Está en la página 1de 8

Journal of Colloid and Interface Science 276 (2004) 392399

www.elsevier.com/locate/jcis

Electrospray emission from nonwetting flat dielectric surfaces


Paulo Lozano , Manuel Martnez-Snchez, Jose M. Lopez-Urdiales
Department of Aeronautics and Astronautics, Massachusetts Institute of Technology, 77 Massachusetts Avenue, Room 37-438, Cambridge, MA 02139, USA
Received 5 January 2004; accepted 8 April 2004
Available online 10 May 2004

Abstract
Electrosprays are devices in which nanometer sized droplets and/or solvated ions are electrically extracted from a liquid surface and
accelerated to high velocities. They are usually constructed from conductive capillaries with one of the ends tapered down to a sharp tip
with the purpose of enhancing the local electric field that produces the instability that develops into the Taylor cone structure from which
emission occurs. In an alternative configuration, the conductive needles are replaced by small holes through a dielectric, nonwetting flat
block. An electrostatic model shows that if the emitter material has low dielectric constant then the local electric field near the emission site
is enhanced in a very similar way as with sharp metallic needles. Furthermore, the combination of the nonwetting property of the material and
the sharp corner formed in the holesurface interface effectively anchors the Taylor cone to the edge of the hole, thus simplifying the process
of manufacturing. The possible microfabrication of this configuration makes it especially attractive for producing arrays of large numbers of
individual emitters. Such arrays may find use as space propulsion thrusters and in the analytical industry to improve the characteristics of
mass-spectrometric analyses.
2004 Elsevier Inc. All rights reserved.
Keywords: Electrospray; Colloid thruster; Micropropulsion; Wetting; Mass spectrometry; Arrays

1. Introduction
Electrospray science has become, in a relatively short
time, a prolific activity that has found or is seeking applicability in very distinct technological areas. In its most
simple form, an electrospray emitter is a device that extracts
and accelerates charged particles, i.e., liquid droplets and/or
solvated ions, by means of an electric field generated after
applying a potential difference between a conductive liquid
and an extraction electrode. As the electric pressure balances
the surface tension, the shape of the liquid interface becomes
initially unstable and then quickly deforms into a stable conical structure. It was Taylor [1] who first recognized that,
in the ideal case where the cone surface is an equipotential,
the normal electric field increases toward the tip as r 1/2
(r is the radius from the cone apex) and the cone semi-angle
has a constant value of 49.3. In an actual electrospray, the
singularity is avoided after the region close to the apex deforms into a cylindrical jet that, depending on the nature of
the liquid and the operating conditions, either breaks up into
* Corresponding author. Fax: +1-617-258-5940.

E-mail address: plozano@mit.edu (P. Lozano).


0021-9797/$ see front matter 2004 Elsevier Inc. All rights reserved.
doi:10.1016/j.jcis.2004.04.017

droplets, with possible ion emission from the cone-jet neck


region [24], or forms a protrusion from which ions are emitted as is the case for liquid metal ion sources (LMIS) [5] or
sources using some types of ionic liquids [6,7]. The liquid
is usually transported using a conductive capillary with one
of its ends (the emission end) carefully sharpened to provide adequate electric field enhancement such that voltages
required to induce the surface instability are as small as possible.
Electric space propulsion was identified as one of many
applications for electrosprays and it was actively pursued
during the 19601980 period by several research groups
[811]. Thrust is simply generated by ejecting charged particles at high speeds. These colloid engines were in the most
part developed to provide main propulsion for different types
of spacecraft. Since the thrust magnitude that can be obtained from a single emitter is very small in comparison
to most other propulsion alternatives, there was a need to
use several emitters in parallel, forming arrays. Mainly because of manufacture limitations, the number of emitters was
in general insufficient to provide adequate levels of thrust.
To compensate, some of these devices were operated with
very high extraction voltages in excess of 10 kV. These and

P. Lozano et al. / Journal of Colloid and Interface Science 276 (2004) 392399

other issues like the increased competition from other electric propulsion devices made colloid thrusters impractical
and their development slowed down considerably.
Recently, a renewed interest in colloid thrusters has
emerged mainly because (a) they seem to be ideal for missions that require very little thrust to compensate for orbital
perturbations such as high altitude atmospheric drag or solar pressure, or missions involving small, power-limited
satellites [12,13], and (b) given the intrinsically small dimensions of individual emitters, it appears to be feasible to
construct very dense arrays taking advantage of the considerable know-how and variety of microfabrication techniques
well developed for the electronics and MEMS industries
[1416] and exploit modularity to increase thrust to competitive levels.
In addition to space propulsion, the microfabrication of
electrospray arrays is of particular interest for the analytical
industry to increase the reliability and magnitude of signals in mass spectrometry. Xue et al. [17] recognized that
electrospray emission was possible from a flat edge formed
when bonding two glass wafers with one of them containing large aspect ratio parallel channels (length of 3.5 to 5 cm
and noncircular cross sections of 60 m wide and 25 m
deep). Glass, however, is highly hydrophilic and it was difficult to avoid wetting of the emitting surfaces. A hydrophobic
reagent coating was applied to the emitter surface and the
device worked but at relatively high voltages (4.2 kV). They
suggested that using plastic materials instead of glass might
help to eliminate the wetting problem. Ramsey and Ramsey [18] proposed a similar concept also made out of glass.
The flat emission surface was polished and stable emission
was obtained at voltages 3 kV using water. Zhang et al.
[19] mentioned that a tip is not necessary in these type of
designs suggesting without further analysis that the field
strength appeared to be high enough to form a Taylor cone.
They also recognized after experimenting also with glass
surfaces that avoiding wetting (or, as they refer to, droplet
formation over the flat edge) is necessary to produce a useful device for mass spectrometry. Schultz et al. [20] opted
for a two-dimensional silicon based featured array, arguing
that previous flat designs are flawed for two reasons: (a) liquid spreading over the flat surface adjacent to the channel
openings is unavoidable to some degree, and (b) flat surfaces
require impractically high voltages to overcome the surface
tension. These reasons may be applicable to previous designs in glass but, as we will see, may not be generalized
to every type of flat emitter. Tang et al. [21] developed a featured electrospray array similar to that in [20] but fabricated
on polycarbonate substrates. They also found that wetting is
an important issue, even for their plastic material. The polycarbonate emitter was treated with a CF4 plasma to increase
hydrophobicity.
The main objective of this work is to understand in more
detail under what general conditions emission from such featureless flat surfaces can be produced. An electrostatic model
is developed and discussed in the following section with the

393

purpose of identifying the different parameters that determine the magnitude of the emission voltage. In Section 3 we
present some experimental results showing that the voltages
required to generate Taylor cones of a given size are comparable and even smaller in this flat configuration than for
cones formed on the tip of a metallic capillary, provided that
the flat surface is part of a nonconductive material with low
dielectric constant and that wetting by the working liquid is
low enough such that the resulting cone is firmly anchored to
the edge of the surface aperture. Model predictions are compared against observed values in that same section. Finally,
concluding remarks are presented in Section 4.

2. Electrostatic model
Fig. 1 shows a schematic of an electrospray emitter made
out of a flat featureless dielectric block in which a small perforation has been made creating a cylindrical cavity in which
the liquid flows from an upstream reservoir (not shown.) At
some distance from the surface lies an extraction electrode.
A voltage Vm is applied to the conductive liquid through an
internal electrode. Suppose that the liquid forms a meniscus
at the surface of the block as shown in Fig. 1a, then, as Vm is
increased, electric forces on the surface of the liquid start to
deform the meniscus until a critical value Vc is reached and
a stable Taylor cone forms (Fig. 1b).
An analytical representation of the nonemitting configuration of Fig. 1a can be constructed using the sphere-on-cone
(SOC) model [22] with proper boundary conditions (Fig. 2).
In this model, a sphere with radius a is superimposed to the
origin of an infinite cone with half-angle . Both geometrical
structures are considered to be perfect conductors such that
they simulate the conductive liquid inside the dielectric material. The objective is to compute the potential distribution
in the region outside of the conical and spherical surfaces, inside and outside the dielectric. In particular, one can choose
an equipotential surface that, although not perfectly cylindrical, closely matches the liquid surface (liquid column and
meniscus) in the real device shown in Fig. 1a.

Fig. 1. Dielectric colloid thruster emitter. (a) Vm < Vc and (b) Vm > Vc .

394

P. Lozano et al. / Journal of Colloid and Interface Science 276 (2004) 392399

 = 0 for the electric


Furthermore, solving Gauss law D
 on a pill-box surface over the interface bedisplacement D
tween the vacuum and the dielectric yields
i = o ,

(5)

where the primes denote / . Equation (5) represents the


reduction of the normal electric field by due to the presence
of the dielectric. It follows that
A 0
P = BP 0 + CQ0 ,
(6)

Fig. 2. The SOC model with > 1 for x < 0 for a sphere with radius a
and an infinite cone with half-angle . An equipotential surface is chosen
to represent the conductive liquid.

Taking the origin in spherical coordinates as shown in


Fig. 2 and considering axial symmetry (no dependence),
Laplaces equation 2 = 0 can be solved using the method
of separation of variables in terms of its radial and angular
components,
(r, ) = R(r)( ).

(1)
r

r 1 ,

and
Solutions of the radial part are of the form
where is the separation constant, while the angular part depends on Legendre functions P (cos ) and Q (cos ). The
complete solution is an infinite superposition of the individual solutions, but the presence of the equipotential cone
restricts the number of these functions to a single value of
> 0 for a given cone angle . Furthermore, it is required
that the solution remains finite when = 0. Because Q (1)
is not defined, the potential is separately solved in two different regions. We call o = o (r, ) the potential distribution in the region outside of the dielectric (vacuum, = 1)
where < /2, while i = i (r, ) is defined as the solution inside of the dielectric ( > 1) where /2 < < c with
c = . Defining (for convenience) as zero the potential
over the SOC structure, we obtain
 
 +1 
r
a
o = AP (cos )

,
a
r
 
 +1 

 r
a
, (2)

i = BP (cos ) + CQ (cos )
a
r

where P 0 = P / |0 and Q0 = Q / |0 are given by


the following expressions [23]:
   +2 
2
2
0
P = sin
,

2 +1

2
   +2 

2
0
Q = cos
(7)
.

2 +1
2
The potential over the surface of the cone is defined as
i ( = c ) = 0, therefore
BP (cos c ) + CQ (cos c ) = 0.

(8)

Equations (3) and (6) can be solved directly to find the values
of constants B and C in terms of A,
 0
   0

Q
Q
Q0
Q0
B =A
,

P 0 P0
P 0
P0


   0
1
Q0
Q
C = A
(9)

P 0
P0
We use these values in (8) to find the form of the equation
that determines = (, ),



 0
Q0
1
Q
Q (cos c ) = 0. (10)
P (cos c )

P 0 P0

(3)

The presence of the infinite cone is revealed by the term r ,


which makes the potential decrease as r increases, provided
that the constant in front of this term is negative. The potential, however, does not reach a constant value. We therefore
need to establish a reference potential downstream from the
SOC structure. This can be accomplished by assuming that
the equipotential (V ) representing the extraction electrode
is fixed at an axial distance R from the sphere center, then
o (R, 0) = V . Since P (1) = 1, we find that
 
 +1 1
a
R

.
A = V
(11)
a
R

where P0 = P (0) and Q0 = Q (0). The values of these


quantities are given by [23]
   +1 
1
2
0
P = cos
,

2 +2

   +1 
2

0
Q =
(4)
sin
.

2
2 +2
2

The final form of the potential distribution is then


 r   a +1

o = V  a   r +1 P (cos ),
R
Ra
a


  0
 
o
Q0
Q
1 Q (cos )

.
i =

Q0
Q0

P (cos )
P 0 P0
P 0
P 0

(12)

where A, B and C are constants. At = /2, the solutions


in (2) must be continuous along the boundary; then,
AP0 = BP0 + CQ0 ,

P. Lozano et al. / Journal of Colloid and Interface Science 276 (2004) 392399

395

It is seen that the solutions are identical in the case where


= 1, i.e., there is no dielectric but vacuum in the region
x < 0 of Fig. 2. As the dielectric constant increases, the solutions depart from each other.
The relevant parameter that controls the onset of the
meniscus instability is the local electric field. To estimate
the magnitude of the field we should first select an appropriate equipotential surface that matches as close as possible
the liquid/meniscus shape as shown in Fig. 2. We take the
radius of the meniscus as rm = f a, where f > 1 is the factor that determines the selected equipotential. If we assume
that the meniscus is spherical in shape and that the pressure
inside the liquid is negligible, then the electric forces will
overcome the surface tension when
2
1
0 En2 >
,
(13)
2
rm
where is the liquidair interfacial surface tension and 0 =
8.854 1012 F/m is the permittivity of vacuum. Assuming,
as is generally the case, that a  R, the normal electric field
En = o /r|r=rm can be obtained from (12),
 

V rm 
+ ( + 1)f 21 P (cos ).
En =
(14)
rm R

Fig. 3. Map of = (, ) according to Eq. (10) for five dielectric constant


values: 1, 2, 6.5, 20, and .

It can be seen from (14) that the electric field depends on


the angular position. This means that, under a fixed configuration, condition (13) could be satisfied at some points over
the liquid surface while not at others. Since the instability
growth into a Taylor cone is a dynamic phenomenon it is
likely that the meniscus instability will start as soon as the
condition is met anywhere over the liquid surface. We assume that this is indeed the case and take the maximum value
of the Legendre function P (1) = 1, which occurs over the
x axis. Condition (13) can be written as

rm
Vm >
F (rm , R, , f ),
0
F =2

(R/rm ) + f 21 1
,
+ ( + 1)f 21

(15)

where Vm is the potential difference between the meniscus


(represented by the selected equipotential) and the extraction
electrode, i.e., the applied potential.
2.1. Model discussion
For a fixed geometry, function F in (15) determines the
magnitude of the voltage required to produce the liquid surface instability and it depends strongly on the value of .
Fig. 3 shows the variation of = (, ) for different dielectric constants.
It is relevant to note that the sensitivity of decreases notably fast with increases in , producing very little difference
for dielectric materials characterized by values of > 20 for
angles ( < 10 ) that result in the best approximation to
the actual liquid/meniscus geometry. It is seen also that all
curves collapse to a value of = 1 for = 90 , although

Fig. 4. The effect of a dielectric on the starting voltage. Example with


= 4 and f = 1.15. Curves are normalized to F1 (10) = 2.24, F1 (20) =
2.52, F1 (50) = 2.93, and F1 (100) = 3.28.

for
1 such collapse practically occurs at an angle of approximately 33 . It can be further noted that for vacuum
conditions the cone angle is 49.3 when = 1/2 consistent
with a normal electric field that varies as r 1/2 as predicted
by Taylors analysis [1].
Fig. 4 shows the variation of F (rm , R, , f ) as a function
of dielectric constant for different values of R/rm , a cone
angle of = 4 and an equipotential geometric factor of
f = 1.15. These values were selected because they provide
a good approximation to the cylindrical channel and meniscus of the experimental device described in the next section.
Each curve in Fig. 4 is normalized by its respective value
when = 1 (F1 in the figure), thus showing the way in which
the minimum meniscus voltage (Vm ) required to produce the
instability in (15) varies because of the presence of the di-

396

P. Lozano et al. / Journal of Colloid and Interface Science 276 (2004) 392399

electric. It is seen that the effect of using materials with high


dielectric constants is to reduce the electric field, therefore
increasing the required voltage. Here we note again that the
rate at which the voltage changes decreases as increases
until the voltage becomes basically flat for large dielectric
constants. For example, we see in Fig. 4 that the voltage
needs to be increased by 37% for a material with 12
(silicon) with respect to a material with 2 (Teflon) if
the extractor is placed relatively far away from the meniscus
(R/rm = 100).
The net effect of introducing an insulator material with
low dielectric constant is almost as having the liquid surrounded by vacuum. The dimensions of this floating liquid
column is what determines the starting voltage. If the dielectric material is substituted by a conductor, the voltage
required to generate the meniscus instability would be much
larger, since a flat conductive surface will behave much like
a parallel plate capacitor with just a modest field increase in
the region close to the meniscus. In this case = 90 , corresponding to a value of = 1 (independent of ) as shown
in Fig. 3. This means that function F in (15) increases much
more rapidly with R/rm as it becomes linear in it. This result
places a strong limitation on the development of flat emitters
in the sense that any residual electrical conductivity must
be very low to make the charge relaxation time r = 0 /
as large as possible and delay the subsequent behavior of the
material as a conductor. Most pure plastics have large resistivities of about 1/ > 1016  m (r = 1.8 105 s for
Teflon), but in silicon (the preferred microfabrication material) the resistivity is limited to 1/ < 104  m (r = 1 s)
for extremely pure samples. The high end in resistivity for
most commercial Si wafers, however, lies below 10  m
(r = 1 ns).
Most liquids of interest for space propulsion have conductivities around 1 1 m1 , the corresponding charge
relaxation time is orders of magnitude shorter than the
inertia-dominated time required to form a typical micrometer sized Taylor cone. The latter is consequently the dominant time scale in the liquid system. If this cone formation
time is somewhat longer than the charge relaxation time in
the emitter material, then the electric field enhancement effect will be lost and the voltage required to produce the
liquid instability will be much larger. We note that this is
indeed the case for silicon as long as the time to form a cone
is longer than r . Furthermore, even with good insulator materials, the field enhancement effect will decrease with time
due to the finite nature of r until it is lost. It is doubtful
that the hysteresis behavior observed in Taylor cones (see
Section 3) will be enough to sustain the conical structure.

This means that continuous operation over times comparable


to r would not be possible unless the voltage is increased
considerably, which is of course undesirable and, in some instances, not feasible. This situation is of critical importance
if these emitters are to be used as space thrusters since most
spacecraft are expected to last for several years, while r is
in general much lower than that.
Electrosprays, however, can be operated in bipolar mode
by alternating the applied voltage. In the case of flat dielectric emitters, voltage alternation would avoid the polarization of the material surface and the subsequent loss of the
electric field enhancement provided that the alternation frequency is higher than 1/r . Furthermore, the problem of
spacecraft electric neutralization that arises when ejecting a
single polarity (positive ions in most electric propulsion engines) can be solved without using external electron sources.
Voltage alternation is also the only way to avoid electrochemical decomposition in some ionic liquid thrusters [7].
On the other hand, many other electrospray applications require continuous operation for not more than a few hours,
in those cases an appropriate selection of an insulator material would provide adequate emission without recurring to
voltage alternation.

3. Experimental observations
A series of experiments were carried out at atmospheric
pressure conditions in which an electrospray emitter was
positioned at some controlled distance R from a grounded
metallic plate, which served as the extraction/collector electrode. The working liquid was ethylene glycol (EG, purity
as purchased), made conductive by adding 0.6% W of LiCl.
Room temperature properties of the EG mixture are listed in
Table 1. Conductivity was measured (Jenway 4320) and verified using a couple of probes with different cell constants
[1,10]. Measurements were in agreement to about 5%.
The emitters were made out of flat dielectric blocks 7
mm thick. Holes were drilled using standard drill bits gauges
80 (343 m OD), 78 (406 m OD) and 76 (508 m OD).
Drilling was performed as carefully as possible but given the
small sizes involved, the nature of the drilled materials and
possible misalignments, the actual bore diameter was not always the same as the drill bit size. Instead, each hole was
independently measured with an accuracy of 15 m using an optical videomicroscope. The distance between a hole
aperture and the block edges is large enough compared to the
bore size that the region adjacent to each opening can be regarded as an infinite plane. Four different dielectric blocks

Table 1
Physical properties of ethylene glycol + LiCl
Density
(kg/m3 )
1113

Surface tension
(N/m)

Viscosity
(cP)

Dielectric constant

Conductivity
(Si/m)

0.048

21.0

38.66

0.077

P. Lozano et al. / Journal of Colloid and Interface Science 276 (2004) 392399

397

Table 2
Dielectric constant and contact angle with the EGLiCl mixture for different materials
Material

Dielectric constant

Contact angle
w (deg)

Teflon
Polyethylene (UHMW)
Polycarbonate
PVC (plasticized)
Stainless steel

2
2.3
3.17
6.5

96.7 1.3
65.2 1.3
55.4 1.7
68.2 2.1
43.5 3.1

were used; their materials and relative dielectric constants


can be found in Table 2. For reference, a manually sharpened stainless steel capillary with an OD of 1.6 mm and an
ID of 178 m was also used in our experiments.
Each material is subject to some degree of wetting from
the EGLiCl mixture, which can be characterized if the contact angle w of the liquidsolidair interface is known.
Larger angles indicate better nonwetting properties. The
contact angle was determined by depositing a small EG drop
(circular cross section, submillimeter in size) over a material sample and measuring the diameter w of its base along
the liquidsolid interface and its height h over the solid surface with the help of a microscope. The contact angle was
then approximated using w 2 tan1 (2h/w). Several measurements were performed and the results are summarized
in Table 2.
Our objective is to establish the potential difference at
which the liquid meniscus becomes unstable and deforms
into an emitting Taylor cone. It is desirable that, except for
the emitter type and size, conditions remain as fixed as possible between observations. One useful way to achieve this
is to control the flow rate Q such that all experiments are
performed with a unique electrospray current I . The flow
rate is controlled by the liquid injection system. There are
several ways by which the liquid can be injected, for example using a syringe pump or a pressurized system. The
syringe pump is generally inadequate, especially for controlling very small flow rates [24]. For this work we have
selected a simple variation of a pressurization system using gravity as the driving force. The system is schematically
depicted in Fig. 5. The pressure is basically controlled by
adjusting the height h of the liquid reservoir over the emission surface. The reservoir is large enough that the liquid
level remains practically constant as liquid is consumed. The
flow rate is determined by the interplay of h, the viscous
transport through a PEEK tubing feed line (1.6 mm OD,
127 m ID and length of 273 mm) and the nonzero, but practically constant, pressure in the cone itself. Fig. 6 shows a
typical emitting Taylor cone formed at the tip of the stainless steel sharpened capillary. The electrospray current was
measured with an electrometer in series with the electrical
circuit.
Fernndez de la Mora et al. [25] investigated the electrospray emission from a number of different liquids and found

Fig. 5. Schematic of the experimental setup.

Fig. 6. Taylor cone formed on a stainless steel emitter as seen through a


microscope, with = 1.7.

that current and flow rate are related by



Q
,
I = f ()
(16)

where the factor f () is fluid dependent, taking a value of


about 17 for EG [25]. The flow rate was not measured directly in our experiments, but was calculated from Eq. (16)
using the measured current. It was also established by Fernndez de la Mora that emission becomes unstable and eventually stops when the value of the nondimensional quantity
( is the liquid density),

Q
,
=
(17)
0
drops below unity. In our characterization experiments, the
cone remains stable for values of ranging from 0.73
(6.24 nl/min, 54.5 nA) to 3.54 (144 nl/min, 263 nA). Below the lower limit, the cone collapses and emission is no
longer possible even if the voltage is increased, while for
values higher than the upper end, the conical structure is

398

P. Lozano et al. / Journal of Colloid and Interface Science 276 (2004) 392399

Table 3
Cone base (bore) diameters and extraction and extinction voltages for R =
3.5 mm and = 1.7
Material

2rm 15 (m)

R/rm

Voff (kV)

Von (kV)

Stainless steel
PVC
Polyethylene
Teflon

366
379
409
333
388
417
638

19.1 1.6
18.5 1.5
17.1 1.2
21.0 1.9
18.0 1.4
16.8 1.2
11.0 0.5

3.6
3.1
2.52
2.4
2.6
2.5
3.15

3.85
3.3
2.92
2.6
2.9
2.9
3.5

Fig. 7. Some microscope images from flat dielectric surfaces. (a) Static
meniscus over PVC with no applied voltage. Stable Taylor cones over
(b) PVC and (c) UHMW-PE with = 1.7 and (d) over Teflon with = 2.

no longer capable of extracting the amount of fluid supplied, the surface oscillates violently ejecting macroscopic
droplets toward the collector, many times causing electrical breakdown. A value of = 1.7 (33.8 nl/min, 127 nA)
was selected for all our experiments to guarantee the current stability that a well-behaved Taylor cone provides. The
collector electrode was positioned using a micrometer translational stage at a distance of R = 3.5 mm from the flat
emitter surfaces (or from the metallic capillary tip) for all
measurements.
The sequence of events toward the formation of Taylor cones is as follows. First, enough pressure through h
is provided to counteract the surface tension such that a
semispherical liquid meniscus forms over the surface of the
emitter, as shown in Fig. 7a. Then, a potential difference Vm
between the conductive liquid and the collector electrode is
applied and increased to generate an electric traction over
the liquid surface that also acts against the surface tension.
An effort is made to keep the shape of the meniscus as fixed
as possible through reductions in h as Vm is increased. At
some particular voltage Vm = Vosc , the shape of the meniscus oscillates rapidly and finally a stable Taylor cone forms
when the potential is increased further to Vm = Von . Hysteresis can be observed as the voltage is reduced from Vm = Von
through Vm = Vosc and the Taylor cone remains well formed
until a lower value of Vm = Voff is reached. It appears that
the energy needed to distort the liquid surface into a conical
structure is larger than the energy to maintain it. This behavior introduces some ambiguity in determining the value of
the emission voltage.
Figs. 7b7d show some typical images of cones formed
over dielectric surfaces as seen through a microscope. Even
though the surfaces were meticulously cleaned, it can be observed that, due to the mechanical drilling process, some
burr persisted close to the bore edges, which was difficult
to eliminate. These distortions produced little or no effect
in Teflon (the least wetting material) but were significant

Fig. 8. Extraction potential curves as given by [15] with R = 3.5 mm,


= 4 and f = 1.15, and experimental operational voltages obtained for
different dielectric materials as functions of R/rm . The upper curve represents Eq. (18). Also shown are voltages for the stainless steel emitter.

in others, particularly in samples of polycarbonate, as they


changed the wetting properties of the region adjacent to the
hole. No stable cone was obtainable with this material since
the size of the static meniscus was larger than the bore of the
hole. Apparently, such behavior destroys part of the electric
field enhancement effect and breakdown occurred for larger
values of Vm before a stable cone was formed. Wetting is
also important for stainless steel. Because of the way the
capillary was sharpened, the tip resembled a truncated cone
with Taylors structure covering the truncated cross section
(Fig. 6). The diameter of this section was about twice the
capillary ID. The reservoir height h needed to be carefully
controlled to avoid wetting of the outer surface of the capillary.
Stable cones were obtained for stainless steel, PVC,
UHMW-PE and Teflon. The resulting emission (Von ) and
extinction (Voff ) voltages, cone base diameters (2rm ) and
R/rm ratios (including the 15 m uncertainty in measurements using the microscope) are summarized in Table 3.
These results are plotted in Fig. 8 together with curves
obtained from the electrostatic model as given by Eq. (15)
using the dielectric constants of plasticized PVC, UHMWPE and Teflon and a geometric factor of f = 1.15 and a cone

P. Lozano et al. / Journal of Colloid and Interface Science 276 (2004) 392399

half-angle of = 4 , values chosen to match as close as possible the equipotential surface (see Fig. 2) of the model to
the cylindrical shape of the liquid inside the dielectric and
the semispherical meniscus protruding from it. Also shown
in Fig. 8 are the operational voltages of the sharpened stainless steel capillary and a theoretical prediction of its starting
voltage based on the following expression obtained from a
geometrical analysis [26,27],



R
rm
ln
,
Vstart =
(18)
0
rm
where the shape factor takes values between 2 and 4
( = 2 was used in Fig. 8). The accuracy of (18) in predicting the starting voltage of needle-type emitters has been previously established [28] for LMIS and electrospray sources.
These experimental results, together with the model predictions, suggest that voltages required to produce stable
cones of a given size over dielectric surfaces are lower than
those for the metallic emitter. The experimental effect of dielectric constant on voltage cannot be clearly distinguished
between Teflon and UHMW-PE, but it is definitively evident
for PVC, following the tendency anticipated by the model.
It is also observed that experimental values for Von and Voff
decrease with increasing R/rm , as expected. Nevertheless
there are some marked differences for a couple of Teflon
cases. These differences are likely to be attributed to the
model loss of accuracy as the a  R approximation losses
validity for the case when R/rm = 11, while an increased
experimental uncertainty in determining rm could contribute
to the discrepancy for larger distance ratios, in particular
that with R/rm = 21. In any event, the static model introduced in this work provides an adequate first approximation to the starting voltage of Taylor cones on flat dielectric
surfaces, even though the phenomenon is dynamic in nature.
4. Conclusion
We have presented some experimental evidence along
with an analytical model to study Taylor cone formation over
featureless dielectric flat surfaces. These results show that
emission is possible at relatively low voltages as long as the
material used has a low dielectric constant and wetting from
the working liquid is poor. The mathematical model developed in Section 2 can be used to estimate the magnitude of
the voltage required to produce emission provided that its
parameters, in particular the geometric factor f and the cone
angle , are properly selected.
As mentioned in the introduction, an important effort is
being carried out to build dense emitter arrays in order to
provide higher thrusts for space propulsion or higher signals
for mass spectrometry. Constructing arrays of holes over dielectric materials can be relatively easy to do, especially if

399

no particular shaping of the region surrounding each hole is


required. For example, pulsed UV laser drilling can be used
to produce high aspect ratio holes down to about 10 m in
diameter. Burr formation, which was an issue for our mechanically drilled holes, can be minimized by decreasing the
pulsing period, therefore reducing heat transfer and material
stresses. Small sizes would bring the emission voltage down
if the extractor electrode is properly designed. For example,
using a geometric factor f = 1.15, a cone angle of = 4 ,
rm = 5 m and taking R/rm = 50, meaning that the extractor is placed 250 m from the emitter, we calculate that the
instability voltage as predicted by Eq. (15) using Teflon and
EG is just 557 V.

References
[1] G.I. Taylor, Proc. R. Soc. London A 280 (1964) 383.
[2] M. Gamero-Castao, J. Fernndez de la Mora, J. Chem. Phys. 113
(2000) 815.
[3] M. Gamero-Castao, Phys. Rev. Lett. 89 (2002) 14.
[4] P. Lozano, M. Martnez-Snchez, AIAA Meeting Paper 2002-3814
(2002).
[5] W. Driesel, Ch. Dietzsch, J. Vac. Sci. Technol. B 14 (1996) 5.
[6] I. Romero, R. Bocanegra, J. Fernndez de la Mora, M. GameroCastao, J. Appl. Phys. 94 (2003) 3599.
[7] P. Lozano, M. Martinez-Sanchez, in preparation.
[8] V.E. Krohn Jr., Prog. Astr. Rocket. 5 (1961).
[9] P.W. Kidd, H. Shelton, in: 10th AIAA Electric Propulsion Conference,
1973.
[10] H.L. Daley, J.F. Mahoney, J. Perel, in: 10th AIAA Electric Propulsion
Conference, 1973.
[11] M.N. Huberman, S.G. Rosen, J. Spacecraft 11 (1974) 7.
[12] J.G. Reichbach, R.J. Sedwick, M. Martnez-Snchez, AIAA Meeting
Paper 2002-3646 (2001).
[13] M. Gamero-Castao, V. Hruby, D. Spence, N. Demmons, R. McCormick, C. Gasdaska, P. Falkos, AIAA Meeting Paper 2003-4543
(2003).
[14] L. Velasquez, J. Carretero, A.I. Akinwande, M. Martnez-Snchez,
AIAA Meeting Paper 2003-4850 (2003).
[15] J. Stark, B. Stevens, M. Alexander, AIAA Meeting Paper 2003-4852
(2003).
[16] J. Xiong, Z. Zhou, X. Ye, X. Wang, Y. Feng, Y. Li, Microelectron.
Eng. 6162 (2002) 1031.
[17] Q. Xue, F. Foret, Y.M. Dunayevskiy, P.M. Zavracky, N.E. McGruer,
B.L. Karger, Anal. Chem. 69 (1997) 426.
[18] R.S. Ramsey, J.M. Ramsey, Anal. Chem. 69 (1997) 1174.
[19] B. Zhang, H. Liu, B.L. Karger, F. Foret, Anal. Chem. 71 (1999) 3258.
[20] G.A. Schultz, T.N. Corso, S.J. Prosser, S. Zhang, Anal. Chem. 72
(2000) 4058.
[21] K. Tang, Y. Lin, D.W. Matson, T. Kim, R.D. Smith, Anal. Chem. 73
(2001) 1658.
[22] J.W. Ward, R.L. Seliger, J. Vac. Sci. Technol. 19 (1981) 4.
[23] CRC Standard Mathematical Tables.
[24] P. Lozano, Ph.D. thesis, MIT, Cambridge, MA, 2002.
[25] J. Fernndez de la Mora, I.G. Loscertales, J. Fluid Mech. 260 (1994)
155.
[26] P.D. Prewett, G.L.R. Mair, Focused Ion Beams from Liquid Metal Ion
Sources, Research Studies Press, 1991.
[27] V. Hruby, et al., Final report, Phase I SBIR NASA, 1999.
[28] G.L.R. Mair, J. Phys. D. Appl. Phys. 17 (1984) 2323.

También podría gustarte