Colour-Specific Microfluidic Droplet Detection for Molecular Communication
Max Bartunik, Marco Fleischer, Werner Haselmayr, Jens Kirchner
CColour-Specific Microfluidic Droplet Detection for MolecularCommunication
Max Bartunik ∗ Marco Fleischer ∗ [email protected]@fau.deInstitute for Electronics Engineering,Friedrich-Alexander-UniversityErlangen-Nuernberg (FAU)Erlangen, Germany Werner Haselmayr [email protected] for CommunicationsEngineering and RF-Systems,Johannes Kepler University Linz (JKU)Linz, Austria
Jens Kirchner [email protected] for Electronics Engineering,Friedrich-Alexander-UniversityErlangen-Nuernberg (FAU)Erlangen, Germany
ABSTRACT
Droplet-based microfluidic systems are a promising platform forlab-on-a-chip (LoC) applications. These systems can also be used toenhance LoC applications with integrated droplet control informa-tion or for data transmission scenarios in the context of molecularcommunication. For both use-cases the detection and characteri-sation of droplets in small microfluidic channels is crucial. So far,only complex lab setups with restricted capabilities have been pre-sented as detection devices. We present a new low-cost and portabledroplet detector. The device is used to confidently distinguish be-tween individual droplets in a droplet-based microfluidic system.Using on-off keying a 16-bit sequence is successfully transmittedfor the first time with such a setup. Furthermore, the devices ca-pabilities to characterise droplets regarding colour and size aredemonstrated. Such an application of a spectral sensor in a mi-crofluidic system presents new possibilities, such as colour-codeddata transmission or analysis of droplet content.
CCS CONCEPTS • Hardware → Sensor devices and platforms . KEYWORDS
Droplet detection, microfluidic environment, molecular communi-cation, colour, optical sensor
ACM Reference Format:
Max Bartunik, Marco Fleischer, Werner Haselmayr, and Jens Kirchner. 2020.Colour-Specific Microfluidic Droplet Detection for Molecular Communica-tion. In
NanoCom ’20: ACM International Conference on Nanoscale Computingand Communication, September 23–25, 2020, College Park, MD.
ACM, NewYork, NY, USA, 6 pages. https://doi.org/10.1145/nnnnnnn.nnnnnnn ∗ Both authors contributed equally to this research.Permission to make digital or hard copies of all or part of this work for personal orclassroom use is granted without fee provided that copies are not made or distributedfor profit or commercial advantage and that copies bear this notice and the full citationon the first page. Copyrights for components of this work owned by others than ACMmust be honored. Abstracting with credit is permitted. To copy otherwise, or republish,to post on servers or to redistribute to lists, requires prior specific permission and/or afee. Request permissions from [email protected].
NanoCom ’20, September 23–25, 2020, College Park, MD © 2020 Association for Computing Machinery.ACM ISBN 978-x-xxxx-xxxx-x/YY/MM.https://doi.org/10.1145/nnnnnnn.nnnnnnn
Molecular communication uses biochemical signals or other in-formation carriers in the nanoscale to achieve data transmissionin use-cases where either wired or wireless connections are notfeasible. This approach has been experimentally investigated in pre-vious work on centimeter- and decimeter-scale (e.g. [1, 2, 7, 9, 18]).For smaller geometries, it can be implemented in a microfluidicsetup. To this end, we use droplets to transmit data in a microfluidicchannel. The theoretical principles of droplet-based communicationhave been presented in [5] and a first practical study was conductedin [11].To enable the practical realisation of droplet-based communi-cation the automated generation and detection of droplets is vital.Challenges are posed by the small size of the microfluidic channels,the requirement to not introduce obstacles into the channel andthe goal of a compact and portable system. A promising methodfor droplet generation has recently been introduced in [10]. In thiswork we aim to develop a simple, versatile and portable dropletdetection device.Droplets may be detected either by change of conductivity be-tween two electrodes [6, 12, 14] or with devices based on opticalsensors. As the use of electrodes results in a complicated fabricationprocedure and inherently necessitates changing the microfluidicchannel to accommodate the electrodes, implementation of an opti-cal sensor principle was chosen.Various optical sensors utilising cameras [4, 8, 10] or fluorescentagents [3, 16] were successfully applied. However, these approachesrequire a complex lab setup or sophisticated image processing. Theaim of this work is the development of a simple and portable dropletdetector. Hence, we initially extend the detection principle based ona light-emitting diode (LED) and a photodetector presented in [15]and [17] to enable droplet-based data transmission. In particular,we eliminate the need to feed optical fibres into the transmissionchannel, significantly simplifying the detector design and improv-ing its portability as well as facilitating an uncomplicated exchangeof the mounted microfluidic chip. This device is further used to, forthe first time, transmit a stream of 16 bit using droplets, which isan important step towards droplet-based data transmission.In addition to the simple droplet detection using a photodetector,we implement droplet colour detection using a spectral sensor. Thisenables a more precise analysis of droplets and allows for moresophisticated data transmission schemes. So far only a rudimentaryapproach to colour detection based on changes in transparency a r X i v : . [ c s . ET ] J un anoCom ’20, September 23–25, 2020, College Park, MD Max Bartunik, Marco Fleischer, Werner Haselmayr, and Jens Kirchner has been presented [17]. Hence, this is the first device allowingfor differentiation between a variety of colours in a droplet-basedmicrofluidic system.Finally, both sensor systems are simultaneously used to deter-mine crucial droplet parameters. In particular, we demonstratethe capability to measure a droplets size inside the transmissionchannel.We present our work in three sections. Section 2 describes theindividual components of the sensor devices design. In Section 3 thedevices capabilities regarding droplet detection, colour characterisa-tion, colour intensity variation and determination of droplet lengthare evaluated. We end the article in Section 4 with a conclusion. The constructed device consists of two separate optical sensors. Ananalog infrared photodiode to detect the presence of an individualwater droplet in an oil carrier medium and a digital 6-channelcolour detector to further characterise the droplet. Both sensorswere implemented on a common circuit board together with asuitable microcontroller. Furthermore, a measurement setup tomount microfluidic chips was constructed. The design allows asimple exchange of the microfluidic chip.
The infrared sensor consists of a light source (LED) that is focusedtowards the channel (submillimeter-scale) and a photodetector onthe opposite side. The result is a light path that gets interruptedwhen the optical absorption of the material (i.e. ink droplets) insidethe microfluidic channel increases. The wavelength of the lightsource was chosen to be in the infrared spectrum (940 nm) as thisis close to the maximally absorbed wavelength (approx. 970 nm) inwater [13].As the signal variation at the photodiode is very small and there-fore also the change in diode current, a transimpedance amplifier(TIA) was used to amplify the photodiode output in the range of0 V to 20 V. The TIA output is then offset corrected and furtheramplified using a differential amplifier with a tunable referencevoltage. The resulting signal output is adjusted to the range of 0 Vto 5 V and can be acquired with the analog-to-digital converter of amicrocontroller. Adequate overvoltage protection at the microcon-troller input was implemented. A simplified circuit diagram of theresulting analog setup is shown in Fig. 1.
To further characterise droplets, a spectral sensor (AS7262) fromams AG was implemented. The sensor has a very compact designand incorporates six photosensors for various wavelengths, cov-ering the range from 430 nm to 670 nm. Each photosensor has abandwidth of 40 nm. As the sensor has integrated digital data acqui-sition, no analog amplification, as in Fig. 1 for the infrared sensor,is required.The recorded spectral data can be retrieved via a standard I Cinterface with an appropriate microcontroller. Due to the timerequired to measure the spectral value and transmit the result withthe I C interface the sample rate is restricted to 20 Sa s −1 . A highersample rate, although possible, would result in reduced sensitivity. Figure 1: Circuit diagram for the infrared sensor. The cur-rent through the photodiode is amplified in two stages (TIAand differential amplifier). The reference voltage at the pos-itive input of the differential amplifier is tunable. V out is inthe range of to . For measurements in the microfluidic setup a white LED was placedopposite to the sensor, analogous to Section 2.1.
The circuitry for both sensors was implemented on a common cir-cuit board, together with a microcontroller (ATmega32U4) fromAtmel Corporation, a USB connector and power converters to sup-ply the required voltages for the amplification stages.The resulting device was constructed as a four layered printedcircuit board (PCB), which was designed using Altium Designer.The PCB was printed by Multi Leiterplatten GmbH in Germanyand assembled using the facilities of the Institute for ElectronicsEngineering at FAU, Germany.Fig. 2 shows the completed PCB from above. Components wereplaced on both sides of the circuit board with a size of 6.6 cm by5.6 cm.To mount the microfluidic chips for testing a custom-made hous-ing was constructed. This allows for a simple exchange of the usedchip and has an adjustable positioning system to accurately placethe microfluidic channel on the sensor. A beam across the top ofthe microfluidic chip acts as a mount for the two light sources.Fig. 3 shows the complete droplet detection device with a mi-crofluidic chip. Moreover, a microfluidic chip with a stream of inkeddroplets is shown in Fig. 4.
Various experiments were conducted to evaluate the individual ca-pabilities of the two implemented sensors. First, the infrared sensorwas used as a presence/absence detector for individual droplets.The transmission of a sample data sequence using on-off keyingwas achieved. Second, the colour sensors spectral sensitivity wasinvestigated using differently coloured droplets. Third, the responseof both sensors to various ink concentrations was examined. Finally, olour-Specific Microfluidic Droplet Detection for Molecular Communication NanoCom ’20, September 23–25, 2020, College Park, MD
Microcontroller (cid:63)
Colour Sensor (cid:54)
Infrared Sensor (cid:54)
Figure 2: Assembled PCB with integrated infrared and spec-tral sensors as well as a microcontroller. A serial connectionis provided via a USB port that is also used to power the de-vice. The attached cables connect to the separate LEDs.
LED-Beam (cid:45)
Chip Fixture (cid:54) (cid:54)
PCB-Housing (cid:54)
Figure 3: Droplet detection device with positioning systemfor the microfluidic chip. The light sources (i.e. infrared andwhite LED) are attached to the beam across the top of the mi-crofluidic chip. The PCB with the sensors is inside the box.The spring-loaded screws can be used to precisely place themicrofluidic channel above the sensors. the characterisation of droplet size and speed using both sensorswas achieved.The microfluidic chips are made of two overlapping polymethylmethacrylate (PMMA) layers, whereby channels are laser engravedinto one of the layers [11]. In all cases the chip was operated witha pressure controlled pump to inject individual droplets into aconstant background flow via a T-junction as proposed in [10]. Thisis achieved by increasing the pressure at the injection point relativeto a constant pressure in the transmission channel. The backgroundflow consists of regular household oil and the injected droplets ofink diluted with tap water. Channel (cid:54)
Channel Outlet (cid:63)
T-Junction (cid:63)
Injection Tubes (cid:63)(cid:63)
Figure 4: Microfluidic chip of PMMA with a straight trans-mission channel and a T-junction used for data transmis-sion. The blue ink is injected at the T-junction creating adroplet stream which can be detected by the proposed de-vice shown in Fig. 2. The channel boundaries were redrawnin the figure for better visibility.
To achieve data transmission, on-off-keying was implemented. Anindividual bit was transmitted every second (1 bit s −1 ), whereas thebit value ’1’ was coded as the injection of a droplet and the value ’0’as no injection. The random 16-bit sequence ’10111100 01011001’was used for transmission.To detect individual bits, the received signal was offset correctedto get a 0 V baseline and a fixed threshold of 0.2 V, significantlyhigher than the noise level, was applied. Symbol intervals with aknown duration were derived from the first threshold pass. A bitwas detected as ’1’ if the threshold was met for at least 30 % of thesymbol interval . To account for inaccurate symbol intervals due tovarying droplet speeds the symbol intervals were re-synchronisedat every detected rising edge.Fig. 5 shows the receive signal for the transmitted sample bitsequence. The detection threshold and the relevant symbol intervals,as derived from the the rising edges, are also shown. As can be seeneach transmitted bit was successfully received with the device andthe scheme described above. The spectral sensors main function is to determine the colour of adetected droplet. However, it can also be used as an extension tothe analog infrared sensor. Possible use-cases for colour detectionare more sophisticated data transmission schemes and the charac-terisation of droplet content (e.g. with a coloured reagent) in LoCapplications.To assess the colour-coded droplet detection the measured spec-tral signal for ink droplet of various colours was recorded. Fig. 6shows the normalised sensor values for the individual droplets. Onedroplet for each of the six different ink colours (violet, blue, green,yellow, orange and red) was transmitted. Significantly differentcolours can clearly be identified due to their spectral values (e.g.blue and violet) using the constructed sensor. Spectral values forrelated colours (e.g. yellow and orange) are very similar and re-quire further investigation to ensure distinct detection. An in-depth Please note that due to the good signal quality various threshold values lead to asuccessful data transmission. anoCom ’20, September 23–25, 2020, College Park, MD Max Bartunik, Marco Fleischer, Werner Haselmayr, and Jens Kirchner . . . . . V o l t a g e [ V ] Figure 5: Receive signal for the transmitted bit sequence’10111100 01011001’ (shown in blue) together with the signalthreshold at (shown in red) and the individual symbolintervals (dashed lines). A symbol was detected as ’1’ if thethreshold was met for at least
30 % of a symbol interval. All16 bits of the sample sequence were decoded correctly. analysis of colour sensitivity may be achieved by investigating thesensor data for gradual transitions between colours with a series ofdilutions.Fig. 7 shows the spectral values that were recorded by the coloursensor for the transmission of the 16-bit sequence introduced inSec. 3.1. The data points are set at the center wavelength of the sixphotosensors.Each bit can clearly be identified with an equal colour distribu-tion equivalent to a red droplet in Fig. 6. The sensor can thereforeconsistently characterise droplet colour throughout a transmissionsequence, allowing for colour coded data transmission by differen-tiation between droplet spectra (as shown in Fig. 6).
In order to assess the sensitivity of the sensors, a series of dilutionswith varying ink concentrations was performed. Furthermore, theinfrared sensors operation based on absorption in water (indepen-dent of colouring) is demonstrated with clear water droplets.Water was diluted with blue ink in concentrations ranging from0 % to 25 % in steps of 5 %. For each measurement a droplet of 5 mmlength was generated.As can be seen in Fig. 8, the measured signal amplitude of theinfrared sensor is constant for various ink concentrations. The cal-culated mean signal is 0.73 V with a maximal deviation of 0.13 Vfrom the average value for the individual concentrations. As ex-pected, the sensors receive signal is independent of colouring as itrelies on the absorption in water. 450 500 550 600 65000 . . . .
81 Wavelength [nm] N o r m a l i s e d I n t e n s i t y Red Blue GreenViolet Orange Yellow
Figure 6: Spectral comparison of various droplet colours us-ing the constructed colour sensor. The data is normalised tothe spectral profile recorded without droplets. Colours canclearly be differentiated. . . . .
81 Time [s] N o r m a l i s e d I n t e n s i t y
450 nm 500 nm 550 nm570 nm 600 nm 650 nm
Figure 7: Recorded spectral values for the sample transmis-sion using red ink in Fig. 5. Each ’1’ bit can be seen as anindividual notch. The colour values are consistent for thewhole transmission.
Fig. 9 shows the measured signal for different ink concentrationsusing the colour sensor. As the ink concentration increases a moredistinct colour distribution can be observed. The light absorptionincreases with the used amount of ink.The measurements show, that the sensor system can be usedto distinguish between different colour concentrations (i.e. colour olour-Specific Microfluidic Droplet Detection for Molecular Communication NanoCom ’20, September 23–25, 2020, College Park, MD . . V o l t a g e [ V ] Figure 8: Measurements of water droplets with varying con-centrations of blue ink in the range of to
25 % usingthe infrared sensor (shown in blue). The mean value foreach droplet (shown in green) as well as the overall average(shown in red) was calculated. A similar signal can be ob-served in all cases, showing the specificity to absorption inwater independent of colouring intensities). Furthermore, the infrared sensor can be used for pres-ence/absence detection of water droplets, independent of colouring.In future applications this may be used to differentiate betweendifferent droplet substances.
The sensor can also be used to measure the size of an individualdroplet by observing the infrared and colour sensor values on atime scale. The observed duration of a droplet is proportional tothe travel speed inside the microfluidic channel and the volume ofthe droplet. Considering the speed of the droplet in the channel v chan the droplet length can be calculated from the observed dropletduration t drop as l drop = v chan t drop = ∆ d ∆ t t drop (1)where ∆ d = .
35 mm is the distance (fixed on the PCB) and ∆ t thetime delay of the received signal edge between the two sensors.Either of the sensors can be used to measure the droplet duration. Inthis case, the colour sensor was chosen and t drop was set to the timespan for which the recorded signal surpassed half of the maximalamplitude (full width at half maximum).Droplets of varying length (1 mm to 5 mm) were generated byadjusting the duration of a rectangular pressure pulse from 1 s to 0 400 . . . . N o r m a l i s e d I n t e n s i t y Figure 9: Recorded signal of the spectral sensor for long droplets with varying concentrations of blue ink. Theobserved sensor intensity decreases, as light absorption in-creases, proportional to the amount of ink in the droplet.
A low-cost and portable sensor for droplets in microfluidic sys-tems was constructed. The device was successfully used for pres-ence/absence detection and to characterise droplet colour and size.Furthermore, a distinction between different ink concentrationswas achieved. It therefore facilitates LoC use cases that require so-phisticated data transmission schemes or colour reagent detection.With the current setup a bit rate of approximately 1 bit s −1 wasachieved. Limiting factors can be found both in parameters ofdroplet generation (precision of volume and timing) and detec-tion (local resolution and sample rate). The precision of dropletgeneration may be improved with other pump systems, such as amicropump or a peristaltic pump. Furthermore, implementing a anoCom ’20, September 23–25, 2020, College Park, MD Max Bartunik, Marco Fleischer, Werner Haselmayr, and Jens Kirchner M e a s u r e d L e n g t h [ mm ] Figure 10: Measurements to evaluate precision of dropletsize values. The average determined droplet size and stan-dard deviation for droplets ranging from to isshown. feedback loop to control a pressure pump, based on the retrievedsensor data, may improve droplet generation consistency. If neces-sary, the sample rate of the sensor device (currently 20 Sa s −1 ) couldbe increased by use of an array with multiple sensors in parallel. Byaddressing these limitations the bit rate could be increased furtherin an optimised setup.In future work various functions of the sensor device may be ex-tended. Specifically, for the data transmission scenario, the bit ratecould be increased by combining coding schemes such as codingof droplet position, colour and size. Furthermore, the developmentof a standardised interface for microfluidic chips could simplifyinteroperability for future devices. Such a standard could includededicated positions for in- and output connectors, sensing regionsand a scalable set of microfluidic chip dimensions. Finally, the sen-sors capabilities regarding sensitivity and more complex codingschemes will be investigated further. ACKNOWLEDGMENTS
The authors would like to express their sincere gratitude to DominikLehner for his support with the microfluidic setup.
REFERENCES [1] Max Bartunik, Maximilian Lübke, Harald Unterweger, Christoph Alexiou, Sebas-tian Meyer, Doaa Ahmed, Georg Fischer, Wayan Wicke, Vahid Jamali, RobertSchober, and Jens Kirchner. 2019. Novel Receiver for Superparamagnetic IronOxide Nanoparticles in a Molecular Communication Setting. In
Proc. Sixth Annu.ACM Int. Conf. Nanoscale Comput. Commun. - NANOCOM '19 . ACM Press, Dublin,1–6. https://doi.org/10.1145/3345312.3345483[2] Max Bartunik, Harald Unterweger, Christoph Alexiou, Robert Schober, Maximil-ian Lübke, Georg Fischer, and Jens Kirchner. 2020. Comparative Evaluation ofa New Sensor for Superparamagnetic Iron Oxide Nanoparticles in a MolecularCommunication Setting. In . ACM Press. toappear. [3] Binyu Chen, Xiaoming Han, Zhen Su, and Quanjun Liu). 2018. Fluorescencedetection system for microfluidic droplets. In
Sixth AIP Int. Conf. Comput.-AidedDesign, Manuf., Model. Simul. (CDMMS 2018) , Vol. 1967. AIP Publishing, Nanjing,1–5. https://doi.org/10.1063/1.5039044[4] Bas de Heij, Bart van der Schoot, Hu Bo, Joseph Hess, and Nicolaas F. de Rooij.2000. Characterisation of a fL droplet generator for inhalation drug therapy.
Sensors and Actuators A: Physical
85, 1–3 (Aug. 2000), 430–434. https://doi.org/10.1016/S0924-4247(00)00400-3[5] Elena De Leo, Lidia Donvito, Laura Galluccio, Alfio Lombardo, Giacomo Morabito,and Laura M. Zanoli. 2013. Communications and Switching in MicrofluidicSystems: Pure Hydrodynamic Control for Networking Labs-on-a-Chip.
IEEETrans. Commun.
61, 11 (November 2013), 4663–4677. https://doi.org/10.1109/TCOMM.2013.100813.130332[6] Caglar Elbuken, Tomasz Glawdel, Danny Chan, and Carolyn L. Ren. 2011. Detec-tion of microdroplet size and speed using capacitive sensors.
Sensors and ActuatorsA: Physical
PLoS ONE
8, 12 (Dec.2013), 1887–1919. https://doi.org/10.1371/journal.pone.0082935[8] Alexander Golberg, Gregory Linshiz, Ilia Kravets, Nina Stawski, Nathan J. Hillson,Martin L. Yarmush, Robert S. Marks, and Tania Konry. 2014. Cloud-Enabled Mi-croscopy and Droplet Microfluidic Platform for Specific Detection o fEscherichiacoli in Water.
PLoS ONE
9, 1 (Jan. 2014), 1–9. https://doi.org/10.1371/journal.pone.0086341[9] Laura Grebenstein, Jens Kirchner, Renata Stavracakis Peixoto, Wiebke Zimmer-mann, Florian Irnstorfer, Wayan Wicke, Arman Ahmadzadeh, Vahid Jamali, GeorgFischer, Robert Weigel, Andreas Burkovski, and Robert Schober. 2019. BiologicalOptical-to-Chemical Signal Conversion Interface: A Small-Scale Modulator forMolecular Communications.
IEEE Trans. Nanobiosci.
18, 1 (Jan. 2019), 31–42.https://doi.org/10.1109/tnb.2018.2870910[10] Medina HamidoviÄĞ, Uli Marta, Helen Bridle, Damir HamidoviÄĞ, Gerold Fink,Robert Wille, Andreas Springer, and Werner Haselmayr. 2020. Off-chip ControlledDroplet-On-Demand Method for Precise Sample Handling.
ACS Omega (2020).to appear.[11] Medina HamidoviÄĞ, Uli Marta, Gerold Fink, Robert Wille, Andreas Springer, andWerner Haselmayr. 2019. Information Encoding in Droplet-Based MicrofluidicSystems: First Practical Study. In
Proc. Sixth Annu. ACM Int. Conf. NanoscaleComput. Commun. - NANOCOM '19 . Association for Computing Machinery,Dublin, 1–6. https://doi.org/10.1145/3345312.3345482[12] Pelin Kubra Isgor, Merve Marcali, Mert Keser, and Caglar Elbuken. 2015. Mi-crofluidic droplet content detection using integrated capacitive sensors.
Sensorsand Actuators B: Chemical
210 (April 2015), 669–675. https://doi.org/10.1016/j.snb.2015.01.018[13] Jerry Workman Jr. and Lois Weyer. 2007.
Practical Guide to Interpretive Near-Infrared Spectroscopy . CRC Press, USA.[14] Xize Niu, Mengying Zhang, Suili Peng, Weijia Wen, and Ping Sheng. 2007. Real-time detection, control, and sorting of microfluidic droplets.
Biomicrofluidics
1, 4(Sept. 2007), 1–12. https://doi.org/10.1063/1.2795392[15] Vijay Srinivasan, Vamsee K. Pamula, and Richard B. Fair. 2004. Droplet-basedmicrofluidic lab-on-a-chip for glucose detection.
Analytica Chimica Acta
J. mechanics medicine biology
11, 1 (March 2011), 1–14. https://doi.org/10.1142/S0219519410003617[17] Varun Trivedi, Ankur Doshi, G. K. Kurup, E. Ereifej, and P. J. VandevordbandAmar S. Basu. 2010. A modular approach for the generation, storage, mixing,and detection of droplet libraries for high throughput screening.
Lab on a Chip