PAPER www.rsc.org/loc | Lab on a Chip

Dow

nloa

ded

by U

NIV

ER

SIT

Y O

F T

EN

NE

SSE

E A

T K

NO

XV

ILL

E o

n 06

Oct

ober

201

0Pu

blis

hed

on 1

9 A

ugus

t 201

0 on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

0LC

0012

0AView Online

On-demand generation of monodisperse femtolitre droplets by shape-inducedshear†

Seung-Yong Jung,a Scott T. Rettererab and C. Patrick Collier*a

Received 16th June 2010, Accepted 28th July 2010

DOI: 10.1039/c0lc00120a

We describe a method for creating discrete femtolitre-scale water-in-oil droplets on demand, based

solely on a geometrically induced reduction in oil/water interfacial area at microfabricated junction

orifices. This on-demand generation method is driven by self-shear of droplets due to interfacial tension

induced forces resulting from a localized transition in microchannel height. The magnitudes of shear

stresses involved appear to be significantly less than the shearing instabilities used to split off daughter

droplets from aqueous mother plugs at microfabricated junctions in continuous water-in-oil segmented

flows, which implies that this method may be better suited for studying biochemical reactions and

reaction kinetics in droplets of decreased volume without loss of chemical reactivity due to

redistribution of surfactant density used to passivate the oil/water interface. Predictable droplet

generation rates under constant pressure conditions or the gated formation of one, two or more

droplets at a time with fixed pressure pulses have been demonstrated in a similar manner to active

on-demand droplet generation strategies, but with a simpler system not needing actuation and sensing

equipment beyond a pressure regulator.

Introduction

Droplet-based microfluidic platforms offer many opportunities

to confine chemical and biochemical reactants in discrete ultra-

small reaction volumes, and investigate the effects of increased

confinement on reaction kinetics.1,2 Most droplet-based systems

rely on generation of continuous streams of droplets in multi-

phase segmented flows, either via a ‘‘squeezing’’ mechanism

involving pressure fluctuations related to periodic blocking of oil

flow in a channel by aqueous plugs, or by ‘‘dripping’’ or ‘‘jetting’’

mechanisms involving shearing of the aqueous phase by the oil

phase.3,4 Droplets are generated in such flows at high frequencies

and transported downstream at high flow rates, which compli-

cates efforts to initiate chemical reactions with a well-defined

time zero, analyze reaction kinetics in real time, and further

manipulate droplets to carry out sequential multistep reactions.

In addition, generation of the smallest droplet sizes (1–10 mm in

diameter) due to droplet splitting generally requires strong shear

stresses, which can adversely affect the distribution of surfactant

stabilizing the oil–water interface and the passivation of the

interface against nonspecific adsorption of biomolecules such as

enzymes.5 On-demand generation of droplets allows more precise

temporal control of reactions inside droplets since each droplet

can be individually triggered, tracked and manipulated.

aCenter for Nanophase Materials Sciences, Oak Ridge NationalLaboratory, P.O. Box 2008, MS-6493, Oak Ridge, Tennessee, 37831,USA. E-mail: colliercp@ornl.govbBiological and Nanoscale Systems Group, Biosciences Division, Oak RidgeNational Laboratory, Oak Ridge, TN, 37831-6445, USA

† Electronic supplementary information (ESI) available: Sequences ofbright field images captured with CCD camera and high-speed CMOScamera of droplet formation dynamics, data files of droplet formationintervals and droplet diameter distributions, device fabrication usinghPDMS. See DOI: 10.1039/c0lc00120a

2688 | Lab Chip, 2010, 10, 2688–2694

Methods to control droplet splitting on demand involve

thinning or breaking the aqueous thread connecting a growing

droplet to the water pore to which it is attached, due to local

extensional and shear stresses at the orifice.6,7 The resulting size

of the droplet, or whether or not multiple droplets or even

aqueous jets are injected into the oil phase, will depend on

competition between the rates of thinning of the aqueous thread

versus inflation of the droplet by fluid flow through the orifice.8

For water-in-oil droplets created rapidly under steady-state

conditions with continuous segmented flows, shear is provided

by the cross-flow of the oil phase.

To form individual droplets on demand without cross-flow,

some other mechanism is necessary to create sufficient shear.

Recent examples of on-demand droplet generation methods in

the absence of steady-state cross-flows include the use of

programmable microinjectors,9 syringe pumps,10,11 piezoelectric

actuators,12 high-voltage pulses,13,14 electrowetting on dielec-

trics,15 and use of dielectrophoretic pressure.16 These methods

rely on actively controlled mechanical displacements of the

water–oil interface sufficient to split off aqueous droplets one at

a time at junctions of aqueous and oil channels. On the other

hand, continuous multiphase flows in microchannels generally

rely on passive means for forming water-in-oil droplets, based on

flow instabilities induced by interfacial forces.

This paper describes a method for creating individual femto-

litre scale aqueous droplets on demand at a junction between an

aqueous and oil channel in the absence of cross-flow, based on an

abrupt increase in the oil channel height relative to that of the

aqueous channel. This increase in channel height permits

a droplet to attain its lowest energy spherical shape, which

creates an interfacial tension induced force on the droplet suffi-

cient to detach it from the orifice. Similar mechanisms have been

described in reports of surface tension guided droplet storage in

nanolitre wells,17,18 droplet formation in microchannel

This journal is ª The Royal Society of Chemistry 2010

Dow

nloa

ded

by U

NIV

ER

SIT

Y O

F T

EN

NE

SSE

E A

T K

NO

XV

ILL

E o

n 06

Oct

ober

201

0Pu

blis

hed

on 1

9 A

ugus

t 201

0 on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

0LC

0012

0AView Online

emulsification membranes19 and arrays,20 and droplet formation

by Rayleigh-Plateau instabilities generated across a micro-

fabricated step.21 Monodisperse droplets were created at regular

intervals, up to several seconds between successive droplets,

simply by using constant backing pressure slightly higher than

the capillary pressure. This slow and predictable droplet

formation rate allows each individual droplet to be tracked and

manipulated separately and continuously in the main channel in

real time, in a similar manner to active on-demand droplet

generation strategies described above, but with a simpler system

not needing bulky actuation and sensing equipment. The only

control element that is needed outside the microfluidic chip is

a pressure regulator. In addition, pulsed output of the pressure

regulator allows droplet formation to be gated, due to the

predictable droplet generation rates at the junction. One, two, or

more droplets per pulse could be produced at a time, depending

on the width and magnitude of the pressure pulse.

Fig. 1 (A) Schematic of the microfluidic device used to produce

monodisperse femtolitre water-in-oil droplets on demand, consisting of

PDMS replica bonded to PDMS-coated glass coverslip. (B) Detail of the

silicon master used to form the PDMS replica, which highlights the

difference in channel height at the water–oil junction. The 200 mm wide

by 18 mm high oil channel was formed from patterned SU8 photoresist.

The 1 mm wide by 1 mm high aqueous channel (equivalent radius 500 nm)

was formed from combination of electron-beam lithography and reactive

ion etching of silicon.

Experimental

Deionized water was used as the aqueous phase. Soybean oil

(Sigma Aldrich) was purified of surface-active contaminants

(mainly monoglycerides) by gravity filtration through a column

packed with a 1 : 1 mixture of fluorisil and silica gel (100–

200 mesh, Sigma-Aldrich) until the equilibrium interfacial

tension at the oil–water interface matched that of reported values

for purified soy oil in the literature (g ¼ 31 mN m�1).22 4-Non-

ylphenyl-polyethylene glycol (NP-PEG) surfactant (Sigma-

Aldrich) was diluted 0.1% to 1.0% v/v in the purified soy oil in

some of the experiments. Interfacial tensions of oil/water, with

and without NP-PEG surfactant, and contact angles on PDMS

surfaces were determined from analyses of captured digital

images of pendant and sessile drops, as described previously.5,23

The microfluidic devices were fabricated in poly-

dimethylsiloxane (PDMS) using multilayer soft-lithographic

techniques. A combination of electron beam lithography and

photolithography was used to fabricate silicon masters for

fabricating PDMS devices by micromolding. The aqueous

channel was defined on the master with 300 nm thick electron

beam resist (ZEP-520A, Zeon Corp., Tokyo, Japan), which was

spin coated onto a 100 mm Si wafer at 6000 rpm for 45 seconds

and then soft-baked for 2 minutes at 180 �C. The resist was

exposed to an electron beam from a JEOL 9300-FS Electron

Beam Lithography System with a dose of 500 mC cm�2 and

patterns were developed in xylene for 30 seconds, rinsed with

isopropanol (IPA) and dried with nitrogen gas. After a brief

oxygen plasma treatment (Technics RIE, 100 sccm O2,

150 mTorr, 100 W for 6 seconds), a chromium layer 15 nm thick

for use as an etch mask was evaporated onto the patterned Si

wafer with a custom dual gun electron beam evaporator at

a deposition rate of 1 �A s�1, followed by lift-off by sonication in

acetone. The desired height of the aqueous microchannel was

achieved in bas relief by dry etching Si with an Oxford Plasmalab

100 inductively coupled plasma reactive ion etching system at an

etch rate �200 nm min�1 (Oxford Instruments, Concord, MA).

The oil channel was defined using SU8 2015 negative-tone

photoresist (Microchem Corp., Newton, MA). The photoresist

was spin-coated at 2000 rpm and soft-baked for 6 minutes at

90 �C. The photoresist-coated wafer was aligned and exposed to

This journal is ª The Royal Society of Chemistry 2010

UV light under a photomask at �13 mW cm�2 for 18 seconds on

a contact aligner (Neutronix-Quintel, Morgan Hill, CA). After

a 6 minute post-exposure bake at 90 �C, the resist was developed

in SU8 developer (Microchem Corp., Newton, MA), rinsed with

IPA and dried with nitrogen. The height of the etched Si feature

defining the aqueous channel was 1 mm, and the height of the

SU8 photoresist feature defining the oil channel on the master

was 18 mm, as measured with a Dektak profilometer (Veeco,

Malvern, PA). The profiles of both the oil and aqueous channel

features were rectangular.

The microfluidic device was fabricated by bonding a PDMS

replica, with the microchannels molded into it, onto a PDMS-

coated glass coverslip, so that all channel walls are PDMS, as

shown in Fig. 1. For the PDMS replica, the Si mold was silanized

with trimethylchlorosilane vapor (Aldrich) for 30 minutes in

order to facilitate release of the PDMS from the mold after

curing. Sylgard 184 PDMS (Dow Corning, Midland, MI) with

a 10 : 1 mass ratio of base to curing agent was thoroughly mixed,

degassed, poured onto the mold and degassed again, followed by

Lab Chip, 2010, 10, 2688–2694 | 2689

Dow

nloa

ded

by U

NIV

ER

SIT

Y O

F T

EN

NE

SSE

E A

T K

NO

XV

ILL

E o

n 06

Oct

ober

201

0Pu

blis

hed

on 1

9 A

ugus

t 201

0 on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

0LC

0012

0AView Online

curing for 30 minutes at 120 �C. The cured PDMS replica was

peeled off from mold and holes were punched with a 0.75 mm

hole-puncher (Harris Uni-Core, Ted Pella, Inc. Redding, CA).

The PDMS replica was bonded onto a #1 glass coverslip (Erie

Scientific Co., Portsmouth, NH) that had a 10 mm thick layer of

PDMS spin coated onto it (6500 rpm), followed by curing for 30

minutes at 120 �C. Bonding between the top PDMS replica and

the bottom PDMS-coated glass coverslip was activated by

plasma treatment of both bonding surfaces in an inductively

coupled plasma cleaner at 10.5 W for 20 seconds (Harrick,

Ithaca, NY). In order to render all the channel surfaces of the

completed device sufficiently hydrophobic, the bonded chips

were heated at 120 �C for an additional 48 hours to ensure

hydrophobic recovery of the PDMS.

4 mL glass vials with PTFE/silicone septum lids (C4015-17W,

National Scientific, Rockwood, TN) were used as sample reser-

voirs for the aqueous and oil phases, and were connected to high

precision closed-loop voltage–pressure transducers (Marsh

Bellofram, Newell, WV) by 24 gauge PTFE tubing (Small Parts,

Miramar, FL). The reservoirs were connected to the inlets of the

PDMS device by 23 gauge stainless steel tubing (Technical

Innovations, Brazoria, TX). Male-to-male luer lock adapters

(Qosina, Edgewood, NY) holding two 23 gauge needles, one

penetrating the septum of the vial cap, and the other connecting

to the 24 gauge PTFE tubing, were used for access into and out of

the sample vials. The pressure regulators were controlled by

a custom Matlab script (Mathworks, Natick, MA) through an

analog output board (16 bit resolution, 0–10 V range, USB3103,

Measurement Computing, Norton, MA), and were calibrated

using a Dwyer Series 475 Mark III digital manometer (Michigan

City, IN). Bright field images were acquired with an inverted

optical microscope (Eclipse TE 300, Nikon Instruments,

Melville, NY), using either a CCD camera (CoolSNAP-HQ,

Roper Scientific, Tucson, AZ) controlled with Metamorph

software (Universal Imaging Corp., Downing Town, PA), or

a high-speed CMOS camera (EPIX SV643, Buffalo Grove, IL).

Images were analyzed with ImageJ software (National Institutes

of Health).

Fig. 2 (A) Series of bright field images spaced at 82 ms intervals of the

formation and detachment of an individual 5.7 mm diameter droplet from

an aqueous channel (1 � 1 mm) into the oil phase, at a constant backing

pressure of 130.3 kPa. (B) Cross-sectional schematic of steps involved

during droplet formation corresponding to bright field images in (A).

Results and discussion

In order to form only one droplet from an orifice at a time,

actively controlled methods have been used to transiently bring

the internal pressure at the curved interface above the capillary

hold-off value without allowing subsequent drops to form until

needed.9–16 Regulation of just the backing hydrostatic pressure at

the orifice with enough precision for this degree of control has

not been demonstrated to date. This degree of control may be

achieved, however, by increasing the resistance to fluid flow in

the channel while providing a mechanism for shearing the

nascent droplet at the orifice.

At these scales, consideration of capillary effects due to the

surface energetics of the water and oil fluids with the hydro-

phobic PDMS channel walls become important. To form discrete

aqueous droplets on demand in an oil phase, the channel walls

must be kept hydrophobic; otherwise, the aqueous phase would

wet the walls of the oil channel without forming discrete droplets.

On the other hand, for hydrophobic channel walls wetted with

oil, a resistive capillary pressure must be overcome by the applied

2690 | Lab Chip, 2010, 10, 2688–2694

backing pressure in order to fill the aqueous channel with water.

For a rectangular channel, this pressure will depend on the

interfacial tension between the two immiscible phases

(g¼ 31 mN m�1 for purified soybean oil and water), the cosine of

the contact angle of the wetting fluid with the channel wall

(qoil with purified soybean oil was measured to be �20 to 30� on

hydrophobic PDMS), and the two radii of curvature associated

with one-half of the width, a¼ 1.0 mm, and height, b¼ 1.0 mm, of

the channel,24

PC ¼ 2g cos qoil

�1

aþ 1

b

�z110� 117 kPa (1)

Experimental values for the backing pressure required to fill

the 1 � 1 mm aqueous channel, PC0, were slightly higher than this

estimate, about 124–131 kPa, due to a combination of factors

increasing local flow resistance in the channel, including surface

roughness of the channel walls, and minor defects/obstructions in

the channel and junction orifice in the PDMS replica formed

during molding. These non-uniform variations in flow resistance

prevented control of fluid flow with the precision and accuracy

required for forming droplets on demand for channel lengths

greater than 7 mm. This also prevented precise positioning the oil/

water interface as a function of backing pressure exactly at the

opening into the oil channel without uncontrolled injection of

multiple droplets into the oil phase. Instead, the closest stable

position of the oil/water interface to the opening tended to be

slightly behind it, inside the aqueous channel. For this reason, the

backing pressure PC0 was adjusted such that the oil/water inter-

face was positioned as close as possible to, but not at the channel

opening. This position was stable indefinitely and was a repro-

ducible function of backing pressure. For the images in

Fig. 2A, the backing pressure necessary to fill the aqueous

channel and maintain the oil/water interface at this position was

This journal is ª The Royal Society of Chemistry 2010

Fig. 3 (A) Plot of droplet formation intervals from a 1 � 1 mm orifice as

functions of the pressure difference DP across the water–oil interface. The

inset shows that at and above DP ¼ 1.0 kPa, the droplet formation

interval is affected by previously formed droplets. (B) Droplet diameter

size distribution histogram corresponding to steady-state droplet

formation over the range of DP values in (A).

Dow

nloa

ded

by U

NIV

ER

SIT

Y O

F T

EN

NE

SSE

E A

T K

NO

XV

ILL

E o

n 06

Oct

ober

201

0Pu

blis

hed

on 1

9 A

ugus

t 201

0 on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

0LC

0012

0AView Online

PC0 ¼ 130.1 kPa. Because the flow resistances depended sensi-

tively on local variations in width in the aqueous injection

channel, the precise location of the pinned interface varied from

one device to the next. From numerous bright field images,

including those shown in Fig. 2A, we determined the mean and

standard deviation of the stable position of the interface to be

1.3 mm � 0.6 mm from the orifice.

Fig. 2A shows a series of bright field images, captured with

a CCD camera every 82 ms (corresponding to the maximum frame

transfer rate for the imaged pixel area), of an individual 5.7 mm

diameter droplet forming and detaching from the orifice at the

junction of a 1 mm wide� 1 mm high� 7 mm long aqueous channel

with a 200 mm wide � 18 mm high main oil channel. The corre-

sponding movie can be found in the ESI†. This sequence was not

triggered, but instead represents a series of successive ‘‘snapshots’’

taken under steady-state conditions, at a constant applied backing

pressure of 130.3 kPa. Fig. 2B shows a schematic of the proposed

droplet formation mechanism from a side-view perspective.

In the first panel of Fig. 2A and B, the curved oil/water

interface has started to protrude from the junction orifice with

a hemispherical shape, which indicates the internal pressure

acting on the water/oil interface is at or very close to the capillary

hold-off value corresponding to spontaneous growth of the

aqueous phase into the oil channel.16 By the next frame (within

82 ms), a fully grown droplet has formed at the interface, con-

nected to the water channel by an aqueous neck. Bright field

images captured with a fast CMOS camera (frame rate 841 ms)

indicate that the droplet formation process, beginning with the

oil/water interface at the capillary hold-off value, is complete

within 2.5 ms (see ESI†).

The distance from the floor of the device to the centerline

of the aqueous channel (500 nm) is significantly less than that

from the ceiling of the device to this centerline (17.5 mm). We

believe the key enabling process for droplet detachment on

demand for our device is the abrupt change in channel height

across the junction orifice, from the 1 mm height of the aqueous

channel to the 18 mm height of the oil channel. This increase in

height allows the rapidly growing droplet room to expand both

vertically and horizontally away from the orifice in order to

minimize the surface area of the droplet by approximating

a spherical shape above the centerline of the orifice. However, at

the hydrophobic floor and wall of the channel near the opening,

the droplet shape is distorted from spherical due to steric

hindrance.

This can be seen in the second panel of Fig. 2A and B, which

show an inflated droplet, the shape of which is clearly distorted

from spherical, near the channel wall. The local Laplace pressure

at the nose of the droplet is less than that at the neck due to

differences in the radii of curvature. The resulting pressure

gradient results in local extensional and shear stresses at the oil–

water interface that drives the growth of an oil film and thinning

of the aqueous neck at the orifice.

This process results in the droplet splitting off from the orifice

and recovering its lowest energy spherical shape.6,7 The newly

freed droplet then drifts away from the orifice in the remaining

panels, presumably due to slow fluid flow in the main oil channel

from a slight hydrostatic pressure imbalance between the inlet

and outlet of the oil channel. If needed, this slow drift in the oil

channel could be controlled with pressure regulators.

This journal is ª The Royal Society of Chemistry 2010

Fig. 3A is a plot of steady-state droplet formation intervals for

this channel geometry as a function of the pressure drop across

the water–oil interface in the aqueous channel, taken from over

50 bright field images. These data are included in the ESI†. We

assigned the pressure drop across the oil/water interface to be

equal to the difference between the applied backing pressure and

an effective capillary pressure of the channel, DP ¼ Pa � PC0. As

described above, PC0 corresponds to the minimum backing

pressure required to pin the interface at its closest stable position

behind the channel opening without flooding the oil channel;

DP ¼ 0 refers to an applied backing pressure equaling this

effective capillary pressure. The difference between neighboring

DP values in the plot, 0.2 kPa, corresponds to the resolution of

the voltage–pressure transducers used to regulate the backing

pressure, which was 0.1% of full scale (200 kPa).

Analyses of time dependent sequences of bright field images

taken at each value of DP were used to generate the mean and

standard deviation for each data point in Fig. 3A. Included in the

Lab Chip, 2010, 10, 2688–2694 | 2691

Fig. 4 Still images of droplet formation during pulsed operation from

four different movie sequences: (A) 10 ms pulse, Pa ¼ 134.5 kPa, (B) 10

ms pulse, Pa ¼ 138.6 kPa, (C) 15 ms pulse, Pa ¼ 134.5 kPa and (D) 20 ms

pulse, Pa ¼ 134.5 kPa.

Dow

nloa

ded

by U

NIV

ER

SIT

Y O

F T

EN

NE

SSE

E A

T K

NO

XV

ILL

E o

n 06

Oct

ober

201

0Pu

blis

hed

on 1

9 A

ugus

t 201

0 on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

0LC

0012

0AView Online

plot is a linear fit (R2 ¼ 0.98) to the data except for the last data

point at DP ¼ 1.0 kPa, which did not follow the linear trend due

to blocking of the aqueous channel opening by previously

formed droplets, as shown by the image in the inset.

Bright field image sequences for the range of DP values used in

Fig. 3A show that for most of the time between successive droplet

generation events (1 to 3.5 seconds, depending on DP), the water/

oil interface in the aqueous channel was pinned at its stable initial

position within the aqueous channel. The growth and detach-

ment of a droplet were rare, transient events within each droplet

generation cycle, occurring rapidly (within 2.5 ms) once the

interface moved to the channel opening and started to protrude

into the oil phase. As each droplet split off from the orifice, the

water/oil interface of the remaining water column recoiled back

to its initial stable position within the aqueous channel, due to

interfacial tension, to begin a new cycle. A detailed description of

the complex time-dependent behavior is beyond the scope of this

paper, but the hydrodynamics should be similar to that of

a macroscale dripping faucet, albeit with surface interactions

such as interfacial tension and contact angles dominating volume

effects such as gravity or inertial forces.

Fig. 3B is a plot of the droplet diameter size distribution

observed from bright field image sequences corresponding to 0.2

to 1.0 kPa, the range of DP values shown in Fig. 3A. The mean

and standard deviation from this distribution was 5.7 � 0.2 mm,

resulting in a coefficient of variation (COV ¼ std/m � 100%) of

3.5%. This diameter corresponds to a volume of 97 femtolitres

for a spherical droplet. The fact that the droplet size distribution

was relatively independent of the backing pressures used and

frequencies of droplet formation are consistent with the fact that

the ultimate droplet diameter is controlled more by interfacial

tension than flow rate at this length scale.

The slow, predictable droplet generation intervals shown in

Fig. 3A represent a different droplet formation methodology

than steady-state approaches to produce droplets at high

formation frequencies (up to kHz) based on continuous

segmented flows, and allows interrogation, tracking and

manipulation of individual droplets in the same way as active-

control schemes for on-demand generation. This slow, predict-

able steady-state rate also enables droplet formation to be gated,

by application of pressure pulses instead of a constant applied

pressure.

Fig. 4 shows still images of droplet formation occurring after

the application of a short pressure pulse (10–20 ms pulse dura-

tion). The number of droplets per pulse could be controlled either

with the magnitude (Fig. 4A and B) or the duration (Fig. 4C and

D) of the pressure pulse. Comparison of Fig. 4A and B shows

that an increase in DP of just 4.1 kPa, for the same pulse duration

(10 ms), results in a transition from one droplet per pulse for

Fig. 4A to numerous droplets per pulse for Fig. 4B. For Fig. 4B,

there were actually two sets of four-droplet injections per pulse,

due to transient pressure oscillations from the Marsh-Bellofram

voltage-to-pressure transducer.

Varying the duration of the pressure pulse was an easier way to

control the number of droplets per pulse in a digital manner than

varying the magnitude of the pressure pulse. Comparison of

Fig. 4A, C and D shows that by increasing the duration of the

pulse from 10 ms to 15 ms to 20 ms, all at Pa ¼ 134.5 kPa, one,

two or three droplets could be produced on demand

2692 | Lab Chip, 2010, 10, 2688–2694

reproducibly. Oscillations in the positions of the oil/water

interface in the channel occurred after each pulse due to the

pressure transducer, but the oscillations did not extend all the

way to the orifice of the channel as they did for Fig. 4B because

the pressure was lower.

Our hypothesis that the droplet formation mechanism is based

on the discontinuity in channel height at the oil/water channel

junction was tested by increasing the height of the aqueous

channel from 1.0 to 3.3 mm, corresponding to an increase in the

channel aspect ratio from 1 : 1 to 1 : 3. We found that the

structural strength of the commercial PDMS used (Sylgard 184)

for the 1 : 1 aspect ratio aqueous channel was not sufficient for

the 1 : 3 aspect ratio channels, which were prone to collapse

during pattern transfer. Therefore, a harder PDMS formulation

(hPDMS) with increased modulus was used for these channels.25

Our prediction was that a lower backing pressure would be

needed for filling the 1 : 3 aspect ratio channels, corresponding to

a lower expected capillary pressure in eqn (1). Our prediction was

verified; the backing pressure needed for forming droplets from

the 1 : 3 aspect ratio channels was 56.6 kPa, compared to

130.3 kPa for the 1 : 1 aspect ratio channels. In addition, we

predicted that the final droplet diameter would be larger for the

1 : 3 aspect ratio channel compared to the 1 : 1 aspect ratio

channel, since the distance from the axial centerline of the

aqueous channel, and hence, the droplet center, to the channel

floor was greater. This was also validated by experiment; the

mean and standard deviation of the droplet diameter distribution

from 1 : 3 aspect ratio channels were 10.6� 1.0 mm, compared to

5.7 � 0.2 mm for the 1 : 1 aspect ratio channels (ESI†).

Another test involved reducing the oil channel height from

18 mm to 1 mm, which is the same height as the 1 : 1 aspect ratio

aqueous channel. Without the room to inflate vertically into the

18 mm high oil channel, which would allow the droplet to assume

its lowest energy spherical shape and shear-off from the orifice, the

aqueous flow assumed an oblate, plug-like shape, bounded by the

reduced 1 mm channel height. Under these conditions, the aqueous

plug ultimately filled the entire volume of the oil channel (ESI†).

This journal is ª The Royal Society of Chemistry 2010

Dow

nloa

ded

by U

NIV

ER

SIT

Y O

F T

EN

NE

SSE

E A

T K

NO

XV

ILL

E o

n 06

Oct

ober

201

0Pu

blis

hed

on 1

9 A

ugus

t 201

0 on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

0LC

0012

0AView Online

The droplet formation mechanism was highly dependent on

interfacial tension induced forces. When the interfacial tension

was lowered with the addition of 0.1% v/v NP-PEG surfactant,

which has been shown to effectively passivate the oil/water

interface against nonspecific adsorption and inactivation of

enzymes,5 a lower backing pressure was required to fill the 1 : 1

aspect ratio channel and form droplets, 122.1 kPa, consistent

with eqn (1). In addition, the time interval between successive

droplets increased (to about 10 seconds). This was because the

time-dependent diffusion of surfactant molecules from the oil

phase and their dynamic adsorption at the three-phase interface

(oil/water/PDMS) in the aqueous channel became the rate-

limiting steps.26,27 When the surfactant concentration was

increased to 1.0% v/v, the interfacial tension was reduced enough

to the point where discrete droplets could no longer be formed;

instead the aqueous phase flooded the oil channel (ESI†).

Conclusions

This paper describes a method that enables precise temporal

control of individual droplet formation. This capability allows

each individual droplet to be tracked and manipulated separately

and continuously in real time. The mechanism for droplet

formation was controlled by interfacial tension coupled to an

abrupt change in channel height at a microfabricated junction

that allowed a reduction in the droplet surface area and inter-

facial energy of the system. This resulted in an interfacial tension

based force that sheared the aqueous thread connecting

a growing droplet to the water column, and drove detachment of

the droplet from the orifice. Individual water-in-oil droplets were

created at constant backing pressure with well-defined regular

intervals between successive droplets as long as several seconds,

without cross-flow of the oil phase. These predictable droplet

generation rates allowed the use of fixed pressure pulses to gate

the formation of one, two, or more droplets at a time, depending

on the length and magnitude of the pulse.

Because droplet formation was controlled predominately by

interfacial tension, rather than flow rate or applied pressure, the

formation of droplets on demand will be very sensitive to the

choice of surfactant, as well as oil and aqueous phase composi-

tion. We used purified soybean oil and deionized water to

demonstrate the method, but fluorinated and silicone-based oils

are also commonly used.1,2 Aqueous phases may contain surface-

active molecules, particularly for biological applications. As long

as careful calibrations are carried out for the effects of different

phases, surfactants and surface-active analytes, this method

should be broadly applicable to a wide range of practitioners in

microfluidics, including in biological fields.

Shearing mechanisms play important roles in droplet forma-

tion in multiphase microfluidics. We have shown previously that

strong shear stresses can adversely affect the distribution of

surfactant stabilizing the oil–water interface and the passivation

of the interface against nonspecific adsorption of biomolecules

such as enzymes.5 Although the droplet detachment mechanism

described here is driven by (self) shear due to interfacial tension

induced forces acting on the droplet, the magnitude of the shear

stresses involved appears to be significantly less than the shearing

instabilities used to split off droplets in continuous segmented

flows. This comparison can be made based on bright field

This journal is ª The Royal Society of Chemistry 2010

analyses of droplet deformation during the splitting off process.28

For the droplets formed on demand described here, the distor-

tion of the droplet from a spherical shape is far less in scope and

is more localized (at the aqueous neck connecting the droplet to

the orifice) than for droplets that are split off in rapid steady-

state flows. This implies that this method may be better suited for

studying biochemical reactions and reaction kinetics in droplets

of decreasing volume without loss of chemical reactivity due to

redistribution of surfactant density at the interface. At the same

time, this methodology is simpler than active control methods for

on-demand droplet generation, requiring only a pressure

regulator.

Acknowledgements

The authors acknowledge support for device fabrication from

the Center for Nanophase Materials Sciences, which is sponsored

by the Scientific User Facilities Division, Office of Basic Energy

Sciences, US Department of Energy. Device characterisation and

analysis sponsored by the Laboratory Directed Research and

Development Program of Oak Ridge National Laboratory,

managed by UT-Battelle, LLC, for the US Department of

Energy under Contract No. DE-AC05-00OR22725.

Notes and references

1 D. T. Chiu and R. M. Lorenz, Acc. Chem. Res., 2009, 42, 649–658.2 H. Song, D. L. Chen and R. F. Ismagilov, Angew. Chem., Int. Ed.,

2006, 45, 7336–7356.3 P. Garstecki, M. J. Fuerstman, H. A. Stone and G. M. Whitesides,

Lab Chip, 2006, 6, 437–446.4 A. G€unther and K. F. Jensen, Lab Chip, 2006, 6, 1487–1503.5 Y. Liu, S.-Y. Jung and C. P. Collier, Anal. Chem., 2009, 81, 4922–

4928.6 W. L. Olbricht, Annu. Rev. Fluid Mech., 1996, 28, 187–213.7 T. M. Tsai and M. J. Miksis, J. Fluid Mech., 1994, 274, 197–217.8 A. S. Utada, A. Fernandez-Nieves, H. A. Stone and D. A. Weitz,

Phys. Rev. Lett., 2007, 99, 094502.9 R. M. Lorenz, J. S. Edgar, G. D. M. Jeffries and D. T. Chiu, Anal.

Chem., 2006, 78, 6433–6439.10 R. M. Lorenz, G. S. Fiorini, G. D. M. Jeffries, D. S. W. Lim, M. He

and D. T. Chiu, Anal. Chim. Acta, 2008, 630, 124–130.11 M. He, J. S. Edgar, G. D. M. Jeffries, R. M. Lorenz, J. P. Shelby and

D. T. Chiu, Anal. Chem., 2005, 77, 1539–1544.12 J. Xu and D. J. Attinger, J. Micromech. Microeng., 2008, 18, 065020.13 M. He, J. S. Kuo and D. T. Chiu, Appl. Phys. Lett., 2005, 87, 031916.14 M. He, J. S. Kuo and D. T. Chiu, Langmuir, 2006, 22, 6408–6413.15 R. Fair, Microfluid. Nanofluid., 2007, 3, 245–281.16 P. R. C. Gascoyne, J. V. Vykoukal, J. A. Schwartz, T. J. Anderson,

D. M. Vykoukal, K. W. Current, C. McConaghy, F. F. Becker andC. Andrews, Lab Chip, 2004, 4, 299–309.

17 J. Shim, G. Cristobal, D. R. Link, T. Thorsen, Y. Jia, K. Piatelli andS. Fraden, J. Am. Chem. Soc., 2007, 129, 8825–8835.

18 H. Boukellal, �S. Selimovi�c, Y. Jia, G. Cristobal and S. Fraden, LabChip, 2009, 9, 331–338.

19 I. Kobayashi, M. Nakajima, K. Chun, Y. Kikuchi and H. Fujita,AIChE J., 2002, 48, 1639–1644.

20 S. Sugiura, M. Nakajima, S. Iwamoto and M. Seki, Langmuir, 2001,17, 5562–5566.

21 D. Saeki, S. Sugiura, T. Kanamori, S. Sato, S. Mukataka andS. Ichikawa, Langmuir, 2008, 24, 13809–13813.

22 A. G. Gaonkar and R. P. Borwankar, J. Colloid Interface Sci., 1991,146, 525–532.

23 H. Jung, C. K. Dalal, S. Kuntz, R. Shah and C. P. Collier, Nano Lett.,2004, 4, 2171–2177.

24 E. Delamarche, A. Bernard, H. Schmid, A. Bietsch, B. Michel andH. Biebuyck, J. Am. Chem. Soc., 1998, 120, 500–508.

25 H. Schmid and B. Michel, Macromolecules, 2000, 33, 3042–3049.

Lab Chip, 2010, 10, 2688–2694 | 2693

Dow

nloa

ded

by U

NIV

ER

SIT

Y O

F T

EN

NE

SSE

E A

T K

NO

XV

ILL

E o

n 06

Oct

ober

201

0Pu

blis

hed

on 1

9 A

ugus

t 201

0 on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

0LC

0012

0AView Online

26 M. J. Geerken, R. G. H. Lammertink and M. Wessling, J. ColloidInterface Sci., 2007, 312, 460–469.

27 B. Huang, H. Wu, S. Kim and R. N. Zare, Lab Chip, 2005, 5, 1005–1007.

28 G. I. Taylor, Proc. R. Soc. London, Ser. A, 1934, 146, 0501–0523.The length (L) and breadth (B) of a deformed droplet subjectedto shear stress by an external flow field can provide a measure ofthe capillary number Ca h shear stress/Laplace pressure fora stationary droplet (one not moving with external flow).

2694 | Lab Chip, 2010, 10, 2688–2694

A dimensionless parameter F h (L � B)/(L + B) can be derivedwhich is equivalent to the capillary number for a stationarydroplet. This condition should be approximately satisfied fora new droplet within the first few ms as it splits from the orifice,as seen in the images in Fig. 2 and in the ESI†. These imagesshow considerably less distortion of droplets from their lowestenergy spherical shape compared to the daughter droplets split offfrom aqueous mother plugs in rapid segmented flows described inref. 5.

This journal is ª The Royal Society of Chemistry 2010