development of a miniaturized chromatographic analytical device … · requiring high-throughput...
TRANSCRIPT
Development of a miniaturized chromatographic analytical
device for food safety applications
Catarina Isabel Pereira Bombaça
Thesis to obtain the Master of Science Degree in
Biotechnology
Supervisors: Prof. Dr. João Pedro Estrela Conde and Dra. Ana Margarida Fortes
Examination Committee
Chairperson: Prof. Dr. Luís Joaquim Pina da Fonseca
Supervisor: Prof. Dr. João Pedro Estrela Conde
Members of the Committee: Dr. Ana Margarida Nunes da Mata Pires de Azevedo
November 2017
ii
iii
Acknowledgments
In first place, I would like to thank for the support, motivation and mentorship given by my
supervisors Prof. Dr. João Pedro Conde and Dra. Ana Margarida Fortes. I really would like to emphasis
my gratitude in allowing me to work on this project. I would also like to thank Dra. Virginia Chu for the
support given throughout my year in INESC-MN.
I would like to thank my colleagues from INESC-MN for the amazing experience that has been
working together this last year. I will never forget the supportive environment provided during work, with
a special thanks to Inês and Ruben who were dedicated and amazing mentors in every phase of this
project. Not only them, but everyone else who contributed in some way to help me grow during this
project, and for all the fun times and cheerful spirit in the lab!
I cannot forget my friends from NOVA with whom I shared the most important moments and
experiences during this academic journey, particularly Gonçalo, my favourite Azorean and partner on
the many adventures we shared in Czech Republic and the rest of Europe. To my friends from Board of
European Students of Technology, for all the hard-working days and nights when we did not even slept
to pursue this dream we shared. A special thanks to Palmeiro, Tomás and Sara who were very
supportive in all projects I pursued, including this one. And, of course, a special thanks to Ľuboš and
Sachin for motivating me to work, while I was doing my thesis, when it was the hardest to keep
everything on track: I hope we have the opportunity to meet again as we did in Latvia.
To my childhood friends, Magda and Margarida, for proving that despite the absence, our friendship
prevails and even becomes stronger.
To my parents, for understanding why I was never home and for always being happy for my
achievements, even when they did not quite understand what it meant. Especially, I want to thank my
mom who proved this year to be the strongest woman I ever known and for teaching me that a disease
cannot define who you are.
Finally, to Eurico, for being my favourite person in the world and for supporting me in all the good
and bad times, for the friendship and love, and for dreaming with me all the adventures we are yet to
live throughout this world.
iv
v
Resumo
Existe uma procura cada vez maior em miniaturizar a instrumentação usada em separações
analíticas, exigindo-se respostas que providenciem soluções de elevado rendimento, mas mais
importante ainda, análises rápidas e on-spot. É, portanto, uma prioridade desenvolver um dispositivo
analítico miniaturizado e cromatográfico que atende não só ao objetivo de total integração, mas também
aponta para um processo de fabricação mais simples (economicamente viável), uma operação mais
user-friendly e uma versatilidade para uma ampla variedade de analitos/aplicações.
O desenvolvimento de um dispositivo de microfluídica, consistindo num injetor de geometria cross-
channel com válvulas integradas e atuadas pneumaticamente (que oferecem quase zero dead-volume),
e num canal de 12 cm de comprimento de separação, permitindo o empacotamento de meio
cromatográfico disponível comercialmente, foi alcançado. O canal de separação foi testado para efeitos
de proof-of-concept com a separação de uma mistura de azul de metileno e corante alimentar amarelo,
seguido de regeneração, provando o grande potencial para aplicações em purificação e separações
analíticas. Além disso, a aplicabilidade deste dispositivo para análises de segurança alimentar foi
avaliada usando o canal de separação para cromatografia de fase-reversa, indicando que com
condições cromatográficas futuramente otimizadas e otimização também do dispositivo, a separação e
deteção de toxinas como a aflatoxina B1 (AFB1) e ocratoxina A (OTA) será possível. Finalmente, a
injeção e progressão de um plug de isotiocianato de fluoresceína (FITC), num injetor cross-channel
com válvulas integradas e atuadas pneumaticamente e adaptado ao canal de separação, foi estudada.
Este estudo ofereceu uma avaliação sobre as otimizações futuras ser feitas no dispositivo de
microfluídica desenvolvido.
Palavras-chave:
Microfluídica, cromatografia, separação, ocratoxina A, aflatoxina B1
vi
vii
Abstract
There is an increased demand for smaller and smaller instrumentation for analytical separations,
requiring high-throughput answers to its problems but more importantly, fast and on-spot analyses of
products of interest. Therefore, it is a priority to develop a miniaturized chromatographic analytical device
which not only meets the goal of fully integration but also aims for a simpler fabrication process (and
therefore more economically viable), user-friendly operation and versatility for a broad range of
analytes/applications.
The development of a microfluidic device which comprises a cross-channel injector with integrated
pneumatic actuated valves that offers almost zero dead-volume and a 12 cm-long separation channel
enabling the packing of most commercial available chromatographic beads was achieved. The
separation channel was tested for a proof-of-concept separation that allowed the separation of a mixture
of methylene blue and yellow food colouring and further regeneration of the channel, proving the great
potential for future purification and analytical applications. Moreover, the applicability of this device for
food safety applications was assessed using a reverse-phase chromatographic mode for the separation
channel, indicating that under the optimal chromatographic conditions and further optimization of the
device design, separation and detection of toxins like aflatoxin B1 (AFB1) and ochratoxin A (OTA) is
possible. Finally, a fluorescein isothiocyanate (FITC) plug injection and progression using the cross-
channel injector with pneumatically actuated valves coupled to the separation channel was studied,
which offered an assessment on further optimizations to be done on the microfluidic device developed.
Keywords:
Microfluidics, chromatography, separation, ochratoxin A, aflatoxin B1
viii
ix
Table of Contents
ACKNOWLEDGMENTS ......................................................................................................................... III
RESUMO ................................................................................................................................................. V
PALAVRAS-CHAVE: ............................................................................................................................... V
ABSTRACT ........................................................................................................................................... VII
KEYWORDS: ......................................................................................................................................... VII
TABLE OF CONTENTS ......................................................................................................................... IX
LIST OF TABLES ................................................................................................................................. XIII
LIST OF FIGURES ................................................................................................................................XV
LIST OF ACRONYMS ..........................................................................................................................XIX
1. INTRODUCTION ............................................................................................................................. 1
1.1. MOTIVATION ................................................................................................................................... 1
1.2. CHROMATOGRAPHY ........................................................................................................................ 2
1.2.1. General concepts............................................................................................................. 2
1.2.1.1. Elution .......................................................................................................................... 3
1.2.1.2. Chromatograms ........................................................................................................... 4
1.2.1.3. Retention time .............................................................................................................. 4
1.2.2. Column performance ....................................................................................................... 5
1.2.2.1. The theoretical plate model of chromatography .......................................................... 5
1.2.2.2. The rate theory and the van Deemter equation ........................................................... 6
1.2.3. HPLC ............................................................................................................................... 6
1.2.3.1. Instrumentation ............................................................................................................ 7
1.2.3.2. Chromatographic modes ............................................................................................. 9
1.3. MICROFLUIDICS............................................................................................................................. 11
1.3.1. General concepts........................................................................................................... 11
1.3.2. Microfabrication ................................................................................................................... 13
1.3.2.1. Photolithography ........................................................................................................... 13
1.3.2.2. Soft lithography technologies ....................................................................................... 14
1.3.3. Chromatography-on-chip ..................................................................................................... 16
1.3.3.1 Column design and fabrication considerations ............................................................. 16
1.3.3.2 Injection and sample volume considerations ................................................................ 17
1.3.3.3 Literature examples on chromatography-on-chip .......................................................... 18
1.4. FOOD SAFETY APPLICATIONS ......................................................................................................... 19
2. MATERIALS AND METHODS .......................................................................................................... 21
x
2.1. MICROFLUIDIC DEVICE FABRICATION ............................................................................................... 21
2.1.1. Hard mask fabrication .......................................................................................................... 21
2.1.2. Master mold fabrication ....................................................................................................... 22
2.1.2.1. Silicon substrate cleaning ............................................................................................. 23
2.1.2.2. 20 μm layer ................................................................................................................... 23
2.1.2.3. 35 μm layer ................................................................................................................... 23
2.1.2.4. 50 μm layer ................................................................................................................... 24
2.1.2.5. 100 μm layer ................................................................................................................. 24
2.1.3. Fabrication of PDMS channels ............................................................................................ 25
2.1.3.1. Process for iteration 1, 2 and microcolumns ................................................................ 26
2.1.3.2. Process for iteration 3 and 4A ...................................................................................... 26
2.1.3.3. Process for iteration 4B ................................................................................................ 26
2.2. MICROFLUIDIC DEVICE MANIPULATION ............................................................................................ 27
2.2.1. Separation channel packing .......................................................................................... 27
2.2.1.1. Beads preparation protocols ...................................................................................... 28
2.2.1.2. Packing method ......................................................................................................... 28
2.2.2. Separation and elution studies ...................................................................................... 29
2.2.2.1. Food colouring separation (proof-of-concept) ........................................................... 29
2.2.2.2. OTA and AFB1 microchannel adsorption and elution ............................................... 30
2.2.2.3. OTA and AFB1 cross-channel injection and elution .................................................. 30
2.2.2.4. Plug optimization ....................................................................................................... 31
3. RESULTS AND DISCUSSION ...................................................................................................... 32
3.1. MICROFLUIDIC DEVICE DESIGN CONSIDERATIONS ............................................................................ 32
3.2. FABRICATION PROCESS CONSIDERATIONS ...................................................................................... 37
3.2.1. Valves thickness optimization ......................................................................................... 37
3.2.2. Microfluidic structure sealing optimization ....................................................................... 38
3.3. SEPARATION CHANNEL PACKING .................................................................................................... 39
3.4. FOOD COLOURING SEPARATION (PROOF-OF-CONCEPT) ................................................................... 43
3.5. TOXINS SEPARATION FOR FOOD SAFETY APPLICATIONS ................................................................... 45
3.5.1. Toxins concentration optimization ....................................................................................... 45
3.5.2. OTA and AFB1 elution studies in microchannels ................................................................ 46
3.5.3. OTA and AFB1 cross-channel injection and elution ............................................................ 48
3.6. PLUG PROGRESSION ASSESSMENT AND OPTIMIZATION .................................................................... 51
4. CONCLUSION AND FUTURE PROSPECTS ............................................................................... 55
REFERENCES ...................................................................................................................................... 58
APPENDIX SECTION ........................................................................................................................... 62
A. OPTIMIZED PROCESSES RUNSHEET ................................................................................................... 62
xi
B. SOLVENT EFFECT INFLUENCE ON FLUORESCENCE SIGNAL FOR OTA AND AFB1 ELUTION IN
MICROCHANNELS ................................................................................................................................. 66
xii
xiii
List of Tables
Table 1. HPLC chromatographic modes are summarized with attached examples present in the
literature. ................................................................................................................................................ 10
Table 2. Materials and equipment required for microfluidic device fabrication. .................................... 21
Table 3. Runsheet comprising the sequential steps involved in each mold fabrication process for the
different design iterations, including iterations where two different master molds are necessary (3 and
4). ........................................................................................................................................................... 22
Table 4. Materials and equipment required for microfluidic device manipulation. ................................ 27
Table 5. The different packing solutions required for each beads type. ............................................... 40
xiv
xv
List of Figures
Figure 1. Column chromatography. Left - the standard elements of a chromatographic column include
a matrix supported inside a column, generally made of plastic or glass. A protein solution to be separated
can percolate the column (starts on top). Proteins migrate through the column at different rates because
of different interactions established with the matrix. Right - diagram representing the separation of the
chromatography procedure illustrated on the left. The detected signal at the various stages of elution is
shown. Adapted from the literature16,17.................................................................................................... 3
Figure 2. Chromatogram. Left - elution of components A, B and C is represented by plotting the detector
signal intensity in millivolts versus time of elution in minutes. Right - two-component chromatogram
exemplifying two approaches for improving separation: (a) original with overlapping peaks, (b)
improvement by increase in band separation, (c) improvement by decrease in band widths. Adapted
from the literature16,17............................................................................................................................... 4
Figure 3. Chromatogram of a two-component (I and II) mixture. t0 represents the dead time, tR1 and tR2
the retention times of the components I and II, respectively. W1 and W2 are the band widths of each
component and Δt is the difference in elution time between the two components. Adapted from
Thermopedia. .......................................................................................................................................... 5
Figure 4. Diagram of the chromatographic methods. GSC – gas solid chromatography, GC – gas
chromatography, SFC – supercritical fluid chromatography, LC – liquid chromatography, LLC – liquid-
liquid chromatography, MEKC – micellar electrokinetic chromatography, LSC – liquid solid
chromatography, SEC – size exclusion chromatography, IEC – ion exchange chromatography, AC –
affinity chromatography and BPC – bonded phase chromatography. .................................................... 7
Figure 5. HPLC system diagram. Adapted from the literature18. ............................................................ 8
Figure 6. Taylor-Aris dispersion in laminar flow where (a) initial analyte plug and (b) the plug after time
t in the absence of diffusion; (c) is the plug after time t with a finite axial diffusion where the effect of
diffusion is minor as compared to the dispersive effect of the flow. (d) represents the plug after time t
with finite radial/transverse diffusion. Adapted from the literature34. ..................................................... 12
Figure 7. Fabrication process of: (1) aluminium hard masks; (2) SU-8 negative photoresist two level
mold; and (3) PDMS structures. The photoresists 1 and 2 should be selected to allow the spin-coating
of a thinner and thicker SU-8 layer, respectively. .................................................................................. 14
Figure 8. Structure of poly(dimethylsiloxane). ...................................................................................... 15
Figure 9. A 3D scale diagram of several pneumatically actuated valves in a peristaltic pump
configuration. The channels are 100 µm wide and 10 µm high. Adapted from the literature39. ............ 15
Figure 10. Chemical structure of ochratoxin A. .................................................................................... 19
Figure 11. Chemical structure of aflatoxin B1....................................................................................... 20
Figure 12. Sequence of steps involved in the fabrication of the aluminium hard masks. .................... 22
Figure 13. Sequence of steps involved in the fabrication of the SU-8 negative photoresist mold (iteration
1, 2 and valves layer molds) and of SU-8 negative and AZ 40 XT positive photoresist molds (iteration 3
and 4 fluidic layer molds). ...................................................................................................................... 24
xvi
Figure 14. Sequence of steps involved in the fabrication of the PDMS channels for the different
iterations, using distinct fabrication methods: soft lithography with irreversible sealing done by plasma
treatment (iteration 1 and 2); multilayer soft lithography and sealing using an adhesive PDMS layer
(iteration 3 and 4A) or plasma treatment (iteration 4B). ........................................................................ 25
Figure 15. Experimental setup for compressed air packing and valves actuation. (A) Switch buttons for
on/off actuation on the valves. (B) and (C) PCB for controlling of valves actuation and respective power
source. (D) Syringe pump. (E) Microfluidic device connected to the system by capillary tubing. ......... 29
Figure 16. Comparison between the first two CAD iterations created using AutoCAD. (1-A) corresponds
to a close-up representation of the bead trapping feature near the outlet where the channel has two
different heights: 100 and 20 µm. (1-B) corresponds to a microscopic image of the master mold
fabricated with the serpentine separation channel with a total length of 30 cm. (1-C) corresponds to the
microscopic image of the master mold where the bead trapping can be seen. (2-A, B) highlight the
simple cross-channel injector representation introduced into the design near the inlet and a microscopic
image of the injector on the master mold, with the two channels of different heights (100 and 20 µm).
(2-C) Microscopic image of the bead trapping feature and serpentine separation channel with a total
length of 12 cm. ..................................................................................................................................... 33
Figure 17. Comparison between the third and fourth CAD iterations created using AutoCAD. (3-A)
corresponds to a close-up representation of the cross-channel injector with integrated pneumatically
actuated valves. (3-B) corresponds to a microscopic image of the valves mold fabricated showing a
single valve channel. (3-C) corresponds to the microscopic image of the master mold where the
separation channel and cross-channel injector can be seen. (4-A, C) A close-up representation and
microscopy image, respectively, highlights the connection channel (height: 20 µm) designed between
the new cross-channel injector (height: 35 µm) and the separation channel (height: 100 µm) which
entraps beads within the separation channel. (4-B) Microscopic image of the new incorporated cross-
channel injector with integrated pneumatically actuated valves............................................................ 35
Figure 18. Comparison between the two injectors from iteration 3 and iteration 4. (A) Microscopic image
of the cross-channel injector and separation channel from iteration 3, packed with silica beads. (B)
Microscopic image of the cross-channel injector, connection channels and separation channel from
iteration 4, packed with silica beads. ..................................................................................................... 36
Figure 19. Microscopic images PDMS microfluidic structures (iteration 3) comprising valves of different
thicknesses originated by different spin-coating speeds, and its actuation. (A, B and C) are microscopic
images of valve channels spin-coated at 300, 400 and 500 rpm originating decreasing valves layer
thickness. (D and E) are microscopic images of the actuation of valve channels at P = 0 and 0.1 MPa,
respectively, for a 300 rpm spin-coated layer. (F) is a microscopic image pf the actuation of valve channel
at P = 0.1 MPa for a 400 rpm spin-coated layer. Green food colouring was flowed through the injector
channel for better visualization of valves actuation. .............................................................................. 37
Figure 20. Schematics illustrating the two strategies used for sealing of the microfluidics structures
comprising valves (iteration 3 and 4), namely bonding through half-cure and PDMS glue or bonding with
conventional plasma cleaner treatment. ................................................................................................ 38
xvii
Figure 21. Schematics comprising a timeframe of the beads trapping feature with packed silica beads
(left); a comparison between the different packing methods used throughout the experimental work
(centre); and a set of microscopic images of a fully packed 12 cm length separation channel with C18
bonded silica beads (right, A-D). ........................................................................................................... 41
Figure 22. Separation of a green coloured mixture. (A-C) corresponds to an illustration of the sequential
steps previously done to insertion of the mobile phase, including beads packing (A), green mixture
solution loading (B) and elution with Milli-Q water as the mobile phase (C). The microscopic images (i-
iv) show the progression of the separation throughout 10 minutes. (i) Initial sample plug starts to be
eluted with Milli-Q water as mobile phase. (ii) Methylene blue starts to be retained in the column. (iii)
Yellow plug travels through the column. (iv) Eluted components leave the column in a yellow plug with
traces of green. ...................................................................................................................................... 43
Figure 23. Schematics illustrating the two employed strategies for separation channel regeneration: (1)
increasing the ionic strength of the mobile phase or (2) protonation of the carboxymethyl group of the
stationary phase. ................................................................................................................................... 44
Figure 24. Plots comprising the fluorescence signal variance in function of the concentration of OTA
(top) and AFB1 (bottom) toxins. The correspondent microscopic images for the different concentrations
for each toxin are displayed on the left of the graphics. Fluorescence images were acquired at the
beginning of the microchannel with an exposure time of 1 second and a gain of 1x. ........................... 46
Figure 25. Elution profiles for toxins OTA and AFB1 previously adsorbed in a C18 bonded silica packed
microchannel. Elutions were performed during 10 minutes and fluorescence images acquired every 30
seconds at the beginning of the microchannel with an exposure time of 1 second and a gain of 1x. .. 47
Figure 26. Cross-channel injection and elution of the toxins AFB1 and OTA in reverse-phase separation
channel. (A) illustrates the mode of operation for loading and elution of the toxins using iteration 2 as a
microfluidic structure. (B) are fluorescence acquisitions in the beginning of the separation channel
comprising the cross-channel injector, of the AFB1 6 µg/mL elution with 100% CH3CN (v/v) during the
first 2 minutes. (C) are fluorescence acquisitions in the beginning of the separation channel, of the
elution with CH3CN:water (25:75 v/v) for the separation of OTA and AFB1, at a concentration of 6 µg/mL
Fluorescence images were acquired at the beginning of the microchannel with an exposure time of 1
second and a gain of 1x. ....................................................................................................................... 48
Figure 27. Schematics showing the operation of the cross-channel injector with integrated valves for
the injection and elution of the FITC plug by illustrating the sequential steps: washing (step 1), loading
(step 2), injection (step 3) and elution (step 4). Arrows indicate the direction of fluid flow and dark lines
the valves that are closed. ..................................................................................................................... 51
Figure 28. Sequential fluorescent image acquisitions of a FITC plug injection (i-iii) and progression (iv-
viii) throughout the separation channel without beads packed. Images were acquired with an exposure
time of 800 milliseconds and 3x gain. ................................................................................................... 52
Figure 29. Sequential fluorescent image acquisitions of a FITC plug injection (i-iii) and progression (iv-
viii) throughout the separation channel with beads packed. Images were acquired with an exposure time
of 800 milliseconds and 3x gain. ........................................................................................................... 53
xviii
xix
List of Acronyms
AFB1 - aflatoxin B1
OTA – ochratoxin A
FITC – fluorescein isothiocyanate
CE – capillary electrophoresis
μTAS – micro total analysis system
LOC – lab-on-chip
LC – liquid chromatography
HPLC – high performance liquid chromatography
GSC – gas solid chromatography
GC – gas chromatography
SFC – supercritical fluid chromatography
LLC – liquid-liquid chromatography
MEKC – micellar electrokinetic chromatography
LSC – liquid solid chromatography
SEC – size exclusion chromatography
IEC – ion exchange chromatography
AC – affinity chromatography
BPC – bonded phase chromatography
RPC - reversed-phase chromatography
NPC - normal-phase chromatography
NARP - non-aqueous reversed phase chromatography
HILIC - hydrophilic interaction chromatography
IPC - Ion-pair chromatography
MEMS - microelectromechanical systems
PDMS – polydimethysiloxane
DWL - direct laser writer
MSL - multilayer soft lithography
IPA – isopropanol
DI – deionized
PGMEA - propylene glycol ether acetate
CAD - computer assisted design
PEG - polyethylene glycol
PCB - printed circuit board
xx
1
1. Introduction
1.1. Motivation
Separation of mixtures is very important in synthesis, industrial chemistry, biomedical field and
chemical analyses. A separation process can be defined as the method to achieve any phenomenon
that converts a mixture of chemical substances into two or more distinct product mixtures, at least one
enriched in one or more of the mixture’s constituents. It is also normally associated with the removal of
impurities to achieve purification of the product, depending on the purpose: either preparative in the
case of purification - as just exemplified - or analytical. The goals of an analytical separation are usually
to eliminate or reduce interferences so that quantitative analytical information can be obtained from
complex mixtures. Many separation methods exist nowadays, ranging from precipitation1 and extraction2
to various chromatographic techniques3,4. All of them differ in their separation principle that can be based
on chemical or physical properties such as size, shape, mass, density or chemical affinity between the
components of the mixture. Apart from the separation itself, either analytical or preparative separations
require a detection mode and later translation of signal into real information. To note that in a technique
such as chromatography, quantitative information is obtained nearly simultaneously with the separation
unlike other procedures already mentioned above.
Analytical separations can occur at any scale: from a laboratory scale to the industrial scale.
However, there is an increasing demand for smaller and smaller instrumentation for analytical
separations5, requiring high-throughput answers to its problems but more importantly, fast and on-spot
analyses of products of interest. This implies small sample and set-up size, low costs and easy disposal
of wastes.
Microfluidic devices have been the answer for these demands in the last years. Although some
work had already been reported6, it was since capillary electrophoresis (CE) was demonstrated in
microfabricated channels in the year of 1992 for the first time7, that a spurt of interest in microfluidic
devices has been generated in analytical separations field. The reason miniaturized systems are gaining
popularity is obvious: size matters. A miniaturized system can attain a high compactness of components
and this simplifies multiplexed systems for providing high sample throughput. Single analysis can as
well be considerably accelerated with the reduced setup sizes because distances for molecules to travel
are drastically shortened – sample size is therefore minimized. The same applies for costs of the device
as less sample is required. More importantly, all the steps involved in the process – from sample
preparation to signal acquisition – can be integrated onto a single device. This concept was first
described in 1990 as micro total analysis system (μTAS)8, nowadays more known as Lab-on-Chip (LOC)
devices9.
2
One of the most popular separation methods – liquid chromatography (LC) – has been one of the
most prolific applications in LOC devices. However, efforts directed on developing pressure driven
separations (analogous to conventional liquid chromatography columns and HPLC – High Performance
Liquid Chromatography) in a chip based platform have been pursued by a minor number of research
groups10,11,12. This can largely be explained by the much different technical challenges involved in
successful execution, such as on-chip integration of the porous stationary phase, ultra-low volume
sample injection, adequate chip sealing, on-chip pumping, and on-chip detection. Nevertheless, non-
stop developments in microfabrication and nanotechnology, as well as propagation of wafer level
processing well outside the microelectronics industry, created an abundant ground for exploration of
such possibilities with high chance of future innovations. Given the current circumstances of the field
stated above, it is imperative to take advantage of the new technologies available to create a
chromatographic chip which not only meets the goal of fully integration like attempted previously12, but
also aims for a simpler fabrication process (and therefore more economically viable), user-friendly
operation and versatility for a broad range of analytes/applications.
Within the range of relevant analytes for this device, mycotoxins are a particularly important
application since they contaminate the diet of a large proportion of the world’s population, especially in
low-income countries where the contamination levels are the highest13. They represent a serious global
health issue as per most known mycotoxins, like OTA and AFB1, are potential carcinogenic (Group 2B)
and confirmed carcinogenic (Group 1), respectively, for humans14. The fact that such a device can have
a broad range of applications, including critical areas such as food safety, is of upmost importance for
the motivation of this work.
1.2. Chromatography
Chromatography is a broadly used method for the separation, identification, and determination of
chemical components in complex mixtures. No other separation method is as powerful and largely
applicable as chromatography. It is such a diverse method that it can be applied to several systems and
techniques3,4,15. However, all of them have common concepts and features associated that define what
is a chromatographic method. Therefore, a brief guide to the most important concepts will be addressed
in the following sections and represent an summary of the literature16.
1.2.1. General concepts
Chromatography is a technique in which the components of a mixture are separated based on
differences in the rates at which they are carried through a fixed stationary phase by a gaseous or liquid
mobile phase. The stationary phase is the phase that is fixed either in a column or on a planar surface.
On the contrary, the mobile phase is the one that moves through the stationary phase carrying with it
the analyte mixture. The mobile phase might be a gas, a liquid, or a supercritical fluid.
3
1.2.1.1. Elution
Elution is a procedure in which solutes are washed through a stationary phase by the movement
of a mobile phase. The mobile phase that departures the column is labelled the eluate.
16,17
In Figure 1, an example of a protein mixture being eluted by column chromatography is shown.
The column is packed with a matrix (solid, porous material) consisting on the stationary phase. Initially,
the mobile phase (eluent) is introduced onto the column, occupying the empty spaces existent in the
matrix. An eluent is a solvent used to transport the components of a mixture through a stationary phase.
Then, the solution containing the protein mixture is inserted as a narrow plug, corresponding to the time
t0 in the diagram. Next, elution happens by compelling the mixture through the column and continuously
adding new mobile phase. Given that the solute (in this case, the proteins) only moves with the help of
the elution, it is usually said that the average rate at which a solute migrates depends on the fraction of
time it spends in that phase, being small for solutes that are strongly retained by the stationary phase
(example: solute B in Figure 1). As shown in the previous figure, these different rates of migration will
result in individual bands along the column. The separation of each component involves
collecting/detecting these different bands as can be seen in the diagram above.
Figure 1. Column chromatography. Left - the standard elements of a chromatographic column include a matrix
supported inside a column, generally made of plastic or glass. A protein solution to be separated can percolate the
column (starts on top). Proteins migrate through the column at different rates because of different interactions
established with the matrix. Right - diagram representing the separation of the chromatography procedure illustrated
on the left. The detected signal at the various stages of elution is shown. Adapted from the literature16,17.
4
1.2.1.2. Chromatograms
The separated components can be collected at the end of the column and/or detected. But how
does this happen? By simply placing a detector that responds to solute concentration at the end of the
column (or other support, depending on the chromatography) when elution occurs, a series of peaks
can be obtained, if the signal is plotted as a function of time or added eluent volume. This plot in
chromatography is usually called chromatogram (Figure 2) and serves as both qualitative and
quantitative analysis.
16,17
The identification of each component on the mixture can be done by analysing the position of the
peaks maxima on the time axis. Also, the peak areas provide quantitative information of the amount of
each species present in the mixture. However, several times the chromatogram does not turn out so
well, due to band broadening, as seen in Figure 2 (right illustration). There are many possible methods
to overcome this difficulty that will be addressed further on: some of them consist in improving the band
separation rate or decrease the rate of band spreading by manipulating several variables that influence
solute migration rates.
1.2.1.3. Retention time
𝑡𝑅 = 𝑡𝑆 + 𝑡0 (Equation 1)
The retention time, 𝑡𝑅, (Equation 1) is the time between injection of a sample and the presence of
a solute peak at the detector of a chromatographic column: different solutes have different retention
times. In Figure 3, a chromatogram of a two-component (I and II) mixture is shown. As seen below, t0,
the dead time, is the time it takes for an unretained species to pass through a chromatographic column.
Figure 2. Chromatogram. Left - elution of components A, B and C is represented by plotting the detector signal
intensity in millivolts versus time of elution in minutes. Right - two-component chromatogram exemplifying two
approaches for improving separation: (a) original with overlapping peaks, (b) improvement by increase in band
separation, (c) improvement by decrease in band widths. Adapted from the literature16,17.
5
All components spend at minimum this total of time in the mobile phase. Separations are based on the
different times 𝑡𝑆, that components spend in the stationary phase.
1.2.2. Column performance
The most important factor when designing or assessing a chromatographic column is its
performance. More important than simply separating the components of a mixture, it is desirable that
this separation is efficient. To obtain optimal separations, sharp, symmetrical chromatographic peaks
must be obtained. This means that band broadening must be limited. However, there are many factors
imposing barriers to a highly efficient separation in chromatography. These factors will be addressed in
the following sections by briefly highlighting the models behind column efficiency assessment. More
information about this can be found on the literature18, since a more profound mathematical approach
to the models is not within the scope of this work.
1.2.2.1. The theoretical plate model of chromatography
There are two essential parameters in this model approach used to measure chromatographic
column efficiency: plate height, 𝐻, and number of theoretical plates, N. They are correlated by the
following equation:
𝑁 = 𝐿
𝐻 (Equation 2)
Where 𝐿 is the length of the column packing. As it can be deduced, the smaller the value for 𝐻
the higher the value for 𝑁, meaning a higher number of plate count, therefore achieving more column
efficiency. The plate model assumes that the column comprises many distinct layers, termed theoretical
Figure 3. Chromatogram of a two-component (I and II) mixture. t0 represents the dead time, tR1 and tR2 the
retention times of the components I and II, respectively. W1 and W2 are the band widths of each component and
Δt is the difference in elution time between the two components. Adapted from Thermopedia.
6
plates. Separate equilibrations of the sample between the stationary and mobile phase occur in these
"plates". The analyte moves down the column by transfer of the mobile phase from one plate to the next.
1.2.2.2. The rate theory and the van Deemter equation
Theoretical studies of zone broadening in the years around 195019 led to the van Deemter
equation (Equation 3). A more correct explanation of what occurs inside a column considers the time
taken for the solute to equilibrate among the stationary and mobile phase (contrasting the plate model,
which accepts that equilibration is infinitely fast). This means that the band shape depends on the rate
of elution. It also may be affected by the diverse paths accessible to solute molecules as they travel
between the stationary phase. This all translates into Equation 3, where 𝐶𝑆 and 𝐶𝑀 are mass-transfer
coefficients for the stationary and mobile phases, respectively, and 𝑢 is the linear velocity of the mobile
phase.
𝐻 = 𝐶𝑀𝑢 (𝑜𝑟 𝐴) + 𝐵
𝑢+ 𝐶𝑆𝑢 (Equation 3)
The term (𝐵𝑢⁄ ) represents longitudinal diffusion. This factor causes band width to rise with time,
and it happens whether or not the mobile phase is flowing. The time spent by the band during its way
through the column differs inversely with the flow rate, so the influence of longitudinal diffusion on band
width decreases for increased flows. In liquid chromatography, diffusion rates are much smaller in value
so this term has an almost insignificant influence on band broadening. On the other hand, the term (𝐶𝑆𝑢)
represents the stationary phase mass-transfer. Among the analyte molecules, there will be ones that
enter further into a particle pore than others, spending in it a given and varied time. Molecules that spend
less time in the particle will move further along the column, with a resulting increase in band width. This
contribution to band broadening increases as the flow rate increases. Also, when the band leaves the
column has to pass through the detector, resulting in additional peak broadening. Finally, the term A or
(𝐶𝑀𝑢) represents the mobile phase mass-transfer. Zone broadening in the mobile phase is caused by
the multiple paths a molecule can take through a packed column, especially when the lengths of these
paths can differ significantly. This phenomenon is usually called eddy diffusion. At low mobile phase
velocities, the molecules are not significantly dispersed due to the contribution of regular diffusion effects
that attenuates the eddy diffusion. At high velocities, there is no sufficient time for diffusion to act, and
then band broadening due to the different path lengths is seen.
1.2.3. HPLC
There are diverse kinds of chromatographic methods and these are classified according to the
shape of bed (planar or column), physical state of the mobile phase and separation mechanisms.
Furthermore, there are other adapted arrangements of these chromatographic techniques relating
7
different mechanisms and are therefore considered as modified or specific chromatographic
techniques20. In the following diagram (Figure 4), the techniques are organized by the mobile phase
physical state, stationary phase and mechanism of separation.
As seen in Figure 4, LC is the one with broader range of available separation mechanisms. This
versatility and the rapid expansion of HPLC since the early 1970s21 has put liquid chromatography on
top of the most important analytical separation techniques.
Early liquid chromatography used to be performed in columns with an internal diameter from 10
to 50 mm and were around 50-500 cm long, therefore particle size used to be around 150-200 µm to
ensure a reasonable flow rate was achieved without a high demand on pressure. Together with the early
development of the theory of liquid chromatography, it was understood that in order to decrease the
plate height smaller particle size (3-10 µm) was in need21. Of course, this new feature required higher
pumping pressures along with the development of new detectors for continuous monitoring. Throughout
that time, each procedure presented key enhancements in convenience, speed, resolving power
detection, quantification and applicability to new sample natures. It is easy to see how HPLC embodies
the modern culmination of the progress of liquid chromatography. Consequently, the equipment for
HPLC nowadays tends to be considerably more elaborate and therefore expensive, time-consuming
and requiring specialized personnel for operation.
1.2.3.1. Instrumentation
Within the scope of this work, it is important have in mind the main features of HPLC
instrumentation to understand the need of several components integrated in the microfluidic device and
even assess the superiority of such device when compared to HPLC.
Figure 4. Diagram of the chromatographic methods. GSC – gas solid chromatography, GC – gas chromatography,
SFC – supercritical fluid chromatography, LC – liquid chromatography, LLC – liquid-liquid chromatography, MEKC
– micellar electrokinetic chromatography, LSC – liquid solid chromatography, SEC – size exclusion
chromatography, IEC – ion exchange chromatography, AC – affinity chromatography and BPC – bonded phase
chromatography.
8
18
In Figure 5 a HPLC system diagram is presented where the main components are highlighted,
such as the solvent reservoir followed by an high-pressure pump and a autosampler (injector) just before
the chromatographic column. Sometimes, it is important to maintain a specific temperature for enhanced
column performance, therefore we can observe a column oven placed on the HPLC system diagram.
Following the column, there is a detector connected to a system that will analyse the signal and
transform it into a chromatogram.
More specifically, the mobile phase reservoirs are often accompanied with components placed
before the pump to remove dissolved gases (in-line degassing) which normally lead to irreproducible
flow rates and band spreading. Primarily, it is important to improve the pump operation to ensure fixed
flow rates, but dissolved oxygen can also be a source of background noise for the detector22. Also,
provisions should be included to remove dust and bubbles that usually interfere with the signal
acquisition of most detectors, especially when HPLC-grade solvents are not used. Regarding tubing
included on the HPLC equipment, care should be taken to ensure that tubing length and diameter are
chosen to minimize peak-broadening contributions, most precisely in tubing where the sample is
present: between the autosampler and the column, and between the column and detector23.
There are several requirements for most HPLC conventional pumping systems: they must
generate high pressures up to 6000 psi, have a pulse-free output and have a good flow reproducibility.
Most commercial instruments use the reciprocating pump type which meets the requirements stated
above: capacity for high output pressure (up to 10000 psi), achieves constant flow rates independent of
column back-pressure and solvent viscosity, in addition to small internal volumes and good adaptability
to gradient elution. Finally, most of these pumps are computer-controlled devices to measure the flow
rate and rapidly adapting the settings to maintain a constant flow rate18.
Figure 5. HPLC system diagram. Adapted from the literature18.
9
Before injection of the sample and mobile phase into the column, it is important to consider the
type of elution needed for the given separation. In an isocratic elution, the solvent composition remains
constant throughout the separation and can either be a single solvent or a mixture. However, gradient
elution is known for improving the separation efficiency and consists in changing the composition of the
solvent throughout the separation either continuously or in a step-wise fashion. Although for isocratic
elution the mobile phase can be hand-mixed, most of the times the HPLC system can have on-line
mixing integrated in the system both for isocratic elution and gradient elution, where on-line mixing is
required. In many HPLC systems, proportioning valves are used for this purpose. Sample injection onto
the column is a very important aspect since sample size can impact peak broadening as well the need
for an automated injection system (autosampler) to increase the throughput of analysis. A feature widely
also widely use is a sample loop that provides a choice of sample sizes ranging from 1 to 100 µL or
more18,23.
Most analytical columns used in HPLC nowadays have a length from 5 to 25 cm, being the most
common lengths 10-15 cm long with an internal diameter of 4.6 mm and packed with 5 µm beads. Apart
from the analytical column, HPLC systems usually have precolumns: placed between the mobile phase
reservoir and injector, a scavenger column minimizes losses of stationary phase from the analytical
column by ensuring the mobile phase is already saturated with silicic acid from the silica packing. In
addition, another precolumn commonly used is a guard column – placed between the injector and the
column - that has a similar packing to the analytical column and therefore prevents impurities to
contaminate the analytical column and increases its lifetime. Finally, post-column we have a detector
that ideally would be small, compatible with liquid flow and universal. The restriction in size advents for
the need for low internal volume in order to minimize extra-column band broadening. Unfortunately,
there is no detector that meets all the requirements and is at the same time highly sensitive, mostly
because the detector used will depend on the nature of the sample under analysis18.
1.2.3.2. Chromatographic modes
In Table 1, several types of operation modes on HPLC are shown. One of the most relevant is
the reversed-phase chromatography (RPC) and its opposite, normal-phase chromatography (NPC).
They are distinguishable based on the relative polarities of the mobile and stationary phases.
In the beginning, most separations were based on highly polar stationary phases (such as silica)
and a relatively nonpolar solvent (hexane) was the mobile phase. This type of chromatography is now
called normal-phase chromatography. The less polar molecule is eluted first and increasing the polarity
of the mobile phase will decrease the elution time. On the other hand, in reversed-phase
chromatography, the stationary phase is nonpolar, often a hydrocarbon brushlike surface, and the
mobile phase is a relatively polar solvent (such as water, methanol, acetonitrile). The most polar
component elutes first, and decreasing the mobile phase polarity will decrease the elution time16.
Nowadays, the majority of HPLC separations are performed in this chromatographic mode due to both
10
the variety of stationary phases available and the broad range of molecules that can benefit from the
separation4,24.
Table 1. HPLC chromatographic modes are summarized with attached examples present in the literature.
Mode Description Literature example
Reversed-phase chromatography (RPC)
The column is nonpolar and the mobile phase is a polar mixture of water plus organic solvent; RPC is the most widely used mode, especially for water-soluble samples.
Bonfatti et al.4
Normal-phase chromatography (NPC)
The column is polar and the mobile phase is a mixture of less-polar organic solvents; NPC is used mainly for water-insoluble samples, preparative HPLC, and the separation of isomers.
Panfili et al.25
Non-aqueous reversed phase chromatography
(NARP)
The column is nonpolar, and the mobile phase is a mixture of organic solvents; NARPC is used for very hydrophobic, water-insoluble samples.
Lísa et al.26
Hydrophilic interaction chromatography (HILIC)
The column is polar, and the mobile phase is a mixture of water plus organic; HILIC is useful for samples that are highly polar and therefore poorly retained in RPC.
Bajad et al.27
Ion-exchange chromatography (IEC)
The column contains charged groups that can bind sample ions of opposite charge, and the mobile phase is usually an aqueous solution of a salt plus buffer; IEC is useful for separating ionizable samples such as acids or bases, and especially for the separation of large biomolecules.
Rea et al.28
Ion-pair chromatography (IPC)
RPC conditions are used, except that an ion-pair reagent is added to the mobile phase for interaction with sample ions of opposite charge; IPC is useful for the separation of acids or bases that are weakly retained in RPC.
Ibáñez et al.29
Size-exclusion chromatography (SEC)
An inert column is used with either an aqueous or organic mobile phase; SEC provides separation based on molecular weight and is used mainly for large biomolecules or synthetic polymers.
Gellein et al.30
Chiral chromatography
A chiral agent is fixed on the surface of a solid support and then modes of interaction can happen, as for example, based on attractive forces between each other or the fitting into chiral cavities, forming inclusion complexes.
Peng et al.31
Another relevant chromatographic mode for this work and present in Table 1 is the ion-
exchange chromatography. The separation principle is based on molecule’s charged groups: the column
retains molecules based on the electrostatic interactions with opposite charges from functional groups
present in the stationary phase. Analyte molecules will have to compete with counterions present in
solution for binding to the functional groups so that initially, molecules that do not bind or bind weakly to
the stationary phase are washed away. Elution of bonded molecules will happen when the mobile phase
is altered to provide an increase in the concentration of the counterions that will compete with the bonded
molecules. This alteration in counterions concentration can be done by changing either the ionic strength
or the pH of the mobile phase. The stationary phase can be classified in two types: when it features
positive groups, it is called an anion exchanger and when there are negative groups it is a cation
exchanger. Finally, it can also be classified as strong or weak exchanger based on whether it can
maintain its charge within a wide range of pH (strong) or only maintaining the charge in small pH
interval15.
The other listed HPLC chromatographic modes also play an important role in the widespread
application of HPLC in chemical analysis, but will not be the focus of this work.
11
1.3. Microfluidics
MEMS (microelectromechanical systems) is a field that was born after the need for
miniaturization of all kinds of systems surged, allowing the creation of devices that are normally between
the range of 1-300 µm in size. Fluid flow in these conditions revealed phenomena worth exploring a new
field that we today know as microfluidics which can be defined as “(…) the study of flows that are simple
or complex, mono- or multiphasic, which are circulating in artificial microsystems (…)”, according to the
literature32. Since technologies for MEMS were established (mostly silicon-based devices) - around the
1980s – there was already enough knowledge to start working on microfluidic devices. However, it was
only after 1992 that microfluidics was recognized for numerous potential applications, particularly in
chromatography and CE7. Nowadays, scientists are very interested in microfluidics, given that it is a way
simpler than MEMS, by not almost never including moving components and being “soft-based”, as
polydimethysiloxane (PDMS) is the main component of nearly all devices32. In this section several
aspects of microfluidics will be addressed, such as general concepts of physics at the microscale and
fabrication as well as the implementation of chromatographic separations at the microscale.
1.3.1. General concepts
To understand the physical phenomena prevailing in miniaturized systems the scaling laws must
be studied. A scaling law can be defined as how a certain physical quantity varies with the size of the
system under analysis. Analysing Equation 4, the concept can be understood:
𝑈 ~ 𝑏2∆𝑃
𝜇𝐿 (Equation 4)
Where 𝜇 is the fluid viscosity, Δ𝑃 the pressure difference, 𝐿 the length of the channel and 𝑏 the
transverse dimension. Flow velocity, 𝑈, is equal to l1. Because there are so many variables into
consideration, some assumptions must be made in order to predict the scaling law applicable. In this
case, pressure difference of a microfluidic channel was considered to be constant. Everything else must
be arranged considering the microfluidic context, as for example the laminar flow regime that is
characteristic of microfluidic operation. Generally, the physical quantity that has a lower exponent when
compared to others is the one that will become dominant at the microscale (for example, capillary forces
winning over gravitational forces). Regarding microhydrodynamics, it is also important to highlight that
generally, liquid flow in microchannels is controlled by the dimensionless Reynolds number (Equation
5):
𝑅𝑒 = 𝜌𝑈𝐿
𝜇 (Equation 5)
12
Where 𝑈 is the velocity of the fluid, 𝜌 the density, 𝐿 the characteristic linear dimension and 𝜇 the
viscosity of the fluid. Considering that fluid velocities are very small and that dimensions of channels are
around the micrometre range, the values for Reynolds number are usually between 1 to 10 units which
falls under the laminar flow regime (Re < 2000).
As already mentioned, the need for smaller particle sizes in chromatography led to higher
operating pressures, creating the concept of HPLC. Such correlation can be addressed by Equation 6:
∆𝑃 ~ 𝐿2𝜇
𝑏2𝑡𝑅 (Equation 6)
Where L is the length of the column and b the typical size of the space between grains or pores.
𝑡𝑅 is the retention time of a given analyte. If we assume that all the variables are constant except for b
(particle size), one can observe that ΔP will increase substantially as b decreases.
Finally, when dealing with pressure-driven flows within a microchannel, an interesting effect is
the one of Taylor dispersion, where a shear flow can increase the effective diffusivity of a given molecule.
Shear acts disperse the concentration distribution in the direction of the flow, increasing the rate at which
it spreads. The overall effect is an enhancement of axial dispersion over molecular diffusion alone in the
absence of flow. There is also a significant difference in the spreading behaviour of the solute shortly
after its injection and after several diffusion times have elapsed after its injection33. In Figure 6, the
spreading behaviour shortly after its injection time can be observed. The cross-channel diffusion is yet
to happen and the distance between the fastest moving molecules at the center and the slow particles
on the wall stretches linearly with time. After some time, enough to let cross-channel diffusion to act, the
edges of the paraboloid change to form a blob which is nearly uniform across the channel and the plug
spreads.
34
Figure 6. Taylor-Aris dispersion in laminar flow where (a) initial analyte plug and (b) the plug after time t in the
absence of diffusion; (c) is the plug after time t with a finite axial diffusion where the effect of diffusion is minor as
compared to the dispersive effect of the flow. (d) represents the plug after time t with finite radial/transverse diffusion.
Adapted from the literature34.
.
13
However, chromatographic processes, such as mass exchange between phases or nonuniform
fluid flow – which were previously described in the van Deemter equation - can greatly affect the
effectiveness of molecular diffusion in causing dispersion, so each microfluidic separation channel has
to be regarded individually and not much generalizations can be done in assessing the role of this effect
on separation efficiency. Additionally, it is not only the effects of pressure-driven flow, packed bed and
chosen chromatographic conditions that can influence plug spreading along the channel, but also factors
such as channel geometry, with reported data on turns35,36 and corners37 and different cross-sections
geometries33 influencing plug spreading.
1.3.2. Microfabrication
In this section, some of the most common microfabrication techniques currently used for
microfluidic operation will be addressed. Special focus will be given to the technologies related to the
work in progress, namely photolithography and soft-lithography technologies.
1.3.2.1. Photolithography
Photolithography is a procedure used in microfabrication to pattern on a substrate. By the action
of light, it transfers a pattern to a photosensitive chemical, usually called a photoresist, previously coated
on the substrate. This pattern can be defined by direct laser writer (DWL) system or by a previously
patterned mask. Optical lithography is the most used technology and comprises wavelengths between
300 and 450 nm32. Masks are generally made of glass substrate on which aluminium (or a metal with
similar purpose) is usually deposited along with a photoresist to be further patterned using a DWL
system. Finally, it is required to etch by wet etching the aluminium that is not protected by the photoresist
to obtain the final features. The mask should be the object with the most precision possible, as the
following substrates will never have more precision than the mask itself. Masks are very useful to create
reusable molds that contain the same features as the masks and are further used in soft-lithography
fabrication.
To fabricate a mold, a photoresist is spin-coated on a silicon substrate. While the resist spreads
on the substrate, the solvent does not evaporate totally, making the film look like a soft solid. To remove
the remaining solvent, the polymer is heated before the exposure step. For exposure, the photoresist
coated substrate is aligned with the mask and the incident light starts chemical reactions in the
photoresist, which will alter the solubility in certain solvents. For example, for a positive resist the zones
exposed to the light become soluble and disappear while for a negative resist the zones exposed to the
light become insoluble in solvent, therefore staying in the substrate. It is important that the photoresists
used have a great disparity between the solubility constants of the exposed and unexposed portions,
14
high photosensitivity and high resistance to some chemicals used in afterward steps of
microfabrication32. The full fabrication of mold (including mask fabrication) is highlighted in Figure 7:
1.3.2.2. Soft lithography technologies
PDMS plays a very important role in microfluidics as one of the elastomeric materials in
popular use in soft lithography due to great capacity for rapid prototyping and easy fabrication without
expensive equipment. PDMS belongs to a family of polymers that especially contain silicon oils as seen
in Figure 8. When PDMS reaches a temperature higher than the polymerization temperature (around
70 ºC) and is mixed with a reticulating agent, it forms an elastomer whose main properties are: optical
transparency (the UV spectrum permits the visualization of flows), elasticity (allows the fabrication of
valves and pumps), good insulation, low surface energy (easier to peel off the PDMS form the mold),
chemical inertia, non-toxicity and low permeability to water (but permeable to gas)38. Also, untreated
PDMS is hydrophobic, and becomes temporarily hydrophilic after being submitted to oxidation of the
surface by oxygen: plasma creates silanes by oxidation of methyl groups. This oxidised PDMS adheres
to glass, silicon or even PDMS itself, if the surfaces receive the same treatment38. To achieve structures
like microchannels a technique called replica molding is used: a mold made of a hard material is
necessary (normally made of silicon or deposited photoresist) and a mixture of PDMS and reticulating
agent is poured on the mold. The mixture is baked and the PDMS polymerizes/reticulates, becoming
Figure 7. Fabrication process of: (1) aluminium hard masks; (2) SU-8 negative photoresist two level mold; and
(3) PDMS structures. The photoresists 1 and 2 should be selected to allow the spin-coating of a thinner and
thicker SU-8 layer, respectively.
15
solid. The last step consists on peeling off the PDMS structure from the mold and sealing against glass
or a PDMS slab to allow liquid flow in the channels.
After molding, the precision of PDMS structures is quite good – around submicrometric
values – but if the elastomeric character of the material and aging phenomena is taken into account, the
best channel dimensions are usually between 5 and 500 µm. Indeed, it is possible to fabricate structures
with higher aspect ratios but a risk is taken in whether or not these structures will collapse due to the
deformability of PDMS32.
Besides the advantages already mentioned, PDMS also allows the fabrication of integrated
pumps and valves by a technique developed by Unger et al designated multilayer soft lithography
(MSL)39,40. This technique combines soft lithography with the capability to bond multiple patterned layers
of elastomer and sealing them through half-curing of each patterned layer and baking them again
stacked together to complete the curing process. The principle is to have two opposite (the ratio of
element A and B in the elastomer mixture is substantially unalike) layers in contact with each other since
reactive molecules remain at the interface between the layers and can be further cured to produce a
sealing. More layers can be added by simply repeating this procedure. The valves were fabricated using
a cross-channel architecture, as seen in Figure 9, and had an actuation area of 100 x 100 µm in thin a
layer with a thickness of about 30 µm, designated control layer. When pressure was applied to the
control channel filled with water, this membrane would deflect to close the lower fluidic channel
beneath39. This new technology allowed for the incorporation of valves within PDMS microchips with
almost zero dead volume, a very interesting aspect in many applications, as for example the integration
of chromatographic separations in microfluidic devices. More recently, many iterations of these valves
or even new type of features have been integrated into chips using multilayer soft lithography, from
magnetic actuated valves41 to pneumatic valves in folded 2D and 3D fluidic devices made from plastic
films and tapes42.
39
Figure 8. Structure of poly(dimethylsiloxane).
Figure 9. A 3D scale diagram of several pneumatically actuated valves in a peristaltic pump configuration. The
channels are 100 µm wide and 10 µm high. Adapted from the literature39.
16
1.3.3. Chromatography-on-chip
Chromatography on chip is not something new but not an entirely finished business either. Since
CE was demonstrated in microfabricated channels in 1992 for the first time, a burst of interest in
microfluidic devices has been generated in analytical separations field7. LC has been one of the most
prolific applications in LOC devices, along with CE. However, efforts directed on developing pressure
driven separations (analogous to conventional liquid chromatography columns and HPLC) in a chip
based platform have been pursued by a minor number of research groups, even less when referred to
packed columns11,12,43,44,45 in the past two decades. This is mainly due to the technical challenges
involved in successful execution, such as on-chip integration of the porous stationary phase, on-chip
ultra-low volume sample injection and pumping and on-chip detection which will be addressed in this
section as well as the recent advances that are now enabling the chromatographic integration on
microfluidic devices to rise again.
1.3.3.1 Column design and fabrication considerations
In a HPLC-like microfluidic device, column and particles sizes must be scaled down at the same
time. Yet, to accomplish speed and separation efficiency that equals the ones from conventional HPLCs
the chip has to be submitted to high pressures, not only for the small particles and channel dimensions
used, but also for the need of increased channel length, which will cause more hydraulic resistance.
This demand causes enormous technical complications in execution of HPLC-like microchips that can
support such high pressures without leaking. Alternatively, if pressure stays identical, the separation
efficiency is worsened in the same degree and no enhancement in speed is gained, when related to
conventional HPLC. This is only one of the reasons why since the expansion of microfluidics in
chromatography, mainly CE has been integrated in microchips. Most importantly, the big advantage of
pressure-driven flow is that it is largely independent of the chemistry of the separation principle46.
To achieve separation efficiency, the channel size, shape and in particular, the aspect ratio must
be carefully designed not only considering the pressure requirements. For example, in packed channels,
the ratio of channel width/depth to particle size must be considered to avoid the known “wall effect”
where quality/density of packing or flow behaviour will be severely different from the bulk channel46,47.
This effect is more pronounced in microfluidic devices where the ratio is normally around 10. For values
much higher than 10 the wall region will only constitute a small part of the channel, and therefore the
impact of this effect will be diminished. On the other hand, for values much lower than 10 the wall and
the bulk region become very similar, therefore such effect loses its relevance. Different cross-sectional
shapes can also affect the separation efficiency by causing dispersion, either by impacting the packing
density and order or the fluid flow as corner regions tend to be of stagnant47.
17
Relatively to stationary phase considerations, there are three main ways to insert a stationary
phase into a channel: covering the walls of the channel with active elements of the stationary phase
(open channel chromatography)48; prepare packed beds of particles that will retain the analytes and
finally, creating monoliths which either already include the retentive elements or can be fixed with it later.
Open channel chromatography is the simplest way to perform the integration of chromatography into
chips, the big advantage is the ease of preparation of the column and also the low pressure required to
flow the mobile phase through the channel, even though small depths are required to compensate the
low diffusion coefficients and mass transfer to occur. However, a big disadvantage is the low surface to
volume ratio, which will decrease the capacity of processing a reasonable volume of sample and
therefore only the most sensitive methods of detection can be coupled to the device49.
There is also the possibility of simply packing the channel with porous media as usually done in
conventional LC. The big benefit of using packed channels is that essentially all the particles that have
been developed over the years for LC can be employed. The particle size should never be the only
factor in consideration, some other important aspects when designing the packed bed are: size
dispersity, porosity of the particles, and additionally, the pressure capacity of the microchip. Considering
all these factors, it seems that the degree of order in the packing is the biggest challenge on the way to
perfectly packed beds and the answer lied within the packing technique49. Also, some structure/material
is necessary to keep the beads trapped inside the channel, both during packing and operation: many of
the solutions used until now can cause band broadening and, consequently, reduce the performance of
the separation. Some examples of currently used designs are a cage-like PDMS device50, renewable
and pneumatically controlled pillar array51 or a double weir design52.
Instead of pressuring particles into a channel, a different alternative is to synthetize a porous
monolithic structure inside the channel. The main advantage is to avoid high packing pressures as well
as tricks to keep the stationary phase within the channel. However, the batch-to-batch reproducibility of
these monoliths is very hard to achieve within the acceptable range, not to mention the monolith
fabrication technology is still way too complex and undiscovered to be reasonably dominated soon and
widely implemented on chips49.
1.3.3.2 Injection and sample volume considerations
Another consideration to take into account is whether or not valves can be integrated on chip,
which will for sure impact on how “portable” can such a device be. Many approaches have been
proposed both for valves incorporation to optimize and integrate the plug injection into the
chromatographic column. One of the first components to tackle is the injector: there are many types of
designs for injectors that can be used nowadays. The most simple ones are the “double-T” junction53 or
a cross-channel design43 at the column inlet. They are preferable since they do not require valves and
can generate small sample plugs. The main disadvantages of these injectors are their propensity to
display sample dilution and leaks that can cause broadening of the plug previously to the insertion on
18
the channel due to backpressure issues. Considering the difficulties, more sophisticated designs of on-
chip injectors are available12 which often incorporate valves within the device or bulky instrumentation
outside the device. Ideally, the incorporation of valves would be preferable since it would offer zero dead
volume injections.
1.3.3.3 Literature examples on chromatography-on-chip
Some examples of different approaches to the chromatography-on-chip approach present in the
literature, are referred in this section. The first example is a highly efficient and ultra-small volume
separation by pressure-driven liquid chromatography in extended nanochannels10. In this work, the first
pressure-driven liquid chromatography system that allows separation of atto- to femtoliter sample
volumes was designed. Furthermore, the separation efficiency was very high and only within a few
seconds. In comparison to a conventional packed HPLC column, the separation is much faster (by 2
orders of magnitude), has a smaller injection volume (9 orders), and a higher separation efficiency by 1
order of magnitude. However, this microchip uses an open-column which means flexibility in relation to
the stationary phase is lost and also the fabrication process is more complicated than soft lithography
processes.
The next example is the work reporting the fabrication of high-quality microfluidic solid-phase
chromatography columns44. In this work, the authors present a column geometry that allows strong and
high-yield packing of chromatographic channels fabricated using MSL technique that can be
incorporated in an integrated microfluidic system for HPLC, providing several prospects for analytical
and preparative applications where low sample consumption, low cost, and automation are important.
However, this packing method adds complexity to the design of the chromatography-chip by
incorporation of elastomeric valves that relieve the pressure accumulated during packing, and are then
closed to restore the column geometry. Adding to these columns an efficient injector geometry which
will also require valves may hamper the fabrication process quality and had complexity to the process.
Other example, also using a PDMS microfluidic structure is this simple microfluidic chip design
for fundamental bioseparation11. The liquid chromatographic column was packed with mesoporous silica
beads. Separation of a dye mixture (fluorescein and rhodamine B) and a biopolymer mixture (dextran
and BSA) was accomplished with good column performance and the chip design was tested. However,
the did not included an on-chip injection system, being handled by off-chip valves and pumping
apparatus.
Finally, the most successful example is the HPLC-chip developed by Agilent Laboratories which
is already available commercially12. This chip includes an enrichment and separation column and a
nanoelectrospray tip all incorporated on-chip for peptide analysis applications. Enrichment and
separation columns were packed with conventional reversed-phase chromatography particles.
19
However, the fabrication process of this chip is impressively more complex and expensive than any
other solution presented so far. Also, the injection system of this device is still considered quite bulky
since a valves system is not fully incorporated.
1.4. Food safety applications
Within the range of relevant analytes for this device, mycotoxins are a particularly important
application since they contaminate the diet of a large proportion of the world’s population, especially in
low-income countries where the contamination levels are the highest13. They represent a serious global
health issue as per most known mycotoxins, like OTA and AFB1, are potential carcinogenic (Group 2B)
and confirmed carcinogenic (Group 1), respectively, for humans14. The fact that such a device can have
a broad range of applications, including critical areas such as food safety, is of upmost importance for
the motivation of this work.
A mycotoxin is a toxic secondary metabolite produced by filamentous fungi, like Aspergillus,
Fusarium and Penicillium. Mycotoxins can appear in the food chain because of fungal infection of crops,
either by being eaten directly by humans or by being used as livestock feed and they can invade crops
directly in the field but also during storage, especially in high humidity and warmer environments.
Aflatoxins, ochratoxins, trichothecenes, zearalenone, fumonisins, tremorgenic toxins, and ergot
alkaloids are the mycotoxins of greatest agroeconomic importance and often more than one mycotoxin
is found in the same contamination site since they are nowadays considered the most dangerous food
contaminants54.
OTA (Figure 10) is produced by fungi of the genera Aspergillus and Penicillium. OTA is a
frequent natural contaminant of many foodstuffs such as cocoa beans, coffee beans, cassava flour,
cereals, fish, peanuts, dried fruits, wine, poultry eggs and milk55.
OTA has a molecular weight of 404 Da and is a weak organic acid (pKa around 4.3 and the
phenol group has a pKa of 7.1) It is soluble in polar organic solvents like alcohols, ketones and
chloroform, but also slightly soluble in water. OTA is a relatively stable molecule under high
Figure 10. Chemical structure of ochratoxin A.
20
temperatures and survives most cooking processes to some extent. Also survives brewing and
winemaking and therefore can be found in a variety of processed consumer food products55.
OTA is considered a nephrotoxic molecule which causes both acute and chronic effects on
kidneys from mammalian species and is also considered genotoxic and teratogenic. Finally, OTA is also
a possible carcinogen, causing renal carcinoma and other cancers in a number of animal species,
although the mechanism for this is uncertain55.
Aflatoxins are poisonous carcinogens that are produced by Aspergillus which grow in soil,
and grains and are regularly found in storage consumables like corn, cotton seed, peanuts, rice, wheat,
and some spices. Animals fed with contaminated food can transmit aflatoxin into products like eggs,
milk derivates, and meat13.
AFB1 has a molecular weight of 312 Da and it is soluble in water and in polar organic
solvents, such as methanol, chloroform, acetone and acetonitrile. AFB1, like OTA, is also a relatively
stable molecule under high temperatures and survives most cooking processes to some extent. AFB1
is considered the most toxic toxin among the other existing variants and it is a known potent genotoxic
and carcinogenic molecule through both damage to DNA and increased oxidative damage13.
A miniaturized chromatographic analytical device where the separation channel constitutes
a packed bed with commercially available beads has a wide range of possible applications, similarly to
conventional HPLC. One of them can be the detection and separation of these toxins for food safety
purposes where fast and cheap analysis could be performed, even with small volume of samples
available. Therefore, this work focus not only on the development but also on the application of the
designed microfluidic for food safety purposes.
Figure 11. Chemical structure of aflatoxin B1.
21
2. Materials and methods
2.1. Microfluidic device fabrication
The fabrication of the microfluidic devices comprised several steps, namely hard mask
fabrication, mold fabrication and replication of PDMS structures. The reagents and materials used for
all these steps and the required equipment are listed in detail in Table 2.
Table 2. Materials and equipment required for microfluidic device fabrication.
2.1.1. Hard mask fabrication
The different design iterations of the microfluidic structure required several hard masks to be
fabricated following the experimental procedure described subsequently in this section and illustrated in
Figure 12. First, a computer assisted design (CAD) of the microfluidic structure was designed using
AutoCAD software for fabrication of the aluminium masks. The substrate used for mask fabrication was
5x5cm glass previously cleaned by washing with acetone, then DI water and lastly immersed in an
Alconox solution for 15 minutes at 65 °C. Completed the washing step, the glass substrate was rinsed
with DI water again (to remove the Alconox solution and remaining dirt) followed by drying with
compressed air. The, a 200 nm aluminium layer was deposited on the clean glass by magnetron
sputtering using a Nordiko 7000 equipment. Afterwards, a positive photoresist (PFR 7790G) layer of 1.5
Step of the process
Reagents/Materials Equipment/Facilities
Hard Mask Fabrication
- Alconox solution, Alconox Inc. (White Plains, NY/USA) - Acetone (99,6%), LabChem Inc. (Zelienople, PA/USA) - isopropanol (IPA), (99,9%), LabChem Inc. (Zelienople, PA/USA) - deionized (DI) water - Glass substrate, Corning Inc, (Corning, NY/USA) - Photoresist PFR 7790G, JSR (Sunnyvale, CA/USA) - Silicon wafer (150 mm diameter), University Wafer (South Boston, MA/USA) - TechniEtch A180 Aluminium etchant, Microchemicals (Ulm, DE)
- AutoCAD software (Autodesk Inc., Mill Valley, CA/USA) - Kerry Ultrasonic Cleaning Bath, Guyson (Skipton, North Yorkshire, UK) - Automatic Dicing SAW DAD-321, Disco Corporation, (Tokyo, JP) - Nordiko 7000 magnetron sputtering system, Nordiko Technical Services Ltd (Havant, Hampshire, UK) - SVG Resist coater and developer track, Silicon Valley Group Inc. (San Jose, CA/USA) - DWL lithograph, Heidelberg Instruments (Heidelberg, DE)
Mold Fabrication
- Silicon substrate, University Wafer (South Boston, MA/USA) - Alconox solution, Alconox Inc. (White Plains, NY/USA) - Acetone (99,6%), LabChem Inc. (Zelienople, PA/USA) - IPA, (99,9%), LabChem Inc. (Zelienople, PA/USA - SU-8 50 photoresist, Microchem Corp. (Newton, MA/USA) - SU8 2015 photoresist, Microchem Corp. (Newton, MA/USA) - AZ 40 XT photoresist, MicroChemicals Corp. (Ulm, Germany) - propylene glycol ether acetate (PGMEA) (99,5%), Sigma-Aldrich (St. Louis, MO/USA) - AZ 400 K developer, MicroChemicals Corp. (Ulm, German) - DI water
- Vertical laminar airflow cabinet, FASTER-BSC-EN (Cornaredo, IT) - Spinner, Laurel Corp. - UVO cleaner 1444AX-220, Jelight Company, Inc. (Irvine, CA/USA) - Hotplate, Stuart (Stafforshine, UK) - UV light (254 nm, 400 W), UV Light Technology Limited (Birmingham, UK) - Stereo microscope, AmScope (Irvine, CA/USA) - Alpha-step 200 profilometer, Tencor Instruments
Fabrication of PDMS
channels
- Sylgard 184 PDMS and curing agent, Dow Corning (Midland, MI/USA) - Micro slides 0.8 mm, Assistent (Sondheim, Germany) - Rounded syringes tips (20 and 18 Gauge), Instech Laboratories, Inc. (Plymouth Meeting, PA/USA)
- Analytical scale d=0.0001g, Scientech (Bradford, MA/USA) - Vacuum desiccator, Bel-Art Products (South Wayre, NJ/USA) - Spin coater, Laurel Corp. - Oven loading model 100-800, Memmert (Schwabach, DE) - Expanded oxygen plasma cleaner PDC-002-CE, Harrick Plasma (Ithaca, NY/USA)
22
μm thick was spin-coated onto the recently deposited aluminium layer. The designs in the AutoCAD file
were then converted and transferred to a DWL laser direct write lithography system (diode laser: 405
nm, blue) that performed a photolithographically transfer of the pattern in the CAD design to the
photoresist. A development step of the photoresist took then place, leaving exposed portions of the
aluminium layer, subsequently removed by a wet chemical etching process with aluminium etchant. To
finish, the remaining photoresist was removed creating a patterned aluminium mask. All the
microfabrication steps previously described were completed under class 100 (100 000 particles over 1
μm, per m3) cleanroom conditions with exception of the photolithography step, which was performed in
class 10 (10 000 particles over 1 μm, per m3) conditions.
Figure 12. Sequence of steps involved in the fabrication of the aluminium hard masks.
2.1.2. Master mold fabrication
The master molds fabrication consisted on several stages of deposition and development of
layers of both different photoresists and heights which were based on previous work56,57. For a clearer
understanding of the experimental procedure, Table 3 shows the runsheet for each master mold and
Figure 13 shows the schematics of different steps involved in the fabrication.
Table 3. Runsheet comprising the sequential steps involved in each mold fabrication process for the different design
iterations, including iterations where two different master molds are necessary (3 and 4).
Fluidic layer mold
Fabrication Steps Iteration 1 Iteration 2 Iteration 3 Iteration 4
Step 1 Silica substrate
cleaning Silica substrate
cleaning Silica substrate
cleaning Silica substrate cleaning
Step 2 20 µm layer 20 µm layer 100 µm layer 20 µm layer
Step 3 100 µm layer 100 µm layer 35 µm layer 100 µm layer
Step 4 Hard bake at 150 ºC
for 15 minutes Hard bake at 150 ºC
for 15 minutes
Final bake by ramping up from 100 to 125 ºC, followed by 5 minutes
baking
35 µm layer
Step 5 - - - Final bake by ramping up
from 100 to 125 ºC, followed by 5 minutes baking
Valves layer mold
Step 1 - - Silica substrate
cleaning Silica substrate cleaning
Step 2 - - 50 µm layer 50 µm layer
- - Hard bake at 150 ºC
for 15 minutes Hard bake at 150 ºC for 15
minutes
23
2.1.2.1. Silicon substrate cleaning
To start the process, a silicon 5x5 cm substrate was cleaned by successive washing steps with
acetone, IPA and DI water to eliminate any residues of photoresist (during the dicing process from the
silicon wafer, photoresist was used to protect the substrate) or other impurities on the surface.
Afterwards, the substrate was immersed in an Alconox solution for 15 minutes at 65 °C, followed by
cleaning with DI water and drying with compressed air. To finish, the substrate was put in a UVO cleaner
for 20 minutes in order to eliminate any lasting organic material on the surface.
2.1.2.2. 20 μm layer
SU-8 2015 (negative photoresist) was spin-coated on top of the substrate for 10 seconds at 500
rpm with an acceleration of 100 rpm/s, followed by a step of 34 seconds at 1700 rpm with an acceleration
of 300 rpm/s, that originated a 20 μm thick layer as measured in the profilometer. Then, a pre-exposure
bake (95 ºC) for 4 minutes was performed in a hot plate with a later step for cooling down of 1 minute.
The hard mask correspondent to the 20 μm design was positioned over the recently deposited resist
layer, carefully placing the aluminium surface facing down so that loss in resolution due to scattering
effects would not happen. The stack was then exposed to UV light during approximately 30 seconds.
One more baking step took place for 5 minutes at 95 °C followed by a 2 minutes’ step for cooling down
to room temperature. Immersion of the substrate in a PGMEA solution was then performed for
development of the pattern exposed on the photoresist with a duration of 2 minutes under manual
agitation. After the development, the substrate was rinsed with IPA and dried with compressed air.
2.1.2.3. 35 μm layer
AZ 40XT (positive photoresist) was spin-coated on top of the substrate for 10 seconds at 500
rpm with an acceleration of 100 rpm/s, followed by a step of 21 seconds, at 2000 rpm with an
acceleration of 1000 rpm/s, that originated a 35 μm thick layer, checked in the profilometer. The spin-
coated photoresist was allowed to settle for 30 minutes to prevent bubbles from being entrapped in the
bulk of the photoresist during the subsequent baking steps. Then, the spin-coated substrate was placed
on a hot plate at 100 ºC, slowly ramping up to 125 ºC, where it continued baking for 5 minutes. The hard
mask correspondent to the 35 μm features was placed on top of the substrate, facing down. The
exposition step took then place during approximately 3 minutes and 30 seconds. Then, a post-exposure
bake was performed at 105 °C for 2 minutes. The mold was then developed by immersion for 10 min in
AZ 400 K developer previously diluted in DI water to a ratio of 1:3, and then washed with DI water.
Finally, a reflow step was performed by ramping up the temperature from 100 °C to 125 °C and then
baking for 5 minutes.
24
Figure 13. Sequence of steps involved in the fabrication of the SU-8 negative photoresist mold (iteration 1, 2 and
valves layer molds) and of SU-8 negative and AZ 40 XT positive photoresist molds (iteration 3 and 4 fluidic layer
molds).
2.1.2.4. 50 μm layer
A SU-8 50 photoresist (negative photoresist) was spin-coated on top of the substrate during 10
seconds at 500 rpm with an acceleration of 100 rpm/s, then followed by 37 seconds at 2300 rpm with
an acceleration of 300 rpm/s. Profilometer measurements were done to confirm the desired height. The
pre-exposure bake step in this stage included sequential baking by heating up to 65 ºC during 3 minutes,
then further heating at 95 ºC for 8 minutes, and finally to cooling down for 1 minute. The hard mask with
the 50 μm features was then manually aligned with the aluminium surface facing down. Exposition to
the UV light was performed during about 25 seconds. Afterwards, a post-exposure bake was done at
65 °C for 1 minute, followed by 7 minutes of heating at 95 °C and, lastly, a 2 minutes’ step for cooling
down. Again, the photoresist layer was developed in PGMEA for about 6 minutes with manual agitation,
washed with IPA, and air dried.
2.1.2.5. 100 μm layer
SU-8 50 photoresist was spin-coated on top of the substrate during 10 seconds at 500 rpm with
an acceleration of 100 rpm/s, followed by 30 seconds at 1000 rpm with an acceleration of 300 rpm/s.
25
The pre-exposure bake step in this stage included sequential baking by heating up to 65 ºC during 10
minutes, then further heating at 95 ºC for 30 minutes, and finally cooling down for 1 minute. The hard
mask with the 100 μm features was then manually aligned with the aluminium surface facing down.
Exposition to the UV light was performed during about 70 seconds. Afterwards, a post-exposure bake
was done at 65 °C for 1 minute, followed by 10 minutes of heating at 95 °C and, lastly, a 2 minutes’ step
for cooling down. Again, the photoresist layer was developed in PGMEA for about 10 minutes with
manual agitation, washed with IPA, and dried.
2.1.3. Fabrication of PDMS channels
The finished master molds were then used for the fabrication of the PDMS channels. For the
microfluidic structure iterations that included valves, besides different master molds they also included
different fabrication processes for the PDMS channels which are described in this section along with the
standard fabrication process and were based on recent published work from the group57. A scheme
highlighting the major differences in the process is shown in Figure 14.
Figure 14. Sequence of steps involved in the fabrication of the PDMS channels for the different iterations, using
distinct fabrication methods: soft lithography with irreversible sealing done by plasma treatment (iteration 1 and 2);
multilayer soft lithography and sealing using an adhesive PDMS layer (iteration 3 and 4A) or plasma treatment
(iteration 4B).
26
2.1.3.1. Process for iteration 1, 2 and microcolumns
The first step was to fix the mold in a Petri dish with the patterns facing up with little pieces of
tape. PDMS elastomer preparation was done by the following experimental procedure: a curing agent
and PDMS mixture was prepared in a plastic cup in a 10:1 ratio, followed by a degasification step for 30
minutes in a desicator. After this, the mixture was poured on top of the mold (inside the Petri dish) until
it fills about half of the height of the latter. The PDMS was cured for 90 minutes at 70 ºC for curing. The
cured PDMS was afterwards cut by means of a scalpel and peeled off from the mold with appropriate
tweezers. Holes were punched on the structures with rounded 20 and 18 Gauge needles for the outlets
and inlets, respectively. A 500 μm thick PDMS membrane was prepared with the purpose of sealing the
PDMS structures, by spin-coating PDMS from the mixture previously prepared on top of a silicon wafer
at 250 rpm for 25 seconds with an acceleration of 100 rpm/s. A baking step was performed as described
above and cut in parts with an appropriate size. Both the structures and the membranes were taken into
an oxygen plasma cleaner in order to seal the two against each other, immediately after plasma
treatment.
2.1.3.2. Process for iteration 3 and 4A
The procedure for PDMS preparation was the same as described in the previous section, in
addition that PDMS on a ratio of 20:1 was also prepared. The PDMS preparation of 10:1 ratio was
poured on top of the fluidic layer mold and baked at 70°C for 50 min while the PDMS preparation of 20:1
ratio was spin-coated on top of the valves layer mold for 25 seconds at 300 (400 μm thickness), 400 or
500 rpm with an acceleration of 100 rpm/s and baked at 70 ºC for 40 minutes. After the half-curing
process, the PDMS was taken from the mold and inlets/outlets were punched using 20 Gauge blunt
syringe tips. The fluidic layer was then aligned on top of the valves layer and the stack was baked
together at 70 °C for 90 minutes. Completed the curing process, the structures were cut and
subsequently taken from the mold with both layers irreversibly sealed against each other. The inlets
corresponding to the valves channels were punched using again 20 Gauge blunt syringe tips. Instead
of an oxidation treatment, sealing was achieved by gently wet the PDMS stack on a PDMS adhesive
layer obtained by spin-coating PDMS 10:1 on top of a coverglass (1 mm thick) for 15 seconds at 500
rpm, followed by 30 seconds at 3000 rpm and, finally, 4 minutes at 6000 rpm. Then, the stack was
sealed against a 100 μm thick glass slide in which PDMS was previously spin-coated for 5 seconds at
500 rpm, followed by 20 seconds at 4200 rpm, resulting in a PDMS-PDMS sealing after a final bake at
70 °C for 90 minutes.
2.1.3.3. Process for iteration 4B
Both the PDMS preparation and casting was done as described in the previous section only
varying the PDMS curing time from 50 (fluidic layer) and 40 (valves layer) minutes to 90 minutes for
27
each layer. The two different PDMS layers were taken into an oxygen plasma cleaner in order to seal
the two against each other, immediately after plasma treatment, creating a PDMS stack that was further
sealed against a 1 mm-thick glass slide in which PDMS was previously spin-coated for 5 seconds at
500 rpm, followed by 20 seconds at 4200 rpm, resulting in a PDMS-PDMS sealing using again the
oxygen plasma cleaner. Between each plasma treatment, the structures were allowed to stabilize a
minimum of two hours before being used in order to increase the robustness of the bonding and avoid
eventual sealing of the valves channels to the surrounding PDMS, upon actuation with the compressed
air.
2.2. Microfluidic device manipulation
The fabrication of the microfluidic devices comprised several steps, namely separation channel
packing and separation and elution studies. The reagents and materials used for all these steps and the
required equipment are listed in detail in Table 4.
Table 4. Materials and equipment required for microfluidic device manipulation.
2.2.1. Separation channel packing
Different beads were used for each experiment were used throughout the experimental work as
well different microfluidic structure configurations which demanded both tailored bead preparation
protocols and packing methodologies. Both subjects are further described in this section.
Step of the process
Reagents/Materials Equipment/Facilities
Separation channel packing
- Insulin syringe 1 mL U-100 Luer-Lock, Codan (Lensahn, DE) - Tubing couplers SC20/15, Instech Laboratories, Inc. (Plymouth Meeting, PA/USA) - Polyethylene tubing (BTPE-90), Instech Laboratories, Inc. (Plymouth Meeting, PA/USA) - Milli-Q water 18 MΩ CM, Millipore - Polyethylene glycol (PEG) 8000 MW, Sigma-Aldrich (USA) - Ethanol (96%), Sigma-Aldrich (USA) - CM-Sepharose Fast Flow Beads 90 μm, GE Healthcare Life Sciences 17-0719-05, (USA) - Spherical C18 bonded flash silica beads 45-75 μm, 97727-U Supelco Analytical (Bellefonte, PA) - Spherical flash silica beads 45-75 μm, 97728-U Supelco Analytical (Bellefonte, PA) - Triton X-100, Sigma-Aldrich (USA)
- Syringe pump NE-1002X, New Era Pump Systems, Inc. (Farmingdale, NY/USA) - 170 mesh cell strainer Alfa Aesar, ThermoFisher (Karlsruhe, GE) - Leica DMLM Microscope, Leica Microsystems (Wetzlar, DE)
Separation and elution
studies
- Insulin syringe 1 mL U-100 Luer-Lock, Codan (Lensahn, DE) - Tubing couplers SC20/15, Instech Laboratories, Inc. (Plymouth Meeting, PA/USA) - Polyethylene tubing (BTPE-90), Instech Laboratories, Inc. (Plymouth Meeting, PA/USA) - Milli-Q water 18 MΩ CM, Millipore - Methylene blue stain, Sigma-Aldrich (USA) - Yellow food colouring E104, E122, E260, 5160414 Globo, (UK) - Phosphate pH 7.5 1M NaCl at 25 ºC solution, Sigma-Aldrich (USA) - Phosphate pH 2.5 2M NaCl at 25 ºC solution, Sigma-Aldrich (USA) - Milli-Q water (18 MΩ CM), Millipore - HCL 1M solution, Sigma-Aldrich (USA) - FITC F427 (98%), Sigma Aldrich (Switzerland) - Aflatoxin B1 from Aspergillus flavus, Sigma-Aldrich (USA) - OTA from Aspergillus ochraceus, Sigma-Aldrich (USA) - Acetonitrile solution (99,8%), Sigma-Aldrich (USA) - Methanol solution (99.9%), Sigma-Aldrich (USA) - white polystyrene 96 well plates, Corning (NY/USA)
- Syringe pump NE-1002X, New Era Pump Systems, Inc. (Farmingdale, NY/USA) - Fluorescence spectrometer Varian, Cary Eclipse - Leica DMLM Microscope, Leica Microsystems (Wetzlar, DE) - ImageJ software (USA)
28
2.2.1.1. Beads preparation protocols
Commercially available beads were supplied in an ethanol solution (20% v/v). For CM-
Sepharose Fast Flow Beads, the first step was sieving the beads, due to their large particle size
dispersion (45 μm-165 μm), using a 170 mesh. The next step was to homogenize the bead suspension
and transfer a certain volume of the bead stock solution into a PEG 8000 20% (w/w) solution in a
proportion of 1:4. PEG 20% was used to ensure beads were homogeneously dispersed and to prevent
settling, which would cause clogging of the microchannel with clumps of beads before an adequate
packing being achieved. For the C18 beads a different protocol was followed to prevent beads
aggregation in solution due to their high hydrophobicity and assure an easy packing inside the
separation channel. From a commercial dry stock of C18 beads, approximately 10 mg were weighed in
an analytical balance and further mixed in 600 μL of previously prepared PEG 20% and 2 μL of Triton
X-100. The prepared solution was mixed continuously for about 30 seconds in the vortex, repeating the
step 2-3 times until complete homogenization was achieved. For silica beads and different beads used
throughout the experimental work, when not specified otherwise, the following preparation protocol was
used: beads were always prepared in stock solution of 20% ethanol and further homogenised in vortex
to then collect a certain volume of stock solution that was added to a PEG 8000 20% (w/w) solution in
a proportion of 1:4 and finally used in the packing and separation experiments.
2.2.1.2. Packing method
In order to pack the beads rapidly and efficiently, three different packing methodologies were
used throughout the work: either a syringe pump on a pushing or pulling mode or a compressed air
packing. To pack the separation channel using the pushing mode (positive pressure applied) the flow
applied to fill the separation channel varied in a range of 10-20 μL/min by action of a syringe pump.
Beads suspension entered the column in a 20 Gauge inlet via a capillary tubing pre-filled with water and
beads solution and connected to the structure using a metal coupler. Caution measures were taken in
order to the capillary tube to be free of air bubbles. To avoid channel clumping during filling, continuous
gentle peristaltic movements were performed on the channels structure with a tweezers, again to avoid
clumping on the channels. Alternatively, using the syringe pump in pulling mode the flow was driven by
applying a negative pressure at the outlet, via capillary tubing pre-filled with water and connected to the
microfluidic structure with a metal coupler inserted into the outlet. A pipette tip containing the bead
suspension was inserted into the inlet and the packing was performed at a flow rate between 10 to 15
µL/min. Finally, the compressed air packing consisted in filling the separation channel with beads
solution using capillary tubing connected to compressed air lines and subjected to a pressure of 0.15
MPa. The compressed air lines were switched on or off via solenoid valves connected to a compressed
air supply that were controlled through a printed circuit board (PCB) supplied by a 24 V power source
(Figure 15).
29
Completed the filling of the channels, inlets and outlets were sealed using closed metal plugs
to avoid solution loss and the structure was stored on the fridge immersed in DI water, approximately at
4 ºC for further uses.
2.2.2. Separation and elution studies
Experimental work in order to validate the functionality of the microfluidic structures and assess
the potential application on food safety analysis was done and the methodologies are described below
in this section.
2.2.2.1. Food colouring separation (proof-of-concept)
Filtered CM-Sepharose Fast Flow bead suspension was prepared and packed inside the
separation channel as described in section 2.2.1.1 and 2.2.1.2. Then, Milli-Q water was flowed during
about 10 minutes at a flow rate of 15 µL/min in order to wash the separation channel from the
components of the beads packing solution. A green solution resulting from the mixture of methylene
blue and yellow food colouring in a proportion of 1:1 was prepared. A volume of about 10 μL of green
solution was used to insert in a capillary tubing pre-filled with water and connected to the structure using
a metal coupler inserted on the cross-shaped injector inlets. To insert a small plug in the channels, using
a flow rate of 3 μL/min, until near 1/3 of the total volume of the green solution entered the channel.
Afterwards, Milli-Q water was used as mobile phase, with a constant flow rate of 8 μL/min in pushing
mode (positive pressure at the inlet) by action of a syringe pump. Elution of the methylene blue retained
on the beads was tested with 3 different buffers/solutions: (i) phosphate buffer at pH 7.5 containing 1 M
NaCl, (ii) phosphate buffer at pH 7.5 containing 2 M of NaCl, (iii) 1 M HCl solution. The flow rate applied
Figure 15. Experimental setup for compressed air packing and valves actuation. (A) Switch buttons for on/off
actuation on the valves. (B) and (C) PCB for controlling of valves actuation and respective power source. (D)
Syringe pump. (E) Microfluidic device connected to the system by capillary tubing.
30
for elution steps was always a constant value of 13 μL/min during at least 20 minutes, in order to
completely remove the methylene blue retained on the beads. After each elution assay, the column was
washed with Milli-Q, during about 10 minutes at 15 µL/min, in order to remove any residual
buffer/solution flowed previously.
2.2.2.2. OTA and AFB1 microchannel adsorption and elution
OTA and AFB1 stock solutions at 100 µg/mL were used to prepare diluted solutions in Milli-Q
water of 10, 6, 3 and 1 µg/mL each. Adsorption was performed in already packed microchannels
designed in previous work56 packed with C18 beads. Milli-Q water was flowed during about 5 minutes
at a flow rate of 15 µL/min in order to wash out the components of the bead packing solution. A pipette
tip containing 10 µL of 6 µg/mL of each toxin solution was inserted in the inlet and the solution was
flowed at a flow rate of 15 µL/min, by applying a negative pressure at the outlet. Acetonitrile and
methanol solutions of 100, 75, 50, and 25% (v/v) in Milli-Q water were prepared and tested with respect
to the ability of eluting the adsorbed toxins elution. Acetonitrile and methanol solutions were flowed
inside the microchannel by applying a positive pressure (pushing) at the inlet at a flow rate of 5 µL/min
for 10 minutes. For image acquisition during elution, auto fluorescence of the toxins was monitored
every 30 seconds using a DFC300FX digital camera coupled to a Leica DMLM microscope with UV light
excitation filter. The UV light excitation filter has a band pass illumination path at wavelengths between
355 and 425nm and a long pass observation path above 470nm. Images were taken with an exposure
time of 1 second and 1x gain and the average fluorescence intensity was measured using ImageJ
software.
To assess the solvent effect in the fluorescence of the toxins measured with the miscroscope,
solutions of 6 µg/mL of both OTA and AFB1 in 100, 75, 50, 25%, 0% (v/v) of either acetonitrile or
methanol in Milli-Q water were prepared and placed on 96 well plates. Measurements were performed
in a fluorescence spectrometer with an excitation and emission wavelengths of 333nm and 460 nm,
respectively, and excitation/emission slits of 5 nm.
2.2.2.3. OTA and AFB1 cross-channel injection and elution
In order to study the feasibility of AFB1 elution within a packed column with a cross-channel
injector (iteration 2) and also the separation of AFB1 and OTA within the same column, the fluorescence
of the toxins was followed in the packed channel. C18 bead suspension was prepared and packed inside
the separation channel as described in sections 2.2.1.1 and 2.2.1.2. Milli-Q water was flowed during
about 10 minutes at a flow rate of 15 µL/min in order to wash out the components of the bead packing
solution. A pipette tip containing 60 µL of 6 µg/mL of AFB1 solution (for the elution) or 60 µL of 6 µg/mL
of AFB1 + OTA solution (for the separation) was inserted in the inlet and the solution was flowed at a
flow rate of 15 µL/min, by applying a negative pressure at the outlet of the cross-channel injector.
31
Acetonitrile 100% v/v (for the elution) and 25% v/v (for the separation) in Milli-Q water solutions were
flowed inside the microchannel by applying a positive pressure (pushing) at the inlet at a flow rate of 2
µL/min for approximately 10 minutes. For image acquisition during elution, auto fluorescence of the
toxins was monitored using a DFC300FX digital camera coupled to a Leica DMLM microscope with UV
light excitation filter. The UV light excitation filter has a band pass illumination path at wavelengths
between 355 and 425nm and a long pass observation path above 470nm. Images were taken with an
exposure time of 1 second and 1x gain.
2.2.2.4. Plug optimization
In order to study plug formation and progression inside a microfluidic structure with a cross-
shaped injector actuated by pneumatic valves (iteration 4B), plug optimization assays were performed.
An ethanol 96% (v/v) solution was flowed by applying a positive pressure (pushing) using a syringe
pump at a flow rate of 15 µL/min during 20 minutes. All inlets should be closed using a pipette tip filled
with liquid to avoid bubble trapping inside the injector and separation channel. Microfluidic structures
with and without packed silica beads were used for assessment of the effect of beads in plug
progression. FITC at 1 mg/mL solution in ethanol 96% was prepared and a volume of 5 µL was loaded
onto the cross-shaped injector by pulling at 15 µL/min and subsequent actuation of the valves. The
ethanol 96% solution was flowed inside the channel at a flow rate of 0.5 µL/min by pushing with one or
more syringe pumps. For image acquisition, during operation, the fluorescence of the FITC solution was
monitored using a DFC300FX digital camera coupled to a Leica DMLM microscope with a blue light
excitation filter. The blue light excitation filter has a band pass illumination path at wavelengths between
450 and 490 nm and a long pass observation path above 515nm. Images were acquired with an
exposure time of 800 milliseconds and 3x gain.
32
3. Results and Discussion
3.1. Microfluidic device design considerations
The first step into the development of the chromatographic microfluidic device was the
conception of the design of both the separation channel and later, the injector geometry. Many factors
must be taken into consideration, from the aspect ratio height/width to particle size to the channel length
and beads trapping within the column46. Conventional columns typically comprise a glass, plastic or
metal cylinder with a packed stationary phase, and various plumbing arrangements to allow introduction
of the sample and collection of fractions with minimal mixing, peak distortion, dilution and sample loss.
In a HPLC, the most common columns have a diameter around 4.6 mm and a length from 10 to 15 cm,
packed with 5 µm beads, although many other sizes can be found16. In microfluidics, microchannels
(analogous to the column) are typically around 100-150 μm in height and can have varied widths and
lengths, only restrained by the scale of wafers and lithography processes.
In Figure 16, the comparison between the two initial CAD iterations for the design of the
microfluidic device are shown. Both iterations have the same separation channel height and width, 100
and 400 µm, respectively. The ratio of channel width (or height, depending on the characteristic length)
to particle size must be considered to avoid the known “wall effect” where quality/density of packing or
flow behaviour will be severely different from the bulk channel, which is very common in microfluidic
devices where aspect ratios of width to particle size are usually around 1046,47. In the case of our
separation channel dimensions, the aspect ratio will be of 4.4 and 6.9 for the beads with mean diameters
of 90 and 58 µm, respectively, that were used throughout the experimental work. For values much lower
than 10 the wall and the bulk region become very similar, therefore such effect loses its relevance.
However, the aspect ratios values are still very close to ignore the effect on quality/density of packing
and flow behaviour which will diminish the column efficiency. To overcome this limitation, several
strategies could be employed on the design: (i) decrease the width of the channel, (ii) decrease the
particle size or (iii) increase the length of the channel. Decreasing the width of the channel would lead
to values much lower than 10 and therefore the effect would not be significant, as previously mentioned.
However, this would also reduce the surface area available for interaction with the analytes and mobile
phase since the number of beads packed would severely decrease. On the other hand, decreasing the
particle size would increase the value of the aspect ratio to 80 if, for example, 5 µm beads were used
and therefore the wall region would only constitute a small part of the channel, reducing the impact of
this effect. This has been a strategy used both in conventional and on-chip HPLC44, however with the
consequence of increased pressure demand on the devices. Also, due to the bead trapping feature
being based on the different heights of the channel and outlet, such a small outlet cannot be fabricated
with the current instrumentation and other bead trapping strategies would have to be employed. Finally,
there is also the opportunity of increasing the length of the channel. Based on theoretical knowledge on
chromatography, it is well known that an increase in length of the column will provide a higher column
33
efficiency value (N) as shown previously in Equation 2 (𝑁 = 𝐿 𝐻⁄ ), where N is the number of theoretical
plates, L the length of the column and H the plate height. This would compensate that loss in separation
efficiency due to the wall effect.
Therefore, an initial length of 30 cm was chosen for the separation channel, being possible by
the serpentine design geometry chosen (Figure 16, 1-B) that maximizes the area available within the
5x5 cm mold. However, due to restrictions in the applied pressures (see section 3.3) during beads
packing, in iteration 2 the separation channel length had to be reduced on more than 50% (12 cm) in
order to be possible to pack the whole channel. The impact of this reduction will be further addressed in
the next sections.
Figure 16. Comparison between the first two CAD iterations created using AutoCAD. (1-A) corresponds to a close-
up representation of the bead trapping feature near the outlet where the channel has two different heights: 100 and
20 µm. (1-B) corresponds to a microscopic image of the master mold fabricated with the serpentine separation
channel with a total length of 30 cm. (1-C) corresponds to the microscopic image of the master mold where the
bead trapping can be seen. (2-A, B) highlight the simple cross-channel injector representation introduced into the
design near the inlet and a microscopic image of the injector on the master mold, with the two channels of different
heights (100 and 20 µm). (2-C) Microscopic image of the bead trapping feature and serpentine separation channel
with a total length of 12 cm.
34
Another important aspect common to both designs is the bead trapping feature which plays with
the difference in the heights of the channels to entrap the beads (Figure 16, 1-A, C). The main purpose
was to retain the bead packed bed by creating a outlet channel which has a height much smaller than
the average bead size, working as a barrier for beads. Since the beads to be used in further experimental
work had average particle size of no more than 100 μm, channel heights were defined as 100 and 20
μm for the larger and smaller channel, respectively. The trapping feature proved to be efficient as it can
be seen further ahead in section 3.3, also probably allowing better column performance in terms of band
broadening since it does not require so much dead volume and disturbance in flow as other solutions51.
The spacing between channels was always above 400 µm to prevent problems in the microfabrication
process (for example, detaching of the photoresist layers from the substrate due to extreme small
features) which proven to be enough for robust fabrication as seen in the microscopic images of Figure
16.
One difference to note between the two iterations are the inlets and outlets, at two levels: first,
both inlet and outlets were reduced in diameter (Figure 16, iteration 2) due to collapsing of the inlet on
the PDMS membrane that was used to seal the structures; secondly, the formation of clogs at the inlet
when inserting the bead suspension led to alterations in the inlet shape so that beads flow would be
facilitated at the beginning of the separation channel.
Finally, the single inlet geometry was replaced by a cross-channel injector in iteration 2 (Figure
16, 2-A, B) to allow smaller and well-defined sample plugs. The importance of the injector as a
component in a microfluidic separation system arises from the fact that it defines the shape and quantity
of analyte that will be used for separation and analysis. The cross-channel injector is a simple
intersection of two channels that also have different heights, similarly to the beads trapping feature, to
ensure beads are kept within the separation channel. It is operated by first loading the beads solution
into the central inlet (oval-shaped), followed by insertion of sample solution on the perpendicular channel
inlet (rounded-one). A defined sample plug will be originated and travel through the separation channel
by loading the mobile phase on the central inlet. However, this cross-channel geometry controlled off-
chip with syringe pumps originated still an uncontrolled and irreproducible loading of sample and if not
carefully handled by an experienced user, the plug created could be as large as one defined without a
injector. These problems arouse mainly due to the rapid plug axial diffusion with time and pressures
differences created from insertion and removal of metal plugs to control fluid flow.
Currently, attempts to adapt pressure-driven separations to a microfluidic device requires the
use of conventional, off-chip sample definition or very complex on-chip injectors12. The large volume
and dead time introduced by an off-chip injection hampers the ability of a microfluidic separation to
deliver rapid and sensitive analyses. A new technique developed by Unger et al. designated multilayer
soft lithography39,40 allows the fabrication of integrated pumps and valves using elastomeric material,
such as PDMS. In Figure 17, a comparison between iterations 3 and 4 is shown where the major
35
difference between the two designs is the injector geometry and location and how the pneumatic
actuated valves, based on the work of Unger et al., were integrated within the microfluidic device.
In iteration 3, the cross-channel injector geometry was maintained, still using the height
difference feature to entrap the beads within the separation channel. However, the inlets for beads
insertion and mobile phase flow were separated (Figure 17, 3-C) so that once totally packed, the
separation channel would be blocked to prevent the unwanted movement of the packed bed and sample
plug when pressure differences were created from the insertion/removal of the metal plugs to control
fluid flow. Valves were incorporated underneath the cross-channel injector inlet and outlets and on the
separation channel outlet with an actuation area of 200 x 200 µm. Controlling the fluid loading and flow
using integrated valves provided the generation of well-defined and controlled plugs and injections.
However, elastomers deform in response to pressure, creating fluidic capacitance that allied to the
cross-channel geometry caused the same problems related to pressure differences from the injector of
Figure 17. Comparison between the third and fourth CAD iterations created using AutoCAD. (3-A) corresponds to a
close-up representation of the cross-channel injector with integrated pneumatically actuated valves. (3-B) corresponds
to a microscopic image of the valves mold fabricated showing a single valve channel. (3-C) corresponds to the
microscopic image of the master mold where the separation channel and cross-channel injector can be seen. (4-A,
C) A close-up representation and microscopy image, respectively, highlights the connection channel (height: 20 µm)
designed between the new cross-channel injector (height: 35 µm) and the separation channel (height: 100 µm) which
entraps beads within the separation channel. (4-B) Microscopic image of the new incorporated cross-channel injector
with integrated pneumatically actuated valves.
36
iteration 2 as seen in Figure 18-A where beads are invading the cross-channel injector and leaving the
packed bed.
To overcome the difficulties encountered in the cross-channel geometry, the injector was
detached from the separation channel to not intersect with the packed beads bed and interfere with
beads packing and plug formation. In iteration 4, the injector consists of two perpendicular channels
(Figure 17, 4-B) with the same height (35 µm) that were pneumatically actuated with the valves channels
underneath (actuation area of 400 x 400 µm). From the microscopic image it is possible to note that the
injector channels on the mold have rounded edges, a requirement for every channel that will be actuated
with the pneumatic valves, since they can only fully close if the channel has a round cross-section. A
connection channel (Figure 17, A-C) of 20 µm height was designed to connect the injector and
separation channel and prevent beads invasion to the injector, using the same principle of the beads
trapping feature. In Figure 18-B, it is possible to see that although the beads can invade the connection
channel at packing pressures higher than 0.2 MPa, they do not reach the cross-channel injector. Finally,
the valve actuating on the outlet of the separation channel was removed in iteration 4 since the
resistance to fluid flow created from the packed bed was so high that any pressure difference effect that
could compromise plug injection was not relevant.
Figure 18. Comparison between the two injectors from iteration 3 and iteration 4. (A) Microscopic image of the
cross-channel injector and separation channel from iteration 3, packed with silica beads. (B) Microscopic image of
the cross-channel injector, connection channels and separation channel from iteration 4, packed with silica beads.
37
3.2. Fabrication process considerations
Besides improving the design from iteration to iteration, many improvements were also needed
regarding the fabrication process itself, especially related to the valves optimization and the sealing of
the microfluidic devices comprising iteration 3 and 4.
3.2.1. Valves thickness optimization
The working principle of the pneumatic valves is based on the deflection of a PDMS membrane
(valves layer) beneath the cross-channel injector, caused by the injection of compressed air in the valve
channels filled with water. It is important to highlight again that the actuated channels need to be
fabricated with a round cross-section, to ensure their complete closure. Based on previous work from
the group57, the needed PDMS valves’ layer thickness for complete closure of the channels by the valves
at 0.1 MPa was 400 µm (300 rpm spin-coating speed). Therefore, the valves layer was first fabricated
using this spin-coating speed. However, as seen from Figure 19-E, the actuation of this valve at the
same pressure (0.1 MPa) in the microfluidic device was not enough to fully close the channel. The
different results can be caused by the difference in the actuation area of the valve or by the pressure
that device accumulates, which is much higher than the device from the reported work due to more
complexity in the design and a higher separation channel length. Applying higher pressures to close the
Figure 19. Microscopic images PDMS microfluidic structures (iteration 3) comprising valves of different thicknesses
originated by different spin-coating speeds, and its actuation. (A, B and C) are microscopic images of valve channels
spin-coated at 300, 400 and 500 rpm originating decreasing valves layer thickness. (D and E) are microscopic
images of the actuation of valve channels at P = 0 and 0.1 MPa, respectively, for a 300 rpm spin-coated layer. (F)
is a microscopic image pf the actuation of valve channel at P = 0.1 MPa for a 400 rpm spin-coated layer. Green
food colouring was flowed through the injector channel for better visualization of valves actuation.
38
valves eventually led to rupture of the microfluidic structure, consequently decreasing the valves layer
thickness by increasing the spin-coating speed was the followed strategy to overcome this limitation.
Spin-coatings speeds of 400 and 500 rpm were further tested as seen in Figure 19-B and C, respectively.
At 500 rpm, the valves layer eventually became too thin causing the valve channels distortion due to the
pressure applied when sealing the several layers of PDMS into one microfluidic structure. On the other
hand, the 400 rpm spin-coating speed allowed for a good fabrication quality without channel distortion
(Figure 19-B) and a P = 0.1 MPa was enough to fully close the channel, as seen in Figure 19-F.
Therefore, the 400 rpm spin-coating speed was the chosen one for further fabrications using either
iteration 3 or 4 microfluidic structures.
3.2.2. Microfluidic structure sealing optimization
Sealing of the PDMS microfluidic structures is an important step during fabrication, and many
techniques have been employed in order to achieve really robust bonding onto PDMS membranes58 or
glass59. During the experimental work done, both the conventional plasma cleaning strategy and
bonding with PDMS glue and half-curing strategy were employed on the different iterations (Figure 20).
Plasma cleaner treatment (oxidation) creates silanol groups by converting –O2Si(CH3)2 to –
O2Si(OH)2. The surfaces undergo an irreversible bonding because of the siloxane covalent bonds (Si–
O–Si) that are formed during condensation reactions. The oxidative effect in PDMS is temporary owing
to the diffusion of oligomers from the bulk to the surface. Hence, the bonding step must be performed
Figure 20. Schematics illustrating the two strategies used for sealing of the microfluidics structures comprising
valves (iteration 3 and 4), namely bonding through half-cure and PDMS glue or bonding with conventional plasma
cleaner treatment.
39
within one minute after the oxidation treatment. This fact hinders the large-scale integration and steps
of alignment for complex devices.
Plasma cleaner treatment technique was employed for bonding glass or PDMS membranes to
microfluidic structures that comprised the iteration 2 design or the microcolumns due to its simplicity,
robustness and easy execution. However, for iterations 3 and 4 that included a multilayer microfluidic
structure, bonding between layers promised to be more complex. Based on fabrication troubleshooting
reports from previous work in the group, bonding the valves thin layer to the fluidic layer using plasma
cleaner treatment lead to collapsing of the valves during the sealing process (when both surfaces were
hydrophilic), causing irreversible bonding of the valves to the fluidic layer and consequently, the valves
membrane did not respond to actuation. Alternatively, MLS technique offers a bonding without oxidative
treatment, based on the capability to bond multiple patterned layers through half-curing of each layer
and baking them again stacked together to complete the curing process. The principle is to have two
opposite (the ratio of element A and B in the elastomer mixture is substantially unalike) layers in contact
with each other since reactive molecules remain at the interface between the layers and can be further
cured to produce a sealing. Finally, for iteration 3 and 4, besides the interlayer bonding, the final stack
had to be sealed against a glass surface which conferred more robustness to the final structure. Since
MLS cannot be employed to glass surfaces, a thin glass with a half-cured PDMS layer was used and
bonding was performed using PDMS glue (Figure 20). The PDMS glue method consists in wetting the
surface of the microfluidic stack onto a uncured 10:1 PDMS layer spin-coated on a glass. This adhesive
layer created on the microfluidic stack will then be aligned on top of the thin glass with the half-cured
PDMS layer and baked together to finish the curing process.
However, bonding the several layers and glass using MLS and PDMS glue offers low robustness
of the microfluidic structure, withstanding badly high pressures that lead to rupture and leaks of the
structure during operation. Therefore, the bonding using plasma cleaner treatment for iteration 4 was
also employed and compared with the alternative, as seen in Figure 20. Surprisingly, the bonding with
plasma treatment did not interfere the actuation of the valves, as seen in section 3.6, where the
microfluidic structures sealed through plasma treatment are being used without impediments.
Additionally, the structures fabricated with this process (iteration 4B) were more robust when compared
to iterations 3 and 4A and did not had leaks or ruptures as often as the previous ones, as expected. This
result may be explained by the fact that actuation areas of the valves reported on the work were larger
(700 x 500 µm), and therefore more easily collapsed during the sealing process unlike the ones used
on this microfluidic device (400 x 400 µm.)
3.3. Separation channel packing
One of the most important steps throughout the development of the microfluidic device was the
packing of the separation channel with beads. From the bead packing solution to, most importantly, the
packing methodology employed, many factors had to improved.
40
Table 5. The different packing solutions required for each beads type.
Beads type Packing solution
CM-Sepharose, 90 µm Beads suspension of ethanol 20% (v/v) on a PEG
8000 20% (w/w) solution in a proportion 1:4 Silica beads, 58 µm
C18 bonded silica beads, 58 µm 10 mg of dry C18 beads in a PEG 8000 20%
(w/w) solution + 2 µL of Triton X-100
In Table 5, the different beads types used throughout this work is listed and the corresponding
packing solutions used for each one. PEG 20% was used to ensure beads were homogeneously
dispersed and to prevent settling, which would cause clogging of the separation channel with clumps of
beads before an adequate packing being achieved. Due to its hydrophilic nature, PEG was a good
dispersant for silica and CM-Sepharose beads which are also more hydrophilic. However, a different
packing solution had to be used for C18 bonded silica beads since they were very hydrophobic and
would create aggregates, aggravating severely the packing of the beads into the channel. Additionally
to PEG 20%, Triton X-100 was added to the packing solution for C18 beads. Triton X-100 is a non-ionic
detergent, considered a mild surfactant with a hydrophobic character that helps to homogeneously
disperse beads and avoid aggregation to PDMS channel walls. Besides the packing solution, one factor
to consider during the beads preparation procedures was the particle size distribution of the beads. CM-
Sepharose beads have a particle size distribution of 45-165 µm, although the mean size is of 90 µm
which led to clogging problems in the channels. The maximum height of the channels was 100 μm,
being inevitable that beads got stuck in the channel, not allowing correct column packing. Therefore,
beads sieving was necessary to be included in the beads preparation procedure for CM-Sepharose
where they were sieved on 170 mesh that excluded every bead that was above 90 µm in size. After
sieving, beads were successfully packed within the separation channel as seen in 3.4, where
experiments were done using a microfluidic structure packed with sieved CM-Sepharose beads.
Three main strategies were employed to pack the beads within the separation channel as
illustrated in Figure 21 (centre). The first one was to apply a negative pressure at the outlet of the
separation channel (packing by pulling) at a constant flow rate which would flow the beads from pipette
tip into the channel. The second one was to directly flow the beads solution by applying a positive
pressure (packing by pushing) at the inlet of the separation channel, maintaining the flow rate constant.
Finally, the third strategy was to pack the beads by directly flowing the beads solution into the inlet of
the separation channel at a constant pressure applied by a compressed air line. The pressure applied
to flow the beads could be controlled up to 0.3 MPa. The first two strategies are based on packing by
maintaining the flow rate constant, unlike the latter, which packs the separation channel by controlling
the pressure applied. This difference was very important to be able to pack the very long separation
channels used in this microfluidic device.
41
The “packing by pulling” strategy would be preferential since the use of a negative pressure
provides a gradual increase in liquid flow velocity with minor distortion of the PDMS channel or
deformation of the beads, contrary to what is observed when a positive pressure was used, where there
is also a higher likelihood of trapping more air bubbles. However, due to the extended length of the
separation channel, a high pressure difference is required to enable the flowing of beads solution within
the channel. Therefore, it was not possible to overcome this limitation and packing by this method was
disregarded.
Regarding the “packing by pushing” strategy, although it has some of the previously mentioned
drawbacks, it can easily overcome the limitation imposed by the “packing by pulling” strategy, and
several microfluidic structures were packed within 40 minutes using sieved CM-Sepharose beads further
used for the experiments in section 3.4. Packing separation channels longer than 12 cm with this method
proved to be impossible with any type of beads due to the high pressure requirements that the
microfluidic structure could not withstand, provoking leaking and rupture, mainly at the cross-channel
injector area. This phenomenon is easily deduced from equation 6:
Figure 21. Schematics comprising a timeframe of the beads trapping feature with packed silica beads (left); a
comparison between the different packing methods used throughout the experimental work (centre); and a set of
microscopic images of a fully packed 12 cm length separation channel with C18 bonded silica beads (right, A-D).
42
∆𝑃 ~ 𝐿2𝜇
𝑏2𝑡𝑅
where the increase of the separation channel length is directly correlated to the increase of the
pressure requirements for the structure.
However, this strategy revealed to not be satisfying when using C18 bonded silica beads, which
due to their hydrophobic nature offered a lot more resistance into being packed within the separation
channel, even when being aided by the correct packing solution. The packing of both C18 bonded silica
and silica beads revealed to be impossible when also using this method, since although silica beads are
more hydrophilic, they are also smaller beads (58 µm average particle size and with a particle size
distribution of 40-75 µm) which will also reinforce the severely the increase in the pressure requirements
as seen in Equation 6, where b represents the particle size. To overcome this difficulty, a novel packing
method was employed: in order to pack the beads within the channel without the structure’s rupture or
leaking, the packing pressure had to be controlled and maintained within levels the microfluidic structure
could handle (between 0.15 and 0.2 MPa). This was achieved by pushing the beads suspension into
the channel by applying a constant pressure of 0.15 MPa using a compressed air line. Packing was
achieved after 30 minutes with two consecutive insertions of beads slurry since maintaining the packing
pressure leads to an eventual decrease in the flow velocity until packing stops and needs to be
reinitiated. Between insertions a PEG 20% solution is flowed to tightly pack the beads inside the channel.
Additionally, packing the separation channel two hours after performing a plasma cleaner treatment
demonstrated to facilitate the insertion particularly of the C18 beads, since the PDMS walls are still less
hydrophobic then they would be without any plasma cleaner treatment done to the surface. This packing
was adopted for further experiments with silica and C18 bonded silica beads since it was the fastest and
most efficient one, as seen in Figure 21 (A-D) where we can observe a fully packed 12 cm in length
separation channel.
In Figure 21 (left) we can observe a timeframe of the initial 12 seconds of silica beads packing
into the separation channel, highlighting the beads trapping feature already mentioned previously
(Section 3.1). This design provides trapping efficiencies of nearly 100% (Figure 21, timeframe and D).
However, beads are often heterogeneous in size which results, on rare occasions, in the smaller beads
being able to pass through the gap at the interface of the two channels as seen in Figure 21 (left,
timeframe) and therefore the extent to which this occurs does not impair the performance of the packing
step, since most of the time beads are entirely trapped in the separation channel (Figure 21, D). This
design allows the reduction of air bubbles formed inside the channels and provides mechanical
robustness, therefore allowing the device to withstand higher pressures than, for example, cage-like
PDMS device50, renewable and pneumatically controlled pillar array51 or a double weir design52, which
also increase the complexity of fabrication. Finally, the separation channels after packing could be used
even several weeks after, if properly stored under a DI water bath at around 4 ºC, to prevent the bed
from drying and formation of air bubbles.
43
3.4. Food colouring separation (proof-of-concept)
In order to test the chromatographic separation channel in a simple and straightforward way, a
colour-based separation was performed using the microfluidic structure from iteration 2 (simple cross-
channel injector without valves). A mixture of methylene blue, which holds a positive charge, and yellow
food colouring (which contains a azorubine and acetate, negatively charged) was successfully
separated (Figure 22, i-iv) with a packed column of sieved CM-Sepharose beads. CM-Sepharose is a
weak cation exchanger meaning that it is itself negatively charged.
Filtered CM-Sepharose bead suspension was prepared and packed inside the separation
channel. Then, Milli-Q water was flowed during about 10 minutes at a flow rate of 15 µL/min in order to
wash the separation channel from the components of the beads packing solution (Figure 22-A). A green
solution resulting from the mixture of methylene blue and yellow food colouring in a proportion of 1:1
was inserted on the cross-shaped injector inlet by flowing 3 μL/min perpendicularly to the separation
channel (Figure 22-B). Afterwards, Milli-Q water was used as mobile phase, with a flow rate of 8 μL/min.
(Figure 22-C). In Figure 22-(i), the green mixture had just been loaded and elution with Milli-Q water
Figure 22. Separation of a green coloured mixture. (A-C) corresponds to an illustration of the sequential steps
previously done to insertion of the mobile phase, including beads packing (A), green mixture solution loading (B)
and elution with Milli-Q water as the mobile phase (C). The microscopic images (i-iv) show the progression of the
separation throughout 10 minutes. (i) Initial sample plug starts to be eluted with Milli-Q water as mobile phase. (ii)
Methylene blue starts to be retained in the column. (iii) Yellow plug travels through the column. (iv) Eluted
components leave the column in a yellow plug with traces of green.
44
took place on the following steps (ii, iii, iv). As expected, the methylene blue was being retained by the
stationary phase. However, yellow food colouring (which contains quinoline yellow, azorubine and
acetate) is not being eluted alone and some of the methylene blue travels with the plug (Figure 22-iii),
partially acquiring a green colour. This may be explained by the fact that interaction between the
positively charged methylene blue and negatively charged azorubine can be partially happening. Still, a
great portion of the methylene blue was retained, proving the functionality of the chip. In Figure 22-iv,
the eluted components are leaving the column after about 10 minutes from the initial moment of elution.
It is important to note that as seen in Figure 22-i, the plug formed with the cross-channel injector
was indeed too wide for further separations on a more HPLC-like approach, which is the goal of this
work. This justifies the integration of valves within the optimized injector, that lead to iteration 4, later
used for other experiments in section 3.6.
Two main strategies were approached to proceed to the regeneration step (Figure 23) of the
separation channel: the first one was to try to equilibrate the charges of the methylene blue with buffers
of phosphate and different NaCl concentrations. At low ionic strength, competition between the buffer
ions and methylene blue for charged groups on the ion exchanger is minimal and methylene blue binds
strongly. Increasing the ionic strength increases the competition and so reduces the interaction between
the carboxymethyl group and methylene blue, causing it to elute. However, it revealed ineffective
washing the separation channel with phosphate buffer 1 M or 2 M NaCl solution in order to elute the
methylene blue since a continuous washing during 20 minutes at a flow rate of 13 µL/min did not wash
out the methylene blue from the channel.
Figure 23. Schematics illustrating the two employed strategies for separation channel regeneration: (1) increasing
the ionic strength of the mobile phase or (2) protonation of the carboxymethyl group of the stationary phase.
45
Anion and cation exchangers are classified as strong or weak, depending on how much the
ionization state of the functional groups varies with pH. A strong ion exchanger has the same charge
density on its surface over a broad pH range, whereas the charge density of a weak ion exchanger
changes with pH. Therefore, another approach was to protonate the carboxymethyl group present on
the beads by changing the pH, since it is classified as a weak cation exchanger. The carboxymethyl
group carries negative charge only above pH 4.5 (pKa: 4.0). Therefore, by changing the pH of the mobile
phase to below the pKa of the functional group with a HCl 1M solution flowed during 20 minutes at a
flow rate of 13 µL/min, the separation channel regeneration was possible. In Figure 23, the caption of
the microfluidic structure confirms this by showing the clean separation channel and a blue droplet at
the outlet, marked with an arrow.
3.5. Toxins separation for food safety applications
The developed microfluidic device was tested for food safety applications, namely the
separation of OTA and AFB1 toxins. In order to study the feasibility of separating these two toxins using
the microfluidic structure developed throughout this work, many factors had to be investigated, from the
toxins concentration to be used, the elution profiles of the toxins with different mobile phases, to the
actual elution using the 12 cm length separation channel and cross-channel injector.
3.5.1. Toxins concentration optimization
For the purpose of this experiment, the goal was to assess the optimum concentration of toxins
to use in further experiments, there was no need to use the long separation channels used so far for
this kind of assessment. Microchannels previously developed on the group56 were used for a rather fast
analysis of the parameters under study and packing, loading and elution were done as reported on the
literature. In Figure 24, the plots describing the variance of the fluorescence signal from the toxins OTA
and AFB1, and corresponding microscopic acquisitions are shown. OTA and AFB1 at different
concentrations were flowed by pulling into the microchannels previously packed with C18 bonded silica
beads and interacted differently with the beads based on the polarity of each toxin. It can be observed
that for OTA, a significative fluorescence signal can only be seen from a concentration higher than 3
µg/mL and that at 10 µg/mL the fluorescence signal is reasonably good, not being saturated. However,
for AFB1 the fluorescence signal is already relatively higher at 3 µg/mL when compared to the value of
OTA at the same concentration. At 10 µg/mL, the signal seems to be already saturated for AFB1.
Therefore, a compromising concentration of 6 µg/mL of both OTA and AFB1 was chosen to proceed to
the next experiments done with this microchannels.
46
3.5.2. OTA and AFB1 elution studies in microchannels
Acetonitrile and methanol are widely used in HPLC separations, mainly when using the RPC
mode, which also uses C18 bonded silica beads as a stationary phase. Acetonitrile is a polar aprotic
organic solvent, used as a medium-polarity solvent that is miscible with water and a range of organic
solvents. In reverse-phase separations, acetonitrile is generally considered the strongest eluent, above
methanol. However, for each separation, conditions have to be studied since the elution strength of a
compound is only a guideline to what will actually happen during the separation. In order to understand
which conditions of the mobile phase would be good to separate the two toxins in a separation channel
with the C18 bonded silica beads, again the microchannels were used to perform the experiments.
Similarly, to the previous section, solutions of OTA and AFB1 at 6 µg/mL were loaded and adsorbed to
the beads. Acetonitrile and methanol solutions of 75, 50, and 25% (v/v) in Milli-Q water were tested with
respect to the ability of eluting the adsorbed toxins. Acetonitrile and methanol solutions were flowed
Figure 24. Plots comprising the fluorescence signal variance in function of the concentration of OTA (top) and AFB1
(bottom) toxins. The correspondent microscopic images for the different concentrations for each toxin are displayed
on the left of the graphics. Fluorescence images were acquired at the beginning of the microchannel with an
exposure time of 1 second and a gain of 1x.
47
inside the microchannel by pushing and images were acquired every 30 seconds to analyse the
fluorescence signal. The plots obtained for the elution profiles can be seen in Figure 25 for both OTA
and AFB1.
Analysing the OTA elution profiles, it is clear that acetonitrile was stronger than methanol, eluting
almost entirely the OTA adsorbed on the beads, with the fluorescence signal reaching the same or lower
values when compared to the initial fluorescence at t=0 on any solvent ratio. The same analysis can be
applied for AFB1 elution profiles, where acetonitrile revealed to be a stronger eluent than methanol with
exception for the solvent ratio 25:75, where acetonitrile and methanol had similar elution strength.
Overall, the difference between the elution strength of acetonitrile and methanol was more accentuated
on OTA than in AFB1. Curiously, when observing the OTA elution plots for both acetonitrile and
methanol it is noticeable that the fluorescence signal seems to increase relatively to the initial loading
fluorescence during the first 1-2 minutes. This may be explained by a “wave-like” behaviour when
starting elution, since OTA molecules adsorbed on the beads that clogged on the inlet could travel to
the beginning of the microchannel and increase the fluorescence signal momently as they pass. To
overcome this from happening, caution should be taken to only pack beads on the microchannel and
avoid bead packing at the inlet. The presented fluorescence values on the plots were normalized
Figure 25. Elution profiles for toxins OTA and AFB1 previously adsorbed in a C18 bonded silica packed microchannel.
Elutions were performed during 10 minutes and fluorescence images acquired every 30 seconds at the beginning of
the microchannel with an exposure time of 1 second and a gain of 1x.
48
considering the background noise from the beads fluorescence. Also, the solvent in which toxins are
solubilized can alter the fluorescence signal of the toxins, increasing or decreasing it. This effect was
also considered and the presented plots are normalized taking that into account, although there was not
a significant change on the fluorescence signal that would alter the analysis of the plots (see section B,
Appendix 2)
Finally, a significant difference between the elution profile for OTA and AFB1 was observed for
a solvent ratio of CH3CN:water (25:75 v/v), where AFB1 was eluted more gradually than OTA, which
tended to stay on the mobile phase and quickly leave the microchannel. Therefore, this condition
promised some separation between the two toxins and was chosen to separate the toxins on the
experiments of the next section.
3.5.3. OTA and AFB1 cross-channel injection and elution
In order to study the feasibility of AFB1 elution within a packed column with a cross-channel
injector (iteration 2) and also the separation of AFB1 and OTA within the same column, the fluorescence
Figure 26. Cross-channel injection and elution of the toxins AFB1 and OTA in reverse-phase separation channel.
(A) illustrates the mode of operation for loading and elution of the toxins using iteration 2 as a microfluidic structure.
(B) are fluorescence acquisitions in the beginning of the separation channel comprising the cross-channel injector,
of the AFB1 6 µg/mL elution with 100% CH3CN (v/v) during the first 2 minutes. (C) are fluorescence acquisitions in
the beginning of the separation channel, of the elution with CH3CN:water (25:75 v/v) for the separation of OTA and
AFB1, at a concentration of 6 µg/mL Fluorescence images were acquired at the beginning of the microchannel with
an exposure time of 1 second and a gain of 1x.
49
of the toxins was followed in the reverse-phase packed channel (Figure 26). Before moving to directly
experimenting the toxins separation onto a microfluidic structure with a separation channel and cross-
channel injector with incorporated valves (iteration 4), it was important to assess the behaviour of the
toxin’s elution profiles within a long separation channel like the one developed throughout this work
before adding to the experiments the complex mechanism of the cross-channel injector with
incorporated valves. Therefore, the design from iteration 2 was used for this experiment, similarly to the
proof-of-concept coloured separation done previously. In Figure 26-A, the mode of operation for the
loading and elution of the toxins is shown. A volume of 60 µL of 6 µg/mL of AFB1 solution (for the elution)
or 60 µL of 6 µg/mL of AFB1 + OTA solution (for the separation) was loaded by pulling in the channel
perpendicular to the separation channel as seen in Figure 26-A. Acetonitrile 100% v/v (for the elution)
and 25% v/v (for the separation) in Milli-Q water solutions were flowed inside the microchannel by
pushing at the inlet for approximately 10 minutes.
In Figure 26-B we can observe the elution of AFB1 within the first 2 minutes with 100%
acetonitrile as mobile phase. The goal of this first experiment was to assess the behaviour of the
fluorescent plug while travelling through the separation channel, and therefore the strongest possible
eluent was used in pure form to ensure elution would happen. B-(i) refers to the moment where flowing
of the mobile phase had just started while in B-(ii, iii) the movement of the plug can already be observed.
In B-(iv), around two minutes after elution started the plug starts to spread and the fluorescence signal
to gradually faint until it was almost not distinguishable with the inherent fluorescent of the beads. This
limitation made following the fluorescent plug along the separation channel a really difficult task,
although a large volume of AFB1 solution (60 µL) was loaded in order to increase the initial fluorescent
of the plug and tackle this limitation. The plug continued to spread through the separation channel and
reached the outlet after around 10 minutes (results not shown), although the fluorescence signal was
so faint that the difference in the fluorescence signal cannot be regarded as 100 % sure belonging to
the plug. Nevertheless, the separation between OTA and AFB1 was also tried on this microfluidic
structure as seen in Figure 26-C. The expected outcome from this experiment was to at least observe
the initial plug starting to divide into two different fluorescent fronts somewhere at the beginning of the
separation channel while flowing CH3CN:water (25:75 v/v) as a mobile phase, since from the results of
the previous experiment it would be unlikely to be able to monitor the elution of the toxins at the end of
the separation channel due to the faint fluorescence signal. The mobile phase was chosen considering
the results from the experiments of section 3.5.2, where AFB1 would be more retained in the stationary
phase than OTA, at these chromatographic conditions. However, as seen in Figure 26-C, the initial plug
containing the two toxins gradually loses its fluorescent signal, not being clear if there were two fronts
of fluorescence being eluted although more than two minutes had passed between the first and last
image.
These results indicate three main problems that need to be overcome: first, the chromatographic
conditions need to be further optimized in order for the toxins to be effectively separated, especially the
mobile phase composition and flow rate used. An interesting approach would be adding acetic acid to
50
the composition of the mobile phase as demonstrated in the literature60. Secondly, the experimental
design to perform these experiments needs to be improved, since many limitations were experienced in
following the plug travelling through the separation channel and assess the actual difference between
the fluorescent signal of the toxins and the beads. Lastly, and most importantly, there is a clear problem
with both the design and packing of the separation channel causing band broadening as seen from the
dispersion the plug suffers while travelling through the packed channel. Recalling the van Deemter
equation (Equation 3):
𝐻 = 𝐶𝑀𝑢 (𝑜𝑟 𝐴) + 𝐵
𝑢+ 𝐶𝑆𝑢
One can speculate on which terms are probably affecting more the band broadening of the
separation channel given some known characteristics of the separation channel. For example, the term
A (eddy diffusion) refers to the mobile phase mass-transfer. Zone broadening in the mobile phase is
caused by the multiple paths a molecule can take through a packed column, especially when the lengths
of these paths can differ significantly. At low mobile phase velocities usually used in microfluidics, the
molecules are not significantly dispersed due to the contribution of regular diffusion effects that
attenuates the eddy diffusion. However, these multiple paths arise due to inhomogeneities in column
packing and small variations in the particle size of the packing material which is something to improve
in our separation channel, as discussed in section 3.1. For future work, is advised to use smaller particles
and with narrower size distribution that will lead to a better bed uniformity, to reduce the effect of this
parameter on band broadening. The term (𝐵𝑢⁄ ) represents longitudinal diffusion. This factor causes
band width to rise with time, and it happens whether or not the mobile phase is flowing. The time spent
by the band during its way through the column differs inversely with the flow rate, so the influence of
longitudinal diffusion on band width increases for low flow rates. Although in liquid chromatography,
diffusion rates are much smaller in value, so this term has an almost insignificant influence on band
broadening, this correlation may not be so straightforward in microfluidics as discussed in further section
3.6. Finally, the term (𝐶𝑆𝑢) represents the stationary phase mass-transfer: some analyte molecules will
enter further into a particle pore than others, spending in it a given and varied time. Molecules that spend
less time in the particle will move further along the column, with a resulting increase in band width. By
decreasing the particle size because of term A, the impact of this term will also be reduced, although
this term is probably the one least contributing to band broadening since dead volume is also a
contribution included in this effect and our microfluidic device was designed to be low dead volume
device.
Optimizing the chromatographic conditions and have a better experimental design are problems
that can be later addressed, independently of the design and packing of the separation channel.
However, studying the plug progression in order to assess the most important improvements to be done
in the future is imperative before tackling more advanced problems. Having this in mind, further
51
experiments were done in the next sections referring to the plug progression through the separation
channel.
3.6. Plug progression assessment and optimization
In order to study plug formation and progression inside a microfluidic structure with a cross-
shaped injector actuated by pneumatic valves (iteration 4B), plug optimization assays were performed
using FITC, a green fluorescent dye.
Figure 27 illustrates the necessary sequential steps in order to inject the plug and start elution.
In step 1 (washing), an ethanol 96% (v/v) solution was flowed by pushing using syringe pumps at a flow
rate of 15 µL/min for 20 minutes, from both the injector and separation channel inlets as indicated by
the black and white arrows, respectively. During this step, the valves from the loading channel of the
injector should be closed at both the inlet and outlet, as indicated by the black lines. This step is essential
to get rid on any air bubbles that may be entrapped between the injector and the separation channel
and, of course, remove any unwanted components from the channels. Next, the loading of 5 µL the
FITC 1mg/mL solution proceeded as seen in step 2 (loading) by pulling at 15 µL/min and subsequent
actuation of the loading channel valves while the other two were already closed (black lines). The plug
had a defined size, corresponding to the length between the two valves that give access to the
Figure 27. Schematics showing the operation of the cross-channel injector with integrated valves for the injection
and elution of the FITC plug by illustrating the sequential steps: washing (step 1), loading (step 2), injection (step
3) and elution (step 4). Arrows indicate the direction of fluid flow and dark lines the valves that are closed.
52
separation channel. Then, for the injection of the defined plug into the separation channel (step 3), the
valves that were previously closed were opened. Flow of the mobile phase should not start before
opening these valves that give access to the separation channel, since it was observed that pressure is
accumulated and when the valves are open, the accumulated pressure injects the plug too fast. Finally,
step 4 took place: keeping only the valves from the loading channel closed, the mobile phase (ethanol
96%) was flowed by pushing at a rate of 0.5 µL/min from both the injector and separation channel inlets,
illustrated by the white and black arrows, respectively. To note that in order for the elution to start, flow
had to be driven both from the injector and the separation channel inlets because flowing only from the
inlet, at the given flow rate, was not enough to win the accumulated fluid resistance in that zone of
intersection between the connection and separation channel. Therefore, one of the improvements in
future work is that the injector should not be placed perpendicular to separation channel in order to
ensure appropriate flow.
In Figure 28, a set of microscopic images are shown, that correspond to the sequential steps
described in Figure 27. From image (i) to (iii), the formation and injection of FITC plug can be observed.
Then, elution started (iv) in the separation channel without beads with ethanol 96% as the mobile phase
at a flow rate of 0.5 µL/min: the formation of a parabolic flow profile can immediately be seen. This
phenomenon is typical of pressure-driven separations and is regarded as the Taylor dispersion, where
a shear flow can increase the effective diffusivity of a given molecule. Shear acts to disperse the
concentration distribution in the direction of the flow, increasing the rate at which it spreads. The overall
effect is an enhancement of axial dispersion over molecular diffusion alone. From (v) to (vi) the evolution
of the plug observed corresponded to the expected evolution of plug under the effect of Taylor
dispersion. In (v) the spreading behaviour shortly after its injection time can be observed where the
cross-channel diffusion is yet to happen and the distance between the fastest moving molecules at the
centre and the slow particles on the wall stretches linearly with time. After some time (vi), enough to let
cross-channel diffusion to act, the edges of the paraboloid change to form a blob which is nearly uniform
Figure 28. Sequential fluorescent image acquisitions of a FITC plug injection (i-iii) and progression (iv-viii)
throughout the separation channel without beads packed. Images were acquired with an exposure time of 800
milliseconds and 3x gain.
53
across the channel and the plug spreads. In (vii) it can be seen the extension of the plug length after
traveling more than half of the separation channel, from the gradually decreased fluorescence signal in
each turn, which indicates that the plug can be several centimetres in length. Finally, in (viii) we can see
the front of the plug arriving at the outlet of the separation channel, after approximately 11 minutes had
passed since the beginning of the elution. From this experiment we could confirm the effect of Taylor
dispersion in our microfluidic device without beads packed within the separation channel. Additionally,
from image (v) to (vi) was a critical moment for the plug to disperse even more. This can be a
confirmation of the hypothesis reported in the literature, that factors such turns35,36 and corners37 can
influence plug spreading.
The question arises if this behaviour can also be observed when the separation channel is
packed with beads. Therefore, the same experiment was performed within a separation channel packed
with silica beads (Figure 29).
In Figure 29, a set of microscopic images are shown, that correspond to the sequential steps
described in Figure 27. From image (i) to (iii), the formation and injection of FITC plug can be observed.
Then, elution started (iv) in the separation channel with silica beads with ethanol 96% as the mobile
phase at a flow rate of 0.5 µL/min: the formation of a parabolic flow profile even in a packed channel
can be seen, although not as accentuated as in Figure 28-(iv). From (iv) to (v) the evolution of the plug
observed reached the “blob” state earlier in the separation channel, probably because of the beads
present within the channel. In (vi) and (vii), one can observe that similarly to the results without beads,
the spreading of the plug seems to be on the order of the several centimetres by analysing the different
intensities of fluorescence in each turn of the channel. Finally, in (viii) the front of the FITC plug reaches
the outlet around 13 minutes after elution started. This delay when compared to the same elution without
beads may be attributed to the beads itself, although FITC will interact shortly since it prefers more
Figure 29. Sequential fluorescent image acquisitions of a FITC plug injection (i-iii) and progression (iv-viii) throughout
the separation channel with beads packed. Images were acquired with an exposure time of 800 milliseconds and 3x
gain.
54
hydrophobic phases than silica. Although in liquid chromatography, diffusion rates are much smaller in
value, so this term has an almost insignificant influence on band broadening, this correlation may not
be so straightforward as demonstrated in this experiment. If a Taylor dispersion effect occurs in the
presence of beads, even if only in the beginning of the elution, it can hinder the separation efficiency of
the separation channel. Therefore, each microfluidic separation channel has to be regarded individually
and not much generalizations can be done in assessing the role of this effect on separation efficiency
and on the plug behaviour.
55
4. Conclusion and Future Prospects
Analytical separations can occur at any scale: from a laboratory scale to the industrial scale.
However, there is an increased demand for smaller and smaller instrumentation for analytical
separations5, requiring high-throughput answers to its problems but more importantly, fast and on-spot
analyses of products of interest. This implies small sample and set-up size, low costs and easy disposal
of wastes. In this experimental work, the main goal was to develop a miniaturized chromatographic
analytical device for food safety applications, which not only meets the goal of fully integration like
attempted previously12, but also aims for a simpler fabrication process (and therefore more economically
viable), user-friendly operation and versatility for a broad range of analytes/applications.
The main outcome of this work was the development of a microfluidic device which comprises a
cross-channel injector with integrated pneumatic actuated valves that offers almost zero dead-volume
and a 12 cm-long separation channel enabling the packing of most commercial available
chromatographic beads. The separation channel was tested for a proof-of-concept separation that
allowed the separation of a mixture of methylene blue and yellow food colouring proving the great
potential for future purification and analytical applications. Moreover, the applicability of this device for
food safety applications was assessed using a reverse-phase chromatographic mode for the separation
channel, indicating that under the optimal chromatographic conditions and further optimization of the
device design, separation and detection of toxins like AFB1 and OTA is possible.
Throughout the experimental work, five different microfluidic structures were developed from which
two represented the achievement of two of the main milestones of this project: first, iteration 2 for
establishing the 12 cm-long separation channel with an efficient beads trapping feature and lastly,
iteration 4B for providing an optimized cross-channel injector with integrated valves coupled to the
separation channel in a microfluidic structure that provided robustness for operation under relatively
high pressures for a soft-polymer based microfluidic device. Furthermore, optimizations regarding the
fabrication process were done that lead to optimized pneumatic actuation of the valves and the
application of a robust oxygen plasma cleaner interlayer bonding for multilayer soft lithography
fabricated devices. Nevertheless, although separation channel packing was achieved with different
commercially available chromatographic beads using a controlled-pressure packing method if smaller
beads or longer length for the separation channel might be needed, the microfluidic structure needs new
fabrication methodologies providing robustness to withstand the new pressure requirements. The beads
trapping feature adapted from previous work from the group56 revealed great performance in trapping
the beads of nearly 100%, even under a pressure accumulation from a 12-cm long packed channel
situation. However, again if smaller beads are aimed to be used this feature might need some adaption
to efficiently trap these small particles too. Regarding separations within the separation channel, a
mixture of methylene blue, which holds a positive charge, and yellow food colouring (which contains a
azorubine and acetate, negatively charged) was successfully separated with a packed column of sieved
56
CM-Sepharose beads (weak cation exchanger) with regeneration of the column possible by flowing HCl
1 M solution inside the channel. This proved the potential of the separation channel to provide both
preparative and analytical separation in several different applications, although some need for
improvements in the plug broadening right after insertion was observed because a valve-less cross-
channel injector was used. Nevertheless, the same separation channel was also tested for food safety
applications using a reverse-phase chromatographic mode for the separation channel, more precisely,
by eluting and separating OTA and AFB1. For this, separated elution studies for each toxin in
microchannels were performed to choose a mobile phase able to provide separation of the toxins. The
results from these experiments indicated that a deeper study of the chromatographic conditions needs
to be done and also further optimization of the device design since plug broadening was observed to
hinder the efficiency of separation. Regarding the plug broadening limitation observed, it was
hypothesized that the main players on this effect may be the mobile phase mass-transfer and
longitudinal diffusion phenomena present in the van Deemter equation. However, from studying the
FITC plug injection and progression in a separation channel with and without silica packing, it was
hypothesized that these limitations can be easily attained by providing a well-packed, smaller particle
packing and a more careful design as for example, a straight separation channel without turns or curves
in to dimming the effect of Taylor-dispersion.
Therefore, for future work, further optimization of the microfluidic device will be able to provide
better separations for reversed-phase chromatographic applications. More precisely, integration of
smaller particles (in the order of 5 µm), development of straight separation channel and consequently,
a fabrication process that delivers a device that can attain higher operation pressures while maintaining
the fabrication process simple and economically viable. For full integration, it will also be interesting to
integrate at the end of the separation channel a sensor in order to achieve the portability and fast
analysis aimed in LoC devices.
57
58
References
1. Huei-Li, C., Chen, Y.-S. & Ruey-Shin, J. Separation of surfactin from fermentation broths by acid precipitation and two-stage dead-end ultrafiltration processes. J. Memb. Sci. 299, 114–121 (2007).
2. Lee, J.-Y., Yoo, C., Jun, S.-Y., Ahn, C.-Y. & Oh, H.-M. Comparison of several methods for effective lipid extraction from microalgae. Bioresour. Technol. 101, S75–S77 (2010).
3. Lisec, J., Schauer, N., Kopka, J., Willmitzer, L. & Fernie, A. R. Gas chromatography mass spectrometry–based metabolite profiling in plants. Nat. Protoc. 1, 387–396 (2006).
4. Bonfatti, V., Grigoletto, L., Cecchinato, A., Gallo, L. & Carnier, P. Validation of a new reversed-phase high-performance liquid chromatography method for separation and quantification of bovine milk protein genetic variants. J. Chromatogr. A 1195, 101–106 (2008).
5. Lavrik, N. V., Taylor, L. T. & Sepaniak, M. J. Nanotechnology and chip level systems for pressure driven liquid chromatography and emerging analytical separation techniques: A review. Analytica Chimica Acta 694, 6–20 (2011).
6. Terry, S. C., Herman, J. H. & Angell, J. B. A gas chromatographic air analyzer fabricated on a silicon wafer. IEEE Trans. Electron Devices 26, 1880–1886 (1979).
7. Harrison, D. J., Manz, A., Lüdi, H., Widmer, H. M. & Fan, Z. Capillary Electrophoresis and Sample Injection Systems Integrated on a Planar Glass Chip. Anal. Chem. 64, 1926–1932 (1992).
8. Manz, A., Graber, N. & Widmer, H. M. Miniaturized total chemical analysis systems: A novel concept for chemical sensing. Sensors Actuators B Chem. 1, 244–248 (1990).
9. Whitesides, G. The lab finally comes to the chip! Lab Chip 14, 3125 (2014).
10. Ishibashi, R., Mawatari, K. & Kitamori, T. Highly efficient and ultra-small volume separation by pressure-driven liquid chromatography in extended nanochannels. Small 8, 1237–1242 (2012).
11. Chan, A. S., Danquah, M. K., Agyei, D., Hartley, P. G. & Zhu, Y. A simple microfluidic chip design for fundamental bioseparation. J. Anal. Methods Chem. 2014, (2014).
12. Yin, H. et al. Microfluidic chip for peptide analysis with an integrated HPLC column, sample enrichment column, and nanoelectrospray tip. Anal. Chem. 77, 527–533 (2005).
13. Wild, C. P. & Gong, Y. Y. Mycotoxins and human disease: A largely ignored global health issue. Carcinogenesis 31, 71–82 (2009).
14. IARC/WHO. Some naturally occurring substances: food items and constituents, heterocyclic aromatic amines and mycotoxins. IARC Monogr. Eval. Carcinog. risks to humans 56, 362 (1993).
15. Jungbauer, A. & Hahn, R. Ion-Exchange Chromatography. Methods Enzymol. 463, 349–371 (2009).
16. Skoog, Douglas A.; West, Donald M.; Hollar, J. F. Fundamentals of Analytical Chemistry, 9th Edition. Brooks/Cole. (1992). Belmont.
17. Nelson, D. L. & Cox, M. M. Lehninger Principles of Biochemistry, 4th Edition. Freeman and Company. (2008). New York.
18. Snyder, L. R., Kirkland, J. J. & Dolan, J. W. Introduction to Modern Liquid Chromatography, 3rd Edition. Wiley. (2010). USA.
59
19. Pressltd, P. & Deemter, V. a N. Diffusion and Resistance To Mass Transfer Nonideality in Chromatography. Chem. Eng. Sci. 6, 271–289 (1956).
20. D’Orazio, G., Asensio-Ramos, M., Fanali, C., Hernández-Borges, J. & Fanali, S. Capillary electrochromatography in food analysis. TrAC - Trends in Analytical Chemistry 82, 250–267 (2016).
21. Bidlingmeyer, B. A., Hooker, R. P., Lochmuller, C. H. & Rogers, L. B. Improved Chromatographic Resolution from Pressure-Induced Changes in Liquid—Solid Distribution Ratios. Sep. Sci. 4, 439–446 (1969).
22. Bakalyar, S. R., Bradley, M. P. T. & Honganen, R. The role of dissolved gases in high-performance liquid chromatography. J. Chromatogr. 158, 277–293 (1978).
23. Gritti, F., Sanchez, C. A., Farkas, T. & Guiochon, G. Achieving the full performance of highly efficient columns by optimizing conventional benchmark high-performance liquid chromatography instruments. J. Chromatogr. A 1217, 3000–3012 (2010).
24. Dall’Asta, C., Galaverna, G., Dossena, A. & Marchelli, R. Reversed-phase liquid chromatographic method for the determination of ochratoxin A in wine. J. Chromatogr. A 1024, 275–279 (2004).
25. Panfili, G., Fratianni, A. & Irano, M. Normal phase high-performance liquid chromatography method for the determination of tocopherols and tocotrienols in cereals. J. Agric. Food Chem. 51, 3940–3944 (2003).
26. Lísa, M. et al. Characterization of fatty acid and triacylglycerol composition in animal fats using silver-ion and non-aqueous reversed-phase high-performance liquid chromatography/mass spectrometry and gas chromatography/flame ionization detection. J. Chromatogr. A 1218, 7499–7510 (2011).
27. Bajad, S. U. et al. Separation and quantitation of water soluble cellular metabolites by hydrophilic interaction chromatography-tandem mass spectrometry. J. Chromatogr. A 1125, 76–88 (2006).
28. Rea, J. C., Moreno, G. T., Lou, Y. & Farnan, D. Validation of a pH gradient-based ion-exchange chromatography method for high-resolution monoclonal antibody charge variant separations. J. Pharm. Biomed. Anal. 54, 317–323 (2011).
29. Ibáñez, M., Sancho, J. V. & Hernández, F. Determination of melamine in milk-based products and other food and beverage products by ion-pair liquid chromatography-tandem mass spectrometry. Anal. Chim. Acta 649, 91–97 (2009).
30. Gellein, K. et al. Separation of proteins including metallothionein in cerebrospinal fluid by size exclusion HPLC and determination of trace elements by HR-ICP-MS. Brain Res. 1174, 136–142 (2007).
31. Peng, L., Jayapalan, S., Chankvetadze, B. & Farkas, T. Reversed-phase chiral HPLC and LC/MS analysis with tris(chloromethylphenylcarbamate) derivatives of cellulose and amylose as chiral stationary phases. J. Chromatogr. A 1217, 6942–6955 (2010).
32. Tabeling, P. Introduction to Microfluidics. Oxford University Press. (2005). New York.
33. Datta, S. & Ghosal, S. Characterizing dispersion in microfluidic channels. Lab Chip 9, 2537–2550 (2009).
34. Kirby, B. Micro- and nanoscale fluid mechanics : transport in microfluidic devices. Cambridge University Press. (2010). New York.
60
35. Molho, J. I. et al. Optimization of turn geometries for microchip electrophoresis. Anal. Chem. 73, 1350–1360 (2001).
36. Paegel, B. M., Hutt, L. D., Simpson, P. C. & Mathies, R. A. Turn geometry for minimizing band broadening in microfabricated capillary electrophoresis channels. Anal. Chem. 72, 3030–3037 (2000).
37. von Heeren, F., Verpoorte, E., Manz, a & Thormann, W. Micellar electrokinetic chromatography separations and analyses of biological samples on a cyclic planar microstructure. Anal. Chem. 68, 2044–2053 (1996).
38. Kuncová-Kallio, J. & Kallio, P. J. PDMS and its suitability for analytical microfluidic devices. in Annual International Conference of the IEEE Engineering in Medicine and Biology - Proceedings 2486–2489 (2006).
39. Unger, M. A. Monolithic Microfabricated Valves and Pumps by Multilayer Soft Lithography. Science (80-. ). 288, 113–116 (2000).
40. Thorsen, T. Microfluidic Large-Scale Integration. Science (80-. ). 298, 580–584 (2002).
41. Cai, Z., Xiang, J., Zhang, B. & Wang, W. A magnetically actuated valve for centrifugal microfluidic applications. Sensors Actuators, B Chem. 206, 22–29 (2015).
42. Cooksey, G. A. et al. Pneumatic valves in folded 2D and 3D fluidic devices made from plastic films and tapes. Lab Chip 14, 1665 (2014).
43. Heiland, J. J. et al. On-chip integration of organic synthesis and HPLC/MS analysis for monitoring stereoselective transformations at the micro-scale. Lab Chip 17, 76–81 (2017).
44. Huft, J., Haynes, C. A. & Hansen, C. L. Fabrication of High-Quality Micro fl uidic Solid-Phase Chromatography Columns. Anal Chem 85, 1797–1802 (2013).
45. Xie, J., Miao, Y., Shih, J., Tai, Y. C. & Lee, T. D. Microfluidic platform for liquid chromatography-tandem mass spectrometry analyses of complex peptide mixtures. Anal. Chem. 77, 6947–6953 (2005).
46. Kutter, J. P. Liquid phase chromatography on microchips. Journal of Chromatography A 1221, 72–82 (2012).
47. Jung, S. et al. Impact of conduit geometry on the performance of typical particulate microchip packings. Anal. Chem. 81, 10193–10200 (2009).
48. McEnery, M. et al. Liquid chromatography on-chip: progression towards a μ-total analysis system. Analyst 125, 25–27 (2000).
49. Faure, K. Liquid chromatography on chip. Electrophoresis 31, 2499–2511 (2010).
50. Jeong, Y. et al. PDMS micro bead cage reactor for the detection of alpha feto protein (AFP). Sensors Actuators, B Chem. 128, 349–358 (2008).
51. Shao, G. et al. Design, fabrication and test of a pneumatically controlled, renewable, microfluidic bead trapping device for sequential injection analysis applications. Analyst 141, 206–215 (2016).
52. Oleschuk, R. D., Shultz-Lockyear, L. L., Ning, Y. & Harrison, D. J. Trapping of Bead-Based Reagents within Microfluidic Systems: On-Chip Solid-Phase Extraction and Electrochromatography. Anal. Chem. 72, 585–590 (2000).
53. Ishida, A., Yoshikawa, T., Natsume, M. & Kamidate, T. Reversed-phase liquid chromatography
61
on a microchip with sample injector and monolithic silica column. J. Chromatogr. A 1132, 90–98 (2006).
54. Zain, M. E. Impact of mycotoxins on humans and animals. J. Saudi Chem. Soc. 15, 129–144 (2011).
55. el Khoury, A. & Atoui, A. Ochratoxin A: General Overview and Actual Molecular Status. Toxins (Basel). 2, 461–493 (2010).
56. Pinto, I. F. et al. The application of microbeads to microfluidic systems for enhanced detection and purification of biomolecules. Methods 116, 112–124 (2017).
57. Pinto, I. F. et al. A regenerable microfluidic device with integrated valves and thin-film photodiodes for rapid optimization of chromatography conditions. Sensors Actuators B Chem. (2017).
58. Eddings, M. A., Johnson, M. A. & Gale, B. K. Determining the optimal PDMS–PDMS bonding technique for microfluidic devices. J. Micromechanics Microengineering 18, 67001 (2008).
59. Xiong, L., Chen, P. & Zhou, Q. Adhesion promotion between PDMS and glass by oxygen plasma pre-treatment. J. Adhes. Sci. Technol. 28, 1046–1054 (2014).
60. Tafuri, A., Meca, G. & Ritieni, A. A rapid high-performance liquid chromatography with fluorescence detection method developed to analyze ochratoxin A in wine. J. Food Prot. 71, 2133–2137 (2008).
62
Appendix Section
A. Optimized processes runsheet
Appendix 1. Runsheet for the final optimized process of the fabrication of iteration 4B, which includes the fluidic
layer and valves layer mold fabrication and also the whole optimized process for the PDMS structure fabrication.
STEP DESCRIPTION
SILICON SUBSTRATE
CLEANING
1. Silicon substrate 5x5 cm must be rinsed with acetone, followed
by IPA and finally DI water
2. Cleaned silicon substrate must be placed on a Petri dish filled
with Alconox solution, enough to fully cover the substrate
3. Put the Petri dish on bath at 65 ºC during 15 minutes
4. Clean the silicon substrate with DI water, followed by drying with
compressed air
5. Place the substrate on the UVO cleaner for 20 minutes
FLUIDIC LAYER MOLD
FABRICATION
6. SU-8 2015 photoresist spin-coating on top of the substrate
STEP 1: 10 seconds, 500 rpm, 100 rpm/s
STEP 2: 34 seconds, 1700 rpm, 300 rpm/s
7. Pre-exposure bake in hot plate at 95 ºC during 4 minutes with 1
minute for cooling down
8. Place the 20 µm design hard mask on top of the recently
deposited photoresist with the aluminium surface facing down
9. Exposure step to UV light during 30 seconds on the further way
slot
10. Post-exposure bake in a hot plate at 95 ºC during 5 minutes
with 2 minutes for cooling down
11. Development of the pattern by immersing the substrate in a
PGMEA solution during 2 minutes with manual agitation
12. Cleaning with IPA, followed by drying with compressed air
13. SU-8 50 photoresist spin-coating on top of the substrate
STEP 1: 10 seconds, 500 rpm, 100 rpm/s
STEP 2: 30 seconds, 1000 rpm, 300 rpm/s
14. Pre-exposure bake in a hot plate by sequential baking
STEP 1: 65 ºC during 10 minutes
63
STEP 2: ramping up to 95 ºC
STEP 3: 95 ºC during 30 minutes
STEP 4: cooling down for 1 minute
15. Place the 100 µm design hard mask on top of the recently
deposited photoresist with the aluminium surface facing down
16. Exposure step to UV light during 70 seconds on the further
way slot
17. Post-exposure bake in a hot plate by sequential baking
STEP 1: 65 ºC for 1 minute
STEP 2: ramping up to 95 ºC
STEP 3: 95 ºC for 10 minutes
STEP 4: cooling down for 2 minutes
18. Development of the pattern by immersing the substrate in a
PGMEA solution during 10 minutes with manual agitation
REPEAT STEP 12
20. AZ 40XT spin-coating on top of the substrate
STEP 1: 10 seconds, 500 rpm, 100 rpm/s
STEP 2: 21 seconds, 2000 rpm, 1000 rpm/s
21. Let the spin-coated layer settle for 30 minutes
22. Pre-exposure bake in a hot plate by sequential baking
STEP 1: Slowly ramping from 100 to 125 ºC
STEP 2: 125 ºC during 5 minutes
STEP 3: cooling down for 2 minutes
23. Place the 35 µm design hard mask on top of the recently
deposited photoresist with the aluminium surface facing down
24. Exposure step to UV light for 3 minutes and 30 seconds on the
further away slot
25. Post exposure in a hot plate at 105 ºC for 2 minutes with 2
minutes for cooling down
26. Development of the pattern by immersing the substrate in a AZ
400 K developer 1:3 DI water solution during 10 minutes with
manual agitation
27. Cleaning with DI water, followed by drying with compressed air
REPEAT STEP 22
VALVES LAYER MOLD
FABRICATION
28. SU-8 50 spin-coating on top of the substrate
STEP 1: 10 seconds, 500 rpm, 100 rpm/s
64
STEP 2: 37 seconds, 2300 rpm, 300 rpm/s
29. Pre-exposure bake in a hot plate by sequential baking
STEP 1: 65 ºC during 3 minutes
STEP 2: ramping up to 95 ºC
STEP 3: 95 ºC during 8 minutes
STEP 4: cooling down for 1 minute
30. Place the 50 µm design hard mask on top of the recently
deposited photoresist with the aluminium surface facing down
31. Exposure step to UV light for 25 seconds on the nearest slot
32. Post-exposure bake in a hot plate by sequential baking
STEP 1: 65 ºC during 1 minute
STEP 2: ramping up to 95 ºC
STEP 3: 95 ºC during 7 minutes
STEP 4: cooling down for 2 minutes
33. Development of the pattern by immersing the substrate in a
PGMEA solution during 6 minutes with manual agitation
REPEAT STEP 12
34. Hard bake step in a hot plate at 150 ºC during 15 minutes
35. Slowly cooling down until 50 ºC
PDMS PREPARATION
36. PDMS elastomer and curing agent mixed in a 10:1 ratio
37. PDMS elastomer and curing agent mixed in a 20:1 ratio
38. Degasification step for 30 minutes in a desicator
PDMS CASTING
39. Casting for the fluidic layer channels by pouring the PDMS
preparation of 10:1 on top of the fluidic layer mold inside a Petri
dish
40. Casting for the valves layer channels by PDMS 20:1
preparation spin-coating on top of the valves layer mold at 25
seconds, 400 rpm and 100 rpm/s
41. PDMS curing by baking both in the oven at 70 ºC during 90
minutes
PDMS LAYERS BONDING
42. PDMS fluidic layer is cut and peeled off from the fluidic layer
mold
43. Holes are punched using 18 Gauge needles for the inlets and
outlets
44. Oxygen plasma bonding by placing both the PDMS fluidic layer
and PDMS valves layer in a oxygen plasma cleaner during 60
65
seconds at MED intensity, and sealed against each other
immediately after the plasma treatment
PDMS STACK SEALING
45. Holes are punched using 18 Gauge needles for the inlets of
the valves channels
46. PDMS 10:1 preparation spin-coating on top of a 1 mm-thick
glass slide
STEP 1: 5 seconds, 500 rpm, 100 rpm/s
STEP 2: 20 seconds, 4200 rpm, 1000 rpm/s
47. Oxygen plasma bonding by placing both the PDMS stack and
the PDMS spin-coated glass in a oxygen plasma cleaner during 60
seconds at MED intensity, and sealed against each other
immediately after the plasma treatment
66
B. Solvent effect influence on fluorescence signal for OTA
and AFB1 elution in microchannels
Appendix 2. Elution profiles for toxins OTA and AFB1 and respective curves normalized for the solvent effect on the
fluorescence signal of the toxins, previously adsorbed in a C18 bonded silica packed microchannel. Elution was
performed during 10 minutes and fluorescence images acquired every 30 seconds at the beginning of the microchannel
with an exposure time of 1 second and a gain of 1x. The elution profiles taking into account the solvent effect were
calculated by dividing the fluorescence values for each curve by the ratio between the toxin fluorescence signal at a given
mobile phase concentration and the same fluorescence signal in water. The curves were also normalized for the
background noise originated from the beads fluorescence.
67