Sensors and Actuators B 117 (2006) 431–436
`
`Optical detection for droplet size control in microfluidic
`droplet-based analysis systems
`∗
`
`, Sumantri Lassemono, Franck Alexis Chollet
`Nam-Trung Nguyen
`School of Mechanical and Aerospace Engineering, Nanyang Technological University, 50 Nanyang Avenue, Singapore 639798, Singapore
`
`Received 11 May 2005; accepted 2 December 2005
`Available online 18 January 2006
`
`Abstract
`
`This paper reports on a hybrid polymeric microfluidic device with optical detection for droplet-based systems. The optical part of the device
`is integrated by a hybrid concept. The microfluidic structures were fabricated using CO2 laser on poly methylmethacrylate (PMMA) substrate.
`The microfluidic network consists of two microchannels for forming droplets of an aqueous liquid in an immiscible carrier liquid. The optical
`component consists of two optical fibers for guiding laser light from the source, through the detection point, to a photo diode. The formed droplets
`pass the detection point and diffract the incoming laser light. The detected signal at the photo diode can be used for evaluating droplet size, droplet
`shape, and droplet formation frequency. The device can detect very high formation frequencies, which are not detectable using a conventional CCD
`camera/microscope setup.
`© 2005 Elsevier B.V. All rights reserved.
`
`Keywords: Droplet microfluidics; Polymeric micromachining; Lab on chip; Optical detection
`
`1. Introduction
`
`Droplet-based microfluidics has been emerging in the recent
`years because of its potential and apparent advantages. One of
`the key advantages of this concept is the small sample volume
`on the order of picoliters and nanoliters. A number of concepts
`such as thermocapillary [1], electrowetting [2,3], or multi-phase
`flow [4] can be used for generating and controlling of droplets.
`Droplet formation based on multi-phase flow is easy to imple-
`ment in a continuous-flow system. In this case, the fluidic system
`consists of two immiscible phases such as an aqueous liquid and
`an oil. The balance between the shear stress of the carrier flow
`and the interfacial tension between the two liquid phases leads
`to the formation of droplets [5].
`Systems generating micro-droplets have been successfully
`used as microreactors for chemical analysis and protein crystal-
`lization [4]. Furthermore, dispersed droplets of one liquid in a
`second liquid can form an emulsion, which have many appli-
`cations in food industries and cosmetic industries. Emulsion is
`important for packaging small amounts of fluid and other active
`
`∗
`
`Corresponding author. Tel.: +65 67904457; fax: +65 67911859.
`E-mail address: mntnguyen@ntu.edu.sg (N.-T. Nguyen).
`
`0925-4005/$ – see front matter © 2005 Elsevier B.V. All rights reserved.
`doi:10.1016/j.snb.2005.12.010
`
`ingredients such as drugs. Encapsulation of nanoliter droplets
`can also be achieved with a double emulsion. An intermediate
`fluid layer works as an additional barrier between the inner fluid
`and the carrier fluid. Recent interest of the research community
`on double emulsion in micro-scale shows its potential signifi-
`cance [6,7]. Since the droplet’s size and its other properties are
`important for the actual application, a detection system for the
`micro-droplets is vital for providing a feedback signal to the
`droplet formation process.
`Most of the recent works on micro-droplets only report
`devices made of polydimethylsiloxane (PDMS) and glass using
`external syringe pumps. A microscope and a CCD-camera are
`usually used for characterization of the droplets. The whole setup
`is rather bulky and the collected data are difficult to analyze
`automatically. Furthermore, expensive high-speed camera and
`synchronized strobe illumination are needed for capturing pro-
`cesses with droplet frequencies on the order of tens Hertz and
`above.
`In this paper, we present a poly methylmethacrylate (PMMA)
`device for droplet formation. The device has a hybrid-integrated
`optical system for evaluating parameters of the droplet forma-
`tion process such as the formation frequency, droplet size, and
`contact angles of the receding and advancing edges. Very high
`formation frequency can be detected by this device. The paper
`
`1
`
`

`
`432
`
`N.-T. Nguyen et al. / Sensors and Actuators B 117 (2006) 431–436
`
`The effective drag interfacial AD grows with the droplet.
`Assuming that the droplet is a sphere, the effective drag sur-
`face at the detachment moment is:
`AD = πD2
`2
`where Dd is the diameter of the generated droplet. Initially, the
`interfacial tension is large enough to keep the small droplet at
`the injection port. At the detachment moment, the continuous
`droplet growth makes the drag force large enough to release the
`droplet. Substituting (3) into (1) results in the droplet diameter:
`
`d
`
`(cid:1)
`
`(3)
`
`Dd = 2
`
`CS
`CD
`
`Di
`
`σ
`ρcU2
`c
`
`(4)
`
`The formation frequency can be estimated from the flow rate
`of the aqueous liquid ˙Qd and the droplet volume Vd as:
`f = ˙Qd
`(5)
`Vd
`Using the droplet diameter Dd and the relation ˙Qd = α ˙Qc,
`the formation frequency in (5) can be expressed as:
`
`(6)
`
`U4
`c
`
`32c
`
`ρ
`
`34
`
`σ
`
`32
`
`3αD2
`c
`16((CS /CD)Di)
`
`f =
`
`where Dc is the diameter of the carrier channel. Eq. (6) shows
`a nonlinear relation between the formation frequency and the
`average carrier’s velocity (f ∝ U4
`c ) or flow rate (f ∝ ˙Q4
`c). Fur-
`thermore, if the Marangoni force is considered in Eq. (1), and if
`the surfactant concentration is high, there will be an additional
`term for such force in the numerator of (6), resulting in a higher
`formation frequency.
`Our droplet-based microfluidic device consists of two parts: a
`microchannel system for droplet formation and an optical detec-
`tion system, Fig. 1. The microfluidic network consists of a large
`carrier channel and a small injection channel. The aqueous liq-
`uid enters through the injection channel, while an immiscible
`carrier liquid is introduced into the carrier channel. The two
`channels form a T-junction, at which droplet formation occurs.
`After droplets are formed and stabilized, they can be detected
`at a downstream position. The detection system is based on the
`optical concept. Laser light is guided into the microchannel by an
`optical fiber. After passing through the microchannel, the light
`is received at the other side by a second optical fiber which leads
`the laser to an optical sensor. The passing-by droplets change
`the intensity of the light due to diffraction and absorption. Thus
`the size and shape of the droplet can be well recorded as the time
`signal of the optical sensor.
`On the receiving side of the microchannel, the laser light
`diffracted by the propagation and by the different interfaces must
`enter the fiber within a fixed angle. This angle depends on the
`optical fiber’s numerical aperture (NA) and the refractive indices
`of the fiber’s core and cladding. In our later experiments, the
`optical fiber had a numerical aperture of 0.22. Theoretically,
`◦
`the maximum angle of the incident light in our device is 12.7
`.
`
`Fig. 1. Concept of formation and detection of liquid droplets in a microchannel.
`
`first discusses a simple theory on droplet formation to identify
`the key parameters of this process. Next, the fabrication of the
`device and experimental results are presented and compared with
`the theory.
`
`2. Formation and optical detection of micro-droplets
`
`Fig. 1 depicts a simple model of the formation process of
`a liquid droplet in another immiscible carrier fluid. The fol-
`lowing model only serves the purpose of understanding the
`relations between key parameters such as droplet size, forma-
`tion frequency, flow rates, and most importantly the interfacial
`tension between the two liquid phases. The model assumes a
`fixed flow rate ratio between the aqueous liquid and carrier liq-
`uid (α = ˙Qd/ ˙Qc). We further assume that the droplet size is
`small (α(cid:2) 1). Since the droplets are formed in micro-scale and
`the flows are in steady state, mass related forces such as inertial
`force, momentum force and buoyancy force are neglected in this
`model. If the aqueous liquid contents a surfactant, the surfactant
`concentration at the droplet surface is not uniformly distributed
`during the process of droplet growth. The distributed surfactant
`concentration leads to a gradient of interfacial tension on the
`droplet surface. This interfacial tension gradient in turn induces
`a Marangoni force on the droplet. If the surfactant solution is
`diluted, the Marangoni force is assumed to be small and neg-
`ligible. The injection channel and the carrier channel are both
`assumed to be cylindrical.
`Considering all the above assumptions, the force balance
`includes only the drag force of the carrier flow and the inter-
`facial tension at the injection port:
`= Finterfacial tension
`AD = CSπDiσ
`where ρc, Uc, AD, Di, and σ are the density of the carrier fluid, the
`average velocity of the carrier flow, the effective drag surface,
`the diameter of the injection port, and the interfacial tension,
`respectively. In addition, CD and CS are the drag coefficient
`and the coefficient for the interfacial tension. The coefficient CS
`depends on the contact angle and the shape of the injection port.
`In this model CS is assumed to be constant. We assume for CD
`the drag coefficient of a hard sphere at a low Reynolds number
`Re:
`CD = 24
`
`(1)
`
`(2)
`
`Fdrag
`CDρU2
`c
`
`12
`
`.
`
`Re
`
`2
`
`

`
`N.-T. Nguyen et al. / Sensors and Actuators B 117 (2006) 431–436
`
`433
`
`Fig. 2. Cross section of the laser machined microchannel.
`
`Fig. 4. Microfluidic device for droplet formation and optical detection.
`
`Beyond this angle, light would be refracted away from the fiber
`and would not reach the photo detector.
`
`3. Device fabrication and experimental setup
`
`Our polymeric technology is based on direct writing on
`PMMA using a CO2 laser beam. Because the substrate mate-
`rial is ablated by thermal energy, the microchannel cross section
`has the same shape as the beam intensity distribution, which has
`a typical Gaussian shape, Fig. 2. The width and depth of the
`ablated channel depend on the laser power and the beam speed.
`In our device, the injection channel and the guides for inserting
`the optical fibers are 175 ␮m in width and 205 ␮m in depth. The
`carrier channel is larger with a width of 340 ␮m and a depth of
`340 ␮m. Fig. 3 shows the T-junction of the injection channel and
`the carrier channel.
`Our device has three fluidic interconnects, two for the inlets
`and one for the outlet. Close to the outlet, two guides are
`machined into the substrate for aligning the two glass fibers.
`These fibers are used for optical detection of the micro-droplets.
`The optical fiber (AFS105/125Y from THORLABS Inc.) has a
`
`core diameter of 105 ␮m, a clad diameter of 125 ␮m, a buffer
`diameter of 250 ␮m, a numerical aperture of 0.22. After posi-
`tioning the fibers, the whole device is bonded thermally at
`a temperature above the glass temperature of PMMA. Fig. 4
`depicts the complete device.
`For detecting the droplets, one glass fiber is positioned and
`aligned to a laser source (laser diode, λ = 635 nm), the other
`fiber is connected to an avalanche photodiode module (APD,
`C5460-01, Hamamtsu, Japan). The complete experimental setup
`is shown in Fig. 5.
`In our experiments, oil with a viscosity of 6.52× 10
`−2 Pa s
`enters the channel through the inlet for carrier liquid. The other
`inlet is for the aqueous liquid. The aqueous liquids are pure DI-
`−3 Pa s) or diluted detergent
`water (viscosity of approximately 10
`solutions. A liquid detergent was diluted in DI water with differ-
`ent volume ratios to form our test samples. Both carrier liquid
`and aqueous liquid are driven by a syringe pump. The diameters
`of the syringes have a fixed ratio of 1:2. Thus, the total flow rate
`of oil is two times the flow rate of water. To relate the output
`signal to the droplet parameters, the images of droplet forma-
`
`Fig. 3. Microchannels at the T-junction.
`
`Fig. 5. Experimental setup for droplet formation and detection in a microfluidic
`device.
`
`3
`
`

`
`434
`
`N.-T. Nguyen et al. / Sensors and Actuators B 117 (2006) 431–436
`
`Fig. 6. Droplet formation in the microchannel (flow rate of the aqueous liquid is 30 ␮L/h): (a) pure DI water, (b) detergent diluted in DI water, volume ratio of
`12.5× 10
`−3.
`
`tion are captured with a CCD camera and a microscope system.
`The CCD sensor has a fixed exposure time of 20 ms. Thus, only
`formation frequencies on the order of 1 Hz can be observed and
`recorded clearly. The output signal of the optical detection sys-
`tem is recorded with a digital oscilloscope and transferred to a
`PC for further evaluations.
`
`4. Results and discussion
`
`4.1. Droplet shapes
`
`Fig. 6 shows the typical micro-droplet formed inside the
`microchannel. The advancing and receding edge of the droplet
`have different radii of curvature. Using the CCD camera, the
`advancing and receding edge of a water droplet can not be clearly
`differentiated. The difference between the two edges is more
`apparent when the interfacial tension decreases with detergent.
`Firstly, the receding side of the droplet becomes concave, while
`the advancing side of the droplet remains convex, Fig. 6(b). Sec-
`ondly, the weaker interfacial tension allows the droplets to be
`formed more frequently. The formation frequency can be mea-
`sured accurately using the optical detection system.
`Fig. 7 shows the typical signals of the droplets detected by the
`APD. The micro-droplets are well recognized as a pulse in the
`time signal. In the case of pure water, the pulse is well defined
`with two peaks representing the two edges of the micro-droplet.
`While measurement with the CCD camera can not differentiate
`the two edges of the water droplet, a clear difference in peak
`heights can be observed clearly in the optical signal, Fig. 7(a).
`Decreasing the interfacial tension between the aqueous liquid
`and the carrier liquid changes the shape of the droplet sig-
`nificantly. The detected time signal confirms the deformation
`observed previously with the CCD camera. At higher flow rates,
`the kinetic energy dominates over the surface energy of the
`droplet. Fig. 7(b) depicts the signal of droplets of diluted deter-
`gent (volume ratio of 12.5× 10
`−3) at a flow rate of 50 ␮L/h. The
`instability in the droplet shape and even satellite droplets can be
`clearly observed in the time signal.
`In order to investigate the impact of interfacial tension on the
`droplet shape, aqueous liquids was prepared with different vol-
`ume ratios between detergent and DI water. The higher the ratio,
`the lower is the interfacial tension between the aqueous liquid
`
`and the carrier liquid. Fig. 8 shows the effect of detergent/water
`ratio on the interfacial tension and subsequently on the radius
`of curvature at the two ends of the droplet. The droplets are
`formed at the same flow rate of 10 ␮L/h. From images recorded
`by the CCD camera we can clearly observe, that both sides of a
`pure-water droplet are convex, Fig. 6(a). This fact is represented
`well in the detected time signal shown in Fig. 8(a). The slight
`difference at the two edges is caused by the dynamic effect of a
`moving droplet.
`With a high interfacial tension in the case of pure water,
`the difference between the advancing and receding sides of
`the droplet is minimum. As shown in Fig. 8(b–d), this differ-
`ence increases with decreasing interfacial tension. The droplet
`transforms from an almost symmetric shape into a “bullet-like”
`shape as shown in Fig. 6(b). Fig. 9 depicts the corresponding
`time-differential signal ds/dt of the original signal s(t) shown in
`Fig. 8. In practice, the time-differential signal can be obtained
`using a differentiator circuit. The positive and negative peaks
`of the time-differential signal represent the maximum slopes at
`the receding edge and advancing edge, respectively. The time-
`differential signal can be used for evaluating the shape of each
`
`Fig. 7. Time signals of micro-droplets received by the APD (flow rate of the
`aqueous liquid is 50 ␮L/h): (a) pure DI water, (b) detergent diluted in DI water
`with a volume ratio of 12.5× 10
`−3.
`
`4
`
`

`
`N.-T. Nguyen et al. / Sensors and Actuators B 117 (2006) 431–436
`
`435
`
`Fig. 8. Time signal s(t) of sample droplet with different surfactant concentration
`(flow rate of the aqueous liquid: 10 ␮L/h).
`
`droplet edge. The gap between the positive peak and the nega-
`tive peak of the signals shown in Fig. 9 is a measure of the size
`of the droplet.
`
`4.2. Droplet size
`
`As mentioned above, the droplet size can be measured by
`the gap between the positive peak and the negative peak in
`the time-differential signal. From the analytical model, the size
`of the droplets is determined mainly by the interfacial tension
`and the flowrate. Eq. (4) shows that the droplet diameter is
`inversely proportional to the mean velocity of the carrier liquid
`(Dd ∝ 1/Uc), and proportional to the square root of the interfa-
`cial tension (Dd ∝ √
`σ). The interfacial tension is supposed to
`be inversely proportional to the detergent concentration. Thus
`droplet size will decrease with increasing detergent’s volume
`ratio.
`
`Fig. 9. The time-differential signal ds/dt of the signals shown in Fig. 8 (flow
`rate of the aqueous liquid: 10 ␮L/h).
`
`Fig. 10. Detected droplet size as function of (a) flow rate of the aqueous liquid
`and (b) volume ratio of detergent to pure water (solid lines are sixth order
`polynomial fitting functions).
`
`The results shown in Fig. 10 confirm the behavior of droplet
`size expected from the theory. Droplet size or the pulse width of
`the time signal decrease with increasing flow rate of the aqueous
`liquid, Fig. 10(a). Since the flow rate ratio between the aqueous
`liquid and the carrier liquid was kept constant at 1:2, the droplet
`size is inversely proportional to the flow rate or the mean velocity
`of the carrier flow. Increasing the volume ratio between detergent
`and water decreases its interfacial tension to the carrier liquid
`and allows the droplets to form at smaller diameters, Fig. 10(b).
`
`4.3. Formation frequency
`
`The droplet formation process results from the balance
`between the drag force and the interfacial tension force. If the
`drag force on the droplet is higher then the interfacial tension, the
`droplet is released. Thus on the one hand, a higher drag force or
`a higher flow rate of the carrier flow leads to a higher formation
`frequency. On the other hand, a higher surfactant concentration
`or a lower interfacial tension also lead to a higher formation
`frequency. Eq. (4) predicts that the droplet formation frequency
`is proportional to the forth power of the mean velocity of the
`carrier liquid (f ∝ U4
`c ), and inversely proportional to the 3/2-th
`power of the interfacial tension (f∝ 1/σ3/2). Fig. 11(a) shows the
`relation between the measured formation frequency and the flow
`rate of the aqueous liquid. The solid lines are fourth order poly-
`nomial fitting functions. Higher detergent’s volume ratio leads
`to lower interfacial tension, and consequently higher formation
`frequency, Fig. 11(b). We can also observe that the error bars are
`wider at higher flow rates and higher detergent’s volume ratios.
`
`5
`
`

`
`436
`
`N.-T. Nguyen et al. / Sensors and Actuators B 117 (2006) 431–436
`
`Acknowledgment
`
`This work was supported by the academic research fund of
`the Ministry of Education Singapore, contract number RG11/02.
`
`References
`
`[1] K. Handique, M.A. Burns, Mathematical modeling of drop mixing in a
`slit-type microchannel, J. Micromech. Microeng. 11 (2001) 548–552.
`[2] P. Paik, V.K. Pamula, R.B. Fair, Rapid droplet mixers for digital microflu-
`idic systems, Lab Chip 3 (2003) 253–259.
`[3] T. Taniguchi, T. Torii, T. Higuchi, Chemical reactions in microdroplets by
`electrostatic manipulation of droplets in liquid media, Lab Chip 2 (2002)
`19–23.
`[4] H. Song, M.R. Bringer, J.D. Tice, C.J. Gerdts, R.F. Ismagilov, Experi-
`mental test of scaling of mixing by chaotic advection in droplets moving
`through microfluidic channels, Appl. Phys. Lett. 83 (2003) 4664–4666.
`[5] T. Nisisako, T. Torii, T. Higuchi, Droplet formation in a microchannel
`network, Lab Chip 2 (2002) 24–26.
`[6] S. Okushima, T. Nisisako, T. Torii, T. Higuchi, Controlled production of
`monodisperse double emulsion by two-step droplet breakup in microflu-
`idic devices, Langmuir 20 (2004) 9905–9908.
`[7] A.S. Utada, E. Lorenceau, D.R. Link, P.D. Kaplan, H.A. Stone, D.A.
`Weitz, Monodisperse double emulsions generated from a microcapillary
`device, Science 308 (2005) 537–541.
`
`Biographies
`
`Nam-Trung Nguyen was born in Hanoi, Vietnam, in 1970. He received
`his Dip-Ing, Dr Ing and Dr Ing Habil degrees from Chemnitz University
`of Technology, Germany, in 1993, 1997 and 2004, respectively. In 1998 he
`worked as a postdoctoral research engineer in the Berkeley Sensor and Actu-
`ator Center (UC Berkeley, USA). Currently he is an associate professor with
`the School of Mechanical and Aerospace Engineering of the Nanyang Tech-
`nological University in Singapore. His research is focused on microfluidics
`and instrumentation for biomedical applications. He published a number of
`research papers on microfluidics. The second edition of his book “Funda-
`mentals and Applications of Microfluidics” co-authored with S. Wereley will
`be published in summer 2006.
`
`Sumantri Lassemono was born in Indonesia. He is currently competing his
`undergraduate study at the School of Mechanical and Aerospace Engineering
`of the Nanyang Technological University in Singapore.
`
`Franck Alexis Chollet received his electronics engineering degree from the
`ENSERB, Bordeaux, France in 1991, and his doctorate degree in Sciences
`Pour lIng´enieur from the Universit´e de Franche-Comte, Besanc¸on, France,
`in 1995 for his work on integrated optics devices for the CNET (France
`Telecom). He was then with LIMMS, a French Japanese laboratory, for two
`years at the IIS, University of Tokyo, Japan, as a JSPS post-doctoral fellow,
`where he developed optical MEMS. After a short stay at LPMO, Besanc¸on,
`France, working on micro-cooler and LIGA process, he became a research
`associate at the IMRE, Singapore, where he worked on MEMS optical sensors
`and switches. Since September 1999 he is associate professor in the school of
`Mechanical and Aerospace Engineering at Nanyang Technological University,
`Singapore, and is the Vice-Director of the MicroMachines Centre where he
`pursues his work on optical MEMS and micro/nano fabrication technology.
`
`Fig. 11. Droplet formation frequency as function of (a) flow rate of the aqueous
`liquid and (b) volume ratio between detergent and pure water (solid lines are
`fourth order polynomial fitting functions).
`
`In these two cases, the kinetic energy dominates over the sur-
`face energy of the droplets. Thus, the noise in the time signal is
`caused by the instability of the droplet shape and the formation
`of satellite droplets as shown earlier in Fig. 7.
`
`5. Conclusions
`
`This paper reports a microfluidic device for droplet forma-
`tion and detection. The device has a microchannel network to
`form the droplets. The droplets are detected optically by two
`optical fibers. A droplet passing by the detection point diffracts
`a part of the incoming laser light and can be detected by the
`fiber placed on the other side of the channel. The device has
`a potential for feedback control in a droplet-based microfluidic
`device. Currently, droplet size, shape, and formation frequency
`can be measured accurately. The device can detect droplets at
`very high formation frequencies, where observation with a CCD
`camera and a microscope is not possible. The system can pos-
`sibly detect the extent of mixing processes as well as chemical
`reactions inside droplets, which may be useful for adsorption
`analysis of biochemical samples.
`
`6