Prediction and control of drop formation modes in microfluidic generation of double emulsions by single-step emulsification.

HYPOTHESIS
Predicting formation mode of double emulsion drops in microfluidic emulsification is crucial for controlling the drop size and morphology.


EXPERIMENTS AND MODELLING
A three-phase Volume of Fluid-Continuum Surface Force (VOF-CSF) model was developed, validated with analytical solutions, and used to investigate drop formation in different regimes. Experimental investigations were done using a glue-free demountable glass capillary device with a true axisymmetric geometry, capable of readjusting the distance between the two inner capillaries during operation.


FINDINGS
A non-dimensional parameter (ζ) for prediction of double emulsion formation mode as a function of the capillary numbers of all fluids and device geometry was developed and its critical values were determined using simulation and experimental data. At logζ>5.7, drops were formed in dripping mode; the widening jetting occurred at 5<logζ<5.7; while the narrowing jetting was observed at logζ<5. The ζ criterion was correlated with the ratio of the break-up length to drop diameter. The transition from widening to narrowing jetting was achieved by increasing the outer fluid flow rate at the high capillary number of the inner fluid. The drop size was reduced by reducing the distance between the two inner capillaries and the minimum drop size was achieved when the distance between the capillaries was zero.


Introduction
Core-shell capsules attract great interest for potential applications ranging from controlled release of drugs and nutrients [1][2][3][4], encapsulation of fragrances [5], to CO 2 capture and energy storage [6][7][8]. The conventional fabrication strategies for core-shell capsules such as complex coacervation [9], internal phase separation [10], layer-by-layer electrostatic deposition [11], interfacial polymerisation [12], and spray drying [13] often require multistage processing and are associated with low encapsulation efficiency, low reproducibility and a lack of control over the size of the capsules and the shell thickness. In addition, the synthesised particles are usually highly polydispersed [14].
Microfluidic emulsification is a promising strategy for production of monodispersed double emulsions with controlled drop size and morphology. Double emulsion drops can be created through single-step or two-step emulsification methods, using flowfocusing [15,16], cross-flow [17,18] or co-flow [19,20] drop makers. Utada et al. [21] developed a three-phase 3D glass capillary device for single-step generation of double emulsions by combining co-flow and counter-current flow focusing flow patterns. The device consists of two tapered round capillaries, coaxially aligned within a square capillary and separated at some distance from each other, Fig. 1a and b. The device provides precise control over the droplet size and shell thickness. The wettability of each capillary can be independently modified, which is advantageous compared to polydimethylsiloxane (PDMS) devices that suffer from poor wettability control [22]. However, the distance between the round capillaries is fixed during operation and the device is difficult to clean after use.
The drop formation in capillary devices occurs in two distinct instability modes, absolute and convective [23]. Dripping is the result of an absolute instability; the perturbations that lead to drop pinch-off grow at a fixed spatial location and at a frequency that is intrinsic to the system, leading to monodispersed drops. Jetting occurs due to convective instability, which is associated with the advection of the perturbations along the interface of the jet, which causes random variation in the pinch-off location. Therefore, the drops generated in jetting mode are polydispersed [23,24]. The mechanism of drop formation in each mode is governed by the interaction between viscous, inertial, interfacial, and gravitational forces [25]. However, the effect of gravity is negligible when the drop diameter is in the micrometre range [26,27].
In dripping mode, the interfacial force is dominant and the inertial forces are negligible. The drops are formed in the vicinity of the injection nozzle (in co-flow geometry) or within one orifice diameter downstream of the orifice of the collection capillary (in flowfocusing geometry) once the viscous force, exerted by the continuous phase, exceeds the pinning force arising from the interfacial tension [27,29].
Two types of jetting mode have been observed in capillary devices, narrowing and widening [30]. Narrowing jetting is associated with formation of a relatively thin and long jet that eventually breaks into small drops; this mode occurs when the viscous force generated by the continuous phase is significantly greater than the interfacial force. On the other hand, widening jetting occurs when the inertial forces of the jet dominate the interfacial tension force. The widening shape of the jet has been attributed to the deceleration of the jet [31].
The formation of single emulsion drops in two-phase glass capillary devices is well characterised [27,[30][31][32][33]. In three-phase glass capillary devices, inner and outer drops can be formed in the same or different modes and thus, versatile drop morphologies can be achieved. Due to complexity of the coaxial jet  [21,29,34,35]. Utada et al. [21] studied the formation of core-shell drops in both dripping and widening jetting modes and found that double emulsions were not formed when the outer drops were formed in dripping mode and the inner drops in jetting mode. Although double emulsion formation by narrowing jetting has been reproduced numerically [29,35], it has not yet been observed experimentally in glass capillary devices. In addition, the variation of the distance between the inner capillaries changed the drop formation mode in numerical simulations [29], but no experimental evidence has been provided so far. Lee and Weitz [34] observed the formation of multi-core double-emulsion drops in capillary devices in an inertia-induced dripping-to-widening jetting transition mode at relatively high flow rates of the inner and middle fluids. To the best of our knowledge, no parameter capable of predicting the drop formation regime in capillary microfluidics has been developed so far. In this study, a novel dimensionless criterion, log f, that can be used for prediction of drop formation mode in three-phase glass capillary devices, was introduced and its critical values at which transitions occur were determined using a large amount of simulation and experimental data. Different transitions of drop formation modes were observed experimentally and a high ability of log f to accurately predict the drop formation behaviour was confirmed. In addition, a glue-free, detachable and reusable threephase glass capillary device was fabricated, with the ability to adjust the distance between the opposing inner capillaries during operation. The device consists only of round capillaries and has a true axisymmetric geometry, unlike conventional capillary devices consisting of round inner capillaries placed within a square outer capillary, which have a quasi-axisymmetric geometry. The device was used to reveal the significant impact of the distance between the two tapered inner capillaries on droplet formation behaviour.

Materials and methods
where V and P are the velocity and pressure, respectively, while t, l, and q are the time, dynamic viscosity and density. F b is a source term which includes interfacial force, F r and gravitational force.
Since the drop diameter is in the order of micrometres, the gravitational force is negligible and F b ¼ F r [26,27]. In the VOF method, a momentum balance equation is solved for all the phases and the advection of interface is calculated by solving a transport equation, Eq. (3), for volume fraction, f 0 : The volume fraction gives the portion of the cell which is filled with either phase, where The CSF method was used to calculate F b in Eq.
(2) at the interface: where r is the interfacial tension and j is the local curvature of the interface, given by: andn is the unit normal, given by Eq. (9):

Numerical method
An unsteady-flow pressure-based segregated algorithm, established in finite-volume-based commercial software Ansys Ò Fluent v. 14.0, was used to solve the governing equations. The discretised moment equation was approximated using a second order upwind scheme. The PRESTO scheme was used to interpolate the pressure term. Although the PRESTO scheme is computationally expensive, it directly calculates the pressures at cell faces and avoids interpolation errors. The pressure and velocity were coupled using the SIMPLE scheme. The interface interpolation was obtained by the Geo-Reconstruct algorithm. A variable-time-step method based on the maximum Courant number of 0.35 was used to reduce the computational cost. To evaluate grid independence, five meshes with a resolution of 8 mm, 4 mm, 3 mm, 2 mm, and 1 mm were constructed. The difference in simulation results obtained with the resolutions of 1, 2, and 4 mm was negligible. Therefore, a mesh size of 2 mm was used around the injection orifice and inside the collection tube where the drops were formed, while a coarser mesh was used for the rest of the domain. Further details on the numerical model can be found elsewhere [29,38].

Dimensionless numbers
Non-dimensional numbers used in this study are the capillary numbers of inner, Ca 1 , middle, Ca 2 , and outer phase, Ca 3 , given by: ci =4Þ. Subscripts 1, 2 and 3 stand for the inner, middle and outer phase, respectively, D ci is the outer diameter of the injection capillary, and D co is the internal height of the square capillary, Fig. 1b.

Materials
Glycerol was supplied by Fisher Scientific (UK) and used as a viscosity modifier. Polyvinyl alcohol (PVA, M w = 13,000-23,000 g mol À1 , 87-89% hydrolysed) was purchased from Sigma-Aldrich (UK) and used as a hydrophilic stabiliser. Dow Corning Ò 749 Fluid (DC 749) was supplied by Univar (UK) and used as a lipophilic surfactant. Polydimethyl siloxane (PDMS) fluid (Dow Corning 200/10cs fluid) was obtained from VWR (UK) and used as the main constituent in the middle phase. 2-[methoxy(polyethyle neoxy)propyl]trimethoxysilane and octadecyltrimethoxysilane (OTMS) were purchased from Fluorochem, Ltd. (UK) and used for hydrophilic and hydrophobic treatment of glass capillaries. The Millipore Milli-Q Plus water purification system was used to produce water for preparing all aqueous solutions. The density, dynamic viscosity and interfacial tension were measured using pycnometer, capillary viscometer, and Krüss DSA-100 pendant drop tensiometer, respectively.

Droplet production
Fabrication of microfluidic devices: Two different kinds of threephase glass capillary device were used in this study: standard and novel. In a standard device, inner capillaries were manually centred within a square capillary and fixed on a microscope slide using a two-part epoxy (Devcon 5-Minute Epoxy Ò ) [29,38]. In this device, the capillaries are fixed in place and the distance between inner capillaries cannot be changed. This fabrication process is tedious and the device is difficult to clean; after several experiments, when the capillaries become fouled, a new device must be made. Hence, a novel glue-free separable device was designed and manufactured. The device has an adjustable distance between inner capillaries, which can be changed during operation and can be dismantled, cleaned and reassembled. The device is easier to fabricate since all the capillaries are placed into an engineered holder to avoid the manual centring. The holder was made of acetal copolymer with a density of 1.41 g/cm 3 , and manufactured by Computer Numerical Control (CNC) machining using a fully automated HAAS Super Mini Mill machine (HAAS Automation, Norwich, UK), Fig. 1c. The holder is similar to the 3D-printed holder developed by Martino et al. [39], but CNC milling offers a wide range of materials for manufacturing and provides a superior surface quality. First, round borosilicate glass capillaries of 1/0.58 mm (O.D./I.D.) (Intracel, UK) were pulled using a Flaming/Brown micropipette puller (P-97, Sutter Instrument Co., Linton Instrumentation, Norfolk, UK) to obtain injection and collection capillary tubes. The pulled capillaries were polished to the desired tip size using an abrasive paper (Black Ice Waterproof T402 Paper, Alpine Abrasives, UK) under visual control with the aid of a Narishige Model MF-830 microforge (Japan). A borosilicate round capillary of 2/1.56 mm (O. D/I.D.) (World Precision Instruments, UK) was then placed in the holder and used as the outer capillary. Then, the injection and collection capillaries were inserted into the outer capillary at both ends. Finally, Omnifit connectors were attached to the holder and used for introduction of the three fluids into the device, Fig. 1c.
Double emulsion formation: Three gas-tight glass syringes with Luer-lock fitting (VWR Catalyst Company, UK) were filled with the three fluids and placed on Harvard Apparatus 11 Elite syringe pumps. The inner phase was 5-85 wt% aqueous glycerol solution, the middle phase was 2 wt% DC 749 in PDMS fluid, and the outer phase was an aqueous solution containing 40-70 wt% glycerol and 2 wt% PVA. The fluids were delivered into the device using polyethylene medical tubing of 1.52/0.86 mm (O.D./I.D.) (Fisher Scientific, UK). Double emulsion drops were formed upon breakup of the coaxial jet composed of inner and middle fluids in the collection tube, Fig. 1b. The drop generation was monitored and recorded using Phantom V9.0 high-speed camera (Vision Research, UK) attached to an inverted microscope (XDS-3, GX Microscopes, UK) and interfaced to a PC computer. The ImageJ program developed at the National Institute of Health was used to analyse the captured frames.

Validation of numerical model
The developed numerical model was validated with several analytical models in terms of the inner and outer drop sizes, D 1 and D 2 , respectively. In the dripping mode, the numerical D 2 values were validated using analytical models developed by Erb et al. [28] and Utada et al. [21]. Erb h i D j;2 and D j;1 are the diameter of outer and inner jets at the orifice of the collection capillary, and D orif;ef is the effective orifice diameter, which corresponds to the diameter of the collection tube at the location where the double emulsion drops are pinched off. The model developed by Utada et al. [21] for dripping mode is given by the equation: where k ¼ ð3p=2inÞ 1=3 and D j;2 can be calculated using Eq. (12): The proportionality constant k in Eq. (11) depends on the maximum instability (in), which is a function of the viscosity ratio, l dispersed =l 3 [40,41], where l dispersed is the effective viscosity of the mixture of inner and middle phases. In current simulations, since the shell was relatively thick and the middle phase was much more viscous than the inner phase (l 2 ) l 1 Þ, l dispersed was approximated with the viscosity of pure middle phase, l 2 and in = 0.55. In jetting mode, D 2 can be predicted using Eq. (13) developed by Utada et al. [21]: where C 0 is a constant, which depends on the viscosity ratio, l 2 =l 3 [42]. In this work, C 0 = 79.4 and D j,2 can be estimated by solving Eq.
(14) for a (a ¼ D j;2 =D orif ) [21]: For both regimes, D 1 can be calculated from D 2 using the mass balance equation: predict the drop size in the dripping-to-jetting transitional regime, at Q 1 =ðQ 2 þ Q 3 Þ between 5 and 12 (the blue region in Fig. 2). Fig. 3 shows the evolution of cross-sectional normalised stream-wise velocity profiles along the collection tube in different drop formation modes. Starting from the dripping mode, Fig. 3a, the widening jetting was obtained by increasing Q 2 from 0.92 to 2.47 mL/h, Fig. 3b, and the narrowing jetting was achieved by increasing Q 3 from 2.47 to 24 mL/h, Fig. 3c, while keeping all other parameters constant. In each graph the two red diamonds indicate the locations of the inner and outer interface. At the entrance of the collection tube (X ¼ 0), the velocity profiles in dripping and widening jetting are flat and equal to the centreline velocity (V at r ¼ 0), except within the boundary layer where the velocity decreases sharply to zero at the tube wall. The centreline velocity in the widening jetting mode was larger than in the dripping mode, due to higher Q 2 . At X ¼ 0, a saddle-shaped velocity profile was observed in the narrowing jetting regime and the maximum velocity was observed in the continuous phase stream at r=r c = 0.26. Due to the strong elongational flow of the outermost fluid that pulls the jet downstream, the centreline velocity was much larger in the narrowing jetting than in the widening jetting and dripping modes, although Q 1 was the same. The two red diamonds almost overlap each other due to small shell thickness and both interfaces are very close to the centreline due to jet thinning by strong shear and pressure forces.

Evolution of velocity profile in collection tube in different drop formation modes
In the dripping regime, the velocity profile further downstream remained almost flat, Fig. 3a, but at X ¼ 1:13D orif the velocity of the outermost fluid was higher than the velocity of the middle and inner fluids, due to jet widening at the tip as a result of drop formation. The centreline velocity decreases along the tube due to an increase in the tube cross-sectional area. In the widening jetting mode, the velocity profiles further downstream approach a parabolic shape with the maximum velocity located at the centreline. The velocity at the tube axis increases at X ¼ 4:58D orif due to necking of the inner jet. In the narrowing jetting regime, the velocity profile changes from a saddle shape at X ¼ 0 to parabolic at X ¼ 3:85D orif . In addition, almost the same velocity profile with a small decrease in the centreline velocity was observed until the end of the computational domain.

Prediction of double emulsion formation mode
Each drop generation mode can be characterised with a certain range of D 2 =L 2 values, where L 2 is the break-up length, i.e., the distance from the orifice of the injection capillary to the point of drop pinch-off. In dripping mode, D 2 is similar to D orif , while L 2 is very small, thus D 2 =L 2 is relatively large (D 2 =L 2 > 0:3). In narrowing jetting mode, D 2 is small and L 2 is large and consequently, D 2 =L 2 is small (D 2 =L 2 < 0:1). In widening jetting, D 2 is larger than in narrowing jetting and dripping mode, but L 2 is much larger than in dripping mode and thus, D 2 =L 2 is between the limits for the dripping and narrowing jetting mode (0:1 < D 2 =L 2 < 0:3). The drop formation mode can be altered by changing fluid properties and flow rates [29], which can be expressed in terms of Ca 1 , Ca 2 , and Ca 3 . Moreover, the variations in D orif can impact the drop formation regime with higher D orif values favouring the dripping mode [29]. The numerical results collected over a wide range of fluid properties, flow rates and device geometries were analysed to develop an empirical parameter, f, that can be used for prediction of the double emulsion formation regime:  Fig. 4. Good agreement between the measured and calculated f values was achieved. For log f > 5:7 (green region), double emulsion was formed in dripping mode, while log f < 5:7 (yellow region) was  associated with the jetting mode. The widening jetting regime was observed for 5 < log f < 5:7, while the narrowing jetting mode occurred at log f < 5. The transition between the regimes was not sharp but occurred gradually near the critical log f values, which might cause inaccurate regime predictions near the critical f values. The ability of log f to predict the drop formation regime was tested experimentally in Figs. 5 and 6. In most cases, log f calculated from Eq. (16) was able to accurately predict the drop formation regime. In addition, log f was calculated for the numerical data points in Fig. 2 and the results are shown in Fig. S1 in the supplementary material. The dripping-to-jetting transition occurred at log f ¼ 5:7. The D 2 =L 2 values were plotted on the same graph and the dripping-to-jetting transition occurred at D 2 =L 2 ¼0:3, as predicted. It should be highlighted that Eq. (16) was developed for primary drops and does not take into account the formation of satellite drops.

Transitions of drop generation modes
In this section, the effect of Ca 1 , Ca 2 , and Ca 3 on the transitions of drop generation regimes has been investigated experimentally. The inner phase was 85 wt% aqueous solution of glycerol. The middle phase consisted of 2 wt% DC 749 in PDMS fluid, and the outer phase was a mixture of 70 wt% glycerol and 2 wt% PVA in water. Core-shell drops produced using this formulation were stable for at least several hours after production [8,43]. Fig. 5 shows the transition from the dripping to widening jetting achieved by increasing Q 2 and thereby Ca 2 at constant values of Ca 3 and Ca 1 . An increase in Ca 2 from 0.7 Â 10 À4 to 3.6 Â 10 À4 (Fig. 5a-d) caused an increase in D 2 from 361 to 504 mm and a decrease in D 1 from 266 to 245 mm, which resulted in 2.7 times thicker shell. The reduction in D 1 was due to higher shear force exerted by the middle phase on the inner liquid, which led to faster  generation of inner drops. According to the mass balance equation: , higher drop generation frequency, f 1 , at Q 1 ¼ constant results in smaller D 1 . An increase in D 2 can be attributed to the larger inertia of the middle phase at higher Ca 2 and introduction of a higher amount of the middle phase into the outer drop prior to drop pinch-off [29].
In Fig. 5a-c, the outer drops were formed in dripping mode (log f > 5.7), but in Fig. 5d, the value of log f was at the boundary between dripping and widening jetting. At Ca 2 = 4.1 Â 10 À4 (Fig. 5e), the outer drops were formed in the widening jetting regime (log f = 5.6); two smaller inner drops were encapsulated within each outer drop and the diameter of the outer drops was larger than the diameter of the collection tube. At Ca 2 = 6.7 Â 10 À4 (Fig. 5f), the outer drops were formed by pinching of a long, wide jet (D 2 =L 2 < 0.3), while the inner drops were formed in the dripping mode. At Ca 2 < 0.7 Â 10 À4 , a thin oil shell burst during drop formation and the inner phase leaked into the continuous phase. At Ca 2 > 7.9 Â 10 À4 , the jet diameter reached the orifice diameter causing wetting of the orifice and the formation of double emulsion drops failed.
The dripping-to-narrowing-jetting transition could not occur by changing Q 2 at constant Ca 3 and Ca 1 . However, the widening-tonarrowing jetting transition, which was not reported earlier, was achieved by gradually increasing Ca 3 at relatively high inner phase flow rate. Fig. 6 shows the transition from widening to narrowing jetting achieved by increasing Ca 3 at constant Ca 1 and Ca 2 , starting from a Ca 1 value which was 13 times larger than in Fig. 5. At low Ca 3 ( Fig. 6a and b), both inner and outer drops were formed in the widening jetting mode (log f = 5.4-5.5). The drops were large and the jets occupied a significant portion of the tube cross section. At high Ca 3 ( Fig. 6e and f), both inner and outer drops were formed in the narrowing jetting mode (log f = 4.7-4.8). For intermediate Ca 3 values (Fig. 6c-d), the regime was not clearly defined and log f was close to the critical value of 5.0 at which the transition from widening to narrowing jetting occurs (log f= 4.9-5.2). In Fig. 6c, monodisperse core-shell drops with thin shells were formed at relatively high flow rates of the inner and outer phases.
By increasing Ca 3 from 4.6 Â 10 À3 to 9.2 Â 10 À3 , the length of the jet decreased due to increasing shear exerted by the outer fluid, causing a faster jet pinch-off and formation of smaller drops.
At the transition between the widening and narrowing jetting (log f % 5, Fig. 6d), the inner drops reached the minimum size and the neck of the middle phase between the first and second inner drop was stable, causing the formation of dual-core double-emulsion drops with tightly packed core drops. Interestingly, two core drops were also formed at the transition between dripping and widening jetting (Fig. 5e), but the outer drops were much larger. Finally, the increase in Ca 3 from 15.3 Â 10 À3 to 21.5 Â 10 À3 (Fig. 6d-f) resulted in shear-driven stretching of the inner phase to an increasingly long and narrow jet, which led to small drops. Fig. 7 shows the effect of distance between the two orifices, DL, on the drop formation behaviour at constant fluid flow rates. Without interrupting the drop generation process, the injection capillary was gradually moved in the downstream direction. The inner phase was 5 wt% aqueous solution of glycerol, the middle phase was 2 wt% DC 749 in PDMS fluid, and the continuous aqueous phase contained 40 wt% glycerol and 2 wt% PVA.

Effect of distance between inner capillaries on drop generation
At DL=D orif = À1, due to low shear from the surrounding middle phase, the inner drops grew to a large size before detachment from the tip of the injection tube. An increase in DL=D orif from À1 to À0.2 caused a 2.3-fold reduction in D 1 from 406 to 179 mm. By approaching the orifice of the collection tube, the growing inner drops experienced an increasing shear from the middle fluid pulling the inner jet through the orifice at higher force. Therefore, increased shear accelerated the pinch-off of inner drops causing a reduction in D 1 . At DL ¼ 0, the double emulsion drops reached the minimum size, because the jet experienced the highest shear. At DL > 0, due to increasing cross-sectional area of the collection tube, the shear exerted by the continuous phase was smaller, causing jetting of the middle phase and formation of outer drops with two inner drops [34,43]. However, both inner and outer drops were highly uniform.
The formation and advection of inner drops can considerably destabilise the outer interface and suppress the middle phase jetting. At DL=D orif = À1, when the inner drops were formed relatively far from the outer interface, the middle phase jet was very stable. Therefore, a long middle phase jet was developed within the collection tube, which eventually broke up into large drops. When the inner drops were formed very close to the orifice, e.g., at DL=D orif = À0.2, the growing inner drops strongly interacted with the outer interface. In this geometry, the outer interface was significantly deformed by the inner drops, which led to hydrodynamic instabilities in the jet and caused its early break-up into smaller drops. At DL=D orif ¼ 0.2 or 0.4, a very thin stream of the middle phase was formed at the entrance of the collection tube, due to the relatively long distance between the drop generation and the entry section of the tube. Further thinning of the middle phase stream around the injection tube at DL=D orif > 0:4 caused interruption of the middle phase flow and double emulsion drops failed to form.

Conclusions
The developed Volume of Fluid-Continuum Surface Force (VOF-CSF) numerical model was successfully validated with existing analytical solutions [21,28]. The velocity profile at the entry section of the tube was flat in dripping and widening jetting mode, while a saddle-shaped profile was observed in narrowing jetting mode. The numerical results were used to develop a dimensionless parameter, f, for prediction of drop formation mode, which is a function of the capillary numbers of the three fluids and the device geometry. For log f > 5:7, double emulsion drops were formed in dripping mode, the widening jetting was observed for 5 < log f < 5:7, while the narrowing jetting occurred at log f < 5.
The dimensionless criterion f was correlated with the ratio of the break-up length to drop diameter and tested against a large number of numerical and experimental results to show its high accuracy in predicting the drop formation mode.
Narrowing jetting reported in numerical simulation studies [29,35] was achieved starting from widening jetting by increasing the flow rate of the outer fluid at constant flow rates of the inner and middle phases. Near the transition between widening jetting and narrowing jetting, it was possible to generate monodispersed core-shell drops with thin shells at much higher flow rates of inner fluid than in the dripping mode.
A glue-free reusable capillary device was used to change the separation distance between the capillary inlets during operation to manipulate the drop formation behaviour. The device consists of two coaxial round capillaries, and therefore, has a true axisymmetric geometry, which differs from the quasi-axisymmetric geometry of a recently fabricated reusable capillary device [44]. The size of core-shell drops was significantly affected by changing the distance between the two tapered capillaries. The minimum drop size was achieved when the orifice of the injection tube was placed exactly at the entry section of the collection tube. In future work, the capillary device will be equipped with a thermoelectric generator in order to study the effect of temperature on generation of double emulsion drops.