Vous êtes sur la page 1sur 8

REVIEW OF SCIENTIFIC INSTRUMENTS

VOLUME 74, NUMBER 2

FEBRUARY 2003

Piezoelectric droplet ejector for ink-jet printing of fluids and solid particles
Gokhan Percina) and Butrus T. Khuri-Yakubb)
ADEPTIENT, 410 Sheridan Avenue, Apartment 216, Palo Alto, California 94306-2021

Received 16 May 2002; accepted 30 October 2002


In this article, we present a technique for the deposition of inks, organic polymers, and solid
particles, using a fluid ejector. The ejector design is based on a flextensional transducer that excites
axisymmetric resonant modes in a clamped circular plate. It is constructed by bonding a thin
piezoelectric annular disk to a thin edge clamped circular plate. Liquids or solid particles are placed
behind one face of the plate which has a small orifice 50200 m diameter at its center. By
applying an ac signal across the piezoelectric element, continuous or drop-on-demand ejection of
photoresist Shipley Microposit S1400-21, S1400-27, S1805, and S1813, oil-based ink, water, or
talcum powder Mg3 Si4 O10(OH) 2 has been achieved. Successful deposition of a photoresist has
been accomplished without spinning, and thus without waste. A boundary integral method was used
to numerically simulate drop formation from the vibrating orifice. Simulations have been used to
optimize ejection performance. 2003 American Institute of Physics.
DOI: 10.1063/1.1532839

size ranges in diameter from 5 m at 3.5 MHz to 150 m at


7 kHz, the corresponding ejected drop volume ranges from
65 fl to 1.5 nl, and the corresponding flow rate ranges from
0.2 to 10 l/s.4,5 The unique features of the device are that
the fluid is not pressurized, it is not thermally actuated, drops
are uniform in size, the ejector does not damage sensitive
fluids, the fluid container is biologically and chemically compatible with most fluids, and the vibrating plate contains the
orifice as the actuation source. In Percin et al.,4,5 the device
is manufactured by silicon surface micromachining and
implemented in the form of two-dimensional arrays where
each array element has its own actuation.

I. INTRODUCTION

There is a continuing need for alternative deposition


techniques of organic polymers in precision droplet-based
manufacturing and material synthesis, such as the deposition
of doped organic polymers for organic light-emitting devices
of flat panel displays, the deposition of low-k dielectrics, and
the deposition of photoresist without spinning on large or
oddly shaped substrates.1,2 There is also a need for alternative deposition and sample preparation techniques of chemical and biological fluids in biomedical and biotechnological
applications, such as drug delivery, drug discovery, high
throughput screening, assaying, and combinatorial chemistry.
In addition, small particle ejectors are necessary for the study
of heating and combustion behavior of small solid particles,
such as coal and metals. To date, there has been no report of
a drop-on-demand solid particle ejector that can eject fine
solid particles with spatial control, although a continuous
mode pneumatically operated solid particle ejector has been
reported.3
In this article, we present a technique for the deposition
of inks, toners, organic polymers, low-k dielectrics, fuels,
small solid particles, and biological and chemical fluids, using a fluid ejector. The ejector design is based on a flextensional transducer that excites the axisymmetric resonant
modes of a clamped circular plate. It is constructed by depositing a thin piezoelectric annular plate onto a thin edge
clamped circular plate. Liquids or small solid particles are
placed behind one face of the plate which has a small orifice
at its center. By applying an ac signal across the piezoelectric
element, capillary waves are set up at the liquidair interface, and continuous or drop-on-demand ejection of fluids
and small solid particles has been achieved. The ejected drop

II. DEVICE DESIGN

A schematic of the ejector is shown in Fig. 1. A thin


sheet of brass, a shim, with a small orifice is bonded to a
piezoelectric annular disk. A cylinder attached to the shim
serves both as a fluid reservoir and to clamp the ends of the
compound plate formed of the shim and piezoelectric. The
reservoir is open, and the fluid is at atmospheric pressure. An
ac voltage is applied to the plate to set it into vibration. At
the resonant frequencies of the fluid loaded compound plate,
the displacement at the center is large. The fluid behind the

Electronic mail: percin@adeptient.com


Also at: Edward L. Ginzton Laboratory, Stanford University, Stanford,
CA 94305-4085.

0034-6748/2003/74(2)/1120/8/$20.00

FIG. 1. A schematic of the device: The lateral extent is 9 mm.


1120

2003 American Institute of Physics

Downloaded 08 Nov 2004 to 171.64.86.217. Redistribution subject to AIP license or copyright, see http://rsi.aip.org/rsi/copyright.jsp

Rev. Sci. Instrum., Vol. 74, No. 2, February 2003

Piezoelectric droplet ejector

TABLE I. Physical dimensions of the device.


Dimension

Value

Radius of brass carrier plate


Inner radius of piezoelectric Murata SWa material
Outer radius of piezoelectric Murata SW material
Thickness of brass carrier plate
Thickness of piezoelectric Murata SW material

4 mm
1 mm
3 mm
25 m
20 m

SW indicates surface wave.

orifice is accelerated as the plate moves. When the inertial


force is larger than the surface tension force that holds it to
the orifice, a drop is ejected from the orifice. The size of the
drop and its initial speed depend on the fluid, the size of the
orifice, and the energy supplied to the transducer. A unique
feature of the device is that the fluid is not pressurized, and
the vibrating plate contains the orifice as the actuation
source. Thus, the device can be manufactured by surface
micromachining and is amenable for implementation in the
form of two-dimensional arrays. Indeed, a silicon micromachined version of the device is presented in Percin et al.4,5
We designed the transducer to have maximum displacement at the center of the plate at the resonant frequency.
Similar ejector designs can be found in Strom,6 Maehara
et al.,710 Ueha et al.,11 Tetsuo,12 and Ivri.13 However, the
ejector designs in literature consist of only one ejector element and some of them use multiple orifices per element and
pressurized fluid. The highest operation frequency reported

1121

in literature is 100 kHz. Some of the devices reported operate


as a nebulizer or atomizer rather than droplet generator, and
they generate nonuniform drop sizes. The ejector design
shown in Fig. 1 utilizes only one ejection source orifice per
element, i.e., each orifice has its own actuation source. A tiny
vibrating carrier plate sets up capillary waves at the liquid
air interface near the orifice, and this tiny plate, that has a
thickness much smaller than the ejected drop size, causes the
uniform breakup of the drops from the surface. The complexity of the structure and the fact that the piezoelectric used is
an annular disk rather than a full disk necessitate the use of a
more complicated analysis to determine the resonant frequencies of the structure, the input impedance of the transducer, and the normal displacement of the surface as presented in Percin et al.14 By using the model developed in
Percin et al.,14 the electrical input impedance, average displacement, and mode shape simulations for a device dimensions and materials given in Table I are presented in Figs. 2
and 3. As shown in Fig. 2, the measured real and imaginary
parts of the electrical input impedance resemble those obtained by the model. The device has multiple resonances.
The model is also capable of predicting electrical input impedance and average displacement of the fluid loaded device,
and the corresponding simulations are also presented in Figs.
2 and 3. As presented in these plots, the resonant frequencies
are decreased by fluid loading on one side of the plate, and

FIG. 2. Device electrical input impedance measurements and simulations.

Downloaded 08 Nov 2004 to 171.64.86.217. Redistribution subject to AIP license or copyright, see http://rsi.aip.org/rsi/copyright.jsp

1122

Rev. Sci. Instrum., Vol. 74, No. 2, February 2003

G. Percin and B. T. Khuri-Yakub

FIG. 3. Device average displacement and mode shape simulations.

the second mode has higher quality factor than the third
mode. By looking at the mode shapes shown in Fig. 3, one
can conclude that the first mode has higher coupling to the
surrounding medium than the other modes, since the volume
velocity for the first mode is higher than the other modes.
The average displacement of the plate decreased by fluid
loading on one side of the plate. The third mode has higher
center displacement value compared the other modes, as a
result, it is more suitable for the droplet ejector operation
than for the transducer operation.
III. DEVICE OPERATION

The vibrating plate sets up capillary waves at the liquid


air interface, and raises the pressure in the liquid above at-

mospheric as high as 15 atm during part of a cycle, and if


this pressure rise stays above atmospheric pressure long
enough during a cycle, and this is high enough to overcome
inertia and surface tension restoring forces, drops are ejected
through the orifice. If the plate displacement amplitude is too
small, the meniscus in the orifice simply oscillates up and
down. If the frequency is too high, the pressure in the fluid
does not remain above atmospheric long enough to eject a
drop. An estimate of the correct balance can be found by
considering the Kelvin equation15,16
c

2
f 2

1/3

which gives the wavelength c of linear capillary waves


driven at frequency f on a plane interface. The parameters
and are surface tension and density of the liquid. If c is
much smaller than the diameter of the orifice, the vibrating
plate will simply set up short ripples on the liquid surface.
This suggests that a dimensionless surface tension parameter,
S

FIG. 4. Droplet ejection simulation: Top, simulation; bottom, actual ejection.

2
,
r 3h f 2

where r h is the radius of the orifice, should be on the order of


1 or greater. This is required for the breakup of the drop from
the liquidair interface, otherwise, the wavelength of the
ripples is too small for the drop to pinch off from the liquid
surface.
A computational model which simulates droplet ejection
has been developed17 using a boundary integral method similar to one used previously.18 21 The previous work did not
have solid boundaries, therefore, modifications were necessary: Singular potential flow dipole solutions were distributed along the liquidair interface in an axially symmetric
configuration, as in the previous work, and singular source
solutions were distributed on the solid plate surface. Enforcing boundary conditions on these surfaces, pressuresurface
tension balance on the liquid interface and specifying the
velocity on the solid plate gives integral equations to determine the dipole density and the source density functions. The
use of both dipoles and sources in this manner results in
coupled Fredholm equations of the second kind which can be
solved simultaneously by an iterative process.

Downloaded 08 Nov 2004 to 171.64.86.217. Redistribution subject to AIP license or copyright, see http://rsi.aip.org/rsi/copyright.jsp

Rev. Sci. Instrum., Vol. 74, No. 2, February 2003

Piezoelectric droplet ejector

1123

FIG. 5. Simulation of water ejection


at 16.4 kHz through 60 m diameter
orifice for different modes and displacements. The drop formation is
shown at every 6.1 s intervals.

These equations were made dimensionless using the radius of the orifice as the characteristic length and the period
of plate oscillation as the characteristic time. The only parameter which remains in the equations is the surface tension
parameter, S, introduced previously. This provides a scaling
law for drop ejection. All other things being equal, such as
amplitude and plate mode shape, this says that droplet size
and shape are only a function of this single dimensionless
parameter. For instance, if f is made larger, then either the
orifice radius should be smaller or made larger. It should
be emphasized that viscous effects have been neglected in
this analysis given that the ejection velocity scales with r h f .
A Reynolds number,
Re

r 2h f

which describes the ratio of inertial forces to viscous forces,


should be sufficiently large for the analysis to be valid,
where is kinematic viscosity of the liquid.
In the computation, drop ejection is initiated by pushing
the plate downward from an elevated stationary position to a
depressed position where it is stopped. That is, it moves

through a half cycle with a motion which produces a flowrate


through the orifice of qq max sin(2t) with 0t0.5 and
q max3.1 l/s for the simulation shown in Fig. 4. The simulated droplet ejection closely resembles the actual water
droplet ejection picture with a similar elongated tail. This is
seen in Fig. 4 where the simulated droplet ejection is shown,
with actual dimensions, at three times during a cycle. The
orifice diameter is 60 m, the ejection frequency is 16.4 kHz
period 60.97 s, and the ejected fluid is water. The diameter of the drop is 82% of the orifice diameter and the final
velocity is about 1.64 m/s based on the distance between the
heights of the last two frames, this value is a reasonable
approximation to 1.54 m/s that was measured in the experiments, at this amplitude and frequency. In the experiment,
the energy dissipated per cycle was 0.55 J, the peak energy
stored was 3.5 J, and the kinetic energy of the drop was
76.8 pJ. In the computation, the drop pinch off time was
41.95 s, Re14.70, and the surface tension parameter S
20. We have used 1.004106 m2 /s, v 1.54 m/s,
72.7103 N/m, and 1 g/cm3 for calculating these
parameters.
Figure 5 shows the ejection simulation of water droplets

Downloaded 08 Nov 2004 to 171.64.86.217. Redistribution subject to AIP license or copyright, see http://rsi.aip.org/rsi/copyright.jsp

1124

Rev. Sci. Instrum., Vol. 74, No. 2, February 2003

G. Percin and B. T. Khuri-Yakub

FIG. 6. Simulation of water ejection at different frequencies through 60 m diameter orifice. The drop formation is shown at every 0.1/f intervals.

by using different modes of operation of the device and excitation displacement amplitudes. The excitation frequency is
16.4 kHz, the orifice diameter is 60 m, and the drop formation is shown at every 0.1/f , 6.1 s, intervals. The required
excitation displacement amplitude for the first mode of operation of the device is smaller than the one for the second
mode. For the first mode and an excitation displacement amplitude of 0.3 m, the plate displacement amplitude is so
small that the meniscus in the orifice simply oscillates up and

down. Figure 6 shows the ejection simulation of water droplets for different excitation frequencies. The orifice diameter
is 60 m, the second mode of operation of the device is
used, the excitation displacement amplitude is 3 m, and the
drop formation is shown at every 0.1/f intervals, where f is
the excitation frequency. As seen in Fig. 6, the water droplet
cannot pinch off from the liquidair interface for the excitation frequencies less than 14.4 kHz, and as the excitation
frequency increases, the ejection velocity and the pinch off

Downloaded 08 Nov 2004 to 171.64.86.217. Redistribution subject to AIP license or copyright, see http://rsi.aip.org/rsi/copyright.jsp

Rev. Sci. Instrum., Vol. 74, No. 2, February 2003

Piezoelectric droplet ejector

1125

FIG. 9. An example of ink-jet printing by using the device.

FIG. 7. Water ejection at 16.4 kHz through 60 m diameter orifice. Note


that the camera is tilted.

FIG. 10. Small solid particle ejection: Linewidth is 260 m, and individual
particles are 918 m in diameter.

FIG. 11. Small solid particle ejection through 150 m orifice: Left-hand
side, at 2.9 kHz; right-hand side, at 5.5 kHz.

TABLE II. Properties of Shipley Microposit photoresists used in the experiments.

FIG. 8. Water ejection at different frequencies through 60 m diameter


orifice. Note that the camera is tilted.

Photoresist

Dynamic viscositya
mPa s

Specific gravity

Volatile organic
compound %w

S1805
S1813
S1400-21
S1400-27

5
20
8
18

1.01
1.04
1.02
1.02

83
73
79
73

Water 20 C has a dynamic viscosity of 1.002 mPa s.

Downloaded 08 Nov 2004 to 171.64.86.217. Redistribution subject to AIP license or copyright, see http://rsi.aip.org/rsi/copyright.jsp

1126

Rev. Sci. Instrum., Vol. 74, No. 2, February 2003

G. Percin and B. T. Khuri-Yakub

FIG. 12. S1813 photoresist ejection


simulations and picture at 7.15 kHz
through 110 m diameter orifice.

time also increase. For example, the pinch off time is around
0.8/f for the excitation frequency of 22.4 kHz, 0.7/f for the
excitation frequency of 18.4 kHz, and 0.6/f for the excitation
frequency of 14.4 kHz. Note that, according to Eq. 2, the
dimensionless surface tension parameter, S, decreases as the
excitation frequency increases. For the excitation frequencies
less than 14.4 kHz, the excitation frequency is too low, and
the formed droplet in the orifice simply collapses onto itself.
FIG. 13. Ejected photoresist droplets through 110 m diameter orifice by
using the device. Note that the camera is tilted.

IV. EXPERIMENTS

FIG. 14. Photoresist covered 3 in. wafers.

Scale model devices were fabricated using the optimum


configuration. The reservoir was a brass cylinder of height 8
mm. A 25 m thin shim was bonded to the 25 m thick
piezoelectric annular disk. The inner and outer diameters of
the annular disk were 2 mm and 7 mm, respectively. The
orifice diameter ranged from 50 to 200 m and was made
using either a drill in a small lathe, or by chemical etching.
Figures 7 and 8 show water ejection at different time
frames and different resonant modes, respectively. As seen in
Fig. 8, the water droplet cannot pinch off from the liquidair
interface for the first two resonant modes at 1.35 kHz and
3.13 kHz. This is because the radius of the orifice is too
small for the given excitation frequency and liquid to provide
an optimal dimensionless surface tension parameter for ejection. Furthermore, the reader should note that the ejection
pictures presented in this article were taken using stroboscopic imaging rather than high speed photography. Therefore, each drop shown in the ejection pictures is actually
hundreds of drops overlapped on each other. Given that the

FIG. 15. The scan of individual photoresist drops. The vertical axis is in k,
and the horizontal axis is in m.

FIG. 16. Drop-on-demand direct write with photoresist: The lines are 350
m wide.

Downloaded 08 Nov 2004 to 171.64.86.217. Redistribution subject to AIP license or copyright, see http://rsi.aip.org/rsi/copyright.jsp

Rev. Sci. Instrum., Vol. 74, No. 2, February 2003

drop images are very sharp without any blur, these data show
that each drop has same diameter and leaves the orifice with
same velocity. Thus, the ejection is not only monodispersed
but also very controllable. Figure 9 shows the drop-ondemand ejection of oil-based ink on a white paper, where
individual letters are written by overlapping the single drops
ejected per tone burst applied to the ejector. Figure 10 shows
the drop-on-demand direct write of fine solid particles on a
sticky tape. Talcum powder Mg3 Si4 O10(OH) 2 with a particle size of 918 m was ejected through a 60 m diameter
orifice. The deposited solid particle line width is 260 m.
The deposited linewidth can be reduced by decreasing the
ejector to the sample spacing. Figure 11 also shows the ejection of talcum powder at different frequencies through 150
m diameter orifice. As shown in Fig. 11, a cluster of talcum
powder with a particle size of 918 m is ejected through a
150 m diameter orifice and forms a jet of agglomerated
particles. The width of this jet of agglomerated particles increases with increasing frequency of oscillations of the vibrating plate.
Shipley Microposit S1400-21, S1400-27, S1805, and
S1813 photoresists were ejected using the ejector under the
conditions described, with a 200 V peak-to-peak voltage.
Table II shows the physical properties of the photoresists
ejected. The ejected photoresist drop size is 85% of the orifice size as shown in Figs. 12 and 13, where the orifice size
is 110 m. Figure 12, where the simulated droplet ejections
are shown with actual dimensions at every 0.1/f intervals,
also shows that the simulated droplet ejection closely resembles the actual photoresist droplet ejection picture with a
similar elongated tail. The orifice diameter is 110 m, the
ejection frequency is 7.15 kHz period 139.86 s, and the
ejected fluid is Shipley S1813 photoresist. The final velocity
is about 1.79 m/s for the first mode and 0.98 m/s for the
second mode. In the computation, the drop pinch off time
was 123.0 s and the surface tension parameter S11.3. We
have used 50.0103 N/m and 1.04 g/cm3 for calculating these parameters. Figure 14 shows square coating of
3 in. silicon wafers with zero waste. Figure 15 shows profilometer alpha-step scans of the deposited photoresist
spots on the wafer. Figure 16 demonstrates the ability to
deposit lines of photoresist that are 350 m wide. Narrower
lines can be deposited with smaller drops for microelectromechanical systems where critical dimensions are in the order of several microns. Coating in a clean environment will
allow the lithography of circuits for microelectronic applications.

Piezoelectric droplet ejector

1127

V. DISCUSSION

In summary, we have developed a fluid and solid particle


ejector, and a photoresist coating technique. The ejector was
also demonstrated with water, ink, and talcum powder. The
ejector design, based on that of a flextensional transducer,
was optimized using a finite-element analysis. Simulation
software for computing drop formation from a vibrating orifice was developed by using a boundary integral method.17
This ejector has been silicon micromachined into twodimensional arrays.4,5

ACKNOWLEDGMENTS

This research was conducted during the Ph.D. study of


one of the authors G.P. at Stanford University. This research was supported by the Defense Advanced Research
Projects Agency of the Department of Defense and was
monitored by the Air Force Office of Scientific Research
under Grant No. F49620-95-1-0525.

P. Calvert, Chem. Mater. 13, 3299 2001.


S. Holdcroft, Adv. Mater. 13, 1753 2001.
3
C. C. Hwang, Rev. Sci. Instrum. 51, 581 1980.
4
G. Percin and B. T. Khuri-Yakub, IEEE Trans. Ultrason. Ferroelectr. Freq.
Control 49, 756 2002.
5
G. Percin and B. T. Khuri-Yakub, Rev. Sci. Instrum. 73, 2193 2002.
6
L. Strom, Rev. Sci. Instrum. 40, 778 1969.
7
N. Maehara, S. Ueha, and E. Mori, Jpn. J. Appl. Phys., Suppl. 26, 215
1987.
8
N. Maehara, K. Hashido, and H. Hirata, European Patent No. 0077636 27
April 1983.
9
N. Maehara, S. Ueha, and E. Mori, Rev. Sci. Instrum. 57, 2870 1986.
10
N. Maehara, U.S. Patent No. 4,605,167 12 August 1986.
11
S. Ueha, N. Maehara, and E. Mori, J. Acoust. Soc. Jpn. 6, 21 1985.
12
I. Tetsuo, Japan Patent No. JP59073963 26 April 1984.
13
Y. Ivri, International Patent No. WO 93/01404 21 January 1993.
14
G. Percin and B. T. Khuri-Yakub, IEEE Trans. Ultrason. Ferroelectr. Freq.
Control 49, 573 2002.
15
S. C. Tsai, P. Luu, P. Childs, A. Teshome, and C. S. Tsai, Phys. Fluids 9,
2909 1997.
16
G. Brenn and U. Lackermeier, Phys. Fluids 9, 3658 1997.
17
G. Percin, T. S. Lundgren, and B. T. Khuri-Yakub, Appl. Phys. Lett. 73,
2375 1998.
18
S. A. Elrod, B. Hadimioglu, B. T. Khuri-Yakub, E. G. Rawson, E. Richley,
C. F. Quate, N. N. Mansour, and T. S. Lundgren, J. Appl. Phys. 65, 3441
1989.
19
T. S. Lundgren and N. N. Mansour, J. Fluid Mech. 194, 479 1988.
20
T. S. Lundgren and N. N. Mansour, J. Fluid Mech. 224, 177 1991.
21
N. N. Mansour and T. S. Lundgren, Phys. Fluids A 2, 1141 1990.
1
2

Downloaded 08 Nov 2004 to 171.64.86.217. Redistribution subject to AIP license or copyright, see http://rsi.aip.org/rsi/copyright.jsp