a digital microfluidic device for antibiotic testing in milk
TRANSCRIPT
A Digital Microfluidic Device for Antibiotic Testing in Milk
by
Rahul Eswar
A Thesis
presented to
The University of Guelph
In partial fulfilment of requirements for the degree of
Master of Applied Science
in
Engineering
Guelph, Ontario, Canada
© Rahul Eswar, January, 2021
ABSTRACT
A DIGITAL MICROFLUIDIC DEVICE FOR ANTIBIOTIC TESTING IN MILK
Rahul Eswar
University of Guelph, 2021
Advisor:
Dr. Christopher Collier
Co-Advisor:
Dr. Ashutosh Singh
Worldwide, milk is tested to meet limits for antibiotic residue content which can
invoke allergic reaction and antibiotic resistance in human consumers. Dairy farmers
are responsible for collecting milk and shipping it to off-site laboratories for testing. Non-
compliant milk results in penalties including fines and loss of licensure for offending
dairy farmers. Modern on-site testers require long response times and manual
preparation procedures which are unavailable during milk collection. An appealing
technology to address this issue is digital microfluidics which handles sample in discrete
droplets to leverage reconfigurability and automation. This thesis demonstrates the
fundamental functionalities required by an on-site digital microfluidic dairy tester. Design
of an integrated digital microfluidic device for milk droplet actuation is detailed. A label-
free method for antibiotic detection in milk droplets is demonstrated using lock-in
amplified fluorescence spectroscopy. This work provides the foundation for
development of a full on-site digital microfluidic dairy testing solution.
iii
ACKNOWLEDGEMENTS
There are a number of individuals I must thank as this work would not have been
possible without their collaboration.
I want to first extend the utmost gratitude to my advisor, Dr. Christopher Collier.
His guidance, support and confidence in me have been invaluable to my growth and to
the completion of this work. I feel very fortunate to have worked as part of his team.
I want to also thank my co-advisor, Dr. Ashutosh Singh, for providing access to
his Food Research Lab and always encouraging my work.
For their willingness to serve on my advisory committee, I am grateful to Dr.
Christopher Collier, Dr. Ashutosh Singh and Dr. Abdallah Elsayed. I greatly appreciate
their time and commitment to oversee my progress.
Many thanks go to Larry Wood and Norwell Dairy Systems for providing the
inspiration for this work. This project is a direct product of their support.
For laying the foundation on which this work is built, I want to thank Rick Bosma.
His advice and technical assistance has been instrumental in accomplishing this work.
I am very grateful to David Wright and Jacqueline Fountain for their technical and
administrative support of my work. I feel fortunate to have completed this work at the
University of Guelph due in large part to exceptional staff like them.
iv
Special thanks go to Harrison Brodie, Jasen Devasagayam, Iasmin de Franca
Albuquerque, Michelle Del Rosso and Sydney Lepard for their collaboration and advice.
I would like to thank all my colleagues in the Applied Optics and Microsystems
Laboratory for their advice and friendships throughout this undertaking.
For their technical support, I want to thank Saipriya Ramalingam and Benjamin
Dyer. Their ever present willingness to lend their expertise has been critical to the
completion of this work.
Lastly, thank you to my father, mother and brother, Eswar, Rajani and Aditya.
They have inspired and believed in me always.
v
TABLE OF CONTENTS
Abstract ............................................................................................................................ii
Acknowledgements ......................................................................................................... iii
Table of Contents ............................................................................................................ v
List of Figures ................................................................................................................ viii
1 Introduction .............................................................................................................. 1
1.1 Background ........................................................................................................ 1
1.2 Thesis Scope ..................................................................................................... 3
2 Literature Review ..................................................................................................... 4
2.1 Introduction ........................................................................................................ 4
2.2 Regulated Antibiotics in Dairy ............................................................................ 4
2.3 Laboratory Detection Methods ........................................................................... 5
2.3.1 High Pressure Liquid Chromatography ........................................................ 5
2.3.2 Immunoassays ............................................................................................ 7
2.4 On-site Detection Methods ................................................................................. 9
2.4.1 Lateral Flow Assay ...................................................................................... 9
2.4.2 Prepared Incubated Test ........................................................................... 11
2.4.3 Microchip Capillary Electrophoresis ........................................................... 12
2.5 Actuation on Digital Microfluidic Devices .......................................................... 15
2.5.1 General Digital Microfluidic Design ............................................................ 15
2.5.2 Biofouling ................................................................................................... 16
2.5.3 Filler Medium ............................................................................................. 17
2.6 Detection on Digital Microfluidic Devices ......................................................... 18
vi
2.6.1 Fluorescence Spectroscopy ...................................................................... 18
2.6.2 Surface Plasmon Resonance .................................................................... 20
2.6.3 Impedance Spectroscopy .......................................................................... 21
2.6.4 Electrochemical Detection ......................................................................... 21
3 An Integrated Digital Microfluidic Device for Dairy Applications ............................. 23
3.1 Chapter Introduction ......................................................................................... 23
3.2 Background ...................................................................................................... 24
3.3 Electrowetting Actuation ................................................................................... 28
3.4 General DMF Design ....................................................................................... 30
3.5 DMF Prototype Design ..................................................................................... 32
3.6 Closed Configuration Experiments ................................................................... 34
3.7 Open Configuration Experiments ..................................................................... 39
3.8 Methods ........................................................................................................... 44
3.8.1 Reagents ................................................................................................... 44
3.8.2 Digital Microfluidic Device Printed Circuit Board ........................................ 44
3.8.3 Dielectric Coating ...................................................................................... 45
3.8.4 Statistical Analysis ..................................................................................... 45
3.9 Chapter Conclusion .......................................................................................... 46
4 Label-free Antibiotic Detection in Milk Droplets ...................................................... 47
4.1 Chapter Introduction ......................................................................................... 47
4.2 Lock-In Amplification ........................................................................................ 51
4.3 Experimental Setup .......................................................................................... 53
4.4 Rhodamine Pilot Experiment ............................................................................ 55
4.5 Antibiotic Detection .......................................................................................... 57
vii
4.6 Effect of Droplet Volume .................................................................................. 59
4.7 Chapter Conclusion .......................................................................................... 60
5 Conclusions ........................................................................................................... 62
5.1 Summary of Contributions ................................................................................ 62
5.2 Future Works .................................................................................................... 65
References .................................................................................................................... 66
viii
LIST OF FIGURES
Figure 1: Cross-sectional views of a DMF device is presented for (a) a closed system configuration, and (b) an open system configuration. A closed DMF system configuration constitutes a continuous upper electrode tied to ground and is coated in a hydrophobic material. The open DMF system configuration system is without the top electrode. ...................................................................................................................... 30
Figure 2: Hardware design of an integrated DMF device is presented. (a) The block diagram of the DMF device comprises of three hardware components, being: (1) the PCB housing firmware and electrodes, (2) personal computer running the control application, (3) a high voltage supply. (b) The modular PCB design using physically separated boards for control logic and electrodes connected by ribbon cables. (c) Photograph of the fully assembled DMF device. ........................................................... 33
Figure 3: Graphical user interface of the DMF device is presented featuring real-time control of 60 electrodes. Buttons and switches provide the capacity to automate recorded sequences and enable AC-like PFM actuation............................................... 34
Figure 4: Droplet velocity versus voltage in a closed configuration is presented for both (a) PDMS and (b) parafilm dielectric layers using water. The mean velocity (circle) and corresponding standard deviation (vertical bar) of n = 3 repeated measures are shown as a function of DC voltage. The closed microfluidic system is configured having 240 micron spacing between top and bottom planes for a fixed droplet volume of 4.2 µL. .. 37
Figure 5: Droplet velocity versus actuation frequency in a closed configuration is presented for both (a) PDMS and (b) parafilm dielectric layers using water. The mean velocity (circle) and corresponding standard deviation (vertical bar) of n = 3 repeated measures are shown as a function of frequency. The closed microfluidic system is configured having 240 micron spacing between top and bottom planes for a fixed droplet volume of 4.2 µL and a 280 V actuation potential. ............................................ 38
Figure 6: Droplet velocity versus voltage in an open configuration is presented, as a comparison between milk (red) and water (blue), for both (a) PDMS and (b) parafilm dielectric layers. The mean velocity (circle/square) and corresponding standard deviation (vertical bar) of n = 3 repeated measures are shown as a function of DC voltage. The open microfluidic system performance is characterised for a fixed droplet volume of 9.0 µL. ........................................................................................................... 40
Figure 7: Milk droplet actuation in open configuration DMF device is presented with parafilm dielectric layer and silicone oil. Electrode pitch is 2.75 mm, droplet volume is 9.0 µL, actuation voltage is 280 V. ................................................................................ 42
Figure 8: Droplet velocity versus actuation frequency in an open configuration is presented, as a comparison between milk (red) and water (blue), for both (a) PDMS and
ix
(b) parafilm dielectric layers. The mean velocity (circle/square) and corresponding standard deviation (vertical bar) of n = 3 repeated measures are shown as a function of frequency. The open microfluidic system performance is characterised for a fixed droplet volume of 9.0 µL and a 280 V actuation potential. ............................................ 43
Figure 9: Schematic of the LIA fluorescence spectroscopy experiment. An LED, sample droplet, optical filter and photodiode are aligned in the vertical plane. The LED is driven by square wave from a function generator. TTL output from the function generator is given as reference signal to the LIA. Output current from the photodiode is the raw input to the LIA. ...................................................................................................................... 55
Figure 10: Lock-in amplified photocurrent against rhodamine B concentration in 5.0 µL aqueous droplets. The mean photocurrent and corresponding standard deviation of n = 3 repeated measurements are shown as a function of rhodamine B concentration. ..... 56
Figure 11: Lock-in amplified photocurrent against ciprofloxacin concentration in 5.0 µL milk droplets. The mean photocurrent and corresponding standard deviation of n = 5 repeated measurements are shown as a function of ciprofloxacin concentration. ........ 58
Figure 12: Lock-in amplified photocurrent against ciprofloxacin concentration in milk droplets of volumes: (a) 7 µL, (b) 9 µL, (c) 11 µL and (d) 13 µL. The mean photocurrents and corresponding standard deviations of n = 3 repeated measurements are shown as a function of ciprofloxacin concentration. ...................................................................... 60
1
1 Introduction
This thesis demonstrates the fundamental functionalities of actuation and detection
required to produce a digital (droplet-based) microfluidic device for dairy testing.
This introductory chapter details the motivation for developing a digital microfluidic
dairy tester and states the scope of this work.
1.1 Background
In the interest of public health, the concentration of antibiotic residues in
consumable milk is strictly regulated [1]. In Canada, having antibiotic contaminated milk
arrive at a dairy processing plant yields severe consequences for the offending
producer [2]. The milk arrives at processing plants in common loads transporting
several bulk tanks from multiple producers [3]. If any bulk tank fails antibiotic screening,
the full volume is dumped and the offending producer bears the full penalty [4].
Moreover, a producer whose milk tests positive twice within a 12-month period is
indefinitely suspended from the milk market [4]. In the 12-month period ending October
2019, 30 penalties were assessed to dairy producers in Ontario due to antibiotic
contamination resulting in millions of dollars in economic loss [3].
Dairy farmers face an urgent need for development of an on-site sensor capable of
measuring antibiotic concentration within the approximate 5 minutes that each cow is
milked. The common practice of testing the milk after collection in a bulk tank is
inadequate because a single cow’s contaminated milk is sufficient to render the full tank
unusable [5]. Current commercially available on-site tests such as the Delvotest T (DSM
2
Food Specialties, Netherlands) and Charm ROSA test (Charm Sciences Incorporated,
United States) are unsuitable for use during the milk collection process due to long test
times [6], [7] and manual procedures [8]–[10].
A potential solution to on-site antibiotic testing is offered by microfluidics.
Microfluidic systems represent the miniaturization of analytical chemistry processes
onto a device with micron-scale features. Miniaturization of chemical processes yields
advantages including portability, minimal reagent consumption and fast response time
[11]–[13].
A subset of microfluidics known as digital microfluidics (DMF) specializes in highly
scalable, automatable and parallel operations [14], [15]. A DMF chip uses a grid of small
electrodes to handle discrete droplets of sample [16]–[18]. Using this technology,
multiple chemical analyses processes can be automated and reconfigured on a single
device. Thus, the capability for automation and versatility of DMF devices makes them
an attractive potential solution to on-site dairy testing.
The application of digital microfluidics to dairy testing is immediately faced with two
fundamental challenges: (1) biofouling and (2) antibiotic detection in droplets. Biofouling
refers to the adherence of fat and protein molecules to hydrophobic surfaces on the
device [19]. Biofouling causes irreversible damage to the device surface and is a
significant issue in microfluidics applications concerning biological samples [14], [20],
[21]. To handle milk droplets along with their fat and protein constituents, a DMF
solution to on-site dairy testing must overcome biofouling. The second fundamental
3
challenge for a DMF dairy tester is to detect antibiotic concentrations in milk droplets.
By comparison to traditional microfluidics, DMF is very limited in its ability to separate a
droplet into constituent particles. Thus, a method must be developed to detect antibiotic
contaminants within the bulk droplet. The research presented in this thesis addresses
these two fundamental challenges in actuation and detection to enable future
development of an integrated DMF solution to on-site dairy testing.
1.2 Thesis Scope
The contents of this thesis are described here. Chapter 2 will present a literature
review on the current techniques for antibiotic detection used in off-site laboratories and
on-site at the dairy farm. This chapter also provides background about droplet actuation
on digital microfluidic devices, and will include a review of relevant detection techniques
which have been implemented on digital microfluidic devices.
Chapter 3 presents a fully integrated digital microfluidic device which
demonstrates milk droplet actuation. The design considerations to overcome biofouling
and provide reconfigurability are reported.
Chapter 4 demonstrates a label-free method using fluorescence spectroscopy and
lock-in detection for antibiotic sensing in droplets. The ability to detect antibiotic
concentration in the droplet without need for separation from the bulk medium is well-
suited to digital microfluidic devices.
Chapter 5 reviews the main contributions of this thesis. It also speculates on
important future research.
4
2 Literature Review
2.1 Introduction
This chapter is a literature review which will begin by detailing common antibiotic
contaminants regulated in the dairy industry. The methods used to detect these
antibiotics in the lab and on-site will be discussed. Specifically, the advantages and
limitations of traditional microfluidic dairy testers will be discussed. DMF technology is
then detailed with attention to droplet actuation and compatible detection methods.
2.2 Regulated Antibiotics in Dairy
The use of antibiotics to treat livestock has been practiced for decades to promote
health and growth of farm animals [5], [22]. There is strong evidence that the quantity of
antimicrobials used worldwide to treat farm animals is up to four times greater than the
quantity used in human treatment [23], [24]. When these antibiotics are used in food
producing animals, such as cattle, measures must be taken to prevent residues from
appearing in food products.
The presence of antibiotic contaminants in food can have adverse health effects
on human consumers including chemical poisoning, allergic reactions and development
of antimicrobial resistance [25], [26]. It is for this reason that the concentration of
antibiotic residues allowed in consumable milk is regulated in many countries around
the world [5], [27], [28]. In Canada, there exist 96 maximum residue limits (MRLs)
5
enforced in the animal food products industry; 36 of these MRLs pertain to milk from
cattle [29].
Antibiotics from four families known as tetracyclines, sulfonamides, β-lactams and
fluoroquinolones are often used in combination for treatment of disease in dairy cows
[24]. The ciprofloxacin antibiotic used in Chapter 4 of this thesis is from the
fluoroquinolones family.
2.3 Laboratory Detection Methods
Since the first reported detection of antibiotics in milk in 1965 [22], there has been
exponential growth in the number of publications related to antibiotic detection in milk
each decade [5]. High pressure liquid chromatography (HPLC) and immunoassays
account for the majority of published research in this field [5]; both methods will be
detailed in this section.
2.3.1 High Pressure Liquid Chromatography
High pressure liquid chromatography (HPLC) is a method of molecular
separation. This technique uses a thin column which is filled with some absorbent
material referred to as the stationary phase [30]–[32]. Outside the column, the sample to
be detected is mixed into a liquid solvent known as the mobile phase [30]–[32]. The
substances used as the stationary and mobile phases are selected with consideration to
chemical properties of the target analyte(s) [30]–[34]. The liquid solution is then pushed
through the thin column under high pressure typically generated by a pump [33]–[36].
6
The constituents of the mobile phase will pass through the column at different speeds
depending on their interaction with the stationary phase. This is to say, the greater a
particle’s attraction to the stationary phase, the longer it will take to pass through the
column. Hence, the HPLC column acts as a filter to separate the sample.
It is important to note that HPLC only accomplishes separation. This system is
coupled with some detection method to read concentration of the separated analyte.
Methods including UV-Vis fluorescence [34], [36]–[40] and mass spectrometry [31]–[33],
[35], [41]–[43] have been demonstrated for HPLC coupled antibiotic detection. Using
either of these methods with HPLC separation will allow a time-resolved signal to be
generated. If separation is correctly achieved, signal peaks corresponding to
constituents of the sample will appear after a repeatable duration of time [31]–[40].
Among published works on antibiotic detection in milk, HPLC has been the most
frequently used method [5]. Using HPLC, studies have achieved antibiotic limits of
detection meeting or exceeding regulatory limits [31]–[40].
The limitations of HPLC with respect to on-site antibiotic testing are the need for
bulky lab equipment, extensive sample preparation and the relatively slow response
time. Even under high pressures up to 50 MPa, the fastest antibiotic tests with HPLC
take about 10 minutes [33]. This excludes time needed for centrifuging the sample and
mixing analytes into the mobile phase solution. Hence, HPLC is the golden standard of
antibiotic testing in the lab but has inherent limitations which do not make it adaptable
for on-site testing.
7
2.3.2 Immunoassays
The immunological approach to antibiotic detection takes advantage of the highly
selective binding that occurs between antigens and antibodies produced by animal
immune systems [44]. In animal immune systems, the antigen is a compound which
incites an immune response [44]. An antibody is a protein produced by the immune
system to combat a specific antigen [44]. Likewise, an antibiotic is a synthetically made
compound to combat an antigen [44], [45]. An immunoassay for antibiotic sensing uses
the corresponding antigen to label the target antibiotic [44], [45].
An immunoassay is carried out in a well plate. To prepare the well plate, the well
surfaces must first be coated with the antigen or antibiotic that will bind to the target
analyte. For antibiotic testing, the well plates are typically coated with the required
antigen [46], [47]. The antigen free sites on the well-plate are coated with a blocking
protein [47]–[49].
To allow for optical detection after the immunological binding, a fluorescent or
coloured molecule is also conjugated with the antigen or to a secondary antibody [10],
[44], [45], [47]. Commonly used fluorescent labels for antibiotic detection include
horseradish peroxidase, gold nanoparticles, and in more recent work, quantum dots
[10], [46], [47]. Quantum dots are synthetic semiconductor nanoparticles that can be
manufactured to fluoresce at different wavelengths when stimulated with UV light. The
versatile optical behavior of quantum dots makes them well suited to multiple antibiotic
8
detecting immunoassays where each antigen can be conjugated with its own coloured
quantum dot [10], [47].
When the well plate surfaces have been prepared, the sample solutions are
dispensed into the well plates. Typically, a 96-well plate is used and immunoassay tests
can be run in duplicate or triplicate [49]. The sample solutions must often be incubated
in the wells at constrained temperature conditions [6], [7], [44], [50].
The completed immunoassay is typically washed with a buffer solution which
removes any unbound antigens from the surface [44], [45]. To read the concentration of
antibiotics, each micro-well in the well plate is interrogated by fluorescence or
absorbance spectroscopy [10], [47], [51], [52]. Comparing to blank samples and known
concentrations included in the immunoassay, the antibiotic concentrations are
determined.
Immunoassays show excellent selectivity and versatility for antibiotic detection.
One study by Song et al. used quantum dot fluorescent labels to achieve simultaneous
detection of three antibiotic families in milk samples and achieved a 5 ppb limit of
detection [47]. However, the immunoassay took a 1 hour incubation time in addition to
the preparation time. Another work showed a rapid immunoassay coupled with
chromatography separation to get concentrations of 27 antibiotics in approximately 20
minutes [10]. In summary, the immunoassay is a scalable and accurate analysis method
but its dependence on label molecules and extensive preparation process make it
impractical to implement outside a traditional laboratory.
9
2.4 On-site Detection Methods
This section details methods for dairy testing which have demonstrated the
portability to be used on-site. The limitations in automation, scalability, response time,
and robustness which prevent these methods from being used during milk collection are
discussed.
2.4.1 Lateral Flow Assay
The lateral flow assay is a type of immunoassay analysis which uses capillary
forces to pull the sample fluid across a test strip [53]. At specified lengths, the test strip
is lined with antigens which will selectively bind to their target antibody [9], [50], [51],
[53]–[56]. Thus, the target antibody binds to the labelled antigen and the newly formed
conjugate is carried the length of the strip [9], [50], [51], [53]–[56]. By measuring the
intensity of colour produced at the detection end of the strip, the concentration of the
conjugate and therefore the antibiotic can be quantified [9], [50], [51], [53]–[56].
Research studies have shown the capability of lateral flow assay sensors to
detect up to five antibiotics on a single strip [54]. This is accomplished by creating
multiple reaction zones on the strip where the surface of each zone is functionalized
with the complementary antigen for the targeted antibiotic [54]. The antigen-antibody
conjugates which form continue down the chip and travel through multiple reaction
zones which selectively immobilize the conjugates [9], [51], [53]–[55]. The final result on
the strip is multiple lines appearing near the detection end; the colour intensity of each
10
line can be read by a spectrometer to identify concentration of the specified antibiotic
[9], [51], [53]–[55].
Lateral flow assays have demonstrated the ability to detect antibiotics from the
four major antibiotic families of beta-lactams, tetracyclines, sulfonamides and
fluoroquinolones [51], [56]. Furthermore, limits of detection on the scale of 0.1 – 10 ppb
have been achieved [51], [53], [54], [56].
The fastest lateral flow assay test time reported in literature was found to be five
minutes [53]. However, this does not account for the time the milk sample had to be
centrifuged prior to testing or the time for optical testing required after the assay.
The lateral flow assay has been used in commercial products for on-site
antibiotic testing. Examples of such products are the Charm ROSA Test and BetaStar
Plus Test (Neogen, United States). These tests have been studied to perform with
strong specificity and accuracy to detect beta-lactam and tetracyclines below their
MRLs [9], [55]. Moreover, some of the Charm ROSA Tests show a relatively fast
response time of 8 minutes [9].
Although lateral flow assays show strong portability, sensitivity and specificity,
there are limitations in their design that prevent their adaption for use during the milk
collection process. For most antibiotics, the test is advertised to take between 8 – 15
minutes which is beyond the approximate 5 minute period it takes to milk a cow [57].
Reading the test strip often requires trained personnel [9]. Also, the manual
11
implementation and single use nature of the technology make it expensive and
challenging to automate.
2.4.2 Prepared Incubated Test
The prepared incubated test uses a colourimetric chemical reaction to indicate the
presence of antibiotics in a milk sample [6], [7]. An example of a commercial product
using this method for on-site antibiotic testing is the DelvoTest T. This product provides
farmers with an incubated well-plate holding small vials of the reactive solution which
appear purple in colour. The user adds 100 µL of raw milk to the vial and allows the
incubator to set for three hours at 64 ºC. After this time, the solution will turn to a green
colour if antibiotic residues are present. A sample without residues will remain purple.
The results of the incubated test are meant to be visually interpreted as positive or
negative depending on colour of the solution. However, a study has shown that
spectroscopic analysis of the sample can be used to quantify concentration of
antibiotics [6]. It should be noted that such analysis can only give an idea of general
antibiotic concentration in the milk but does not indicate concentration of any specific
antibiotics [7].
To characterize accuracy of the DelvoTest, an experiment has been conducted to
test 27 antibiotic residues in raw milk [7]. The researchers found that the DelvoTest
consistently detected 23 out of the 27 antibiotics when they were present at levels
exceeding the MRL mandated by the European Union. The remaining 4 antibiotics had
12
a false-negative rate between 1.7% and 4.9%. The DelvoTest has been reported in
other studies to be unreliable for detection of tetracycline antibiotics [6], [7].
The advantage of the prepared incubated test is that there is no preprocessing of
the milk required. The readings also show accuracy suitable for on-site testing and
multiple antibiotics can be tested simultaneously. However, the test requires hours to
complete. There is also responsibility of the user to maintain the temperature of the
incubation and read the final output. These limitations make the method infeasible for
an automated, high throughput dairy sensor.
2.4.3 Microchip Capillary Electrophoresis
A more recent advancement to dairy testing has been the application of
continuous flow microfluidics. These continuous flow microfluidic devices leverage an
actuation technique termed microchip capillary electrophoresis (MCE) [37]–[40], [42],
[43], [58]–[61]. MCE uses the interaction between actuation forces of electroosmotic
flow and electrophoresis to achieve milk sample separation [58].
An MCE device features fluid channels that are microns wide and usually
centimeters long to transport the sample solution. Electrodes are placed at the ends of
the fluid channel so that voltage can be applied [59]. The walls of the microfluidic
channels are made from borosilicate glass, polydimethylsiloxane (PDMS) or a similar
material. It is important that the material of the channels should become deprotonated
by contact with aqueous electrolyte solution and therefore take a negative charge [62].
This deprotonation is required to enable electroosmotic flow [62], [63].
13
The electroosmotic flow is the dominant driving force in MCE. The electroosmotic
flow results from the interaction between the negatively charged channel surface, polar
aqueous solution and external electric field [63]. Firstly, the negatively charged channel
surface causes the polar aqueous molecules to adhere to them. Thus, what is known as
the electric double layer is formed by the channel walls and water molecules [64]. When
an external DC voltage is applied across the fluid channel, a uniform electric field is
created [59]. Within this electric field, the negative charges on the channel walls will
migrate towards the cathode at an approximately constant velocity [64]. Subsequently,
the electric double layer mobilizes toward the cathode. Because the channels are only
microns wide, viscous effects in the fluid are sufficient to pull the bulk solution towards
the cathode [62]. By this mechanism, electroosmosis creates an approximately constant
fluid velocity towards the cathode.
The separation of constituents in the sample solution is achieved by
electrophoresis. The electrophoretic force acting on a particle is the force induced on a
particle by an external electric field [64]. The velocity of a particle due to electrophoresis
is proportional to its charge and size [65]. This is to say, charges with different charge
and mass properties will move at different velocities in an electric field.
In MCE applications, the electroosmotic force acting on the sample solution will
cause a net flow towards the cathode. Simultaneously, the electrophoretic force acting
on each particle will result in varied force of attraction to the cathode or anode. The
cumulative result is the net flow of fluid towards the cathode while constituents of the
fluid spread apart due to electrophoresis [62], [64], [65].
14
Once separated, the fluid constituents can be detected. MCE devices have
successfully been coupled with fluorescent labels [66], electrochemical sensors [65],
UV-Vis fluorescence spectroscopy [37], [38], [40], [58], [60], [61] and mass spectrometry
[41]–[43], [61] to measure concentration of antibiotics in milk. MCE has strength in
sensitivity as several of these previous works demonstrated limits of detection
exceeding regulatory limits [37]–[43], [60], [61], [66]. However, MCE faces significant
challenges in robustness and automation.
The narrow channels in continuous flow microfluidic are extremely prone to
becoming clogged by fat and protein biomolecules. For this reason, MCE applications
using milk must centrifuge or filter the milk before loading onto the device [37], [38],
[40]–[43], [58], [60], [61], [67]. The channels must also be thoroughly rinsed with custom
buffer solutions between each trial [37], [38], [40]–[43], [58], [60], [61], [67]. This
requirement for preprocessing and manual rinsing requires equipment and time that is
unavailable during an on-site milk collection process.
Additionally, MCE dairy applications typically require voltages in the range of 14
kV – 30 kV [37], [38], [40]–[43], [60], [61], [67]. Although one previous work by Bosma et
al. reduced the voltage level to 500 V, sacrifices were made in response time (18
minutes) and limit of detection [58]. These voltage levels may be challenging to
generate in an on-site device with respect to miniaturization and safety.
In summary, traditional microfluidic devices have demonstrated the portability and
sensitivity for dairy testing. However, the need for milk sample preprocessing, time
15
consuming manual rinsing and particularly high voltage requirements make MCE
unviable as an on-site dairy tester.
2.5 Actuation on Digital Microfluidic Devices
This section provides background on the design and actuation of droplets on a
digital microfluidic device.
2.5.1 General Digital Microfluidic Design
A DMF device uses a grid of isolated electrodes to transport droplets of sample
fluid in two dimensions. The electrodes typically have length and width on the scale of
millimeters and can be patterned in a variety of geometries [68]. These electrodes are
used to generate localized electric fields which drive the movement of liquid droplets. As
such, each electrode must have a switchable connection to ground and voltage high. An
insulating dielectric layer is coated over top the electrodes. It is common for digital
microfluidic devices to use voltages up to 400 V [69] and so the dielectric layer is
needed to prevent arcing between electrodes. A hydrophobic layer is then coated over
the dielectric to minimize adhesion to the droplet. This hydrophobic layer eases droplet
movement over the electrodes.
There are two main actuation forces used to drive droplet movement:
dielectrophoresis and electrowetting-on-dielectric (EWOD) [14]. Dielectrophoresis
applies a non-uniform dielectric field to drive movement of a polarizable liquid [70], [71].
The direction of movement of a droplet under dielectrophoresis is dependent on how
polarizable the droplet is relative to the surrounding medium [72]. If the droplet has
16
greater dielectric constant than the filler medium, it is pulled towards the cathode [71]. A
droplet with dielectric constant less than the filler medium will be pulled towards the
anode [71]. Dielectrophoresis is used to handle insulating droplets such as deionized
water [71] and oils [70], [72]–[74].
EWOD is the more common actuation force in DMF applications [14], [15].
EWOD applies an external electric field to drive movement of a conductive liquid [16],
[17], [73], [74]. Most reported DMF applications use conductive aqueous droplets and
thus, EWOD is more commonly used [14]. The presence of the electric field causes
polarization of the dielectric layer and droplet. Consequently, an electric double layer
develops between the dielectric and droplet which causes an increased attraction over
the polarized surface [16]. Using this EWOD effect, the droplet is pulled in the direction
of the electric field. Further detail on EWOD and design of an EWOD driven DMF device
is given in Chapter 3.
2.5.2 Biofouling
Biofouling is an important effect to consider in DMF devices working with food
sample droplets such as milk. Biofouling refers to the effect of biomolecules including
fats and proteins adhering to the surfaces of the DMF devices [19], [21]. Even
hydrophobic coatings are prone to biofouling when they come in contact with
biomolecules [75]. For DMF devices, droplet actuation depends on a homogenous
hydrophobic coating across the device surface. By irreversibly contaminating the device
surface, biofouling causes droplets to become stuck at the site of fouling thereby
disabling droplet actuation [21].
17
One previous work mitigated biofouling by adjusting the sample solution pH,
using AC voltage actuation and limit the time that voltage is applied [76]. However, this
requires manual preparation of the reagent solution and does not allow the droplet to sit
stationary as would be required for any detection measurements.
Another work addresses biofouling in EWOD with etched nanostructures on the
device surface [77]. This approach has the disadvantage of complex and expensive
device fabrication.
The most common solution to biofouling has been the use silicone oil filler
mediums [15], [78], [79]. This approach surrounds droplets with low viscosity silicone oil
which encapsulates droplets and prevents physical contact with the device surfaces.
The disadvantages of using a silicone oil filler medium are discussed in the following
section.
2.5.3 Filler Medium
As described above, DMF is performed with droplets positioned above a grid of
electrodes. Thus, the droplets are surrounded by a fluid filler medium. The requirements
of the filler medium to facilitate droplet movement are low viscosity, electrically
insulating and immiscibility with the sample droplet [15]. The two most commonly used
filler mediums are air and low viscosity silicone oil [15].
The motivations for using a silicone oil filler medium are lowered actuation
voltages [80], faster droplet speeds [17] and reduced physical contact between the
droplet and solid surfaces [15]. The reduced physical contact is used in applications
18
where biofouling is a concern. Using a silicone oil filler medium, DMF devices have
been demonstrated to handle bio-fluids including human blood and saliva [78].
However, the use of an oil filler medium comes with the problem of emulsification and
interference with the sample droplets [11], [15], [74], [78]. To mitigate an oil filler
medium’s interference with sample droplets, chemical modifications to the sample
solution are needed [15]. Hence, the use of an air filler medium is preferred for
applications where accurate chemical analyses of the sample are valued.
2.6 Detection on Digital Microfluidic Devices
To adapt digital microfluidics for dairy testing, a method to detect antibiotic
residues in milk droplets must be determined. In this section, the available chemical
detection methods which have been previously integrated with digital microfluidic
devices are described.
2.6.1 Fluorescence Spectroscopy
Fluorescence spectroscopy is an optical detection technique that relies on
exciting a molecule by exposure to electromagnetic radiation and subsequently
measuring the lower-energy electromagnetic radiation emitted by the molecule [81].
This molecule, known as a fluorophore, absorbs the initial light and an electron is
elevated to a higher energy state [81]. The electron then falls to a lower energy without
emitting any light through a process termed internal conversion [81]. Finally, the
electron returns to its ground state and emits light at a wavelength proportional to the
difference in energy levels [82].
19
The energy level of the incident light used to excite the molecule and energy
level of the emitted light are dependent on the chemical properties of the molecule [82].
This is to say, the wavelength used for excitation and emitted wavelength are
dependent on the target analyte. Additionally, in cases where the fluorescence
characteristics of the target analyte are undesirable, a fluorescent label can be used. A
fluorescent label is a molecule which binds to the target analyte and has a known
fluorescent response [10], [47]. Thus, applying fluorescence spectroscopy to measure
concentration of the label is analogous to directly measuring concentration of the
analyte it binds to.
Optical detection techniques such as fluorescence spectroscopy are well suited
to microfluidic devices due to the small footprint of optoelectronics and the ability of light
to travel in small spaces. Moreover, the common use of transparent materials in
microfluidic fabrication such as PDMS and indium tin oxide (ITO) make the devices fit
for optical coupling. Using LEDs, photodiodes and optical filters, past works have
produced fully integrated digital microfluidic devices offering fluorescent sensing [83],
[84]. Optical spectroscopy in microfluidics is traditionally performed in the vertical plane
(perpendicular to sample motion). However, some works have used fiber optic cables to
enable fluorescence spectroscopy in the horizontal plane thereby increasing sample
path length and sensitivity [52], [85]. Thus, attractive features for use of fluorescence
spectroscopy with digital microfluidics include minimal space requirement and
compatibility with traditional fabrication methods.
20
2.6.2 Surface Plasmon Resonance
Surface plasmon resonance (SPR) is an optical detection technique which
requires a thin reflective surface placed between a prism and a medium with lower
refractive index than the prism [86]. Common configurations use a gold film tens of
microns thick between a prism and an air or solution medium [87]. A beam of polarized
light is shot on the gold surface through the prism and photons meeting the resonant
energy of the surface will be absorbed while other wavelengths are reflected [86], [88].
Using a sensor to capture the reflected light, the resonant wavelength can be identified
as a dark band in the spectrum. SPR leverages this phenomenon by immobilizing
molecules which bind to the target analyte onto the gold surface; when binding occurs
between the immobilized surface molecules and analyte, the resonant energy of the
surface changes [88]. Thus, when the immobilized molecule is correctly selected,
measuring the resonant wavelength as analyte passes over the surface gives time-
resolved concentration data [87].
SPR has been demonstrated in traditional microfluidic systems to detect multiple
antibiotic families in milk with limits of detection below Canadian MRLs [89], [90]. This
method has also been integrated with digital microfluidics [91]. SPR offers strong
sensitivity. Weaknesses include complex and expensive fabrication of the sensor
surface.
21
2.6.3 Impedance Spectroscopy
Impedance spectroscopy (IS) is a technique where alternating current (AC) low
voltage signal is applied across the sample while the response is measured in form of
current or impedance [92]. From measuring the magnitude and phase responses as a
function of frequency, the real and imaginary components of the sample’s impedance
are obtained [92]. The real component of impedance represents the resistive element of
the sample while the imaginary component represents the capacitive element [92], [93].
By comparing known IS responses with the signature given by a sample, properties of
the sample such as concentration of an analyte can be inferenced [92]–[94].
Previous work has integrated IS with digital microfluidic [93], as well as traditional
microfluidic systems where the presence of DNA and antibodies were detected with
limits of detection on the nanomolar scale [94]. IS is relatively easy to integrate with
digital microfluidics as it only requires two electrodes able to simultaneously contact the
sample. Disadvantages include the fact that the electrodes usually have a hydrophilic
surface making them prone to biofouling [19], [95].
2.6.4 Electrochemical Detection
Electrochemical detection techniques require three separate electrodes: the
working electrode, reference electrode and counter electrode [96]. The reference
electrode houses a redox reaction independent from the analyte solution in order to
provide a known reference potential [96]. Thus, a potential can be applied or measured
in reference to this known potential. The counter electrode is available as a path for
22
current to flow between itself and the working electrode when voltage is applied [96].
Using this three electrode configuration, three main measurement methods can be
used: potentiometry, amperometry and voltammetry.
Potentiometry measures the voltage between the working electrode and reference
electrode while immersed in the sample solution [97]. Amperometry measures the
current through the sample when a known voltage is applied to the working electrode
[97]. Lastly, voltammetry is the most versatile method of the three as it measured
voltage and current [97]. Voltammetry uses a device called a potentiostat to sweep
through a range of voltages applied to the working electrode and measure the resulting
currents in the sample; this generates a plot known as a voltagramm [97]. These
measurements can be performed with or without the use of labels. From each of these
methods, the measured response from a sample can be compared against calibration
data to determine concentration of an analyte [96], [98], [99].
Previous works have used potentiometry and label molecules to detect tetracycline
antibiotics in milk at concentrations low as 1 ppb [100]. Moreover, studies have
integrated the three electrode configuration on digital microfluidic chips and
demonstrated the capacity for potentiometric sensing [96], [98]. A significant drawback
of potentiometric sensing is vulnerability to biofouling due to the hydrophilic electrode
surfaces [19].
23
3 An Integrated Digital Microfluidic Device for Dairy Applications
This chapter addresses the problems of overall device integration and milk droplet
actuation for a digital microfluidic dairy testing device.
3.1 Chapter Introduction
The study of microfluidics represents the miniaturization of traditional chemical
analyses processes onto a single integrated device [59], [101]–[103]. Microfluidics offers
advantages in portability, fast response time and minimal reagent consumption over
their full-scale laboratory counterparts [14]. This technology has been adapted to on-site
applications including clinical diagnostics [104], environmental testing [105] and
biological cell culturing [106]. Traditional microfluidic devices transport reagent fluids
through micron-scale width channels driven by hydraulic pressure [66], [107] or electric
field gradients [58]. This method of microfluidics is broadly termed continuous flow
microfluidics.
More recently, a subset of microfluidics known as digital microfluidics (DMF) has
garnered increased favor due to its unique ability to handle samples in discrete droplets
over an array of electrodes rather than through continuous fluidic channels [16]. Where
traditional microfluidic devices are limited in flexibility by the fabrication of rigid fluid
channels, DMF devices offer the potential to be reprogrammed to achieve
reconfigurability, automation and highly parallel operations [108].
The optimal design of an integrated DMF device, having fully leveraged the
aforementioned capabilities, is an on-going effort. As such, there is a need for further
24
automation of DMF devices. This is particularly true to on-site applications, such as on-
farm dairy analyses [58], [59].
A significant challenge faced in the microfluidics domain, being relevant to
applications concerning biological samples, is the phenomenon known as biofouling.
Biofouling refers to the adhesion of fat and protein biomolecules to the fluid handling
surface of the microfluidic device, resulting in damage and the degradation of system
performance [19].
In this chapter, we address the issues of overall device integration and biofouling
within a DMF device. An integrated DMF device controlled through a versatile graphical
user interface (GUI) computer program is demonstrated to actuate milk droplets. The
velocities of milk and water droplets are compared for different dielectric coatings and
device configurations to address the effect of biofouling. The results of this chapter are
important in demonstrating an integrated DMF design suited to bio-fluid actuation and
opening the use of DMF devices in dairy testing applications.
3.2 Background
Dairy testing is an important field of research which has seen exponential growth
in published literature over the past several decades [5]. Around the world, veterinary
drugs are administered to dairy cattle to promote growth, control breeding and treat
disease [26], [109]. Consequently, the administered antibiotics and their metabolites are
prone to appearing in the milk produced by these cattle in the form of antibiotic residues
[26]. The presence of antibiotic residues in the dairy milk can be directly toxic to human
consumers [26] or cause the development of unwanted antibiotic resistance [110]. To
25
address this problem, worldwide regulatory bodies conduct laboratory tests for
identification of antibiotic residues in dairy products which are made available for
consumption [111].
There is an important need for the development of an on-site dairy tester which
has not been met by existing technology. Traditionally, testing milk for antibiotic
residues occurs at an off-site milk processing laboratory [2]. The farmer is not equipped
to test milk on-site at the dairy farm before bulk tank collection and shipping to the
processing facility [58]. Consequently, the farmer’s dairy enterprise is burdened if
contaminated milk is inadvertently collected. By submitting non-compliant milk to supply
chain stakeholders, the farming enterprise is penalized with monetary fines and loss of
licensure after multiple offences [2]. Modern commercially available methods for on-site
dairy testing are unviable to use during milk collection because of their manual
requirements, long test times and need for trained personnel [7], [9].
A promising move towards on-site dairy testing is offered by microfluidic
technology. Previous studies applying continuous flow microfluidics to dairy testing have
demonstrated the capacity to separate and detect antibiotic residues in milk [37]–[39],
[43], [58], [66], [67], [112]. However, continuous channel microfluidic systems face
significant challenges in robustness and automation which limit their suitability for on-
site dairy testing. To prevent fat and protein molecules in milk from clogging the narrow
channels, the milk samples must be centrifuged or filtered before loading onto the
microfluidic device [37]–[39], [43], [58], [66], [67], [112]. The fluid channels must also be
carefully rinsed for several minutes with custom buffer solutions between each trial [37]–
26
[39], [43], [58], [66], [67], [112]. These manual processing steps require time and
equipment which is not available during a dairy farm’s milk collection process.
DMF is an alternate technology which addresses the limitations in robustness
and scalability faced by conventional continuous microfluidics. DMF handles liquid
samples in discrete droplets across an array of addressable electrodes as opposed to in
narrow channels. By applying voltage to the individual electrodes, a dielectric surface is
polarized and attracts droplets of conductive liquid by a force known as electrowetting-
on-dielectric (EWOD) [108]. DMF devices have been demonstrated to transport, split
and mix liquid droplets thereby showing their capacity for lab-on-chip functionality [113].
Scalable designs for droplet actuation [18] and position sensing [103] allow DMF
devices to handle multiple droplets in parallel. Hence, DMF devices offer capability for
automation, reconfiguration and parallel throughput required by an on-site dairy tester.
DMF encounters two challenges in being adapted for dairy testing. These
challenges are as follows: (1) integration of automation and reconfigurability and (2)
overcoming biofouling. A DMF device for dairy testing must feature ease of
reconfiguration and automation to be used for detecting multiple antibiotic residues in an
on-site solution. The device also has to overcome biofouling to prevent biomolecules in
milk from adhering to the chip surface and causing the droplets to become immovable
[21].
Important previous works reporting integrated DMF devices are the DropBot by
the Wheeler Microfluidics Laboratory [12] and the OpenDrop by Gaudi Labs [11]. The
DropBot is a DMF fixture which allows droplet position sensing, instantaneous droplet
27
velocity measurements and droplet actuation. The DropBot is an excellent research tool
but requires a complex and bulky assembly not suitable to on-site applications. By
contrast, the OpenDrop is a simpler DMF assembly which uses via-hole routing to
construct a densely packed array of electrodes. Users control the droplet position
through a button and joystick interface. The drawback of this embedded user interface
is that complex functions such as multiple droplet handling and automation are non-
intuitive and challenging to achieve for the user.
In this work, we address DMF integration challenges as we have built an
integrated and modular printed circuit board (PCB) design allowing individual control of
60 electrodes connected by through-hole via routes. The electrodes are controlled
through a user-friendly GUI application which allows real-time electrode switching and
automation through recorded sequences. Thus, parallel handling of multiple droplets
and automation are easily achieved and reconfigured.
To address the issue of biofouling, we develop a novel DMF chip with enhanced
properties of a silicone oil hydrophobic layer. The velocity of milk droplets is quantified
in relation to water when device configuration and dielectric coating material are varied.
The actuation of milk droplets is demonstrated within an air filler medium as opposed to
an oil filler medium. Oil filler mediums reduce the voltage needed for droplet actuation
[80], however, they require more complex device fabrication and interfere with biological
samples [11], [15]. For this reason, silicone oil was used only as a hydrophobic layer
while the droplets were surrounded by air. The velocity of milk and water droplets were
measured in closed and open DMF system configurations. The closed DMF system is
28
differentiated by presence of a top electrode which sandwiches the droplet. The effect of
dielectric material on biofouling was also studied by measuring droplet velocity when a
gel-based polydimethylsiloxane (PDMS) or thin-film parafilm dielectric were used in the
DMF device.
3.3 Electrowetting Actuation
The EWOD effect used to actuate droplets in a DMF system is based on the
manipulation of surface energy at the interface between a polarizable solid layer and a
conductive liquid droplet [16]. The Lippmann equation expresses this relationship as
𝛾𝑆𝐿(𝑉) = 𝛾𝑆𝐿(0) −휀𝑉2
2𝑑 ,
(1)
where 휀 and 𝑑 are the permittivity and thickness of the solid layer respectively, 𝑉 is the
potential applied across the dielectric layer, 𝛾𝑆𝐿(0) is the solid-liquid surface tension at
rest and 𝛾𝑆𝐿(𝑉) is the solid-liquid surface tension in the presence of an electric field
applied across the dielectric. The Lippmann equation is complemented by Young’s
equation which relates the surface tensions at a three phase solid-liquid-gas interface
as
𝑐𝑜𝑠𝜃𝐶 =𝛾𝑆𝐿 − 𝛾𝑆𝐺
𝛾𝐿𝐺 , (2)
where 𝛾𝑆𝐺 and 𝛾𝐿𝐺 are surface tensions at the solid-gas and liquid-gas interfaces
respectively and 𝜃𝐶 is the contact angle formed between the surface and inner edge of
the liquid as labelled in Figure 1. By substituting (1) into (2), the resulting Young-
Lippmann equation shows
29
𝑐𝑜𝑠𝜃𝑉 = 𝑐𝑜𝑠𝜃0 +휀𝑉2
2𝛾𝐿𝐺𝑑 ,
(3)
where 𝑐𝑜𝑠𝜃0 and 𝑐𝑜𝑠𝜃𝑉 represent the solid-liquid contact angle at rest and in the
presence of an electric field.
The Young-Lippmann equation shows that the contact angle at a dielectric solid-
liquid interface decreases when an electric field is applied; this phenomenon is known
as EWOD. In DMF devices, EWOD is leveraged by switching localized electric fields to
create an unbalanced interfacial tension across a liquid droplet [16]. The resulting
pressure gradient across the droplet causes it to translate in the direction of the
activated electrode [113].
Although actuation of aqueous droplets has been repeated across many works,
handling milk samples is more complex. This is due to the biofouling effect which
causes the device’s surface to be irreversibly contaminated by fats and proteins [19].
Additional opposing forces to EWOD include surface roughness and contact angle
hysteresis [114]. In this chapter, we overcome these opposing forces to demonstrate
milk droplet actuation.
30
Figure 1: Cross-sectional views of a DMF device is presented for (a) a closed system configuration, and (b) an open system configuration. A closed DMF system configuration constitutes a continuous upper electrode tied to ground and is coated in a hydrophobic material. The open DMF system configuration system is without the top electrode.
3.4 General DMF Design
General design of a DMF system includes the material layers shown in Figure 1.
The bottom layer is composed of isolated electrodes which should always be pulled low
to ground or high to the supply voltage. A dielectric layer over electrodes provides the
high impedance needed to prevent leakage currents between alternately switched
electrodes and create the electric field gradient needed for EWOD. The dielectric layer
also serves as the polarizable solid surface for EWOD. A hydrophobic coating is used to
reduce surface forces and ease droplet translation. Note from Figure 1 that the droplet
volume must be large enough that the droplet overlaps with adjacent electrodes. This
arrangement is necessary to facilitate actuation by EWOD [17].
31
DMF devices can be designed in closed and open configurations as pictured in
Figure 1(a) and Figure 1(b) respectively. By contrast to an open configuration, the
closed configuration uses a continuous top electrode to sandwich the droplet with the
bottom layer. The closed configuration allows smaller droplet volumes than an open
configuration as the sandwiched droplet more readily spreads onto adjacent electrodes.
Moreover, the presence of the top electrode creates a strong vertical component of the
electric field gradient. This arrangement allows improved control and alignment of
droplets over the activated electrodes when compared to the open configuration [17]. To
perform more complex operations of droplet splitting and dispensing, a closed
configuration is needed [115]. In the open configuration, frictional force between the
droplet and top layer is eliminated. The open configuration is also advantageous in
providing access to the droplet for sensors [116] and for handling larger volume
droplets. The device demonstrated in this work can be operated in both closed and
open configurations.
The choice of material and thickness of the dielectric layer are also important
parameters which affect droplet mobility in a DMF system. As given by equation (3), a
thinner dielectric layer results in a greater change of contact angle during EWOD.
Simultaneously, the dielectric must have sufficient dielectric strength and thickness to
withstand the applied voltages without breaking down. Previous works have shown
dependence of droplet velocity on dielectric thickness and material [103].
32
To experimentally determine the effect of DMF device configuration and dielectric
material on biofouling, an easily modifiable DMF design was required. We address this
need with a modular integrated DMF design detailed in the following section.
3.5 DMF Prototype Design
The DMF device PCB pictured in Figure 2(b) was divided into two separate
boards: the electrode board and the control board. The electrode board incorporated 60
interdigitated electrodes of varied pitch and spacing. For our experiments, a 6 by 6
electrode grid of 2.75 mm dimensioned square electrodes with 100 micron spacing was
used. The electrode board featured four-layers and via-holes to reach each of the
individually addressable electrodes. Creepage and clearance between routes was
managed to safely accommodate up to 300 V in adherence with IPC-2221 guidelines
when a conformal coat was applied [117]. The physical separation of the electrode
board from the control board allowed experimenting with electrode surface treatments
without consequence to the control electronics.
The control board architecture incorporated an Arduino Nano microcontroller
running a finite state machine firmware. This event driven firmware was responsible for
receiving commands from the GUI application and relaying signals to an HV507 serial-
to-parallel converter (Microchip Technologies Inc.). The HV507 is a push-pull high-
voltage driver with 64 channels and rated up to 300 V. Commands from the GUI
application were sent to the microcontroller over a universal asynchronous receiver
transmitter (UART) serial connection. Commands for toggling electrodes on-off were
sent with a 64-bit integer argument where each bit corresponded to the switch status of
33
an individual electrode. The microcontroller parsed the 64-bit argument and drove the
corresponding electrode channels through the HV507. At a UART communication rate
of 9600 bits per second and microcontroller clock rate of 16 MHz, this method allowed
switching times less than 15 milliseconds.
Figure 2: Hardware design of an integrated DMF device is presented. (a) The block diagram of the DMF device comprises of three hardware components, being: (1) the PCB housing firmware and electrodes, (2) personal computer running the control application, (3) a high voltage supply. (b) The modular PCB design using physically separated boards for control logic and electrodes connected by ribbon cables. (c) Photograph of the fully assembled DMF device.
The GUI desktop program used to control electrodes and automate sequences is
shown in Figure 3. This application was developed for Windows environments using NI
LabVIEW. From the GUI, users have control and visualization of the actuated
electrodes. Electrodes can be spatially selected via computer peripherals to switch in
real-time or recorded sequences could be played with defined time duration between
steps. The recorded sequences enable easy automated handling of one or many
34
droplets simultaneously. An actuation frequency could also be set to drive the
electrodes with an AC-like pulse frequency modulated (PFM) signal with fixed duty cycle
of 50% and user defined frequency. Note that driving a true AC signal would require an
inverter circuit on the PCB.
Figure 3: Graphical user interface of the DMF device is presented featuring real-time control of 60 electrodes. Buttons and switches provide the capacity to automate recorded sequences and enable AC-like PFM actuation.
3.6 Closed Configuration Experiments
The closed configuration experiments were performed with an indium tin oxide
(ITO) top layer as pictured in Figure 2(b). The ITO slide was tied to ground by wrapping
the edges in a copper tape and connecting to a ground wire. The top layer was spaced
35
approximately 240 microns apart from the bottom layer by four layers of 60 micron thick
tape. A commercially available fluoropolymer hydrophobic coating called Fluoropel
PFC1601V (Cytonix Inc.) was spin coated onto to the ITO slide.
Closed configuration droplet velocities were calculated using the minimum
electrode switching time required by droplets to travel an automated path. The droplets
were actuated in a square circuit of eight electrodes while the switching time was
reduced until the droplet could not keep pace. Thereby, velocities were calculated using
the electrode pitch of 2.75 mm and the minimum achievable switching time. Beginning
at a voltage of 300 V, the voltage was decremented in steps of 10 V until the droplet
could not be actuated between successive electrodes. Measurements were made in
triplicate at each voltage level with a fixed droplet volume of 4.2 µL.
The droplet velocities versus voltage are plotted in Figure 4 where Figure 4(a)
uses a PDMS dielectric and Figure 4(b) uses a parafilm dielectric. Only water droplet
velocities are plotted in Figure 4 as milk droplets were immoveable in the closed
configuration. The milk droplets left behind residues on the fluoropel coated ITO which
caused droplets to stick to these spots. This biofouling effect could only be corrected by
recoating the ITO with a fresh fluoropel layer. In the closed configuration, milk droplets
could be moved only when the droplet was encapsulated in a silicone oil shell.
Consequences of the silicone oil shell include potential emulsification with the sample
fluid which is detrimental to sensing applications [15]. Our results using an air filler
medium showed that the biofouling effect is dominant in a closed configuration.
36
Closed configuration water droplet velocities are plotted against actuation
frequency in Figure 5 with voltage fixed at 280 V. The trend shows droplet velocity is
maximized at DC and decreases to a plateau as frequency is increased. Previous
studies show a similar trend of decreasing EWOD force with actuation frequency [74],
[118], [119]. The complex permittivity of the droplet is given by
휀∗ = 휀 − 𝑗𝜎
2𝜋𝑓 , (4)
where 휀 is dielectric permittivity, 𝜎 is conductivity and 𝑓 is frequency. Equation (4)
shows that the contribution of the droplet’s conductivity to the complex permittivity is
reduced as signal frequency increases [118]. Thus, at high frequencies a conductive
droplet behaves as an insulating droplet. The EWOD force which requires the droplet to
be conductive is then weakened at higher frequencies [15]. The frequency weakened
EWOD force results in a smoother droplet motion which can be leveraged for finer
control over droplet velocity [120].
37
Figure 4: Droplet velocity versus voltage in a closed configuration is presented for both (a) PDMS and (b) parafilm dielectric layers using water. The mean velocity (circle) and corresponding standard deviation (vertical bar) of n = 3 repeated measures are shown as a function of DC voltage. The closed microfluidic system is configured having 240 micron spacing between top and bottom planes for a fixed droplet volume of 4.2 µL.
38
Figure 5: Droplet velocity versus actuation frequency in a closed configuration is presented for both (a) PDMS and (b) parafilm dielectric layers using water. The mean velocity (circle) and corresponding standard deviation (vertical bar) of n = 3 repeated measures are shown as a function of frequency. The closed microfluidic system is configured having 240 micron spacing between top and bottom planes for a fixed droplet volume of 4.2 µL and a 280 V actuation potential.
39
3.7 Open Configuration Experiments
The open configuration eliminates friction between the droplet and the top planar
surface thereby enabling increased droplet velocities when compared to closed
configurations. The weakness of the open configuration is alignment of the droplet over
activated electrodes. Because EWOD is a polarity independent force, there exists a
position of equal electric field gradient exactly in the middle of an electrode at high
potential and an electrode at low potential [17]. Without the vertical field gradient
component given by a top planar electrode, the droplet is prone to becoming “stuck”
between electrodes. There are two solutions for this problem. The first is to increase the
ratio of droplet volume to electrode pitch [17]. The other is to have a densely packed 2D
electrode array whereby the droplet can be recovered when positioned between two
electrodes. In this work, we apply the latter solution using via routing to construct the
electrode array.
Open configuration droplet velocities were measured using video frame analysis
to determine time for the droplet edge to move from one electrode to the center of an
adjacent activated electrode. The droplet volumes were fixed at 9.0 µL. Actuation
voltage was decremented in steps of 10 V until the droplet could not be actuated
between successive electrodes. Figure 6 plots the velocities of water and milk droplets
on PDMS and parafilm dielectric layers respectively. By eliminating droplet contact with
the fluoropolymer hydrophobic layer used in the closed configuration, biofouling of the
fluoropolymer is overcome. The silicone oil layer on the bottom plane forms a thin film
preventing biofouling of the bottom plane.
40
Figure 6: Droplet velocity versus voltage in an open configuration is presented, as a comparison between milk (red) and water (blue), for both (a) PDMS and (b) parafilm dielectric layers. The mean velocity (circle/square) and corresponding standard deviation (vertical bar) of n = 3 repeated measures are shown as a function of DC voltage. The open microfluidic system performance is characterised for a fixed droplet volume of 9.0 µL.
From Figure 6 the maximum observed milk droplet velocities were 4.6 mm/s and
8.8 mm/s on the PDMS and parafilm respectively. Similarly, maximum observed water
droplet velocities were 10.9 mm/s and 11.0 mm/s on PDMS and parafilm respectively.
41
Milk droplet velocities are significantly less than those of water likely due to the effect of
fat and protein constituents on physical parameters not limited to mass, viscosity and
conductivity. It should be noted that the surface roughness of the PDMS coating had a
significant effect in open configuration experiments. Water and milk droplets on the
PDMS dielectric regularly became stuck between electrodes thereby making continuous
actuation challenging. The PDMS surface roughness was overcome in a closed
configuration due to the improved control of droplet alignment but was a prevalent issue
in the open configuration. By contrast, the parafilm dielectric reliably allowed continuous
actuation of water and milk droplets in an open configuration. To mitigate surface
roughness, more sophisticated coating techniques such as chemical vapour deposition
(CVD) can be used instead of spin coating. Similarly, surface roughness of the PCB
electrodes can be mitigated using the “via in pad plated over” (VIPPO) technique during
fabrication [121].
Snapshots of an automated actuation sequence of a milk droplet are shown in
Figure 7. Droplet velocities at voltages greater than 280 V could not be accurately
measured by the 20 frame per second video camera to be included in Figure 6.
However, open configuration droplet velocities on parafilm of up to 9.2 mm/s for milk
and 27.5 mm/s for water were measured at 300 V by automated actuation of droplets in
a circuit pattern (Figure 7) with minimum switching times of 300 ms and 100 ms
respectively.
42
Figure 7: Milk droplet actuation in open configuration DMF device is presented with parafilm dielectric layer and silicone oil. Electrode pitch is 2.75 mm, droplet volume is 9.0 µL, actuation voltage is 280 V.
Open configuration droplet velocities against actuation frequency shown in
Figure 8 give a similar trend to that observed in a closed configuration. As predicted by
equation (4), the EWOD force is maximized at DC actuation for conductive droplets.
43
Figure 8: Droplet velocity versus actuation frequency in an open configuration is presented, as a comparison between milk (red) and water (blue), for both (a) PDMS and (b) parafilm dielectric layers. The mean velocity (circle/square) and corresponding standard deviation (vertical bar) of n = 3 repeated measures are shown as a function of frequency. The open microfluidic system performance is characterised for a fixed droplet volume of 9.0 µL and a 280 V actuation potential.
44
3.8 Methods
3.8.1 Reagents
5 cSt silicone oil is purchased from Sigma Aldrich (Missouri, United States).
PDMS is purchased from Dow Chemicals (Michigan, United States). Fluoropel 1601V is
purchased from Cytonix LLC (Maryland, United States). Two percent milk is purchased
from a local vendor.
3.8.2 Digital Microfluidic Device Printed Circuit Board
The electrode board and control board were fabricated by JLCPCB (JiaLiChuang
Company Limited, China). The AQW216 solid state replay (Panasonic Corporation,
Japan) provided isolation of high and low voltage. The HV507 serial to parallel high
voltage driver was used to switch voltages to the 60 electrodes. The Arduino Nano
microcontroller firmware was a customized version of the open-source LINX Toolkit
(National Instruments, United States). The GUI control program was built using
LabVIEW 2017 (National Instruments, United States). The PCB was coated in 419D
acrylic conformal coating (M.G. Chemicals, Canada) to meet electrical spacing
requirements up to 300 V given by the IPC-2221 design standard [117]. Conformal
coating was applied by aerosol on the assembled PCB in four coats with four minutes to
air dry between coats. For conformal coat curing, the PCB was left on a hot plate at
80°C for 20 minutes. This process was repeated for both sides of the PCB. The
conformal coat was verified by a wet-gauge comb to have thickness less than 10
microns.
45
3.8.3 Dielectric Coating
PDMS coating was applied using the Sylgard 184 silicone elastomer (Dow
Chemicals Company, United States). The PDMS was mixed at a 7:1 ratio of elastomer
to curing agent. The mixed solution was spin coated onto the cleaned electrode PCB at
a maximum speed of 2000 rpm for 20 seconds. Note that conformal coating was not
applied to the electrode boards where PDMS was to be coated. The coated PCB was
left to cure on a hot plate at 105°C for 1 hour and then air cooled until the PDMS was
dry to touch. Two drops of silicone oil (Sigma Aldrich, United States) were applied and
spread gently by gloved finger.
Parafilm coating first required stretching of the parafilm to thin the dielectric layer.
A piece of parafilm was stretched across a flat surface by hand and held in the
stretched position for 10 minutes. The stretched parafilm was secured to a custom 3D
printed carrier by double sided tape. One drop of silicone oil was spread on the
electrode board to prevent air from being trapped between the PCB and parafilm. The
3D carrier was then secured to the electrode board by screws. Two drops of silicone oil
were applied and spread gently by gloved finger.
3.8.4 Statistical Analysis
Measurements of droplet velocity were made in triplicate. The data points plotted
in figures represent the mean of the n = 3 repeated measures. The vertical chart bars
represent distribution of the standard deviation between repeated measures.
46
3.9 Chapter Conclusion
In this chapter, an integrated DMF device with a modular PCB design, user-
friendly software interface, reconfigurability and capacity for automation was developed
for bio-fluid handling in an air filler medium. Experiments using closed and open
configurations determined that the open configuration is best suited to mitigate
biofouling due to its removal of physical contact between the sample and the
hydrophobic coating on the top plane. The use of silicone oil on the bottom plane
prevented any biofouling from milk droplets. Comparison of the gel-based PDMS and
thin-film parafilm dielectrics showed that the thin-film dielectrics are advantageous in
reducing surface roughness. Frequency modulation of actuation voltage supported the
understanding that EWOD force is maximized at DC while finer control over droplet
velocity is enabled by AC. Our DMF device demonstrated automated continuous
actuation of milk droplets using an open configuration, parafilm dielectric and silicone oil
coating. Thus, our results demonstrate applicability of DMF to dairy testing by
addressing the fundamental obstacles of device integration and biofouling. These
results respond to an urgent need for an on-site dairy testing technology that provides
robustness and scalability not offered by current methods.
47
4 Label-free Antibiotic Detection in Milk Droplets
This chapter addresses the problem of antibiotic detection in milk droplets required
for a digital microfluidic dairy testing device.
4.1 Chapter Introduction
It is a common practice around the world to administer veterinary drugs to cattle
for promoting growth and regulating breeding. Furthermore, antibiotics are routinely
used to treat infections in livestock such as cattle [26], [109]. Consequently, residues of
these antibiotics and their metabolites are prone to appearing in dairy products which
can then have negative health effects on human health including allergic reactions [26]
and antibiotic resistance [110].
To protect public health, regulatory bodies enforce the testing of dairy milk for
antibiotic residues [111]. Dozens of regulated antibiotics must fall under specified
concentration limits in order for the milk to be permissible for human consumption [2].
The tests for antibiotic residues are traditionally performed at a laboratory where dairy
farmers must ship their milk in large containers [58]. The critical flaw in this method is
that dairy farmers are unable to test the milk as they collect it and are therefore
vulnerable to collecting large batches of contaminated milk [58]. Modern commercially
available on-site dairy testing devices have long response times [6], [7] and complex
manual test procedures [9] which make them unviable to use during the milk collection
process. To address this urgent need for an on-site dairy tester, microfluidics represents
a promising research technology.
48
Microfluidic devices are miniaturized sensors that aim to replicate chemical
analyses processes performed in a laboratory onto an integrated device; thus, these
devices are often referred to as lab-on-a-chip devices. Advantages of implementing
chemical analyses processes on a microfluidic device include portability, minimal
reagent consumption and high throughput [14], [15]. Traditional microfluidic devices are
based on continuous flow of reagent fluid through micron-wide channels.
In the past two decades, a newer branch of microfluidics known as digital
microfluidics (DMF) has been actively studied [16]. The ability of DMF devices to
transport, split and mix individual droplets of liquid samples across a digitized grid of
electrodes provides a highly scalable lab-on-a-chip architecture [18], [113]. The
reconfigurability of DMF devices yields the potential to perform multiple parallelized
chemical analyses on a single integrated device [122]. DMF devices have found
applications including point-of-care clinical diagnostics [79], [123], environmental testing
[105] and biological cell culturing [106], [124]. In the previous chapter, we overcame
biofouling and presented a fully integrated DMF device for milk droplet actuation.
Hence, the robustness and capacity for automation of DMF devices make them an
appealing option for on-site dairy testing.
For DMF to be adapted for on-site dairy testing, the functionality to detect
antibiotic residues from milk droplets must be demonstrated. Previous works applying
continuous microfluidics to antibiotic detection in milk have used chromatography [38],
[125] and electrophoresis [37], [39]–[41], [58], [60] to separate sample particles based
on mass, charge and chemical structure. The separated particles flow through narrow
49
channels on the device for subsequent detection by UV-Vis spectroscopy [37]–[40],
[58], [60] or mass spectrometry [41]–[43], [61]. Although continuous flow devices are
effective in sensing antibiotics, they face significant limitations including need for
preprocessing of milk samples, very high voltages and need for custom buffer solutions
[37]–[43], [60], [61]. To prevent clogging of narrow channels, milk samples must be
centrifuged or filtered using laboratory equipment [37]–[43], [60], [61]. The channels
must then be rinsed for several minutes between each trail [37]–[43], [60], [61]. The
DMF device in the previous chapter is able to actuate milk droplets without any
preprocessing. To achieve fast response times and low detection limits, continuous
microfluidics rely on high voltages in range of 14 kV – 30 kV [37]–[43], [60], [61].
Although voltages low as 500 V have been demonstrated, they sacrifice performance
and sensitivity. These kilovolt scale potentials are challenging to generate in a
miniaturized on-site dairy tester. By contrast, the DMF device in the previous chapter
operates at 280 V which can be reliably produced by a voltage boost converter circuit
[11]. Continuous flow devices also require carefully prepared and stored buffer
solutions to be loaded in the channels for each trial [37]–[43], [58], [60], [61].
To summarize, modern continuous flow devices are unsuitable for on-site dairy
testing due to dependence on laboratory equipment, long preparation times and manual
processes. DMF devices provide valuable improvements in these areas but struggle in
antibiotic separation and detection. In this chapter, we address the functionality of
antibiotic sensing for an on-site DMF dairy tester.
50
The most widely used detection method in DMF biosensors is optical detection
including absorbance and fluorescence spectroscopy [15]. Optical detection is well-
suited to DMF devices due to being contact-free, miniaturize-able and accessible. Being
contact-free allows optical detectors to work with biological sample droplets without the
risk of biofouling [19]. Optical equipment including light emitting diodes (LED),
photodiodes and optical filters have the small footprints required to integrate with
microfluidic systems. Compatibility with fiber optic waveguides has further shown
miniaturization of optical detection in DMF devices [103], [126]. Optical detection is also
well-suited to dairy testing as multiple regulated antibiotics are measurable by UV-Vis
spectroscopy [37], [38], [40], [61]. However, without the utility for separation, DMF
sensors commonly rely on label molecules which selectively bind to the target analyte
and are more readily detected [15], [79], [123].
The use of label molecules is undesirable due to cost and complexity. Label
molecules are highly specific and therefore a new label molecule must be used for each
target analyte [45]. As a result, labelled detection of multiple analytes requires multiple
expensive label molecules which are consumed in each trial. The preparation of label
molecules is also commonly customized to each application and requires trained
personnel in a laboratory environment [127]. This dependence on personnel and lab
facilities for label preparation is not well-suited to on-site applications. Thus, the
development of a label-free method of antibiotic detection in DMF devices will be
valuable.
51
In this chapter we present a label-free optical detection method for antibiotics in
milk droplets. The presented method is enabled by lock-in amplified fluorescence
spectroscopy. Detection is performed in milk droplets without need for chemical
separation thereby demonstrating compatibility with modern DMF devices. This method
is demonstrated by detection of the ciprofloxacin antibiotic in droplets of milk. The
relationship between droplet volume and fluorescent intensity of ciprofloxacin is also
investigated. The results of this chapter demonstrate a viable detection method for
integration in a DMF dairy tester.
4.2 Lock-In Amplification
The lock-in amplifier (LIA) is a scientific instrument designed early as 1941 and
used to measure weak signals from a noisy background in applications where signal-to-
noise ratio (SNR) is extremely low [128]. The LIA is able to take a noisy signal and
compare it to a reference signal to selectively extract the weak signal oscillating at the
reference frequency [129]. The instrument is fundamentally composed of two
components: a phase sensitive detector (PSD), and a low pass filter.
The PSD is the critical component in a LIA. The two input signals to the PSD are
the raw signal and the reference signal. The raw signal is the noisy signal where there
exists some weak component signal of interest at a target frequency. The reference
signal is a separate sinusoidal signal whose frequency is equivalent to the target
frequency. Thus, the general form of the raw signal and reference signal can be
represented by the expressions
52
𝑋(𝑡)𝑟𝑎𝑤 = 𝑉𝑟𝑎𝑤𝑠𝑖𝑛 (2𝜋𝑓𝑟𝑎𝑤𝑡 + 𝜃𝑟𝑎𝑤) (5)
and
𝑋(𝑡)𝑟𝑒𝑓 = 𝑉𝑟𝑒𝑓 𝑠𝑖𝑛(2𝜋𝑓𝑟𝑒𝑓𝑡 + 𝜃𝑟𝑒𝑓) , (6)
where 𝑉 is amplitude, 𝑓 is frequency and 𝜃 is phase shift for the raw and reference
signals respectively. Equation (5) and equation (6) can also be given in the frequency
domain as
𝑋(𝑓)𝑟𝑎𝑤 =𝛿(𝑓 − 𝑓𝑟𝑎𝑤) − 𝛿(𝑓 + 𝑓𝑟𝑎𝑤)
2
(7)
and
𝑋(𝑓)𝑟𝑒𝑓 =𝛿(𝑓 − 𝑓𝑟𝑒𝑓) − 𝛿(𝑓 + 𝑓𝑟𝑒𝑓)
2 ,
(8)
respectively. Note that frequency domain equations are shown without accounting for
phase shift for simplicity. The PSD passes the raw and reference signals through a
frequency mixer. This is analogous to multiplying the time domain signals or performing
a convolution in the frequency domain. The resulting output of the PSD is
53
𝑌(𝑡) =1
2𝑉𝑟𝑎𝑤𝑉𝑟𝑒𝑓[𝑐𝑜𝑠(2𝜋𝑓𝑟𝑎𝑤𝑡 − 2𝜋𝑓𝑟𝑒𝑓𝑡 + 𝜃𝑟𝑎𝑤 − 𝜃𝑟𝑒𝑓)
− 𝑐𝑜𝑠(2𝜋𝑓𝑟𝑎𝑤𝑡 + 2𝜋𝑓𝑟𝑒𝑓𝑡 + 𝜃𝑟𝑎𝑤 + 𝜃𝑟𝑒𝑓)]
(9)
in the time domain or
𝑌(𝑓) =1
4(𝛿(𝑓 − 𝑎) + 𝛿(𝑓 + 𝑎) − 𝛿(𝑓 − 𝑏) − 𝛿(𝑓 + 𝑏)) , (10)
in the frequency domain where 𝑎 = 𝑓𝑟𝑎𝑤 − 𝑓𝑟𝑒𝑓 and 𝑏 = 𝑓𝑟𝑎𝑤 + 𝑓𝑟𝑒𝑓 . Considering the
condition that 𝑓𝑟𝑎𝑤 = 𝑓𝑟𝑒𝑓 , equations (9) and (10) show the output from the PSD has
two frequency components; one at frequency of zero, and another at the sum frequency
of the inputs. This output signal is conditioned by a low pass filter which attenuates
signal components at the sum frequency and its harmonics. Thus, the final output from
the LIA is a DC signal where amplitude is proportional to the reference frequency
amplitude in the raw signal.
4.3 Experimental Setup
A schematic of the experimental setup used in this chapter is shown in Figure 9. A
printed circuit board housing an LED is held by a xyz-stage (i.e., a three axis micro-
adjustable platform structure). The LED is vertically aligned with the sample droplet, an
indium tin oxide (ITO) slide, and a photodiode with an optical filter. To minimize distance
between the LED and photodiode, the LED is positioned as close to the sample droplet
as possible without physical contact.
54
Indium tin oxide is commonly used in DMF optical sensing applications due to its
conductivity and transparency [15]. In these experiments, the ITO slide was prepared by
dispensing a 5 cSt silicone oil droplet (Sigma Aldrich, United States) and spreading
gently by gloved finger. The thin silicone oil layer provided the hydrophobic coating that
would be present in DMF devices such as that presented in Chapter 3. The silicone oil
also allowed sample droplets to form a high contact angle and prevent from spreading
onto the ITO slide.
The photodiode used in these experiments was the DET 36A/M (Thorlabs,
United States) silicon based photodiode. This photodiode is sensitive to light
wavelengths in the range of 350 nm – 1100 nm. The photodiode was operated in
photovoltaic mode to minimize dark current. The optical filter is placed between the ITO
slide and photodiode to block the excitation wavelength and pass the longer fluorescent
wavelength. The LED and optical filters are chosen depending on fluorescent
characteristics of the analyte.
The SR830 LIA (Stanford Research Systems, United States) was used for the
presented lock-in based florescence spectroscopy experiments. The LIA took input
current from the photodiode to its internal trans-impedance amplifier configured with a
gain of 106 V/A. Sensitivity of the LIA was set as 50 mV/nA. Time constant was set as
30 milliseconds. Slope of the internal low pass filter was set as 18 dB.
55
Figure 9: Schematic of the LIA fluorescence spectroscopy experiment. An LED, sample droplet, optical filter and photodiode are aligned in the vertical plane. The LED is driven by square wave from a function generator. TTL output from the function generator is given as reference signal to the LIA. Output current from the photodiode is the raw input to the LIA.
4.4 Rhodamine Pilot Experiment
A pilot experiment using a fluorescent dye called rhodamine B (Sigma Aldrich)
was performed to validate the experimental setup. The rhodamine B fluorescent dye
emits visible light at a wavelength of 627 nm (i.e., red) when excited at 553 nm (i.e.,
green). The rhodamine dye was mixed in aqueous droplets to test concentrations
ranging 0.5 µM to 67 µM. Droplets were pipetted onto the ITO slide at a fixed volume of
5.0 µL. A green LED was driven by a 2500 Hz square wave from a function generator.
The FEL0600 long-pass filter (Thorlabs, United States) with cut-on wavelength of 600
nm was used to pass only fluorescent light onto the photodiode. Between each trial the
sample droplet was removed and a new droplet was pipetted. The photocurrent
56
measured by the LIA versus rhodamine B concentration in aqueous droplets is plotted
in Figure 10.
The plot in Figure 10 shows a positive correlation between fluorescent dye
concentration and photocurrent. The logarithmic trend is characteristic of photodiodes
operated in photovoltaic mode. The response can be made increasingly linear by
reverse-biasing the photodiode and operating in the photoconductive mode although the
drawback is increased dark current. The results of this pilot experiment validated
performance of the experimental setup.
Figure 10: Lock-in amplified photocurrent against rhodamine B concentration in 5.0 µL aqueous droplets. The mean photocurrent and corresponding standard deviation of n = 3 repeated measurements are shown as a function of rhodamine B concentration.
57
4.5 Antibiotic Detection
The following experiments measure concentration of ciprofloxacin in milk
droplets. Ciprofloxacin is a widely used antibiotic to treat cattle and is strictly regulated
in dairy products [130]. Ciprofloxacin hydrochloride (Fisher Scientific, United States),
was dissolved in two percent milk bought from a local vendor to produce solutions of
concentrations ranging 0.01 mM to 1.00 mM. Previous studies have demonstrated
fluorescence of ciprofloxacin to emit at 440 nm (i.e., blue) when excited at 280 nm (i.e.,
ultraviolet) [131], [132]. The LED used in this work was the commercially available
VLMU60CL00-280–125 (Vishay Semiconductors, United States) with peak wavelength
of 280 nm and output power of 2.4 mW. The LED was driven by a 2500 Hz square
wave. The Edmund Optics 86–339 band-pass filter (Edmund Optics, United States)
centered at 440nm with bandwidth of 40 nm was placed above the photodiode. The
photocurrent measured by the LIA is plotted against ciprofloxacin concentration in
Figure 11.
58
Figure 11: Lock-in amplified photocurrent against ciprofloxacin concentration in 5.0 µL milk droplets. The mean photocurrent and corresponding standard deviation of n = 5 repeated measurements are shown as a function of ciprofloxacin concentration.
The results in Figure 11 show an increasing photocurrent with increasing
ciprofloxacin concentration in 5.0 µL milk droplets. The lowest ciprofloxacin
concentration experimentally measured is 0.01 mM. The slope of the response curve in
Figure 11 corresponds to a measurement sensitivity of 1.39 nA/mM. The limit of
detection can then be calculated as the ratio of three times the standard deviation of the
blank reading to detection sensitivity [37]. This yields a limit of detection of 0.015 mM.
This 0.015 mM limit of detection achieved with LIA measurement is improved from a
similar experiment using continuous microfluidics and electrophoresis separation of
filtered milk to achieve a limit of detection of 0.19 mM [58].
59
4.6 Effect of Droplet Volume
The LIA fluorescent detection of ciprofloxacin in milk was performed with varied
droplet volumes to investigate the effect on fluorescent output. These results are shown
in Figure 12. The results showed a general increase in LIA measured photocurrent with
droplet volume which suggested increased fluorescent intensity with droplet volume.
Previous studies have demonstrated that increased fluorescence intensity is not directly
proportional to droplet volume [133]. Instead, the relation between droplet fluorescence
intensity and volume is characterized by a complex interaction between scattering, ray
focusing and absorbance within the droplet [133], [134]. The trend shown in Figure 12 is
very likely to reverse at larger droplet volumes where absorbance of the bulk medium
becomes more dominant and the effect of ray focusing is minimized [134].
The increasing fluorescent intensity of ciprofloxacin with milk droplet volumes on
the microlitre scale is well-suited to open configuration DMF devices such as that
demonstrated in Chapter 3. By comparison to closed configuration DMF, the open
configuration is preferred to handle larger droplet volumes [135]. Thus, the use of an
open configuration DMF device for dairy testing shows advantages in overcoming
biofouling and optimizing fluorescence detection.
60
Figure 12: Lock-in amplified photocurrent against ciprofloxacin concentration in milk droplets of volumes: (a) 7 µL, (b) 9 µL, (c) 11 µL and (d) 13 µL. The mean photocurrents and corresponding standard deviations of n = 3 repeated measurements are shown as a function of ciprofloxacin concentration.
4.7 Chapter Conclusion
In this chapter we demonstrated an LIA fluorescence spectroscopy method for
measuring concentration of ciprofloxacin in milk droplets. This detection method is label-
free and does not require any separation of the antibiotic from the bulk milk droplet.
Using UV-Vis fluorescence spectroscopy, a limit of detection of 0.015 mM for
ciprofloxacin in milk droplets was achieved. At the molar mass of 331.35 g/mol and milk
61
density of 1.03 kg/L [136], this limit of detection translates to 5 ppm which is comparable
to the required regulatory limit of detection of 0.1 ppm [130]. This UV-Vis spectroscopy
detection can be extended to multiple antibiotics regulated in dairy milk [37], [38], [40],
[61].
The presented method demonstrates label-free optical detection which is well-
suited to integration on DMF devices. The trend between droplet volume and
fluorescent intensity also lends itself to potential optimization on a DMF device. Thus,
the results of this work are important to demonstrating a viable method for antibiotic
detection in milk droplets which is necessary for future development of a DMF dairy
testing device.
62
5 Conclusions
This final chapter provides a review of the contributions of this thesis. Each chapter
is briefly summarized. The impact of this experimental work is stated along with
recommendations for future work.
5.1 Summary of Contributions
This thesis demonstrated the fundamental functionalities required by a digital
microfluidic device for on-site dairy testing. The functionalities of automated milk droplet
actuation and antibiotic detection in milk droplets were developed and studied.
The first chapter of this thesis detailed the problem faced by dairy farmers due to
the current practice of antibiotic residue testing. The value of an on-site dairy tester to
use during milk collection was shown to protect the farmer and dairy industry from loss
of money and milk. Current commercially available methods were shown to be unviable
for use during milk collection due to manual processes and long response times. Digital
microfluidic technology was shown to be an appealing solution due to strengths in
automation, parallelization and reconfigurability. The application of digital microfluidics
to dairy testing was met with two immediate challenges in overcoming biofouling to
actuate milk droplets and antibiotic detection in milk droplets. The work scope of this
thesis was to overcome these two immediate challenges and thereby enable the
adaption of digital microfluidics to dairy testing.
The second chapter provided a detailed literature review of the existing technology
used in dairy testing and also about DMF. The importance of antibiotic regulation in
63
dairy milk was discussed. The current testing methods used in laboratories and their
limitations for adaption in on-site solutions were stated. Next, commercially available on-
site dairy sensors and their underlying technologies were detailed. Specifically, the
advancement of traditional microfluidic MCE devices was discussed. These on-site
dairy sensors were shown to be unsuitable for dairy testing during milk collection due to
need for manual procedures, long response times and lack of robustness. DMF
technology and its advantages were introduced. The theoretical actuation of droplets
and complications presented by biofouling were described. DMF compatible detection
methods and the particular appeal of fluorescence spectroscopy for this work were
shown.
The third chapter addressed issues of biofouling and integration faced in
developing a DMF device for dairy testing applications. A fully integrated DMF device
was designed to leverage advantages in automation, parallelization and
reconfigurability. A modular PCB hardware, event driven firmware and versatile control
software was demonstrated in a single integrated DMF device. Experimental studies
measuring droplet velocities were conducted to determine an optimized architecture for
actuation of milk droplets in an air filler medium. The use of a closed DMF configuration
was found unsuitable for bio-fluid actuation due to contact between the sample and
hydrophobic layer of the top plane of the device. Thus, the open configuration was
proven to be the desired configuration. Experiments comparing performance of a gel-
based PDMS dielectric and stretched parafilm dielectric showed the surface roughness
of PDMS was especially disadvantageous in open configurations. The developed DMF
64
device showed automated continuous milk droplet actuation when using an open
configuration, parafilm dielectric and silicone oil coating.
The fourth chapter addressed the issue of antibiotic detection in milk droplets. This
issue was the second required functionality of a digital microfluidic device for dairy
testing. Digital microfluidic technology was shown to be disadvantaged compared to
traditional microfluidics in the domain of sample separation. Digital microfluidic detection
applications were shown to often rely on label molecules which are undesirable due to
cost and complexity. To address this, a lock-in amplifier instrument was used to
enhance the fluorescence spectroscopy detection method which is widely compatible
with digital microfluidic devices. Using lock-in amplified UV-Vis fluorescence
spectroscopy, the ciprofloxacin antibiotic was detected label-free in milk droplets without
any chemical separation. Experiments investigating fluorescent intensity by droplet
volume showed a trend favourable for open configuration digital microfluidics. Hence, a
method for antibiotic detection in milk droplets compatible with digital microfluidics was
demonstrated.
In summary, the work of this thesis has successfully demonstrated the most
fundamental functionalities, with respect to actuation and detection, for adaption of
digital microfluidics to dairy testing applications. Thus, this thesis provides the
foundation for designing a digital microfluidic dairy device for antibiotic detection.
65
5.2 Future Works
The linearity and sensitivity of the fluorescent intensity versus antibiotic
concentration response shown in Chapter 4 has potential for improvement by operating
the photodiode in photoconductive mode. By reverse-biasing the photodiode, the width
of the semiconductor depletion region is increased. Consequently, the response has
greater sensitivity and linearity. Operation in photoconductive mode often has the
drawback of greater dark current and therefore a greater noise floor. However, the dark
current effect may be negated by the lock-in amplified measurement.
Having had this thesis demonstrate milk droplet actuation and milk droplet antibiotic
detection on separated systems, an intriguing future work will be the combination of
these functionalities in a single device. A custom fixture should be developed to
minimize the distance between an LED, milk droplet and photodiode. Use of multiple
LEDs and photodiodes has potential for detecting multiple regulated antibiotics in milk
as has been done in some traditional microfluidic devices [38]–[40], [61]. This proposed
future work will be very impactful in realizing a complete digital microfluidic on-site dairy
tester.
66
REFERENCES
[1] D. F. of Ontario, Raw Milk Quality Programs Procedure. Mississauga, Ontario: Dairy Farmers of Ontario, 2020.
[2] J. M. Mitchell, M. W. Griffiths, S. A. McEwen, W. B. McNab, and A. J. Yee, “Antimicrobial Drug Residues in Milk and Meat: Causes, Concerns, Prevalence, Regulations, Tests, and Test Performance,” J. Food Prot., vol. 61, no. 6, pp. 742–756, 1998.
[3] D. F. of Ontario, Dairy Statistical Handbook 2015-16. Mississauga, Ontario: Dairy Farmers of Ontario, 2016.
[4] D. F. of Ontario, 2018-2019 Annual Report. Mississauga, Ontario: Dairy Farmers of Ontario, 2019.
[5] S. Sachi, J. Ferdous, M. H. Sikder, and S. M. A. Hussan, “Antibiotic residues in milk: Past, present, and future,” J. Adv. Vet. Anim. Res., vol. 6, no. 3, pp. 315–332, 2019.
[6] R. L. Althaus, A. Torres, A. Montero, S. Balasch, and M. P. Molina, “Detection Limits of Antimicrobials in Ewe Milk by Delvotest Photometric Measurements,” J. Dairy Sci., vol. 86, pp. 457–463, 2003.
[7] C. Bion, A. Beck Henzelin, Y. Qu, G. Pizzocri, G. Bolzoni, and E. Buffoli, “Analysis of 27 antibiotic residues in raw cow’s milk and milk-based products – validation of Delvotest® T.,” Food Addit. Contam. Part A, vol. 3, no. 1, pp. 54–59, 2016.
[8] R. D. Watters, R. M. Bruckmaier, H. M. Crawford, N. Schuring, Y. H. Schukken, and D. M. Galton, “The effect of manual and mechanical stimulation on oxytocin release and milking characteristics in Holstein cows milked 3 times daily,” J. Dairy Sci., vol. 98, no. 3, pp. 1721–1729, 2015, doi: http://dx.doi.org/10.3168/jds.2014-8335.
[9] M. C. Beltrán Martínez, R. Rueda, T. Althaus, R. Lisandro, and M. Pons, “Evaluation of the Charm maximum residue limit beta-lactam and tetracycline test for the detection of antibiotics in ewe and goat milk,” J. Dairy Sci., vol. 96, no. 5, pp. 2737–2745, 2013, doi: http://dx.doi.org/10.3168/jds.2012-6044.
[10] W. Jiang et al., “Development of a multiplex flow-through immunoaffinity chromatography test for the on-site screening of 14 sulfonamide and 13 quinolone residues in milk,” Biosens. Bioelectron., vol. 66, pp. 124–128, 2015, doi: 10.1016/j.bios.2014.11.004.
67
[11] M. Alistar and U. Gaudenz, “Opendrop: An integrated do-it-yourself platform for personal use of biochips,” Bioengineering, 2017, doi: 10.3390/bioengineering4020045.
[12] R. Fobel, C. Fobel, and A. R. Wheeler, “DropBot: An open-source digital microfluidic control system with precise control of electrostatic driving force and instantaneous drop velocity measurement,” Appl. Phys. Lett., vol. 102, no. 19, p. 193513, 2013.
[13] A. R. Vollertsen et al., “Modular operation of microfluidic chips for highly parallelized cell culture and liquid dosing via a fluidic circuit board,” Microsystems Nanoeng., vol. 6, no. 1, 2020.
[14] E. Samiei, M. Tabrizian, and M. Hoorfar, “A review of digital microfluidics as portable platforms for lab-on a-chip applications,” Lab Chip, vol. 16, pp. 2376–2396, 2016, doi: http://dx.doi.org/10.1039/c6lc00387g.
[15] M. G. Pollack, V. K. Pamula, V. Srinivasan, and A. E. Eckhardt, “Applications of electrowetting-based digital microfluidics in clinical diagnostics,” Expert Rev. Mol. Diagn., vol. 11, no. 4, pp. 393–407, 2011.
[16] M. G. Pollack, R. B. Fair, and A. D. Shenderov, “Electrowetting-based actuation of liquid droplets for microfluidic applications,” Appl. Phys. Lett., vol. 77, no. 11, pp. 1725–1726, Sep. 2000, doi: 10.1063/1.1308534.
[17] M. G. Pollack, A. D. Shenderov, and R. B. Fair, “Electrowetting-based actuation of droplets for integrated microfluidics,” Lab Chip, vol. 2, pp. 96–101, 2002.
[18] C. M. Collier, M. Wiltshire, J. Nichols, B. Born, E. L. Landry, and J. F. Holzman, “Nonlinear dual-phase multiplexing in digital microfluidic architectures,” Micromachines, 2011, doi: 10.3390/mi2040369.
[19] E. Samiei, G. S. Luka, H. Najjaran, and M. Hoorfar, “Integration of biosensors into digital microfluidics: impact of hydrophilic surface of biosensors on droplet manipulation,” Biosens. Bioelectron., vol. 81, pp. 480–486, 2016.
[20] G. Perry, V. Thomy, M. R. Das, Y. Coffinier, and R. Boukherroub, “Inhibiting protein biofouling using graphene oxide in droplet-based microfluidic microsystemsn,” Lab Chip, vol. 12, no. 9, pp. 1601–1604, 2012.
[21] V. N. Luk, G. C. H. Mo, and A. R. Wheeler, “Pluronic Additives: A Solution to Sticky Problems in Digital Microfluidics,” Langmuir, vol. 24, no. 12, pp. 6382–6389, 2008.
[22] S. IH, K. I. Loken, and H. HH, “Antibiotic residues in milk transferred from treated to untreated quarters in dairy cattle,” J. Am. Vet. Med. Assoc., vol. 146, pp. 589–
68
593, 1965.
[23] T. P. Van Boeckela et al., “Global trends in antimicrobial use in food animals,” Proc. Natl. Acad. Sci. U. S. A., vol. 112, no. 18, pp. 5649–5654, 2015.
[24] F. Conzuelo et al., “Rapid screening of multiple antibiotic residues in milk using disposable amperometric magnetosensors,” Anal. Chim. Acta, vol. 820, pp. 32–38, 2014.
[25] C. Capita and C. Alonso-Calleja, “Antibiotic-resistant bacteria: a challenge for the food industry,” Crit. Rev. Food Sci. Nutr., vol. 53, no. 1, pp. 11–48, 2013.
[26] A. Nisha, “Antibiotic Residues - A Global Health Hazard,” Vet. World, vol. 2, no. 2, pp. 375–377, 2008.
[27] B. K. Khanal, M. B. Sadiq, M. Singh, and A. K. Anal, “Screening of antibiotic residues in fresh milk of Kathmandu Valley, Nepal,” J. Environ. Sci. Heal. Part B, vol. 53, no. 1, pp. 57–86, 2018.
[28] M. M. Moghadam, M. Amiri, H. R. Riabi, and H. R. Riabi, “Evaluation of Antibiotic Residues in Pasteurized and Raw Milk Distributed in the South of Khorasan-e Razavi Province, Iran,” J. Clin. Diagnostic Res., vol. 10, no. 12, pp. 31–35, 2016.
[29] H. Canada, List of Maximum Residue Limits (MRLs) for Veterinary Drugs in Foods. Government of Canada, 2018.
[30] S. Gao et al., “Ionic liquid-based homogeneous liquid–liquid microextraction for the determination of antibiotics in milk by high-performance liquid chromatography,” J. Chromatogr. A, vol. 1218, no. 41, pp. 7254–7263, 2011.
[31] R. V Pereira, J. D. Siler, R. C. Bicalho, and L. D. Warnick, “Multiresidue screening of milk withheld for sale at dairy farms in central New York State,” J. Dairy Sci., vol. 97, no. 3, pp. 1513–1519, 2014.
[32] C. Díez et al., “Aminoglycoside analysis in food of animal origin with a zwitterionic stationary phase and liquid chromatography–tandem mass spectrometry,” Anal. Chim. Acta, vol. 882, pp. 127–139, 2015.
[33] X. Zhang and X. Cai, “Rapid simultaneous determination of 53 beta-lactam antibiotics and their metabolites in milk by ultra-performance liquid chromatography coupled with triple quadrupole mass spectrometry,” Chinese J. Chromatogr., vol. 32, no. 7, pp. 693–701, 2014.
[34] M.-C. Verdier, O. Tribut, P. Tattevin, Y. Le Tulzo, C. Michelet, and D. le Bentue´-Ferrer, “Simultaneous Determination of 12 β-Lactam Antibiotics in Human Plasma by High-Performance Liquid Chromatography with UV Detection: Application to
69
Therapeutic Drug Monitoring,” Antimicrob. Agents Chemother., vol. 55, no. 10, pp. 4873–4879, 2011.
[35] Q. Wang et al., “Simultaneous determination of zeranols and chloramphenicol in foodstuffs of animal origin by combination immunoaffinity column clean-up and liquid chromatography-tandem mass spectrometry,” Chinese J. Chromatogr., vol. 32, no. 6, pp. 640–646, 2014.
[36] I. Meisen et al., “Normal silica gel and reversed phase thin-layer chromatography coupled with UV spectroscopy and IR-MALDI-o-TOF-MS for the detection of tetracycline antibiotics,” Anal. Bioanal. Chem., vol. 398, pp. 2821–2831, 2010.
[37] G. Mu, H. Liu, L. Xu, L. Tian, and F. Luan, “Matrix solid-phase dispersion extraction and capillary electrophoresis determination of tetracycline residues in milk,” Food Anal. Methods, vol. 5, no. 1, pp. 148–153, 2011.
[38] V. H. Springer and A. G. Lista, “Micellar nanotubes dispersed electrokinetic chromatography for the simultaneous determination of antibiotics in bovine milk,” Electrophoresis, vol. 33, no. 13, pp. 2049–2055, 2012.
[39] L. Vera-Candioti, A. C. Olivieri, and H. C. Goicoechea, “Development of a novel strategy for preconcentration of antibiotic residues in milk and their quantitation by capillary electrophoresis,” Talanta, vol. 82, no. 1, pp. 213–221, 2010.
[40] S. M. Santos, M. Henriques, A. C. Duarte, and V. I. Esteves, “Development and application of a capillary electrophoresis based method for the simultaneous screening of six antibiotics in spiked milk samples,” Talanta, vol. 71, no. 2, pp. 731–737, 2007.
[41] F. J. Lara, A. M. García-Campaña, F. Alés-Barrero, J. M. Bosque-Sendra, and L. E. García-Ayuso, “Multiresidue Method for the Determination of Quinolone Antibiotics in Bovine Raw Milk by Capillary Electrophoresis−Tandem Mass Spectrometry,” Anal. Chem., vol. 78, no. 22, pp. 7665–7673, 2006.
[42] C. Blasco, Y. Picó, and V. Andreu, “Analytical method for simultaneous
determination of pesticide and veterinary drug residues in milk by CE‐MS,” Electrophoresis, vol. 30, no. 10, pp. 1698–1707, 2009.
[43] D. Moreno-González, A. M. Hamed, B. Gilbert-López, L. Gámiz-Gracia, and A. M. García-Campaña, “Evaluation of a multiresidue capillary electrophoresis-quadrupole-time-of-flight mass spectrometry method for the determination of antibiotics in milk samples,” J. Chromatogr. A, vol. 1510, pp. 100–107, 2017.
[44] P. B. Luppa, L. J. Sokoll, and D. W. Chan, “Immunosensors—principles and applications to clinical chemistry,” Clin. Chim. acta, vol. 314, no. 1, pp. 1–26, 2001.
70
[45] H. K. Hunt and A. M. Armani, “Label-free biological and chemical sensors,” Nanoscale, vol. 2, no. 9, pp. 1544–1559, 2010.
[46] Z. Wang, R. C. Beier, and J. Shen, “Immunoassays for the detection of macrocyclic lactones in food matrices - A review,” Trends Anal. Chem., vol. 92, pp. 42–61, 2017.
[47] E. Song et al., “Multi-color quantum dot-based fluorescence immunoassay array for simultaneous visual detection of multiple antibiotic residues in milk,” Biosens. Bioelectron., vol. 72, no. 15, pp. 320–325, 2015, doi: 10.1016/j.bios.2015.05.018.
[48] V. Gaudin, C. Hedou, C. Soumet, and E. Verdon, “A review of enhancers for chemiluminescence enzyme immunoassay,” Food Addit. Contam., vol. 35, no. 12, pp. 2348–2365, 2018.
[49] W. Yu, M. Knauer, C. Kunas, U. Acaroz, R. Dietrich, and E. Mrtlbauer, “A gold nanoparticles growth-based immunoassay for detection of antibiotic residues,” Anal. Methods, vol. 9, no. 2, pp. 188–191, 2017.
[50] Q. Shi et al., “A SERS-based multiple immuno-nanoprobe for ultrasensitive detection of neomycin and quinolone antibiotics via a lateral flow assay,” Microchim. Acta, vol. 185, no. 2, pp. 1–8, 2018.
[51] Y. Chen et al., “Near-infrared fluorescence-based multiplex lateral flow immunoassay for the simultaneous detection of four antibiotic residue families in milk,” Biosens. Bioelectron., vol. 219, pp. 430–434, 2016.
[52] F. Ceyssens, D. Witters, T. Van Grimbergen, K. Knez, J. Lammertyn, and R. Puers, “Integrating optical waveguides in electrowetting-on-dielectric digital microfluidic chips,” Sensors Actuators, B Chem., vol. 191, pp. 166–171, 2013, doi: 10.1016/j.snb.2013.01.078.
[53] L. Naik, R. Sharma, B. Mann, K. Lata, Y. S. Rajput, and B. Surendra Nath, “Rapid screening test for detection of oxytetracycline residues in milk using lateral flow assay,” Food Chem., 2017, doi: 10.1016/j.foodchem.2016.09.090.
[54] J. Peng, Y. Wang, L. Liu, H. Kuang, A. Li, and C. Xu, “Multiplex lateral flow immunoassay for five antibiotics detection based on gold nanoparticle aggregations,” RSC Adv., vol. 6, no. 10, pp. 1–8, 2016.
[55] M. Abouzied et al., “Validation Study of the BetaStar Plus Lateral Flow Assay for Detection of Beta-Lactam Antibiotics in Milk,” J. AOAC Int., vol. 95, no. 4, pp. 1211–1221, 2012.
[56] M. Han et al., “An octuplex lateral flow immunoassay for rapid detection of antibiotic residues, aflatoxin M1 and melamine in milk,” Sensors Actuators B
71
Chem., vol. 292, pp. 94–104, 2016.
[57] R. Bosma, “Detection of Ciprofloxacin Antibiotics in Milk Through Microchip Capillary Electrophoresis Actuation and Fluorescence Spectroscopy Sensing,” University of Guelph, 50 Stone Road East, 2019.
[58] R. Bosma, J. Devasagayam, A. Singh, and C. M. Collier, “Microchip capillary electrophoresis dairy device using fluorescence spectroscopy for detection of ciprofloxacin in milk samples,” Sci. Rep., vol. 10, no. 1, p. 13548, 2020.
[59] R. Bosma, J. Devasagayam, R. Eswar, I. de F. Albuquerque, and C. M. Collier, “Voltage control for microchip capillary electrophoresis analyses,” Electrophoresis, vol. 41, no. 23, pp. 1961–1968, 2020.
[60] H. Sun, W. Zhao, and P. He, “Effective Separation and Simultaneous Determination of Four Fluoroquinolones in Milk by CE with SPE,” Chromatographia, vol. 68, no. 5, pp. 425–429, 2008.
[61] S. Wang, P. Yang, and Y. Cheng, “Analysis of tetracycline residues in bovine milk by CE‐MS with field‐amplified sample stacking,” Electrophoresis, vol. 28, no. 22, pp. 4173–4179, 2007.
[62] V. Adam and M. Vaculovicova, “Capillary electrophoresis and nanomaterials– Part I: Capillary electrophoresis ofnanomaterials,” Electrophoresis, vol. 38, pp. 2389–2404, 2017.
[63] J. L. Felhofer, L. Blanes, and C. D. Garcia, “Recent developments in instrumentation for capillary electrophoresis and microchipcapillary electrophoresis,” Electrophoresis, vol. 31, no. 15, pp. 2469–2486, 2010.
[64] M. R. Hossan, D. Dutta, N. Islam, and P. Dutta, “Review: Electric field driven pumping inmicrofluidic device,” Electrophoresis, vol. 39, pp. 702–73, 2018.
[65] Y. Ding, L. Bai, X. Suo, and X. Meng, “Post separation adjustment of pH to enable the analysis of aminoglycoside antibiotics by microchip electrophoresis with amperometric detection,” Electrophoresis, vol. 33, pp. 3245–3253, 2012.
[66] G. Suarez et al., “Lab-on-a-chip for multiplexed biosensing of residual antibiotics in milk,” Lab Chip, vol. 9, no. 11, pp. 1625–1630, 2009.
[67] C. Long, B. Deng, S. Sun, and S. Meng, “Simultaneous determination of chlortetracycline, ampicillin and sarafloxacin in milk using capillary electrophoresis with electrochemiluminescence detection,” Food Addit. Contam. Part A, vol. 34, no. 1, pp. 24–31, 2017.
[68] V. Jain, V. Devarasetty, and R. Patrikar, “Effect of electrode geometry on droplet
72
velocity in open EWOD based device for digital microfluidics applications,” J. Electrostat., 2017, doi: 10.1016/j.elstat.2017.02.006.
[69] V. ; R. T. P. ; R. T. P. ; D. R. ; D. R. ; P. R. ; P. R. H. W.-H. ; H. W.-H. Jain Vandana ; Jain, “Design, fabrication and characterization of low cost printed circuit board based EWOD device for digital microfluidics applications,” Microsyst. Technol., vol. 23, no. 2, pp. 389–397, 2017.
[70] T. B. Jones, “On the Relationship of Dielectrophoresis and Electrowetting,” Langmuir, vol. 18, no. 11, pp. 4437–4443, 2002.
[71] C. M. Collier, K. A. Hill, and J. F. Holzman, “A dielectrophoresis microjet for on-chip technologies,” RSC Adv., vol. 3, no. 45, p. 23309, 2013.
[72] T. P. Hunt, D. Issadore, and R. M. Westervelt, “Integrated circuit/microfluidic chip to programmably trap and move cells and droplets with dielectrophoresis,” Lab Chip, vol. 8, no. 1, pp. 81–87, 2008.
[73] S.-K. Fan, T.-H. Hsieh, and D.-Y. Lin, “General digital microfluidic platform manipulating dielectric and conductive droplets by dielectrophoresis and electrowetting,” Lab Chip, vol. 9, no. 9, pp. 1236–1242, 2009.
[74] D. Chatterjee, B. Hetayothin, A. R. Wheeler, D. J. K. King, and R. L. Garrell, “Droplet-based microfluidics with nonaqueous solvents and solutions,” Lab Chip, vol. 6, no. 2, pp. 199–206, 2006.
[75] C. E. Giacomelli and W. Norde, “Influence of Hydrophobic Teflon Particles on the Structure of Amyloid β-Peptide,” Biomacromolecules, vol. 4, no. 6, pp. 1719–1726, 2003.
[76] J.-Y. Yoon and R. L. Garrell, “Preventing Biomolecular Adsorption in Electrowetting-Based Biofluidic Chips,” Anal. Chem., vol. 75, no. 19, pp. 5097–5102, 2003.
[77] T. N. Krupenkin, J. A. Taylor, T. M. Schneider, and S. Yang, “From Rolling Ball to Complete Wetting: The Dynamic Tuning of Liquids on Nanostructured Surfaces,” Langmuir, vol. 20, no. 10, pp. 3824–3827, 2004.
[78] V. Srinivasan, V. K. Pamula, and R. B. Fair, “An integrated digital microfluidic lab-on-a-chip for clinical diagnostics on human physiological fluids,” Lab Chip, vol. 4`, pp. 310–315, 2004, doi: 10.1039/b403341h.
[79] R. Sista et al., “Development of a digital microfluidic platform for point of care testing,” Expert Rev. Mol. Diagn., vol. 8, no. 12, pp. 2091–2104, 2008.
[80] H. Moon, S. K. Cho, R. L. Garrell, and C.-J. Kim, “Low voltage electrowetting-on-
73
dielectric,” J. Appl. Phys., vol. 92, no. 7, pp. 4080–4087, 2002.
[81] R. Karoui and C. Blecker, “Fluorescence Spectroscopy Measurement for Quality Assessment of Food Systems—a Review,” Food Bioprocess Technol, vol. 4, no. 3, pp. 364–386, 2011.
[82] S. T. Hess, S. Huang, A. A. Heikal, and W. W. Webb, “Biological and Chemical Applications of Fluorescence Correlation Spectroscopy: A Review,” Biochemistry, vol. 41, no. 3, pp. 697–705, 2002.
[83] V. Srinivasan, V. K. Pamula, and R. B. Fair, “An integrated digital microfluidic lab-on-a-chip for clinical diagnostics on human physiological fluids,” Lab Chip, 2004, doi: 10.1039/b403341h.
[84] P. A. L. Wijethunga, Y. S. Nanayakkara, P. Kunchala, D. W. Armstrong, and H. Moon, “On-chip drop-to-drop liquid microextraction coupled with real-time concentration monitoring technique,” Anal. Chem., vol. 83, no. 5, pp. 1658–1664, 2011.
[85] K. Choi, J. M. Mudrik, and A. R. Wheeler, “A guiding light: spectroscopy on digital microfluidic devices using in-planeoptical fibre waveguides,” Anal. Bioanal. Chem., vol. 407, no. 24, pp. 7467–7475, 2015.
[86] C. L. Wong and M. Olivo, “Surface Plasmon Resonance Imaging Sensors: A Review,” Lab Chip, vol. 9, no. 4, pp. 809–824, 2014.
[87] F. Fernández, D. G. Pinacho, F. Sanchez-Baeza, and M.-P. Marco, “Portable Surface Plasmon Resonance Immunosensor for the Detection ofFluoroquinolone Antibiotic Residues in Milk,” J. Agric. Food Chem., vol. 59, no. 9, pp. 5036–5043, 2011.
[88] X. D. Hoa, A. G. Kirk, and M. Tabrizian, “Towards integrated and sensitive surface plasmon resonance biosensors: A review of recent progress,” Biosens. Bioelectron., vol. 23, no. 2, pp. 151–160, 2007.
[89] E. Gustavsson, J. Degelaen, P. Bjurling, and A. Sternesjo, “Determination of β-Lactams in Milk Using a Surface Plasmon Resonance-Based Biosensor,” J. Agric. Food Chem., vol. 52, no. 10, pp. 2791–2796, 2004.
[90] F. Fernández, K. Hegnerová, M. Piliarik, F. Sanchez-Baeza, J. Homola, and M.-P. Marco, “A label-free and portable multichannel surface plasmon resonance im-munosensor for on site analysis of antibiotics in milk samples,” Biosens. Bioelectron., vol. 26, no. 4, pp. 1231–1238, 2010.
[91] L. Malic, T. Veres, and M. Tabrizian, “Two-dimensional droplet-based surface plasmon resonance imaging using electrowetting-on-dielectric microfluidics,” Lab
74
Chip, vol. 9, no. 3, pp. 473–475, 2009.
[92] F. Fasmin and R. Srinivasan, “Review—Nonlinear Electrochemical Impedance Spectroscopy,” J. Electrochem. Soc., vol. 164, no. 7, pp. 443–455, 2017.
[93] T. Lederer, S. Clara, B. Jakoby, and W. Hilber, “Integration of impedance spectroscopy sensors in a digital microfluidic platform,” in Microsystem Technologies, 2012, doi: 10.1007/s00542-012-1464-6.
[94] D. R. Santos et al., “Label-Free Detection of Biomolecules in Microfluidic Systems Using On-Chip UV and Impedimetric Sensors,” IEEE Sensors, vol. 19, no. 18, pp. 7803–7812, 2011.
[95] A. H. C. Ng, K. Choi, R. P. Luoma, J. M. Robinson, and A. R. Wheeler, “Digital Microfluidic Magnetic Separation for Particle-Based Immunoassays,” Anal. Chem., vol. 84, no. 20, pp. 8805–8812, 2012.
[96] A. Farzbod and H. Moon, “Integration of reconfigurable potentiometric electrochemical sensors into a digital microfluidic platform,” Biosens. Bioelectron., vol. 106, pp. 37–42, 2018.
[97] L. Su, W. Jia, C. Hou, and Y. Lei, “Microbial biosensors: A review,” Biosens. Bioelectron., vol. 6, no. 5, pp. 1788–1799, 2011.
[98] M. D. M. Dryden, D. D. G. Rackus, M. H. Shamsi, and A. R. Wheeler, “Integrated digital microfluidic platform for voltammetric analysis,” Anal. Chem., vol. 85, no. 18, pp. 8809–8816, 2013.
[99] N. Ruecha et al., “Paper-Based Digital Microfluidic Chip for Multiple Electrochemical Assay Operated by a Wireless Portable Control System,” Adv. Mater. Technol., vol. 2, no. 3, pp. 1–8, 2017.
[100] F. Conzuelo, M. Gamella, S. Campuzano, A. J. Reviejo, and J. M. Pingarrón, “Disposable amperometric magneto-immunosensor for direct detection of tetracyclines antibiotics residues in milk,” Anal. Chim. Acta, vol. 737, pp. 29–36, 2012.
[101] S. Prasad, M. Del Rosso, J. R. Vale, and C. M. Collier, “Optofluidic lenses with horizontal-to-vertical aspect ratios in the subunit regime,” Appl. Opt., vol. 57, no. 19, pp. 5474–5482, 2018, doi: 10.1364/ao.57.005474.
[102] M. Del Rosso, S. Lepard, R. Eswar, and C. M. Collier, “Microfluidic fabrication with silver nanowires for optofluidic structures with three-dimensional operation,” Sensors actuators. A. Phys., vol. 315, p. 112297, 2020.
[103] I. Spotts, D. Ismail, N. Jaffar, and C. M. Collier, “Fibre-optic sensing in digital
75
microfluidic devices,” Sensors Actuators, A Phys., vol. 280, pp. 164–169, Sep. 2018, doi: 10.1016/j.sna.2018.07.039.
[104] A. Trautmann, G.-L. Roth, B. Nujiqi, T. Walther, and R. Hellmann, “Towards a versatile point-of-care system combining femtosecond laser generated microfluidic channels and direct laser written microneedle arrays,” Microsystems Nanoeng., vol. 5, no. 1, 2019.
[105] Q. Zhang et al., “A feedback-controlling digital microfluidic fluorimetric sensor device for simple and rapid detection of mercury (II) in costal seawater,” Mar. Pollut. Bull., vol. 144, pp. 20–27, 2019.
[106] J. Zhai et al., “A digital microfluidic system with 3D microstructures for single-cell culture,” Microsystems Nanoeng., vol. 6, no. 1, pp. 1–10, 2020.
[107] W. Liu, Z. Zhang, and Z. Liu, “Determination of -lactam antibiotics in milk using micro-flow chemiluminescence system with on-line solid phase extraction,” Anal. Chim. Acta, vol. 592, pp. 187–192, 2007.
[108] H. R. Holmes and K. F. Böhringer, “Transporting droplets through surface anisotropy,” Microsystems Nanoeng., vol. 1, no. 1, p. 15022, 2015.
[109] O. A. Arikan, “Degradation and metabolization of chlortetracycline during the anaerobic digestion of manure from medicated calves,” J. Hazard. Mater., vol. 158, no. 2, pp. 485–490, 2008.
[110] M. M. Hassan, K. Bin Amin, A. Mahabub Alam, S. Al Faruk, and I. Uddin, “Antimicrobial Resistance Pattern against E. coli and Salmonella in Layer Poultry,” Res. J. Vet. Pract., vol. 2, no. 2, pp. 30–35, 2014.
[111] H. Zhang, Y. Ren, and X. Bao, “Simultaneous determination of (fluoro)quinolones antibacterials residues in bovine milk using ultra performance liquid chromatography–tandem mass spectrometry,” J. Pharm. Biomed. Anal., vol. 49, no. 2, pp. 367–374, 2009.
[112] V. Gaudin, C. Hedou, C. Soumet, and E. Verdon, “Multiplex immunoassay based on biochip technology for the screening of antibiotic residues in milk: validation according to the European guideline,” Food Addit. Contam., vol. 35, no. 12, pp. 2348–2365, 2018.
[113] S. K. Cho, H. Moon, and C. J. Kim, “Creating, transporting, cutting, and merging liquid droplets by electrowetting-based actuation for digital microfluidic circuits,” J. Microelectromechanical Syst., vol. 12, no. 1, pp. 70–80, 2003, doi: 10.1109/JMEMS.2002.807467.
[114] J. Berthier, P. Dubois, P. Clementz, P. Claustre, C. Peponnet, and Y. Fouillet,
76
“Actuation potentials and capillary forces in electrowetting based microsystems,” Sensors Actuators A, vol. 134, no. 2, pp. 471–479, 2007.
[115] Y. Guan, B. Li, and L. Xing, “Numerical investigation of electrowetting-based droplet splitting in closed digital microfluidic system: Dynamics, mode, and satellite droplet,” Phys. fluids, vol. 30, no. 11, p. 112001, 2018.
[116] C. Karuwan et al., “Electrochemical detection on electrowetting-on-dielectric digital microfluidic chip,” Talanta, vol. 84, no. 5, pp. 1384–1389, 2011, doi: 10.1016/j.talanta.2011.03.073.
[117] T. I. for Interconnecting and P. E. Circuits, IPC-2221: Generic Standard on Printed Board Design. Institute of Printed Circuits, 1998.
[118] N. Kumari, V. Bahadur, and S. V. Garimella, “Electrical actuation of electrically conducting and insulating droplets using ac and dc voltages,” J. Micromechanics Microengineering, vol. 18, no. 10, Oct. 2008, doi: 10.1088/0960-1317/18/10/105015.
[119] S. Dash, N. Kumari, and S. V Garimella, “Frequency-dependent transient response of an oscillating electrically actuated droplet,” J. micromechanics microengineering, vol. 22, no. 7, p. 75004, 2012.
[120] M. A. Murran and H. Najjaran, “Direct current pulse train actuation to enhance droplet control in digital microfluidics,” Appl. Phys. Lett., vol. 101, no. 14, p. 144102, 2012.
[121] T. Hoang et al., “Minimal microfabrication required digital microfluidic system toward point-of-care nucleic acid amplification test application for developing countries,” Microsyst. Technol., vol. 26, no. 9, pp. 1863–1873, 2020.
[122] K. F. Böhringer, “Modeling and Controlling Parallel Tasks in Droplet-Based Microfluidic Systems,” IEEE Trans. Comput. Des. Integr. Circuits Syst., vol. 25, no. 2, pp. 334–344, 2006.
[123] D. Millington, S. Norton, R. Singh, R. Sista, V. Srinivasan, and V. Pamula, “Digital microfluidics comes of age: high-throughput screening to bedside diagnostic testing for genetic disorders in newborns,” Expert Rev. Mol. Diagn., vol. 18, no. 8, pp. 701–712, 2018.
[124] I. Barbulovic-Nad, S. H. Au, and A. R. Wheeler, “A microfluidic platform for complete mammalian cell culture,” Lab Chip, vol. 10, no. 12, pp. 1536–1542, 2010.
[125] Y. Zhao, M. Tang, F. Liu, H. Li, H. Wang, and D. Xu, “Highly Integrated Microfluidic Chip Coupled to Mass Spectrometry for Online Analysis of Residual
77
Quinolones in Milk,” Anal. Chem., vol. 91, no. 21, pp. 13418–13426, 2019.
[126] K. Choi, J. M. Mudrik, and A. R. Wheeler, “A guiding light: spectroscopy on digital microfluidic devices using in-plane optical fibre waveguides,” Anal. Bioanal. Chem., vol. 407, no. 24, pp. 7467–7475, Sep. 2015, doi: 10.1007/s00216-015-8913-x.
[127] K. A. Neilson et al., “Less label, more free: Approaches in label-free quantitative mass spectrometry,” Proteomics, vol. 11, no. 4, pp. 535–553, 2010.
[128] W. C. Michels and N. L. Curtis, “A Pentode Lock‐In Amplifier of High Frequency Selectivity,” Rev. Sci. Instrum., vol. 12, no. 9, pp. 444–447, 1941.
[129] Y. Sukekawa and T. Nakamoto, “Odor biosensor system based on image lock‐in measurement for odorant discrimination,” Electron. Commun. Japan, vol. 102, no. 4, pp. 57–64, 2019.
[130] A. Aresta, P. Cotugno, and C. Zambonin, “Determination of Ciprofloxacin, Enrofloxacin, and Marbofloxacin in Bovine Urine, Serum, and Milk by Microextraction by a Packed Sorbent Coupled to Ultra-High Performance Liquid Chromatography,” Anal. Lett., vol. 52, no. 5, pp. 790–802, 2019.
[131] R. Yang, Y. Fu, L. Di Li, and J. M. Liu, “Medium effects on fluorescence of ciprofloxacin hydrochloride,” Spectrochim. Acta - Part A Mol. Biomol. Spectrosc., 2003, doi: 10.1016/S1386-1425(03)00059-3.
[132] T. Mahmood, M. Abbas, S. Ilyas, N. Afzal, and R. Nawaz, “Quantification of Fluoroquinolone (enrofloxacin, norfloxacin and ciprofloxacin) residues in cow milk,” IJCBS, vol. 10, pp. 10–15, 2016.
[133] R. Domann, Y. Hardalupas, and A. R. Jones, “A study of the influence of absorption on the spatial distribution of fluorescence intensity within large droplets using Mie theory, geometrical optics and imaging experiments,” Meas. Sci. Technol., vol. 13, no. 3, pp. 280–291, 2002.
[134] B. Frackowiak and C. Tropea, “Numerical analysis of diameter influence on droplet fluorescence,” Appl. Opt., vol. 49, no. 12, pp. 2363–2370, 2010, doi: 10.1364/AO.49.002363.
[135] I. Moon and J. Kim, “Using EWOD (electrowetting-on-dielectric) actuation in a micro conveyor system,” Sensors actuators. A. Phys., vol. 130, pp. 537–544, 2006.
[136] P. Walstra, J. T. M. Wouters, and T. J. Geurts, Dairy Science and Technology, Second Edition. Boca Raton, Florida: CRC/Taylor & Francis, 2006.