Modeling, Realization and Characterization of Microreactors in Lab-on-a-Chip Devices PDF Free Download
1 / 124/124
100%
Budapest University of Technology and Economics
Faculty of Electrical Engineering and Informatics
Modeling, Realization and Characterization
of Microreactors in Lab-on-a-Chip Devices
PhD Thesis
Author: Ferenc Ender
Advisor: Prof. Vladimír Székely, member of HAS
Department of Electron Devices
Budapest, 2016.
Nyilatkozat önálló munkáról, hivatkozások
átvételéről
Alulírott Ferenc Ender kijelentem, hogy ezt a doktori értekezést magam készítettem és abban csak a
megadott forrásokat használtam fel. Minden olyan részt, amelyet szó szerint, vagy azonos tartalomban,
de átfogalmazva más forrásból átvettem, egyértelműen, a forrás megadásával megjelöltem.
Budapest, 2016. május 19.
Nyilatkozat nyilvánosságra hozatalról
Alulírott Ferenc Ender hozzájárulok a doktori értekezésem Interneten történő nyilvánosságra hozata-
lához az alábbi formában:
•korlátozás nélkül
•elérhetőség csak magyarországi címről
•elérhetőség a fokozat odaítélését követően 2 év múlva, korlátozás nélkül
•elérhetőség a fokozat odaítélését követően 2 év múlva, csak magyarországi címről
Budapest, 2016. május 19.
Contents
1 Introduction 6
1.1 Objectives ........................................... 7
2 Principles 9
2.1 Microuidics and related technologies . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.1.1 Introduction to microuidics . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.1.2 Technologyoverview................................. 11
2.1.3 Systems ........................................ 12
2.1.4 System level modelling of microsystems . . . . . . . . . . . . . . . . . . . . . . 13
2.2 Theory and modelling of two phase ows . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.2.1 Classication of two phase ows . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.2.2 Segmented ow in microuidics . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.2.3 Dropletformation .................................. 17
2.2.4 Dropletmerging ................................... 18
2.3 Modelling of heat transfer in microchannels . . . . . . . . . . . . . . . . . . . . . . . . 18
2.3.1 Convective heat transfer in tubes . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.3.2 Thermal compact modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.4 Catalytic reactions in microscale . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.4.1 Interaction between macromolecules and ligands . . . . . . . . . . . . . . . . . 22
2.4.2 Enzymekinetics ................................... 23
2.4.3 Enyzmes in packed bed microreactors . . . . . . . . . . . . . . . . . . . . . . . 26
2.4.4 Phenylalanine ammonia-lyase (PAL) . . . . . . . . . . . . . . . . . . . . . . . . 27
2.5 Analysis methods of reaction kinetics . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.5.1 Analysis of the reaction kinetics by calorimetry . . . . . . . . . . . . . . . . . . 28
2.5.2 Spectroscopic analysis of reaction kinetics . . . . . . . . . . . . . . . . . . . . . 28
2.6 Magneticnanoparticles.................................... 31
2.6.1 Magnetic properties of MNPs . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.6.2 Spherepackingtheory................................ 33
3 State-of-the-art 34
3.1 Thermal design of microreactors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.2 Nanoparticlecounting .................................... 36
3.2.1 Ensemblemethods .................................. 37
3.2.2 Single particle counting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.3 Microreactors ......................................... 41
1
CONTENTS 2
4 Thermal compact model for droplet microreactors 46
4.1 Introduction .......................................... 47
4.2 Modellingmethods ...................................... 47
4.2.1 Governingequations................................. 48
4.2.2 Switched capacitor approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
4.2.3 Simplications .................................... 50
4.2.4 Setting up the model solver . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
4.2.5 BuildingtheAENmodel............................... 54
4.2.6 CFD simulation settings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.3 Results ............................................. 56
4.3.1 Performanceanalysis................................. 56
4.3.2 Model validity range analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
4.3.3 CFD comparison for constant heat ux boundary . . . . . . . . . . . . . . . . . 58
4.3.4 Comparing FEM and AEN solutions . . . . . . . . . . . . . . . . . . . . . . . . 59
4.4 Applicationexample ..................................... 59
4.4.1 Modelling the enzyme reaction . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
4.4.2 Temperature prole analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.5 Conclusion........................................... 61
4.6 Summaryofscienticresults................................. 62
5 Equipment 63
5.1 Microuidictestbench..................................... 63
5.1.1 uFLUStudioFramework............................... 63
5.1.2 Controller....................................... 66
5.1.3 FluidControlUnit .................................. 66
5.2 MagneChip........................................... 67
5.2.1 Characteristics .................................... 68
5.3 Construction methods of microuidic chips . . . . . . . . . . . . . . . . . . . . . . . . . 69
5.3.1 4cellMagneChip................................... 69
5.3.2 2 cell MagneChip with integrated magnetosensor . . . . . . . . . . . . . . . . . 69
5.4 Unit operations of MagneChip . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
5.4.1 Filling of the Magne-Chip microreactors with MNPs . . . . . . . . . . . . . . . 71
5.4.2 Cleaningthechip................................... 72
5.5 Method of single parameter experiments in-chip . . . . . . . . . . . . . . . . . . . . . . 72
5.6 Method of multi parameter experiments in-chip . . . . . . . . . . . . . . . . . . . . . . 73
5.6.1 Calibration ...................................... 74
5.6.2 Fluid handling during the Experiment Cycles . . . . . . . . . . . . . . . . . . . 74
5.6.3 Variants of experiment cycles . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
6 In-situ quantication of magnetic nanoparticles in a microchamber 76
6.1 Introduction .......................................... 76
6.2 MaterialsandMethods .................................... 77
6.2.1 Magnetic nanoparticle suspensions . . . . . . . . . . . . . . . . . . . . . . . . . 77
6.2.2 Immobilization PcPAL onto MNPs . . . . . . . . . . . . . . . . . . . . . . . . . 78
6.2.3 Oscillator circuit and frequency measurement . . . . . . . . . . . . . . . . . . . 78
6.2.4 Calibration of the sensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
6.2.5 Measurement of the entrapped particle quantity . . . . . . . . . . . . . . . . . . 80
6.3 Results ............................................. 81
CONTENTS 3
6.3.1 Characterization of the sensors . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
6.3.2 Measurements of the MNP quantity in the chamber . . . . . . . . . . . . . . . . 82
6.3.3 Particle volume fraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
6.4 Conclusions .......................................... 84
6.5 Summaryofscienticresults................................. 85
7 Lab-on-a-Chip microreactor platform 86
7.1 Introduction .......................................... 87
7.2 Materials............................................ 87
7.2.1 Phenylalanine ammonia-lyase from parsley (Petroselinum crispum) . . . . . . . 87
7.2.2 Chemicals....................................... 88
7.3 Methods ............................................ 88
7.3.1 UV characterization of the substrates . . . . . . . . . . . . . . . . . . . . . . . . 88
7.3.2 Referencemeasurements............................... 88
7.3.3 Chip selection and uid handling methods . . . . . . . . . . . . . . . . . . . . . 88
7.3.4 Summary of MagneChip parameter settings for enzymatic reactions . . . . . . 89
7.3.5 Optical inspection of the chambers . . . . . . . . . . . . . . . . . . . . . . . . . 89
7.3.6 Numerical modeling of the chambers . . . . . . . . . . . . . . . . . . . . . . . . 90
7.3.7 Calculation of kinetic parameters . . . . . . . . . . . . . . . . . . . . . . . . . . 90
7.4 Results ............................................. 91
7.4.1 General assumptions on the reliability of MagneChip experiments . . . . . . . 91
7.4.2 Failures of MNP layers detected by visual inspection . . . . . . . . . . . . . . . 92
7.4.3 Assessment of the reliability of multiparameter measurements . . . . . . . . . 94
7.4.4 Referencemeasurements............................... 94
7.4.5 Eect of particle size on the enzymatic activity . . . . . . . . . . . . . . . . . . 96
7.4.6 Inuence of the ow rate on biotransformation with 1a ............. 98
7.4.7 Calculation of the kinetic parameters of the transformation of l-1a to 2a . . . . 98
7.4.8 Substrate screening with MNP biocatalyst in the MagneChip system . . . . . . 100
7.5 Conclusion........................................... 101
7.6 Summary of the scientic results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
8 Utilization of the results 104
8.1 Compact model for the nutrient transport in blood capillary vessels . . . . . . . . . . . 104
8.2 Finding a new operation mechanism of PAL enzyme . . . . . . . . . . . . . . . . . . . . 104
CONTENTS 4
Köszönetnyilvánítás
Szeretném köszönetemet kifejezni mindazoknak, akik segítettek a disszertációban bemutatott tudo-
mányos munkám során. Elsősorban köszönettel tartozom Poppe László professzornak a BME Szerves
Kémia és Technológia Tanszékről, akitől rengeteg szakmai támogatást kaptam az enzimes kísérletek
kidolgozása és eredményeim publikálása során. Köszönöm a segítséget Weiser Diánának a BME Szerves
Kémia és Technológia Tanszékről, hogy úgy mint kiváló szakember és mint barát, mellettem állt és
minden helyzetben támogatott. Köszönöm hallgatóimnak, különösen Drozdy Andrásnak, Pálovics Péter-
nek, Vitéz Andrásnak és Sallai Gábornak a lelkes munkájukat és a rutinmérések során nyújtott segít-
ségüket. Köszönöm a szakmai és emberi támogatást a BME Elektronikus Eszközök Tanszéke vezetőinek,
Dr. Rencz Mártának és Dr. Poppe Andrásnak. Köszettel tartozom Dr. Sántha Hunornak az Elektronikai
Technológia Tanszékről, aki elindított a tudományos pályán. Végül tiszteletemet fejezem ki konzu-
lensemnek Dr. Székely Vladimírnek, akinek a tudományos életművét példaértékűnek tekintem. Kö-
szönöm az MTA-EK-MFA a mikrouidikai chipek elkészítésében nyújtott támogatását.
Köszönöm a támogatást családomnak. Köszönöm Édesapámnak, hogy mindig mögöttem állt és támog-
atott a legnehezebb döntésekben. Végül köszönöm páromnak, hogy példát mutatott munkabírásból és
kitartásból.
CONTENTS 5
List of Abbreviations
AEN Analogous Elecrical Network
CCD Charge Coupled Device
CFD Computational Fluid Dynamics
CTM Compact Thermal Model
DLS Dynamic Light Scattering
FEM Finite Element Model
HOMO Highest Occupied Molecular Orbital
LoC Lab-on-a-Chip
LUMO Lowest Unoccupied Molecular Orbital
MEMS Micro Electromechanical Systems
MNP Magnetic Nanoparticle
MNP-PAL MNP immobilized with PAL
MOR Model Order Reduction
NMR Nuclear Magnetic Resonance
PCR Polymerase Chain Reaction
PDE Partial Dierential Equation
PDM Physical Device Models
PoC Point-of-Care
RCM Resonant Coil Magnetometer
ROM Reduced Order Modelling
SNR Signal to Noise Ratio
SPICE Simulation Program with Integrated Circuit Emphasis
SQUID Superconducting Quantum Interference Device
SVR Surface to Volume Ratio
UV-Vis Ultraviolet and Visible Spectral Range (also ∼spectroscopy)
VHDL Very High Speed Integrated Circuit Hardware Description Language
VHDL-AMS Analog and Mixed Signal Library for VHDL
Materials
COC Cyclic Olefyn Copolymer
GOD Glucose Oxidase Enzyme
IPA Isopropyl Alcohol
PAL Phenylalanine Ammonia Lyase
PDMS Polydimethylsiloxane
PEG Polyethylene Glicol
PTFE Polytetrauoroethylene
TEOS Tetraethyl orthosilicate
Chapter 1
Introduction
Gordon Moore’s prediction on the development of integrated circuits [1] became a standard measure
in the technology progress of microsystems. This unbroken development in the recent 50 years led to a
technology breakdown in the related elds such as microtechnology and precision engineering. Short
after the rst presented micro-electromechanical systems (MEMS) e.g. pressure sensors in the 70’s,
chip-scale manipulation of biological samples became reality, establishing the basics of Lab-on-a-Chip
technology. Microscale working with biosamples and bio related data, such as DNA, have undergone a
steady development that nowadays requires a specialized and devoted infrastructure of robotics, bioin-
formatics, computer databases and instrumentation [2]. In the progress of its development, the cost
per reaction of DNA sequencing has fallen with a Moore’s Law precision, and, most recently, has been
developing even faster [3].
Microuidics and Lab-on-a-Chip technology have its own role in chemical and especially in biochem-
ical analysis and became even more signicant in microreactor technology where small size makes
enzymatic processes more eective and economical [4, 5, 6].
Microreactors are usually dened as miniaturized reaction systems fabricated by using methods of
microtechnology and precision engineering. The term ’microreactor’ is the proposed name for a wide
range of devices, having typically sub millimetre channel dimensions which can be further divided into
submicron sized components e.g. micro and nanoparticle carriers[7].
Before the evolving of microreactor technology, the traditional way to conduct solution phase syn-
thesis and analysis was the conventional batch mode in stationary reactors with stirring or shaking as
the only means of mix reactants. Today, micro structured devices oer greatly enhanced performance
compared to conventional batch systems due to eects arising from the microscale domain:
•pBatch processes are space-resolved therefore the process must be readjusted in each demand for
larger product quantities. In contrast, ow-microreactor processes are time-resolved therefore
the output of the reaction is determined by the ow rate and the operation time and no further
optimization is needed. This also leads to accelerated process development and enhanced safety
due to smaller reactor volumes.[8]
•Microreactors with high surface to volume ratios (SVR) are able to absorb heat created from
a reaction more eciently than any batch reactor. Therefore the reactions are subjected to a
homogeneous temperature distribution inside the reactor volume. In contrast, small SVR usually
leads to uneven temperature distribution in large scale batch reactors, decreasing the product
yield.[8]
•Mixing quality is crucial for many reactions where the molar ratio between the reactants needs
to be controlled precisely. Short diusion paths provide ecient mixing in microreactors, which
6
1. CHAPTER. INTRODUCTION 7
overrides the achievable mixing eciency of batch reactors. [8]
•In biocatalytic applications, the eciency of the microreactor can be further improved by immob-
ilization of enzymes on nanoscale carriers accommodated in the reactor volume. Re-usability of
the biocatalyst makes the process economical and more environmental friendly.
•The possibility of performing similar analyses in shorter reaction time-scale even in parallel is
an attractive feature for screening and routine use [9] in protein and enzyme research. A desir-
able goal is the high throughput screening of enzymes and their substrates and inhibitors. The
prospective elds of application of microreactors are quite wide, include biotechnology, as well
as combinatorial chemistry and enzyme targeted drug search [10].
•Analytical systems which comprise microreactors are characterized by outstanding repeatability
and reproducibility, due to replacing batch iterative steps and discrete sample treatment by ow
injection systems [10]. Beneting from system automation, this also eliminates errors associated
with manual protocols.
•The above benets lead to a more exible response to market demands gaining a high poten-
tial of using microreactors in industry, as the development results can be faster transferred into
production at lower costs.
Despite of the rapid development of enzymatic microreactors in the recent decade, important design
questions still need to be answered.
Reaction kinetics is a key parameter of device design. Widely used kinetic parameters are deduced
from the Michaelis-Menten model, which has a limited validity to batch reactions only. In ow systems
the eects introduced by the ow itself should be also considered. Further complication in modelling
can be expected from the immobilized enzymes. On one hand, immobilization may aect the kinetic
parameters, on the other hand, the kinetic model should also be changed as the liquid and solid phases
are moving related to each other.
Microreactors are built of a reaction chamber which may be lled by an appropriate carrier of the
catalyst. The reproducible loading of this carrier is not always straightforward especially in micro scale.
Even more challenging the determination of the actual loaded quantity of the carriers and biocatalysts
Long-time stability of the reactor and the reproducibility of the measurements may be aected by the
ow rate, the substrate concentration, the immobilized biocatalyst morphology etc.
Micro scale may arise issues also in thermal design as the eects of axial heat conduction, viscous
dissipation and low Reynolds numbers should be also considered.
1.1 Objectives
Principles of related elds such as single and two phase microuidics, heat transfer, enzyme kinetics
and nanoparticles are summarized in the next chapter. Recent developments and the state-of-the art
are summarized; the challenges are also identied in each related eld. The following objectives of this
dissertation work intend to make an impact in relation with the challenges outlined below.
•Thermal aspects of device design may play a key role in some bioMEMS devices such as droplet
polymerase chain reaction (PCR) microreactors and nanocalorimeters. Approaches on thermal
modelling have been demonstrated recently but no general solution has been presented so far
for eective assistance of droplet based LoC devices design.
Objective 1: To construct a thermal compact model which provides direct input for a subsequent
1. CHAPTER. INTRODUCTION 8
transient analysis and handles transient chemical (e.g. enzyme) reactions taking place inside the
droplets and resulting in a temperature eld as output.
•Possible applications of packed bed microreactors were demonstrated in biocatalysis and in en-
zyme screening, however, some questions still need to be claried. An ever arising question is
the accurate determination of the biocatalyst concentration in the reactor. Although several ad-
hoc methods were already presented, there is no standardized method to measure the quantity
of immobilized enzymes in micro chambers.
Objective 2: To implement a method for accurate, in-situ and on-line determination of the
amount of biocatalyst particles in a microuidic reaction chamber.
•A layer built up of biocatalyst carriers (e.g. nanoparticles) may change its ne structure due to
viscous eects caused by the owing medium in the reactor. Changes in the layer structure may
aect the activity of the biocatalyst.
Objective 3: To analyse the eects of structural changes in the layer structure on the biocatalytic
activity. Furthermore, to create a measure to describe the structural changes and investigate the
requirements of making reproducible measurements with enzyme biocatalysts.
•Particle size and distribution undoubtedly aect the achievable enzyme activity in microreactors.
However, the eect of dierent particle sizes or using a mixture of dierent particles have not
been analysed before.
Objective 4: To investigate the eect of using dierent particle sizes on the enzymatic activity
and on the possible loading capacity of the microchambers.
•Due to their re-usability, working with enzymes in microreactors makes the process environ-
mentally friendly and economical.
Objective 5: To analyse the kinetics of immobilized enzymes in a chip sized microreactor sys-
tem and to present a method to carry out multi-parametric measurements providing that the
biocatalyst is reused during the measurements.
Chapter 2
Principles
2.1 Microuidics and related technologies
2.1.1 Introduction to microuidics
At the microscale, the forces which may be negligible in macroscale (everyday life), can become dom-
inant and vice versa. Because of downscaling, shrinking existing large devices and expecting them to
function well at the microscale is often counter-productive, although proper design enables function-
alities which are unreachable at macroscale [11].
Microuidics has the potential to change the way modern biology is performed due to the possibility
of working with small reagent volumes, shorter reaction times, excellent controllability and the pos-
sibility of parallel operation and therefore gaining higher throughput. Ecient technologies may be
utilized for the cheap and economical mass production of microuidics devices.
In the past decade two main areas have prevailed in microuidics research and development. One is
related to the precise handling of biological liquid samples and detecting analytes. The eld of Lab-on-
a-Chip (LoC) devices [12] has emerged from in-line or on-chip detection. The other eld has emerged
from the enhanced heat transport in microscale, which created the possibility to design and devise
cooling devices on-chip for integrated circuits [13].
A general aim of microuidics design is the intended use of the scaling eects in order to achieve
the above mentioned functionalities. Such eects becoming dominant in microuidics include laminar
ow, diusion, uid resistance, high surface area to volume ratio and surface tension [11].
Reynolds Number is dened as the ratio of the inertial and viscous forces of a uid ow. Re is a
dimensionless property which describes the ow pattern of the ow. One of the basic concept related
to Reynolds number is the dynamic similarity theorem of uid mechanics. It states that two vessels
with the same boundary conditions and same Re will exhibit the same uid ow. Besides the uid ow
velocity v, the Reynolds Number Re depends on geometry and material properties only.
Re =ρvDh
µ(2.1)
where ρis the density and µis the dynamic viscosity of the uid. Geometry constraint is dened by
the Dhhydraulic diameter which is related to the cross sectional geometry of the channel. At high Re
(typically greater than Re > 2500) inertial forces override viscous forces and the ow may turn into
turbulent which is a complicated ow structure of interacting vortices. Taking the typical geometry
and ow considerations of microuidic devices, viscous forces become dominant resulting in low Re.
9
2. CHAPTER. PRINCIPLES 10
In this case laminar ow develops which is smooth and predictable [14].
Laminar Flow is a condition in which the velocity distribution of a uid ow, far enough from the
uid entrance is invariant in space and time. One important consequence of this invariability is that the
analytical solution of the velocity prole can be obtained directly by solving the Navier-Stokes equations
[15]. In case of circular pipes the velocity prole is parabolic and depends on the geometry and the axial
pressure gradient only. This follows to the Hagen–Poiseuille equation
∆P=32µLv
d2=128µLQ
πd4=QR (2.2)
where ∆Pis the pressure gradient between the ends of the pipe, Lis the length, dis the diameter,
R=128µL
πd4is the uid resistance of the pipe, respectively. The uid ow can be characterized even by
the ow velocity vor by the volumetric ow rate Q. The above approach suggests the electrical circuit
analogy (Ohm’s Law) therefore basic uid ow calculations can be done with ease.
Diusion is the dominating radial mass transfer method in laminar ow. Note that viscous forces
constrain the uid’s molecules to move on the axial direction only, according to the direction of the
pressure gradient. Taking the derivation of the Fick’s equation in one dimension the moving distance d
of a particle over ttime is given by
d=p2Dtdiff (2.3)
where Dis the diusion constant of the particle. By rearranging, the diusion time can be deduced
tdiff =d2
2D(2.4)
Surface Area to Volume Ratio (SV R) is another factor that becomes important at the microscale
and dened as
SV R =A
V(2.5)
Table 2.1. Scaling eect of some related physical quantities
Quantity Scaling factor (of size) Microscale example
(d= 50 µm)
Macroscale example
(d= 5 mm)
Re [L1]Re = 0.5Re = 50
∆P[L−2] 6.4 kPa 0.64Pa
tdiff [L2] 5 s 14 days
SV R [L−1] 2.7·105m−1842 m−1
Scaling of the related quantities A comparison of two similar but dierent sized devices is shown
in Table 2.1.
The macroscopic device is a circular duct with a diameter of 5 mm and length of 5 cm. Water is owing
through the duct with a ow rate of v= 1 cm s−1.
The microscopic device is a circular duct with a diameter of 50 µmand length of 5 cm. Water is owing
2. CHAPTER. PRINCIPLES 11
through the duct with a ow rate of v= 1 cm s−1.
Relevant quantities such as Re,∆P,tdiff and SV R are calculated for both devices. Scaling factor is
denoted by the Trimmer’s notation [Ls][16].
Diusion mixing time is considered as the homogeneous mixing of a solution of 30 bp DNA strands.
Taking typical geometry sizes of microchannels, Re certainly falls into the laminar ow regime, how-
ever high driving pressures are expected in the range of kPas. Pressure drop found to be an important
design parameter of blood plasma separator devices using the Zweinfach-Fung eect [17]. The squared
scaling of diusion time provides the using diusion mixers in microscale. Eorts had been made in
the recent decade to improve mixing eciency by using zig-zag channel layouts [18, 19]. Surface to
volume factor increases linearly as the size reduces, which provides more eective surface reactions
and also reduced reaction time compared to the macroscale e.g. in growing complex organisms, such
as bacteria [20].
2.1.2 Technology overview
The current technologies used for fabricating microfuidic devices include micromachining (a.k.a. MEMS
technology), soft lithography, embossing, injection moulding and laser ablation [11].
Micromachining is originated from the integrated circuit technology and typically uses silicon as
the construction material [21]. Distinguishable to surface and bulk micromachining, the technology is
widely used to construct micro-electromechanical systems (MEMS) and one of the rst approaches to
fabricate microuidic devices. Using bulk micromachining, which denes the structures using selective
etching, microchannels can be constructed. Nanopillars were formed in silicon microchannels by using
DRIE (Deep Reactive Ion Etching) method [22]. Microchannels were formed in a microcooler struc-
ture by anisotropic wet chemical etching using potassium hydroxide (KOH) [23]. In contrast, surface
micromachining relies on the deposition of structural thin lms on the wafer surface, resulting in relat-
ively easy integration of the micromachined structures with on-chip electronics. SU-8 is a typical choice
to construct thin lm layers for microuidics [24]. As a supplementary technique, substrate bond-
ing can be used either for integrating functionalities or for packaging. Wafer bonding (silicon–silicon,
silicon–glass, and glass–glass) is frequently used to fabricate complex 3-D structures. The two most
important bonding techniques are silicon–silicon fusion (or silicon direct bonding) and silicon–glass
electrostatic (or anodic) bonding [21].
Soft lithography is faster, less expensive, and more suitable for most biological applications than
glass or silicon micromachining [11].
PDMS (polydimethylsiloxane) is a popular material of choice for microuidic devices due to its low
cost, ease of fabrication, oxygen permeability and optical transparency [25]. PDMS microchips can be
fabricated through microscale moulding processes. For laboratory use, a silicon wafer with patterned
photo resist can be used as a mould master. To have a relatively thick structure of microchannels and mi-
crochambers for transportation and/or incubation of the reagents and samples, an ultra thick photores-
ist, SU-8 was adopted. After the patterning, prepolymer of PDMS is poured into the mould master. Cured
PDMS is peeled o from the master to be pasted on a at plate, i.e. PMMA (polymethylmethacrylate),
glass, etc. On PDMS channel body access ports for introduction of the reagents and samples should be
drilled in advance [26]. Ports can be also created by punching the PDMS body. The at plate and the
PDMS body can be bonded through free standing siloxane bonds. The process is initiated by oxygen
or air plasma treatment [27]. However PDMS is generally treated to be biocompatible, in some cases
biocompatibility has to be improved by surface modications such as covalent linking of hyaluronic
2. CHAPTER. PRINCIPLES 12
acid on the PDMS surface [28]. PDMS’s hydrophobicity and fast hydrophobic recovery after surface
hydrophilization is a general issue in many bioMEMS applications therefore attempts have been made
by many research groups to devise longer lasting surface modications of PDMS [25].
Other techniques Injection moulding is a very promising technique for low cost and production
of microuidic devices [29]. Thermoplastic polymer materials are heated past their glass transition
temperature to make them soft and pliable. The molten plastic is injected into a cavity that contains
the master. Because the cavity is maintained at a lower temperature than the plastic, rapid cooling of
the plastic occurs, and the moulded part is ready in only a few minutes [11]. Similarly, hot embossing
is a exible, low-cost microfabrication method for polymer microstructures, which uses the replica-
tion of a micromachined embossing master to generate microstructures on a polymer substrate [30].
Another method of forming microuidic devices is laser ablation of polymer surfaces such as PMMA
(poly(methyl methacrylate)) or COC (cyclic olen copolymer) [31].
2.1.3 Systems
The ultimate goal of microuidic systems is a “Lab-on-a-Chip” (LoC) — the incorporation of multiple
aspects of modern biology or chemistry labs on a single microchip. One of the rst mention of the term
’Lab-on-a-Chip’ was found in the work of Ramsey et. al. (1995) [32]. In the state-of-the-art of LoC tech-
nology semiconductor sequencing had one of the greatest impact [3] and stands here as an example of
the cutting edge but already commercialized LoC technology. This remarkable method broke through
the $1,000 limit of genome sequencing cost for the rst time and therefore opened up the perspectives
of next-generation sequencing. The device is the clever integration of semiconductor technology, mi-
crofabrication, electrochemical sensing and microuidics in a chip-sized device and uses a variety of
the advances of microtechnology discussed so far.
Figure 2.1. a) Simplied drawing of a well of the semiconductor sequencer, a bead containing DNA tem-
plate, and the underlying sensor and electronics. Protons (H+) are released when nucleotides (dNTP)
are incorporated on the growing DNA strands, changing the pH of the well (∆pH). This induces a
change in surface potential of the metal-oxide-sensing layer, and a change in potential (∆V) of the
source terminal of the underlying eld-eect b) Die in ceramic package wire bonded for electrical con-
nection, shown with moulded uid lid to allow addition of sequencing reagents. Figure reprinted from
[3]
2. CHAPTER. PRINCIPLES 13
DNA-template is prepared using droplet microreactor based PCR on high surface-to-volume ratio poly-
mer micro-beads which are spread homogeneously in the micromachined well array of a silicon chip.
Ion-sensitive (ISFET) chemical sensors are integrated below the wells, together with the seamlessly in-
tegrated CMOS readout electronics in the same silicon substrate (Figure 2.1,a). Highly parallel reaction
occurs in each well (microreactor) when the mixture of a selected dNTP and polymerase enzyme is
driven through the chip (’nucleotide ow’) chamber which is designed to maintain laminar ow on
the whole chip surface. The process is repeated sequentially using dierent nucleotide in the ow, en-
abling the detection of the polymerisation of the forthcoming nucleotide bind in each well (depending
on the chip size, 1.1 million-11 million wells per chip). Due to short diusion times each cycle is car-
ried out within 5 seconds. One of the rst demonstration attempt of the usage the new device was the
sequencing of Gordon Moore’s genome, author of Moore’s law [3].
2.1.4 System level modelling of microsystems
System level modelling is a top-down design methodology which became one of the common design
methods of complex microsystems [34]. The basic concept of system modelling is to provide the beha-
vioural description of the system using a high level language e.g. System-C [35] or VHDL-AMS1[36].
Design steps follow each other [33] from the higher level downwards as follows (Figure 2.2)
•Device (LoC) description lies at the highest, ’system’ level and represents the highest level of
abstraction as a pure behavioural model [37, 38]. The structural part of the system description is
aschematic, a textual (e.g. System C [35]) or graphical (e.g. LabVIEW, SimuLink) representation
of the interconnection network of system components. The layout of the device can be generated
directly from the description. Detailed simulation provides information to validate if the system
fulls the design constraints and layout modications can be done on request. In contrast, system
1VHDL-AMS stands for Very High Speed Integrated Circuit Hardware Description Language for Analog and Mixed
Signals
LoC description
Schematic (RTL)
Behavioral
simulation
Meet
the
Spec?
Fabrication
Outputs
Element
Library
Layout
generation
Detailed
simulation
Meet
the
Spec?
No
No
Figure 2.2. An integrated top-down design automation environment for microuidic biochips (after
[33]). Compact models and behavioural description provides a shorcut in the design process (red path)
2. CHAPTER. PRINCIPLES 14
components may be represented by the library elements of compact models, which are connected
to each other through algebraic equations.
•Compact models or Reduced Order Models (ROMs) are derived from Physical Device Models (PDMs)
by negligating, approximating or linearising secondary physical eects act on the device while
the primary eects are described by partial dierential equations or algebraic equations. Com-
pact models are considered as a behavioural description of the device, therefore even multi do-
main functionalities can be handled together at a high level of abstraction. Detailed numerical
modelling usually provides high accuracy, requires high computation times, though. In contrast,
reduced order models focus on the key behaviour of the device and require only moderate com-
putation time. Therefore bypassing the numerical modelling step the design iteration time can
be signicantly reduced. Figure 2.2, red track shows a possible shortcut in the design ow by
utilizing compact models. Dierent methods were developed for ROM generation.
–Set of Algebraic Equations can describe the device’s behaviour with a given accuracy. The
equations can be solved directly or by numerical methods [39]. Initial values are coming
from preceding calculations, even from other models. Calculated values are passed towards
to subsequent simulations.
–Model Order Reduction (MOR) is an automated method to generate a set of Ordinary Dif-
ferential Equations from the original set of Partial Dierential Equations e.g. by the Modal
Superposition Method [40], used especially for mechanical systems. These models are usu-
ally sharing a common interface with FEM elements.
–Analogous electrical network representation. Components of electrical networks may rep-
resent analogous quantities in mechanical [36], ow [33, 41] or thermal systems [42]. An
important advance of this models is the straightforward integration with electrical subdo-
mains of the modelled system.
•Detailed simulation, also referred as Finite Element Analysis (FEM) module, deals with the 2D
or 3D geometrical representation of the device which is discretized into elements dened by the
’mesh’ grid. After dening the boundary conditions, loadings and initial values, a set of partial
dierential equations (PDEs) describe the corresponding physics domain (mechanics, uidics,
thermal, electrostatics etc) are solved numerically [43, 44, 45]. Coupling of the elds (e.g. electro-
thermal) is also possible. The result of the analysis is a set of Degree of Freedom (DoF) values
(e.g. temperature, displacement etc.) in each nite element.
2.2 Theory and modelling of two phase ows
2.2.1 Classication of two phase ows
Multiphase ows provide several mechanisms for enhancing the performance and extending the func-
tionalities of single phase microuidic systems. Even the shorter diusion times related to microscale
often limits the throughput of single-phase ow systems. Diusion time can be reduced by adding a
second, immiscible, uid stream that enhances mixing and transverse channel transport by inducing a
recirculation motion in the liquid [48]. The resulting multiphase system can also prevent a liquid from
direct contact with microchannel walls and thereby eliminate or reduce the unwanted deposition of
material on wall surfaces. Multiphase ows are created when two of more partially or immiscible uids
(liquids or a liquid and a gas phase) are brought in contact[49, 50, 44].
2. CHAPTER. PRINCIPLES 15
(a) (b) (c) (d) (e) (f) (
g
) (h)
Slug or Annular
Annular
Slug
Bubble
ReGs
ReLs
ReLs
Ca
10−310−210−1100101102103
107
106
105
104
Figure 2.3. Left: Sketch of observed ow patterns in capillary channels. (a,b): bubbly ow, (c,d) segmen-
ted ow (a.k.a. bubble train ow, Taylor ow, capillary slug ow), (e) transitional slug/churn ow, (f)
churn ow, (g) lm ow (downow only), (h) annular ow. Right: ’Flowmap’ from Jayawardena et al.
(after Jayawardena et al., 1997) [46]. Figures reprinted from [47]
Multiphase microows are characterized by the ratio of viscous to surface forces, the capillary number
(Ca), and by the ratio of the Reynolds numbers Re of the uids [46]. Capillary number is the ratio of
interfacial tension and viscous forces. Ca can be dened as follows:
Ca =µv
γ(2.6)
where γis the surface tension between the two uids. Based on the ratio of the Reynolds number of the
carrier liquid and the capillary number (ReLs
Ca ) and the ratio the Reynolds numbers of the two immiscible
uids (ReGs
ReLs ) dierent ow patterns may develop [47] as it it shown in (Figure 2.3).
•In bubbly ow, when small bubbles dispersed in the continuous, wetting liquid (Figure 2.3,a,b).
In this regime Ca < 1therefore the interfacial tension dominates resulting in spherical bubbles.
•Taylor ow, sometimes called plug ow, slug ow, bubble train ow, segmented ow or inter-
mittent ow is the ow pattern of large long bubbles that span most of the cross-section of the
channel. The relevant lengths are mainly determined by the inlet conditions [51] (Figure 2.3,c,d).
In this region Ca is typically larger than 1resulting in elongated and asymmetric bubbles.
•At higher velocities, small satellite bubbles appear at the rear of the slug, the pattern is called
churn ow. (Figure 2.3,e,f).
•At high velocities and low liquid fraction, the annular ow pattern exists only of a thin liquid
lm owing along the wall, while the core volume is occupied by the owing gas (Figure 2.3,g,h).
Bond number Bo is the ratio of gravitational and surface forces and dened as follows
Bo =∆ρgL2
γ(2.7)
In microuidic devices Bo 1can be found indicating gravitational forces can be neglected and
should be not considered in device design.
Finally Weber number W e is the ratio of inertial to surface forces and dened as follows
W e =ρv2L
γ(2.8)
2. CHAPTER. PRINCIPLES 16
Typical W e in microuidic devices tends to zero indicating the predomination of surface forces. In
deed, ne tuning of surface forces is the key point of droplet manipulation device design.
2.2.2 Segmented ow in microuidics
In the Lab-on-a-Chip usage of such reactors the term ’microdroplets’ usually refers to the above men-
tioned second type (Figure 2.3,b,c). Note, that droplets can be also manipulated based on the electrowet-
ting eect, sometimes referred as digital microuidics.
The key features of microdroplets in microuidics are as follows [50]. Microdroplets, forming a mi-
croreactor
1. provide a compartment in which species or reactions can be isolated,
2. are monodisperse and therefore potentially suitable for carrying out quantitative studies,
3. provide the possibility to work with extremely small volumes and single cells or molecules and,
4. oer the ability to perform very large numbers of experiments.
Taking this approach, each droplet is analogous to the traditional chemist’s ask [54], with the
added physical advantages of reduced reagent consumption, rapid mixing, automated handling, and
continuous rather than batch processing [50]. Multiphase microchemical systems (Figure 2.5) take ad-
vantage of the large interfacial areas (SVR), fast mixing and reduced mass transfer limitations to achieve
0
0.2
0.4
0.6
0.8
1
0 0.5 1 1.5 2 2.5 3 3.5 4
r/R
z/R
Figure 2.4. Bubble shape (bold line) of gas-liquid segmented ow and relative streamlines. Solid line,
clockwise circulation; dashed line, anti-clockwise circulation. Figure reprinted from [52]
Figure 2.5. Droplets formed within microuidic channels can serve as microreactors. The reactions
are performed within aqueous droplets, which contain reagents (A and C), and a separating stream
containing buer (B). The droplets are encapsulated by a layer of a carrier uid (D) and transported
through the microchannels. Figure reprinted from [53]
2. CHAPTER. PRINCIPLES 17
increased performance compared to conventional bench scale systems [48]. The use of isolated, well
mixed droplets enables kinetic studies of organic reactions at millisecond time scales [55], synthesis
of nanoparticles [56], biological applications include DNA analysis [57], droplet-based PCR [58], cell
encapsulation, and cell stimulus and lysis [48]. Manipulation techniques like well controlled droplet
formation, merging, splitting and incubation of the droplets were worked out to build up a complete
set of functions for Lab-on-a-Chip usage [59].
The velocity stream structure of the segmented ow was rst investigated by Taylor (Taylor, 1961)
[60], advancing the development of high performance numeric methods, CFD analysis was done later
by others [43, 61, 44, 45, 52] and optical investigation was done by Micro-Particle Image Velocimetry
(µPIV) [62, 49].
In Figure 2.4 the typical streamlines can be seen developed inside the bubbles and droplets. The elong-
ated anti-clockwise circulation is analogous to a caterpillar track and signicantly enhance the mass
and heat transfer inside the segmented compartments [52]. Enhanced mixing inside of liquid droplets
was proved experimentally [63, 64]. Enhanced heat transfer was investigated experimentally [65] and
by CFD analysis [44, 43].
2.2.3 Droplet formation
Droplet formation is the process of in-situ production of monodisperse droplets within the microuidic
device. Droplet formation can be passive and active by means of alternate external eld is applied on
the device corresponding the droplet generation frequency.
a
b
c
d
e
f
Figure 2.6. Droplet formation techniques: a) T-Junction dripping b) T-junction squeezing c) ow-focus
dripping d) ow-focus squeezing; Droplet merging techniques: e) merging in wide channels f) merging
with pillars
The review article of Gu et al. [59] summarized the most common techniques utilized nowadays:
•Passive T-junction devices utilize a T-shaped channel junction where the uid phase to be dis-
persed is brought into the microchannel while the immiscible carrier uid is driven independ-
ently through the other branch. These two phases meet at the junction, where the interface of
the two immiscible uids is deformed by the forces dictated by the ow conditions and geometry.
The competion between the viscous shear stress and capillary pressure results in the pinch-o
of the droplet, i.e. viscous shear stress override interfacial tension. Based on the ratio of these
forces the following scenarios could be developed:
–In case of dripping, typically develops at higher Ca droplet breakup happens before the
droplet would obstruct the channel (Figure 2.6,a).
–Alternatively, in the squeezing regime, the growing droplet may reach the channel wall and
restricts the carrier uid resulting in rapid pressure increment. This induces the pinch-o
of the droplet (Figure 2.6,b).
2. CHAPTER. PRINCIPLES 18
•Passive Flow-Focusing (FF) devices consist of three inlet channels converging into a main channel
via a narrow orice. The dispersed phase is squeezed by the continuous phase ows from the
sides. The laminar stream of the two phases ow through the orice than the narrowed dispersed
phase breaks apart into droplets. The droplet size is entirely determined by the ow rate ratio
and the orice geometry. Similarly to T-junction devices, droplet breakup can develop also in the
dripping and squeezing, furthermore in jetting regimes (Figure 2.6,c,d).
•Active devices utilize external elds to aect interfacial forces on demand e.g. by electrowetting.
2.2.4 Droplet merging
Merging of droplets is a key step for triggered start of chemical reactions. The perquisite of merging is
the touch of the two droplets and the overcome of stabilizing forces by surface tension and lubrication.
Gu et. al. dierentiated the following types [59]:
•In passive merging the channel is widened at a certain section. In this geometry the droplet velo-
city decreases in the widening channel because of drainage of the continuous phase. Due to this
changing ow eld, two subsequent droplets are allowed to come close and coalescence happens
(Figure 2.6,e).
•Droplets can be also merged by slowing down or stopping the leading droplet at widening channel
with an array of pillar elements, where the carrier phase is drained letting the subsequent droplet
come close (Figure 2.6,f).
•Droplets can be merged actively e.g. by applying voltages with opposite sign to the droplets. The
oppositely charged surfaces will attract each other as soon as the droplets come close to each
other and merging occurs.
2.3 Modelling of heat transfer in microchannels
2.3.1 Convective heat transfer in tubes
Convective heat transfer in tubes The Navier-Stokes equation describes the general energy, mo-
mentum and continuity relations of the uid ow. The following deduction of the energy equation is
strongly built on the 9th chapter of Convective Heat and Mass Transfer by W. M. Kays [15]. In case
of laminar ow, after a transient section where the velocity prole is space dependent and referred as
r
r0
x
u
q''
.
u
t0
t
Hydrodyamic entry region
Figure 2.7. Development of the velocity prole in the hydrodynamic entry region of a circular pipe, the
fully developed velocity prole and the fully developed temperature prole under constant wall heat
ux condition. (After [15])
2. CHAPTER. PRINCIPLES 19
hydrodynamic entry region, a parabolic velocity distribution is developed along the radial axis of the
channel (Figure 2.7). Assuming circular pipe with a radius of r, laminar ow with a ow velocity of
u, symmetrical and constant heat ux from the walls, hydrodynamically fully developed ow and no
axial conduction the following equation describes the energy state of the ow:
1
r
∂
∂r r∂t
∂r =uρc
k
dtm
dx (2.9)
where cis the specic heat and kis the thermal conductivity of the uid, respectively. tmis the mixed
mean uid temperature and dened as
tm=1
AcVZAc
utdAc(2.10)
where Acis the cross sectional area of the tube.
The uid temperature is necessary equals to the wall temperature at the wall i.e.
t=t0at r =r0(2.11)
As the temperature prole must be symmetrical once the heat ux is also symmetrical, the local max-
imum of the temperature prole is at the centreline of the channel
∂t
∂r = 0 at r = 0 (2.12)
After a transient section where the temperature prole may vary along the tube axis, the prole be-
come invariant with the tube length and thereafter is called fully developed temperature prole. The
hydrodynamic and thermal development regions are not necessarily the same.
Heat transfer coecient (h) is dened as
h=˙q00
t0−tm
(2.13)
where ˙q00 is the heat ux and t0is the local (axial position dependent) wall temperature. A heat current
will be also developed due to the temperature gradient along the radial axis of the pipe:
˙q00 =−k∂t
∂r r=r0
(2.14)
Nusselt Number Assuming fully developed velocity and temperature prole and by combining the
equations 2.9, 2.13 and 2.14, the following can be deduced for the ratio of hand k:
Nu =2hr0
k= 4.364 (2.15)
where Nu is the Nusselt number, which is simply the non-dimensional version of the heat-transfer
coecient. For non-fully developed ows, Nu may be dierent. Note, that by using Nu, the uid and
wall temperatures can be easily calculated at any length of the channel. The theoretical values of Nu
were determined for a great variety of boundary conditions and geometries. Empirical data is also
available for Nu(x)plots for the thermal development region. On one hand, despite its simplicity
the Nu model is quite well usable for engineering calculations. On the other hand, by increasing the
Reynolds number or decreasing the channel size, experimental heat transfer values show great deviance
from the theoretical values.
2. CHAPTER. PRINCIPLES 20
10−1100101102103104
0
1
2
3
4
5
6
7
8
Re
Nu
Conjugate
effects
Entrance
effects
Viscous
dissipation
Nu0
Figure 2.8. Scaling eects on the mean value of the Nusselt number for water, based on [66]
Scaling eects Several special problems related to heat transfer in micro-channels were observed
and analysed in depth: the eect of axial conduction in the channel wall [67], entrance eects and the
viscous dissipation eect (Figure 2.8) [68].
•at low Reynolds numbers the heat conduction along the solid wall of the channel is coupled to the
convection heat transfer inside the channel: this eect tends to reduce the mean Nusselt number
and is referred to as axial conduction or conjugate heat transfer. Due to small hydraulic diameters
of the channels and thick walls compared to the channel diameter, axial heat conduction in the
walls has to be taken into consideration. This phenomenon is generally neglected in the standard
macro-analysis stems from the small size of the systems. Disregarding this eect can lead to a
very large bias in the experimental estimation of heat transfer coecients. [67]. Consequently,
modelling of small scale uid systems requires the introduction of heat conductive shell regions
around the channel volume.
•Although this dissertation intends to focus on ows with small Reynolds numbers, entrance ef-
fects and viscous dissipation should be undoubtedly mentioned here. Entrance eects become
important at high values of the Reynolds numbers and tend to increase the mean value of the
Nusselt number [69]. In contrast, viscous heat generation at high Reynolds numbers [68] tends
to reduce the mean value of the Nusselt number.
2.3.2 Thermal compact modeling
Transient behaviour of electrical systems The transient behaviour of an electrical transmission
line can be described using the telegrapher’s equation [70]. Each innitesimally short segment of the
distributed system can be modelled by theoretical elementary components, represented by a two-port
circuit (Figure 2.9,a):
∂
∂xu(x, t) = −Ri(x, t)−L∂
∂ti(x, t)(2.16)
2. CHAPTER. PRINCIPLES 21
u(x, t)Cdx
RdxLdx
i(x, t)
Gdx
i(x+dx, t)
u(x+dx, t)
u(x, t)
Rdx
i(x, t)
Cdx
i(x+dx, t)
u(x+dx, t)T(x, t)
Rth
q(x, t)
Cth
a)
b) c)
Figure 2.9. Schematic representation of the elementary component of a transmission line a) detailed
model b) simplied model c) analogous representation of the elementary thermal model
∂
∂xi(x, t) = Gu(x, t)−C∂
∂tu(x, t)(2.17)
If the eect of the distributed magnetic eld and shunt conductance between the lines should not be
considered then a simplifying boundary condition can be applied i.e. L= 0 and G= 0 (Figure 2.9,b)
∂
∂xu(x, t) = −Ri(x, t)(2.18)
∂
∂xi(x, t) = −C∂
∂tu(x, t)(2.19)
Transient behaviour of thermal systems The distribution of heat in given region over time is
described by the heat equation. Assuming one dimensional problem
k∂2
∂x2T(x, t)−ρc ∂
∂tT(x, t) = 0 (2.20)
where kis the thermal conductivity, cis the specic heat, ρis the density and Tis the temperature.
Temperature gradient causes heat current of qas described by Fourier’s law:
q=−k∂
∂xT(x, t)(2.21)
where, again, the problem is assumed to be one dimensional. Substituting Eq. 2.21 into Eq. 2.20
−∂
∂xq(x, t)−ρc ∂
∂tT(x, t) = 0 (2.22)
by rearranging
∂
∂xT(x, t) = −(1
k)q(x, t)(2.23)
∂
∂xq(x, t) = −(ρc)∂
∂tT(x, t)(2.24)
The formal analogy is obvious between Eqs. 2.18 and 2.23 and similarly Eqs. 2.19 and 2.24. The analogy
can be summarized as follows and also represented as an elementary component in Figure 2.9,c.
2. CHAPTER. PRINCIPLES 22
Table 2.2. Electrical-thermal analogy of lumped thermal models
Electrical quantity Symbols Units Thermal equivalent Symbols Units
Electrical voltage u[V] Temperature T[◦C,K]
Electrical current i[A] Heat current q[W]
Electrical resistance R[Ω] Thermal resistance Rth =1
k[K W−1]
Electrical capacitance C[F] Thermal capacitance Cth =ρc [J K−1]
2.4 Catalytic reactions in microscale
2.4.1 Interaction between macromolecules and ligands
This and the following Section 2.5 are merely based on chapters 1.2, 2.1 and 2.2, 3.3, 3.4 of H. Biss-
wanger’s Enzyme Kinetics book [71].
Chemical reactions are initiated by the collision of molecules having sucient energy to react with
each other, resulted in conversion into products. In living matter this process is strictly controlled by
enzymes and only those compounds (the ligands) are converted into products, which were previously
selected from others. Selectivity is provided by the binding site of the enzyme and the ligand is a so
called substrate. The latter is selectively bond to the binding site and the enzyme catalyses the trans-
formation of the substrate to product.
In an arbitrary system, the substrate molecules are free to move by diusion obeying the Fick’s law,
characterized by the Ddiusion constant. Enzymes can be also free to move (free enzyme system) or
they can be attached to a x surface (immobilized enzyme system). In both cases, the probability of the
event that the substrate meets the macromolecule is characterized by the association rate constant,ka.
Assuming that both the enzyme and and the substrate are spherical and their distance is r, and the
binding site is regarded as a circular area, forming angle αwith the center of the enzyme, the following
relationship can be written to ka(known as Smoluchowski limit)
α
r
Enzyme
Binding center
Substrate
[E]+[S] [ES] [E]+[P]
Figure 2.10. Schematic illustration of the interaction of a substrate molecule with its binding site on
the enzyme, and the three steps of the catalytic reaction sequence of the product formation. : non-
substrate, : substrate, : product
2. CHAPTER. PRINCIPLES 23
ka= 4πrD sin α(2.25)
This approach suggests that every ligand should react whichever once met the binding site. Such un-
specic binding could not distinguish between the specic ligand and other metabolites. The gating
model, however assumes the binding site opened or closed like a gate by changing the physical con-
formation of the enzyme, thus modulating the accessibility of the binding site.
Taking the gating mechanism also into account, the kaassociation constant is limited by kcat, the so
called turnover number.
2.4.2 Enzyme kinetics
The order of a chemical reaction with respect to the individual components is dened as the power of
the component concentration included in the rate equation.
Zero order reaction is a reaction which is independent on the reactant concentrations. In this case
the reaction rate is dictated solely by the very limited amount of free enzyme which remains unchanged
during the reactions
v=−d[S]
dt =d[P]
dt =k(2.26)
Integration with respect to time gives a linear relationship
[S]=[S0]−kt (2.27)
Therefore zero order reactions can be identied by the linear progression of the substrate decay and
subsequent product formation (Figure 2.11).
0 5 10 15 20 25 30 35
0
1
2
3
4
5
6
7
Time [s−1]
Concentration
[S]
[E]
[ES]
[P]
12 3
Figure 2.11. Time-related changes of the reactants of an enzyme-catalysed reaction. 1) Pre-steady phase
2) steady-state phase 3) substrate depletion phase. (kinetic parameters: k1= 2 mol−1s−1k−1=
0.15 s−1k2= 0.5 s−1)
2. CHAPTER. PRINCIPLES 24
First order reaction is the conversion of a substrate Sinto a product P
Ak1
−→ P(2.28)
The reaction rate vcan be determined either from the time-dependent decrease of Sor the increase of
Pformed from Sby the k1rate constant
v=−d[S]
dt =d[P]
dt =k1[S](2.29)
As one can notice k1has a dimension of s−1i.e. independent on the concentration. By integration, the
decrease of the substrate can be written as
[S] = [S0]e−k1t(2.30)
Where [S0]is the initial substrate concentration. Decrease in substrate or increase in product proceeds
exponentially with time (Figure 2.11).
Let us assume that product Pis formed irreversibly from substrate Sby the enzyme Ein a steady
system with diusion limited mass transfer, and with no compound in - and outow.
E+Sk1
−−
−−
k−1
ES k2
−−→ E+P(2.31)
The time-dependent variations of the individual reactants are expressed by the following dierential
equations:
d[S]
dt =−k1[S][E] + k−1[ES](2.32)
d[E]
dt =−k1[S][E]+(k−1+k2)[ES](2.33)
d[ES]
dt =k1[S][E]−(k−1+k2)[ES](2.34)
d[P]
dt =k2[ES] = v(2.35)
The turnover rate vis dened as the product formation. This depends on, and is therefore directly
proportional to the amount of the enzyme-substrate complex ES. Moreover, [ES]depends on the
concentration of the reactants.
By solving Eqs. 2.32-2.35 the time-dependent concentration changes of the reactants can be calculated.
This solution merely describes the batch reaction, where initial concentrations S0and E0are given. For
solution in ow systems, see Section 2.4.3. Figure 2.11 shows the solution of changes of concentrations
done in MATLAB. Three phases can be dierentiated:
1. A short initial (pre-steady) phase, where the [ES]complex is formed and free enzyme [E]de-
creases.The turnover rate is low in this region
2. A medium (steady-state) phase, where the concentration of [ES]is nearly constant. Here the
turnover rate vattains its highest value.
3. Depletion phase, where the substrate [S]becomes exhausted, the [ES]complex decays and
turnover rate vtends to zero
2. CHAPTER. PRINCIPLES 25
Stability
Temperature stability
pH stability
Ingredient/byproduct stability
Solvent stability
Specificity
Substrate range
Substrate Specificity (Κm,kcat/Κm
Substrate regioselectivity and
enantioselectivity
Substrate conversion (%) yield
Producibility/expression yield
Byproduct/ingredient inhibition
Product inhibition
Efficiency
Space-time yield
PH profile
Activity
Turnover frequency (kcat)
Specific activity (kat/kg, U/mg)
Temperature profile
)
Figure 2.12. Construction of a multi-parameter decision matrix for an ecient candidate enzyme selec-
tion. Catalytic reaction rate (kcat), Michaelis-Menten constant (Km), Biocatalytic activity (U) [72]
Since in the steady-state [ES]is nearly constant, the zero order kinetics can be applied therefore
d[ES]
dt = 0 and d[E]
dt = 0. The corresponding rate equations can be simplied therefore the turnover rate
can be written as
v=d[P]
dt =k2[ES] = k2[E0][S]
k−1+k2
k1+ [S](2.36)
In practical cases the individual rate constants are not directly accessible by measurement, therefore
they are converted into kinetic constants:
•Michaelis-Menten constant Km, units mor mm, gives and indication of the anity of the substrate,
low Kmvalues indicating high anities and vica-versa
•Catalytic constant kcat, units s−1, is a measure of the turnover rate of the enzyme i.e. the maximum
number of chemical conversions of substrate molecules per second, kcat =Vmax/[E]0
•Maximum velocity Vmax, units ms−1is the saturation turnover rate of the reaction
•Catalytic eciency, the ratio of kcat
Km, dimension m−1s−1, large values indicate high specicity
Using the kinetic constants, the turnover rate can be written as
v=Vmax[S]
Km+ [S](2.37)
Therefore by knowing the kinetic constants from measurements, the product quantity can be calculated
assuming a given substrate concentration [S]and conversely, measuring the product concentration [P]
with respect to the initial substrate concentration [S0], the kinetic constants can be determined. For the
determination of the constants, several graphical representations are widely used e.g. direct diagram
2. CHAPTER. PRINCIPLES 26
(a.k.a. Michaelis-Menten plot), Eadie-Hofstee diagram, Lineweaver-Burk diagram etc.
Taking enzymatic reactions, the trade-o between activity,stability,specicity and eciency have
to be taken into consideration. (Figure 2.12). This decision matrix reveals the strengths and weaknesses
of every candidate enzyme, so that the most promising candidate enzymes from diverse enzyme lib-
raries can be selected for further process development by re-screening, protein engineering or directed
evolution methods [72]. For instance, immobilization of enzymes may increase their stability but the
activity may be reduced [73]. Specicity of enzymes or enzyme-cascades could improve by genetic
modication (mutation) of the protein, however stability and activity could be aected as well [74].
2.4.3 Enyzmes in packed bed microreactors
The kinetics of enzyme catalysed reactions in reactor columns lled with enzymes linked to insoluble
carriers were rst studied by Lilly and Hornby[75] and their proposed method is referred as Lilly-
Hornby method thereafter. The reactor has an initial volume of Vtand as a given space is occupied by
the carrier, the void volume is Vl(Figure 2.13).
The insoluble enzyme constituting the microreactor chambers may be considered as a suspension of
enzyme in a volume equal to the total volume of the reactor:
[E] = E
Vt
(2.38)
Let us consider an enzymatic catalysis where the substrate ows through the reactor by a ow rate of
˙
Qand spends tresidence time in the reactor. The initial substrate concentration [S0]has been changed
to [St]as [S0]−[St]concentration was transformed to product over the residence time t=Vl/˙
Q(Figure
2.13). Let us assume that the reaction obeys the Michaelis-Menten kinetics, therefore the amount of the
reacted substrate can be obtained by integration the Michaelis-Menten equation:
[S0]−[St] = kcat[E]t−Kmlog([St]/[S0]) (2.39)
The model assumes that a horizontal cross-section of the liquid moves like an imaginary piston in the
column and the ow may be referred to as ’piston ow’. Under these conditions Eq. 2.39 applies to each
innitesimal cross-sectional piston volume and therefore expresses the total reaction taking place in
log(1−P)
*
Ps0
*
s0
Q
st
Q
Vt
VlP = s0−st
s0
P s0= K mlog(1−P∗
) + R/ Q
Figure 2.13. Interpretation of enzyme kinetics in lled reactors and the plot of Lilly-Hornby method to
determine the kinetic constants
2. CHAPTER. PRINCIPLES 27
each such volume during its passage through the column [75].
Let P∗be the fraction of the substrate reacted in the column:
P∗=[S0]−[St]
[S0](2.40)
Using the above notations Eq. 2.39 can be written as
P s0=Kmlog(1 −P∗) + R/Q (2.41)
where R=kcatEVl/Vt. If the values of Pare measured when reacting with dierent initial [S0]con-
centrations of the substrate, then P∗[S0]plotted against log(1 −P∗)will give a straight line if Km,˙
Q
and Rare constants. The slope of the line will be equal to Kmand the intercept on the P∗s0axis will
be equal to R/Q.
2.4.4 Phenylalanine ammonia-lyase (PAL)
PAL enzyme was used as a model biocatalyst in the experiments presented in this work. PAL is a
member of the ammonia-lyase family, which catalyses non-oxidative deamination of it natural sub-
strate (l-phenylalanine , denoted by l-1a hereinafter). In nature phenylalanine ammonia-lyase (PAL;
E.C.4.3.1.24) catalyses the biotransformation of l-1a to (E)-cinnamic acid (denoted by l-2a hereinafter),
a precursor for the lignin and avonoid biosynthetic pathways [76]. The reaction scheme including the
subsequent steps can be described as follows:
[E]+[l−1a]k1
−−
−−
k−1
[E.l−1a]k2
−−
−−
k−2
[E.NH3−l−2a]k3
−−
−−
k−3
[E.NH3]+[l−2a]k4
−−
−−
k−4
[E]+[l−2a]+[NH3]
(2.42)
Here ’.’ denotes the primary, strong bonds while ’-’ denotes weaker, secondary bonds. During the meas-
urements, only the concentration of l-2a was measured directly therefore there was no information
collected regarding the 2nd and 3nd reaction steps.
Figure 2.14 illustrates the ammonia elimination of l-1a and the formation of l-2a by the biocatalyst.
Here, the term biocatalyst is referred as the biofunctionalized particle (MNP), carrying the PAL enzyme.
MNP-PAL
L-phenylalanine trans-cinnamic acid
PcPAL
MNP
L-1a L-2a
Figure 2.14. Schema of ammonia elimination of l-phenylalanine by MNP-PAL biocatalyst
Since its discovery, much knowledge has been gathered with reference to the enzyme’s catabolic
role in micro-organisms and its importance in the phenylpropanoid pathway of plants[77]. PAL has
2. CHAPTER. PRINCIPLES 28
recently been studied for possible therapeutic benets in humans aicted with phenylketonuria[78].
Analogous to how aspartame is synthesized, PAL is also used to synthesize enantiopure unnatural
L-amino acids from various substituted cinnamic acids[76],[79].
2.5 Analysis methods of reaction kinetics
Many methods have been developed to analyse binding methods in macroscale chemistry and the ma-
jority of them was successfully adopted to microuidic devices.
Electrochemical methods such as pH sensing [80], cyclic voltammetry [81], chronoamperometry [82]
were demonstrated in LoC devices. A point-of-care approach was presented using paper-based micro-
uidics [83]. Electrochemical detection was also applied in multiphase ow devices using microdroplets
[84] and magnetic particles [85].
Physical detection methods such as calorimetry [86, 87, 88, 89] were also demonstrated in microuidics,
however, downscaling often limits the achievable accuracy of the device [90].
Besides electrochemical methods, examples of using spectroscopic methods can be frequently found in
the literature.
Spectroscopic methods, especially absorption and uorescence spectroscopy, are widely used for study-
ing biological reactions and were also adopted to microuidics. An exhaustive review in this topic sum-
marizes the recent achievements [91]. Due to its high accuracy and selectivity, uorescence is widely
used [92] even with integrated LED for excitation [93].
Despite its lower accuracy and more challenging integration, UV-VIS absorbance spectroscopy was also
demonstrated in microuidic devices [94, 95].
2.5.1 Analysis of the reaction kinetics by calorimetry
Most chemical and biological processes are accompanied by release or uptake of heat from the envir-
onment. Development of heat is directly related to the reaction process. Because of this, calorimetry
represents a method with a broad potential of applications and has the advantage of directly studying
systems without external inuences or modications, there are also no special requirements for purity.
Calorimetric titrations of the macromolecule (e.g. enzyme) with the ligand (e.g. substrate) yield the
dissociation constant Kdand the binding enthalpy ∆H. The heat Qis determined by the calorimetric
equipment which is linearly dependent on the ligand concentration on the double reciprocal plot:
1
Q=1
Qm
+Kd
Qm[S](2.43)
where Qmis the heat quantity at saturation. [S]eq is the concentration of the ligand at saturation and
Qm= ∆H[A]eq.
Calorimetry method is capable of detecting antibody-antigen, protein-ligand or enzyme-ligand interac-
tions at the bedside for any biomedically relevant case, such as cancer, neurological disorders, diabetes,
metabolic diseases, etc.
2.5.2 Spectroscopic analysis of reaction kinetics
A possible interaction of a photon with a molecule is illustrated in Figure 2.15, inset. The free electrons
of the πand σorbitals of the molecule may accept the energy from photon irradiation and goes form
the bonding-πstate to excited state even to anti-bonding π∗or σ∗orbitals. The energy of the exciting
2. CHAPTER. PRINCIPLES 29
photon is expended to the excitation of the electron and therefore cannot pass towards the medium as
light. This phenomenon can be observed as weakened transmittance at the wavelength of the corres-
ponding electron state. The molecule cannot remain in the excited state, though. Mostly the excited
state becomes deactivated in a radiationless process and the excitation energy is dispersed as phonon
vibrations (i.e. heat) to the medium. In other cases the deactivation causes direct light emitting in lower
energy e.g. in uorescence spectroscopy.
Figure 2.15, inset shows the bonding orbitals of ethene. Ethene contains a simple isolated carbon-carbon
double bond, but the other two have conjugated double bonds. In these cases, there is delocalisation
of the πbonding orbitals over the whole molecule. Excitation happens most probably on the highest
occupied molecular orbital (a.k.a. HOMO), in this case the πorbital, to the lowest unoccupied molecu-
lar orbital (a.k.a. LUMO), in this case the π∗orbital. Energy gaps between the corresponding orbitals
typically fall in the 200 nm to 800 nm range, often referred as the UV-Vis range.
Absorbance spectroscopy setup A typical ow-absorbance spectroscopy setup can be seen in Fig-
ure 2.15. The uid sample to be analysed ows through a ow-cell consisting of channels forming
Z-shape, while the channel runs into two opposing cavities at both ends. Light is attached in one of
the cavities thus light can pass through the uid while the transmitted light is measured at the other
π (bonding)
σ (bonding)
π* (anti-bonding)
σ* (anti-bonding)
n (lone)
Entrance slit
Grating
Array detector
Mirror 1
Mirror 2
Sample inflow
Sample outflow
Z-Flowcell
Light source
Spectrometer
Data acquisition Data analysis
I0
Imeas
Figure 2.15. Basic setup of ow-through UV-VIS spectroscopy with the schematic of a CCD spectro-
meter. Inset: Energy diagram of light excitation in molecules and the bond structure of ethene
2. CHAPTER. PRINCIPLES 30
end of the channel. The excitation light is typically produced by a tungsten-halogen light source. At
the detection side a spectrometer is used. The components are connected by bre light guides.
The typical (so called linear) arrangement of CCD spectrometers can be seen in Figure 2.15. The light
enters through a narrow slit, typically in the range of 25 µmto 100 µm. The width of the slit is a trade-o
between the light intensity (higher light intensity gains higher SNR) and spectral resolution (narrower
slit gains better resolution). The light is directed towards an optical grating which splits the light into
colour components. An other mirror focuses the light onto a linear CCD array. Therefore the spectral
components are projected on the detector. The intensity of the light falls on the corresponding area of
the detector correlates the spectral intensity of the entering light at a given wavelength. The recording
of the spectral intensity data happens at once and the data is transferred towards a computer for further
analysis.
The Lambert-Beer Law The ratio of the excitation light intensity I0and the transmitted light Ias
the light passes through the solution having a molar concentration c, while travels in the solution a
distance of dis dened by the Lambert-Beer law:
I
I0
= e−εdc (2.44)
εis the molar absorption coecient (m−1cm−1). I/I0is usually referred as transmittance. In linear
CCD spectrometers I(λ)/I0(λ)is recorded at every available wavelength at once. Transmittance can
be converted into linear dependence of the solution concentration applying the negative logarithm of
the transmittance
AU =−log (I/I0) = εdc (2.45)
AU is the absorbance unit, the measure of light absorption. Taking the linearity of AU, concentration
can be calculated if a reference value of I0is known. The reference intensity (I=I0) is usually recor-
ded using a sample solution and represents AU = 0.
The other end of the AU scale (AU =∞) is represented by the virtually impermeable solution, mim-
icked by a light shutter inserted in the light path. The calibration process is made of the two above
measurements.
Once the calibration is done, the transmitted intensity Iis recorded and the corresponding AU is cal-
culated continuously.
Following a biotransformation by UV-VIS spectroscopy Biotransformations can be analysed by
UV-VIS absorbance spectroscopy i.e. the concentration decrement of the substrate or the concentration
increment of the product can be measured if the following conditions are met:
1. At least the substrate or the product has a characteristic spectrum in the UV-VIS range
2. The substrate and the product can be dierentiated at least at one wavelength i.e. the measured
AU(λ)value corresponds to one compound only
Figure 2.16 shows the UV absorbance spectra of the substrate and product compounds of the biotrans-
formation of phenylalanine to trans-cinnamic-acid. It can be clearly seen that these compounds full
the conditions above as both of them have dierent and characteristic spectra and for instance at 290nm
the AU(λ)value is characteristic for the tranc-cinnamic acid only.
2. CHAPTER. PRINCIPLES 31
200 220 240 260 280 300 320 340 360
−0.1
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
Wavelength (nm)
Absorbance unit (AU)
trans−cinnamic acid
L − Phenylalanine
NH2
290 nm
Figure 2.16. UV absorbance spectra of substrate (phenylalanine) and product (trans-cinnamic acid) of
a biotransformation by phenylalanine ammonia-lyase enzyme
Concentration values can be calculated if εis known a-priori at the given wavelength or characterized
by taking a dilution series of the compound:
c=A(λ)
ε(λ)d(2.46)
2.6 Magnetic nanoparticles
Nano(bio)technology oers the use and manipulation of the living matter even at the molecular and
cellular levels.
The use of nanoparticles oers major advantages due to their unique size and physiochemical proper-
ties. An important class of nanoparticles whose cores are ferro or superparamagnetic, thus the particles
are moveable by magnetic forces. Because of the widespread applications of magnetic nanoparticles
(MNPs) in biotechnology, biomedical and material science, more and more attention has been paid
to the synthesis techniques of dierent kinds of MNPs. An exhaustive review dealing with synthesis
techniques (e.g. co-precipitation, micro emulsion, thermal decomposition, solvothermal, sonochemical,
microwave assisted, chemical vapour deposition, com- bustion synthesis) can be read elsewhere [96].
One of the most important challenges that will determine the shape, the size distribution, the particle
size, the surface chemistry of the particles, and consequently their magnetic properties. Magnetic nan-
oparticles are used in many research and commercial applications where dierent aspects could be
optimized [97]. In data storage and in magnetic resonance contrast imaging in cancer treatment, the
magnetic properties are of the highest importance. For in vivo applications, the surface chemistry is
emphasized i.e. the magnetic nanoparticles must be encapsulated with a biocompatible polymer during
or after the preparation process to prevent toxication. The nanoparticle coated with polymer will also
allow binding of drugs by entrapment on the particles, adsorption, or covalent attachment. In vitro
application the superparamagnetic composites can be easily prepared with functionality by biofunc-
tionalization. The particles can be separated from the uid phase with ease by using a permanent or
electromagnet.
2. CHAPTER. PRINCIPLES 32
2.6.1 Magnetic properties of MNPs
Five basic types of magnetism can be described: diamagnetism, paramagnetism, ferromagnetism, anti-
ferromagnetism and ferrimagnetisms. Dealing with magnetic nanoparticles, ferromagnetism has of the
greatest importance. For a ferromagnetic material, the magnetic dipoles always exist in the absence and
presence of an external magnetic eld and exhibit long-range order. Macroscopically, such a material
displays a permanent magnetic moment. Ferromagnets are described using the magnetization curve
(Figure 2.17) where magnetization Mis plotted against the magnetic eld strength H[98]. From the
magnetization curve, one can easily nd the saturation magnetization Ms, remanence magnetization
Mrand the coercitivity Hc. The coercitivity Hcequals to the external eld required to reduce the mag-
netization back to zero. This property causes the macroscopic attractive or repulsive phenomena of
hard magnets.
Magnetism is highly volume and temperature dependent because this property arises from the collect-
ive interaction of magnetic dipoles [98]. Figure 2.17 shows the coercitivity Hcin respect to the particle
size r. Based on the particle size, dierent magnetic behaviours can be observed:
Ferromagnetic regime Above the particle diameter of r= 500nm, the material is ferromagnetic
and displays remanence magnetism.
Multi-domain regime Above a critical radius rc, typically r= 50 nm −500 nm several magnetic
domains can develop therefore this regime is referred as the multi-domain regime. Here the material
displays soft ferromagnetic properties with typically low coercitivity and low remanence magnetism.
Single-domain regime Below a critical size (rc<50 nm for the common materials) a particle would
consist of a singular magnetic domain only and the regime is referred as the single-domain regime.
Therefore such a particle would remain in a state of uniform magnetization at any eld [97]. The critical
radius rcfor dierent particles dier based on shape, temperature and crystalline magnetoanisotropy
[96]. An applied external magnetic eld changes the state of this solely domain by rotation. This results
in a large coercivity Hc.
Superparamagnetic regime Experiments proved that below the radius r0(<15 nm) and below
the so called blocking temperature, the coercitivity is zero. Particles fall in this size region are referred
Particle radius r
Coercitivity (Hc)
Super -
Paramagnetic Single domain Multi domain M
H
Mr
Hc
Ms
r0rc
Figure 2.17. Schematic representation of the size dependency of magnetic properties of MNPs; and the
magnetization curve of a ferromagnetic material with the coercitivity value Hc
2. CHAPTER. PRINCIPLES 33
as superparamagnetic particles. For superparamagnetic colloids, the M−Hcurve does not show any
hysteresis.[98].
2.6.2 Sphere packing theory
2r r
Figure 2.18. Illustration of the sphere packing theory: the ratio of the box volume to the spheres volume
independent on the sphere radius as they obey the same scaling law
Highly monodisperse nanoparticles are usually modelled as perfect spheres. When the particles
tightly ll a closed volume (e.g. a chamber) the ratio (µ) of the sum volume of the particles (Vp) and the
chamber volume (Vch) is given by the sphere packing theory as follows:
µ=Vp
Vch
,π
6<µ< π
3√2(2.47)
This ratio (µ) depends on the arrangement of the particles relative to each other and it is independent
on the actual radius of the particles [99]. Therefore, halving the radius of the spheres inside of a box
will not aect µbut the number of the spheres will be eight times (23) more. In general, when a reaction
chamber lled up by tightly packed particles, the total mass of them should remain the same even if
the particle sizes are dierent case by case. Of course, the area to volume ratio and the total number of
particles will change, what makes sense when surface reaction may occur. This is the typical case of
immobilized enzymes onto nanoparticles.
However, the above theoretical limit could be overcome by using unequal-(e.g. binary)-sized spheres
([100]) as the smaller particles can t in the voids between the larger particles.
Chapter 3
State-of-the-art
The following sections summarize the state-of-the-art of the three groups of scientic challenges on
which the results of this dissertation were worked out. The work presented in the rest of the dissertation
intends to answer these challenges.
3.1 Thermal design of microreactors
Thermal design is a crucial step in the design process of various electrical [101] and electromechanical
microsystems [102]. Several approaches have been developed for reducing computational capacity re-
quirements and calculation time to obtain appropriate results for thermal analysis such as successive
node reduction (SUNRED) [103] or compact modelling with lumped elements [42].
Although thermal design and modelling is a key factor for such biological microsystems as polymerase
chain reaction (PCR) devices [104] or chip sized (nano-)calorimeters [105], no unied thermal design
methodology has been developed yet. Thermal design considerations of biochips can be divided into
the following cases:
1. The (bio-)reaction is strongly temperature dependent or the reaction is triggered by temperature
changes. A typical example is the polymerase chain reaction (PCR) where the sample is cyclically
heated and cooled to three target temperatures while a dened part of the sample DNS is mul-
tiplied exponentially [106]. Recent developments in PCR technology have focussed on reducing
the sample volume while increasing the throughput by utilizing droplet reactors [107, 108, 109].
Here, thermal design focuses on the precision temperature control of the device. In this case, the
ultimate goal of thermal design is to achieve the shortest possible temperature transition inside
the microreactors [106].
2. The biological reaction itself generates heat and the enthalpy change due to the reaction should
be measured. The miniaturized implementations of calorimeters (a.k.a. micro and nanocalorimet-
ers) are intended to detect nano Joule heat changes whilst using submicrolitre sample amount
only [110, 90, 88, 89]. Thermal modeling is a coupled domain problem here as both the thermo-
dynamics of the enzyme reaction and the thermal behaviour of the device should be taken into
consideration [110, 111]. In this case, the ultimate goal of thermal design is to obtain the minimal
heat loss towards the ambient. Smaller heat loss tends to increase the signal to noise ratio and
detection limit.
Further classication can be done based on the microreactor type:
34
3. CHAPTER. STATE-OF-THE-ART 35
void write_liquid()
{
item_write=fluid_read;
… }
{
fluid_write=item_write;
…}
void thermal_cycle()
{…}
};
a)
b)
c)
d)
e)
f)
Axial distance
Nusselt number
Figure 3.1. Dierent approaches of thermal modelling in microsystems: a) CFD simulation of segmen-
ted ow to determine local temperature and b) Nusselt number [44], c) Thermal compact model for
a single phase microcooler system and d) its single cell lumped R-C thermal model [42], e) System-C
representation of a PCR biochip [35] f) Static FEM thermal analysis of a microcalorimeter device [110]
1. In stationary reactors the reaction occurs in a designated volume of the chip. If the uid stands
still during the reaction, no uid dynamics should be taken into consideration and the problem
can be reduced to the thermal domain only [112, 105]. If enzyme kinetics is also modelled, the
problem becomes multi-domain.
2. If the uid ows continuously through the reactor, the eect of convective heat loss should be
also taken into consideration. In this case the problem becomes multi-domain and both the uid
dynamics and heat transfer should be calculated.
3. In droplet reactors the discrete reaction volumes move relative to the microchannel and therefore
no local thermal equilibria will be developed. In such cases both the movement of the uid (uid
dynamics domain) and the transient thermal behaviour of the device (thermal domain) should
be taken into consideration i.e. a coupled problem is formed [113]. The number of the coupled
domains could be further increased if the time dependent heat generation of the bioreaction is
also modelled.
Dierent approaches of thermal modelling are summarized in Figure 3.1. Static FEM thermal ana-
lysis of a microcalorimeter device was performed [110] to approximate the expected temperature of the
heat sensing element. The complex thermal analysis of segmented ow in a microchannel was done
[44] using a CFD simulation tool. Local temperature and heat transfer coecients (Nusselt number)
were determined (Figure 3.1,a,b). A major drawback of time dependent CFD simulations is the expens-
ive computational requirements (50-100 hours of simulation time is usual). Computational time can
be dramatically reduced by using thermal compact models. Shridhar et al.[42] demonstrated a lumped
model consisting of resistances and capacitances. Convective heat transfer due to uid ow is taken
into account by using controlled current sources (Figure 3.1,d). The discretized geometry of a micro-
cooler system was analysed using an instance of such a compact model in each nite element (Figure
3.1,c) and 1000 fold speed-up was achieved compared to the detailed CFX simulation. Thermal com-
3. CHAPTER. STATE-OF-THE-ART 36
pact models can be also integrated with behavioural system descriptions (Figure 3.1,e). Zhang et al. [35]
demonstrated a schematic representation of the system level description of a PCR biochip.
Challenges
A reasonably fast thermal model has not been implemented for droplet microreactor devices. Heat
transfer models and experimental investigations are traditionally based on the determination of the
Nusselt number. This method is obviously practical when the heat transfer properties of the individual
droplets can be neglected and an overall description of the ow is feasible. This is widely used e.g. for
modelling uids in electronics cooling applications such as integrated circuit cooling.
In contrast, for bio-analytical investigations the thermal interactions should be analysed in detail as
they happen inside or aect directly the individual droplets, each accommodating an individual bio-
chemical reaction. A major challenge of this type of investigation is the high computational time. Typ-
ical simulation run-times of the droplet-ow problem vary in wide scale from roughly 1.5 days to about
60 days [44, 43]. That is, there is a strong need to reduce the computational time e.g. by using reduced
models. A thermal compact model for integrated circuits with cooling microchannels was presented
[114], which consists of thermal resistances and controlled voltage sources. This model can predict the
temperature distribution of the channel wall but it is not suitable for the analysis of individual droplets.
Conclusion Recent approaches in thermal modelling of biochips have been demonstrated but no
general solution has been presented so far for eective design assistance of droplet based LoC devices.
Emulsion PCR or ow-through nanocalorimeters demand accurate thermal design. The modelling
needs of these devices could be fullled by a novel thermal compact model, which
•takes into account the conjugate heat transfer eects which develop in micro ow systems at
very low Reynolds numbers
•describes the heat transfer inside the microchannel containing one or more liquid slugs separated
by an insoluble media such as gas bubbles,
•provides direct input for a subsequent transient analysis and handles transient chemical (e.g.
enzyme) reactions taking place inside the droplets and results in a temperature eld as output
•has an interface to attach to the representative thermal solid model of the device package
•has a common behavioural level representation like the one used in electrical subsystems (e.g.
VHDL-AMS or an electronic circuit as an analogous system representation)
•results in execution time in the order of magnitude of minutes allowing reasonable iteration times
in the LoC design ow.
The above needs are addressed by Objective 1 (Section 1.1), the realization of the model is presented in
Chapter 4.
3.2 Nanoparticle counting
Functional nanomaterials e.g. nanoparticles have emerged as a new generation of building blocks bey-
ond conventional chemicals and are being incorporated in a variety of exciting opportunities in elec-
tronics, photonics, (bio-)catalysis and medicine. Despite these advances in the past several decades,
3. CHAPTER. STATE-OF-THE-ART 37
Resonance coil
magnetometer
Magnetoresistive
sensor
c
de
Nanoparticle
Ensemble methods
f
Resonanant cantilever
Single particle counting
Total weight
Single particle
weight
DLS
UV-Vis
Turbidimetry
Light
TEM
Light scattering
Microscope
Figure 3.2. Summary of the methods to measure nanoparticle concentrations: a,b,c: ensemble methods;
d,e,f: single particle counting. Based on [115]
some basic issues still remain to be tackled. One of these challenges is how to determine the number
or molar concentration of nanoparticles accurately [115]. The signicance of this simple question is
obvious. For instance, in biocatalysis, the transformation occurs on the MNP’s surface therefore for
quantitative reaction analysis at least the concentration of the particles should be known.
Two approaches can be distinguished: in ensemble methods a statistically large amount of nanoparticles
is contained in the measurement system and their amount is measured at once. The size or weight of a
single particle must be known a-priori for concentration calculations.
In contrast, single particle counting methods interact with a single particle at once and the total amount
of the carrier uid must be known as a prerequisite.
3.2.1 Ensemble methods
Gravimetric measurements are based on the weighting the ensembled particles (mtotal) by an ana-
lytical balance or alternatively by quartz crystal microbalance (QCM). In this approach the particle
is assumed as an articial molecule and based on its composition the weight of the single articial
molecule (m∗
particle) can be approximated [115]. Then the concentration is given by
c=mtotal
m∗
particleNAV(3.1)
where NAis the Avogadro number and Vis the total volume of the suspension.
Methods based on light absorption Nanoparticles made of certain materials may absorb light at
specic wavelengths. Based on the Lambert-Beer law (see detailed 2.5.2), the amplitude of the unique
excitation UV-Vis spectra at a specic wavelength is proportional to the particle concentration. The
molar extinction coecient εof gold nanoparticles has been estimated by both theoretical calculations
and experimental measurements [115] and a correlation was found between the dparticle size and ε
as follows:
log ε= 3.22 log d+ 10.8(3.2)
3. CHAPTER. STATE-OF-THE-ART 38
Turbidimetry measures the decreasing of the intensity of the incident light (τ) caused by light scattering
of nanoparticle suspensions. Turbidity is propotional to the concentration of the particles c
τ= 0.25πKd2c(3.3)
where Kis the scattering coecient, to be determined experimentally [115].
Dynamic light scattering, short DLS, measures the intensity of the light scattered by the suspended nan-
oparticles. As nanoparticles undergo Brownian motion in solution, the intensity of the scattered light
demonstrates a time-dependent uctuation pattern. The uctuations are related to the mobility and dif-
fusion of the particles which is particle size dependent. Therefore, DLS is usually used to determine the
size distribution of the particles. In monodisperse solutions, the light intensity is also proportional to
the number of particles and through the Rayleigh scattering theory the concentration can be determined
[115].
Reaction vessel
Sensing coil
MNP quantitiy
Resonance frequency
Sensing coil
a) b)
Figure 3.3. a) The concept of magneto immunosensor. Paramagnetic particles are attached to the de-
tecting antibodies while the the same kind of antibodies are immobilized to the bottom of the reaction
vessel. The frequency shift of the resonant coil magnetometer placed below the vessel is proportional
to the particle count of the sandwich-assay [116] b) Detection principle of the resonant coil magneto-
meter. The inductivity of the resonant coil is changed due to the magnetic particles interact its electric
eld and the resonance frequency of the circuit is shifted
Resonant coil magnetometer Magnetite (F e3O4) or maghemite (F e2O3) nanoparticles exhibit su-
perparamagnetic or soft ferromagnetic behaviour with saturation magnetization values ranging from
20 A m2kg−1to 60 A m2kg−1resulted in high permeability values [117]. This advantageous magnetic
behaviour led to the idea of inductive measurement of the amount of nanoparticles. In general, the
inductance (L) of a conductor coil depends on the relative permeability of the core material. Therefore
the inductance depends on the MNP density of the given volume (e.g. a sample chamber), which is
surrounded by the coil. The determination of the resonance frequency shift of a serial RLC oscillator
is a straightforward method of measuring inductance changes. The change in the resonance frequency
with respect to the inductance change is given by
dω
dL =−1
2√CL3/2(3.4)
where Cis the capacitance of the capacitor in the RLC circuit (note that in practical cases, the capa-
citance of the inductor coil is negligible compared to the external capacitor). The resulted RLC circuit
is a resonant coil magnetometer (RCM), which was reported earlier by others [118, 119]. The measured
3. CHAPTER. STATE-OF-THE-ART 39
frequency shift ∆fis a measure of the total particle quantity.
The weight of the single particle has to be known a-priori
mtotal =∆f
φ(3.5)
where φis the frequency sensitivity factor, which has to be determined experimentally by characteriz-
ation.
The ability of measuring MNP quantity by magnetometers arised the concept of magneto-immunosensors.
In such assays magnetic particles are attached to analytes (Figure 3.3). One of the rst approach was
reported by Kriz et al. [118]. The quantity of ferromagnetic particles suspended in a 5 mL cuvette
was determined by measuring the inductance of a coil of wire when the sample was placed in it. A
sandwich assay structure utilizing biofunctionalized paramagnetic microparticles was developed and
reported by Richardson et. al [119]. The particle content of the test strip was quantied by a resonant
inductor coil in which the strip was placed. The resonant inductor coil technique was further reported
in a magneto-immunosensor for the sensitive detection of intracellular proteins by Sharif and Luxton
[120]. The sample was stored in a cuvette and the quantity of the particles was determined by a at
inductor coil placed below it. Eveness et al. [120] studied the sensitivity limits of the resonant coil mag-
netosensor using paramagnetic particles from dierent vendors in dierent sizes. Doubtless, the main
advantage of the resonant coil technique is that the method is non-destructive, its accuracy is relative
high and the cost of the equipment is signicantly cheaper than other magnetometer techniques e.g.
SQUID (superconducting quantum interference device) [121].
3.2.2 Single particle counting
Flow cytometry uses optical or impedimetric detection of ow-through particles in a narrow chan-
nel. Optical technique often employs a photodetector to measure the ash of the scattered light (Figure
3.2,d) [115]. By counting the number of ashed light pulses, the number of particles can be quanti-
ed. In contrast, impedance changes induced by the passing through particles could be measured by
electrode pairs [122].
Magnetoresistive sensor Such sensors can be integrated within microuidic channels to detect
magnetically labelled cells. Like in ow cytometry, the particles ow through a sensor area. Mag-
netoresistive sensors are composed of a non-magnetic metal (spacer layer) between two layers of fer-
romagnetic metals, one of which (the pinned layer) has its magnetization xed by an adjacent antifer-
romagnetic layer, while the other (the free layer) is free to rotate (Figure 3.4,a). The relative magnetic
orientation of the ferromagnetic layers are changed to be antiparralel. [123]. Small variations of the ex-
ternal magnetic eld (e.g. the passing-through of a magnetized particle) will change the angle between
the free layer and the pinned layer therefore changing the sensor resistance. With this conguration
the resistance changes linearly with respect to the external eld. The detector is sensitive enough to
detect single particles [124].
Micromechanical resonator Cantilever type microstructures enable the measurement of mass with
zeptogram sensitivity[125]. Viscous damping of the cantilever can be eliminated by placing the uid
solution inside the hollow resonator that is surrounded by vacuum. A suspended microchannel trans-
lates mass changes into changes in resonance frequency (Figure 3.4,b). Particles ow through the can-
tilever having an eective mass of m∗and spring constant of k, and the observed signal depends on
3. CHAPTER. STATE-OF-THE-ART 40
Resistance [Ω]
Magnetoresistance %
1
Resonance frequen cy [Hz]
2
3
Time
Magnetic field [Oe]
a) b)
Figure 3.4. The concepts of a) the magnetoresistive cytometer: the resistance of the sensor depends on
the disturbance in the magnetic eld caused by the magnetized particle passing through the cell [116]
b) micromechanical resonator: the resonance frequency of the resonator depends on the mass of the
particle owing through the hollow [125]
the position of particles along the channel. The exact mass excess of a particle ∆mcan be quantied
by the peak frequency shift ∆finduced [125] as follows:
∆f=1
2πrk
m∗+α∆m(3.6)
where αis a characterization constant.
Challenges
For quantitative analysis of bioreactions, at least the concentration of the biocatalyst should be known.
In case of such biocatalysts that are immobilized onto the surface of (nano)particles, the direct de-
termination of the catalyst concentration is not feasible. Therefore, nanoparticle quantication has an
important rule in such systems. During the immobilization process, the biocatalyst is added in a given
mass ratio to the particle suspension. The ecacy of the immobilization procedure is usually qualied
by the Bradford method [126]. Finally, the concentration of the biocatalyst can be determined if the
total weight or total count of the particles taking part in the reaction is known.
State-of-the art methods are summarized in Table 3.1. Some methods (Gravimetry, RCM) provide a
direct measure of the total particle weight, other methods (UV-Vis, Turbidimetry, DLS) give the con-
centration. Single particle methods can provide the particle count in a given volume and then the
concentration can be deduced.
•If the total weight is known, the total amount of the biocatalyst can be calculated directly by
taking the mass ratios.
•If the particle concentration is known, the biocatalyst concentration can be calculated based on
the mass ratio of the macromolecule and the approximated weight of a single particle.
•If the particle count is known, the particle concentration can be deduced with ease.
If the biocatalyst amount should be known, Gravimetry and RCM methods produce the most accurate
results. In contrast, for other methods the single particle weight should be taken into account. In prac-
tical cases, the weight of a single particle can be approximated only.
3. CHAPTER. STATE-OF-THE-ART 41
Table 3.1. Summary of nanoparticle counting methods
Method Particle
type
Particle
size range Prerequisites Microu.
integration
Conc.
accuracy
Quant.
accuracy
Technol.
complexity
Ensemble methods
Gravimetric Any Any Composition - + ++ +
UV-Vis Specic UV spectra Any Extinction coe. + ++ + ++
Turbidimetry Non-absorbant,
concentration limit Any Scattering coe. + ++ + ++
DLS Non-absorbant at
laser WL Submicron Calibration - ++ ++ ++
RCM Magnetic particles Any Freq. sensitivity +++ + ++ +
Single particle counting methods
Light scattering Strong scattering 24 nm for
gold particles Calibration +++ + +++ ++
Magnetoresistive Magnetic particles Micrometer Calibration ++ + +++ ++
Micro cantilever Any 10 fg Calibration + + +++ +++
Using ensemble methods, the particle quantity can be measured right in the reaction chamber (in-
situ), which enables even on-line measurements. If the measurement principle does not aect the re-
action, the particle quantity can be monitored during the reaction as well.
Magnetoresistive and resonant coil magnetometer methods are highly selective in terms of the particle
core material. Therefore, any change in the surrounding medium (e.g. reactants ow in the chamber)
will not aect the measurement accuracy signicantly. Optical methods, however, may become inac-
curate if any of the reaction compounds inuence the investigated spectra.
Microuidic integration is possible in most cases. The gravimetry method can not be integrated though.
Laboratory balances are not exible enough to measure the weight of the microuidic chip and the
surplus mass of the nanoparticle at once. DLS is a complex equipment itself and microuidic integra-
tion is not an alternative. Although the micro cantilever structure consists of a microchannel hollow,
its integration to a more complex Lab-on-a-Chip system could be challenging. The magnetoresistive
technique also requires complex technology but the microuidic integration has already been demon-
strated [124, 123]. Chip sized implementation of optical cytometry was demonstrated by many authors
[127, 128, 129, 130].
Conclusion The RCM method provides a direct measure of the total particle weight and also al-
lows on-line monitoring as the reaction compounds do not aect the measurement. Magnetic nano-
particle quantication based on resonant coil magnetometry was demonstrated earlier in magneto-
immunosensor applications [120, 131, 116]. In the demonstrated devices, the resonant coil was placed
under the macroscopic reaction vessel. Although the integration of a at coil below a microuidic
chamber is feasible, no such attempt had been published before.
The above needs are addressed by Objective 2 (Section 1.1), the realization of the integrated RCM device
and subsequent measurements are presented in Chapter 6.
3.3 Microreactors
Analytical systems which comprise microreactors are characterized by outstanding repeatability and
reproducibility, due to replacing batch iterative steps and discrete sample treatment by ow injection
3. CHAPTER. STATE-OF-THE-ART 42
systems [10]. Immobilization of enzymes on beads, particles or on monoliths has been used for separ-
ation and recycling of enzymes. Such technique is also utilized in microreactors [136]. The possibility
of performing similar analyses in parallel is an attractive feature for screening and routine [9] use. Mi-
croreactors have been integrated into automated analytical systems, and as well as providing benets
from system automation this also eliminates errors associated with manual protocols [10].
The importance of magnetic nanoparticles (MNPs) as potential carriers of biomolecules is growing rap-
idly in biotechnology and biomedicine. In LoC systems nano-sized magnetic particles provide quasi-
homogeneous systems, high dispersion, high reactivity, low diusion limits and the possibility of mag-
netic separation. The MNPs are usually collected in micro-sized reaction chambers. The collection is
compiled by magnetic separation from the uid stream. Such microreactors were found to be highly
eective in biodetection [137], biocatalytic [138] and bioanalytic [S4] applications.
The following immobilized enzymes are used on an industrial scale: glucose isomerase, sucrose mutase,
β-galactosidase, penicillin acylase, d-amino acid oxidase, glutaryl amidase, thermolysin, nitrilase, amino
acylase and hydantoinases [139].
Applications of microreactors can be divided into three classes [10]:
•Organic synthesis, when a target molecule is formed from components in ow
•Analytical usage for biocatalysts in order to transform an analyte dicult to measure to an easy
to measure form
•Analytical usage for screening of substrates, enzymes and examine their kinetic characteristics
0
0.2
0.4
0.6
0.8
1
1.2
0 5 10 15 20 25 3
0
Days
Relative activity
Immobilized GOD
GOD in solution
Relative activity
0
0.2
0.4
0.6
0.8
1
1.2
1 2 3 4 5 6 7 8 9 10
Runs
Immobilized Asparaginase
0.5 mm
40 mm
60 mm
a) b) c)
d) e)
Immobilized Asparaginase
Figure 3.5. Lab-on-a-Chip microreactors a) as monolith silica reactor [132] b) packed bed silica reactor
[133] c) packed bed MNP [134]. Stability of the immobilized enzymes e) after re-use of asparaginase in
10 cycles, 10 minutes each [135] f) long term stability of immobilized GOD [132]
3. CHAPTER. STATE-OF-THE-ART 43
Classication of microreactors
The majority of the commercial ow synthesis systems utilize laminar ow with soluble components
and enzymes [140]. Losing the catalyst is a major drawback of this technique. In contrast, microreactors
utilizing immobilized enzymes have many advantages over the traditional ow reactors:
•the biocatalyst (enzyme) can be recycled after usage therefore the process is more economical
and environmentally friendly. It was reported that immobilized asparaginase enzyme retained
the 95.7% of its activity after 10 cycles of usage [135] (Figure 3.5,d) while immobilized glucose
oxidase enzyme (GOD) retained 97% of its original activity after cyclic regeneration and re-usage
[132].
•analytical steps are highly reproducible as the biocatalyst concentration is xed in the system
•immobilization often extends the long-term stability and temperature resistance of the enzymes,
in several cases even the catalytic activity is increased compared to the soluble form. Immobilized
asparginase retained the 72.6% of its original activity for 10 weeks, immobilized GOD retained
the 95% of its activity [132] (Figure 3.5,e) for 30 days.
One of the major drawback of the immobilization technique, that enzymes often subjected to a decre-
ment in biocatalytic activity and choosing the appropriate immobilization technique is challenging
[141],
Immobilized reactors can be further divided into two groups according to the type of the supporting
material of the biocatalyst:
•The reactor is dened as a monolith reactor where the supporting material is xed in the re-
actor volume and the biocatalyst is non-removable therefore the chip is single use. Examples
include silica monolith reactors [132] (Figure 3.5,a). The silica-monoliths were prepared from
two precursors, using a sol–gel method. The resultant liquid mixture containing the precursors
was loaded into the monolith channel and allowed to polymerize[132]. Most recently nanobrous
reactors made by electrospinning [142, 143]. The PDMS channel structure was glued on the top
of the nanobrous layer, which was made separately [143].
•The reactor is dened as apacked bead reactor where the supporting materials are micro- or nan-
obeads and the biocatalyst is immobilized onto their surfaces. The reactor can be loaded with the
suspension of the beads (Figure 3.5,b) and viscous [133] or magnetic [144, 145, 146] forces are
utilized to keep the particles xed in the reaction chamber.
Magnetic manipulation of MNPs
Magnetic manipulation techniques of magnetic beads provides various methods for precise handling of
biocatalysts inside the chips [144]. Do et al. [145] utilized a magnetic bead separator array. The magnetic
eld is concentrated between permalloy patterns (50 ×100) to produce a high magnetic eld gradient
over the edges of them, thus being able to trap the magnetic beads. Li et al. [146] used an external hard
magnet to develop a concentrated magnetic eld perpendicular to the channel at a certain position of
the chip. The particles accumulated at the designated place. Slovakova et al. [147] used, however, a pair
of hard neodymium magnets positioned in a given angle to develop a magnetic eld parallel to the
channel structure. It was reported that in this case the particles were arranged parallel with the chan-
nel axis and also the reaction eciency was reasonably higher than in orthogonal congurations. Lien
3. CHAPTER. STATE-OF-THE-ART 44
et al.[134] used an integrated electromagnet with active cooling for the entrapment of the magnetic
particles in the reaction chamber (Figure 3.5,c).
Challenges
Table 3.2. Lab-on-a-Chip microreactors with immobilized enzymes
Reference Method Reusability Stability Particle [E] measurement
Mu (2014)[135] Michaelis-Menten, LB plot 10 cycles 100 min 10 weeks MNP Out of chip
Asparginase Km,Vmax 95.70% 72.60% 12 nm Supernatant
CLE-CE out of chip
He (2010)[132] Michaelis-Menten, EH plot 97% 30 days Monolith reactor Out of chip
GOD Km,kcat 95% silica Absorbance
Amperometry, on-chip
Kerby (2006)[148] Lilly-Hornby N/A N/A Silica microbead In chip
Alkaline phospatase km,kcat Optical
Fluorescent imaging
Slovakova (2005) [147] Michaelis-Menten, LB plot 80% N/A MNP Approximated
Trypsin Km,kcat 600 nm
Seong (2003) [133] Lilly-Hornby N/A N/A Microspheres Optical
Km15 um
Fluorescent imaging
Kinetic studies could be carried out with ease in microreactors by changing the attributes of the re-
action e.g. the inow substrate concentration. As the generic properties of the Michaelis-Menten model
cannot be applied to ow reactors, the Lilly-Hornby model was used in several cases [148, 133]. Even
if the kinetic parameters were interpreted traditionally, at least the ow rate dependency of the kinetic
parameters was reported [132, 147, 148, 133].
Kinetic parameters Kmand kcat were determined in every investigated experiment while the on-chip
detection of the product quantity was realized in various ways such as capillary electrophoresis (CE)
[135], amperometry [132], and uorescent imaging [147, 133].
Though the accurate determination of the enzyme concentration [E]is crucial for the calculation of
kcat, no direct method has been reported so far. In most cases the enzyme concentration was deduced
by measuring its value in the supernatant of the waste. In-chip determination of the quantity of relat-
ively large (15 µm) microbeads was obtained by counting the particles on the microscopic image of the
reaction cell [133].
Conclusion Although various applications of microreactors have already been demonstrated, the
following attributes may aect the reactor performance and therefore need consideration.
•Taking advantage of the reusability of enzymes, multi -parameter experiments can be carried out
without changing the biocatalyst in the chambers. In such arrangement the repeatability of the
measurements should be investigated rst. Substrate concentration, substrate ow dependency,
even substrate material can be changed periodically, gaining multiple characteristics with ease.
3. CHAPTER. STATE-OF-THE-ART 45
•The traditional approach of Michaelis-Menten kinetics is not applicable for packed bed microre-
actors therefore obtaining kinetic parameters from measurement data should be carried out by
involving the eect of ow. This issue is addressed by Objective 5 (Section 1.1). Chapter 5 describes
the actual microreactor device which is capable of carrying out multi parameter investigation of
the enzyme system, Chapter 7 presents the subsequent measurements and results.
•Kinetic data cannot be obtained without knowing the exact amount of biocatalysts in the reac-
tion chamber. Determination of [E]is not obvious, though. One possible method is the in-situ
measurement of the quantity of the entrapped particles in the chamber. A possible method is
presented in Chapter 6.
•Upon loading the chamber with particles or even forming a monolith layer, the physical structure
of the carrier is a subject of change due to time, viscous forces of the uid ow, air bubbles etc.
Particularly in case of magnetic trapping techniques an amount of the particles can be drifted
away from the chamber. Such changes of the carrier structure may have an eect on the biocata-
lytic activity. This issue was addressed by Objective 3 (Section 1.1). A possible implementation of
the chamber observation algorithm and conclusions based on actual measurements are presented
in Chapter 7.
•Biocatalytic activity may depend on the particle type therefore the following factors should be
considered
–A chamber loaded with smaller particles exhibits larger bioactive surface due to larger sur-
face to volume ratio. As the biocatalysts occur on the surface of the particles, particle size
may also aect the achievable biocatalytic activity.
–Using binary sized particles (i.e. a mixture of bigger and smaller particles) even more of
them can be loaded in the chamber as the smaller particles may t in the voids between the
larger ones. Having more biocatalysts inside the chamber, the biocatalytic activity may also
be changed. The latter two issues were addressed by Objective 4 (Section 1.1). Actual meas-
urements in enzymatic microreactors with dierent particle sizes and consequent results
are presented in Chapter 7.
Chapter 4
Thermal compact model for droplet
microreactors
Abbreviations and Symbols
Symbol Units
AArea m2
CHeat capacity J K−1
cVVolumetric specic heat capacity J m−3
dnDistance between nodes m
δFFluid lm thickness m
fsFrequency Hz
hEnthalpy-density J m−3
∆HrReaction enthalpy J mol−1
kThermal conductivity W m−1K−1
n Normal vector (to surface)
QHeat J
˙
QHeat current W
˙qVolumetric heat generation W m−3
˙q00 Heat ux W m−2
RRadius of the gas droplet m
Re Reynolds number
Rth Thermal resistance K W−1
ρDensity kg m−3
TTemperature K
tTime s
∆titer Iteration time s
u Vector of velocity m s−1
uavg Average vector of velocity m s−1
uavg Average scalar of velocity magnitude m s−1
VVolume m3
γSurface tension N m−1
µDynamic viscosity mPa s
τCharacteristic time s
46
4. CHAPTER. THERMAL COMPACT MODEL FOR DROPLET MICROREACTORS 47
4.1 Introduction
Current design methodologies for microuidics-based biochips are typically full-custom and bottom-up
in nature. Detailed device level simulations are used extensively to design and optimize the component
and device, and help to create custom compact models for them. Once the devices are optimized using
detailed physical simulation, they can be used to assemble a complete microuidics-based biochip. Only
at this stage, the system-level simulations and optimizations can be carried out [33]. Since the system
behaviour can only be veried at this late stage, costly and time-consuming redesign eort is required
if the system does not satisfy design constraints. That is, a strong need of developing more generalized
compact models for microuidic systems [39], ready to use even at the earlier design stages.
This chapter introduces a reduced order thermal model for droplet microreactors. Being the model
compatible with the standard SPICE description, it can be embedded into the top-down design ow of
a microsystem (Figure 2.2). The model enables the thermal analysis of microchannels consisting of con-
tinuously moving microdroplets with biologically active content inside (see detailed in Section 2.2.2).
Thermal design is a key design aspect of micro sized calorimeters (see also Section 2.5.1). Reaction heat
measurement based biodetection seems to be one of the promising concepts in such micro scale (chip
size) laboratory devices [149]. Heat measurements constitute a direct means of determining enthalpy
changes, which appear in all cases of protein- protein or protein-ligand interactions. Although calor-
imetry is traditionally a low throughput method, the micro-scale realization enables higher reaction
speeds and multiple measurements, therefore higher throughput can be achieved [150].
Droplet size, reaction kinetics, droplet velocity, material of the channel etc. aect the heat transfer
from the droplet towards the ambient. These properties should be considered together with the signal
processing solutions (e.g. on- chip integrated analog or digital circuits) at behavioural level. Thermal
compact models are widely used in design practice, where such reduced order models (in contrast to
detailed numerical models) yield results quickly in the early design phase.
In the current model, convective heat transfer of the moving droplets is modelled by the switched ca-
pacitor approach. The model results the temperature prole of the channel in minutes compared to
conventional numerical techniques that requires days or weeks.
4.2 Modelling methods
The compact model was compared to actual detailed CFD simulations as a reliable benchmark. For this
purpose the work of Gupta et al [44] was reproduced 1. Their CFD simulation results were validated by
actual experiments [61]. For the present work the same multi-physical CFD tool was used that Gupta
et al have used. Therefore the reproduced detailed CFD model was considered as a reliable benchmark.
1Work of A. Drozdy, TDK Scientic contest, BME, 2013
Table 4.1. Summary of material properties used in simulations
Fluid ρ µ k c γ
kg m−3mPa s W m−1K−1J kg−1K−1N m−1
Water at 25oC997.0 1.002 0.6 4182 0.072
Nitrogen at 25oC1.145 0.018 0.0242 1040
PDMS 965 -0.15 1460 -
4. CHAPTER. THERMAL COMPACT MODEL FOR DROPLET MICROREACTORS 48
Using this detailed model the simulation parameters (uid ow velocity, type of uids) were changed
and the capability of modelling chemical reactions was also added to our baseline detailed model. With
this CFD model the compact model was checked in two relevant cases: for constant heat ux boundary
condition and for internal heat generation by an enzyme reaction.
The baseline model was implemented in ANSYS Multiphysics and its solver was used to calculate the
problem. Once this model was compared to the detailed CFD model and a good match was achieved,
the model was translated into a SPICE compatible analogous electrical network (AEN ) description. The
network can be calculated directly with an electrical network solver.
4.2.1 Governing equations
The modelling approach applied is based on the suggestion of Sridhar et al [42]. Here the most essential
parts of their work are summarized; further details can be found in their paper. The energy conservation
equation for heat transfer in a nite control volume of liquid Vcan be written as
d
dtZV
hdV−ZA
k∇T·ndA+ZA
hu ·ndA=ZV
˙qdV+ZA
˙q00dA(4.1)
where n is the normal vector of Asurface.
Taking the limit of V→0and applying the Stokes theorem we get
cV
dT
dt−∇·(k∇T) + cVu · ∇T= ˙q+∇ · ˙q00 (4.2)
It is deduced by Sridhar et al. that the third, convection term can be calculated by using temperature
controlled heat sources. By applying the nite dierence approximation we get a compact model built up
of thermal conductances,thermal capacitances and heat generators. Thermal capacitance Cth and thermal
resistances Rth are calculated from the material properties of the given volume using an axisymmetrical
transformation. The heat generator element ˙q00dA is utilized for constant heat generation; the controlled
heat generator element ˙q(t)represents the time dependent heat generation of the biochemical reaction
(Figure 4.1.). The term ˙
Qconv =cVuavg,y∆Ay(Ts2−Ts1)(4.3)
for convection was introduced by Sridhar et al., where uavg,y is the uid mean velocity in mainstream
direction while Ts2−Ts1is the temperature dierence between the opposite downstream surfaces of the
cell. The last term of this equation is the control potential of the temperature controlled heat source
while the rst terms are assumed to be constant. Therefore the controlled source can be written as
−cconv(TA−TB), where TAand TBare the geometrical projections of the surface temperatures to the
centres of the cells.
4.2.2 Switched capacitor approach
Let us assume a moving uid which is owing through a stationary channel. Take two cross sections of
the channel perpendicular to the channel axis. As these two surfaces may have dierent temperatures,
heat may pass through the surfaces while heat is stored in the enclosed volume. As it was deduced in Eq.
4.3., the heat passes through the volume is proportional to the temperature dierence of the opposite
surfaces and it can be represented by temperature controlled heat uxes. As long as the uid properties
do not change along the channel (i.e. heat ux depends only on the temperature dierence) the system
can be modelled by a static network of controlled sources,thermal resistances and capacitances.
4. CHAPTER. THERMAL COMPACT MODEL FOR DROPLET MICROREACTORS 49
Figure 4.1. Representative R-C model of the unit cell
Figure 4.2. Switched capacitor approach of convective heat transfer
Assuming segmented ow, the material properties alternate along the channel as the phase changes.
Nevertheless the material properties are also alternating along the channel wall. Assuming a stationary
channel wall the model should be changed dynamically too as the slug and bubble sections follow each
other. The stationary domain can be divided into jsub-domains (rows) and every row can be divided
into further ncells. Σdenotes the network (Figure 4.2) which represents the stationary domain of the
channels and the droplets. The network contains ninstances of the unit-cell (Figure 4.1) in every row.
The initial temperatures of the rst cells (where the uid enters into the channel) are T0, therefore
a proportional amount of heat ∆qis accumulated in every rst capacitor. Let us assume that as the
droplet is moving, this amount of heat ∆qis transferred from the nth cell of the stationary domain
towards to the (n+ 1)th cell having a heat capacitance of Cand initial temperatures of Tnand Tn+1,
respectively. The sum heat transferred is
∆Q=Qn+1 −Qn=C(Tn+1 −Tn)(4.4)
Let us assume a capacitor which is switched between two nodes (n+ 1 and n) with a frequency of fs.
This results in a heat current qequ as ∆qheat is transferred in each switching cycle.
˙
Qequ =fs∆Q=fsC(Tn+1 −Tn)(4.5)
It follows that an equivalent thermal resistance can be dened as the ratio of the temperature dierence
and the heat current.
On the contrary, in the novel model the capacitors are switched towards to the next downstream node
4. CHAPTER. THERMAL COMPACT MODEL FOR DROPLET MICROREACTORS 50
(1..n)in every iteration (Figure 4.2). This approach represents a temperature controlled current source.
Assuming a downstream mean uid velocity of uavg and a cell geometrical size of d, the switching
frequency of the cells is fs=uavg 1
dwhich yields an equivalent heat current ˙
Qequ of
˙
Qequ =uavg
dC(Tn+1 −Tn) = cVuavg∆A(Tn+1 −Tn)(4.6)
In this case the equivalent heat current through a given ∆Asurface can be formulated on the basis of
the one phase case as follows:
˙
Qequ =˙
Qconv =Cuavg (Tn+1 −Tn)(4.7)
which gives formally the same equation which describes the temperature controlled heat current source
as it was shown in Eq. 4.3.
Once the transient solution is calculated, the amount of heat stored in each capacitance is determined.
In the next iteration the capacitors are switched towards the next node and a transient solution is
calculated again. The iterative stepping of the capacitances therefore represent the heat transfer in the
channel.
4.2.3 Simplications
Constant bubble shape Once the segmented ow developed, the slug shape, the slug-slug distances
and the velocity prole remain stable because the boundary conditions are periodicals, and the system
has shift invariance. Therefore we assume that the phase boundaries and initial slug positions are
known a priori and are not calculated throughout the simulation which leads to a linear heat conduction
problem considering only a single time step.
Figure 4.3. Channel cross section with two gas droplets for the investigation of the internal heat transfer
Negligible internal convection Wall friction establishes mass circulation inside the slugs which
can be easily visualized as caterpillar tracks (Figure 4.3). At higher ow rates two kinds of heat transfer
need to be concerned inside the liquid slugs:
1. Conductive heat transfer from the channel wall to the channel axis ( ˙qdiff )
2. Convective heat transfer along the channel wall by laminar ow and along a curved path down
to the channel axis by the internal circulation ( ˙qconv)
First the characteristic times of the two processes will be investigated. Next, the temperature dierences
of the upstream and downstream sections (both of the axis and the wall) will be derived. It will be shown
that the characteristic time of the conductive term is independent of the ow velocity but that of the
4. CHAPTER. THERMAL COMPACT MODEL FOR DROPLET MICROREACTORS 51
convective heat transfer is not. Finally, a critical ow velocity can be dened. Below this velocity the
convective term can be neglected and in this regime the proposed compact model is valid.
The conductive term is investigated at rst. A constant heat ux along the channel wall is assumed
which causes heat diusion from the channel wall to the axis ( ˙qdiff ).
∂T (r, t)
∂t =k
cV
1
r
∂
∂r T(r, t)r∂T (r, t)
∂r (4.8)
By solving the dierential equation for nearly constant temperature walls we get a simple dierential
equation for the time part, and a Bessel’s dierential equation for the radial part, R(r):
T(r, t) = R(r)e−2.4052kt
cVR2=R(r)e−t
τcond (4.9)
where τcond is the characteristic time of the ˙qdiff conduction heat transfer
τcond =cVR2
2.4052k,(4.10)
which does not depend on the ow velocity. In contrast, the transfer times of the convective transferred
heat in the liquid slug along the wall (ttr,w) and from the wall along a curved path to the channel axis
(ttr,c) can be determined as follows:
ttr,c =Rπ
uavg
, ttr,w =lslug
uavg
,(4.11)
where Ris the radius of the curved velocity eld, lslug is the length of the liquid slug, uavg is the mean
axial velocity of the uid.
In the next step the temperature changes along the wall and axis will be determined, respectively.
∆TW,up can be derived from the heat transfer in fully developed laminar ow. For the calculation of
∆TW,down, the tiny part of uid between the curve of the gas bubble and the wall is investigated. The
calculation is based on the ratio of the energy inow through the wall AWand the heat capacity of the
uid.
∆TW,up =3 ˙q00R
4k,∆TW,down =Aδ˙q00∆t
VδcV
,(4.12)
where Aδis the wall surface, and Vδis the uid volume around the wall. The ratio of the temperature
changes can be calculated hereinafter based on the conductive heat loss over the given transfer time.
For the up and downstream axial temperatures we get:
∆TA,up = ∆TW,downe−ttr,w +ttr,c
τcond + ∆TW,upe−ttr,c
τcond (4.13)
∆TA,down = ∆TW,up(1 −e−ttr,v +ttr,w
τcond )(4.14)
To determine the validity range of the compact model, equations 4.12 and 4.13-4.14 are combined and
solved numerically for ∆TA,up = ∆TA,down
∆TA,up −∆TA,down = 0 at v=vlimit (4.15)
If the ow velocity is above vlimit, which results in TA,up > TA,down because ttr,w < τcond, therefore the
convective heat transfer overrides the conductive one. On the contrary, if the ow velocity is under
vlimit, it follows that TA,up < TA,down. As the internal mass circulation (which causes convective mass
transfer) is not handled in this model, the compact model is only valid under vlimit.
4. CHAPTER. THERMAL COMPACT MODEL FOR DROPLET MICROREACTORS 52
Fluid inflow Fluid inflow
Heat flux boundary
Heat flux boundary
a) b)
Figure 4.4. Symmetry models used in the simulations: a) 2-D planar model with uniform thickness b)
axisymmetric model
Symmetry models Two symmetry conditions were used in the simulations. The 2-D planar model
(Figure 4.4a) represents the virtual centreline cross section of the channel with an uniform material and
temperature distribution on the perpendicular plane . Heat ux boundary condition was set on the top
and bottom surfaces while the sidewalls are treated to be adiabatic. This approach is feasible when the
channel is embedded into a thick PDMS body.
In case of axisymmetric modelling (Figure 4.4b), the boundary conditions, the loads and the results
are obeying an axial symmetry. The axis of symmetry is in the center line of the channel (Figure 4.6.)
Because of this axial symmetry, two-dimensional elements are used so 4 thermal resistances are con-
nected to a node. Material properties and results are interpolated to three dimensions. This approach
is typical for circular channels, and as a reliable benchmark, this was used for the comparison of the
model with the results of the detailed CFD simulations in FLUENT.
4.2.4 Setting up the model solver
A single element – consisting of four nodes – represents a nite heat capacitance (Figure 4.6.) and four
thermal resistances in each direction - referred as two-dimensional therma solid. The working principle
of the dynamic model is shown by the owchart presented in Figure 4.5. First, the two-dimensional
geometry was dened. Channel geometry, droplet size and length were input parameters. Material
and uid parameters such as thermal conductivity, heat capacitance, number of droplets, uid velocity
were also dened here. Any type of boundary conditions could be applied like constant wall heat ux,
constant wall temperature,time dependent volumetric heat generation (e.g. enzyme reaction) inside the
droplets. A stationary computational domain was used which includes a part of the channel and a
given number of droplets. In order to preserve the energy balance the number of droplets was kept
constant during the simulation (see Figure 4.6, the volumetric sum of the droplets was always constant).
The model parameters of each element (cell) and the cross-element connections were represented in a
netlist le. The above steps were repeated until the desired simulation time was reached. Local values
of the channel wall temperature and uid mean temperature were available in every iteration step.
The discretization of the space was done manually by quadratic non-uniform meshing. Meshing divides
the geometry into non-uniform elements with volumes of Vcell,n. The radial distribution of the nodes
corresponded to the reverse Gaussian distribution, i.e. the density of the nodes was higher at the wall-
liquid interface region than in the middle of the channel. This non-uniformity of the mesh reduced the
total node number and therefore decreases the simulation time.
Material parameters were assigned to each element. Gas and liquid phase, channel wall and liquid lm
layers were treated as dierent materials. A rst, steady-state run was performed with all body force
4. CHAPTER. THERMAL COMPACT MODEL FOR DROPLET MICROREACTORS 53
loads deactivated and the nodal temperature solution was saved. The mass transfer was modelled by a
moving mesh. The mesh of the uid area was shifted by one node in each iteration step in the direction
of uid ow (Figure 4.6.). The nodes had a consecutive numbering from left to right and from bottom
to top. Therefore the value of the nodes being shifted was formulated through a modulo function
shifteff = mod(iterationn,cellnum) (4.16)
Where iterationnis the number of the actual iteration while cellnum is the number of nodes in axial
direction. For the ow simulation periodic boundary condition was applied in the downstream direction
while for the thermal simulation Dirichlet condition was applied at the inlet and Neumann condition
was applied at the outlet. This later was calculated as follows:
˙qout =
A·
n
P
i=1
cnode ·Tnode
∆t(4.17)
Once the model had been built and meshed and the boundary conditions were set, the solution could be
done by a SPICE solver or the linear transient solver of any FEM based program. The results presented
in this chapter were solved by the linear transient thermal solver of ANSYS 14.
The solution was obtained by the sparse matrix direct solver using in-core memory mode. The sparse
solver found a non-negative pivot range between 0.67 and 1.87 ·10−6which provided convergence in
every iteration step. The solver found the solution in 5-6substeps in each iteration.
The nodal temperature results of the last run was loaded again as an initial condition before the follow-
ing step was being solved. Body force loads were applied and a transient simulation was performed.
Iteration time depended on uid velocity as
∆titer =dn,ax
uavg
(4.18)
Figure 4.5. Flowchart of the ANSYS implementation
4. CHAPTER. THERMAL COMPACT MODEL FOR DROPLET MICROREACTORS 54
Figure 4.6. Moving mesh technique and applied boundary conditions
where dn,ax is the axial distance of the nodes.
A liquid lm layer is observed between the slug and the wall. The thickness of the lm was calculated
on a basis of the correlation proposed by Aussillous and Quere:
δF
R0
=1.34Ca2/3
1+2.5(1.34Ca2/3),(4.19)
where Ca is the capillarity number. A mesh resolution of at least δF/2was set within the lm area.
The lm layer was therefore modelled by a conductive liquid layer.
4.2.5 Building the AEN model
Parsing
Model File
Fluid
element
Heat gen.
element
Temperature
element
Terminating
element
Building
geometry
Calculate
Rth and C th
values
Generating
Model File
Building
AEN
Runnig
network solver
Results
File
Model Library
ANSYS
Parameter file
xle
yle
w
ARth,N
Rth,S
Rth,E
Rth,W
Single Finite Element
Figure 4.7. Method of obtaining results with a commercial electrical network solver; Geometry was
dened in ANSYS and a model le was generated. The parser transformed the model le into the
analogous electrical network (AEN). The network was calculated by the ELDO solver and the results
were saved in each iterations
In order to enhance the compatibility of the model with the existing tools of the EDA design toolkit,
the nite element model was converted into analogous electrical network (AEN) model. The network
model could be directly solved by a network model solver ELDO (Mentor Graphics Inc, USA). As ELDO
has an application programming interface (API) which allows the control of the solver even in runtime,
4. CHAPTER. THERMAL COMPACT MODEL FOR DROPLET MICROREACTORS 55
it was a preferable choice. Based on the parameter settings (Figure 4.7, Parameter le) the model geo-
metry was automatically generated by ANSYS. While a nite volume is represented by an element with
four nodes (Figure 4.7, Single Finite Element) in ANSYS, it is equivalent to a network of ve nodes con-
sisting of four thermal resistances and a heat capacitance. The corresponding values were calculated
as follows:
Cth =Awρc ˙q=1
Aw
1
i
4
X
i
˙qi(4.20)
Rth,S =Rth,N =yle
k2xlewRth,W =Rth,E =xle
k2ylew(4.21)
Here w, xle, yle are the geometry sizes of a unit cell (see Figure 4.7), Ais the area of the upstream face of
the unit cell, k,ρ,care the thermal conductivity, density and specic heat of the unit cell, respectively.
The calculated values were stored in the Model le, which was transformed into a SPICE compatible
network description (AEN) (Figure 4.7, Parsing). The AEN consisted of four subcircuit elements (Figure
4.7, Model Library), such as
•Fluid element: represents wall, liquid or gas without heat generation
•Heat generator element: represents wall, liquid or gas with volumetric heat generation. Heat
current is analogous to electrical current.
•Temperature element: represents wall or uid inlet with a x temperature. Temperature is ana-
logous to voltage.
•Terminating element: represents the external geometry boundaries as a high thermal impedance
towards to the ambient
A framework program was constructed in C for controlling the network solver, using the standard
C API of ELDO. Once the network description has been built up the network solver was executed.
The voltage results of the previous iteration were set up as initial conditions to each node. In the next
iteration the capacitors were switched to the next upstream node and the network was solved again.
The nodal voltage results were summarized in each iteration in a Results le (Figure 4.7, Results le).
Taking the electrical-thermal analogy, temperatures are represented by the taken voltages.
4.2.6 CFD simulation settings
Table 4.2. Summary of the simulation parameters
Test case Channel length Radial mesh Axial mesh Droplet length Re
mm µmµmµm
CFD CHF 1.5 1.85 −2.7 2.7 300 1
CTMF EM CHF 5.7 0.5−20 60 300 1
CTMAEN CHF 5.7 0.5−20 60 300 1
CFD ENZ 2.5 2.72 −5.4 5.4 530 2.88
CTM ENZ 2.5 0.5−20 38 530 2.88
4. CHAPTER. THERMAL COMPACT MODEL FOR DROPLET MICROREACTORS 56
Four simulations were performed by the ANSYS FLUENT (CFD) tool and by our novel compact
thermal model (CTM) with constant heat ux (CHF) boundary conditions (Figure 4.8a). As an applica-
tion example the case of heat generation by enzyme reaction (ENZ) was also demonstrated (Figure 4.8b).
For one case (Re = 1), the CTM was solved both by the linear transient solver of ANSYS (CTMF EM )
and after parsing, by an industrial electrical network solver, ELDO (CTMAEN ). Further six simulations
were done for the model validity range analysis in CHF case. The simulation parameters are summar-
ized in Table 4.2.
The CFD simulation was built according to the experimentally validated recommendations of Gupta
et. al. [44] and was used as a benchmark for comparing the results with the compact model. The axial
mesh distribution was always equidistant while the radial distribution was varied from a rare to a
more dense mesh from the center line of the channel towards the walls. In FLUENT the coupled pres-
sure based solver was used with Courant number: Co = 0.25. Discretization was done by Green-
Gauss node based and body force weighted methods. Phase-phase interactions were concerned by the
volume-of-uid (VoF) approach both for the CHF and ENZ cases. In the CHF case a constant heat ux
of ˙q00 = 100W/m2was set on the channel wall. The channel wall was considered hydrophilic therefore
the gas phase form spherical bubbles (Figure 4.8a).
4.3 Results
4.3.1 Performance analysis
For the performance test the linear solver of ANSYS 14 was run on an Intel Core i5 4 core CPU with 4
GB RAM under 64 bit Debian GNU/Linux (kernel 3.2.0-2-amd64). The solver has a parallel computing
capability but parallelism did not improve the solution speed as the steps must be processed sequen-
tially. In order to achieve reasonable results an axial resolution of 80 µm was applied for a channel
Figure 4.8. Boundary conditions for the simulation a) for the constant heat ux (CHF) case b) for the
volumetric heat generation due to enzyme reaction (ENZ) case
4. CHAPTER. THERMAL COMPACT MODEL FOR DROPLET MICROREACTORS 57
diameter of 320 µm which yields a sum of 2736 nodes (radial mesh resolution ranges from 0.5µm to
20 µm. For the channel length of 6mm a sum of 75 iterations were needed to achieve the thermal
steady state. Meshing the model required 4.96 s(the time in each iteration step the nodes are set up)
while the model solver ran 13.29 sin each iteration. The solver/meshing execution time ratio was thus
2.68 which grew rapidly for more complex problems, for a problem size of 21888 nodes the ratio be-
came 3.94. Therefore the compact model resulted in a total run-time of 20 minutes. In contrast, the
CFD simulation for the same problem (see later as CFD comparison case) took about 150 hours on a 3
GHz i5 2 core CPU with 4 GB RAM. Thus, thee compact model runs about 450 times faster than the
traditional FEM model.
4.3.2 Model validity range analysis
Figure 4.9. Comparison of the temperature contours of the channel axial cross section resulted in FLU-
ENT (above) and by the current model (below) in CHF case at Re = 1. The contour lines are interpolated
between the nodes in both cases
It was suggested in Eq. 4.15 that above vlimit the convective heat transfer will be dominant and
therefore the compact model could not be valid any longer. Assuming the parameters of Table 4.2.
for CHF case and for the droplet length of 300µm we get vlimit = 1.75 cm/s, (Re = 6.35). Total
temperature rise along the channel and axial temperature proles in the slug sections were investigated
at dierent Reynolds numbers (ranging from Re = 1 to Re = 7). The results are summarized in Table
4.3. We found that the slope of the temperature rise along the channel ts very well in FLUENT and by
solving the compact model in every case under the critical velocity.
The accuracy of the compact model was investigated at dierent ow rates. The axial temperature
distribution was calculated in the slug sections both in FLUENT and by solving the compact model. The
resulting functions were compared and the variance (σ2) of the two thermal proles were calculated as
follows:
σ2=1
NX
i∈slug
(∆TFLT(xi)−∆TCTM(xi))2,(4.22)
where Nis the number of nodes to sum. At Re = 1 the variance between the thermal prole calculated
by FLUENT and by the compact model was only 1.02 ·10−5K2. This case is also shown as a contour
plot in Figure 4.9. which also suggests a good match. Between Re = 3 and Re = 7 the comparison
resulted in a manageable error with a higher variance of an order of magnitude. The model accuracy
was strongly decreased above Re = 7 as the theoretical calculations suggested. While the slope error
increased to around 30%, the variance of the thermal proles increased with 1.5order of magnitude.
In the particular channel geometry under investigation in this chapter, the validity range of the compact
model was found under Re = 7.
4. CHAPTER. THERMAL COMPACT MODEL FOR DROPLET MICROREACTORS 58
4.3.3 CFD comparison for constant heat ux boundary
00.2 0.4 0.6 0.8 11.2
0
0.1
0.2
0.3
0.4
0.5
Current model
FLUENT
Wall temperature (shifted by 0.2 K)
Fluid temperature
Figure 4.10. CFD comparison of axial temperature distribution in segmented ow (FLUENT and the
current model). On the phase map light=gas phase, dark=liquid phase
For the comparison of the model with detailed CFD simulations, a two phase, axisymmetric, tran-
sient simulation was performed in ANSYS FLUENT 14 under constant heat ux boundary condition
(Figure 4.8.). Figure 4.10 compares the axial distribution of the wall and uid center-line temperatures
calculated by FLUENT and calculated by our present model. Note, that the wall temperature is shif-
ted up by 0.2 K for the better visibility. Temperatures calculated by the two models match fairly well
(variances of the temperature proles are 1.02 ·10−5K2and 1.24 ·10−5K2for the uid and for the
wall, respectively). Gas temperatures ran together (humps) while some discrepancy can be seen in the
slug region. The uid temperature calculated by compact model is slightly higher at the upstream re-
gion and is lower at the downstream region than the values provided by the CFD model since inner
circulation transfers heat towards the wall in the upstream region which leads to a higher temperature
while the opposite occurs downstream. This kind of temperature dierence however, was negligible
Table 4.3. Model accuracy (FLUENT vs. CTM and ANSYS vs. ELDO) for dierent ow rates under
constant wall heat ux
Re Slope, FLU Slope, CTM Slope error σ2of thermal prole
K mm−1K mm−1K
1 0.13 0.137 5.11% 1.02 ·10−5
3 0.038 0.0401 5.24% 9.22 ·10−5
5 0.0214 0.0228 6.14% 1.23 ·10−4
7 0.017 0.0163 4.29% 1.30 ·10−4
12 0.0092 0.0128 28.1% 4.07 ·10−3
Re Slope, ANS Slope, ELD Slope error σ2of thermal prole
K mm−1K mm−1K
1 0.13 0.138 2.64% 1.77 ·10−5
4. CHAPTER. THERMAL COMPACT MODEL FOR DROPLET MICROREACTORS 59
this particular case; the inner circulation should be considered only at higher ow rates. The same was
found for the wall temperatures. Abrupt changes of the temperature in the gas region – calculated by
FLUENT – was not shown by our compact model.
4.3.4 Comparing FEM and AEN solutions
0 0.5 11.5 22.5 3 3.5 4
0
0.02
0.04
0.06
0.08
0.1
0.12
ANSYS solver
Eldo solver
Figure 4.11. CFD comparison of centreline axial temperature distribution in segmented ow (the same
model represented by FEM and solved by ANSYS (red), and represented by AEN and solved by ELDO
(blue)
For the CHF case of Re = 1, the model was solved both by the linear transient solver of ANSYS
and, after parsing to AEN, by the electrical network solver ELDO. Despite of the analogous model de-
scription, a minor dierence in the results could be still observed. Figure 4.11 depicts the temperature
prole in the two cases. The further analysis revealed that the slope error was 2.64% and the variance
of the temperature proles was 1.77 ·10−5K2(Table 4.3). The deviation stems in the slightly dierent
discretisation of the geometry. In the network model, the center node of the unit cell served as temper-
ature probe. In contrast, in the nite element model, the temperatures were determined on the corner
nodes.
4.4 Application example
As an application example of the proposed compact model, an enzyme reaction was studied taking
place inside the moving droplet microreactors. Due to the enzyme reaction, heat was released which
resulted in temperature increment inside the droplets.
4.4.1 Modelling the enzyme reaction
It was assumed that the reaction occurs inside the droplets can be handled by the following system
of dierential equations. Of course for more detailed analysis more complex models (including e.g.
inhibition) can be also implemented. Let us assume that the reaction follows the Michaelis-Menten
4. CHAPTER. THERMAL COMPACT MODEL FOR DROPLET MICROREACTORS 60
kinetics as it was explained in Section 2.4.2 The dierential equation system of the Michaelis-Menten
model can be reduced using the following considerations at t= 0:
[P] = 0,[ES] = 0,[E] = [E0],[S]=[S0](4.23)
The dierential equation for [P]for a particular solution:
−
¨
[P]
k2−˙
[P] =
=−k1 [S0]−[E0]−[P] + k2[S0]−˙
[P]
k2!×(4.24)
× k2[S0]−˙
[P]
k2!+k−1[S0]−k−1
k2[S0]−˙
[P]
k2
,
The equation was solved using an analytical approximation. The solution then yields the time derivative
of product concentration P(t). This quantity multiplied by the molar reaction enthalpy gives the heat
generation during the reaction.
Eq. 4.24. is a non-linear second-order dierential equation. The solution was done by Wolfram Research
Mathematica with NDSolve command using the predictor-corrector "Adams" method. The precision
could be up to 5digits with 10000 maximal steps. For demonstration, the carbonic anhydrase enzyme
was chosen as it is considered a fast and eective catalyst.
The reaction enthalpy can be calculated from the standard state enthalpy of water, carbon dioxide and
carbonic acid. The reaction-rate constant k3is equal to kcat (which can be measured).
The Michaelis-Menten constant of carbonic anhydrase is
2.6·10−2, the reaction enthalpy is ∆Hr= 20.1kJ/mol [151], k1and k2are arbitrary chosen to 1.5·107
and 1500, respectively. In the compact model case the time dependent product concentration [P]of
the enzyme reaction was calculated by solving the Eq. 4.24. applying the aforementioned boundary
conditions. Once the P(t)concentration is known, the volumetric heat generation ˙qn(t)(Figure 4.1)
for the nth cell can be calculated as follows:
˙qn(t) = ∆Hr·˙
P(t)∆titer∆Vcell,n (4.25)
This equation was solved assuming water droplets as reactors with volume of 45 nl and length of
530 µmeach. The solution revealed that the substrate was consumed in about 0.3 s and 2010 J l−1heat
was generated in each droplets. The supercial ow velocities in the consequent ow simulations were
arbitrary chosen in order to provide the thermal transients to be completed whilst a single droplet has
been own through the channel. Further details of the results can be seen in [S1].
For comparison, the same reaction was also modelled in ANSYS FLUENT. The reaction occurs in the
whole volume of the water phase. The channel wall is the widely used PDMS (polidimethylsyloxane).
The initial concentrations of the solution which enters into the channel were
E0= 2 µmol/l, S0= 0.1mol/l (4.26)
for the enzyme, and the substrate concentrations. Both the initial enzyme-substrate complex and the
product concentrations are zeros. FLUENT utilized its own solver to handle the reaction kinetics. Con-
cerning these quantities the relaxation factor was set to 0.9instead of the default 1as the latter caused
divergence during the solution.
4. CHAPTER. THERMAL COMPACT MODEL FOR DROPLET MICROREACTORS 61
00.2 0.4 0.6 0.8 11.2 1.4 1.6 1.8 2
0.1
0.2
0.3
0.4
0.5
0.6
Current model
FLUENT
Figure 4.12. CFD comparison of axial temperature distribution of moving liquid slugs with an en-
zyme reaction inside (FLUENT model and the compact model) On the phase map above red=gas phase,
blue=liquid phase
4.4.2 Temperature prole analysis
The total heat generated inside the droplet reactors was calculated in each iteration step and was ap-
plied to the droplet volumes. The enzyme reaction started when the droplet entered the channel. The
supercial ow velocity was 0.9cm/s, assuming Re = 2.88 Reynold’s number in the channel. So the
substrate uid had already been consumed and the reaction had been ended by the time the rst droplet
reached the end of the channel. As there was no heat generation inside the N2bubbles, the liquid slugs
were thermally separated. The axial thermal prole at the wall-uid interface is shown in Figure 4.12.
The enzyme reaction was still active in the rst and in the second droplet, resulting in an increased
temperature at the wall. In the third droplet the reaction was already ended however the accumulated
heat still kept the third droplet warmer than the former ones.
The compact model quantitatively followed the temperature prole calculated by FLUENT, however
the mean error was larger than in the constant heat ux case (variance of the two thermal proles is
1.08 ·10−3K2). The dierence was caused by two factors. On one hand the components of the enzyme
reaction had an innite diusion constant which caused an inhomogeneous heat generation inside of
the droplet. This also resulted in an inhomogeneous temperature eld in FLUENT, which of course
– in average – equalled to the one calculated by the compact model. In contrast, the compact model
assumed homogeneous concentrations. On the other hand the enzyme reaction itself was described
by a simplied model too, which assumed homogeneous concentration as well. As this was not the
case, especially not in the rst droplet where the distribution of the enzymes was uneven and therefore
resulted in a larger temperature error between results provided by the CFD simulation by the compact
model.
4.5 Conclusion
Based on the general properties of the segmented ow, a simplied model was introduced to describe
heat transfer in microchannels. Thermal analysis was performed by a novel compact model and the res-
4. CHAPTER. THERMAL COMPACT MODEL FOR DROPLET MICROREACTORS 62
ults were compared with detailed CFD simulations done by ANSYS FLUENT. The presented compact
model provided a 450 fold reduction in execution with little compromise for accuracy. Relative errors
of the temperature proles were calculated in order to compare the results of the two methods. The
relative errors of the temperature proles were 5.33% and 5.3% for the uid and for the wall, respect-
ively. It was also demonstrated, that by utilizing a straightforward parsing method, the model could be
also solved by a standard SPICE like network solver. This feature comprises the possible integration of
this compact model into the design toolchain of integrated microsystems.
The model was also capable of handling enzymatic reactions that occur inside of the droplet microreact-
ors. The same problem was solved by FLUENT and the result comparison yielded a reasonable match.
The calculated relative error was found to be 13.8%.
Considering the reduced simulation times (in the order of magnitude of ten minutes) the compact
model can be eectively used in the complex Lab-on-a-Chip design ow. The model could be used in
any cases where the co-design of the electric and uid parts is needed. The iterative optimization of
the enzyme reactions and the ow properties could be accelerated with this novel technique. With an
appropriate interface the compact model presented in this chapter can also be coupled to the thermal
model of the LoC package.
4.6 Summary of scientic results
Thesis 1 Thermal compact model for droplet microreactors
I established a novel thermal compact model for two phase ow microchannels. The model provides the time
and space dependent temperature eld of the microchannel. During the calculation the heat generated on
the channel wall with a constant ux or inside the droplets and the velocity of the owing medium are
considered. The heat generated in the microreactor may be a consequence of a time dependent chemical
reaction (e.g. enzymatic catalysis). [S1],[C1],[C2]
1. A cell unit of the model is built up of heat conducting elements. Convective heat transfer is
modelled by the switched capacitor’s theorem. The amount of the heat stored in the capacitors is
conducted through thermal conducting elements while the rest of the heat is switched towards
the next capacitor at the rate of the uid ow. The switching frequency depends on the ow
velocity. The elements are aligned on the rectangular mesh of the channel geometry. I showed
that the model can be realized either by nite 2D axisymmetric heat conduction elements aligned
to the mesh or by an equivalent electrical network aligned to the mesh nodes. I showed that the
two realizations are practically equivalent (the dierence was smaller than 2.65% in case of the
investigated benchmark problems). The equivalent electrical network representation can be also
calculated by industrial standard network solvers. I created a parser algorithm for the conversion
between the two realizations.
2. I determined the range of Reynolds numbers where the model is valid. I provided an analytical
relationship to calculate the upper limit of the Reynolds number. I found that the current model
is valid under Re = 7. The results provided by the compact model were compared to the results
obtained by detailed CFD simulations of the identical problem. It was found that the relative
error of the calculated temperature proles remained below 7% under Re = 7 and increased
rapidly above that. The computational time required for the calculation of the compact model
was reduced by two orders of magnitude compared to the detailed CFD model.
Chapter 5
Equipment
5.1 Microuidic testbench
Enzymatic biotransformations and magnetic nanoparticle quantication experiments were carried out
using an in-house developed microuidic testbench. The testbench consists of three simultaneously
working subsystems:
•uFLU Studio Framework is a software toolset designed a) to plan and to carry out measurements,
b) to store and to analyse measurement data. The framework has access to every part of the
microuidic testbench system and operates its subunits.
•Controller is a collection of several hardware components. These components are connected to
the System Bus, or connected to the Serial Bus (USB), or connected to the local Ethernet.
•Fluid Control Unit is a uidic platform consisting of the microuidic device under test, sample
holder, valves, microscope camera and the MNP quantier. O-chip sensors are also connected
to the unit.
5.1.1 uFLU Studio Framework
The uFLU Studio Framework is the software interface of the microuidics testbench. The entire frame-
work was implemented in National Instruments LabVIEW and the units are currently compiled to
Windows PCs to Apple OS X platforms. The main units and their functions are:
Measurement planner unit enables to design a measurement sequence which is later evaluated
by the Controller, which operates the Fluid Control Unit (FCU). The microuidic chip under test is
accommodated in the FCU.
•Measurement sequences are built of operation steps, which contain commands controlling the
following units:
–Syringe pump controller, whereas the dispension ow rate can be set in the range of 1.13 µl min−1
to 7 ml min−1.
–Additionally a dened dilution ratio can be set therefore the sample is diluted to a specied
concentration using two syringe pumps and a micromixer element. The ow rates are ad-
justed automatically according to the desired concentration.
63
5. CHAPTER. EQUIPMENT 64
–Sample holder unit which contains at most 8sample tubes. By appropriate switching of the
valves any sample can be selected and loaded into the syringe automatically.
–Valve control unit Valves can be operated as request by manual programming
•Three stepping modes are available for controlling the sequence:
–Time stepping evaluates the next step after the desired time has been passed
–Pressure controlled stepping jumps the next step if the pressure drop of the chip reaches a
System Bus
Syringe pump
controller
Thermostat
controller
Valve
controller
Data
acquisition
Serial Bus
Ethernet
Optical
Inspection Unit
LCR
Meter Unit
Laboratory
Data Storage
Sample
Identification
Measurement
controller
Measurement
planner
Measurement
analyzer
Spectra
analyzer
Measurement
reporter
uFLU Studio Framework
Spectrometer
Unit
Pressure
sensor
Flow rate
sensor
Bubble
detector
Microfluidic
device under test
Sample
Holder
Valve
Control
Unit
Camera MNP
quantifier
Fluid Control Unit
Controller
Figure 5.1. Functional diagram of the microuidic testbench system
5. CHAPTER. EQUIPMENT 65
dened range
–Absorbance controlled stepping jumps the next step if the product concentration of the en-
zyme reaction (measured by the absorbance at a given wavelength) reaches a dened range.
–For the last two stepping modes a time-out value can be also dened
•Sample names, initial concentrations, dened syringe sizes can be recorded as a supplementary
information of the measurement sequence which are displayed and available during the evalu-
ation of the sequence.
•Measurement planner enables to create a schematic representation of the uidic connections of
the measurement
•Sample Identication is a subsystem which identies samples by QR code and stores relevant
sample data in a cloud database. This database is available in the Measurement planner software
tool therefore the measurement can be planned by taking the stock information into account.
Measurement controller unit has two operating modes.
•In direct control mode the actuators of the Fluid Control Unit can be operated directly from the
PC or by a remote controller.
•In sequence mode a measurement sequence is loaded and evaluated. Actual Flow rates and sample
volumes are displayed which provides an on-line overview of the system state during the meas-
urements. The system also allows email reporting for longer term measurements.
The data of the sensors connected to the Fluid Control Unit are acquired and available on-line during
the measurement and also stored for further analysis.
Measurement analyser unit enables the post-processing of measurement data (actuator settings,
event reports, sensor data) and has two sub-units
•Spectra analyser is a software tool for post-processing of spectral data acquired from the linear
CCD spectrometer. The tool has the following options:
–Displaying the spectral response of the chip output at a given time
–Displaying the time dependent response of the chip output at a given wavelength
–Baseline and oset compensation of the measured spectra
–Integration and averaging of the spectral response
–Calculation of the time dependent product concentration [P]and turnover Vof the reaction
•Measurement reporter is a tool which analyses the data of the sensors collected during the meas-
urements, and on request plots and summarizes these data into measurement report les which
are contains all relevant information of the measurement. One of these data is the photo image of
the reaction cells which is stored automatically and later compared to a reference image to obtain
quality check information about the measurement. These measurement reports are automatically
migrated to the Laboratory Data Storage.
5. CHAPTER. EQUIPMENT 66
5.1.2 Controller
The main structure of the Controller consists of three buses, namely the System Bus, the Serial Bus and
Ethernet. Dierent hardware devices are connected to the Controller and each of them are available in
the uFlu Studio Framework.
•System Bus is an in-house developed platform which accommodates Card Modules. The system
can be easily upgraded with new functions by installing new Card Modules on request.
–Syringe pump controller is able to operate two syringe pumps
–Thermostat controller operates a circulating thermostat (Julabo Inc. type F-25) whose target
temperature can be set and the actual temperature of the bath can be acquired on request.
The heat exchanger is installed right below the microuidic chip.
–Valve controller is able to operate 4 bistable or 4 monostable solenoid valves and four pro-
portional solenoid valves
•Serial Bus uses the Universal Serial Bus (USB) 2.0 protocol and operates the following devices:
–Spectrometer unit, a linear CCD spectrometer for the UV-VIS range (Ocean Optics USB
2000+, slit=25 µm). Light excitation is provided by a Tungsten-Halogen light source (DT-
mini-2-GS, Ocean Optics). Flow-through absorbance measurement occurs in a Z-ow cell
(Micro-Flow-Z-cell 10, Avantes Inc., d= 10 mm).
–LCR Meter Unit is a network analyser by Agilent (model E4980A)
–Data acquisition unit is a data sampler by National Instruments (model NI-6008). Available
sensors are: pressure sensor (Freescale MPX4250), and an in-house developed optical bubble
detector. On request an in-house developed ow rate sensor can be also attached.
•Ethernet bus consists of devices connected to the local area network of the laboratory
–Sample identication is a subsystem consisting of a camera and a sample holder. Each sample
tube is identied by a QR code. The code is read by the device and the relevant data can be
accessed through a cloud sample database.
–Optical inspection unit is a Smart Camera by National Instruments (NI-1772) consisting of an
embedded processor for on-line image analysis. The morphology of the MNP layer is detec-
ted and analysed by the unit to provide on-line information for the Measurement Controller.
–Laboratory Data Storage is a virtual cloud service to store measurement data
5.1.3 Fluid Control Unit
Fluid Control Unit accommodates the microuidic device under test and several uidic sensors and
actuators
•Microuidic chip holder is based on the platform of Micronit Inc (Fluidic Connect Pro). The struc-
ture was modied in several points to be able to accommodate the neodymium magnets for the
MNP accumulation.
•Zooming microscope aims the observation of the reaction chambers during the MNP loading and
during the reactions. The Smart Camera is attached to the microscope and its image is stored
automatically by the Measurement Controller, and analysed later by the Measurement Analyser.
5. CHAPTER. EQUIPMENT 67
•MNP quantier is a sensor device which is able to accurately measure the MNP quantity in the
reaction chambers by using a resonant coil magnetometer (RCM) (see in chapter 6).
5.2 MagneChip
MagneChip is a microuidic chip consisting of several reaction chambers to accumulate magnetic nan-
oparticles. The idea behind the magnetic microreactor chip lies on the excellent separation ability of
MNPs in magnetic eld. In the practical usage of such particles, biologically active molecules (e.g. biore-
ceptors) are immobilized on their surfaces. A suspension of the particles is forced to ow through the
microchip whereas the microchannel is prolated in certain positions forming microreactors. Magnetic
techniques enable separating of the particles [152] e.g. inside the microreactors where the magnetic
particles are accumulating then form a dense layer. In the consecutive step the reagents are owing
through the chip, while bioreaction occurs inside the microchambers. The Magne-Chip platform is a
possible realization. Thanks for the magnetic separation the biocatalyst remains in the microreactors
while the resulted product ows through the chip, which can be collected and/or quantied outside the
chip, for instance by absorbance method.
MagneChip reaction cells are designed to accomplish the following requirements
•Chemical ow-reaction occurs inside the chambers therefore a relative homogeneous ow velo-
city distribution should be provided. This condition is fullled if laminar ow is developed in the
chamber
•Magnetic nanoparticles are accumulated in the chambers and their drifting is prevented by a
directed magnetic eld. The critical ow rate and the accumulated amount of the particles can be
increased by using channels prolated to chambers. Nanoparticle amount shall be increased also
in order to provide enough biocatalysts for the reaction. Previous measurements in shake vials
revealed that at least a few hundred micrograms of particles are required for reasonable results.
The minimal required volume for this particle amount was approximated to be a few microlitres.
0.1 s 0.5 s 0.9 s 1.3 s
1.7 s 2.1 s 2.3 s 2.5 s
1 mm
Figure 5.2. Phase exchange in the reaction chamber (CFD simulation); air (white), water (gray); ow
rate Q= 28.64 µl min−1
5. CHAPTER. EQUIPMENT 68
Reaction cells (1.1 μl)
Inlet port Outlet port
10 mm
a)
b)
1.6
0.8
c)
d) e)
100101102
10−1
100
101
102
Flow rate (µL min−1)
Residence time (sec)
20 40 60 80
1
2
3
4
5
6
7
8
9
Flow rate Q (µL min−1)
P(Q)=0.059907Q+0.4303
P(Q)=0.086857Q+1.2783
MNP filled chamber
Empty chamber
drop
Figure 5.3. MagneChip characteristics; Layouts a) for enzymatic reactions b) for MNP quantication c)
ow velocity prole inside the chamber (units mm s−1), CFD simulation d) Residence time over ow
rate e) Pressure drop over ow rate
•Air bubbles stuck in the channels or chambers might cause serious measurement and uid hand-
ling errors. Channel prolongations especially foster the blockage of air bubbles. Therefore the
chamber should be small enough to prevent the spontaneous development of bubbles during the
initial lling by water.
Parametric simulations using a CFD solver revealed that the chamber diameter of about 4 mm and
height of about 100 µmgives the maximal chamber volume where the internal viscous forces keep the
uid front safely together even at reasonably high ow rates of 200 µl min−1to 500 µl min−1[153].
The chambers were designed to be 1µlin volume (i.e. 3.6 mm of diameter and 100µmof height). Final
volume of the chambers in realized chips were slightly larger as the mould master was thicker (110 µm)
instead of the planned (100 µm).
Transient CFD simulation of such structure can be seen in Fig. 5.2 as the air phase is exchanged by water.
The sequence clearly shows the eect of the viscous forces which nally provide the entirely change
of the gas to liquid. Simulation at 100×ow velocity although (not shown) caused the demerging of
the uid front and residual gas bubbles remained in the chamber after the uid passage 1.
5.2.1 Characteristics
MagneChip is a magnetic-nanoparticle reactor microchip, consisting of four chambers, 1.1µLof volume
each. The chips were made of PDMS (polydimethylsiloxane, Sylgard 184, Dow Corning Ltd, Germany),
plasma bonded on a glass substrate. Microchannel dimensions: width 300 µmand height 110 µm; re-
action cell dimensions: volume of 1.1µL, diameter of 3.6mm, height 110 µm[15].
The chip was xed in a microuidic chip holder (Fluidic Connect Pro, Micronit Inc, The Netherlands).
Figure 5.3 summarizes the main characteristics of MagneChip.
Layout type a) consists of four reaction chambers with meander channel sections between them. The
meandered sections are intended to provide the better observability of air bubble movements.
Layout type b) consists of two reaction chambers and was designed to use in the MNP quantication
experiments where one chamber is equipped with the resonant coil magnetometer and the other acted
as a reference.
Fig. 5.3c shows the uid velocity prole of the chamber at a typical ow rate conditions (28.64 µl min−1).
1Result of my student Peter Palovics
5. CHAPTER. EQUIPMENT 69
It can be seen that excluding the channel in and outow regions the uid velocity distribution falls in a
narrow range (the velocity dierence between the lowest and highest ow velocity area is about twice).
Fig. 5.3d shows the dependence of the residence time of the reaction compounds in the chamber on the
input ow rate. The residence time ranges from 1 s to 100 s.
Fig. 5.3e shows the pressure drop of the chip with and without MNP loading in the chambers. It can be
seen that the hydrodynamical resistance of the chip is increased in the presence of the MNP layer.
5.3 Construction methods of microuidic chips
5.3.1 4 cell MagneChip
Four chamber microuidic chip layout was used in the current study. Figure 5.3,a shows the four cham-
bers chip arrangement used for the enzymatic reaction measurements.
The chip were constructed by PDMS moulding technology. SU-8 photoresist (Microchem Inc., USA)
structures were prepared as a moulding master2, resulting in a channel height of 110 µm. PDMS was
poured on the master and was kept on room temperature for one day. After crosslinking the PDMS
replica was released.
Fluid ports were prepared by punching, using an in-house built puncher tool. PDMS channel bodies
and standard microscope glasses were bonded together after oxygen plasma treatment (Power stage:
“HI”, 90 sec) (Harrick Plasma, Ithaca, NY, US). The PDMS body was roughly 1 mm thick while the depth
of the recessed channel structure was 110µm. Therefore the enclosed chambers for MNP accumulation
had a volume of 1.1µl, diameter of 3.6 mm and a height of 110 µm.
5.3.2 2 cell MagneChip with integrated magnetosensor
Figure 5.3,b shows the two chambers chip arrangement used for the MNP quantity measurements.
The chip were constructed by PDMS moulding technology as it was described in section 5.3.1.
Inductor coils (20 turns, d= 4.25 mm) were custom made using a 20 µmcopper wire. Liquid PDMS was
poured on the surface of a glass plate and the coils were put on the mold positioned on the middle of the
glass plate. Glass spacer beads with a size of 500 µmwere sprinkled homogeneously in the mold and a
second glass layer was put on the top of the mold. A exible COC (cyclic olen copolymer, h= 100 µm)
lm was put between the glass cover and the PDMS to avoid the unwanted bonding of the cover glass
(Figure 5.4, step 1). Due to the spacer beads the resulted sandwich structure had a uniform thickness.
After thermal treatment (90 ◦C) on a hot plate for 40 minutes, the mold polymerized, then both the
COC lm and the cover glass plate were removed (Figure 5.4, step 2). PDMS channel bodies and PMDS
membranes were bonded together after oxygen plasma treatment (Power stage: “HI”, 90 sec) (Harrick
Plasma, Ithaca, NY, US). The PDMS body was roughly 1 mm thick while the depth of the recessed
channel structure was 110µm. Therefore the enclosed chambers for MNP accumulation had a volume
of 1.1µl, diameter of 3.6 mm and a height of 110 µm(Figure 5.4, step 3).
In the resulted structure the at coil was placed directly under the chamber i.e. the particles accumu-
lated right above the core of the coil. This way the nanoparticles interact only with the external eld
of the at coil, which may somewhat decrease the sensitivity.
Only the rst one (left one in the gures) of the two chambers was prepared for sensing, the other
chamber (without inductor coil) acted as a control chamber for further optical investigations (Figure
2SU-8 master was fabricated in Institute of Technical Physics and Materials Science, Hungarian Academy of Sciences
5. CHAPTER. EQUIPMENT 70
5.5). MNP trapping was provided by 3×4mm neodymium magnet (type N48, EuroMagnet Inc., Bud-
apest, Hungary) cylinders located directly below the chambers (Figure 5.5,a). The magnet cylinders
were placed in moveable drawers enabling “on/o” switching of the magnetic eld inside the cham-
bers by pushing in and pulling out the magnets. In Figure 5.5 the upper two magnet drawers were
“opened” i.e. the magnets were pulled out while the below two drawers were “closed” therefore MNP
was accumulated in the left chamber. The control chamber remained empty until the saturation of the
measurement one.
CoC film
Glass beads (500 μm)
Glass plates
Coil
Coil
CoC film
Microchamber (110 μm)
PDMS body (≈1 mm)
3Measurement
Chamber (1.1 μL)
Inlet
Outlet
PDMS film
Figure 5.4. Chip construction technology (sizing of elements are dier from real scale due to better
visibility): 1) PDMS layer formation for encapsulating the coil 2) delamination of the CoC layer before
bonding 3) bonding of the PDMS microchannel layer resulting in chambers of 1.1µl
1 mm
Coil
Coil
Magnet
1 mm
MagneChip
Input
Output
Chip holder Inspection window
1 cm Control
Chamber
a) b)
Figure 5.5. 2 cell Magne-Chip, left cell lled with MNP ; a) magnied image of the empty cell with the
neodymium magnet positioned below the cell and the inductor coil turns around the cell b) the same
cell lled with MNP
5. CHAPTER. EQUIPMENT 71
5.4 Unit operations of MagneChip
5.4.1 Filling of the Magne-Chip microreactors with MNPs
In an Eppendorf-tube a suspension (total volume: 1.5mL) was prepared from the MNP biocatalyst
(3.3 mg mL−1and PEG-4000 (3.3 mg mL−1in a 5 : 1 mixture of ultrapure water and 2-propanol. The
mixture was sonicated for 15 min in an ultrasonic bath (EMAG EMMI 15HC) kept on 30 ◦C. Right
before the lling up process the suspension was sonicated again for 10 min. During the lling process
the chip was kept on 25 ◦C. To avoid sedimentation gentle shaking (700 rpm) was maintained by an
orbital shaker (4.5mm of orbital diameter) during the chip loading procedure. Neodymium magnets
(type N48, 3×4mm) were placed in moveable drawers enabling “on/o” switching of the magnetic
eld when placed directly below the magnetic cells.
When lling up the chip, the prepared MNP suspension was driven through the chip by applying a
slight air pressure (0.2−0.3bar) to the suspension vial connected to the inlet of Magne-Chip via a
PTFE tube (Figure 5.6, Chip lling) while the temperature of the chip was kept on 25 ◦C. During the
lling process, the MNPs were accumulated in the reaction chambers due to the permanent magnets
placed in moveable drawers enabling “on/o” switching of the magnetic eld. Once one chamber was
saturated (Figure 5.6, Chamber #4 was loaded at rst), the magnet of the previous cell was turned on.
The same procedure was repeated (Figure 5.6, #3,#2,#1) until the required number of chambers was
lled up. Each cell of the Magne-Chip device could capture ca. 250 µgof MNP biocatalyst (see later
in section 6.3.2). For the optical inspection of the chip the NI-1772 (National Instruments, USA) smart
camera was used mounted to a stereo zoom microscope (10×magnication).
Vsis the total volume of the reaction chambers. Vcis the void volume, dened as the volume which
is not occupied by the particles thus the uid can ow through. The void fraction of the chambers is
dened as
β=Vv
Vs
(5.1)
Substrate(s)
Magnet drawers
Cell
#1
Cell
#2
Cell
#3
Cell
#4
#1 #2 #3 #4
Magne-Chip device
Suspension of MNP +
immobilized biomolecule
MNP
W
A
S
T
E
Time
#4 #3#2 #1
Reaction
Chip filling
Figure 5.6. The Magne-Chip concept – the MNP suspension ows through the chip while the con-
secutively actuated magnet drawers activate the magnetic eld in the reaction chambers (“Cell 1-4”),
therefore the MNP accumulates in the cells and form a dense biocatalytic layer. Once the cells are lled
with MNP, the reagents (e.g. substrates) ow through the chip and reaction occurs in the cells
5. CHAPTER. EQUIPMENT 72
T=30 oC
Magne-Chip
Detector
Deuterium
Light
Source
Z-Cell
Waste
Thermostat
T-Cell
Substrate
(1 ml)
Washing Buffer
(10 ml)
Optical
Inspection
P
Magnet drawers
#1 #2 #3 #4
#4 #3 #2 #1
Calibration Chip
Cleaning
Chip filling Experiment phase
MNP suspension
Figure 5.7. Schematic diagram of the uid control system during single parameter experiments
5.4.2 Cleaning the chip
Detergent solution (1 : 99 mixture of EMAG EM-080 and ultrapure water) was used for the cleaning
of microchannels. At the end of the reaction, by positioning the moveable drawers to “o” position,
the magnetic eld is deactivated and TRIS buer (0.1m, pH 8.8), solution of detergent, nally ultrapure
water were driven through the chip, each for 2minutes at 300 µl min−1ow rate while the particles
were ushed out completely.
5.5 Method of single parameter experiments in-chip
Biocatalytic activity of the magnetically entrapped particles was measured in order to demonstrate the
eectiveness of dierent particle size congurations. Phenylalanine ammonia-lyase (PAL) enzyme was
immobilized onto the surface of the MNPs. Suspension of MNPs was prepared as it was described in
section 6.2.1. Four chambers of a magnetic microreactor system were loaded with MNPs (Figure 5.7,
Chip lling) as it was described in section 5.4.1.
Then, 20 mM of l-1a substrate was dispensed through the chip in absorbance-measurement congur-
ation (Figure 5.7) with a constant ow rate of 6µl min−1on 30 ◦C. The enzyme catalyses the conversion
of l-1a to trans-cinnamic acid (l-2a), which has a specic absorbance at 295 nm. During the absorbance
measurement the solution ows through a Z-shaped channel whereas a collimated light beam crosses
the ow path and the outgoing light intensity is measured. This intensity was measured rst while TRIS
buer was passing through the cell, acting as reference intensity (Figure 5.7, Calibration). Secondly, the
5. CHAPTER. EQUIPMENT 73
T=30 oC
Magne-Chip
Detector
Deuterium
Light
Source
Z-Cell
Waste
Thermostat
T-Cell
Substrate
(1 ml)
Washing Buffer
(10 ml)
Valve
Optical
Inspection
S
I
F
E
D
C
B
A
Substrate stocks
A B C E F
D
P
Magnet drawers
#1 #2 #3 #4
#4 #3 #2 #1
Calibration
Cycle
#1
Cycle
#2
Cycle
#3
Cycle
#n-1
Cycle
#n
Chip
Cleaning
Reaction Re-init
Chip filling Experiment cycles
Bypass
MNP suspension
Control
Figure 5.8. Schematic diagram of the uid control system for performing multiparameter experiments
substrate was owing through the chip and trans-cinnamic acid was formed as the enzyme acted on
the l-1a substrate (Figure 5.7, Experiment phase). The measurement was interrupted when the specic
product absorbance signal (based on the method described in section 2.5.2, with equipments described
in 5.1) turned into saturation for 5minutes. After the measurement the chip was cleaned (Figure 5.7,
Chip cleaning).
5.6 Method of multi parameter experiments in-chip
Once the Magne-Chip was lled with MNPs, the dosage of liquids owing through the chip could be
controlled in a programmed manner through uFlu-Studio (see section 5.1). During the measurements
the product quantity was monitored by an UV-Vis equipment. As the MNP biocatalysts are reusable, the
chip could be cyclically re-initialized by washing after each biotransformation experiment. This enabled
the switching of the substrate or changing the working parameters cycle by cycle without changing
the biocatalyst in the system i.e. screening conditions or various substrates following an automated
sequence.
Magne-Chip experiments involved four steps: 1) lling up the chip with MNPs, 2) absorbance cal-
ibration 3) experiment cycles 4) chip cleaning.
During the forthcoming steps (Figure 5.8, Calibration, Experiment cycles), the switch valve at the inlet
of the Magne-Chip was switched to the substrate (reagent) circuit. The ow controller maintained the
5. CHAPTER. EQUIPMENT 74
AU
tstart,n tend
tsat
5%
5%
Δtsat,min
Δtref,min
tstart,n+1
Reaction step Re-initialization step
Figure 5.9. Typical one cycle absorbance plot over time of multi-parameter measurements in-chip at a
given wavelength. Time markers show the integration limits of kinetic calculations. Absorbance mark-
ers show the 5% acceptance range of stationary product concentration
dosage of the substrate and other chemicals following the programmed sequence.
At the nal, chip cleaning step (Figure 5.8, Chip cleaning) the magnetic drawers were pulled out and
washing solution was driven through the chip while the particles were ushed.
5.6.1 Calibration
Washing buer was driven through the chip and the Z-cell (Figure 7.1) at 440 µl min−1ow rate. Once
the ow conditions are stable (i.e. no air bubbles in the Z-cell) the total intensity spectra and the dark
spectra were stored as references and the calculated absorbance spectra was stored for the total avail-
able wavelength range (200 −1100 nm) in every second.
5.6.2 Fluid handling during the Experiment Cycles
One syringe (the “substrate syringe”) was lled up through the switching valve (Figure Figure 5.8, S-
A, S-B, ... S-F positions) by the adequate substrate from the substrate stocks A-F in accordance to the
programmed measurement sequence. Once the substrate syringe (1 mL of volume) was fully loaded the
valve switched to S-I position. Substrate, washing buer or a mixture of them (mixing ratio is variable
from 1:1to 1 : 1600) was driven through the chip at a determined ow rate. The internal pressure
of the uid circuitry was continuously monitored through the connector to the other syringe (the
“washing buer syringe”). The chip was kept on a predened temperature through the measurement.
5.6.3 Variants of experiment cycles
Each experiment cycle involves a Reaction step and a Re-initialization step (Figure 5.8, Experiment
Cycles). Reaction step – (Figure 5.9, tstart,n −tend). The dosage of the substrate follows a predened
sequence in accordance to one of the variants listed below. The absorbance value at a previously selec-
ted wavelength was continuously monitored. The cycle step ended when the designated step time had
been passed or when the reaction had been saturated at least for ∆tsat,min = 10 min (i.e. absorbance
changes fall into the 5% range of the saturation absorbance level) whichever happened earlier.
5. CHAPTER. EQUIPMENT 75
(a) The dosage of the substrate started (1st cycle) or continued at unchanged ow rate.
(b) The dosage of the substrate started (1st cycle) or continued while the ow rate may change cycle
by cycle.
(c) The dosage of the substrate and the washing buer ran parallel on a designated ratio resulted in
a predened dilution of the substrate at the chip inlet. The dilution ratio may be dierent cycle
by cycle.
(d) The remaining substrate was drained through a bypass valve (Figure 5.8). The consequent sub-
strate from the substrate stock (A-F) was loaded into the substrate syringe and the dosage of the
substrate began at a predened ow rate. Re-initialization step – (Figure 5.9, tend −tstart,n+1) The
dosage of the substrate stopped while the washing buer was driven through the system at a
designated ow rate. The step ended when the designated step time had been passed or when
the absorbance value reached the reference value (i.e. absorbance changes fall into the 5% range
of the reference absorbance level) at least for ∆tref,min = 10 min, whichever happened earlier.
The reaction and re-initialization steps constitute one cycle. Cycles are repeated several times ac-
cording to the predened measurement sequence.
Chapter 6
In-situ quantication of magnetic
nanoparticles in a microchamber
Abbreviations and Symbols
Symbol Units
cEEnzyme coverage -
dcc Particle-to-particle distance m
ffrequency Hz
nNumber of particles -
CCapacitance F
LInductance H
TTemperature ◦C
RResistance Ω
QQuality factor -
βaVolume fraction -
φFrequency sensitivity factor Hz/100µg
µbVolume ratio Vp
Vch -
σFrequency noise Hz
ΓCell capacity µg
aSee section 2.4.3
bSee section 2.6.2
6.1 Introduction
The accurate determination of the biocatalyst amount in the chamber is crucial for obtaining reaction
kinetics data (see also in Section 3.3). Maximizing the reactive surface to volume ratio is crucial in
microreactors even by using smaller particles or nding ways to collect more particles in the chamber.
In this chapter a method is demonstrated for the accurate in-situ determination the quantity of the
entrapped magnetic nanoparticles in the reaction chamber of a microuidic device. The measurement
was based on the resonance frequency shift of a passive electrical resonant circuit where a at inductor
coil integrated in a silicone elastomer lm acted as a sensor. The particle amount inside the chamber
76
6. CHAPTER. IN-SITU QUANTIFICATION OF MAGNETIC NANOPARTICLES IN A MICROCHAMBER77
Table 6.1. Composition of the suspensions used in the experiments
Composition
No
MNP250 : MNP600
mass
ratio
MNP
[mg ml−1]
PEG
[mg ml−1]
IPA
[µL]
mMNP a
[mg]
mEb
[µg]
cEc
1 1:0 1
2 0:1 0.16
3 1:1
10 10 - 5 750
0.58
4 1:0 1
5 0:1 0.16
6 1:1
2.1 2.1 150 1.65 240
0.58
atotal mass of MNP powder added to the suspension
btotal PAL enzyme content in the suspension
cenzyme coverage on the MNP surface relative to MNP250
aected the inductivity therefore the resonance frequency was changed. The method also enabled the
on-line monitoring of the actual particle quantity in the chamber.
The maximum amount of accumulated particles in the chamber was measured in several cases using
dierent particle sizes or a mixture of dierent sized particles.
6.2 Materials and Methods
6.2.1 Magnetic nanoparticle suspensions
MNP250 and MNP600 used in this study were prepared from magnetite (Fe3O4) core magnetic nano-
particle carriers (MNP-250-EP14 and MNP-600-EP14, products of SynBiocat LLC, Budapest, Hungary)
by bio-functionalization. The nominal diameters of the biofunctionalized particles were 250 nm and
600 nm, for MNP250 and MNP600 , respectively. Particle size distribution was also characterized by dy-
namic light scattering (DLS, Brookhaven BI-200SM Laser Light Scattering Instrument). Samples were
sonicated in ethanol for 20 minutes in room temperature, then analysed by laser beam (λ= 490 nm)
at 25◦C. The initial MNP particle dimensions were thought to be large enough as suitable carriers
of phenylalanine ammonia-lyase (PAL) which is a relatively large multimeric enzyme (d= 10 nm)
but still small enough to form stable suspensions behaving like nanouids [154]. Briey, the cores for
MNP-250-EP14 and MNP-600-EP14 were prepared using a solvothermal method [155], which resulted in
monodisperse magnetite MNPs. Then, with the aid of TEOS (tetraethyl orthosilicate) and epoxy func-
tionalization, the MNP cores were coated by an epoxy-functionalized SiO2-layer. Bio-functionalization
with PAL (15%, relative to the mass of MNP-250-EP14 and MNP-600-EP14) resulted in particles ready
to use in the experiments.
After the binding process, the supernatant was analysed by the Bradford method [126], suggesting that
all PALs were bounded on the surface of the particles.
Dierent MNP suspensions were made from the PAL-functionalized MNP powder and distilled water
(quantities are indicated in Table 6.2.1). MNP quantity was measured by a standard 0.1 mg laboratory
balance (PW124, Adam Equipment, US). Polyethylene-glycol (PEG, Mw= 4000 g mol−1) was added to
the suspension (PEG/MNP=1/1w/w) and sonicated for 10 minutes to avoid the aggregation of MNPs.
After 10 minutes of vortex stirring a highly homogeneous suspension was obtained.
The morphology and the surface element composition of the particles was analysed by JEOL JSM-
6. CHAPTER. IN-SITU QUANTIFICATION OF MAGNETIC NANOPARTICLES IN A MICROCHAMBER78
5500LV scanning electron microscope (SEM) and the element analysis was carried out with energy
dispersive spectroscopy/energy dispersive X-ray analysis (EDS/EDAX with Si(Li) detector) applying
20 kV accelerating voltage and sampling time of 60 sec. Six dierent suspensions were prepared as
summarized in Table 6.2.1. Suspensions 1−3were used for the sensor characterization measurements,
while suspensions 4−6were used for chip loading both in quantity measurement and in biocatalytic
experiments. Isopropyl alcohol was added to the suspensions (quantities indicated in Table 6.2.1, res-
ulting 8% IPA solution), which enhanced the transfer of the particles from the sample tube to the chip.
Once the chip was lled with particles, TRIS solution (2 ml buer, 0.1m, pH 8.8) was own through
the chip to ensure the washing out of IPA. Preliminary in-chip investigations with – and without IPA
showed negligible changes in enzyme activity, thus short time (≈10 minutes) interaction with 8% IPA
did not aect the enzyme activity during the measurements.
6.2.2 Immobilization PcPAL onto MNPs
Epoxy-MNPs [MagneCat-250GP14 (epoxy-functionalized magnetic nanoparticles with an average dia-
meter of 250 nm) : 108 mg] were added to TRIS buer (3mL, 0.1M, pH 8.8) and dispersed by ultrason-
ication (35kHz,20min). The MNP suspension was added to the PcPAL solution (3 mg mL−1), in TRIS
buer: 4mL, 0.1M, pH 8.8) and the mixture was shaken for 24 h (25 ◦C,450 rpm). The PAL-coated
MNP particles were xed at the bottom of the ask with a neodymium magnet and the supernatant
was decanted. The MNP preparation was washed – by a re-suspension - magnetic xing - decantation
sequence in each washing steps – with TRIS buer (3×4mL, 0.1M, pH 8.8) and with ethanol (4mL).
The MNP biocatalyst was dried in vacuum at room temperature for 2h. After immobilization, negli-
gible protein contents in the supernatant of the washing procedure were determined by the Bradford
assay[14].
6.2.3 Oscillator circuit and frequency measurement
Figure 6.1. Equivalent circuit of the measurement setup
Resonance frequency measurement was performed by an Agilent E4980A LCR meter in frequency
sweep mode in a narrow range (≈10 kHz) around the resonance frequency. A four-wire measurement
arrangement was used with Kelvin-clips to ensure proper electrical connection. The measurement setup
was sensitive to spread impedances therefore the clips were xed by screws on the measurement plat-
form. In the resonance region the phase plot decreases sharply and crosses the 0°at the resonance
frequency. The phase shift data was acquired using a supplementary controller software developed in
LabVIEW (National Instruments, USA). A linear t was applied to the sampled points and the solution
of the linear equation gave the resonance frequency. The frequency shift was recorded with a sampling
6. CHAPTER. IN-SITU QUANTIFICATION OF MAGNETIC NANOPARTICLES IN A MICROCHAMBER79
rate of 10 samples/sec and this data was analysed further. In serial RLC oscillators the quality factor
(Q) is dened as follows:
Q=1
RrL
C(6.1)
where Ris practically the serial resistance of the coil. The higher the quality factor is, the steeper the
phase plot will be at the resonance frequency resulting in higher sensitivity during measurement. The
LCR analyser used during the measurements has a bandwidth limitation at 2 MHz therefore 1.6 MHz
natural resonance frequency was chosen for the experiments. The 20 turn at coils showed 3µHself-
inductance and 250 fF parallel capacitance in wide frequency range between 0.001 MHz to 1 MHz. The
real part of the impedance was about 4 Ω . Figure 6.1 shows the equivalent circuit of the measurement
setup. An external lm capacitor of 3.3 nF was also installed to ensure the desired resonance frequency
and quality factor (fr= 1.57 MHz and Q= 7.4). Film capacitors have small temperature dependence,
therefore no particular attention had to pay the thermal stability of the environment. Nevertheless,
during the initial sensor characterization the chip and the discrete elements were attached to a cold
plate (T= 25◦C). The total frequency shift values (∆f) were determined by seeking the local minima
and maxima of the sampled data.
6.2.4 Calibration of the sensors
10 mg/ml of MNP in 1 ml
Coolant out Coolant inTo the LRC meter
100 μg per drop
Shield
Chip holder
Figure 6.2. Initial characterization of the inductive sensors, pipetting from the MNP suspension (1),
dispensing on the PDMS lm surface above the integrated coil (2)
Calibration was performed using open membranes (coils without bonded channels, Figure 6.2, step
1.). A neodymium magnet was placed under the coil therefore the particles accumulated inside the open
coil area. 1 ml suspension containing MNP of 10 mg ml−1was prepared. Droplets of 10 µlcontaining
100 µgof nanoparticles were dispensed on the membrane surface (Figure 6.2, step 2) using a 10 µl
digital pipette in 10 steps. The frequency-shift response was recorded and the frequency sensitivity
factor (frequency shift due to 100 µgchange in particle quantity) was determined for each particle
type.
The chip was placed on a cold-plate to ensure thermal stability and a grounded stainless steel cover (wall
thickness 1 mm) was used for electric shielding. The frequency shift data were acquired and the average
step-size was determined on each chip for each particle sizes and mixtures. After this characterization
step the chip preparation was nished by bonding the PDMS channel bodies.
6. CHAPTER. IN-SITU QUANTIFICATION OF MAGNETIC NANOPARTICLES IN A MICROCHAMBER80
6.2.5 Measurement of the entrapped particle quantity
Eppendorf tube
with MNP suspension
Measurement chamber
Control chamber
time
Vortex stirrer
Signal processing
Waste
f
Figure 6.3. Arrangement for the MNP quantity measurement and the three sequential steps of the
chamber lling with MNPs
Particle quantity measurements were done in a magnetic microreactor system with a 2chambers
layout (Figure 5.3,a). MNP suspension (see Section 6.2.1) was prepared in a standard centrifuge tube
(1.5 ml volume). The chip was ushed with air before the lling up procedure i.e. the channels were
empty. The tube with the suspension was placed in a vortex stirrer, and then it was pressurized which
forced the particle suspension to ow through the chip (Figure 6.3, step 1,2). The particles were accu-
mulated in the measurement chamber due to the neodymium magnet positioned under the reaction
cell.
Once the measurement chamber saturated (Figure 5.5,b) some particles drifted towards the downstream
control chamber and became observable there (Figure 6.3, step 3).
The accumulating nanoparticle mass increased the relative permeability of the inductor which resulted
in a decrement of the resonance frequency.
When the chamber became full, the magnet was released and the particles ushed away. The process
could be repeated 7-8 times as long as there was enough suspension in the tube.
The total frequency shift values (∆f) were averaged measurement by measurement. Cell capacity
(Γ) is given as the ratio of the particle dependent frequency sensitivity factor (φ) and the measured
frequency shift:
Γ[µg] = (∆f/φ)×100 (6.2)
Noise of the measurement was determined by taking a quasi linear part of the recorded frequency
6. CHAPTER. IN-SITU QUANTIFICATION OF MAGNETIC NANOPARTICLES IN A MICROCHAMBER81
shift data of about 100 samples. The baseline drift was removed and the standard deviation ([Hz]) was
calculated.
Signal to noise (SNR) was calculated by taking the typical value of the frequency value shift during a
chamber lling measurement as the signal, over the calculated noise, SNR = ∆ffull/δnoise. Resolution
of the sensor was determined as the minimal detectable change in resonance frequency ∆fmin =δnoise.
6.3 Results
6.3.1 Characterization of the sensors
Typical sensor characterization plot can be seen in Figure 6.4, left. Some uncertainty was observed at
the rst drop, therefore the rst step was neglected in each experiment. The consequent step-by-step
dispensing of particle suspension resulted in a frequency-step function. The deposition time of the
particles somewhat increased from the second dispensing step because of the accumulated water drop
around the coil. We found that pure water dispensing somewhat aects the resonance frequency but
the eect was negligible compared to the normal usage.
0 50 100 150
−250
−220
−190
−160
−130
−100
−70
−40
−10
0
Sample
Resonance frequency shift (Hz)
0 20 40 60 80 100 120 140
−80
−70
−60
−50
−40
−30
−20
−10
0
Time (s)
Resonance frequency shift (Hz)
Accumulation
Saturation
100 µg
100 µg
100 µg
100 µg
100 µg
100 µg
100 µg
250 µg
Figure 6.4. Typical frequency shift responses during characterization by adding 100 µgstocks of MNP600
by pipetting (left) and during measurement while the chamber is being lled with MNP600 (right)
The frequency steps (n= 9) were averaged chip by chip and particle by particle, and the resulted
frequency shift values were normalized with respect to the total particle mass to get particle dependent
frequency sensitivity factor (φ).
It was found that the sensitivity depends on the actual particle type. The sensitivity varied between
φ600 = 30.3±0.63Hz/100µgto φ250 = 34.1±0.71Hz/100µgfor MNP600 and MNP250 respectively,
while the 1 : 1 mixture of the two types resulted in the mean value at φ600:250 = 32.1±0.73Hz/100µg
in line with the expectations. Table 6.2, sensitivity (φ) column summarizes these results.
The observed sensitivity of the sensor was far below the ones used in magneto-immunosensors (e.g.
125 Hz µg−1[116]). Though, such sensitivity is not needed in the current application as the expected
sum mass of the particles fell in the hundreds of micrograms range. Moreover, no implementation of
resonant coil magneto sensor was found with microuidic integration. Such realization enabled to work
with sample volumes in the microlitre range.
6. CHAPTER. IN-SITU QUANTIFICATION OF MAGNETIC NANOPARTICLES IN A MICROCHAMBER82
Table 6.2. Calculated frequency sensitivity, SNR and sensitivity limit values for the dierent particle
types
Particle Sensitivity φSNR aResolution [µg] b
MNP600 30.3±0.63 34.57.98
MNP250 34.1±0.71 30.72 7.04
MNP250:600 32.1±0.73 36.97.48
aSNR was calculated as the ratio of the frequency shift at full chamber and the measurement noise σnoise = 2.4Hz
bThe resolution is the smallest detectable change is particle mass when ∆f=σnoise
241.67
248.32
283.59
0
50
100
150
200
250
300
350
MNP600
MNP250
Mixture
MeasuredMNPquantity[μg]
n=5
n=7
n=5
Figure 6.5. Summary of in-situ MNP quantity measurement in the reaction chamber. The mixture of
MNP600 and MNP250 resulted in higher reaction cell capacity by 15%
6.3.2 Measurements of the MNP quantity in the chamber
Initially, the magnet drawer was pushed in, i.e. the magnetic eld was activated under the cell. The
induced eld of the moving magnet caused a single frequency shift peak and a notably oset in the
resonance frequency of about 27 kHz. The oset shift was compensated in the controller software then
pressure was applied to the MNP suspension tube which forced the particles to ow in the chip. Due
to the activated magnetic eld, the MNPs began to accumulate in the measurement chamber.
Typical frequency-shift response is plotted while lling the chamber with MNPs (Figure 6.4, right). It
was observed that the accumulation of the particles in the chamber resulted in decreasing resonant fre-
quency until the chamber saturated with particles and the frequency shift plot attened. To avoid the
obstruction of the microchannels, the magnet holder drawer was pulled out thereafter. After pushing
in the drawer the accumulation could begin again.
The results of in situ MNP quantity measurements are summarized in Figure 6.5. It is clearly visible that
the cell capacity was approximately the same for both homogeneous particle sizes (MNP600 or MNP250 ).
This result reects the sphere packing theory, as the dierence of 2% falls within the expected meas-
urement error.
According to the measurements the binary mixture (MNP250:600 ) of the particles resulted in a signic-
antly higher MNP mass (17%) captured in the magnetic chamber as compared to the chamber capacity
6. CHAPTER. IN-SITU QUANTIFICATION OF MAGNETIC NANOPARTICLES IN A MICROCHAMBER83
Table 6.3. Summary of particle parameters and comparison of measurement results
Relative changes
Particle Size (DLS) [nm] Avg. Dista[nm] S to V ratiobCell capacity
MNP600 562 1286 1 1
MNP250 230 500 2.5 1.02
MNP250:600 Mixture 770; 300 2.06 1.17
aCalculation is based on the approximated total number of particles (see Section 6.3.3)
bSurface to Volume ratio, calculation is based on the method described in Section 6.3.3
Table 6.4. Material composition of the magnetic nanoparticles
Element Atomic Weight (%)
Core Core+TEOS
C27% 13%
O60% 64%
Si 0% 11%
Fe 13% 12%
lled with MNPs of uniform particle sizes. The relative changes of chamber capacities compared to the
MNP600 case are summarized in Table 6.3 (Cell capacity).
The average noise level during the measurements was found to be σnoise = 2.4 Hz. Calculated SNR
and resolution of the measurements with respect to the dierent particle types are shown in Table 6.2.
Please note, that all measurements were done with the same chip, and the indicated values correspond
to the given chip.
6.3.3 Particle volume fraction
DLS analysis provided the size distribution of the particles in dierent phases of the production. Initial
particle dimensions were 210 nm and 548 nm of mean diameters (Figure 6.6a, blue trace) for MNP-250-EP14
and MNP-600-EP14, respectively. After building the silicon dioxide layer on the surface, the measured
mean diameters were 562 nm and 230 nm, respectively (Figure 6.6a, red trace, see also the SEM image
in Figure 6.6b). Taking the size of the PAL enzyme (≈10 nm) bonded onto the surface of the particles,
the nal dimensions were approximated to be about 582 nm (MNP600 ) and 250 nm (MNP250 ).
The size of the particles suggests that they fall into the multidomain magnetic regime with moderate
coercivity and some remanent magnetization. The immobilized enzymes represent a repulsive force
against the other particles therefore remanent magnetization eects were not observed.
Based on the material composition data of the particles (resulted by EDS analysis, Table 6.4), the density
and mass of the individual particles can be calculated.
Taking also the measured particle sizes, the mass of the silica shell can be estimated taking into ac-
count its thickness of 20 nm and 14 nm for the MNP250 and MNP600 particles, respectively. More 20 nm
should be added to this value representing the enzyme layer, however its contribution in the total mass
is only 15%. The estimated masses of the single particles are m250 = 25 fg and m600 = 426 fg. The
ratio of the measured quantity and the single particle mass reveals the total number of particles in the
6. CHAPTER. IN-SITU QUANTIFICATION OF MAGNETIC NANOPARTICLES IN A MICROCHAMBER84
0
20
40
60
80
100
0
1000
2000
Intensity(%)
Partcicle diameter (nm)
548 nm
562 nm
0
20
40
60
80
100
0
200
400
600
Intensity(%)
210nm
230nm
Partciclediameter(nm)
a)
b)
c)
Figure 6.6. a-b) DLS characteristics and mean diameters of the investigated MNPs (a) MNP250 ; b)
MNP600 ), magnetite cores (blue), cores with TEOS coating (red) c) SEM image of the MNP suspen-
sion (MNP250 )
chamber: n250 = 9.04 ·109±2.8%,n600 = 5.3·108±2.8%, taking the resolution of the measurement
into account. Based on the above numbers the surface to volume (S/V) ratio can be estimated both for
homogeneous and mixed suspensions (Table 6.3). Taking the theoretical limits of the particle volumet-
ric ratio the following equation provides the average center-to-center distance between the particles
(also included in (Table 6.3):
dcc = 2 ·3
sµVp
4
3πn (6.3)
As it was found that dcc is about twice the particle diameter, the above results suggest that the
emplacement of the particles in the chamber is sparse. This may be due to the internal repulsive forces,
solvating of the enzyme layer by water and PEG and the “stickiness” of the enzyme coating which led
to the enlargement of the possible voids. On one hand, the volume fraction of the particles was about
7% regardless to the particle size and it was increased to 10% in case of the mixed particles. Therefore
the void fraction of the chambers is about β= 0.93 -0.9. On the other hand the further accumulation
of the particles resulted in channel obstruction as the new particles at the inlet force the ones at the
outlet to drift out the chamber making an obstacle in the narrow outlet channel. This implies that the
practical limit of the cell capacity had been already reached.
6.4 Conclusions
A novel method was reported for in situ measurement of the amount of magnetic nanoparticles in
micro-sized reaction chambers. The method is based on the measurement of the resonance frequency
shift of a serial RLC circuit where the inductor coil acts as a sensor (referred as resonant coil magne-
tometer), the inductance of which is aected by the accumulating nanoparticles. Doubtless, the main
advantage of the resonant coil technique is that the method is non-destructive, its accuracy is relat-
ive high and the cost of the equipment is signicantly cheaper than other magnetometer techniques
6. CHAPTER. IN-SITU QUANTIFICATION OF MAGNETIC NANOPARTICLES IN A MICROCHAMBER85
e.g. SQUID. The presented in-chip integrated inductor coil enables the measurement in-situ; the ar-
rangement and quantity of the magnetic nanoparticles remains the same during the bioanalytic or
biocatalytic measurements. The technique also enables the continuous monitoring of the entrapped
particle quantity which would not feasible with ow-type counters.
The main ndings of this chapter can be concluded as follows.
•A resonant inductor coil magnetometer was integrated into a microuidic chip. It was success-
fully demonstrated that the sensor had sucient resolution to determine the quantity of the
accumulated magnetic nanoparticles in the microliter volume reaction chamber of the chip.
•Particle size aected the sensitivity of the resonant coil magnetosensor, therefore it should be
characterized for each particle type.
•Magnetic nanoparticles suspended homogeneously in a wet medium and behaved like perfect
spheres while being accumulated in micro sized chambers. The arrangement of these spheres
seemed to be invariant of the sphere size in the 200 nm to 600 nm range, as the measured particle
quantity was found to be equal regardless the particle size. Consequently, the arrangement of the
particles approaches the sphere-packing theory.
•Filling the chamber by an 1:1 mixture of dierent sized particles (size ratio 1:2.5) indicated
that the smaller particles may be wedged between the larger ones, which was resulted in higher
lling quantity (by ≈17%) compared to the cells lled with uniformly sized MNPs.
6.5 Summary of scientic results
Thesis 2 In-situ quantication of mangetic nanoparticles in a microchamber
I developed a novel method for the reproducible lling of the chambers of the chip sized reactor device with
magnetic nanoparticles. The particles are separated and accumulated in the chambers by using permanent
magnets. I developed an in-situ and on-line method for the quantication of magnetic nanoparticles located
in a micro chamber. The measurement setup consists of a modied PDMS microuidic structure encapsu-
lating a resonant coil. With external passive components attached the arrangement forms a resonant coil
magnetometer. I found that the resonance frequency of the magnetometer depends on the overall quantity
of magnetic particles in the chamber.[S2], [C3]
1. I measured the total mass of particles accumulated in the chamber of the microuidic device
using particles of 250 nm and 600 nm diameter, respectively. I found that the total mass of the
accumulated particles were nearly identical in case of the two investigated particle types (dier-
ence was 2.7%). I veried, that the total mass of the 1:1mixture of two dierent sized particles
(250 nm and 600 nm) accumulated in the chamber was signicantly (by 17%) higher than the
total mass of single sized particles in the same chamber.
Chapter 7
Lab-on-a-Chip microreactor platform
Abbreviations and Symbols
Symbol Units
EEnzyme quantity µmol
[E]Enzyme concentration mm
kcat Turnover number s−1
kcat/KmSpecicity constant s−1m
KmMichaelis-Menten constant mm
pPressure bar
PProduct quantity mol
[P∗]Product fraction ([S0]−[St])/[S0]-
RReaction capacity mol/sec
RDViscous resistance m−2
SC Cell dierence score -
[S0]Initial substrate concentration mm
[St]Outow substrate concentration mm
TTemperature ◦C
tTime s,min
UBSpecic activity µmol g−1min−1
UVVolumetric activity g l−1min−1
VcReaction chamber volume µl
VvVoid chamber volume µl
˙
QFlow rate µl min−1
βVoidity (Vv/Vc) -
εMolar extinction coecient m−1cm−1
λWavelength nm
νReaction velocity nmol min−1
ΠPlan view image of the chip -
86
7. CHAPTER. LAB-ON-A-CHIP MICROREACTOR PLATFORM 87
7.1 Introduction
Microuidic application of immobilized enzymes oers novel possibilities in diagnostics [156, 157],
synthetic [158] or analytical [132, 159] applications (see detailed in Section 3.3). When PAL is immob-
ilized in magnetic nanoparticles (MNPs) [S4], the MNPs are applicable to ow in microuidic systems
together with the liquid [156], or can be anchored at pre-designed positions so that the ow [157] of
the uid in the system can be chosen freely compared to them. This creates a unique opportunity to
develop modular micro-systems with the ability of exible variation of biocatalysts [S4].
This chapter presents experiments carried out by the MagneChip platform, used as a microreactor
system. In each experiment, phenylalanine ammonia lyase (PAL) from Petroselinum crispum was im-
mobilized onto the surface of magnetic nanoparticles and one or more chamber of the chip were loaded
with them.
Biotransformation of the natural substrate l-phenylalanine (l-1a) of PAL and some unnatural sub-
strates (rac-1b-f) were carried out under dierent conditions to study
•the reproducibility of biotransformations occur in the microreactor system,
•the eects of the long term and cyclic re-usage of the biocatalyst on the biocatalytic activity,
•the nding of the optimal substrate concentration and ow rate of the in-chip biotransformation
and the calculation of the reaction kinetic constants,
•the eects of the particle size on the biocatalytic activity.
7.2 Materials
7.2.1 Phenylalanine ammonia-lyase from parsley (Petroselinum crispum)
Phenylalanine ammonia-lyase (PAL) from parsley (Petroselinum crispum) was overexpressed in E. coli
and puried according to the method described by Poppe, et al [76].
MNP PAL
PAL
PAL
PAL
PAL
PAL
PAL PAL
Ar Ar Ar
(L-1a)
L-Phe 2a
NH4
+
NH4
++
(S)(S)
NH2
COOH
NH2
COOH COOH
COOH
(R)(R)
NH2
COOH
Cl
Ar =
Cl
Br O S
PAL
b c d e f
rac-1b-f 2b-f D-1b-f
Scheme 7.1. Ammonia elimination from dierent amino acids (1a-f) catalyzed by PAL immobilized on
MNPs
7. CHAPTER. LAB-ON-A-CHIP MICROREACTOR PLATFORM 88
7.2.2 Chemicals
Functionalized magnetic nanoparticles for covalent binding of PAL (MagneCat-250GP14) were obtained
from SynBiocat LLC (Budapest, Hungary). Detergent EMAG EM-080 was obtained from EMAG AG
(Mörfelden-Walldor, Germany).
l-phenylalanine , cinnamic acid, tris(hydroxymethyl)aminomethane (TRIS), polyethylene glycol 4000
(PEG 4000), 2-propanol, DL-2-amino-3-(4-bromophenyl)-propanoic acid (rac-1b) and
(E)-3-(4-bromophenyl)acrylic acid (2b) were purchased from Sigma Aldrich (St. Luis, MO, USA). DL-
2-amino-3-(furan-2-yl)propanoic acid (rac-1c), DL-2-amino-3-(thiophen-2-yl)propanoic acid (rac-1d),
DL-2-amino-3-(2-chlorophenyl)propanoic acid (rac-1e), DL-2-amino-3-(4-chlorophenyl)propanoic acid
(rac-1f), (E)-3-(furan-2-yl)acrylic acid (2c), (E)-3-(2-chlorophenyl)acrylic acid (2e), (E)-3-(thiophen-2-
yl)acrylic acid (2d) and (E)-3-(4-chlorophenyl)acrylic acid (2f) were synthetized as previously described
[S3].
7.3 Methods
7.3.1 UV characterization of the substrates
UV-spectra of l-1a,rac-1b-f and the corresponding acrylic acids (2a-f) were measured by the ow-cell
spectrometer setup in continuous ow mode at ow rate (48 µl min−1). Absorbance spectra were taken
in 4 dierent molar concentrations from 0.05 mg mL−1to 0.006 mg mL−1and the molar extinction
coecient (ε) was determined in each case by linear tting. The spectrum of the TRIS solution (0.1M,
pH 8.8) acted as reference for each measurements.
7.3.2 Reference measurements
Homogeneity of the MNP biocatalyst
Reproducibility of the MNP experiments were studied prior to the quantity and enzyme activity meas-
urements. The homogeneity of the MNP biocatalyst suspension was tested in the ammonia elimination
of l-phenylalanine in shake vials in nine parallel trials. To ensure full homogeneity, MNP biocatalyst
(2.0 mg) and PEG 4000 (2.0 mg) were suspended in TRIS buer (1 ml,0.1mM, pH 8.8) by sonication
for 30 min. In each test reaction 500 µlMNP suspension was added to 500 µlsolution of l-1a (40 mM
in TRIS buer; 0.1M, pH 8.8) then the mixtures were shaken in Eppendorf thermoshaker (at 850 rpm,
30 ◦C) for 20 min. MNPs were collected with magnet, then the product content of the supernatants
were determined at λ= 290 nm and 30 ◦Cby UV-VIS spectrophotometer (Milton Roy, Genesys 2).
Reproducibility of the MNP ll-up of the chambers
One of the 4chambers of the chip was lled with MNPs and the ammonia elimination of l-phenylalanine
by PAL was performed. The reaction was performed at 40 µl min−1ow rate on 30 ◦C. The resulted
product was measured in a ow-through absorbance cell at 290 nm. After saturation, the measurement
was repeated three times in the three consecutive chambers.
7.3.3 Chip selection and uid handling methods
All experiments were carried out by using a 4 cell MagneChip layout (see detailed in Section 5.3.1).
Loading the chambers with MNPs was done as it was described in Section 5.4.1.
7. CHAPTER. LAB-ON-A-CHIP MICROREACTOR PLATFORM 89
Fluid handling methods were chosen as follows:
•The eect of the particle size on the biocatalytic activity was investigated using the Single para-
meter uid handling method (see detailed in Section 5.5)
•Flow rate optimization, determination of kinetic parameters and substrate screening experiments
were carried out using the Multiple parameter uid handling method (see detailed in Section 5.6)
After a careful washing process (as described in Section 5.4.2) the chips were re-used.
7.3.4 Summary of MagneChip parameter settings for enzymatic reactions
Single parameter measurements
Table 7.1 summarizes the settings of the MagneChip platform which were applied during the single
parameter experiments.
Table 7.1. Summary of the settings for single parameter measurements
Experiment MNP type Sum MNP mass Substrate ow rate
Particle size dependency No1MNP600 966.6µg 6 µl min−1
Particle size dependency No2MNP250 993.28 µg 6 µl min−1
Particle size dependency No3MNP250:600 1134.36 µg 6 µl min−1
Multiple parameter measurements
Table 7.2 summarizes the settings of the MagneChip platform which were applied during the multi-
parameter experiments.
7.3.5 Optical inspection of the chambers
During the measurement the chip was optically inspected through a zooming microscope and a mono-
chrome hi-speed smart camera. Before evaluating the measurement sequence, the plan view of the chip
was stored as a reference (Πref,i). At the end of the step iof the measurement sequence, the plan view
of the chip was sampled again (Πseq,i) and it was compared to the reference as follows:
Πdiff (j, k) = (Πref (j, k),if Πref (j, k)−Πseq,i(j, k)<0
0,if Πref (j, k)−Πseq,i(j, k)≥0(7.1)
where (j, k) are the pixel coordinates of the plan view image, therefore the changes in accordance
to the reference image are indicated by white pixels. The total sum of white pixels ranged from 0(the
original image) to 11612 (empty chamber) and it was normalized in a range of 0to 1. The normalized
value is the Optical Inspection Value (OIV ), which was used as a marker for describing the changes of
the MNP layer arrangement. Therefore, the changes compared to the image of the rst cycle (reference)
were indicated by white areas during the consecutive cycles of the measurement.
7. CHAPTER. LAB-ON-A-CHIP MICROREACTOR PLATFORM 90
Table 7.2. Summary of the settings for multiple parameter measurements
Experiment Substrate
stocksa
Parameter
change
Substrate
ow rate
(µl min−1)
Washing
ow rate
(µl min−1)
No. of
cyclesb
MNP massc
(mg)
Multiple lling
of MNP load A – l-1a Const. 48.6
(for 10 min)
440
(for 5 min)
4
(in 1 h)
0.25
(in 1 cell)[Var. a)]
Single lling,
single parameter A – l-1a Const. 28.6 28.6
(for 1 h)
7
(in 14 h)
1
(in 4 cells)[Var. a)] (for 1 h)
Flow rate
optimization A – l-1a
Flow rate:
Varied 28.6
(for 30 min)
7
(in 7 h)
1
(in 4 cells)
3.6-
28.6 µl min−1
[Var b)]
Kinetic
param. calc.
A – l-1a [S]
28.6
(for 30 min)
28.6
(for 30 min)
10
(in 7 hours)
1
(in 4 cells)
3.3 mM 0.19 -
B – l-1a 43.4 mM
65 mM [Var. a)]
Saturation
of chambers
A – l-1a
28.6
(for 30 min)
300
(for 5 min)
4
(in 1 h)
0.25x4
(in 1-4 cells)
Chamber
Release
Substrate
screening
A – l-1a Substrate
48.6
(for 10 min)
440
(for 10 min)
7
(in 2.3 h)
0.5
(in 2 cells)
B – rac-1b
[Var d)]
C – rac-1c
D – rac-1d
E – rac-1e
F – rac-1f
aWhen otherwise not stated concentration of substrate was 20 mM. Control measurement in the last cycle was done by
the substrate indicated with bold.
bOne cycle constituted Reaction and Re-Initialization steps
cBiocatalyst mass was determined as described earlier [S2]
7.3.6 Numerical modeling of the chambers
The velocity distribution of the reaction cell structure was investigated by numerical methods in ANSYS
FLUENT (ANSYS Inc., USA). Stationary simulation was performed for a three dimensional pressure-
based, laminar ow problem with absolute velocity formulation. Velocity inlet was set up for the chan-
nel inlet while outow boundary for the outlet. The SIMPLE simulation method was applied with Least-
Squares Cell-Based gradient, Second Order Pressure, Second Order Upwind Momentum calculations.
The inlet ow rate was set to 28.62 µl min−1.
7.3.7 Calculation of kinetic parameters
Product concentration was calculated by taking the integral of the time dependent absorbance plot at
the product specic wavelength:
[P] = 1
ε∆tZtend
tstart
AU(t)dt (7.2)
7. CHAPTER. LAB-ON-A-CHIP MICROREACTOR PLATFORM 91
where [P]is the product concentration, AU is the absorbance unit, εis the molar extinction coecient,
∆tis the dierence of the integration limits.
Product quantity was calculated from the product concentration
P= [P]˙
Q∆t(7.3)
where ˙
Qis the inlet ow rate.
Specic biocatalytic activity was calculated by taking the integral of the saturation region of the ab-
sorbance plot
UB=
1
εRtsat
tend AU(t)dt ˙
Q
m∆t(7.4)
where mis the total mass of the MNP biocatalyst.
Reaction velocity (ν) was dened as
ν=P
∆t(7.5)
Volumetric productivity was dened as
PV=MP
Vc∆t(7.6)
where Mis the molar mass, Vcis the void volume of the reaction chambers.
Referring to Section 6.3.3, the void fraction ranges from 0.93 to 0.9. For measurements with monod-
isperse particle sizes β=.93 was used for kinetic calculations.
The insoluble enzyme constituting a lled reactor may be considered as a suspension of enzyme protein
in a volume equal to the total volume of the microreactor. Thus:
[E] = E
Vv
(7.7)
where Eis the total amount of enzymes in the chamber in moles.
Let [S0]and [St]are the initial substrate concentration and substrate concentration after the residence
time of the substrate solution in the chip.
For kinetic calculations the Lilly-Hornby model was used as it was described in Section 2.4.3. Therefore
Kmand kcat was calculated based on Eq. 2.41.
7.4 Results
7.4.1 General assumptions on the reliability of MagneChip experiments
As the MagneChip system is a quite unexploited tool so far, rst the reliability assessment of the meas-
urements was performed. A series of subsequent measurements performed by the system can be con-
sidered as reliable if all the following conditions are met:
1. the biocatalyst (in this case the MNP and the enzyme immobilized onto its the surface) is homo-
genous and its activity is unchanged during the measurements (see in Section 7.4.4)
2. independent loadings of the reaction chambers with the same biocatalyst load yield the same
biocatalytic activity (see in Section 7.4.4)
7. CHAPTER. LAB-ON-A-CHIP MICROREACTOR PLATFORM 92
3. the product of the enzyme reaction can be measured selectively in the UV-VIS range (see in
Section 7.4.8);
4. product and substrate can be completely removed through the washing steps and
5. the enzymatic activity of the MNP biocatalyst remains unchanged during the measurement (see
in Section 7.4.4)
6. and last but not least, the MNP layer in the magnetic reactors remain unharmed during the meas-
urement cycles (see in Section 7.4.2).
In order to validate the fullment of the rst group of conditions, a control measurement was done
after each experiment i.e. the rst step of the sequence was repeated in the last step under the same
conditions and the specic activity of the immobilized biocatalyst (UB) at saturation concentrations of
l-phenylalanine were compared.
7.4.2 Failures of MNP layers detected by visual inspection
The ow velocity prole of the reaction chamber aects the achievable biocatalytic activity, as the
continuous exchange of substrate and the simultaneous transport of the product molecules are provided
by the ow. Having the dense MNP layer higher viscous resistance than the empty chamber, the ow
velocity prole and the pressure drop of the chip may be aected by the accumulated particles in the
chambers. Ideally, the chambers are fully and homogeneously loaded by particles. Though the MNP
layer can be damaged due to the following reasons:
•internal drag and viscous forces at higher ow rates
•air bubble development during the reaction
•unintended air bubbles from external sources, which may enter the chip during the measurements
To investigate the eects of such unintended changes in the layer structure, the ow prole of the
chambers was analysed by a CFD model. During the actual measurements, the chambers were visually
inspected by a zooming microscope camera and the pictures were analysed o-line.
CFD analysis of the reaction chambers
The 3D model of a single reaction chamber was analysed. In a certain volume, porous cell zone condition
was applied to model the space occupied by the MNPs. Pressure drop was measured in 7ow rates
between 10 µl min−1and 80 µl min−1with and without MNPs in the chambers (see detailed in Section
5.2.1), which resulted in 0.086 kPa µL−1min−1and 0.059 kPa µL−1min−1ow specic pressure drop,
respectively.
Being laminar ow in the channels, the pressure is proportional to the velocity:
gradp=−µDv (7.8)
where µis the dynamic viscosity of the uid, pis the measured pressure and vis the ow velocity
at the inlet, respectively. Dis the viscous resistance of the porous media (MNP layer). Assuming the
parameters of water from the software’s material database, it was found that D= 2.071 ·1010m−2
CFD simulations were carried out as described in Section 7.3.6. CFD results are presented in two rel-
evant cases in Figure 7.2c, d; in the rst case the nanoparticle layer fully lled in the chamber, while in
the second case the MNP layer is partially destroyed by the unintended passage of an air bubble.
7. CHAPTER. LAB-ON-A-CHIP MICROREACTOR PLATFORM 93
Visual inspection of the reaction chambers
Visual inspection of the chambers provides a continuous measure of signicant changes (e.g. bubble
passage) of the MNP layer structure. Optical Inspection Value (OIV ) characterizes the actual arrange-
ment of the layer compared to the initial (reference) image. OIV ranges from 0to 1and it was calculated
as described in Section 7.3.5.
The eect of the air bubble passage can be seen in Figure 7.2b,d,f, resulting in OIV = 0.436. Air bubbles
usually don’t split at the channel entrance, rather pass on one side along the chamber wall. While
passing through, the bubble elongates therefore the middle section of the developed tunnel is nar-
rower (≈300µm). Numerical simulations revealed (Figure 7.2c, d, e, f) that the velocity prole became
asymmetric due to the bubble passage and the overall mass ow rate through the porous MNP layer
signicantly decreased (28.61 µl min−1to 20.66 µl min−1, roughly 72% of its original value) while the
OIV=0
OIV=0.436
1.6
0.8
1.1
0.5
0.5 1 1.5
−2
−1
0
1
2
Velocity (mm s
−1)
Axial position (mm)
18.56
0.5 1 1.5
−2
−1
0
1
2
Velocity (mm s
−1)
Axial position (mm)
28.61
μl min-1
7.95 μl min-1
1 mm
a)
b)
c)
d)
e)
f)
1st cycle
4th cycle
1st cycle
MagneChip
Input
Output
Chip holder Inspection window
20.66 μl min-1
1 cm 500 nm
Figure 7.2. MagneChip device with four microchambers loaded with MNPs, SEM image of the MNP
layer (right). a)-f): The eect of air bubble passage through the reaction cell a) photograph, before
passage b) dierence image (OIV = 0.436), after passage c) calculated ow velocity eld before and d)
after the passage (units mm s−1), e) velocity prole in the middle cross section of the chamber before
and f) after the passage
7. CHAPTER. LAB-ON-A-CHIP MICROREACTOR PLATFORM 94
remaining uid passed through the developed tunnel. In the relatively decreased velocity eld the sub-
strate concentration may fall below the saturation level (see further in Section 7.4.7). Additionally, the
passing bubble may also drift towards particles, consequently decreasing the sum mass of biocatalysts.
Therefore, the biocatalytic activity of the damaged cell decreased and the consequent measurements
were no longer reliable.
7.4.3 Assessment of the reliability of multiparameter measurements
Based on the measured values of OIV and the error of the reference measurement, the following
empirical limits were determined:
•Average 0< OIV < 0.017, the noise level of the image processing algorithm
•Average 0.017 < OIV < 0.2, implies negligible changes in the layer structure. The measurement
is accepted if the error of the reference measurement is below 5%.
•Average 0.21 < OIV < 0.35, may imply remarkable changes in the layer structure, it is the
operator’s decision whether the test can be continued or terminated. If a bubble breakthrough is
observed, the test should be terminated. Moreover, the measurement can be only accepted if the
error of the reference measurement is below 5%.
•Average OIV > 0.35, implies serious changes in the layer, the measurement is declined
Table 7.3 summarizes the assessment evaluation of the measurements.
Table 7.3. Summary of reliability assessment measures in multi parameter measurements
% Error of the OIV OIV
Experiment experimentaaverage maximum Assessment
Multiple llings, MNP load 3.1- - Accepted
Single parameter cyclic (attempt 1) 1.4 0.1 0.14 Accepted
Single parameter cyclic (attempt 2) 32.0 0.37 0.51 Declined
Flow rate optimization 3.2 0.017 0.03 Accepted
Substrate screening 1.5 0.15 0.18 Accepted
aCalculation is based on the ratio of the saturation UBvalues of the rst and last (control) measurement
7.4.4 Reference measurements
Homogeneity of the MNP biocatalyst
Biotransformation of l-phenylalanine (l-1a) to (E)-cinnamic acid (2a) by MNP biocatalyst suspen-
sion (Scheme 7.1) was performed in shake vial in three parallel independent cases and resulted in
UB= 2.91 ±0.08 µmol g−1min−1(for working conditions see Section 7.3.2) ensuring that the homo-
geneity of the MNP suspension was sucient.
7. CHAPTER. LAB-ON-A-CHIP MICROREACTOR PLATFORM 95
Reproducibility of the chamber llup
Biocatalytic activity of the biotransformation of l-1a to 2a was measured in four consecutive runs in
four independent chamber ll-ups (for working conditions see Section 7.3.2 and Table 7.2, Multiple
lling of MNP load). After the ll-up of the rst chamber, a reaction cycle was performed. After the
reaction turned into saturation, the chamber’s magnetic eld was deactivated while activated simul-
taneously in the next chamber i.e. the same particles were accumulated there.
After lling up the next chamber, the reagents were washed out by TRIS puer and then the same
enzymatic reaction was performed.
The measurement was repeated three times in the consecutive three chambers, resulting in UB=
8.01 ±0.14 µmol g−1min−1.
Concluding the results of the reference measurements, neither that the homogeneity of the MNP sus-
pension nor the lling procedure of the chambers had remarkable eect on the reproducibility of the
measurements.
The signicant dierence between the UBvalues of MNP biocatalyst in shake vials and in MagneChip
indicated increased eectivity of the biocatalysts in MagneChip device. The volumetric productivity
of the ammonia elimination of l-1a catalysed by the MNP biocatalyst were compared in the Magne-
Chip and in shake vials indicating more than three orders of magnitude higher value in MagneChip
8.82 g l−1h−1than that in the shake vial 3.13 mg l−1h−1.
Re-initialization of the chip and re-use of the MNP biocatalyst in cyclic measurements
MagneChip was lled with MNP biocatalyst and biotransformation of l-phenylalanine l-1a to (E)-
cinnamic acid 2a was performed in 7consecutive cycles, while the chip was re-initialized during the
steps by washing out the substrate and product completely (For working conditions see Table 7.2,
Single lling, single parameter). The absorbance plot in Figure 7.3 shows the concentration change of
(E)-cinnamic acid 2a at the specic wavelength of 290 nm It can be clearly seen that the chip was
successfully re-initialized in every cycle throughout the experiment and the reaction was repeated
seven times, following the same kinetics.
The product quantity was calculated in each cycle (for details see Section 7.3.7). Two independent
experiments are shown in Figure 7.4. In the rst attempt (Figure 5, blue bars) the MNP layer remained
stable throughout the measurement, resulting in an average product quantity of P= 0.12±1.5% µmol.
0 200 400 600 800
0
0.2
0.4
0.6
0.8
1
1.2
time (min)
AU
Control measurementRe initialization
Figure 7.3. Time plot of the periodic absorbance change during the cyclic measurement (attempt 1, stable
layer). The chip is re-initialized between the reaction steps (reaching zero absorbance) by washing out
substrate and product completely. The last measurement acts as a control in each experiment
7. CHAPTER. LAB-ON-A-CHIP MICROREACTOR PLATFORM 96
1 2 3 4 5 6 7
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
0.18
0.2
(SC=68)
(SC=1438)
(SC=1456)
(SC=1558)
(SC=1559)
(SC=1567)
(SC=1609)
(SC=419)
(SC=2605)
(SC=3941)
(SC=5073)
(SC=5663)
(SC=5669)
(SC=5742)
Cycle number
Product amount per cycle (µmol)
Cyclic measurement, stable layer
Cyclic measurement, bubble development
Figure 7.4. Product amount P per reaction cycle during a cyclic measurement with unmodied para-
meter settings. Blue bars indicate the attempt with unharmed MNP layer (OIV < 0.18) with excellent
reproducibility while air bubble passage damaged the layer (OIV > 0.36) during the second attempt
and demolished the biocatalytic activity (red bars)
The moderate mean value of the cell dierence score OIV = 0.1(0.14 max) also reects negligible
changes in the MNP layer. However, the signicantly higher mean OIV = 0.37 (0.51 max) in the
second attempt (Figure 7.4, red bars) indicates a damaged MNP layer structure due to an air bubble
passage. It is also remarkable that the cyclic product quantity decreased abruptly as the OIV value
elevated above the critical value. In fact, the air bubble previously stalled (3rd cycle) in the reaction cell,
nally drifted away leaving a tunnel behind in the 4th cycle (Figure 7.2b).
The above results indicated the excellent reproducibility of the reactions in a periodical sequence within
the chip when the MNP layer remained undamaged.
7.4.5 Eect of particle size on the enzymatic activity
Preliminary investigations (see detailed in Section 6.3.2) revealed that the sum volume fraction of the
loaded particles in the chamber is independent on the particle size, however it can be reasonably in-
creased by loading the chamber with an 1 : 1 mixture of the MNP250 and MNP600 particles. Total
number of particles and the average distance between the particle cores were approximated in Section
6.3.3, in Table 6.3.
According to the calculations, the total surface area increased both in the MNP250 and the mixture cases
(Table 7.4) compared to the reference case (MNP600 ). Although larger surface usually allows higher en-
zyme loading during the immobilization and thus higher specic biocatalytic activity, in this study the
15% enzyme to MNP mass ratio was kept constant – and below the maximum enzyme binding capa-
city – during the PAL immobilization. Consequently, in this study the larger total surface area of the
MNP250 -lled chamber compared to the MNP600 -lled one does not mean more active enzyme. There-
7. CHAPTER. LAB-ON-A-CHIP MICROREACTOR PLATFORM 97
Table 7.4. Eect of particle size on the enzymatic activity
Relative changes in
Particle S to V ratioaCell capacity bbiocatal. activity
MNP600 1 1 1
MNP250 2.5 1.02 1.46
MNP250:600 2.06 1.17 1.86
aSurface to Volume ratio, calculation is based on the method described in Section 6.3.3
bDetermined earlier in Section 6.3.2
fore, dierences in biocatalytic activity can be expected only due to changes of transport limitations or
in case of lling with MNPs of mixed sizes due to the elevated quantity of trapped particles.
Measurement results are summarized in Fig 7.5. In fact, the MNP600 -lled chambers yielded the lowest
nal concentration of product as indicated by the lowest specic absorbance (AU = 1.07, at 295 nm)
at the chip outlet. Filling the chip by MNP250 resulted in an increase of the measured absorbance by
46% (AU = 1.56, at 295 nm). The dierence between the MNP250 - and the MNP600 -lled chamber
cannot be attributed to the dierences in total surface area (as they contained the same lling mass and
therefore the same enzyme amount).
The results of Section -6.3.3 suggested that the packing of the particles was sparse (see ) i.e. substrate
and product molecules were transported through microtunnels formed between the particles.
The major dierence can stem from the remarkably smaller average microtunnel diameters (see Table
6.3, Average distance) between the particles within the MNP250 -lled cell as compared to the MNP600
-lled one. This can result in shortened diusion path and therefore better mass transport. An addi-
tional 40% increment was achieved by using the 1 : 1 particle mixture, which was obviously resulted as
a synergy of the higher enzyme content (17%) due to the higher cell capacity and enhanced transport
phenomena due to the small average microtunnel diameter.
1.07
1.56
1.99
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
1.6
1.8
2.0
2.2
MNP600
MNP250
Mixed
Absorbance Unit
Figure 7.5. Specic absorbance at 295nm of cinnamon acid at the chip outlet using MNP600 , MNP250
and 1:1mixture of the two kind of particles in the chip
7. CHAPTER. LAB-ON-A-CHIP MICROREACTOR PLATFORM 98
7.4.6 Inuence of the ow rate on biotransformation with 1a
MagneChip was lled with MNP biocatalyst and biotransformations of l-1a to 2a at various ow rates
were performed in 7consecutive cycles, while the chip was re-initialized at the end of each cycle and
a new substrate ow rate was set (for working condition see Table 7.2, Flow rate optimization). The
lowest ow rate was chosen to 3.6µl min−1and increased in the forthcoming steps up to 28.6µl min−1
(Figure 7.6). The negligible dierence in specic biocatalytic activity (UB) between the reference and
control measurements was found to be only 3%, which suggests that the shear forces develop even at
the highest chosen ow rate (28.6µl min−1) had not caused irreversible changes on the enzyme activ-
ity.
The reaction velocity was calculated for each cycles (for details see Section 7.3.7). By increasing the
ow rate, the calculated reaction velocity was increased too, until turned into saturation at about
25 µl min−1.
The uneven ow velocity eld inside the reaction cells (Figure 7.2c) implies that the middle section of
the cell reached the saturation velocity even at lower ow rates while the whole cell area turned into
saturation at roughly 25 µl min−1inlet ow rate.
0 5 10 15 20 25 30
0
0.5
1
1.5
2
Flow rate (µL min 1)
Reaction velocity (nmol min 1)
Figure 7.6. Flow rate dependency of the reaction rate in MagneChip under the transformation of l-1a
to 2a acid by MNP-PAL. Saturation value was achieved at 25 µl min−1
7.4.7 Calculation of the kinetic parameters of the transformation of l-1a to
2a
MagneChip was lled with MNP biocatalyst and biotransformations of l-1a to 2a at various concen-
trations of l-1a [S]0were performed in 10 consecutive cycles, while the chip was re-initialized at the
end of each cycle and a new substrate concentration was set (Figure 7.6, top; for working conditions
see Table 7.2, Kinetic param. calc.). It was found that the reaction followed the rst order kinetics up
to [S] =3 mmand saturated roughly at [S] =20 mm.
The linear tting method proposed by Lilly, et al. [75] was applied for the calculation of the kinetic con-
stants of the biotransformation of l-1a to 2a (Figure 7.7, bottom). The values of the kinetic constants
are summarized in 7.5. It was found that the apparent Kmvalue was reasonably smaller in MagneChip
(2.5 mm) than in shake vial (9.1 mm). Turnover number (kcat) and specicity constant (kcat/Km) were
determined also for both reaction modes. While in the shake vial the turnover number was somewhat
higher (3.2·10−2s−1) than the in chip (2.8·10−2s−1), the specicity constant turned out to be signic-
antly higher in chip (11.3 s−1m−1) as compared to the shake vial (3.5 s−1m−1). This may be attributed
to the smaller Km value in the MagneChip indicating signicant contribution of diusion eects to the
higher apparent Kmvalue in shake vial.
7. CHAPTER. LAB-ON-A-CHIP MICROREACTOR PLATFORM 99
Table 7.5. Kinetic constants in biotransformation of l-1a to 2a with MNP in shake vial and in Magne-
Chip
Kinetic parameter MagneChip Shake vial
Km[mm] 2.5 9.1
kcat [s−1] 2.8·10−23.2·10−2
kcat/Km[ s−1m−1] 11.3 3.5
−0.03 −0.025 −0.02 −0.015 −0.01 −0.005 0
log(1−P)
0.01
0.02
0.03
0.04
0.05
0.06
0.07
Ps0
y=2.48x+0.0704 (R2=0.976)
0 5 10 15 20 25 30 35 40 45
0
0.5
1
1.5
2
2.5
[S] (mM)
Reeaction velocity (nmol min−)
a) b)
*
*
Figure 7.7. a) Dependency of the substrate concentration on reaction velocity in MagneChip for the
transformation of l-1a to 2a by MNP biocatalyst. Saturation concentration was reached at 20 mmb)
Linear t based on the Lilly-Hornby model to determine Km(resulting in Km=2.5 mm)
1234
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
Number of chambers active
Absorbance unit (AU)
0.76
0.56 (− 1 %)
0.36 (− 3 %)
0.18 (−3%)
Figure 7.8. Contribution of the chambers to the overall product quantity. Initially 4 chambers were
active then the biocatalyst content of 1,2,3 chambers were released
7. CHAPTER. LAB-ON-A-CHIP MICROREACTOR PLATFORM 100
Once the substrate concentration had been reached its saturation value, the contribution of the
chambers to the overall biocatalytic activity was studied. Four chambers were lled by MNPs and the
biotransformation of l-1a to 2a was carried out until the reaction saturated (For working conditions see
Table 7.2, Saturation of chambers). After saturation, the magnetic eld of the last downstream chamber
was deactivated therefore the particles were released. With this new conguration, the same reaction
was carried out. The process was repeated three times. Figure 7.8 shows the measured absorbance (AU)
when four, three, two, one chambers were active. The results suggest that being the reaction saturated,
the contribution of the chambers to the overall biocatalytic activity was nearly equal. Only a slight
decrement (≈3%) could be observed compared to the expected value, in case of two and one active
chambers.
7.4.8 Substrate screening with MNP biocatalyst in the MagneChip system
The screening experiments were performed with the natural substrate (l-1a), four dierent phenylalan-
ine analogues as known substrates of PcPAL (rac-1c-f) and 4-bromophenylalanine (rac-1b) which has
never been tested as substrate for PcPAL (Figure 7.1). For the substrate screening, the same stock of
MNP biocatalyst was as for the previous experiments presented.
First, the extinction coecients (Table 7.6) of the corresponding elimination products (2a-f, Scheme
7.1) were determined at selected wavelengths (Figure 7.9: preferably at wavelengths resulting in high
absorbance of the acrylic acid 2 while practically zero of the amino acid 1a-f) using the ow-cell spec-
trometer of the MagneChip platform. The linear regressions with the measured molar concentration
dependence of absorbance values were rather well tted.
200 250 300 350
0
0.2
0.4
0.6
0.8
1
Wavelength (nm)
Absorbance unit (AU)
L 1a
2a
200 250 300 350
0
0.2
0.4
0.6
0.8
1
Wavelength (nm)
Absorbance unit (AU)
rac 1b
2b
200 250 300 350
0
0.2
0.4
0.6
0.8
1
Wavelength (nm)
Absorbance unit (AU)
rac 1c
2c
200 250 300 350
0
0.2
0.4
0.6
0.8
1
Wavelength (nm)
Absorbance unit (AU)
rac 1d
2d
200 250 300 350
0
0.2
0.4
0.6
0.8
1
Wavelength (nm)
Absorbance unit (AU)
rac 1e
2e
200 250 300 350
0
0.2
0.4
0.6
0.8
1
Wavelength (nm)
Absorbance unit (AU)
rac 1f
2f
Figure 7.9. Absorbance spectra of l-1a and rac-1b-f (green lines) and the spectra of the correspond-
ing acrylic acids 2a-f (dotted purple lines). Extinction coecients were determined at the chosen
wavelengths indicated by blue markers
7. CHAPTER. LAB-ON-A-CHIP MICROREACTOR PLATFORM 101
Table 7.6. Extinction coecients of the investigated acrylic acids 2a-f
Acrylic acid Wavelength (nm) Extinction coecient (M−1cm) Linear regression coecient
2a 290 8800 0.991
2b 300 10 200 0.998
2c 300 7919 0.988
2d 280 14 721 0.993
2e 280 9172 0.991
2f 280 15 327 0.998
The substrate screening experiments were performed with single chip loading and the dierent sub-
strate solutions own through the chip according to a predened sequence (for working condition see
Table 7.2, Substrate screening). The intensive washing procedure intended to reach the zero absorbance
unit at every relevant wavelengths. In the rst cycle the ammonia elimination of rac-1a, was measured
which is the natural substrate of PAL. This case was chosen as reference point of the comparison of the
other elimination reaction of PAL substrates. Surprisingly, at high ow rate in the MagneChip device,
higher biocatalytic activities (UB) were observed for four of the unnatural substrates (rac-1b,c,e,f),
than for the natural substrate l-phenylalanine l-1a (Figure 7.10).
Noteworthy, all the four unnatural substrates showed higher biocatalytic activity (UB) than l-phenylalanine
l-1a contained slightly more electron-withdrawing aromatic moieties than the phenyl group. This sig-
nicant deviation from the productivity ranks observed with homogenous PcPAL so far1 may be due to
the reduced contribution of the reverse reaction (equilibrium eect) to the apparent forward reaction
rates in the continuous-ow system at high ow rates.
7.5 Conclusion
A microuidic device was characterized which consisted four microliter volume reaction chambers
lled with PcPAL-coated magnetic nanoparticles (MNP) as functional biocatalyst. In the chambers of
1.00
2.69
4.27
0.92
3.67
2.73
0.98
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
2a
2b
2c
2d
2e
2f
2a*
U
B
/
U
B,2a
Figure 7.10. Comparison of the biocatalytic activity of PcPAL enzyme immobilized on MNPs with sub-
strates l-1a and rac-1b-f in MagneChip system ([S] = 20 mM, ow rate: 48.6µl min−1). * Control
measurement
7. CHAPTER. LAB-ON-A-CHIP MICROREACTOR PLATFORM 102
the chip MNPs could be captured or released with the aid of moveable external permanent magnets.
The experiments proved that the microuidic device was capable of cyclical operation by re-using the
biocatalyst.
It was shown that neither the homogeneity of the MNP suspension nor the lling procedure of the
chambers had remarkable eect on the reproducibility. The lling procedure as well as the reactions
in the chambers were found highly reproducible (1.5% variation around the mean within cyclic oper-
ations a 14htime-frame, if no MNP layer-degradation happened).
Biotransformation of l-1a to 2a was carried out in a ow-through microreactor arrangement using dif-
ferent sized particles. It was found that the biocatalytic activity increased as the particle size decreased
due to the better mass transport of substrate and product molecules around the particles.
It was shown earlier (see in Section 6.3.2) that lling the chamber by an 1 : 1 mixture of dierent
sized particles resulted in higher amount of biocatalysts in the chamber, compared to the cells lled
with uniformly sized MNPs. Biotransformation experiments were also resulted in a signicantly large
enhancement in biocatalytic activity (by 46% to MNP250 and by 86% to MNP600 ), indicating also better
mass transport situation for the enzymatic catalysis due to the decreased average microtunnel diamet-
ers.
During cyclic operations, optical inspection of the MNP layer and repeated control measurements
oered measures to assess the reliability of the process. The visible degradation of the layer was de-
scribed by the arbitrary OIV value. The eect of bubble passage was investigated experimentally and
by numerical methods as well. It was found that air bubbles had serious eect on reproducibility which
was also reected in increased OIV values. The deviation of the biocatalytic activity of rst and the
last cycles was also evaluated in each series of experiments (the dierence was only in the 1.4%−3.6%
range in the accepted cases).
Saturation ow rate and substrate concentration of the in-chip transformation with l-phenylalanine
l-1a to (E)-cinnamic acid 2a were found to be 25 µl min−1and 20 mm, respectively.
The MagneChip system equipped with an on-line UV-VIS ow cell could be applied for automated
activity screening of four known (rac-1c-f) and a novel (rac-1b) unnatural substrate of PcPAL com-
pared to the natural substrate l-phenylalanine by a cyclic test series using the same MNP lling.
Our results proved that the MagneChip microuidic device is a reliable, reproducible and ecient tool
which was capable of fast, reliable and fully automated screening of PcPAL substrates using minimal
solvent (≈500 µL) and biocatalyst (≈1 mg MNP) amounts for a test compound. Compared to shake
vial, the volumetric productivity of the MNP biocatalyst in the chip exceeded the one of the shake vial
by more than three orders of magnitude.
7.6 Summary of the scientic results
Thesis 3 Characterization of packed bed microreactors
I developed a Lab-on-a-Chip packed bed microreactor platform aiming the investigation of the reaction kin-
etics of enzymes immobilized onto magnetic nanoparticles. The platform enables the real time, optical (UV-
VIS) in-line following of enzyme catalysed biotransformations taking place in the reaction chambers.[S3],[B1]
1. I developed a method to perform multi-parameter enzyme catalysed reactions applicable in Lab-
on-a-Chip devices. The method is based on subsequent measurement cycles where every cycle
represent a dierent parameter setting. During the cycles the nanoparticles are magnetically
trapped in the chambers so the biocatalyst load remains unchanged. The rst phase of the cyc-
lic operation is the release of the reagents by washing. The subsequent phase is the enzymatic
reaction with the actual modied parameters.
7. CHAPTER. LAB-ON-A-CHIP MICROREACTOR PLATFORM 103
2. I investigated the eect of the chamber lling on the reproducibility of the measurements. I
found that the lling of the chambers with a given suspension of enzyme coated nanoparticles
was resulted in more than 98% of reproducibility related to the immobilized biocatalytic activity.
3. I experimentally characterized the repeatability of the enzymatic biotransformation from one
parametrically identical cycle to another. I found that the repeatability of these measurements
was more than 98% related to the immobilized biocatalytic activity.
4. I developed a method for the optical inspection of the particle lled chambers. Two subsequent
cycles were treated to be comparable only if the change of their normalized optical inspection
values fell below 0.35 on the range of 0−1. I validated by measurements that every measure-
ment cycles within a single experiment can be treated to be independent, if the dierence of the
measured biocatalytic activity between the rst and last cycle was under 5%.
Thesis 4 Kinetic characterization of microreactors
I characterized the reactions catalysed by the phenylalanine-ammonia-lyase (PAL) enzyme immobilized
onto magnetic nanoparticles by using a multiple reaction chamber Lab-on-a-Chip device. Using this meas-
urement set-up, I experimentally characterized the kinetic properties of the PAL-catalyzed reactions in
micro-ow systems.[S3],[B1]
1. I found that the biocatalytic activity was decreased by decreasing the particle diameter. For the
experiments 600 nm and 250 nm diameter particles were used. I experimentally validated that
the biocatalytic activity can be further increased by using a binary mixture
(250 nm+600 nm) of the particles.
2. I characterized the biocatalytic activity of the ammonia elimination reaction of phenylalanine by
PAL enzyme. Furthermore I characterized the eect of the substrate ow rate on the biocatalytic
activity in the same reaction by multiparameter measurements. I validated that by increasing
ow rate the biocatalytic activity increased until a saturation value of 20 µl min−1. I found that
the viscous forces of higher ow rates do not have an eect on the enzymatic activity or on the
reproducibility of the measurements.
3. I experimentally determined the kinetic constants of the above reaction conducted in ow mi-
croreactors at a xed substrate ow rate of 20 µl min−1. I found that the reaction kinetic followed
the Lilly-Hornby model and the biocatalytic activity was found to be 2.75 fold higher, the Km
value was found to be 3.64 fold lower and the specicity constant was found to be 3.22 fold
higher than ones of the same reaction with identical parameters conducted in traditional shake
vial.
Chapter 8
Utilization of the results
8.1 Compact model for the nutrient transport in blood capil-
lary vessels
LRBC LPlasma Flow direction
RBC RBC
Figure 8.1. Illustration of the mass circulation in the plasma droplet formed in a capillary vessel between
two red blood cells
Developed on the basis of the thermal compact model for droplet microreactors (see detailed in
chapter 4), Márton et al. proposed a reduced order model for the description of mass transfer in capillary
vessels. The model idea lies upon the similarity of the two phase droplet ow and the ow of red blood
cells (RBC) in capillary vessels (Figure 8.1).
Modeling capillaries is among the deeply studied issues in many branches of physiology since with
a reliable capillary model the capillary-organ interactions can be better understood. The tissues with
higher rate of metabolism require higher density of capillaries, therefore, understanding the capillary
mass transfer is very important in cases of certain diseases such as cancer where the ever increasing
malignant tumours require more and more nutrients. In droplet ow, internal mass circulations are
developed which may aect the nutrient transport through the capillary wall.
In the proposed model, the volume was discretised like in the thermal compact model and the moving
mesh technique was utilized also. Unlike in the thermal compact model, the internal mass circulation
of the droplets was also considered by introducing a mass ow buer [C4].
8.2 Finding a new operation mechanism of PAL enzyme
By a reaction performed in the MagneChip device (see detailed in chapter 7), it was rst demonstrated
that PAL can catalyse the ammonia elimination from the acyclic dl-propargylglycine (PG) to yield (E)-
pent-2-ene-4-ynoate indicating new opportunities to extend the MIO-enzyme toolbox towards acyclic
substrates. Deamination of PG, being acyclic, cannot involve a Friedel-Crafts-type attack at an aromatic
ring. Therefore a novel operation mechanism of the PAL enzyme was proved.
104
8. CHAPTER. UTILIZATION OF THE RESULTS 105
Figure 8.2. Ammonia elimination from dl-propargylglycine in MagneChip lled with PAL immobilized
on MNPs and equipped with in-line UV-Vis detector. The progress of the reaction was followed by full
UV-spectra.
MagneChip, lled by PAL-MNPs, was used for the microscale biotransformation of dl-propargylglycine
in sodium carbonate-buered D2O. The device enabled to detect the formation of (E)-pent-2-en-4-
ynoate at 242 nm and to produce measurable quantities of the product for recording 1H-NMR spectra
without any work-up. Besides the signicant increase of the UV-signal at 242 nm (up to AU = 1.2) in
the in-line UV-cell (Figure 8.2), the appearance of olen hydrogen signals in the 1H-NMR spectrum of
the reaction mixture indicated unambiguously the formation of (E)-pent-2-en-4-ynoate. On the other
hand, emergence of the UV signal at 274 nm during the process indicated the formation of further
by-product(s) apart from (E)-pent-2-en-4-ynoate (Figure 8.2). [S4]
Scientic Publications
[S1] Ferenc Ender, Márton Németh, Péter Pálovics, Andras Drozdy, and András Poppe. Thermal com-
pact modeling approach of droplet microreactor based Lab-on-a-Chip devices. Microelectronics
Journal, 45(12):1786–1794, 2014.
[S2] Ferenc Ender, Diána Weiser, András Vitéz, Gábor Sallai, Márton Németh, and László Poppe. In-situ
measurement of magnetic nanoparticle quantity in a microuidic device. Microsystem Technolo-
gies, 21(12), 2015.
[S3] Ferenc Ender, Diána Weiser, Botond Nagy, László Csaba Bencze, Csaba Paizs, Péter Pálovics, and
László Poppe. Microuidic multiple cell chip reactor lled with enzyme-coated magnetic nano-
particles - An ecient and exible novel tool for enzyme catalyzed biotransformation. Journal of
Flow Chemistry, Accepted, 2015.
[S4] Diána Weiser, László Csaba Bencze, Gergely Bánóczi, Ferenc Ender, Róbert Kiss, Eszter Kókai,
András Szilágyi, Beáta G Vértessy, Ödön Farkas, Csaba Paizs, and László Poppe. Phenylalanine
Ammonia-Lyase-Catalyzed Deamination of an Acyclic Amino Acid: Enzyme Mechanistic Studies
Aided by a Novel Microreactor Filled with Magnetic Nanoparticles. ChemBioChem, 16(16):2283–
2288, 2015.
106
Conference Proceedings
[C1] Ferenc Ender. Modelling of heat transfer in Taylor ow in microchannels. In Design, Test, Integ-
ration and Packaging of MEMS/MOEMS (DTIP), 2012 Symposium on, pages 164–170. IEEE, 2012.
[C2] Ferenc Ender and Gusztav Hantos. Modelling of heat transfer in microdroplets as microreactors.
In 18th International Workshop on THERMal INvestigation of ICs and Systems, 2012.
[C3] Ferenc Ender, András Vitez, Gábor Sallai, Diána Weiser, and Márton Nemeth. In-situ measure-
ment of nanoparticle quantity in microchambers. In Design, Test, Integration and Packaging of
MEMS/MOEMS (DTIP), 2015 Symposium on, pages 1–6, apr 2015.
[C4] Márton Németh, Ferenc Ender, and András Poppe. Modeling of Circular Mass Transport of Nutri-
ents in Capillary Vessels Using Microuidic Approach. In Proceedings of First European Biomed-
ical Engineering Conference for Young Investigators: ENCY2015, pages 102–105, Budapest, 2015.
Springer Singapore.
107
Edited Books
[B1] Ferenc Ender, Diána Weiser, and László Poppe. Microuidic multiple cell chip reactor lled with
enzyme-coated magnetic nanoparticles. In Margarita Stoytcheva, editor, Lab on a Chip (in press).
InTech Open, Rijeka, 2016.
108
Unrelated Publications
[N1] Ferenc Ender, Gusztav Hantos, Dirk Schweitzer, and Peter Gabor Szabo. Thermal characterization
of multichip structures. In Thermal Investigations of ICs and Systems (THERMINIC), 2013 19th
International Workshop on, pages 319–322, sep 2013.
[N2] Ferenc Ender, Hunor Santha, and Vladimir Szekely. Optimization of microuidic ow sensors for
dierent ow ranges by FEM simulation. In Electronics Technology (ISSE), 2010 33rd International
Spring Seminar on, pages 308–313, may 2010.
[N3] Ferenc Ender, Hunor Santha, and Vladimir Szekely. Flow sensor for microuidic applications;
Based on standard PWB technology. In Electronics Technology, 2009. ISSE 2009. 32nd International
Spring Seminar on, pages 1–6, may 2009.
[N4] Ferenc Ender and Vladimir Szekely. Thermal transfer impedance variations by forced convect-
ive heat transfer in microchannels. In Design, Test, Integration and Packaging of MEMS/MOEMS
(DTIP), 2012 Symposium on, pages 119–124, apr 2012.
[N5] Ferenc Ender, Gusztav Hantos, Andras Vitez, and Diana Weiser. In-situ thermal conductivity
measurement of magnetic nanoparticle layers in Lab-on-a-Chip devices. In Thermal Investigations
of ICs and Systems (THERMINIC), 2014 20th International Workshop on, pages 1–6, sep 2014.
[N6] Márton Németh, Ferenc Ender, and András Poppe. Heat and mass transfer reduced order mod-
eling approach of droplet microreactor based Lab-on-a-Chip devices. Microelectronics Journal,
46(12):1152–1161, 2015.
[N7] Dirk Schweitzer, Ferenc Ender, Gusztáv Hantos, and Péter G. Szabó. Thermal transient charac-
terization of semiconductor devices with multiple heat sources - Fundamentals for a new thermal
standard. Microelectronics Journal, 46(2):174–182, 2015.
109
Bibliography
[1] Gordon Moore. Cramming More Components Onto Integrated Circuits, Electronics,(38) 8, 1965.
[2] Elaine R. Mardis. The impact of next-generation sequencing technology on genetics. Trends in
Genetics, 24(3):133–141, 2008.
[3] Jonathan M. Rothberg, Wolfgang Hinz, Todd M. Rearick, Jonathan Schultz, William Mileski, Mel
Davey, John H. Leamon, Kim Johnson, Mark J. Milgrew, Matthew Edwards, Jeremy Hoon, Jan F. Si-
mons, David Marran, Jason W. Myers, John F. Davidson, Annika Branting, John R. Nobile, Bernard P.
Puc, David Light, Travis A. Clark, Martin Huber, Jerey T. Branciforte, Isaac B. Stoner, Simon E.
Cawley, Michael Lyons, Yutao Fu, Nils Homer, Marina Sedova, Xin Miao, Brian Reed, Jerey Sa-
bina, Erika Feierstein, Michelle Schorn, Mohammad Alanjary, Eileen Dimalanta, Devin Dressman,
Rachel Kasinskas, Tanya Sokolsky, Jacqueline A. Fidanza, Eugeni Namsaraev, Kevin J. McKernan,
Alan Williams, G. Thomas Roth, and James Bustillo. An integrated semiconductor device enabling
non-optical genome sequencing. Nature, 475(7356):348–352, 2011.
[4] Rajesh Munirathinam, Jurriaan Huskens, and Willem Verboom. Supported Catalysis in
Continuous-Flow Microreactors. Advanced Synthesis & Catalysis, 357(6):1093–1123, 2015.
[5] Thomas Wirth. Microreactors in Organic Chemistry and Catalysis. John Wiley & Sons, 2013.
[6] Shakeel Ahmed Ansari and Qayyum Husain. Potential applications of enzymes immobilized on/in
nano materials: A review. Biotechnology Advances, 30(3):512–523, 2012.
[7] Wolfgang Ehrfeld, Volker Hessel, and Verena Haverkamp. Microreactors. WILEY-VCH Verlag
GmbH, Weinheim, rst edition, 2000.
[8] Andreas Weiler and Matthias Junkers. Using Microreactors in Chemical Synthesis: Batch Process
versus Continuous Flow. Sigma Aldrich ChemFIles, 9(4), 2009.
[9] Travis R Besanger, Yang Chen, Anil K Deisingh, Richard Hodgson, Wen Jin, Stanislas Mayer, Mi-
chael a Brook, and John D Brennan. Screening of inhibitors using enzymes entrapped in sol-gel-
derived materials. Analytical chemistry, 75(10):2382–91, 2003.
[10] Pawel L. Urban, David M. Goodall, and Neil C. Bruce. Enzymatic microreactors in chemical ana-
lysis and kinetic studies. Biotechnology Advances, 24(1):42–57, 2006.
[11] David J Beebe, Glennys a Mensing, and Glenn M Walker. Physics and applications of microuidics
in biology. Annual review of biomedical engineering, 4(November):261–286, 2002.
[12] Carl Hansen and Stephen R. Quake. Microuidics in structural biology: Smaller, faster... better.
Current Opinion in Structural Biology, 13(5):538–544, 2003.
110
BIBLIOGRAPHY 111
[13] Vamsee K Pamula and Krishnendu Chakrabarty. Cooling of Integrated Circuits Using Droplet-
based Microuidics. In Proceedings of the 13th ACM Great Lakes Symposium on VLSI, GLSVLSI ’03,
pages 84–87, New York, NY, USA, 2003. ACM.
[14] J.P. Brody, P. Yager, R.E. Goldstein, and R.H. Austin. Biotechnology at low Reynolds numbers.
Biophysical Journal, 71(6):3430–3441, 1996.
[15] M. Kays, W. and E. Crawford, M. Convective Heat and Mass Transfer, 1993.
[16] Gad-el-Hak Mohamed. MEMS Introduction and Fundamentals. Taylor & Francis, 2005.
[17] Sung Yang, Akif Ündar, and Jerey D. Zahn. A microuidic device for continuous, real time blood
plasma separation. Lab on a Chip, 6(7):871, 2006.
[18] Nam-Trung Nguyen and Zhigang Wu. Micromixers—a review. Journal of Micromechanics and
Microengineering, 15(2):R1–R16, feb 2005.
[19] S. Hardt, K. S. Drese, V. Hessel, and F. Schönfeld. Passive micromixers for applications in the
microreactor andµTAS elds. Microuidics and Nanouidics, 1(2):108–118, apr 2005.
[20] Chia Hsiang Chen, Yi Lu, Mandy L Y Sin, Kathleen E Mach, Donna D Zhang, Vincent Gau, Joseph C
Liao, and Pak Kin Wong. Antimicrobial Susceptibility Testing Using High Surface-to-Volume Ratio
Microchannels. Anal. Chem., 82(3):1012–1019, 2010.
[21] Babak Ziaie, Antonio Baldi, Ming Lei, Yuandong Gu, and Ronald a. Siegel. Hard and soft mi-
cromachining for BioMEMS: Review of techniques and examples of applications in microuidics
and drug delivery. Advanced Drug Delivery Reviews, 56(2):145–172, 2004.
[22] Pradeep Dixit, Nay Lin, Jianmin Miao, Wai Kwan Wong, and Teo Kiat Choon. Silicon nanopillars
based 3D stacked microchannel heat sinks concept for enhanced heat dissipation applications in
MEMS packaging. Sensors and Actuators, A: Physical, 141(2):685–694, 2008.
[23] G. Takács, P.G. Szabó, and Gy. Bognár. Thermal management in System-on-Package structures by
applying microscale heat sink. Part I: Consideration of the appropriate channel length of microscale
heat sink(s). Microelectronics Journal, pages 1–6, 2015.
[24] Che-Hsin Lin, Gwo-Bin Lee, Bao-Wen Chang, and Guan-Liang Chang. A new fabrication process
for ultra-thick microuidic microstructures utilizing SU-8 photoresist. Journal of Micromechanics
and Microengineering, 12(5):590, 2002.
[25] Jinwen Zhou, Amanda Vera Ellis, and Nicolas Hans Voelcker. Recent developments in PDMS
surface modication for microuidic devices. Electrophoresis, 31(1):2–16, 2010.
[26] Teruo Fujii. PDMS-based microuidic devices for biomedical applications. Microelectronic Engin-
eering, 61-62:907–914, 2002.
[27] Shantanu Bhattacharya, Arindom Datta, Jordan M. Berg, and Shubhra Gangopadhyay. Studies on
surface wettability of poly(dimethyl) siloxane (PDMS) and glass under oxygen-plasma treatment
and correlation with bond strength. Journal of Microelectromechanical Systems, 14(3):590–597, 2005.
[28] Johan G. Alauzun, Stuart Young, Renita D’Souza, Lina Liu, Michael a. Brook, and Heather D. Shear-
down. Biocompatible, hyaluronic acid modied silicone elastomers. Biomaterials, 31(13):3471–3478,
2010.
BIBLIOGRAPHY 112
[29] UsamaM. Attia, Silvia Marson, and JereyR. Alcock. Micro-injection moulding of polymer micro-
uidic devices. Microuidics and Nanouidics, 7(1):1–28, 2009.
[30] Holger Becker and Ulf Heim. Hot embossing as a method for the fabrication of polymer high
aspect ratio structures. Sensors and Actuators, A: Physical, 83(1):130–135, 2000.
[31] Henning Klank, Jrg P. Kutter, and Oliver Geschke. CO2-laser micromachining and back-end pro-
cessing for rapid production of PMMA-based microuidic systems. Lab on a Chip, 2(4):242, 2002.
[32] J Michael Ramsey, Stephen C Jacobson, and Michael R Knapp. Microfabricated chemical meas-
urement systems. Nat Med, 1(10):1093–1095, oct 1995.
[33] Krishnendu Chakrabarty and Jun Zeng. Design automation for microuidics-based biochips. ACM
Journal on Emerging Technologies in Computing Systems, 1(3):186–223, 2005.
[34] Francesca Sapuppo, Florinda Schembri, Luigi Fortuna, and Maide Bucolo. Microuidic circuits
and systems. IEEE Circuits and Systems Magazine, 9(3):6–19, 2009.
[35] T. Zhang, K. Chakrabarty, and R.B. Fair. Behavioral Modeling and Performance Evaluation of
Microelectrouidics-Based PCR Systems Using SystemC. IEEE Transactions on Computer-Aided
Design of Integrated Circuits and Systems, 23(6):843–858, 2004.
[36] Michael Schlegel, Fouad Bennini, Jan E Mehner, Göran Herrmann, Dietmar Müller, and Wolfram
Dötzel. Analyzing and Simulation of MEMS in VHDL-AMS Based on Reduced-Order FE Models.
IEEE Sensors Journal, 5(5):1019–1026, 2005.
[37] Fei Su and Krishnendu Chakrabarty. High-level synthesis of digital microuidic biochips. ACM
Journal on Emerging Technologies in Computing Systems, 3(4):1–32, 2008.
[38] Fei Su, Krishnendu Chakrabarty, and Richard B. Fair. Microuidics-based biochips: Techno-
logy issues, implementation platforms, and design-automation challenges. IEEE Transactions on
Computer-Aided Design of Integrated Circuits and Systems, 25(2):211–223, 2006.
[39] Yi Wang, Qiao Lin, and Tamal Mukherjee. Composable behavioral models and schematic-based
simulation of electrokinetic lab-on-a-chip systems. Design Automation Methods and Tools for
Microuidics-Based Biochips, 25(2):109–142, 2006.
[40] Evgenii B. Rudnyi and Jan G. Korvink. Review: Automatic Model Reduction for Transient Simu-
lation of MEMS-based Devices. Sensors Update, 11(1):3–33, 2002.
[41] Fabien Jousse, Guoping Lian, Ruth Janes, and John Melrose. Compact model for multi-phase
liquid-liquid ows in micro-uidic devices. Lab on a chip, 5(6):646–56, jun 2005.
[42] Arvind Sridhar, Alessandro Vincenzi, Martino Ruggiero, Thomas Brunschwiler, and David
Atienza. 3D-ICE: Fast compact transient thermal modeling for 3D ICs with inter-tier liquid cooling.
2010 IEEE/ACM International Conference on Computer-Aided Design (ICCAD), pages 463–470, nov
2010.
[43] Azadeh N. Asadolahi, Raghvendra Gupta, David F. Fletcher, and Brian S. Haynes. CFD approaches
for the simulation of hydrodynamics and heat transfer in Taylor ow. Chemical Engineering Science,
66(22):5575–5584, nov 2011.
BIBLIOGRAPHY 113
[44] Raghvendra Gupta, David F. Fletcher, and Brian S. Haynes. CFD modelling of ow and heat
transfer in the Taylor ow regime. Chemical Engineering Science, 65(6):2094–2107, mar 2010.
[45] T Taha and Z Cui. CFD modelling of slug ow inside square capillaries. Chemical Engineering
Science, 61(2):665–675, jan 2006.
[46] Subash S Jayawardena, Vemuri Balakotaiah, and Larry C Witte. Flow pattern transition maps for
microgravity two-phase ows. AIChE Journal, 43(6):1637–1640, 1997.
[47] M Kreutzer, F Kapteijn, J Moulijn, and J Heiszwolf. Multiphase monolith reactors: Chemical reac-
tion engineering of segmented ow in microchannels. Chemical Engineering Science, 60(22):5895–
5916, nov 2005.
[48] Axel Günther and Klavs F Jensen. Multiphase microuidics: from ow characteristics to chemical
and materials synthesis. Lab on a chip, 6(12):1487–503, dec 2006.
[49] T.C. Thulasidas, M.a. Abraham, and R.L. Cerro. Flow patterns in liquid slugs during bubble-train
ow inside capillaries. Chemical Engineering Science, 52(17):2947–2962, sep 1997.
[50] Ashleigh B Theberge, Fabienne Courtois, Yolanda Schaerli, Martin Fischlechner, Chris Abell,
Florian Hollfelder, and Wilhelm T S Huck. Microdroplets in microuidics: an evolving plat-
form for discoveries in chemistry and biology. Angewandte Chemie (International ed. in English),
49(34):5846–68, aug 2010.
[51] M T Kreutzer, W Wei, F Kapteijn, J A Moulijn, and J J Heiszwolf. Pressure Drop of Taylor Flow
in Capillaries: Impact of Slug Length. In S G Kandlikar, editor, First International Conference on
Microchannels and Minichannels, pages 519–526, 2003.
[52] Koji Fukagata, Nobuhide Kasagi, Poychat Ua-arayaporn, and Takehiro Himeno. Numerical sim-
ulation of gas–liquid two-phase ow and convective heat transfer in a micro tube. International
Journal of Heat and Fluid Flow, 28(1):72–82, feb 2007.
[53] Helen Song, Joshua D Tice, and Rustem F Ismagilov. A Microuidic System for Controlling Reac-
tion Networks in Time. Angewandte Chemie, 115(7):792–796, 2003.
[54] Helen Song, Delai L. Chen, and Rustem F. Ismagilov. Reactions in Droplets in Microuidic Chan-
nels. Angewandte Chemie (International ed. in English), 45(44):7336–7356, 2007.
[55] Linas Mazutis, Jean-Christophe Baret, Patrick Treacy, Yousr Skhiri, Ali Fallah Araghi, Michael
Ryckelynck, Valerie Taly, and Andrew D Griths. Multi-step microuidic droplet processing: kin-
etic analysis of an in vitro translated enzyme. Lab Chip, 9(20):2902–2908, 2009.
[56] Chun Xia Zhao, Lizhong He, Shi Zhang Qiao, and Anton P J Middelberg. Nanoparticle synthesis
in microreactors. Chemical Engineering Science, 66(7):1463–1479, 2011.
[57] Albert Tsung Hsi Hsieh, Patrick Jen Hao Pan, and Abraham Phillip Lee. Rapid label-free DNA ana-
lysis in picoliter microuidic droplets using FRET probes. Microuidics and Nanouidics, 6(3):391–
401, 2009.
[58] Yolanda Schaerli, Robert C Wootton, Tom Robinson, Viktor Stein, Christopher Dunsby, Mark A A
Neil, Paul M W French, Andrew J DeMello, Chris Abell, and Florian Hollfelder. Continuous-Flow
Polymerase Chain Reaction of Single-Copy DNA in Microuidic Microdroplets. Analytical Chem-
istry, 81(1):302–306, jan 2009.
BIBLIOGRAPHY 114
[59] Hao Gu, Michel H G Duits, and Frieder Mugele. Droplets formation and merging in two-phase
ow microuidics. International journal of molecular sciences, 12(4):2572–97, jan 2011.
[60] G I Taylor. Deposition of a viscous uid on the wall of a tube. Journal of Fluid Mechanics,
10(02):161–165, 1961.
[61] Azadeh N. Asadolahi, Raghvendra Gupta, Sharon S.Y. Leung, David F. Fletcher, and Brian S.
Haynes. Validation of a CFD model of Taylor ow hydrodynamics and heat transfer. Chemical
Engineering Science, 69(1):541–552, feb 2012.
[62] Ralph Lindken and Massimiliano Rossi. Micro-Particle Image Velocimetry ([small micro]PIV):
Recent developments{,} applications{,} and guidelines. Lab Chip, 9(17):2551–2567, 2009.
[63] Donata M. Fries and Philipp Rudolf von Rohr. Liquid mixing in gas–liquid two-phase ow by
meandering microchannels. Chemical Engineering Science, 64(6):1326–1335, mar 2009.
[64] Axel Günther, Manish Jhunjhunwala, Martina Thalmann, Martin a Schmidt, and Klavs F Jensen.
Micromixing of miscible liquids in segmented gas-liquid ow. Langmuir : the ACS journal of surfaces
and colloids, 21(4):1547–55, feb 2005.
[65] Sharon S.Y. Leung, Yang Liu, David F. Fletcher, and Brian S. Haynes. Heat transfer in well-
characterised Taylor ow. Chemical Engineering Science, 65(24):6379–6388, dec 2010.
[66] Gian Luca Morini. Scaling Eects for Liquid Flows in Microchannels. Heat Transfer Engineering,
27(4):64–73, may 2006.
[67] G Maranzana. Mini- and micro-channels: inuence of axial conduction in the walls. International
Journal of Heat and Mass Transfer, 47(17-18):3993–4004, aug 2004.
[68] G.P. Celata, M. Cumo, V. Marconi, S.J. McPhail, and G. Zummo. Microtube liquid single-phase
heat transfer in laminar ow. International Journal of Heat and Mass Transfer, 49(19-20):3538–3546,
sep 2006.
[69] G P Celata. Single and Two Phase Flow Heat Transfer in Micropipes. In Meer van de Steenhoven,
editor, Proceedings of the 5th European Thermal-Sciences Conference, pages 1–15, Eindhoven, 2008.
[70] Károly Simonyi. Villamosságtan. Akadémiai Kiadó, Budapest, 1983.
[71] Hans Bisswanger. Enzyme Kinetics. Wiley-vch Verlag Isbn, Weinheim, 2008.
[72] Boutros Sarrouh, Tulio Marcos Santos, Anderson Miyoshi, Rodrigo Dias, and Vasco Azevedo. Up-
To-Date Insight on Industrial Enzymes Applications and Global Market. Journal of bioprocessing &
biotechniques, S4(002):1–10, 2012.
[73] Sumitra Datta, L. Rene Christena, and Yamuna Rani Sriramulu Rajaram. Enzyme immobilization:
an overview on techniques and support materials. 3 Biotech, pages 1–9, 2012.
[74] Sebastian Bartsch and Uwe T Bornscheuer. A Single Residue Inuences the Reaction Mechanism
of Ammonia Lyases and Mutases. Angewandte Chemie International Edition, 48(18):3362–3365, 2009.
[75] M D Lilly, W E Hornby, and E M Crook. The kinetics of carboxymethylcellulose–cin in packed
beds. The Biochemical journal, 100(3):718–723, 1966.
BIBLIOGRAPHY 115
[76] László Poppe, Csaba Paizs, Klaudia Kovács, Florin-Dan Irimie, and Beáta Vértessy. Preparation of
Unnatural Amino Acids with Ammonia-Lyases and 2,3-Aminomutases. In Loredano Pollegioni and
Stefano Servi, editors, Unnatural Amino Acids SE - 1, volume 794 of Methods in Molecular Biology,
pages 3–19. Humana Press, 2012.
[77] D S Hodgins. Yeast phenylalanine ammonia-lyase. Purication, properties, and the identication
of catalytically essential dehydroalanine. The Journal of biological chemistry, 246(9):2977–2985, may
1971.
[78] Christineh N. Sarkissian and Alejandra Gámez. Phenylalanine ammonia lyase, enzyme substitu-
tion therapy for phenylketonuria, where are we now? Molecular Genetics and Metabolism, 86:22–26,
2005.
[79] A. Hughes. Amino Acids, Peptides and Proteins in Organic Chemistry Volume I. Weinheim Germany:
Wiley VCH, 2009.
[80] Chen Fu Lin, Gwo Bin Lee, Chih Hao Wang, Huei Huang Lee, Wei Yin Liao, and Tse Chuan Chou.
Microuidic pH-sensing chips integrated with pneumatic uid-control devices. Biosensors and Bio-
electronics, 21:1468–1475, 2006.
[81] Kamrul Islam, You-Cheol Jang, Rohit Chand, Sandeep Kumar Jha, Hyun Ho Lee, and Yong-Sang
Kim. Microuidic Biosensor for <I>β</I>-Amyloid(1-42) Detection Using Cyclic Voltammetry.
Journal of Nanoscience and Nanotechnology, 11(7):5657–5662, 2011.
[82] Julaluk Noiphung, Temsiri Songjaroen, Wijitar Dungchai, Charles S. Henry, Orawon Chailapakul,
and Wanida Laiwattanapaisal. Electrochemical detection of glucose from whole blood using paper-
based microuidic devices. Analytica Chimica Acta, 788:39–45, 2013.
[83] Wijitar Dungchai, Orawon Chailapakul, and Charles S Henry. Electrochemical Detection for
Paper-Based Microuidics. Analytical Chemistry, 81(14):5821–5826, jul 2009.
[84] Shujuan Liu, Yunfeng Gu, Rudolph B Le Roux, Sinead M Matthews, Daniel Bratton, Kamran Yunus,
Adrian C Fisher, and Wilhelm T S Huck. The electrochemical detection of droplets in microuidic
devices. Lab Chip, 8(11):1937–1942, 2008.
[85] Jin-Woo Choi, Kwang W Oh, Jennifer H Thomas, William R Heineman, H Brian Halsall, Joseph H
Nevin, Arthur J Helmicki, H Thurman Henderson, and Chong H Ahn. An integrated microuidic
biochemical detection system for protein analysis with magnetic bead-based sampling capabilities.
Lab Chip, 2(1):27–30, 2002.
[86] B. S. Kwak, B. S. Kim, H. H. Cho, J. S. Park, and H. I. Jung. Dual thermopile integrated microuidic
calorimeter for biochemical thermodynamics. Microuidics and Nanouidics, 5(2):255–262, dec
2007.
[87] Bin Wang and Qiao Lin. A MEMS dierential scanning calorimeter for thermodynamic char-
acterization of biomolecules. 2011 IEEE 24th International Conference on Micro Electro Mechanical
Systems, pages 821–824, jan 2011.
[88] T. Adrega and a.W. van Herwaarden. Chip calorimeter for thermal characterization of bio-
chemical solutions. Sensors and Actuators A: Physical, 167(2):354–358, jun 2011.
BIBLIOGRAPHY 116
[89] Wonhee Lee. Development and applications of chip calorimeters as novel biosensors. Nanobio-
sensors in Disease Diagnosis, page 17, apr 2012.
[90] Johannes Lerchner, Thomas Maskow, and Gert Wolf. Chip calorimetry and its use for biochem-
ical and cell biological investigations. Chemical Engineering and Processing: Process Intensication,
47(6):991–999, jun 2008.
[91] Frank B. Myers and Luke P. Lee. Innovations in optical microuidic technologies for point-of-care
diagnostics. Lab on a Chip, 8(12):2015, 2008.
[92] Peng Liu, Tae Seok Seo, Nathaniel Beyor, Kyoung Jin Shin, James R. Scherer, and Richard a. Math-
ies. Integrated portable polymerase chain reaction-capillary electrophoresis microsystem for rapid
forensic short tandem repeat typing. Analytical Chemistry, 79(5):1881–1889, 2007.
[93] A Bhattacharyya and C M Klapperich. Design and testing of a disposable microuidic chemilu-
minescent immunoassay for disease biomarkers in human serum samples. Biomedical Microdevices,
9(2):245–251, 2007.
[94] Mark L. Adams, Markus Enzelberger, Stephen Quake, and Axel Scherer. Microuidic integration
on detector arrays for absorption and uorescence micro-spectrometers. Sensors and Actuators, A:
Physical, 104(1):25–31, 2003.
[95] Rebecca J Jackman, Tamara M Floyd, Reza Ghodssi, Martin A Schmidt, and Klavs F Jensen. Mi-
crouidic systems with on-line UV detection fabricated in photodenable epoxy. Journal of Mi-
cromechanics and Microengineering, 11(3):263, 2001.
[96] M Faraji, Y Yamini, and M Rezaee. Magnetic nanoparticles: Synthesis, stabilization, functionaliz-
ation, characterization, and applications. Journal of the Iranian Chemical Society, 7(1):1–37, 2010.
[97] Abolfazl Akbarzadeh, Mohammad Samiei, and Soodabeh Davaran. Magnetic nanoparticles :
preparation , physical properties , and applications in biomedicine. Nanoscale Research Letters,
7(144):1–13, 2012.
[98] U. Jeong, X. Teng, Y. Wang, H. Yang, and Y. Xia. Superparamagnetic Colloids: Controlled Synthesis
and Niche Applications. Advanced Materials, 19(1):33–60, 2007.
[99] Siming Wang, Ping Su, Jun Huang, Jingwei Wu, and Yi Yang. Magnetic nanoparticles coated
with immobilized alkaline phosphatase for enzymolysis and enzyme inhibition assays. Journal of
Materials Chemistry B, 1:1749, 2013.
[100] Gavin W Marshall and Toby S Hudson. Dense Binary Sphere Packings. Contrib. Algebra Geom.,
51(2):337–344, 2010.
[101] Wei Huang, Student Member, Shougata Ghosh, Siva Velusamy, Karthik Sankaranarayanan, Kevin
Skadron, Senior Member, and Mircea R Stan. HotSpot : A Compact Thermal Modeling Methodology
for Early-Stage VLSI Design. IEEE Transactions on VLSI Systems, 14(5):501–513, 2006.
[102] Jin Xie, Chengkuo Lee, and Hanhua Feng. Design, Fabrication, and Characterization of CMOS
MEMS-Based Thermoelectric Power Generators. Microelectromechanical Systems, Journal of,
19(2):317–324, apr 2010.
BIBLIOGRAPHY 117
[103] László Pohl, Ernő Kollár, András Poppe, and Zsolt Kohári. Nonlinear electro-thermal modeling
and eld-simulation of {OLEDs} for lighting applications I: Algorithmic fundamentals. Microelec-
tronics Journal, 43(9):624–632, 2012.
[104] Seungkyung Park, Yi Zhang, Shin Lin, Tza-Huei Wang, and Samuel Yang. Advances in micro-
uidic PCR for point-of-care infectious disease diagnostics. Biotechnology advances, 29(6):830–839,
2011.
[105] J. Lerchner, a. Wolf, G. Wolf, and I. Fernandez. Chip calorimeters for the investigation of liquid
phase reactions: Design rules. Thermochimica Acta, 446(1-2):168–175, jul 2006.
[106] Janak Singh and Mayang Ekaputri. PCR thermal management in an integrated Lab on Chip.
Journal of Physics: Conference Series, 34:222–227, apr 2006.
[107] Ryan Tewhey, Jason B Warner, Masakazu Nakano, Brian Libby, Martina Medkova, Patricia H
David, Steve K Kotsopoulos, Michael L Samuels, J Brian Hutchison, and Jonathan W Larson.
Microdroplet-based PCR enrichment for large-scale targeted sequencing. Nature biotechnology,
27(11):1025–1031, 2009.
[108] N Reginald Beer, Benjamin J Hindson, Elizabeth K Wheeler, Sara B Hall, Klint A Rose, Ian M
Kennedy, and Bill W Colston. On-chip, real-time, single-copy polymerase chain reaction in picoliter
droplets. Analytical Chemistry, 79(22):8471–8475, 2007.
[109] Wei Wang Yang, Zhi-Xin Li, Rong Luo, Shu-Hai Lü, Ai-Dong Xu, and Yong-Jun. Droplet-based
micro oscillating-ow PCR chip. Journal of Micromechanics and Microengineering, 15(8):1369, 2005.
[110] Wonhee Lee, Warren Fon, Blake W Axelrod, and Michael L Roukes. High-sensitivity micro-
uidic calorimeters for biological and chemical applications. Proceedings of the National Academy
of Sciences, 106(36):15225–15230, 2009.
[111] Cindy Hany, Helene Lebrun, Christophe Pradere, Jean Toutain, and Jean-Christophe Batsale.
Thermal analysis of chemical reaction with a continuous microuidic calorimeter. Chemical En-
gineering Journal, 160(3):814–822, jun 2010.
[112] J. Lerchner, a. Wolf, G. Wolf, V. Baier, E. Kessler, M. Nietzsch, and M. Krügel. A new micro-uid
chip calorimeter for biochemical applications. Thermochimica Acta, 445(2):144–150, jun 2006.
[113] N Reginald Beer, Elizabeth K Wheeler, Lorenna Lee-Houghton, Nicholas Watkins, Shanavaz Nas-
arabadi, Nicole Hebert, Patrick Leung, Don W Arnold, Christopher G Bailey, and Bill W Colston.
On-chip single-copy real-time reverse-transcription PCR in isolated picoliter droplets. Analytical
Chemistry, 80(6):1854–1858, 2008.
[114] Lian Zhang, Jae-mo Koo, Linan Jiang, Mehdi Asheghi, Kenneth E Goodson, Associate Member,
Juan G Santiago, and Thomas W Kenny. Measurements and Modeling of Two-Phase Flow in Mi-
crochannels With Nearly Constant Heat Flux Boundary Conditions. Measurement, 11(1):12–19,
2002.
[115] Jing Shang and Xiaohu Gao. Nanoparticle counting: towards accurate determination of the molar
concentration. Chemical Society reviews, 2014.
BIBLIOGRAPHY 118
[116] John Eveness, Janice Kiely, Peter Hawkins, Patrick Wraith, and Richard Luxton. Evaluation
of paramagnetic particles for use in a resonant coil magnetometer based magneto-immunoassay.
Sensors and Actuators, B: Chemical, 139(2):538–542, 2009.
[117] Jing Xu, Haibin Yang, Wuyou Fu, Kai Du, Yongming Sui, Jiuju Chen, Yi Zeng, Minghui Li, and
Guangtian Zou. Preparation and magnetic properties of magnetite nanoparticles by sol-gel method.
Journal of Magnetism and Magnetic Materials, 309:307–311, 2007.
[118] C B Kriz, K Rådevik, and D Kriz. Magnetic permeability measurements in bioanalysis and bio-
sensors. Analytical chemistry, 68(11):1966–70, 1996.
[119] Julie Richardson, Andrew Hill, Richard Luxton, and Peter Hawkins. A novel measuring system
for the determination of paramagnetic particle labels for use in magneto-immunoassays. Biosensors
and Bioelectronics, 16(9-12):1127–1132, 2001.
[120] Elham Sharif, Janice Kiely, and Richard Luxton. Novel immunoassay technique for rapid meas-
urement of intracellular proteins using paramagnetic particles. Journal of Immunological Methods,
388(1-2):78–85, 2013.
[121] Satoru Enpuku, Keiji and Minotani, Tadashi and Gima, Takemitsu and Kuroki, Yukinori and Itoh,
Yuzuru and Yamashita, Makiko and Katakura, Yoshinori and Kuhara. Detection of magnetic nan-
oparticles with superconducting quantum interference device (SQUID) magnetometer and applic-
ation to immunoassays. Japanese journal of applied physics, 38(10A), 1999.
[122] Mamas I Prodromidis. Impedimetric immunosensors—A review. Electrochimica Acta,
55(14):4227–4233, 2010.
[123] Ana Fernandes, Carla Duarte, Filipe Cardoso, Ricardo Bexiga, Susana Cardoso, and Paulo Freitas.
Lab-on-Chip Cytometry Based on Magnetoresistive Sensors for Bacteria Detection in Milk. Sensors,
14(8):15496–15524, 2014.
[124] J Loureiro, P Z Andrade, S Cardoso, C L Da Silva, J M Cabral, and P P Freitas. Magnetoresistive
chip cytometer. Lab on a Chip, 11(13):2255–2261, 2011.
[125] Thomas P Burg, Michel Godin, Scott M Knudsen, Wenjiang Shen, Greg Carlson, John S Foster,
Ken Babcock, and Scott R Manalis. Weighing of biomolecules, single cells and single nanoparticles
in uid. Nature, 446(7139):1066–9, apr 2007.
[126] M M Bradford. A rapid and sensitive method for the quantitation of microgram quantities of
protein utilizing the principle of protein-dye binding. Analytical biochemistry, 72:248–254, 1976.
[127] Xiaole Mao, Ahmad Ahsan Nawaz, Sz-Chin Steven Lin, Michael Ian Lapsley, Yanhui Zhao, J Philip
McCoy, Wak S El-Deiry, and Tony Jun Huang. An integrated, multiparametric ow cytometry chip
using “microuidic drifting” based three-dimensional hydrodynamic focusing. Biomicrouidics,
6(2):24113, 2012.
[128] Taek Dong Chung and Hee Chan Kim. Recent advances in miniaturized microuidic ow cyto-
metry for clinical use. Electrophoresis, 28(24):4511–20, dec 2007.
[129] Nicholas N Watkins, Umer Hassan, Gregory Damhorst, Hengkan Ni, Awais Vaid, William Rodrig-
uez, and Rashid Bashir. Microuidic CD4+ and CD8+ T Lymphocyte Counters for Point-of-Care
HIV Diagnostics Using Whole Blood. Science translational medicine, 5(214):214ra170, dec 2013.
BIBLIOGRAPHY 119
[130] Canjun Mu, Feiling Zhang, Zhiyi Zhang, Min Lin, and Xudong Cao. Highly ecient dual-channel
cytometric-detection of micron-sized particles in microuidic device. Sensors and Actuators B:
Chemical, 151(2):402–409, jan 2011.
[131] Elham Sharif, Janice Kiely, Patrick Wraith, and Richard Luxton. The dual role of paramagnetic
particles for integrated lysis and measurement in a rapid immunoassay for intracellular proteins.
IEEE Transactions on Biomedical Engineering, 60(5):1209–1216, 2013.
[132] Ping He, Gillian Greenway, and Stephen J. Haswell. Development of enzyme immobilized mono-
lith micro-reactors integrated with microuidic electrochemical cell for the evaluation of enzyme
kinetics. Microuidics and Nanouidics, 8(5):565–573, 2010.
[133] Gi Hun Seong, Jinseok Heo, and Richard M Crooks. Measurement of Enzyme Kinetics Using a
Continuous-Flow Microuidic System. Analytical Chemistry, 75(13):5206–5212, 2003.
[134] Kang-Yi Lien, Wan-Chi Lee, Huan-Yao Lei, and Gwo-Bin Lee. Integrated reverse transcription
polymerase chain reaction systems for virus detection. Biosensors & bioelectronics, 22(8):1739–1748,
2007.
[135] X Mu, J Qiao, L Qi, P Dong, and H Ma. Poly(2-vinyl-4,4-dimethylazlactone)-functionalized mag-
netic nanoparticles as carriers for enzyme immobilization and its application. ACS Applied Materials
and Interfaces, 6(23):21346–21354, 2014.
[136] Yuya Asanomi, Hiroshi Yamaguchi, Masaya Miyazaki, and Hideaki Maeda. Enzyme-Immobilized
Microuidic Process Reactors. Molecules, 16(12):6041–6059, 2011.
[137] Nicole Pamme. On-chip bioanalysis with magnetic particles. Current Opinion in Chemical Bio-
logy, 16(3-4):436–443, 2012.
[138] Caterina G C M Netto, Henrique E. Toma, and Leandro H. Andrade. Superparamagnetic nano-
particles as versatile carriers and supporting materials for enzymes. Journal of Molecular Catalysis
B: Enzymatic, 85-86:71–92, 2013.
[139] Klaus Buchholz, Volker Kasche, and Uwe Theo Bornscheuer. Biocatalysts and enzyme technology.
John Wiley & Sons, 2012.
[140] Damien Webb and Timothy F. Jamison. Continuous ow multi-step organic synthesis. Chemical
Science, 1(6):675, 2010.
[141] S Nisha, S Arun Karthick, and N Gobi. A Review on Methods, Application and Properties of
Immobilized Enzyme. Chemical Science Reiew and Letters, 1(3):148–155, 2012.
[142] Bin Ding, Moran Wang, Xianfeng Wang, Jianyong Yu, and Gang Sun. Electrospun nanomaterials
for ultrasensitive sensors. Materials Today, 13(11):16–27, 2010.
[143] Yingyi Liu, Dayong Yang, Tao Yu, and Xingyu Jiang. Incorporation of electrospun nanobrous
PVDF membranes into a microuidic chip assembled by PDMS and scotch tape for immunoassays.
Electrophoresis, 30(18):3269–3275, 2009.
[144] Moser Yves. Dynamic Actuation of Magnetic Beads for Immunoassays on-chip. PhD thesis, École
Polytechnique Fédérale De Lausanne, 2010.
BIBLIOGRAPHY 120
[145] Jaephil Do and Chong H Ahn. A polymer lab-on-a-chip for magnetic immunoassay with on-chip
sampling and detection capabilities. Lab on a chip, 8(4):542–549, 2008.
[146] Yan Li, Xiuqing Xu, Bo Yan, Chunhui Deng, Wenjia Yu, Pengyuan Yang, and Xiangmin Zhang.
Microchip reactor packed with metal-ion chelated magnetic silica microspheres for highly ecient
proteolysis. Journal of proteome research, 6(6):2367–75, 2007.
[147] Marcela Slovakova, Nicolas Minc, Zuzana Bilkova, Claire Smadja, Wolfgang Faigle, Claus Füt-
terer, Myriam Taverna, and Jean-Louis Viovy. Use of self assembled magnetic beads for on-chip
protein digestion. Lab on a chip, 5(9):935, 2005.
[148] Matthew B. Kerby, Robert S. Legge, and Anubhav Tripathi. Measurements of kinetic parameters
in a microuidic reactor. Analytical Chemistry, 78(24):8273–8280, 2006.
[149] Francisco E Torres, Peter Kuhn, Dirk De Bruyker, Alan G Bell, Michal V Wolkin, Eric Peeters,
James R Williamson, Gregory B Anderson, Gregory P Schmitz, Michael I Recht, and Others. En-
thalpy arrays. Proceedings of the National Academy of Sciences of the United States of America,
101(26):9517–9522, 2004.
[150] Francisco E Torres, Michael I Recht, Joseph E Coyle, Richard H Bruce, and Glyn Williams. Higher
throughput calorimetry: opportunities, approaches and challenges. Current opinion in structural
biology, 20(5):598–605, oct 2010.
[151] Shihan Zhang, Yongqi Lu, and Xinhuai Ye. Catalytic behavior of carbonic anhydrase enzyme
immobilized onto nonporous silica nanoparticles for enhancing CO2 absorption into a carbonate
solution. International Journal of Greenhouse Gas Control, 13:17–25, mar 2013.
[152] Martin A M Gijs, Frédéric Lacharme, and Ulrike Lehmann. Microuidic Applications of Magnetic
Particles for Biological Analysis and Catalysis. Chemical Reviews, 110(3):1518–1563, 2010.
[153] Péter Pálovics. Chipméretű laboratórium mikroreaktor cellájának szimulációs vizsgálata. In
Diplomaterv. BME, Budapest, 2014.
[154] Rosa Mondragon, J. Enrique Julia, Antonio Barba, and Juan Carlos Jarque. Determination of
the packing fraction of silica nanoparticles from the rheological and viscoelastic measurements of
nanouids. Chemical Engineering Science, 80:119–127, 2012.
[155] Hong Deng, Xiaolin Li, Qing Peng, Xun Wang, Jinping Chen, and Yadong Li. Monodisperse Mag-
netic Single-Crystal Ferrite Microspheres. Angewandte Chemie International Edition, 44(18):2782–
2785, 2005.
[156] Jin Sheng, Lei Zhang, Jianping Lei, and Huangxian Ju. Fabrication of tunable microreactor with
enzyme modied magnetic nanoparticles for microuidic electrochemical detection of glucose.
Analytica Chimica Acta, 709:41–46, 2012.
[157] Min S. Wang, Joshua C. Black, Michelle K. Knowles, and Scott M. Reed. C-reactive protein (CRP)
aptamer binds to monomeric but not pentameric form of CRP. Analytical and Bioanalytical Chem-
istry, 401(4):1309–1318, 2011.
[158] Yoon Seok Song, Hyun Yong Shin, Jin Young Lee, Chulhwan Park, and Seung Wook Kim. β-
Galactosidase-immobilised microreactor fabricated using a novel technique for enzyme immobil-
isation and its application for continuous synthesis of lactulose. Food Chemistry, 133(3):611–617,
2012.
BIBLIOGRAPHY 121
[159] Martin a M Gijs and Ulrike Lehmann. Microuidic Applications of Magnetic Particles for Biolo-
gical Analysis and Catalysis. Chem. Rev., pages 1518–1563, 2010.
