Está en la página 1de 4

Ideal hydrogen termination of the Si(111) surface

G. S. Higashi, Y. J. Chabal, G. W. Trucks, and Krishnan Raghavachari



Citation: Appl. Phys. Lett. 56, 656 (1990); doi: 10.1063/1.102728
View online: http://dx.doi.org/10.1063/1.102728
View Table of Contents: http://apl.aip.org/resource/1/APPLAB/v56/i7
Published by the American Institute of Physics.

Related Articles
Specific channels for electron energy dissipation in the adsorbed system
J. Chem. Phys. 138, 104201 (2013)
Conduction band offset at GeO2/Ge interface determined by internal photoemission and charge-corrected x-ray
photoelectron spectroscopies
Appl. Phys. Lett. 102, 102106 (2013)
Antiferromagnetic ordering of dangling-bond electrons at the stepped Si(001) surface
J. Chem. Phys. 138, 104702 (2013)
Contribution of radial dopant concentration to the thermoelectric properties of core-shell nanowires
Appl. Phys. Lett. 102, 103101 (2013)
The role of dimensionality in the decay of surface effects
J. Chem. Phys. 138, 084707 (2013)

Additional information on Appl. Phys. Lett.
Journal Homepage: http://apl.aip.org/
Journal Information: http://apl.aip.org/about/about_the_journal
Top downloads: http://apl.aip.org/features/most_downloaded
Information for Authors: http://apl.aip.org/authors
Downloaded 18 Mar 2013 to 31.147.117.245. Redistribution subject to AIP license or copyright; see http://apl.aip.org/about/rights_and_permissions
Ideal hydrogen termination of the 51 (iii) surface
G. S. Higashi, Y. J. Chabal, G. W. Trucks, and Krishnan Raghavachari
AT& T Bell Laboratories, Murray Hill, New Jersey 07974
(Received 29 June 1989; accepted for publication 4 December 1989)
Aqueous HF etching of silicon surfaces results in the removal of the surface oxide and leaves
behind silicon surfaces terminated by atomic hydrogen. The effect of varying the solution pH
on the surface structure is studied by measuring the SiH stretch vibrations with infrared
absorption spectroscopy. Basic solutions (pH = 9-10) produce ideally terminated Si( 111)
surfaces with silicon monohydride C-Si-H) oriented normal to the surface. The surface is
found to be very homogeneous with low defect density ( < 0.5%) and narrow vibrational
linewidth (0.95 cm I).
HF acid etching is a key step in producing silicon sur-
faces which are contamination-free and chemically stable for
subsequent processing in the semiconductor industry. 13 In-
deed, chemical oxidation and subsequent HF treatment of Si
surfaces are used prior to the growth of gate oxide, where
surface contamination and interface control are crucial to
device performance.
4
The chemical stability of the HF-treat-
ed surface was long thought to be explained by fluorine ter-
mination. Recently, however, it has become understood that
fluorine is a minority species on the surface and that the
remarkable surface passivation achieved is explained by H
termination of silicon dangling bonds, 3,5,6 protecting the sur-
face from chemical attack. The structure of the H-terminat-
ed surface also plays a role in the initial stages of gate oxide
formation and possibly affects the resulting SiiSiO
z
interfa-
cial properties. Recently, it was shown
7
,8 that aqueous HF
solutions (HF in H
2
0 of various concentrations) induce mi-
croscopic roughness on both Si ( 111) as well as Si ( 100). In
this study, we find that varying the pH of the HF solutions
drastically alters the microscopic roughness and the nature
of the associated H termination of the surface. In particular,
basic solutions produce ideally terminated SiC 111) surfaces
that are microscopically smooth.
Our approach is to measure the infrared absorption in
the SiH stretch region ( ~ 2 1 0 0 cm I) using a multiple in-
ternal reflection geometry which gives nearly equal sensitiv-
ity to components of vibrations parallel and perpendicular to
the surface.
7
-
1o
From these spectra the relative concentra-
tions of different hydride species and their orientation can be
determined. In particular, for well-ordered surfaces, the de-
fect density can be quantified.
Experiments are performed on 0.5 X 19 X 38 mm
3
n-
doped (p-lOO-300 n em) SiC I 11} and Si( 100) samples
with 45 bevels on each of the short sides. The infrared radi-
ation exiting the interferometer is focused at normal inci-
dence onto the input bevel, is internally reflected ~ 75 times,
and exits the output bevel to be collected and refocused onto
a liquid-N
2
-cooled InSb photodetectof. Samples are pre-
pared by conventional thermal oxidation ( 1000 A oxide)
to move the interface away from any residua! polishing dam-
age. After HF acid removal of the thermal oxide, the Si sam-
ples are placed into the HF solution of interest. They are then
removed, placed into a dry N 2 purged environment, and
spectra are recorded. Background reference spectra are ob-
tained by measuring the transmission of chemically oxidized
8i surfaces using HCl:H
2
0
z
or H
2
S0
4
:H
z
O
z
solutions.
In the course of our work on HF etching, we have noted
that the concentration of aqueous HF alters the observed
spectra. In contrast to the well-defined (narrow - 10 cm -1 )
vibrational features when dilute HF (1-10% in H
2
0) is
used,7.8 the vibrational lines are broad ( ~ 3 0 cm - 1) when
concentrated HF (49% in H
2
0) is used, and the band is
characterized by a higher density of defect modes (dihy-
dride). Since the pH varies from 1.0 to 2.0 with water dilu-
tion (relative concentrations ofR + and OH-) and appears
to affect the morphology of the H-passivated surfaces, a
study of the effects of pH was undertaken. A common way to
increase the pH and maintain it constant at p H ~ 5.0 is to use
ammonium fluoride (NH
4
F) as a buffering agent (7:1
NH
4
F:HF). Thus, we used this buffered HF solution as a
starting point and varied its pH by adding ammonium hy-
droxide (NH
4
0H) to raise the pH (pH=5 to 12) orhy-
drochloric acid (HeI) to lower the pH (pH = 5-0).
For pH greater than 4.0, we observe a dramatic change
in the hydrogen termination, compared to solutions pre-
viously prepared with lower pH. 7,8 A comparison of the SiH
stretch spectra is shown in Fig. 1 (a). The spectrum shown in
dashed lines corresponds to a Si ( 1 I 1) surface prepared in
dilute HF (100:1 H
2
0:HF) and is characterized by a large
density of defect modes (coupled monohydride 1\1 and dihy-
dride D). 7,8 The (111) terraces are terminated by both mon-
ohydride (M) and trihydride (n. In contrast, the spectrum
of the surface prepared in pH modified buffered HF
(pH = 9-10) is dominated by one very sharp vibrational
line at 2083.7 cm - 1, perpendicular to the surface. The obser-
vation of a single, narrow mode, oriented perpendicular to
the surface, makes its assignment as the monohydride of an
ideally terminated SiC 111) surface unambiguous. Our ear-
lier studies
7
,8 on surfaces with high defect densities mista-
kenly assigned this mode to the monohydride on adstruc-
tures because of its frequency position. 11
The defect density can be estimated from the area under
the s-polarized spectrum because only H modes at defects
have vibrational components parallel to the surface. After
correction for misalignment (-2) resulting in an apparent
contribution at 2083.7 cm -1 in thes-polarized spectrum and
for screening of the perpendicular mode, 12 the defect density
is found to be -0.5% (every 700 A for one-dimensional
656 AppL Phys. Lett. 56 (7), 12 February 1990 0003-6951/90/070656-03$02,00 @ 1990 American Institute of PhYSics 656
Downloaded 18 Mar 2013 to 31.147.117.245. Redistribution subject to AIP license or copyright; see http://apl.aip.org/about/rights_and_permissions
roughness or every 30 A for two-dimensional roughness).
This number is comparable to what one would expect for the
misorientation of the sample or 360 A between
steps).
Another measure of surface homogeneity is the
linewidth of the SiH stretch vibration. The inset of Fig. 1
shows an expanded view of this line, measured with higher
resolution (0.25 em -I) to resolve its linewidth (0.95 cm--
1
).
Since the homogeneous linewidth of the SiH stretch is ex-
pected to be - 1 em - 1 at room temperature, due to anhar-
monic coupling to lower frequency modes, 13 the inhomogen-
eous contribution to the linewidth is likely to be small ( 0.1
cm-
1
). It should be stressed that, since the frequency of the
SiH stretch vibration depends on how the Si is bonded to its
neighbors,14 and since its weak dynamic dipole moment
(e* - 0.1) 15 precludes line narrowing due to dipole interac-
tions,I6 the SiR measured width is a sensitive measure of
inhomogeneities. I}
The mechanism responsible for producing these dra-
matically different surface morphologies is elucidated by the
observation that one can go reversibly from a rough surface
(Fig. 1, dashed curve) to a smooth, ideally terminated sur-
face (Fig. 1, solid curve) by dipping the H-passivated sur-
face in dilute aqueous HF or the pH modified HF solution,
without an intervening chemical oxidation. Since addition of
silicon atoms to the surface from the solution is unlikely, the
surface chemical changes must be occurring via etching of
the H-terminated surface. In contrast, concentrated HF so-
lutions do not alter the surface morphology (no change in
concentration of defect modes), implying that the HF acid
itself does not etch the surface.
A possible explanation of the observed etching in dilute
solutions is the slow oxidation of the H-passivated sur-
face,1718 followed by a fast removal of the surface oxide by
HF. The second step involved in this etching process can be
eliminated by removing the HF from the aqueous solution.
(0) ,1
I
LlR/R '
5,,0-4
PER
REFLECTION
8"45'
SI(HH
S-POLARIZATION
P- POLARIZATION
...
--0-- i
o 25cm-' PER i
RESOLU-ION I
2062 2086
-'- -
k-__ __ __ __ ___ -LI __ __
2020 2060 2,40 2'80
FREQUENCY (em-')
FIG. 1. Internal icficction spectra ofHF-treated Si(1l1) surfaces: Ca) sur-
face treated with pH modified buffered HF (pH = 9-10) (solid curve: res-
olution = 0.5 cm-- ') and with dilute HF (100:1 H
2
0:HF) (dashed curve:
resolution = I em-'); (b) s polarization for surface treatment with pH
modified buffered rtF C pH = 9-10) (resolution = 2 cm - ;). Inset: High-
resolution spectrum of Si ( Ill) surface treated with pH modified buffered
HF( pH = 9-10).
657 Appl. Phys. Lett, Vol. 56, No.7, 12 February 1990
Confirming this idea, the spectra in Fig. 2 show the attack of
a RF prepared Si(111) surface by water. Surface prepara-
tion in buffered HF (pH = 5.0) results in a more defective
surface, with coupled monohydride, dihydride, and trihy-
dride clearly visible [Fig. 2 (a) ]. Immersion in water for 5
min results in the reduction in the amount of coupled mono-
hydride and trihydride and complete removal of the dihy-
dride, accompanied by a narrowing and strengthening of the
line at 2083.7 cm '. The difference spectrum (Fig. 2, top)
shows these effects more clearly. The observed stability of
the ideally terminated Si( 111) terrace (2083.7 cm- i mode)
accounts for the preferential production of the 8i ( 111 ) sur-
faces.
Given that (111) planes are produced preferentially, it
would be expected that this phenomenon would lead to
roughening of the Si(1oo) surface. While concentrated HF
removes silicon oxide leaving a rough H-passivated SiC 100)
surface mostly covered with dihydride [Fig. 3(a)],8 treat-
ment with buffered HF (pH = 5.0) still results in a rough
surface [Fig. 3(b)] but with an increase of monohydride
modes (both coupled and uncoupled), indicating the forma-
tion of small (111) facets.
The mode at 2083.7 em -1 [arrow in Fig. 3 (b) ], charac-
teristic of the ideally terminated (111) terraces, appears
both ill s- and p-polarized spectra (dashed and solid curves,
respectively) with similar intensities, showing that the SiH is
tilted from the (100) surface normal as expected for (111)
facets. The other peaks in the range 2070-2090 cm-- I are
characteristic of other monohydrides (coupled
H--Si--Si-H or H attached to Si atoms with Si-Si bond-
ing arrangements different than for bulk 8i) .
The startling result of this study is the formation of a
homogeneous rnonohydride phase on the ( 111) surface us-
ing these pH-modified buffered HF solutions. To our knowl-
edge, this is the first time the ideally terminated Si ( 111 )
surface has been produced, whether in solution
7

R
or by
atomic H exposure of clean surfaces in ultrahigh vacuum.
I9
To understand how this termination is achieved, one must
understand why the surface is H terminated in the first place.
SI HH)
2020 2060
I
l>R/R.
5. iO-
4
PER

-t--
0.5 cm-
i
RESO,-UTION
(0)
(b)
2100 2140 2180
FREQUENCY (0",-1)
FIG. 2. (a) Internal reflection spectrum of Si( Ill) treated with buffered
HF (pH = 5.0) and (b) subsequently rinsed in water. (b/a) Difference
spectrum showing the decrease in the intensity of defect modes and the in-
crease of the ideally terminated terrace mode.
Higashi et at 657
Downloaded 18 Mar 2013 to 31.147.117.245. Redistribution subject to AIP license or copyright; see http://apl.aip.org/about/rights_and_permissions
ibJ
51 (100)
2020 2060 2.00 2140
FREQUENCY (cm-'i
--

Rf:SOLUTION
I
AR/R'
2"
PER
REFLECTION
2<60
FrG. 3. (a) Internal reflection spectrum of Si( 100) upon oxide removal
with concentrated HF; (b) p-po1arization (solid) and s-polarization
(dashed) spectra of the surface subsequently treated in buffered HF
(pH=SoOj.
Indeed, previous authors have argued that fluorine termin-
ates HF-treated surfaces
20
because of the relative strength of
the SiF versus the SiH bonds (-6 eV vs -3.5 eV). Further-
more, F termination of the silicon interface is expected to be
the final step of oxide removal, given the accepted mecha-
nism of oxide etching. Therefore, the observation of atomic
H on the clearly indicates that the termination is
largely determined by reaction kinetics rather than thermo-
dynamics alone.
A simple explanation for the observed H termination
was first proposed by Ubara et al. 5 on the basis of infrared
measurements of HF-treated microcrystalline silicon sam-
ples. They postulate that F-terminated silicon complexes are
unstable in HF solutions due to strong polarization of the
Si--Si back bonds (e.g., + F
3
) facilitating their
attack by HF molecules. This results in reactions such as
=Si-SiF
3
+ HF- SiR + SiF
4
which release silicon
fluorides into solution leaving a H-terminated surface be-
hind. Accurate quantum chemical calculations of the activa-
tion barriers for these types of reactions support their gen-
eral hypothesis.
21
In addition, the question of why the SiH
bonds formed are not attacked by HF is also answered by
these studies. The principal reason is that a reaction such as
=SiH + HF -+ _SiF + H
2
, though exothermic, has an acti-
vation barrier which is significantly higher than that for the
Si-Si bond cleavage reaction discussed above.
The mechanism described above is valid for all forms of
H termination (mono-, di-, and tri-hydride) and depends
only on the initial structure of the Si/ Si0
2
interface, since
only the first Si atom is removed in the process. However,
our best surfaces are achieved in solutions where etching of
the silicon surface itself occurs, as evidenced by the reversibi-
lity of the spectra obtained with different solutions. Highly
basic solutions (e.g., KOH) are known to etch silicon sur-
faces anisotropically [etch rates of Si ( 100) vs Si ( 111) of
greater than 10 to 1], suggesting that the OH concentration
in these solutions plays a role in determining the observed
surface morphologies. The results of Fig. 2 support the idea
that the H-terminated steps and defects characterized by
658 Appl. Phys. Lett., Vol. 56, No.7, 12 February 1990
coupled monohydride, dihydride, and trihydride are prefer-
entially attacked by water.
If this attack results in Si-O bonds (e.g., Si-OH for-
mati on) , preferential removal of those silicon atoms will oc-
cur in the presence ofHF. Given this proposed mechanism,
the formation of ideal H-terminated (111) terraces in the
highly basic HF solutions results from the accelerated etch-
ing of the surface defects due to the high OH concentration.
In contrast to the defective areas, the ideally terminated
Si ( Ill) terraces are extremely stable in highly acidic as well
as highly basic solutions (e.g., from Hel to NH
4
0H).
Our experimental observations of pH modified HF etch-
ing (pH>5) of Si(111) surfaces leads immediately to the
idea that anisotropic etching will occur when surfaces other
than the (111) are used. The results obtained on the SiC 100)
surfaces (Fig. 3) clearly show that the more basic the HF
solution, the more developed the ( 111) facets, although the
size of these facets remains limited, This finding stresses that
the pH of the HF solutions used to clean Si(100) prior to
oxidation must be chosen carefully to minimize roughness.
S. B. Christman, E. E. Chaban, and R. D. Yadvish are
gratefully acknowledged for their technical support.
'w. Kern, Semicond. Int. April, 94 (1984).
2F. J. Gnmthaner and P. J. Grunthaner, Mater. Sci. Rep. 1, 69 (1986).
3M. Grundner and H. Jacob, App!. Phys. A 39,73 (1986).
"Trace contamination at the 10 ppm level can reslIlt in threshold voltage
shifts, device instability, lowered transconductance, etc.
'R. Ubara, T. Imura, and A. Hiraki, Solid State Commun. 50, 673 (1984).
"E. Yab1onovitch, D. L. AHara, C. C. Chang, T. Gmitter, and T. B. Bright,
Phys. Rev. Lett. 57, 249 (1986).
"V. A. Burrows, Y. J. Chabal, G. S. Higashi, K. Raghavachari, and S. B.
Christman. App!. Phys. Lett. 53, 998 (1988).
"Y. J. Chahal, G. S. Higashi, K. Raghavachari, and V. A. Burrows, J. Vac.
Sci. Techno!. A 7, 2104 (1989).
"N. J. Harrick, I mernal Reflection Spectroscopy (Wiley, New York, 1967),
second printing (Harrick, Ossining, 1979).
lOY. J. Chabal, Surf. Sci. Rep. 8, 211 (1988).
"Origillally it was thought that the mode observed at 2077 em -I for H on
Si( Ill) 7X7 might be characteristic of the ideal terrace monohydride. It
appears, however, that the strain around the terrace atoms may be respon-
sible for shifting the frequency down by 6-7 em - I.
ICy' J. Chabal and K. Raghavachari, Phys. Rev. Lett. 53, 282 (1984).
13This is the smallest linewidth observed for any adsorbate on any substrate
at room temperature. In particular, the monohydride on Si( 100) is char-
acterized by a 3 em - I width, with 2 em - I due to dephasing and 1 em - I
due to inhomogeneous broadening. See J. C. Tully, Y. J. Chabal, K. Ragh-
avachari, J. M. Bowman, and R. R. Lucchese, Phys. Rev. B 31, 1184
(1985).
'"Yo J. Chabal and K. Raghavachari, Phys. Rev. Lett. 54, 1055 (1985).
"Y. J. Chahal, in Chemistry and Physics a/Solid Surfaces VII, edited by R.
Vanselow and R. F. Howe, Springer Series in Surface Science. Vol. 10
(Springer, Berlin, 1988) pp. 109-150.
1('See, for example, Z. Schesinger, L. H. Greene, and A. J. Sievers, Phys.
Rev. B 32, 2721 (1985), and references therein.
17D. Graf, M. Grundner, and R. Shultz, J. Vac. Sci. Techno!. A 7, 808
(1989).
IMM. Grundner and R. Schulz, AlP Conf. Proc. 167, 329 (1987).
19H. Kobayashi, K. Edamoto, M. Onchi, and M. Nishijima, J. Chern. Phys.
78, ;429 (1983); Y. J. Chabal, G. S. Higashi, and S. B. Christman, Phys.
Rev. B 28, 4472 (1983).
ellE. R. Weinberger, G. G. Peterson, T. C. Eschrich. and H. A. Krasinski, J.
App!. Phys. 60, 3232 (1986).
"G. W. Trucks, K. Raghavachari, G. S. Higashi, and Y. 1. Chahal (unpub-
lished).
Higashi et al. 658
Downloaded 18 Mar 2013 to 31.147.117.245. Redistribution subject to AIP license or copyright; see http://apl.aip.org/about/rights_and_permissions

También podría gustarte