You are on page 1of 11

Experimental Thermal and Fluid Science 86 (2017) 149–159

Contents lists available at ScienceDirect

Experimental Thermal and Fluid Science


journal homepage: www.elsevier.com/locate/etfs

Dynamics of monodisperse micrometre-sized metal droplets at low non-


dimensional wavenumbers
Alexander F.R. Sanders ⇑, Mario Nakhle dit el Ghorr, Reza S. Abhari
Lab for Energy Conversion, ETH Zurich, Sonneggstrasse 3, 8092 Zurich, Switzerland

a r t i c l e i n f o a b s t r a c t

Article history: This work studies the break-up process of a jet into droplets at low non-dimensional wavenumbers. The
Received 1 September 2016 work is based on a novel droplet generator suitable for high temperature fluids including molten metals
Received in revised form 15 March 2017 employing a piston resonant structure capable of generating high amplitude pressure excitations in the
Accepted 19 March 2017
fluid medium. The relation between the excitation strength and the resulting jet break-up characteristics
such as break-up length and in particular the temporal stability of the resulting droplet stream is studied.
It is found that a given minimum excitation voltage is required for uniform droplet generation and that
Keywords:
the temporal stability ameliorates further with increasing excitation levels. This analysis is comple-
Droplet generator
Monodisperse droplet stream
mented with in situ measurements of the generated pressure excitation signal in a molten metal. It is
Molten metal droplets shown that the novel droplet generator can produce a continuous monodisperse stream of
Droplets for EUV sources micrometre-sized molten tin droplets at low non-dimensional wavenumber of ’0.12. This work is of par-
Temporal droplet stream stability ticular importance for Laser Plasma produced (LPP) Extreme Ultraviolet (EUV) light sources, which
Jet break-up at low non-dimensional require a highly stable stream of monodisperse micrometre-sized molten tin droplets as well as for 3D
wavenumber metal printing.
Ó 2017 Elsevier Inc. All rights reserved.

1. Introduction excitation. This work, however, focuses on extending the achiev-


able range of uniform droplet formation with a non-modulated
The break-up of a jet into stable uniform droplets with low tem- waveform signal and discusses the requirements thereof.
poral timing variation employing external excitation mechanisms The work is based on the droplet generator shown in Fig. 1,
has been a major focus of research for many decades. Similarly, which is novel in that its actuation system is based on a piston res-
the generation of high temperature molten metal droplets in the onant structure forming a standing wave at resonant frequency f 0 .
micrometre range has attracted high research interest and is a This is shown in Fig. 2, where the piston tip and backing mass
common challenge found in various modern technologies such as attachment form points of maximum displacement, while the
soldering of circuit boards [1], rapid prototyping [2] and in partic- flange and the lower step change form points of minimum dis-
ular regarding EUV light sources [3–5]. Particularly from a fluid placement. The objective of this resonant actuation system is to
science perspective the generation of molten metal droplets for maximise the strength of the generated acoustic pressure waves
EUV sources poses a highly interesting research problem in that in the liquid medium exciting the jet in an attempt to achieve
uniform jet break-up is required in a low non-dimensional stable droplet generation even at low non-dimensional wavenum-

wavenumber regime (k < 0:3). This regime is generally associated bers. The droplet generator is explained in detail in Section 3.1. The
with unstable jet break-up due to the small growth rate of the exci- design of the droplet dispenser was filed for priority application
tation signal in this range [3,6]. Methods such as droplet-merging (EP16169161) and will soon be submitted for patent application.
[7] circumvent this problem by exciting the jet with a modulated Most quantitative assessments of the droplet stream stability
waveform resulting in droplet merging and in a droplet stream only refer to the non-dimensional wavenumber and omit the influ-
exhibiting a lower non-dimensional wavenumber than the initial ence of the excitation strength [9–11]. Though studies exist on how
the break-up is influenced by the excitation strength of the exter-
nal actuation mechanism, often only the influence on the break-up
⇑ Corresponding author. length is analysed [12,13] or only qualitative statements on the
E-mail addresses: sanders@lec.mavt.ethz.ch (A.F.R. Sanders), namario@student. influence on the break-up stability are made [14,13]. Only one
ethz.ch (M. Nakhle dit el Ghorr), rabhari@lec.mavt.ethz.ch (R.S. Abhari).

http://dx.doi.org/10.1016/j.expthermflusci.2017.03.024
0894-1777/Ó 2017 Elsevier Inc. All rights reserved.
150 A.F.R. Sanders et al. / Experimental Thermal and Fluid Science 86 (2017) 149–159

Nomenclature

Symbols c speed of sound in liquid (m s1)


k

non-dimensional wavenumber = 2pa=k (dimensionless) z0 specific acoustic impedance = qc ¼ p0 =u (kg m2 s1)
a unperturbed jet radius (m) V voltage applied to actuator (kg m2 A1 s3)
k wavelength of excitation (m) V0 critical voltage at f 0 at which transition from low to
v jet velocity (m s1) high temporal stability occurs (kg m2 A1 s3)
f excitation frequency (s1) tb break-up time (s)
f0 resonant frequency (s1) Dh initial perturbation (m)
c growth rate of disturbance (s1) c maximum possible growth rate of disturbance (s1)
q density of liquid (kg m3) lb break-up length (m)
rs surface tension (N m1) k1 constant (dimensionless)
l dynamic viscosity (kg m1 s1) k2 a/k1 (dimensionless)
In nth element of 1st Bessel function (dimensionless) R2 coefficient of determination, indicating amount of vari-
Kn nth element of 2nd Bessel function (dimensionless) ance in data explained through regression model
g gravitational acceleration 2
pffiffiffiffiffiffiffi (9.81 m s ) Dt time period between subsequent droplets
Fr Froude number = v/ g lb (dimensionless) r standard deviation
p0 amplitude of induced pressure signal (Pa, kg m1 s2) l mean value Timing jitter = rðDt 1;...;n Þ/ lðDt 1;...;n Þ (dimen-
u displacement velocity of piston tip (m s1) sionless)

account is found describing the requirement of a minimum excita- holistic assessment of how the break-up characteristics are related
tion amplitude for uniform droplet break-up and showing quanti- to the acoustic excitation signal. Conclusions are drawn about the
tatively how this minimum excitation amplitude changes with the requirements for uniform droplet generation at low non-
non-dimensional wavenumber [15]. However, to the author’s dimensional wavenumbers and how elevated excitation ampli-
knowledge the current literature does not provide any quantitative tudes influence various jet break-up phenomena.
measurements of how the temporal stability of the droplet stream
is influenced by the excitation amplitude. 2. Background
This paper analyses how the break-up characteristics and in
particular the resulting temporal stability are affected by the exci- 2.1. Theory of droplet break-up and motivation
tation voltage both using high resolution images of the break-up
region as well as precise measurements of the temporal stability Break-up of a jet into droplets is generally based on the growth
with a laser-diode system. This analysis is complemented with of disturbances due to considerations of surface tension and sur-
time-resolved measurements of the acoustic excitation signal gen- face energy minimisation [16,17]. Break-up occurs when the dis-
erated in the liquid by the resonant excitation system giving a turbance has grown to equal the jet radius. Fig. 3 shows the very
similar dispersion curves (growth rate c versus the non-

dimensional wavenumber k ) both for including and neglecting
viscous effects of tin [18,19] and shows clearly the strong decrease
 
in growth rate of the low k regime of k < 0:3 analysed in this
Backing mass 
paper compared to the maximum at k  0:7. The non-
attachment / actuator dimensional wavenumber is defined as (a being the jet radius, k
and f the wavelength and the frequency of the excitation respec-
Cover tively and v the jet velocity):

Pressurised  2pa 2paf


k ¼ ¼ ð1Þ
gas
k v

Piston

Molten Flange
Longitudinal Curve indicating
displacement longitudinal
metal
displacement
Heater Nodal point
Piston

Reservoir
Piston
Nodal point / tip
Step change
Backing mass
attachment /
actuator
Nozzle
Fig. 2. Resonant actuation system showing the longitudinal displacement at
resonant frequency f 0 . The piezoelectric actuator is placed on outside of high
Fig. 1. Novel droplet dispenser design, EP16169161 [8]. pressure and temperature reservoir.
A.F.R. Sanders et al. / Experimental Thermal and Fluid Science 86 (2017) 149–159 151

Dispersion curve tin 2.2. Importance of stable droplet stream at low  values for EUV
sources
1 non-viscous
viscous EUV radiation centred at 13.5 nm is the main candidate for the
next generation semiconductor lithography process as it offers
0.8 higher resolution capabilities and smaller semiconductor pattern
Growth Rate (norm.)

sizes than currently achievable with 193 nm immersion lithogra-


phy. It also offers high resolution microscopy applications
0.6 [24,25]. Fig. 4 shows an EUV light source employed to generate
EUV radiation. It consists of a vacuum chamber equipped with a
droplet dispenser and EUV radiation is generated by a laser-
0.4 produced plasma (LPP) that is obtained by irradiating
micrometre-sized tin droplets with a high power (1 kW) laser
[26]. An EUV collector is then used to focus the EUV radiation onto
0.2 the intermediate focus, where various devices can be installed to
utilise the EUV radiation. The instability of this EUV radiation is
directly linked to the instability of the droplet stream and a high
0
0 0.2 0.4 0.6 0.8 1 pulse-to-pulse stability of the EUV radiation is of utmost impor-
tance particularly for the above stated applications [27,28]. The
Non-dimensional Wavenumber k*
industry requirement is often stated with an integrated EUV
Fig. 3. Viscous and non-viscous dispersion relation for tin. energy stability of 3r, 0.2% over 50 EUV pulses [4]. In particular
for reasons of droplet debris reduction which increases the opera-
tion lifetime of the EUV collector of the source the usage of tin dro-
plets with very small diameters (10 lm) is envisaged [4]. Further,
The reason for the occurrence of droplet jitter is related to high droplet velocities are desirable to achieve a sufficiently large
small variations in the curvature of the ligament and swelling droplet separation distance in order to avoid interaction of the
drop just before break-up occurs due to non-linear effects becom- expanding plasma with the subsequent droplet. The small droplet
ing dominant as well as due to the presence of noise sources diameter together with the high droplet velocities imply very small
[11,13]. This results in variations in the momentum exchange wavenumbers, at which stable droplet formation is required and
between droplet and ligament [16] and thus to ‘unpredictable serves as a motivation for the novel droplet dispenser and the stud-
speed fluctuations connected with the break-up process’ [20]. ies in this paper.
These timely variations in droplet spacing have been shown to
be smallest at the maximum of the dispersion relation
[11,16,20], where fast growth of the principal disturbance close 2.3. Methods for droplet generation

to the optimal k value can suppress any interfering disturbances
[11,21]. Several authors have shown experimentally that there While several designs specifically suited to the generation of

exists a limited range around the optimum k value of roughly molten micrometre-sized metal droplets (e.g. pneumatic or elec-
 tromagnetic pinch force mechanisms) are presented in the avail-
k = 0.3–0.97, over which stable droplet formation occurs
[10,11,22]. This limited operating range can be explained by the able literature, very few of them allow the generation of a highly
relative exponential increase in the detrimental influence of the temporally stable droplet stream [29,30] and in particular not at
noise source on the break-up process when moving away from low non-dimensional wavenumbers or frequencies in the kHz
 range [2,31,32].
the optimal k value.
However, many of the above mentioned studies do not take Various designs based on a piston structure are presented in the
the effect of the excitation amplitude into account. High initial literature, where the piston provides a spatial separation and tem-
amplitudes of the principal excitation should facilitate achieving perature barrier between the actuator and the hot molten metal.
stable droplet break-up even if the principal disturbance is at Piston based excitation system are designed both for drop-on-

low k values. Yoshida, for example, shows that the excitation demand (DoD) droplet break-up [33,34] as well as for continuous
amplitude needs to be increased for uniform droplet generation droplet break-up [35]. While the DoD systems exhibit low droplet

at k values larger or smaller than the optimal value [15]. This speeds and frequencies with the achievable temporal stability of
is also found by Rollinger numerically [3]. Amplitude modulated
excitations disturbing the jet have also been shown to improve
the achievable temporal stability of the resulting droplet stream
and to extend in particular the lower range of attainable Droplet
wavenumbers for uniform droplet generation [9,11]. This paper, dispenser
however, focuses solely on the improvement of the temporal
droplet stream stability via the maximisation of the acoustic Vacuum
pressure strength of the excitation signal. Effectively, the goal chamber Intermediate
is to achieve a high signal-to-noise ratio (SNR) regarding the focus
EUV
respective initial amplitudes such that even if the noise sources
have a more favourable growth rate the principal disturbance Drive laser
& beamline
remains being the dominant influence on the jet. This effect
has not been studied in detail in the literature yet. The effect
of gravity on the break-up can be neglected for the current study, EUV
as the Froude number representing the ratio of inertial to gravi- Collector Plasma site
tational forces, evaluated using the break-up length of the jet is
Fr J 2400 [23]. Fig. 4. EUV light source.
152 A.F.R. Sanders et al. / Experimental Thermal and Fluid Science 86 (2017) 149–159

the droplet stream remaining unclear [33,34], the continuous jet stream ca. 5 cm further downstream can be captured. Measure-
break-up systems based on a piston structure and a piezoelectric ments of the temporal stability of the jet are conducted in a sepa-
actuator are shown to generate a uniform and stable droplet rate vacuum chamber facility by measuring the time of flight of
stream [4,35,36]. each individual droplet as it passes between two laser sheets direc-
ted onto an array of diodes downstream of the nozzle (see Fig. 6) at
3. Experimental setup and method a similar vertical position as the downstream images are taken. As
a droplet passes through a laser sheet it creates a shadow on the
3.1. Droplet dispenser design diode array, leading to a dip in the diode signal. The temporal
instability, referred to as temporal jitter (or mean jitter), of the dro-
The resonant structure of the presented actuation system is plet stream is then computed in real time as the coefficient of vari-
shown in Fig. 2 and consists of a piston, a piezoelectric actuator ation (standard deviation rðtÞ divided by the mean lðtÞ) in the
and a backing mass - the latter two are placed outside the high time period Dt between subsequent minima of the diode signal
temperature and pressure reservoir where it is shielded from the for a period of 500 ms. This allows a highly accurate computation
heat load. The piezoelectric actuator and the backing mass are of the temporal instability, referred to as temporal jitter. The sys-
assembled in the same way as outlined in [37,38]. The piston is tem used in this study consists of two laser sheets that are sepa-
actuated on the one side by the piezoelectric actuator, while its rated by a small distance along the axis of flight of the droplet
tip is immersed in the liquid in the fuel reservoir where it trans- train. The system also computes the smallest standard deviation
mits acoustic pressure waves into the molten metal reservoir and in the time period between any two subsequent minima of the
onto the jet. The length scales of the backing mass, piezoelectric diode signal per 500 ms, referred to as minimum jitter. The laser
actuator and piston are chosen such that at the desired resonant diode measurement system does not take satellite droplets into
frequency the actuating system forms a multiple of k=4 such that account for the computation of the temporal stability.
a standing wave is formed. This is shown in Fig. 2, where the flange
and the step change form a nodal point of maximum stress but 3.3. Pressure measurements
minimum displacement. At the same time both the end of the
backing mass and the piston tip represent points of maximum dis- The strength and time signature of the generated pressure
placement, with the piston tip having a higher displacement than waves was measured using a high temperature fast response pres-
the backing mass due to the change in cross-sectional area of the sure probe developed by Rollinger et al. [39]. The probe itself was
piston at the lower nodal point [37]. On the basis of this resonant directly attached to the dispenser just below the vibrating piston
structure, it is aimed to maximise the strength of the generated where normally the nozzle orifice is placed. The details of the mea-
acoustic pressure waves with the presented design. surement setup are described elsewhere [4,39]. Due to tempera-
ture limitations and reliability concerns the measurements were
3.2. Droplet measurements conducted with indium as a target liquid, which has a specific
acoustic impedance (z0 ) only 5.6% smaller than that of tin. The
The droplet dispenser is placed in a vacuum chamber shown in specific acoustic impedance is defined as:
Fig. 5, which allows the elimination of effects such as drag forces or
p0
wind-induced phenomena when studying the jet break-up charac- z0 ¼ q  c ¼ ð2Þ
teristics. The droplet stream is captured with high resolution imag- u
ing using back light illumination and a CCD camera. By varying the This is important as the acoustic impedance of the material the
vertical distance between the droplet generator and the imaging piston resonant structure is immersed into at its tip influences the
system, images both of the break-up region and of the droplet resonant characteristics of the actuation system in terms of damp-
ing and frequency response. Further, it was tried to keep the tem-

vacuum chamber
amplifier signal generator
Droplet
piezo Generator

press.
controller

t1 t2 t3
LED control system
camera

Ar
tank
Timing Real
Jitter time
=X% analysis
LED camera

Fig. 6. Laser Diode system measuring the temporal variation in the time period
between subsequent droplets Dt. The temporal jitter is calculated as the coefficient
Fig. 5. Droplet test facility showing the shadowgraphy system for imaging of the of variation (standard deviation rðtÞ divided by the mean lðtÞ of Dt 1 ; Dt1 ; Dtn ) for a
break-up region and the droplet stream further downstream (5 cm downstream of time period of 500 ms. The figure shows the manipulated and inverted photodiode
the nozzle). signal created when droplets pass through the laser diode system.
A.F.R. Sanders et al. / Experimental Thermal and Fluid Science 86 (2017) 149–159 153

perature difference between the droplet dispenser filled with 100


indium and tin as small as permissible by the temperature limita-
tion of the pressure sensor. Hence, the results obtained with the
pressure measurements conducted in indium can be extrapolated
to tin. All experiments with droplets are conducted with high pur-
ity molten tin.

Jitter (%)
4. Results and discussion 10

4.1. High resolution break-up imaging

Fig. 7 depicts the generated droplet stream ca. 5 cm down-


stream of the nozzle and shows that stable droplet generation at Mean Jitter

k values as small as ’0.12 is feasible and thus below the limit of Min. Jitter
0.3 found by [6] for non-modulated excitation signals. The follow- 1
ing sections try to give an answer for the requirement of the stable 0 1 2 3 4
droplet generation at these low non-dimensional wavenumbers.
Particularly, the importance of the excitation amplitude is exam- Excitation Voltage (V/V0)
ined, combined with an analysis of the frequency content of the
Fig. 8. Temporal jitter level versus excitation voltage at 0.95 f 0 .
pressure excitation signals.

4.2. Temporal stability versus excitation amplitude Fig. 6 and described in the experimental setup. Measurements of
the minimum and mean temporal jitter were taken at each voltage
Quantitative measurements of the temporal stability versus the level for an average time period of 15 s with the shortest time per-
excitation voltage are conducted as shown in Fig. 8 with molten tin iod being 3 s for the highest excitation voltage level. The droplet
and a ’24 lm diameter nozzle. Measurements are taken with two dispenser was operated at 0.95 f 0 and data points are collected
laser-diode systems simultaneously and in real time, as shown in going up and down in voltage underlining the repeatability of
the measurement.
It can be observed that the plot exhibits a sudden jump in tem-
poral stability from a low to high uniformity regime of the droplet
stream around 2.3V 0 (V 0 is the approximate voltage level at reso-
nant frequency f 0 , at which the jump from low to high temporal
stability occurs). Furthermore, it can be seen that the temporal sta-
bility ameliorates on both sides of the jump with an increase in the
excitation voltage. The smallest measured temporal instability cor-
responds to a jitter value of 1.7%. This result shows the existence
of a critical voltage below which the temporal stability of the dro-
plet stream is poor. Further, it highlights the beneficial effect of a
strong excitation signal for further improving the temporal stabil-
ity. Interestingly, this improvement can be observed both for the
low and high stability regime. Fig. 9 shows another quantitative
temporal stability series conducted at the resonant frequency f 0 .
It can be seen that the jump from low to high temporal stability
regime occurs at an approximately 2.3 times smaller voltage level

100
Mean Jitter
Min. Jitter
Jitter (%)

10

(a) (b) 1
0.5 1 1.5 2 2.5
Fig. 7. Molten tin droplets generated at resonance frequency f 0 ; LHS: nozzle
diameter ’25 lm, low pressure level, droplet diameter ’57 lm, droplet spacing Excitation Voltage (V/V0)
’256 lm; RHS: nozzle diameter ’13 lm, medium pressure level, droplet diameter
’42 lm, droplet spacing ’296 lm and k ’ 0:12.

Fig. 9. Temporal jitter level versus excitation voltage at resonance frequency f 0 .
154 A.F.R. Sanders et al. / Experimental Thermal and Fluid Science 86 (2017) 149–159

than at 0.95 f 0 . The lower voltage level is expected as for the same High resolution shadowgraphy images of the break-up process
pressure amplitude a smaller excitation voltage is required at res- are taken to study the effect of the excitation amplitude on the
onant frequency and concurrently the regime change occurs at a shape and break-up process of the jet. This is shown in Fig. 11, in
lower excitation voltage. which the excitation voltage of the droplet generator operating
The difference between the minimum and mean computed jit- at resonant frequency f 0 is increased successively from left to right.
ter is an inverse indicator for the stability and consistency of the While at the lowest excitation voltage the jet break-up is highly
droplet formation process. A strong discrepancy can be observed unstable and irregular, it becomes more and more uniform with
for values close to the regime change from low to high temporal increasing excitation voltage. Besides the break-up length succes-
stability indicating the instability of the operating point associated sively decreasing towards higher excitation voltages as expected,
with this regime change. The difference between the two laser one can further observe that the formation of satellite droplets
sheets separated by a small vertical distance is likely to come from generally gets more suppressed towards higher excitation levels,
sensitive changes in the droplet structure with the vertical distance which is in good agreement with the droplet stream images further
at this operating point. downstream (see Fig. 10).
Fig. 10 shows images of the droplet stream taken downstream Additionally, one can notice the unusual shape of the jet ema-
of the nozzle exit for both scenarios below and above the critical nating from the nozzle prior to the break-up of the jet in Fig. 11.
voltage at resonant frequency f 0 , taken during another test at the This does not resemble classical Rayleigh break-up, which would
droplet test facility. It can be clearly seen that below the critical exhibit a continuous increase of the swell and a concurrent
excitation V 0 the droplet stream exhibits satellite droplets and decrease of the neck. Another peculiar occurrence is the non-axis
the droplet spacing is not always uniform, while above the critical symmetric shape of the jet with respect to the longitudinal jet axis,
voltage satellite droplets are fully suppressed and the droplet spac- for which the author has not found a similar report in the litera-
ing is highly uniform. Hence, the droplet stream images corrobo- ture. Section 4.6 states some possible hypotheses for this
rate the existence of a critical excitation level for a mono- occurrence.
disperse droplet stream observed in the quantitative stability mea- The requirement of a strong excitation amplitude blends in with
surements in Figs. 8 and 9. Further, according to Fig. 10 this critical the qualitative finding from Donnelly [40] that e.g. for viscous flu-

voltage level is between 0.67V 0 and 1.33V 0 , and thus in good ids or for operation at non-optimal operating points (k – 0:7) the
agreement with the critical excitation range from the quantitative excitation mechanism needs to be improved in order to still be able
temporal stability measurements at resonant frequency f 0 (see to generate stable droplets. Also other authors conclude on the
Fig. 9). The observation of a required minimum excitation ampli- importance of an increased excitation strength for regular droplet

tude for the generation of a uniform droplet stream fits the findings break-up at k values well below or above the optimum of

by Yoshida [15]. k ¼ 0:7 [3,15,41,42]. This effect was further corroborated by
Dressler [14] who validated the dependence of the break-up mode
on the excitation strength. In particular, Dressler showed that for
disturbances at non-optimal parameters the perturbation would

(a) (b) (c)


Fig. 10. Molten tin droplets downstream of a ’24 lm diameter nozzle generated at Fig. 11. Images of jet break-up versus excitation voltage at resonant frequency f 0
resonance frequency f 0 ; droplet diameter ’65 lm, droplet spacing ’465 lm, and k ’ 0:14, droplet diameter ’65 lm, nozzle diameter ’24 lm (increasing



droplet velocity ’10.3 m/s, k ’ 0:14; (a) and (b) at 0.67V 0 (below critical voltage from left to right: 0.17V 0 , 0.33V 0 , 0.67V 0 , 1.33V 0 , 2.0V 0 , 2.67V 0 , 3.33V 0 )
excitation), (c) at 1.33V 0 (above critical excitation). exhibiting non-axis symmetric shape (with respect to longitudinal axis).
A.F.R. Sanders et al. / Experimental Thermal and Fluid Science 86 (2017) 149–159 155

decay at low excitation values while at higher excitation values


they grow and lead to droplet break-up of the jet. This is similar
to the minimum required excitation voltage threshold for stable
jet break-up observed in the presented experiments.

4.3. Break-up length versus excitation amplitude

In a next step the relation between the break-up length and the
excitation strength is examined, particularly towards whether it
exhibits an unexpected behaviour around the critical excitation
voltage. This analysis is greatly facilitated by the observation that
the strength of the generated pressure waves follows a linear trend
(R2 of 0.97) with respect to the applied excitation voltage as shown
in Fig. 12 (R2 is the coefficient of determination and indicates the
amount of variance in the data that can be explained by the linear
model). A more detailed analysis of the pressure signal induced in
the molten metal (including the analysis of higher orders) is con-
ducted in Section 4.5.
Fig. 13. Break-up length versus excitation voltage at resonant frequency f 0 .
The dependence of the break-up length and break-up time on
the excitation voltage can be found in numerous reports both
experimentally [13,30,43] and numerically [44,45]. Several authors c tb
a ¼ Dhe ð3Þ
confirm the logarithmic relation between the break-up length and
the amplitude of the initial disturbance as predicted by linear the- For a constant jet velocity v ; v t b ¼ lb is satisfied and it follows [49]:
ory [13,43,46]. In addition, Donnelly and Glaberson find an expo- v a
lb ¼ ln ð4Þ
nential relationship between the distance of neighbouring radii c
 Dh
on the jet contour and the break-up time [40] also matching the
predictions of linear theory. It needs to be noted that most of these Considering the difference in break-up length for different exci-
experiments were conducted close to the maximum growth rate tation amplitudes, the following relationship between lb and Dh is
and for small perturbations, for which linear theory is applicable found [48,50]:
opposed to large disturbances [47]. Linear theory is derived on c
lnDh1  lnDh2 ¼  ðlb;1  lb;2 Þ ð5Þ
the assumption that the initial perturbation was infinitesimal v
small and provides no energy for the jet breakup [14], while with
the given setup it is specifically aimed to generate large perturba- Based on the linear correlation between the applied excitation volt-
tions and supply energy for the jet breakup. Nevertheless, a com- age and the generated pressure wave strength (see Fig. 12), it fol-
parison is made between the measured relation of break-up lows that [48]:
length to excitation strength with the presented setup (see Dh ¼ k 1 V ð6Þ
Fig. 13) and the one predicted by linear theory.
Linear theory predicts an exponential growth of the initial per- Hence, the break-up length can be formulated in terms of the exci-
tation voltage (suspending the condition of c to be at its maximum;
turbation Dh, and once it equals the jet radius a, breakup occurs
after time t ¼ t b (with c being the maximum growth rate of the k2 ¼ a=k1 ) [51]:
dispersion curve, [48,49]): v
lb ¼ ½lnðk2 Þ  lnðVÞ ð7Þ
c
Fig. 13 shows the break-up length versus the logarithmic exci-
tation strength for the presented droplet dispenser with a small
1.2
selection of the corresponding jet break-up images shown in
Fig. 11. The break-up length is defined as the point where the main
Pressure (bar, ptp, norm.)

1 ligament detaches from the jet, and not where a satellite droplet
breaks off in front of the main droplet (only the case for low volt-
ages - see Fig. 15b). The standard deviation in the measurement of
0.8 the break-up length at each voltage level is only between 2 and 6%
as shown by the small size of the error bars in Fig. 13. The plot fol-
0.6 lows a straight line perfectly up to an excitation voltage of 2V 0 as
predicted by the linear theory Eq. (5) with a high R2 of 0.99958.
This result is surprising as the high resolution break-up images
0.4 shown in Fig. 11 do not resemble the classical Rayleigh break-up
even for lower excitation voltages. Furthermore, it is surprising
0.2 to notice that even at low excitation amplitudes the relationship
of break-up length versus perturbation strength is not disturbed
by a relatively higher influence of noise. Interestingly, the change
0 from unstable to stable break-up regime between 0.67V 0 and
0 2 4 6 8 1.33V 0 is not reflected in the break-up length versus voltage curve.
Piezo Voltage (V/V0) Beyond 2V 0 a deviation from the exponential relation can be
observed, and the rate of decrease of the break-up length with an
Fig. 12. Pressure level measured in molten indium with respect to excitation increase in the excitation amplitude falls below the prediction by
voltage. linear theory. Fig. 14 shows the growth rate versus the excitation
156 A.F.R. Sanders et al. / Experimental Thermal and Fluid Science 86 (2017) 149–159

1.2

1
Growth rate (norm.)

0.8

0.6

0.4

0.2

0
0 1 2 3 4
Piezo Voltage (V/V0)
(a) Break-up series at 3.33V0 at resonant frequency f0 (above
Fig. 14. Growth rate based on linear model versus excitation voltage. critical excitation amplitude)

voltage, calculated in a similar approach to Lopez and Crane


[48,50] employing Eq. (7) and using the gradient and intercept of
the linear fit. As expected, the growth rate falls significantly for
the measurement points beyond the threshold excitation voltage
of 2V 0 as shown in Fig. 14. In Section 4.5 FFT spectra of the mea-
sured pressure signal are presented, indicating that this deviation
is purely due to the increase in the amplitude level.
A deviation from the exponential relation towards higher exci-
tation levels is also found by Kalaaji et al. and Cline and Anthony
[49,52]. Cline and Anthony estimate that the initial vibration
amplitude corresponding to the lowest excitation amplitude that
deviates from the linear theory is about 10% of the jet radius,
though no further mention about the estimation procedure is made
[52]. Employing the linear model described in Eq. (6) and fitting it
to the data points corresponding to the linear theory, it is found (b) Break-up series at 0.67V0 at resonant frequency f0 (below
that for the presented data the first data point deviating from the
critical excitation amplitude)
linear model occurs at an initial perturbation corresponding to
8.8% of the jet radius, while the last data point still matching the Fig. 15. Break-up series using phase lock technique; nozzle diameter ’24 lm,
linear model corresponds to 6.6% of the jet radius. Assuming that droplet diameter ’65 lm.
Cline and Anthony [52] employed the same break-up model and
considering potentially unknown effects of the fluid such as viscos-
breaks up first before the main droplet forming a satellite droplet,
ity on the occurrence of this non-linear behaviour a difference of
which eventually merges into the main droplet (rear-merging).
1.2% is a very good match.
This fits in with findings of Pimbley and Lee as well as of Vassallo
and Ashgriz [13,53] who observe rear-merging satellite droplets at
4.4. Forward and rear-merging satellite droplets low excitation amplitudes, forward-merging satellite droplets at
high excitation amplitudes and even completely suppressed satel-
Break-up time series are shown in Figs. 15(a) and (b) for both lite formation when further increasing the excitation strength.
below and above the critical voltage level defining the regime However, similarly to the case at higher excitation levels, at low
change between low and high temporal uniformity of the droplet excitation levels in Fig. 15(b) the remainder of the main droplet
stream. These images are taken using the phase-locking technique, (after the satellite droplet had brake off in front of the main dro-
for which there is a slight mismatch in frequencies between the plet) first breaks off the jet before it then breaks further up into
imaging frequency (at around 8 Hz) and the break-up frequency. smaller droplets which eventually catch up again with the main
From these break-up time series (Figs. 15(a) and (b)) it can be droplet (forward-merging). Thus at lower excitation voltage we
observed that jet break-up at lower excitation levels occurs before observe both forward and rear-merging satellite droplets. The
a future main droplet, while at higher excitation levels it occurs downstream images of the droplet stream in Fig. 10 show, how-
behind a future main droplet. This fits in with findings by Kalaaji ever, that at 0.67V 0 the merging of the satellite droplets into the
and Pimbley [13,49]. Kalaaji states that while at low initial excita- main droplet does not occur consistently. In contrast above the
tion amplitudes ‘jet rupture first occurs behind a future main dro- critical voltage complete merging takes place and satellite droplets
plet, it occurs before at higher excitation amplitudes’ [49]. In his are suppressed consistently.
case said rupture occurs at either end of a thin thread connecting
subsequent main droplets. 4.5. Analysis of excitation signal - pressure measurements
Further, a long tail is formed at higher excitation levels, which
after the first jet rupture further breaks down into smaller droplets, Fig. 16 shows the Fast Fourier Transform of the pressure signal
which subsequently catch up with the main droplet and merge measured at different excitation amplitudes at the position of the
with it (forward-merging). At lower excitation voltages, the jet nozzle orifice. While at low excitation voltages (e.g. 0.167V 0 ), the
A.F.R. Sanders et al. / Experimental Thermal and Fluid Science 86 (2017) 149–159 157

1 break-up in Fig. 15 somewhat resembles this classical droplet


merging at lower excitation voltages (see Fig. 15(b)) it does not
0.8 at higher excitation amplitudes (see Fig. 15(a)). For higher excita-
Pressure (bar)

tion voltages, the tail behind the main droplet that breaks up into
0.6 one or two smaller satellite droplets that further down merge with
the main droplet. Despite this strong difference in the observed
0.4 pattern to classical droplet merging at higher elevation amplitudes,
the higher order components may play a role, besides the imposi-
0.2 tion of a high amplitude excitation, in enabling the observed uni-
form droplet generation at non-dimensional wavenumbers as
0 
0 1 2 3 4
low as k ’ 0:12.
Response frequency (f/f 0)
4.6. Modes of perturbation energy transfer
(a) 0.167V0 excitation voltage
1 Higher orders, however, do not explain the non-symmetric jet
shape seen in Fig. 11, for which possible hypotheses are mentioned
0.8 in this section. There are principally two modes via which the exci-
tation can be transferred from the piston into the liquid medium:
Pressure (bar)

0.6 As depicted in Fig. 17(b) through acoustic pressure waves resulting


in the classical Rayleigh break-up of the jet. On the other hand per-
0.4 turbation energy can also be transferred onto the jet through the
active displacement of the liquid as indicated in Fig. 18. The latter
0.2 is generally referred to as a drop-on-demand excitation system.
What is interesting to notice is that such drop-on-demand systems
0 often exhibit non-axisymmetric jet profiles after the nozzle exit
0 1 2 3 4
similar to the ones observed in this paper [32,54]. While for the
Response frequency (f/f0)

(b) 1.0V0 excitation voltage


Growth Rate & Pressure amplitude

0.8 Static
back-
0.6 pressure

0.4

0.2

0
0 1 2 3 4
Frequency (f/f0)

(c) 2.67V0 excitation voltage, overlaid with (a) (b)


growth rate plot Fig. 17. (a) Irregular jet break-up when only static backpressure is applied but no
excitation signal and (b) when a static backpressure is applied together with
Fig. 16. FFT plots for resonance region. acoustic pressure waves generated by the oscillating piston tip resulting in Rayleigh
break-up.

pressure signal measured in liquid indium is highly linear and fol-


lows the voltage excitation signal applied to the piezoelectric actu-
ator perfectly, it starts exhibiting higher orders at higher excitation
voltages.
Overlaying the pressure signal at 2.67V 0 with the growth rate
according to Weber, one can see that the higher frequency contri-
butions are associated with higher growth rates though smaller in
absolute amplitude. Nevertheless, this could imply that they signif-
icantly influence the break-up process. These higher order compo-
nents are also present with similar relative magnitudes in the
pressure signal at 1V 0 as shown in Fig. 16. Thus, it can be con-
cluded that the deviation of the measurement data from the expo-
nential relation between break-up length and excitation voltage
according to linear theory at voltages of 2.67V 0 or higher as shown
in Fig. 13 is only due to the elevated amplitude and not due to the
effect of the higher order components.
In the classical droplet-merging pattern described by Orme and t1 t2 t3
others [20] two or more droplets exhibit a relative velocity compo-
nent to each other and eventually merge. While the observed jet Fig. 18. Active displacement of fluid.
158 A.F.R. Sanders et al. / Experimental Thermal and Fluid Science 86 (2017) 149–159

presented system the energy transfer through acoustic pressure actuation system and for the provision of the pressure sensor as
waves is certainly present (case of Fig. 17(b)), there might be an well as the ALPS group members Duane Hudgins, Markus Brandstt-
additional energy transfer through the active displacement of fluid ter and Marco Weber for helpful discussions and support. The
similar to classical drop-on-demand systems that are often also authors would also like to thank F. Alickaj, R. Rüttimann, and T.
based on a piston design (case of Fig. 18) [40]. The active displace- Künzle for the support with the electronic setup and the manufac-
ment of the fluid could lead to a periodic pushing and pulling back turing of the droplet dispenser.
of the jet after having emanated from the nozzles through viscous
forces. This could hypothetically lead to asymmetries of the nozzle
orifice being translated into a non-axisymmetric jet profile via a References
non-uniform velocity profile over the jet cross-section.
[1] D.J. Hayes, D.B. Wallace, W. Royall Cox, Microjet printing of solder and
polymers for multi-chip modules and chip-scale packages, Proc. SPIE, vol.
3830, 1999, pp. 242–247.
5. Summary and conclusion [2] A. Tropmann, N. Lass, N. Paust, T. Metz, C. Ziegler, R. Zengerle, P. Koltay,
Pneumatic dispensing of nano-to picoliter droplets of liquid metal with the
In this work a novel actuation mechanism based on a resonating starjet method for rapid prototyping of metal microstructures, Microfluid.
Nanofluid. 12 (1–4) (2012) 75–84, http://dx.doi.org/10.1007/s10404-011-
piston structure is presented, capable of generating a stable stream

0850-1.
of droplets as low as k ’ 0:12 and suitable for high melting point [3] B. Rollinger, R. Abhari, Excitation and dynamics of liquid tin micrometer
metals. Measurements based on an accurate laser-diode system droplet generation, Microfluid. Nanofluid. 28 (7) (2016) 074105, http://dx.doi.
org/10.1063/1.4955114.
show that droplet timing jitter of less than 2% can be achieved with [4] B. Rollinger, Droplet Target for Laser-Produced Plasma Light Sources, Ph.D.

the presented actuation system at an approximate k value of thesis, ETH Zurich, 2012.
 
k ’ 0:14 (Note: the measurement of the k value was conducted [5] J.M. Algots, O. Hemberg, A. Bykanov, Liquid metal micro-droplet generator for
laser produced plasma target delivery used in an extreme ultra-violet source,
at a different facility while running the droplet dispenser at the Proc. SPIE, vol. 5751, 2005, pp. 885–891.
same operating conditions and might therefore differ slightly). It [6] G. Brenn, U. Lackermeier, Drop formation from a vibrating orifice generator
has been found that a minimum excitation voltage exists for the driven by modulated electrical signals, Phys. Fluids 9 (12) (1997) 3658–3669,
http://dx.doi.org/10.1063/1.869503.
generation of a droplet stream with a uniform droplet spacing [7] M. Orme, E. Muntz, New technique for producing highly uniform droplet
and the suppression of satellite droplets. Further, it is found that streams over an extended range of disturbance wavenumbers, Rev. Sci.
the temporal stability ameliorates with increasing the excitation Instrum. 58 (2) (1987) 279–284, http://dx.doi.org/10.1063/1.1139322.
[8] A. Sanders, Droplet Dispensing Device, Method for Providing a Monodisperse
voltage both in the low-uniformity and high-uniformity regime.
Stream of Droplets, and Light Source for Providing uv or x-ray Light, Patent
By roughly doubling the excitation voltage beyond the critical volt- Application: EP16169161, May 13 2016.
age V 0 a further improve of ca. 1.5% in temporal jitter can be [9] M. Orme, E. Muntz, The manipulation of capillary stream breakup using
amplitude-modulated disturbances: a pictorial and quantitative
achieved. Thus, it is clearly shown that the strong pressure waves
representation, Phys. Fluids A 2 (7) (1990) 1124–1140, http://dx.doi.org/
generated by the resonant actuation system are beneficial in 10.1063/1.857612.
improving the droplet stream quality. [10] P.E. Frommhold, A. Lippert, F.L. Holsteyns, R. Mettin, High-speed monodisperse
A time-resolved analysis of the pressure signal induced in mol- droplet generation by ultrasonically controlled micro-jet breakup, Exp. Fluids
55 (4) (2014) 1–12, http://dx.doi.org/10.1007/s00348-014-1716-6.
ten indium has been conducted exhibiting higher orders above a [11] M. Rohani, F. Jabbari, D. Dunn-Rankin, Breakup control of a liquid jet by
certain voltage level inherent to the presented excitation mecha- disturbance manipulation, Phys. Fluids 22 (10) (2010) 107103, http://dx.doi.
nism. It thus remains unclear, whether the stable regular jet org/10.1063/1.3494610.
[12] B. Cheong, T. Howes, Effect of initial disturbance amplitude in gravity affected
break-up can be extended towards lower non-dimensional jet break-up, Chem. Eng. Sci. 60 (13) (2005) 3715–3719, http://dx.doi.org/
wavenumbers merely on the basis of a strong excitation signal or 10.1016/j.ces.2005.02.014.
whether the existence of higher orders is also required besides [13] W. Pimbley, H. Lee, Satellite droplet formation in a liquid jet, IBM J. Res. Dev.
21 (1) (1977) 21–30, http://dx.doi.org/10.1147/rd.211.0021.
the presence of a strong excitation strength. Despite the jet [14] J.L. Dressler, Two-Dimensional, High Flow, Precisely Controlled Monodisperse
break-up not resembling classical Rayleigh jet break-up, very good Drop Source, Tech. rep., DTIC Document, 1993.
agreement with linear theory based on the correlation between [15] T. YOSHIDA, Effects of nozzle amplitude on production of uniformly sized
liquid droplets, in: Proc. of 14th Int. Conf. on the Properties of Water and
break-up length and excitation voltage is found up to a given exci-
Steam, Kyoto, Japan, 2004, pp. 765–770.
tation voltage beyond which it deviates. Based on the time- [16] E. Goedde, M. Yuen, Experiments on liquid jet instability, J. Fluid Mech. 40 (03)
resolved pressure data analysis it can be concluded that this devi- (1970) 495–511, http://dx.doi.org/10.1017/S0022112070000289.
[17] G. Amini, M. Ihme, A. Dolatabadi, Effect of gravity on capillary instability of
ation is only due to the elevated excitation level and not the exis-
liquid jets, Phys. Rev. E 87 (5) (2013) 053017, http://dx.doi.org/10.1103/
tence of higher orders in the excitation signal. No correlation PhysRevE.87.053017.
between this deviation and the critical voltage level between low [18] C. Weber, Zum zerfall eines flüssigkeitsstrahles, Z. Angew. Math. Mech. 11 (2)
and high uniformity regime could be observed. Further, findings (1931) 136–154, http://dx.doi.org/10.1002/zamm.19310110207.
[19] J. Eggers, E. Villermaux, Physics of liquid jets, Rep. Prog. Phys. 71 (3) (2008)
of other authors regarding the observation of rear merging satellite 036601, http://dx.doi.org/10.1088/0034-4885/71/3/036601.
droplets at low excitation voltages and forward merging satellite [20] M. Orme, On the genesis of droplet stream microspeed dispersions, Phys.
droplets at high excitation voltages could be confirmed. It is spec- Fluids A 3 (12) (1991) 2936–2947, http://dx.doi.org/10.1063/1.857836.
[21] K. Chaudhary, The Nonlinear Capillary Instability of a Jet Ph.D. thesis,
ulated that the non-axisymmetric jet profile could be related to an University of Southern California, 1977.
active displacement of the fluid similar to the one in a drop-on- [22] G. Brenn, On the controlled production of sprays with discrete polydisperse
demand system. drop size spectra, Chem. Eng. Sci. 55 (22) (2000) 5437–5444, http://dx.doi.org/
10.1016/S0009-2509(00)00167-6.
[23] H. González, F. García, The measurement of growth rates in capillary jets, J.
Funding source Fluid Mech. 619 (2009) 179–212, http://dx.doi.org/10.1017/
S0022112008004576.
[24] L. Rymell, H. Hertz, Droplet target for low-debris laser-plasma soft x-ray
This research did not receive any specific grant from funding generation, Opt. Commun. 103 (1–2) (1993) 105–110, http://dx.doi.org/
agencies in the public, commercial, or not-for-profit sectors. 10.1016/0030-4018(93)90651-K.
[25] B. Rollinger, N. Gambino, A.Z. Giovannini, L.S. Bozinova, F. Alickaj, K. Hertig, R.
S. Abhari, F. Abreau, Clean and stable LPP light source for HVM inspection
Acknowledgements applications, Proc. SPIE, vol. 9048, 2014, pp. 90482K–90482K.
[26] A. Giovannini, Droplet-Based Laser-Produced Plasma Sources - Photon
Emission, Matter Expansion and Mitigation, Ph.D. thesis, ETH Zurich, 2014.
The authors would like to thank Dr. Bob Rollinger for the valu- [27] M.J. Partlow, M.M. Besen, P.A. Blackborow, R. Collins, D. Gustafson, S.F. Horne,
able discussions during the design conception phase of the novel D.K. Smith, Extreme-ultraviolet light source development to enable pre-
A.F.R. Sanders et al. / Experimental Thermal and Fluid Science 86 (2017) 149–159 159

production mask inspection, J. Micro/Nanolith. MEMS MOEMS 11 (2) (2012), [40] R. Donnelly, W. Glaberson, Experiments on the capillary instability of a liquid
http://dx.doi.org/10.1117/1.JMM.11.2.021105, pp. 021105-1. jet, Proc. R. Soc. Lond. A, vol. 290, 1966, pp. 547–556.
[28] H. Feldmann, U. Mller, Light Sources for EUV Mask Metrology, Tech. Rep., [41] K. Chaudhary, L. Redekopp, The nonlinear capillary instability of a liquid jet.
International Workshop on EUV ad Soft X-Ray Sources, Dublin, 2012. Part 1. Theory, J. Fluid Mech. 96 (02) (1980) 257–274, http://dx.doi.org/
[29] S. Shimasaki, S. Taniguchi, Formation of uniformly sized metal droplets from a 10.1017/S0022112080002108.
capillary jet by electromagnetic force, Appl. Math. Model. 35 (4) (2011) 1571– [42] K. Chaudhary, T. Maxworthy, The nonlinear capillary instability of a liquid jet.
1580, http://dx.doi.org/10.1016/j.apm.2010.09.033. Part 2. Experiments on jet behaviour before droplet formation, J. Fluid Mech.
[30] H. Duan, W. Yang, C. Li, B. Lojewski, W. Deng, Scalable generation of strictly 96 (02) (1980) 275–286, http://dx.doi.org/10.1017/S0022112080002108.
monodisperse droplets by transverse electrohydrodynamic excitations, [43] J. Wissema, G. Davies, The formation of uniformly sized drops by vibration-
Aerosol Sci. Technol. 47 (11) (2013) 1174–1179, http://dx.doi.org/10.1080/ atomization, Can. J. Chem. Eng. 47 (6) (1969) 530–535, http://dx.doi.org/
02786826.2013.829208. 10.1002/cjce.5450470609.
[31] R. Zamora, J.H. Ortega, J. López, F. Faura, J. Hernández, Development of a facility [44] J.H. Hilbing, S.D. Heister, Droplet size control in liquid jet breakup, Phys. Fluids
for molten metal micro-droplets generation. application to microfabrication 8 (6) (1996) 1574–1581, http://dx.doi.org/10.1063/1.868931.
by deposition, Proc. Eng. 132 (2015) 110–117, http://dx.doi.org/10.1016/j. [45] N. Ashgriz, F. Mashayek, Temporal analysis of capillary jet breakup, J. Fluid
proeng.2015.12.486. Mech. 291 (1995) 163–190, http://dx.doi.org/10.1017/S0022112095002667.
[32] S. Cheng, S. Chandra, A pneumatic droplet-on-demand generator, Exp. Fluids [46] T.W. Petersen, G. van den Engh, Stability of the breakoff point in a high-speed
34 (6) (2003) 755–762, http://dx.doi.org/10.1007/s00348-003-0629-6. cell sorter, Cytometry A 56 (2) (2003) 63–70, http://dx.doi.org/10.1002/cyto.
[33] T.-M. Lee, T.G. Kang, J.-S. Yang, J. Jo, K.-Y. Kim, B.-O. Choi, D.-S. Kim, Drop-on- a.10090.
demand solder droplet jetting system for fabricating microstructure, IEEE [47] J. Eggers, Nonlinear dynamics and breakup of free-surface flows, Rev. Mod.
Trans. Electron. Packag. Manuf. 31 (3) (2008) 202–210, http://dx.doi.org/ Phys. 69 (3) (1997) 865, http://dx.doi.org/10.1103/RevModPhys.69.865.
10.1109/TEPM.2008.926285. [48] L. Crane, S. Birch, P. McCormack, The effect of mechanical vibration on the
[34] H. Sohn, D. Yang, Drop-on-demand deposition of superheated metal droplets break-up of a cylindrical water jet in air, Brit. J. Appl. Phys. 15 (6) (1964) 743.
for selective infiltration manufacturing, Mater. Sci. Eng. A 392 (1) (2005) 415– [49] A. Kalaaji, B. Lopez, P. Attane, A. Soucemarianadin, Breakup length of forced
421, http://dx.doi.org/10.1016/j.msea.2004.09.049. liquid jets, Phys. Fluids 15 (9) (2003) 2469–2479, http://dx.doi.org/10.1063/
[35] M. Orme, R.F. Smith, Enhanced aluminum properties by means of precise 1.1593023.
droplet deposition, J. Manuf. Sci. Eng. 122 (3) (2000) 484–493, http://dx.doi. [50] B. Lopez, A. Soucemarianadin, P. Attané, Break-up of continuous liquid jets:
org/10.1115/1.1285914. effect of nozzle geometry, J. Imag. Sci. Technol. 43 (2) (1999) 145–152.
[36] M. Orme-Marmerelis, R. Smith, High-Speed Fabrication of Highly Uniform [51] C. Bruce, Dependence of ink jet dynamics on fluid characteristics, IBM J. Res.
Metallic Microspheres, Patent US 6,562,099, May 13 2003. Dev. 20 (3) (1976) 258–270, http://dx.doi.org/10.1147/rd.203.0258.
[37] Y. Bar-Cohen, S. Sherrit, B.P. Dolgin, N. Bridges, X. Bao, Z. Chang, A. Yen, R.S. [52] H. Cline, T. Anthony, The effect of harmonics on the capillary instability of
Saunders, D. Pal, J. Kroh, et al., Ultrasonic/sonic driller/corer (usdc) as a liquid jets, J. Appl. Phys. 49 (6) (1978) 3203–3208, http://dx.doi.org/10.1063/
sampler for planetary exploration, Aerosp. Conf., 2001, IEEE Proc., vol. 1, IEEE, 1.325267.
2001, pp. 1–263. [53] P. Vassallo, N. Ashgriz, Satellite formation and merging in liquid jet breakup,
[38] A. Abdullah, M. Shahini, A. Pak, An approach to design a high power Proc. R. Soc. Lond. A, vol. 433, 1991, pp. 269–286.
piezoelectric ultrasonic transducer, J. Electroceram. 22 (4) (2009) 369–382. [54] A. Amirzadeh, S. Chandra, Small droplet formation in a pneumatic drop-on-
[39] B. Rollinger, M. Mansour, R. Abhari, High temperature fast response pressure demand generator: experiments and analysis, Exp. Therm Fluid Sci. 34 (8)
probe for use in liquid metal droplet dispensers, Rev. Sci. Instrum. 83 (6) (2010) 1488–1497, http://dx.doi.org/10.1016/j.expthermflusci.2010.07.013.
(2012) 065002, http://dx.doi.org/10.1063/1.4730046.

You might also like