, in Advances 28. O. Prakash, P. Bichebois, Y. BreÂchet, F.
in Applied Mechanics 34, E. van der Giessen, Louchet, J. D. Embury, Phil. Mag. 1996, A73,
T. Y. Wu (eds), Academic Press, New York 739 751.
1998, p. 171 302. 29. T. Daxner, H. J. BoÈhm, F. G. Rammerstor-
6. S. Torquato, L. V. Gibiansky, M. J. Silva, L. J. fer, in Metal Foams and Porous Metal Struc-
Gibson, Int. J. Mech. Sci. 1998, 40, 71 82. tures, J. Banhart, M. F. Ashby, N. A. Fleck
7. L. J. Gibson, Mater. Sci. Engng. 1989, A110, (eds), MIT Verlag, Bremen 1999, p. 283 288.
1 36. 30. M. J. Silva, S. C. Hayes, L. J. Gibson, Int. J.
8. M. F. Ashby, C. J. Seymour, D. Cebon, in Mech. Sci. 1995, 37, 1161 1177.
Proc. MetallschaÈume, J. Banhart (ed.), MIT 31. T. J. Lu, C. Chen, Acta Mater. 1999, 47,
Verlag, Bremen 1997, p. 199 216. 1469 1485.
9. J. Zhang, M. F. Ashby, Int. J. Mech. Sci. 1992, 32. T. Daxner, R. Denzer, H. J. BoÈhm, F. G.
34, 491 509. Rammerstorfer, M. Maier, Mater.-wiss. Werk-
10. E. W. Andrews, L. J. Gibson, M. F. Ashby, stofftechn. 2000, 31, 447 450.
Acta Mater. 1999, 47, 2853 2863. 33. R. E. Williams, Science 1968, 161, 276 277.
11. J. L. Grenestedt, Int. J. Sol. Struct. 1999, 36, 34. W. E. Warren, A. M. Kraynik, J. Appl. Mech.
1471 1501. 1997, 64, 787 794.
12. R. M. Christensen, Int. J. Sol. Struct. 2000, 35. A. Ableidinger, ªSome Aspects of the
37, 93 104. Fracture Behavior of Metal Foamsº, Diploma
13. H. X. Zhu, J. F. Knott, N. J. Mills, J. Mech. Thesis, Vienna University of Technology,
Phys. Sol. 1997, 45, 319 343. Vienna 2000.
14. H. X. Zhu, N. J. Mills, J. F. Knott, J. Mech. 36. J. L. Grenestedt, F. Bassinet, Int. J. Mech.
Phys. Sol. 1997, 45, 1875 1904. Sci. 2000, 42, 1327 1338.
15. N. J. Mills, H. X. Zhu, J. Mech. Phys. Sol. 37. V. Shulmeister, M. W. D. Van der Burg, E.
1999, 47, 669 695. Van der Giessen, R. Marissen, Mech. Mater.
16. J. L. Grenestedt, J. Mech. Phys. Sol. 1998, 46, 1998, 30, 125 140.
29 50. 38. H. X. Zhu, J. R. Hobdell, A. H. Windle,
17. C. Chen, T. J. Lu, N. A. Fleck, J. Mech. Phys. Acta Mater. 2000, 48, 4893 4900.
Sol. 1999, 47, 2235 2272. 39. A. M. Kraynik, M. K. Nielsen, D. A. Reinelt,
18. S. P. Santosa, T. Wierzbicki, J. Mech. Phys. W. E. Warren, in Foams and Emulsions, J. F.
Sol. 1998, 46, 645 669. Sadoc, N. Rivier (eds), Kluwer, Dordrecht
19. S. D. Papka, S. Kyriakides, J. Mech. Phys. 1999.
Sol. 1994, 42, 1499 1532. 40. R. Denzer, M. Maier, in Proc. ECCM 99,
20. Z. Hashin, J. Appl. Mech. 1983, 50, Munich 1999.
481 505. 41. S. J. Hollister, J. M. Brennan, N. Kikuchi,
21. S. D. Papka, S. Kyriakides, Acta Mater. 1998, J. Biomech. 1994, 27, 433 444.
46, 2765 2776. 42. R. MuÈller, P. RuÈegsegger, J. Med. Engng.
22. A. E. Simone, L. J. Gibson, Acta Mater. Phys. 1995, 17, 126 133.
1998, 46, 3929 3935.
43. E. Maire, F. Wattebled, J. Y. BuffieÁre, G. 60. J. Schluppkotten, R. Paûmann, C. Cheng,

Peix, in Metal Matrix Composites and Metallic M. Maier, in ABAQUS Anwendertreffen 1999,
Foams, T. W. Clyne, F. Simancik (eds), Wiley- ABACOM, Aachen 1999.
VCH, Weinheim 2000, p. 68 73. 61. M. Seitzberger, F. G. Rammerstorfer, H. P.
44. H. X. Zhu, J. R. Hobdell, A. H. Windle, Degischer, R. Gradinger, Acta Mech. 1997,
J. Mech. Phys. Sol. 2001, 49, 857 870. 125, 93 105.
45. X. E. Guo, L. J. Gibson, Int. J. Mech. Sci. 62. A. G. Hanssen, M. Langseth, O. S. Hop-
1999, 41, 85 105. perstad, Int. J. Mech. Sci. 1999, 41, 967 993.
46. J. M. Albuquerque, M. FaÂtima Vaz, M. A. 63. M. Seitzberger, F. G. Rammerstorfer, R.
Fortes, Scr. Mater. 1999, 41, 167 174. Gradinger, H. P. Degischer, M. Blaimschein,
47. V. S. Deshpande, N. A. Fleck, J. Mech. Phys. C. Walch, Int. J. Sol. Struct. 2000, 37,
Sol. 2000, 48, 1253 1283. 4125 4147.
48. ABAQUS Theory Manual Version 5.8 64. S. Santosa, T. Wierzbicki, Comput. Struct.
(section 4.4.6), Hibbitt, Karlsson & Sorensen, 1998, 68, 343 367.
Inc., Pawtucket 1998. 65. W. Abramowicz, T. Wierzbicki, Int. J. Mech.
49. J. Y. Chen, Y. Huang, M. Ortiz, J. Mech. Sci. 1988, 30, 263 271.
Phys. Sol. 1998, 46, 789 828. 66. S. Santosa, T. Wierzbicki, Int. J. Mech. Sci.
50. F. G. Rammerstorfer, H. J. BoÈhm, in 1999, 41, 995 1019.
Computational Mechanics for the Twenty-First 67. M. Seitzberger, S. Willminger, in Proc.
Century, B. H. V. Topping (ed), Saxe-Coburg ICRASH 2000, E. C. Chirwa, D. Otte (eds),
Publications, Edinburgh 2000, p. 145 164. Bolton Institute, Bolton 2000, p. 458 469,
51. V. P. W. Shim, B. Y. Tay, W. J. Stronge, 68. Crash Cad. Impact Design, Europe, Inc.,
J. Engng. Mater. Technol. 1990, 112, 398 405. Michalowice, Poland.
52. R. Gradinger, F. G. Rammerstorfer, Acta 69. A. G. Hanssen, M. Langseth, O. S. Hop-
Mater. 1999, 47, 143 148. perstad, Int. J. Mech. Sci. 2001, 43, 153 176.
53. T. Daxner, H. J. BoÈhm, F. G. Rammerstor- 70. A. G. Evans, J. W. Hutchinson, M. F. Ashby,
fer, Comput. Mater. Sci. 1999, 16, 61 69. Prog. Mater. Sci. 1999, 43, 171 221.
54. M. F. Ashby, A. G. Evans, J. W. Hutchinson, 71. J. W. Hutchinson, M. Y. He, Int. J. Sol.
N. A. Fleck, Metal Foams: A Design Guide. Struct. 2000, 37, 6777 6794.
Technical Report CUED/C-MICROMECH/ 72. T. Reiter, F. G. Rammerstorfer, in Optimal
TR.3, Dept. of Engineering, Cambridge Uni- Design with Advanced Materials, P. Pedersen
versity, Cambridge 1998. (ed), Elsevier, Amsterdam 1993, p. 25 36.
55. C. Chen, Manual for a UMAT User Sub- 73. T. J. Reiter, Functional Adaptation of Bone
routine. Technical Report CUED/C- MICRO- and Application in Optimal Structural Design.
MECH/TR.4, Dept. of Engineering, Cam- VDI Fortschrittsberichte, Reihe 17, Nr. 145,
bridge University, Cambridge 1998. VDI Verlag, DuÈsseldorf 1996.
56. H. L. Schreyer, Q. H. Zuo, A. K. Maji, 74. T. Daxner, H. J. BoÈhm, F. G. Rammerstor-
J. Engng. Mech. 1994, 120, 1913 1930. fer, Mater. Sci. Technol. 2000, 16, 935 939.
57. W. Ehlers, A. Droste, Techn. Mech. 1999, 19, 75. W. K. Vonach, T. Daxner, F. G. Rammers-
341 350. torfer, in Sandwich Construction 5, H. R.
58. R. E. Miller, Int. J. Mech. Sci. 2000, 42, Meyer-Piening, D. Zenkert (eds), EMAS,
729 754. Solihull 2000, p. 291 300.
59. G. Gioux, T. M. McCormack, L. J. Gibson,
Int. J. Mech. Sci. 2000, 42, 1097 1117.
6.2
Mesomodel of Real Cellular Structures
B. Foroughi, B. Kriszt, and H. P. Degischer
6.2.1
Introduction
Cellular metals are a class of material that offers special properties as described in
Chapter 5. Their successful introduction into industrial application requires the
development of design methods that predict the responses of these materials to
external loads.
The discrete mass distribution of cellular metals causes discontinuities in their
local properties. To determine the response of a highly porous solid to external
loading, such a cellular structure can be modeled as a network of curved beams
(open-cell foam) or shells (closed-cell foam). This model can be analyzed using
available numerical solutions such as finite elements or boundary elements meth-
ods. The 3D cellular structure of a real, foamed metal (see Chapter 4) is too com-
plex to be easily handled with such a model due to excessive calculation time.
Another method is to approximate the cellular structure by a continuum model.
Such a homogenization method was used by some of the theoretical studies [1]. In
this approach the cellular structure is homogenized over the whole sample or com-
ponent, a scale much larger than the typical microstructure.
Theoretical studies [1] suggest that the mechanical properties of an ideally homo-
geneous foam are scaled by rn. In these relations that are also called ªscaling lawsº
the r is the relative density of the foam and n is a constant that depends on the
mechanism governing the deformation of the cell walls. For example, the elastic
modulus of an open-cell foam with low relative density ( r J 0.1) is governed by
bending of struts and scales with r2. At higher relative densities (0.1 J r J 0.3)
the effect of shear and axial deformation of struts shall be considered [2]. Correc-
tion of the scaling law shows that the influence of these effects on the final
response of an open-cell foam in the elastic regime is small. Prediction of such
scaling laws and their correction at medium densities are compared in Fig. 6.2-1.
In closed-cell foams, bending of struts is accompanied by stretching of the
cell faces, which is described by an additional linear term. The equation for the
theoretical elastic modulus of a closed-cell foam is expressed as [1]
r)2 S C2 (1 s f)
E=Es C1 (f r (1)
where Es is the elastic modulus of the solid material of struts and cell faces. f is the
volume fraction of solid contained in the cell edges, the remaining fraction (1 f) is
in the cell faces. The uniaxial compression (plateau) strength, s, of low-density open-
cell foams is scaled with r3/2. At higher relative densities the influence of the thick-
ness of the struts and the size of the corner becomes considerable [2].
282 6.2 Mesomodel of Real Cellular Structures
As shown in Fig. 6.2-1b the correction of the scaling law at medium densities
has a significant effect on the prediction of the value of plateau strength. The
theoretical value of plateau strength is modified for a closed-cell foam with the
stretching effect of cell faces as
r)3=2 S C4 (1 s f)
s=s ys C3 (f r (2)
where s ys is the yield stress of solid material and C1, ... C4 are constants depending
on the cell geometry [1]. These equations describe the behavior of a material that is
assumed to be continuous and homogeneous. In this study, our interest lies in
local variations of density and other dependent fields over scales not very much lar-
ger than the typical microscale. This may be accomplished with the help of a me-
soscale window (averaging volume), which becomes the classical representative vol-
ume element (RVE) in the infinite medium. Mechanical properties at such a me-
soscale can not be uniquely approximated. But they depend on the essential or nat-
ural boundary conditions of the averaging volume, which bounds the material re-
sponse from above and below [3]. This bound will be tighter with increasing size of
mesoscale or, in other words, with decreasing effect of boundary conditions on the
response of the averaging volume. In the mesoscale, the relative density can be de-
fined as a function of a local point X in an undeformed stress-free reference con-
figuration, r r(X). This allows us to define an inhomogeneous mass distribution
in the foamed samples. Studies on the effects of inhomogeneous or gradient
distributions of mass on the behavior of metallic foams are limited. Beals and
Thomson reported that density gradients have a significant effect on compression
properties of Alcan aluminum foams [4]. Huschka et al. modeled foam materials
with a stacking of layers of different densities [5]. Gradinger and Rammerstorfer
developed a 1D model to analyze the effect of density gradient in loading direction
[6]. The effect of 1D inhomogeneities on the stress strain response and on the
crush energy absorption of aluminum foams has been studied. Daxner et al.
improved this method to a 2D model [7]. In this model, the material is re-
Figure 6.2-1. Correction of scaling laws at medium densities: a) elastic modulus; b) plateau
strength.
0.05
Alulight cast
Alulight wrought alloy Scaling law (Correc.)
0.04 Alporas φ = 0.96
σ /σ ys
0.03
Scaling law (Correc.)
φ = 0.9
0.02
0.01 Scaling law

φ =1
0
0.08 0.1 0.12 0.14 0.16 0.18 0.2
Figure 6.2-2. Comparison of measured and predicted plateau strength of Alporas and Alulight
samples. The plateau strengths have been normalized by the yield stress of the cell wall.
presented as an array of point masses connected by nonlinear springs and rigid

cross-bridges.
X-ray computed tomography (XCT) revealed the 3D inhomogeneous mass distri-
bution in metallic foams [8,9]. The influence of the inhomogeneities on the defor-
mation and the elastic and plastic behavior of metallic foams was also studied as a
mesoscopical imperfection [10 12].
The manufacturing process of foamed metals has a major effect on the architec-
ture of the cellular structure exhibiting inhomogeneous pore distribution, cell
nodes, and walls. The plateau strength of two types of aluminum foam samples
measured in uniaxial compression tests were compared with the predictions of
scaling laws (Eq. 2) in Fig. 6.2-2. Two different routes are used to produce the
Alporas [13] and Alulight [14] foams, respectively. The homogenized scaling laws
can clearly predict the experimental results for Alporas material better than
those of Alulight samples. This phenomenon can be explained by the distribution
of mass in the cellular structure: the Alporas material is quite uniform while the
Alulight material shows high variation in the local density [9].
A continuum model is implemented using the finite element (FE) method to si-
mulate the effect of local mass distribution on the overall behavior of cellular sam-
ples. Three constitutive laws are used for modeling the plastic deformation of ele-
ments: classical J2 plasticity, the crushable foam model, the simplified self-similar
model. It is shown that the inhomogeneous density distribution leads to plastic
strain localization. The simulated plastic strain localization is compared with the
real observations and the influence of inhomogeneity on calculated elastic modulus
and plateau strength are studied.
6.2.2
3D Mesomodel
The density mapping method was used to approximate the cellular structure by a
continuous, 3D density distribution at the mesolevel. The density recorded by
XCT is homogenized over an averaging volume. The dimension of the averaging
volume specifies the size of mesoscale where the cellular structure is assumed
to be homogeneous. Overlapping of averaging volumes allows to increase the reso-
lution of continuous density distribution up to the resolution of used XCT data.
Details of the density mapping method have been explained in Chapter 4.3 and
previously [8,10]. The influence of an averaging volume size (AVS) on the result
of the developed mesomodel will be discussed in Section 6.2.4.
The obtained inhomogeneous continuum body is then implemented by FE anal-
ysis. The mean density of each element is obtained by averaging the values of den-
sity mapping for all XCT voxels within the FE. Figure 6.2.-3. shows 2D XCT pixels
overlapping with a triangular FE. The next step is to describe the inhomogeneous
body as a multiple sub-domain problem. The material properties (elastic modulus
and Poisson's ratio in the elastic regime and plateau strength in the plastic regime)
depend on the local relative density of each element. In a general case, each ele-
ment will have a mechanical behavior that is different from the behavior of neigh-
boring elements. This high variation in the material properties can complicate the
analysis of the model. To overcome this difficulty the relative density variation,
rmin, rmax], is divided into N intervals. All elements, in which the relative densities
[
lie in the interval i (rFE [ri, ri 1]) form a sub-domain Vi (i 1,2,....,N). Therefore
the material consists of N sub-domains each of which is assumed to be homoge-
neous and isotropic. Their mechanical properties depend on the average value of
the relative densities of the elements lying within the sub-domain.
Although the mechanical properties in each sub-domain are assumed to be con-
stant, they show discontinuities across any interface between sub-domains. The in-
terfacial conditions can lead to deviation from uniaxiality of the loading condition
in the sub-domains even in the case of uniaxial compression of samples. Because
of this, the general form of constitutive laws is used for the description of elastic
and plastic behavior of the sub-domains.
Figure 6.2-3. Overlapping the XCT pixels with a

finite element.
6.2.2.1 Elastic Regime

The generalized Hooke's law is used to calculate the linear elastic strain tensor in
the form
eei Sij s j i, j 1, 2, . . . 6 (3)
where eie, s j are the elastic strain and stress tensor, respectively, and Sij is the com-
pliance matrix. This matrix is characterized by a set of moduli. Two independent
moduli describe an isotropic and homogeneous material. The Young's modulus
and the Poisson's ratio are used for this purpose. The Young's modulus of each
sub-domain is derived from the scaling law (Eq. 1) with the assumption that the
material behaves like a closed-cell foam. The volume fraction, f, should be ob-
tained by fitting the experimental results.
The Poisson's ratio is assumed to be 0.35 in the elastic regime. There is no ex-
perimental evidence for this parameter but the computational modeling of a reg-
ular cellular structure (Kelvin cells) suggests 0.35 as a reliable value [15].
6.2.2.2 Plastic Regime

The linear elasticity is limited by the yield criterion. Plastic yielding is the most
usual mechanism for ªfailureº of ductile metals. This mechanism can be described
by rate-dependent incremental plasticity theory. According to this theory, the elastic
theory breaks down as a certain function of stress reaches a certain value. It is
known as a yield surface within which the material is elastic and on which the plas-
tic flow may take place. Various models have been developed to describe the shape
of a yield surface for a homogeneous cellular material [16 18]. The present inves-
tigation focuses on three models: classical J2 plasticity [19], the crushable foam
model [20], and the simplified self-similar model [16].
Classical J2 Plasticity
The classical J2 plasticity is the simplest and most common model to predict the
plastic deformation in dense metals. The yield surface in this model is expressed as
f se s Y 0 (4)
where Y is the yield strength of the material under uniaxial load and s e is the von
Mises effective stress defined by
r
3
se sij sij (5)
2
sij is the deviatoric stress tensor. This yield surface is independent of the mean
stress. Such a material can not fail due to hydrostatic load and is plastically incom-
pressible (epvol) when an associate flow rule is assumed [19]. Because of plastic com-
pressibility of cellular structure, the J2 plasticity is used to estimate the mechanical
behavior of foams at small global strain values and to determine the position where
local deformation begins to form a band. Strictly, J2 plasticity is not valid for
description of plastic deformation of foams under multiaxial loads, but in uniaxial

compression of an inhomogeneous material when the effects of interfacial condi-
tions are small, this model can provide reasonable results. The model is calibrated
against the response of material to the uniaxial loading. We assume, here, that each
sub-domain behaves in a linearly elastic and perfectly plastic manner. The scaling
law (Eq. 2) with the correction for medium densities is used to predict the plateau
stress in each sub-domain. The volume fraction, f, is considered to be the same as
for the prediction of elastic moduli.
Crushable Foam Model

ABAQUS [20] has also offered a model for simulation of plastic deformation of cel-
lular structure. There are two advantages of this model in comparison with the J2
plasticity: the ability to model volumetrically plastic deformation, and the different
responses of the material in tension and compression. Note that in tension, cell
walls break and as a result the tension-bearing capacity of cellular structure is con-
siderably smaller than its compressive strength. This effect has been considered in
this model. The yield surface is defined, here, in the term of the invariant stresses
with an elliptical dependence of deviatoric stress on pressure (mean) stress.
" !2 !2 #1=2
pt s pc t pt S pc
f Sp S s 0 (6a)
2 M 2
where
" ! !3 #
1 1 1 r
t se 1 S s 1s (6b)
2 K K se
where pc and pt are the strength of material in hydrostatic compression and ten-
sion, respectively. M is the slope of critical state line that depends on pt and on the
initial yield strength under uniaxial load, Y, as well as under initial hydrostatic com-
pression strength, pc0. K is a material parameter that defines the shape of yield surface
in the deviatoric plane (Fig. 6.2-4a), and r is the third deviatoric invariant of sij.
The yield surface is depicted in Fig. 6.2-4b. Non-associated flow is assumed for
this constitutive model. The flow potential, g, is defined in term of hydrostatic and
deviatoric effective stress, and chosen in this model as
r
9 2
g p S s 2e (7)
2
This potential gives the zero plastic Poisson's ratio. This means that loading in
any direction causes insignificant deformation in the other directions perpendicu-
lar to the loading. Experiments don't always confirm this property for foamed me-
tals, specially those that have high relative densities [16]. This is a limit of the
crushable foam model to describe the plastic deformation of foamed metals. An-
other restriction is the calibration of the model with the experimental results.
S1 t hardened surface
K=1.0
K=0.8
pt pc|0 pc p
S3 softened surface
S2
original surface
(a) (b)
Figure 6.2-4. Crushable foam model: a) effect of material parameter K; b) yield surface in t p
plane.
Here, the hardening/softening response is controlled by the value of the hydro-

static compression strength pc. In other words, compaction or dilation of the ma-
terial, which is modeled by variation of the yield surface's shape as shown in
Fig. 6.2-4b, is calibrated with the response of the material in a hydrostatic compres-
sion test. It is assumed that pt remains fixed throughout any plastic deformation.
Generally, we define the pressure pc versus total plastic volume strain epl vol as
follows.
a) plateau regime
pl pl
pc k1 evol S pc0 evol J eD (8a)
b) densification regime
pl pl
pc k2 eaevol eD J evol (8b)
k1, k2, and a are material parameters that depend only on relative density. They
must be evaluated by fitting the p e curves obtained by hydrostatic compression
tests. Parameter eD (densification strain) is adjusted to satisfy the continuity of
the function pc(epl
vol) in the passage from zone (a) to zone (b).
Simplified Self-Similar Model

The model was developed by Deshpande and Fleck [16] assuming the hardening
response is given directly by uniaxial compression responses. Hence the model
can exactly predict the compression response of a homogeneous material that is
loaded uniaxially. Multiaxial loading causes disagreement of the model predictions
with experimental results.
In uniaxial loading of an inhomogeneous material, the interface forces cause a
multiaxial loading condition in the sub-domains. This can also lead to deviation
of material behavior from the prediction of the model. This deviation depends
on the order of inhomogeneity (difference between lowest and highest density)
as well as the shape of sub-domains in the foamed sample. This constitutive law
is employed for modeling the uniaxial compression (including plateau and densi-
fication regimes) of real foamed samples as an inhomogeneous material. For this
purpose, the hardening responses of each sub-domain to uniaxial loading should

be specified until full densification. This means that the stress strain response
of homogeneous cellular structures should be available for the whole density
range as a function of relative density. In this study, we introduce a material
model for this purpose, the aim of which is to specify the ªCauchy stress true
plastic strainº of an isotropic homogeneous metallic foam.
The model assumes that the hardening response of an ideally homogeneous cel-
lular structure in the plateau regime is only governed by work hardening of cell wall
material. However this hardening effect is small, consequently a homogeneous
foam shows no hardening response in the plateau regime. In the densification re-
gime, the opposing walls of cells crush together and the cell wall material itself is
compressed. When this happens the stress strain curve rises rapidly. This harden-
ing response should be approximated using experimental data, since there is no
physical model for this regime. Accordingly, the material model is defined as
a) plateau regime
pl
s true s etrue J eD
true (9a)
b) densification regime
pl pl
s true cedetrue eD
true J etrue (9b)
where s is the compression strength of homogeneous material that is determined

using scaling laws (Eq. 2). The material parameter d and densification strain, eDtrue,
are obtained by fitting the experimental data. Finally, the parameter c is determined
so that the curve remains continuous in the transition from plateau to densifica-
tion regime
D
c s es detrue (10)
6.2.3
Modeling of Uniaxial Compression
Uniaxial compression tests are simulated with respect to the presented model. The
compression samples are rectangular prisms with dimensions of 20 q 20 q
40 mm3. ABAQUS/Standard is employed for computation of the field variable (dis-
placement, strain, and stress). 3D solid (continuum) elements with eight nodes of
type C3D8 are used to model the sub-domains. The nodes of this element have
three translational DOFs (degrees of freedom) but no rotational DOFs. The bound-
ary conditions are chosen so that they represent the conditions of uniaxial com-
pression tests. For this purpose, one clamping surface (bottom face) is assumed
to be fully fixed within the clamping plane, so all nodes are fixed in this plane
(x-, y-, z- translational displacements are assumed to be zero). Note that because
of the friction effect, in practice, the lateral displacement of a sample on the clamp-
ing surface is also negligible. Nodes in the other clamping surface (top face) are
fixed in the lateral directions (x and y) whereas the displacement in the axial direc-
tion (z) is given as loading.
6.2.3.1 Deformation Band

The compression experiments indicate that the inhomogeneous density distribu-
tion leads to localization of deformation and the formation of deformation
band(s) [11,12]. Figure 6.2-5 depicts the lateral surfaces of an Alulight sample re-
corded by a digital camera at 0 and 4 % nominal strain. The sample has an average
density of 350 kg/m3. The deformed cells are marked by arrows. It can be seen that
the deformed cells form a band in the sample. The edge length of the averaging
volume chosen to homogenize the density surrounding a XCT voxel was 6.4 mm
(11 pixel). The edge length of elements is chosen to be three times the size of
the CT pixel in each slice.
Figure 6.2-5. Surface images of an Alulight cells are marked with white arrows; c) calcu-
cast alloy sample of r 0.13: a) cellular struc- lated local plastic strains in loading direction
ture before deformation; b) cellular structure using presented model at 4 % nominal strain,
after deformation, 4 % nominal stain, deformed yielding the first deformation band.
Figure 6.2-6. Results of deformation modeling using the crushable foam model for densification
of first deformation band and forming the second band.
Figure 6.2-7. Simulation of large deformation of an Alulight sample using the simplified self-
similar model.
To compute the effect of inhomogeneity the investigated sample was divided into
nine sub-domains. The densities of the sub-domains were 220, 240, 260, 280, 330,
400, 500, 620, and 720 kg/m3. The calculated plastic strains in the loading direction
happening at a nominal strain of 4 % are depicted in the Fig. 6.2-5c.
Very good agreement is seen between the position of the observed and the com-
puted deformation band. Densification of this band and forming the second defor-
mation band was modeled using the crushable foam model as shown in Fig. 6.2-6.
Large deformation of another Alulight specimen was simulated using the simpli-
fied self-similar model as shown in Fig. 6.2-7. It should be noted that field para-
meters (such as stress, strain, and displacement) are computed for a continuum.
They are average values and must not be considered as microscopical field para-
meters related to deformation of the cell walls.
6.2.3.2 Mechanical Properties

Figure 6.2-8 shows the predicted stress strain curve of a typical Alporas specimen.
The simplified self-similar model described above was used to model the harden-
ing response during uniaxial loading. Although no strain hardening was assumed
for sub-domains in the plateau regime, the response of the whole specimens shows
hardening, which follows the measured response accurately. This phenomenon
can be described by large deformation of a sub-domain or of some elements in
a sub-domain, which can cause local densification. Rapidly increasing stress in
densification regime for these elements can lead to slow hardening of the whole
specimen in the plateau regime. With an increasing number of elements that
are densified the slope of stress strain curve increases quickly. The material para-
meters in the Eq. 9 are obtained by fitting the measured data and are considered in
this simulation as follows [21]:
Figure 6.2-8. Comparison of the measured stress strain curve of a typical Alporas sample with
the prediction of the mesomodel.
s 0:507r1:56, d 2:0, eD
true 1:0 s 2:13 r (11)
The effect of inhomogeneous mass distribution on the mechanical properties can
also be calculated using the developed mesomodel. For example, the Young's mod-
uli and compressive strengths of various Alulight samples have been predicted. As
described in Section 6.2.2, the mechanical properties of each sub-domain were ob-
tained from Eq. 1 and Eq. 2 assuming f 0.94 [22]. Computed and measured val-
ues as well as the prediction of scaling laws are depicted in Fig. 6.2-9. Comparison
of computed values with measured ones suggests the effect of inhomogeneity. It
0.08
0.07
0.06
0.05
E/Es
0.04
0.03
Exp.
0.02 mesomodel
Phi=1
Phi=0.9
0.01
Phi=0.94
(a) 0
0.08 0.10 0.12 0.14 0.16 0.18 0.20 0.22 0.24 0.26
0.06
0.05
0.04
σp /σ ys
0.03
0.02
Exp.
mesomodel
Phi=1
0.01
Phi=0.94
Phi=0.94 (correc.)
(b)
0
0.08 0.10 0.12 0.14 0.16 0.18 0.20 0.22 0.24 0.26
Figure 6.2-9. Comparison of measured values with predication of the mesomodel based on the
simplified self-similar model: a) elastic modulus; b) plateau strength.
can be seen that this effect on elastic modulus is small, while the plateau strength
can be significantly affected from inhomogeneous mass distribution. However, the
overall response of foamed samples depends on the size and shape of soft (sub-do-
mains with low densities) and hard (sub-domains with high densities) regions. The
plateau strength will be decreased specially when a soft region is oriented perpen-
dicular to the external compressive load [10,12].
6.2.4
Discussion
The transformation of a discretely heterogeneous structure to an approximated

continuum was one of the major challenges in this study. The discrete XCT data
(microscopical density field) have been averaged over a certain volume. However,
the overall response of samples is affected by the size of the AVS, which should
contain a cellular structure that behaves like a ªfoamº. Figure 6.2-10a shows the
calculated plateau strength of an Alulight sample versus the size of the averaging
volume. The sample had an average density 419 kg/m3 and the measured plateau
strength was 3.0 MPa, which is about 12 % below the predicted value of scaling laws
applied to the average density. It is seen that the difference between measured and
calculated values (mesomodel) is decreasing with increasing AVS (Fig. 6.2-10a). The
relative error is less than 12 % for an averaging volume equal 33 pixel (19.5 mm) at
0.2 % nominal strain and less than 5 % at 4 % nominal strain (see Fig. 6.2-10).
Andrews et al. [23] reported that the compressive strength of closed-cell samples
reaches the plateau values if the ratio of sample size to cell size is greater than five. The
average cell size of investigated Alulight samples was measured around 3 4 mm.
Therefore it can be assumed that at an average volume size of about 19 mm, the
effect of boundary conditions on response of the averaging volume is negligible
and the homogenization requirement is fulfilled. If AVS 11 pixels (6.4 mm) the
calculated stress at 0.2 % stain is 20 % below the measured plateau stress, at 4 % the
difference is reduced to 8 %. The comparison of predicted strength depends on the
strain at which the plateau strength is taken. The effect of AVS on the stress strain
curve is relevant at the beginning of loading as shown in Fig. 6.2-10b, revealing
a higher resolution of the strengthening of the cellular structure in the beginning
of compression. Anyhow the first loading to plateau stress could not be simulated.
Figure 6.2-11 demonstrates the difference in the resolution of the density mapping
on increasing the AVS. But it is remarkable that the inhomogeneity is still very
clearly imaged at an AVS of 19.5 mm (33 pixel) edge lengths. However, averaging
over very large volumes leads to elimination of inhomogeneity effects.
Three constitutive laws have been used to model the plastic deformation of sub-
domains. Because of incompressibility of J2 plasticity model it is only suitable to
predict the position of the first deformation. The applicability of the crushable
foam model to cellular metals should be demonstrated with more accuracy. In ad-
dition, this model must be calibrated with the results of hydrostatic compression
tests. Therefore to simulate the hardening response of an inhomogeneous cellular
sample with crushable foam model, the stress strain curves under hydrostatic
compression as a function of density is needed. The tests should be performed

until full densification to allow the fitting of material parameters for this model,
which was presented in Eq. 8. Insufficient material data lead to divergence of
the method and stopping the FE analysis. The simplified self-similar model [17]
assumes that the hardening is given directly by the response of a homogeneous
cellular metal under uniaxial compression. Eq. 9 can model the response of
Alporas material accurately. More experiments are necessary to calibrate the
material parameters in these equations for Alulight specimens specially in the
densification regime.
3.5
2.5
Stress [MPa]
1.5
Exp.
0.5 AVS=11 pixel
AVS=23 pixel
ADS=33 pixel
0
(b) 0 1 2 3 4 5 6
Nominal strain [%]

Figure 6.2-10. Effect of averaging volume size (AVS): a) on the computed plateau strength; b) on
the predicted stress strain response.
Figure 6.2-11. Influence of averaging volume size (AVS) on the density distribution derived from
XCT data (0.588 mm/pixel).
6.2.5
Conclusions
A mesomodel has been developed to explain the localized deformation behavior and
to calculate the mechanical properties on the basis of the mass distribution of the
cellular metals measured by CT data. Density mapping method is employed to ap-
proximate the mesoscopical continuous distribution of density in investigated sam-
ples. The model confirms the correlation between the inhomogeneous density dis-
tribution and the localization of plastic strains yielding deformation bands. The po-
sition of forming and propagation of the deformation band is predicted accurately.
The influence of inhomogeneity on the elastic modulus (the stiffness of the cel-
lular structure) and yield strength of metallic foams can be demonstrated. The in-
fluence on the elastic modulus is small while the plateau strength can be affected
significantly. The discrepancy between the calculated and the measured plateau
strength is caused by the uncertainties for the parameters used for the scaling
laws for Alulight and effects due to the assumption of isotropy.
References
1. L. J. Gibson, M. F. Ashby, Cellular Solids: 12. B. Kriszt, B. Foroughi, K. Faure, H. P.
Structure and Properties, 2nd edn, Cambridge Degischer, Mater. Sci. Technol. 2000, 16,
University Press, UK 1997, Chapter 5. 792 796.
2. L. J. Gibson, M. F. Ashby, Proc. R. Soc. Lond. 13. T. Miyoshi, M. Itoh, S. Akiyama, A. Kita-
1982, A382, 43 59 hara, Adv. Eng. Mater. 2000, 2, 179 183.
3. M. Ostoja-starzewski, Int. J. Sol. Struct. 1998, 14. J. Banhart, J. Baumeister, J. Mater. Sci.
35, 2429 2455. 1998, 33, 1431 1440.
4. J. T. Beals, M. S. Thomson, J. Mater. Sci. 15. A. Ableidinger, Thesis, University of
1997, 32, 3595 3600. Vienna, 2000.
5. S. Huschka, S. Hicken, F. J. Arendts, in Proc. 16. V. S. Deshpande, N. A. Fleck, J. Mech. Phys.
MetallschaÈume, J. Banhart (ed.), MIT, Bremen Solids 2000, 48, 1253 1283.
1997, p. 189±197. 17. R. E. Miller, Int. J. Mech. Sci. 2000, 42,
6. R. Gradinger, F. G. Rammerstorfer, Acta 729 754.
Mater. 1999, 47, 143 148. 18. W. Ehlers, H. MuÈllerschoÈn, O. Klar, in
7. T. Daxner, H. J. BoÈhm, F. G. Rammerstorfer, Metal Foams and Porous Metal Structures,
Comput. Mater. Sci. 1999, 16, 61 91. J. Banhart, M. F. Ashby, N. A. Fleck
8. H. P. Degischer, A. Kottar, in Metal Foams (eds), MIT Verlag, Bremen 1999,
and Porous Metal Structures, J. Banhart, M. F. p. 255 262.
Ashby, N. A. Fleck (eds), MIT Verlag, Bremen 19. J. Lubliner, Plasticity Theory, Macmillan,
1999, p. 213 220. New York 1990.
9. A. Kottar, H. P. Degischer, B. Kriszt, Mater- 20. HKS ABAQUS/Standard user manual,
wiss. Werkstofftechn. 2000, 6, 465 469. Version 5.8, Hibbit Karlsson & Sorensen Inc.,
10. B. Kriszt, B. Foroughi, K. Faure, H. P. Nantucket, RI 1998.
Degischer, in Metal Foams and Porous Metal 21. B. Foroughi, Thesis Vienna, University of
Structures, J. Banhart, M. F. Ashby, N. A. Fleck Technology, Austria 2001.
(eds), MIT Verlag, Bremen 1999, p. 241 246. 22. K. Y. G. McCullough, N. A. Fleck, M. F.
11. B. Kriszt, B. Foroughi, A. Kottar, H. P. Ashby, Acta Mater. 1999, 47, 2323 2330.
Degischer, Proc. Euromat 99, Vol. 5, Wiley- 23. E. W. Andrews, P. R. Oneck, L. J. Gibson,
VCH, Weinheim, 2000, p. 4 82. Int. J. Mech. Sci. 2001, 43, 701 713.
7
Service Properties and Exploitability
7.1
The Range of Applications of Structural Foams Based on Cellular Metals
and Alternative Polymer Solutions
C. Haberling
Lightweight construction is gaining importance in automobile technology. Along-

side improvements in wind resistance and tribology, weight reduction measures
are of primary importance. This is necessary because for each particular automo-
bile class there are increasing customer demands for safety, comfort, performance,
and available internal space. Using conventional methods of construction to pro-
vide these customer-oriented features leads to a considerable increase in weight,
giving rise to strong pressures to reverse the weight gain. It is therefore regarded
as an imperative development objective to achieve a reduction in weight relative to
axle loads.
7.1.1
Introduction
Apart from using single materials in lightweight construction, increasing use is

made of mixed construction methods and various forms of composite materials
to achieve weight reduction. However, since these materials frequently incur
higher costs, it is of primary importance to ensure that they play a multifunctional
role [1].
This also applies to structural foams in the form of cellular metals based on alu-
minum or alternative systems based on expanding plastics, such as epoxy or poly-
urethane resin foams. These are required to take over functions such as energy ab-
sorption, stiffness, thrust-field reinforcement, and sound absorption in different
methods of construction [2].
300 7.1 The Range of Applications of Structural Foams Based on Cellular Metals
Plate structures:
p
3

E Z flexual stiffness
r
(plate)
Rp0,2
Z tensile strength
r
p
E
Z bending stiffness
r (bar)
Hollow structures:
q
3
Rp0,22
Z bending strength
r
(bar)
Rp0,2
Z tensile strength
r
Figure 7.1-1. Typical loading of car body structure (Audi A8): strategy for structural foam rein-
forcements.
7.1.2
Potential Areas of Use
The areas where these kinds of material are of potential use can be derived at an
early stage in design using a finite element (FE) analysis of typical stresses in a car
body. The result makes it possible to evaluate which structural components in the
body structure can play a multifunctional role and what measures can be adopted
to achieve the desired result cost-effectively, taking into account space availability
and the factors inherent in lightweight construction. As illustrated in the car
body of the Audi A8 (Fig. 7.1-1), there are many areas in the bodywork where struc-
tural foams can be used. They are particularly suitable for hollow structures where
space is too restricted for other more cost-effective measures. Particular areas
where the multifunctional applications profile can be exploited to beneficial effect
are the side-members, the boundaries of the floor of the passenger compartment,
and in the region of the sills and pillars.
7.1.3
Material Properties
The application profile of a structural foam can be derived from the density of the
material, its compressive strength, and its elastic modulus. These are illustrated
(Fig. 7.1-2) for a single construction method. It is quite possible to adapt the behav-
ior of the material specifically to the local stress relationships within the structure
of the vehicle owing to the wide range of material densities, types of material, and
Figure 7.1-2. Mechanical properties of structural foams.
heat treatment methods that are available. Typically, metallic foams have a higher
elastic modulus, whereas epoxy foams exhibit a greater compressive strength at
similar densities for temperatures up to 80 hC.
7.1.4
Main Component Configurations
Using lightweight construction criteria [3], it is possible to use these material char-
acteristics to assess which type of material is most suited for what kind of stress
conditions (Fig. 7.1-1). Compared with compact materials, however, the efficiency
of foam materials in single construction methods where the available building
space is not an issue is somewhat limited. It is only where a differential method
of construction is used with compact materials and in limited spaces, which is
the predominant case encountered in automobile construction, where the effective-
ness of these types of foam materials comes into its own. Even greater advantages
are gained when they are used in conjunction with compact materials. In the first
place, all foam materials exhibit an extremely variable behavior under tensile and
compressive loads. While the compressive range is characterized by a broad, uni-
form, and even deformation behavior (plateau) under constant load, under tension,
structural foams inherently become brittle and fail at very low loads (Fig. 7.1-3). In
addition, inhomogeneous regions are formed in the material that do not particu-
larly affect the overall behavior under compressive stress (reaching the ultimate
strength either later or at a lower level) under purely axial loads, but may cause
buckling or failure of the foam structure if there is an uneven distribution across
Figure 7.1-3. Deformation behavior of cellular metals under compressive and tensile stresses.
Figure 7.1-4. Buckling due to non-uniform loads on cellular metals: testing and FE simulation.
the stressed cross section, and thus lead to a premature total failure (Fig. 7.1-4).
This would represent a risk particularly in the case of the complex stress patterns
in a car body that can certainly be avoided by using a carrier structure.
The foam materials on the market today that are used for structural stiffening
are ideally used inside hollow components or inside a sandwich composite struc-
ture (Fig. 7.1-5). The function of these carrier materials is to avoid high local ten-
sile stresses in the foam and to apply the forces acting externally into the core of
the foam, directing them during deformation. It is of primary importance, particu-
larly in circumstances where energy absorption is required, to obtain the correct
geometric balance in the carrier/foam system. A completely foamed tubular struc-
ture that is subject to bending can serve as an example. While the empty tube de-
forms considerably under a low load, when it is filled completely with foam, its
flexural resistance is increased considerably, but, at the same time, it will fail earlier
with smaller deformations. This can be attributed to excessive local stresses in the
outer tube structure that is overloaded owing to the restricted plastic deformation
locally, causing brittle fracture. It is more advisable to provide a partially foam-filled
structure where material is saved in the region of the neutral axis or inside the less
active tensile zones in the case of structural foams. In this way, weight can be
Figure 7.1-5. Use of cellular metals in combination with compact materials.
Figure 7.1-6. Foam-filled tubes: variation of foam geometries in impact bending tests.
reduced while, at the same time, increasing energy absorption significantly

(Fig. 7.1-6).
7.1.5
Application and Attachment Techniques
In general, there are many ways of applying and attaching structural foams. The
effectiveness depends essentially on the attachment of the foam materials to the
surrounding or adjacent components. This is particularly important in structures
that derive their properties from a composite sandwich structure. Frequently, how-
ever, these requirements conflict with production-related conditions such as man-
ufacturing tolerances. Several additional factors affect the stiffening of hollow
bodywork structures in car construction. In one such instance, it must be ensured
that the process liquids in the paint shop flow through the hollow cross sections of
the bodywork, so that, in general, this objective is defeated if larger hollow sections
are completely filled with foam. To prevent corrosion zones, narrow gaps must be
avoided between foam elements and the basic structure, particularly where mixed
construction methods are used, such as the employment of aluminum foams in a
steel body. Also, incomplete contact must be avoided for acoustic reasons. The con-
tact must be such that either it is so intimate that contact is permanently assured
over the entire surface, or both parts must be protected separately against corrosion
before assembly, for example by painting. However, this solution generally incurs
considerable extra costs. In the case of cellular metals, both cementing and casting
are feasible means of bonding surfaces. Local joining can be done using thermal
and mechanical bonding processes.
7.1.5.1 Casting
The casting of aluminum foams using sand and permanent molds can be regarded
as state of the art. Cast-in molding in pressure die-casting is possible using die-
casting machines controlled in real-time to avoid pressures in the molten metal be-
coming too high. Most die-cast parts used in car construction have a typical wall
thickness of less than 5 mm. In order to deal with all aspects of lightweight con-
struction in this range of applications, it is necessary to explore at what maximum
pressures and casting speeds aluminum foam inserts of minimum density can be
cast-in with the thinnest possible casting skin while avoiding both foam infiltration
and incomplete filling of the mold (Fig. 7.1-7). This is particularly important for
components with large hollow sections (such as swivel bearings) that are normally
produced by gravity molding with a lost-sand core (see Section 3.3).
Apart from casting, gluing is a process that can be used to join together the sur-
faces of foam parts. In car production it is especially important to achieve a suffi-
cient adherence and durability against corrosion and aging. Expanding adhesives
can be used to compensate for even greater manufacturing tolerances, so that, for
instance, it is possible to insert a foam part in an extruded section (Fig. 7.1-7)
(see Section 3.2).
Figure 7.1-7. Attachment techniques: casting and gluing.
7.1.5.2 Thermal Joining Processes

To produce a high quality joint, the thermal joining process is restricted to applica-
tions involving identical materials (see Section 3.2). Aluminum foams can be incor-
porated in the structure using soldering or welding methods. While soldering pro-
duces a joint across the entire surface, welding will result in a localized joint only.
Laser-beam welding in particular (Fig. 7.1-8) places high demands on the toler-
Figure 7.1-8. Attachment techniques: welding.

ances of the parts being joined. Other welding methods require appropriate pre-
paration of the carrier system, such as localized perforations, since the depth of
the weld is shallow. Welding with and without filler materials has been checked
in numerous ways. When employed in fabricating car bodies, there still remain
questions regarding the corrosion behavior in narrow gaps along the weld seam.
Furthermore, the positions of the seams must be adjusted to match the stress pro-
file of the component. This is necessary to ensure the maximum effectiveness of
the foam inlay, and, at the same time, to minimize as much as possible the weak-
ness of the surrounding sheet metal along the seam, since typically in aluminum
welds the zone around the weld is subject to softening.
7.1.5.3 Mechanical Joining Processes

It is not possible to achieve a joint over a large surface between the foam part and
the structure using a mechanical joining process (see Section 3.2). However, these
types of joining methods can compensate for broad tolerances and, if appropriate
corrosion protection measures are taken, mixed construction methods can be
adopted. Screw connections and riveting can be considered as well as localized
crimping of the component where it is more practical when access is possible
from one side only. These local joining methods are of particular benefit for pre-
assembly of components in combination with other joining techniques.
7.1.5.4 Three-dimensional sandwich

A further solution is offered by the use of a prefabricated three dimensional sandwich
structure. The most notable example of this type of construction is the aluminum
sandwich foam (AFS) sheet [4], in which the joint between the foam core and the
outer plate is produced in a roll-bonding cladding process before foaming in situ
(see Section 3.4). Since the foaming process does not take place until after the con-
version process at temperatures over 500 hC, 3D sandwich contours can also be fab-
ricated using this material. These sandwich structures are either screwed or welded
to a surrounding aluminum structure using the cover plates.
7.1.5.5 Alternative Cellular Materials Based on Polymers

Thermally-activated foaming is also a feature of alternative materials based on plas-
tics. The activation temperature is lower, however, falling in the range 150 190 hC
and can be incorporated into car production without incurring added expense in
the paint process. Furthermore, when foamed, the plastics adhere across the full
face of the surrounding outer structure, ensuring a complete joint over the
whole surface even with broader tolerance bands without requiring supplementary
joining measures while, at the same time, avoiding corrosion-prone gaps.
Figure 7.1-9. Test piece for stiffness and impact testing.
7.1.6
Effectiveness
Physical characteristics can be checked using simple test pieces for bending, im-
pact, torsional, and axial stresses that mirror the geometry of how hollow struc-
tures are used in a vehicle. In doing so, the assumption is made that the outer
dimensions of the construction space that is being optimized are not allowed to
change since an increase in the cross section of the carrier profile generally repre-
sents the most effective lightweight construction solution. This practical necessity
occurs frequently in vehicle production owing to space restrictions. The test piece
selected was a closed spot-welded hat section made from DC-04 sheet steel, filled
with U- and O-shaped foam pads with a density r 0.55 g/cm3 made from alu-
minum or epoxy foam. The properties evaluated were stiffness under bending,
or flexural strength, impact strength behavior, torsional stiffness, and torsional
strength. This was compared with a section of the same weight with thickened
walls (Fig. 7.1-9).
7.1.6.1 Bending and Torsional Stress

Regarding the flexural strength of the initial section with a 0.9 mm wall thickness,
although increasing the wall thickness results in an increase in the bending mo-
ment under elastic conditions, when the additional weight is taken into account,
this adjustment has a negative effect on the lightweight construction factor
s b,elast./m for the sections being studied. Tests on a hat section starting with a
wall thickness of 0.85 mm and a 4 mm thick U-section structural epoxy foam re-
inforcement result, in contrast, related to the weight used, to a 95 % improvement
in the flexural strength compared with an unreinforced section (Fig. 7.1-10).
A U-section reinforcement is not recommended for increasing the torsional stiff-
ness, since, in this case, the unreinforced wall fails under similar loads to those
applied to unreinforced sections. On the other hand, torsional stiffness can be in-
Figure 7.1-10. Bending and torsion testing of foam-reinforced test specimen.
creased by using an O-shaped structural foam insert [5]. Nevertheless, the effective-
ness is considerably less than when a bending stress is applied. Removing the ef-
fect of weight (t max,elast/m), reinforcement using 4 mm thick epoxy foam produces a
20 % improvement compared with an initial section without the insert. Using an
8 mm thick aluminum foam pad (this reflects the minimum wall thickness cur-
rently possible under production conditions when producing more complex
parts) results in a torsional stiffness that is scarcely any different than if one
used a 4 mm thick epoxy foam of the same density, with adhesion across the
complete surface. This is despite the higher elastic modulus and the greater
wall thickness, and it occurs because the surfaces for gluing the upper and
lower sides are reduced for technical reasons. In this case, therefore, owing to
the total additional weight, the aluminum foam does not offer an advantage
for lightweight construction. This applies also to unreinforced sections with
increased wall thicknesses used for this application, where the increased weight
make them unsuitable for lightweight construction since, again, they offer no
advantages.
7.1.6.2 Impact Stresses

The obvious advantage offered by a structural foam insert in relation to impact
strength is the strengthening and energy absorption that it confers on a hollow
structure even where there are large dynamic bending deformations. The behavior
can be modeled with great precision in an FE simulation. Calculations and tests on
a hat section show that, at an impact speed of 3 m/sec using a mass of 600 kg on a
section reinforced with a U-shaped structural foam insert joined to the surface of
Figure 7.1-11. Loading of empty and foam-reinforced test pieces in an impact bending test.
the external plate, the failure load and the energy absorption are more than
doubled compared with an unreinforced hollow section of the same weight and
with thickened walls. This applies also to epoxy foam variants and for aluminum
foam elements glued on all sides. If the same aluminum foam element is glued
only locally, however, such as along the cover plate, the stiffening effect is reduced
considerably (Fig. 7.1-11). If the energy absorption is spread through the body of
the composite component, the same pattern emerges. However, it is also evident
that the effectiveness in relation to the mass used falls off again when the foam
thicknesses are too high.
7.1.6.3 Axial Load

Tests have been carried out on different sections in many different ways to study
their effectiveness under axial loads with and without foaming in place. The buck-
ling associated with folding that is usually used for this type of loading applied to
hollow structural parts can also be used in cases where structural foams provide
stiffening and can bring about significant improvements in its efficiency [6]. A de-
cisive element in this is a reduction of the force levels as the folding is taking place.
Nevertheless this improvement is largely offset in many instances when one takes
the added weight into account [7]. Furthermore, the effectiveness of the empty
structure can be improved significantly by making certain changes in shape or
by using alternative composite fiber structures (carbon/aramid fibers in an epoxy
matrix) (Fig. 7.1-12) with little increase in weight [2].
Figure 7.1-12. Different construction methods of axial load tubes.
7.1.6.4 Acoustics
Acoustic effectiveness has been tested on different models (see also Section 3.4.3.1).
Large AFS panels display good absorption qualities below 500 Hz [4]. Other com-
pact foam parts also provide good absorption characteristics but only within a lim-
ited frequency range. Tests performed on foam-filled engine supports (Fig. 7.1-13)
A: diecast B: aluminium
engine support foam-insert
20 mm
m n
C: combination D: alternative
of A and B solution
Figure 7.1-13. Aluminum-foam filled engine support for acoustic means and alternative solution.
revealed high acoustic effectiveness, but were considerably heavier than alternative
solutions using organic materials since the minimum thickness of the foam was
8 mm. At present, therefore, the acoustic efficiency provided by closed pore cellular
metals when used in car construction can only be regarded as an added benefit
when taken in the context of a multifunctional role.
7.1.7
Outlook
It is important to make a clear distinction between different modes of transport

when describing the cost/benefit relationship of these types of materials. Whereas
structural foams in car construction represent quite a high-cost solution combined
with a high degree of effectiveness (Fig. 7.1-14), in the case of aviation the evalua-
tion turns out to be the other way round. In this case, owing to the completely dif-
ferent circumstances, the foam offers good value for the cost, but its efficiency is
somewhat modest compared with other materials used [8].
Up to now, despite the numerous possible applications in automobile technol-
ogy, aluminum-based cellular metals have not yet been used in series production.
This is understandable to the extent that most vehicle bodies are based on ferrous
materials, so that, owing to the problems of joining, tolerances, and corrosion as-
sociated with currently available methods, an aluminum foam core cannot be in-
corporated within cost budgets. The only applications where they could be used
at present are in those places in vehicles that are less prone to corrosion, or inside
aluminum bodies or aluminum structures. For components that are subject to
higher thermal stresses, such as heat-shield panels, metal foam is unsuitable for
weight reasons today owing to its minimum wall thickness of over 8 mm when
formed into complex shapes in production. Regarding its acoustic efficiency, it
competes with organic insulating materials that are very effective at low costs.
,
ls
Costs for lightweigt construction $/kg
e ria s al
at e ti
m ss en on Cellular
e w roce iffer ucti metals and
n p d str
w d n structural
ne - an f co foams
o o
on d
M t ho Mixed
e
m materials
lighter
materials
Structural
Starting modification
point I=bh /12
W=bh /6
weight reduction, meeting requirements

Figure 7.1-14. Foam usage as function of cost and weight-saving for car bodies.
On the other hand, structural foams based on plastics are now found in numer-
ous applications in car manufacture and are used in the structural regions illustrat-
ed in (Fig. 7.1-1). Large-scale use of these materials is, however, prevented in the
current cost situation, although many of the additional costs associated with metal-
lic foams, such as extra joining operations and corrosion protection do not apply in
the case of these materials. It is possible to use them when the space available ex-
cludes any other solutions, or where immediate additional stiffening measures
need to be taken, or when, as in the new Audi A4, additional costs or variants
can be avoided in another place.
AFS sandwich panels in prototypes have shown that there is a potential for using
them also [7]. Even for this material, however, series production has raised new
questions about joining in mixed construction methods. Furthermore, vehicle de-
sign must be adapted to the AFS sandwich in order to extract the maximum effi-
ciency from this material.
First and foremost, the quality of the material has to be improved before cellular
metals can be used more in vehicle technology. It is already possible, however, to
use them inside sandwich structures or in conjunction with carrier materials. De-
velopment effort is important to produce an efficient cost-effective joining method,
as well as investigations on corrosion protection for using them inside steel struc-
tures that are subject to corrosive conditions. Furthermore it is essential to reduce
further the present minimum wall thicknesses and the manufacturing costs in
order to be able to compete with comparable polymer-based materials and other
approaches.
References
1. H. G. Haldenwanger, ªKonzeptioneller 5. T. Wierzbicki et al., ªCrush Behaviour of Box

Leichtbau mit neuen Werkstoffen: Verfahren, Columns filled with Honeycombs or foamsº
ein Zielkonflikt bei der Verifizierung der Proc. Fraunhofer USA Metal Foam Symp. 7 8
Entwicklungsprozeûkette im Karosseriebauº, Oct 1997, Stanton, DW, MIT Publishing,
in Proc. Aachener Kolloquium, Fahrzeug und 1997.
Motortechnik, RWTH Aachen, IKA, VKA, 6. H. P. Degischer et al., ªOn the Non-De-
4 6 Oct 1999. structive Testing of Metal Foamsº in Metal
2. H. G. Haldenwanger et al., ªLeichtbau mit Foams and Porous Metal Structures, J. Banhart,
alternativen Werkstoffen und Verfahrenº In- M. F. Ashby, N. A. Fleck (eds), MIT Verlag,
genieur-Werkstoffe 1999, 8(1). Bremen 1999.
3. B. LuÈdge, ªFunktionaler Rohkarosserie- 7. Ch. Haberling et al., ªAluminum foams for
Leichtbauº, in Proc. Aachener Kolloquium, energy absorbing structures under axial
Fahrzeug und Motortechnik, RWTH Aachen, loadingº Metal Foams and Porous Metal
IKA, VKA, 4 6 Oct 1999. Structures, J. Banhart, M. F. Ashby, N. A.
4. H. W. Seelinger, ªApplication Strategies for Fleck (eds), MIT Verlag, Bremen 1999.
Aluminum-Foam-Sandwich Parts (AFS)º in 8. O. Schultz, R. Schindler,
Metal Foams and Porous Metal Structures, ªAluminiumschaum: Dynamische Festigkeit
J. Banhart, M. F. Ashby, N. A. Fleck (eds), und Anwendung im Hubschrauberº, Mater-
MIT Verlag, Bremen 1999. wiss. Werkstofftechn. 2000, 6, 511 514.
7.2
Functional Applications
J. Banhart
Cellular metallic materials are finding an increasing range of applications.

Whether a suitable porous metal or metal foam can be found to solve a given prob-
lem depends on many factors, summarized here by the following keywords.
x Morphology: type of porosity needed (open versus closed), amount of porosity
needed, size of porosity desired, total internal surface area of cellular material re-
quired.
x Metallurgy: metal, alloy, or microstructural state required.
x Processing: possibilities for carrying out secondary operations on the foam or cel-
lular solid, such as shaping, cutting, joining, coating.
x Economy: cost issues, suitability for large-volume production.
7.2.1
General Considerations
The first point is, in particular, crucial for any evaluation of applications for cellular
metallic materials. Many applications require that a medium, either liquid or gas-
eous, be able to pass through the cellular material. There may be a need for various
degrees of ªopennessº, ranging from ªvery openº for high rate fluid flow to
ªcompletely closedº for load-bearing structural applications, and appropriate mate-
rials satisfying these conditions have to be found. Figure 7.2-1 shows which types
of porosity the various application fields require. Normally, a difference is made be-
tween whether an application is ªfunctionalº or ªstructuralº, but there is consider-
able overlap between these two notions.
The question of from which metals or alloys a given type of cellular structure can
be manufactured is also important. Structural load-bearing parts have to be light
because otherwise they would be made from conventional massive metals or alloys.
Therefore, aluminum, magnesium, or titanium cellular or porous metals are pre-
ferred for such applications. For medical applications, titanium may be preferred
because of its compatibility with tissue. Stainless steel or titanium is required
for applications in which aggressive media are involved or high temperatures
occur.
Finally, processing and cost issues have to be considered. The technology must
be available to bring the selected cellular metal into the required shape and to in-
corporate it into the machine or vehicle where it has its function. A technology for
making cellular metal will be futile if the required component cannot be manufac-
tured at a reasonable price.
In the following sections, applications for cellular metals are discussed that are
predominantly ªfunctionalº in the sense discussed. Traditional powder metallurgy
314 7.2 Functional Applications
structural load-bearing
components
bio-medical energy
implants absorbers
catalyst
kind of supports
silencers
application
sound
absorbers
filters bearings
heat
exchangers
functional
open partially open closed
type of porosity
Applications of cellular metals grouped according to the degree of open porosity

Figure 7.2-1.
needed and whether the application is functional or structural.
(PM) has created porous sintered metals for a wide range of applications [1 3]. It
is therefore not surprising to find very similar applications for the cellular metals
described in the present text, provided they have a certain degree of open porosity.
7.2.2
Biomedical Implants
Biomedical applications often have both a structural and a functional aspect, and
are therefore very challenging. Titanium or cobalt chromium alloys are used for
prostheses or dental implants because of their biocompatibility. To ensure ingrowth
of tissue, one usually produces a porous layer of the same or another biocompatible
material on the prosthesis by thermal spraying or other methods [4]. Alternatively,
one could use porous titanium or titanium foam for such applications and tailor
the density distribution to meet the required strength and moduli of such compo-
nents. There is no unanimity about how implants should be designed to ensure
maximum durability and functionality. According to one opinion, the modulus
of, say, dental implants should match the modulus of the jaw bone. Knowing
the relationship between modulus and density of metallic foams, one could easily
manufacture implants with an appropriately adapted modulus, ensure bio-

compatibility, and stimulate bone ingrowth into the (open) porosity [5]. Strength
and design criteria have to be met as well. Magnesium foams could be used as
biodegradable implants that serve as a load-bearing structure as long as the bone
still grows but are gradually absorbed by the body in a later stage of recoalescence
[4].
7.2.3
Filtration and Separation
There are two types of filters: filters holding back and separating solid particles or
fibers dispersed in a liquid (suspensions) or filters holding back solid or liquid par-
ticles dispersed in a gas (smoke or fog). Examples of the first type are filters for
cleaning recycled polymer melts, for removing yeast from beer, or for contami-
nated oil. The second type includes filtration of diesel fumes or water removal
in air lines. Important filter properties are fine filtration capacity, good particle re-
tention, cleanability, mechanical properties, corrosion resistance, and cost. Some of
the cellular metals described in this text possess a combination of properties not
covered by the traditional PM materials and might therefore be considered as com-
plementary to these, such as the materials described in Sections 2.1.2, 2.3 and 2.4.
7.2.4
Heat Exchangers and Cooling Machines
Highly conductive foams based on copper or aluminum can be used as heat ex-
changers [6]. In this case open-cell structures are needed such as those described
in Sections 2.3 and 2.4. Heat can be removed from or added to gases or liquids
by letting them flow through the foam and cooling or heating the foam at the
same time. Owing to the open porosity, pressure drops can be minimized (see
Fig. 7.2-2). An example of such an application is a compact heat sink for cooling,
for example, in microelectronic devices with a high power dissipation density such
as computer chips or power electronic components. Nowadays, fin-pin arrays are
the standard solution in such cases. Metal foams or ordered cellular metals of
the type shown Fig. 7.2-3 can perform better if they are selected so that thermal
conductivity is kept as high as possible with their flow resistance maintained
low. These two requirements are contradictory. Therefore, an optimization problem
has to be solved [7,9,10]. The variables are the dimensions of the individual struts
and their spacing in the example given. A denser arrangement of metal naturally
gives rise to a better heat conduction but also increases pressure drop. The opti-
mum determined is represented by a certain area in the plane of the diagram
spanned by the heat dissipation and the pressure drop, both expressed as dimen-
sionless quantities, in Fig. 7.2-3.
Another application field for open-cellular materials is transpiration cooling. The
high surface area, low flow resistivity, and good thermal conductivity of some of the
materials used make them promising candidates for such purposes.
Figure 7.2-2. Pressure drop in partially open-porous cellular solids of the type shown in the
middle of Fig. 7.2-1 [8].
Figure 7.2-3. A diagram of an

open-cell metal used as a heat
dissipation medium for cooling
high-power electronic compo-
nents. Also shown is the trade-
off between pressure drop and
heat flux, with the preferred ma-
terial domain indicated [7].
7.2.5
Supports for Catalysts
The effectiveness of catalysis critically depends on a high interfacial surface area

between the catalyst and the gases or liquids to be reacted. Therefore, the catalyst
is either processed into a highly porous structure or, if this is not possible, applied
to another porous system such as a porous ceramic material. Cellular metals could
replace such ceramics even if they cannot compete with them on surface area be-
cause they exhibit other useful properties such as high ductility and thermal con-
ductivity. One application concept includes the preparation of a thin sheet of corro-
sion-resistant metal foam, filling this foam with a slurry containing the catalytic
substance, by rolling for example, and finally curing at elevated temperatures
[11]. The resulting catalyst has good mechanical integrity: even after many tem-
perature cycles the catalyst is not separated from the metal foam support. One ap-
plication for such catalysts is for removing nitrogen oxides NOx from the exhaust
fumes of power plants.
7.2.6
Storage and Transfer of Liquids
One of the oldest applications of porous PM materials is as self-lubricating bear-

ings in which oil is stored in the interstices between particles and is allowed to
flow slowly out, thus replacing the used oil. Of course, some of the cellular materi-
als described in this book could fulfil the same function but with the advantage of
having a higher storage volume than traditional PM parts. The application is not
limited to oil: water can be kept and slowly released for automatic humidity con-
trol. Perfume can be stored and allowed to evaporate slowly. Porous rolls can
hold and distribute water or adhesives to surfaces. Transport of the liquid can be
driven by capillary action alone or by excess pressure in the roll. Finally, very
open metallic structures can be used to store fluids at a constant and uniform
temperature, for example, in cryogenic conditions. Moreover, the foam can
reduce undesired movements of the liquid in partially filled tanks (ªanti-sloshingº)
[10,12].
7.2.7
Fluid Flow Control
Porous materials can be used for controlling the flow of liquids and gases [1]. It is
known that PM flow restrictors are more reliable and accurate than conventional
micrometering valves. Because so many degrees of ªopennessº of cellular metals
are available, one could find tailored solutions for even more applications by appro-
priately selecting a cellular metal. Metal foams have already been used as flow
straighteners in wind tunnels [12] or flow distributors in valves [13].
Figure 7.2-4.Different types of silencers (ap-

proximate diameter of components is 10 cm).
7.2.8
Silencers
Components for damping sound, pressure pulses, or mechanical vibrations are

also common in industrial PM applications [1]. Materials with a certain degree
of open porosity can be tailored to damp some frequencies selectively while they
pass others. Sudden pressure changes occurring in compressors or pneumatic de-
vices, for example, can be damped with porous sintered elements. Materials such
as the investment cast foams described in Section 2.3 or the foams made by deposi-
tion (Section 2.4) could replace such traditional elements for reasons of cost and
effectivity [13]. Figure 7.2-4 shows some silencers. The maximum diameter of
the components shown is about 10 cm.
7.2.9
Spargers
Some applications require that a gas be introduced into a liquid homogeneously

and at a constant rate. An example of such an application is the carbonation of bev-
erages. This operation requires a porous part that creates sufficiently small gas
bubbles and satisfies criteria such as corrosion, heat, or shock resistance. Porous
metals can be a superior solution compared to other materials such as porous cera-
mics.
7.2.10
Battery Electrodes
Lead foams could serve as supports for the active material in lead acid batteries in
replacement of conventional lead gratings, thus allowing for constructing very
light electrodes [14]. The electrochemically active mass, a paste containing very
fine lead oxide powders, could be filled into the open voids of a lead foam
where it would contact the electrolyte (sulfuric acid). The lead foam acts as a highly
conductive lattice leading to a low internal resistance of the battery. The nickel
foams described in Section 2.4 are already used as electrodes in pasted rechargeable
NiCd or NiMeH batteries where weight savings and a higher energy density can be
achieved [15 17]. Porous PM materials with their extremely high surface area are
being used in fuel cells [1,15].
7.2.11
Electrochemical Applications
Nickel foams can be used as electrode material in electrochemical reactors. In fil-

ter-press electrodes, for example, a stack of isolated metal plates is used. The plates
are separated by a turbulence-promoting plastic mesh and insulating membranes.
If these meshes are replaced by sheets of cellular nickel with open channels (with
each sheet attached to one of these plates), one increases the electrode surface
while maintaining the turbulence promotion [18]. The reactors can be built
more compactly this way. Nickel foams can also be used to improve electrocatalytic
processes such as the electrooxidation of benzyl alcohol assisted by NiOOH, which
is electrogenerated on nickel anodes. Packed beds of nickel foams were shown to
improve the performance of such reactors [19].
7.2.12
Flame Arresters
Cellular metals with high thermal conductivities of the cell-wall material can be
used to stop flame propagation in combustible gases. Open-cell foams of the
type described in Section 2.4 have been shown to be capable of arresting flames
even when they were travelling at velocities up to 550 m/s. In practice, long
runs of pipes transporting combustible gases are protected close to possible
sources of ignition so that, if ignition does occur, the flame cannot accelerate to
high velocities [13].
7.2.13
Water Purification
Cellular metallic materials could be used to reduce the concentration of undesired

ions dissolved in water. In this application the contaminated water flows through a
highly porous cellular metal with an open structure. The ions react with the matrix
metal of the cellular structure in a redox reaction. An electroless reduction of
Cr(VI) ions by cast aluminum foams has been investigated [20]
7.2.14
Acoustic Control
A sound-wave control device can be obtained if one creates a lens- or prism-shaped

part from a rigid open-cell material, such as a metal foam. The sound waves will
then be guided and redirected by this acoustic device [21]. Moreover, closed-cell
320 7.3 Machinery Applications
foams have been studied for their suitability as impedance adapters for ultrasound
sources. Noise reduction functions are described in Section 5.3.
References
1. M. Eisenmann, in Metal Powder Technologies Metal Foams: A Design Guide, Butterworth-

and Applications, ASM Handbook Vol. 7, ASM Heinemann, Oxford 2000.
International, Materials Park, OH 1998, 11. H. Swars, German Patent Application
p. 1031 1042. 36 19 360, 1987.
2. P. Neumann, in Metal Foams and Porous 12. ERG Inc., Oakland, USA, Product infor-
Metal Structures, J. Banhart, M. F. Ashby, mation of ªDuocelº 1996, and http://www.er-
N. A. Fleck (eds), MIT Verlag, Bremen 1999, gaerospace.com
p. 167 170. 13. SEAC International BV, Krimpen, Nether-
3. W. R. Johnson, M. Shenuski, Machine Design lands, Product data sheet of ªRecematº 1998,
1987, Jan, 89 91. and http://www.seac.nl
4. W. U. Bende, Guo Fuhe, Advances in Powder 14. J. Banhart, German Patent Application
Metallurgy and Particulate Materials, Vol. 6, 100 15 409, 2000.
J. M. Capus, R. M. German (eds), Metal 15. Inco Ltd., Canada, Product data sheet of
Powder Industries Federation, Princeton, NJ ªIncofoamº 1998 and http://www.inco.com
1992, p. 145. 16. V. Ettel, in Proc. NiCad `98, Prague 21 22
5. K. R. Wheeler, M. T. Karagianes, K. R. Sept 1998.
Sump, in Titanium Alloys in Surgical Implants, 17. I. Matsumoto, T. Iwaki, N. Yanagihara, US
H. A. Luckey, F. Kubli (eds), ASTM, Phila- Patent 4 251 603, 1981.
delphia 1983, p. 241 254. 18. A. Montillet, J. Comiti, J. Legrand, J. Appl.
6. W. Frischmann, European Patent Applica- Electrochem. 1993, 23, 1045 1050.
tion EP 0 666 129, 1995. 19. P. Cognet, J. Berlan, G. Lacoste, P.-L. Fabre,
7. A. G. Evans, J. W. Hutchinson, M. F. Ashby, J.-M. Jud, J. Appl. Electrochem. 1996, 26,
Prog. Mater. Sci. 1998, 43, 171 221. 631 637.
8. J. Banhart, Aluminium 1999, 75, 1094 1099. 20. J. G. Ibanez, A. Fresan, A. Fregoso, K. Ra-
9. T. J. Lu, H. A. Stone, M. F. Ashby, Acta. jeshwar, S. Basak, Proc. Electrochem. Soc.
Mater. 1998, 46, 3619 3635. 1995, 12, 102 108.
10. M. F. Ashby, A. Evans, N. A. Fleck, L. J. 21. K. Iida, K. Mizuno, K. Kondo, US Patent
Gibson, J. W. Hutchinson, H. N. G. Wadley, 4 726 444, 1988.
7.3
Machinery Applications
Th. Hipke and R. Neugebauer
Cellular metals have existed as particularly innovative new materials for several
years [1 3]. Since the development of foaming technology, a number of prototypes
for use in the car and machine tool industries have been demonstrated [4 6].
This Section describes possible applications of foamed metals in mechanical en-
gineering.
7.3.1
Parameters
For applications in mechanical engineering, design engineers need the necessary

information to enable them to dimension cellular components. These design para-
meters are required for composites such as sandwich panels, foamed profiles, or
special combinations of foamed and solid materials, as well as for components
completely made of foamed metal. We describe two examples in which powder-
compacted foamed aluminum was used.
7.3.1.1 Thermal Behavior

To evaluate the thermal characteristics of metal foams, the response of sandwich
sheets with a foamed aluminum core to thermal impacts was investigated. The
sheets (400 mm q 300 mm) were clamped on one side and warmed by a heating
mat placed on the top cover plate.
The room temperature was kept constant at 20 hC by a thermal cell. The sheet
temperature was recorded by two probes, one attached to each of the bottom
and top plates.
The thermal expansion was recorded by path sensors and was transformed into
the resulting linear expansion coefficient. The expansion coefficients of foamed
sheets covered with aluminum range from 0.020 to 0.024 mm/K mm. These values
are slightly higher than those of the AlSi12 alloy (0.0199 to 0.0206 mm/K mm) that
had been foamed. The face plate expansion coefficient is 0.0235 mm/K mm, so the
cover plates dominantly influence the total expansion. The lower expansion coeffi-
cient of steel (0.012 mm/K mm) also diminishes the resulting strain on the sand-
wich sheet (see Fig. 7.3-1). The expansion coefficients of the plates covered by
steel sheet range from 0.0105 to 0.014 mm/K mm.
0,030
Coefficient of thermal expansion
0,025
x-direction
0,020 z-direction
[µm/(K*mm)]
0,015
0,010
0,005
0,000
Alu Alu Steel Steel
1 mm 2 mm 1 mm 2 mm
Figure 7.3-1. Expansion coefficients as a function of cover plate material and plate thickness for
a two-face sandwich (density 0.65 g/cm3; sandwich 25 mm).
7.3.1.2 Pull-Out Strength of Detachable Joints

Foamed parts have to be joined to other components and assemblies. One essential
fastening is the screwed joint. A study carried out at the Fraunhofer Institute for
Machine Tools and Forming Technology (IWU Chemnitz) determined the tensile
load limits of such a joint.
The failure behavior depends on the type of joint. For adhesive joints, there is a
huge growth in force. The slope of the curve reflects the elasticity of all the inte-
grated components such as the tensile testing machine, the tongs, the connection
element, and the foam strip. For high-density foams, the joint pulls out after the
adhesive bond has failed, but for lower-density foams the foam structure gives
way. For screwed elements, the force-path progression includes a wide range of
plastic deformation before the maximal force is achieved. These connections also
pull out in several steps, each interrupted by the interlocking of cell walls.
It was shown that in some cases very high forces can be transmitted. The foams
vary widely in their ability to transmit large forces at a screwed joint. Figure 7.3-2
shows that screwed joints have a smaller load carrying capacity than glued and fas-
tened joints. At a metal foam density of 0.7 g/cm3, which is relatively low, the pores
are quiet big and with partially open cell walls, so that the adhesive fills the entire
environment of the joint. Thus, an entirely new material is created locally, which
has a large surface area so the applied force can be well distributed.
The advantages of the adhesively bonded joints are not apparent at higher foam
densities. With high-density foam the adhesive cannot flow through the material
into the pore structure because the cell structure collapsed on drilling. In this
foam the screwed fastenings intersect enough material to make a loadable thread
Sheet metal screw, screwed
Wood screw, screwed
Recoil, screwed
Hexagon screw, screwed
Hexagon screw, glued
Hexagonal nut, glued
Knurled nut, glued
Sonic Lok 862, glued
Ensat 302, glued
Figure 7.3-2. Maximal forces for different connecting elements (diameter M6, Al foam density
0.7 g/cm3).
shape. The joint length with the screws was always 20 mm, but the nuts and in-
serts had a joint length that ranges between 6 and 20 mm.
7.3.2
Examples of Application
7.3.2.1 Foamed Steel Pipes

To make stiff, light structures hollow profiles of various shapes and dimensions are
often used. To make such hollow sections very stiff, very large cross sections are
required, but then the walls tend to oscillate. Filling hollow sections, or pipes,
with metal foam is one possible way of influencing frequency and damping vibra-
tions.
The following pipe profiles were used, being rectangular steel pipes of various
cross sections and lengths, with different wall thicknesses:
Height q Width q Wall thickness
160 q 80 q 4 mm length: 2000 mm
160 q 80 q 3 mm length: 1000 mm
160 q 80 q 4 mm length: 1000 mm
160 q 80 q 5 mm length: 1000 mm
120 q 60 q 4 mm length: 1000 mm
80 q 40 q 4 mm length: 1000 mm
The purpose of these investigations was to find out about damping as a result of
foaming of pipe profiles with aluminum. The influence of the foams on the fre-
quency spectrum can also be clarified. The investigation was performed to find
out how far the natural frequencies are shifted by foaming, and to determine
the dominant frequency and the sequence of the natural frequencies. For the de-
sign engineer, it is important to know whether the first natural frequency causes
a locally limited transverse vibration or a bending vibration of the entire structure.
Reference values were determined by finite element (FE) analysis and measured
on the hollow profiles. To minimize the influence of the fixture, at the nodal points
(the Bessel points), the pipes were laid on two thin wires (see Fig. 7.3-3). This in-
stallation is more advantageous than a one-end fixture.
Vibrations were excited by hitting the pipe on a free end using an impulse ham-
mer. The attenuation of the freely damped vibration was recorded by an accelera-
Figure 7.3-3. Sketch of the beam for vibration studies.

tion sensor. From this curve, we obtain the logarithmic damping decrement and
consequently damping, D
!
X^1 1
D ln (1)
X^z 2p(z s 1)
Natural Frequencies and Characteristic Types

The simulation runs showed that the frequency spectrum depends on all three
pipe geometry variables: cross section, length, and wall thickness. Consequently,
the first natural frequency of long pipes is not necessarily lower than for shorter
pipes. Assuming identical cross sections, pipes with thin walls are characterized
by lower natural frequencies (see Table 7.3-1).
The first natural frequencies are mostly bending vibrations in the x direction,
which is the direction of minimum area moment of inertia. In two pipes, the
first frequency is a local vibration of the pipe wall. This only occurs if the wall is
very thin or if the wall-thickness to cross-section ratio is small. Such wall vibrations
are almost excluded by the metal foam, since the transverse motions are hardly
hindered. For that reason, these frequencies are shifted towards higher values
(Table 7.3-2). For bending vibrations, frequency can only be improved if the stiff-
ness grows more than the weight due to foam filling, but this cannot be achieved
for thick-walled pipes.
As shown, the first bending frequency for the hollow pipe appears as the 5th
frequency only. For the foam-filled pipe, wall vibration can totally be avoided and
only appears in conjunction with bending vibrations of the entire pipe at higher
frequencies.
The frequencies determined experimentally were in good agreement with the
simulation values, which is crucial for damping measurement. If the measured
Table 7.3-1 Natural frequencies of hollow profiles and foam-filled profiles. Vibration type: W
local oscillation of pipe walls; X oscillation in x direction (direction of minimum area moment
of inertia).
Pipe type Hollow profile Foam filled

Cross-section Length First natural Character- First natural Character-
[mm] [mm] frequency [Hz] istic type frequenzy [Hz] istic type
160 q 80 q 4 2000 151.3 X 134.4 X
160 q 80 q 3 1000 324.9 W 493.4 X
160 q 80 q 4 1000 432.7 W 506.3 X
160 q 80 q 5 1000 531.2 X 513.8 X
120 q 60 q 4 1000 430.4 X 397.1 X
80 q 40 q 4 1000 290.8 X 271.9 X

Table 7.3-2. Frequencies and characteristic types in comparison: pipe 160 q 80 q 3, length
1000, foam-filled/hollow.
Frequency Hollow profile Foam filled

no.
Frequency Characteristic type Frequency Characteristic type
[Hz] [Hz]
1 325.0 Walls oscillate 493.4 First bending in x
2 369.6 Walls oscillate 805.7 First bending in y
3 459.2 Walls oscillate 1135.0 First torsion
4 476.6 Walls oscillate 1222.5 Second bending in x
5 479.0 First bending in x 1879.5 Second bending in y
6 530.8 Walls oscillate 2102.5 Second bending in x

whereby the walls are
slightly buckled
damping values are to be integrated into the simulation runs, it is of particular im-
portance that the calculated and the measured values coincide. This is the only way
making sure that the measured damping values can be used for the calculation of
component dimensions
Damping of Structure
Damping was determined at the same time as the frequency was measured. Con-
trarily to expectations, the location of the supports did not influence the frequency
and the damping parameters. This is because, in a dynamic understanding, the
ropes used are soft. So the ropes would only influence a much lower frequency
range than that investigated here.
Having kept all the other conditions constant, even the scanning rate was with-
out any obvious influence: when comparing scanning procedures at 2.5 and
1.25 kHz, the difference between both damping values was merely 0.46 %.
The evaluation showed that the damping factor was slightly higher for the first
vibrations than for more attenuated vibrations. In one experiment, damping be-
tween peak 200 and 700 was D 0.66 q 10 5, and between 1000 and 1500, it
was D 0.64 q 10 5. This effect is caused by the higher damping influence of
the surrounding air for high amplitudes. To minimize this influence of about
3 %, the evaluation range was selected as widely as possible.
In the follow-up investigation of pipes, the damping values shown in Fig. 7.3-4
were determined for the corresponding frequencies. In the investigation, the de-
pendency of damping on the pipe cross sections or the wall thickness could not
be recognized. Instead the damping values are related to the natural frequency.
It makes sense to compare the damping of equal pipe frequencies before and
after foaming.
800
Growth of damping [%]

700
Tube 1
600
Tube 2
500
400
300
200
100
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
Test no.
Figure 7.3-4. Damping growth of the individual pipes (at 300 % change in mechanics assumed).
The increase in damping by foaming varied: the mean change ranged from 133
to 645 %, or 799 % for actual, rather than mean, values. Figure 7.3-4 shows that the
majority of measured values are located below 300 % (see line); few values lie
above, but these are notable. We suppose that the damping mechanisms that
acted in both types of pipes (above and below the 300 % line) are different. For
the types below the line, material damping is the active mechanism, assuming
that the foamed cores and the steel pipes do not move relative to each other,
thereby excluding any frictional or joint damping. These conditions are obtained
for an optimal foaming procedure with clean steel pipes, so a direct metal connec-
tion between the steel pipe and the foam core is brought about, without any gap
that would permit a relative motion.
If the steel pipes are slightly rusty, then metal-to-metal contact between the pipe
and the foam core may be incomplete, and some frictional damping may then
occur. Frictional damping depends on many factors such as pressure, surface
roughness, the medium in the gap, and others. In pipes lying above the 300 %
limit, a combination of material and frictional vibration can be assumed.
We cut sample sections from the foamed pipes, which confirmed that in some
there was good metal-to-metal contact and in others there was a visible gap
between the foam and the pipe wall. It is too simple to classify the foamed
pipes into ªgoodº and ªpoorº categories, but the following statements can be pos-
tulated.
x For foam-filled pipes in which the foam adheres to the inner steel wall in an op-
timum manner, damping is increased by up to 300 % in comparison with hollow
pipes.
x Damping beyond this limit suggests that the foam has not been completely
bonded. This phenomenon could be used as the basis of a nondestructive quality
inspection of the foaming process.
Comparing the measured damping values with those taken from literature, we
note that a damping for steel of D 3.7 q 10 4 [7] is in the range of our steel
pipes. The values for metal foam range from D 9 q 10 4 to 2.1 q 10 3 and
are clearly higher than those of steel. Taking both materials together, a
higher total damping should be provided. However, a direct comparison is not
possible, since damping does not only depend on the material and load, but also
the shape of the specimen [8]. For that reason, the damping measured on a work-
piece is sometimes defined as the workpiece vibration instead of the material
vibration.
7.3.2.2 Machine Table

On the occasion of the CPK `98 conference, the first functioning assembly filled
with metal foam was demonstrated. The part size was 800 mm q 720 mm q
250 mm, so it can be justifiably called an assembly relevant for practice. A conven-
tional machine table (Fig. 7.3-5) on the linear motor test bench of the IWU in
Chemnitz was replaced a steel framework filled with aluminum foam, optimized
by FE analysis. The static and dynamic properties of the conventional assembly,
the steel framework, and the foam-filled framework were measured (Fig. 7.3-6).
The main goal was to reduce weight whilst maintaining stiffness and strength.
The weight of the foamed structure was 72 % of the conventional table (Fig. 7.3-7),
but deflection only decreased by 19 %. These results were obtained by increasing
the table thickness and having a higher area moment of inertia. The static behavior
of the foamed table is much better than the hollow steel frame, because of the im-
proved stiffness and hence minimal deflection (Fig. 7.3-8).
Figure 7.3-5.Original design of machining

table made of steel.
Figure 7.3-6. Design of hollow steel profile

to be filled with Al-foam (800 mm q 720 mm
q 250 mm).
500 Figure 7.3-7. Comparison of weights.

400
Mass [kg]
Solid table
300
Stell frame
200
Foamed table
100
25
20
Deflection [µm]
Solid table
15
Steel frame
10 Foamed table
5
Figure 7.3-8. Comparison of maximum
0 deflection.
600
Natural frequency [Hz]
500
400 Solid table
300 Steel frame
200 Foamed table
100
Figure 7.3-9. Frequencies of the first charac-
0 teristic type.
Very good results were obtained for dynamic parameters: the first natural fre-
quency was increased by 80 % (Fig. 7.3-9), which is mostly due to wall vibrations
being almost eliminated by the foam. Such vibrations can only be detected at very
high frequencies.
The damping was multiply increased for the foamed version: for the first natural
bending frequency, a 10 times improvement was achieved (Fig. 7.3-10). Increased
damping arises from the material damping of the metal foam and from the proper-
ties of the adhesive that bonds the steel and the foamed parts.
This result is of special interest for applying the technique in mechanical engi-
neering. In present machine tools, assemblies move with enormous accelerations.
To meet the increased demands for manufacturing accuracy, surface quality, and
tool life, the vibrations resulting from deceleration, acceleration, and the manufac-
turing process, have to be damped in an efficient manner.
2
Damping [%]
1,5
Solid table
1 Steel frame
Foamed table
0,5
Figure 7.3-10. Damping of the first natural
0 frequency.
7.3.2.3 Cross-Slide
As a second example, the cross-slide of a high-speed milling machine constructed
from steel and foamed aluminum is shown after fabrication (Fig. 7.3-11). The
cross-slide represents the connecting element between the z slide spindle and
the machine bed.
Obvious improvements compared with conventional castings can be anticipated
from simulations (Fig. 7.3-12). Thus, weight is reduced by 46 %. The deformation
deteriorated by 12 %. This worsening could be fully compensated by increasing the
wall thickness. Even then, an evident mass reduction would remain.
Nevertheless, the first natural frequency is increasing by 35 % in any case, which
means a clear enhancement in the field of mechanical engineering. Apart from
this, the position of the natural frequencies can be selected by varying the foam
density.
Figure 7.3-11. Cross-slide of a high-speed milling machine.
140
120
100
Change [%]
80
135
60 112
40
54
20
0
Mass Total deformation 1. natural frequency
Figure 7.3-12. Comparison of variants: original design (light) and foamed construction (dark).
330 7.4 Prototypes by Powder Compact Foaming
7.3.3
Conclusions
Metal foams are characterized by very useful physical and mechanical characteris-
tics. Examples from mechanical engineering show some beneficial properties of
hollow components and sandwiches when filled the metal foam material and
how to make use of them in innovative products.
References
1. L. J. Gibson, M. F. Ashby, Cellular Solids: 5. H.-W. Seeliger, ªApplication strategies for

Structure and Properties, Pergamon Press, aluminium-foam-sandwich parts (AFS)º in
Oxford 1997. Metal Foams and Porous Metal Structures,
2. J. Baumeister, ªÛberblick: Verfahren zur J. Banhart, M. F. Ashby, N. A. Fleck (eds),
Herstellung von MetallschaÈumenº in Proc. MIT Verlag, Bremen 1999, p. 29
MetallschaÈume, J. Banhart (ed.), MIT, Bremen 6. R. Neugebauer, C. Harzbecker, T. Hipke,
1997, p. 3. ªLeichtbaustrukturen im
3. Alulight ein neuer Werkstoff aus Alumi- Werkzeugmaschinenbauº in Proc. Werkzeug-
niumschaum, Mepura Prospectus, Metallpul- maschinenkolloquium 2000, Bern, CH 21 Nov
vergesellschaft mbH, Ranshofen, Austria 2000.
1995. 7. M. Weck, Werkzeugmaschinen und Ferti-
4. T. Hipke, ªMaschinentisch in gungssysteme, Vol. 2, Konstruktion und Berech-
Leichtbauweiseº in Jahresbericht des Fraun- nung, 4th edn, VDI, DuÈsseldorf 1991, p. 3
hofer-Institutes fuÈr Werkzeugmaschinen und 8. F. Holzweiûig, H. Dresig, Lehrbuch der
Umformtechnik 1998, 1998, p. 56. Maschinendynamik, 3rd edn, Fachbuchverlag,
Leipzig-KoÈln, 1992, p. 52, p. 286.
7.4
Prototypes by Powder Compact Foaming
R. Kretz
When metallic foams were revived by new manufacturing technologies such as pow-
der metallurgy, investment casting, and other processes in the early nineties, the
new material was introduced to the industry and to potential customers in a very en-
thusiastic manner and with great advertising expense. Information was widely
spread by companies and institutes who worked on the development of metallic
foams. Metallic foams became highlights at industrial fairs and at many conferences
researchers, designers, and materials engineers discussed the structures and the
properties of cellular metals that are relevant to industrial application. The foam de-
velopers missed no chance to present their work and to introduce their new product,
although the first material samples were only demonstrators: they were small, had
very irregular pore sizes and pore distributions, and a high degree of scattering in
their properties. They were all ªhand madeº and their producers were very proud
of their creations from porous metals. With the growing knowledge and experience,
manufacturing methods were further improved and very soon the presented foams
were of a respectable quality with a certain reproducibility. Whenever people held the
new material in their hands, they were impressed and almost everybody had a new
idea concerning an application of metal foams.
7.4.1
Introduction
But very soon the discussion came round to real parts or components and then the
foam manufacturers had to think about manufacturing net-shape parts, consider-
ing particular problems of geometry or function. Since real net-shape parts are not
flat plates, cubes, or cylinders, new manufacturing methods using mold and fur-
nace technologies had to be introduced. The manufacturing of the powder metal-
lurgical (PM) precursor material as well as the foaming technique for the com-
pacted powder are called the powder compact foaming technique (PCF), which
is practiced by IFAM, Bremen, Germany, Leichtmetallkompetenzzentrum, Rans-
hofen (LKR), Austria, and the Slovak Academy of Science (SAS), Bratislava, Slova-
Figure 7.4 -1. Crucible melting furnaces.

Figure 7.4 -2. Simple foam parts.
kia. The manufacturing of prototypes for component testing or demonstration is

described in the following pages by examples from a bilateral research project at
LKR.
Step-by-step prototype manufacturing methods were developed, starting with the
foaming of parts in crucible melting furnaces (Fig. 7.4-1) and the use of steel
molds. The geometry of the first aluminum-foam parts was rather simple, as
shown in Fig. 7.4-2.
7.4.2
Methods, Machines, and Molds
From the very beginning the application of the PCF route for making net-shape
parts was the goal. This method, using compacted aluminum powder and titanium
hydride as a foaming agent, is described by several authors [1,2].
7.4.2.1 Manufacturing Methods for Precursor Material

The first step in the PM process is to make foamable precursor material from alu-
minum powder. Small amounts of aluminum powder (50 kg) were mixed by me-
chanical stirring, then poured into aluminum liners, then closed and extruded by a
conventional aluminum extrusion machine. This method was easily performed
using existing facilities and an extrusion plant at AMAG in Ranshofen. However,
the bulky liner material in some cases caused insufficient foaming.
An alternative route for producing precursor material for PCF is the Conform
process (Section 2.1.1.1) [3]. Small amounts of precursor material made in this
way were supplied by Holton Ltd. and trials showed that the Conform route
could be a suitable manufacturing method for strip like dimensions of precursor
material. Generally, the small batches of manufactured aluminum-foam prototype
parts made in this stage of development led to a large scatter in the foam proper-
ties. The reproducibility of the whole process and the properties of the components
were relatively poor. This was due to the handwork involved in producing the parts,
which typically resulted in varying process parameters. However, samples with uni-
form pore sizes and good pore distribution could be achieved. These undesired
property variation could only prevented by a ªseries production likeº manufactur-
ing route of the precursor material. With growing demand for precursor material
and with an increasing volume of foamed prototypes the process flow and the qual-
ity of the prototypes could be improved by an approach to series production condi-
tions. The batch sizes reached 1000 kg and more. Industrial partners, such as Me-
tallwerk Plansee, Tyrol and Peak, Germany carried out the cold isostatic pressing
(CIP) process, and three staff members performed the prototype manufacturing.
Although this process led to good results, the economic disadvantages were the
high transportation costs and varying process parameters caused at different pro-
cess steps in different companies. A further improvement in process flow was
achieved following the installation of a CIP machine, so that all steps from the
powder production to the production of extrusion billets were achieved within
one company. From this day on the powder was mixed with the foaming agent
in an industrial scale impulse mixer and compacted in a special CIP. Then it
was transferred to an aluminum extrusion machine to produce the precursor ma-
terial by this ªconventionalº extrusion route.
The process steps as it is currently performed in PCF are:
x manufacturing of Al powder (alloyed or unalloyed) and grain size classification,
x mixing Al powder with foaming agent (TiH2),
x CIP of the mixture to form foamable extrusion billets,
x heating and extrusion of the billets to form precursor material,
x filling the precursor in steel or cast-iron molds,
x foaming in a suitable foaming furnace (heating rate, temperature above solidus),
and
x cooling and removal of the foamed parts from the mold.
7.4.2.2 Foaming Process

For successful manufacturing of aluminum-foam parts in the PCF, it is necessary to
observe furnace temperature, heating rate, temperature distribution within the
mold, foaming time, and cooling conditions that have to be kept in certain ranges
to achieve aluminum-foam components of high quality. These parameters are not
the same for all types of component; they depend on their geometry and dimensions
(wall thickness), on the alloy, and also on the properties and design of the mold.
In prototype manufacturing it is not possible to modify the whole equipment or
to change the layout of machines whenever a new component with a new geometry
has to be foamed. Compromises concerning the furnace used and the process para-
meters must be made to enable the best possible technical and economic manufac-
turing conditions to be achieved. Nevertheless, it is possible to make parts in a
wide range of different geometries with the use of adequate equipment. From
the beginning of aluminum-foam development, prototype manufacturing was
the main goal. This policy led to a large number of prototype projects producing
components with very different geometries and applications. The development of
the foaming process, applied to a particular component, was carried out along
with component testing to verify material properties.
In series production it would be necessary to adapt the furnace and the mold to
the product to achieve the optimized foaming parameters as well as the most eco-
nomic production conditions for the particular product. The number of parts that
has to be produced is also an important process parameter to be considered in the
design of an aluminum-foam production machine and the necessary additional
equipment.
7.4.2.3 Foaming Furnaces

One of the most important pieces of equipment in the aluminum-foam manufac-
turing route is the foaming furnace. The foaming parameters have to be kept in
their optimized range, the precursor material in the mold has to be heated to
the foaming temperature in the foreseen time, and the temperature distribution
within the mold must be kept within small tolerances to avoid irregular foaming
within the mold.
The heat transfer from the furnace to the mold with the precursor material can
occur by three different mechanisms: conduction, convection, and radiation.
For the development of prototype manufacturing methods various furnaces
working with these three physical principles were tested. Basic calculations con-
cerning heat capacity and heat transfer were carried out before testing the different
furnace systems. The heating rate in the foamable precursor material is the key
parameter, influenced by the foam part dimensions and foaming mold properties.
Considering the described intentions, the very first trials to make simple foam
parts were performed by immersion of the mold into an aluminum melt. To
achieve a high heating rate by intensive heat transfer by conduction, the molds
containing the precursor material were submerged into liquid aluminum that
was kept in a conventional holding furnace. The conductive heat transfer was
very high and a uniform temperature distribution could be achieved. This method
worked quite well and the aluminum-foam samples yielded good properties, but
chemical attack of the molds by the Al melt caused short lifetimes of the molds
and the mold attachments. Furthermore, the handling of the molds in liquid alu-
minum is dangerous and problematic.
High-convection furnaces (air-circulating furnaces) are used in industry for rapid
heating processes. A ventilator in the furnace effects very high heat transfer from
the heat source to the mold. The rapid heat transfer, caused by high flow velocity,
causes high heating rates, but the gas flow is influenced by the geometry of the
mold and that leads to uneven temperature distribution within the mold. This ef-
fect can be avoided by the installation of baffle plates, which must be adapted for
every new mold that is used in the furnace. Therefore air-circulating furnaces can
be employed in series production, where high numbers of aluminum-foam parts
have to be made with the same mold geometry, but they are not suitable for pro-
totype manufacturing, where flexible and universal furnace systems are required.
For prototype manufacturing it is essential to cover a wide range of different
components. For this purpose a special chamber furnace with adequate heating ca-
pacity had to be developed. One example of a foaming furnace especially built for
foaming of big components, is the 100 kW Nabertherm Chamber furnace at LKR,
Figure 7.4 -3. Nabertherm aluminum-foam chamber furnace.

having internal dimensions 900 mm q 900 mm q 1800 mm and a modified con-

trol unit to keep the required temperature distribution tolerances within the cham-
ber. This furnace is shown in Fig. 7.4-3.
7.4.2.4 Foaming Molds

In the PCF process the mold giving the aluminum-foam part its shape contains the
foamable precursor material during the heating cycle. The thermal energy that is
necessary to start the foaming of the precursor material has to be transmitted
through the mold. Consequently the thermal conductivity of the mold should be
as high as possible and the thermal capacity of the mold as low as possible. On
the other hand the molds must be stable enough to keep the required dimensional
tolerances of the net-shape foam part. The first molds for the PCF process were
made from common steel sheets. The very first aluminum-foam parts were very
simple, regular shaped pieces such as small plates, cubes, or cylinders and the de-
velopment was focussed mainly on the foaming behavior of the precursor material
and the production of test samples. But when customers required 3D net-shape
prototypes, the mold became an important part of the prototype development.
The experience of foundry die makers was combined with the experience in foam-
ing precursor material. Conventional casting dies are very heavy and bulky. A lot of
energy is wasted on heating the whole die during the foaming process. The precur-
sor material would not be heated fast enough to achieve sound foam components.
An economic route is to use cast iron, which allows casting of thin walled foaming
dies, using wood or plastic patterns as well as the lost foam casting process. De-
pending on the molding sand used, the surface in the die may require machining.
Table 7.4 -1. Comparison of various mold materials.
Mould material Advantages Disadvantages
Steel easy machining mold lifetime

5 10 mm thick low material costs distortion
heat conductivity
Cast iron heat conductivity high material costs

free-form shapes high production costs
mold lifetime
High-temperature mould lifetime high material costs

tool steel low distortion
Copper heat conductivity high material costs

chemical reactions
Ceramics shapeability heat conductivity

free-form shapes mold lifetime
no distortion
This method was used for foaming car structural parts, the largest prototypes so far
manufactured in net shape.
Besides structural steel and cast iron, high-temperature tool steel, copper, and
ceramics were also used as mold materials in a large number of prototype devel-
opment projects. Table 7.4-1 shows an overview of different mold materials and
their properties.
To avoid sticking of the aluminum-foam product in the molds, the cavity of the
die was coated with common facing materials.
7.4.3
Prototypes and their Applications
It is a long way from the first idea to the first prototype, and it is even further from
the prototype to a series product. Some of the product ideas had to be rejected be-
cause the aluminum-foam properties were over-estimated. The outstanding me-
Figure 7.4-4. Aluminum-foam prototypes.

chanical properties of closed-cell metal foams are their energy absorbing potential
and specific stiffness. Figure 7.4-4 shows a couple of aluminum-foam samples
made for testing mechanical and technological properties of the material.
The feasible applications are spread in a wide range of various industrial fields.
The automotive and transportation industries are interested in metallic foams as
well as mechanical engineering, architecture, and building construction companies.
7.4.3.1 Automotive Applications

Owing to their particular properties, metal foams are considered to be an appropri-
ate material for impact-energy absorbing components. Depending on foam type
(PCF or melt-based foam), alloy, and density, the energy absorbing behavior can
be influenced within a certain range. This property makes metal foams a candidate
Figure 7.4 -5. Energy-absorbing proto-

types.
Figure 7.4 -6. A-column absorbing element.

material for crash-relevant components in cars as well as in trucks, trains, and

tramways. Specific stiffness is also an important property in automotive applica-
tions. The energy-absorbing behavior is described in many papers and presenta-
tions [4 6].
Manufacturing of aluminum-foam prototypes at LKR included a couple of crash-
relevant components that were tested either as specimens or as real components.
The crash elements can be made of aluminum foam only, or in combination with
other materials such as steel tubing or aluminum castings. Figure 7.4-5 shows a
steel tube specimen [7] after performing a compression test to investigate its
energy-absorbing behavior.
A component testing project was carried out with the Austrian car manufacturer
Steyr Daimler Puch [8]. After the original absorbing element in the A-column was
replaced by an aluminum-foam absorbing element, the A-column cover was
mounted and head impact tests were performed. Figure 7.4-6 shows the A-column
absorbing component.
Figure 7.4 -7. Structural

parts.
Figure 7.4 -8. Aluminum-foam front panel
High specific structural parts such as intermediate panels or front panels are fea-
sible applications. Figure 7.4-7 shows a couple of structural part prototypes includ-
ing an intermediate panel [9] that was made for DaimlerChrysler in 1995.
The largest aluminum-foam prototype part made by LKR was a front panel that
was foamed by order of Opel in 1999. The dimensions of this panel, shown in
Fig. 7.4-8, are about 140 cm q 50 cm; it is a net-shape foamed in a cast-iron
mold [10].
7.4.3.2 Construction and Architecture

In constructional and architectural applications, aluminum foams can be applied
as panels for fireproof doors and walls, light and stiff covers of installations, tech-
nical equipment, and other attachments. Architects seem to approve of the appear-
ance of the as-cast aluminum-foam skin. Plate prototype manufacturing is very dif-
ferent from the production of complex shaped aluminum-foam parts. In most
cases, it is difficult to fulfill the required thickness tolerances of the plates. Special
furnace designs and mold materials have been developed at SAS. For all outdoor
applications they must be protected against moisture penetration and corrosion,
which can easily be achieved using common surface treatment processes such
as painting or powder coating. Figure 7.4-9 shows surface-treated aluminum-
foam plates.
7.4.3.3 Other Technical Applications

Two additional groups of feasible systems have a good chance of realization in the
future: aluminum sandwich sheets, consisting of aluminum-foam plates, covered
with high strength aluminum face sheets, and aluminum-foam cores in aluminum
or magnesium castings.
Figure 7.4 -9. Surface treated foam plates.
Up until now sandwich sheets have had to be made in two process steps. A foam
plate is made and the face sheets are applied by adhesives. Manufacturing of sand-
wich parts and foam filled aluminum profiles by forming the foam between two
aluminum sheets is reported in Section 3.3.
Aluminum-foam filled steel profiles can also be used as energy absorbing
components in cars, trucks, trains, and other vehicles. Manufacturing of these
profiles is relatively easy because the steel profile can be used as a foaming mold.
In Fig. 7.4-10 a couple of foam-filled profiles and sandwich specimens are shown.
Aluminum or magnesium castings with permanent aluminum-foam cores
should meet the requirements of improved energy absorbing potential, higher
stiffness than hollow castings and good vibration damping behavior. The use of
foam cores can also improve the manufacturing process in a number of ways: no
sand cores have to be removed from the casting, complicated quasi-hollow shapes
can be cast, and improved solidification conditions can be achieved. Target applica-
tions for castings with foam cores, as shown in Fig. 7.4-11, are suspension and struc-
Figure 7.4 -10. Foam-filled profiles and sandwich parts.
Table 7.4 -2. Feasible aluminum-foam applications.
Aluminum foam Application field/Customer Focus/property

prototype part
Plates construction, architecture weight, design

furniture industry
street furniture
Cylinders, cubes car component suppliers impact energy absorption
Tube physical research (DESY) physical properties: conduc-

tivity, permeability
Foam cores, cubes, ring foundries, cores in castings casting conditions, damping
behavior, weight,
energy absorption,
Table 7.4-2. Continues
Aluminum foam Application field/Customer Focus/property

prototype part
Compact structural parts automotive industry, stiffness, damping, weight

electrical appliance
camera casings
Structural parts intermediate panels, Stiffness, damping behavior

automotive structures
(DaimlerChrysler, Opel)
Structural parts automotive structures Stiffness, weight

sunroofs
Structural parts ships, boats, yachts weight
Sandwich structures transportation, rail vehicles stiffness
Compact structural parts weapons, stock weight
Foam-filled profiles engine mount (CRFiat) crash behavior

structural car components,
rail vehicles, aircraft,
windmills,
Foam-filled profiles machine tool (IWU) damping properties

robot brackets
building industry
Compact structural parts objects of art, design optical appearance
3-dimensional parts sculptures, art optical appearance
tural parts with improved energy absorbing properties, and wheels. Also mechanical
engineering components with good damping behavior and good castability [11].
In addition to the applications already described a large variety of other potential
technical applications of aluminum foam have been discussed. Table 7.4-2 con-
tains an overview of feasible aluminum-foam applications.
7.4.3.4 Improbable Applications

As mentioned earlier, many of the possible applications of aluminum foams were
discussed with researchers and designers who had to solve very specific technical
problems, by use of suitable materials. People were fascinated by metal foams and
sometimes they over-estimated their potential. For instance, aluminum foam was
thought to provide the highest strength combined with lowest weight, which is
not the case. Table 7.4-3 shows a couple of improbable applications and ideas,
which could not be considered as applications for metal foams.
Figure 7.4 -11. Aluminum-foam cores for castings.
A very important task of foam developers and foam manufacturers must there-
fore be to give their customers and potential material users up-to-date and accurate
information about the material properties and convenient applications, but also its
limits. Otherwise they run the risk of losing interest in this material altogether.
Table 7.4 -3. Improbable applications.
Aluminum foam Application/Customer Focus

prototype part
Plates bulletproof jackets, plates safety, military
Plates skateboard, snowboards weight
Plates milk tanks weight

Table 7.4-3. Continues
Aluminum foam Application/Customer Focus

prototype part
Compact structural parts gear box weight
Structural parts musical instruments acoustic properties
Plates, cubes speakers and headphones acoustic damping
Gas-permeable plates plastic industry molds weight
Gas-permeable cylinders filters and catalysts porosity, weight

exhaust system
Gas-permeable cylinders heat exchangers porosity
Foam cylinders latent heat accumulator heat capacity
Foam cylinders, spheres floats density
Foam-filled cylinders drums, rollers textile machines

paper industry
Buttons clothes design
Precursor nails construction fastening device joining by foaming precursor

nails
References
1. H. P. Degischer, U. Galovsky, R. Gradinger, 7. R. Gradinger, M. Seitzberger, H. P. De-

R. Kretz, F. Simancik, in Proc. MetallschaÈume, gischer, M. Blaimschein, Ch. Walch, Metal
J. Banhart (ed.), MIT, Bremen 1997, p. 79. Foams and Porous Metal Structures, J. Banhart,
2. H. Cohrt, F. BaumgaÈrtner, D. Brungs, H. M. F. Ashby, N. A. Fleck (eds), MIT Verlag,
Gers, in Proc. MetallschaÈume, J. Banhart (ed.), Bremen 1999, p. 313.
MIT, Bremen 1997, p. 15. 8. B. GoÈtzinger, R. Kretz, in Vom Werkstoff zum
3. H. P. Degischer, H. WoÈrz, DE Patent Bauteilsystem, Proc. LKR, H. Kaufmann, P. J.
4206303 1992. Uggowitzer (eds), ISBN 3-902092-009, 1999,
4. H. P. Degischer, U. Galovsky, R. Gradinger, p. 185.
R. Kretz, F. Simancik, in Proc. MetallschaÈume, 9. R. Kretz, E. Hombergsmeier, K. Eipper, in
J. Banhart (ed.), MIT, Bremen 1997, p. 79 Metal Foams and Porous Metal Structures,
5. L. Lorenzi, M. Parisi, M. Valente, A. Zanoni, J. Banhart, M. F. Ashby, N. A. Fleck (eds),
in Metal Matrix Composites and Metallic MIT Verlag, Bremen 1999, p. 23.
Foams, Proc. Euromat, Vol. 5, T. W. Clyne, F. 10. R. Kretz, E. Wolfsgruber, in MetallschaÈume,
Simancik (eds), Wiley-VCH, Weinheim 1999, H. P. Degischer (ed), Wiley VCH, Weinheim
p. 96. 2000, p. 400.
6. J. Siebels, in Metal Foams and Porous Metal 11. R. Kretz, H. Kaufmann, in Metal Matrix
Structures, J. Banhart, M. F. Ashby, N. A. Composites and Metallic Foams, Proc. Euromat,
Fleck (eds), MIT Verlag, Bremen 1999, Vol. 5, T. W. Clyne, F. Simancik (eds), Wiley-
p. 13. VCH, Weinheim 1999, p. 63.
346 7.5 Applying the Investment Methodology for Materials (IMM) to Aluminum Foams
7.5
Applying the Investment Methodology for Materials (IMM) to Aluminum Foams
E. M. A. Maine and M. F. Ashby
Innovations in material development are high-risk investments [1], characterized

both by long gestation periods between invention and widespread market
adoption [2], and by uncertainty of ultimate success. For these and other
reasons, they have generally been driven by large enterprises [3] and national
governments. An investment methodology for new materials (IMM) has been
developed [2,4] that could both reduce risk and shorten gestation time. IMM
provides a structured, informed procedure for assessing the attractiveness of
investing in the industrial scale-up of the production of a new material. This
section presents the results of an IMM analysis on the viability of aluminum
foams.
7.5.1
Introduction: The Investment Methodology for Materials (IMM)
IMM consists of three linked segments: viability analysis, market assessment, and
value capture (Fig. 7.5-1). A material is viable in an application if the balance be-
tween its technical and economic attributes is favorable. Assessing viability in-
volves technical modeling of the application, cost modeling of manufacturing,
input from the market assessment, and value analysis. Market assessment deploys
techniques for identifying promising market applications and for forecasting fu-
ture production volumes. Likelihood of value capture is assessed through an
Figure 7.5-1. The investment methodology

for materials (IMM).
analysis of industry structure, organizational structure, IP issues, appropriability,

and the planned market approach. The most desirable investment opportunities
are those with both a large viable market size and a high likelihood of value capture
[4].
7.5.2
Initial Market Scan for Potential Applications for Al Foams
Aluminum foams have several unusual features, among them, exceptional en-
ergy absorbed per unit volume and, particularly, energy absorbed per unit weight [5].
Potential applications are those in which these functions are important; those
that also exploit secondary attributes of metal foams (such as flame resistance,
heat dissipation, noise reduction, water resistance, blast amelioration) are parti-
cularly attractive. The strongest opportunities for short-term take up of alu-
minum foam lie with substitution into existing markets, since for these the
design requirements are well known. Radically new designs taking best advan-
tage of their properties are slowed by lack of designer awareness and confi-
dence, and by current manufacturing difficulties with achieving high toler-
ances.
There are existing aerospace and automotive applications that require high en-
ergy absorption per unit volume. The large-scale production of aluminum foams
cannot yet meet the tolerances required for aerospace applications (other than
the expensive Duocel, which finds a market in defense and space vehicles); thus,
automotive applications are prioritized, specifically, the ªA-pillarº designed for oc-
cupant safety, and the redesigned hood and front bumper regions to meet new pe-
destrian protection legislation.
7.5.3
Material Assessment
7.5.3.1 Technical Performance

The stress strain curve of aluminum foams (Fig. 7.5-2) approximates the ide-
alized stress strain curve for energy absorption, suggesting their use in en-
ergy-absorbing applications. As a screening mechanism to eliminate prototyp-
ing and testing with uncompetitive materials, a range of aluminum foams
was compared with all other materials along combinations of technical para-
meters. IMM utilizes CES (Cambridge Engineering Selector) software for this
purpose [6]. Once a material is established as a non-dominated candidate for
an application, prototyping and testing are required to substantiate technical
viability. This technique is explained in detail in published books and papers
[7,8].
Figure 7.5-2. Stress strain curves of closed-cell Alporas and open-cell ERG metal foams
(Duocell).
7.5.3.2 Cost of Production

Economic analysis aims to establish the manufacturing cost differential between a
component made with a novel process or material and the incumbent processes
and materials. To this purpose, technical cost models were constructed for three
processes for manufacturing aluminum foam.
x Liquid-state foaming of aluminum,
x TiH2 expansion via powder metallurgical processing as a batch process,
x TiH2 expansion via powder metallurgical processing as a continuous process.
The cost models themselves include sub-models to capture the effect of component
size on production rate, the die and equipment costs, allowance for scrap, die-life,
Figure 7.5-3. Production costs for processing aluminum foam by three methods.
and capital write-off. The output of the models (Fig. 7.5-3) shows the way in which
the cost of the manufactured component depends on production volume. Cost dri-
vers can also be determined. In this instance the batch powder metallurgical pro-
cess is the most economical option for annual production volumes of up to 20 000
parts [2,9].
By comparing the cheapest aluminum foam component (at the required produc-
tion volume) with an incumbent material for any given application, the cost differ-
ential between a new and an incumbent material solution can be established. This
provides input into the value assessment.
7.5.3.3 Co-Minimizing Volume and Cost in Energy-Absorbing Applications

In most applications cost is a strong consideration and this raises the issue of co-
minimizing two objectives: volume and cost (or mass and cost). Used as an energy
absorber in the form of a simple panel or slab application, the cost C of the fin-
ished block of foam is essentially that of the foam itself, CmM, where Cm is the
cost per kg of the foam and M Vr is the mass of the panel, where r is the
foam density and V is the panel volume. Modeling the production cost of shaping
is discussed elsewhere [9].
The co-minimization is best done by constructing trade-off plots. This method
reveals that the cheapest solution is polystyrene foam with a density of 0.05 Mg/
m3, but the volume (and thus thickness) required is large. The solutions with
the lowest volume are offered by the Hydro (liquid-state) or the Alulight (PM) alu-
minum foams, both with densities of about 0.25 Mg/m3 (Fig. 7.5-4).
Figure 7.5-4. Selection of metal foams and polymer foams by volume and cost with s pl 3.5 MPa.
7.5.4
Market Forecast
A technical assessment indicates that metal foams are technically viable as energy
absorption elements within the A-pillar [8]. The next step is to examine the market
size and rate of take-up.
7.5.4.1 Market Size for Aluminum Foam

A pair of A-pillar energy absorbers made of aluminum foam weighs about 1 lb
(0.4 kg). Assuming 1 lb of aluminum foam per car for a mass-produced passenger
car at a finished cost of $10/lb ($25/kg) allows an estimated market size of $5 mil-
lion annually if installed in a single, large-volume model of passenger car; if it
becomes the industry standard, the market size increases to some $100 million
annually. The substitution mode is either lower cost/higher performance, or higher
cost/higher performance. The former mode would be followed when the displaced
energy absorption element is complex, where minimizing volume is a priority, and
when flammability is a concern, eliminating most polymer foams. In some other
A-pillar designs, the Al foam would be replacing a cheaper existing solution that is
unable to meet new occupant safety legislation.
7.5.4.2 Market Timing for Aluminum Foams

A method of comparison with historical substitutions, such as Abernathy and
Clark's transilience map [10], aids a substitution-timing forecast. Aluminum
foam for energy absorbing automotive applications is a revolutionary innovation,
in that it overturns established technical and production competencies, but does
not overturn customer linkages nor require a company to sell into different mar-
kets. Thus the substitution timing of aluminum foam for A-pillars can be modeled
on the historical examples of a revolutionary innovation in the automotive industry
that was both lower in cost and higher in performance. Historical examples of
these include SMC (sheet molded compound) hoods (bonnets) and polymer com-
posite fenders. SMC hoods are lighter than those made of steel, and are cheaper to
design and manufacture for low-volume platforms; substitution has occurred for
these and for vehicles that are close to the legislated CAFE (corporate average
fuel economy) limit, where weight saving is highly valued [11]. Polymer composite
fenders are lighter than those made of steel and better able to resist impact without
damage. For these, substitution has followed a similar curve to that for SMC hoods
[12]. If aluminum foams in energy absorption automotive applications follow a
similar substitution pattern, they could expect to capture 50 % of their potential
market within 12 years (see Fig. 7.5-5).
Automotive platforms with a flexible Body-In-White structure exact less demand-
ing constraints on the A-pillar energy absorber. For these, metal foams, with their
ease of shaping, offer aesthetic design improvements (for example, thinner A-pil-
lars with improved lines of sight), but cost more than cheaper alternatives: a higher
Figure 7.5-5. Scenarios for aluminum foam substitution into A-pillar market.
cost/higher performance mode of substitution. Historical examples of these in-

clude aluminum alloy wheels and air bags. With aluminum wheels, a combination
of aesthetic qualities and lighter weight drove substitution in sporty platforms to a
50 % take-up after 18 years [13]. In the case of airbags, consumer safety concerns,
the viability of the innovation, and automotive regulation drove substitution much
more quickly [13]: an unlikely scenario in the case of metal foams in the A-pillar.
We conclude that, if all other conditions are favorable, the time to 50 % take-up
would be in the range 12 18 years.
7.5.5
Value Capture
In this investment assessment, we consider the position of a small firm, producing

medium quality, medium priced, 3D, aluminum foam parts by a powder metallur-
gical process. The viability assessment and market assessment indicate a potential
future market for aluminum foams in the automotive market starting at $5 million
per annum and going to $100 million per annum. In this section, tools to assess
industry attractiveness, appropriability, and organizational structures are utilized to
predict the likelihood of the small companies who are at present in the process of
commercializing the material capturing the value created by this innovation.
7.5.5.1 Industry Structure

Porter's methodology for assessing industry attractiveness [12] directs attention to
the competitive threats and to buyer/supplier pressures that might reduce value-
capture by the innovating firm. In the present context we find that several compa-
nies are able to make metal foams and that none of them has tight control of the
• Few potential entrants but

weak IP position
Potential
• Substitutes for metal foams: Entrants
polymeric foams, shaped
aluminium sections, fibre
reinforced polymer
composites, wood Competitive
Suppliers Buyers
Rivalry
• Strong buyer power in most
mass applications (e.g.,
automotive)
• Overall, medium-low Substitutes

attractiveness
Figure 7.5-6. Porter's ªfive forcesº as a means of assessing industry attractiveness.
rights to the process, which is not difficult to reproduce. Their position is made
less attractive because they must fight both against competitors who are commer-
cializing alternative processes for making metal foams (the lower end liquid alumi-
num foam process, for instance) and against substitutes for energy absorption in
automotive applications, such as polymeric foams, shaped aluminum sections, and
fiber-reinforced polymer composite sections. Additionally, the automotive compa-
nies exert very strong buyer power that will be difficult for the small producer to
counter (Fig. 7.5-6).
7.5.5.2 Appropriability of Profits

Teece's concept of appropriability [14] is useful in establishing the potential for
value capture. Here we seek to establish the appropriability position of a small
company seeking to sell metal foam automotive components. The intellectual
property (IP) position does not appear to be strong. Specialized assets exist with
the process, but most could be assembled by a die-casting competitor without ex-
cessive difficulty. There is a possibility for co-specialized assets, if certain methods
of automotive design propagate. Aluminum foam for energy absorbing automotive
applications is a revolutionary innovation, in that it overturns established technical
and production competencies, but does not overturn customer linkages nor require
a company to sell into different markets. New product cycle time is slow, giving a
longer period over which to appropriate value, and the structure of the automotive
industry allows for protection of IP. The conclusion is that the appropriability posi-
tion is not strong, but might be classified as medium.
7.5.6
Conclusions: Applying IMM to Aluminum Foams
x Target market: energy absorption in automotive applications: the A-pillar, and

front-end pedestrian protection.
x Viability: yes, in energy absorption applications in the automotive industry. No in
pedestrian protection applications currently.
x Market assessment: market size of up to $100 million annually, with 50 % take-
up in 12 18 years.
x Competitive position: poor, uncertain IP protection, vulnerable to extreme buyer
power.
x Value capture: medium to low chance.
As the chances of value capture are relatively low and the payback period is rela-
tively long, a small firm would be disadvantaged in commercializing this innova-
tion unless as a joint venture with a metal supplier or automotive producer. A lar-
ger company might be interested in pursuing this opportunity if it was in a good
position to capture the value created. Alternatively, a government-sponsored initia-
tive for pedestrian or occupant safety might subsidize the commercialization of
such an innovation.
Acknowledgements
We wish to acknowledge the support of the KoÈrber Foundation, the Cambridge Ca-
nadian Trust, the UK EPSRC through the support of the Engineering Design Cen-
tre at Cambridge, and Granta Design, Cambridge, who developed the software
illustrated in this article.
References
1. D. Wield, R. Roy, ªR&D and corporate stra- 5. M. F. Ashby, A. Evans, N. Fleck, L. J. Gibson,
tegies in UK materials-innovating J. W. Hutchinson, H. N. G. Wadley, Metal
companiesº Technovation 1995, 15(4). Foams: A Design Guide, Butterworth-Heine-
2. E. M. A. Maine, ªInnovation and Adoption of mann, Oxford 2000.
New Materialsº PhD Thesis, Cambridge 6. CES, The Cambridge Engineering Selector,
University 2000. Granta Design, Cambridge 1999, http://
3. C. Freeman, L. Soete, The Economics of In- www.granta.co.uk
dustrial Innovation, 3rd ed, Pinter, London 7. M. F. Ashby, Materials Selection in Mechanical
1997, p. 237. Design, Pergamon Press, Oxford 1992.
4. E. M. A. Maine, M. F. Ashby, ªAn invest- 8. E. M. A. Maine, M. F. Ashby, ªAn invest-
ment methodology for materials. Part 1: The ment methodology for materials. Part 2: Ap-
methodologyº Mater. Design, Elsevier 2002 plications of the methodologyº Mater. Design,
in press. Elsevier 2002 in press.
9. E. M. A. Maine, M. F. Ashby, ªCost estima- 12. M. E. Porter, Competitive Advantage: Creat-

tion and the viability of metal foamsº in Proc. ing and Sustaining Superior Performance, Har-
Int. Conf. Metal Foams and Porous Metal vard Business School Press, Boston, MA
Structures, MIT-Verlag, Bremen 1999, Vol. 1, 1985.
p. 63 70. 13. R. Grant, Contemporary Strategy Analysis,
10. W. J. Abernathy, K. B. Clark, ªInnovation: Blackwell, Oxford 1998, p. 245.
mapping the winds of creative destructionº 14. D. J. Teece, ªProfiting from technological
Research Policy 1985, 14, 3 22. innovation: implications for integration,
11. E. M. A. Maine, ªFuture of Polymers in collaboration, licensing and public policy.º
Automotive Applicationsº Master's thesis, Research Policy, 15, p. 285±305. (Elsevier)
MIT 1997. 1986.
8
Strengths, Weaknesses, and Opportunities
H. P. Degischer
The research interest in cellular metals and the activities of manufacturers in var-
ious market sectors illustrate the appeal of these materials: they are attractive be-
cause of their unusual specific properties and appealing property profiles. Expecta-
tions for so-far unachievable metallic light-weight structures have created ideas for
the substitution of existing materials in weight-sensitive components such as
floating metals, light stiff panels for automotive and aerospace applications, crash-
absorbing safety components, electromagnetic shielding, structural components
with low absorption for nuclear particle radiation, housings with high acoustic
damping, and decorative design objects. Many different markets have been
addressed, but how long will it take for a market breakthrough? Tables 8-1 8-3
resulted from a discussion at the end of the Symposium ªMetallschaÈumeº in
Vienna, 29 February 2000. The tables show the advantages of cellular metals and
the achievements of research as well as their disadvantages, and the remaining un-
answered questions: i. e. cellular metals' strengths, weaknesses, and opportunities.
8.1
Processing
The formation of porous aluminum in welding zones of conventionally pressure

die-cast components is well known, but very much unwanted, requiring special ef-
forts to be suppressed. The controlled production of porous structures and the ar-
chitecture of cellular metals appeared to be complicated. The processing of semi-
products and shaped parts requires additional efforts resulting in higher cost of
production for cellular structures than for the parent metal. Even more complex
processing techniques have to be developed to improve the regularity of cell struc-
tures with reasonable reproducibility. A rule of thumb became apparent: that the
price per unit volume of the material remains at least the same! Therefore specialty
products may find the appropriate market for the introduction of industrial produc-
tion of cellular metals.
In Table 8-1, achievements in various production routes (see Chapter 2) are listed
for the general categories of cellular metals. Structures with open porosity can have
356 8.1 Processing
Table 8-1. Achievements in the processing of cellular metals and remaining questions.
Type Achievements in production Subjects of concern, R&D demand
x Mass densities in the range 10 1 10 2 x Rather slow processes;

of the bulk metal; x no large series production;
Open cell structures (sponges)
x liquid metal replication, deposition, x specifications not yet generally estab-
and solid compaction processes; lished;
x some brands of different metals com- x rather small dimensions;
mercially available; x joining techniques in development,
x uniformity of architecture and good specific system solutions are required;
reproducibility; x damage hazard due to handling and
x net-shape production is possible, machining.
machining with some precautions;
x integrated massive-cellular compo-
nents can be cast or sintered;
x recyclability about 80 %.
x Mass densities about 10 1 of the bulk x Other metals only in laboratory scale;
metal (mainly aluminum alloys); x relation of processing parameters with
x powder compact foaming and melt cell architecture not fully understood,
foaming processes; therefore unsatisfactory control of
x various brands of foamed aluminum uniformity of cell architecture;
Foamed metals
available; x permeability to be quantified (cells not

x complex net-shape products with skin tight);
in dimensions between 5 and x poor surface quality (skin not tight);
2000 mm possible in series up to x tolerances in shape and skin quality
thousands; not specified; mold technology not
x reproducible semi-product blocs of optimized;
relatively uniform, foamed Al melts x large series or continuous production
available; not yet realized;
x products with foamed cores (sand- x pore-size variation is limited, generally
wich, filled profiles, inserts) in proto- irregular distributions, occasionally
type stage; oversized pores and high density
x various joining techniques and ma- clusters;
chining feasible; x control of bonding for integrated parts
x recyclability about 80 %. to be improved (junctions, brazing,
welding);
x foaming of hollow long products
desirable;
x joining techniques in development,
specific system solutions required;
x damage hazard in handling and
machining;
x remnants of blowing agents tolerable?
Table 8-1. Continues
Type Achievements in production Subjects of concern, R&D demand
x Mass densities about 10 1 of the bulk x Ceramic inclusions;

With closed cell
metal are realized; x no process industrially realized;

substrates
x melt, coating, and solid compaction x consistence and stability of substrates;

techniques available; x recycling requires separation from
x controllable architecture with preform substrate.
technology.
x Duocel open-cell aluminum; x Various other techniques mostly in

Incofoam deposited nickel sponge; laboratory status for research and
Industrial production
x Alporas foamed aluminum melt; prototype production (for instance

x Cymat foamed aluminum melt; COMBAL).
x venture status: Alulight, Alufoam,
Foaminal foamed powder compacts;
AFS sandwiches with Al-foam core;
replication technique ªm-poreº; Gasar
eutectics; LDC, LBM solid-state pro-
cessed.
very low densities with the highest degree of regularity and reproducibility of cell
architecture. They seem to be only appropriate for small volume components for
high-tech and high-price applications.
Foaming of metals in the liquid state promises the most economic processing
route with the drawback of difficult control of cell architecture. Products of remark-
able size and complexity have been made for structural applications. Although
foaming produces closed bubbles in the liquid, during solidification the cell
walls are usually damaged giving the components a certain permeability and neg-
ligible mechanical stiffening. Even the characteristic surface skin of foamed metal
parts can not be considered as defect-free and sufficiently protective. Processing of
closed-cell structures with enclosed hollow substrates or sintered hollow spheres is
still in the prototyping stage.
There are no defined specifications for cellular metals. They are essentially char-
acterized by their processing method. The dependence of material properties on
the different techniques of production remain the subject of research and develop-
ment. There is great potential in improving reproducibility by scaling-up produc-
tion, allowing better control of processing parameters. Some successful examples
of industrial production are quoted in Table 8-1.
The chances of any of the cellular metals being used in components depends
mainly on the availability of appropriate bonding and joining techniques. Second-
ary processing like forming, machining, bonding, and surface treatment require
further technological developments, which have to be governed by the service re-
quirements of each case. A significant advantage of cellular metals compared
with polymers is their recyclability and the potential of up-grading scrap metal,
which may gain increasing importance in the future.
358 8.2 Properties
8.2
Properties
The processing method for a cellular metal component defines the achievable
property profile in the first instance. The density achieved by a given production
route determines the magnitude of the properties. Physical principals govern
those property relations, which reveal the limits as well. The introduction of
pores into a metal cannot improve its mechanical properties even when related
to the specific mass. Tensile strength is limited by fracture mechanics due to inho-
mogeneous stress distributions within the cellular architecture containing many
crack-initiation sites. Properties depending nonlinearly on the dimension of the
part offer benefits for design application, for example specific stiffness for bending
of beams or plates.
The close relation between processing and properties is enhanced by the impor-
tance of the mesoscopic 3D-structure formed by the metal matrix with its specific
microstructure. The scatter of material properties is assumed to be caused by irre-
gularities in the cellular architecture of real products. Some of these can be toler-
ated, but there is still the ambiguity of extremely poor property values not yet cor-
related with structural parameters. Sensitivity studies are required to relate the in-
fluence of meso- and microscopic defects to the properties. Consequently the reli-
ability of the properties needs verification by testing series of products.
The properties depend on the variations in cellular architecture and their unifor-
mity. The actual achievements are contrasted in Table 8-2 with the deficiencies still
existing for most of the available cellular metals. The average density does not
seem sufficient for the characterization of a cellular component. The properties
of cellular materials are not only determined by the matrix metal, but essentially
Table 8-2. Achieved properties and characterization methods compared with demands.
Topic Available property data Characterization deficiencies
x Average mass densities in the range x Uniformity only for cellular structures
10 1 10 2 of the bulk determined by with porous substrates, and solid state
classic densitometry; processed;
x medium and high resolution 2D- x architectural defects' assessment re-
microscopy; quired; oversized pores as well as
x thermo-acoustic nondestructive test mass concentrations cause degrada-
Cellular structure
methods applicable; tion of specific properties;

x radiography or 3D X-ray computed x labor intensive and time consuming
tomography (for light metals human determination of 3D pore-size distri-
tomography applicable) determining butions;
local mass distribution; x architectural quality criteria to be
x center of gravity; developed;
x die-penetrant testing of surface skins; x quantified nondestructive test meth-
x permeability determination for open ods to be established.
porosity character.
Topic Available property data Characterization deficiencies
x Static and dynamic compression x Dependence on test procedure and

behavior generally tested; specimen geometry;
x high specific crash energy absorption x scatter of properties to be related to
of high efficiency; irregularities of the material;
x synergy in crash energy efficiency for x limited reliability of properties;
filled folding hollow structures; sensitive to load introduction;
Mechanical properties
x definitions of initial stiffness, peak x tensile properties very poor;

stress, plateau stress, densification x multi-axial load conditions to be
strain available; considered;
x high specific stiffness for beams and x elastic behavior depends on architec-
plates; ture (stiffness of cellular structure);
x some data on fatigue, crack growth, x matrix of Poisson ratio to be deter-
and creep behavior published; mined for elastic and plastic load
x brittle and ductile cellular metals cases;
available; x specification of test procedures and
x properties mainly related to average evaluation required;
mass density. x life-time predictions for components
not yet established;
x properties of the solid material in the
cellular structure difficult to deter-
mine;
x dependence on thermomechanical
treatment to be investigated.
x High internal surface; open cell x Higher reproducibility required;

Physical properties
structures beneficial for heat transfer x controversial frequency dependence

and catalytic functions; for sound absorption;
x potential for acoustic absorption; x not suitable for isolating functions.
x low electrical and thermal conduc- x frequency and intensity dependence
tivity, but higher than of polymers on material conditions not fully
and ceramics; understood.
x electromagnetic shielding effects.
x Thermal stability usually better than x High thermal gradients due to low
Environmental properties
that of bulk metal; not inflammable; conductivity;

x space compatibility of open-cell struc- x high initial desorption rate of foamed
tures, no radiation damage; metals; closed-cell structures entrap
x no low-temperature embrittlement for gas;
thermally stable alloys; x high specific surface and high capil-
x corrosion resistance similar to that of larity enhances corrosive attack (high
bulk metal. sensitivity to galvanic corrosion);
x protective coating rather complicated.
360 8.3 Design and Application
by the local material distribution: the architecture of the cell structure. X-ray com-
puted tomography offers the possibility of 3D characterization, but requires signif-
icant effort. Nondestructive testing is one of the main activities of the research ne-
cessary to correlate properties with cellular architecture. Furthermore, the proper-
ties of the matrix metal itself cannot be simply transferred from bulk material, but
need to be investigated for each processing technique. In the case of age-hardening
metals, the influence of thermal treatments during production has been indicated,
but may offer chances for further material development.
Table 8.2 lists the most appealing properties, especially attractive in relation with
the low specific mass. The potential of cellular metals depends on combinations of
these specific properties according to particular service requirements. The mechan-
ical properties for structural applications have to be defined specifically for these
cellular structures. Elastic properties, for example, have to be considered as the
stiffness properties of the architecture of the metal; similarly plastic deformation
and damage mechanisms. Complications arise from the internally multi-axial
load transfer, which require special attention regarding the test conditions provid-
ing material data. Both test conditions and the significance of the results have to be
defined properly for cellular metals.
Life-time predictions require further research to complete the achieved results
for exposure conditions simulating long-term service. Damage mechanisms are re-
vealed qualitatively and first attempts of quantification need verification in use. The
physical properties deserve considerable attention to underline the perspectives of
multi-functional application. They have to be determined quantitatively individu-
ally for each type of cellular metal, based on the principally existing potential.
There are essential advantages of cellular metals in comparison with polymers re-
garding environmental properties. Of course, the environmental service conditions
have to be considered for specific cases. The high fire resistance and the space
compatibility have not yet been exploited for aerospace applications.
8.3
Design and Application
The technical advantages of cellular metals cannot be fully exploited just by substi-
tution of materials, but require material-appropriate engineering solutions and de-
sign concepts. The simulation of the performance of components relies on the ma-
terial laws and constitutive equations appropriate to cellular metals and their load-
ing conditions. The actual situation in modeling of cellular metals and in design-
ing components thereof is summarized in Table 8-3. Material models for cellular
metals are required, which have to reflect the 3D behavior of the highly porous,
compressible material. Theoretical studies help to identify the important para-
meters. The transfer of the theoretical results to real architectures requires experi-
mental calibration of the material laws. As stated before, even micromechanical
computations require the appropriate metal matrix properties, which are very dif-
ficult to determine in the processing condition of cells. Reasonable approximations
Table 8-3. State of applicability of modeling, simulation, and design methods for different
products.
Topic Actual achievements Potential for improvements
x 3D-models for geometrically perfect x Only for regular microgeometrics

periodic cell structures for elastoplas- (processed by replication, LBM ...)
tic behavior; applicable;
x scaling laws related to apparent mass x neglect local non-uniformities (deple-
densities for estimation of stiffness, tion or concentration of mass) of real
plateau stress and densification cellular structures;
strains; x deformation concentrations not pre-
x mesomechanical methods based on dictable by models of regular struc-
continuum mechanics of multi-phase tures;
material structure to predict elasto- x prerequisite for mesoscopic simula-
plastic behavior of real cellular parts; tions is the existence of a common cell
x finite element methods applied to architecture and uniform properties of
Modeling
simulate the micro-/meso-/marco- the solid metal;

scopic material behavior; x averaging procedure has to represent
x micromechanical studies of elasto- the characteristic microgeometry in
plastic deformation of cell elements; each volume element; size of FEs have
x models for crush behavior of regular to reflect cellular material of known
cell structures; material laws;
x sensitivity studies on defects in regu- x constitutive equations restricted to
lar cell structures; idealized prerequisites;
x rules for qualitative crash behavior of x input data for material behavior and
real parts. constitutive equations not fully avail-
able;
x calibration by macroscopic testing
necessary;
x micromechanical procedures only for
a few cells applicable owing to the
excessive computing time.
x Design based on macroscopic behav- x No safety margins established coping

Design and simulation
ior according to experimental tests; with inhomogenities of real products;

x design methods for phase and shape x sensitivity studies on tolerable defects
optimization for non-uniform cellular are to be developed; reliable material
structures in development. data for design incomplete;
x life time prediction rules not available;
x mechanical properties for multi-axial
load conditions not available;
x appropriate design experience is
missing; conflict between optimiza-
tion and realization;
x joining solutions not yet verified for
transfer into design rules.
362 8.3 Design and Application
Topic Actual achievements Potential for improvements
x Light-weight structural parts of high x No service experience; service tests

bending stiffness; required;
x crash energy absorption by light x integration of cellular metals into
weight components; systems unsatisfactory;
x additional requirements of thermal x established alternative solutions avail-
stability or/and non-inflammability; able;
x light weight components exhibiting x special small series cases;
acoustic absorption and mechanical x multi-functional benefits required
Application
damping; (functional integration, reduction of

x open porosity for heat exchangers, parts, assembling advantages);
catalysts and filters; x processing too complicated and of
x multi-functional applications exploit- limited reproducibility.
ing as well physical properties;
x compounds with face sheets (AFS) or
inserts in hollow sections of extru-
sions or castings;
x a multitude of prototypes offered.
x Applications where the value of weight x Price estimations actually more than
saving is high (means of transport, 15 U/kg or 5 U/liter (which is more
highly accelerated machine parts and than the automotive market is pre-
sporting goods); pared to pay);
x aerospace applications; x advanced material image to be im-
x small series rail, public transport and proved;
Market
utility vehicles; x additional benefits required;

x selective use of cellular metals; x increased importance of recycling;
x value of recyclability; x experience in part integration
x chances for learning curve effects on required;
processing and price. x no large series production existing,
therefore price estimations difficult;
price mainly governed by processing
effort.
x specialty markets with progressive
demands to be selected.
based on averaged or characteristic structures are available. The hierarchical ap-

proach of macro-, meso-, and micromechanical simulations according to the
local sensitivity of a structural feature in relation to the load case of the component
seems very promising.
The close interaction of development of models with experimental determination
of properties is leading to useful interpretations of the general material behavior.
The access to the material data required for input to the different models is a spe-
cific topic of research. Finite-element calculations have to take care of the appropri-
ate choice of mesh geometry owing to the generic heterogeneous structure dealt
with, whether for the micromechanical considerations or for the mesomechanical
approach, where representative cellular material volumes have to be found. Design

optimization for cellular metals refers not only to shape, but also to the arrange-
ment of the varying structures to be interpreted as different material phases.
Load transfer from the surrounding is an essential design element for which
experience has to be broadened and solutions unconventional for metals may be
beneficial.
The influence of scatter in material properties locally within a part and between
different samples of the same component needs further attention in design and
simulation. The sensitivity to irregularities in the cellular architecture has to be
known in order to establish tolerance margins and service-related quality criteria
for products. The discussion on ªhow uniform a cellular metal has to be to be re-
liableº needs quantified arguments related to the intended application. Progress is
expected by performing case studies that enable correlation of processing para-
meters, material properties, modeling requirements, performance simulation,
and service experience. The expected general benefits are listed in Table 8.3. Var-
ious prototypes exist for testing service performance, but unfortunately most of
them are designed for substitution of existing material solutions and the systems
are usually not optimized for cellular metals. Any how, each practical case in-
creases the experience in potential application of cellular metals. Prototypes of spe-
cialty products, like open-cell Duocel or Incofoam electrodes, or noise-damping
Alporas plates, may serve to demonstrate the potential of tailored property profiles
for specific requirements. Multifunctionality and part integration increase the
chances for industrial application of high-value materials.
The market is mainly governed by cost benefit relations. Applications experi-
encing frequent accelerations, where weight savings are of interest, are indicated
in Table 8-3. The question of what price can be paid for such weight savings,
has to be asked periodically. The answer depends mainly on the changing eco-
nomic boundary conditions like energy costs and ecological restrictions. Global
trends seem to support further efforts for weight reductions. Opportunities in
small-series specialty applications should be seized because they provide means
of gaining experience in all aspects: engineering, processing, performance, and ser-
vice assessment. The first series product for automotive applications has been an-
nounced and a successful market introduction is expected. The introduction of a
learning curve in the whole production chain seems to stimulate the most impor-
tant progress step necessary for the development of market-penetrating innova-
tions made of cellular metals.
Index
a anion-stimulated ATPase 380

A-type ATPases 387 anthrax protective antigen 218
ABC (ATP Binding Cassette) 77 anthrax toxin 218
ABC secretion 184 ff antimonite transporters 377 ff
diversity of proteins secreted 186 antiporters 8 ff
ABC secretion pathway, schematic aquaglyceroporins 251
representation of 191 aquaporin AQP1 248
ABC subunit, sequence 97 ff aquaporin Z, of E. coli 250, 255
ABC systems, definition 79 aquaporin 1 254
ABC transporters 77 ff aquaporins 247 ff
associated proteins 84 MIP-related- 248
binding proteins 85 ff plant- 249
components of 85 ff AraF 89
composition of 80 ff Archaeoglobus fulgidus 49
crystal structure 88 ff ARN1 transporter 469
integral transmembrane domains ars operon
(TMDs) 91 ff in bacteria 392
listing of 89 of E. coli 378, 391
metal- 316 arsA 379
ABC-type efflux permeases 13 ATPase activity of 380
ABC-type nickel transporter 407 crystal structure of 389
ABC-type uptake permeases 12 DTAP domain 386
Acidithiobacillus ferrooxidans 332 nucleotide binding sites 381 ff
ACR (arsenic compounds resistance) of E. coli 388
family 393 unisite and multisite catalysis in 384
ACR genes, in yeast 393 ArsAB pumps 392
aerolysin 215 ff arsB 379, 390
Aeromonas hydrophila, channel formation by ArsC 379, 391
aerolysin 216 arsenite efflux 392
AFT1 in yeast 466 arsenite resistance 392
alignment studies, on TonB-dependent arsenite transporters 377 ff
receptors 282 ArsR 379
alkaliphilic Bacillus 50 asperchrome B1 470
ALLBP 89 asperchrome D1 470
a-toxin 217 ATP hydrolysis 98 ff
a-type channels 5 by FhuC 298
d-aminolevulinic acid, transport of 154 ATP synthases 23 ff
482 Index
ATPase activity of ArsA 380 copA 368

ATPases, phylogram 363 copB 368
Atx1 metallochaperone 427 cop operon of Enterococcus hirae 368
Atx2p 456 copY 368
copZ 368
b CopA, in copper uptake 369
b-barrel porins 6 CopB, in copper excretion 369
Bacillus stearothermophilus 49 CopB copper ATPase, of Enterococcus
Bacillus subtilis, CitM protein 404 hirae 363
saturable Mn permease 330 copper ATPases
Bacillus thuringiensis Cry toxins 223 of Helicobacter pylori 373
bacterial copper transport 361 ff of Listeria monocytogenes 373
bacterial cytochrome c oxidases 436 copper chaperones 366
bacterial iron transport 289 ff copper excretion, CopB 369
bacterial zinc transport 313 ff copper ion transport, mitochondrial- 419 ff
binding protein-dependent Zn2 uptake copper resistance
in gram-negative bacteria 320 in Escherichia coli 371
in gram-positive bacteria 316 ff plasmid-encoded- 374
Bradyrhizobium japonicum (HupN) 413 copper transport 361 ff
Bsd2p and manganese ions 453 copper transport system, in S. cerevisiae 426
copper uptake, CopA 369
c coprogen transporter 472
Ca2 -ATPase, of sarcoplasmic reticulum 364 CopY repressor protein 370
Ca(II) ATPase 428 CorA
Ca-ATPase calcium pump, structure of 428 of Methanococcus jannaschii 353
Caloramator fervidus 49 of Salmonella typhimurium 352 ff
Candida albicans peptide transport gene CorA magnesium transporter 351 ff
(Ca Ptr2) 155 Corynebacterium glutamicum 53
carbon monoxide dehydrogenases [Ni] 399 COS16 457
catalysis within the F1 complex 27 ff Cox polypeptides 425
cation diffusion facilitator 315 Cox1 polypeptide 424
Ccc1p: a manganese homeostasis protein Cox2 subunit 424
456 Cox11 sequences 434
Ccc2 P-type ATPase transporter 427 Cox17 metallochaperone 429
CCC2, intracellular transporter of copper Cox17 sequences 430
449 CPx motif 364
Ccc2p 456 CPx-type ATPases 362
CDC1 gene product 456 sequence motifs 365
cell surface permease FTR1 449 crystal structures of DtxR 305
channel formation steps 212 ff CTR1, cell surface copper permease 449
channel forming colicins 222 Ctr1/Ctr3 system 426
channel forming protein toxins 209 ff Cu metallation, of cytochrome c
a-toxin 217 oxidase 431 ff
aerolysin 215 ff Cu metallation, of mitochondrial cytochrome c
channel forming proteins 209 ff oxidase 441
classification of 211 Cu-specific ATPase pumps 428
cholesterol-dependent toxins 219 CxxC motifs 366
CitM protein of Bacillus subtilis 404 cytochrome c oxidase, structure of 424
citrate fermentation 51 cytochrome c oxidase
classes of transporters 4 ff Cu metallation of 431 ff
Clostridium fervidus 49 Cu metallation of mitochondrial- 441
Cnr (cobalt nickel resistance) 406 mitochondrial- 422 ff
coenzyme F430 399 cytochrome c oxidases, bacterial- 436
colicins 222 Czc (cobalt zinc cadmium resistance) 406
Index 483
d F-type ATPases 23 ff
des(diserylglycyl)ferrirhodin 470 families of transporters 1 ff, 5 ff
dipeptide permease (Dpp) 144 Fe3 -citrate 306
diphtheria toxin 304 Fe3 -dicitrate 300
DMT1 329 Fe3 regulatory mechanism 306
Dpp-type 159 Fe3 -siderophore transport 291 ff
DppA 89 Fe-enterobactin transporter 473
DTAP domain, in ArsA 386 FecI and FecR proteins 306
DtxR, crystal structures 305 feo genes 304
DtxR/IdeR family 338 FepA 264 ff
b-barrel 265 ff
e conserved residues 284
EI 118 extracellular loops 270 ff
enantio-ferrichrome 470 N-terminal domain 267 ff
enantio-rhizoferrin 468 ribbon diagram 264
ENB1 transporter 472 structure determination of 275
energy transduction, ATP synthase 31 ff FepA mutants 285
enteric bacteria, peptide transport in 143 ff Ferric Uptake Repressor (Fur) 337
Enterococcus hirae cop operon 368 ferric citrate transport system 306
Enterococcus hirae CopB copper ATPase ferrichrome 470
363 ferrichrome A 470
Escherichia coli HlyA 221 ferrichrome transporter 471
aquaporin Z 250, 255 ferrichrysin 470
ars operon 378, 391 ferricrocin 470
ABC transporters 80 ff ferrirhodin 470
ArsA 388 ferrirubin 470
ATP synthase 24 ff, 36 ff ferrous iron transport system 304
ATP transporter binding proteins 90 FET3 448
copper transport in 371 FET4 450
FhuABCD activities 292 FET5 450
FhuD protein 297 FhuA 264 ff
LamB 240 barrel structure 266
MalK 105 conserved residues 284
melibiose transporter 54 extracellular loops 270 ff
Na/melibiose transporter 58 N-terminal domain 267 ff
Na/proline transporter 55, 59 ff, 65 ff structures with ligand 272 ff
Na/substrate transport in 52 FhuA protein 265, 292
Nik system 407 FhuABCD activities, of E. coli 292
OmpF 234 FhuB transport protein 298
peptide transport 143 ff FhuC, ATP hydrolysis by 298
Tat pathway in 174 FhuD 89, 295 ff
TolC trimer 242 crystal structure 297
extracellular loops Fibrobacter succinogenes 50
in FepA 270 ff FRE genes, of Saccharomyces cerevisiae
in FhuA 270 ff 448
FRE reductases, in siderophore
f transport 474 ff
F1 subunits 25 ff FRE1 447
F1/FO interface 31 ff FRE2 447
F1FO ATP synthase 23 ff FRE3 448
model of energy transduction 39 ff FRE4 448
FO subunits 33 ff FTR1 448
proton translocation pathway 42 fungal ornithine-N5 -oxygenase 466
F-ATPases 23 ff fungi, siderophore transport in 463 ff
484 Index
g iron metabolism, in eukaryotic cells 305

GATA family of transcription factors 466 iron transport 261 ff, 289 ff
GBP 89 Saccharomyces cerevisiae genes involved
General Diffusion Porins 234 in 458
GlnBP 89 regulation of 451
GlpFs 251 iron transporters, in yeast 447 ff
glucose transport 115 ff iron uptake, low-affinity- 450
glutathione transport 154 iron-dependent regulatory protein
glycerol conducting channels 247, 256 (IRP) 305
glyoxalase I 399
group translocation 115 k
Klebsiella pneumoniae
h citrate fermentation in 51
H cycle 52 melibiose transporter 54
Haemophilus influenzae Na/citrate transporter 56, 59, 61, 69
ATP transporter binding proteins 90
heme transport systems 301 l
hFBP 297 lactic acid bacteria, peptide
Halobacteriales 50 transport in 148 ff
Halobacterium 50 Lactobacillus plantarum 328
HbpA 89 Lactobacillus plantarum MntA, gene 331
heavy metal ATPases 362 Lactococcus lactis
heavy metals, transport of 452 ABC transporters 81 ff
Helicobacter pylori, nickel/cobalt peptide transporters 148 ff
transporters 409 lactoferrin 299
Helicobacter pylori copper ATPases 373 LamB, of E. coli 240
heme, bacterial use of 300 LamB channel 238
heme transport systems 301 ff LAO 89
hexahydroferrirhodin 470 LbpA 299
HgAtx1 complex, structure of 427 LbpB 299
high-affinity nickel uptake systems 406 lipid bilayer membranes 231
HisJ 89 Listeria monocytogenes, copper ATPase 373
HoxN 410 LivJ 89
hpCopA 373 LivK 89
HPr 118 low-affinity iron uptake 450
hydrogenases [NiFe] 399 low-affinity Zn2 uptake systems 321
hydroxamate reductases 476 Lys7 metallochaperones 429
i m
IIA 118 magnesium transporters 347 ff
IIAGlc subunit 120 CorA 351 ff
regulatory role 131 MgtE 350
IIB 118 Major Intrinsic Proteins (MIPs)
IIC 118 superfamily 247
IICBGlc mutants 124 ff malK dimer, asymmetry within 105 ff
listing of 125 ff MalK from Thermococcus litoralis 101 ff
IICBGlc subunit 121 ff maltoporin 237
regulatory role 132 maltose binding protein from E. coli 90
topology of 122 maltose binding protein MBP or MalE 239
insecticidal crystal (Cry) proteins 223 manganese, in bacteria 326
integral transmembrane domains (TMDs) manganese accumulation 325 ff
ABC transporters 91 ff manganese homeostasis 452
sequence 96 manganese homeostasis protein,
ion selectivity of porins 233 Ccc1p 456
Index 485
manganese transport n
Saccharomyces cerevisiae genes involved N-terminal domain
in 458 of FepA 267 ff
in Saccharomyces cerevisiae 452 ff of FhuA 267 ff
in bacteria 330 ff Na cycle 52
manganese transport ATPase, Pmr1p 455 Na/citrate transporter (CitS) of
manganese transporters, in yeast 447 ff K. pneumoniae 56, 59, 61, 69
manganese uptake, in bacteria 325 Na/H antiporter 52
MBP 89 Na/K ATPase 23
MBP (MalE) 89 Na/melibiose transporter of E. coli 58
melibiose transporter of E. coli (EcMelB) Na/proline transporter 53
54, 62 ff Na/proline transporter (PutP) of E. coli
melibiose transporter of K. pneumoniae 59 ff, 55, 65 ff
(KnMelB) 54, 62 Na/substrate transport in Escherichia coli
membrane topology of CPx-type ATPases 363 52
metal ABC transporters 316 Na/substrate transport systems 48 ff
metalloenzymes, nickel as cofactor 398 ff Natural Resistance-Associated Macrophage
Methanobacterium thermoautotrophicum 49 Protein (Nramp1) 329
Methanococcus jannaschii 49, 353 Ncc (nickel cobalt cadmium resis-
MFS-type nickel exporter (NrsD) 406 tance) 406
Mg2 transporter families: MgtE, CorA, NhlF 410
and MgtA/B 347 Nic1p, of Schizosaccharomyces pombe 412
MgtA/MgtB Mg2 transporters 355 ff nickel, as a cofactor of metalloenzymes
MgtC protein 357 398 ff
MgtE magnesium transporters 350 nickel/cobalt transporter family 408 ff
microbial nickel transport 397 ff hydropathy profile alignment 409
microbial siderophores 465 in Bradyrhizobium japonicum 412
MIP-like channel genes 252 nickel homeostasis 403 ff
MIP-related aquaporins 248 nickel resistance 401
MIP-related sequences, phylogenetic nickel toxicity 401
tree 252 nickel transport 397 ff
mitochondrial copper ion transport nickel transporter (UreH, UreI)
419 ff of thermophilic Bacillus sp. 412
mitochondrial cytochrome c oxidase nickel transporters, ABC-type- 407
422 ff nickel uptake, high-affinity systems
MMT1 451 406
MMT2 451 nickel/cobalt transporters, metabolic
Mn(II) ABC transporters 328 functions 410
Mn(II) as a Fur co-repressor 338 NiCoT in Bradyrhizobium japonicum,
Mn(II) primary transporter 328 nickel/cobalt transporter in 412
Mn(II) permeases 328 Nik system, Escherichia coli 407
Mn(II) transport in S. aureus 330 Nik-related transporters 408
MntABC uptake system 333 nikABCDE operon 407
MntH 329 nikR gene 407
MntH proteins 336 nitrous oxide reductase (N2OR),
ModA 89 in P. stutzeri 439
molecular recognition templates NixA 410
(MRT) 156 ff (Nramp)/Divalent Metal Transporter
molecular recognition templates (MRTs), (DMT) 329
optimal features of 159 Nramp family 452
MRP1 (multidrug resistance-associated Nramp1 protein 339
protein) 393 Nramp2 329
multi-copper oxidase 448 nucleotide binding sites, in ArsA
Mycobacterium tuberculosis (MtNicT) 413 381 ff
486 Index
o of Pelobacter venetianus 235

oligopeptide permease (Opp) 145 ff solute selectivity of 230
oligopeptide transport in sporulation 151 specific- 237 ff
oligopeptide transporter family (OPT) 142 Porphyromonas gingivalis, heme transport
OmpF of E. coli 234 system 302
Opp-type 159 porters 8 ff
OppA 89 PotD 89
ornithine-N5 -oxygenase, fungal- 466 PotF 89
osmoregulator 53 Propionigenium modestum,
osmosensor 53 ATP synthase 36 ff, 40
osmotic stress 53 protein export 165 ff
Sec pathway 168 ff
p Tat pathway 173 ff
P-type ATPase, superfamily 331 protein secretion 165 ff
P-P-bond-hydrolysis-driven transporters 12 pathways found in gram-negative
P-type ATPase superfamily 355 bacteria 167
P-type ATPases 315, 362 Sec-dependent pathway 178 ff
sequence motifs 365 Sec-independent pathways 184 ff
P. aeruginosa exotoxin A 304 type III secretion pathway 192 ff
Pelobacter venetianus porins 235 type IV secretion systems 198 ff
PepT family 141 proton motive force (pmf) 48
peptide-acetyl-CoA transporter (PAT) fa- proton-dependent manganese
mily 142 transporter 329
peptide transport 139 ff proton-dependent oligopeptide transporter
in eukaryotic microorganisms 155 family (POT) 141
peptide transport systems, classification PsaA and TroA proteins 332
of 140 ff Pseudomonas stutzeri, nitrous oxide reductase
peptide-uptake permease (PUP) family 142 (N2OR) 439
perfringolysin O 219 PstS 89
periplasmic Zn2/Mn2/Fe? binding pro- PTR (Peptide Transport) 155
teins 319 PTR2 transporter 155
peroxide stress regulator PerR 338 PUP (peptide-uptake permease) 152
pertussis CyaA 221 Pyrococcus furiosus 86
phage adsorption sites 293 ATP transporter binding proteins 90
Phosphotransferase System (PTS) 115 ff Pyrococcus horikoshii 49
components 117 ff
PTS proteins 118 r
PTS transporters 119 ff Ralstonia eutropha, nickel/cobalt trans-
regulation of 129 ff porters 409
plant aquaporins 249 Ralstonia eutropha (HoxN) 413
plasma membrane ferric reductase 447 RbsB 89
plasma membranes, transport systems reconstitution, of porin pores 232
in 403 reduction of arsenate 391
plasmid-encoded ars operon of E. coli 391 regulation
plasmid-encoded copper resistance 374 by Fe3 306
Pmr1p: a manganese transporting by Fe3 -siderophores 306
ATPase 455 repressor protein CopY 370
pore-forming toxins 6 rhizoferrin 468
porin channels, function of 230 ff Rhodobacter capsulatus porin 234 ff
porin pores, reconstitution of 232 Rhodococcus rhodochrous, nickel/cobalt
porins 227 ff transporters 409
Rhodobacter capsulatus 234 ff RND family of exporters 313 ff
ion selectivity of 233 RND-type Ni2 export systems 406
isolation of 229 RTX toxins 220
Index 487
s Staphylococcus aureus, Mn(II) transport 330

S. gordonii ScaABC system 333 staphyloferrin 468
S. marcescens cytotoxin 304 Streptococcus bovi 50
Saccharomyces cerevisiae Streptococcus pyogenes 328
FRE genes 448 subunit g rotation 29 ff
copper transport system 426 subunits IIAGlc and IICBGlc 119 ff
genes involved in iron and manganese subunits IIABMan, IICMan and IIDMan 119
transport 458 Sulfolobus solfataricus, ABC transporters
iron transport in 447 ff 81 ff
manganese transport in 452 ff superoxide dismutases [Ni] 399
Salmonella SitA periplasmic protein 334 symporters 8 ff
Salmonella typhimurium 52, 88, 352 ff synechococcal copper ATPases 372
peptide transport 143 ff
sap (sensitivity to antimicrobial peptides) t
152 TAF1 transporter 469
sarcoplasmic reticulum, Ca2 -ATPase 364 Tat (twin arginine transfer) pathway 173 ff
SBP 89 Tat pathway, in Escherichia coli 174
ScaABC system, of S. gordonii 333 Tat proteins 175
Schizosaccharomyces pombe, Nic1p 412 Tat signal peptide 176
Sco protein sequences 432 TbpA 299
Sec-dependent pathway:type II secretion TbpB 299
pathway 178 ff Thermococcus litoralis 86
Sec pathway 168 ff ATP transporter binding proteins 90
schematic representation of 171 MalK 101 ff
Sec proteins 166 Thermoplasma acidophilum 332
Sec translocase 169 ff TIM complex 421
secretion proteins 166 TMBP 89
Selenomonas ruminantium 50 TolC, of E. coli 241
SfuA 300 TOM complex 420
Shigella toxin 304 TonB, functions of 281
shuttle mechanism 295 TonB box 294
siderophore classes 464 ff TonB-dependent phages 294
Siderophore Iron Transport (SIT) TonB-dependent transport 281
family 463 Tpp-type 159
siderophore receptors 261 ff transferrin 299
genetic and biochemical studies on 276 ff transport, of Fe3 -siderophores 291 ff
mechanism 279 ff transport proteins, for heme 301 ff
siderophore transport transport systems, in bacterial and eukaryotic
FRE reductases 474 ff plasma membranes 403
in fungi 463 ff transporter, for ferrichromes 471
siderophores, representative microbial- 465 Transporter Classification (TC) 1 ff
SIT1 transporter 468 transporter for coprogens 472
SitA periplasmic protein, of Salmonella 334 Treponema, TroA protein 332
Smf1p 452 tripeptide permease (Tpp) 147
Smf2p 452 TroA protein from Treponema 332
SMF3 451 type I secretion pathway 184 ff
Smf3p 452 schematic representation of 191
sodium motive force, (smf) 48 type II secretion pathways 178 ff
sodium/substrate transport 47 ff schematic representation of 183
solute selectivity, of porins 230 type III secretion pathway 192 ff
specific porins 237 ff schematic representation of 197
SRE 467 type IV secretion pathways, schematic
SREA 467 representation of 200
SREP 467 type IV secretion system 198 ff
488 Index
u y
ubiquinol oxidase 436 yeast
uniporters 8 ff ACR genes 393
Urbs1 in Ustilago 466 AFT1 in 466
ureases 399 iron transporters 447 ff
Ustilago, Urbs1 466 manganese transporters 447 ff
yeast oxidase complex 422
v yeast vacuole and manganese 457
Vibrio alginolyticus 50 Yersinia pestis YfeABCD system 334
Vibrio parahaemolyticus 50 YfeABCD system, of Yersinia pestis 334
virulence factors 304
vitamin B12, binding of 294 z
Zinc transport 313 ff
w Zinc uptake, low-affinity systems 321
Walker A and B sequences 98 ZIP transporters 405
Walker A sequence 386 Zn2 export 314
water channels 247 Zn2 transport systems, regulators of 323
Wilson copper ATPase (ATP7B) 367 Zn2 uptake, binding protein-dependent-
316 ff
znuA (zinc uptake) gene 321

Handbook of Cellular Metals

Cargado por

Información del documento

Derechos de autor

Formatos disponibles

Compartir este documento

Compartir o incrustar documentos

Opciones para compartir

¿Le pareció útil este documento?

¿Este contenido es inapropiado?

Copyright:

Formatos disponibles

Handbook of Cellular Metals

Cargado por

Copyright:

Formatos disponibles

Handbook of Cellular Metals: Production, Processing, Applications.

Edited by H.-P. Degischer and B. Kriszt

Related titles from Wiley-VCH

W. Schneider, K. Ehrke (Eds.)

Handbook of Cellular Metals

Production, Processing, Applications

Editors This book was carefully produced. Never-

Dr. Brigitte Kriszt British Library Cataloguing-in-Publication

c WILEY-VCH Verlag GmbH

All rights reserved (including those of

printed in the Federal Republic of Germany

Composition Hagedorn Kommunikation,

B. Kriszt and H. P. Degischer

1. H. P. Degischer (ed), ªMetallschaÈumeº 3. M. F. Ashby, A. G. Evans, N. A. Fleck, L. J.

1. B. Sosnick, US Patent 2 434 775, 1948. 4. V. Gergely, H. P. Degischer, T. W. Clyne,

1 Introduction: The Strange World of Cellular Metals 1

2 Material Definitions, Processing, and Recycling 5

2.6 Infiltration and the Replication Process for Producing Metal

3 Secondary Treatment of Cellular Metals 71

3.1.3.3 Thermal Sprayed Composites from Metal Foams 81

4 Characterization of Cellular Metals 127

4.3.5.1 Geometry of Cellular Structure 165

5 Material Properties 179

6 Modeling and Simulation 243

7 Service Properties and Exploitability 299

7.1.5.5 Alternative Cellular Materials Based on Polymers 306

7.5 Applying the Investment Methodology for Materials (IMM) to

8 Strengths, Weaknesses, and Opportunities 355

T. Daxner Th. Hipke

C. San Marchi S. Stanzl-Tschegg

AFS aluminum foam sandwich

PVD physical vapour deposition

In spite of disappointments and mistrust, interest in cellular metals is growing.

A great variety of cellular metals is produced by research laboratories and industrial

architecture within a considered sample or part (degree of uniformity), and the

1. L. J. Gibson, M. F. Ashby, Cellular Solids: 7. J. Banhart, M. F. Ashby, N. A. Fleck (eds),

C. KoÈrner and R. F. Singer

The Cymat/Alcan and Norsk Hydro melt-foaming process is a continuous, gas-in-

Figure 2.1-2. Average cell size as a

2.1.2.1 The Shinko Wire Process [2]

Figure 2.1-4. Manufacturing process for Alporas foams [2].

2.1.2.1 The FORMGRIP Process

1. J. T. Wood, ªProduction and Application of 6. V. Gergely, T. W. Clyne, ªA Novel Melt-Based

J. Banhart and F. BaumgaÈrtner

Figure 2.2-1. Foam production by powder-compact foaming technique.

Figure 2.2-2. Aluminum foam

Table 2.2-1. Characteristics of powder-compact foaming method: advantages and disadvantages

Advantage Problem Disadvantage

Net-shape foaming possible Uniformity of pore structure Cost of powders

Composites can be Process control must be Very large volume parts

Parts are covered by metal skin Permeable (holes) Coating process

Graded porosity can be Difficult to control

Flexibility in alloy choice

No stabilising particles have to be

Ceramics and fibers can be added

Figure 2.2-3. Rear wall of an automobile made of

2.2.2.1 Powder selection

be of sufficient quality. However, powders from different manufacturers led to no-

2.2.2.4 Further processing of foamable material

1. J. Banhart, ªManufacture, characterisation 8. T. Miyoshi, M. Itoh, S. Akiyama, A. Kitahara,

G. Rausch and J. Banhart

Zinc can be foamed by a straight-forward modification of the powder-compact tech-

Figure 2.3-3. Pore structure of open porous titanium with 67 % porosity.

2.3.4.1 Powder-Compact Foaming Technique

where Pbubble bubble pressure, PT equilibrium pressure in the liquid, s sur-

where g gravity constant, P0 ambient pressure, h depth.

where h nr dynamic viscosity.

where T temperature of the melt.

where c0 concentration of gas molecules, f0 frequency factor of gas molecules

where c1 concentration of heterogeneous nucleation sites, f1 frequency factor

where n total number of cells, i cell number, li cell-boundary length, ai

where x cell-wall thickness, 1/RPB curvature of the Plateau border, R radius