university of groningen microfluidic tools for …...cover design: margaryta a. ianovska layout...

University of Groningen

Microfluidic tools for multidimensional liquid chromatographyIanovska, Margaryta

IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite fromit. Please check the document version below.

Document VersionPublisher's PDF, also known as Version of record

Publication date:2018

Link to publication in University of Groningen/UMCG research database

Citation for published version (APA):Ianovska, M. (2018). Microfluidic tools for multidimensional liquid chromatography. University of Groningen.

CopyrightOther than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of theauthor(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).

Take-down policyIf you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediatelyand investigate your claim.

Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons thenumber of authors shown on this cover page is limited to 10 maximum.

Download date: 13-01-2021

https://www.rug.nl/research/portal/en/publications/microfluidic-tools-for-multidimensional-liquid-chromatography(43ecb46e-8815-44af-8c1a-31c39c8c13a4).html

https://www.rug.nl/research/portal/en/publications/microfluidic-tools-for-multidimensional-liquid-chromatography(43ecb46e-8815-44af-8c1a-31c39c8c13a4).html

Microfluidic Tools

for

Multidimensional Liquid Chromatography

Margaryta A. Ianovska

Paranimphs:

Viktoria Starokozko

Nashwa Soliman

Cover design: Margaryta A. Ianovska

Layout design: Margaryta A. Ianovska

Printed by: Ipskamp Printing

The research presented in this thesis was financially supported by The Netherlands Organization for

Scientific Research (NWO) in the framework of the Technology Area-COAST program, project no.

(053.21.102) (HYPERformance LC) and the University of Groningen. Printing of this thesis was

supported by the University of Groningen, Faculty of Science and Engineering and the University

Library.

© Margaryta A. Ianovska, 2018

ISBN (printed version): 978-94-034-1221-4

ISBN (digital version): 978-94-034-1220-7

No parts of this thesis may be reproduced or transmitted in any form or by any means, electronic or

mechanical, including photocopying, recording or any information storage and retrieval system,

without permission of the author.

Microfluidic Tools for

Multidimensional Liquid

Chromatography

PhD thesis

to obtain the degree of PhD at the

University of Groningen

on the authority of the

Rector Magnificus Prof. E. Sterken

and in accordance with

the decision by the College of Deans.

This thesis will be defended in public on

Friday 7 December 2018 at 16:15 hours

by

Margaryta Ianovska

born on 21 June 1990 in Kiev, Ukraine

Supervisors Prof. E.M.J. Verpoorte Prof. P.J. Schoenmakers

Assessment committee Prof. S. Eeltink Prof. R.A.H. Peters Prof. R.C. Chiechi

Can one ever predict the influence of his work on the world?

It might glow in the future as stars in the sky.

(The cover to this thesis was inspired by the most recognized painting in the

history of Western culture - The Starry Night by Vincent van Gogh, who

referred to the painting as a "failure").

At the same time, the cover represents the mixing process that was observed by

the author in the channel with herringbone grooves.

Table of Contents

Introduction ........................................................................... 11

Introduction ........................................................................................................... 13

Scope of the thesis ................................................................................................. 30

References ..............................................................................................................32

Novel micromixers based on chaotic advection and their application —a review ........................................................... 37

Abstract ................................................................................................................. 38

1. Introduction ...................................................................................................... 39

2. Theory ............................................................................................................... 42

3. Passive micromixers based on chaotic advection............................................. 46

Table 1. Micromixers based on chaotic advection. ................................................... 59

4. Application of the passive micromixers based on chaotic advection ............... 62

5. Discussion .......................................................................................................... 76

6. Conclusions ....................................................................................................... 80

References .............................................................................................................. 81

Development of small-volume, microfluidic chaotic mixers for future application in two-dimensional liquid chromatography .................................................................... 85

Abstract ................................................................................................................. 86

Introduction ........................................................................................................... 87

Material and Methods ........................................................................................... 89

Results and Discussion ......................................................................................... 96

Supplementary information ................................................................................ 106

Conclusions ........................................................................................................... 111

Acknowledgements ............................................................................................... 111

References ............................................................................................................ 112

Fabrication of a pressure-resistant microfluidic mixer in fused silica using Selective Laser-Induced Etching ........................ 115

Abstract ................................................................................................................ 116

Introduction .......................................................................................................... 117

Material and Methods .......................................................................................... 121

Results and Discussion ........................................................................................ 126

Evaluation of mixing performance ......................................................................... 134

Conclusions .......................................................................................................... 139

Acknowledgements .............................................................................................. 140

References ............................................................................................................ 141

Microfluidic micromixer as a tool to overcome solvent incompatibilities in two-dimensional liquid chromatography ................................................................... 143

Introduction ......................................................................................................... 145

Material and Methods .......................................................................................... 150

Results and Discussion ........................................................................................ 157

Supplementary Information ................................................................................ 168

Conclusions .......................................................................................................... 172

Acknowledgements .............................................................................................. 173

References ............................................................................................................ 174

General discussion, conclusions and future perspectives .... 177

General discussion, conclusions and future perspectives ................................... 179

References ............................................................................................................ 185

Samenvatting ....................................................................... 187

Acknowledgements .............................................................. 193

Curriculum Vitae .................................................................199

List of publications: ............................................................. 200

Chapter I

Introduction

Introduction

13

Introduction

Over the last centuries the world has undergone breathtaking changes that are

unprecedented in human history. Humanity has never been so far advanced in knowledge, in

so many different areas, and over such a wide spectrum. We have progressed through the age

of technology to the information and digital age. Driven by insatiable curiosity, we are gaining

a lot of insight into biological processes in living systems, including the functions of the human

body at different levels, which holds great promise for the improvement of public health.

Moreover, with limited resources on a limited planet, we have become aware of the influence

we have on our planet, and the environmental problems that arise as a consequence of our

actions.

Further progress in improving the quality of human life on the one hand and controlling

environmental change on the other is not possible without reliable and high-resolution

instrumental methods. Today in many fields, such as environmental and food analysis, and the

analysis of biological material in proteomics and metabolomics, scientists have to deal with

samples that may contain literally thousands of constituents. This has generated a need for

improving existing, or even creating new, methods for analysis of larger numbers of

compounds in more complex samples in the most comprehensive way.

Nowadays, one of the most powerful separation techniques is liquid chromatography

(LC), in which a dissolved sample flows through a column packed with a solid adsorbent. Each

constituent of the mixture interacts differently with the column material, which leads

eventually to different elution times from the column and, therefore, the separation of

components. Liquid chromatography has assumed a very important position among the modern

analytical separation techniques, mainly for the analyses of samples with low complexity

(active pharmaceutical ingredients, food industry, etc.). However, LC often does not provide

sufficient resolving power (the ability to distinguish different compounds from each other) for

the separation of complex samples. For example, biological samples of interest in proteomics

and metabolomics can contain thousands of components; in a more extreme example, samples

that are analyzed in experiments investigating expression of the human proteome in different

tissues may contain up to 200,000 components after digestion with trypsin.1,2 However, there

are few examples to be found in the literature where calculated theoretical peak capacities (the

maximum number of peaks that can be separated per single run) reach even a few thousand

Intr

od

uct

ion

Microfluidic Tools for Multidimensional Liquid Chromatography

14

peaks.3–5 Under typical circumstances in an LC separation, theoretical peak capacities tend to

be on the order of 200, whereas the actual separation will yield a more modest number of peaks.

Approximately 50 - 75 peaks are usually recorded, which means that observed peaks are likely

to contain a number of overlapping single-componentpeaks.2,6

A good alternative for improving the separation power of LC is to develop a multi-

dimensional system that combines two or more separation mechanisms in series to significantly

improve resolving power and achieve peak capacities that are greater than in one-dimensional

LC.

Basic principles of two-dimensional liquid chromatography

In 1978, Erni and Frei7 introduced a two-dimensional liquid-chromatography system,

constructed by coupling two LC columns, for the separation of complex plant extracts. Their

intention was to first separate samples on a gel permeation stationary phase, followed by the

complete transfer of all the eluate coming from the first column or “first dimension” (1D) to a

second column (second dimension, 2D) containing a reverse-phase stationary phase. The 1D

eluate was collected in fractions and re-injected to the 2D for a second separation. Though both

the separation and the set-up were imperfect, this early work is regarded as “pioneering” in the

area of comprehensive LC, as the idea was to subject all of the 1D eluate to a second

separation.8 A certain level of automation was incorporated into this first 2D-LC system in the

form of an 8-port switching valve equipped with two sampling loops of identical volume for

sampling and storing eluate for transfer between dimensions. In 1990, Jorgenson and co-

workers9 utilized comprehensive 2D-LC for protein separations. However, two-dimensional

gas chromatography (2D-GC)10 had reached a higher degree of maturity by the late 1990s6 than

2D-LC, due to inherent advantages such as faster separation as a result of faster diffusion and

mass transfer, easier coupling between dimensions and no need for thermal re-equilibration

between successive second-dimension runs.

Two-dimensional LC has found its practical applications now, two decades later, as the

required instrumentation nowadays is much more advanced. State-of-the-art 2D-LC exists in

three forms: off-line, on-line “heart-cutting” (LC-LC) and comprehensive (LC×LC).11 In the

off-line techniques, the fractions that elute from the first column first are collected off-line,

with subsequent re-injection into the second dimension,8 This approach does not require

additional equipment, as the effluent portions are simply collected, but it does take a lot of

Introduction

15

work and time. In the on-line version of 2D-LC, two dimensions are coupled via an interface

(also called “modulator”). In many cases, it consists of a switching valve with a varying number

of ports and two identical sampling loops, connected to the valve. By switching back and forth

between loops, these loops first collect, store and later re-inject the effluent from the first

column (1D) into the second dimension (2D), where the second separation takes place. The

process of sampling, storing and re-injection of small 1D effluent fractions into the second

dimension is called modulation. The separation of fractions injected onto the 2D should be fast

and in most cases complete before the next transfer occurs. In the “heart-cutting” mode, only

one or a few chosen 1D fractions are collected and immediately re-injected into the second

column for separation, while the remaining effluent is directed to waste.

Figure 1. (A) The principle of peak separation in comprehensive two-dimensional liquid chromatography: (a) the

separation of 1D-overlapping peaks in the second dimension and (b) their position in the resulting 2D

chromatogram (modified from6). (B) (a) Sampling of consecutive fractions from the 1D effluent, which are then

subjected to re-injection and separation in the second dimension; (b) a series of chromatograms resulting from the

2D separation of individual fractions from the 1D. Each chromatogram is separated from its neighbours by dashed

lines (modified from12).

In a comprehensive mode, the entire 1D effluent is transferred in fractions for analysis on

the second column. Due to the different separation mechanisms applied in each dimension,

poorly resolved peaks from one column may be completely separated on another column

Intr

od

uct

ion


16

(Fig.1A.).13,14 The entire 1D effluent is collected on-line in small-volume fractions, which are

transferred one by one to the 2D column in multiple repeated cycles.15 Every second-dimension

run is recorded by a detector as a separate chromatogram (for each individual 1D fraction). A

fraction may contain just one component, or several components, as the case may be (Figure

1Ba). Thus, one sample component may be observed in several successive chromatograms,

spanning over the width of one first-dimension peak (Figure 1B)15. All two-dimensional data,

recorded by the detector, can be presented in the form of a colour plot (similar to Figure 2).

Theoretical aspects

In a two-dimensional separation process three steps can be distinguished: the first-dimension

separation; sampling, storing and re-injection of small 1D effluent fractions into the second

column (“modulation”); and the second-dimension separation. Each of these steps makes a

contribution to the overall resolution of a two-dimensional liquid chromatography experiment.6

The separation in the individual dimensions of the 2D-LC system is dictated by the same

parameters as those in separations performed in one-dimensional systems. However, specific

for the 2D systems is the need for selection of suitable “orthogonal”-phase-system

combinations in both dimensions,16 and for selecting conditions for the fraction transfer

between dimensions15 that result in as little loss of resolution as possible.

Sampling rate (modulation time)

The modulation interface determines whether the resolution obtained in the first separation can

be maintained in the 2nd dimension and is hence the heart of any LC×LC system. The frequency

at which the interface performs operations with each 1D fraction indicates how often the

fractions are transferred into the second dimension (the so-called sampling rate or modulation

time, tR, Figure 1B).15 In practice, re-mixing or re-dispersion of the collected fraction before

its transfer to the second dimension is the main reason for the resolution loss in overall LC×LC

separation. The volume of the 1D fraction contributes to the total band broadening in the second

dimension, which decreases at shorter modulation times (higher sampling rate).15 Hence, the

sampling rate strongly affects the quality of the 2D separation, and a fast and reliable transfer

of the 1D effluent is needed.8 Moreover, the internal volumes of all the components making up

the interface determine the extra-column band broadening; as such, the volume of these

components should also be minimized.

Introduction

17

Choosing the appropriate sampling rate (number of samples per 1D peak, ns) is a major

decision that must be made when designing a comprehensive 2D-LC separation method.6 A

fundamental rule (Murphy-Schure-Foley rule)17 for comprehensive 2D separation is to sample

each 1D peak 3 or 4 times. The effect of different sampling rates on the resolution of peaks in

close proximity to each other in 2D-LC is shown in Figure 2. Where the sampling rate (ns)

decreases from 4 (in this case the sampling time tR is 30 s) to 1 (tR=120 s), three peaks that

were observed at ns = 4 combine into a single peak. Furthermore, Murphy, Schure and Foley17

showed that even if the initial 1D peak is narrow, the effective width of the peak as it enters the

second column will depend on the sampling rate.6 Recent research indicates that it is possible

to simplify the technique and still obtain adequate resolution with only 2.5 - 3 cuts per 1D

peak.18,19

In practice, it is difficult to achieve both good resolution and high throughput. For that

reason, the first-dimensional column is very often operated at a very low flow rate in order to

sample a sufficient number of fractions across the 1D peak, which often leads to some

broadening.8 Moreover, a high sampling rate requires that the second-dimension separations

should not be longer than the duration of a modulation period to ensure an adequate number of

samplings. Thus, the 2D column should be very short, which leads to a relatively low number

of theoretical plates.15

Intr

od

uct

ion


18

Figure 2. Simulated demonstration of the effect of the first dimension sampling time (tR) and sampling rate (ns)

on the peak resolution when: (A) ns=4, (B) ns=2, (C) ns=1.5 and (D) ns =1 (modified from6).

Over the last decades, many different interface configurations have been reported. In

most cases, they consist of a two-position/10-port20–22, several 6-port23,24 or a two-position/12-

port valve25,26 having two sampling loops (with the same volume).25,27,28 One example of a 2D-

LC set-up with an 8-port switching valve is presented in the Figure 3. While Loop 1 collects

the 1D effluent, Loop 2 is simultaneously being emptied, transferring its content to the second

dimension.15 The sampling loop can be replaced with a trap-column.26,27,29–31 In this case,

focusing (pre-concentration) of the solutes occurs on a trap column prior to their analysis in

the second dimension.8 The same principle is used for other reported configurations, but using

two or more columns operating in parallel in the second dimension.28,31–33 Of course, such an

approach would increase the amount of sample to be analysed. However, the interfaces with

trap- and parallel-columns are difficult to implement in practice, due to the need for additional

instrumentation, which increases the complexity of the system.8 For these reasons, the

interfaces that contain a valve with loops are the most common in 2D-LC.34

Introduction

19

Figure 3. Scheme of a 2D-LC system with two positions of an 8-port valve: (A) Loop 1 acts as a storage loop

and is filled with 1D effluent. (B) When the valve switches, the analytes are transferred from Loop 1 to the 2D,

while Loop 2 takes on the role of storage loop, to be filled with 1D effluent.

Orthogonality

In 2D chromatography, the orthogonality of the two dimensions (i.e. of the two separation

mechanisms) is of prime importance. Orthogonality refers to the differences between the

properties of the coupled dimensions.15 If the two separation mechanisms exhibit minimum

correlation in separation selectivity, a 2D separation is considered to be orthogonal.1,8

Orthogonality not only depends on the chemical properties of the stationary and mobile phases

(i.e. the structure, polarity, charge, etc. of the stationary surface and the eluent), but also on the

properties of the solutes being separated (hydrophobicity, charge, etc.). Since solute properties

will of course vary from one solute to the next, the universal orthogonal 2D-LC combination

does not exist. Nowadays, a large variety of stationary phases is available. They differ in pore

size, surface chemistry, support material used, etc. At the same time, properties of the mobile

phase can be changed by modifying pH, adding ion-pair agents or adjusting the temperature.8,16

(In)compatibility of the mobile phases

It is important to make an appropriate choice of both stationary and mobile phases, especially

from the perspective of mobile-phase compatibility between dimensions, as even slight

incompatibility may adversely affect the overall separation. Incompatibility can mean that

either two liquid phases are simply immiscible, or a so-called solvent-strength mismatch is

observed. The latter occurs when a ‘strong’ eluent having high elution strength in one mode is

a ‘weak’ eluent having a low elution strength in the other mode. In both cases, not only

Intr

od

uct

ion


20

retention selectivity may be significantly affected, but also band broadening and/or peak shape

in the second dimension.

In order to maximize the reduction of the fraction volume to improve resolution as

discussed above, the sample should be retained more strongly on the 2D column than on the 1D

column. For this reason, the 1D mobile phase should preferably act as a weak solvent (having

low elution strength) in the second column. In this situation, a transferred fraction is retained

(focused) at the beginning of the second column in a more or less compressed narrow zone

before its elution with 2D mobile phase.15 This effect, called “on-column fraction focusing”,35,36

improves resolution, peak capacity and detection sensitivity, due to suppression of the band

broadening that occurs when the fraction is transferred to the 2D. However, in practice, it is

rather difficult to achieve. For instance, the aqueous-organic mobile phases that are used in 2D

reversed-phase (RP) - hydrophilic interaction liquid chromatography (HILIC) systems are

miscible. However, the RP×HILIC combination still presents significant solvent-strength

mismatch problems. In this case, an organic-solvent-rich 1D mobile phase would act as a strong

solvent in a RP 2D, whereas a water-rich mobile phase used as a weak eluent in a RP 1D would

act as a strong eluent in a HILIC 2D. This would cause reduced retention, increased band

broadening and asymmetrical or even split peaks.20,37

Moreover, if the solvents used as mobile phases are not completely miscible, serious

difficulties could arise, resulting in some separation modes being almost impossible to

combine.8 For instance, coupling normal-phase (NP) and reversed-phase (RP) liquid

chromatography – a combination that in terms of orthogonality is very useful – is generally not

easy to achieve due to mobile-phase immiscibility. In this case, if RP were to constitute the 1D,

the high concentrations of water in the RP mobile phase would almost certainly deactivate the

polar adsorbent of the 2D NP column, since NP utilizes non-aqueous mobile phase (e.g. n-

hexane).15

The design of the interface is of primary importance for systems in which the separation

modes are difficult to combine. Over the last decade, a number of interfaces have been

suggested, including interfaces incorporating trapping columns,26,27,29–31,38,39 or collection of

low-volume fractions from a 1D capillary monolithic column.37 More sophisticated approaches

utilize a vacuum-evaporation interface for on-line evaporation of the 1D solvents from the

loop,40 or thermally-assisted modulation exploiting the influence of temperature on analyte

retention.41,42 While each of these approaches has unique advantages for solving the problem

Introduction

21

of mobile-phase incompatibility, these technologies are still not developed to the extent where

they could be used universally in routine LC×LC analyses.

An approach that is more generally applied is the use of an additional solvent flow (so-

called make-up flow) to allow modification of the solvent composition between dimensions by

diluting the collected 1D fraction with a weaker 2D mobile phase. However, this requires a good

mixing device at the interface between the two columns. Referring to the above discussion, in

order to be able to cope with solvent incompatibility and efficiently perform all required

manipulations with the 1D effluent at the interface between two dimensions, the mixing device

must satisfy three strict conditions. First, it should provide fast mixing in-line at different ratios

over a wide range of flow rates compatible with typical flow rates used in 2D-LC; thus, the

efficiency of the mixing mechanism should not depend on the flow rate, under which the

mixing is happening. Second, a mixer for modulation should have a small volume so as not to

contribute to the extra column-band broadening. Finally, the mixer must be able to withstand

brief pressure pulses of up to a few hundred bar, due to its connection to switching valves used

to shuttle sample from the 1D to the 2D. This is why we chose to use small-volume microfluidic

devices as mixers that are placed in the interface between two dimensions.

Intr

od

uct

ion


22

Microfluidics

Microfluidic technology is an area of scientific research which involves the manipulation of

small (µL and nL) amounts of fluids in micrometer-size channels. It has received growing

interest over the last twenty years, due to its promising implementation in both industrial and

academic fields, especially in applications related to analytical chemistry, and cell or medical

biology. Microfluidic systems are often called “miniaturized total analysis systems” (μTAS) or

“lab-on-a-chip” devices. The main advantages of these systems include a significant reduction

in the amount of samples, reagents and waste products, faster analysis, limited costs, good

sensitivity, small size (portability), and minimal dead volume.43,44

Nowadays, microchip technology has entered the arena of scientific research as an

attractive tool to improve or even replace conventional ‘macro’ analytical techniques. The

development of chip-based microfluidic devices for integration in LC or LC-MS systems has

increased especially in the last decade45–50 as a result of the miniaturization trend in separation

science. In addition to the many benefits of device miniaturization mentioned above,

increased separation efficiency is an expected consequence of the incorporation of

microfluidic components into multi-dimensional LC separation systems. Performance in

many other applications can be enhanced by the increased surface-to-volume ratios of

microchannels, which facilitate high-speed reactions or interactions due to the increased

surface available. Processes can be run at length scales that are more relevant for normal

biological conditions (e.g. microfluidic channels can mimic blood capillaries). Throughput

(defined as samples per unit time) can also be improved if large numbers of samples can

be processed in parallel.51

Nowadays, there are a great variety of materials, fabrication methods, and techniques

available for the development of microfluidic devices. The most important of these will be

described in more detail below.

Materials for microfluidic-device fabrication

Because microfabrication techniques for the first microfluidic devices were adapted from the

microelectronics industry, they were fabricated in glass and silicon using a combination of

planar fabrication techniques (photolithography, thin-film metallization, and chemical

etching).52,53 Both Si and glass possess important characteristics, such as chemical inertness

and excellent thermal stability, and can be used if the application of high temperatures or

organic solvents are required.54,55 Glass is an important material for the fabrication of

Introduction

23

microfluidic devices due to its optical transparency, and the fact that it is available in various

compositions (e.g. fused silica, Pyrex, soda lime glass). Moreover, it is widely used as a

substrate for microchannels, as well as for device covers, often in combination with other

materials, as it allows microchannels and their contents to be directly observed under a

microscope.

The surface characteristics of oxidized silicon and glass can be beneficial for many

applications, due not only to the chemical inertness of these materials, but also because of the

possibility of chemically modifying surfaces using a host of different silane chemistries. The

high electrical resistance of glass also allows the application of high electric fields for induction

of electro-osmotic flow, an easily implemeneted mechanism for fluid propulsion in

microchannels, However, devices fabricated in these materials are not always easily

implemented for applications with living mammalian cells.55 The fact that glass and silicon are

not gas-permeable, for instance, means that perfusion media must be pre-equilibrated with

oxygen and other gases before introduction into a micro cell-culture device. Moreover,

conventional optical detection methods cannot be used for devices fabricated in silicon,

because silicon is opaque to visible and ultraviolet light.56 Besides, the fabrication of devices

from these materials is a time-consuming process that requires a cleanroom environment.53

Nowadays silicon and glass have largely been displaced by polymers (elastomers, such

as PDMS1) and thermoplastics (e.g. PMMA, COC, PC2) that have advantages such as optical

transparency, non-toxicity and lower costs. Besides, they are chemically quite inert (though

they are susceptible to surface softening and swelling in certain organic solvents), and have

good mechanical properties (they are not fragile).54,57 In general, the components required for

lab-on-a-chip devices are easier to fabricate in elastomers than in rigid, thermoplastic materials.

This is because the former materials can be easily cast in solution form onto molds to replicate

microchannels.56

Polydimethylsiloxane (PDMS), an elastomeric silicone rubber, has widened the

possibilities for utilization of microfluidic devices and has sped up their development in the

academic microfluidics community. This is due to its low cost, robustness and the

straightforward fabrication by replication of devices that it enables.52,58,59 PDMS cures at low

temperatures, is flexible (it is a soft elastomer), and is optically transparent down to 280 nm

1 PDMS - poly(dimethylsiloxane) 2 PMMA - polymethylmethacrylate; PC – polycarbonate; COC – cyclic olefin copolymer.

Intr

od

uct

ion


24

(which makes UV/Vis absorbance and fluorescence detection feasible).52,53 In addition, it is

non-toxic (allowing cultivation of mammalian cells in unmodified devices), commercially

available, reasonably inert from a chemical perspective, and durable. All of these qualities have

made PDMS a material of choice for many microfluidic applications, especially in the

exploratory stages of research projects involving device development.59

Of course, some properties of PDMS may be disadvantageous for certain applications.

For example, the elastomeric nature of PDMS may cause microchannels to expand or even tear

at high flow rates or under high pressures.60,61 The utilization of PDMS is also limited by its

incompatibility with many organic solvents.62,63 In addition, non-specific adsorption to the

relatively hydrophobic PDMS surface may occur when working with biological samples,

leading to fouled surfaces having undefined compositions.62

Table 1. Comparison between the most used materials for fabrication of microfluidic devices.

Material Advantages Disadvantages

Silicon chemically inert;

excellent thermal stability53

Opaque to visible and ultraviolet light;55

fabrication is a time-consuming process

that requires a cleanroom environment52

Glass

(pyrex and

fused silica)

chemically inert; optically

transparent; wide availability

in various sizes and chemical

compositions53

relatively expensive material; device

fabrication is a time-consuming process52

requiring a cleanroom environment

PDMS

optically transparent; soft

elastomer; ease in fabrication;

gas permeability;

inexpensive, biocompatible

and non-toxic51,57,58

not pressure-resistant due to elastomeric

nature;58 incompatibility with some

organic solvents;61,62 absorption of small

molecules into the matrix;58 modified

surfaces are generally unstable over time

as the modification wears off58

Thermoplastics

(PMMA, PC,

COC etc.)

low material cost; can easily

be adapted for mass

production;63,64 optically

clear; non-toxic; excellent

chemical inertness; superior

mechanical qualities53,56

surface chemistry control required; often

incompatible with organic solvents and

low-molecular-weight organic solutes;51

generally incompatible with temperatures

greater than 170°C65

Thermoplastics3 are gaining more interest, due to the broad range of material parameters

and surface chemical properties offered, allowing for optimal material selection and, thus,

tailoring of a device to the required application.64,65 Compared to silicon and glass, they are

3 A polymer material that becomes pliable or moldable above a specific temperature and solidifies upon cooling.

Introduction

25

less expensive and can be rapidly implemented in manufacturing processes for mass

production. The need to control the chemistry of the polymer surface makes these materials a

somewhat disadvantageous choice sometimes for microfluidic devices (compared to silicon or

glass). Many thermoplastics also exhibit incompatibility with organic solvents, which tend to

adsorb onto or absorb into the polymer substrate, creating some issues in real applications.

Thermoplastics also generally cannot be used at high temperatures (greater than 170 C66), as

they will tend to soften and deform.52

Table 1 summarizes the most-used materials for microfluidic devices together with their

advantages and disadvantages.

Fabrication techniques

The selection of the microfabrication method is a crucial step in the development of any

microfluidic system. This choice depends on the compatibility of the material with the (reagent)

solutions used, as well as the requirements set for the final product, e.g. the feature resolution,

thermal or pressure resistance, and the time available for device fabrication. The costs for

device fabrication is also considered as an important factor, because for some fabrication

methods the manufacturing costs for one device can be higher (e.g. due to the initial cost of

making the molds) than for the mass production of this device for commercial purposes.

Today the most common fabrication techniques for microfluidic devices are replica

molding (soft lithography),43,52 injection molding,67–69 hot embossing,70,71 and

stereolithography.72 The first three approaches can be regarded as indirect fabrication

techniques, in that the actual fabrication involves making a mold for replicating microchannels.

Once the mold is made, polymer solutions are cast onto it and allowed to cure (as is done with

elastomeric compounds) to form microchannels, or the mold is pressed into a hard polymer

layer which has been softened at elevated temperature to form microfluidic features. In

contrast, stereolithography involves the direct formation of microchannels through patterning

in the center of a mass of light-sensitive material with a focused laser. Other direct fabrication

methods also exist, such as laser micromachining (laser ablation),73,74 wet/dry etching after

photolithographic patterning75 and micromilling.76 In all of these latter examples, microchannel

formation is achieved by removing material from a substrate. An overview of these methods is

given in Table 2 and they are schematically presented in Figure 4.

Intr

od

uct

ion


26

Figure 4. Overview of main fabrication methods for microfluidic devices with their advantages (+) and

disadvantages (-).Modified from 59.

Certain methods are more suited than others to the rapid prototyping of devices, an

important step in many projects involving the use of microfluidics for new applications.

Injection molding and hot embossing can be considered as fast, but they are expensive methods

for polymer-device prototyping, due to the high initial cost of making the molds. Having molds

made can also result in significant waiting times of at least several weeks before a design can

be tested. On the other hand, glass/silicon micromachining processes based on wet/dry etching

create high-precision structures, but are technically demanding and time consuming.59

Although it is possible to achieve the widest range of features for making complex 3D

structures with micromilling, stereolithography (3D-printing) and laser ablation, these

techniques offer relatively low resolution (around 20-60 µm), generate surface roughness, and

yield limited numbers of produced devices, due to the inherent slowness of the sequential

fabrication process.59 Besides, 3D-printers with the highest resolution usually need supporting

materials to fill the void spaces (i.e., a channel) during the printing process, and the removal of

this materials from a very small channel (less than 60 µm in depth) is very difficult and

sometimes even impossible process for designs with complex channel structures.77

To date, however, the most common technique used for prototyping remains soft

lithography, a rapid-prototyping process which is based on replica molding. As was

mentioned above, it is this method that primarily makes use of PDMS. This is thanks to

the extremely precise replication of all the features on a mold surface with resolution in

Introduction

27

the nm range, but without the need for expensive equipment or advanced skills in

microfabrication.78

All the methods that are listed in Table 2, except stereolithography, suffer from one

inherent drawback. They enable the creation of an open 2D channel network in the substrate

surface that has to be hermetically sealed (closed or bonded to a second chip acting as a cover)

in order to obtain a microfluidic channel. Bonded interfaces between chips tend to form a weak

point for any high-pressure application, as it is typically the bond itself that fails first when

higher pressures are applied. Developed more than a decade ago, Femtosecond Laser

Irradiation followed by Chemical Etching (FLICE),79 also called Selective Laser-Induced

Etching (SLE),80,81 has appeared as a novel powerful alternative approach for direct fabrication

of complex 3D structures inside a solid transparent material, such as fused silica. Being a direct

fabrication technique inside the solid piece of material, SLE provides an appealing solution for

avoiding the chip-sealing step. The SLE technique also allows the fabrication of glass devices

with higher resolution (10-20 µm) than is possible with wet etching, allowing the unique

properties of glass (transparency, rigidity, inertness and so on) to be exploited in devices with

finer structuring.

Intr

od

uct

ion


28

Table 2. An overview of the most used methods for microfluidic devices fabrication.59,76

Hot-embossing Injection

molding Micromilling

3D-printing

(stereolithography)

Wet/dry

etching

Laser micromachining

(laser ablation)

Rapid prototyping

(soft-lithography)

Type Indirect Indirect Direct Direct Direct Direct Indirect

Materials polymeric materials thermoplastics metals and

plastics

Photo curable

polymeric resins,

ceramics

glass/silicon ceramics, metals, and

polymers

elastomers; epoxy

resins

General

polymer substrate is

softened at elevated

temperature and then

pressed against a mold

under pressure to

transfer the desired

features from the mold

to the polymer

melted polymer

is injected into a

microstructured

mold and

thendemolded

after cooling

microsize

features are

structured by

removing bulk

material using

special cutting

tools (end mills)

3D parts are created

in bulk photosensitive

material using a light

source (e.g. a laser)

remove the

substrate

material from a

wafer by means

of etching with

abrasive

chemical or

plasma gases

a high-powered pulsed

laser is used to remove

material from the

polymer substrate

(through a mask or by

using a laser direct-write

process)

a prepolymer solution

is cast onto a mold

and is cross-linked

with heat or

ultraviolet (UV)

light; the resulting

polymer is peeled off

the mold

Advantages

fast; high-throughput;

precise; rapid

replication of small

(low micrometer) to

medium-size features

with high aspect ratio;

mass production

Fast, high-

throughput when

the final design

has been

established;

enables complex

3D geometry;

highly automated

fast transfer of

designs into

prototypes;

cost-accessible;

enabling

complex 3D

geometry

enables complex 3D

features (accuracy of

20-60 µm) that may

be impossible with

other methods

enables small

features with

good resolution

cost-accessible;

enables complex 3D-

multilayer structures;

mass production

easy; economical;

fast; high resolution

(few nm); possibility

to create 3D features

Dis-

advantages

expensive due to the

initial cost of making

the molds; restricted to

thermoplastics,

difficult to fabricate

complex 3D structures;

relatively long cycle

times

expensive due to

the initial cost

and time of

making the

molds; limited to

thermoplastics

limited

throughput due

to the inherent

serial nature59;

low resolution

(around 100

µm); surface

roughness

expensive equipment

and material;

limited material

availability; materials

are brittle

technically

demanding;

time-consuming

limited number of

produced devices due to

the inherent feature of

the sequential process;

low resolution (~50 µm);

surface roughness; the

surface chemistry of the

final products is very

different due to laser

treatment

devices are

vulnerable under

higher applied

pressures due to

microchannel

deformation

Introduction

29

Mixing at the microscale

As described above, miniaturization provides many attractive features that separation science

can use for its own benefit. However, some phenomena that are not very relevant in the macro

world, play an important role in micrometer-sized devices. For example, on a large scale, fluids

mix convectively resulting in turbulent flow patterns (think here of milk when it is swirled into

coffee). In this example inertia is more important than viscosity, which is true for most fluids

in the macro world. However, at the microscale this situation is reversed, and viscosity

dominates in microchannels, leading to laminar flow due to the small masses of fluids involved.

This means that turbulence tends not to be exhibited in microsystems, except at flow rates

which are so extremely high that many devices would not survive the high pressures generated

as a result. When two fluids meet in the microchannel under laminar flow conditions, they flow

side-by-side, without agitation or disruptions.56 This makes mixing at the microscale a

challenging task. In such systems, the mixing can be achieved only by diffusion, which is a

passive and slow transport process where molecules of two fluid streams move across their

interface.82

To overcome the problem of mixing at the microscale, a large number of micromixers

have already been developed. All of these work on the fundamental principle of increasing

contact areas between the solutions to be mixed, so as to reduce distances that molecules need

to diffuse to achieve mixing.82–84 Micromixers can be divided into two big groups, comprising

either passive or active mixers. This classification is based on the mechanism by which the

solution interfaces are disrupted to achieve larger contact areas. In active mixers, the need for

integration of elements for transferring energy from an external source into the mixing chamber

complicates the fabrication process, limiting the implementation of such devices. In addition,

the external forces involved in these mixers can negatively influence the samples studied (e.g.

acoustic waves can generate heat, which could lead to unwanted reactions or damage of

biological samples).82 Passive micromixers, on the other hand, generally function either by

splitting solution flows into multiple thinner streams in a branched channel network, or by

placing fixed obstacles in the flow to perturb laminar flow patterns. This makes passive

micromixing, which requires no moving parts and, thus, no external energy (other than that

required to displace solutions through microchannels), a more preferable choice for many

applications. The reader is referred to Chapter 2 of this thesis for a more in-depth discussion

about mixing on the microscale in micromachined devices.

Intr

od

uct

ion


30

Scope of the thesis

In this thesis, we have aimed to improve the performance of two-dimensional liquid

chromatography by using microfluidic devices fabricated by different microfabrication

techniques. We developed a small-volume microfluidic mixer that was implemented in the

interface between two columns in a 2D-LC instrument. The problem of the mobile phase

incompatibilities between dimensions was addressed by fast in-line mixing of the 1D effluent

with a weaker eluent inside the micromixer before it reached the second column. Because of

the small inner volume of our device (˂ 5 µL), we did not introduce extra dispersion to the 1D

effluent fraction.

In Chapter 2 we give a broad overview of already existing micromixers based on

chaotic advection, as well as chaotic advection combined with other principles, that have been

proposed over the last decade. We have emphasized the link between channel geometry,

operating flow conditions and the mixing mechanism adopted. We also describe the most

common application areas of passive chaotic micromixers using real examples. We discuss the

connection between channel geometry and possible areas of application under different flow

conditions, as this influences mixing efficiency.

In Chapter 3 we describe the development of microfluidic chaotic mixers with a small

volume for future applications in two-dimensional liquid chromatography. The PDMS

micromixer contains staggered herringbone grooves with an optimized geometry for fast

modification of mobile phases at different flow-rate ratios (1:2, 1:5 and 1:10). The

microchannel is 5 cm long and complete mixing is achieved within the first 3 cm of the channel.

The mixing is efficient over the whole range of flow rates tested (4-1000 μL/min).

The research described in Chapter 4 was aimed at the fabrication of a pressure-resistant

microfluidic mixer inside a solid piece of fused silica using Selective Laser-Induced Etching

(SLE). We report a chip containing herringbone grooves for chaotic advective mixing in a

channel with lengths up to 33 mm fabricated using SLE. The pressure tests showed that fused

silica chips can withstand pressures up to 85 bar.

In Chapter 5, we successfully implemented a microfluidic micromixer in the interface

of a two-dimensional liquid chromatograph for analysis of real samples. For this research we

used a microfluidic mixer with herringbone grooves fabricated in COC using micromilling.

Using a custom-designed, robust, low-dead-volume interface, a chip was directly coupled to

Introduction

31

the chromatographic equipment. This design could withstand pressure pulses up to 150 bar. A

microfluidic mixer was implemented in a 2D HILIC×RP-LC system and an improved

separation of nylon polymers was obtained compared to the system without a mixer.

In Chapter 6 we summarize and discuss the findings of the research presented in this

thesis. We present future perspectives on the use of microfluidic technology for improving

conventional multidimensional chromatographic techniques.

Intr

od

uct

ion


32

References

1. Dugo, P., Cacciola, F., Kumm, T., Dugo, G. & Mondello, L. Comprehensive multidimensional liquid

chromatography: Theory and applications. J. Chromatogr. A 1184, 353–368 (2008).

2. Wehr, T., Resources, L. C., Group, T., Llc, R. & Creek, W. Liquid Chromatography in Proteomic Studies.

LC GC Eur. 20, 2–8 (2003).

3. Shen, Y. et al. Automated 20 kpsi RPLC-MS and MS/MS with chromatographic peak capacities of 1000-

1500 and capabilities in proteomics and metabolomics. Anal. Chem. 77, 3090–3100 (2005).

4. Shen, Y. et al. Packed capillary reversed-phase liquid chromatography with high-performance electrospray

ionization fourier transform ion cyclotron resonance mass spectrometry for proteomics. Anal. Chem. 73,

1766–1775 (2001).

5. Eeltink, S. et al. High-efficiency liquid chromatography-mass spectrometry separations with 50mm,

250mm, and 1m long polymer-based monolithic capillary columns for the characterization of complex

proteolytic digests. J. Chromatogr. A 1217, 6610–6615 (2010).

6. Stoll, D. R. et al. Fast, comprehensive two-dimensional liquid chromatography. J. Chromatogr. A 1168,

3–43; discussion 2 (2007).

7. Erni, F. & Frei, R. W. Two-dimensional column liquid chromatographic technique for resolution of

complex mixtures. J. Chromatogr. A 149, 561–569 (1978).

8. François, I., Sandra, K. & Sandra, P. Comprehensive liquid chromatography: Fundamental aspects and

practical considerations-A review. Anal. Chim. Acta 641, 14–31 (2009).

9. Bushey, M. M. & Jorgenson, J. W. Automated instrumentation for comprehensive two-dimensional high-

performance liquid chromatography of proteins. Anal. Chem. 62, 161–167 (1990).

10. Phillips, J. B. & Beens, J. Comprehensive two-dimensional gas chromatography: A hyphenated method

with strong coupling between the two dimensions. J. Chromatogr. A 856, 331–347 (1999).

11. Schoenmakers, P., Marriott, P. & Beens, J. Nomenclature and conventions in comprehensive

multidimensional chromatography. LCGC Eur. 25, 1–4 (2003).

12. Groskreutz, S. R., Swenson, M. M., Secor, L. B. & Stoll, D. R. Selective comprehensive multi-dimensional

separation for resolution enhancement in high performance liquid chromatography. Part I: Principles and

instrumentation. J. Chromatogr. A 1228, 31–40 (2012).

13. Mondello, L., Lewis, C. & Bartle, K. D. Multidimensional chromatography. Analytical and Bioanalytical

Chemistry 407, (Biddles Ltd, Guildford and King’s Lynn., 2002).

14. Stoll, D. R. et al. Fast, comprehensive two-dimensional liquid chromatography. J. Chromatogr. A 1168,

3–43 (2007).

15. Jandera, P. Comprehensive two-dimensional liquid chromatography — practical impacts of theoretical

considerations. A review. Cent. Eur. J. Chem. 10, 844–875 (2012).

16. Pirok, B. W. J., Gargano, A. F. G. & Schoenmakers, P. J. Optimizing separations in online comprehensive

two-dimensional liquid chromatography. J. Sep. Sci. 41, 68–98 (2018).

17. Murphy, R. E., Schure, M. R. & Foley, J. P. Effect of Sampling Rate on Resolution in Comprehensive

Two-Dimensional Liquid Chromatography. Anal. Chem. 70, 1585–1594 (1998).

18. Horie, K. et al. Calculating optimal modulation periods to maximize the peak capacity in two-dimensional

HPLC. Anal. Chem. 79, 3764–3770 (2007).

19. Vivo-Truyols, G., van der Wal, S. & Schoenmakers, P. J. Comprehensive Study on the Optimization of

Online Two-Dimensional Liquid Chromatographic Systems Considering Losses in Theoretical Peak

Capacity in 1st- and 2nd-Dimensions A Pareto-Optimality Approach.pdf. Anal. Chem. 82, 8525–8536

(2010).

20. Dugo, P., Favoino, O., Luppino, R., Dugo, G. & Mondello, L. Comprehensive Two-Dimensional Normal-

Phase (Adsorption)-Reversed-Phase Liquid Chromatography. Anal. Chem. 76, 2525–2530 (2004).

21. Jandera, P. et al. Two-dimensional liquid chromatography normal-phase and reversed-phase separation of

(co)oligomers. J. Chromatogr. A 1119, 3–10 (2006).

22. Van Der Horst, A. & Schoenmakers, P. J. Comprehensive two-dimensional liquid chromatography of

polymers. J. Chromatogr. A 1000, 693–709 (2003).

23. Ikegami, T. et al. Two-dimensional reversed-phase liquid chromatography using two monolithic silica C18

columns and different mobile phase modifiers in the two dimensions. J. Chromatogr. A 1106, 112–117

(2006).

24. Gray, M. J., Dennis, G. R., Slonecker, P. J. & Shalliker, R. A. Utilising retention correlation for the

separation of oligostyrenes by coupled-column liquid chromatography. J. Chromatogr. A 1073, 3–9

(2005).

Introduction

33

25. Venkatramani, C. J. & Zelechonok, Y. An automated orthogonal two-dimensional liquid chromatograph.

Anal. Chem. 75, 3484–3494 (2003).

26. Venkatramani, C. J. & Patel, A. Towards a comprehensive 2-D-LC-MS separation. J. Sep. Sci. 29, 510–

518 (2006).

27. Cacciola, F., Jandera, P., Hajdú, Z., Česla, P. & Mondello, L. Comprehensive two-dimensional liquid

chromatography with parallel gradients for separation of phenolic and flavone antioxidants. J.

Chromatogr. A 1149, 73–87 (2007).

28. François, I. et al. Tryptic digest analysis by comprehensive reversed phasextwo reversed phase liquid

chromatography (RP-LCx2RP-LC) at different pH’s. Journal of separation science 32, 1137–44 (2009).

29. Holm, A. et al. Combined solid-phase extraction and 2D LC-MS for characterization of the neuropeptides

in rat-brain tissue. Anal. Bioanal. Chem. 382, 751–759 (2005).

30. Pepaj, M., Wilson, S. R., Novotna, K., Lundanes, E. & Greibrokk, T. Two-dimensional capillary liquid

chromatography: pH Gradient ion exchange and reversed phase chromatography for rapid separation of

proteins. J. Chromatogr. A 1120, 132–141 (2006).

31. Cacciola, F., Jandera, P., Blahová, E. & Mondello, L. Development of different comprehensive two

dimensional systems for the separation of phenolic antioxidants. J. Sep. Sci. 29, 2500–2513 (2006).

32. François, I., de Villiers, A., Tienpont, B., David, F. & Sandra, P. Comprehensive two-dimensional liquid

chromatography applying two parallel columns in the second dimension. J. Chromatogr. A 1178, 33–42

(2008).

33. Wagner, K., Miliotis, T., Bischoff, R. & Unger, K. K. An Automated On-Line Multidimensional HPLC

System for Protein and Peptide Mapping with. Anal. Chem. 74, 809–820 (2002).

34. Li, D., Jakob, C. & Schmitz, O. Practical considerations in comprehensive two-dimensional liquid

chromatography systems (LCxLC) with reversed-phases in both dimensions. Anal. Bioanal. Chem. 407,

153–167 (2015).

35. Jandera, P., Hájek, T. & Česla, P. Effects of the gradient profile, sample volume and solvent on the

separation in very fast gradients, with special attention to the second-dimension gradient in comprehensive

two-dimensional liquid chromatography. J. Chromatogr. A 1218, 1995–2006 (2011).

36. Hoffman, N. E., Pan, S. L. & Rustum, A. M. Injection of eluites in solvents stronger than the mobile phase

in reversed-phase liquid chromatography. J. Chromatogr. A 465, 189–200 (1989).

37. Jandera, P., Hájek, T., Staňková, M., Vyňuchalová, K. & Česla, P. Optimization of comprehensive two-

dimensional gradient chromatography coupling in-line hydrophilic interaction and reversed phase liquid

chromatography. J. Chromatogr. A 1268, 91–101 (2012).

38. Cao, L. et al. The development of an evaluation method for capture columns used in two-dimensional

liquid chromatography. Anal. Chim. Acta 706, 184–190 (2011).

39. Gargano, A. F. G., Duffin, M., Navarro, P. & Schoenmakers, P. J. Reducing Dilution and Analysis Time

in Online Comprehensive Two-Dimensional Liquid Chromatography by Active Modulation. Anal. Chem.

88, 1785–1793 (2016).

40. Tian, H.-Z., Xu, J. & Guan, Y.-F. Vacuum-evaporation Interface of Comprehensive Two-dimensional

Liquid Chromatography and Its Application. Chinese Journal of Analytical Chemistry 36, 860–864 (2008).

41. Verstraeten, M., Pursch, M., Eckerle, P., Luong, J. & Desmet, G. Thermal modulation for

multidimensional liquid chromatography separations using low-thermal-mass liquid chromatography

(LC). Anal. Chem. 83, 7053–7060 (2011).

42. Van de Ven, H. C., Gargano, A. F. G., Van der Wal, S. J. & Schoenmakers, P. J. Switching solvent and

enhancing analyte concentrations in small effluent fractions using in-column focusing. J. Chromatogr. A

1427, 90–95 (2016).

43. Duffy, McDonald, J., Schueller, O. & Whitesides, G. Rapid prototyping of microfluidic systems in poly

(dimethylsiloxane). Anal. Chem (1998).

44. Nguyen, N.-T. & N.-T. Nguyen. Micromixers: Fundamentals, Design and Fabrication. Igarss 2014

(William Andrew Publishing, 2011). doi:10.1016/B978-1-4377-3520-8.00001-2

45. Lin, S.-L., Bai, H.-Y., Lin, T.-Y. & Fuh, M.-R. Microfluidic chip-based liquid chromatography coupled to

mass spectrometry for determination of small molecules in bioanalytical applications: An update.

Electrophoresis 35, 1275–1284 (2014).

46. Dietze, C., Hackl, C., Gerhardt, R., Seim, S. & Belder, D. Chip-based electrochromatography coupled to

ESI-MS detection. Electrophoresis 37, 1345–1352 (2016).

47. Yin, H. et al. Microfluidic chip for peptide analysis with an integrated HPLC column, sample enrichment

column, and nanoelectrospray tip. Anal. Chem. 77, 527–533 (2005).

48. Sung, W. C., Makamba, H. & Chen, S. H. Chip-based microfluidic devices coupled with electrospray

ionization-mass spectrometry. Electrophoresis 26, 1783–1791 (2005).

Intr

od

uct

ion


34

49. Ro, K. W., Liu, J. & Knapp, D. R. Plastic microchip liquid chromatography-matrix-assisted laser

desorption/ionization mass spectrometry using monolithic columns. J. Chromatogr. A 1111, 40–47 (2006).

50. Bai, H.-Y., Lin, S.-L., Chan, S.-A. & Fuh, M.-R. Characterization and evaluation of two-dimensional

microfluidic chip-HPLC coupled to tandem mass spectrometry for quantitative analysis of 7-

aminoflunitrazepam in human urine. Analyst 135, 2737 (2010).

51. Ehrnström, R. Miniaturization and integration: challenges and breakthroughs in microfluidics. Lab Chip

2, 26N–30N (2002).

52. McDonald, J. C. et al. Fabrication of microfluidic systems in poly(dimethylsiloxane). Electrophoresis 21,

27–40 (2000).

53. Whitesides, G. M. & Stroock, A. D. Flexible methods for microfluidics. Phys. Today 54, 42 (2001).

54. Lavrik, N. V., Taylor, L. T. & Sepaniak, M. J. Nanotechnology and chip level systems for pressure driven

liquid chromatography and emerging analytical separation techniques: A review. Anal. Chim. Acta 694,

6–20 (2011).

55. Wu, J. & Gu, M. Microfluidic sensing: state of the art fabrication and detection techniques. J. Biomed.

Opt. 16, 080901 (2011).

56. Whitesides, G. M. The origins and the future of microfluidics. Nature 442, 368–373 (2006).

57. Wu, J. & Gu, M. Microfluidic sensing: state of the art fabrication and detection techniques. J. Biomed.

Opt. 16, 080901 (2011).

58. Ng, J. M. K., Gitlin, I., Stroock, A. D. & Whitesides, G. M. Components for integrated

poly(dimethylsiloxane) microfluidic systems. Electrophoresis 23, 3461–3473 (2002).

59. Sollier, E., Murray, C., Maoddi, P. & Di Carlo, D. Rapid prototyping polymers for microfluidic devices

and high pressure injections. Lab Chip 11, 3752 (2011).

60. Schneider, F., Draheim, J., Kamberger, R. & Wallrabe, U. Process and material properties of

polydimethylsiloxane (PDMS) for Optical MEMS. Sensors Actuators, A Phys. 151, 95–99 (2009).

61. Martin, A., Teychené, S., Camy, S. & Aubin, J. Fast and inexpensive method for the fabrication of

transparent pressure-resistant microfluidic chips. Microfluid. Nanofluidics 20, 1–8 (2016).

62. Stroock, A. Components for integrated poly (dimethylsiloxane) microfluidic systems. Electrophoresis 23,

3461–3473 (2002).

63. Favre, E. Swelling of crosslinked polydimethylsiloxane networks by pure solvents: Influence of

temperature. 32, 1183–1188 (1996).

64. Jena, R. K., Yue, C. Y. & Lam, Y. C. Micro fabrication of cyclic olefin copolymer (COC) based

microfluidic devices. Microsyst. Technol. 18, 159–166 (2012).

65. Tsao, C. W. & DeVoe, D. L. Bonding of thermoplastic polymer microfluidics. Microfluid. Nanofluidics 6,

1–16 (2009).

66. Topas. TOPAS- Cyclic Olefin Copolymers. 4 (2015).

67. Heckele, M. & Schomburg, W. K. Review on micro molding of thermoplastic polymers. J.

Micromechanics Microengineering 14, (2004).

68. Giboz, J., Copponnex, T. & Mélé, P. Microinjection molding of thermoplastic polymers: a review. J.

Micromechanics Microengineering 17, R96–R109 (2007).

69. Attia, U. M., Marson, S. & Alcock, J. R. Micro-injection moulding of polymer microfluidic devices.

Microfluid. Nanofluidics 7, 1–28 (2009).

70. Peng, L., Deng, Y., Yi, P. & Lai, X. Micro hot embossing of thermoplastic polymers: A review. J.

Micromechanics Microengineering 24, (2014).

71. Becker, H. & Heim, U. Hot embossing as a method for the fabrication of polymer high aspect ratio

structures. Sensors Actuators, A Phys. 83, 130–135 (2000).

72. Melchels, F. P. W., Feijen, J. & Grijpma, D. W. A review on stereolithography and its applications in

biomedical engineering. Biomaterials 31, 6121–30 (2010).

73. Locascio, L. E., Ross, D. J., Howell, P. B. & Gaitan, M. Fabrication of polymer microfluidic systems by

hot embossing and laser ablation. Methods Mol Biol., 339, 37–46 (2006).

74. Malek, C. G. K. Laser processing for bio-microfluidics applications (part I). Anal. Bioanal. Chem. 385,

1362–1369 (2006).

75. Misra, A., Hogan, J. D. & Chorush, R. Wet and Dry Etching Materials. Handb. Chem. Gases Semicond.

Ind. 1–5 (2002). doi:10.1002/0471263850.mis056

76. Guckenberger, D. J., de Groot, T. E., Wan, A. M. D., Beebe, D. J. & Young, E. W. K. Micromilling: a

method for ultra-rapid prototyping of plastic microfluidic devices. Lab Chip 15, 2364–2378 (2015).

77. Chen, C. et al. 3D-printed Microfluidic Devices: Fabrication, Advantages and Limitations—a Mini

Review. 8, 6005–6012 (2017).

Introduction

35

78. Whitesides, G. M., Ostuni, E., Takayama, S., Jiang, X. & Ingber, D. E. Soft Lithography in Biology and

Biochemistry. Annu. Rev. Biomed. Eng. 3, 335–373 (2001).

79. Matsuo, S., Tabuchi, Y., Okada, T., Juodkazis, S. & Misawa, H. Femtosecond laser assisted etching of

quartz: Microstructuring from inside. Appl. Phys. A Mater. Sci. Process. 84, 99–102 (2006).

80. Gottmann, J., Hermans, M., Repiev, N. & Ortmann, J. Selective laser-induced etching of 3D precision

quartz glass components for microfluidic applications-up-scaling of complexity and speed. Micromachines

8, (2017).

81. Hermans, M., Gottmann, J. & Riedel, F. Selective, laser-induced etching of fused silica at high scan-speeds

using KOH. J. Laser Micro Nanoeng. 9, 126–131 (2014).

82. Capretto, L., Wei Cheng, M. H. & Zhang, X. Micromixing Within Microfluidic Devices. TripleC 304, 27–

68 (2011).

83. Nguyen, N.-T. & Wu, Z. Micromixers—a review. J. Micromechanics Microengineering 15, R1–R16

(2005).

84. Lee, C. Y., Wang, W. T., Liu, C. C. & Fu, L. M. Passive mixers in microfluidic systems: A review. Chem.

Eng. J. 288, 146–160 (2016).

Intr

od

uct

ion

Chapter II

Novel micromixers based on chaotic

advection and their application —a

review

Margaryta A. Ianovska1,2, Patty P.M.F.A. Mulder1, Elisabeth Verpoorte1

1Pharmaceutical Analysis, Groningen Research Institute of Pharmacy, University of

Groningen, The Netherlands

2 TI-COAST, Amsterdam, The Netherlands

Manuscript in preparation

Abstract

Over the last twenty years, microfluidic technology has received growing interest in a diverse

set of fields, including clinical diagnostics, genetic sequencing, chemical synthesis and

proteomics, all of which are applications in which mixing plays a central role. However, mixing

at the micrometer scale is not easily achieved, due to the dominance of laminar flow, a well-

ordered flow regime characterized by fluid streams flowing parallel to each other. Mixing of

the dissolved species in two neighbouring solution streams occurs by diffusion only. Given that

diffusion is inherently a slow process, and the contact area between laminarly flowing solutions

is limited to their contact interface, mixing in such a system is not particularly efficient. Thus,

specially designed micromixers that are used to overcome the challenges related to mixing in

laminar flows are an important part of many microfluidic platforms. All micromixers ultimately

have the same objective, namely to increase contact areas between the solutions to be mixed,

in order to shorten diffusion lengths and thus promote more efficient mixing. Chaotic advection

is one of the most efficient mechanisms to induce mixing, as it involves the generation of flow

patterns which dramatically thin solution layers. In this chapter we describe passive

micromixers that were proposed within the last decade, based on chaotic advection and its

combination with other mixing principles (e.g. split and recombination (=SAR)). We also

discuss the applications of different types of chaotic micromixers in chemical industry, biology,

and analytical chemistry. Furthermore, we draw the connection between the design and

potential application of recently reported micromixers.

Keywords: Microfluidics; Micromixing; Passive micromixers; Chaotic advection; Combined

principles; 3D convoluted channels; Application of the mixers.

Novel micromixers based on chaotic advection and their application —a review

39

1. Introduction

Microfluidic technology has received growing interest due to its promising application as an

enabling technology in both industrial and and academic science. The key advantage of

microfluidic systems is their small size, which means only small (µL or less) quantities of

chemicals are required for the (bio)chemical process or analysis in question.1 However, if we

introduce two liquids from neighbouring inlets into a single microfluidic channel, we will

observe that these two streams flow parallel to each other. Even if the microchannel has turns

integrated into it, these streams will pass through the turn without any visible mixing occurring

(that can continue for a distance of several meters at the flow rates used typically). This regime

is called laminar flow and it exists in all micrometer-size channels that operate under flow rates

of a few to hundreds of µL/min. In order to use such devices for applications in clinical

diagnostics, genetic sequencing and chemical synthesis, where mixing is central to the

application, this problem should be first overcome.

Basically, mixing can only be achieved by means of one process, molecular diffusion,

which is driven by the gradient formed between highly-concentrated and less-concentrated

regions of the molecules to be mixed. Diffusion results in mixing without requiring directed

bulk motion, and it is faster if the contact area between two regions is larger. However, in most

cases the fluids in the microchannel are introduced by means of a pump at a constant flow rate

and the molecules experience advection – molecular mass transfer by bulk motion of fluid that

occurs parallel to the direction of the main flow. Due to the laminar flow and constant movement

of fluids along the channel, the contact area between two streams is very small and the mixing

(diffusion) happens to a minimal degree only at the interface. With an increase in the flow rate

(faster movement of fluids), the residence time, or time that molecules spend in the channel,

will decrease further, leading to a further decline in both the degree and efficiency of mixing.

These effects will be discussed in more detail later (Sec.2.2.).

To overcome a problem with mixing at the microscale, a large number of micromixers

have been already developed.2–4 In general, the purpose of all micromixer designs is to increase

the contact area between fluid streams, and in this way, decrease the diffusion length, which

makes mixing by diffusion faster. Depending on the basic mixing principle being exploited,

micromixers can be divided into either the passive or active category. Active micromixers

utilize external energy to perturb flow patterns and achieve mixing. For this, an external power

Ch

ap

ter

II


40

source has to be integrated into the system, which complicates the fabrication process, and

possibly limits the implementation of such devices. In addition, the external forces involved in

this type of mixer can negatively influence the samples studied (e.g. acoustic waves can

degrade synthetic polymers or generate heat, which could lead to unwanted reactions or damage

if biological samples are involved).2 This makes passive micromixers, which do not require an

external source of energy beyond that needed for advective flow, a more preferable choice for

a wide range of applications.

Passive micromixers can be further classified according to one of the following mixing

mechanisms: 1) parallel lamination and 2) sequential lamination (split and recombination

(=SAR)), 3) focusing-enhanced (injection), 4) chaotic mixing and 5) droplet micromixers.2,3

Parallel and serial lamination micromixers first split the inlet flows of the solutions to be mixed

into n sub-streams and later recombine them into one flow. In the focusing-enhanced

micromixer, a single solute flow is split by injecting it into several solvent flows. In chaotic

advection, mixing is achieved through generation of chaotic flow patterns formed at an angle

to the main flow, as a result of special microchannel geometries. Passive micromixing in

droplets exploits an internal recirculating flow field induced by their transport in non-miscible

carrier phases.3

Micromixers based on chaotic advection provide for fluid stretching and folding over

the cross-section of the channel, and are especially effective in microfluidic devices.1 A

relatively new trend in mixer designs is the combination of chaotic advection with the SAR

principle, which utilizes so-called 3D convoluted channels that provide efficient mixing over a

large range of Reynold numbers (Re). In this chapter we will primarily describe and discuss

passive micromixers based on chaotic advection. The combination of chaotic advection with

other flow processing approaches to achieve fast microfluidic mixing over extended flow rate

ranges will also be briefly presented.

In our experience, designing a mixing device can be a time-consuming process, due to

the many design parameters that need to be taken into account, as well as the choice of material

and fabrication method, depending on the final application. Before endeavouring to make a new

micromixer from scratch, one should possess appropriate knowledge and a good understanding

of mixing at the microscale, as a lot of designs that work well have been already proposed.5–28

Our main goal in this chapter is to help the reader in that process by providing him or her with

a wide overview of existing micromixers based on chaotic advection and combined principles


41

that can be applied to a variety of fields. We present reported applications of these devices,

which include examples in chemistry, biology and analytical chemistry, to name but a few.

In Section 3 we will discuss the micromixers that have been the most used over the last

decade, presented according to geometric classification. We place an emphasis on the channel

geometry, flow conditions (described by Re) and the mechanism of mixing. In Section 4 we

will describe the most common application areas of passive chaotic micromixers with real

examples. In the Discussion section we will focus on the link between channel geometry and

possible area of application, at different flow conditions known to influence mixing efficiency.

Ch

ap

ter

II


42

2. Theory

2.1. Viscosity, inertia and the Reynolds number

There are two major forces that play an important role in the microchannels, namely viscous

and inertial forces. Both of them can be seen as a measure of resistance. In the case of viscosity,

this resistance appears due to frictional shear forces that arise during the motion of molecules.

When the fluid moves through a channel as the result of an applied pressure gradient, the

molecules of the fluid generally move more quickly in the region around the central axis of the

channel than near the walls. This difference in relative motion of the fluid layers results in

differing amounts of friction being manifested between layers. Informally, viscosity is said to

be related to the “thickness” of liquids and their resistance to flow. For example, water flows

more easily than honey because it has a lower viscosity than honey. Inertia, on the other hand,

is the resistance of a volume of fluid to change its state of motion or its velocity (the fluid

prefers to continue moving in a straight line at a constant velocity). The magnitude of the inertial

force in a fluid flow depends on the mass of the fluid, increasing as fluid mass increases.

The interplay of these two forces determines the flow regime at a given flow rate in any

type of channel and can be expressed as the Reynolds number (see Equation 1). The flow

regimes that govern the behaviour of fluids in channels can be broadly divided into laminar or

turbulent. The Reynolds number predicts the range of flow rates at which flow in a

microchannel changes from laminar to turbulent. It is expressed as a measure of the ratio of

inertial forces to viscous forces for a given set of flow conditions:

𝑅𝑒 =Inertial Forces

Viscous Forces=

v𝑑ℎρ

μ (1),

where dh denotes the hydraulic diameter of the channel (see Eqn. 2), v is average linear velocity

(m/s), ρ equals the density of the fluid (kg/m3) and μ represents the dynamic viscosity of the

fluid (kg/(ms)). In case of heterogeneous flow, an average density and an average viscosity

based on the proportion of each fluid in the mixture are calculated. The fully turbulent regime

starts at Re > 3000 (depending on channel diameter).

Using the Reynolds number (Re) makes it is possible to compare different designs under

the same flow conditions.

The hydraulic diameter can be calculated with the following equation:

𝑑ℎ =4𝐴

𝑃 (2),


43

where A is the cross-sectional area of the flow (mm2) and P is the perimeter of the cross-section

(mm).

For a channel with circular cross-section the hydraulic diameter is calculated using the radius

of the circular pipe (r, mm), yielding the following familiar relationship:

𝑑ℎ =4𝜋𝑟2

2𝜋𝑟= 2𝑟 (3),

The hydraulic diameter of a rectangular duct is:

𝑑ℎ =2𝑤ℎ

𝑤+ℎ (4),

where h is the channel height (mm) and w is the channel width (mm).

2.2. Forms of mass transport to achieve the mixing

In general, there exist four types of mass transport in miccrochannels, namely molecular

diffusion, eddy diffusion, advection, and Taylor dispersion.29 Eddy diffusion is the transport

of large solutes by turbulent flow, where turbulent flow is characterized by chaotic changes in

flow velocity. However, the dominance of viscous forces at the microscale at the flow rates

typically used makes turbulence difficult to achieve (Re ≤ 2000) and, hence, this type of mixing

is not relevant for micromixers.

Taylor dispersion refers to the dispersion of solutes at the front of an advancing solution

flow in a microfluidic channel. When a new solution is introduced into an already-filled

microchannel under pressure-driven flow conditions, the solution front quickly adopts the

parabolic velocity profile in the channel. As a result, the front of the new solution is drawn out

into the back end of the solution in front of it, creating concentration gradients of dissolved

compounds across the channel in a direction perpendicular to flow. Diffusion of species

between streamlines having different velocities (due to the parabolic profile of the flow),30

serves to further smear out the sharp concentration profile at the solution front. Because Taylor

dispersion occurs in the direction of flow, it can be seen as an interplay between advection and

diffusion. In the situation when the microchannel is already fully filled with a given solution,

and no new solutions need to be introduced, Taylor dispersion no longer is a parameter, which

needs to be taken into consideration when describing flows and concentration gradients. In fact,

concentration gradients will cease to exist once advective flow has served to fill the

microchannel entirely with one solution having a constant composition. However, it is worth

Ch

ap

ter

II


44

noting that at the interface between two fluid streams, Taylor dispersion is also dictated purely

by molecular diffusion.

The most important forms of mass transport at the microscale remain molecular

diffusion and advection. Molecular diffusion involves the random motion of molecules,

whereas molecular transport by advection sees molecules being carried in bulk flow. Diffusion

is a mass transfer phenomenon that causes the distribution of dissolved (bio)chemical species

to become more uniform in space as time passes. The driving force for diffusion is the thermal

motion of molecules, where molecules migrate from a region of high concentration to a region

of low concentration. Fick's first law of diffusion states that the magnitude of this molecular

flux is proportional to the concentration gradient thus formed, as expressed in the following

equation:

𝐽 = −𝐷∇𝑛 (5),

where J is the diffusion flux per unit area per unit time (mol/(m2×s)), D is the diffusion

coefficient (m2/s) and ∇n represents the relevant concentration gradient (mol/m4). The diffusion

coefficient, D, is a measure of the rate of the diffusion process. The average distance that a

molecule travels by diffusion in a given amount of time can be calculated using the Einstein-

Smoluchowski equation, given below, which was derived from Fick’s law of diffusion by the

two scientists after which it is named.

𝑑𝑑𝑖𝑓𝑓 = √2𝐷𝑡 (6),

In this equation, ddiff is the distance a dissolved species travels in a time, t (s). Usually, diffusion

is a very slow process. For instance, a molecule of glucose with a diffusion coefficient of 5 ×

10-6 cm2/s requires more than 27 h to travel a distance of 1 cm (the total path-length).

Diffusion is superimposed on advection, the mass transport that occurs in a direction

parallel to the main flow as a result of dissolved molecular species being carried by the flow.

Advection determines the flow conditions under which diffusion takes place. In fact, advection

is not very useful in microfluidics for the mixing process, given the predominance of laminar

flow at the flow rates typically used in microfluidics (L/s to L/min). However, advection that

occurs in directions that are not parallel to the net flux of the solution, secondary flows also

known as chaotic advection, can facilitate mixing dramatically.1 In chaotic advection simple

regular velocity fields produce chaotic molecular or particle trajectories.31 This results in an

exponential growth of the interfacial area and an accompanying decrease in the thickness of the

fluids layers over which diffusion must occur to complete homogenization of two or more


45

solutions. Thus, chaotic advection is a very promising mechanism to improve mixing at the

microscale.2

It is important to note that chaotic advection is not turbulence. For a flow system under

steady state, the velocity components in chaotic advection at each point in space remain

constant over time, while the velocity components in turbulent flow vary over time at each point

in space.1 A necessary condition for chaos is that streamlines should cross each other at different

times, causing particles to change their paths. Thus, chaotic advection can occur in a time-

periodic flow or a spatially time-independent periodic flow.1 The first type can be implemented

by setting boundaries into motion through application of external forces (e.g. electric field).

These micromixers fall into the active category, and are based on effects such as electrokinetic

instability, EKI, a phenomenon which ca be induced in a microchannel using an applied electric

field.32,33 Chaotic advection in a spatially time-independent periodic flow can be achieved by

using 2D curved channels, for example, which will be described in Section 3.1.

Many authors20,22,34–36 report that there exists a critical value of the Reynolds number

(Recr) for every micromixer based on chaotic advection. Below this critical value, mixing is

dominated by diffusion and because the Re is proportional to the linear velocity in the system

(i.e. flow rate), the mixing efficiency is reduced with increase in flow rate. Above Recr the

mixing process is advection-dominated and mixing efficiency increases with increase in flow

rate. A probable explanation for this observation is that at low flow rates, the strength of these

secondary flows is not sufficient to significantly disturb the laminar flow profile, and mixing

by diffusion occurs between two neighbouring parallel streams. When secondary flow patterns

become more pronounced at higher flow rate, mixing by diffusion is facilitated by resulting

increases in contact area and thinner solution streamlines. Each particular micromixer design

has its own critical value of Re, above which mixing is especially efficient. For the end-user

looking for an appropriate mixer for a specific application, the critical value of Reynolds

number should be an important indicator whether a chosen micromixer design will work in the

most efficient way under the required conditions of the particular application.

Ch

ap

ter

II


46

3. Passive micromixers based on chaotic advection

In this section we will describe designs and mixing mechanisms of the mixers that have been

proposed within the last decade. The classification of these micromixers is based on their

geometry and includes simple channels (spiral, zig-zag and serpentine), obstacles or wall

modifications, and 3D convoluted channels. The possible mechanism of mixing depends on the

channel geometry. Flow conditions (described by Re), under which the mixer is operated,

dictate the type of phenomenon that governs mixing and, thus, the efficiency of mixing. Thus,

each design can provide different mixing performances at different Re.

3.1 Simple geometries: spiral, zig-zag and serpentine channels

The easiest design for creating chaotic advection is the T-mixer, where two streams collide at a

T-junction. Due to the sharp 90° angle at the entrance, the inertial force is large enough to cause

vortices at the junction (so-called Dean vortices), which lead to chaotic advection.37 T-mixers

have been investigated extensively by many researchers.38–40 However, the efficient application

of T-mixers require Re>150, which is Re at which vortices inside the T-mixer become

asymmetric and real chaotic advection occurs.1 Thus, many research groups used the T-junction

for introducing streams of liquids in combination with other channel modifications, for instance,

the spiral6,35 zig-zag-shaped36 and serpentine22 microchannels. In these designs, similar to the

T-mixers, the chaotic advection is induced by the appearance of Dean vortices when the fluids

experience centrifugal effects when traveling along a curved path of the pipe or at turns in the

channel.6 Dean flow can be intensified by introducing larger numbers of repetitive turns (Fig.

1A), and mixing by chaotic advection will be improved when the flow rate used is increased.

Sundarsan et al.35 tested mixing in spiral channels (Fig. 1A) using five different designs

(the four-arc, six-arc, eight-arc and ten-arc spiral channels) for Reynolds numbers between 0.02

and 18.6. The mixing efficiency of all designs improves with increased flow rate (Re>10) and

with increase in length of individual spiral contours together with decrease of their curvature

radius. This effect illustrates the correlation between mixing efficiency and the flow rate (Re,

De). Li et al.5 designed a planar labyrinth micromixer (PLM) (Fig. 1Ba) consisting of ten

successive in-line “S-shaped” mixing units (Fig. 1Bb) that are compactly arranged within a

confined circular area. Using such micromixers the range of Reynolds number, at which

efficient mixing occurs, can be expanded to Re 30. A design with a short straight channel


47

between two consecutive semicircles arranged with a 180°-turn provides continuous rotation of

the fluid, repeatedly distorts the interface between two streams, and breaks up unmixed regions

due to a complete position switching of the two streams.

Figure 1. Passive micromixers with simple geometries: (A) The spiral channel network incorporating three mixing

sections (Modified from35); (B) (a) a scheme of the planar labyrinth micromixer (PLM) with (b) “S-shaped” mixing

unit (Modified from 5); (C) ILSC mixer and (D) Ω mixer (Modified from6). (E) Microscopy image of a zig-zag

microchannel (Taken from36; (F) C-shaped micromixer micromixer with baffles (Taken from27); (G) Mixer with

(a) staggered and (b) symmetric obstructions along the microchannel (Modified from 22; (H) A passive alcove-

based mixer (Taken from23). See Table 1 for geometric dimensions.

Recently, inspired by the mixing results in the spiral channels, Al-Halhouli et al.6

presented computational simulations and experimental results for two new mixers composed of

units shaped as interlocking-semicircle (ILSC) and omega (Ω) channels. The ILSC mixer (Fig.

1C) consists of several mixing modules, which are composed of two offset mirrored

interlocking semi-circles (ILSC) whereas the second design consists of series of Ω-shaped

modules (Fig. 1D). Both designs enable a simultaneous rapid 90°-change in the flow direction

(and the direction of Dean vortices formed) four and six times within each Ω- and ILSC mixing

module, respectively. Both micromixers can be used over the entire range of 0.01<Re<50;

however, complete mixing is achieved only at Re>10. It should be mentioned that the strength

Ch

ap

ter

II


48

of Dean vortices in the Ω design is expected to be less than those in the ILSC design for a given

Reynolds number, because the Ω-mixer has 1.67-times larger mean radius of curvature.

Another simple design that was exploited decades ago, is a zig-zag microchannel (Fig.

1E), where periodic turns cause chaotic advection. Mengeaud et al.36 made simulations in the

Reynolds number range of 26-267. They found that there exists a critical Reynolds number of

80, under which the mixing relies entirely on molecular diffusion. At higher Re, mixing was

improved by recirculation generated at the channel turns.

Tsai and Wu 27 introduced radial baffles to the curved microchannel and named this

design a curved-straight-curved (C-shaped, CSC) micromixer (Fig. 1F). Dean vortices due to

the curved channel appear after the baffles, and the converging-diverging flow profile between

the baffle and the channel wall enhance mixing at Re≥27.

Another approach was taken by Sahu et al.22, who investigated mixing in a microchannel

integrating short narrow channel sections. Two types of obstructions were studied, namely a

staggered (Fig. 1Ga) and symmetric (Fig. 1Gb) arrangement. It was observed that the staggered

arrangement provided slightly higher (5%) mixing performance due to the presence of a cross-

stream velocity component. It was shown that mixing efficiency increases quadratically with

the number of obstructions due to increased residence times in the obstruction region. A larger

depth and width of the obstruction leads to larger turns of the flow, introducing larger secondary

flow that leads to higher mixing efficiency. The pressure drop is observed to be significantly

higher in the case of symmetric arrangements. In this type of mixer, a relatively high critical

value of Recr ~ 100 was found.

A sophisticated design termed an “alcove-based mixer” was proposed by Egawa et al.23

(Fig. 1H). The mixer consists of a T-junction, followed by three repeats of an alcove or cavity,

adjusted to the channel and arranged in a zig-zag manner. This mixer is capable efficiently of

mixing solvents with different viscosities (1.04-1.17 cP), due to recirculation of solution within

the alcoves to promote fluid mixing.

3.2 Microchannels with wall modifications

Another simple way to induce transverse flow in the microchannel is to insert obstacles or to

modify the channel wall with grooves. Special attention will be paid to mixers with grooves

fabricated in the channel walls.


49

Placing obstructions within a microfluidic channel offers a simple approach to enhance

mixing by chaotic advection. Obstacles alter the direction of flow, and the resulting swirling

flows and recirculation create transverse mass transport. The barriers are placed asymmetrically

in an alternating way inside the microchannel41 to provide even more chaotic flow patterns, due

to changing flow directions that force fluids to merge.42,43

Wang et al.42 numerically investigated different layouts of cylindrical pillars in a mixing

channel. This work showed that obstacles cannot generate eddies or recirculation at low Re.

However, mixing performance can be improved at high Reynolds numbers (Re ≥ 200). One of

the important findings was that an increase in the number of obstacles in the channel led to the

enhancement of the mixing. Later, Chen et al.44 reported a microfluidic mixer containing a high-

density array of pillars (Fig. 2A) that can provide fast mixing at very low Reynolds numbers

(Re≤1). The micropillars cause multiple splitting and reunification of laminar flows in the

channel. At a low flow rate of 0.1 µL/min, almost complete mixing was obtained due to this

‘‘split-and-recombination’’ effect (discussed more in Section 3.3), that decreases the thickness

of each fluid layer and provides shorter characteristic diffusional lengths. However, this effect

is highly reduced at higher flow rates and more clusters of obstructions are needed for complete

mixing. When the flow rate was increased to 5–15 µL/min, the mixing process starts to be

dictated by chaotic advection, and mixing performance is slightly enhanced. The mixer was

tested for mixing solutions with different viscosities (phosphate-buffered solution, gold

nanocolloids and 20% glycerol with Rhodamine 6G) at various flow rates (0.1-10 µL/min). As

expected, glycerol/Rhodamine 6G, due to its higher dynamic viscosity (1.76 cP), shows a

relatively lower mixing efficiency than the other solutions, and requires a distance of 35 mm

compared to 21 mm with phosphate buffer solution to obtain completet mixing.

C

ha

pte

r II


50

Figure 2. Micromixers with obstacles in the mixing channel: (A) A pillar obstruction micromixer: (a) schematic

diagram and (b) SEM image of micropillars in poly(dimethylsiloxane)(Modified from25); (B) An obstruction-based

micromixer with rectangular ribs (Taken from26); (C) T-shaped (a) simple, (b) wavy and (c) converging–diverging

micro-channels with rectangular ribs and (d) magnified rectangular rib placed on the channel floor (Modified

from17); (D) Micromixer with incorporated 2D and 3D baffles (a) 2D mixer with triangle-shaped mixing elements

and (b) 3D mixer with trapezoidal mixing units (Modified from28); (E) Mixer with cylindrical alcoves (Modified

from24). See Table 1 for geometric dimensions.

Another obstruction-based micromixer with optimized rectangular ribs was reported by

Bhagat et al.26(Fig. 2B). It provides ∼90% fluid mixing within 5 mm and is capable of achieving

particle dispersion with a wide range of particle sizes (190 nm - 1.9 µm), showing a 30%

increase in particle dispersion over a modified Tesla design45 (discussed in Section 3.3).

Hsieh and Huang 17 proposed mixers that can work at very low Re (0.027≤Re≤0.081)

(Fig.2C). Several T-shaped designs with rectangular ribs with simple (Tr) (Fig.2C-a), wavy

(Twr) (Fig.2C-b) and converging–diverging microchannels (Tcdr) (Fig.2C-c) were proposed.

Although all micromixers perform better at low Re, there was an established performance

superiority as follows: Twr＞Tcdr＞Tr. The periodically positioned ribs improve mixing

performance by altering the flow direction. However, the fact that better mixing is achieved at

lower Re indicates that the mixing is governed mainly by diffusion, which requires longer

residence time to occur. Probably, as in many other cases, there exists an Recr, above which the

mixing will become more efficient by increasing the flow rates.

Conlisk and Connor28 designed 2D- and 3D micromixers with triangle- (Fig. 2Da) and

trapezoidal-shaped (Fig. 2Db) baffles. The characterization within Re range 0.1–20 showed

(Recr = 1.0) that 90% of the mixing was achieved in 32 and 7 mm for the 2D and 3D mixer


51

respectively. The mixing is enhanced due to the focus-and-diverging effect. The 3D mixer

showed a significant increase in mixing efficiency (82% mixing homogeneity compered to a

simple T-mixer) by introducing transverse flow recirculation due to the shape of 3D baffles.

Figure 3. Micromixers with structures on channel walls: (A) Schematic diagram of slanted groove micromixer

(SGM) and (B) (a) Staggered herringbone mixer (SHM) and (b) chaotic mixing patterns in the channel (Modified

from 52); (C) Micromixer with both slanted and herringbone grooves (Taken from53); (D) Connected-groove

micromixer (CGM): (a) CGM-1; (b) CGM-2 (Modified from34); (E) Mixer with alternating slanted ridges on the

top and bottom of the channel: (a) Slanted Ridge Mixer Mirrored (SRM-M) and (b) Slanted Ridge Mixer Opposite

(SRM-O) and (c) 3D view (Taken from7); (F) Three-dimensional staggered herringbone mixer (3D SHM) (Taken

from8). See Table 1 for geometrical dimensions.

Recently, Wang et al.24 proposed designs with cylindrical alcoves extending from

microchannel walls (Fig. 2E) that varied in radius. In general, the design with smaller

cylindrical alcoves gave a 15% and 37%-increase in mixing performance compared to the

straight channel for Re 0.1 and 100, respectively. On the other hand, with the increase in Re the

efficiency of mixing decreased in all the mixers.

Modifying the channel wall is a powerful tool for creating chaotic advection, especially

at low Re numbers. This approach benefits from the low pressure drop and relatively easy

fabrication techniques due to the planar structure.46 Probably the most well-known examples of

patterned a wall of the channel are those micromixers incorporating slanted (SG) (Fig. 3A) and

staggered herringbone (SHG) grooves placed on the bottom wall (Fig. 3B, 3C). They have been

studied extensively.47–51

Grooves can generate transversal secondary flow similar to Dean vortices.52 In a slanted-

groove-micromixer (SGM, Fig. 3A) flow over the groove array assumes a chaotic helical or

Ch

ap

ter

II


52

corkscrew pattern, in which two solution streams twist around each other close to the bottom

of the channel. Fluid elements are stretched in a transverse direction due to the oblique position

of the groove with respect to the channel walls. This increases the contact area between two

adjacent solutions dramatically and facilitates mixing by diffusion. A detailed description of

the mechanism is given elsewhere.51 However, helical flow in a channel alone does not give

rise to chaotic mixing. In order to induce chaotic advection, it is necessary to superimpose

different recirculation patterns.54 This can be achieved with array of staggered herringbone

grooves (SHG) (Fig.3B), as described by Stroock et al.52 These structures generate a pair of

counter-rotating vortices that stretch and fold the mixing liquids, reducing the striation

thickness significantly.50 Repetition of these patterns leads to chaotic advection. A detailed

description of the mechanism can be found elsewhere.47,50,51

A variety of designs have been derived from this basic concept. For instance, Howell et

al.53 proposed a micromixer with both slanted and herringbone ridges, whereas some designs

employ grooves on both top and bottom walls (Fig. 3C) 53,55 or on the side and top walls (Fig.

3D-3E).8,34 The design proposed by Howell et al.53 (Fig. 3C) with both slanted and symmetric

herringbone ridges (chevrons) aims to improve the mixing using the combined mechanism: the

chevrons generate two equally-sized vortices that drive fluid upward in the center of the channel

and downward toward the sidewalls. On the other hand, the SG creates two vortices, one above

the other. Such a design allows the formation of a pair of counter rotating vortices in vertical

and horizontal planes, which creates far more rapid mixing than previous designs. Later, Floyd-

Smith et al.55 showed that grooves on the top and bottom of channel improve mixing by 10%

over micromixers with grooves placed only on one channel wall.

In designs where connected grooves are composed of bottom grooves and sidewall

grooves conjoined across the adjacent walls, the sidewall grooves assisted in inducing an

intensive helical motion. This situation was observed in connected-groove micromixer with

slanted grooves on the bottom and sidewall grooves (CGM, Figure 3D).34 From the bottom

grooves the fluid is guided along the sidewall grooves, then to the top and back to the main

stream. Such design can increase the helical intensity by 20%. Recently, Van Schijndel et al.7

proposed a mixer with alternating slanted ridges on the top and bottom of the channel (Fig. 3E).

Adding mixing elements to both walls promoted lateral mass transport and assisted in the

formation of advection patterns, which increased mixing efficiency. Lin et al.8 theoretically

and experimentally showed a micromixer with staggered herringbone grooves patterned on both


53

bottom and side walls (3D Staggered herringbone mixer, SHM, Figure 3F) that reduced the

mixing length by almost half as compared with the originally reported SHM mixer.52

The simulations confirmed that the flow pattern in the mixers with staggered

herringbone grooves is almost independent of Reynolds number.1 SHG improve mixing for a

wide range of Re from 1 to 100.34,52 However, a dependence of efficiency of mixing on flow

rate (different Re) is observed, implying the existence of an Recr (that was mentioned before)

can be observed for grooved mixers as well. It was shown that in the connected-groove

micromixer,34 the distance required for complete mixing for Re>10 decreased with increasing

flow rate because the inertial forces start to dominate over viscosity.

3.3 3D convoluted channels (combined principles)

As shown previously in this Section, simple channels can generate chaotic advection at higher

Re. However, the mixing at low Re (<1) remains a problem in these designs. To overcome this,

a large number of novel three-dimensional serpentine (3D convoluted, 3D twisted) designs

based on planar micromixers (Section 3.1 and 3.2) have been proposed over the last decade.

The mixing in such micromixers is enhanced by the superposition of several mechanisms,

mostly the combination of chaotic advection and the splitting-and-recombination principle

(SAR). The complex 3D geometry of such mixers causes continuous splitting, recombination

and collision of flows at the same time. In general, chaotic advection in this type of micromixer

can be induced at high flow rates, Re˃70, while the SAR mechanism works well at lower Re,

decreasing the operational range of such mixers to 5<Re<30. Due to the combination of mixing

principles, the distance required for complete mixing in these mixers is much shorter than in

mixers based only on chaotic advection.

A good example of such a micromixer is the serpentine laminating micromixer (SLM)

developed by Kim et al.56 The mixer consists of ‘‘F’’- shaped units arranged in two layers (Fig.

4A) that cause continuous splitting and recombination, keeping the same flow path length for

the two split streams. This SAR principle governs mixing at lower Re. As Re increases, the

serpentine channel design starts to induce chaotic advection. Thus, efficient mixing in the SLM

can be achieved for a wide range of Re (0.44<Re<12.3). Compared to a T-micromixer, the SLM

design requires a 20-times shorter distance to achieve complete mixing. Later, an improved

serpentine laminating micromixer (ISLM) was developed within the same group by Park et

al.57. It was shown that the reduced cross-sectional area in the recombination region enhanced

Ch

ap

ter

II


54

the advection effect, which helped achieve better vertical lamination. This change results in

improved mixing performance: at Re 0.2 and 20 at least a 1.2-fold shorter distance was required

to achieve complete mixing for the ISLM compared to the SLM.

Xia et al.58 designed and investigated several configurations of two-layer crossing

channels in the micromixers (TLCCM, Fig. 4B). All three mixers have a two-layer structure. It

is thought that the complex 3D geometry of the microchannels would impose perturbations on

the flow. However, Model 1 fails to generate chaotic advection at Re<1, which can be attributed

to a lack of fluid inertial effects. Model 2 was found to be only a partial chaotic mixer, exhibiting

incomplete mixing at Re=0.01. On the other hand, rapid mixing can be achieved at Re<1 for

Model 3. When Re increased to 10, the mixing became even better due to promotion of chaotic

advection. Further improvement was observed at Re=60. Recently, several similar designs,

namely a tangentially crossing channel mixer (Fig. 4C)9 and a micromixer with XH-shaped and

XO-shaped elements (Fig. 4D),10 both utilizing the combination of SAR and chaotic advection,

were proposed. Both of these designs give a good performance for mixing fluids are a wide

range of Reynolds numbers, 0.1<Re<10 and 0.3<Re<60, respectively.

Another micromixer with 3D square-wave structures and cubic grooves (Fig. 4E), that

expands Re to a wider range (30<Re<220), was proposed by Lin et al.11 The main flow path of

the micromixer has a square-wave shape in order to facilitate laminar flow recirculation by

vortex generation, followed by stretching of these vortices in the cubic groove. The mixer shows

good performances in the range of 0.675 - 4 mL/min flow rates. In addition, the proposed

micromixer featured a stainless steel body, making it resistant to high temperature, high

pressure, and strong corrosion, which can be beneficial in many analytical applications.


55

Figure 4. 3D convoluted channels: (A) Serpentine laminating micromixer (SLM) (Taken from56); (B)

Configurations of two-layer crossing channels in the micromixer design: (a) Model 1, (b) Model 2 and (c) Model

3 (Modified from58); (C) Tangentially crossing channel (TCC) mixer (Modified from9); (D) SAR micromixer with

(a) XH and (b) XO elements (Modified from10); (E) The micromixer with 3D square-wave structures and cubic

grooves (Modified from11); (F) SAR µ-reactor (a) side view and (b) mixing unit (Modified from59); (G) Horizontal

and vertical weaving micromixer (HVW mixer)(Modified from12). See Table 1 for geometric dimensions.

Fang and Yang59 designed a SAR µ-reactor (Fig. 4F) suitable for mixing fluids with

viscosities over a wide range (0.9–186 cP) for 0.01<Re<100. The 3D structures inside the mixer

cause stream cutting, separation and recombination utilizing the SAR principle. On the other

hand, the mass transfer of fluids between upper and lower halves of the channel induces a 3D-

counter-clockwise flow. The repetitive overlapping of flows forces them to collapse and stretch,

which is a characteristic of chaotic advection. Results showed that at high flow rates, such as at

Re>50, mixing becomes dominated by inertial forces and the complete mixing of fluids can be

achieved within the first 6 mm of the length of the mixer. Furthermore, authors assessed the

mixing behavior of fluorescent proteins (C-phycocyanin and R-phycoerythrin) in 88% glycerol

with a confocal microscope. Results revealed that the SAR µ-reactor exhibit only a small

difference (10–15%) in mixing efficiency when mixing highly viscous fluids (186 cP) as

compared to slightly viscous fluids (0.9 cP). This difference for the micromixer with slanted

grooves was 40–45%,59 which indicates that the mixing of viscous fluids can be achieved more

efficiently using a SAR µ-reactor.

Ch

ap

ter

II


56

Recently, a horizontal and vertical weaving micromixer (HVW mixer, Fig. 4G) with

crossed barriers inside a microchannel was proposed.12 Barriers cause two fluids to be divided

into upper and lower layers followed by the generation of clockwise and counter clockwise

motion both vertically and horizontally. The unique feature of this mixer is that only a very

short distance of 450 µm is required, to obtain 89.9% mixing efficiency at a Reynolds number

of 5. The overall channel width is 300 µm, channel depth is 200 µm and barrier dimensions

were 50×100 µm (width by depth).

Another mixer for mixing fluids with widely different viscosities (in ratios of up to 104)

has been reported by Xia et al.15 The mixer with interconnected multi-channel network (Fig.

5A) also employs two mechanisms to improve the mixing. First, through splitting and

recombination, the bulk fluid volumes are broken into thinner streams and chaotically

recombined together. Afterwards, the multiple fluid streams enter a circular expansion chamber,

where viscous flow instabilities lead to turbulent fluid motion. At flow rates higher than 0.20

mL/min, the initial occurrence of flow instability is observed. However, at lower flow rates, no

flow instability occurs, which reduces the quality of mixing. The mixer was tested for mixing

glycerol (680 cP) and other viscous samples (5440 cP, 17300 cP and 54600 cP) with aqueous

solutions (~1 cP). As expected, the mixer becomes less efficient at increased viscosity ratios.

However, complete mixing is still obtained by the end of the mixer (after 8 mixer units) for all

tested mixtures.

Li et al.13 developed an overbridge-shaped micromixer (OBM, Figure 5B) that was used

for mixing two fluids under both isocratic and gradient conditions with Re values of 0.01-200

(Recr=10), corresponding to 0.0045 - 900 µL/min flow rates. The mixer was compared to the

previously discussed SLM micromixer with F-shaped units [Fig. 4A],56 which revealed that

mixing performance of the OBM was always higher (>90%) comparing to F-shaped mixer

(<60%) at the same Reynolds number. Numerical simulation showed that a mixing efficiency

of more than 90% can be achieved for mixing fluids with different flow rate ratios ranging from

1:9 to 9:1, which can be useful in analytical and biological applications. The success of the

OBM mixer can be explained by the combination of different designs used. The mixer consists

of overbridge-shaped (OB) and square-wave (SW) channels. The OB channel has a branched

structure, which split a single fluid stream into two sub-streams. One sub-stream flow together

with the second fluid stream through the main SW channel, where the interface between streams

is stretched at sharp turns. The other sub-stream is transported to the other side of the channel


57

and collided with the main stream at a 90° angle, which will increase the contact area between

fluids.

Liu et al.18 proposed a novel cross-linked dual helical micromixer (CLDH, Fig. 5C) that

consists of double helical channels rotating in opposite directions to create repeated crossing

regions. This mixer employs flow collision to stretch, split and fold streams that recombine in

the crossing regions. Chaotic advection is enhanced with the sharply twisting streams on the

basis of helical flow and flow collision where Re>1. The simulation and experimental results

show that 99% mixing can be achieved in four cycles (320 µm) over a wide range of Re (0.003–

30).

Figure 5. 3D convoluted micromixers. (A) (a) Plain view and (b) a profile of the mixer (Taken from15); (B) (a) 3-

D overbridge-shaped micromixer (OBM) with (b) its mixing unit (Modified from13); (C) 3D cross-linked dual

helical micromixer (CLDH) (Taken from18); (D) 3D Tesla micromixer (Modified from19); (E) Micromixer with

shifted trapezoidal blades (STB) (Modified from60); (F) H-C passive micromixer (Modified from20); (G) “Twisted”

3D microfluidic mixer (Modified from21). See Table 1 for geometric dimensions.

Another possible approach for the creation of chaotic advection is the combination of

Taylor dispersion with Dean vortices. Hong et al.61 proposed to use an in-plane micromixer

with modified Tesla structures. This mixer exploits the Coanda effect, which enhances

convective mixing of the fluids by producing transverse Taylor dispersion. Recently, Yang et

al.19 designed a micromixer with three-dimensional Tesla structures (Fig. 5D). A repetitive

distortion and squeezing of flow occurs at the turning joints of the Tesla structures that generate

transverse dispersion. Moreover, an added layer of Tesla structures provides more flow

disturbances, which improves mixing. The efficiency of mixing reached 94% in the flow rate

Ch

ap

ter

II


58

range of 0.9 - 900 µL/min (0.1<Re<100). The application of this mixer will be discussed in

Section 4.2.2. Recently, another two micromixers were proposed for mixing in the similar range

of Re (1<Re<100): a micromixer with shifted trapezoidal blades (STB, Figure 5E)60 and an H-

C micromixer (Fig. 5G)20. The mixing efficiency was 80% and 90%, respectively.

Sivashankar et al.21 proposed a new “twisted” 3D microfluidic mixer with a two-layered

quadrant of circles (Fig. 5F). Mixing is enhanced due to chaotic advection through generation

of vortices at the edge of the arc-shaped channels, with additional splitting and recombination

of flows. These micromixers can operate at low (1.0 µL/min) and high (1.0 mL/min) flow rates

without reduction in the mixing performance. Moreover, the proposed mixer showed a good

mixing efficiency at high flow rates for mixing 98% glycerol (919 cP) with water (1 cP), making

this mixer ideal for a variety of applications where highly viscous solutions have to be mixed

at high flow rates (~ 1.0 mL/min).

Table 1 summarizes different types of micromixers based on chaotic advection with their

dimensions and material/fabrication methods. We highlighted the mixers from the current

Section applications, which will be shown in Section 4. The lines in red colour marks the mixers

that found a real application that was proposed in the original paper. The lines in blue highlights

the application in which the original or modified mixer designs from the original study were

used.


59

Table 1. Micromixers based on chaotic advection.

Name of the mixer Re* Dimensions Material/fabrication method

Ref Application area

CHAOTIC MICROMIXERS WITH SIMPLE GEOMETRIES

Spiral S-shaped channel 0.02< Re < 18.6 w=150 µm; h = 29 µm SEBS/ printed circuit technology; single planar soft lithography

35 A size-based particle filtration device;86 a microreactor

Planar labyrinth micromixer (PLM) with S-shaped geometry

Re = 2.5; 30 h = 267 µm; w = 220 µm; the spacing - 240 µm

PDMS/single-step soft-lithography

5

Spiral-shaped, interlocking-semicircle and Ω channel designs

0.01< Re < 50 Recr = 10

h = 230 µm; w = 200 µm; L = 22 mm

PDMS bonded to a glass/soft-lithography

6 for systems working under continuous flow conditions

Zig-zag channel 80< Re < 267 Recr = 80

h = 48 µm; w = 100 µm; L = 2 mm; s = 100-800 µm (s - periodic step)

Polyethyleneterephthalate (PET)/an excimer laser

36 A microreactor: polymerizations of styrene in cyclohexane; ultrasensitive trace analysis62

Curved-straight-curved (CSC) micromixer

Re = 1; 9; 81 w = 130 µm; h = 130 µm; L=1.95 mm; baffle thickness 40 µm; w(radial baffles) = 97.5 µm

PDMS bound to glass/soft lithography

27 As microreactor

Microchannels with lateral obstructions Recr = 100 w = 50 µm; h = 50 µm (total); L = 66 mm;

SU-8 - PMMA/photolithography and micro-milling

22 In DNA hybridization analysis

Alcove-based mixer with a triangular obstruction

Re ˂ 400 h = 82 µm; w = 20 µm; alcove: w = 30 µm; l = 40 µm

Silicon/standard photolithographic techniques

23 For handling complex biochemical and chemical reactions in parallel; mixing fluids with different viscosities

CHAOTIC MICROMIXERS WITH OBSTACLES IN THE MIXING CHANNEL

Pillar obstruction channels

Re: 0.289-0.354 (0.1–15 µL/min) Recr ≥ 5 µL/min

h = 45 µm; w = 200 µm; L= 35 mml; pillars: h = 45 µm; w = 15 µm;


44 to mix solutions with different viscosities; 44 capturing bioparticles on the immobilized surfaces85

The obstruction micromixer with rectangular ribs

Re = 0.05 h = 50 µm; w = 100 µm PDMS/soft lithography 26 Particle dispersion with a wide range of particle sizes26

Simple T-shaped-, T-shaped wavy- and T-shaped micro-channel with rectangular ribs

0.027 ˂ Re ˂ 0.081

w = 200 µm; h = 200 µm (total); L = 10.1 mm; obstacle sizes: w = 50 µm; l = 100 µm; h = 80 µm

PDMS – PDMS/two-step soft lithography

17 Capturing bioparticles on the immobilized surfaces

Mixers incorporating 2D and 3D baffles 0.1 < Re < 20 Recr = 1

w = 100 µm; h = 50 µm; L = 5.23 mm

PMMA- PMMA/an excimer laser beam

28 As microreactors; in DNA hybridization analysis

Microfluidic mixer with cylindrical alcoves or grooves (CG)

1 < Re < 10 w = 200 µm; h = 100 µm; rCG = 100; 200; 300 µm; L = 20 mm


24 For biochemical and medical diagnosis

GROOVES IN THE CHANNEL

Slanted groove micromixer (SGM) 0 < Re < 100 w = 200 µm; h = 70 µm; grooves: d = 14 µm

PDMS bound to glass/2-step soft lithography

52 A microreactor: synthesis of a statistical-copolymer-brush composition


60

gradient;63 continuous glucose monitoring;72 on-line chemical modification of peptides and direct ESI-MS analysis94

Staggered herringbone mixer (SHM) w = 200 µm; h = 77 µm; grooves: d = 17.7 µm

For parallel screening in situ click chemistry;65 as microreactor: production of siRNA-LNPs67–69 and continuous glucose monitoring;71 trapping of particles and DNA hybridization;758283,84 trace analysis: sarin in blood89 and cobalt (II) ions and hydrogen peroxide;93

changing mobile phase composition between dimensions in LC×LC;99 enzymatic digestion, one of the key functions of the gastrointestinal tract100.

Grooves placed on the top and bottom of the channel

0.06 < Re < 10 w = 3.175 mm; h = 0.76; 1.02; 1.27 mm; d = 0.94 mm

PMMA – Plexiglas/milling 53

Binding reactions (for DNA extraction); trapping of particles; enrichment and focusing of beads and cells; in immunoassays (trapping cancer cells on the antibody-coated surface); in environmental analysis

Re ≤ 30 w = 200 µm; h = 60 µm PDMS/soft lithography 55

Connected-groove micromixer (CGM) 0.28 < Re < 112 w = 200 µm; h = 70 µm; L = 1.7 mm; grooves: w = 50 µm; d = 30 µm

PDMS bound to glass/two standard photolithography

34

Slanted ridge mixer (SRM) Re ~ 1 (10 µL/min)

w = 185 µm (bottom); w = 120 µm (top); h = 90 µm; L = 43 mm; ridges: w = 70 µm; h = 20 µm

a glass plate bound to glass/two-step SU-8 process

7

Three-dimensional staggered herringbone mixer (3D SHM)

Re ~ 0.7 w = 200 µm; h = 80 µm; grooves: h = 20 µm (bottom); h = 40 µm (side); w = 60 µm

fused silica bound to PDMS/femtosecond-laser-assisted chemical wet etching

8

3D CONVOLUTED CHANNELS Serpentine laminating micromixer (SLM) with ‘‘F’’- shape units

0.44 < Re < 12.3 w = 250 µm; h = 60 µm; L = 10 mm

COC/hot embossing; injection molding

56 In diagnostic devices (for blood typing);88 in analytical chemistry and separation science (e.g., for gradients formation); as microreactor

Improved serpentine laminating micromixer (ISLM) with ‘‘F’’- shape units

Re =0.2; 2; 20 w = 500 µm; h = 300 µm PDMS bound to glass/soft lithography

57

Two-layer crossing channels: TLCCM, model A and model B

0.01< Re < 0.2 w = 300 µm; h = 1500 µm PMMA/the laser ablation method

58

As microreactors Tangentially crossing channel (TCC) mixer

0.1 < Re < 10 w = 100 µm; h = 50 µm PDMS-PDMS/soft lithography

9

Micromixer with self-rotated contact surface (XH and XO models)

0.3 < Re < 60 w = 450 µm; h = 150 µm; L =10.25 mm

PDMS-PDMS/multilayer soft lithography

10

SAR µ-reactor 0.01 < Re < 100 w = 300 µm; h = 100 µm; L =6.15 mm

PDMS/standard photolithography

59 Mixing of fluids with different viscosities59


61

Micromixer with 3D periodic perturbation

30 < Re < 220 h = 300 µm; L = 50 mm stainless steel/conventional machining

11,101 In analytical chemistry (liquid chromatography); operations under pressure and temperatures

3D structures resembling teeth (alligator teeth-shaped micromixer)

0.08 < Re < 16

w = 300 µm; h = 100-300 µm, L = 20 mm; the triangular structures: w = 300 µm, h = 300 µm; d = 50, 100, 150 µm

PDMS-PDMS/soft lithography

64

as microreactors for continuous glucose monitoring,64 DNA hybridization assays;73,74,76,77 for an ultrasensitive trace analysis of cyanide90

3-D overbridge-shaped micromixer (OBM)

0.01 < Re < 200 Recr=10

w = 100 µm; h = 50 µm; L = 2 mm

Three layers of PDMS.single-step soft lithography

13 Formation of gradients (at different flow rate ratios)13

Horizontal and vertical weaving micromixer (HVW mixer)

Re = 5 w = 300 µm; h = 200 µm; L = 1.2 mm; barriers: w = 50 µm; d = 100 µm

PDMS – PDMS/soft lithography

12 Binding reactions (for DNA extraction); trapping of particles

A micromixer with interconnected multi-channel network

Re ~ 2.8 (400 µl/min)

w1 = 600 µm, w2 = 450 µm, w3= 750 µm; h = 400 µm; dchamber = 3.45 mm.

PMMA – PMMA/CNC micro-milling

15 In analytical chemistry and separation science (e.g., for gradients formation)15

Micromixer with shifted trapezoidal blades (STB)

0.5 < Re < 100 Recr = 5

w = 210 µm; h = 200 µm PDMS-glass/soft lithography 60 In clinical and environmental analyses or diagnostic systems

3D cross-linked dual helical micromixer (CLDH)

0.003 < Re < 30 D(helical)=60 µm; P(helical)=80 µm, separation distance: 21 µm

fused silica/femtosecond laser wet etching (FLWE) technology

18 -

Micromixer with modified Tesla structures

0.1 < Re < 100 w = 200 µm; h = 100 µm; L = 11.2 mm

PDMS/soft lithography 19

In immunofluorescence experiments (for binding reaction of antibodies for detecting antigens of lung cancer cells);19 a microreactor: for fabrication of homogenous lipid-polymeric and lipid-quantum dot nanoparticles;66 formation of gradients in liquid chromatography95

H-C passive micromixer 1, 30, 50, 100 Recr = 30

wmax= 600 µm, wmin= 400 µm; hmax= 1300 µm, hmin= 400 µm

PC/micromilling 20 -

“Twisted” 3D microfluidic mixer 0.02 < Re < 20 (1, 5, 10, 100, 1000 µL/min)

w = 200 µm; h = 200 µm; L = 30 mm

PMMA – PMMA/CO2 laser system, thermal bonding

21

The mixing of various viscous fluids For diagnostic devices (cell analysis);21 integrated systems for study of reaction kinetics, sample dilution, and improved reaction selectivity.

* Experimental values unless stated different.

The lines in red colour marks the mixers that found a real application that was proposed in the original paper. The lines in blue highlights the application in which the original or modified mixer designs

from the original study were used. Other applications indicated in black are possible applications for each particular design.

PDMS - Poly(dimethylsiloxane); PC – Polycarbonate; COC – Cyclic olefin copolymer; PMMA – Poly(methyl methacrylate)

PETG - Polyethylene terephthalate glycol; SEBS - Polystyrene-Polyethylene-Polybutylene-Polystyrene


62

4. Application of the passive micromixers based on chaotic

advection

Microfluidic systems are widely used in biology, biotechnology and chemistry. Most of these

applications involve complicated (bio)chemical reactions that require mixing.19 Micromixers

based on chaotic advection have found their application as microreactors;62–72 and in biological

applications in the analysis of DNA,73–81 sorting of particles and cells,19,25,82–86 improvement

of diverse cell culture platforms87 and in full integrated lab-on-the-chip devices for blood

typing88 or for detecting a trace amount of sarin in whole blood.89 In analytical chemistry

chaotic micromixers have been used for analysis of hazardous compounds (e.g. cyanide,

pesticides, malachite green);89–93 on-line chemical modification of peptides in an LC-MS

interface;94 mixing liquids with different viscosities15,21,44,59 and for gradient formation.13,95

4.1. Microreactors for chemical reactions

Micromixers as microreactors possess some unique features that are advantageous for using

them for performing various chemical reactions. First, the microscale mixing time is usually

equal to or even less than the reaction time. Of course, micromixers can not produce a large

amount of product comparing to the macroscale production, however, the relative reaction

yield can be higher and the synthesis can be performed in a more controllable way. Besides, in

the micromixers the small thermal inertia and the uniform temperature provide improved

control over mass and heat transfer.1,65 This allows the synthesis of more homogeneous highly

reproducible reaction products. The small volume of the microreactors also provides an

opportunity for green syntheses by reducing the use of hazardous reagents, which makes the

production more cost effective, and safe.96 At the same time, the larger surface-to-volume ratio

provides more surface for catalyst incorporation.

There are a few examples of utilizing chaotic mixers for polymerization reactions: a

synthesis of a statistical-copolymer-brush composition gradient using a mixer with slanted

grooves63 and polymerizations of styrene in cyclohexane in zig-zag microchannels.62 The flow

rate in these applications was relatively high: ~0.15 - 0.3 mL/min. Both studies showed that

the passive mixing induces by flow only allows more controllable processes in the

microchannels, either for obtaining polymers with narrow molecular mass distribution62 or for

the fabrication of surface materials with well-defined composition gradients.63


63

Figure 6. (A) Schematic representation of a chemical reaction circuit used for the parallel screening of

an in situ click chemistry library (Adapted from65); (B) Schematic illustration of nanoprecipitation of

lipid polymeric NPs (a) in microchannel with Tesla structures (discussed in Section 3.3) and (b)

micrograph of mixing process between fluorescent dye and water at total flow rate 55 µL/min (Modified

from66); (C) Schematic illustration of lipid nanoparticle (LNP) small interfering RNA (siRNA)

formulation inside staggered herringbone micromixer (SHM) (Modified from67,68); (D) Schematic

illustration of LNP formation in channel with groove structures for rapid mixing (Modified from69); (E)

The ceramic microreactor design for the synthesis of core-shell nanocrystals with a three-dimensional

serpentine micromixer for the formation of the core quantum dots and a longitudinal channel for the

shell formation (Taken from70).

In 2006 Wang et al.65 described a new type of microfluidics-based chemical reaction

circuits for the parallel screening of 32 in situ click chemistry reactions. This approach allows

to synthesize a library of high-affinity protein ligands from the complementary building block

reagents via irreversible connection chemistry. In this work click reactions between acetylene

and azide was chosen as a model system. Figure 6A shows how this performed in practice.

First, a nanoliter-level rotary mixer (nL-Level mixer with a volume of 250 nL) selectively

sample nL-quantities of reagents - acetylene and azides with/without inhibitors - for each

screening reaction. Then, reagents enter the microliter-level chaotic mixer (µL-Level mixer)

and mixed with mL-quantities of bovine carbonic anhydrase II (bCAII) solution by means of

chaotic advection inside the 37.8-mm long microchannel. Afterwards, the homogeneous

reaction mixtures are guided by microfluidic multiplexer into one of the 32 individually

addressable microvessels for storing. As a result of these manipulations, 10 different binary

azide/acetylene combinations are obtained: 1) ten in situ click chemistry reactions between

Ch

ap

ter

II


64

acetylene and azides in the presence of bCAII; 2) ten control reactions performed th same as

in (1), but in the presence of inhibitor; 3) ten thermal click chemistry reactions performed as

in (1), but in the absence of bCAII. The total volume of the system is only 4 µL, which allows

to reduce the consumption of reagents in 2.5-11 times compared to the conventional method

using 96-well plates.

A very good example of utilizing the micromixers as microreactors is their application

for synthesis of lipid nanoparticles (LNP)66 and their complexation with small interfering RNA

(siRNA).67–69 In order to obtain monodisperse LNP siRNA systems with minimum sizes that

exhibit better gene silencing potency, faster mixing rates (higher flow rates) are required.68 The

conventional techniques for encapsulation of nucleic acids require milliliters of expensive

nucleic acid solution and do not provide good homogeneity and reproducibility.69 To overcome

this, Valencia et al.66 have developed a PDMS-based microfluidic mixer consisting of Tesla

structures for fabrication of monodisperse homogenous lipid-polymeric and lipid-quantum dot

nanoparticles (Fig. 6B). Other studies67–69 have utilized a staggered herringbone mixer for

production of siRNA-LNPs (Fig. 6C-D). Later, Rungta et al.97 showed the efficient silence

neuronal gene expression in cell culture and in vivo in the brain using LNPs produced this way.

Pedro et al.70 have reported an automatic microreactor for the easy and controlled

synthesis of water soluble quantum dots (CdS and CdS/ZnS) for in situ optical characterization.

Homogeneous, stable and highly reproducible nanocrystals have been obtained due to a

hydrodynamic focusing of reagents and the introduction of three-dimensional micromixers for

efficient mixing (Fig. 6E).

Several studies used micromixers with staggered herringbone grooves,71 slanted grooves72

and three-dimensional structures resembling teeth64 as microreactors for continuous glucose

monitoring. For these experiments relatively low flow rates in the range of 0.37-75.0 μL/min

were used. However, when the sample flow rate increases from 10 to 70 μL/min in a SHG

mixer71, a decrease in the detected signal was observed, apparently due to insufficient reaction

times. On the other hand, in micromixers with three-dimensional structures64 a mixing

efficiency between 81% to 92% was determined for the full range of the tested flow rates (0.37-

74.6 μL/min).

4.2. Biological applications

Biological processes, such as cell activation, enzyme reactions and protein folding, often

involve reactions that require mixing of reactants for their initiation.78


65

4.2.1. DNA analysis

Nucleic acid (NA) probe assays have an enormous scope of applications in biotechnology and

medicine in order to identify genes and mutants, to map their correlations, and to analyze their

expression.75,78 DNA microarrays involve multi-component biochemical reactions that use

thousands of oligonucleotides, complementary DNA (cDNA) clones or polymerase-chain-

reaction (PCR) products.75 Therefore, the sample and reagents should be completely mixed in

order to achieve good results. However, the fact that reagents are immobilized means that

hybridization in the conventional way may take 8–24 hours due to the diffusion-limited

kinetics.73,75

Recently, microfluidic devices started to attract attention for DNA probe assays due to

their low costs, good performances, and ability to be used for different assays by just changing

the nature of the reagents.78 However, the fundamental problem faced by DNA-microarray in

microfluidic devices remains: slow transport of DNA molecules to the probes at low Reynolds

numbers.75 To overcome this, many researchers have used microfluidic mixers based on

chaotic advection. Several different designs of micromixers have been used for this application,

including a three-dimensional serpentine mixer,78,79 mixer with overlapping channels,80 an

alligator teeth-shaped micromixer73,74,76,77 and mixer with herringbone grooves.75

Very often, modern diagnostic techniques require the isolation and purification of

nucleic acids directly from patient samples. Several studies78,79 reported utilization of three-

dimensional serpentine micromixers for DNA extraction based on binding reaction to the glass

surfaces. Lee et al. reported a DNA purification from a biological sample using a microfluidic

mixer for a stepwise change in salt concentration.79 Under high-salt buffer DNA, which is

negatively charged, is strongly adsorbed on the glass surface. Afterwards, under a low-salt

buffer conditions, adsorbed DNA was eluted from the glass. Due to the fact that other

components of the sample (e.g. proteins or sugars) are weakly charged, DNA absorption occurs

in a selective manner and allows its purification.

The group of S. Lee73,74,76,77 had been working on the development of DNA

hybridization assays using an alligator teeth-shaped PDMS microfluidic mixer (Figure 7A).

The channel of this micromixers contains an array of upper and lower teeth that are responsible

for the fluid dispersion of confluent streams in both transversal and vertical direction.64 First,

Park et al.77 investigated the rapid and highly sensitive detection of duplex dye-labelled DNA

sequences. A mixer was used for efficient mixing between DNA oligomers and aggregated

Ch

ap

ter

II


66

silver colloids and was operated under low flow rate of 0.37 - 74.6 µL/min. It should be noted

that the mixing under the flow rate of 74.6 µL/min was not complete.

Later, Yea et al.73 used an alligator teeth-shaped PDMS microfluidic channel for the

lab-on-a-chip-based DNA hybridization analysis (Fig. 7B). The micromixer was used to obtain

efficient mixing between the probe and target DNA oligomers at a flow rate of 1 µL/min. Kim

et al.76 and Jung et al.81 then used a molecular beacon, a stem–loop DNA oligonucleotide

labelled with two fluorescent dyes as a probe DNA to analyze a target DNA with 20 base pairs.

Finally, Chen et al.74 reported a fast and sensitive online detection technique for label-free

target DNA based on changes in the FRET (Fluorescence Resonance Energy Transfer) signal

resulting from the sequence-specific hybridization between two fluorescently labelled nucleic

acid probes and target DNA in a PDMS microfluidic channel (Fig. 7C).

Figure 7. (A) Scheme of alligator teeth-shaped micromixer;77 (B) Microfluidic channel for DNA hybridization

with marked boxes for the FRET measurement areas (Adjusted from73); (C) (a) An alligator teeth-shaped mixer

with (b) schematic drawing of an alligator-teeth-shaped channel,91 that was used for DNA hybridization: two

fluorescently labelled nucleic acid probes were mixed first, and then a target DNA oligonucleotide was added;

(Modified from74); (D) (a) Optical micrographs of the PDMS device with two identical chambers, loaded half

with red and half with blue solution, and (b) the situation when the pump start to circulate solution clockwise

between chambers and the bridge channels with herringbone grooves (HG) provide mixing of red and blue

solutions(Modified from75).

Another system for DNA hybridization (Fig. 7D) was constructed by Liu et al.75 A

microfluidic chip consisted of two identical hybridization chambers (6 × 6.5 × 65 mm, 5 mL)

for solution circulation, which were connected to each other through the bridge channels with

herringbone structure. When chambers are loaded with a sample and a peristaltic pump starts

to circulate solution between chambers, the fluids passing through the bridge channels with


67

herringbone grooves experience chaotic advection which leads to their mixing. It takes about

16 min to complete the circulation between two chambers. In this device DNA hybridization

was performed under dynamic and static (control) conditions using two identical PDMS

devices. The cDNA samples were loaded into the PDMS devices and by actuating the

peristaltic pumps in one of the devices a dynamic hybridization was performed, while static

hybridization was performed in the other device as a control. After hybridization at 52°C for 2

h, PDMS devices were peeled away and scanned to recieve fluorescence images. It was

observed that the hybridization in the dynamic conditions when microfluidic chaotic mixer was

used, produced stronger signals and better sensitivity than under static conditions. Besides, it

was demonstrated that a mixer with herringbone grooves enhances DNA hybridization signals

3-8-fold and improves sensitivity by nearly one order of magnitude relative to the conventional

methods over the same length of time. The device is disposable and compatible with high-

density microarray slides with a potential to hybridize about 135 000 array features in a single

experiment.

4.2.2. Sorting/separation of particles and cells

Trapping, manipulation and separation of bioparticles (such as viruses, DNA molecules,

bacteria, and cells) are important goals for biotechnology in order to develop a good therapeutic

understanding of many diseases.86 One way to achieve this is by using microconcentrators

based on dielectrophoresis (DEP) for collecting these particles on electrodes. However, this

approach is only effective, when the sample particles are brought close to electrode surfaces.

Microfluidic mixers that generate a complex helical flow can recirculate particles and bring

them closer to electrode surfaces (usually on the bottom of the channel) increasing the

percentage of particles that get trapped. Lee and Voldman82 designed a microconcentrator with

patterned grooves to enhance the trapping of particles at high flow rate (100-500 µL/min),

which also means that particle interaction with the electrode surface was strong (Fig. 8A).

Different types of grooves – the slanted groove, staggered herringbone, and symmetrical

herringbone mixers – were investigated. The amount of trapped particles (1-µm-size magnetic

microspheres) in the staggered herringbone mixer increased by a factor of ~1.5x compared to

a smooth channel. Here, the generation of vertical flow is even more important than achieving

the mixing itself. Later, for enrichment and focusing of microbeads (6, 10, and 20 μm in

diameter) and mouse dendritic cells, Chen et al.25 used single slanted and V-shaped grooves

under very low flow rate of 0.27 µL/min. Such a difference in the flow rate used is probably

due to the lower efficiency of the symmetrical V-shaped grooves to generate vertical fluid

Ch

ap

ter

II


68

motion (and, thus, a longer residence time is required for trapping particles) comparing to the

staggered herringbone mixer).

Figure 8. (A) (a) Illustration of the assembled device with a channel and gold interdigitated electrodes (IDEs) on

the Pyrex substrate; (b) particle trapping in chaotic flow (Modified from82); (B) The HB-Chip with a microfluidic

array of channels illustrating (a) the uniform blood flow through the device and (b) a micrograph of the grooved

surface with a side view (Modified from84). (C) An integrated device for capturing circulating tumor cells (CTCs):

(a) a patterned silicon nanopillar (SiNP) substrate with anti-EpCAM-coating, and (b) an overlaid microfluidic

chaotic mixing chip (Modified from85); (D)(a)Schematic illustration of an integrated lab-on-a-disc system: two

inlets, mixing unit and the connected inertial flow-focusing channel; (b) an upstream section of the mixing unit

processing blood-spiked and microbead solutions (Taken from86).

Interestingly enough, a number of micromixers were used for the circulation of tumor

cells by modifying the mixer surface. Idea of trapping particles come back in different

applications. The group of Mehmet Toner83,84 has reported a microvortex manipulator (MVM)

with patterned chevrons (symmetrical V-shaped grooves) and herringbone grooves

(asymmetric chevrons) on the upper wall. First, it was used for passive parallel focusing and

guiding beads and cells83 and then, for a platform of circulating tumor cells (CTC) isolation

(“HB-Chip”, Fig. 8B).84 The HB-Chip design applies passive mixing of blood cells through

the generation of microvortices in order to increase the number of interactions between target

CTCs and the antibody-coated chip surface. The most unexpected finding of this study was the


69

isolation of CTCs clusters from the blood of patients with metastatic cancer, which may

provide insight into the process of metastasis in human cancer.

Another CTC-capture platform was designed by Wang et al.85 (Fig. 8C). It consists of

a patterned silicon nanopillar (SiNP) substrate coated with cancer-cell capture agents

(epithelial cell adhesion molecule antibody, anti-EpCAM) for recognizing/capturing EpCAM-

expressing cells, and a PDMS chip with a serpentine mixer that contains chevron-shaped

grooves. The last one induces a vertical flow in the microchannel that increases cell–substrate

contact frequency resulting in enhanced CTC capture on SiNP substrate. The device was tested

for blood samples from a prostate cancer patient and it showed significantly higher CTC

numbers compared to the commercial CellSearch® assay.

Recently, Aguirre et al.86 have demonstrated for the first time an integrated device for

changing the properties of cells (by mans of complex creation) in order to separate them

according to this property (size). The proposed device contains a micromixer for the creation

of a cancer cell–microbead complex, and a flow-focusing region for separation (Fig. 8D). In

addition to the mixing, provided by Dean flow, a chaotic advection is induced by turning the

microchannel contents 180º at each turn (termed “flipping” by the authors). Depending on a

size of a cancer cell–microbead complex, its trajectory will vary. Thus, the proposed system

works as a size-based particle filtration device.

4.2.3. Fully-Integrated lab-on-the-chip devices

Mixers based on chaotic advection also found their application in microfluidic perfusion

systems for cell culture studies.87 The design of such generic systems with multiple

functionalities that allows tuned and controllable (medium compositions, flow rates, gradients

etc.) analysis of cell behaviour87 is an attractive but challenging idea. In order to address the

need for perfusion of large numbers of cells with the ability to change cell culture conditions,

Cooksey et al.87 developed a sophisticated system shown in Figure 9. It consists of closed-at-

rest microvalves with a multiplexer (for combining of fluids), an on/off chaotic micromixer

circuit, a bypass channel (to divert the flow to the waste output), a central chamber (containing

three exits) and a novel microfluidic resistor (to control fluid flow rates). To achieve mixing,

the device has a separate channel with deflectable herringbone-shaped PDMS membranes.

Under negative pressure, membranes are deflected downward and become grooves that induce

chaotic mixing (‘‘on/off chaotic mixer’’). If the chaotic mixer is bypassed, very little diffusive

mixing occurs between the two coloured fluids. If the chaotic mixer is off (the grooves are not

Ch

ap

ter

II


70

actuated), the wide gradient is formed; if the mixer is turned on, the streams are completely

mixed by the time they enter the chamber. This approach allows using the same device for

generation of gradient as well as homogenized mixtures.

Figure 9. (a) Photograph of the device loaded with alternating colored inlets when all inlets are open; (b) different

states and parts of on/off micromixer: herringbone mixer is bypassed (d,g,j)), herringbone membranes are

undeflected (e,h,k), and grooves are activated (f,i,l) at 44 µL/min (Modified from87).

Kim et al.88 designed an integrated microfluidic biochip for blood typing using both red

blood cells (RBCs) and serum (Fig. 10). The reported lab-on-a-chip device consists of 4 parts:

flow-splitting microchannels, chaotic micromixers, reaction chambers and detection

microfilters. The sample blood (RBCs) and the standard serum are injected into the biochip

through the flow splitting microchannels and are introduced into each connected chaotic

micromixer for mixing. This induces the agglutination of RBCs with the corresponding

antibodies. Then the mixture enters a reaction microchamber and stays there during the

appropriate reaction time (1–3 min) for finishing the agglutination of RBCs. The reacted (or

non-reacted) mixture of blood and serum passes through detection microfilters: the large size

agglutinated RBCs are effectively filtered, whereas non-reacted RBCs easily pass through. The

separation allows the visual detection of blood groups A, B, and AB with 3 µL of the whole

blood within 3 min. A chaotic mixer used in this work is named serpentine laminating

micromixer (SLM), that consists of eight ‘‘F’’-shaped mixing units (Section 3.3) that works

under the flow rate of 30 µL/min.


71

Figure 10. (a) Image of a blood typing microfluidic biochip and (b) schematic with flow splitting microchannels,

chaotic micromixers, reaction microchambers and detection microfilters are fully integrated (Modified from88).

Tan et al.89 fabricated a fully integrated lab-on-a-chip device for detecting a trace

amount of sarin in a small volume of whole blood (Fig. 11). The chip consists of regeneration

reactor for nerve agent regeneration from whole blood (realised as a micromixer); a channel

with rectangular pillars for both reaction and filtration of precipitated blood proteins; a filter

chamber with microbeads for removal of fluoride ions; an inhibition reactor (with the

herringbone structures) for the enzyme-based hydrolysis reaction. The last section contains

herringbone grooves to improve the transport of reagents to the immobilized surface coated

with cholinesterase chromophore (enzyme) for the induced optical detection. For a longer

shelf-life, the enzyme is protected by a coating layer. In this case, the coating layer is removed

with an extra washing step before the measurement. If nerve agent exists in the blood sample,

the enzyme is inhibited, hence hydrolysis of substrate is prevented and chromophore remained

unconverted. The colour was determined by an absorbance measurement. The optimal flow

rate was 20 µL/min, because enzymatic reaction requires longer time. When the flow rate was

increased to 48 µL/min, only 5% of the enzymes were inhibited due to either inefficient mixing

in the first two stages or insufficient residence time for the regeneration reaction. The rapid

decrease in the efficiency is related to a relatively small volume of reaction chamber (63 µL

comparing to the total volume of ~ 800 µL).

Ch

ap

ter

II


72

Figure 11. (a) Design of a lab-on-a-chip device for detection of nerve agent in blood consisting of (1) nerve agent

regeneration reactor, (2) reaction and filtration channel; (3) a filter chamber with microbeads, (4) inhibition reactor

with the herringbone structures and (5) optical detection region; and (b) the fabricated PMMA device. (Taken

from89).

4.3. Analysis of hazardous compounds (environmental analysis)

Several groups had developed sensitive analytical systems for the detection of cyanide90 and

methyl parathion pesticides91 in water, cyanide in the living cells92 and cobalt (II) ions and

hydrogen peroxide93 using micromixers based on chaotic advection.

An already-mentioned micromixer – zig-zag-shaped microfluidic mixer with alligator

teeth-shaped structures (Sec. 4.2.1)– was applied for development of an ultrasensitive trace

analysis of cyanide90 and methyl parathion pesticides91 in water. In both studies, hazardous

molecules were effectively adsorbed onto silver nanoparticles while flowing along the upper

and lower parts of the channel (Fig. 12A). Both devices were fabricated in PDMS and

investigated using confocal surface-enhanced Raman spectroscopy (SERS). The determined

detection limits for methyl parathion pesticides and cyanide were 0.1 ppm and 0.5–1.0 ppb

respectively. Later, Lee et al.98 used the same system for the sensitive trace analysis for

determining malachite green with limit of detection below the 1–2 ppb.

Kwon et al.92 developed a fluorescent chemosensor, which exhibits a selective green

fluorescence upon the addition of cyanide. In order to mix compounds, the mixing channel has

herringbone-shaped obstacles on the wall that caused chaotic advection (Fig. 12B).

Developing a sensitive analytical system with a shorter analysis time, Lok et al.93

presented a microchip containing a chaotic micromixer for luminol-peroxide

chemiluminescence (CL) detection of cobalt (II) ions and hydrogen peroxide (as an artificial

model system). The micromixer design was adopted from Tan et al.’s work89 and consists of


73

forward and reversed staggered herringbone grooves (Figure 1CA), which allowed better

catalytic interaction between the reactants. With a total volume of 145 µL, the corresponding

residence time of the system ranges from 1.45 hours to 8.72 minutes at the flow rates from 1.67

to 16.7 µL/min respectively. The chip was also tested at 100 µL/min and 163 µL/min, which

reduced the analysis times to 1.5 min and 1 min respectively. It was shown that a higher flow

rate increases the CL intensity. However, it also leads to the excessive usage of reagents and

leakages due to the increased pressure. As a compromise, the authors decided to use an optimal

flow rate of 16.8 µL/min (analytical time of 8.65 min).93

Figure 12. (A) An alligator teeth-shaped microfluidic mixer for mixing silver colloids and cyanide solution

(Taken from90); (B) (a) Optical images of a fluorescent chemosensor (b) that exhibits a selective green

fluorescence upon the addition of cyanide at a flow rate of 10 µL/min (Taken from92). (C) Schematic layout of

the microchip with of forward and reversed staggered herringbone grooves (SHG) (Taken from 93).

4.4. Analytical techniques

Recently, several interesting applications in the area of analytical chemistry were reported.

Abonnenc et al.94 explored the application of a new electrospray emitter microchip for on-line

chemical modification of peptides and direct ESI-MS analysis (Fig. 13A). This microchip

comprises a passive mixer with integrated photoablated slanted grooves to carry out on-chip

chemical derivatization. In order to apply the voltage to the solution and consequently induce

the ESI process, a carbon microelectrode is also integrated in the microchip. To illustrate the

feasibility of the on-line tagging reaction directly on the electrospray microchip and to assess

Ch

ap

ter

II


74

the performance of such micromixer, on-chip chemical tagging of three cysteines with 1,4-

benzoquinone was performed. As an ultimate application, the electrospray micromixer was

implemented in an LC-MS workflow. It was shown that online derivatization of albumin

tryptic peptides after a reversed-phase separation enhances the protein identification. Finally,

on-line RPHPLC-tagging-ESI-MS of tryptic cysteinyl peptides of bovine serum albumin was

achieved with the electrospray micromixer chip. In the present study, total flow rates from 4

to 6 µL/min were applied, at which good spray stability was demonstrated.

Figure 13. (A) Electrospray microchip with a mixing unit (Taken from94); (B) A gradient elution system for

pressure-driven liquid chromatography on a chip and a photograph of the fabricated microchip with the cross-

Tesla mixer (Modified from95).

In practice, biological and chemical samples vary over a wide range of their properties

and some fluids that have to be mixed differ in viscosity. The viscosity differences can limit

many processes in the microchannels (e.g. continuous chemical synthesis in microreactors,

polymer formulation), and, thus, the utilization of a micromixer is essential. Several

examples15,21,44,59 of chaotic micromixers that were tested for mixing viscous fluids were

presented in the Section 3.2 and 3.3. The utilization of these micromixers can be also useful in

applications, where the differences in viscosity influence mixing dramatically (e.g. the high-

viscous fluids can cause a dramatic increase of pressure in HPLC).

Another way to use micromixers for improving analytical techniques is in the formation

of gradients.13,95 One promising example - an overbridge-shaped micromixer13 - was already

presented in Section 3.3. Another mixer with integrated Tesla structure for formation of a

gradient, a channel of pillar array columns for separation and a sample injection channel

(Figure 13B) was proposed by Song et al.95. It was used for gradient elution of four aliphatic

amines in pressure-driven reversed-phase liquid chromatography separation, that were

successfully separated within 110 s at a total flow rate of 1.0 μL/min.. Due to the gradient

elution conditions, the separation was shorter with sharper peaks than under isocratic elution


75

conditions. These micromixers have great potential in the improvement of conventional

techniques combined with microfluidic technology, especially in the analysis of complex

biological samples.

In our group, we developed a pressure-resistant microfluidic mixer with herringbone

grooves for changing the mobile phase composition between two columns in on-line

comprehensive two-dimensional liquid chromatography (LC×LC).99 The device is meant to

mix in-line rapidly in the wide range of flow rates (0.1-1 mL/min). This chip was micromilled

in rigid cyclic-olefin copolymer (COC) and can withstand pressures of 200 bar due to a

specially designed low-dead-volume interface. Using standardized HPLC connectors, the chip

is directly connected to LC×LC system. It was successfully implemented for improved

separation and identification of various oligomeric series in polyamide samples and showed

similar performance to the commercially available mixing units.

Ch

ap

ter

II


76

5. Discussion

In the Section 3 the reader was already introduced to the Table 1, which summarizes different

types of micromixers based on chaotic advection with their dimensions and material/fabrication

methods. This table can serve as a good guideline for helping scientists to make a right choice

of appropriate mixer. However, in this chapter we are also providing the reader with the detailed

process of choosing the micromixer taking into account many aspects of the requirements set

by the particular application (Figure 14).

There are a lot of different micromixers proposed in the literature that were developed

for a specific application, however there is no a straightforward approach for choosing an

appropriate micromixer for a particular application. Besides, the same micromixer can be

successfully used for a variety of applications (Table 1, last column). However, it is possible to

draw some general criteria that should be taken into account while choosing an appropriate

mixer design among a big number of existing micromixers. The first thing that should be done

is to define clear requirements for the particular application. It is important to know what kind

of samples (chemical, biological etc.), ranges of flow rates (or residence times) and detection

methods will be used (Fig. 14). The application will also dictate choice of substrate material or

fabrication of the mixer. For example, some hydrophobic molecules can be absorbed in PDMS.

In combination with three-dimensional convoluted micromixers, the produced insoluble

materials in chemical synthesis102 or biological material in bio-applications can easily clog the

PDMS chip.

Another important criterion is the required flow rate range. It also dictates the material

of the mixer (and, thus, a fabrication method) and the type of connection with the macro world.

Most microfluidic systems are operated under low flow rates (0.2-75 µL/min). Therefore, those

micromixers can be fabricated in PDMS/PDMS or PDMS/glass by soft lithography (Table 1).

However, in cases where higher flow rates (0.9 - 4 mL/min) and/or the connection to

conventional equipment (e.g. LC, MS) are necessary, microfluidic systems have to be fabricated

from more rigid materials (e.g. PMMA, silicon, stainless steel) and appropriate pressure-

resistant connectors are needed, to resist higher pressures. The choice of the material influences

the type of fabrication method that should have a sufficient resolution for fabrication of a

particular design.


77

The detection method that is planned to be used, also determines the material (e.g.

devices should be transparent in case of optical detection), the required flow rate and the type

of connection with other equipment (if applicable). For example, the signal intensity in

chemiluminescent detection relies on the speed of mixing that determines the sensitivity of the

method. The higher flow rate in this case provides not only the faster analysis, but what is more

important a control over degree of dispersion of reactive species to localize the reaction region.

This increases the intensity of the chemiluminescent signal and, therefore, the method

sensitivity. This requires the utilization of micromixers that can mix at relatively high flow rates

(>100 µL/min).

When the flow rate range is set and the material is chosen, it is possible to decide which

type of micromixer (simple geometry, obstacles/wall modifications or 3D convoluted channels)

will be the most efficient under the defined flow conditions. As mentioned in the beginning of

this chapter, flow conditions (described by Re), under which the mixer operates, dictate the type

of phenomena that govern mixing (dominance of diffusion or advection). This means that each

type of mixer has its own range of flow rates under which it shows the best performance. There

are similar patterns in mixing performances in all mixers: below some critical value of Re, the

mixing efficiency decreases with the increase of flow rate (dominance of diffusion); and above

this value the mixing efficiency increases with the increase of flow rate (dominance of chaotic

advection).

Besides those patterns some micromixers have inherent features that are especially

beneficial for particular applications. For example, micromixers with patterned grooves can be

used in applications, where reactants have to be trapped on the immobilized surfaces (e.g., on

electrodes82). In this case grooves on both top and bottom channel walls facilitate the binding

reactions85,86 not only because of the ability to mix reactants, but also due to the creation of the

vertical flow (in the z-direction) that provides better transport of reactants to the immobilized

surfaces. This approach can be used for analysis of bioparticles with lower diffusivity (e.g.

proteins, DNA),75 which have to be transported faster to the probes than just by diffusion. Other

examples are micromixers with simple geometries (e.g. spiral, zig-zag, serpentine). They can

work under a wide range of Reynolds numbers (1˂<Re<800) due to the appearance of vortices,

which resulted in mixing by chaotic advection at high Re and molecular diffusion at low Re.

This type of mixers contains long channel paths and, thus, provides long residence time for

efficient mixing of reactants to form the final product. These features make them suitable to be

used as microreactors or for DNA hybridization.

Ch

ap

ter

II


78

Figure 14. The approach for choosing the chaotic micromixer for a particular application.

Although mixers with obstructions in the channel are used in a smaller range of Re

(0.01<Re< 80), they can also be used for the same purposes as mixers with simple geometries

due to the increase of the reactants residence time in the obstruction region.22 Besides, the ability

to modify the obstructions (e.g. micropillars) inside the mixer with reacting agents opens a wide

range of possibilities for separation and extraction processes, for example when the analytes of

interest are trapped on micropillars (e.g. DNA capturing or detection of affinity binding in

immunoassays).

However, the same design features that are beneficial to some applications can impose

negative effects at the same time. For instance, very deep grooves can create a large dead

volume or force reactant to stay too long in the groove, which distorts peak shape in the

chromatographic separation. Furthermore, the perpendicular bends of 3D convoluted designs

can create dead volume zones (the fluid at the corner of these bends is stagnant, the only mass

transport is diffusion).


79

After choosing the type of chaotic micromixer, it is important to draw an appropriate

design taking into account all nuances related to the channel parameter and the design of the

structures (obstructions, grooves etc). The geometry of the structures placed in the mixing

channel is described by their depth, width, the angle position, symmetrical or asymmetrical

arrangment etc. A larger depth and width of obstructions leads to a higher chance for inducing

chaotic advection by creating larger vortices in the flow streamlines, which increases larger

contact area between two flows. For instance, studies50,85 showed that the mixing in

micromixers with staggered herringbone grooves improves with deeper grooves. This can be

explained by the increased fluid entrainment in the grooves leading to an increase of the vertical

motions of the fluid at the side edges of the groove.47

The number of obstructions placed in the channel also influences the mixing

performance.9 For instance, Wang et al.42 investigating cylindrical pillars in a mixing channel

found that the mixing improves with the increase of the number of obstacles in the channel

(within same area). Sahu et al.22 also observed that mixing efficiency increases quadratically

with the number of obstructions. The authors explained this observation by fluids staying in the

obstruction region for a longer time (with larger number of obstruction), which provides more

time to finish mixing by diffusion.

Hence, in order to make a choice of an appropriate micromixer that performs the best

way possible for a particular application, many criteria should be considered. This choice

should be done in consideration with all nuances related to the type of the sample, the flow rate

range and the unique features of the application. However, the amount and the variety of already

available designs ease this task. There is no need to develop completely new micromixers, it is

enough to choose a right design and alter it towards the needs of a specific application.

C

ha

pte

r II


80

6. Conclusions

In this chapter, we showed novel designs of passive micromixers based on chaotic advection

that were proposed within the last decade. We classified them according to their geometry:

simple geometries, mixers with obstacles in the channel or wall modification, and 3D

convoluted channels. The chaotic micromixers with simple planar geometries provide better

mixing at Re>10, because the Dean flow that governs mixing is intensified with the increase of

the flow rate. The introduction of obstructions to the channel and pattering the channel wall

with grooves provide the efficient mixing at lower Re (0.01<Re<80 and 1<Re<100,

respectively). The mixing in a wider range of Reynolds numbers (0.1˂Re<260) is achieved,

when the mixers with 3D convoluted channels are used due to the combination of two mixing

mechanisms: SAR and chaotic advection.

We showed successful applications of passive chaotic mixers in chemical industry,

biology and analytical chemistry. Most of applications described in this chapter use the range

of flow rates from 0.01 µL/min to 4 mL/min. Micromixers based on chaotic advection have

found their application as microreactors, in analysis of DNA, in sorting of particles and cells,

to improve diverse cell culture platforms and in the full integrated lab-on-the-chip devices. In

analytical chemistry chaotic micromixers were used for analysis of hazardous compounds (e.g.

cyanide, pesticides, malachite green), for an on-line chemical modification of peptides in a LC-

MS interface, for mixing liquids with different viscosities, for gradient formation or improving

the performance of conventional analytical techniques such as LC×LC.

As it was shown in this chapter, very often, a mixer having a particular design finds

diverse applications in a number of different areas. As an example we can name a mixer with

herringbone grooves52 (we report here 13 different applications)65,67–69,71,75,82–84,89,93,99,100 or an

alligator teeth-shaped micromixer64 (with 6 applications).64,73,74,76,77,90 These micromixers can

be used in a such big variety of ways due to the fact that they work efficiently in a wide range

of flow rates.

We hope that this chapter will prove to be useful for the scientists in their endeavours

with respect to choosing and implementation of appropriate micromixers to the real-world

applications.


81

References




68 (2011).


(2005).

4. Lee, C. Y., Wang, W. T., Liu, C. C. & Fu, L. M. Passive mixers in microfluidic systems: A review. Chem.

Eng. J. 288, 146–160 (2016).

5. Li, P., Cogswell, J. & Faghri, M. Design and test of a passive planar labyrinth micromixer for rapid fluid

mixing. Sensors Actuators, B Chem. 174, 126–132 (2012).

6. Al-Halhouli, A. et al. Passive micromixers with interlocking semi-circle and omega-shaped modules:

Experiments and simulations. Micromachines 6, 953–968 (2015).

7. Van Schijndel, T. et al. Toward gradient formation in microfluidic devices by using slanted ridges.

Macromol. Mater. Eng. 296, 373–379 (2011).

8. Lin, D. et al. Three-dimensional staggered herringbone mixer fabricated by femtosecond laser direct

writing. J. Opt. 15, 025601 (2013).

9. Lee, D. & Lo, P. H. On the enhancement of mixing in tangentially crossing micro-channels. Chem. Eng.

J. 181–182, 524–529 (2012).

10. Feng, X., Ren, Y. & Jiang, H. An effective splitting-and-recombination micromixer with self-rotated

contact surface for wide Reynolds number range applications. Biomicrofluidics 7, 1–10 (2013).

11. Lin, Y., Yu, X., Wang, Z., Tu, S. T. & Wang, Z. Design and evaluation of an easily fabricated micromixer

with three-dimensional periodic perturbation. Chem. Eng. J. 171, 291–300 (2011).

12. Yoo, W.-S., Go, J. S., Park, S. S.-H. & Park, S. S.-H. A novel effective micromixer having horizontal and

vertical weaving flow motion. J. Micromechanics Microengineering 22, 5007 (2012).

13. Li, X., Chang, H., Liu, X., Ye, F. & Yuan, W. A 3-D Overbridge-Shaped Micromixer for Fast Mixing

Over a Wide Range of Reynolds Numbers. J. Microelectromechanical Syst. 24, 1391–1399 (2015).

14. Chen, Z. et al. Performance analysis of a folding flow micromixer. Microfluid. Nanofluidics 6, 763–774

(2009).

15. Xia, H. M., Wang, Z. P., Koh, Y. X. & May, K. T. A microfluidic mixer with self-excited ‘turbulent’ fluid

motion for wide viscosity ratio applications. Lab Chip 10, 1712–1716 (2010).

16. The, H. Le et al. Geometric effects on mixing performance in a novel passive micromixer with trapezoidal-

zigzag channels. J. Micromechanics Microengineering 25, 094004 (2015).

17. Hsieh, S.-S. & Huang, Y.-C. Passive mixing in micro-channels with geometric variations through µPIV

and µLIF measurements. J. Micromechanics Microengineering 18, 065017 (2008).

18. Liu, K. et al. Design and analysis of the cross-linked dual helical micromixer for rapid mixing at low

Reynolds numbers. Microfluid. Nanofluidics 19, 169–180 (2015).

19. Yang, A. S. et al. A high-performance micromixer using three-dimensional Tesla structures for bio-

applications. Chem. Eng. J. 263, 444–451 (2015).

20. Viktorov, V., Mahmud, M. R. & Visconte, C. Design and characterization of a new H-C passive

micromixer up to Reynolds number 100. Chem. Eng. Res. Des. 108, 152–163 (2016).

21. Sivashankar, S. et al. A ‘twisted’ microfluidic mixer suitable for a wide range of flow rate applications.

Biomicrofluidics 10, 034120 (2016).

22. Sahu, P. K., Golia, A. & Sen, A. K. Investigations into mixing of fluids in microchannels with lateral

obstructions. Microsyst. Technol. 19, 493–501 (2013).

23. Egawa, T., Durand, J. L., Hayden, E. Y., Rousseau, D. L. & Yeh, S. R. Design and evaluation of a passive

alcove-based microfluidic mixer. Anal. Chem. 81, 1622–1627 (2009).

24. Wang, L., Liu, D., Wang, X. & Han, X. Mixing enhancement of novel passive microfluidic mixers with

cylindrical grooves. Chem. Eng. Sci. 81, 157–163 (2012).

25. Chen, H.-H., Sun, B., Tran, K. K., Shen, H. & Gao, D. A microfluidic manipulator for enrichment and

alignment of moving cells and particles. J. Biomech. Eng. 131, 074505 (2009).

26. Bhagat, A. A. S. & Papautsky, I. Enhancing particle dispersion in a passive planar micromixer using

rectangular obstacles. J. Micromechanics Microengineering 18, 085005 (2008).

27. Tsai, R. T. & Wu, C. Y. An efficient micromixer based on multidirectional vortices due to baffles and

channel curvature. Biomicrofluidics 5, 1–13 (2011).

28. Conlisk, K. & Connor, G. M. O. Analysis of passive microfluidic mixers incorporating 2D and 3D baffle

geometries fabricated using an excimer laser. Microfluid. Nanofluidics 12, 941–951 (2012).

29. Malvern Instruments Worldwide, Understanding Taylor Dispersion Analysis (2015).

Ch

ap

ter

II


82

30. Pidugu, S. B. & Bayraktar, T. FLOW PHYSICS IN MICROCHANNELS. in Proceedings of IMECE2005

2005 ASME International Mechanical Engineering Congress and Exposition 1–6 (2005).

31. Aref, H. Stirring by chaotic advection. J. Fluid Mech. 143, 1–21 (1984).

32. Tai, C. H. et al. Micromixer utilizing electrokinetic instability-induced shedding effect. Electrophoresis

27, 4982–4990 (2006).

33. Yan, D., Yang, C., Miao, J., Lam, Y. & Huang, X. Enhancement of electrokinetically driven microfluidic

T-mixer using frequency modulated electric field and channel geometry effects. Electrophoresis 30, 3144–

3152 (2009).

34. Yang, J. T., Fang, W. F. & Tung, K. Y. Fluids mixing in devices with connected-groove channels. Chem.

Eng. Sci. 63, 1871–1881 (2008).

35. Sudarsan, A. P. & Ugaz, V. M. Fluid mixing in planar spiral microchannels. Lab Chip 6, 74–82 (2006).

36. Mengeaud, V., Josserand, J. & Girault, H. H. Mixing processes in a zigzag microchannel: Finite element

simulations and optical study. Anal. Chem. 74, 4279–4286 (2002).

37. Nguyen, N. T. & Wereley, S. Fundamentals and applications of microfluidics. Artech House 111 (2006).

doi:http://doi.wiley.com/10.1002/1521-

3773%252820010316%252940%253A6%253C9823%253A%253AAID-

ANIE9823%253E3.3.CO%253B2-C

38. Bothe, D., Stemich, C. & Warnecke, H. J. Fluid mixing in a T-shaped micro-mixer. Chem. Eng. Sci. 61,

2950–2958 (2006).

39. Hoffmann, M., Schlüter, M. & Räbiger, N. Experimental investigation of liquid-liquid mixing in T-shaped

micro-mixers using μ-LIF and μ-PIV. Chem. Eng. Sci. 61, 2968–2976 (2006).

40. Ait Mouheb, N., Malsch, D., Montillet, A., Solliec, C. & Henkel, T. Numerical and experimental

investigations of mixing in T-shaped and cross-shaped micromixers. Chem. Eng. Sci. 68, 278–289 (2012).

41. Park, J. M., Seo, K. D. & Kwon, T. H. A chaotic micromixer using obstruction-pairs. J. Micromechanics

Microengineering 20, 015023 (2010).

42. Wang, H., Iovenitti, P., Harvey, E. & Masood, S. Optimizing layout of obstacles for enhanced mixing in

microchannels. Smart Mater. Struct. 11, 662 (2002).

43. Kang, T. G., Singh, M. K., Kwon, T. H. & Anderson, P. D. Chaotic mixing using periodic and aperiodic

sequences of mixing protocols in a micromixer. Microfluid. Nanofluidics 4, 589–599 (2008).

44. Chen, L. et al. Evaluation of passive mixing behaviors in a pillar obstruction poly(dimethylsiloxane)

microfluidic mixer using fluorescence microscopy. Microfluid. Nanofluidics 7, 267–273 (2009).

45. Hossain, S., Ansari, M. A., Husain, A. & Kim, K. Y. Analysis and optimization of a micromixer with a

modified Tesla structure. Chem. Eng. J. 158, 305–314 (2010).

46. Hessel, V., Löwe, H. & Schönfeld, F. Micromixers—a review on passive and active mixing principles.

Chem. Eng. Sci. 60, 2479–2501 (2005).

47. Lynn, N. S. & Dandy, D. S. Geometrical optimization of helical flow in grooved micromixers. Lab Chip

7, 580–587 (2007).

48. Tóth, E., Holczer, E., Iván, K. & Fürjes, P. Optimized Simulation and Validation of Particle Advection in

Asymmetric Staggered Herringbone Type Micromixers. Micromachines 6, 136–150 (2015).

49. Chen, F. et al. Process for the fabrication of complex three-dimensional microcoils in fused silica. Opt.

Lett. 38, 2911–4 (2013).

50. Yang, J.-T., Huang, K.-J. & Lin, Y.-C. Geometric effects on fluid mixing in passive grooved micromixers.

Lab Chip 5, 1140–1147 (2005).

51. Yun, S., Lim, G., Kang, K. H. & Suh, Y. K. Geometric effects on lateral transport induced by slanted

grooves in a microchannel at a low Reynolds number. Chem. Eng. Sci. 104, 82–92 (2013).

52. Stroock, A. D. et al. Chaotic mixer for microchannels. Science 295, 647–651 (2002).

53. Howell Jr., P. B. et al. A microfluidic mixer with grooves placed on the top and bottom of the channel.

Lab Chip 5, 524–530 (2005).

54. Hardt, S., Drese, K. S., Hessel, V. & Schönfeld, F. Passive micromixers for applications in the microreactor

and ??TAS fields. Microfluid. Nanofluidics 1, 108–118 (2005).

55. Floyd-Smith, T. M., Golden, J. P., Howell, P. B. & Ligler, F. S. Characterization of passive microfluidic

mixers fabricated using soft lithography. Microfluid. Nanofluidics 2, 180–183 (2006).

56. Kim, D. S., Lee, S. H., Kwon, T. H. & Ahn, C. H. A serpentine laminating micromixer combining

splitting/recombination and advection. Lab Chip 5, 739–747 (2005).

57. Park, J. M., Kim, D. S., Kang, T. G. & Kwon, T. H. Improved serpentine laminating micromixer with

enhanced local advection. Microfluid. Nanofluidics 4, 513–523 (2008).

58. Xia, H. M., Shu, C., Wan, S. Y. M. & Chew, Y. T. Influence of the Reynolds number on chaotic mixing

in a spatially periodic micromixer and its characterization using dynamical system techniques. J.

Micromechanics Microengineering 16, 53–61 (2006).


83

59. Fang, W. F. & Yang, J. T. A novel microreactor with 3D rotating flow to boost fluid reaction and mixing

of viscous fluids. Sensors Actuators, B Chem. 140, 629–642 (2009).

60. Le The, H. et al. An effective passive micromixer with shifted trapezoidal blades using wide Reynolds

number range. Chem. Eng. Res. Des. 93, 1–11 (2015). 61. Hong, C.-C., Choi, J.-W. & Ahn, C. H. A novel in-plane passive microfluidic mixer with modified Tesla

structures. Lab Chip 4, 109–113 (2004).

62. Iida, K. et al. Living anionic polymerization using a microfluidic reactor. Lab Chip 9, 339–345 (2009).

63. Xu, C. et al. Solution and surface composition gradients via microfluidic confinement: Fabrication of a

statistical-copolymer-brush composition gradient. Adv. Mater. 18, 1427–1430 (2006).

64. Kim, D. J., Oh, H. J., Park, T. H., Choo, J. B. & Lee, S. H. An easily integrative and efficient micromixer

and its application to the spectroscopic detection of glucose-catalyst reactions. Analyst 130, 293–298

(2005).

65. Wang, J. et al. Integrated microfluidics for parallel screening of an in situ click chemistry library. Angew.

Chemie - Int. Ed. 45, 5276–5281 (2006).

66. Valencia, P. M. et al. Single-step assembly of homogenous lipid-polymeric and lipid-quantum dot

nanoparticles enabled by microfluidic rapid mixing. ACS Nano 4, 1671–1679 (2010).

67. Leung, A. K. K. et al. Lipid nanoparticles containing siRNA synthesized by microfluidic mixing exhibit

an electron-dense nanostructured core. J. Phys. Chem. C 116, 18440–18450 (2012).

68. Belliveau, N. M. et al. Microfluidic Synthesis of Highly Potent Limit-size Lipid Nanoparticles for In Vivo

Delivery of siRNA. Mol. Ther. Nucleic Acids 1, e37 (2012).

69. Chen, D. et al. Rapid discovery of potent siRNA-containing lipid nanoparticles enabled by controlled

microfluidic formulation. J. Am. Chem. Soc. 134, 6948–6951 (2012).

70. Pedro, S. G. et al. A ceramic microreactor for the synthesis of water soluble CdS and CdS/ZnS nanocrystals

with on-line optical characterization. Nanoscale 4, 1328 (2012).

71. Xu, Z. R. & Fang, Z. L. Composite poly(dimethylsiloxane)/glass microfluidic system with an immobilized

enzymatic particle-bed reactor and sequential sample injection for chemiluminescence determinations.

Anal. Chim. Acta 507, 129–135 (2004).

72. Moon, B. U., De Vries, M. G., Cordeiro, C. A., Westerink, B. H. C. & Verpoorte, E. Microdialysis-coupled

enzymatic microreactor for in vivo glucose monitoring in rats. Anal. Chem. 85, 10949–10955 (2013).

73. Yea, K., Lee, S., Choo, J., Oh, C.-H. & Lee, S. Fast and sensitive analysis of DNA hybridization in a

PDMS micro-fluidic channel using fluorescence resonance energy transfer. Chem. Commun. (Camb).

1509–1511 (2006). doi:10.1039/b516253j

74. Chen, L. et al. DNA hybridization detection in a microfluidic channel using two fluorescently labelled

nucleic acid probes. Biosens. Bioelectron. 23, 1878–1882 (2008).

75. Liu, J., Williams, B. A., Gwirtz, R. M., Wold, B. J. & Quake, S. Enhanced signals and fast nucleic acid

hybridization by microfluidic chaotic mixing. Angew. Chemie - Int. Ed. 45, 3618–3623 (2006).

76. Kim, S. et al. Rapid DNA hybridization analysis using a PDMS microfluidic sensor and a molecular

beacon. Anal. Sci. 23, 401–405 (2007).

77. Park, T. et al. Highly sensitive signal detection of duplex dye-labelled DNA oligonucleotides in a PDMS

microfluidic chip: confocal surface-enhanced Raman spectroscopic study. Lab Chip 5, 437–442 (2005).

78. Yun, K.-S. & Yoon, E. Microfluidic Components and Bio-reactors for Miniaturized Bio-chip Applications.

iotechnology Bioprocess Eng. 2004 9, 86–92 (2004).

79. Lee, N. Y., Yamada, M. & Seki, M. Development of a passive micromixer based on repeated fluid twisting

and flattening, and its application to DNA purification. Anal. Bioanal. Chem. 383, 776–782 (2005).

80. Fang, W. F., Hsu, M. H., Chen, Y. T. & Yang, J. T. Characterization of microfluidic mixing and reaction

in microchannels via analysis of cross-sectional patterns. Biomicrofluidics 5, 1–12 (2011).

81. Jung, J. et al. Fast and sensitive DNA analysis using changes in the FRET signals of molecular beacons in

a PDMS microfluidic channel. Anal. Bioanal. Chem. 387, 2609–2615 (2007).

82. Lee, H. Y. & Voldman, J. Optimizing micromixer design for enhancing dielectrophoretic

microconcentrator performance. Anal. Chem. 79, 1833–1839 (2007).

83. Hsu, C.-H., Di Carlo, D., Chen, C., Irimia, D. & Toner, M. Microvortex for focusing, guiding and sorting

of particles. Lab Chip 8, 2128–2134 (2008).

84. Stott, S. L. et al. Isolation of circulating tumor cells using a microvortex-generating herringbone-chip.

Proc. Natl. Acad. Sci. 107, 18392–18397 (2010).

85. Wang, S. et al. Highly Efficient Capture of Circulating Tumor Cells by Using Nanostructured Silicon

Substrates with Integrated Chaotic Micromixers. Angew. Chemie Int. Ed. 50, 3084–3088 (2011).

86. Aguirre, G. R., Efremov, V., Kitsara, M. & Ducrée, J. Integrated micromixer for incubation and separation

of cancer cells on a centrifugal platform using inertial and dean forces. Microfluid. Nanofluidics 18, 513–

526 (2015).

Ch

ap

ter

II


84

87. Cooksey, G. A., Sip, C. G. & Folch, A. A multi-purpose microfluidic perfusion system with combinatorial

choice of inputs, mixtures, gradient patterns, and flow rates. Lab Chip 9, 417–26 (2009).

88. Kim, D. S., Lee, S. H., Ahn, C. H., Lee, J. Y. & Kwon, T. H. Disposable integrated microfluidic biochip

for blood typing by plastic microinjection moulding. Lab Chip 6, 794–802 (2006).

89. Tan, H. Y., Loke, W. K., Tan, Y. T. & Nguyen, N.-T. A lab-on-a-chip for detection of nerve agent sarin

in blood. Lab Chip 8, 885–891 (2008).

90. Yea, K. et al. Ultra-sensitive trace analysis of cyanide water pollutant in a PDMS microfluidic channel

using surface-enhanced Raman spectroscopy. Analyst 130, 1009–1011 (2005).

91. Lee, D. et al. Quantitative analysis of methyl parathion pesticides in a polydimethylsiloxane microfluidic

channel using confocal surface-enhanced Raman spectroscopy. Appl. Spectrosc. 60, 373–377 (2006).

92. Kwon, S. K. et al. Sensing cyanide ion via fluorescent change and its application to the microfluidic

system. Tetrahedron Lett. 49, 4102–4105 (2008).

93. Lok, K. S., Kwok, Y. C. & Nguyen, N.-T. Passive micromixer for luminol-peroxide chemiluminescence

detection. Analyst 136, 2586–91 (2011).

94. Abonnenc, M., Dayon, L., Perruche, B., Lion, N. & Girault, H. H. Electrospray micromixer chip for on-

line derivatization and kinetic studies. Anal. Chem. 80, 3372–3378 (2008).

95. Song, Y. et al. Integration of pillar array columns into a gradient elution system for pressure-driven liquid

chromatography. Anal. Chem. 84, 4739–4745 (2012).

96. Krishna, K. S., Li, Y., Li, S. & Kumar, C. S. S. R. Lab-on-a-chip synthesis of inorganic nanomaterials and

quantum dots for biomedical applications. Adv. Drug Deliv. Rev. 65, 1470–1495 (2013).

97. Rungta, R. L. et al. Lipid Nanoparticle Delivery of siRNA to Silence Neuronal Gene Expression in the

Brain. Mol. Ther. Nucleic Acids 2, e136 (2013).

98. Lee, S. et al. Fast and sensitive trace analysis of malachite green using a surface-enhanced Raman

microfluidic sensor. Anal. Chim. Acta 590, 139–144 (2007).

99. Ianovska, M. A. et al. Microfluidic micromixer as a tool to overcome solvent incompatibilities in two-

dimensional liquid chromatography.

100. Haan, P. de et al. "Digestion-on-a-Chip: A Continuous-Flow Modular Microsystem for Enzymatic

Digestion for Gut-on-a-Chip Applications. (2018).

101. Lin, Y. Numerical characterization of simple three-dimensional chaotic micromixers. Chem. Eng. J. 277,

303–311 (2015).

102. Bally, F., Serra, C. A., Hessel, V. & Hadziioannou, G. Micromixer-assisted polymerization processes.

Chem. Eng. Sci. 66, 1449–1462 (2011).

Chapter III

Development of small-volume,

microfluidic chaotic mixers for future

application in two-dimensional liquid

chromatography

Margaryta A. Ianovska1,2, Patty P.M.F.A. Mulder1, Elisabeth Verpoorte1




RSC Adv. 2017, 7, 9090-9099

Abstract

We report a microfluidic chaotic micromixer with staggered herringbone grooves having a

geometry optimized for fast mobile-phase modification at the interface of a two-dimensional

liquid chromatography system. The volume of the 300 µm-mixers is 1.6 microliters and they

provide mixing within 26 sec at flow rate of 4 μL/min and 0.09 sec at flow rate of 1000 μL/min.

Complete mixing is achieved within a distance of 3 cm along the 5 cm-long microchannel over

the whole range of flow rates. The mixers can be used to mix aqueous phosphate-buffered saline

solutions with methanol or acetonitrile at different ratios (1:2, 1:5 and 1:10). We also describe

in detail a fabrication protocol for these mixers using a two-step soft photolithographic

procedure. Mixers are made by replication in poly(dimethylsiloxane).

Keywords: microfluidics, micromixers, chaotic advection, herringbone grooves, SHG

Development of small-volume, microfluidic chaotic mixers for future application in two-dimensional liquid chromatography

87

Introduction

The increasing demand for analysis of more complex samples is stimulating the development

of high-resolution multidimensional separation techniques, such as two-dimensional (2D)

liquid chromatography (LC).1,2 Coupling different separation mechanisms in 2D LC has two

important consequences. First, as the separation mechanism in LC is determined by the nature

of stationary and mobile phases, coupling two columns (two dimensions) with different

stationary phases necessarily means that each dimension requires a different mobile phase. This

leads to a major issue in 2D LC, namely how to deal with solvent incompatibility between

dimensions. This often means that a solvent in the first dimension (1D) becomes a strong eluent

in the second dimension (2D), rapidly eluting analytes. This results in so-called breakthrough

on the second column, and poor separation of analytes as a result. Additionally, viscosity

differences and immiscibility of solvents can cause flow instability (viscous fingering effect) in

situations where mobile phases of mixed composition are required (e.g. gradient elution). This

can lead to distortion of the peak shape in the second dimension.3

The second consequence of coupling two columns is the requirement of a specially

designed interface to maintain the resolution of the separation in the first dimension for the

second dimension separation. It should provide for the efficient fast transfer of 1D effluent to

the 2D and allow modification of the solvent composition between dimensions. The interface

usually consists of a 10-port valve with either two loops for cutting 1D effluent into small

fractions4 or trap-columns for pre-concentration of analytes before re-injection onto the second

column,5 or both 6.

A dilution of the 1D effluent with 2D mobile phase improves the sample focusing in the

2D which is crucial for an overall good performance of 2D LC. For the purpose of solvent

modification between dimensions, an additional pump and a mixer unit are required. As such a

dilution can lead to peak broadening, the mixer should have a small internal volume (low µL-

range) to obtain the desired dilution ratios in minimal volumes. Additionally, the small volume

of the mixer should enable fast modification (20-30 sec) and maintain small sampled portions

of 1D effluent. The most used mixing unit in the area of LC nowadays is the T-piece, in which

the two streams are simply collided with each other, with optimal mixing obtained at higher

flow rates. Another commercially available mixer for LC applications is the so-called static

Ch

ap

ter

III


88

mixer (S-mixer, e.g. HyperShear™ HPLC).7 S-mixers are usually composed of two periodically

repeated elements in the axial direction. Each element consists of two pairs of four crossed bars

perpendicular to the orientation of the fluid stream.8 Thus, the fluid interface experiences

stretching and folding eight times while moving through each element. The mixing efficiency

of the S-mixer improves with higher flow rates and bigger volumes,7 making it inherently

unsuitable for 2D LC purposes. In order to obtain mixing in small volumes and over a wide

range of flow rates, we propose to use chip-based microfluidic technologies,9 which focus on

the development of tools for manipulation of small volumes of fluids. Perhaps the best example

of attempts to implement microfluidic technologies in an LC system is the commercially

available Jet Weaver mixer.10 This device employs a network of multi-layer microfluidic

channels (120 μm x 120 μm), and uses the split-and-recombine principle to ensure an efficient

solvent gradient formation. It is incorporated into the HPLC pumping system (1290 Infinity

Binary pump) and is available in volumes of 35 µL, 100 µL and 380 µL. Our mixer differs

substantially from this device, as it has a much smaller internal volume and is based on chaotic

mixing, which ensures fast mixing in small volumes over a wide range of flow rates.

Mixing at the micrometer scale is a challenge because of the existence of well-defined

laminar flow under typical flow conditions in microchannels. A number of approaches to

overcome this limitation have been proposed, including passive and active micromixers that

can rapidly mix small amounts of fluids.11,12,13 Passive micromixers are generally preferred

since they are easier to fabricate and do not require the application of an external force to

achieve mixing, which makes them more robust and stable. The approach chosen for this work

was first described by Stroock et al.14 and is based on passive chaotic mixing. Mixing is

achieved through the incorporation of microgrooves into a microchannel wall. Grooves can be

positioned in arrays at an oblique angle to the wall (slanted grooves, SG), or take the shape of

asymmetric chevrons or herringbones in staggered arrays (herringbone grooves, HG). These

grooves work as obstacles placed in the path of the flow and alter the laminar flow profile. This

leads to a dramatic increase of the contact area between the two streams, and facilitates mixing

by diffusion. Herringbone grooves generate two counter-rotating vortices (perpendicular to the

direction of the flow) whereas slanted grooves create a helical or corkscrew pattern flow.14

Chaotic mixers with embedded microgrooves have been found to work well for systems

with Reynolds numbers from 1 to 100.14 Several studies report the utilization of mixers to

improve a surface electrochemical reaction,15,16 perform on-line chemical modification of


89

peptides,17 and provide mixing for direct and sandwich immunoassays.18 There are other

alternative applications in the area of surface interactions, such as binding of DNA on magnetic

beads;19 focusing, guiding, sorting particles;20 and the binding of proteins21 and circulating

tumor cells to functionalized surfaces.22,23 Most of these applications utilize the same

dimensions of the mixer reported in the original study,14 not altering them to better satisfy the

demands of the current application or optimizing them based on numerical computational

studies available in the literature. This often leads to the implementation of non-optimal

micromixer designs and suboptimal performance.

The aim of this work was to develop a chaotic mixer for fast mixing performance in a

given small volume for future application in 2D LC for solvent modification between columns.

For this we used an approach taken from the literature to design optimized grooved microfluidic

mixers with internal volumes on the order of just 1 or 2 microliters. We also characterized the

mixer in order to ensure its applicability to the 2D LC system. We demonstrated the possibility

of using small-volume micromixers for flow rates compatible with 2D LC (300-1000 μL/min).

Also, devices were tested for mixing solutions with different compositions and viscosities, such

as phosphate buffered saline/acetonitrile and phosphate buffered saline/methanol mixtures,

which are the most common solvents used in liquid chromatography. In addition, the fabrication

process of mixers is described in detail. We believe that our approach represents one further

step in the implementation of microfluidic technologies for mixing in conventional LC.

Material and Methods

All chemicals were analytical reagent-grade. Fluorescein was purchased from Sigma-Aldrich

(NL) and used to prepare separate 5.0 μM fluorescein solutions in 10.0 mM phosphate-buffered

saline with pH 7.4 (PBS; Gibco, UK). Acetonitrile (HPLC-S) and methanol were both obtained

from Biosolve, The Netherlands. The pH was measured using pH-indicator strips (Neutralit,

MERCK). All solutions were prepared with 18 M-ohm ultrapure water (Arium 611, Sartorius

Stedim Biotech, Germany). Both acetonitrile and methanol were degassed for 15 min prior to

experiments. There was no deformation or swelling observed for PDMS when acetonitrile or

methanol were used.

Ch

ap

ter

III


90

Mixer parameters and optimization

The mixer has a Y-shaped channel with two inlets and one outlet (Figure 1A). The

mixing channels are 50 mm long (from the Y-junction) and 300 or 400 µm wide. A ruler

is located along the channel to show the distance from the Y-junction. The total volume

of the mixing channel is about 1.6 μL and 2.2 μL for widths of 300 and 400 µm,

respectively.

The geometry of the grooves is determined by their depth (d), width (a) and

groove spacing (b) (Fig. 1B). These parameters are the same for the HG and SG tested.

Additional parameters for the HG are the asymmetry index, p, between long and short

groove arms (p is the fraction of channel width occupied by the long arm of a HG i.e. p

= wl / w) and groove intersection angle (θ). The groove depth-to-channel depth ratio (d/h)

(hereafter known as “groove-depth ratio”) for both slanted and herringbone grooves and

p were found to have the greatest influence on mixing efficiency.23

All geometric ratios – groove-depth ratio (d/h), groove spacing-to-channel width

ratio (b/w) and channel-aspect ratio (h/w) - were found to be interdependent, and there

exists an optimal groove width-to-channel width ratio (a/w) that maximizes mixing

efficiency.25 Table 1 compares optimal channel and groove parameter values taken from

the literature that maximize mixing efficiency25 with measured values of these

parameters for fabricated devices (actual parameters).

Table 1. Optimal channel and groove parameter values taken from the literature that maximize mixing efficiency25

compared with measured values of these parameters for fabricated devices (actual parameters).

Channel parameters Optimal (based on 25)

Channel 1 Channel 2

w - channel width (chosen), µm 300/400 300 400 h/w - channel aspect ratio 0.2/0.15 0.2 0.15

h - channel height, µm 60 60 60 d/h - groove depth to channel height

ratio ≥ 1.6 0.8 0.8

d - groove depth, µm 96 50 50 p - asymmetry index 0.58-0.67 0.62 0.62

θ - groove intersection angle, ° 90 90 90 a - groove width, µm 120/160 105±5 120±2

b - groove spacing, µm 45/60 50±2 65±2 n – number of grooves per half cycle 5-6 6 6


91

One of the most important parameters for mixer design is the groove depth ratio (d/h).

Previous studies24,25 showed that mixing performance of both slanted and herringbone grooves

improves with an increase in the value of d/h, achieved using deeper grooves with respect to

channel height. This can be explained by the increased fluid entrainment in the grooves leading

to an increase of the vertical motions of the fluid at the side edges of the groove.25 The influence

of d/h on mixing was investigated experimentally; channel heights were varied from 60 to 90

µm while groove depths were varied from 50 to 20 µm deep, respectively, to achieved d/h of

0.83 down to 0.22. Results will be discussed in the section 3.1. Note that the optimal d/h is 1.6

for the given h/w, according to Lynn and Dandy. This would lead to a groove depth of 96 µm,

which could pose problems from a fabrication perspective as well as introduce excessive dead

volume, adversively affecting chromatographic performance.

Another important parameter is the groove asymmetry (p). The effect of p on the

mixing performance was investigated by Li and Chen using the Lattice-Boltzmann

method for computational simulation and optimization of chaotic micromixers based on

particle mesoscopic kinetic equations.26 The long groove arm is believed to transport

fluid to the other side of the channel. The stirring effect generated in this way is increased

through the interchange of the positions of short and long groove arms every half cycle

(Fig. 1C). Such alteration of the flow motion causes a change in the position of

asymmetric vortices that appear in each half cycle.27 The optimal value of p was found

to be 0.6.26 The same result was shown by Lynn and Dandy,25 and Stroock.14

Several groups have studied the effect of the number of grooves per half cycle (n)

on the mixing performance. Li and Chen found that the mixing depends on n as long as

n≥4.26 The optimal number of grooves per half cycle was found to be 5-6 grooves.26

Another study showed that more mixing cycles lead to better mixing efficiency than

more grooves per cycle.28 Also, previous experiments reported by Stroock14,30 showed

that grooves with an oblique angle of 45° (SG) and an intersection angle of 90° (HG)

can generate maximum transverse flows.

Lynn and Dandy showed for SG that wider grooves (larger a) with smaller groove

spacing (smaller b) increase of the magnitude of secondary flow by up to 50% compared

to the case where 𝑎 = 𝑏.25 However, increasing the width of the groove will result in

more pronounced helical motion only to some extent. According to Du et al.30, the

Ch

ap

ter

III


92

mixing length (the distance along the channel at which two solutions are well mixed)

decreases sharply as a/w is increased from 0.2 to 0.25., However, the mixing

performance is hardly improved when the a/w is further increased to 0.4. Decreasing the

groove spacing also allows an increase in the number of cycles within the same channel

length.

Chip Fabrication and Assembly

The microchannels were constructed by standard microfabrication and replicated in the

silicone rubber, poly(dimethylsiloxane) (PDMS) (Sylgard 184, Dow Corning, U.S.). The

PDMS channels were sealed by bonding to glass. The chip layout and design were drawn

using the software Clewin (Wieweb software, Hengelo, The Netherlands). SU-8 masters

were fabricated in a similar way to that used by Stroock,14 through two steps of standard

photolithography. To the best of our knowledge, no detailed description of the


93

fabrication has been presented in the literature, though a number of papers refer generally

to the fact that two-step photolithography is used. We therefore present a more detailed

description of the process used to fabricate the masters in the ESI.

Grooved microchannels were fabricated by casting a solution of PDMS

prepolymer onto the master. PDMS resin and curing agent were mixed at a weight ratio

of 10:1 and manually stirred to mix thoroughly. The stirred solution was exposed to mild

vacuum for 30 min to remove air bubbles. After curing on a hot plate for an hour at 70°C,

the PDMS layer was cut into individual devices and peeled off the master (there were

two microchannels on one wafer).

Holes were punched (1.5 mm (od)) into the PDMS device at the locations of the

inlets and outlet, and the glass slides were cleaned with acetone and 96% ethanol. In

order to bond the PDMS channel to the glass slide, PDMS chips and glass slides were

exposed to oxygen plasma for 20 sec. Afterwards, the treated surfaces were immediately

brought into contact with each other. The assembled chips were placed on a hot plate for

30 min at 70°C to enhance the formation of a chemical bond, after which chips were

taken from the hotplate to cool down to room temperature. Teflon tubing (0.8 mm (id),

1.6 mm (od), Polyfluor Plastics, The Netherlands) was inserted directly into the punched

holes in the PDMS layer (Fig. 1A).

Experimental Setup

In order to characterize the degree of mixing, fluorescence detection was used.

Fluorescein (5 μM) in phosphate buffer and phosphate buffer were introduced from

separate inlets into the Y-junction of the channel at different flow rates using syringe

pumps with 5-mL syringes (Prosense, The Netherlands).

The Péclet number (Pe) was used to calculate the flow rates required in channels

with different widths to perform experiments under the same conditions of molecular

mass transport. The Péclet number is a dimensionless parameter that characterizes

molecular mass transport in flow conduits as a ratio of advective transport (flow) rate to

diffusive transport rate:

Pe = v𝑑ℎ

D (1)

Ch

ap

ter

III


94

Here v is average linear velocity (mm/s) and D represents diffusion coefficient

(mm2/s); dh denotes hydraulic diameter for rectangular duct (e.g. equivalent diameter of

a channel, mm):

𝑑ℎ =2𝑤(ℎ+𝑑)

𝑤+ℎ+𝑑 (2),

where h is channel height (mm), d, groove depth (mm), and w, channel width.

Mixing was then tested under constant Péclet-number conditions rather than constant

flow rates to ensure the same mass transport conditions in devices with different dimensions

(Table 2).

The Reynolds number (Re) was also calculated in order to confirm that laminar flow

conditions were used for experiments. Re is a dimensionless number that gives a measure of

the ratio of inertial forces to viscous forces for given flow conditions:

Re = v𝑑ℎρ

μ (3),

where dh denotes relevant length (see Equation 2), v is average linear velocity (m/s), ρ

equals the density of the fluid (kg/m3) and μ represents the dynamic viscosity of the fluid

(kg/(m*s)). All experiments were performed under laminar flow conditions (Re ≪

2000).

Table 2. Tested flow rates based on Péclet-number calculation for channels with different

widths; d+h = 110 µm; dh = 0,161 mm (w = 300 µm), dh = 0,173 mm (w = 400 µm), ρ = 103

kg/m3, µ = 10-3 kg/(m*s), D = 2.6×10-10 m2/s (for fluorescein).32

Channel width, µm

Pe, 103 300 400

Re Total flow rate, µL/min

1.0 3.7 4.6 0.3

10.0 37 46 3.0

30.0 111.0 138.0 9.0

50 185.0 230.0 15.0

100 370.0 460.0 30.1

150 556.0 691.6 45.2

200 740.0 920.5 60.1

300 1112.5 1383.2 90.4


95

The chip was placed under the fluorescent microscope (model “DMIL”, Leica

Microsystems, The Netherlands), equipped with a 4×objective, an external light source

for fluorescence (EL6000, Leica Microsystems, The Netherlands), and a CCD camera.

For visualization of fluorescence, a fluorescein filter set (488 nm excitation, 518 nm

emission) was used. Images were captured at different positions along the channel with

a CCD camera connected to a computer using a 4x objective magnification with a field

of view of 1.8 mm, a 1-sec exposure time, a gamma setting of 1.75, and a gain of 3.5.

To investigate the mixing mechanism and monitor the mixing behavior over the

cross-sections of the mixing channel, we utilized a confocal microscope (LEICA TCS

SP8, Leica Microsystems B.V.). Devices were mounted on the moving microscope stage

and syringes from syringe pumps were connected to the inlets of the devices. More

information about these experiments may be found in the ESI, Section 3.

All experiments were performed in triplicate using different chips from different

masters which were fabricated using the same procedure.

Data analysis

The degree of mixing was quantified by determining the standard deviation (SD) in

fluorescence intensity across the width of the channel at different locations along its

length. The SD was calculated using the following equation (4)32:

SD = √1

𝑁∑ (𝑥𝑖

𝑁𝑖=1 −)2 (4)

Here, xi is the gray-scale intensity value of pixel i, and is the mean intensity

value of pixels across the entire channel. In order to be able to compare different parts

of the channel, normalized fluorescent intensity was used:

𝑆𝐷𝑛𝑜𝑟𝑚 = 𝑆𝐷

∑ 𝑥𝑖𝑁𝑖=1

(5)

For this, SD (equation (4)) was normalized by the total intensity value of pixels

across the channel (xi). In order to compare different chips, the value of SDnorm for the

position 0 mm was set as 0.5, and the SD values for the other positions were recalculated

respectively. A normalized SD of 0 represents completely mixed solutions (when the

intensity is uniform across the channel) where a value of 0.5 indicates unmixed solutions.

Ch

ap

ter

III


96

To calculate SD, images were analyzed using LispixLx85P free software (Allegro

Common LISP v. 8.0, (c) 2004 Franz Inc.) by determining SD of the intensity distribu-

tion across the width of the channel. It should be mentioned that a SD value of 0.01,

which corresponds to 98% mixing, can be considered as corresponding to a completely

mixed situation, as introduction of premixed solutions in the channel yields the value of

SD 0.01. Thus, SD cannot reach a value of 0. This relates to the uniformity of pixel

intensity values on the image itself captured by the CCD camera. We define mixing

efficiency as the ability to accomplish mixing with a minimum time and length. Mixing

within 20-25 mm of the channel is efficient. We consider 98% (SD 0.01) as

corresponding to complete mixing.

Results and Discussion

Optimization of mixing channel design

For application in the 2D LC interfaces, it is important that the passive chaotic

micromixers under consideration possess small internal volumes (on the order of just a

few μL). At the same time, these components should not contribute significantly to

overall pressure drops in the system at the flow rates typically used in 1D (100’s of μL

per minute). Hence, we chose for devices having internal volumes of 1.6 μL and 2.2 μL

(1.5 to 2 times larger than those reported by Stroock et al. 14) with total channel length

of 50 mm in order to maintain pressure drops below 1 bar.14 However, the dimensions

used by Stroock et al.14 cannot simply be multiplied by a constant to achieve mixers with

bigger volumes exhibiting the same performance (mixing efficiency). Particularly

important is the selection of groove depth, width and spacing in relation to altered

channel widths and depths. Optimized channel and groove parameters used in this study

for SG and HG mixers (Table 1) were thus selected or calculated based on a previously

described numerical study by Lynn and Dandy.25

In order to increase the inner volume of the mixer with respect to the original

report by Stroock et al.14, microchannels having widths, w, of 300 and 400 μm and a

height, h, of 60 μm were used for this study. After choosing the values of w and h to

establish channel aspect ratios, h/w, of 0.20 (w = 300 μm) or 0.15 (w = 400 μm), other


97

groove parameters (groove depth, d; groove width, a; and groove spacing, b) were

selected or calculated based on h/w (Table 1).25 θ, n and p were kept constant in this

study; d/h, h/w, a/w and b/w were varied.

To test the influence of d/h on mixing, microchannels with HG having different

h and d were fabricated. Three different HG mixers were realized, with d/h = 0.22 (d =

20 μm, h = 90 μm), d/h = 0.37 (d = 30 μm, h = 80 μm), and d/h = 0.83 (d = 50 μm, h =

60 μm). The results obtained are shown in Figure 2, where a definite increase in mixing

efficiency is observed as d/h is increased. SD decreased the further down the channel

mixing efficiency was determined.

For d/h= 0.83 in Fig. 2, complete mixing has been essentially achieved at a

Figure 2. Influence of the groove depth-to-channel height ratio, d/h, on mixing performance in an HG

mixer (n = 3 chips). The total flow rate is 20 µL/min; 300-μm-wide channel; fluorescein (5 μM) in

PBS was mixed with PBS in a 1:1 flow rate ratio; total channel length is 50 mm. The variations in

standard deviation are due in part to the fact that these experiments were carried out over a period of

several months, during which time the lab environment varied somewhat and final adjustments to the

fabrication protocol were being made.

Ch

ap

ter

III


98

channel length of 20 mm from the Y-junction. In contrast, mixing has only been partially

achieved at 20 mm for d/h. = 0.22 and 0.37. This is consistent with observations made

in other studies.14,25,26 A probable explanation is related to the two counter-rotating

vortices vortices generated in HG mixers. The size of the larger vortex, formed above

the longer arms of the herringbone grooves, grows as d/h increases28,34. (Supplementary

Figure 2 of the Supplementary Information shows confocal microscopic images of the

cross-sectional flow profile recorded along the length of a HG mixer with w = 300 μm.)

Thus, deeper grooves provide an enhancement in mixing. However, Du et al.31 showed

that increasing the d/h-value is only effective for enhancing mixing within a limited

range of d/h. Optimum values of d/h may be found in a range of 0.28 to 0.7 if h is

decreased or for values between 0.25 and 0.4 if d is increased. A further increase in d

in this latter case does not lead to faster mixing. This can be explained by considering

the location of the transverse fluid transport caused by grooves. In the microchannel,

mixing occurs above the grooves where the vortices are to be found, and chaotic mixing

proceeds rapidly as a result. Mixing also occurs within the grooves; however, mixing in

this region is much less rapid, as it is dictated by laminar flows and slow diffusion. When

d is increased, a large quantity of fluid (more than 60%) enters the grooves, and the slow

diffusional mixing inside the groove becomes significant with respect to the overall

mixing inside the channel.31 A deeper groove could result in a bigger dead volume, in

which molecules could be retained for inordinately long periods of time in real

applications, making mixing inefficient.27

The optimal d/h value for the chosen h/w ratios was 1.6 or larger, according to

Lynn and Dandy.25 However, for h = 60 µm, this implies grooves that are at least 96 µm

deep. This would introduce a large dead volume to the mixer which could adversely

influence chromatographic results through increased band broadening in future

applications. For this reason, d/h=0.8 was chosen (d = 50 µm h=60 µm). This value still

provides enhanced mixing, but does not contribute a large dead volume as discussed

below.


99

Mixing performance in the different types of microchannels with different groove arrays

In order to determine which mixer exhibits the most suitable performance for the

application at hand, three types of microchannels were investigated: 1) channels with

slanted grooves (SG) and 2) herringbone grooves (HG), and 3) channels with no grooves

(NG). In addition, channels having w = 300 μm or 400 μm were studied. The efficiency

with which PBS and PBS-fluorescein solutions are mixed in these types of channels is

compared in Figure 3. The standard deviation of fluorescence intensity across the

channel is plotted versus distance along the channel for a wide range of flow rates.

Experiments were carried out in channels with w = 300 µm (Fig. 3A) and w = 400 µm

(Fig. 3B) in the range of Pe values from 103 to 3×105 (Table 2). Data is shown only for

low (Pe = 103, solid line) and high (Pe = 105, dashed line) flow rates in Fig. 3.

Incomplete mixing is observed in the NG channel at low flow rate (Pe = 103) at

a channel length of 50 mm for both channels widths studied. The mixing in these

channels relies entirely on diffusion of molecules between side-by-side flows, which is

a slow process. (Molecules would require more than 10 seconds to diffuse from the

interface between solution streams at the middle of the channel to the sides. This is a

long time when compared to the residence time of molecules in the channel at even low

flow rates, see Table 2 and Fig. 3D.) In fact, mixing at the lower flow rate in the 300-µm-

wide ungrooved (NG) channel (Fig. 3A), though incomplete at the end of the 50-mm

channel, is more complete than in the 400-µm-wide ungrooved (NG) channel (Fig. 3B).

This is in keeping with the longer radial distance that solutes need to travel by diffusion

for mixing to occur in the wider channel.

Increasing the flow rate by a factor of 100 (Pe = 105) would lead to decreased

residence times of solutes in the microchannel and thus to negligible mixing or no mixing

at all. Introducing a mixer with slanted or herringbone grooves results in more efficient

mixing, as presented in Stroock’s original study.14

For the grooved channels, we observe a similar decrease in mixing efficiency, especially

at low flow rate, for the 400-µm-wide channel compared to the 300-µm-wide channel, which

means that, perhaps, similar diffusional effects as discussed above could still be playing a role.

We tried to compensate for the increase in w/h by maintaining the a/w ratio, widening the

grooves from 105 µm (as in the 300-µm-wide channel in Fig. 3A) to 120 µm for the 400-µm

Ch

ap

ter

III


100

wide channel in Fig. 3B. Based on experimental data which is not shown, we assume that even

wider and deeper grooves would improve the mixing efficiency further in the 400-µm-wide

channel. Considering the better performance of the 300-μm-wide channel (Fig. 3A and B) and

its smaller volume (1.6 µL compared to 2.2 µL of the 400-μm-wide channel), we selected the

300-μm-wide channel for further studies.

If we look at channels which are 300 µm wide, it can be seen that in the HG mixer (red

dots on Fig. 3A), 98% of mixing is completed by a distance of 10 mm and 15 mm for Pe of 103

and 105, respectively. These findings are in good agreement with values obtained by Stroock14

for the same Péclet number but for a channel with smaller cross-sectional area. For the SG

mixer(blue dots, Fig. 3A), the required distances for complete mixing are 20 and 35 mm for Pe

of 103 and 105, respectively. From these data, we can conclude that herringbone grooves

provide better mixing performance than slanted grooves for all the flow rates tested. This is

consistent with observations from other studies.25,35 The HG mixer is 30 and 55 times more

efficient than the NG channel, and 2.0 and 3.8 times more efficient than the SG mixer, at 3.7

µL/min and 370 µL/min, the HG mixer lies in the difference between the processes involved in

mixing. In general, grooves enhance mixing because of the additional motion of fluids

(stretching and folding), which leads to an increased contact area between the solutions to be

mixed, thereby decreasing diffusion lengths. Stretching and folding of solution volumes inside

the mixers proceeds exponentially as a function of the distance travelled along the channel.14,36

In the SG mixer, mixing happens through generation of a single helical flow along the axis of

flow (a more detailed mechanism for SG is reported elsewhere).29 This requires a longer

distance to complete mixing. In contrast, mixing in the HG mixer occurs as a result of the

formation of two oppositely rotating vortices across the channel. This makes the HG mixer

more efficient.


101

In order to investigate the influence of Péclet number on mixing efficiency, we

tested the 300-µm-wide HG mixer over a wide range of the Péclet numbers, namely 103

to 3×105 which corresponds to a flow rate range of 3.7 to 1114 µL/min (Table 2). As

Figure 3. Comparison of microfluidic mixers having no grooves (NG), slanted grooves (SG) and

herringbone grooves (HG) as a function of distance from the Y-junction for channel widths of 300 µm

(A) and 400 µm (B) at different flow rates: Pe=103 (solid line) and Pe=105 (dashed line). The flow rate

in each case is the total flow rate in the mixing channel, with a 1:1 flow rate ratio of PBS (fluorescein)

- PBS; n = 3 chips. For grooved channels: d = 50 µm and h = 60 µm; for ungrooved channels, h = 110

µm; a=105 µm for the 300-µm-wide channel, a=120 µm for the 400-µm-wide channel. (C) Standard

deviation versus position along the channel for the 300-µm-wide HG mixer for Pe in the range of 103

to 3×105, which corresponds to the flow-rate range of 3.7 to 1114 µL/min (Table 2). Microphotographs

are presented to show mixing at (a) 0 mm; (b) 5 mm; (c) 15 mm; (d) 35 mm. (D) Residence times at

different flow rates (Pe = 103-105) for HG mixer; flow rate ratio of PBS ( (fluorescein)-PBS was 1:1;

total channel length is 50 mm for all channels. The observed range of standard deviation is 0.05-0.15 at

5 mm.

Ch

ap

ter

III


102

seen in Figure 3C, the HG mixer performed well over the whole chosen range of Pe.

Initially, the intensity decreased sharply (decrease in SD, Figure 3C (a-b)) within the

first 10 mm of channel length and then quickly leveled off to approach a constant value,

which corresponded to complete mixing. As expected, the efficiency of mixing

decreased with the increase of the flow rate but only over the first 15 mm of the channel.

The observed SD varies from about 0.25 to about 0.05 at 5 mm for flow rates from 1114

to 3.7 μL/min, whereas it varies from 0.01 to 0.008 at 40 mm for the same flow-rate

range in this 300-μm-wide HG channel). Complete mixing was achieved by 15 mm,

independent of the flow rate. This can be explained by the compensation of shorter

residence (and diffusion) times by increased agitation of the flows, which leads to more

chaotic flow patterns. Such effects make the HG mixer efficient over a wide range of

flow rates. The observed variation in fluorescence intensity was the same as in Stroock’s

study14, who concluded that the form of the flow remains qualitatively the same for

0<Re<100 (Pe ＞ 106).

As Pe increases by a factor of 300 (from 103 to 3×105), the distance required for

98% mixing (SD=0.01) increases by a factor of 1.5 (from 10 mm to 15 mm). Complete

mixing requires a relatively longer distance (additional 5-10 mm) at higher flow rate

(Pe≥103). Shorter residence times, leading to shorter diffusion times, account for this

observation, as already alluded to above (Figure 3D). Residence time (Rt, sec) was

calculated as the centre-line length of the channel (cm) divided by the average flow

velocity (cm/sec). The calculated values of Rt underline the speed of mixing, particularly

at higher flow rates. As seen from Figure 3D, mixing can be achieved in the 300-μm-

wide channel within a distance of 20 mm in 10.7 sec, 1.1 sec and 0.11 sec at total flow

rate 3.7 μL/min (Pe 103), 37 μL/min (Pe 104) and 370 μL/min (Pe 105), respectively.

With herringbone grooves, then, the increased flow rate leads to almost the same mixing

distance but in a much shorter period of time, which is beneficial for fast solvent

modification in 2D LC. Also, it is clear that potential dead volumes in the grooves

themselves is not an issue.


103

Mixing of different solvents

Micromixers designed in this study will be implemented for the modification of mobile

phase eluting from the first dimension before entering the second dimension in 2D LC.

This application requires mixing of different solvents to tune the ability of a mobile

phase to elute analytes from a stationary phase. In order to investigate the efficiency of

the HG mixer, two of the most commonly used solvents in liquid chromatography,

acetonitrile (ACN) and methanol (MeOH), were chosen for further experiments. First,

these solvents were mixed with phosphate-buffered saline (PBS) at equal (1:1) flow rate

ratios. Figure 4 shows images obtained with a fluorescent microscope which have been

stitched together to show the first 20 mm of the 300-μm-wide, 60-μm-deep channel with

herringbone grooves (d = 50 µm). The solution of fluorescein in PBS (green color) from

the left inlet (upper inlet in images) and solution of PBS (Fig. 4A), ACN (Fig. 4B) or

MeOH (Fig. 4C) from the right inlet (lower inlet in images) were introduced at equal

flow rates. As the mixing proceeds along the channel, the observed fluorescence

gradually expands to cover the whole channel width, and an almost equal distribution of

fluorescence can be observed at the 20-mm mark in the channel, indicating almost

complete mixing. Here, as in all previous experiments, the absolute intensity of the

fluorescence decreases, which is related to the dilution effect. The same chaotic flow

patterns, observed with confocal microscopy in the channel cross-section (ESI, Fig. S2),

appear as striations when viewed from above in Figure 4.

Ch

ap

ter

III


104

Figure 4. Fluorescence images taken from above of HG micromixers in which a solution of fluorescein in PBS is

mixed with (A) PBS solution, (B) ACN, (C) MeOH at a 1:1 flow rate ratio; images have been stitched together to

show the first 20 mm of the 300-μm-wide channel, Pe = 2.7 ×105, Re = 81 (same channel used in Fig. 3).

Figure 5. Efficiency of mixing at different flow ratios (1:1, 1:2, 1:5 and 1:10) in HG micromixer of PBS (5

µM fluorescein) and (A) ACN; or (B) MeOH; total flow rate, 1000 µL/min; channel width, 300 µm; n=3 chips;

Pe = 2.7 ×105, Re = 81 (same channel as used in Fig.3). The observed range of SD is 0.02-0.04 at 5 mm. This

decreases to a range of 0.001-0.003 along the channel at 45 mm. The viscosities of pure ACN and MeOH at

25°C are 0.334 cP and 0.543 cP, respectively.


105

In order to enable the solvent modification between dimensions in 2D LC, a

relevant solvent (e.g. water) should be mixed with the 1D eluent. In most cases, the 1D

effluent will contain a high percentage of organic solvent which should be diluted five

or ten times. Thus, ACN or MeOH were introduced together with PBS solution at

different flow rate ratios: 1:1, 1:2, 1:5 and 1:10 (Fig. 5). A solution of 5 μM fluorescein

in PBS was used to visualize the mixing. All experiments were designed to maintain a

total flow rate of 1 mL/min (Pe = 2.7×105). In general, for both ACN/PBS and

MeOH/PBS systems no significant difference in mixing efficiency was observed and the

mixing was complete at a distance of 45 mm. The fact that mixing efficiency was

unaffected by buffer-solvent flow rate ratios is noteworthy. Both ACN/PBS and

MeOH/PBS mixtures can exhibit viscosities which are different from the pure solvents

(the viscosities of pure ACN and MeOH at 25°C are 0.334 cP and 0.543 cP,

respectively), as a function of mixing ratios. In fact, a 45:55 MeOH/H2O mixture has a

viscosity of 1.83 cP, which is almost twice that of water alone. For the ACN/PBS system

the maximum viscosity is 1.15 cP (20°C) at 10-30% of ACN.37 However, such changes

in viscosity had no visible effect on the mixing of MeOH and water solution in the HG

micromixer.

It should be mentioned that the appearance of bubbles was observed when mixing

methanol with PBS solution at low flow rate at channel distances greater than 15 mm,

despite the fact that we degassed the methanol prior to experiments. This can be related

to the fact that mixing of methanol and water is an exothermic process38 resulting in a

decrease of gas solubility which leads to the production of air bubbles.

C

ha

pte

r II

I


106

Supplementary information

Chip fabrication and assembly

The microchannels were constructed by standard microfabrication and replicated in the silicone

rubber, poly(dimethylsiloxane)(PDMS)(Sylgard 184, Dow Corning, U.S.). The chip layout and

design were drawn using the software Clewin (Wieweb software, Hengelo, The Netherlands).

Masters were fabricated using two steps of standard photolithography in the negative

photoresists SU-8 50 and SU-8 10 (MicroChem). These resists were optimized for different

layer thicknesses (SU-8 50 was used to pattern microchannels, and SU-8 10 to pattern the

grooves) (Supplementary Figure 1A). The conditions were chosen based on the recommended

parameters described by MicroChem, the manufacturer of SU-8 photoresists.39,40

A 4-inch borofloat wafer (700 µm thickness, Borofloat 33, Handelsagentur Helmut

Teller, Jena, Germany) was employed as a substrate. The wafer was cleaned sequentially with

acetone, isopropyl alcohol, and ultrapure water, dried with N2 gas and baked for 5 min at 150oC

to remove residual water. A first layer of negative photoresist (PR), SU-8 50 (MicroChem,

Newton, MA), was coated on the wafer using 500 rpm for 13 sec followed by different speeds,

depending on the required layer thickness (Table S-1). The wafers were then soft-baked (from

20oC to 65oC in 45 min, 8 min at 65oC, from 65oC to 95oC in 30 min, 25 min at 95oC and cooled

down to room temperature on the hotplate). The coated wafer was illuminated with ultraviolet

(UV) light (365 nm, 10 mW/cm2) from a collimated light source to pattern the microchannels,

using a photomask printed on a transparency (resolution 3,810 dpi; Pro-Art BV, Groningen,

The Netherlands). The illumination was followed by a post-bake step (from 20oC to 65oC in 45

min, 1 min at 65oC, from 65oC to 95oC in 30 min, 8 min at 95oC and cooled down to room

temperature).

Wafers were then exposed to oxygen plasma (Harrick plasma, USA) to ensure adhesion

of the second photoresist layer for 20 sec. The second layer of SU-8 10 photoresist was spin-

coated with SU-8 10 (MicroChem, Newton, MA) at different speeds, depending on the required

layer thickness (Supplementary Table 1), soft-baked at the conditions mentioned above and

illuminated by UV light. The second photomask, which contains the pattern for grooves only,

was aligned manually with the microchannels in the first layer under the microscope using

alignment marks. After the post-bake step, wafers were immersed in SU-8 developer for 15


107

min, rinsed with isopropanol and dried using N2 gas. They were then placed under vacuum in

the presence of hexamethyldisilazane (HMDS) (Sigma-Aldrich, Germany) for at least 30 min

to make the wafer surface more hydrophobic and thus facilitate the peeling of the cured

PDMSfrom the masters. The heights of the channels and grooves on the masters were

determined using a profilometer (Veeco Instrument BV, located in the Zernike NanoLab

Groningen, The Netherlands).

Supplementary figure 1. (A) A schematic diagram of the two-step photolithography fabrication

process. (B) Schematic drawing (top view) of the three types of channels investigated in this study:

channel without grooves (1); channel with slanted grooves (2); channel with herringbone grooves (3).

Supplementary table 1. Spin coating conditions for different photoresist layer thickness.

1st layer (SU-8 50) 2nd layer (SU-8 10)

Speed, rpm Layer thickness, μm Speed, rpm Layer thickness, μm

1600 (40 sec) 60 750 (30 sec) 50

1300 (40 sec) 80 1300 (35 sec) 30

1300 (35 sec) 90 1500 (45 sec) 20

Ch

ap

ter

III


108

Investigation the mixing mechanism

Evaluation of the degree of mixing by conventional microscopy in the manner that we have

done may yield misleading data with respect to mixing in microchannels. Image analysis for

our reported experiments was done using top-view images of the channel. However, during

mixing, fluids change their orientations inside the channel, and fluid zones containing different

dye concentrations can be situated perpendicular or parallel to the camera of the microscope.

This means that an observed equal distribution of fluorescence for the channel top-view may

not necessarily correspond to complete mixing. In order to confirm the reliability of of our

image analysis, and at the same time investigate the mixing mechanism at the cross-sections of

the mixing channel, we utilized confocal microscopy.

Supplementary Figure 2 shows the cross sections of the HG mixer at positions from 0

to 9.5 mm along the channel obtained by confocal microscopy. The first image (position 0 mm)

shows two fluids, PBS and PBS-fluorescein, flowing in the channel before entering the region

with grooves. The second image represents the situation where the solution from the left inlet

hits the sharp edge of the groove and a portion of the solution moves along the long groove

arm. At this point the flow from the left inlet splits into two parts, as was reported previously

by Yang et al.24 One part of the flow moves further to the right side, enters the groove and hits

the next rigid curve backwards, which results in formation of the counterclockwise-rotating

vortex. Another part of the flow rolls out of the groove on the left side and returns to the

mainstream. This process results in clockwise rotation. On the images obtained with confocal

microscopy (Suppl. Fig. 2) the generation of two counter-rotating asymmetrical vortices can be

clearly observed. These vortices change in asymmetry from one half-cycle to another. For this

micromixer with a groove width, a, of 80 µm and channel width, w, of 300 µm, grooves start

at 0.78 mm and one full cycle occupies 2.0 mm (each half-cycle is 0.98 mm plus interval

between half-cycles). Therefore, there are 21 full cycles within the 50 mm channel length.

Moving from a position of 2.3 mm (Supplementary Fig. S-2.B), which indicates the first half-

cycle, to a position of 3.2 mm, which indicates the second half-cycle, we clearly see two

counter-rotating vortices with changed asymmetry in between half-cycles. These results are in

good agreement with both the numerical simulations 24,29,35 and experimental results by Stroock

et al. 14,30


109

Yang et al.24 identified two dominant mechanisms of mixing in the HG mixer: (1) the

stretching and folding of the interface due to the vertical motions of flow at the side edge of the

groove and (2) the increase in contact area between the two fluids due to fluid transportation

inside the groove. Stretching and folding are chaotic processes that lead to the production of

chaotic advection.13 The flow velocity components are constant over space and time in chaotic

advection, in contrast to the turbulent flow where these are considered to be random.29

Supplementary figure 2. Images of the cross sections of the HG mixer obtained using confocal microscopy at

total flow rate (A) 20 μL/min (Pe ~ 5×103, Re 1.6) and (B) 200 μL/min (Pe ~ 50×103, Re 16.3); distance from the

Y-junction: 0 - 11.5 mm of the channel length; w = 300 μm; a = 80 μm; PBS (fluorescein)-PBS 1:1. Mixing is

facilitated by the generation of two counter-rotating asymmetrical vortices. Equal distribution of fluorescein on

the images indicates the complete mixing.

Experiments were carried out using two flow rates: 20 μL/min (Suppl. Fig. 2A) and 200

μL/min (Suppl. Fig. 2B). It is reasonable to expect that mixing should be achieved faster at the

lower flow rate. The influence of the flow rate on the behavior of mixing can be clearly

observed. Compared with the smooth vortices (due to longer residence time) at lower flow rate

(Suppl. Fig. 2A), the high flow rate introduces agitation to the flows creating more chaotic flow

patterns and increasing the contact area between them. Such effects are what make the HG

mixer efficient over the wide range of flow rates tested. As the flow rate increases by a factor

of 10 (from 20 to 200 μL/min), the distance required for complete mixing increases only by a

factor of 1.35 (from 7.0 to 9.5 mm). These results are in good accordance with our previous

Ch

ap

ter

III


110

data (Figure 3) obtained by fluorescence microscopy, where we showed that 10 mm was enough

to complete mixing at Pe < 105.


111

Conclusions

We have successfully demonstrated chaotic micromixers, which are larger than those

originally reported by Stroock et al.,14 with optimized channel and groove geometries,

designed using previously reported numerical studies. The resulting micromixers can be

used at flow rates ranging from 150 to 1000 µL/min without significant differences in

the mixing efficiency. We confirm that the HG mixer works significantly better than the

SG mixer or the NG channel. The HG mixer is 55 times more efficient than the NG

channel and 3.8 times more efficient than the channel with SG at 370 µL/min. Mixing

can be achieved within 45 ms in the 300-μm-wide channel at a flow rate of 1.1 mL/min

at a distance of less than 25 mm.

In this work, we have also demonstrated mixing of different solvents in HG

micromixers. Mixers can rapidly mix aqueous buffers with ACN and MeOH solutions

at different flow rate ratios at flow rates in the range of 5-1000 µl/min, which makes it

possible to use micromixers for applications in 2D LC. Future work will be directed

towards implementation of mixers into 2D LC systems.

Acknowledgements

This work was financially supported by The Netherlands Organization for Scientific Research

(NWO) in the framework of the Technology Area-COAST program, project no. (053.21.102)

(HYPERformance LC). C

ha

pte

r II

I


112

References





3. Mayfield, K. J. et al. Viscous fingering induced flow instability in multidimensional liquid chromatography.

J. Chromatogr. A 1080, 124–131 (2005).

4. Van Der Horst, A. & Schoenmakers, P. J. Comprehensive two-dimensional liquid chromatography of

polymers. J. Chromatogr. A 1000, 693–709 (2003).



88, 1785–1793 (2016).

6. Wang, D. et al. On-line two-dimensional countercurrent chromatography × high performance liquid

chromatography system with a novel fragmentary dilution and turbulent mixing interface for preparation

of coumarins from Cnidium monnieri. J. Chromatogr. A 1406, 215–223 (2015).

7. Brochure provided by Analytical Scientific Instruments US, http://www.hplc-asi.com/static-mixers/. (2014).

8. Singh, M. K., Anderson, P. D. & Meijer, H. E. H. Understanding and optimizing the SMX static mixer.

Macromol. Rapid Commun. 30, 362–376 (2009).

9. Whitesides, G. M. The origins and the future of microfluidics. Nature 442, 368–373 (2006).

10. Agilent 1290 Infinity LC System, Manual and Quick Reference, Agilent Technologies. (2012).


(2005).

12. Lee, C.-Y., Chang, C.-L., Wang, Y.-N. & Fu, L.-M. Microfluidic Mixing: A Review. Int. J. Mol. Sci. 12,

3263–3287 (2011).




15. Moon, B.-U. et al. An enzymatic microreactor based on chaotic micromixing for enhanced amperometric

detection in a continuous glucose monitoring application. Anal. Chem. 82, 6756–6763 (2010).

16. Yoon, S. K., Fichtl, G. W. & Kenis, P. J. A. Active control of the depletion boundary layers in microfluidic

electrochemical reactors. Lab Chip 6, 1516–1524 (2006).

17. Abonnenc, M., Dayon, L., Perruche, B., Lion, N. & Girault, H. H. Electrospray micromixer chip for on-

line derivatization and kinetic studies. Anal. Chem. 80, 3372–3378 (2008).

18. Golden, J. P., Floyd-Smith, T. M., Mott, D. R. & Ligler, F. S. Target delivery in a microfluidic

immunosensor. Biosens. Bioelectron. 22, 2763–7 (2007).

19. Lund-Olesen, T., Dufva, M. & Hansen, M. F. Capture of DNA in microfluidic channel using magnetic

beads: Increasing capture efficiency with integrated microfluidic mixer. J. Magn. Magn. Mater. 311, 396–

400 (2007).

20. Hsu, C.-H., Di Carlo, D., Chen, C., Irimia, D. & Toner, M. Microvortex for focusing, guiding and sorting

of particles. Lab Chip 8, 2128–2134 (2008).

21. Foley, J. O., Mashadi-Hossein, A., Fu, E., Finlayson, B. A. & Yager, P. Experimental and model

investigation of the time-dependent 2-dimensional distribution of binding in a herringbone microchannel.

Lab Chip 8, 557 (2008).

22. Stott, S. L. et al. Isolation of circulating tumor cells using a microvortex-generating herringbone-chip. Proc.

Natl. Acad. Sci. 107, 18392–18397 (2010).




Lab Chip 5, 1140–1147 (2005).

25. Lynn, N. S. & Dandy, D. S. Geometrical optimization of helical flow in grooved micromixers. Lab Chip 7,

580–587 (2007).

26. Aubin, J., Fletcher, D. F. & Xuereb, C. Design of micromixers using CFD modelling. Chem. Eng. Sci. 60,

2503–2516 (2005).

27. Li, C. & Chen, T. Simulation and optimization of chaotic micromixer using lattice Boltzmann method.

Sensors Actuators B Chem. 106, 871–877 (2005).


113

28. Schönfeld, F. & Hardt, S. Simulation of Helical Flows in Microchannels. AIChE J. 50, 771–778 (2004).

29. Tóth, E., Holczer, E., Iván, K. & Fürjes, P. Optimized Simulation and Validation of Particle Advection in

Asymmetric Staggered Herringbone Type Micromixers. Micromachines 6, 136–150 (2015).

30. Stroock, A. D. & Whitesides, G. M. Controlling flows in microchannels with patterned surface charge and

topography. Acc. Chem. Res. 36, 597–604 (2003).

31. Du, Y., Zhang, Z., Yim, C., Lin, M. & Cao, X. A simplified design of the staggered herringbone micromixer

for practical applications. Biomicrofluidics 4, 1–13 (2010).

32. Periasamy, N. & Verkman, A. S. Analysis of fluorophore diffusion by continuous distributions of diffusion

coefficients: application to photobleaching measurements of multicomponent and anomalous diffusion.

Biophys. J. 75, 557–567 (1998).

33. Xia, H. M., Wan, S. Y. M., Shu, C. & Chew, Y. T. Chaotic micromixers using two-layer crossing channels

to exhibit fast mixing at low Reynolds numbers. Lab Chip 5, 748–755 (2005).

34. Ansari, M. A. & Kim, K. Y. Shape optimization of a micromixer with staggered herringbone groove. Chem.

Eng. Sci. 62, 6687–6695 (2007).

35. Aubin, J., Fletcher, D. F., Bertrand, J. & Xuereb, C. Characterization of the mixing quality in micromixers.

Chem. Eng. Technol. 26, 1262–1270 (2003).

36. Kee, S. P. & Gavriilidis, A. Design and characterisation of the staggered herringbone mixer. Chem. Eng. J.

142, 109–121 (2008).

37. Snyder L. R., Kirkland J. J., D. J. W. Introduction to modern liquid chromatography. (A John Wiley &

Sons, Inc., Publication, 2010).

38. Aburjai, T., Muhammed, A. & Al-Hiari, Y. M. Temperature and Pressure Behaviours of Methanol,

Acetonitrile/Water Mixtures on Chromatographic Systems. Am. J. Anal. Chem. 02, 934–937 (2011).

39. Brochure provided by NANOTM SU-8 Negative Tone Photoresist Formulations 50-100.

40. Brochure provided by NANOTM SU-8 Negative Tone Photoresist Formulations 2-25.

Ch

ap

ter

III

Chapter IV

Fabrication of a pressure-resistant

microfluidic mixer in fused silica

using Selective Laser-Induced

Etching

Margaryta A. Ianovska1,2, Jean-Paul S.H. Mulder1, Martin Hermans3, Elisabeth

Verpoorte1




3 LightFab GmbH, Aachen, Germany

Abstract

We report a microfluidic mixer fabricated in a solid block of fused silica using Selective Laser-

Induced Etching (SLE). The micromixer contains herringbone grooves (HG) that induce mixing

based on chaotic advection, as investigated in our previous work. The chip was designed for

utilization in the interface between two columns in a multidimensional liquid chromatography

system, which implies the utilization of pressure-resistant devices. Our first chips were hybrid

devices made from poly(dimethylsiloxane) (PDMS) and glass. These devices could not

withstand pressures higher than 10 bar, due both to the elastic properties of PDMS and a lack

of a robust bond between PDMS and glass. We therefore opted for utilization of a relatively

new technique, namely Selective Laser-Induced Etching (SLE), as a route to making monolithic

devices in a single block of a rigid material. Our material of choice was a block of fused silica.

This eliminates the need to bond a structured chip to a chip acting as the lid for a microfluidic

device. Moreover, fused silica is four orders of magnitude more rigid than PDMS and can thus

withstand higher pressures. We aimed to fabricate in silica, for the first time, microfluidic

mixers with channels up to 33 mm long containing arrays of microgrooves. Fabrication proved

challenging, as removing laser-modified silica from the patterned channels by the introduction

of etchant from the channel ends meant longer exposure to etchant at the beginning and end of

the mixing channel. This resulted in overetching of the channel and grooves in the end regions

with an accompanying loss of groove resolution. In order to solve these fabrication issues and

to account for differences in etch progression in different device regions, we have made use of

an adjusted design that provided improved mixing performance. The pressure tests showed that

the fused-silica chips can withstand pressures of up to 85 bar and can be used in the interface

between two columns of a multidimensional liquid chromatography system to facilitate the fast

adjustment of mobile phase composition.

Keywords: micromixers; pressure-resistant chips; fused silica chips; Selective Laser-Induced

Etching (SLE); Femtosecond Laser Irradiation followed by Chemical Etching (FLICE).

Fabrication of a pressure-resistant microfluidic mixer in fused silica using selective laser-induced etching

117

Introduction

The growing interest in microfluidics over the last few decades has led to the development of

many different techniques and methods for fabrication of microfluidic devices in scientific

settings in both academia and industry. Miniaturization imparts such advantages as reduced

consumption of reagents, shortened analysis times, and the possibility to have good control over

flow conditions, as well as mass and heat transfer.1 This makes microfluidics an attractive field

for flow chemistry, materials sciences and also as a means to realize components to improve

state-of-the-art analytical separation techniques such as high performance liquid

chromatography (HPLC). Very often such applications require utilization of high pressures.

Therefore, development of different types of pressure-resistant microfluidic systems recently

became a new trend in the microfluidics field.1

There is a big variety of materials available for chip fabrication, such as silicon, glass

and elastomeric polymers. Polydimethylsiloxane (PDMS) has gained wide acceptance in the

academic microfluidics community due to its low cost, robustness, route to simple device

fabrication, optical transparency and non-toxicity.2 In previous work,3 we developed a 1.6-μL

microfluidic mixer that provides good mixing within seconds at flow rates compatible with

LC×LC (0.1-1 mL/min). This chip was designed to be placed in the interface between two

columns of a multidimensional liquid chromatography system and so had to withstand high

pressure pulses (up to 200 bar), which arise from valve switching and additional back pressure

from the second column. In order to provide mixing, the device contained an array of

herringbone-shaped grooves (herringbone grooves, HG) with a depth of 50 µm and a width of

110 µm. The presence of these grooves led to the generation of two counter-rotating vortices

by chaotic advection.4 This mixer was fabricated in PDMS bonded to a glass plate. However,

the elastomeric nature of PDMS and its low Young's modulus become a significant problem for

the development of a chip that could withstand high pressures. In our experience, device failure

occurred above pressures of 10 bar either in the PDMS itself (mostly) or at the interface between

the PDMS and glass. Even at low flow rates, significant channel deformation can occur, which

leads to alterations of the flow profile and subsequently to changes in device performance.2

Because of problems mentioned above, other materials are used in order to fabricate

high-pressure resistant chips, such as polymers (e.g. cyclic olefin copolymer, COC),5–7 or glass

Ch

ap

ter

IV


118

and silicon.8–11 Liu et al.5 developed a COC chip containing in situ photopolymerized

polymethacrylate monolithic stationary phases for HPLC separations of fluorescein-labelled

intact proteins. This chip was fabricated by hot embossing and can withstand 200 bar. Another

chip suitable for liquid chromatography was demonstrated by Chen et al.6 with a maximum

burst pressure (the pressure that a device can withstand before failing) of 400 bar. It was also

fabricated using COC but by direct microscale mechanical milling. Several examples of

silicon/pyrex10 and glass/glass11 microreactors that can be used at pressures up to 300 bar have

been presented. Silicon/pyrex chips have been fabricated using deep-reactive-ion etching

(DRIE) followed by anodic bonding of the twp layers,10 whereas for fabrication of glass/glass

microreactors11 wet etching and direct fusion bonding were used. Recently, an alternative

method for inexpensive rapid prototyping based on off-stoichiometry thiolenes (OSTEs) was

demonstrated by Martin et al.1 The chip, fabricated using UV-curable OSTE and bonded to

glass, could withstand 200 bar, and was used to perform multiphase flow visualization studies

in microchannels. However, the chip fabricated in this study had a square cross-sectional

channel of 200 µm without any features inside.

Multiple methods for fabrication of microfluidic devices exist, each with unique

advantages and drawbacks. Injection molding and hot embossing can be considered as fast but

expensive methods for polymer prototyping due to the high initial cost of making the molds.

On the other hand, glass/silicon micromachining processes by wet/dry etching create good-

precision structures but are technically demanding and time consuming to fabricate.2 Direct

fabrication methods such as micromilling and laser ablation, though cost-accessible and

enabling complex 3D-multilayer structures, have low resolution (around 50 µm). They also

generate surface roughness, and fabrication has limited throughput due to the inherent serial

nature of the fabrication process.2 All these methods suffer from one inherent drawback: they

create a 2D-open channel network on one substrate surface that has to be sealed (closed or

bonded) to a second chip in order to obtain a microfluidic channel. This creates a weak point

which manifests itself as bond breakage in any high pressure application, as mentioned above.

Developed more than a decade ago, Femtosecond Laser Irradiation followed by

Chemical Etching (FLICE),12 also called Selective Laser-Induced Etching (SLE),13,14 has

emerged as a novel powerful approach for direct fabrication of complex 3D structures inside a

solid transparent material such as fused silica. The SLE technique consists of two steps: 1) the

exposure of glass to scanning focused ultra-short (fs or ps) pulsed laser radiation, which locally


119

changes glass properties in the focal volume to create self-aligned nanocracks perpendicular to

the laser polarisation direction;15 2) etching of the laser-modified zone by HF or an alkaline

solution such as KOH in water.13,15 During the etching process, the nanocracks created by the

pulsed laser act as channels through which the etching agent diffuses deeper into the fused

silica. Etching takes place where the etching agent comes into contact with modified fused silica

along the diffusion path.15 Being a direct fabrication technique inside a solid piece of material,

SLE provides an appealing solution to avoid microchannel-sealing or chip-bonding steps during

device fabrication. It also doesn’t require complex cleanroom facilities and allows for the

fabrication of complex 3D structures.15 The degree of feature resolution and the aspect ratios

possible in glass and silica are higher with SLE than with wet etching. SLE thus allows the

exploitation of the unique properties of glass (transparency, rigidity, inertness etc) in devices

having smaller and better defined features than previously was possible in glass.

However, the SLE approach has some limitations regarding channel length, shape and

aspect ratio. As the channel is etched after patterning by starting from one end, it is necessary

to continuously remove the reaction products and provide fresh etching agent to diffuse along

the channel. However, as the channel length increases, the amount of fresh acid able to reach

the end of the channel reduces and the etch rate gradually decreases,16 i.e. the etching process

saturates at longer etching periods.17 This saturation leads to microchannels with lower aspect

ratios and/or channels with conical shapes (tapered channels), geometries which become more

pronounced the longer channels get. The longest dead-end channels reported were about 1.8

mm long, and had an aspect ratio (length-to-hydraulic diameter ratio4) of ~ 20.18 This effect can

be reduced by etching the microchannel simultaneously from opposite ends. Vishnubhatla et

al.19 managed to obtain a 4-mm-long microchannel with an aspect ratio of 4 by etching from

both ends of the channel in an ultrasonic bath containing a 20% solution of HF in water for 4.5

hours.

Studies have shown that the depletion of the HF acid toward the center of the etched

microchannel and the difficulty of replenishing it in this region often leads to self-termination

of the etching process.15 Moreover, HF is an isotropic etching agent removing material laterally

at a similar rate to the speed of downward etching, and the selectivity of the HF for the laser-

4An aspect ratio in this work is calculated not as the ratio of the width to the channel height but as the channel

length to hydraulic diameter. This type of ratio was used here in order to compare mixers with channels fabricated

by other authors. The higher this ratio is, the more difficult it is to fabricate a channel (because of its length).

Ch

ap

ter

IV


120

modified region with respect to unmodified material is not sufficient.16 This results in limited

channel lengths (about 1.5 - 2 mm)20 and limited length/diameter aspect ratios. In addition,

water, which is formed during the HF action on silica, dilutes the HF acid, and further impeding

the etching process.21 On the other hand, these effects were not observed when aqueous KOH

was used as the etching agent.16 Having a higher selectivity for modified fused silica due to the

formation of a Si-rich structure16 and being an anisotropic etching agent, aqueous KOH

provides slow etching with almost constant selectivity regardless of the etching period.15,16 It

was reported that the selective etch rate with KOH is even higher (14 times higher) than for

etching with ~2% HF.14 Thus, using prolonged 60-hour etching in KOH, Kiyama et al.16

fabricated 10-mm microchannels with less than 60 μm diameter (an aspect ratio of almost 200).

In this work, we aimed to fabricate a 30-mm-long microfluidic mixing channel having

an aspect ration of more than 100 in fused silica by the Selective Laser-Induced Etching method.

Besides the problems that arise when such a long channel is to be fabricated (as described

above), the need to have herringbone grooves on the channel wall in order to generate mixing

complicate the fabrication process. The fabrication of such chips has never been explored before

using the SLE technique. In the current study, we describe several generations of chips

fabricated in fused silica with different dimensions. In order to solve some fabrication issues

and account for differences in etch times in different device regions, we exploit a modified

micromixer design incorporating compensatin structures to counteract overetching in regions

where the silica is exposed to HF etchant for longer periods. We refer to this design as a

compensation design. Compensation structures incorporated into regions subjected to longer

etchant exposure were designed so that longer HF treatment was required to yield the final

features having the desired dimensions. The fabricated micromixer with this design showed

improved mixing performance compared to previous chip generations where overetching had

(partially) eroded the grooves and made channels undesirably wider. Also, pressure tests

showed that fused silica chips can withstand pressures of up to 85 bar.


121


Chip designs

All chip designs were drawn in SolidWorks© (Waltham, Massachusetts, USA) and saved as a

STEP file for further fabrication. The SolidWorks design of the first fabricated chip with

different relevant dimensions is given in Figure 1A. We refer to it as the 1st generation chip as

it forms the basis upon which other devices were subsequently designed. All dimensions of this

chip are summarized in Table 1. A schematic representation of herringbone grooves (HG) in

the channel with names of different parameters is shown in Figure 1C. The 1st generation chip

had the same herringbone groove dimensions as were used in our previous work.3 However,

the channel dimensions are different. Moreover, the inlets and an outlet were designed as

standard 10-32 female HPLC connectors. The midpoint of the chip channel cross-section is

aligned dead centre with the midpoint of the 1/16” ID peek tubing fixed in the inlets and outlet.

Figure 1. The SolidWorks© design of the (A) 1st and (B) 2nd generation microfluidic mixer with herringbone

grooves; all dimensions are in mm. (C) Schematic drawing of herringbone grooves in the channel.

Ch

ap

ter

IV


122

Table 1. The channel parameters of the 1st and 2nd generation chips.

Parameter 1st generation 2nd generation

Channel length, mm 31 33.6

Length of the inlet, mm 9.3 2.09

Hydraulic diameter, mm5 0.161 0.316

Inlet width, µm 150 215

Channel height, µm 60 150

Groove depth, µm 50 100

Channel width, µm 300 430

Groove width, µm 110 260

Ridge width, µm 50 70

Volume, µL 1.6 3.6

The first fabricated chip revealed that the chosen fabrication method is not suitable for

the fabrication of well-resolved grooves having dimensions smaller than 100 µm. Based on

these results, all dimensions of the second chip (2nd generation chip) were increased (Table 1).

We decided to increase the channel height and channel width to 150 µm and 430 µm,

respectively, in order to provide better access for etching agent to penetrate into the channel.

Based on these values, the other parameters were recalculated following the protocol for

achieving optimized geometry22 (Table 1), which is described in our previous work.3 Moreover,

the Y-junction was replaced by the T-junction (Fig. 1B) and the length of the inlet channel was

shortened significantly, for reasons which will be discussed later.

Chip fabrication

All chips were fabricated in quartz glass (fused silica) by Selective Laser-Induced Etching

(SLE) using the LightFab 3D Printer at LightFab (Aachen, Germany).13,14 The thickness of the

fused silica was 7 mm with optically polished surfaces. The FCPA laser (Satsuma, Amplitude

Systemes, Pessac, France) provided ultrashort laser pulses having a wavelength of 1030 nm,

with a pulse duration of 1000 fs, a pulse energy of about 500 nJ and a writing velocity of 200

mm/s at a repetition rate of 750 kHz. Laser radiation is focused by a 20x microscope objective

with a numerical aperture of 0.45 (LCPLAN N 20x/0.45 IR, Olympus Europa GmbH;

Hamburg, Germany) equipped with a collar for correction of spherical aberrations.

5 The formula for hydraulic diameter will be presented in Section 2.4 (Eq. 2).


123

The chip designs that were saved as STEP files were opened in the CAM software

LightFabScan. The laser parameters, the three linear axes and the three dynamic axes in the 3D

Microscanner were controlled using the same software. Vectors were generated automatically

by SliceLas (available from LightFab) in the CAD software Rhinoceros 3D (from Rhino3D)

and transferred to the CAM software LightFab Scan.

For development of the structures by wet-chemical etching, the laser-heated silica

device was immersed in an aqueous solution of 8 mol/L KOH for 10 days at 85 C with

ultrasonic excitation. The heavy-duty ultrasonic bath was equipped with a 99 h timer, automatic

heating and cooling with temperature control, and automatic water refill to compensate for

evaporation.

Characterization of channels and grooves

Photos of the fabricated grooved channels were obtained using a microscope (model “DMIL”,

Leica Microsystems, The Netherlands) equipped with a 40x magnification and a CCD camera.

Channel and HG dimensions were measured using ImageJ to analyse the photos (U. S. National

Institutes of Health, Bethesda, Maryland, USA). These values were then compared with the

dimensions of the original Solidworks design. The data so obtained was plotted in OriginPro

9.1.0 (OriginLab Corporation, Massachusetts, USA).

Mixing test

In order to visualize mixing, the commercially available solutions of two food dyes, Brilliant

Blue (BB, E133, KoepoE, Indonesia) and Tartrazine (Tz, E102, KoepoE, Indonesia), were

introduced from separate inlets into the channel junction with a 5-mL syringe (B.braun, The

Netherlands) through a 1.59-mm (od), 0.8-mm (id), polyetheretherketone (PEEK) tubing

(Kinesis Ltd, Cambridgeshire, UK) using a syringe pump. Photos were taken at different

positions along the channel with a camera (Canon EOS 700D) that was mounted on the

microscope (Leica S8 APO, Leica Microsystems, Germany).

To ensure that micromixers having different internal dimensions and volumes could be

tested and compared under the same flow conditions, the dimensionless Péclet number (Pe) was

used to calculate appropriate flow rates for mixing experiments. Pe characterizes molecular

Ch

ap

ter

IV


124

mass transport in flow conduits as a ratio of advective transport (flow) rate to diffusive transport

rate, can be calculated according to the formula:

𝑃𝑒 = 𝑣𝑑ℎ

𝐷 (1),

where v is average linear velocity (mm/s), D represents diffusion coefficient (mm2/s)

and dh denotes the hydraulic diameter for a rectangular duct (e.g. equivalent diameter of a

channel, mm). This latter parameter can be calculated according to Equation 2,

𝑑ℎ =2𝑤ℎ

𝑤+ℎ (2),

where w (mm) is channel width, h is channel height (mm).

Because the hydraulic diameter of the 2nd generation chip is 0.316 mm, which is almost

twice as large as in the 1st generation, mixing was tested under constant Péclet-number

conditions rather than constant flow rates to ensure the same mass transport conditions (Table

2).

Table 2. Tested flow rates based on Pe calculation for chips with different dimensions; (d+h)1

= 110 µm; (d+h)2 = 250 µm, where d is groove depth; dh1 = 0.161 mm (w = 300 µm), dh2 =

0.316 mm (w = 430 µm), where dh is hydraulic diameter; ρ = 103 kg/m3, µ = 10-3 kg/(m*s), DBB

= 2.8×10-9 m2 /s (for BB),23 DTz ~10-9 m2 /s (for Tz).24

Chip generation Flow rate, mL/min Pe, 103

1st 0.12 3.5

2nd and 3rd 0.2

1st 0.6 17.5

2nd and 3rd 1.0

All experiments were performed after conditioning the channel with 0.6 mol/L NaOH

(Sigma-Aldrich, Sweden, Missouri, USA) and 18 M-ohm ultrapure water (Arium 611, Sartorius

Stedim Biotech, Germany).

Pressure test

To evaluate the burst pressure, the inlets of the 2nd generation chip were connected to HPLC

pumps (Waters 515, Waters Corporation, Massachusetts, USA) and the outlet was connected

to an HPLC column (HYPERSIL SAX 5U 4.6×150 mm, Alltech) using standard HPLC fitting

(Fingertight Fitting One-Piece PEEK, Upchurch Scientific, IDEX Health & Science, CA,


125

USA), as shown in Figure 2. In the first set of experiments, the column was open and the tested

flow rate range was set to 0.2-2.0 mL/min. Afterwards, the column was closed with threaded

stopper to build up the pressure. The water was pumped at a total flow rate of 0.5 mL/min.

Figure 2. Set-up for the pressure test: A 2nd generation micromixer was connected to HPLC pumps and a sealed

column.

Ch

ap

ter

IV


126


Fabricated fused-silica chips

The fabrication of long channels with uniform cross-sections using the SLE technique has to

date been limited as was mentioned in the Introduction. However, the appealing idea to directly

fabricate buried channels in a solid block of fused silica, thereby circumventing the need to seal

channels in a bonding step, motivated us to explore this technique in more detail. In our previous

work,3 we developed and investigated a microfluidic mixer with herringbone grooves, that was

destined to be used in multidimensional chromatography. Our experiments showed that when

fabricated in PDMS/glass, such chips cannot withstand pressures more than 10 bar, due to the

low Young's modulus of PDMS (a measure of the ability of a material to withstand changes in

length (elasticity) when under lengthwise tension (pressure)). The value of the Young's modulus

lies in the range between 0.57-3.7 MPa for PDMS and 73 GPa for fused silica. Therefore, we

aimed to fabricate the micromixer in fused silica, which is three to four orders of magnitude

less elastic. The first generation chip had the same dimensions as the PDMS/glass chip reported

previously by our group.3 Because of the possible difficulty of fabricating the 50-mm-long

channel from the original design, the channel length was decreased to 30 mm. Thus, the channel

aspect ratio (length to hydraulic diameter ratio) of the 1st generation chip turned out to be 186.

Figure 3. Photos of the 1st and 2nd generation chips with herringbone grooves fabricated in fused silica: (A),(B)

whole chips and (C),(D) their channels imaged using a microscope. Images have been stitched together to show

the full channel; a 40×magnification objective lens was used for (C) and (D). The scale of the images in (C) and

(D) is the same.


127

Figures 3A and 3C show the whole 1st generation chip and its magnified channel,

respectively. A clear difference is observed between the laser-modified (white region) and

overetched channel (black region) material observed (Fig. 3C). Unfortunately, the channel

quality is not particularly good and the resolution of grooves is poor. The channel diameter

varies along the channel, with a larger cross-section observed at the entrance as compared to

the middle part of the channel (width 300 µm and depth 60 µm). As was discussed above, it is

an inherent feature of the etching process itself. Only in the middle part of the channel (Fig. 3C,

marked in red) did we observe channel parameters that correspond to the original design, with

well-resolved grooves. Such results were predictable due to the typical problems associated

with etching. The longer the channel is, the more etching time is needed to reach laser-modified

material in the middle of the channel. This means that both ends of the channel where laser-

modified material had already been removed remain in contact longer with etching solution

than section towards the middle of the channel. Even though we used KOH as an etching

solution, the etching selectivity of which remained almost constant regardless of the etching

time, some excessive etching took place. For comparison, Kiyama et al.16 fabricated 9.2-mm

microchannels with less than 60 μm diameter and with aspect ratio of 153, which is very similar

to the one we were aiming for (186). The etching time was 60 hours when 10 mol/L (35.8%)

aqueous KOH was used. This is 4 times shorter than the time that we used, which is not

surprising given that the microchannel length and hydraulic diameter were ~4 times shorter and

5 times smaller respectively than in our case.

Nevertheless, we decided to increase the cross-section of the 1st generation chip in order

to simplify the introduction of the etching solution into the mixing channel by using wider inlets

(Table 1). New dimensions for the 2nd generation chip were calculated based on the protocol

for optimized geometry.22 Figures 3B and 3D show the 2nd generation chip and a top-view of

its channel under the microscope. This chip has almost twice as large a channel diameter, with

a channel aspect ratio of 104, compared to the 1st generation chip. In general, the channel quality

is better and grooves are visible along the full channel length. It is also clearly visible that the

section with HG having good resolution is significantly longer for the 2nd generation chip (Fig.

3D, outlined by a red rectangular) than for the 1st generation (Fig. 3C).

Ch

ap

ter

IV


128

Figure 4. Micrographs of the (A-C) 1st and (D-F) 2nd generation channels: (A, D) inlet; (B, E) junction and (C, F)

middle part which correspond to the original design with clearly resolved HG. Red lines (B, E) indicate the border

of the channel according to the original design. All photos are made with the same magnification.

In addition to increased channel dimensions, the 2nd generation chip was designed with

T-junction (Fig. 3B, 4D, 4E) instead of Y-junction (Fig. 3A, 4A, 4B). Because the etching

solution enters the chip through inlets, the etching time proportionally increases with their

lengths. The T-junction provided a 4-fold shorter inlet length, providing easier access for KOH

to enter the channel. The clear improvement in etching of the middle part of the 2nd generation

chips can be also seen in Figure 4F compared to the 1st generation chip (Fig. 4C) (photos are

made with the same magnification.).

It is important to mention that besides the difficulties associated with fabrication of

channels longer than few mm using the SLE technique, the chip fabrication in our case is

complicated by the inclusion of herringbone groove arrays on the bottom of the channel, which

are essential for generation of mixing based on chaotic advection. It should be noted that the

fabrication of herringbone grooves having a rectangular cross-section is impossible with

conventional photolithographic patterning and HF etching in glass. This is due to the isotropic

nature of the HF etching process which precludes high-aspect-ratio channels. SLE provides the

possibility to fabricate channels with a high-aspect ratio that have well-defined grooves in the

glass surface. However, to the best of our knowledge, no other groups have fabricated anything

other than long straight or bent channels having no internal features. The only attempt to

fabricate a similar fused-silica herringbone mixer by femtosecond-laser direct writing

combined with wet etching using HF was proposed by Lin et al.25 Several 2D and 3D designs

were fabricated, one of which contained walls that were also patterned with slanted grooves.


129

However, in that study, the surface of the fused silica substrate was irradiated to form the

grooved channel, which was then sealed to PDMS. Lin et al.25 thus demonstrated the use of

SLE to produce arbitrary patterns on the vertical side walls, but didn’t exploit the most attractive

feature of SLE technique, namely, the elimination of the sealing/bonding process by fabricating

the full channel inside a block of fused silica.

Figure 5. Side-view of the 2nd generation chip with regions where ridges are clearly visible (middle part) and

where they disappear (at the beginning and the end of the channel). These are photographs taken of a tilted

channel/device.

Grooves in the micromixer are defined by ridges that separate grooves from each other.

If ridges are “eaten away” by etching solution during fabrication, there will no longer be any

grooves left over by the end of device fabrication to induce any mixing. Figure 5 presents the

side view of the 2nd generation chip to illustrate this situation. Ridges have disappeared at the

beginning and end of the channel, as a result of excessive etching over several days. This means

that mixing based on chaotic advection will occur only in the middle part where the grooves

still have good resolution, i.e. only within 3 mm out the total 33 mm of the channel length (see

Section 3.3).

Compensation design and the 3rd generation chip

In order to obtain channels with constant cross-section, we decided to make a design that

compensates for the excessive etching in the beginning and at the end of the channel. Mixing

channels were designed to have varying widths and depths, with narrower and shallower regions

at the ends to allow for more etching in these regions. Both channel width and depth increased

gradually toward the middle part of the channel. We measured the width in top-view

Ch

ap

ter

IV


130

micrographs of the 2nd generation chip channel (Fig. 6). We made two measurements every 0.4

- 0.45 mm along the channel, as shown in Figure 6. One value was obtained for laser-modified

(white) regions, w1, while the other was made to include etched (dark) regions, w2,

respectively. The measured data was plotted as presented in Figure 7.

Figure 6. Measurements of the channel width at different positions along the channel of the 2nd generation chip

(a) at modified (w1) and (b) etched (w2) areas at (1) junction; (2) 4.5 mm; (3) 10 mm (4) 13 mm; (5) 16 mm (6)

21 mm and (7) 27 mm along the channel. Red line across the channel shows the location of the measurement.

Figure 7 reveals five distinct regions for etching behaviour: (I) a large difference

between w1 and w2 is observed here; (II) exhibits a monotonic decrease in the difference

between w1 and w2 as channel distance increases, (III) w1 ≈ w2; (IV) shows a linear increase

in the difference between w1 and w2 and (V) exhibits almost constant w2. Also noteworthy is

the fact that w1, the width measured across the laser-modified region of the final channel,

remains constant along the entire length of the channel. Only in region (III), where modified

and etched areas overlap, do both channel width and groove parameters equal those in the

original design. This area comprises less than 3 mm (12 %) of the total 30-mm length of the

channel. Based on these data, we created a compensation design (Fig. 8A) with a channel that

has five regions with different channel parameters.

The final channel design has five regions (Fig. 8A) where different changes to the

original design were made, including channel width and channel depth. Table 3 summarizes

dimensions in each region of the compensation design. The maximum difference between w1

and w2 was calculated to be 190 µm. Thus, the channel widths in Regions I and V have been

assigned values of 240 m rather than 430 m as in the original device. This resulted in the

channel with narrower region in the beginning and at the end comparing to the middle region.

In Region II, the channel width increased linearly from 240 m to 430 m, whereas it was


131

reduced linearly from 430 m to 240 m in Region 4. In general, a channel has a narrower

width of 240 µm in the beginning and at the end compared to the 430 µm in middle section,

where the original design was kept unchanged. The channel was deepest in the middle section

of the channel (Region IV) (150 m; Fig.8D) and was designed to become shallower towards

the ends of the channel.

Figure 7. Difference between channel width of modified (red curve, w1) and etched (black curve, w2) regions

based on the channel of the 2nd generation chip: (I) increase in difference between w1 and w2; (II) linear decrease

of difference between w1 and w2, (III) w1 ≈ w2; (IV) linear increase of difference between w1 and w2 and (V)

constant w2.

An important consideration in the compensation design was to adjust groove dimensions

(e.g. groove width, ridge width and groove depth) to counteract effectscaused by excessive

etching in the channel. This is especially true in regions (II) and (IV), where channel width

linearly decreases towards the middle of the channel and increases again afterwards. Channel

depth increases as the middle section of the channel is approached, and decreases again once

the midpoint of the channel has been passed. As with the channel, the HG design resemble the

originally designed HG only in the middle section (Fig. 8B). All grooves touched the walls of

the new channel. In addition, groove depth was also adjusted according to the changing channel

Ch

ap

ter

IV


132

depth. All HG dimensions for each region are summarized in Table 3. Figures 8C and 8D show

the groove depth in the beginning and in the middle section of the channel. The etching process

also influences groove width as can be seen on Figure 6. Therefore, we measured the difference

in groove width the same way as was described above for the channel width (Fig. 6). In the

original design the groove and ridge widths are 260 µm and 70 µm respectively. In order to

keep a groove width of 260 µm after etching, grooves in the compensation design were

narrowed to different extents in all regions except the middle section of the channel (Table 3).

The same approach was taken for ridges that decrease during etching. Therefore, in order to

maintain the same ratios between grooves and ridges, we kept the distance taken up by each set

of groove + ridge along the channel the same, at 330 µm.

Figure 8. SolidWorks© design of the compensation chip design with herringbone grooves representing different

regions of the channel with different: (A) channel width; (B) groove and ridge width (dark regions); groove depth

at the (C) beginning (70 µm) and (D) middle part (100 µm) of the channel.

Table 3. Five regions of the compensation design with different parameters.

Region in

the channel,

mm

Groove

width, mm

Ridge

width, mm

Channel width, µm Groove depth, µm

0 – 6.5 0.2 0.13 240 70

6.5 – 12.7 0.22-0.24 0.11-0.1 240-430 (linear

increase)

70-100 (linear

increase)

12.7 – 15.2* 0.26* 0.07* 430* 100*

15.2 – 25.2 0.23-0.21 0.10-0.12 430-240 (linear

decrease)

100-70 (linear

decrease)

25.2 – 42.3 0.2 0.13 240 70

*The dimensions are the same as in the 2nd generation device.


133

Based on these adjustments, we designed the chip with compensation structures (the 3rd

generation chip) which is presented in Figure 9A. We noticed improvement in terms of

uniformity of the channel width compared to the previous designs. The measurements of

channel widths in the laser-modified and etched parts of the 3rd generation chip (Figure 10)

clearly show that the middle section having the desired width (430 µm) is ~ 6 mm long, which

is larger than for the 2nd generation chip (2.5 mm). Squeezing the two sides of the channel even

more could help obtain channels with an even more uniform cross-section. On the other hand,

narrowing the beginning of the channel too much is undesirable due to the associated reduction

of the etching solution access and, thus, an increase in the etching time. However, etching didn’t

quite proceed as we had assumed or hoped. In regions (I) and (V), the maximal difference

between the modified and etched parts still remains more than 300 µm. These regions better

resemble the compensated design than the desired original design with a narrower channel in

the beginning and at the end of the chip.

Figure 9. Photos of the 3rd generation chip with compensation design: (A) top-view of the whole channel, 40x

magnification objective lens and (B) side-view showing the difference in channel depth.

As with channel width, groove shapes have also been altered to take compensation for

longer etching times into account. compensated design (wider in the beginning and at the end

of the channel). Besides, Figure 9B reveals clear variations in channel depth. Probably, the

difference of 30 µm in depth between the beginning/end and the middle part of the channel was

not sufficient to compensate long etching times and to obtain uniform channel depth. However,

the side-view of the 3rd generation chip shows that the resolution of the HG is better compared

to 2nd generation chip. Ridges are present along the whole channel length, which probably

accounts for the better mixing performance observed (see Section 3.3).

Ch

ap

ter

IV


134

Our approach for compensation of the Solidwrks designdesign for the conical channel

shape is somewhat similar to the method proposed by Vishnubhatla et al.19 That approach

consisted of an irradiation of a reverse conical-shaped channel with respect to the one normally

obtained with the SLE technique. However in our approach we decided to make changes to the

actual device design and not to the SLE procedure. It is also worth mentioning that the method

proposed by Vishnubhatla et al. would be difficult to realize in our chip due to the presence of

herringbone grooves on the bottom.

Another possible solution to decrease the influence of excessive etching could be to

cycle the etching process,17 i.e. by frequently interrupting it. Acid or base would then be

replenished in the narrow modified regions during each cycle to decrease the effect of excessive

etching.

Figure 10. Difference between channel width of laser-modified (red curve, w1) and etched (black curve, w2)

regions of the 3rd generation chip channel. The marked regions represent the five regions of the original design

(Table 3).

Evaluation of mixing performance

All chip generations were tested for their mixing performance. As the mixing is enhanced by

the herringbone grooves on the bottom of the channel, the mixing quality relies on how good

the resolution/quality of the obtained grooves is. In other words, how well the original design

of HG was transferred into the fused-silica chip is of critical importance. We used a qualitative


135

approach with food dyes to visualize the mixing. Because of different channel dimensions, the

mixing performance was tested at different flow rates according to the same Peclet number

representing the same mass transfer conditions (Section 2.4).

Figure 11 and Figure 12 show results for mixing experiments with all three chip

generations at Pe values of 3.5×103 and 17.5×103, respectively. For both Pe, the best mixing is

observed in the 3rd generation mixer, as indicated by the appearance of green colour across the

entire channel much earlier. Such performance can be attributed to several factors. Though the

green colour starts to appear within first few mm of the channel, this is probably caused not

only by the existence of grooves. As was discussed above, the depth of the 3rd generation chip

decreases towards the middle section of the channel, which linearly increases the groove depth-

to-channel height ratio (Fig. 9). Previous studies26,27 showed that mixing performance of

herringbone grooves improves with an increase in the value of this ratio, which is explained in

detail elsewhere.22 Thus, the mixing at the beginning of the channel can probably be attributed

in the first place to the decreased width and depth of the channel at the junction where fluid

streams are physically brought into the contact (Fig.11C). On the other hand, mixing based on

chaotic advection occurs only in the middle section of the channel where grooves correspond

to or closely resemble their original design. This can be clearly seen in Figure 12D. Similar

mixer performance was observed in all studies involving similar herringbone-groove

designs.3,4,28 The 2nd generation chip also has a region where chaotic advection takes place,

however, the overall mixing performance is poor. The 1st generation chip (Fig. 11A, Fig. 12A)

has a very small region with well-resolved grooves, which results in even poorer mixing

performance.

Figure 11. Mixing performance of the (A) 1st, (B) 2nd and (C) 3rd generation fused-silica micromixers; Pe = 3.5×103

(actual flow rates: 0.12 mL/min for (A) and 0.2 mL/min for (B) and (C)); two food dyes; combined photos.

Ch

ap

ter

IV


136

However, the mixing is not complete by the end of the channel for any of chips, which

is obvious from the streaks of unmixed dyes at the channel outlet (Fig.11 and 12). This relates

to the insufficiently long channel region exhibiting the generation of chaotic advection.

Furthermore, at higher flow rates the mixing efficiency decreases as well, as evidenced by the

green-coloured area that appears further down the channel and is less pronounced at lower flow

rates. This effect can be explained by the decrease in the residence time and, thus, decrease in

the mixing that is governed by diffusion.

Clearly then, the fabrication process needs to be optimized to realize all the grooves

initially patterned by the pulsed laser, as only then will these chips function as efficient mixers.

Figure 12. Mixing performance of the (A) 1st, (B) 2nd and (C) 3rd generation fused-silica micromixers; (D) the

middle part of the 3rd generation chip where mixing by chaotic advection is evident; Pe =17.5×103; combined

photos.

Pressure tests

Because of the original intention to place the chip in the interface of a two-dimensional liquid

chromatograph, and the future plan to integrate a monolithic column into the same chip, the

fabricated chips had to be tested for pressure resistance. For this test only the 2nd generation

chip was chosen, as the 1st generation chip was received with a small crack in the inlet.

In the test, the column connected to the chip outlet was left open. The pressure was

increased from 0 to 22 bar for flow rates of 0.2 - 2.0 mL/min. However, no changes in the chip

or connected tubing were observed. Next, the column was closed off. The results of this second


137

experiment are summarized in Table 4. In the first two attempts, the tubing detached from

different inlets as the pressure increased. In the last experiment, when the pressure reached 85

bar, a big part of the outlet region split from the rest of the chip and the chip was destroyed

(Fig.13). Device fracture in the outlet region could be expected, as the used silica was not very

thick and fell subject to high total back pressures. Thus, the pressure test showed that the fused-

silica chip could withstand a pressure of 85 bar. This value is sufficient for utilization of this

device in the interface of two-dimensional liquid chromatography as a separate mixing device.

However it may not be sufficient for applications involving an integrated monolithic stationary

phase on the same chip, as the channel filled with stationary phase would increase the flow

resistance and, thus, the internal pressure drop in the device itself.

Table 4. Pressure test with the 2nd generation chip when the column was sealed at the total flow

rate 0.5 mL/min.

Total max pressure Observation

63,5 bar Different inlets leakage

77.6 bar

85.15 bar The chip split into two part in the outlet region

Figure 13. (A) Top-view and (B) side-view of the two parts of the 2nd generation chip after the pressure test.

In the literature, there is limited information regarding glass-based microfluidic devices

for high-pressure chemistry. In most cases, the main limitation associated with the development

of such devices is their micro-to-macro interface (i.e. quality of the connections). Working

pressures of 50-150 bar were achieved when glass chips were placed in different types of clamp-

holders or similar support units.29,30 Szekely and Freitag29 clamped a glass chip into a Teflon

holder/interface with integrated high pressure connectors and O-rings that withstood back

pressures of 150 bar. The disadvantage of this approach is reduced flexibility of patterning, as

Ch

ap

ter

IV


138

only straight lines and crosses can be fabricated. Tiggelaar et al.11 reached working pressures

of 300 bar using an in-plane, fiber-based interface with glass capillaries connected to a glass

microreactor. However, in this approach epoxy resin fibers were glued to the chip, which

decreased the flexibility for connection of the chip to different equipment. In our work, we also

used in-plane connectors for improved robustness. In contrast to other studies, no extra holders

or gluing is needed in our approach. Having female connectors with standard 10-32 thread

(suitable for standard HPLC connectors) allows easy connection of the chip to conventional

equipment and makes the interface user-friendly. Even though our chip didn’t reach pressures

beyond 85 bar, the obtained result is compatible with other high-pressure resistant chip-based

devices. Besides, it should be noted that the chip was not placed in any clamp-holder or support;

doing so would most likely allow for much higher operating pressures.


139

Conclusions

In this work, we explored the fabrication of microfluidic mixers with integrated arrays of

herringbone grooves in fused silica using the Selective Laser-Induced Etching (SLE) technique.

This approach allowed us to obtain devices with complex features inside solid pieces of fused

silica, eliminating the need for a bonding step. We aimed to fabricate microfluidic mixers with

channels up to 33 mm long and with aspect ratios more than 100, the first time that SLE has

been applied for this type of application. We managed to fabricate several generations of chips

with different dimensions. Our results showed that increasing the channel diameter allows

higher-resolution channels with grooves, due to faster access of etching solution. However, we

didn’t succeed in the fabrication of channels with uniform cross-section. The channel still had

a tapered, conical shape toward the middle of the channel, due to the etching process. To

overcome this effect, we proposed a compensation design that resulted in slightly better

resolutions of grooves in the channel.

All micromixers were tested for mixing performance. Tests revealed that mixing based

on chaotic advection is observed only in the middle part of the channel where herringbone

grooves correspond to or resemble most the original design. The best mixing performance was

observed for the chip with compensation design, due to changes in channel depth/width and

better resolution of herringbone grooves. However, the mixing was still not completed by the

end of the channel at either of the tested flow rates (0.2 mL/min and 1.0 mL/min).

In terms of pressure, the obtained chip can withstand pressures up to 85 bar. This is

within the pressure range reported for the glass chips placed in different types of clamp-holders

or similar support units.29,30 However, in our tests the chip wasn’t placed in any housing, which

means that the pressure resistance could be improved with both utilization of the extra support

and using a thicker block of fused silica. Compared to other studies, no extra holders or gluing

is needed in our approach. The device can be easily connected to any conventional equipment

using standard 10-32 thread (standard HPLC connectors).

We believe that it is worthwhile to further explore the fabrication of microfluidic

devices using SLE. To achieve desired structures, an optimized fabrication procedure using a

compensation design to account for variation in exposure times to etching solution should be

developed.

Ch

ap

ter

IV


140

Acknowledgements



(HYPERformance LC).


141

References



2. Sollier, E., Murray, C., Maoddi, P. & Di Carlo, D. Rapid prototyping polymers for microfluidic devices

and high pressure injections. Lab Chip 11, 3752 (2011).

3. Ianovska, M. A., Mulder, P. P. M. F. A. & Verpoorte, E. Development of small-volume, microfluidic

chaotic mixers for future application in two-dimensional liquid chromatography. RSC Adv. 7, 9090–9099

(2017).


5. Liu, J. et al. Polymer Microchips Integrating Solid-Phase Extraction and High-Performance Liquid

Chromatography Using Reversed-Phase Polymethacrylate Monoliths. 81, 2545–2554 (2009).

6. Chen, C. F. et al. High-pressure needle interface for thermoplastic microfluidics. Lab Chip 9, 50–55

(2009).

7. Mair, D. A. et al. Room-temperature bonding for plastic high-pressure microfluidic chips. Anal. Chem.

79, 5097–5102 (2007).

8. Urakawa, A., Trachsel, F., von Rohr, P. R. & Baiker, A. On-chip Raman analysis of heterogeneous

catalytic reaction in supercritical CO2: phase behaviour monitoring and activity profiling. Analyst 133,

1352–1354 (2008).

9. Trachsel, F., Hutter, C. & von Rohr, P. R. Transparent silicon/glass microreactor for high-pressure and

high-temperature reactions. Chem. Eng. J. 135, 309–316 (2007).

10. Marre, S., Adamo, A., Basak, S., Aymonier, C. & Jensen, K. F. Design and packaging of microreactors

for high pressure and high temperature applications. Ind. Eng. Chem. Res. 49, 11310–11320 (2010).

11. Tiggelaar, R. M. et al. Fabrication, mechanical testing and application of high-pressure glass microreactor

chips. Chem. Eng. J. 131, 163–170 (2007).




quartz glass components for microfluidic applications-up-scaling of complexity and speed.

Micromachines 8, (2017).



15. Osellame, R., Hoekstra, H. J. W. M., Cerullo, G. & Pollnau, M. Femtosecond laser microstructuring: An

enabling tool for optofluidic lab-on-chips. Laser Photonics Rev. 5, 442–463 (2011).

16. Kiyama, S., Matsuo, S., Hashimoto, S. & Morihira, Y. Examination of etching agent and etching

mechanism on femotosecond laser microfabrication of channels inside vitreous silica substrates. J. Phys.

Chem. C 113, 11560–11566 (2009).

17. Hnatovsky, C. et al. Fabrication of microchannels in glass using focused femtosecond laser radiation and

selective chemical etching. Appl. Phys. A-Materials Sci. \& Process. 84, 47–61 (2006).

18. Osellame, R., Maselli, V., Martinez Vazquez, R., Laporta, P. & Cerullo, G. Integration of optical

waveguides and microfluidic channels fabricated by femtosecond laser irradiation. Conf. Lasers Electro-

Optics, 2007, CLEO 2007 231118, 88–91 (2007).

19. Vishnubhatla, K. C., Bellini, N., Ramponi, R., Cerullo, G. & Osellame, R. Shape control of microchannels

fabricated in fused silica by femtosecond laser irradiation and chemical etching. Opt. Express 17, 8685–

8695 (2009).

20. Maselli, V. et al. Fabrication of long microchannels with circular cross section using astigmatically shaped

femtosecond laser pulses and chemical etching. Appl. Phys. Lett. 88, (2006).

21. Venturini, F. et al. Maskless, fast and highly selective etching of fused silica with gaseous fluorine and

gaseous hydrogen fluoride. J. Micromechanics Microengineering 24, 025004 (2014).


7, 580–587 (2007).

23. Ghoreishi, S. M., Behpour, M. & Golestaneh, M. Simultaneous voltammetric determination of Brilliant

Blue and Tartrazine in real samples at the surface of a multi-walled carbon nanotube paste electrode. Anal.

Methods 3, 2842 (2011).

24. DIACU, E., UNGUREANU1, E.-M., ENE, C. P. & IVANOV, A. A. Voltammetric Studies for Detection

and Degradation Assessment of some Synthetic Food Dyestuffs. REV. CHIM. 62, 1085–1089 (2011).

Ch

ap

ter

IV


142

25. Lin, D. et al. Three-dimensional staggered herringbone mixer fabricated by femtosecond laser direct

writing. J. Opt. 15, 025601 (2013).




Lab Chip 5, 1140–1147 (2005).


29. Szekely, L. & Freitag, R. Fabrication of a versatile microanalytical system without need for clean room

conditions. Anal. Chim. Acta 512, 39–47 (2004).

30. Shintani, Y. et al. Polydimethylsiloxane connection for quartz microchips in a high-pressure system. Anal.

Sci. 20, 1721–1723 (2004).

Chapter V

Microfluidic micromixer as a tool to

overcome solvent incompatibilities in

two-dimensional liquid

chromatography

Margaryta A. Ianovska1,2, Arto Heiskanen3, Bert Wouters4, Erik Ritzen5, Ynze

Mengerink5, Peter Schoenmakers4, Jenny Emnéus3, Elisabeth Verpoorte1




3Department Micro- and Nanotechnology, Technical University of Denmark, Denmark

4University of Amsterdam, Van ‘t Hoff Institute for Molecular Sciences, Amsterdam, The

Netherlands

5DSM Resolve, Geleen, The Netherlands

Manuscript in preparation

Abstract

We report a micromilled, pressure-resistant mixing chip for application in on-line

comprehensive two-dimensional liquid chromatography (LC×LC). This microfluidic mixer is

based on chaotic advection generated in grooved microchannels, and can be operated at flow

rates compatible with LC×LC (0.1-1 mL/min). The design of this chip was optimized and tested

in our previous work in the transparent and flexible silicon rubber material,

polydimethylsiloxane (PDMS). In this work, the microfluidic chip was micromilled in rigid

cyclic-olefin copolymer (COC) substrate in order to better withstand the high pressure

environment of LC×LC. Two micromilled parts were bonded using solvent-vapour-assisted

bonding under elevated temperature and pressure. A specially designed, robust, low-dead-

volume interface allows direct connection of the microfluidic chip to an LC×LC system using

standardized HPLC connectors. Thus-fabricated chips can withstand pressures of 200 bar. The

chip was successfully implemented in a comprehensive two-dimensional HILIC×RP-LC

system for improved separation and identification of various oligomeric series in nylon polymer

(polyamide) samples. Initial trials of the micromixer in a simple modulator at the interface

between the two columns yielded chromatographic performance similar to that obtained with

commercially available mixing units. However, with the grooved micromixer, it was possible

to mix rapidly within the same flow rate range in a volume, which was 30 times smaller than

the commercially available micromixers, opening a route to utilization of microfluidic mixers

in the conventional equipment.

Keywords: microfluidic chip-based technologies; cyclic-olefin copolymer (COC) devices;

micromilling; two-dimensional liquid chromatography (LC×LC).

Microfluidic micromixer as a tool to overcome solvent incompetibilities in two-dimensional

liquid chromatography

145

Introduction

With the growing complexity of real samples, there is a need to analyze these in the most

complete way possible. While liquid chromatography is a powerful separation tool, a single

column does not offer sufficient resolution for such samples, which can, for example, contain

many different proteins, peptides or inorganic polymers. This has led to the rapid development

of techniques that employ combinations of columns with different separation mechanisms and

thus different selectivity, such as two-dimensional liquid chromatography (2D-LC).1,2 In order

to maintain the resolution obtained in the first dimension (1D) in the second dimension (2D),

the two dimensions are coupled through a special interface that performs all the required

manipulations of the 1D effluent, such as sampling, storing and re-injection of small 1D effluent

fractions. In comprehensive two-dimensional (2D) liquid chromatography (LC×LC), where all

sample material passes through both columns before reaching the detector,3 it is especially

important that the switching frequency of the interface is high. The smaller the volumetric

sample fractions are and the faster they are transferred to the 2D, the better the 1D resolution is

maintained.

Each separation mechanism in LC×LC requires its own mobile phase for the best

separation of analytes. The utilization of different stationary phases and, as a consequence,

different mobile phases in LC×LC means that the fractions dissolved in the 1D mobile phase

need to be transferred to the second dimension with a mobile phase having a different

composition. Ideally, the second-dimension (2D) mobile phase should be a stronger eluting

mobile phase on the 2D column than the 1D mobile phase at the time of elution. In addition, the

2D column should have a higher retention with this latter mobile phase than the 1D column.

Also, the 1D effluent and the 2D mobile phase at the moment of injection should be fully

miscible. If these conditions are fulfilled the band broadening caused by mobile phase

incompatibility is minimized.4 However, such a combination of mobile and stationary phases

is usually difficult to achieve, especially in highly orthogonal systems, such as hydrophilic-

interaction liquid chromatography (1D) followed by reversed-phase liquid chromatography (2D)

(HILIC×RP-LC). In this case, the 1D effluent is rich in organic solvent, giving it a high elution

strength in the RP dimension. If the order of columns were to be reversed, the water-rich

effluent from the 1D RP column would act as a strong eluent in a 2D HILIC system. Hence, the

solvent in the collected 1D fraction is a stronger eluent on the 2D column than the 2D mobile

Ch

ap

ter

V


146

phase,5 which can cause a decrease in retention, increased band broadening, and non-

symmetrical or even split peaks.6,7

To overcome the solvent-strength mismatch problem between the dimensions in

LC×LC, a number of solutions have been previously suggested, including trapping-column

interfaces8–10 and collection of very small fraction volumes from a 1D capillary monolithic

column.7 More sophisticated approaches utilize a vacuum-evaporation interface for on-line

evaporation of 1D solvent from the loop11 or thermally assisted modulation using the influence

of temperature on the retention of analytes.12,13 While each of these approaches has unique

advantages for solving the problem of mobile-phase incompatibility, these technologies are still

not developed to the extent where they could be used universally in routine LC×LC analyses.

An approach which is more generally applied is the use of an additional solvent flow

(make-up flow) to allow modification of the solvent composition between dimensions by

diluting the collected 1D fraction with a weaker 2D mobile phase. However, this requires the

use of a good mixing device at the interface between the two columns. Moreover, the mixer

must be able to withstand brief pressure pulses up to a few hundred bar, due to its connection

to switching valves used to shuttle sample from the 1D to the 2D. Nowadays, T-pieces14 and

static mixers (S-mixers)15,16 are used for this purpose. A T-piece mixer provides mixing by

causing two streams to collide at a T-junction to create an interdiffusion region where mixing

by diffusion takes place more rapidly.17 S-mixers, on the other hand, are based on two elements

in a circular tube, each consisting of multiple X-shaped cross-bars positioned at 45 with respect

to the axis of the tube, but with the second element rotated 90° with respect to the first.18 These

two elements form a unit which is repeated a number of times. Mixing is achieved by splitting

the main flow into sub-streams and then re-joining them. Due to this split-and-recombine

mechanism, the diffusion length between the two fluids decreases due to an increase in the

contact surface area. Both T-piece and S-mixer approaches exhibit better performance at higher

flow rates and larger mixing-unit volumes.19 For the flow-rate range (100-300 µL/min) that is

used for most 2D-LC separations, the volume of the suitable mixer should be at least 150 µL

(for S-mixer). Such volumes are not in keeping with the low 1D fraction volumes that have to

be manipulated, which makes the application of these mixing units in LC×LC not efficient

enough. 2D-LC thus stands to benefit from the development of fast, small-volume micromixers.

The development of chip-based devices integrated in LC or LC-MS systems using

microfluidic technology has increased substantially in the last decade.20–25 The key advantages

of chip-based microfluidic systems are their small size, resulting in small-volume devices, and



147

their applicability for controlling small amounts of liquids, which is beneficial for application

in LC. However, many chip-based devices implemented in LC and LC-MS systems are

operated in nano-LC mode at flow rates in the 100-300 nL/min range,20,22,24,25 whereas the

conventional LC×LC applications imply utilization of higher flow rates. The connection to

conventional equipment operating at high pressures implies that devices should be also

pressure-resistant. In previous work,26 we developed a 1.6-μL mixer with herringbone grooves

(also referred to grooves having an asymmetric chevron geometry) that provides good mixing

within seconds at flow rates compatible with LC×LC (0.1-1 mL/min). This mixer has a small

internal volume in order to obtain the desired dilution ratios in minimal volumes. The mixing

is achieved through the generation of two counter-rotating vortices (perpendicular to the

direction of the main flow) based on chaotic advection,27 a substantially different mechanism

than in T-piece and S-mixers. However, in that research we employed poly(dimethylsiloxane)

(PDMS), a transparent and flexible silicone rubber, the elastomeric nature and low Young's

modulus of which become a significant problem for development of high-pressure-resistant

chips. At high flow rates, significant channel deformations (resulting from high-pressure flows)

eventually lead to the fatigue and cracking of PDMS.

In order to cope with the pressures used in conventional LC systems, we fabricated our

micromixer with herringbone grooves in cyclic-olefin copolymer (COC) using micromilling.

We decided to use COC, because of its compatibility with typical solvents used in HPLC (e.g.

acetonitrile, methanol, and isopropyl alcohol) and excellent optical and mechanical

properties.28,29 Moreover, COC has substantially lower raw material costs (compared to silicon

and glass) and can be easily used in conjunction with micromilling.29 This fabrication method

was chosen due to its applicability for the rapid translation of designs into prototypes, and it

met our requirements with respect to cost and resolution as well.30

Beside the need for the chip itself to withstand pressure for the LC applications, a

connection between conventional equipment and the chip, the so-called macro-to-micro

interface, should be pressure-resistant. However, there are only a few articles that propose

pressure-resistant interfaces compatible with LC equipment.22,31–34 Wouters et al.33 developed

a macro-to-micro interface for connection of a COC chip to LC equipment using a custom-

made aluminium chip holder with integrated flat-bottom Nanoport connections to connect 360-

μm-o.d. capillary PEEK tubing to the chip. The burst pressure (the pressure that a device can

withstand before failing or coming apart) was determined to be between 15 and 20 bar using

Ch

ap

ter

V


148

this approach (channel cross-section was 500 µm × 500 µm). Though commercial Nanoports

have gained popularity as a straightforward approach to create interconnections between chips

and peripheral equipment, they cannot be used for applications that require pressures above ca.

70 bar.35 Mair et al.31 reported an interface with a threaded mating port for a standard coned

capillary fitting directly fabricated into an injection molded COC chip. The chip withstood 156

bar and was broken due to its delamination rather than interconnect failure. Later this approach

was improved by changing the thermal fusion bonding protocol to solvent-vapour bonding and

subsequent exposure of the bonded chip to UV light, to achieve a burst pressure of 346 bar.32

However, these authors used injection molding, which requires a new mold for each chip design

and can be costly and time consuming in terms of rapid prototyping. Agilent developed a

layered, polyimide microfluidic HPLC chip which incorporated an enrichment column, a

packed analytical column, and a nanospray emitter with a burst pressure of 200 bar.22 The input

capillaries were filled with a slurry of particles in 2-propanol. Each capillary was then coupled

to the chip and isopropanol was used to flush the slurry into the chip under a pressure of 120

bar. The chip was sealed using standard 3.2-mm diameter ring, which enables chip pressure

resistance up to 200 bar. In this system, standard HPLC connectors were used. On the other

hand, Chen et al.34 used a non-standard approach using either a stainless steel needle with or

without the self-tapping thread, that was inserted into a mating hole or screwed into an inlet

hole respectively. In both cases, the flat bottom needle part was directly in contact with the

microchannel. Both interfaces were compatible with pressures up to 400 bar. However, a

potential issue is that such an interface is not universal and thus standardized HPLC tubing

cannot be used. Furthermore, during the process of inserting the needle, a substrate can easily

be cracked.34

Here we report the successful application of a micromilled thermoplastic-polymer

micromixer at the interface of an LC×LC system for overcoming mobile-phase mismatch

between dimensions. The micromixer contains array of microfabricated grooves having so-

called “herringbone” or asymmetric chevron geometry. These grooves perturb the side-by-side

laminar flow profile of two solutions to be mixed such that two counter-rotated vortices are

established. Solution streams are thinned leading to shorter diffusion length and thus faster

mixing. The mixer developed in this study had an internal volume of 4.65 µL, which is much

smaller volume than conventional units have. Using a custom-designed, robust, low-dead-

volume interface, a mixing chip can be directly coupled to the LC×LC equipment using

standardized HPLC connectors. The operating pressure for the chip, which is clamped in a metal



149

holder, is up to 200 bar. We implemented our micromilled, COC microfluidic mixer in a

HILIC×RP-LC system. When our mixer was used to incorporate make-up flow, an improved

separation was obtained compared to the system with no make-up flow, as expected. We

successfully identified various oligomeric series in nylon (Polyamide 46) samples as a

demonstration of the 2D-LC system with implemented micromixer.

Ch

ap

ter

V


150


Chemicals and reagents

All chemicals were analytical-reagent grade. Fluorescein was purchased from Sigma-Aldrich

(NL) and used to prepare separate 5.0 μM fluorescein solutions in 10.0 mM phosphate-buffered

saline, pH 7.4 (PBS; Gibco, UK). The pH was measured using pH-indicator strips (Neutralit,

MERCK). All solutions were prepared with 18 MΩ-ohm ultrapure water (Arium 611, Sartorius

Stedim Biotech, Germany).

Formic acid (FA) and mass-spectrometry-grade acetonitrile (ACN) were purchased

from Merck (Darmstadt, Germany). Milli-Q grade water (Merck Millipore system 0.22 µm)

was used for preparation of mobile phases. Acetone (BASF, Ludwigshafen, Germany) was used

for the dispersion test.

A list of the chemicals (and suppliers) used in this study is reported in the Supplementary

Information (SI).

Fabrication of the COC chip

COC substrates (grade TOPAS 5013L-10) were purchased from Kunststoff-Zentrum

(Leipzig, Germany). This grade was chosen to minimize occurrence of burrs (raised edges or

small chips of plastic that remain attached to the workpiece)30 during the milling process. A

device consists of a top and a bottom plate which are 2 mm and 5.5 mm thick, respectively. All

polymers were used as received in plates and cut to the desired dimensions in-house (30×58

mm). Channels with grooves were designed in SolidWorks (Waltham, MA, USA), and the G-

Code was generated from these designs for the milling machine (Autodesk, San Rafael, CA,

USA). The actual groove and channel dimensions may be found in Figure 3 (Results and

Discussion section). The bottom part of the chip was micromilled using a computer-

numerically-controlled (CNC) micromilling robot (Mini-Mill/3, Minitech Machinery,

Norcross, GA, USA) equipped with a brushless electronic Astro-E500Z spindle (NSK,

Schaumburg, IL, USA) with a maximum rotational speed of 50,000 rpm. MACH3 CNC was

used as control software. The channel containing grooves was milled using 2-flute endmills of

200 µm and 100 µm, respectively. Several bottom parts were milled with only a channel, as test

pieces for establishing the optimal bonding conditions. The spindle speed was set to 15,000

rpm. It took 1 hour for the milling of the channel and 5 hours to mill grooves. In order to prevent



151

melting during milling due to the heating of the substrate, a cooling liquid (water with soap)

was continuously flushed over the COC surface.

The cover plate was polished with a PICOMAX 20 machine (Fehlmann, Seon,

Switzerland) to obtain better flatness for bonding later on. Conical access holes were drilled

into this piece using a high-speed custom step drill. Prior to bonding, the cover plate and the

channel chip were thoroughly cleaned with deionized water and isopropanol, and dried under a

N2 stream. The plates were then exposed to the vapour of the cyclohexane for 8 min at room

temperature (COC activation). For this, plates were placed over a square glass reservoir (appr.

20×10×5 cm) filled with approximately 200 mL of cyclohexane inside a closed container.

Afterwards, the top and bottom plates were aligned using 1.59 mm (1/16“) polyether ether

ketone (PEEK) tubing (Kinesis, Altrincham, Cheshire, UK). Two extra holes (1.6 mm diameter)

were drilled on the opposite side of both plates for this purpose. An assembled chip was placed

between the two chucks of a hot embosser (Dr. Collin, Ebersberg, Germany). The device was

first heated from 25°C to 120°C in 20 min, then compressed with a force of 15 bar (377 N/cm2)

for 15 min at 120°C, and finally passively cooled down to 25°C.

To evaluate bonding strength, each chip was clamped in the metal holder (see following

section) and the inlets were connected to HPLC pumps (Waters 515, Waters, Milford, MA,

USA). An HPLC column (HYPERSIL SAX 5U 4.6×150 mm, Alltech Associates/Applied

Science, Carnforth, UK) was used as a restrictor at the outlet, and water was pumped through

at a total flow rate of 0.5 mL/min. After a linear increase in pressure, a sudden pressure drop

was observed, indicating the burst pressure of the device.

Micro-to-macro interface

There are a few important requirements for the macro-to-micro interfaces of microfluidic chips,

which should be met in order to be applicable in LC: they should be compatible with

standardized connectors for easy implementation, provide adequate pressure stability and not

introduce dead volume.36

In order to connect the micromixer to other instruments, a specially designed micro-to-

macro interface was developed (Fig. 1). The chip was clamped between two brass parts that

were screwed together. For the creation of a female connector, a conical hole in the shape of an

HPLC ferrule (ca. 0.5 mm) was drilled into the top COC plate and a standard 10-32 threaded

hole was fabricated in the metal top part (8 mm diameter) (Fig. 1A). The bonded COC chip and

Ch

ap

ter

V


152

top part of the metal holder were aligned using small pieces of 1.59 mm (1/16“) PEEK tubing,

inserted through the standard 10-32 holes in the top metal part and into the conical holes in the

device to center them around the conical holes drilled into the top of the COC device. The chip

was then sandwiched between top and bottom metal plates, and the metal plates clamped

together at the sides with 6 metal bolts (3 down each long side of the holder) (Figure 1B, 1C).

For experiments, PEEK tubing (1.59 mm o.d., 0.254 mm i.d.) was inserted through connector

holes and standard HPLC fittings were twisted into the connector holes until they were finger

tight in the metal holder.

Figure 1. Micro-to-macro interface for establishing the connection with LC×LC: (A) side view of the COC device

aligned with top and bottom brass plates of the holder; holes for the female connectors were drilled partly in the

top COC plate (conical part) and in metal (10-32 thread). (B) side view and (C) top view of the fully assembled

chip clamped in the metal holder with finger-tight HPLC fittings, fixed into the assembly.

Standardized HPLC finger-tight fittings consist of a male nut (10-32 thread) with conical

ferrule and female tapered ferrule seat.37 Because COC substrate was not commercially

available in thicker plates (more than 10 mm), the fabrication of full female tapered ferrule seats

inside COC top plates was not possible. We decided to fabricate only the conical part for ferrule-

and-pilot depth in the COC top plate. The rest of the female connector – suitable for 10-32

thread - was fabricated in the metal top part (Figure 1A). Dimensions that were used for the

fabrication of the female tapered ferrule seat are in agreement with LC industry standards.

The interface was designed in such a way that HPLC tubing was pushed directly down

onto the channel in the COC chip, and the standard HPLC fittings were introduced and turned

into the metal top part until finger-tight. This design is meant to minimize the dead volume in

the interconnects. Such a macro-to-micro interface allowed direct connection of the



153

microfluidic chip to the LC×LC using standard HPLC fittings. Moreover, clamping between

metal plates adds extra pressure stability to the chip itself. The metal holder is reusable and can

be used every time with a new chip. Fittings can be inserted easily, without extra force.

In order to evaluate the pressure stability of the connections and to measure the maximum

burst pressure, each inlet was connected to a UHPLC pump (Agilent 1290 Infinity, Agilent,

Waldbronn, Germany) and the outlet was connected to a restrictor (GL Sciences, Eindhoven,

The Netherlands, Inertsil ODS-3 column; 250 mm × 3 mm i.d., 3-µm particle size). The total

flow rate was set to 0.5 mL/min and the pressure profile was recorded. In the first pressure test,

the pressure limit of each pump was set to 150 bar and PEEK tubing was used. A second

pressure test was performed using stainless UHPLC tubing. The pressure limit of each pump

was now set to 200 bar. The chip was connected as explained above.

LC×LC setup and chromatographic conditions

The 2D-LC instrument in which the micromixer was tested was an Agilent Technologies

Infinity 1290, consisting of two binary pumps (G4220A and G4220B) (one for each column),

an autosampler (G4226A) with thermostat (G1330B), a thermostatted column compartment

(G1316C) with 2-position/4-port-duo switching valve (G1170A) equipped with two 80-µL

loops (Agilent, 5067-5426), and a UV detector (G4212A). The HILIC separation were carried

out using an Agilent ZORBAX Rx-SILcolumn (150 mm × 4.6 mm i.d., 5-µm particle size,

Agilent, Wilmington, DE, USA). In both HILIC and RP dimensions, water containing 0.1% FA

and 100% ACN were used as eluent components. These solvents were pumped in different

ratios using a 1D pump and a 2D pump. We performed 1D gradient separations at 30 μL/min

using the following program: 0 min 99% ACN, 500 min 5% ACN, 510 min 5% ACN, 512 min

99% ACN, then kept at the initial solvent composition for 88 min.

The second-dimension separation column (Waters Cortecs C18, 30 mm × 4.6 mm i.d.;

2.7 µm particles) was operated at 3 mL/min at 40°C. The following gradient was used: 0 min

1% ACN, 0.3 min 30% ACN, 0.36 min 70% ACN, 0.39 min 1% ACN. The modulation time

(the time between valve switches) was 0.6 min (36 sec).

The operating principle of the interface between the dimensions was as follows (Figure

2A). The microfluidic mixer containing mixes 1D effluent (30 μL/min) with a 0.1% FA aqueous

solution (200 μL/min) using the make-up-flow pump (Gilson 305, Middleton, Wisconsin,

USA). The mixed solution was introduced to an 80-µL storage loop (Loop 1). Each loop was

Ch

ap

ter

V


154

filled 100%, but some part of effluent goes to the waste. The contents of Loop 1 are transferred

with the flow from the 2D Pump to the 2D column upon switching the valve (Fig.2B), while

Loop 2 is being filled. This occurs in multiple alternating cycles during the whole 1D

chromatographic run.

During the sequence, one blank sample (Formic acid solution) was injected after each

sample, and no significant carryover of sample was observed. LC×LC separations and other

tests were performed in triplicate.

Figure 2. Schematic representation of the modulation interface of the LC×LC-Q-ToF-MS system. A two-

position/4-duo switching valve is the central component at this interface. (A) In the first position Loop 1 is filled

with 1D effluent, while the 2D pump pushes the contents of the Loop 2 to the 2D column. (B) In the second valve

position, the analytes are transferred from Loop 1 to the 2D while Loop 2 is being filled. In such interface the

microfluidic mixer is placed before the switching valve. (C) Microfluidic mixer connected to the 8-port valve with

two loops at the interface. 1D effluent is mixed with an aqueous solution of 0.1% FA inside the microfluidic mixer

in order to reduce the eluent strength in the second dimension.

All oligomers were identified manually based on high-accuracy mass-to-charge (m/z)

values measured by Q-ToF MS.



155

Sample Preparation

In the HILIC-UV experiments, 250 mg of Polyamide 46 (PA46) were dissolved in 25 mL of

formic acid. The volume of injected PA46 sample was 2 µL (20 µg of PA46). The polyamide

solution was injected directly into the LC×LC system without additional sample preparation.

Dispersion tests

Dispersion tests were performed using either 1) a zero-dead volume union (316 Stainless Steel,

Swagelok, Solon, OH, USA); 2) a T-piece (316 Stainless Steel, Swagelok); 3) a 10-µL S-mixer

(Ultra HPLC, binary input housing, 316 stainless steel, Analytical Scientific Instruments,

Richmond, CA, USA); 4) 150-µL S-mixer (Standard HPLC, binary input housing, 316 stainless

steel, Analytical Scientific Instruments); or 5) our microfluidic mixer with herringbone grooves.

A zero-dead-volume union was placed between an injector and a UV detector (Agilent

Technologies, G4212A) instead of the 2D column. The inlets of each mixing units were

connected to an injector and an additional pump, and the outlet was connected to the same UV

detector. For all experiments, 1 µL of a 1% solution of acetone was injected at a total flow rate

of 0.5 mL/min. A concentration profile was recorded with a photodiode array UV detector at λ

= 275 nm (the wavelength at which the acetone exhibits the maximum absorbance).38

Mass Spectrometry Parameters

The outlet of the LC×LC system was coupled online to an Agilent 6540 UHD accurate-mass

quadrupole time-of-flight mass spectrometer (Q-ToF MS) equipped with an Agilent Jet Stream

ESI source (Agilent, G1958-65138) through a T-piece (316 Stainless Steel, Swagelok, Ohio,

United States). The Jet Stream ESI source was operated in positive mode and instrument

parameters were set as follows: gas, nitrogen; sheath gas temperature, 350°C; sheath gas flow,

11 L/min; dry gas temperature, 300°C; dry gas flow, 8 L/min; and capillary entrance voltage,

3,500 V. The Fragmentor and Skimmer were operated at 125 and 65 V, respectively. The

normalized collision energy was 750 V (OCT 1RF Vpp). The scans were acquired in MS mode

in a mass range from 50 to 1700 m/z at a rate of 4 spectra/s.

Ch

ap

ter

V


156

Data Analysis

The LC system was operated using Agilent OpenLAB CDS ChemStation Edition, version

C.01.07 SRI. The LC×LC and MS were operated using Mass Hunter Workstation Software

LC/MS Data Acquisition, version B.05.01 (Agilent). LC-MS data were analyzed using LCLC

Software from GC Image(Lincoln, NE, USA; GC Image LC×LC-HPMS, version R2.5b3).

Quantitative evaluation of the degree of dispersion caused by the different mixing units tested

was determined using a chromatogram recorded at the maximum absorbance wavelength of

acetone using the Mass Hunter Workstation Software Qualitative Analysis software (version

B.06.00, Agilent, 2012). The same software was applied to identify oligomeric series. All

extracted ion chromatograms (EICs) were obtained with ± 20 ppm m/z expansion (the m/z value

window).



157


Our previous research was devoted to fabrication of the HG microfluidic mixer in

poly(dimethylsiloxane) (PDMS), with an optimal design for application in LC×LC.26 However,

due to the elastomeric properties of PDMS, microchannels deformed when high flow rates are

used as a result of the higher pressures induced.39,40 PDMS also exhibits poor compatibility

with solvents commonly used in LC, limiting its applicability in LC. To prevent microchannel

deformation, it was decided to use a more rigid material for the chip fabrication. Cyclic-olefin

copolymer (COC) was the material of choice, because of its excellent optical and mechanical

properties, and compatibility with the typical solvents used in HPLC.28,29 Micromilling was

used as fabrication technique, as it offers a low start-up cost and a fast way to translate designs

into prototypes.30 However, even with ongoing advancements in the technology, micromilling

does have some limitations in resolution, dictated by the endmills that are used.30 In this work,

the smallest dimension for the groove design was determined to be 100 µm (groove spacing),

whereas in our earlier PDMS device this parameter was 50 µm. To retain optimized mixing

performance, it was thus necessary to recalculate the other dimensions in the micromixer to

accommodate larger groove widths. This was done using an approach described in the

references 41 for optimized geometries.26,27 The re-calculations were based on ratios between

different channel and groove parameters according to the design protocol for optimized

geometry41 that was described elsewhere.26,41 The dimensions that were used in this work are

shown in Figure 3: the channel width and depth were set to 430 µm and 150 µm, respectively;

the groove width and depth were 260 µm and 100 µm; and the ridge between grooves was 100

µm (determined by the resolution of the milling).

In order to characterize the mixing performance of micromilled COC mixers with new

geometries, mixing experiments with different flow rates and ratios, including the ratio 1:7

(PBS:PBS) that was used in this work in the LC experiments, were performed (Suppl. Fig. 1).

At this ratio with the total flow rate of 230 µL/min, the COC chip showed sufficient mixing

performance (see the Supplementary Information).

Rounding of the internal corners of features by micromilling is caused by the shape of the

endmills used. Rounded corners have a radius equivalent to that of the endmill. Experiments

with a PDMS chip that incorporated rounded features were performed in advance (see the SI,

Ch

ap

ter

V


158

Supplementary Figure 2), but did not reveal any substantial adverse influence on the mixing

performance of these rounded groove structures.

Figure 3. The channel of the micromixer with herringbone grooves: (A) schematic channel side view with channel

and groove dimensions; (B) photograph of the micromilled COC channel at the Y-junction and (C) schematic top

view of the full microchannel.

Optimization of the bonding procedure

To bond the top and bottom COC plates, a solvent-vapour-assisted bonding approach was used.

This method allows the direct bonding of substrates to one another without the use of additional

adhesive materials added to the interface. This type of bonding is especially suited for

applications that require high pressure resistance. Solvent bonding of thermoplastics takes

advantage of polymer solubility in the selected solvents.42 Exposure of the COC surface to a

vapour phase can avoid excessive solvent absorption and allow a more-controlled distribution

of the solvent molecules over the polymer surface.42 Cyclohexane was chosen as a solvent for

the bonding procedure, based on very similar Hildebrandt solubility parameters (which provide

a numerical estimate of the degree of interaction between materials) of COC and cyclohexane

(17.7 and 16.7 J1/2 cm−3/2, respectively).43 When the surface of the COC substrate is solvated,

polymer chains become mobile and can diffuse across the solvated layer into another similarly

solvated COC surface layer. This leads to mechanical interlocking of chains between the two

surfaces and creates an exceptionally strong bond between the two parts.42



159

It is known that bonding at elevated temperature can greatly enhance polymer

entanglement, which can result in stronger bonding.42 For this reason, after exposing both top

and bottom COC substrates to cyclohexane vapour and aligning and bringing them into contact,

the chip was heated up to 120°C (a bit lower than the glass transition temperature, which is

134°C for this COC grade), pressed together and cooled down to room temperature. Different

exposure times to cyclohexane vapour and applied forces were tested in order to find the optimal

bonding conditions (Table 1). Note that only devices containing non-grooved channels were

tested in this part of the study.

Table 1. Parameters that were tested for optimizing the solvent-vapour-assisted bonding

procedure.

Exposure

time, min

Force applied,

bar

Bonding

time, min

Burst

pressure,

bar

Reason for device

failure

1 2 12 15 <5 delamination

2 4 12 15 42 delamination



5 8 15 15 160 connector failed; chip

wasn’t damaged

After each chip was clamped in the macro-to-micro interface (as discussed below), we

performed an evaluation of bonding strength. A sudden pressure drop was observed at some

point during a linear increase of pressure, indicating that the burst pressure of the device had

been reached. All chips were visually inspected and it was noticed that in the first four chips

(Table 1) top and bottom plates were delaminated from each other. Using visualization with a

food dye, we observed that the area of the delamination was inversely proportional to the

exposure time to cyclohexane vapour. When the pressure reached 160 bar during the experiment

with the last chip (number 5), the HPLC tubing failed (due to the pressure limitation of the

PEEK tubing). A visual inspection of the chip did not reveal any delaminated or damaged areas.

Thus, the following bonding conditions were used for further experiments: 8 min of exposure

to cyclohexane vapour and 15 bar of applied pressure for 15 min at 120°C. Pressure tests to

reveal the maximum burst pressure were performed later and described in the following section.

Ch

ap

ter

V


160

Macro-to-micro interface

Even though the chip is meant to be placed in the interface between dimensions (separation

columns) in the low pressure region (10-30 bar), the valve switching can cause unpredictable

pressure pulses. It was thus decided to test the chip for the maximum burst pressure it could

withstand. During the first pressure stability test with an actual mixing device with herringbone

grooves, the pressure reached 150 bar and the flow was stopped. No visual damage or leakage

was observed for 15 minutes. In a repetition of this test, after a linear increase of pressure up to

~180 bar, the pressure dropped, because of leakage at the PEEK tubing. This experiment was

performed with three different chips with the same outcome. No delamination of the chips was

observed. Because of this it was decided to use UHPLC tubing, which can withstand much

higher pressures. When the total pressure reached 200 bar, UHPLC pumps started to pump only

with the flow rate needed to maintain the pressure in the systems. The pressure signal on both

pumps was recorded (Suppl. Fig. 4). After 4 min a pressure drop was observed. Visual

inspection of the chip revealed a crack in the COC substrate that appeared in the region of the

inlets. Thus, two connector holes milled in close proximity (appr. 6 mm) of each other create

the weakest point in the COC substrate.

There are a few other features of the interface that should be mentioned. In order to

assemble the interface, an extra screwdriver and/or wrench for the bolts is needed. Also, precise

alignment between the COC chip and metal parts is required, otherwise the two parts of the

female connector will be misaligned. The current interface does not allow experiments in which

optical data acquisition needs to be performed. This was not required for the present application.

The mixing experiments described in Supplementary Information were performed without the

metal holder, with the tubing simply inserted through the inlets of the COC chip.

Dispersion test

In order to characterize the elution profile of the micromixer was compared with those produced

by other conventional (mixing) components, as well as a zero-dead volume unit. To understand

the effect of the integration of the micromixer into the modulation interface, the Taylor-

dispersion-Analysis was performed.44 A flow rate of 0.5 mL/min was chosen to allow rapid

testing without introducing excessive Taylor dispersion and unnecessary pressure to the system.

The pressure in the system was in the range of 36-38 bar for all mixing units.

First, to obtain the value of the dead volume of the system, either a zero-dead volume

union, a T-piece, one of two S-mixers (10 µL and 150 µL) or a microfluidic mixer (M-mixer)



161

were tested (Suppl. Figure 5). Results are shown in Figure 4. Table 2 presents the quantified

data obtained from the integration of the acetone peaks in the chromatograms (n = 3).

Figure 4. Normalized concentration (UV) profiles for 1% solutions of acetone at a total flow rate of 0.5 mL/min

for the zero-dead volume union, T-piece, microfluidic mixer (M-mixer), 10-µL S-mixer and 150-µL S-mixer.

These profiles were recorded using UV detector.

As was expected, band broadening appears for all mixing units to a different extent.

Among all mixing units, the least dispersion is observed in the T-piece. This can be explained

by the simple construction of the T-piece, which represents a T-junction (with 0.28 µL dead

volume). The geometry of the M-mixer and S-mixers are more complicated, which leads to a

longer fluid path. Besides, they have larger internal volumes compared to the T-piece.

Table 2. Chromatographic data obtained from the integration of the acetone plug profile during

dispersion tests; experiments performed in triplicate; an average value is presented.

Unit Peak height,

mAU

Area under

the peak

Peak width, s Total elution

timef, s

Zero-dead volume union

(0.007 µL)45

89.3 ± 0.22 226.8 ± 2.76 5.4 ± 0.01 6.0

T-piece (0.28 µL)45 79.2 ± 0.56 226.7 ± 2.2 7.5± 0.01 9.0

M-mixer (4.65 µL) 65.6 ± 0.62 222.8 ± 5.9 9.0 ± 0.02 10.8

S-mixer (10 µL) 62.5 ± 1.18 222.7 ± 7.5 9.6 ± 0.06 12.0

S-mixer (150 µL) 11.0 ± 0.30 213.8 ± 9.5 49.8 ± 0.10 58.8

f The elution time from the beginning of acquisition till the last point of the elution of each peak profile on the

chromatogram.

Ch

ap

ter

V


162

M-mixer with the inner volume of 4.65 µL yields a concentration profile similar to the S-

mixer with the volume of 10 µL (the peak width of the S-mixer is slightly larger due to tailing).

The similarity of the M-mixer to the mixing unit with a larger volume can be explained by the

location of the transverse fluid transport caused by grooves in M-mixer. Mixing occurs both

above and within the grooves.46 While rapid chaotic mixing occurs above the grooves due to

vortex generation, mixing within grooves is governed by laminar flows and slow diffusion.

Thus, the molecules that enter a groove spend a longer time in the channel than molecules that

remain in the open channel, resulting in increased dispersion. Acetone elutes from the

micromixer over a time period of 10.8 s compared to 6 s with a zero-dead volume union.

However, the microfluidic mixer provides much lower dispersion than the 150-µL S-mixer

(dispersion peak width of ~ 50 s), which is recommended for the flow rate range that is used in

our work (100-250 µL/min).19 From this perspective, the developed microfluidic mixer

provides a very good alternative to the S-mixer in LC×LC under these flow rate conditions.

Online LC×LC separation of Polyamide 46

Polyamide 46 consists of linear and cyclic oligomers of adipic acid (ADP) and 1,4-

diaminobutane (DAB).

To demonstrate the use of the M-mixer in 2D-LC, we focused on the identification of the

four most important oligomeric series: ADP-(DAB-ADP)n-2-DAB-ADP, cyclic oligomers,

ADP-(DAB-ADP)n-2-DAB and DAB-(ADP-DAB)n-2-ADP-DAB. HILIC and RP were chosen

for the LC×LC separation in the first and the second dimensions, respectively. In order to

minimize eluent mismatch between these dimensions, a modulation interface (Figure 2)

consists of a two-position 8-port switching valve equipped with two loops, a mixing unit and

an additional pump were used. In this approach, a make-up flow is used to reduce the mobile-

phase strength of the 1D effluent on the 2D column using a microfluidic mixer with herringbone

grooves for this purpose.26

To develop the LC×LC separation, we optimized each dimension separately. Afterwards,

a HILIC×RP-LC separation of Polyamide 46 without make-up flow was performed. The results

are presented in Figure 5A. When the fraction of 1D effluent with a high content of acetonitrile

reaches the 2D column, a substantial portion of the sample peak is not able to interact with the

stationary phase and elutes with the solvent front around the column dead-time (2t0 7 s). This

is the so-called ‘‘solvent-plug peak’’ or ‘‘breakthrough peak’’ (Figure 5A).47

To trap oligomers eluting in an ACN-rich mobile phase, the effluent from the 1D (30

μL/min) was mixed with a make-up flow of water containing 0.1% FA (200 μL/min), diluting



163

the 1D eluent almost 7 times. Since the HILIC gradient in the first dimension varies from 99%

to 5% ACN, the mobile-phase composition varies between 14.8% and 0.7% of ACN after

dilution. The results, obtained using this approach, are shown in Figure 5B. No breakthrough is

observed and good separation is obtained.

Due to the 7.67-time dilution and the fact that some of the sample went to the waste in

the experiment when a make-up flow was used, the absolute amounts of injected sample on the

second column was lower that in experiment with no make-up flow (in the last case a loop was

filled only with 1D effluent).

Figure 5. Comparison of the 2D HILIC×RP-LC separation of PA 46 (A) without and (B) with make-up flow

using a microfluidic mixer in the interface between two columns. Separation conditions are reported in the “Setup

and chromatographic conditions” (Material and Method section). (C) Chemical structures of the four identified

series in the polyamide 46 samples: (1) ADP-(ADP)n-2-ADP, (2) cyclic oligomers, (3) ADP-(DAB-ADP)n-2-DAB

and (4) DAB-(DAB)n-2-DAB.

In most cases, a dilution of the sample is seen as something to be avoided, as it can cause

additional band broadening and a decrease in sensitivity. However, when used in-between

dimensions, a weaker sample solvent obtained by dilution not only improves retention of

analytes on the 2D column, but it also tends to limit band broadening due to absorbing analytes

in a narrow zone at the top of the column, a so-called on-column sample-focusing effect.4

Ch

ap

ter

V


164

Figure 5B shows the two-dimensional separation of a Polyamide 46 sample in which four

identified oligomeric series are indicated. Their chemical structures are given in Figure 5C. The

elution behavior of these series obeys simple rules related to the chain length, where the

monomer elutes first and the higher oligomers elute successively in order of increasing chain

length. During the HILIC separation, oligomers elute based on their chemical composition. On

the one hand, the retention increases with increasing sample polarity and the increase in number

of polar functional groups in the molecule usually enhances sample retention.7 On the other

hand, the interactions of basic and acidic analytes with the stationary phase are expected to be

based on both hydrophilic interactions and electrostatic forces.48 The elution sequence on the

HILIC column can be explained by the ionized state of the primary end groups. It is believed

that electrostatic interactions play an important role in HILIC, due to the partially ionized

residual silanol groups on the surface of the silica stationary phase.48 This leads to a negatively

charged surface that creates an electrostatic field in the contacting mobile phase. In the case of

the ADP-(ADP)n-ADP series, the weak retention in the first dimension can be explained by

repulsion between the silica gel surface (which is negatively charged due to the adsorbed water

molecules48) and carboxylic-acid end groups that are partially negatively charged due to the

presence of water (absorbed layer on the stationary phase). However, the main fraction of

carboxylic-acid end groups remain neutral due to the formic acid that is present in the solute

plug. For the cyclic series, which has no terminal functional groups, the slightly better retention

can be explained by other polar interactions, such as hydrogen bonding and dipole–dipole

interactions. In the oligomeric series ADP-(DAB-ADP)n-DAB, which contains both a

carboxylic-acid group and a primary amine group, the retention mechanism is a superposition

of electrostatic interactions and hydrogen bonding to the stationary phase. The last identified

series was DAB-(DAB)n-DAB, which contains only amine end groups that may be partially

protonated in the presence of formic acid. These molecules have the highest retention on the

HILIC column due to the attractive electrostatic interactions. Due to good retention, these

compounds stay the longest in the first dimension, which results in a very broad peak shape,

long tailing and a large number of modulations required to transfer one-dimensional peaks to

the second dimension.

Besides these identified series, other oligomeric series are visible in Figure 5B (not

marked) (e.g. with adipic acid/pyrrolidine, adipic acid/amide and 1,4-diaminobutane/amide end

groups). However, we focused only on the identification of the main four series discussed

above.



165

Figure 6. Extracted-ion chromatograms (EICs) of (A) ADP monomer (m/z value is 345.197), (B) tetramer ADP-

(ADP)2-ADP ((m/z value is 939.597).) and (C) DAB monomer (m/z value is 287.24).

Oligomers with carboxylic-acid end groups are weakly retained on HILIC and are eluted

in the beginning of the gradient at higher concentrations of acetonitrile. The utilization of the

microfluidic mixer for dilution of 1D effluent before the second dimension allows the first series

of oligomers (ADP-(ADP)n-ADP) to provide some focusing effect on the 2D column. Figure 6

shows extracted-ion chromatograms (EICs) for ADP monomer (Figure 6A) and tetramer ADP-

(ADP)2-ADP (Figure 6B). Each peak represents one second-dimension run (60 s) from one

collected fraction and the analyte is distributed over these chromatograms.1 As can be seen from

Figure 6A and 6B, each first-dimension peak was sampled 9-10 times, resulting in the 9-10

peaks shown. The obtained 2D chromatograms are sharp. However, a bit of tailing can be

observed, the cause of which is not completely understood, but may be related to a lack of the

complete focusing at the top of the column for the early eluting analytes. With each succeeding

oligomer in the series the tailing becomes less, to disappear completely for tetramers (Figure

6B).

Ch

ap

ter

V


166

Oligomers with primary amine end groups (DAB-(DAB)n-DAB) are strongly retained on

the HILIC column due to attractive electrostatic interactions between partially negatively

charged silica gel surface and partially protonated amino groups.48 This series exhibits the most

band broadening during the first-dimension separation and it is the most affected by the

intermediate addition of water (aqueous buffer). Figure 6C shows the EIC for the DAB

monomer. This 1D peak is very wide, has substantial tailing and takes more than 29 modulations

to be transferred to the second dimension.

Figure 7. Extracted ion chromatograms (EICs) of the ADP monomer (m/z value is 345.197) obtained with (A) T-

piece, (B) 10-µL S-mixer, (C) 150-µL S-mixer and (D) M-mixer in the interface of 2D HILIC×RP-LC system for

analysis of PA46 sample.

In order to compare the performance of the M-mixer with other mixing units, we

conducted experiments with a T-piece and two S-mixers (10 µL and 150 µL). Although the

overall performance of all mixing units was similar, several effects were observed. We noticed

a shift to longer total elution times for all mixers with respect to the T-piece. This shift was

small for the 10-µL S-mixer and larger for the M-mixer. A clear pattern for the shifted elution

times of the M-mixer starting from 2 min for ADP monomer (Fig. 7) up to 7 min for DAB

monomer (Fig. 8) was observed. The delay in elution can be caused by the change in 1D flow

rate when a mixing unit is placed directly in the fluid flow and acts as a restrictor. In the case

of S-mixers the restriction is small because the inner cartridge has a bigger diameter compared

to the long narrow channel of the M-mixer. However, a delay of 7 min causes only a 1%



167

increase in the total analysis time. Furthermore, the quality of 2D separation, when the M-mixer

is used, is not influenced by this shift.

Figure 8. Extract ion chromatograms (EICs) of the DAB monomer (m/z value is 287.24) obtained with (A) T-

piece, (B) 10-µL S-mixer, (C) 150-µL S-mixer and (D) m-mixer in the interface of 2D HILIC×RP-LC system for

analysis of PA46 sample.

We also noticed that the 150-µL S-mixer, which is recommended by the manufacturer for

the flow-rate range 100-250 μL/min,19 gave rise to split peaks for the first oligomeric series

(Supplementary Figure 6C), for which the mixing is most important. This can be caused by

incomplete mixing and the distribution of sample between different regimes having different

compositions. Additionally, the number of modulations required to transfer one 1D peak was

higher than for any other mixing unit, due to broadening of the peak. Finally, the volume of this

mixer is almost twice the volume of the loop size. In this case, resolution is lost inside the mixer

itself due to its relatively large volume.

Ch

ap

ter

V


168

Supplementary Information

Characterization of the degree of mixing

For characterization of the degree of mixing, fluorescence detection was used. Fluorescein (5

μM) in phosphate buffer (pH 7.4) was introduced from one inlet whereas 10mM phosphate

buffer was introduced from the other inlets into the Y-junction of the channel at different flow

rates using syringe pumps with 5-mL syringes (Prosens NE1000, The Netherlands). The chip

was placed on an inverted fluorescence microscope (model “DMIL”, Leica Microsystems, The

Netherlands), equipped with a 4×objective, an external light source for fluorescence (EL6000,

Leica Microsystems, The Netherlands), and a CCD camera. For visualization of fluorescence,

a fluorescein filter set (488 nm excitation, 518 nm emission) was used. Images were captured

at different positions along the channel with a CCD camera connected to a computer using a 4x

objective magnification with a field of view of 1.8 mm, a 1-sec exposure time, a gamma setting

of 1.75, and a gain of 3.5.

The mixing performance of COC-milled micromixer

After the COC chip was micromilled and assembled, its mixing performance was tested at

different flow rates. The protocol for testing the mixing efficiency was described in our previous

work.26

The degree of mixing was quantified by determining the standard deviation (SD) in

fluorescence intensity across the width of the channel at different locations along the channel

length. The detailed data analysis procedure described elsewhere.26 For the purpose of this

study, sufficient to say that high SD in fluorescence intensity across the mixing channel

correlates with incomplete mixing. The degree of mixing is inversely proportional to SD. The

high SD is measured at the very beginning of the mixing channel (at the Y-junction), where the

two solutions (5µM fluorescein solution in PBS buffer and PBS buffer) meet and enter the

channel under laminar flow conditions (no mixing).

For the purpose of mixing in the interface between two dimensions of LC×LC to adjust

mobile-phase compositions, it was import to test different flow rate ratios as well.

Supplementary Figure 1 shows results for the mixing performance experiments with different

flow rates at 1:1 ratio and with the flow rates used for this work at 1:7 ratio. As usual, the mixing

efficiency increases (standard deviation, SD drops) toward the end of the channel. It was

observed that when the ratio of inlet flows was 1:1, mixing behaviour was similar for all tested



169

flow rates. However, the mixing is not complete by the end of the 40-mm channel. The flow

rate ratio 1:7 (30 µL-min – 200 µL-min) was important to test because that was the flow rate

ratio that we wanted to use further in the application of the mixer. Slightly deteriorated situation

was observed, which can be explained by the difficulties to involve thinner fluid layer into the

general chaotic flow. Similar situation was observed in our previous research too.26

Nevertheless, the COC chip showed sufficient mixing performance for all flow rate ratios.

Supplementary figure 1. Mixing efficiency in the COC-micromilled microfluidic mixer at different flow rates

and different flow ratios (1:1 and 1:7); 5 µM fluorescein in PBS mixed with PBS; channel width 430 µm; channel

depth is 150 µm; groove depth is 100 µm; the total channel length 40 mm.

Experiments with groove rounded angles

The radius of the endmills, used for the micromilling process, do not allow for the creation of

sharp concave corners. All concave have an internal radius of curvature depending on the

endmill used. Previous experiments reported by Stroock27,49 showed that herringbone grooves

with an intersection angle θ of 90°, the angle between long and short groove arms

(Supplementary Figure 2A), results in the generation of maximum transverse flows. In order to

investigate how the geometry of the grooves with rounded concave corners affects mixing,

chips with grooves having this geometry were designed and replicated in PDMS. A detailed

fabrication procedure can be found in our previous report.26 Only one angle in the groove design

Ch

ap

ter

V


170

was not rounded (Supplementary Figure 2B). This angle is an outer angle and was expected to

stay sharp during the micromilling.

Supplementary figure 2. (A) Schematic drawing of grooves in a channel with herringbone grooves showing long

and short groove arms and groove intersection angle θ. (B) Microscope images taken from above the microchannel

with herringbone grooves that have rounded angles. Only one angle of the groove was not rounded (marked in

red).

Supplementary figure 3. Efficiency of mixing in the PDMS microfluidic mixer with rounded angles herringbone

grooves at different flow rates at 1:1; 5 µM fluorescein in 10mM PBS (pH 7.4) mixed with PBS; channel width

430 µm; channel depth is 150 µm; groove depth is 100 µm; n=3 chips, the total channel length 45 mm.

Supplementary Figure 3 shows that SD levels, which implies that the mixing is

complete, after 20 mm of the channel length at all flow rates. Here, an increased optimized

dimensions comparing to our previous work26 were used but obtained results were similar.

Thus, we decided to proceed with the fabrication by micromilling of a microfluidic mixer based

on this design (dimensions, for eventual use under high pressure).



171

Pressure test

Supplementary Figure 4 shows the results for the second pressure test that was performed with

a COC chip containing a grooved channel (dimensions are given in Fig.3). The chip withstood

200 bar for 4 min.

Supplementary figure 4. Pressure test with microfluidic COC chip: (A) position of the chip between two UHPLC

pumps as also given Fig.2 both schematically and in a photo and (B) pressure profile for each pump (red and blue

signals). After 4 min at 200 bar, a pressure drop was observed due to the crack at the inlets.

Tests with different mixing units

In order to investigate the influence of the mixing in the interface between two dimensions,

microfluidic mixer (m-mixer), 10-µL S-mixer, 150-µL S-mixer and T-piece were tested..

Supplementary Figure 5 presents the different mixing components that were tested to obtain the

value of dispersion of the system.

Supplementary figure 5. Units that were used for dispersion test: (1) 4.65-µL microfluidic mixer (M-mixer), (2)

150-µL S-mixer, (3) 10-µL S-mixer and (4) T-piece and (5) a Zero-Dead Volume Union.

Ch

ap

ter

V


172

Conclusions

In this work, the possibility to use microfluidic technology to improve LC×LC separations has

been successfully demonstrated. We fabricated a microfluidic mixer with herringbone grooves

that was implemented in a HILIC×RPLC system for successful separation and identification of

various oligomeric series in Polyamide 46 samples. A microfluidic chip was fabricated in rigid

COC substrate, which has excellent properties for chip-based LC technologies. Thanks to the

solvent-vapour-assisted bonding approach, a strong bond between two COC parts was obtained.

The resuting chips, when clamped in a metal holder, are able to withstand pressures up to 200

bar.

In order to create a connection with an LC×LC system, a very robust, low-dead-volume

interface was developed. Using standard HPLC connectors, the chip can be directly coupled to

any LC equipment. This opens a perspective for the future implementation of the microfluidic

components in conventional instrumentation for improved performance, even at high pressures.

The microfluidic mixer successfully coped with the mobile-phase mismatch between two

dimensions, diluting the 1D effluent before transferring it to the second dimension. As was

shown, good mixing prevents breakthrough and gives good peak shapes in the second

dimension. Due to the small size and ease of design adjustment, the developed microfluidic

micromixer can be integrated with trap columns on one chip in order to provide an active

modulation in the LC×LC interface. This can be an attractive solution to improve LC×LC

modulation even further, applying modulators with smaller inner volumes with the capability

to adjust mobile-phase compositions much more rapidly and efficiently.



173

Acknowledgements



(HYPERformance LC).

The authors would like to thank Gert Salentijn for helpful discussions and his careful reading

of the manuscript.

Ch

ap

ter

V


174

References

1. Jandera, P. Comprehensive two-dimensional liquid chromatography — practical impacts of theoretical

considerations. A review. Cent. Eur. J. Chem. 10, 844–875 (2012).



3. Schoenmakers, P., Marriott, P. & Beens, J. Nomenclature and conventions in comprehensive

multidimensional chromatography. LCGC Eur. 25, 1–4 (2003).

4. Jandera, P., Hájek, T. & Česla, P. Effects of the gradient profile, sample volume and solvent on the

separation in very fast gradients, with special attention to the second-dimension gradient in comprehensive

two-dimensional liquid chromatography. J. Chromatogr. A 1218, 1995–2006 (2011).

5. Jandera, P. & Hájek, T. Utilization of dual retention mechanism on columns with bonded PEG and diol

stationary phases for adjusting the separation selectivity of phenolic and flavone natural antioxidants. J.

Sep. Sci. 32, 3603–3619 (2009).

6. Dugo, P., Favoino, O., Luppino, R., Dugo, G. & Mondello, L. Comprehensive Two-Dimensional Normal-

Phase (Adsorption)-Reversed-Phase Liquid Chromatography. Anal. Chem. 76, 2525–2530 (2004).

7. Jandera, P., Hájek, T., Staňková, M., Vyňuchalová, K. & Česla, P. Optimization of comprehensive two-

dimensional gradient chromatography coupling in-line hydrophilic interaction and reversed phase liquid




9. Cacciola, F., Jandera, P., Blahová, E. & Mondello, L. Development of different comprehensive two

dimensional systems for the separation of phenolic antioxidants. J. Sep. Sci. 29, 2500–2513 (2006).

10. Cao, L. et al. The development of an evaluation method for capture columns used in two-dimensional

liquid chromatography. Anal. Chim. Acta 706, 184–190 (2011).

11. Tian, H.-Z., Xu, J. & Guan, Y.-F. Vacuum-evaporation Interface of Comprehensive Two-dimensional

Liquid Chromatography and Its Application. Chinese Journal of Analytical Chemistry 36, 860–864 (2008).

12. Verstraeten, M., Pursch, M., Eckerle, P., Luong, J. & Desmet, G. Thermal modulation for

multidimensional liquid chromatography separations using low-thermal-mass liquid chromatography

(LC). Anal. Chem. 83, 7053–7060 (2011).

13. Van de Ven, H. C., Gargano, A. F. G., Van der Wal, S. J. & Schoenmakers, P. J. Switching solvent and

enhancing analyte concentrations in small effluent fractions using in-column focusing. J. Chromatogr. A

1427, 90–95 (2016).

14. Vonk, R. J. et al. Comprehensive two-dimensional liquid chromatography with stationary-phase-assisted

modulation coupled to high-resolution mass spectrometry applied to proteome analysis of saccharomyces

cerevisiae. Anal. Chem. 87, 5387–5394 (2015).

15. Murahashi, T. Comprehensive two-dimensional high-performance liquid chromatography for the

separation of polycyclic aromatic hydrocarbons. Analyst 128, 611 (2003).



88, 1785–1793 (2016).


68 (2011).

18. Singh, M. K., Anderson, P. D. & Meijer, H. E. H. Understanding and optimizing the SMX static mixer.

Macromol. Rapid Commun. 30, 362–376 (2009).

19. Brochure provided by Analytical Scientific Instruments US, http://www.hplc-asi.com/static-mixers/.

(2014).

20. Lin, S.-L., Bai, H.-Y., Lin, T.-Y. & Fuh, M.-R. Microfluidic chip-based liquid chromatography coupled

to mass spectrometry for determination of small molecules in bioanalytical applications: An update.

Electrophoresis 35, 1275–1284 (2014).

21. Dietze, C., Hackl, C., Gerhardt, R., Seim, S. & Belder, D. Chip-based electrochromatography coupled to

ESI-MS detection. Electrophoresis 37, 1345–1352 (2016).

22. Yin, H. et al. Microfluidic chip for peptide analysis with an integrated HPLC column, sample enrichment

column, and nanoelectrospray tip. Anal. Chem. 77, 527–533 (2005).

23. Sung, W. C., Makamba, H. & Chen, S. H. Chip-based microfluidic devices coupled with electrospray

ionization-mass spectrometry. Electrophoresis 26, 1783–1791 (2005).

24. Ro, K. W., Liu, J. & Knapp, D. R. Plastic microchip liquid chromatography-matrix-assisted laser

desorption/ionization mass spectrometry using monolithic columns. J. Chromatogr. A 1111, 40–47 (2006).



175

25. Bai, H.-Y., Lin, S.-L., Chan, S.-A. & Fuh, M.-R. Characterization and evaluation of two-dimensional

microfluidic chip-HPLC coupled to tandem mass spectrometry for quantitative analysis of 7-

aminoflunitrazepam in human urine. Analyst 135, 2737 (2010).

26. Ianovska, M. A., Mulder, P. P. M. F. A. & Verpoorte, E. Development of small-volume, microfluidic

chaotic mixers for future application in two-dimensional liquid chromatography. RSC Adv. 7, 9090–9099

(2017).


28. Jena, R. K., Yue, C. Y. & Lam, Y. C. Micro fabrication of cyclic olefin copolymer (COC) based

microfluidic devices. Microsyst. Technol. 18, 159–166 (2012).

29. TOPAS Advanced Polymers. Brochure provided by Polyplastics, Topas COC: Transparent copolymer

with excellent optical properties. (2015).



31. Mair, D. A., Geiger, E., Pisano, A. P., Fréchet, J. M. J. & Svec, F. Injection molded microfluidic chips

featuring integrated interconnects. Lab Chip 6, 1346–1354 (2006).

32. Mair, D. A. et al. Room-temperature bonding for plastic high-pressure microfluidic chips. Anal. Chem.

79, 5097–5102 (2007).

33. Wouters, B. et al. Design of a microfluidic device for comprehensive spatial two-dimensional liquid

chromatography. J.Sep.Sci. 38, 11123–1129 (2015).

34. Chen, C. F. et al. High-pressure needle interface for thermoplastic microfluidics. Lab Chip 9, 50–55

(2009).

35. IDEX & IDEX Health and Science. Catalogue IDEX. (2014).

36. Fredrickson, C. K. & Fan, Z. H. Macro-to-micro interfaces for microfluidic devices. Lab Chip 4, 526

(2004).

37. https://www.vici.com/vfit/vfit.php. No Title. https://www.vici.com/vfit/vfit.php Available at:

https://www.vici.com/vfit/vfit.php.

38. Bayliss, N. S. & McRae, E. G. Solvent effects in the spectra of acetone, crotonaldehyde, nitromethane and

nitrobenzene. J. Phys. Chem. 58, 1006–1011 (1954).

39. Schneider, F., Draheim, J., Kamberger, R. & Wallrabe, U. Process and material properties of

polydimethylsiloxane (PDMS) for Optical MEMS. Sensors Actuators, A Phys. 151, 95–99 (2009).




7, 580–587 (2007).

42. Tsao, C. W. & DeVoe, D. L. Bonding of thermoplastic polymer microfluidics. Microfluid. Nanofluidics

6, 1–16 (2009).

43. Barton, A. F. M. CRC Handbook of solubility parameters and other cohesion parameters. (CRC Press,

1991).

44. Understanding Taylor Dispersion Analysis. (2015).

45. Swagelok.com. Gaugeable Chromatograph and Column End Fittings. www.swagelok.com Available at:

https://www.swagelok.com/downloads/webcatalogs/EN/MS-02-173.PDF.

46. Du, Y., Zhang, Z., Yim, C., Lin, M. & Cao, X. A simplified design of the staggered herringbone

micromixer for practical applications. Biomicrofluidics 4, 1–13 (2010).

47. Jiang, X., Van Der Horst, A. & Schoenmakers, P. J. Breakthrough of polymers in interactive liquid


48. Buszewski, B. & Noga, S. Hydrophilic interaction liquid chromatography (HILIC)-a powerful separation

technique. Anal. Bioanal. Chem. 402, 231–247 (2012).

49. Stroock, A. D. & Whitesides, G. M. Controlling flows in microchannels with patterned surface charge and

topography. Acc. Chem. Res. 36, 597–604 (2003).

Ch

ap

ter

V

Chapter VI

General discussion,

conclusions and future

perspectives

General discussion, conclusions and future perspectives

179


In this thesis, we have demonstrated the development and application of the chaotic

microfluidic mixer for improving conventional analytical technique such as two-dimensional

liquid chromatography (2D-LC). Without doubts, the research described here provides

intriguing possibilities for the application of the microfluidic technology. While many

miniaturized lab-on-a-chip systems trying to replace large laboratory equipment (which in

many cases is not achievable task), this project was focusing on using advantages of the

miniaturized devices (such as small volume) to be connected on-line with conventional

equipment. The research showed in this thesis arose as a need to find more convenient

approaches to improve the performance of the 2D-LC technique and to bring it to a different

level, where analysis of tens thousand and more components in one run become possible.

The research, presented in this thesis, covers many different facets of modern technology.

This work has been done in the interface between two rapidly developing fields –

multidimensional LC and microfluidics. On one hand we talk about huge complicated

equipment, composed of 2 chromatographic systems coupled through the intricate interface. On

the other hand, we have a small 5-cm long chip, with, nevertheless, a potential to solve one of

the major problems that exist in 2D-LC separation. The problem that we are talking about here

is a solvent incompatibility between dimensions and we propose a solution to it such as the

utilization of the mixing device in the interface.

As was discussed in Chapter 1 and Chapter 3, the mixing device in the interface between

two columns in 2D-LC must satisfy three strict conditions: it should provide fast mixing in-line

at different ratios in the wide range of flow rates (compatible with typical flow rates used in

2D-LC); have a small volume (to not contribute to the extra column-band broadening); and

must be able to withstand brief pressure pulses up to a few hundred bar (due to its connection

to switching valves which operated under the pressure from pumps). Therefore, an idea to

develop and use small-volume microfluidic mixers that are able rapidly mix solutions at

different flow rates is justified.

It is important to keep in mind, that microfluidic devices, on the other hand, has a very

important and disadvantageous in our situation feature such as an existence of a laminar flow,

which does not allow two streams to be mixed under normal consequences. Therefore, our first

research goal was to develop a fast micromixer that works under the wide range of flow rates Dis

cuss

ion


180

and able to mix solvents with different viscosity (which is often a case in LC). Based on the

overview we presented in Chapter 2, it was clear that the mixer with herringbone grooves

meets the required conditions. In Chapter 3 we proved this concept. Besides, we believe that

the wide overview of the existing micromixers based on chaotic advection and the approach for

choosing the appropriate type for a particular application presented in Chapter 2, combined

with the fabrication protocol described in Chapter 3 provide a good base for the end-users who

are interested in using microfluidic mixers in their own research.

As Robert M. Pirsig said, “Technology presumes there is just one right way to do things

and there never is”. Seeing the presented research from this perspective, the fabricated devices

were designed with the idea to provide competition for the existing mixers – T-piece and S-

mixers1 – that are currently used for mixing in the conventional LC systems. The developed by

us micromixer possess better characteristics in terms of volume and efficiency of mixing under

lower flow rates. As we showed later in Chapter 5, the developed COC-micromixer (with the

volume of 4.65 µL, M-mixer) shows similar dispersion profile to T-piece and the S-mixer with

the volume to 10 µL. M-mixer is clearly outperforms S-mixer with 150 µL volume in both

dispersion profile and the real application. However, in the real separation T-piece and S-mixer

give similar performance while M-mixer give some time-shift and small tailing. Even if the

mixer will show better performance than commercial mixers at lower flow rate (for which it

was designed), the overall dispersion due to the relatively large volume of the system (loops

and tubing) will vanish the benefit of having a small volume of the M-mixer. We suggest that

the benefit of the developed micromixer would be seen to the full extend if other component

have also smaller volume and integrated on the same chip. In other words, if the modulator and

a second-dimension column are also integrated on the same micromachined device. This will

dramatically decreased not only the dispersion of the system (due to the absence of extra tubing

for connecting different components), but also reduce the time of the second-dimension

separation. Figure 1 represents the situation when a column packed with a stationary phase

(particles or monoliths) is used instead of loops in the modulator. In such system, the effluent

coming from the first dimension will be mixed with the modifying solvent in order to trap

analytes on the pre-concentration column, followed by elution of the analytes after switching

the solvent with higher eluting strength. In this case, this will create so-called trap column that

provide an extra pre-concentration of analytes before entering the second column and it would

provide an active modulation in the LC×LC interface. Afterwards, the analytes will proceed to


181

the second-dimension where the effluent will be mixed with the 2D mobile phase, after which

the second-dimension separation will take place in the channel packed with a stationary phase.

We believe that developing such state-of-the-art device one should be concerned about several

aspects. The uniform packing of the microfluidic channel with particles can be a challenging

task, it require utilization of special frits and high pressure for packaging. The fabrication of

frits is problematic within a microfluidic channel and formation of bubbles and band broadening

is frequently observed. In case of monoliths, during the polymerization process the gaps

between channel walls and the monolithic phase can form. Besides, the monoliths can dislodge

under applied pressure. In both cases, the utilization of the channel with stationary phase will

create additional backpressure that should be taking into account designing the whole separation

system.

Figure 1. A micromachined device with integrated modulator with a mixer and pre-concentration column and

mixer for modifying the mobile phase followed by the second-dimension separation column. Additional pump

assists in the absorption/desorption process from the pre-concentration column.

Another example of the device that our mixer could compete with is a commercially

available Jet Weaver mixer,2 developed by Agilent and incorporated into the HPLC pumping

system (1290 Infinity Binary pump). That is a perfect example of the chip-based microfluidic

device that have already successfully entered the world of “macro-equipment”. As discussed in

Chapter 3, this mixer employs the split-and-recombine mixing principle using fro this a network

of multi-layer microfluidic channels. The minimum volume of this device is 35 µL. In our

approach we use different mixing mechanism – chaotic advection, which was proved to be more

efficient mixing mechanism that allows to obtain mixing using devices with smaller volume, as

discussed in Chapter 2. Besides, our mixer can be separately connected to any LC or MS

equipment, while Jet Weaver mixer is not applicable outside LC pumping system.

Dis

cuss

ion


182

Another important requirement that was set for the mixing device is its pressure-

resistance. The originally reported design by Stroock et al.3 had a very small inner

volume (around 1 µL). In order to maintain pressure drops below 1 bar, we designed

micromixers with increased dimensions (obtaining the inner volumes of 1.6 μL and 2.2

μL). However, the dimensions used by Stroock et al.3 can not simply be multiplied by a

constant to achieve mixers with bigger volumes exhibiting the same mixing efficiency.

Therefore, a lot of work was put into optimization of channel and groove parameters

based on a previously described numerical study.4

As was shown in Chapter 3, our initial devices were fabricated in PDMF using soft-

lithography. While it was sufficient for the prove of concept, these devices didn’t withstand

more than 10 bar of pressure. Starting to search a new fabrication method in order to obtain

pressure-resistant chips revealed that the initial chip features around 50 µm is on the edge of the

resolution of all accessible to us fabrication techniques. Therefore, in order to fabricate a

pressure-resistant chip, we had to modify and optimize the geometrical features inside the

channel again. Using the same approach described in Chapter 3, we obtained chips with

increased dimensions and the minimum features of 100 µm. Thought it is slightly increased the

inner volume of the device from 1.6 to 3.6 µL (Chapter 4), it allowed us to consider several

fabrication methods.

In Chapter 4 and Chapter 5 we were exploring two relatively new techniques, such as

Selective Laser-Induced Etching (SLE) and micromilling. For the work descried in Chapter 4,

a solid block of fused silica was used to create a chip using the laser modification of the fused

silica with the following etching of modified material. Because the whole procedure was

performed in the solid piece of fused silica, the bounding step (bonding of two parts of the chip)

was eliminated. We aimed to fabricate a 30-mm-long microfluidic mixing channel with

herringbone structures. Unfortunately, due to such channel length, the etching time required for

removing the modified material was also too long and we didn’t manage to obtain the equal

cross-section of the channel. Besides, the herringbone grooves were observed only in the middle

part of the channel, which was not sufficient to obtain a good working mixing device. However,

in terms of pressure-resistance, the obtained fused-silica chips can withstand pressure of 85 bar,

which make them applicable in the interface of multidimensional liquid chromatography as a

separate mixing device.


183

We tried to solve these fabrication issues accounting for difference in etching rate in

different device regions (more etching from the channel side and less in the middle). This

brought us to the idea of an adjusted “gradient” design with different channel width/depth and

groove dimensions at the beginning/end and in the middle of the channel (depending on the

time that a particular region stays in contact with the aching solution). The next chip generation

with such adjusted design showed more equal cross section and much larger regions with good-

resolved herringbone grooves, which lead to a better degree of mixing. Unfortunately, a

complete mixing by the end of the channel was still not obtained. In the future, it would be

interesting to optimize design further and obtain a chip with good-resolved grooves along the

whole channel.

Nevertheless, to the best of our knowledge, the fabrication of the fused silica chips with

such length with the complex features inside using the SLE technique has never been explored

before. This makes our work a pioneering in this area of research and makes an interesting topic

for the future.

Due to the failure of the fabrication approach exploited in Chapter 4. we have to search

for another option for the fabrication of pressure-resistant chip. We chose micromilling

technique as a compromise between speed of transferring the sketch into the actual device (it is

a direct fabrication method), its accessibility and appropriate resolution (around 100 µm). The

choice of material was also an important aspect. We decided to use cyclic-olefin copolymer

(COC), due to it is rigidity and compatibility with solvents used in LC applications (acetonitrile,

methanol, etc.). The fabrication method, as described in Chapter 5, consisted of several steps,

which, due to the limited time, were strategically planned. The first part was milling in COC

substrate. It required 5 hour in total to mill one bottom part of the chip with a channel and

herringbone structures. The top part was cut from the sicker piece of COC with the drilled

conical access holes. The next step was the bonding of two parts, which was done using the

solvent-vapour-assisted bonding approach that was used in the group of Eeltink.5 Due to the

mechanical interlocking of polymer chains between two surfaces under the cyclohexane vapor,

exceptionally strong bond between two parts was achieved.6 Without doubts, this approach

proved to be very useful in our application, but can be easily applied for the fabrication of any

pressure-resistance devices. The last step of the fabrication process was development of a

specially designed holder to assist in the pressure-resistance of the device and to connect the

chip to the equipment. The holder consists of two metal parts and the chip was clamped between

Dis

cuss

ion


184

them. The top metal part had an access wholes that had a standard 10-32 thread for standard

HPLC connectors. The final assembly was able to withstand pressures up to 200 bar.

Due to the standard 10-32 thread in the metal top part and the conical parts in the top

COC plate, the holder also served as a low-dead-volume micro-to-macro interface to directly

coupled the chip to 2D-LC system. This approach opens an attractive perspective for the future

implementation of any microfluidic component into conventional instrumentation, even at high

pressures.

The developed COC chip was implemented in a HILIC×RPLC system for successful

separation and identification of various oligomeric series in Polyamide 46 samples. As was

shown in Chapter 5, the microfluidic mixer successfully copes with the mobile-phase

mismatch between two dimensions, mixing the effluent after the first column before

transferring it to the second dimension. The proof of good mixing can be seen as the prevention

of the break-through in the second dimension. We want to emphasize that even though the

developed mixer did not show better performance than T-piece and 10-uL S-mixer, there is no

need to underestimate the developed mixer. Having a very small volume, it proved to be fast,

efficient and robust during a very large number of modulations. Unfortunately, the time of this

project was limited and there was no opportunity to perform more detailed research and to

compare thoroughly the behavior of different mixing units. It would be also interesting to

investigate performance of the microfluidic mixer in the separation of different samples and

therefore, at different flow rates.

In conclusion, the results of this thesis show a very good potential of combining the

microfluidic technology and conventional analytical techniques. In our case, we aimed to

improve the performance of multidimensional liquid chromatography using a microfluidic

mixer. In this work, we provided a comprehensive methodology for choosing an appropriate

micromixer and design adjustments for the need of particular application, its fabrication using

several techniques, its integration into the conventional LC equipment and its successful

application even at high pressures. We believe that the research described in this thesis gives a

good basis for scientists, who has an interest in both microfluidics and multidimensional liquid

chromatography for their own research.


185

References

1. Brochure provided by Analytical Scientific Instruments US, http://www.hplc-asi.com/static-mixers.

(2014).

2. Agilent 1290 Infinity LC System, Manual and Quick Reference, Agilent Technologies. (2012).



580–587 (2007).

5. Wouters, B. et al. Design of a microfluidic device for comprehensive spatial two-dimensional liquid

chromatography. J.Sep.Sci. 38, 11123–1129 (2015).

6. Tsao, C. W. & DeVoe, D. L. Bonding of thermoplastic polymer microfluidics. Microfluid. Nanofluidics 6,

1–16 (2009).

Dis

cuss

ion

Samenvatting

Samenvatting

189

Dit proefschrift is gericht op het uitbreiden van de toepassing van microfluïdische technologie

voor het verbeteren van conventionele analytische scheidingstechnieken, zoals two-dimensional

liquid chromatography (2D-LC), die gebruikt wordt voor het analyseren van complexe

monsters. Het meest voorkomende probleem van de mobiele fase incompatibiliteit tussen

kolommen (dimensies) in 2D-LC vereist de toepassing van een mixing component in de

interface om mobiele fasesamenstellingen aan te passen. Deze mixer moet zorgen voor snelle

menging in lijn bij verschillende flowsnelheidsverhoudingen voor een breed bereik van

stroomsnelheden, het moet een klein volume hebben en in staat zijn om korte drukpulsen tot een

paar honderd bar te weerstaan, vanwege zijn verbinding met gebruikte schakelkleppen om

monster tussen kolommen over te brengen.

Het potentieel van de lab-on-a-chip apparaten voor toepassingen in zowel industriële als

wetenschappelijke gebieden heeft de ontwikkeling van microfluïdische technologie in een hoog

tempo aangestuurd. Het belangrijkste monsterbewerkingsproces in klinische diagnostiek,

genetische sequentiebepaling, chemische productie en proteomica blijft menging, daarom is de

implementatie van effectieve menging op de microschaal één van de belangrijkste aspecten van

veel microfluïdische systemen. In Hoofdstuk 2 geven we een uitgebreid overzicht van de

literatuur van het afgelopen decennium waarin reeds bestaande micromixers worden beschreven

op basis van chaotische advectie en hun combinatie met andere mengprincipes, bijvoorbeeld

splitsen en recombinatie. Bij het onderzoeken en vergelijken van deze micromixers leggen we

de nadruk op kanaalgeometrie, stromingscondities en het mechanisme van mengen. We

beschrijven ook de meest voorkomende toepassingsgebieden van passieve chaotische

micromixers aan de hand van echte voorbeelden, en bespreken de verbinding tussen

kanaalgeometrie, mengmechanisme en mogelijk toepassingsgebied onder verschillende

stroomomstandigheden.

In Hoofdstuk 3 beschrijven we de ontwikkeling van microfluïdische mixers met een

klein volume in poly(dimethylsiloxaan) (PDMS) die groeven op de boven- of onderkanaalwand

bevatten om mengen te induceren op basis van chaotische advectie. Deze benadering werd voor

het eerst beschreven door Stroock et al.1 in 2002. Om echter binnenvolumina in de orde van één

microliter te hebben, geschikt voor 2D-LC, verhoogden we eerder gerapporteerde dimensies op

basis van numerieke studies voor geoptimaliseerde kanaal- en groefgeometrieën. Groeven

werden geplaatst in matrices, hetzij in een schuine hoek ten opzichte van de wand (schuine

groeven, SG), hetzij in de vorm van asymmetrische chevrons, ook bekend als herringbones, in

verspringende reeksen (visgraatgroeven, HG). In onze studies bleken visgraatgroeven efficiënter

Samenvatting

190

te zijn voor het verbeteren van het mengen, wat consistent is met waarnemingen uit andere

onderzoeken.2,3 We hebben met succes de prestaties van HG-micromixers aangetoond voor het

mengen van vloeistoffen met verschillende viscositeiten (acetonitril, methanol en

wateroplossingen) in verschillende verhoudingen (1: 2, 1: 5 en 1:10). De ontwikkelde

micromixer maakt volledige menging mogelijk binnen een afstand van 3 cm binnen het 5 cm

lange microkanaal over een breed bereik van stroomsnelheden (4-1000 μL/min).

De succesvolle toepassing van de ontwikkelde micromixer in de interface tussen twee

dimensies in 2D-LC hangt sterk af van de drukweerstand (om bestand te zijn tegen korte

drukpulsen wanneer bijvoorbeeld multiport valves worden geschakeld). Er werden

verschillende benaderingen genomen om een chip te fabriceren die bestand is tegen drukken tot

200 bar. Het onderzoek in Hoofdstuk 4 beschrijft de fabricage van een gegroefde

microfluïdische mixer in een blok van gesmolten silica met behulp van Selective Laser-Induced

Etching (SLE).4–6 Deze techniek bestaat uit twee stappen: 1) de blootstelling van glas aan

scanning gefocuste ultrakorte (fs of ps) gepulseerde laserstraling, die plaatselijk

glaseigenschappen in het focale volume verandert om zelf-uitgerichte nanoscheurtjes loodrecht

op de laserpolarisatierichting te creëren;7 2) etsen van de met een laser gemodificeerde zone

door HF of een alkalische oplossing zoals KOH in water.5,7 Deze aanpak stelde ons in staat om

een chip te verkrijgen met complexe groefstructuren in een massief stuk materiaal, waardoor de

gebruikelijke verbindingsstap bij de fabricage van microfluïdische apparaten werd

geëlimineerd. We rapporteren een mengchip met visgraatstructuren met een kanaallengte tot 33

mm, die voor de eerste keer gefabriceerd zijn met SLE. We zijn erin geslaagd om drie generaties

chips met verschillende dimensies te fabriceren. Onze resultaten toonden aan dat het vergroten

van de kanaaldiameter het mogelijk maakte om kanalen met een betere resolutie met groeven te

verkrijgen vanwege de gemakkelijkere toegang van de etsoplossing tot het met een laser

behandelde gesmolten siliciumdioxide. Het lukte echter niet om een gegroefd kanaal met

uniforme doorsnede te maken. Het kanaal had nog steeds een conische vorm naar het midden

van het kanaal toe, wat te wijten is aan de aard van het etsproces. Gebieden aan het einde van

het mengkanaal worden langere tijd blootgesteld aan etsmiddel dan gebieden in de richting van

het midden van het kanaal, omdat het etsmiddel vanaf de uiteinden in het kanaal moet werken.

Om dit effect te ondervangen, hebben we een aangepast ontwerp voorgesteld dat een overmatige

ets aan het begin en aan het einde van het kanaal compenseert. Deze aanpak resulteerde in een

iets betere groefresolutie in het kanaal en verschafte verbeterde mengprestaties. De drukproeven

toonden aan dat deze chips met gesmolten siliciumdioxide bestand zijn tegen drukken tot 85 bar.

Samenvatting

191

Bij hogere drukken breken ofwel de slangconnectoren naar het apparaat of breekt het apparaat

zelf bij de schroefdraad waarin de schroefconnectoren zich bevinden.

In Hoofdstuk 5 beschrijven we de fabricage van de microfluïdische mixer in cyclisch

olefinecopolymeer (COC)8 met behulp van micromilling9 als de andere benadering voor het

verkrijgen van een drukbestendige chip. Een in COC gefreesde chip werd geplaatst in een

speciaal ontworpen, robuuste metalen houder met een klein dood volume die een directe

verbinding mogelijk maakte met chromatografische instrumentatie met behulp van

gestandaardiseerde HPLC-connectoren. Dit ontwerp is bestand tegen drukpulsen tot 150 bar.

Een microfluïdische mixer werd geïmplementeerd in een 2D HILIC × RP-LC-systeem voor

analyse van nylonmonsters. Het probleem van de incompatibiliteit van de mobiele fase tussen

de dimensies werd aangepakt door de snelle menging in lijn van het effluent van de eerste

dimensie met een zwakker oplosmiddel in de micromixer voordat het de tweede kolom bereikte.

Wanneer onze mixer werd gebruikt om de make-up flow op te nemen, werd een verbeterde

scheiding (zonder doorbraak en goede piekvormen in de tweede dimensie) verkregen in

vergelijking met het systeem zonder make-up flow. We hebben met succes verschillende

oligomere series in nylonmonsters (Polyamide 46) geïdentificeerd.

Slotopmerkingen

De resultaten in dit proefschrift demonstreren het potentieel van microfluïdische apparaten als

componenten om conventionele "macroscopische" apparatuurprestaties te verbeteren. Ons

onderzoek toont de mogelijkheid om kleine microfluïdische apparaten te fabriceren die bestand

zijn tegen relatief hoge drukken, wat hun toepasbaarheid vergroot. Als toekomstig perspectief

en de volgende stap in het verbeteren van de prestaties van 2D-LC, zou de ontwikkelde

microfluïdische micromixer kunnen worden geïntegreerd met trapkolommen op één chip om

extra analyt pre-concentratie te bieden voordat de tweede kolom wordt betreden. Dit zou een

aantrekkelijke oplossing zijn om de modulatie in 2D-LC nog verder te verbeteren door apparaten

toe te passen met kleinere binnenvolumes met zowel meng- als pre-concentratiefuncties.

Samenvatting

192

Referenties



580–587 (2007).

3. Aubin, J., Fletcher, D. F., Bertrand, J. & Xuereb, C. Characterization of the mixing quality in micromixers.

Chem. Eng. Technol. 26, 1262–1270 (2003).




quartz glass components for microfluidic applications-up-scaling of complexity and speed. Micromachines

8, (2017).



7. Osellame, R., Hoekstra, H. J. W. M., Cerullo, G. & Pollnau, M. Femtosecond laser microstructuring: An

enabling tool for optofluidic lab-on-chips. Laser Photonics Rev. 5, 442–463 (2011).

8. TOPAS Advanced Polymers. Brochure provided by Polyplastics, Topas COC: Transparent copolymer with

excellent optical properties. (2015).



Acknowledgements

Acknowledgements

195

There are lot of people that made an impact on my life during PhD years. Firstly, I would

like to express my sincere gratitude to my Prof. Sabeth Verpoorte who accepted me for the

position of PhD student in her Pharmaceutical Analysis group. Sabeth, you believed in me even

when I was too critical to myself. You provided me with enormous freedom, which was the

main condition to personalize my research and apply my creativity. I am grateful for your

understanding and at the same time critical mindset, for your wisdom and all your

encouragement!

I am also grateful to the following university staff: Patty Mulder and J.P. for their

unfailing support and assistance during my PhD research. Patty, thank you for helping me in

learning everything in the lab. Your input in my PhD is undeniably tremendous. J.P. thank you

for trying so hard to help me with my research and many valuable lessons that you gave me for

life. The hard times, that you have put me through, helped me to grow.

____________

I would like to thank my fellow doctoral students for their feedback, cooperation and of

course friendship. Maureen, Nadiah, Sergio, Hanan and Pim, thank you for our amazing time

together. You all are, first of all, my friends and only than colleagues. Your presence in the

group made me feel that I want to come to the lunchroom and be a part of your joyful company.

I always felt your support and eagerness to help in both research and in my daily life. I hope we

can meet time to time and share cheerful updates about our lives.

Maureen, thank you for being both my colleague and my best friend here, far away

from my home country. I think we are so similar and because of this we are able to understand

each other perfectly. I hope we will support each other in the future and our friendship will

grow. Maybe one day we become partners in business, who knows :-) Hanan, I am very happy

that we became closer this year. I hope we will never loose this! And remember, you will break

any walls on your way, but nobody will break you, I am sure! Sergio, I wish you enormous

achievements in both professional and personal life (don’t forget to invite me for the wedding).

Nadiah, I am happy to met you and I hope you will have a great scientific career! Pim, I am

very happy that my research helped yours and I hope one day we will have another bbq on your

rooftop!

Thank you, Gert and Pieter, for sharing our lovely office and helping me so much,

especially in the beginning. Gert, big thanks for reading and correcting a big part of this thesis.

Maciej S., thank you for helping with flow simulations. Maciej G., thank you for showing me

around in my first evening in Groningen, De Toeter is still one of my favorite places.

____________

Acknowledgements

196

I would like to acknowledge The Netherlands Organization for Scientific Research

(NWO) for providing the funding for this work. I also would like to express my thankfulness

to Ynze Mengerink for arranging the internships at DSM Resolve and Erik Ritzen for helping

me in the lab during that time. This secondment gave me an opportunity to see better the

company life and to obtain good results that I included in this thesis.

In addition, I would like to express my gratitude to my colleagues from the

HYPERformance projec: Peter, Henrik, Anna and Andrea for your support and help.

Henrik, separate thanks for organizing an amazing trip to Iceland, and of course for your help

with some HPLC tests! With a special mention to Arto Heiskanen and Jenny Emnéus from

DTU Nanotech. It was fantastic to have an opportunity to work in your facilities. My pressure-

resistant chip would not be able to be without our collaboration on such a short notice. A very

special gratitude goes out to Bert Wouters, who helped me in the crucial step of my PhD,

namely to assemble a chip. Without your help, I would not make it so fast.

_________

My lovely artistic friend Viktoria, from the moment you entered my office I recognized

a real Ukrainian soul in you. I am so happy than we have met and now sharing so many things

together! I hope our friendship will grow throughout the years and we will have more art

exhibitions together, more stories to tell and more great valuable time.

Nashwa, my colleague and my friends, that is amazing that we finally met (not at the

faculty, not at the salsa parties but in the middle of nowhere, in Ter Apel). You was giving me

strength and positive vibe to continue balancing between work, PhD thesis and Dutch. Thank

you so much for this! And thank you becoming my paranimph, we are a very good team! I am

sure the connection we have will stay our whole lives!

____________

To my scientific family

Спасибо моей семье за то, что всегда направляли меня в нужном направлении.

Де, всегда, когда я тебя слушала, все получалось как надо и имело огромный вклад в

будущее.

Ба, ты всегда так переживаешь за меня, и я никогда не хочу тебя расстраивать,

потому всегда очень стараюсь достичь большего.

Мама, спасибо что отпустила меня к моей мечте.

____________

Acknowledgements

197

But the most important, only two people in this world have seen what was happening

“behind the scenes” of my PhD. First person is you, Andrew. I would never heard about

Groningen if not you. You always helped me with study, science and with private problems.

Too bad we were more than friends. But we learnt so much on the way, even moving in the

wrong direction. Thank you for all you did for me and thank you for being next to me when my

life was a mess.

And of course you, Remi. Thank you for appearing in my life in time, when I needed

someone to be on my side, somebody who actually wanted to be with me in good and bad times,

who «shares my dreams, I hope that someday I'll share a home”. You have seen me struggling

with writing this PhD thesis and always supported me. I am so grateful to have you in my life.

Love you.

Curriculum Vitae

199

Curriculum Vitae

Margaryta Ianovska was born on June 21st 1990 in Kiev, Ukraine. After finishing high school

in 2007 she entered the Taras Shevchenko National University of Kiev, where she obtained first

BSc and then MSc degree (both cum laude) in Analytical Chemistry. After the graduation

Margaryta had been working for several months at the pharmaceutical company Darnitsa in

Kiev. In 2013 she was accepted as a PhD student at the University of Groningen and moved

abroad. Mainly she had been working the Department of Pharmacy, having several internships

at DSM and visiting/working in other labs across the Europe. The results of her PhD project are

presented in this thesis. During her second PhD year, Margaryta gave an oral talk at the

international HPLC conference in Geneva, Switzerland. Being a creative person, she designed

a new group logo that is used on all posters and presentation of the Pharmaceutical Analysis

group. Beside of that, Margaryta is interested in art, photography, baking and dancing salsa.

Updated Linkedin page:

List of publications

200

List of publications:

Published:

Ianovska M.A., Mulder P.P.M.F.A., Verpoorte E. Development of small-volume, microfluidic

chaotic mixers for future application in two-dimensional liquid chromatography, RSC Adv., 2017,

7, 9090.

Submitted:

De Haan P., Ianovska M.A., Mathwig K., van Lieshout G., Triantis V., Bouwmeester H., Verpoorte

E. Digestion-on-a-Chip: A Continuous-Flow Modular Microsystem for Enzymatic Digestion for

Gut-on-a-Chip Applications, Lab on a Chip.

In preparation:

Ianovska M.A., Heiskanen A., Wouters B., Ritzen E., Mengerink Y., Schoenmakers P., Emnéus J.,

Verpoorte E. Microfluidic micromixer as a tool to overcome solvent incompetibilities in two-

dimensional liquid chromatography.

Ianovska M.A., Mulder P.P.M.F.A., Verpoorte E. Novel micromixers based on chaotic advection

and their application —a review.

Ianovska M.A., Mulder J.-P.S.H., Verpoorte E. Fabrication of a pressure-resistant microfluidic

mixer in fused silica using Selective Laser-Induced Etching.

For an updated list of scientific publications one can follow this QR code:

university of groningen microfluidic tools for …...cover design: margaryta a. ianovska layout...

Documents