FREEFORM-BASED OPTOFLUIDIC DEVICES TOWARDS …

Thesis submitted in fulfilment of requirements for the award of the degree of Doctor

of Engineering Sciences (Doctor in de Ingenieurswetenschappen)

FREEFORM-BASED OPTOFLUIDIC DEVICES TOWARDS

SURFACE-ENHANCED RAMAN SPECTROSCOPY (SERS)

QING LIU

Academic Year 2019-2020

Promotors:

Prof. dr. ir. Heidi Ottevaere

Prof. dr. ir. Hugo Thienpont

Faculty of Engineering

Department of Applied Physics and Photonics, Brussels Photonics

Members of the Jury

Prof. Dr. Dominique Maes, president

Department of Bio-engineering Sciences,

Vrije Universiteit Brussel, Belgium

Prof. Dr. Roger Vounckx, vice-president

Department of Electronics and Informatics,


Prof. Dr. Ir. Heidi Ottevaere, promotor

Department of Applied Physics and Photonics,


Prof. Dr. Ir. Hugo Thienpont, promotor



Prof. Dr. Ir. Jürgen Van Erps, Secretary



Dr. Ir. Michael S. Schmidt

Silmeco ApS

Copenhagen, Denmark

Prof. Dr. Eric Ziemons

Centre Interdisciplinaire de Recherche sur le Médicament (CIRM),

Université de Liège, Belgium

Prof. Dr. Peter Vandenabeele

Department of Archaeology,

Universiteit Gent, Belgium

Acknowledgements

I still remember the day when I first time met my promoter prof. Heidi Ottevaere in

Beijing in 2014. The sky was a bit grey and traffic was bad, but we had a pleasant

conversation and one year later, my PhD story began. After 4+ years, at the moment

when I am about to finalize my PhD thesis, I would like to express my sincere

gratitude to those who have helped and supported me from all aspects during my PhD.

First of all, I would like to thank my promotors prof. Heidi Ottevaere and prof. Hugo

Thienpont who have guided me through these years of my doctoral degree. Heidi,

thank you for giving me the opportunity to join your team as a PhD student. You

inspired me to come up with a lot of new ideas and encouraged me to investigate them

in depth and implement them. Your excellent organization and coordination ability

have not only promoted the progress of my project, but also taught me to work in close

liaison with international people. I deeply appreciate the time that you have spent and

the effort you have made to shape me into a scientist. Hugo, thank you for your

consultations and your critical remarks which gave me a deeper insight into the

problem I was facing, as well as proposing many new solutions. This work would not

have been possible without your guidance.

I would like to thank the jury members prof. Dominique Maes, prof. Roger Vounckx,

prof. Jürgen Van Erps, dr. Michael S. Schmidt, prof. Eric Ziemons and prof. Peter

Vandenabeele, who have provided me with relevant comments and advices on how to

improve the quality of my doctoral dissertation. I would like to express my gratitude

to the members of the guidance committee prof. Peter Dubruel and prof. Gert Desmet,

which yearly evaluated my work and they gave me advices to make sure that my thesis

progressed in the right direction.

I would like to thank my colleagues in the Photonics Innovation Center in Gooik for

their help with the fabrication and metrology, and for their precious suggestions to my

research based on their expertise. Thank you, Koen Vanmol, Michael Vervaeke, Kurt

Rochlitz, Dries Rosseel and Julie Verdood. I would also like to acknowledge my

fellow PhD students, colleagues and friends, especially Jens De Pelsmaeker, who

helped me to adapt in a new environment and to deal with scientific and technological

problems in the beginning of my PhD.

Last but not the least, I would like to thank my family and my beloved parents for

their unconditional love, trusts and support. And I would like to thank my beloved

wife Yangzi He for understanding, for supporting and motivating me in the most

difficult moments. Thank you!

Qing Liu

Brussels, May 2020

I

Abstract

Material identification is of great importance in industrial production, scientific

research and daily life of human beings. Optical detection technologies, among all

physical and chemical methods, have become indispensable tools to qualitatively and

quantitatively characterize the composition, concentration and properties of the

sample under test by analyzing the light emitted by the substances, or by refraction,

reflection, scattering or absorption of external sources. Laser induced Raman

spectroscopy, one of the common optical detection technologies, achieve the detection

purposes by providing fingerprint information of the molecular vibrational modes in

a label-free and non-invasive approach, which is essential in various application

domains to eliminate the interference from external factors.

Today, Raman spectroscopy has made a splash in identifying mycotoxins, which are

toxic secondary metabolite compounds naturally produced by certain types of fungi

(molds) in food and have adverse effects on human health ranging from acute

poisoning to long-term effects such as cancers and immune deficiency. Raman

spectroscopy is also a powerful tool in clinical diagnostics such as tumor detection, as

well as biological research of single-cell analysis. Today, many Raman spectroscopy

setups are already commercially available, but most of them are traditional bulky

instruments with considerable requirements in terms of time, volume consumption

and manual sampling of substances of interest. Today, there is a growing demand for

compact, integrated and intelligent Raman spectroscopy systems for various

application domains.

In this PhD, we aim for the miniaturization of traditional bulky Raman spectroscopy

setups and for the integration of several laboratory functionalities in a polymer-based

lab-on-chip for microfluidic detection with high sensitivity. First, a Raman probe

design for a lab-on-chip was developed by miniaturizing and optimizing the optical

components, which allowed remote and robust Raman detection. In addition, freeform

reflector embedded lab-on-chips have been designed for hot embossing mass

II

manufacturing, which could substantially reduce the sample consumption,

experimental complexity and the cost of detection. The performance of the Raman

probe working in combination with the lab-on-chip were assessed by a non-sequential

ray tracing simulation approach, and the simulation results showed a good agreement

with the experimental results. In addition, we fabricated different nanostructures by

two-photon polymerization (2PP) 3D lithography for Surface-Enhanced Raman

Spectroscopy (SERS) applications, such as mycotoxin detection. The performance of

our 2PP printed SERS substrates were evaluated by Finite-Difference Time-Domain

(FDTD) simulations and experimental approaches respectively. Furthermore, we

implemented a segmented freeform reflector-based tunable Raman spectroscopy

setup for microfluidic lab-on-chips that allows both conventional and surface-

enhanced Raman detection. Our tunable Raman spectroscopy setup is compatible with

our mass fabricated polymer lab-on-chips as well as many commercial microfluidic

chips.

III

List of Abbreviations

2PP/TPP Two-Photon Polymerization

AEF Analytical Enhancement Factor

AFM Atomic Force Microscopy

ALD Atomic Layer Deposition

AOM Acousto-Optical Modulator

AOTF Acousto-Optic Tunable Filter

ASAP Advanced Systems Analysis Program

BPF Bandpass Filter

BZMA Benzyl Methacrylate Monomer

CAD Computer-Aided Design

CAM Computer-Aided Manufacturing

CB Conduction Band

CCD Charge-Coupled Device

CL Collection Lens

CLVF Circular Linear Variable Filter

CMOS Complementary Metal–Oxide–Semiconductor

COC Cyclic Olefin Copolymer

CPA Close-Packed Arrays

CRM Confocal Raman Microspectrometer

CTAB Cetyltrimethyl Ammonium Bromide

CV Cyclic Voltammetry

DM Dichroic Mirror

DMF Dimethylformamide

DON Deoxynivalenol

EG Ethylene Glycol

EL Excitation Lens

FDTD Finite-Difference Time-Domain

FEP Fluorinated Ethylene Propylene

FIA Flow Injection Analyses

FT Fourier Transform

FUM Fumonisin

FWHM Full Width at Half Maximum

GC Gas Chromatography

GPC Gel Permeation Chromatography

IV

HCP Hexagonal Close Packed

HDMA Hexanediol Dimethacrylate Crosslinker

HDT Heat Deflection Temperature

HOMO Highest Occupied Molecular Orbital

IC Integrated Circuit

IC Integrated Circuit

IERS Interference Enhancement of Raman Spectra

IR Infrared

LCTF Liquid Crystal Tunable Filter

LED Light-Emitting Diode

LoC Lab-on-Chip

LPF Long-Pass Filter

LSPR Localized Surface Plasmon Resonance

LUMO Lowest Unoccupied Molecular Orbital

LVF Linear Variable Filter

MEMS Micro-Electro-Mechanical System

MF Merit Functions

MFON Metal Film Over Nanospheres

MIP Molecularly Imprinted Polymer

MMF Multi-Mode Fiber

µPAD Micro paper-based analytical device

NEC Noise-Equivalent-Concentration

NF Notch Filter

NiP Nickel Phosphorus

NIR Near-Infrared

NSC Non-Sequential Component

ORC Oxidation-Reduction Cycle

PC Polycarbonate

PCA Principal Component Analysis

PDMS Polydimethylsiloxane

PET Polyethylene Terephthalate

PMMA Poly(Methyl Methacrylate)

PoC Proof-of-Concept

POCT Point-of-care test

PS Polystyrene

PT Plasma Treatment

PVP Polyvinylpyrrolidone

V

R2R Roll-to-Roll

RhB Rhodamine B

RI Refractive Index

RRS Resonance Raman Spectroscopy

SA Sparse Array

SEM Scanning Electron Microscopy

SERRS Surface-Enhanced Resonance Raman Spectroscopy

SERS Surface-Enhanced Raman Scattering/Spectroscopy

SF Suppression Factor

SMEF Single-Molecule Enhancement Factor

SMF Single-Mode Fiber

SNR Signal-to-Noise Ratio

SOP Standard Operating Procedure

SP Surface Plasmon

SR Spheres Removal

STM Scanning Tunneling Microscope

TAS Total Analysis Systems

Tg Transition Temperature

TIR Total-Internal Reflection

TIS Total Integrated Scatter

UV Ultraviolet

VB Valence Band

VIS Visible

VLSI Very-large-scale integration

WCRS Waveguide Confined Raman Spectroscopy

WD Working Distance

Table of Contents

1

Table of Contents

Abstract ......................................................................................................................................................I

List of Abbreviations ............................................................................................................................. III

Table of Contents ...................................................................................................................................... 1

1 General Introduction ...................................................................................................................... 1

1.1 Rationale of this PhD ............................................................................................................ 1

1.2 Objectives and methods of this PhD work ............................................................................. 2

1.2.1 Objectives of this PhD ..................................................................................................... 2

1.2.2 Methods of this PhD ........................................................................................................ 4

1.3 Microfluidic Lab-on-chip Raman State-of-the-Art ................................................................ 5

1.3.1 History of microfluidic Raman spectroscopy lab-on-chip ................................................ 5

1.3.2 LoC materials ................................................................................................................ 14

1.3.3 LoC fabrication methods ................................................................................................ 17

1.4 Overview of this PhD .......................................................................................................... 21

References ........................................................................................................................................... 24

2 Introduction to Raman and Surface-Enhanced Raman Spectroscopy...................................... 31

2.1 Introduction to electromagnetic radiation ............................................................................ 31

2.2 Introduction to Raman spectroscopy ................................................................................... 34

2.2.1 Diatomic molecule: Harmonic oscillator........................................................................ 34

2.2.2 Diatomic molecule under an external electromagnetic field .......................................... 37

2.2.3 Vibrational modes of molecules ..................................................................................... 40

2.2.4 Selection rules of Raman scattering and IR absorption .................................................. 44

2.2.5 Raman spectrum ............................................................................................................ 45

2.2.6 Raman cross section ....................................................................................................... 47

2.2.7 Pros and Cons of Raman spectroscopy .......................................................................... 48

2.3 Instrumentation for Raman spectroscopy ............................................................................ 49

2.3.1 Light source ................................................................................................................... 50

2.3.2 Filters ............................................................................................................................. 51

2.3.3 Lenses ............................................................................................................................ 51

2.3.4 Spectrometer .................................................................................................................. 51

2.4 Surface Enhanced Raman Spectroscopy (SERS) ................................................................. 52

2.4.1 History of SERS ............................................................................................................ 52

2.4.2 Electromagnetic Enhancement ....................................................................................... 54

2.4.3 Chemical Enhancement mechanism ............................................................................... 57

Table of Contents

2

2.4.4 SERS substrates ............................................................................................................. 58

2.5 A brief overview on Raman and SERS applications ........................................................... 67

References ........................................................................................................................................... 70

3 Two-Photon Polymerized Nanostructures for SERS Analysis................................................... 79

3.1 Fabrication of nanostructures with two-photon polymerization and simulations ................. 80

3.1.1 Two-photon polymerization process .............................................................................. 80

3.1.2 Electromagnetic field enhancement simulation of nanostructures with FDTD method .. 82

3.1.3 Influence of fabrication errors ........................................................................................ 87

3.2 Experiments with the printed SERS substrates .................................................................... 89

3.2.1 SERS enhancement analysis .......................................................................................... 89

3.2.2 SERS substrate calibration ............................................................................................. 92

3.3 An application of two-photon polymerized SERS substrates .............................................. 93

3.3.1 Mycotoxin detection with 2PP polymerized SERS substrates ........................................ 93

3.3.2 Principal Component Analysis (PCA) for the spectra of mycotoxins............................. 94

3.4 Conclusions ......................................................................................................................... 96

References ........................................................................................................................................... 97

4 Integrated Confocal Raman Probe Combined with a Freeform Reflector Embedded Lab-on-

Chip ...................................................................................................................................................... 103

4.1 Introduction to a freeform reflector-based confocal Raman spectroscopy lab-on-chip system

103

4.1.1 Confocal principle........................................................................................................ 104

4.1.2 Design of the confocal Raman spectroscopy LoC setup .............................................. 104

4.1.3 Implementation of the confocal Raman spectroscopy LoC setup ................................. 106

4.1.4 Conclusion ................................................................................................................... 107

4.2 Integrated Confocal Raman probe optimized for microfluidic lab-on-chip ....................... 108

4.2.1 Concept of Raman probe and state-of-the-art ............................................................... 108

4.2.2 Design considerations of an integrated Raman probe for confocal Raman lab-on-chip

measurements................................................................................................................................ 111

4.2.3 Experimental performance of the Raman probe combined with a LoC ........................ 119

4.2.4 Conclusion ................................................................................................................... 120

4.3 Mass manufacturing of the LoC ........................................................................................ 121

4.3.1 Drawbacks of our previous Raman LoCs and solutions ............................................... 121

4.3.2 TOPAS COC polymers ................................................................................................ 122

4.3.3 Design of the freeform reflector-based LoC for mass manufacturing .......................... 125

4.3.4 Double-sided hot embossing for the 50µm focus G2 LoC ........................................... 130

4.3.5 Mass fabrication of the COC-based 100µm focus G3 LoC .......................................... 136

4.3.6 Bonding and shear strength tests .................................................................................. 140

Table of Contents

3

4.4 Mass manufactured LoCs in combination with the developed Raman probe .................... 142

4.4.1 Implementation of the proof-of-concept demonstration for confocal measurements.... 142

4.4.2 System calibration........................................................................................................ 142

4.4.3 Proof-of-concept for mycotoxin detection ................................................................... 145

4.5 Conclusion ........................................................................................................................ 147

References ......................................................................................................................................... 149

5 A Tunable Freeform Segmented Reflector in a Microfluidic System for Conventional Raman

and SERS Spectroscopy ....................................................................................................................... 155

5.1 Design of the freeform segmented reflector ...................................................................... 157

5.1.1 An overview of the segmented reflector ...................................................................... 157

5.1.2 The middle concave segment design ............................................................................ 158

5.1.3 The center and marginal segment design ..................................................................... 160

5.2 Non-sequential simulation of the Raman spectroscopy with segmented reflector ............. 162

5.3 Fabrication of the segmented reflector and the microfluidic chip ...................................... 167

5.4 Preparation of the SERS substrates and SERS chip........................................................... 169

5.5 Experiments with the microfluidic Raman setup ............................................................... 170

5.5.1 Suppression factor of microfluidic Raman setup .......................................................... 170

5.5.2 Alignment tolerances ................................................................................................... 172

5.5.3 Limit of detection for conventional Raman analysis .................................................... 173

5.6 Using the microfluidic chip and segmented reflector in combination with a SERS substrate

175

5.7 Conclusions ....................................................................................................................... 176

References ......................................................................................................................................... 177

6 A Compact Conical Beam Shaper and Freeform Segmented Reflector for SERS Analysis .. 183

6.1 Design and fabrication of the freeform segmented reflector and beam shaper .................. 184

6.1.1 Working principle of the conical beam shaper and segmented reflector ...................... 184

6.1.2 Numerical approaches to calculate the surface profile of the segmented reflector ....... 185

6.1.3 Fabrication of the conical beam shaper and segmented reflector ................................. 189

6.2 Optical system design and simulation ............................................................................... 190

6.2.1 Non-sequential ray tracing simulations for the Raman system ..................................... 190

6.2.2 Confocal behavior of the Raman system ...................................................................... 191

6.3 Proof-of-concept demonstration of the Raman system ...................................................... 192

6.3.1 Performance of the beam shaper .................................................................................. 193

6.3.2 SERS measurements for the Rhodamine B .................................................................. 194

6.3.3 Misalignment tolerances of the setup ........................................................................... 195

6.4 Conclusion ........................................................................................................................ 197

References ......................................................................................................................................... 199

Table of Contents

4

7 Conclusions and Perspectives ..................................................................................................... 203

7.1 Conclusions ....................................................................................................................... 203

7.1.1 3D printed nanostructures by two-photon polymerization for SERS analysis .............. 203

7.1.2 Mass fabricated lab-on-chip with integrated freeform reflector in combination with a

Raman probe for microfluidic analysis ......................................................................................... 204

7.1.3 Freeform segmented reflector design for conventional Raman and SERS analysis...... 205

7.2 Perspectives ....................................................................................................................... 206

7.2.1 Optimization of the nanostructures for SERS analysis and mass manufacturing of the

nanostructure ................................................................................................................................. 206

7.2.2 Towards a 2PP fabricated optofluidic lab-on-chip for SERS analysis .......................... 207

7.2.3 Towards a lab-on-chip with integrated micro-excitation source and micro-spectrometer.

208

References ......................................................................................................................................... 209

List of Publications ............................................................................................................................... 213

Appendix 1 List of Tables ..................................................................................................................... 215

Appendix 2 List of Figures ................................................................................................................... 217

Chapter 1

1

Chapter 1

1 General Introduction

1.1 Rationale of this PhD

Material identification is of great importance in industrial production, scientific

research and daily lives of human beings. Optical detection technologies, among all

physical and chemical methods, have become indispensable tools to qualitatively and

quantitatively characterize the composition, concentration and properties of a sample

under test by analyzing the light emitted by the substance itself, or by refraction,

reflection, scattering or absorption of external light. In particular, laser induced

Raman scattering has made a splash in application domains such as biotechnology,

polymer chemistry, life-sciences, clinical diagnostics, drug development, archaeology

and art history, as it can provide the fingerprint information of the molecular

vibrational modes of the sample by non-invasive and label-free detection. In recent

decades, with the rapid development of computer sciences, Very-Large-Scale

Integration (VLSI) and information technologies, material science and other applied

sciences, as well as the in-depth exploration in the field of theoretical physics and

chemistry, there is a growing demand for highly compact, integrated and intelligent

Raman spectroscopy setups in various application domains especially in biochemical

research, since the high energy and sample consumption, high financial cost and

complexity in operation of the traditional bulky Raman spectroscopy setups have

limited the applications of Raman spectroscopy. As a result, it comes down to

miniaturize traditional Raman devices and integrate some of the main components

and key functionalities into a single milli- and micro-scale chip, meanwhile

combining with microfluidic approaches. The main components and functions that

can be integrated include optical components such as objective lenses, optical fibers

and optical filters for collecting, transmitting and splitting lights, and/or mechanical

parts such as translation stages, sample holder, chemical reaction chamber, fluidic

channel for transporting, mixing or sorting analytes, and/or detectors and electronic

circuits for signal acquisition and processing. We refer to these types of miniaturized

Chapter 1 General Introduction

2

and integrated systems for Raman analysis as “Microfluidic Raman spectroscopy lab-

on-chips (LoC)” or “Optofluidic Raman spectroscopy systems”.

Figure 1.1 Sum of times cited per year for 945 results on TOPIC (Raman) AND TOPIC

(microfluidic) from 1989-2019 on the Web of Science data base. (Accessed on Mar. 25, 2020)

Microfluidic lab-on-chip Raman spectroscopy has attracted a great deal of attention

in recent years, as is shown in Figure 1.1 that depicts the sum of times cited per year

for papers on topics of “Raman” and “microfluidic”. However, there are many

challenges to implement in such a microfluidic lab-on-chip Raman spectroscopy setup,

prior to bringing it to the market. How to reduce the overall size of the device while

ensuring the sensitivity and stability of the Raman signal, as well as the choice of

materials and processing methods to manufacture the lab-on-chips must be considered.

This PhD work discusses the key issues in each step of the whole process of

implementing microfluidic lab-on-chip Raman spectroscopy systems, from system

design and fabrication, to application and evaluation. In the process of solving

problems and optimizing the methodology, this dissertation establishes a preliminary

Standard Operating Procedure (SOP) that paves the way for designing and

implementing microfluidic LoC systems for Raman detection.

1.2 Objectives and methods of this PhD work

1.2.1 Objectives of this PhD

The main objectives of this PhD include miniaturization and integration of the optics,

Raman signal enhancement, optimization of the microfluidic lab-on-chips, mass-

manufacturing of the lab-on-chips and the use of microfluidic lab-on-chips in specific

application domains.

Chapter 1

3

Miniaturization and Integration

The philosophy of LoC research is to reduce the cost and complexity of laboratory

analysis by miniaturizing and integrating multiple laboratory processes on a single

device. Conventional Raman spectroscopy for biomedical analysis mainly consists of

an excitation source, the optics, the sample chamber and a spectral analyzer. In the

first step of this PhD work, we aim for the miniaturization and integration of the optics

that collects, transmits or splits light, and the fluidic chamber, channels and in/outlets

to an optofluidic lab-on-chip device. This is one of the major tasks of our ultimate

vision towards a microfluidic Raman spectroscopy system that miniaturizes and

integrates the laser source for Raman excitation, the optics and fluidics, and the

spectral analyzer and all other functionalities in a single device.

Signal Enhancement

Although Raman spectroscopy enables non-invasive and label-free detection to

provide fingerprint information of the molecules, the drawback of poor signal

intensity has drastically restricted its promotion in various application domains.

Raman spectroscopy is a spectral technique that analyzes the inelastic scattered light

absorbed and re-emitted by the molecules. Briefly, the Raman process consists of

three steps including excitation, absorption and re-emitting, and signal acquisition.

Based on the working principle of Raman spectroscopy, the Raman response can be

enhanced during each of the three phases of the Raman process. The direct approaches

are either to increase the power of the laser during excitation or extend the exposure

time during signal acquisition. However, both aforementioned methods will increase

the risk of sample degradation. The intensity of the laser is also restricted by the pump

sources, the gain mediums, the optical coupling process and so on. The good thing is

that we can also boost the Raman signal during the absorption and re-emitting process

by applying various approaches including Surface-Enhanced Raman Spectroscopy

(SERS), interference enhancement of Raman spectra (IERS) and Resonance Raman

Spectroscopy (RRS).1 In this PhD work, we explore the feasibility of the SERS

technique for biochemical applications. The details of SERS will be discussed in

Chapters 2 and 3.

Optimization

The optimization of the Raman spectroscopy setup refers to the action of making

effective use of the existing materials, components, structures and other resources.

With the help of computer-aided design (CAD) software simulation, we can optimize

the optical and mechanical structures and characteristics of the components to obtain

higher excitation and collection efficiencies within a compact structure. Selecting the

proper materials for lab-on-chip manufacturing enables biochemical analysis of more


4

types of analytes in a wide range of solvents. As there exists many factors and

variables that will influence the performance of the whole system, we can still make

a trade-off by allocating the resources between e.g. chemical resistance, detection

efficiencies and SNR.

Mass fabrication

One of the important reasons that limit the popularization of Raman spectroscopy is

its high cost for analysis. The idea of modern industrial design advocates mass

manufacturing to reduce the unit price. To reach this objective, we must consider the

possibility of mass manufacturing from the draft design and allow improvement and

innovation on the product structure and technological process. This also brings the

possibility of reproducible and disposable Raman measurements for biochemical

research and especially for mycotoxin detection.

Applications

Microfluidic lab-on-chip Raman spectroscopy has been used in a great variety of

application domains such as material science, drug development and medical

diagnostics. The main goal of my PhD thesis was to set up a platform for conventional

Raman and SERS detection. The detection of mycotoxins was chosen as an

application to demonstrate the proof-of-concept of the developed optofluidic devices.

1.2.2 Methods of this PhD

The general steps of this PhD can be summarized as illustrated in Figure 1.2. We first

establish some reasonable and yet basic level performance goals for the PhD topic.

Then we investigate the related concepts and theories and discuss the overall frame,

modules and functions of the microfluidic Raman spectroscopy LoC setup. Later we

build up the models for different modules by using relative CAD platforms

respectively and perform the simulations. The simulation results in a specific CAD

platform are introduced as feedback for the internal and inter-platform optimizations.

For instance, we design and calculate the surface profile of a freeform lens via a

numerical approach with the help of Matlab. The data points and fitted polynomial

functions of the freeform lens are introduced to SpaceClaim, a commonly used solid

modeling CAD software for mechanical engineering, and converted to a

stereolithography (STL) file, which can be inserted to Zemax OpticStudio as a Non-

Sequential Component (NSC) for the optical performance simulations. The

dimensions, optical properties, positions, etc. of the other NSCs optimized and

assessed by the Merit Functions (MF) in Zemax OpticStudio are used as references

Chapter 1

5

for the mechanical structure design via SpaceClaim. If there is a misalignment of the

mechanical structure design regarding the optical simulation results, we make an

adjustment and repeat the optical simulations until we reach a trade-off between

different modules and CAD platforms. Once the draft design of the lab-on-chip

Raman spectroscopy system is done, we select the proper materials and tools to

fabricate different parts of the system. A proof-of-concept (PoC) demonstration setup

is implemented by assembling and aligning the mechanical, optical and electronic

parts fabricated after the previous step. Later we conduct the calibration experiments

to assess the preliminary performances of the PoC setup, such as the system response,

signal-to-noise ratio (SNR), limit of detection for certain analytes and so on. In the

end, we introduce our PoC setup to real application domains as set from the beginning

of the workflow and evaluate the work to see if we reach our target.

Figure 1.2 Workflow of this PhD work.

1.3 Microfluidic Lab-on-chip Raman State-of-the-Art

1.3.1 History of microfluidic Raman spectroscopy lab-on-chip

The history of microfluidic lab-on-chip analysis can be traced back to the concept of

flow analysis that emerged in the early 1970s. In 1970, E. Pungor, Z. Feher and G.

Nagy initially developed silicone rubber-based graphite electrodes for analysis in

continuous flowing media through a capillary tube2,3, as shown in Figure 1.3.


6

Figure 1.3 (Left)The first apparatus of flow analysis setup for determination of organic and in-

organic compounds by using (Right) silicone rubber-based graphite electrodes. (C) Valve,

(E1) Indicator electrode, (E2) Reference electrode. Presented by E. Pungor et al.2,3

In 1975, J. Ruzicka and E. Hansen firstly presented the concept of Flow Injection

Analyses (FIA),4 which aim to develop automated and compact analytical systems

that integrate the sample preparation, separation and sensing by injecting continuous

flow to form a Total Analysis Systems (TAS), as shown in Figure 1.4.

Figure 1.4 The first actual FIA system described by J. Ruzicka and E. Hansen.4 (a) Polymer

blocks; (b) Silicone rubber wall; (c) polyethylene tubing.

Since then, with the fast development of micromechanics and microfabrication

technologies, there came the tendency to reduce the size of the TASs towards the

miniaturized TASs (µ-TAS).5 In 1979, C.T. Stephen and co-workers implemented a

miniature gas chromatography (GC) system where the major parts of the capillary

column were fabricated in silicon using photolithography and chemical etching,6 as

shown in Figure 1.5. This is the first reported actual lab-on-chip system, as C.T.

Stephen presented that

Chapter 1

7

“With the addition of a powerful one-chip, battery-powered microcomputer, a

complete gas analysis system can be contained in a volume similar to that of early

pocket calculators”.

Figure 1.5 (Left) Block diagram of the first LoC system developed by C.T. Stephen et al.6

(Right) Photograph of micro capillary column and gas chromatographic system on a 5-cm-

diameter silicon wafer.

The miniaturization of the analytical systems, especially the optical analytical systems,

has continuously been ongoing since then based on the principles of total-internal

reflection (TIR),7 optical interference,8 chemiluminescence,9 fluorescence10 and other

optical processes. In 1982, B. Barry and R. Mathies demonstrated the first

miniaturized Resonance Raman platform for in situ detection of single visual pigment

cells.11 The Raman microprobe consists of a quartz glass sealed chamber and a liquid

nitrogen cold stage for trapping the labelled cells. The Raman setup was also equipped

with dry nitrogen flow to reduce the condensation during cooling, as shown in Figure

1.6.

Figure 1.6 Cross-section of the Raman detection chamber and liquid nitrogen cold stage for in

situ detection, presented by B. Barry and R. Mathies.11

The miniaturization of Raman spectroscopy systems has continuously attracted

increasing but relatively low degree of attention in the 1980s and 1990s.12–14 However,


8

with the rapid development of nanotechnology in the 21st century and the growing

demand for material characterization in biological, chemical and other fields, the

research output of LoC showed an explosive growth. L. Moonkwon and co-workers

demonstrated the feasibility of using glass-based microfluidic Raman chips for in situ

monitoring of imine formation in 200315, as shown in Figure 1.7. The microfluidic

chip was fabricated via photolithography and wet etching and has three inlets for

injecting the chemical reagents and one outlet for dumping. Chemical reagents

including benzaldehyde, aniline and chloroform were injected into a 400µm width,

20µm height microfluidic channel through 3 inlets simultaneously using a micro

pump. The chemical reaction was monitored continuously by focusing a 514.5nm

wavelength laser inside the microfluidic channel using a 10× objective lens. The

experimental results of time-dependent Raman measurements proved that Raman

spectroscopy combined with careful in-depth focusing in a microfluidic channel is a

powerful tool for chemical reaction analysis.

Figure 1.7 (a) Layout of the glass microfluidic chip and (b) photograph of the microfluidic

Raman spectroscopy setup, demonstrated by L. Moonkwon et al.15

In 2006, S. Barnes demonstrated a thiolene-based microfluidic Raman spectroscopy

device for analysis of polymer droplets composition and conversion.16 The structure

of the microfluidic chip is similar to that of Moonkwon’s glass-based microfluidic

chip but with 4 inlets. Benzyl methacrylate monomer (BZMA) and hexanediol

dimethacrylate crosslinker (HDMA) were used as reagents for organic droplet

formation. The polymerization of the droplets was initiated by a UV lamp and

monitored by a Raman probe with 785nm wavelength excitation, as shown in Figure

1.8. The combination of a microfluidic lab-on-chip with fiber optics Raman probe has

allowed real-time screening of polymeric reactions with a relatively low cost and

versatile instrument. However, because of the low Raman response, this application

required a comparatively long acquisition time (3 minutes) to obtain high quality

signals.

Chapter 1

9

Figure 1.8 (Left) schematics of the thiolene-based microfluidic lab-on-chip and (right)

photograph of the Raman spectroscopy system for droplet formulation analysis presented by

S. Barnes et al.16

To overcome the aformentioned low sensitivity and long acquisition time drawbacks

of regular Raman spectroscopy, R. Keir and co-workers presented their work of a

glass-based Surface-Enhanced Resonance Raman Spectroscopy (SERRS) lab-on-chip

device for microflow cell analysis in 200217, as shown in Figure 1.9. The fluidic

channels were fabricated by photolithographic techniques. The inlets and outlets on

the cover glass plate were fabricated by a diamond engraving tool. The flow inside

the channel was characterized with mixing the Ag colloid as SERS substrate and

excited by a 15mW argon ion laser working at 514nm wavelength. Later in 2004, they

developed the first polymer-based Surface-Enhanced Resonance Raman

Spectroscopy (SERRS) microfluidic lab-on-chip for the detection of three dye-

labelled oligonucleotides.18 They fabricated the lab-on-chip by using PDMS material

with a silicon mask fabricated by photolighography and etching. This SERRS lab-on-

chip system enables the Raman detection of oligonucleotides labelled with R6G dye

at a concentration of 0.1mM. The use of PDMS has greatly reduced the cost of sample

analysis, and also paves the way for mass manufacturing of signal-enhanced Raman

lab-on-chip systems.


10

Figure 1.9 Schematic of the glass-based SERRS microfluidic lab-on-chip demonstrated by R.

Keir et al.17

The main function of the early stage microfluidic lab-on-chips is to transport, mix or

separate the analyte flows. The Raman excitation and detection processes of these lab-

on-chip systems are largely dependant on the external optics such as confocal Raman

microscope or Raman probe. One solution to reduce the overall size of the

microfluidic lab-on-chip systems is to integrate the fiber optics in the chip as the key

functional components in the excitation and collection paths. In 2010, P. Ashok and

his co-workers reported on the demonstration of an optical probe-based microfluidic

Raman spectroscopy lab-on-chip.19 The microfluidic device was fabricated with

PDMS using a soft lithography approach. It contains a set of fiber probe channels and

a set of fluidic channels respectively. The diameter of the optical fibers and the fluidic

channels are both 200µm. The physical dimension of the lab-on-chip with inserted

probe heads is approximatly 30mm × 25mm, as shown in Figure 1.10.

Figure 1.10 (a) Schematic and (b) photograph of the optical probe-based PDMS Raman lab-

on-chip device demonstrated by P. Ashok et al.19

The noise-equivalent-concentration (NEC) - the concentration of the analyte when the

signal from the sample of interest is equal to the noise - of this lab-on-chip device is

estimated to be 150mM for urea solution under the 200mW laser excitation at 785nm

wavelength, with an integration time of 5 seconds. In the next year, they reported the

Chapter 1

11

first implementation of a Waveguide Confined Raman Spectroscopy (WCRS) lab-on-

chip with optimized fluidic and fiber channel layout.20 The new configuration further

miniaturized the chip size by replacing the optical probe with directly inserted optical

fibers, as shown in Figure 1.11. The new lab-on-chip setup showed a better NEC

(80mM) compared with their previous optical probe-based lab-on-chip device. The

WCRS lab-on-chip enables bioanalytes detection with minimal sample preparation

and low cost due to the optimized configuration and fabrication process. Microfluidic

lab-on-chip with integrated fiber optics was also investigated by many other

researchers.21–23

Figure 1.11 Schematic of the PDMS-based WCRS microfluidic chip presented by P. Ashok et

al.20

Although integration of fiber-optics with microfluidic chips can greatly reduce the

size and cost of Raman spectroscopy devices, one of the main disadvantages of this

type of devices is their low sensitivity resulting from the low excitation and collection

efficiencies. Generally, there are two approaches to solve this problem to implement

a sensitive microfluidic Raman spectroscopy system.

The first one involves embedding miniaturized lenses and other optics in the system

to increase the excitation and collecting efficiency.24–27 Fiber optics in combination

with other optical components such as mirrors and lenses can be placed on the sample

directly or work in combination with microfluidic chips, as shown in Figure 1.12.25

The integration of a freeform lens with a microfluidic chip enables a high numerical

aperture (NA=1.28), resulting in high excitation and collection efficiencies, and the

possibility of optical trapping applications27, as shown in Figure 1.13. Various types

of tunable lenses with variable foci have been investigated for microfluidic Raman

spectroscopy applications. Unlike the traditional zoom lens with a bundle of lenses,

which controls the focus by changing the relative distance between the individual

lenses, the variable lenses integrated on the microfluidic lab-on-chip systems are

usually made from flexible material, and the dynamic changes of the focal length are


12

realized by controlling the geometries of the lenses according to the change of surface

tensions, as shown in Figure 1.14.26

Figure 1.12 Design of a fiber Raman probe with miniaturized mirrors and lenses for a

microfluidic lab-on-chip system. (a) Non-sequential ray tracing of the microfluidic system. (b)

Schematic drawing of a mold for PDMS probe fabrication. (c) Schematic (left) and

photograph (right) of the PDMS probe. Demonstrated by T. Ngernsutivorakul.25

Figure 1.13 Freeform reflector embedded microfluidic lab-on-chip demonstrated by D. De.

Coster et al.27

Chapter 1

13

Figure 1.14 Schematics of different types of on-chip variable lens systems. (A) Pneumatically

tuned lens. (B) Lens geometry controlled by electrowetting induced changes. (C)

Hydrodynamic or electrokinetic lens. (D) Environmentally responsive lens. (ITO) indium tine

oxide. The dashed lines refer to the limits of physical tunability. Presented by K. Bates el al.26

The second approach to increase the sensitivity of Raman LoC detection involves

incorporating SERS with microfluidic devices by either mixing SERS sensitive

nanoparticles with analytes28–30 or integration of fixed SERS substrates inside the

microfluidic channel31–34, as shown in Figure 1.15 and Figure 1.16. Unlike the first

method, which is only a few times of reinforcement compared to conventional Raman

spectroscopy, the use of SERS is able to enhance the Raman scattering by a typical

factor of 104-109 due to localized surface plasmon resonance (LSPR).35 Besides,

optical fiber-based Raman probes with integrated nanostructures on the fiber facets

are also utilized for fluidic detection.36–39 The principle and other details of SERS will

be discussed in Chapter 2.


14

Figure 1.15 Schematic of a PDMS-based microfluidic channel for the analysis of cyanide

water pollutant with silver colloids presented by K. Yea et al.29

Figure 1.16 Schematics of tip coated multimode fibers (TCMMF) as Raman probe presented

by (left) C. Shi et al.36 and (right) M. Fan et al.,37 respectively.

1.3.2 LoC materials

One of the essential objectives of lab-on-chip systems is to reduce the cost of

equipment through miniaturization and integration. To do this, it is very important to

choose the proper materials and processing approaches to fabricate lab-on-chip

devices.40 In terms of material selection, the main considerations involve a variety of

physical and chemical properties, as well as the cost of the materials. Typical materials

utilized for microfluidic lab-on-chip fabrication include silicon, soda-lime and quartz

glass, and a variety of polymer resins. The general characteristics of some typical

materials used for microfluidic lab-on-chip fabrication are listed in Table 1.1.

Chapter 1

15

Table 1.1 General characteristics of some materials for microfluidic LoC fabrication.

Si Glass Quartz COC PC PMMA PDMS

cured

PS

Transmission

(%)

(400-800nm)

- 91 93 91 88-89 92 93 90

Refractive Index 3.98 1.52 1.46 1.53 1.15-

1.2

1.49 1.38-

1.40

1.59

Glass transition

Temperature Tg

(°C)

- - - 80-180 160-

200

115 - 100

Melting point

Tm (°C)

1700 1400-

1700

1650 290-

310

280-

320

170-190 - 210-

249

Heat Deflection

Temperature

HDT under

1.8MPa load

(°C)

- - - 130-

170

140-

180

99 - 78

Dielectric

constant

11.7 7.86 3.8 2.35 2.8-3 2.8-3.6 2.75 2.56

Density (g/cm3) 2.65 2.48 2.20 1.02 1.15-

1.2

1.18 0.970 1.05

Moulding

Shrinkage (%)

0.5-

0.8

0.2-0.8 0.4-0.7

Tensile Strength

(MPa)

165 70 48 46-63 55-77 69 1.55-9 16

Elongation at

break (%)

- 2.2-2.5 - 1.7-2.7 50-

120

4.5 430-725 >35

Water

absorption (%)

(23°C, 24h)

- 0.0 0.0 <0.01 0.1-

0.2

0.2 0.7 0.03-

0.1

Mohs Hardness

Scale

7 5-7 7 2-3 1.5-2 2.5-3 <141 1.5-2

(The table is a summary based on the data manuals provided by various glass and

polymer material suppliers.)

Since Raman spectroscopy is an optical detection technique that analyzes the

interaction of light with sample molecules42, there is no doubt that the first

consideration is the optical properties of the material for the channel, chamber, seal

and other parts of the chip, such as its transmission and refractive index. Besides, the

Raman responses of the materials should be as weak as possible, or at least avoid

overlap with the main Raman bands of the sample molecules under test. Otherwise,

the Raman scattering of the materials will result in vital interference to the qualitative


16

and quantitative analysis of the sample analytes. The Raman bands and relative

intensities of some polymers are listed in the table below.

Table 1.2 Raman peaks and relative intensities of different types of polymers. (s: strong; m:

medium; w: weak; The bold numbers refer to the main Raman peaks)

Raman Shift (cm-1) COC* PMMA* PC43,44 PS45 PDMS46

0-500 307 (m) 336 (w) 393 (w)

425 (w) 486 (w) 492 (s)

500-1000 512 (w) 602 (m) 636 (m) 621 (w) 618 (w)

747 (w) 814 (s) 705 (m) 796 (w) 689 (w)

837 (w) 988 (m) 734 (w) 711 (m)

889 (m) 829 (w) 791 (w)

932 (s) 888 (s) 862 (w)

919 (w)

1000-2000 1042 (w) 1065 (m) 1006 (w) 1001 (s) 1265 (w)

1072 (w) 1131 (m) 1111 (s) 1032 (m) 1414 (w)

1119 (m) 1297 (s) 1178 (m) 1115 (w)

1225 (m) 1444 (s) 1235 (m) 1451 (w)

1308 (m) 1725 (w) 1445 (w) 1583 (w)

1449 (s) 1464 (w) 1602 (m)

1603 (s)

1774 (w)

2000-3500 2871 (m) 2849 (w) 2472 (w) 2852 (w) 2161 (w)

2952 (m) 2883 (w) 2718 (w) 2904 (w) 2791 (w)

3129 (w) 2766 (w) 3054 (m) 2880 (s)

2873 (m) 2941 (s)

2914 (m)

2942 (m)

2972 (m)

3075 (s)

(*: Raman spectra measured with Bruker Senterra confocal Raman microscope at the Department of

Analytical Chemistry, Ghent University)

Considering the environment of the applications, one should also take into account

the thermal properties or electrical properties of the materials. Especially for the lab-

on-chip systems used in the field of biology and chemistry, the non-negligible

consideration of chip materials involves the chemical resistance and biocompatibility.

The chemical resistance refers to the ability of the material to protect against chemical

attack or solvent reactions. Most of the polymers are stable when directly in contact

with mild aqueous solutions of inorganic chemicals, fats and oils. However, exposure

with some organic solvents will attack the polymers gradually or acutely, resulting in

stress cracking or corrosion of the materials.47,48 The degradation of polymers will

either decay the optical and mechanical performance of the system or contaminate the

Chapter 1

17

samples, which we do not accept. Biocompatibility is of vital importance when

developing a lab-on-chip system for in vitro or in vivo studies, because the

microorganisms, cells or biological molecules have direct contact with the material

for a length of time. The chemical resistances and biocompatibilities of some materials

are listed in Table 1.3.

Table 1.3 Chemical resistance and biocompatibility of some typical materials for lab-on-chip

fabrication.

Si Glass Quartz COC PC PMMA PDMS PS

Soap solution o + + + + + + o

Sodium Hydroxide - - - + - + o +

Ammonia solution o + + + - + + +

Hydrochloric acid (36%) + + + + + + + o

Sulphur acid (40%) + + + + + o - +

Acetic acid (10%) + + + + + + + +

Nitric acid (35%) + + + + + o - o

Methanol + + + + - - + o

Ethanol + + + + + o + +

Isopropanol + + + + + - + +

Acetone + + + + - - - -

Hexane + + + - o + - o

Benzaldehyde + + + o - - - -

Benzene + + + - - - - -

Oleic Acid + + + + + + + +

Biocompatibility Poor Poor Poor Good Good Good Good Good

(+: resistant; o: limited resistant; -: not resistant, @20°C. The table is a summary based on the data manuals

provided by various glass and polymer material suppliers.)

1.3.3 LoC fabrication methods

Fabrication processes are closely in line with the types and properties of the materials.

Commonly used fabrication approaches include conventional photolithography,49–51

chemical etching,52,53 hot embossing,41,54–56 injection molding,40,57–60 soft

lithography,61–64 CNC machining,27 laser printing,32,65 laser etching/engraving,66,67 etc.

Figure 1.17 compares the cost and volume of some common lab-on-chip fabrication

technologies.68


18

Figure 1.17 Cost and volume comparison for common lab-on-chip fabrication technologies.68

(POCT: point-of-care testing, µPAD: micro paper-based analytical device.)

In the early times, photolithography for conventional silicon-based two-dimensional

integrated circuits (IC) has been investigated to fabricate microfluidic devices.6

However, photolithography has high requirements for its processing environment and

equipment because it is very expensive and can only be implemented in specific

institutes with cleanrooms and UV/X-ray lithography infrastructure available. Now it

is mostly used to produce silicon-based master molds for other processes.69 The

master molds can also be produced by micro milling or turning depends on the

roughness of the sidewalls and bottom of the channel required, as shown in Figure

1.18.

Figure 1.18 Illustrations of the typical fabrication techniques: (a) micro-milling, (b) UV

lithography, and (c) X-ray lithography.69

Chapter 1

19

Hot embossing is a widely used chip fabrication technique for rapid replication of

microstructures: the micro- and nano-patterns can be stamped into a thermoplastic by

raising the temperature of the polymer above its glass transition temperature (Tg).54–

56 Hot embossing is very suitable for small batch production of large-scale polymer

wafers with micron-scale features. Injection molding is another replication method

that fabricates microfluidic chips by feeding the material into a heated barrel, then

mixing and forcing material flow into a mold cavity where it cools and hardens to the

structure of the cavity.40,57–59 Micro injection molding can sustainably reduce the

production cost of mass fabricated microfluidic chips. Figure 1.19 illustrates the

scheme of hot embossing and injection molding. Soft lithography, or polymer casting

is the most commonly used fabrication method for the fast prototyping of microfluidic

lab-on-chip setups in a laboratory setting owning to its short lead time, low cost and

low requirements to infrastructure.61–63 The typical materials for soft lithography are

UV sensitive photoresists and PDMS, as shown in Figure 1.20.

Figure 1.19 Schematic illustrations of (left) hot embossing and (right) injection molding.70

Figure 1.20 Main processes of soft lithography for SERS sensor fabrication, presented by S.Z.

Oo et al.71


20

The roll-to-roll (R2R) technique can be regarded as a modified cost-efficient and

compatible hot embossing process for the mass fabrication of microfluidic Raman

spectroscopy lab-on-chips.33,72,73 A. Habermehl et al. presented a custom-made R2R

embossing scheme where the master structures for microfluidic channels on the

pressing cylinders can be transferred to the PS foil by rolling and heating the

cylinders.33 They also introduced the aerosol jet printing approach that allows

deposition of Au nanoparticles for SERS functionality. The fabricated bottom layer

of the microfluidic SERS chip and the polymer lid can be sealed together by another

pair of rolling cylinders, as shown in Figure 1.21.

Figure 1.21 Fabrication process of a microfluidic SERS lab-on-chip by R2R approach,

presented by A. Habermehl el al.33

Fabrication of microfluidic lab-on-chips using laser related techniques have been

investigated with increasing interest recently. Laser cutting and engraving for

microfluidic channel fabrication can be easily completed by commercial laser cutting

machines.74 Femtosecond laser induced two-photon polymerization (2PP) is a novel

3D printing technique that can be used for micro- and nanostructures for microfluidic

lab-on-chips, as shown in Figure 1.22.32 Laser assisted wet etching is a powerful tool

for microstructure fabrication of glass or quartz to generate various optical elements

and microfluidic channels, and to delineate the edge of the glass element, as shown in

Figure 1.23.65

Chapter 1

21

Figure 1.22 Fabrication procedure of a 3D microfluidic SERS lab-on-chip by femtosecond-

laser direct writing, demonstrated by S. Bai et al.32

Figure 1.23 Schematics of (left) femtosecond laser writing and (right) femtosecond laser

assisted wet etching presented by A. Scott et al.65

The post-fabrication procedure, bonding and packaging, should also be considered for

microfluidic lab-on-chip fabrication in line with the material selected. Typically, the

fabricated lab-on-chip parts can be bonded via ultrasonic welding,57,58 laser welding,75

thermal bonding,62 plasma oxidation sealing61,63 and UV curing27.

1.4 Overview of this PhD

In this chapter we have introduced the concept of a microfluidic-based Raman

spectroscopy lab-on-chip in which bulky traditional Raman spectroscopy laboratory

processes are miniaturized and integrated into a single-chip, in combination with

microfluidic techniques for application domains such as material characterization and

clinical diagnostics. The microfluidic Raman spectroscopy lab-on-chip can

substantially reduce the sample consumption and lead time of experiments in terms of


22

the size and structure of the device. We have presented the tendency of the scientific

research in this field with statistical data of journal citations. We have briefly

discussed the steps we have followed for this PhD work and presented the objectives

of our research, including miniaturization and integration of LoCs as well as signal

enhancement, optimization, mass fabrication and their use in applications. We briefly

introduced the history and state-of-the-art of microfluidic Raman lab-on-chip systems.

Finally, the main considerations when developing a Raman-based LoC system are

given in terms of material selection and fabrication processes.

Chapter 2 starts by introducing Raman spectroscopy – electromagnetic radiation –

and the molecular vibrational modes from a classical harmonic oscillator picture. The

process of Raman scattering, where an external electromagnetic radiation interacts

with a vibrational molecule, is explained by introducing the ‘virtual energy states’

using the simplified Jablonski diagram. We also compare the Raman scattering

process with IR absorption and fluorescence in terms of selection rules, cross-sections,

sample applicability, etc. We introduce experimental equipment required for

obtaining a Raman signal. In addition, we introduce the history of surface-enhanced

Raman spectroscopy for acquiring highly sensitive Raman detection. The mechanisms

of SERS including the electromagnetic enhancement and chemical enhancement are

elaborated. We discuss the electromagnetic enhancement factor that measures the

degree of Raman signal boosting of a SERS substrate in terms of permittivity of the

metals and the effective distance. Moreover, we summarize the characteristics of the

commonly used SERS substrates and the fabrication approaches. At the start of this

PhD, SERS was a new topic in our research group, hence part of this chapter is

intended to help future researchers with tentative work on SERS.

Chapter 3 opens with an introduction to two-photon polymerization (2PP)

lithography by which 3D micro- and nanostructures can be printed via two-photon

absorption of the photoresist. In a next step, we discuss the feasibility of 3D printed

nanostructures for SERS applications by the Finite-Difference Time-Domain (FDTD)

approach to mimic the electromagnetic enhancement process. Then we build the

nominal shape modes and the voxel-based modes for the FDTD simulations. Next, we

fabricate the nanostructures by 2PP lithography, and the fabricated nanostructures are

coated with Au layers to make them SERS active. We characterize the surface profiles

of the nanostructures. The CAD models based on the surface data measured are built

and introduced to the FDTD simulations. We also conduct SERS experiments with

Rhodamine B solutions to access the enhancement factors of the nanostructures. As

benchmark, we conduct SERS experiments using some commercial SERS substrates.

The experimental results are compared with the FDTD simulation results, and show

good agreement. Moreover, we perform mycotoxin detection for the discrimination

Chapter 1

23

of Fumonisin b1 (FUM) and Deoxynivalenol (DON) with our 2PP printed SERS

substrates.

Chapter 4 starts by introducing our previously developed optofluidic Raman

spectroscopy setup. In a first step, we optimize the external optics of our previous

Raman spectroscopy system by miniaturizing and integrating the lenses and filters

into a fiber-based Raman probe. The internal tolerances and external tolerances of the

Raman probe are analyzed via a non-sequential ray tracing approach. We implement

a proof-of-concept demonstration setup of the Raman probe. In a next step, we

develop a fabrication process flow for the mass production of the freeform reflector

based microfluidic lab-on-chips by using double-sided hot embossing to reduce the

cost and complexity of Raman detection. The key considerations of this process,

including the material selection for the molds and chips, the structural design of the

molds and chips, the hot embossing parameters and the bonding methods, are

discussed. We conduct the confocal Raman experiments by using our mass fabricated

microfluidic lab-on-chip in combination with the fiber-based Raman probe we

optimized and analyze the experimental results.

In Chapter 5, we further improve our freeform reflector-based lab-on-chip Raman

spectroscopy setup into a tunable segmented freeform reflector-based microfluidic

Raman spectroscopy setup. We calculate the shape of the segmented freeform

reflector by means of numerical approaches. The shape of the segmented reflector, as

well as the flat polymer plate-based microfluidic lab-on-chip are introduced into non-

sequential ray tracing simulations to evaluate the confocality and tolerances of the

system. Next, we fabricate the segmented freeform reflector with ultra-precision

diamond tooling and implement a proof-of-concept demonstration setup. We perform

the confocal Raman measurements with our setup and analyze the experimental

results to access the tolerances and detection limit of our setup. Finally, the concepts

of microfluidic lab-on-chip and SERS that are respectively introduced and

investigated in the previous chapters, are combined in one microfluidic device. We

perform the microfluidic SERS measurements and discuss the experimental results

briefly. In addition to this chapter, we discuss the possibility of using a conical beam-

shaper in combination with an optimized segmented freeform reflector for SERS

measurements.

In Chapter 6, we present a Raman spectroscopy setup containing a conical beam

shaper in combination with a freeform segmented reflector for surface enhanced

Raman scattering (SERS) analysis. The freeform segmented reflector and the conical

beam shaper are designed by numerical approaches and fabricated by means of ultra-

precision diamond tooling. The segmented reflector has a numerical aperture of 0.984


24

and a working distance of 1mm for SERS measurements. We perform systematic

simulations using non-sequential ray tracing to assess the detecting abilities of the

designed SERS-based system. We implement a proof-of-concept setup and

demonstrate the confocal behavior by measuring the SERS signal of Rhodamine B

solution. The experimental results agree well with the simulations concerning the

misalignment tolerances of the beam shaper with respect to the segmented reflector

and the misalignment tolerances of the collecting fiber. In addition, we conduct

benchmark SERS measurements by using a 60× commercial objective lens with a

numerical aperture of 0.85. We find that the main Raman intensity of Rhodamine B

at 1503cm-1 obtained by our segmented reflector working together with the conical

beam shaper is approximately 30% higher compared to the commercial objective lens.

In Chapter 7, we give a summary of the original contributions in this PhD work and

describe the most promising directions of this PhD work for future research.

References

1. Friedrich, D. M. & Exarhos, G. J. Raman Enhancement Methods for

Molecular Structure Characterization Of Optical Thin Films. Thin Solid Films

54, 257–270 (1987).

2. Nagy, G., Feher, Z. & Pungor, E. Application of Silicone Rubber-based

Graphite Electrodes for Continuous Flow Measurements. Part II. Anal. Chim.

Acta 52, 47–54 (1970).

3. Pungor, E., Feher, Z. & Nagy, G. Application of Silicone Rubber-based

Graphite Electrodes for continuous flow measurements. Part I. Anal. Chim.

Acta 51, 417–424 (1970).

4. Ruzicka, J. & Hansen, E. H. Flow Injection Analyses: Part I. A New Concept

of Fast Continuous Flow Analysis. Anal. Chim. Acta 78, 145–157 (1975).

5. Reyes, D. R., Iossifidis, D., Auroux, P. & Manz, A. Micro Total Analysis

Systems . 1 . Introduction , Theory , and Technology. Anal. Chem. 74, 2623–

2636 (2002).

6. Goldstein, Y. et al. A Gas Chromatographic Air Analyzer Fabricated. IEEE

Trans. Electron Devices 26, 1880–1886 (1979).

Chapter 1

25

7. Geddes, J. J. & Hocker, G. B. Fiber Optic Temperature Sensor Using Liquid

Component Fiber. (1980).

8. Fabricius, N., Gauglitz, G. & Ingenhoff, J. A gas sensor based on an integrated

optical Mach-Zehnder interferometer. Sensors Actuators B Chem. 7, 672–676

(1992).

9. Baeyens, W. R. G., Schulman, S. G., Calokerinos, A. C. & Zhao, Y.

Chemiluminescence-based detection : principles and analytical applications

in flowing streams and in immunoassays. J. Pharm. Biomed. Anal. 17, 941–

953 (1998).

10. Kamholz, A. E., Weigl, B. H., Finlayson, B. A. & Yager, P. Quantitative

Analysis of Molecular Interaction in a Microfluidic Channel : The T-Sensor.

Anal. Chem. 71, 5340–5347 (1999).

11. Barry, B. & Mathies, R. Resonance Raman Microscopy of Rod and Cone

Photoreceptors. J. Cell Biol. 94, 479–482 (1982).

12. Dubessy, J., Poty, B. & Ramboz, C. Advances in C-O-H-N-S fluid

geochemistry based on micro-Raman spectrometry analysis of fluid

inclusions. Eur. J. Mineral. 1, 517–534 (1989).

13. Myrick, M. L., Angel, S. M. & Desiderio, R. Comparison of some fiber optic

configurations for measurement of luminescence and Raman scattering. Appl.

Opt. 29, 1333–1344 (1990).

14. Cooper, J. B. Remote Fiber-Optic Raman Analysis of Xylene Isomers in

Mock Petroleum Fuels Using a Low-Cost Dispersive Instrument and Partial

Least-Squares Regression Analysis. Appl. Spectrosc. 49, 586–592 (1995).

15. Lee, M. et al. Applicability of laser-induced Raman microscopy for in situ

monitoring of imine formation in a glass microfluidic chip. J. Raman

Spectrosc. 34, 737–742 (2003).

16. Barnes, S. E., Cygan, Z. T., Yates, J. K., Beers, K. L. & Amis, E. J. Raman

spectroscopic monitoring of droplet polymerization in a microfluidic device.

Analyst 131, 1027–1033 (2006).

17. Keir, R. et al. SERRS . In Situ Substrate Formation and Improved Detection

Using Microfluidics. Anal. Chem. 74, 1503–1508 (2002).

18. Docherty, F. T., Monaghan, P. B., Keir, R., Graham, D. & Ewen, W. The first


26

SERRS multiplexing from labelled oligonucleotides in a microfluidics lab-

on-a-chip. Chem. Commun. 118–119 (2004).

19. Ashok, P. C., Singh, G. P., Tan, K. M. & Dholakia, K. Fiber probe based

microfluidic Raman spectroscopy. Opt. Express 18, 7642–9 (2010).

20. Ashok, P. C., Singh, G. P., Rendall, H. A., Krauss, T. F. & Dholakia, K.

Waveguide confined Raman spectroscopy for microfluidic interrogation. Lab

Chip 11, 1262 (2011).

21. Neugebauer, U. et al. Diagnostics of tumor cells by combination of Raman

spectroscopy and microfluidics. Clin. Biomed. Spectrosc. Imaging II 8087,

80870J (2011).

22. Dochow, S. et al. Raman-on-chip device and detection fibres with fibre Bragg

grating for analysis of solutions and particles. Lab Chip 13, 1109 (2013).

23. Krafft, C. & Popp, J. The many facets of Raman spectroscopy for biomedical

analysis. Anal. Bioanal. Chem. 407, 699–717 (2015).

24. Watson, D. A. et al. A Flow Cytometer for the Measurement of Raman

Spectra. Cytom. Part A 73, 119–128 (2008).

25. Ngernsutivorakul, T. et al. Design and Microfabrication of a miniature fiber

optic probe with integrated lenses and mirrors for Raman and fluorescence

measurements. Anal Bioanal Chem 409, 275–285 (2017).

26. Bates, K. E. & Lu, H. Biophysical Perspective Optics-Integrated Microfluidic

Platforms for Biomolecular Analyses. Biophys. J. 110, 1684–1697 (2016).

27. De Coster, D. et al. Free-form optics enhanced confocal Raman spectroscopy

for optofluidic lab-on-chips. IEEE J. Sel. Top. Quantum Electron. 21, 1–9

(2015).

28. 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).

29. 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 (2005).

30. Pallaoro, A., Hoonejani, M. R., Braun, G. B., Meinhart, C. D. & Moskovits,

M. Rapid Identification by Surface-Enhanced Raman Spectroscopy of Cancer

Chapter 1

27

Cells at Low Concentrations Flowing in a Microfluidic Channel. ACS Nano

9, 4328–4336 (2015).

31. Liu, J., Devoe, D. L. & White, I. M. Microfluidic SERS Using a 3-

Dimensional Porous Monolith as a SERS-Active Solid Phase in a

Microchannel. OSA Conf. 1, 6–7 (2010).

32. Bai, S., Serien, D., Hu, A. & Sugioka, K. 3D Microfluidic Surface-Enhanced

Raman Spectroscopy ( SERS ) Chips Fabricated by All-Femtosecond-Laser-

Processing for Real-Time Sensing of Toxic Substances. Adv. Funct. Mater.

28, 1706262 (2018).

33. Habermehl, A. et al. Lab-on-chip, surface-enhanced Raman analysis by

aerosol jet printing and roll-to-roll hot embossing. Sensors 17, 1–11 (2017).

34. Focsan, M., Craciun, A. M., Astilean, S. & Baldeck, P. L. Two-photon

fabrication of three-dimensional silver microstructures in microfluidic

channels for volumetric surface-enhanced Raman scattering detection. Opt.

Mater. Express 6, 1587–1593 (2016).

35. Huang, T. & Xu, X.-H. N. Synthesis and characterization of tunable rainbow

colored colloidal silver nanoparticles using single-nanoparticle plasmonic

microscopy and spectroscopy. J. Mater. Chem. 20, 9867 (2010).

36. Shi, C. S. C. et al. Fiber surface enhanced Raman scattering (SERS) sensors

based on a double substrate ‘Sandwich’ structure. 2008 Conf. Lasers Electro-

Optics 2008 Conf. Quantum Electron. Laser Sci. 9–10 (2008).

37. Fan, M., Escobedo, C. R., Sinton, D. & Brolo, A. G. Surface-enhanced Raman

scattering (SERS) optrodes for multiplexed on-chip sensing of nile blue A and

oxazine. Lab Chip 12, 1554 (2012).

38. Pisco, M. et al. Nanosphere lithography for advanced all fiber Sers probes.

Proc. SPIE 9916, 99161S–2 (2016).

39. Pisco, M. et al. Nanosphere lithography for optical fiber tip nanoprobes. Light

Sci. Appl. 6, e16229 (2017).

40. Ren, K. N., Zhou, J. & Wu, H. Materials for Microfluidic Chip Fabrication.

Acc. Chem. Res. 46, 2396–2406 (2013).

41. Kim, M., Moon, B.-U. & Hidrovo, C. Enhancement of the thermo-mechanical

properties of PDMS molds for the hot embossing of PMMA microfluidic


28

devices. J. Micromechanics Microengineering 23, 095024 (2013).

42. Calin, M. A., Parasca, S. V., Savastru, R., Calin, M. R. & Dontu, S. Optical

techniques for the noninvasive diagnosis of skin cancer. J. Cancer Res. Clin.

Oncol. 139, 1083–1104 (2013).

43. Zimmerer, C. et al. Nondestructive characterization of the polycarbonate -

octadecylamine interface by surface enhanced Raman spectroscopy. Polym.

Test. 73, 152–158 (2019).

44. Pan, Z. et al. Depolymerization of polycarbonate with catalyst in autoclave

reactors. RSC Adv. 4, 19992–19998 (2014).

45. Serafim, A., Mallet, R., Pascaretti-grizon, F., Stancu, I. & Chappard, D.

Osteoblast-Like Cell Behavior on Porous Scaffolds Based on Poly(styrene)

Fibers. J. Biomed. Mater. Res. B 4, 609319 (2014).

46. Cai, D., Neyer, A., Kuckuk, R. & Heise, H. M. Raman , mid-infrared , near-

infrared and ultraviolet – visible spectroscopy of PDMS silicone rubber for

characterization of polymer optical waveguide materials. J. Mol. Struct. 976,

274–281 (2010).

47. Arnold, J. C. The effects of diffusion on environmental stress crack initiation

in PMMA. J. Mater. Sci. 33, 5193–5204 (1998).

48. Neogi, P. & Zahedi, G. Environmental Stress Cracking of Glassy Polymers.

Ind. Eng. Chem. Res. 53, 672–677 (2014).

49. Lau, A. Y., Lee, L. P. & Chan, J. W. An integrated optofluidic platform for

Raman-activated cell sorting. Lab Chip 8, 1116 (2008).

50. Zhang, Q., Lee, Y. H., Phang, I. Y., Lee, C. K. & Ling, X. Y. Hierarchical 3D

SERS substrates fabricated by integrating photolithographic microstructures

and self-assembly of silver nanoparticles. Small 10, 2703–2711 (2014).

51. Zahra, A. et al. Thermally Actuat ted Microfluidic System for Lab on Chip

Applications. in XVIII AISEM Annual Conference 1–4 (2015).

52. Bu, M., Melvin, T., Ensell, G. J., Wilkinson, J. S. & Evans, A. G. R. A new

masking technology for deep glass etching and its microfluidic application.

Sensors and Actuators 115, 476–482 (2004).

53. Iliescu, C., Chen, B. & Miao, J. On the wet etching of Pyrex glass. Sensors

and Actuators 143, 154–161 (2008).

Chapter 1

29

54. Koerner, T., Brown, L., Xie, R. & Oleschuk, R. D. Epoxy resins as stamps for

hot embossing of microstructures and microfluidic channels. Sensors

Actuators B 107, 632–639 (2005).

55. Goral, V. N., Hsieh, Y., Petzold, O. N., Faris, R. A. & Yuen, P. K. Hot

embossing of plastic microfluidic devices using poly(dimethylsiloxane)

molds. in 14th International Conference on Miniaturized Systems for

Chemistry and Life Sciences 1214–1216 (2010).

56. Gan, Z., Yu, Z., Chen, Z. & Chen, G. Hot embossing of electrophoresis

microchannels in PMMA substrates using electric heating wires. Anal.

Bioanal. Chem. 396, 2715–2720 (2010).

57. Viehrig, M. et al. Injection-Molded Micro fluidic Device for SERS Sensing

Using Embedded Au-Capped Polymer Nanocones. Appl. Mater. Interfaces 10,

37417–37425 (2018).

58. Alessandra, F. et al. Injection molded lab-on-a-disc platform for screening of

genetically modified E. coli using liquid-liquid extraction and surface

enhanced Raman scattering. Lab Chip 18, 869–877 (2018).

59. Matteucci, M. et al. Fiber-Based , Injection-Molded Optofluidic Systems :

Improvements in Assembly and Applications. Micromachines 6, 1971–1983

(2015).

60. Rivet, C., Lee, H., Hirsch, A., Hamilton, S. & Lu, H. Microfluidics for

medical diagnostics and biosensors. Chem. Eng. Sci. 66, 1490–1507 (2011).

61. Duffy, D. C., Mcdonald, J. C., Schueller, O. J. A. & Whitesides, G. M. Rapid

Prototyping of Microfluidic Systems in Poly ( dimethylsiloxane ). Anal. Chem.

70, 4974–4984 (1998).

62. Perozziello, G. et al. Microfluidic device for continuous single cells analysis

via Raman spectroscopy enhanced by integrated plasmonic nanodimers. Opt.

Express 24, 180–190 (2016).

63. Lu, H. et al. Highly uniform SERS-active microchannel on hydrophobic

PDMS : a balance of high reproducibility and sensitivity for detection of

proteins †. RSC Adv. 7, 8771–8778 (2017).

64. Lin, B., Yang, Y., Ho, C., Yang, H. & Wang, H. A PDMS-Based Cylindrical

Hybrid Lens for Enhanced Fluorescence Detection in Microfluidic Systems.

Sensors 14, 2967–2980 (2014).


30

65. Scott, A. J. et al. Optical microsystem for analyzing engine lubricants. Proc.

SPIE 5590, 122–127 (2004).

66. Hnatovsky, C. et al. Polarization-Selective Etching in Femtosecond Laser-

Assisted Microfluidic Polarization-selective etching in femtosecond laser-

assisted microfluidic channel fabrication. Opt. Lett. 30, 1867–1869 (2005).

67. Malek, C. K. et al. Femtosecond laser machining and lamination for large-

area flexible organic microfluidic chips. Eur. Phys. Journal-Applied Phys. 46,

(2010).

68. Jenkins, G., Wang, Y., Lei, Y. & Qiong, X. Printed electronics integrated with

paper-based microfluidics : new methodologies for next-generation health

care. Microfluid. Nanofluidics 19, 251–261 (2015).

69. Wu, J. & Min Gu. Microfluidic sensing : state of the art fabrication and. J.

Biomed. Opt. 16, 080901–1 (2011).

70. What is injection moulding. A Plus Plastics (2019). Available at:

https://aplusplastics.com.au/about/what-is-injection-moulding/.

71. Oo, S. Z. et al. Disposable plasmonic plastic SERS sensor. Opt. Express 21,

18484 (2013).

72. Dixon, C., Ng, A. H. C., Fobel, R., Miltenburg, M. B. & Wheeler, A. R. An

inkjet printed, roll-coated digital microfluidic device for inexpensive,

miniaturized diagnostic assays. Lab Chip 16, 4560–4568 (2016).

73. Lewis, G. G., Ditucci, M. J., Baker, M. S. & Phillips, S. T. High throughput

method for prototyping three-dimensional, paper-based microfluidic devices.

Lab Chip 12, 2630–2633 (2014).

74. Focke, M., Kosse, D., Claas, M., Reinecke, H. & Von, F. Lab-on-a-Foil :

microfluidics on thin and flexible films. Lab Chip 10, 1365–1386 (2010).

75. Pelsmaeker, J. De et al. Clear to clear laser welding for joining thermoplastic

polymers: a comparative study based on physicochemical characterization. J.

Mater. Process. Technol. 255, 808–815 (2017).

Chapter 2

31

Chapter 2

2 Introduction to Raman and Surface-Enhanced

Raman Spectroscopy

2.1 Introduction to electromagnetic radiation

Spectroscopy began with the discovery of the dispersion of the visible (VIS) light by

using a glass prism. It is the study of the interaction of electromagnetic radiation with

matter1 to qualitatively and quantitatively identify the substances and characterize

their chemical compositions, atomic/molecular structures or relative content.

Spectroscopy can mainly be divided into absorption spectroscopy, emission

spectroscopy and scattering spectroscopy according to the principle of analytical

methods. It can also be classified into atomic spectroscopy and molecular

spectroscopy according to the morphology of the sample analytes. Actually, human

beings are performing spectroscopic analyses all the time in their life, for example the

human eyes, working as detectors, are constantly receiving and processing visible

light emitted, scattered or reflected by objects. However, visible light counts only a

small section of the electromagnetic (EM) radiation band. All electromagnetic

radiations are transverse waves consisting of electric field 𝑬 and magnetic field 𝑩 that

are vibrating perpendicular to each other and the vibration of each field is

perpendicular to the direction of propagation.2

The electric and magnetic fields are mainly characterized by three properties, namely

the amplitude, periodicity and the phase. As the electric/magnetic vectors and the

propagation vector are orthogonal, an EM radiation can be expressed by the wave

equation of either the electric field or its magnetic field. Normally, the propagation of

the electric field and magnetic field are given respectively by3

(𝜐𝑝ℎ2 ∇2 −

𝜕2

𝜕𝑡2) 𝑬 = 0 （2.1）

Chapter 2 Introduction to Raman and Surface-Enhanced Raman Spectroscopy

32

(𝜐𝑝ℎ2 ∇2 −

𝜕2

𝜕𝑡2) 𝑩 = 0 （2.2）

Where, 𝜐𝑝ℎ =1

√𝜇𝜀 is the speed of light in a medium with permeability 𝜇 and

permittivity 휀.

For a monochromatic wave, the EM radiation is given by the sinusoidal solution

𝑬(𝒓, 𝑡) = 𝑬𝟎 cos(𝜔𝑡 − 𝒌 ∙ 𝒓 + 𝜑0) （2.3）

𝑩(𝒓, 𝑡) = 𝑩𝟎 cos(𝜔𝑡 − 𝒌 ∙ 𝒓 + 𝜑0) （2.4）

Where, 𝒓 = (𝑥, 𝑦, 𝑧) is the position vector, t is the time, 𝜔 is the angular frequency in

radians, 𝒌 is the wave vector and 𝜑0 is the initial phase angle in radians. |𝑬𝟎| and

|𝑩𝟎| are the amplitude of the electric and magnetic fields which refers to the

maximum amount of displacement the wave makes away from its propagation axis.

The absolute value of wave vector is related to the frequency of the wave:

𝑘 = |𝒌| =𝜔

𝑐=

2𝜋

𝜆 （2.5）

Here 𝑐 is the propagation speed of the wave in vacuum, which is approximately 3 ×

108 m/s. The wavelength 𝜆 is a measurement of the periodicity refering to the distance

between any two consecutive points with the same phase, such as two successive

troughs or crests. The periodicity can also be characterized by the frequency of the

oscillation 𝜈𝑠 or wavenumber 𝜈, which are defined as the number of waves that passes

through a particular point in a unit time and the number of complete wavelengths

contained in a unit length, respectively. Figure 2.1 illustrates the electric and magnetic

fields oscillation in fixed planes in free space.

𝜈𝑠 =𝑐

𝜆 （2.6）

𝜈 =1

𝜆 （2.7）

Chapter 2

33

Figure 2.1 The propagation of radiation.2

In the quantum mechanics theory, the electromagnetic radiation is also described as

photons with certain electromagnetic energy. The energy of each photon 𝐸 is

proportional to its frequency 𝜈:

𝐸 = ℎ𝜈 （2.8）

Where, ℎ is Planck’s constant. A weak ray of electromagnetic radiation can be

regarded as a few photons, while an intense ray consists of a huge number of photons.

Regions of the spectrum

Table 2.1 The electromagnetic spectrum.3

γ-ray X-ray UV Visible Infrared Microwave Radio

𝜆/m 10-11 10-9 10-7 10-6 10-4-10-5 10-3 10-2-10

𝜈/Hz 1019 1017 1015 1014 1012-1013 1011 106-107

The entire electromagnetic radiation band is typically divided into various regions of

the spectrum depending upon its wavelength or frequency as shown in Table 2.1. Each

region is corresponding to a specific energy level of the molecule or atom that interacts

with the light.4

The regions in ascend order of wavelength and the assignments with respect to matter

are:

1. Gamma-ray (γ-ray) radiation: energy changes arise from the nuclear decay or other

nuclear and subnuclear processes.

2. X-ray radiation: energy changes arise from the highly energetic inner atomic

electrons.


34

3. Ultraviolet radiation (UV): energy changes involve the excitation of molecular and

atomic valence electrons.

4. Visible radiation: energy changes involve the excitation of molecular electrons and

plasma oscillations.

5. Infrared radiation (IR): energy changes involve molecular vibration and rotation.

This region can be further divided into the near-IR, mid-IR and far-IR bands.

6. Microwave radiation: energy changes involve molecular rotation.

7. Radio waves: energy changes involve the reversal of nucleus or electron spin.

The boundaries between different regions are not precisely defined since they may

fade into each other just like the bands of the rainbow. The atomic/molecular process

associated with each region is intrinsically different regarding the methodologies and

equipment of the spectroscopic analysis, types of matter under test, etc. Therefore,

spectroscopy can be classified into spectroscopy of relative regions, such as X-ray

spectroscopy, UV spectroscopy or IR spectroscopy, taking into account the band of

the spectrum. In this dissertation, we focus specifically on visible and IR spectroscopy

induced by the vibrational modes of molecules, called Raman spectroscopy (named

after Indian physicist C. V. Raman3).

2.2 Introduction to Raman spectroscopy

2.2.1 Diatomic molecule: Harmonic oscillator

An isolated molecule has various forms of energy due to its different types of motions

and intramolecular interactions. The total energy of a molecule 𝐸𝑡𝑜𝑡𝑎𝑙 is the sum of

the constituent energies3

𝐸𝑡𝑜𝑡𝑎𝑙 = 𝐸𝑡𝑟𝑎𝑛𝑠 + 𝐸𝑟𝑜𝑡 + 𝐸𝑣𝑖𝑏 + 𝐸𝑒𝑙 + ⋯ （2.9）

Where 𝐸𝑡𝑟𝑎𝑛𝑠 is the translational energy due to the motion of the molecule as a whole.

𝐸𝑟𝑜𝑡 is the rotational energy by virtue of the rotation of the molecule around its center

of gravity. 𝐸𝑣𝑖𝑏 is the vibrational energy by virtue of the periodic displacement of the

atoms away from the equilibrium position. 𝐸𝑒𝑙 is the potential and kinetic energy of

the electrons. The molecule also possesses nuclear energy and other forms of

subnuclear energies. The research area in this dissertation, Raman spectroscopy, is the

technology based on the Raman scattering which arises from the interaction of

electromagnetic radiation with vibrations of molecules.

Chapter 2

35

To better understand the vibrational modes of the molecule, we will first discuss the

simplest diatomic molecule consisting of two atoms A-B. The diatomic molecule can

be described as a harmonic oscillator5, as shown in Figure 2.2.

Figure 2.2 Ball and spring model of a diatomic molecule.

The chemical bond between two atoms can be considered as a spring that obeys

Hook’s law with a spring constant 𝐾 for limited dislocations of the atoms. Assume

that the displacements of the two atoms from their equilibrium position are 𝛿𝑥1

and 𝛿𝑥2 , respectively. The restoring force 𝑓 to the displacement 𝑞 from the

equilibrium position of the molecule is

𝑓 = −𝐾𝑞 （2.10）

Here, 𝑞 = 𝛿𝑥1 + 𝛿𝑥2. For the displacement of the two atoms, we have

𝛿𝑥1 ∙ 𝑚1 = 𝛿𝑥2 ∙ 𝑚2 （2.11）

The potential energy 𝐸 of the molecule is then given by

𝐸 = − ∫ 𝑓 𝑑𝑞 = ∫ 𝐾𝑞 𝑑𝑞 =1

2𝐾𝑞2 （2.12）

According to Newton’s second law, we obtain the equation of the harmonic motion

𝑚𝑑2𝑞

𝑑𝑡2= −𝐾𝑞 （2.13）

Where,

𝑚 =𝑚1𝑚2

𝑚1 + 𝑚2

（2.14）

The harmonic motion equation (2.13) has a general solution3

𝑞 = 𝑞0 cos 2𝜋𝜈0𝑡 （2.15）


36

Where 𝑞 is the displacement, 𝑞0 is the amplitude of the vibration that refers to the

maximum extension or contraction of the displacement. The fundamental frequency

of the harmonic oscillator is given by

𝜈 =1

2𝜋√

𝐾

𝑚 （2.16）

According to the potential well equation (2.12), the vibrational energy of the molecule

only depends on the force constant 𝐾 and the displacement 𝑞, apparently any energy

value could be possible. However, the quantum theory states that the interval between

any two adjacent levels is ℎ𝜈 . Therefore, the vibrational energy of a harmonic

oscillator in discrete potential energy level is given by6

𝐸𝑣 = (𝑉 +1

2) ℎ𝜈 = (𝑉 +

1

2) ℎ𝑐𝜈 （2.17）

Where, 𝑉 (𝑉 = 0, 1, 2, …) is the vibrational quantum number. The vibrational energy

levels of a quantum mechanical harmonic oscillator are depicted in Figure 2.3.

Figure 2.3 Potential well of a diatomic harmonic oscillator (dashed red) and the Morse

potential (solid blue) that better approximate the vibrational energy of an actual molecule.

(Figure adapted from en.wikipedia.org/wiki/Morse_potential)

In reality, the potential energy of any diatomic molecule is much more complex than

that of a simple harmonic oscillator. For instance, the displacement of the atoms with

larger value will result in the dissociation of the molecule, therefore there should be a

threshold energy at least. Figure 2.3 also compares the vibrational potential energy of

Chapter 2

37

the diatomic molecule in the classical harmonic oscillator and the actual vibrational

potential energy stated empirically by the Morse function.6

𝐸(𝑟) = 𝐷𝑒[1 − exp(−𝑎(𝑟 − 𝑟𝑒))]2 （2.18）

Where 𝐷𝑒 refers to the dissociation energy of the molecule. 𝑎 is a constant as a

measurement of the internuclear bond in a specific electronic state. It controls the

‘width’ of the potential.

The vibrational energy level of a harmonic oscillator is given by the Schrodinger

approximation5,7 in terms of the frequency 𝜈𝑒 and the anharmonicity constant 𝑥𝑒.

𝐸 = (𝑉 +1

2) ℎ𝜈𝑒 − (𝑉 +

1

2)

2

ℎ𝜈𝑒𝑥𝑒 （2.19）

The potential well of an actual diatomic molecule increases faster when the

internuclear distance decreases from the equilibrium position than that of the classic

harmonic oscillator. On the other hand, the potential energy tends to a constant value

asymptotically at large internuclear distances. The curve of the actual potential well

is asymmetric compared to the harmonic oscillator.

2.2.2 Diatomic molecule under an external electromagnetic field

Assume that a diatomic molecule is located in an electromagnetic field 𝐸 with angular

frequency 𝜔

𝑬 = 𝑬𝟎 cos 2𝜋𝜔𝑡 （2.20）

The electromagnetic radiation will result in the oscillation of dipoles inside the

molecule. The induced dipole moment 𝑷 is given by

𝑷 = 𝛼𝑬 = 𝛼𝑬𝟎 cos 2𝜋𝜔𝑡 （2.21）

Where 𝛼 is the polarizability tensor depending on the displacement of the atoms given

by

𝛼 = 𝛼0 +𝑑𝛼

𝑑𝑞𝑞 + ⋯ （2.22）

As discussed before, the displacement of atoms is given by


38

𝑞 = 𝑞0 cos 2𝜋𝜈0𝑡 （2.23）

Therefore, the dipole moment 𝑷 can be expressed as

𝑷 = 𝛼0𝑬𝟎 cos 2𝜋𝜔𝑡 +𝑑𝛼

𝑑𝑞𝑞𝑬𝟎 cos 2𝜋𝜔𝑡 cos 2𝜋𝜈0𝑡

= 𝛼0𝑬𝟎 cos 2𝜋𝜔𝑡

+1

2𝑬𝟎𝑞0

𝑑𝛼

𝑑𝑞[cos 2𝜋(𝜔 − 𝜈0)

+ cos 2𝜋(𝜔 + 𝜈0)]

（2.24）

We find that there are three frequency components related to the dipole moment,

which are

𝑷 = 𝑷𝒓(𝜔) + 𝑷𝒔(𝜔 − 𝜈0) + 𝑷𝒂𝒔(𝜔 + 𝜈0) （2.25）

The first component represents the Rayleigh scattering, which is an elastic process of

the interaction between incident light and the molecule.8 The frequency of the

Rayleigh scattering is identical with that of the incident electromagnetic radiation.

Rayleigh scattering is the dominant scattering when a molecule interacts with the

electromagnetic radiation.

The second and third terms represent Raman scattering - where the reemitted photons

have different frequencies from the incident radiation - which is an inelastic process.

The process of the second component is called Stokes scattering. When the molecule

interacts with the incident photons, the molecule is excited from the ground

vibrational state to a virtual energy state. The molecule is then falling back instantly

from the virtual energy state to the vibrational energy state but with a higher energy

level. Therefore, the Stokes scattering possesses a lower frequency (𝜔 − 𝜈0). If the

molecule falls back to the vibrational energy state with a lower energy level, the

scattering is called anti-Stokes scattering whereby the re-emitted photon shifts its

frequency to (𝜔 + 𝜈0), as stated by the third term of the equation. The processes of

Rayleigh scattering, Stokes and Anti-Stokes scattering are demonstrated in Figure 2.4.

Figure 2.5 shows the energy levels of Rayleigh scattering and Raman scattering. The

virtual energy state is used to explain the quantum state of Raman scattering because

the absorption and re-emitting of the photons occur almost simultaneously. The

excited-state lifetime of Raman scattering usually lasts less than a picosecond9.

Chapter 2

39

Figure 2.4 A molecule interacting with an incident photon results in Rayleigh scattering,

Raman Stokes and Anti-Stokes scattering.

Figure 2.5 Simplified Jablonski diagram of different transition processes occurring upon the

interaction of an incident photon with a molecule. ‘Virtual energy states’ do not exist in

reality, but they are used to describe the Raman scattering. In reality the electron falls back to

the ground instantaneously.

If the frequency of the incident photon matches the specific transition energy from the

ground vibrational state to a higher vibrational state, the molecule will be excited to

the higher vibrational state directly, and the photon will be absorbed by the molecule

without re-emitting. This process is denoted as infrared absorption (IR absorption),

since it mainly happens when the incident electromagnetic radiation is in the infrared

region.10 The IR absorption spectrum is typically divided into three regions, namely

the near infrared region (0.75~2.5µm), mid-infrared region (2.5~25µm) and far

infrared region (25~300µm). The mid-infrared absorption associated with the

fundamental vibrations of most organic and inorganic molecules, is widely utilized in


40

various research areas. The mid-IR absorption occurs by overtone or harmonic

vibrations of the fundamental vibrations. The far-infrared absorption lying adjacent to

the microwave region is associated with the rotational modes of most molecules.

Resonance Raman scattering occurs when the excitation energy is tuned to a specific

electronic transition of a molecule from the ground vibrational state to the high-level

electronic state.11 The Raman scattering sometimes happens together with

fluorescence, which is another type of absorption and re-emitting process of photons

with molecules. The basic difference between Raman scattering and fluorescence is

related to the intermediate states and the time scale involved. Unlike in the Raman

scattering that the molecule is excited to a virtual state, the molecule will be at the

electronic energy state first for fluorescence. The frequency of the re-emitted photon

is close to the transition energy between two electronic states. The excited energy

states for fluorescence typically last more than one nanosecond.12

2.2.3 Vibrational modes of molecules

In reality, only a small portion of matter consists of diatomic molecules. Even the

simplest protein in a human body, insulin, is composed of 51 amino acids with 788

atoms in total. The deoxyribonucleic acid molecule can be composed of billions of

atoms. The vibrational modes of polyatomic molecules are much more complex than

that of the diatomic molecules, but we can extend the harmonic oscillator model of

the diatomic molecules to understand the vibrational modes of polyatomic molecules.

For a nonlinear molecule containing n atoms, there are 3n-6 degrees of vibrational

freedom, in which the internuclear distance between different atoms change.13 A

linear molecule has 3n-5 degrees of vibrational freedom.4 For example, CH4, as a non-

linear molecule possesses (3×5)-6=9 degrees of vibrational freedom. These degrees

of vibrational freedom can be classified into six basic categories, which are bond

stretching, angle deformation, rocking, wagging, out of plane deformation, and

twisting, as shown in Figure 2.6.

Chapter 2

41

Figure 2.6 Schematic diagram of different vibrational modes of a -CH2 group. The yellow ball

represents the carbon atom and the blue ball represents the hydrogen atom (Figure adapted

from https://en.wikipedia.org/wiki/Molecular_vibration)

a. Bond stretching

Bond stretching is the simplest and general vibrational mode of any molecule. It

denotes the changing in the bond length of a diatomic group during a vibration. The

stretching vibrations are generally denoted by the symbol 𝑣.

The stretching mode can be divided into symmetrical stretching and asymmetrical

stretching depending on the direction of the displacement vectors with respect to the

terminal atom. In a cyclic system such as benzene or pyridine, the symmetrical

expansion or contraction of the bonds may be referred to as a ring stretching mode.14

Figure 2.7 Scheme of the ring breathing vibration.

b. Scissoring (in plane bending, angle deformation or symmetrical bending)

Angle deformation denotes the change of the interbond angle. The displacement

vectors of angle deformation at the terminal atoms are perpendicular to the bond. The

symbol 𝛿 is typically used to indicate the angle deformation of the molecules.


42

c. Rocking (Asymmetrical bending)

The rocking represents the moves of a chemical group as a whole with respect to

another atom. Rocking mode is denoted by the symbol 𝑟.

d. Wagging (symmetrical out-of-plane bending)

The wagging mode represents the change in angle between a bond and a plane, or the

folding of a plane about a line through it. Wagging vibration is typically denoted by

the symbol 𝜔.

f. Twisting (asymmetrical out-of-plane bending)

Twisting mode denotes the restricted rotation of a chemical group with respect to the

overall molecular skeleton, and is indicated by the symbol 𝜏 . An example is the

rotation of -CH3 group in ethanoic acid along the C-C axis.

e. Out of plane bending

Out of plane bending is another basic vibration mode in molecules with multiple

atoms. It represents the movement of an atom perpendicular to the molecular plane as

shown in Figure 2.8. For a planar cyclic molecule with n atoms, there are n-3 out of

plane bending vibrational modes.

Figure 2.8 Schematics of out of plane bending. The yellow, green and blue balls represent

different types of atoms.

Because of the compexity of the vibrational modes of many organic macromolecules,

it is clear that scissoring, rocking, wagging and twisting vibrations are always summed

up as deformation vibrations in literature. Each vibrational mode has its specific

frequency and energy, which corresponds to certain frequency shifts of Raman

scattering. Therefore, Raman scattering can be ultilized to determine the structure of

a molecule. This is also the fundamental principle of Raman spectroscopic analysis.

Table 2.2 lists some typical vibrational modes with respect to the frequency shift of

Chapter 2

43

Raman scattering.15 The vibrational frequency of the same bond can be different for

different molecules. A structural change in the molecule is normally the cause of

frequency shifts, but changes in temperature and stress within the sample may also

affect the band position and shape. The shorter bond length causes a shift towards a

higher wavenumber or vice versa. If a Raman band is sufficiently narrow, the peak

position can be seen to shift with temperature. As the temperature increases, the bond

length will increase and consequently one can expect a decrease in the energy of the

vibrational mode based on a Boltzmann distribution of the ground and first excited

state populations.

Table 2.2 Vibrational modes of some chemical bonds and their spectra intensity.15

Vibrational modea Frequency

(cm-1)

Intensityb

Raman IR

𝜐(𝑂 − 𝐻) 3650 – 3000 w s

𝜐(𝑁 − 𝐻) 3500 – 3300 m m

𝜐(−𝐶 − 𝐻) 3000 – 2800 s s

𝜐(= 𝐶 − 𝐻) 3100 – 2800 s m

𝜐(≡ 𝐶 − 𝐻) 3300 w s

𝜐(−𝑆 − 𝐻) 2600 – 2500 s w

𝜐(𝐶 ≡ 𝑁) 2255 – 2220 m-s 0-s

𝜐(𝐶 ≡ 𝐶) 2250 – 2100 vs 0-w

𝜐(𝐶 = 𝑂) 1820 – 1680 w-s vs

𝜐(𝐶 = 𝑂) 1680 – 1610 s m

𝜐(𝑁 = 𝑁) aliph. derivat 1580 – 1550 m 0

𝜐(𝑁 = 𝑁) arom. derivat 1440 – 1410 m 0

𝜐𝑎((𝐶 −)𝑁𝑂2) 1590 – 1530 m s

𝜐𝑠((𝐶 −)𝑁𝑂2) 1380 – 1340 vs m

𝜐𝑎((𝐶 −)𝑆𝐶2(−𝐶)) 1350 – 1310 0-w s

𝜐𝑠((𝐶 −)𝑆𝑂2(−𝐶)) 1160 – 1120 s s

𝜐((𝐶 −)𝑆𝑂2(−𝐶)) 1070 – 1020 m s

𝜐(𝐶 = 𝑆) 1250 – 1000 s w

𝛿(𝐶𝐻2), 𝛿𝑎(𝐶𝐻3) 1470 – 1400 m m

𝛿𝑠(𝐶𝐻3) 1380 w-m m-s

𝜐(𝐶 − 𝐶) 𝑎𝑟𝑜𝑚𝑎𝑡𝑖𝑐𝑠 1600, 1580

1500, 1450

m-s

w-m

m-s

m-s

𝜐(𝐶 − 𝐶) 𝑎𝑙𝑖𝑐𝑦𝑐𝑙𝑒𝑠

𝑎𝑙𝑖𝑝ℎ𝑎𝑡𝑖𝑐 𝑐ℎ𝑎𝑖𝑛𝑠

1300 – 600 m-s w-m

𝜐𝑎(𝐶 − 𝑂 − 𝐶) 1150 – 1060 w s

𝜐𝑠(𝐶 − 𝑂 − 𝐶) 970 – 800 m-s 0-w

𝜐𝑎(𝑆𝑖 − 𝑂 − 𝑆𝑖) 1110 – 1000

550 – 450

0-w

vs

vs

0-w

𝜐(𝑂 − 𝑂) 900 – 845 s 0-w

𝜐(𝑆 − 𝑆) 550 – 430 s 0-w


44

𝜐(𝑆𝑒 − 𝑆𝑒) 330 – 290 s 0-w

𝜐(𝐶 − 𝑆) 𝑎𝑟𝑜𝑚𝑎𝑡𝑖𝑐𝑠 1100 – 1080 s s-m

𝜐(𝐶 − 𝑆) 𝑎𝑙𝑖𝑝ℎ𝑎𝑡𝑖𝑐 790 – 630 s s-m

𝜐(𝐶 − 𝐶𝑙) 800 – 550 s s

𝜐(𝐶 − 𝐵𝑟) 700 – 500 s s

𝜐(𝐶 − 𝐼) 660 – 480 s s

(a) 𝜐: streching, 𝛿: bending/deformation, 𝜐𝑠: symmetrical stretching, 𝜐𝑎: asymmetrical

stretching, 𝛿𝑠: symmetrical bending/deformation, 𝛿𝑎: asymmetrical

bending/deformation.

(b) vs: very strong, s: strong, m: medium, w: weak, 0: very weak or inactive

2.2.4 Selection rules of Raman scattering and IR absorption

The molecular vibrational theory reveals that the Raman and IR absorption arising is

determined by selection rules.1 The selection rule for IR absorption suggests that if

the oscillations of the atoms in a molecule under external electromagnetic radiation

result in a periodic alternating of the electric and magnetic dipole moments, the

molecule is IR active. If the oscillation of the atoms in a molecule under external

electromagnetic radiation results in changes of the polarizability tensor, the molecule

is Raman active. The changes of the polarizability tensor are associated with the

distortion of electron cloud geometry of a molecule with respect to the displacement

of the atoms from their equilibrium positions.

Figure 2.9 shows the stretching vibration of a diatomic molecule and the changes of

the dipole moment.

Figure 2.9 Stretching vibration of a diatomic molecule and the related fluctuating dipole

moment.3

We take the triatomic molecule CO2 with a center of symmetry as an example, as

shown in Figure 2.10. According to the 3n-5 rule4 for a linear molecule, there are four

Chapter 2

45

fundamental vibrational modes, namely symmetrical stretching, asymmetrical

stretching, in plane bending, out of plane bending v1, v2, v3, v4, respectively, where v3

and v4 are coincident with each other. When the CO2 molecule vibrates at v1, the

shapes of the electron cloud at the inverse locations are different. Therefore, CO2 is

Raman active at v1 because its polarizability changes during this vibrational motion.

In contrast, the charge distributions of CO2 are the same at the inverse dislocations,

but the dipole moments are different. Therefore, the CO2 molecule is IR active at v2,

v3 and v4. This is a special case of the so-called principle of mutual exclusion, where

in molecules with a center of symmetry, the vibrational modes are Raman-active but

are IR-inactive and vice versa. In molecules with different elements of symmetry,

certain vibrational modes may be active in Raman, IR, both or neither. For molecules

with no symmetry, all the normal vibrational modes are active in both IR and Raman.

A general rule is that the symmetric modes have a higher response in Raman scattering,

whereas the asymmetric modes usually have lower intensities in Raman scattering.

Figure 2.10 Changes of the dipole moment or the polarizability of the CO2 molecule during

vibrations. At the symmetrical stretching mode v1, there is a change in the polarizability but

no change in the dipole moment, the molecule is Raman active and IR inactive. At the

asymmetrical stretching mode v2, there is a change in dipole moment, so the molecule is IR

active but Raman inactive. At the deformation modes v3 and v4, there are only changes in

dipole moment, so the molecule is Raman inactive.

2.2.5 Raman spectrum

Raman scattering can be detected by a spectrum analyzer and plotted as Raman

spectrum. The full pattern of a spectrum typically consists of an excitation line and a

set of lines in the lower and higher frequencies. An excitation line is the result of the

Rayleigh scattering that has a peak far more intense than the other lines, the lines in


46

the lower and higher frequencies refer to the Anti-Stokes and Stokes Raman scattering.

The frequencies, or the Raman shift can be calculated by the following equation.16

𝑅𝑎𝑚𝑎𝑛 𝑆ℎ𝑖𝑓𝑡 [𝑐𝑚−1] =107

𝜆𝑒𝑥[𝑛𝑚]−

107

𝜆𝑟[𝑛𝑚] （2.26）

Where 𝜆𝑒𝑥 is the wavelength of the excitation radiation, and 𝜆𝑟 is the wavelength of

the Raman scattering.

The anti-Stokes lines are a mirror of the Stokes lines but the intensities are much lower

than those of the Stokes lines, as shown in Figure 2.11. Therefore, Stokes Raman lines

are preferably utilized to analyze the molecular vibration modes. Figure 2.12 shows

the energy level diagram of all the lines appearing in Figure 2.11.

Figure 2.11 Complete Raman spectrum of carbon tetrachloride (CCl4) excited by a 488nm

wavelength laser.17

Figure 2.12 Jablonski diagram showing the origin of Stokes Raman and Anti-Stokes Raman

lines in the Raman spectrum of CCl4.

Chapter 2

47

2.2.6 Raman cross section

The scattering cross section is the measurement of the scattering efficiency of a

particular molecule.18 It represents the equivalent area that a molecule would absorb

and convert to scattered photons with an intensity of 𝐼𝑚𝑛 under a given uniform

irradiance of 𝐼0

𝜎𝑚𝑛 =

𝐼𝑚𝑛

𝐼0

=27𝜋5

32𝜈0(𝜈0 − 𝜈𝑚𝑛)3𝑔𝑓(𝑇) |∑ 𝛼𝜌𝜎(𝜈0)

𝜌𝜎

|

2

（2.27）

Where, 𝐼𝑚𝑛 ( photons/s ∙ cm2 ) refers to the intensity of the integrated Raman

scattering over 4𝜋 steradians integrated over the bandwidth of the vibrational

transition from state 𝑚 to state 𝑛 . 𝜈0 (cm−1) is the frequency of the incident

excitation, while 𝜈𝑚𝑛 refers to the vibrational frequency of the molecule

corresponding to the specific transition. 𝑔 is a factor corresponding to the degeneracy

of the initial state 𝑚. 𝑓(𝑇) is the temperature dependent Boltzmann weighting factor

specifying the thermal occupancy of the initial state 𝑚. 𝛼𝜌𝜎 is the average 𝜌𝑡ℎ and

𝜎𝑡ℎ (𝜌, 𝜎 = 𝑥, 𝑦, 𝑧) component of the Raman polarizability tensor over all orientations

of the molecule for the excitation frequency of 𝜈0. For the detection with a non-photon

counting detector, such as a Fourier Transform (FT) spectroscopy, the term

𝜈0(𝜈0 − 𝜈𝑚𝑛)3 is replaced with (𝜈0 − 𝜈𝑚𝑛)4.

From the equation (2.27) we observe that the Raman cross section approximately

scales with the frequency of the incident excitation to the fourth power. Therefore,

reducing the wavelength of the excitation beam will generate a higher Raman response.

For example, excitation at 532nm can induce approximately five times more Raman

scattering than excitation at 785nm. This behavior is similar to Rayleigh scattering,

but note that the fluorescence will also be increased as the frequency of the incident

excitation reduces.

The Raman scattering cross section of a sample molecule can be calculated by

comparing the intensity of a Raman band in an integrated area with that of a reference

molecule19

𝜎𝑅𝑆 = 𝜎𝑅

𝑟 (𝐼𝑆

𝐼𝑟) [

𝐸(𝜈0 − 𝜈𝑗𝑟)

𝐸(𝜈0 − 𝜈𝑗𝑆)

](𝑐𝑟

𝑐𝑆) （2.28）

Where 𝜎𝑅 is the Raman cross section of a molecule. 𝐼 refers to the integrated area of

the Raman band. 𝐸(𝜈0 − 𝜈𝑗) is the spectrometer efficiency at the specified Raman


48

shift frequency of 𝜈0 − 𝜈𝑗 . 𝑐 refers to the concentration of the sample molecule.

Superscripts 𝑟 and 𝑠 represent the reference and sample, respectively.

Table 2.3 A comparison of cross sections between different types of spectroscopy.

Transition of the spectroscopy Typical absolute cross section

Electronic (UV-VIS) absorption20 10-20 m2

Fluorescence21 10-20 m2

Vibrational (IR) absorption22 10-23 m2

Resonance Raman scattering23 10-29 m2

Non-resonant Raman scattering19,24 10-33 m2

Surface Enhanced Raman Scattering25 10-21 m2 to 10-27 m2

2.2.7 Pros and Cons of Raman spectroscopy

Raman spectroscopy is sensitive to molecules with homo-nuclear bonds. For example,

C-C, C=C and C≡C bonds can be easily distinguished by Raman spectroscopy.10 IR

spectroscopy, on the other hand, is sensitive to the vibrations of hetero-nuclear

functional groups and polar bonds, especially OH stretching. Fluorescence, however,

is only sensitive to the fluorescent bonds, such as Rhodamine B, Rhodamine 6G,

Dihydrorhodamine, Fluorescein, etc.

Raman spectroscopy can be applied to characterize bulky solids, granular solids, clear

and scattering solutions and gases. It requires little sample preparation, while the

fluorescent method has constraints on the labeling of samples, and the IR method is

limited to sample thickness, uniformity, dilution etc. However, Raman spectroscopy

is not applicable to metals and alloys and it becomes difficult to measure samples with

dark colors or high fluorescence. In Table 2.4 we have summarized the main features

and sample applicability of Raman, IR and fluorescence spectroscopy. Table 2.5

compares the sample applicability of different spectroscopy approaches.

Table 2.4 Features of Raman, Near-IR absorption and fluorescence spectroscopy.

Raman Near-IR absorption Fluorescence

Assignments Molecular

vibrations

Molecular vibrations &

rotations

Fluorescent bonds

Spectral band width Narrow Medium Wide

Intensity Low Low High

Source Laser Xenon arc lamp

Silicon Carbide/Silicon

Nitride emitter

Laser

LED

Mercury-vaper lamp

Labeling Label-free Label-free Labeling required

Chapter 2

49

Table 2.5 Sample applicability of Raman, IR absorption and fluorescence spectroscopy.26–28

Raman IR absorption Fluorescence

Sample applicability ★★★★✰ ★★★★★ ★✰✰✰✰

Sample preparation ★★★★★ ★★★✰✰ ★✰✰✰✰

Darkly colored sample ★✰✰✰✰ ★★★★✰ ★★★★★

Gas, Vaper ★★★✰✰ ★★★★★ ★★★✰✰

Aqueous solutions ★★★★★ ✰✰✰✰✰ ★★★★★

Clear Fluids ★★★★★ ★★★✰✰ ★★★★★

Scattering Fluids ★★★★★ ★★★✰✰ ★★★★✰

Granular Solids ★★★★★ ★✰✰✰✰ ★★★★★

Bulk Solids, Clear ★★★★★ ★★★★✰ ★★★★★

Bulk Solids, Scattering ★★★★★ ★✰✰✰✰ ★★★★★

(1=Poor 5=Excellent)

2.3 Instrumentation for Raman spectroscopy

The original Raman scattering was observed by utilizing the sun light or a mecury

lamp as excitation, and a few colored filters and photographic plates to record the

spectra for hours.29,30 Thanks to technological development, modern Raman

spectroscopic systems provide ways to obtain Raman spectra with tremendous higher

sensitivity and spectral resolution. In general, a Raman spectroscopy system consists

of a light source, an excitation optical path, a sample under test, a collection optical

path and a spectrometer, as shown in Figure 2.13. The optical paths are constructed

with several optical lenses, filters, mirrors, and other optical components.


50

Figure 2.13 Schematic of a dispersive Raman spectroscopy setup.31

2.3.1 Light source

Nowadays most of the Raman spectroscopy systems use lasers as excitation source

because of their high intensity, directionality and monochromatic properties. Table

2.6 lists some laser sources used for Raman detection.

Table 2.6 Typical lasers used in Raman spectroscopy.

Laser

wavelength

(nm)

Laser type Maxium

Power

Raman wavelengths

(nm)

Stokes region,

100~3000 cm-1

Comments

Near-IR

1064 Nd: YAG 3 W 1075~1563 FT spectrometer, risk

of sample thermal

degradation

830 Semi-conductor 300mW 827~980

785 Semi-conductor 500mW 791~1027 Most widely used

excitation source for

Raman spectroscopy

VIS

632.832,33 He-Ne 500mW 637~781 High Raman signal,

high fluorescence 532 Semi-conductor

or Nd: YAG

1.5W 535~632.8

Chapter 2

51

514.5 Argon 517~608 influence, risk of

sample degradation 405-510 Semi-conductor 1W 406~602

UV

213-400 Semi-conductor 1W 213~455 Resonance Raman

spectroscopy

According to equation (2.27) for the Raman cross section and the (𝜈0 − 𝜈𝑚𝑛)4

approximation, incident light with higher frequences will give rise to a higher Raman

response. However, an excitation with higher frequences will result in a higher

fluorescense that interferes with the Raman signal. Nowdays, most of the commercial

Raman spectroscopy setups are equiped with 785nm laser as excitation light to make

a trade-off between the intensity of the Raman signal and the influence of fluorescence.

2.3.2 Filters

As mentioned before, the intensity of Rayleigh scattering is dominant over Raman

scattering. Therefore, the use of a notch filter or long-pass optical filter at the end of

the collection path is usally necessary to reduce the Rayleigh scattering. In some

Raman spectroscopy systems, band-pass optical filters with a narrow full-width half-

maxium (FWHM) bandwidth are placed next to the exciation source to suppress the

ambient light as well as the secondary laser lines. Besides, a dichroic mirror or laser

line mirror with different reflection and transmission characteristics at two different

wavelengths is used to split the excitation and the collection path.

2.3.3 Lenses

The lenses in a conventional Raman spectroscopy system include an excitation lens

for focusing the laser onto the sample and collimating the scattered light emitted by

the sample after excitation, and a collection lens for collecting the Raman signal into

the entrance of the detector. Most of the Raman spectroscopy systems use objective

lenses with a large NA as excitation lenses in order to achieve a high excitation

efficiency. The main rule of thumb for selecting the collection lens is to match the

dimension and the NA of the detector’s entrance, such as the facets of the optical

fibers or the slit of the spectrometer.

2.3.4 Spectrometer

Raman scattering typically can be captured and detected by a dispersion-based or

Fourier Transform (FT) spectrometer.10 As the name suggests, a dispersion-based

spectrometer applies a stationary or dynamic dispersive element to generate spectra

of the incoming radiation. The typical stationary dispersive elements are refraction

prisms and diffraction gratings. The dynamic dispersive elements include acousto-


52

optic tunable filters (AOTF), liquid crystal tunable filters (LCTF), linear variable

filters (LVF) and circular linear variable filters (CLVF).34 Among these elements, the

diffraction grating is the most widely used in modern compact spectrometers in the

visible and NIR region with high spectral resolution and fast response. The dispersed

signal can be measured by a CCD or CMOS array.

A Fourier-transform spectrometer detects the spectra by modulating the radiation in

the time domain through interference and then Fourier-transforming the interferogram.

The Fourier-transform spectrometers with Germanium or InGaAs detectors have

overwhelming advantages over the dispersive spectrometers in the IR region because

of their wide spectral range, high throughput, signal-to-noise ratio and spectral

resolution.

2.4 Surface Enhanced Raman Spectroscopy (SERS)

2.4.1 History of SERS

Although Raman scattering is widely considered to have great advantages for the

study of vibrational and rotational modes of molecules, in the early days, the

advantages of Raman spectroscopy only stayed at the theoretical phase. From

discovery35 to practical applications, Raman spectroscopy was far from being

developed compared with other optical detection techniques, such as X-ray,

fluorescence and infrared spectroscopy. There are many reasons for this, but the key

point is that the Raman scattering intensity itself is very weak. As mentioned in the

previous sections, the intensity of Raman scattering is only one out of thousands to

millions of the Rayleigh scattering. The previous excitation sources used to observe

Raman scattering were usually mercury lamps with monochromatic and poor

directional energy and low intensity, which led to the slow progress of research in

Raman in the 30 years after 1928.36 In the 1960s, with the invention of lasers and the

development of computer technology, the study of Raman spectroscopy began to

accelerate. Especially in 1974, Fleischman et al. accidentally discovered that Pyridine

molecules adsorbed on the surface of electrochemically roughened silver electrodes

produced very strong Raman scattering.37 This phenomenon did not attract special

attention at the beginning but was understood as the roughening of the silver

electrodes resulted in an increase in total surface area, which adsorbed more Pyridine

molecules and led to the increase of the Raman scattering intensity. However, Van

Duyne et al. pointed out that the Raman signal can be enhanced to 106 times which

couldn’t simply be explained by the increment of the surface area.38 They believed

that there must be some connection between the surface properties and the Raman

Chapter 2

53

enhancement, which is now known as Surface Enhanced Raman Scattering. The

discovery of the SERS phenomenon not only greatly expanded the researchers’

perception of Raman scattering, but also brought new impetus to the development of

Raman spectroscopy. However, in the following decades, the use of Raman

spectroscopy in practical applications was still relatively low.39 One of the main

reasons is that SERS is a type of surface sensitive technology where sample molecules

need to be adsorbed on a specific surface but developing a reliable and stable SERS

substrate was quite challenging at that time. More importantly, the understanding of

SERS mechanisms had not been well explored.

Only until the 1990s, with the rapid development of nanotechnologies and

nanomaterials, and the emergence and extensive study of various SERS substrates,

Raman spectroscopy and SERS have been widely utilized in more and more

application domains. The theoretical research on Raman and SERS mechanisms have

also stepped into an entirely new stage.

With the generation and extensive study of various SERS substrates, Raman

spectroscopy and SERS have becoming popular not only in applications, but also in

theoretical research. The combination of SERS with micro/nano sensors has

enormously improved the accuracy, sensitivity and reliability of molecular

characterization both qualitatively and quantitatively.

There have been many assumptions and controversies regarding the mechanisms of

SERS, but nowadays, it is widely accepted that the dominant contributor to most of

the SERS phenomena is the electromagnetic enhancement mechanism. In addition,

the chemical mechanism also explains some of the SERS processes. According to the

equation for Raman response mentioned previously (equation (2.24)), the magnitude

of the induced dipole moment which characterizes the Raman scattering, is relevant

to the electric field E0 and the molecular polarizability. Among these two factors, the

electromagnetic enhancement mechanism corresponds to the local electric field, and

the chemical enhancement corresponds to the molecular polarizability respectively.


54

2.4.2 Electromagnetic Enhancement

Figure 2.14 (a) An Au nanoparticle acts as an antenna by excitation of a dipolar localized

surface plasmon resonance (LSPR).40 (b) FDTD simulated Electromagnetic field enhanced of

an Au nanoparticle due to the LSPR, excited by 785nm wavelength radiation.

Figure 2.14 depicts the electromagnetic enhancement schematically for the case of an

Au nanoparticle. The electromagnetic mechanism is the main enhancement effect of

the electromagnetic field amplification induced by resonance excitations based on

localized condition-electron oscillation on the surface of metallic nanostructures,

which is called surface plasmon (SP).41 This type of resonance, known as dipolar

localized surface plasmon resonance (LSPR), can be derived from the optical

extinction spectrum, including absorption and elastic scattering of specific metallic

nanostructures.16 The oscillation frequency 𝜔𝑚𝑎𝑥 of the plasmons in the metallic

structure depends on the dielectric function of the metal 휀𝑚𝑒𝑡𝑎𝑙(𝜔) and the

surrounding dielectric 휀𝑚(𝜔). The electromagnetic enhancement has two key features.

First of all, the coupling state of the photon and localized surface plasmons associate

with a sharp increase in the amplitude of the electromagnetic field near the rough

metallic surface. Therefore, the molecules adsorbed on the surface are subjected to

much stronger electromagnetic fields. In addition, the molecular dipole radiates

Raman scattering in close vicinity of metal rather than in free space.16 Virtually, not

only the incoming radiation from the excitation source, but also the Stokes Raman

scattering at a frequency of 𝜔𝑠 = 𝜔𝑖𝑛𝑐 − 𝜔𝑣𝑏 can induce LSPR of the metallic

nanostructure for a specific vibration mode. Both the local field enhancement and the

Raman radiation enhancement originating from the electromagnetic field can be

coupled into the metal substrate LSPR. Although the coupling may not be the same

for the two enhancements, the resonance should be at least similar in quality. The

overall intensity of SERS is given by Equation (2.29). It depends on the

Chapter 2

55

electromagnetic field of both the incident radiation 𝜔𝑖𝑛𝑐 and the induced Raman

scattering 𝜔𝑠

𝐼𝑆𝐸𝑅𝑆 = 𝐼𝑖𝑛𝑐(𝜔𝑖𝑛𝑐) ∙ 𝐼(𝜔𝑠) = |𝐸𝑖𝑛𝑐(𝜔𝑖𝑛𝑐)|2 ∙ |𝐸(𝜔𝑠)|2 （2.29）

Therefore, the optimal SERS enhancement requires that the incident EM radiation at

𝜔𝑖𝑛𝑐 and the Raman scattering radiate at 𝜔𝑠 are in resonance with the LSPR peak of

the nanostructure. In general, in order to fulfill the conditions for visible and NIR

excitation sources, it is necessary to utilize nanostructures with typical dimensions

between 20nm and 100nm. LSPR is strongly dependent on the size and shape of the

nanostructures. Due to the coupling from the interaction of LSPR of the individual

nanostructures, LSPR is also stronger for the nanostructures with smaller spaces. The

degree of enhancement by LSPR on the metallic nanostructures compared to

conventional Raman scattering is called the SERS enhancement factor (EF). For a

small metallic nanosphere, the electromagnetic enhancement factor can be explicitly

described as42,43

𝐸𝐹𝑒𝑚(𝜔𝑠) ≅ |

휀(𝜔𝑖𝑛𝑐) − 휀0

휀(𝜔𝑖𝑛𝑐) + 2휀0

|

2

|휀(𝜔𝑠) − 휀0

휀(𝜔𝑠 + 2휀0)|

2

(𝑟

𝑟 + 𝑑)12

（2.30）

Where, 휀(𝜔)is the dielectric function of the metal associated with its frequency, and

휀0 represents the dielectric constant of the surrounding medium. 𝑟 is the radius of the

nanosphere, and 𝑑 is the distance between the molecule and the surface. Index 𝑖𝑛𝑐

represents the incident light and 𝑠 stands for the scattered light. Noble metals such as

Au, Ag and Cu are the proper materials for SERS under the typical visible and NIR

excitation44, as the real part of 휀(𝜔) is close to −2휀0, and the imaginary part is small.

The enhanced local field of a 100nm diameter nanosphere with different base

materials according to the FDTD simulations are listed in Table 2.7. The real part

Re(휀) and imaginary part Im(휀) of the dielectric function of the materials at certain

wavelengths (𝜆) are extracted from the Lumerical FDTD database.

Table 2.7 Permittivity of some materials and maximum amplitude of the electromagnetic field

of a 100nm diameter nanosphere under 785nm excitation. (Lumerical FDTD database &

Simulation results)

Metal 𝜆 (nm) Re(휀) Im(휀) |E|/|E0|

Au 775 -20.787 0.729 4.15

Ag 756 -33.451 3.127 3.88

Pt 775 -15.808 26.716 3.59

Cu 729 -19.686 1.949 4.04

Al 775 -67.017 45.134 3.53


56

Fe 775 -3.96 21.6 3.45

Ni 775 -12.671 20.946 3.64

Ti 775 -3.3824 18.084 3.43

Si 785 13.726 0.0518 2.27

SiO2 707 2.117 0 1.47

(Re(ε): real part, and Im(ε): imaginary part of the dielectric function of the materials)

Figure 2.15 The normalized EF of a nanosphere drops fast down when the distance of the

molecule from the surface increases. The legends refer to the radius of the nanosphere.

According to equation (2.30), the electromagnetic enhancement factor also largely

depends on the radius of the nanosphere and the distance of the molecule from the

surface, as shown in Figure 2.15. The enhancement decreases fast as the distance of

the molecule from the surface becomes larger than 50nm for nanoparticles with a

radius smaller than 100nm. This implies that the sample molecule must be near the

metallic surface to induce the highest Raman scattering. Normally, a maximum

enhancement factor, or single molecule enhancement factor (SMEF) is defined to

characterize the optimal enhancement of the SERS substrate, which is given by the

|𝐸𝑙𝑜𝑐(𝜔𝑖𝑛𝑐)|4 approximation, since the shifted frequency of Raman scattering is not

far from that of the excitation16

𝑆𝑀𝐸𝐹(𝜔𝑖𝑛𝑐 , 𝜔𝑠) ≅|𝐸𝑙𝑜𝑐(𝜔𝑖𝑛𝑐)|2

|𝐸0(𝜔𝑖𝑛𝑐)|2

|𝐸𝑙𝑜𝑐(𝜔𝑖𝑛𝑐)|2

|𝐸𝑙𝑜𝑐(𝜔𝑠)|2

≅|𝐸𝑙𝑜𝑐(𝜔𝑖𝑛𝑐)|4

|𝐸0(𝜔𝑖𝑛𝑐)|4

（2.31）

The electromagnetic field tend to be enhanced more according to the localized surface

plasmon resonance in the area that is close to the sharp tips, edges, in the pitch area

between two nanostructures and other sub-structures on top of the smooth structure,

as illustrated in Figure 2.16.

Chapter 2

57

Figure 2.16 FDTD simulated electric fields of an Au nano-pyramid, cube and nanospheres in

an external EM radiation at 785nm wavelength.

2.4.3 Chemical Enhancement mechanism

The chemical enhancement mechanism refers to the Raman scattering enhanced by

the analyte molecules rather than the interaction of the incident radiation with SERS

substrates. The polarizability of the molecule binding on a noble-metal surface can be

changed owing to different transitions and processes45. This change can be induced

by either the charge transfer transition between the Fermi level of metal and the

highest occupied molecular orbital (HOMO) of sample molecules, or the charge

transfer transition between the valence band (VB) or conduction band (CB) of the

dielectric substrate and lowest unoccupied molecular orbital (LUMO) of analyte

molecules.42 In addition, the charge transfer transition between HOMO and LUMO or

between VB and CB of the substrate material is also helpful to the enhancement of

the Raman signal due to the resonance Raman scattering. These transitions and

processes can be understood as changes in the shape of the internal electron structures,

which leads to altering of the relative strength and frequencies of the molecular

vibrational modes, or even give rise to new vibrational modes. It is obvious that the


58

chemical enhancement has different effects on different molecular-substrate

combinations and each vibrational band within that molecule. The chemical

enhancement generally contributes to an additional enhancement factor of 10-100

accordingly.36

2.4.4 SERS substrates

The SERS phenomenon mainly originates from the electromagnetic enhancement

ascribed to the LSPR. Therefore, to perform a SERS experiment, it is essential to

choose or fabricate proper SERS substrates with an optimal LSPR effect and thus with

a high enhancement factor. The most critical aspects of a SERS substrate are the

shapes and dimensions of surface structures and the type of noble-metal that covers

or forms the surface structures. The typical SERS substrates include conventional

roughened electrodes, colloidal nanoparticles, metal island films and modern

periodical nanostructures.

Roughened electrodes

Electrochemically roughened silver electrodes were the initial type of SERS

substrates in history.37 In 1974, M. Fleischmann and his colleagues discovered that

Pyridine adsorbed on the electrode of an electrochemical cell generated an

enormously high intensity Raman signal. Their electrode was fabricated by the

electrochemical Oxidation-Reduction Cycle (ORC). The process of fabrication is to

put a polished silver plane in a cell and to apply a cyclic linear potential sweeping for

15 minutes. The formation and reduction of silver chloride during each cycle finally

resulted in a fully etched silver surface. The surface area of the silver plane by the

ORC process can be increased by at least a factor of 10. Some kinds of metal

electrodes, such as Cu, Al and Ni can also be easily roughened by chemical corrosion

to be used as SERS substrates.46 But these methods are problematic when treating

other noble metals such as Au, Pt and Rh, because they tend to generate a compact

oxide layer that hinders further dissolution and deposition of the surface.47 To

overcome this obstacle, A.J. Ariva and co-workers implemented a unique roughening

procedure such that a Pt electrode with a large surface area and controlled roughness

can be produced in acidic electrolytes due to the reduction of a previously generated

hydrous Pt oxide layer by applying a high-frequency periodic potential.48 The

roughening can also be realized by controlling the oxidation and a reduction of the

current rather than the potential. The typical surface protrusion of electrochemically

roughened surfaces is 25 to 500nm depending on the electrolyte, the metal and the

ORC potential or current.47 The enhancement factors of typical noble and transition

metal electrodes vary very much, ranging from 1 to 6 orders of magnitude, in terms

Chapter 2

59

of the surface morphology and the nature of the metal. Figure 2.17 shows the STM

image of a Pt electrode and the SEM image of a Rh electrode after an ORC procedure.

Figure 2.17 (Left) The Scanning Tunneling Microscope (STM) image of a Pt electrode

roughened by controlled a ORC potential, (Right) the AFM image of a Rh electrode

roughened by a controlled ORC current.47

Colloidal nanoparticles

Au and Ag nanoparticles in colloids are widely used as SERS substrates in chemical

and biological research.49 These colloidal nanoparticles can easily be synthesized via

chemical reaction without sophisticated lab equipment and processes. C. G.

Blatchford et al. developed a synthesis method for Au colloids by mixing a solution

of 0.27mM gold chloride acid and 17mM tri-sodium citrate solution with a volume

ratio of 19:1. It was rapidly mixed with vigorous stirring and then gentle stirring at

90°C for 40 minutes. The Au nanoparticles in the colloid fabricated by this approach

are between 12nm to 18nm.50 The dimension range of the Au nanoparticles can be

adapted by controlling for example the concentrations of the gold chloride acid and

the tri-sodium, the pH of the solvents, the reaction temperature and the stirring time,

resulting in an average diameter of more than 30nm.51 The synthesis of colloidal Ag

nanoparticles has been reported in many papers. Briefly, an ethylene glycol solution

of 0.25M AgNO3 and an ethylene glycol (EG) solution of 0.25M

Polyvinylpyrrolidone (PVP) were mixed with 1:6 volume ratio at room temperature.

The reaction was taking place at 160°C for 60 minutes with vigorous stirring and

resulted in colloidal Ag nanoparticles with diameters ranging from 100nm to 300nm.52

Adapting the molar ratio of AgNO3 and PVP close to 1:1 can even create Ag

nanowires and nanocubes.53 Silver triangular nanoplates can also be synthesized by

replacing the PVP solution with N,N-dimethylformamide (DMF) solution. In addition

to the chemical approaches, Ag nanoparticles can also be produced physically by a

granular-film approach.54


60

Figure 2.18 (A) SEM image of silver nanowires, (B) Transmission Electron Microscopy

(TEM) image of an individual Ag nanowire, (C) SEM image of silver triangular nanoplates,

(D) TEM image of silver triangular nanoplates, (E) SEM image of Ag nanoparticles, (F) TEM

image of Ag nanoparticles. 52

In recent years, some composite nanoparticles with multilayers and unconventional

shapes have been utilized as SERS substrates. Jin-Qun Xue and co-workers explored

surface-imprinted core-shell Au nanoparticles as a specific functional SERS

substrate.55 The two-layer nanoparticles were fabricated by a method based on a sol-

gel process where a 2nm thick molecularly imprinted polymer (MIP) layer was capped

on top of each Au core with a diameter of 50nm. The MIP material has shown high

affinity for the target molecules thus resulting in a stronger SERS signal than that of

the bare Au nanoparticles. I.I.S Lim and co-workers investigated Au and magnetic

nanoparticles coated with gold shells as bio-functional SERS nanoprobes. The

magnetic core could be either Fe2O3 or MnZn ferrite.56 The diameter of the core-shell

nanoparticles ranges from 30nm to 90nm depending on the conditions during

synthesis. Qiao An et al. developed the process for the fabrication of Ag nanoparticles

towards modified Fe3O4@carbon core-shell nanospheres as SERS probes for organic

pollutants detection.57 The size and shape of the core-shell nanospheres can be

Chapter 2

61

controlled by modifying the concentration of stock AgNO3 solution during the

synthesis.

Figure 2.19 TEM images of (A) AuNPs and (B) MIP-AuNPs.55

Figure 2.20 TEM images of Fe3O4@C@Ag nanospheres with different sizes.57

The enhancement factor of colloidal nanoparticles typically ranges from 104 to 109

accordingly.58–61

Metal island films

Metal island films on substrates are commonly used SERS substrates as they can be

fabricated in large areas at relatively low cost. Physical methods for the fabrication of

metal island films including vacuum thermal evaporation62 and sputtering coating63

are based on the self-assembly of the metal particles, as shown in Figure 2.21 and

Figure 2.22. The mass thicknesses of the metal island film and the dimensions of the

individual metal islands can be controlled by tuning the deposition speed of the

evaporation or sputtering process. Ag island films can also be deposited onto stainless

steel substrates by multiple scan cyclic voltammetry (CV) using an electrolyte

containing 0.1M KNO3, 0.1M KCN and 0.01M AgNO3 in a conventional three-

electrode cell.64 Adjusting the potential scan rate gives rise to Ag island films with

different thicknesses and dimensions.


62

Figure 2.21 Ag island films with mass thicknesses of 5nm, 10nm, 15nm and 20nm on

crystalline silicon substrates. Each image shows an area of 1µm × 1µm.62

Figure 2.22 SEM images of different Ag island films on a stainless steel substrate at potential

CV rates of 50mV/s, 100mV/s, 150mV/s and 200mV/s, respectively. 64 The scale bar is 1µm.

The analytical enhancement factor of electrochemically deposited Ag island films can

be as high as 3.7 × 1012 corresponding to a detection limit of 5 × 10-16 M for R6G

according to the literature.64 But considering the affinity of the molecules with the

SERS substrate and taken into account that the molecules were distributed unevenly

at low concentration, an enhancement factor up to a 9th order of magnitude is

reasonable.

Periodical nanostructures / nanostructure arrays

The progress of modern nanotechnology and nanomaterial science has brought

enormous possibilities for the fabrication of SERS substrates with higher stability,

sensitivity and reproducibility. As suggested by the electromagnetic enhancement

mechanism, periodical nanostructures tend to produce higher and more homogeneous

LSPR compared to unevenly distributed nanostructures in space. The periodical

nanostructures can be fabricated by various approaches such as electrodeposition,

nanoimprinting lithography, e-beam lithography, chemical etching or a combination

of different approaches.

The self-assembly nature of the metal nanoparticles provides a relatively easy way of

fabricating periodical nanostructures. Hui Wang and co-workers reported a

specifically designed approach to deposit Au nanoshells into a 2D periodic array with

sub-10nm interparticle gaps65, as shown in Figure 2.23. The synthesized Au

Chapter 2

63

nanospheres were functionalized with surfactant cetyltrimethyl ammonium bromide

(CTAB) into colloidal nanoshells first. A droplet of this colloidal nanoshells were

placed on a substrate and dried, and finally organized into a hexagonal close packed

(hcp) pattern. Jian Feng Li et al. also reported self-assembled Au nanoparticles as

SERS substrates. But unlike the work of Hui Wang et al. where the Au nanoparticles

were covered with CTAB, Jian Feng Li isolated the Au nanoparticles with a transition

metal shell or silica shell.66

Figure 2.23 (a) Scheme of the nanoshell array fabrication and (b) SEM image of the nanoshell

array. 65

Wonjoo Lee reported self-assembled Au nanospheres with a controlled diameter and

spacing. The fabrication process involved solvent annealing, quaternization, colloid

adsorption and overgrowth, as shown in Figure 2.24.67

Figure 2.24 (a) AFM of the P4VP arrays, and SEM images of AuNP arrays on the quaternized

PS-b-P4VP films after overgrowth time of (b) 0 minutes, (c) 1 minute, (d) 3 minute. And (e)

the scheme of AuNP array fabrication. (i) Solvent annealing. (ii) Quaternization, (iii) Colloid

adsorption, (iv) Overgrowth. (Blue: substrate, green: P4VP, yellow: Polystyrene; red: Au). 67

Metal film over nanospheres (MFON), especially AgFON surfaces (Figure 2.25) are

very robust plasmonic materials for SERS detection.68,69 The procedure is to apply a


64

monolayer of self-assembled polystyrene or a SiO2 nanospheres mask on top of a

substrate, and then covered with a thin layer of noble metal via thermal evaporation

or sputtering. Polystyrene nanobeads and silica nanoparticles can be synthesized via

a half seeded emulsion polymerization method70 and Sol-Gel process respectively.71

The diameter of the nanospheres of the MFON are typically large (400-600nm), but

the atomic scale roughness behavior in combination with this large scale

nanostructures results in an enhancement factor up to 107.

Figure 2.25 SEM images of the AgFON SERS substrate from (A) side and (B) top-down.

J. Haes Amanda and co-workers combined the AgFON method based on self-

assembly of polystyrene nanobeads and nanolithography technique, and fabricated a

nano triangular array as SERS substrate (Figure 2.26).72 To produce the triangular

footprint, the nanospheres and the atomic Ag layer were removed by sonicating the

sample in a solvent. Nanospheres with a diameter of 400nm result in nanotriangles

with an in-plane width of 100nm and interparticle distance of 230nm approximately.

The enhancement factor of the periodical nanotriangles is up to 9 × 107 characterized

by Raman scattering of benzenethiol.

Figure 2.26 (a) AFM image of the nanotriangle array; (b) Raman spectrum benzenethiol from

the nanotriangle arrays enhancement.72

Chapter 2

65

The above method can be further developed to fabricate close-packed arrays (CPA),

sparse arrays (SA), spheres removal (SR) structures on substrate or optical fibers as

SERS sensors,73,74 as shown in Figure 2.27.

Figure 2.27 SEM image of nanostructures on the fiber tip with (A) CPA, (B, C) SA, (D) CPA-

SR and (E) SA-SR patterns using 1µm diameter PS nanospheres. Scale bars are 3µm.73

H. Schneidewind and co-workers investigated Ag-covered polymer based

nanogratings as SERS substrates.75 A 70nm thick electro-sensitive resist layer was

spin coated on top of a quartz substrate and then polymerized by electron-beam

lithography into polymer nanogratings. An Al2O3 layer with a thickness of around

1nm was deposited and then covered with 20nm thick Ag film as effective metal for

SERS. The enhancement factor of this polymer-based SERS substrate is 4 orders of

magnitude.

Figure 2.28 (a) Scheme of the fabrication process of polymer-based nanogratings as SERS

substrate; (1) spin coating and E-beam lithography, (2) Atomic Layer Deposition (ALD), (3)

Sputtering. (b) SEM image of the nanogratings.


66

Applying nano-imprinting lithography with or without chemical etching can also give

rise to periodical nanostructures.76–78 The enhancement factors of the typical

nanopillar arrays are in the order of 104 to 106, as shown in Figure 2.29.

Figure 2.29 Nanopillar arrays fabricated by (a) nano-imprinting lithography76 and (b)

chemical etching.77

Two-photon polymerization (2PP) additive printing is a new technology suitable for

the fabrication of arbitrary 3D micro- and nano-structures. The typical spatial

resolution of the 2PP technique is between 150nm to 200nm, however minimum

feature sizes of 52nm can be realized under specific conditions.79 2PP 3D printed

nanostructures such as a nanocorral complex and nanocorral array80 and a

nanopyramid81 with Ag coating give rise to a SERS enhancement factor in the order

of 104 to 106.

Figure 2.30 SEM image of 2PP printed (a) nanocorral complex, (b) nanocorral array, and (c)

nanopyramid.

In Table 2.8 we have summarized the characteristics of different types of SERS

substrates including the enhancement factors, synthesis and fabrication methods and

typical dimensions.

Chapter 2

67

Table 2.8 Comparison of different types of SERS substrates.

SERS substrate EF/ log Synthesis / Fabrication dimensions

Roughened

electrodes46–48

1 - 6 Electrochemical ORC

Chemical etching

25 – 500nm (surface

protrusion)

Colloidal nanoparticles,

nanocore-shells50-61

4 - 9 Chemosynthesis 10 – 300nm

(diameter)

Metal island films62–64 2 - 12 Thermal evaporation,

Sputtering, Electrochemical deposition

5 - 20nm (thickness)

Periodical

nanostructures65-81

4 - 9 Lithography, Chemical etching, Chemical

deposition, Self-assembly deposition,

ALD, 2PP and etc.

10 – 500nm

(diameter, pitch)

2.5 A brief overview on Raman and SERS applications

In the past decades, the number of publications on topics such as Raman spectroscopy

and Surface Enhanced Raman spectroscopy have grown tremendously, as shown in

Figure 2.31. It is almost unfeasible to provide a comprehensive overview of the fields,

not even in the subfields of Raman and SERS for toxicity or single cell detection.

Nevertheless, it is instructive to get insight into the different applications of Raman

spectroscopy and SERS from various angles rather than specific domains.

Figure 2.31 Number of publications on Raman OR Surface Enhanced Raman spectroscopy

(SERS) per year from 1999 to 2019. (Statistic data extracted from the Web of Science,

accessed on April 7, 2020)


68

Table 2.9 Number of publications on Raman OR Surface Enhanced Raman spectroscopy

(SERS) from 1999 to 2019 categorized according to the fields of application. (Statistic data

extracted from the Web of Science, accessed on April 7, 2020)

Field: Web of Science Categories Record

Count

% of

245,817

MATERIALS SCIENCE MULTIDISCIPLINARY 66,683 27.127

CHEMISTRY PHYSICAL 51,069 20.775

PHYSICS APPLIED 50,893 20.704

PHYSICS CONDENSED MATTER 31,205 12.694

CHEMISTRY MULTIDISCIPLINARY 28,175 11.462

OPTICS 24,594 10.005

NANOSCIENCE NANOTECHNOLOGY 24,422 9.935

SPECTROSCOPY 16,090 6.546

PHYSICS ATOMIC MOLECULAR CHEMICAL 13,140 5.345

ENGINEERING ELECTRICAL ELECTRONIC 12,400 5.044

CHEMISTRY ANALYTICAL 11,262 4.581

MATERIALS SCIENCE COATINGS FILMS 10,199 4.149

ELECTROCHEMISTRY 7,691 3.129

PHYSICS MULTIDISCIPLINARY 7,479 3.043

ENGINEERING CHEMICAL 7,287 2.964

CHEMISTRY INORGANIC NUCLEAR 7,127 2.899

MATERIALS SCIENCE CERAMICS 6,252 2.543

INSTRUMENTS INSTRUMENTATION 5,875 2.39

ENERGY FUELS 5,509 2.241

POLYMER SCIENCE 5,444 2.215

METALLURGY METALLURGICAL

ENGINEERING

5,288 2.151

BIOCHEMISTRY MOLECULAR BIOLOGY 4,169 1.696

CRYSTALLOGRAPHY 3,955 1.609

MULTIDISCIPLINARY SCIENCES 3,733 1.519

CHEMISTRY APPLIED 3,606 1.467

Table 2.9 lists the application domains of Raman and SERS grouped in the Web of

Science Categories. More than 90% of the research output is in the field of chemistry,

materials science, physics and nano-science and technology. Conventional Raman and

Chapter 2

69

SERS Raman spectroscopy can be applied to characterize the structure of the materials

based on their fingerprint frequencies of specific chemical groups or measure the

concentration of certain analytes by analyzing the intensity of the Raman signal.

Raman and SERS spectroscopy are very useful tools for online monitoring the

progress of chemical reactions.82

In crystallography, single crystals can be detected by utilizing Raman spectroscopy

because Raman signals provide unique information about folding, substrate binding

and catalysis of the biomolecules.83 Gergely Katona and co-workers reveal the end-

on peroxide intermediates in a nonheme iron enzyme via Raman-assisted

crystallography.84 Bo Gong et al. detected the innersphere interactions between the

Phosphate backbone of the hepatitis delta virus (HDV) ribozyme and the magnesium

hydrate.85

The advantage of being label-free makes Raman spectroscopy a powerful tool for

toxin detection. Various types of mycotoxins, such as fumonisin, deoxynivalenol and

citrinin in maize, wheat, barley, corn or beverage can be identified by SERS.86–89 K.

Samir and co-workers also traced thiram pesticide residues from fruits by confocal

SERS spectroscopy.90 J. Hongquan and his co-worker investigated the feasibility of

antimicrobial characterization by a Ag nanoparticles based SERS approach.91

Confocal Raman spectroscopy is widely used in biological and medical research. With

the maturity of confocal Raman micro spectrometers (CRM), the field of single cell

investigation has been extended significantly. Intracellular contents, such as

DNA/RNA information from precisely defined positions in living or fixed cells can

be obtained.92,93 The chemical fingerprints of cells by Raman spectroscopy can be

used as evidence for cancer diagnositics.94–96

Raman spectroscopy is becoming an increasingly important analytical tool in the field

of art and archaeology. A. Sanchez et al. combined micro Raman spectroscopy and

X-ray microfluorescence to analyze the pigments in glass and ceramic vessel

fragments found in the Iberian cemetery Tutugi (4th-3rd century, BC).97 R. Wang and

L. Yuesheng investigated multi excitation Raman spectroscopy for identifying

Chinese jade.33 The study of pigments, dyes and binders in antique art paintings by

non-contacting Raman spectroscopy gave new indications of painting techniques in

ancient times.98

Out of the fields discussed here, there is no doubt that Raman spectroscopy and SERS

have been widely used in a lot of other application domains. There is also no doubt

that Raman and SERS will pave the way towards a lot of other applications in the

future.


70

References

1. Pérez‐Juste, I. & Faza, O. N. Interaction of radiation with matter. in Structure

Elucidation in Organic Chemistry: The Search for the Right Tools 1, 1–26

(Wiley, 2014).

2. Born, M. & Wolf, E. Principles of Optics: electromagnetic theory of

propagation, interference and diffraction of light. (Pergamon Press, 1970).

3. Sathyanarayana, D. N. Vibrational spectroscopy: theory and applications.

(New Age International, 2015).

4. Pieczonka, N. P. W. & Aroca, R. F. Inherent complexities of trace detection

by surface-enhanced Raman scattering. ChemPhysChem 6, 2473–2484 (2005).

5. Ikhdair, S. M. Rotation and vibration of diatomic molecule in the spatially-

dependent mass Schrodinger equation with generalized q -deformed Morse

potential. Chem. Phys. 361, 9–17 (2009).

6. Sage, M. L. Morse Oscillator Transition Probabilitles for Molecular Bond

Modes. Chem. Phys. 35, 375–380 (1978).

7. D.Louck, J. Derivation of the Molecular Vibration-Rotation Hamiltonian

from the Schrodinger Equation for the Molecular Model. J. Mol. Spectrosc.

61, 107–137 (1976).

8. Wang, W. & Paliwal, J. Vibrational Spectroscopy for Food Processing. in

Introduction to Advanced Food Process Engineering (ed. Sahu, J. K.) 497–

534 (CRC Press, 2014).

9. Meiselman, S., Cohen, O., DeCamp, M. F. & Lorenz., V. O. Measuring

vibrational coherence lifetimes in liquid methanol using transient coherent

Raman scattering Measuring. J. Phys. Conf. Ser. 497, 012004 (2014).

10. Khan, S. A. et al. Fourier Transform Infrared Spectroscopy: Fundamentals

and Application in Functional Groups and Nanomaterials Characterization. in

Handbook of Materials Characterization 317–344 (SPRINGER, 2018).

doi:https://doi.org/10.1007/978-3-319-92955-2_9

11. Robert, B. Resonance Raman spectroscopy. Photosynth. Res. 101, 147–155

Chapter 2

71

(2009).

12. Lacksberg, J. O. B., Lerstam, E. R. I. K. A., Aruyama, Y. U. K. I. M., Ochrane,

C. O. J. C. & Ossman, G. E. R. R. Miniaturized time-resolved Raman

spectrometer for planetary science based on a fast single photon avalanche

diode detector array. Appl. Opt. 55, 739–748 (2020).

13. Shipp, D. W., Sinjab, F. & Notingher, I. Raman spectroscopy: techniques and

applications in the life sciences. Adv. Opt. Photonics 9, 315 (2017).

14. Berg, E. R., Freeman, S. A., Green, D. D. & Ulness, D. J. Effects of Hydrogen

Bonding on the Ring Stretching Modes of Pyridine. J. Phys. Chem. 110,

13434–13446 (2006).

15. Lin-na, Z., Lin-Jie, H., Su-Hua, C. & Bo, C. Molecular vibrational

Spectroscopy and its Applications in Chemistry. Fujian Anal. Test. 4, 258–

275 (1995).

16. Prochazka, M. Basics of Surface-Enhanced Raman Scattering (SERS). in

Surface-Enhanced Raman Spectroscopy. Biological and Medical Physics,

Biomedical Engineering 21–59 (SPRINGER, 2015).

doi:https://doi.org/10.1007/978-3-319-23992-7_3

17. Musso, M. Raman Spectroscopy: Comparison of Infrared Spectra with those

gathered by Raman Spectroscopy of selected substances. (2000).

18. Meyer, S. A., Ru, E. C. Le & Etchegoin, P. G. Quantifying Resonant Raman

Cross Sections with SERS. J. Phys. Chem. 114, 5515–5519 (2010).

19. Wilcox, P. G., Emmons, E. D. & Pardoe, I. J. Raman Spectra and Cross

Sections of Chemical Warfare Agents, Agent Simulants, and Explosives Using

213nm Deep-Ultraviolet Laser Excitation. ARMY EDGEWOOD CHEMICAL

BIOLOGICAL CENTER APG MD ABERDEEN PROVING GROUND United

States (2018).

20. Volkamer, R., Spietz, P., Burrows, J. & Platt, U. High-resolution absorption

cross-section of glyoxal in the UV – vis and IR spectral ranges. J. Photochem.

Photobiol. A Chem. 172, 35–46 (2005).

21. Faris, G. W., Copeland, R. A., Mortelmans, K. & Bronk, B. V. Spectrally

resolved absolute fluorescence cross sections for bacillus spores. Appl. Opt.

36, 958–967 (1997).


72

22. Burkholder, J. B., Ravishankara, A. R. & Solomon, S. UV/visible and IR

absorption cross sections of BrONO2. J. Geophys. Res. 100, 16793–16800

(1995).

23. Johnson, C. R. & Asher, S. A. Wavelength dependence of the preresonance

Raman cross sections of CH3CN , Wavelength dependence of the

preresonance Raman cross sections of CH3CN ,. J. Chem. Phys. 82, 1985

(1985).

24. Aggarwal, R. L. et al. Measurement of the absolute Raman cross section of

the optical phonon in silicon. Solid State Commun. 151, 553–556 (2011).

25. Kneipp, K. et al. Single Molecule Detection Using Surface-Enhanced Raman

Scattering (SERS). Phys. Rev. Lett. 78, 1667–1670 (1997).

26. Doyle, W. M. & Stage, E. Comparison of Near-IR and Raman Analysis for

potential process applications. Tech. note AN-922, IFPAC-2001 1–20 (2001).

27. Mahadevan-Jansen, A. et al. Optical techniques for diagnosis of cervical

precancers: a comparison of Raman and fluorescence spectroscopies. in

Advances in Fluorescence Sensing Technology II 2388, 110–120 (1995).

28. Dieing, T., Hollricher, O. & Toporski, J. Confocal Raman Microscopy.

(SPRINGER, 2009).

29. Bhagavantam, S. The carbon isotope in Raman scattering. Part I. Proc. Indian

Acad. Sci. A 2, 86–92 (1935).

30. Pattabhiramayya, P. The study of the Raman effect in sodium nitrate. Proc.

Math. Sci. 7, 229–235 (1938).

31. Schmid, T. Raman Microspectroscopic Imaging of Binder Remnants in

Historical Mortars Reveals Processing Conditions. Heritage 2, 1662–1683

(2019).

32. Windom, B. C. & David, W. G. S. A Raman Spectroscopic Study of MoS2

and MoO3 : Applications to Tribological A Raman Spectroscopic Study of

MoS 2 and MoO 3 : Applications to Tribological Systems. Tribol. Lett. 42,

301–310 (2011).

33. Wang, R. Multiexcitation Raman Spectroscopy in Identification of Chinese

Jade. Spectrosc. Lett. 44, 432–439 (2011).

34. Ford, B. K., Volin, C. E., Murphy, S. M., Lynch, R. M. & Descour, M. R.

Chapter 2

73

Computed tomography-based spectral imaging for fluorescence microscopy.

Biophys. J. 80, 986–993 (2001).

35. Raman, C. V. A new radiation. India J. Phys 2, 387–398 (1928).

36. Mosier-Boss, P. Review of SERS Substrates for Chemical Sensing.

Nanomaterials 7, 142 (2017).

37. Fleischmann, M., Hendra, P. J. & McQuillan, A. J. Raman Spectra of Pyridine

Adsorbed at a Silver Electrode. Chem. Phys. Lett. 26, 163–166 (1974).

38. Duyne, V., P., R., L., D., Jeanmaire & Shriver., D. F. Mode-Locked Laser

Raman Spectroscopy-A New Technique for the Rejection of Interfering

Background Luminescence Signals. Anal. Chem. 46, 213–222 (1974).

39. Baia, M., Astilean, S. & Iliescu, T. Raman and SERS investigations of

pharmaceuticals. (SPRINGER, 2008). doi:10.1007/978-3-540-78283-4

40. Schlücker, S. Surface-enhanced raman spectroscopy: Concepts and chemical

applications. Angew. Chemie - Int. Ed. 53, 4756–4795 (2014).

41. Le Ru, E. C., Blackie, E. J., Meyer, M. & Etchegoin, P. G. Surface Enhanced

Raman Scattering Enhancement Factors: A Comprehensive Study. J. Phys.

Chem. C 111, 13794–13803 (2007).

42. Le Ru, E. & Etchegoin, P. Principles of Surfance Enhanced Raman

Spectroscopy and related plasmonic effets. Elsevier 1, (2009).

43. Pilot, R., Signorini, R., Durante, C. & Orian, L. A Review on Surface-

Enhanced Raman Scattering. Biosensors 9, 1–100 (2019).

44. Yang, X., Yang, M., Pang, B., Vara, M. & Xia, Y. Gold Nanomaterials at

Work in Biomedicine. Chem. Rev. 115, 10410–10488 (2015).

45. Kim, J. et al. Study of Chemical Enhancement Mechanism in Non-plasmonic

Surface Enhanced Raman Spectroscopy ( SERS ). Front. Chem. 7, 1–7 (2019).

46. Carron, K. T., Xue, G. & Lewis, M. L. A Surface Enhanced Raman

Spectroscopy Study of the Corrosion-Inhibiting Properties of Benzimidazole

and Benzotriazole on Copper. 7, 2–4 (1991).

47. Tian, Z., Ren, B. & Wu, D. Surface-Enhanced Raman Scattering: From Noble

to transition Metals and From Rough Surface to Ordered Nanostructures. J.

Phys. Chem. B 106, (2002).


74

48. Arvia, A. J., Salvarezza, R. C. & Triaca, W. E. The Development of Metal

overlays with smooth and rough topographies. Electrochim. Acta 34, 1051–

1071 (1989).

49. Wang, A. X. & Kong, X. Review of recent progress of plasmonic materials

and nano-structures for surface-enhanced raman scattering. Materials (Basel).

8, 3024–3052 (2015).

50. Blatchford, C. G. & Campbell, J. R. Plasma resonance - enhanced adsorbates

on gold colloids: aggregation c.g. blatchford,. Surf. Sci. 120, 435–455 (1982).

51. Dou, X., Jung, Y. M., Yamamoto, H., Doi, S. & Ozaki, Y. Near-Infrared

Excited Surface-Enhanced Raman Scattering of Biological Molecules on

Gold Colloid I: Effects of pH of the Solutions of Amino Acids and of Their

Polymerization. Appl. Spectrosc. 53, 133–138 (1999).

52. Zhang, J., Li, X., Sun, X. & Li, Y. Surface Enhanced Raman Scattering

Effects of Silver Colloids with Different Shapes. J. Phys. Chem. B 109,

12544–12548 (2005).

53. Banchelli, M. et al. Controlled Veiling of Silver Nanocubes with Graphene

Oxide for Improved Surface-Enhanced Raman Scattering Detection. ACS

Appl. Mater. Interfaces 8, 2628–2634 (2016).

54. Kawabata, S., Ishikawa, N., Mitsui, M. & Nakajima, A. Fabrication and

characterization of a granular film consisting of size-selected silver

nanoparticles : application to a SERS. Eur. Phys. J. D 158, 155–158 (2007).

55. Xue, J. Q., Li, D. W., Qu, L. L. & Long, Y. T. Surface-imprinted core-shell

Au nanoparticles for selective detection of bisphenol A based on surface-

enhanced Raman scattering. Anal. Chim. Acta 777, 57–62 (2013).

56. Lim, I. I. S. et al. Gold and magnetic oxide/gold core/shell nanoparticles as

bio-functional nanoprobes. Nanotechnology 19, 305102 (2008).

57. An, Q. et al. Silver-coated magnetite–carbon core–shell microspheres as

substrate-enhanced SERS probes for detection of trace persistent organic

pollutants. Nanoscale 4, 5210 (2012).

58. Knauer, M., Ivleva, N. P., Niessner, R. & Haisch, C. Optimized surface-

enhanced Raman scattering (SERS) colloids for the characterization of

microorganisms. Anal. Sci. 26, 761–766 (2010).

Chapter 2

75

59. Kuroda, K., Ishida, T. & Haruta, M. Reduction of 4-nitrophenol to 4-

aminophenol over Au nanoparticles deposited on PMMA. J. Mol. Catal. A

Chem. 298, 7–11 (2009).




61. Lee, B. S., Lin, D. Z. & Yen, T. J. A low-cost, highly-stable surface enhanced

raman scattering substrate by si nanowire arrays decorated with au

nanoparticles and Au backplate. Sci. Rep. 7, 1–7 (2017).

62. Santbergen, R. et al. Application of plasmonic silver island films in thin-film

silicon solar cells. J. Opt. 14, (2012).

63. Mondt, R. De, Baert, K., Geuens, I., Vaeck, L. Van & Hubin, A. Study of the

Affinity of Thermographic Additives for Silver by Time-of-Flight Static

Secondary Ion Mass Spectrometry and Surface-Enhanced Raman

Spectroscopy on Silver Nanoparticles. Langmuir 22, 11360–11368 (2006).

64. Oriňáková, R. et al. Electrochemical deposition of SERS active

nanostructured silver films. Int. J. Electrochem. Sci. 8, 80–99 (2013).

65. Wang, H., Kundu, J. & Halas, N. J. Plasmonic nanoshell arrays combine

surface-enhanced vibrational spectroscopies on a single substrate. Angew.

Chemie - Int. Ed. 46, 9040–9044 (2007).

66. Li, J. F. & Tian, Z. Q. Shell-Isolated Nanoparticle-Enhanced Raman

Spectroscopy (SHINERS). Nature 464, 392–395 (2010).

67. Lee, W., Lee, S. Y., Briber, R. M. & Rabin, O. Self-assembled SERS

substrates with tunable surface plasmon resonances. Adv. Funct. Mater. 21,

3424–3429 (2011).

68. Dick, L. A., McFarland, A. D., Haynes, C. L. & Van Duyne, R. P. Metal film

over nanosphere (MFON) electrodes for surface-enhanced Raman

spectroscopy (SERS): Improvements in surface nanostructure stability and

suppression of irreversible loss. J. Phys. Chem. B 106, 853–860 (2002).

69. Masango, S. S. et al. High-Resolution Distance Dependence Study of Surface-

Enhanced Raman Scattering Enabled by Atomic Layer Deposition. Nano Lett.

16, 4251–4259 (2016).


76

70. Nuruzatulifah, A. M., Nizam, A. A. & Ain, N. M. N. Synthesis and

characterization of polystyrene nanoparticles with covalently attached

fluorescent dye. Mater. Today 3S, 112–119 (2016).

71. Rahman, I. A. & Padavettan, V. Synthesis of Silica Nanoparticles by Sol-Gel :

Size-Dependent Properties , Surface Modification , and Applications in

Silica-Polymer Nanocomposites — A Review. J. Nanomater. 2012, 1–15

(2012).

72. Haes, A. J., Haynes, C. L., Mcfarland, A. D. & Schatz, G. C. Plasmonic

materials for surface-enhanced sensing and spectroscopy. MRS Bull. 30, 368–

375 (2005).

73. Pisco, M. et al. Nanosphere lithography for advanced all fiber Sers probes.

Proc. SPIE 9916, 99161S–2 (2016).

74. Pisco, M. et al. Nanosphere lithography for optical fiber tip nanoprobes. Light

Sci. Appl. 6, e16229 (2017).

75. Schneidewind, H. et al. The effect of silver thickness on the enhancement of

polymer based SERS substrates. Nanotechnology 25, (2014).

76. Gartia, M. R. et al. Rigorous surface enhanced Raman spectral

characterization of large-area high-uniformity silver-coated tapered silica

nanopillar arrays. Nanotechnology 21, 395701 (2010).

77. Schmidt, M. S., Hübner, J. & Boisen, A. Large area fabrication of leaning

silicon nanopillars for Surface Enhanced Raman Spectroscopy. Adv. Mater.

24, OP11–OP18 (2012).

78. Schmidt, M. S., Boisen, A. & Jörg, H. Towards Easily Reproducible Nano-

structured SERS substrates. in IEEE Sensors Conference 1763–1767 (2009).

doi:10.1109/ICSENS.2009.5398468

79. Gan, Z., Cao, Y., Evans, R. A. & Gu, M. Three-dimensional deep sub-

diffraction optical beam lithography with 9 nm feature size. Nat. Commun. 4,

1–7 (2013).

80. Mu, J. et al. Direct laser writing of symmetry-broken nanocorrals and their

applications in SERS spectroscopy. Appl. Phys. B Lasers Opt. 117, 121–125

(2014).


Chapter 2

77



82. Leung, S., Winkle, R. F., Wootton, R. C. R. & Andrew, J. A method for rapid

reaction optimisation in continuous-flow microfluidic reactors using online

Raman spectroscopic detection. Analyst 130, 46–51 (2005).

83. Gong, B. et al. Detection of Innersphere Interactions between Magnesium

Hydrate and the Phosphate Backbone of the HDV Ribozyme Using Raman

Crystallography. J. Am. Chem. Soc. 130, 9670–9672 (2008).

84. Katona, G. et al. Raman-Assisted Crystallography Reveals End-On Peroxide

Intermediates in a Nonheme Iron Enzyme. Science (80-. ). 316, 449–454

(2007).

85. Gong, B. et al. Detection of Innersphere Interactions between Magnesium

Hydrate and the Phosphate Backbone of the HDV Ribozyme Using Raman

Crystallography. Methods 49, 101–111 (2010).

86. Lee, K. & Herrman, T. J. Determination and Prediction of Fumonisin

Contamination in Maize by Surface – Enhanced Raman Spectroscopy

( SERS ). Food Bioprocess Technol 9, 588–603 (2016).

87. Lee, K., Herrman, T. J., Nansen, C. & Yun, U. Application of Raman

spectroscopy for qualitative and quantitative detection of fumonisins in

ground maize samples. IJRS 1, 1–14 (2013).

88. Yuan, J. et al. A rapid Raman detection of deoxynivalenol in agricultural

products. Food Chem. 221, 797–802 (2017).

89. Singh, D. K. et al. Detection of the mycotoxin citrinin using silver substrates

and Raman spectroscopy. J. Hazard. Mater. 265, 89–95 (2014).

90. Kumar, S., Goel, P. & Singh, J. P. Sensors and Actuators B : Chemical

Flexible and robust SERS active substrates for conformal rapid detection of

pesticide residues from fruits. Sensors Actuators B. Chem. 241, 577–583

(2017).

91. Jiang, H., Manolache, S., Wong, A. C. L. & Denes, F. S. Plasma-enhanced

deposition of silver nanoparticles onto polymer and metal surfaces for the

generation of antimicrobial characteristics. J. Appl. Polym. Sci. 93, 1411–

1422 (2004).


78

92. Puppels, G. J. et al. Laser irradiation and Raman spectroscopy of single living

cells and chromosomes: Sample degradation occurs with 514.5 nm but not

with 660 nm laser light. Exp. Cell Res. 195, 361–367 (1991).

93. Kudo, T., Wang, S. F., Yuyama, K. I. & Masuhara, H. Optical Trapping-

Formed Colloidal Assembly with Horns Extended to the Outside of a Focus

through Light Propagation. Nano Lett. 16, 3058–3062 (2016).

94. Agenant, M. et al. Clinical superficial Raman probe aimed for epithelial tumor

detection: Phantom model results. Biomed. Opt. Express 5, 1203 (2014).



80870J (2011).




9, 4328–4336 (2015).

97. Sánchez, A., Tuñón, J. & Parras, D. Micro Raman spectroscopy (MRS) and

energy dispersive x-ray micro fluorescence (µEDXRF) analysis of pigments

in the Iberian cemetery of Tutugi. J. Raman Spectrosc. 43, 1788–1795 (2012).

98. Bersani, D. & Madariaga, J. M. Applications of Raman spectroscopy in art

and archaeology. J. Raman Spectrosc. 43, 1523–1528 (2012).

Chapter 3

79

Chapter 3

3 Two-Photon Polymerized Nanostructures for

SERS Analysis

Surface Enhanced Raman Spectroscopy (SERS) is a surface sensitive detection

technique that can enhance the Raman scattering of molecules substantially due to the

oscillation of localized surface plasmons by using rough metal surfaces or

nanostructures. Nowadays, SERS has been employed in a variety of application

domains. The enhancement of Raman scattering makes SERS a powerful tool for

chemists to analyse the vibrational and rotational modes of single molecules1,2. SERS

has also been increasingly utilized in biological research and life sciences such as

cancer diagnostics3–5, DNA/RNA identification6–8, toxin and drug detection9–11. To

achieve these detection goals, SERS substrates with specific surface profiles having a

boosted electromagnetic field under laser excitation should be considered.

Electrochemically roughened metal electrodes have been used as SERS substrate

since 1974, but the low level of enhancement and lack of reproducibility has restricted

their applications12. Noble metal nanospheres ranging from 10-200nm in colloids or

as deposition are typical SERS substrates which can be synthesized by chemical

approaches13–15. Other kinds of irregularly shaped nanoparticles, such as nano-rod,

nano-triangles, nano-cubes, nano-stars, nano-cookies and core-shells are increasingly

investigated as SERS substrates since they can generate stronger plasmonic oscillation

due to the sharp tips or edges, therefore resulting in higher Raman enhancement16–21.

However, the repeatability and reproducibility of SERS measurements with

This chapter is based on a published article:

Liu, Q.; Vanmol, K.; Lycke, S.; Van Erps, J.; Vandenabeele, P.; Thienpont, H.; Ottevaere, H.

SERS using two-photon polymerized nanostructures for mycotoxin detection. RSC

Advances 2020, 10, 14274.

Chapter 3 Two-Photon Polymerized Nanostructures for SERS Analysis

80

nanoparticles can hardly be guaranteed because of the inhomogeneity of the particles’

distributions and shapes. Therefore, nano-pillar arrays, hemisphere arrays, triangle

arrays, and other periodical nanostructures have been developed as SERS substrates

by means of electrodeposition9 and nanoimprinting lithography22,23 to improve the

repeatability and reproducibility. However, these methods are time consuming and

less flexible. To fabricate these periodic nanostructures with different dimensions,

new polymer or silicon masks must be produced in advance. Chemical/plasma etching

provides a fast approach for large area SERS substrates manufacturing24,25, but only

nano-pillar arrays can be fabricated with this method.

In this chapter we present an additive manufacturing method employing two-photon

polymerization to fabricate periodic nanostructures as SERS sensing platform.

Multiple nano-pillar arrays with different dimensions are printed and characterized.

This approach allows fast and flexible prototyping of SERS substrates26,27. The use of

two-photon polymerization lithography can also greatly reduce the complexity and

lead time of the nanostructure manufacturing for SERS applications by computer-

aided design (CAD) and computer-aided manufacturing (CAM). We simulate the

boosting of electromagnetic fields for different nano-pillar arrays by the Finite-

Difference Time-Domain (FDTD) method.28 The simulation output is compared with

experimental results of Rhodamine B (RhB) solutions. In a proof-of-concept

experiment, we performed SERS measurements on a mix of Fumonisin b1 and

Deoxynivalenol. The spectra of these two types of mycotoxins are analysed with PCA

methods and with respect to their vibrational modes.

3.1 Fabrication of nanostructures with two-photon

polymerization and simulations

3.1.1 Two-photon polymerization process

Two-photon polymerization lithography is a novel 3D additive fabrication technique

based on two-photon absorption of photo resins29. When an ultrashort laser pulse is

highly focused onto the photoresist, it will initiate two-photon polymerization via two-

photon absorption in a small region where the energy is higher than a threshold. This

region is the so called ‘voxel’ which represents the minimum feature size we can

achieve by two-photon polymerization lithography. Typically, the voxel is defined as

an ellipsoid with a specified diameter and aspect ratio which are determined by

parameters such as the physical and chemical properties of the photoresist, the

intensity and illumination time of the femtosecond laser, the magnification and

Chapter 3

81

numerical aperture of the objective lens for focusing. The typical resolution of two-

photon polymerization is between 150nm to 200nm, but a resolution of 52nm can be

realized under certain circumstances with a newly developed photoresin with high

mechanical strength30. By moving the relative spatial position of the voxel with a

galvanometric mirror scanner and piezo stages, periodic or complex 3D structures can

be printed31.

Figure 3.1 Scheme of the two-photon polymerization system we used for manufacturing

nanostructures. (AOM: acousto-optical modulator).

In our process, we employ the Nanoscribe Photonic Professional GT 3D printer to

fabricate our SERS substrates. Figure 3.1 shows the schematic diagram of the GT 3D

printer. The GT printer has a 780nm wavelength femtosecond laser with about 100 fs

pulses. The minimum XY feature size we can achieve with the Nanoscribe GT with

IP-dip photoresin is about 200nm. The typical printing range is 300 × 300 µm2, and

the accessible writing area can be up to 100 × 100 mm2 using stitching. The 3D

printing of an individual nano-pillar array with an area of 50µm × 50µm takes 3 to 5

minutes. The fabrication process can be divided into three steps. First, we use CAD

software to design the nanostructures with nominal shapes, such as the ideal nano-

pillar arrays. These designs can be exported as general stereolithography (STL) files

and be imported to the workstation of the Nanoscribe GT. Next, according to the

system parameters of the 3D printer, such as the magnification of the objective lens,

the type of photoresin and the dimensions of the voxel, the workstation compiles the

STL files into CNC programs which contain the instructions and parameters the

printer will follow. The workstation runs the CNC programs virtually to simulate the

two-photon polymerization process. Finally, the Nanoscribe GT runs the CNC

programs such that the 3D nanostructures are fabricated. We use a 50× objective lens


82

and IP-dip photoresin (Nanoscribe, Eggenstein-Leopoldshafen/BW, Germany) for the

two-photon polymerization manufacturing of the nanostructures.

3.1.2 Electromagnetic field enhancement simulation of nanostructures with

FDTD method

As explained in Chapter 2, it is generally accepted that the enhancement of SERS

stems from two major mechanisms, namely chemical and electromagnetic

enhancement.32 Chemical enhancement corresponds to the intrinsic properties of the

molecule adsorbed on the metallic surface, including the chemical polarity,

chemisorption, orientation with respect to the surface, etc33. The chemical mechanism

is explained only in specific occasions with low Raman enhancement contribution,

and it occurs jointly with the electromagnetic enhancement which is much more

dominant34.

According to the electromagnetic enhancement hypothesis, the interaction of incident

light with the metallic surface of a SERS substrate will generate oscillations of

localized plasma dipoles, thereby boosting the electromagnetic field in the region near

the surface35. The boosted region is interpreted as a ‘hotspot’ of the SERS substrate.

The electromagnetic enhancement factor, or single-molecule enhancement factor

(SMEF) for a molecule located at the hotspot can be expressed in the |𝐸𝑅|2|𝐸𝑆|2

approximation36 as discussed in equation (2.31):

𝑆𝑀𝐸𝐹(𝜔𝑅 , 𝜔𝑆) ≅|𝐸𝑙𝑜𝑐(𝜔𝑅)|2|𝐸𝑙𝑜𝑐(𝜔𝑅)|2

|𝐸0(𝜔𝑅)|2|𝐸𝑙𝑜𝑐(𝜔𝑆)|2 （3.1）

Where 𝜔𝑅 is the frequency of incident light, and 𝜔𝑆 is the Stokes frequency of Raman

radiation enhanced by the local field. 𝐸0 refers to the field of incident light, and 𝐸𝑙𝑜𝑐

refers to the local field induced by the incident light, respectively.

Normally, the Stokes frequency 𝜔𝑆 is quite close to the incident frequency 𝜔𝑅. The

expression can be further simplified to:

𝑆𝑀𝐸𝐹 ≅|𝐸𝑙𝑜𝑐(𝜔𝑅)|4

|𝐸0(𝜔𝑅)|4 （3.2）

This implies that the SMEF is proportional to the 4th power of the local field, which

provides a simple estimation to mimic the SERS performance via a numerical

approach. In this PhD we verify the feasibility of two-photon polymerized SERS

substrates both theoretically with the FDTD method using Lumerical software and

experimentally with Rhodamine B detection.

Chapter 3

83

The key parameters of the nano-pillar arrays are the height (H) and diameter (D) of

each pillar, and the pitch (P) between them, as shown in Figure 3.2. In a preliminary

study, nanopillar arrays with an aspect ratio of 1, i.e. having identical values of height

and diameter, are considered for a range from 200nm to 600nm by taking into account

the ~200nm resolution of two-photon polymerization. We built three different

simulation models and investigate them for their FDTD electromagnetic solutions,

whereby each model corresponds to one step of the nanostructure manufacturing

process. The first nominal shape model consists of cylindrical pillars, whereas the

second is a voxel-based model, in line with the 3D printing process flow of two-

photon polymerization manufacturing, and the third model reflecting the true

fabricated shape as measured by Veeco Dimension 3100 Atomic Force Microscopy

(AFM). FDTD solves Maxwell’s equations in the time domain by calculating the

electromagnetic field values changes at discrete steps (grid or mesh) in time28. We

started the FDTD simulations from 2D models, which ran very fast (<1 minutes) with

Lumerical software. Later we built 3D models of the nanostructures and performed in

depth simulations to assess the Raman enhancement performances, which took a

much longer time for each model as the operand increased exponentially with the

amount of meshes defined in the simulation. For example, the FDTD simulation for

the 200nm pillar array took several minutes, but the 600nm pillar array took several

hours.

Figure 3.2(a) and Figure 3.3(a) are illustrations of nano-pillar arrays for the nominal

shape model and the voxel-based model, respectively. Figure 3.2.(b, c, d) show the

induced electric field |𝐸|/|𝐸0| with 200nm, 400nm and 600nm nano-pillar arrays in

the nominal shape model under 785nm excitation after FDTD simulations. Figure

3.3.(b, c, d) show the induced electric field |𝐸|/|𝐸0| with 200nm, 400nm and 600nm

nano-pillar arrays in the voxel-based model under 785nm excitation after FDTD

simulations. Here 𝐸0 is the amplitude of the excitation electric field, and 𝐸 is the

amplitude of the induced electric field. In the nominal shape model, the leading

enhancement of the electric field appears in the edge of the nanostructures. Basically,

such perfect shapes are difficult to obtain via two-photon polymerization. In contrast,

we notice that the voxel-based model gives rise to complex discrete shapes with

cavities and bulges on top of a smoother profile. These complex structures induce

extended hotspot sections and higher electric field values in general. But it should be

mentioned that the complex structures may also destruct the enhancement in some

local areas which we will discuss later. The maximum electric field of the 200nm

pillar array in both the nominal shape model and the voxel-based model is larger than

the ones of the 400nm and 600nm pillar-arrays, suggesting that the 200nm pillar array

has a higher SMEF according to our FDTD simulations.


84

Figure 3.2. (a) Drawing of a nominal shape model, and (b, c, d) electric field distribution of

the 200nm, 400nm and 600nm pillar arrays using the nominal shape model and simulated by

the FDTD method.

Chapter 3

85

Figure 3.3. (a) Drawing of a voxel-based model, and (b, c, d) electric field distribution of the

200nm, 400nm and 600nm pillar arrays using voxel-based model and simulated by the FDTD

method.

After two-photon polymerization fabrication, we employ a 20nm thick Au layer with

a sputtering coater such that the front surfaces as well as the sides of the nano-pillars

are covered with metal where localized enhancement can be induced by excitation.

The fabricated nano-pillar arrays are characterized with Scanning Electron

Microscopy (SEM) and Atomic Force Microscopy (AFM). Some of the nano-pillar

arrays are shown in Figure 3.4. Each of the periodic nanostructure arrays has an

effective area of 50µm×50µm and a 1µm thick base layer to increase the adhesion to

the silica glass and its stability. The silica glass has been silanized with 3-

(Trimethoxysilyl)propyl methacrylate before two-photon polymerization to further

improve the adhesiveness. The periodic lines visible in both the SEM and AFM

images are due to a stitching process that was used during the 3D fabrication to reduce

the size of the STL files. This effect should not deteriorate the SERS signal in the

experiments as the detection area is within the center area of each block, but will be

avoided by systematic optimization of the two-photon polymerization process in

future work.

Figure 3.4. Morphologies of 200nm, 400nm and 600nm nano-pillar arrays. Measured with

SEM (column 1 and 2) and AFM (column 3).

We obtained the morphological characteristics of the nano-pillar arrays based on an

analysis of SEM images and AFM data, shown in Table 3.1. The measured values of


86

height, diameter and pitch are in good agreement with the designed values. Although

the homogeneity of the 200nm nano-pillar array looks visually less than the ones of

the 400nm and 600nm nano-pillar arrays, the standard errors of the height, diameter

and pitch of the different nanostructures caused by the fabrication errors are similar

according to the SEM and AFM measurements. We also notice that, although the

heights of the 200nm and 300nm structures are still within the confidence interval of

the measurements, they are a little bit smaller than the nominal values. This is

probably due to the limit of the AFM probe which cannot reach the bottom of the

trough when the pitch is very narrow. We build the fabricated shape model based on

the 3D data obtained from the AFM.

Table 3.1 Dimensions of nano-pillar arrays measured with SEM and AFM. (all units in nm)

Pillar/nm Height Diameter Pitch

200 179.7 ± 50.5 217.3 ± 38.1 228.8 ± 36.1

300 275.3 ± 54.5 287.5 ± 21.8 304.5 ± 22.4

400 431.1 ± 76.9 381.9 ± 35.2 358.0 ± 32.0

500 564.4 ± 67.6 474.0 ± 43.0 473.0 ± 46.6

600 625.0 ± 96.2 579.4 ± 35.2 573.1 ± 32.3

(The average and standard deviation for each pillar array are obtained over the measurements

within a 20µm by 20µm area.)

We performed the FDTD simulation on the fabricated shape model and investigated

its electromagnetic enhancement. Figure 3.5 illustrates the fabricated shape model and

the simulated electric field of one cross-section of 200nm, 400 and 600nm nano-pillar

arrays. Unlike the other two models in which the periodic pattern and induced hotspots

are homogeneously distributed, the fabricated model shows a poorer uniformity of the

hotspots because the oscillation of localized plasmonic dipoles is much more intricate

due to fabrication errors. In this case, the location of molecular adsorption and the

detection area are of great importance from a practical point of view. For instance, if

a molecule is adsorbed on the region of the middle pillar in Figure 3.5.(c), a stronger

Raman scattering will be induced compared to the other regions.

Chapter 3

87

Figure 3.5 (a) Drawing of a nominal shape model, and (b, c, d) electric field distribution of the

200nm, 400nm and 600nm pillar arrays using the fabricated model and simulated by the

FDTD method.

3.1.3 Influence of fabrication errors

During the actual fabrication process, displacement of voxel tracks and distortion of

the voxel shape is inevitable due to mechanical vibrations and the accuracy of the

printing device, fluctuation of the femtosecond laser power, a change of

environmental temperature, etc. To understand the influence of fabrication errors, we

performed FDTD simulations considering changes of pitch, height and diameter

values of the nano-pillar array, as shown in Figure 3.6. For a 400nm pillar array, a

pitch of 50nm less (b) than the original design (a) will increase the amplitude of the

electric field from 7 to 7.5, whereas 100nm less pitch (c) will increase the amplitude

of the electric field to 9.6. As a result, according to the |𝐸𝑙𝑜𝑐(𝜔𝑅)|4 approximation,

the SMEF for 400nm pillar array with 50nm and 100nm displacement in pitch are 1.3

times and 3.5 times more than that of the original, respectively. The errors in the height

of the pillar will also change the distribution and intensity of the electric field (d, e

and f). If there is a displacement between two adjacent layers in the fabrication

process, the diameters of the pillars will be different than our expectation. In some

cases, the electric field pattern is similar to that of the pitch errors (g and h), while in

other cases the distribution of hotspots is totally relocated from the troughs to the top


88

areas (i). The simulation results indicate that the SMEF of the nano-pillar arrays is

more sensitive to horizontal fabrication errors than to vertical errors.

Figure 3.6. Electric fields of 400nm pillar array in voxel-based model without (a) and with

fabrication errors (b-i). 50nm and 100nm closer pitch (b, c) increase the electric field. 50nm

height error (d) and 100nm height error (e) of one pillar, and all pillars with 50nm height error

(f) have little impact on the electric field. Diameter errors due to displacement of different

adjacent layers (g-i) have big impact to the relocation and intensity change of electric fields,

as these also reduce the pitch values.

In addition to the errors of the macro profile, the substructures such as nano-cavities,

bulges or spikes on top of the nanostructures may result in much more complex

enhancement patterns.

Chapter 3

89

3.2 Experiments with the printed SERS substrates

3.2.1 SERS enhancement analysis

To experimentally verify the enhancement factors of the two-photon polymerized

SERS substrates, we prepared 10µM Rhodamine B solutions in ethanol and water

respectively and performed Raman measurements utilizing a confocal Raman

microscope (Bruker Optics - Senterra). This spectrometer is equipped with a 785nm

laser which is the same wavelength as has been used in our FDTD simulations. The

experimental enhancement factor, or the analytical enhancement factor (AEF) can be

calculated according to the following equation36:

𝐴𝐸𝐹 = (𝐼𝑆𝐸𝑅𝑆/𝑁𝑆𝐸𝑅𝑆)/(𝐼𝑅𝑒𝑓/𝑁𝑅𝑒𝑓) （3.3）

Where 𝐼𝑆𝐸𝑅𝑆 and 𝐼𝑅𝑒𝑓 are the intensities of the SERS signal and conventional Raman

signal respectively. 𝑁𝑆𝐸𝑅𝑆 stands for the number of molecules that induce a SERS

signal, and 𝑁𝑅𝑒𝑓 is the number of molecules that generate normal Raman scattering.

Figure 3.7. (a) Raman background of the 200nm pillar array; (b) Average Raman spectra of

10µM RhB obtained with 200nm pillar arrays. Baselines are corrected by subtracting the

background Raman signal. Spectra are obtained under a 785nm wavelength and 2.5mW laser

excitation with 1 second integration time. The dashed lines refer to the upper and lower range

of the spectra. Average spectrum is obtained over 16 measurements.

We obtained the background spectrum of the 200nm pillar array for a reference, as

shown in Figure 3.7(a). Next, we placed 30µL of the Rhodamine B ethanol solution

on the SERS substrates, and measured the Raman spectra in the dry state under a

2.5mW excitation with a 50× objective lens. The sample deposited area (glass plate)

is much larger than the SERS sensitive area (50µm by 50µm) and the SERS sensitive

area is in the middle of the base plate where the concentration of the molecule under

test is lower compared to the ring but more uniform. We can estimate 𝑁𝑆𝐸𝑅𝑆 with

reference to the laser spot size and deposition area under the assumption that the


90

Rhodamine B molecules are evenly delivered on the overall SERS substrate. The

reference measurements are conducted by analyzing the Raman scattering of a

Rhodamine B aqueous solution since water gives very weak Raman background. We

calculated the 𝑁𝑅𝑒𝑓 inside the interaction probe volume of the 50× objective lens

under the same excitation. However, we must be aware that the local AEF with

inhomogeneous distribution of the molecules may be different than the overall AEF

of the SERS substrate, which may be influenced by parameters such as the uniformity

of the nano-pillar array, the adsorption of the molecules and the surface tension of the

SERS substrate. Figure 3.7(b) shows the SERS spectra of Rhodamine B on a 200nm

nano-pillar array after subtracting the background of the SERS substrate. Rhodamine

B molecules present distinct enhancement features at 629cm-1, 1203cm-1, 1287cm-1,

1363cm-1, 1511cm-1 and 1600cm-1. The peaks of the Rhodamine B’s spectra

correspond to the molecular vibrational modes as listed in Table 3.2.

Table 3.2 The intensities and peak assignments of main Raman bands of Rhodamine B.

Literature37–39 Raman (cm-1) SERS (cm-1) Assignment

619 S 622.5 S 629.5 S ν (Aromatic C-C)

965-980 W 978 W 969 W δ (Ethylene C-H)

1065-1085 W 1080 M 1086 M δ (Aromatic C-H)

1130 W 1115 W 1126 W δ (Aromatic C-H)

1199 M 1198 S 1203 S δ (Aromatic C-H)

1284 S 1281 S 1287 S δ (C-C)

1360 S 1359 S 1363 S ν (Aromatic C-C)

1508 S 1508 S 1511 S ν (Aromatic C-C)

1591 W 1595 W 1600 S ν (C=C)

1644 S 1647 S 1652 W ν (Aromatic C-C)

(ν: stretching, δ: deformation; W: weak; M: medium; S: strong)

The Raman peak of Rhodamine B at 1363cm-1 shows the highest intensity. Therefore,

we calculate the AEF corresponding to the 𝐼𝑆𝐸𝑅𝑆 and 𝐼𝑅𝑒𝑓 at 1363cm-1. Figure 3.8 is

the comparison of enhancement factors for the three different models obtained by

FDTD simulations and the experimental results. A maximum enhancement factor

close to 104 is achieved by our 200nm nano-pillar array. It is reasonable that smaller

structures give rise to a higher enhancement as they can induce stronger oscillation of

localized surface plasmons. The voxel-based model shows the best consistency with

experimental results. However, one interesting phenomenon we can observe is that

the experimental EFs are larger than the simulated EFs except for the 200nm and

300nm voxel-based models. This can be explained by the relocation and intensity

changes of hotspots due to the fabrication errors mentioned before. It is hard to predict

the precise influence of the fabrication errors over the entire SERS substrate, but we

Chapter 3

91

believe that the FDTD simulation of the fabricated shape model can give rise to a

better reflection of SERS performance if an AFM with higher resolutions and smaller

probe size can be used. In addition, because the major boosted electric field of them

is generated in the edge area, the EFs of the 300nm to 600nm nano-pillar arrays in the

nominal shape model show less differences.

Figure 3.8. Comparison of enhancement factors for the different models obtained by FDTD

simulations and the experimental result. The average and standard deviation of each

enhancement factor are obtained over 16 measurements on a different detection area of a

nano-pillar array.

Figure 3.9. Comparison of 10µM RhB Raman spectra on different substrates and the Raman

spectrum of pure RhB.

We also performed benchmark measurements of Rhodamine B with two types of

commercial SERS substrates from Silmeco (Copenhagen, Denmark) and Horiba, Ltd

(Kyoto, Japan) respectively, see Figure 3.9. The enhancement factor of our 200nm

pillar array is two orders of magnitude less than the 106 EF of Silmeco, which can be

proven by the spectra. It’s noteworthy that our SERS substrate is comparable with


92

Horiba’s substrate and performs even better with a higher sensitivity. Therefore, our

two-photon polymerized SERS substrate can be used as a powerful tool to study the

vibrational modes of molecules.

Additionally, other types of periodic nanostructures, including nano-hemisphere

arrays and nano-grids have been printed by two-photon polymerization as well, as

shown in Figure 3.10. But the Raman performance of these SERS substrates is less

than that of the nano-pillar arrays both in simulations and experiments. Therefore, in

our later experiments we utilize only the 200nm pillar arrays.

Figure 3.10. SEM images of 200nm, 400nm and 600nm hemisphere arrays (a-c) and nano-

grids with 200nm, 400nm and 600nm spacing (d-f).

3.2.2 SERS substrate calibration

In order to further evaluate the performance of our two-photon polymerized SERS

substrates, we measured the Raman spectra of Rhodamine B with different

concentrations from 0.5µM to 1000µM. The calibration curves of the peak intensity

with respect to the concentration are shown in Figure 3.11(a). The peak intensity rises

as the concentration increases, but the rate of increment gradually falls off. From a

certain threshold of concentration (~750µM), the peak stops increasing and starts to

decrease. Because the SERS effective area is just within a few hundred nanometers,

the number of hotspots is limited. The molecules can quickly fill the hotspots of the

SERS substrate and emit enhanced Raman scattering under low concentrations.

However, as the concentration increases, the number of remaining hot spots begins to

decrease with the accumulation of molecules on the metallic surface. And eventually

a thick layer is formed that prevents the SERS signal underneath. This can be proved

by the microscope images of the original SERS substrate and the SERS substrate with

1mM Rhodamine B deposition in Figure 3.11(b, c).

Chapter 3

93

Figure 3.11. Calibration curves of the 200nm nano-pillar array showing the peak intensities at

629cm-1, 1287cm-1 and 1363cm-1 with respect to the RhB concentrations (a). Molecules can

easily adsorb on the SERS substrate in the low concentration condition (b), but a thick layer

of molecules with high concentration will restrain the SERS signal (c). The RhB molecules

cover both the nanostructures and the surrounding flat surfaces, for which the interaction of

molecules with nanostructures has changed the color of the RhB from red violet to green

under white light illumination due to a higher reflectivity at the green light band.

The detection limit of our 200nm pillar array for Rhodamine B is estimated to be

0.55µM (5.7 ppm) according to the linear regression of the calibration curve in the

low concentration region40.

3.3 An application of two-photon polymerized SERS substrates

3.3.1 Mycotoxin detection with 2PP polymerized SERS substrates

SERS is an ideal technique for biochemistry and toxicology because of its low laser

intensity, short integration time and small sampling volume required. As proof of

concept for our two-photon polymerized SERS substrate, we tested two types of

mycotoxins, Deoxynivalenol (DON) and Fumonisin b1 (FUM). Deoxynivalenol, also

known as vomitoxin, is a metabolite of fusarium graminearum which can cause

anorexic effects on humans41. Fumonisin is a metabolite of fusarium verticillioides,

one of the most prevalent seed-borne fungi associated with maize42. Long-term intake

of Fumonisin contaminated food can greatly increase the risk of esophageal cancer.

According to the EU’s Mycotoxins Factsheet43, the limit of deoxynivalenol in

unprocessed cereals is 1.25ppm, and the limit of Fumonisin in unprocessed maize is

4ppm. We measured the Raman spectra of 1ppm deoxynivalenol and 1.25ppm


94

Fumonisin b1 in acetonitrile with the same equipment under the same excitation as

for Rhodamine B. The average spectra of Fumonisin b1 and Deoxynivalenol obtained

with our 200nm pillar arrays are shown in Figure 3.12.

Figure 3.12. Raman spectra of 1ppm DON and 1.25ppm FUM obtained with our 200nm pillar

array SERS substrate under a 785nm wavelength and 2.5mW laser excitation with 1 second

integration time. The concentrations of mycotoxins are close to the detection limit of our

SERS substrates and therefore it is difficult to recognize the mycotoxins directly.

3.3.2 Principal Component Analysis (PCA) for the spectra of mycotoxins

The concentrations of mycotoxins are close to the detection limit of our SERS

substrates. Therefore, it is difficult to recognize the DON and FUM directly by

comparing the spectra, although there are some distinguishing peaks present, such as

the 1452cm-1 and 1140cm-1 peaks of the DON spectrum, and the 873cm-1 and 1775cm-

1 peaks of the FUM spectrum respectively. We employed Principal Component

Analysis (PCA) towards a set of Raman spectra observed to discriminate them44. PCA

is one of the most common methods of multivariate analysis for classification

purposes which uses the spectral decomposition of a correlation coefficient. After

PCA transformation, we find that the first three principal components represent 85.9%

of the variance, and the two mycotoxins can be clearly distinguished by the second

and third principal components, as shown in Figure 3.13 (a). Raman spectra of DON

and FUM show good repeatability with groups of replies, which allow to detect

sensible between-sample differences.

Chapter 3

95

Figure 3.13. (a) Spectra of DON and FUM are clustered by the scores of the second and third

principal components. (b)-(c) Coefficients of principal components 1-3 which represent

85.9% of the variance.

By analyzing the coefficients of the second and third principal components which contains

fingerprint Raman shifts information of the two mycotoxins, as shown in Figure 3.13 (b),

combined with literature results, we obtained the characteristic peaks of the two mycotoxins

and the corresponding assignments, see Table 3.3.

Table 3.3 The characteristic peaks and assignments of main Raman bands of FUM and DON

according to PCA.

Mycotoxin Literature45–47 (cm-1) SERS (cm-1) assignment

FUM 760 754 ν (C-C)

868 873 ν (C-O-C)

1460 1466 δ (-CH3)

1488 1482 δ (C-H)

1776 1775 ν (C=O)

DON 780 787 ν (O-H) + ν (C-H)

855 852 ν (C-H)

923 927 ν (-CH3) + ν (C-H)

1139 1140 ν (C-H)

1293 1287 ν (C-H)

1430 1435 δ (C=C) + ν (-CH3) + ν (C-H)

1449 1452 ν (-CH3)

(ν: stretching, δ: deformation)


96

3.4 Conclusions

We designed and fabricated a series of nano-pillar arrays as SERS substrates ranging

from 200nm to 600nm by using two-photon polymerization and gold sputtering. To

verify the feasibility of these nano-arrays as SERS substrates, we setup a nominal

shape model, a voxel-based model, and simulated the induced electric field under

785nm excitation alongside the nanostructures with the FDTD method. We

characterized the morphologies of the different nano-pillar arrays from a

comprehensive analysis of SEM images and AFM 3D data. The fabricated shape

model based on AFM data is defined as input for the FDTD simulations to determine

the induced electric field. The single molecule enhancement factor (SMEF) of

different models are calculated based upon the FDTD results.

We estimated the experimental enhancement factor (AFE) by analyzing the Raman

scattering of 10µM Rhodamine B solutions in ethanol and water. A maximum

enhancement factor closes to 104 is achieved with the 200nm pillar array. Benchmark

measurements have shown that obtained results of our two-photon polymerized SERS

substrates are comparable with the ones obtained with the commercial SERS

substrates. We compared the experimental enhancement factor with the SMEFs

obtained using different models for FDTD simulations, finding that the voxel-based

model gives the best consistency with the experimental results. In addition, we

analyzed the reasons for the differences between simulations and experiments.

Moreover, we evaluated the detection limit of the 200nm nano-pillar array SERS

substrates using different concentrations of Rhodamine B solutions and found the

limit to be 0.55µM. To demonstrate the proof-of-concept of our SERS substrates in

an application, we detected the Raman spectra of 1ppm deoxynivalenol and 1.25ppm

fumonisin b1 solutions. The two types of mycotoxin are discriminated by principal

component analysis (PCA). Our two-photon polymerized nano-pillar arrays pave the

way for fast prototyping of SERS substrates for biochemical and toxicological

research.

The limitation of two-photon polymerization is the restricted structure-sizes of the

nano-pillar arrays. But the enhancement factor of the SERS substrates can still be

increased by optimizing the nanostructures. Increasing the homogeneity of the

nanostructures should be further investigated with respect to parameters such as the

optimization of femtosecond laser power fluctuations, the photoresins and the voxel

path compiling. Finally, the quantitative analysis capabilities of our 2PP printed SERS

substrates should in future be investigated using other analytes such as crystal violet

(CV), p-mercaptobenzoic acid (p-MBA), 1,2-bis (4-pyridyl)-ethylene (BPE) and

biological samples.

Chapter 3

97

References

1. Michaels, A. M., Nirmal, M. & Brus, L. E. Surface enhanced Raman

spectroscopy of individual rhodamine 6G molecules on large Ag nanocrystals.

J. Am. Chem. Soc. 121, 9932–9939 (1999).

2. Le Ru, E. C., Meyer, M. & Etchegoin, P. G. Proof of single-mokeule

sensitivity in surface enhanced Raman scattering (SERS) by means of a two-

analyte technique. J. Phys. Chem. B 110, 1944–1948 (2006).

3. Lee, S. et al. Rapid and sensitive phenotypic marker detection on breast

cancer cells using surface-enhanced Raman scattering (SERS) imaging.

Biosens. Bioelectron. 51, 238–243 (2014).

4. Chon, H., Lee, S., Son, S. W., Oh, C. H. & Choo, J. Highly sensitive

immunoassay of lung cancer marker carcinoembryonic antigen using surface-

enhanced raman scattering of hollow gold nanospheres. Anal. Chem. 81,

3029–3034 (2009).

5. Sha, M. Y., Xu, H., Natan, M. J. & Cromer, R. Surface-Enhanced Raman

Scattering Tags for Rapid and Homogeneous Detection of Circulating Tumor

Cells in the Presence of Human Whole Blood. J. Am. Chem. Soc. 130, 17214–

17215 (2008).

6. Song, H., Lee, S., Kang, T. & Kim, K. Silver nanoislands based SERS-active

substrates for DNAs detection. Prog. Biomed. Opt. Imaging - Proc. SPIE

9340, (2015).

7. Kang, T., Yoo, S. M., Yoon, I., Lee, S. Y. & Kim, B. Patterned Multiplex

Pathogen DNA Detection by Au Particle-on-Wire SERS Sensor. Nano Lett.

10, 1189–1193 (2010).

8. Bell, S. E. J. & Sirimuthu, N. M. S. Surface-Enhanced Raman Spectroscopy

( SERS ) for Sub-Micromolar Detection of DNA / RNA Mononucleotides. J.

Am. Chem. Soc. 128, 15580–15581 (2006).

9. Zhou, Y., Chen, J., Zhang, L. & Yang, L. Multifunctional TiO2-coated Ag

nanowire arrays as recyclable SERS substrates for the detection of organic

pollutants. Eur. J. Inorg. Chem. 2012, 3176–3182 (2012).

10. Aoki, P. H., Furini, L. N., Alessio, P., Aliaga, A. E. & Constantino., C. J.

Surface-enhanced Raman scattering ( SERS ) applied to cancer diagnosis and

detection of pesticides , explosives , and drugs. Rev. Anal. Chem. 32, 55–76

(2013).


98



12. Fleischmann, M., Hendra, P. J. & McQuillan, A. J. Raman Spectra of Pyridine

Adsorbed at a Silver Electrode. Chem. Phys. Lett. 26, 163–166 (1974).

13. Dou, X., Jung, Y. M., Cao, Z. Q. & Ozaki, Y. Surface-enhanced Raman

scattering of biological molecules on metal colloid II: Effects of aggregation

of gold colloid and comparison of effects of pH of glycine solutions between

gold and silver colloids. Appl. Spectrosc. 53, 1440–1447 (1999).

14. Knauer, M., Ivleva, N. P., Niessner, R. & Haisch, C. Optimized surface-

enhanced Raman scattering (SERS) colloids for the characterization of

microorganisms. Anal. Sci. 26, 761–766 (2010).

15. Dou, X., Jung, Y. M., Yamamoto, H., Doi, S. & Ozaki, Y. Near-Infrared

Excited Surface-Enhanced Raman Scattering of Biological Molecules on

Gold Colloid I: Effects of pH of the Solutions of Amino Acids and of Their

Polymerization. Appl. Spectrosc. 53, 133–138 (1999).




17. Guo, H., Xing, B., Hamlet, L. C., Chica, A. & He, L. Surface-enhanced

Raman scattering detection of silver nanoparticles in environmental and

biological samples. Sci. Total Environ. 554–555, 246–252 (2016).

18. Reznickova, A., Novotna, Z., Kasalkova, N. S. & Svorcik, V. Gold

nanoparticles deposited on glass: Physicochemical characterization and

cytocompatibility. Nanoscale Res. Lett. 8, 1–8 (2013).

19. Zhang, Z., Liao, F., Ma, S., Gao, S. & Shao, M. SERS and FDTD simulation

of gold nanoparticles grafted on germanium wafer via galvanic displacement.

Surf. Interface Anal. 47, 398–402 (2015).

20. Rodrıiguez-Lorenzo, L. et al. Intracellular mapping with SERS-encoded gold

nanostars. Integr. Biol. 3, 922–926 (2011).




22. Wang, H., Lai, K., Lin, Y., Lin, C. & He, J. Periodic Si Nanopillar Arrays

Fabricated by Colloidal Lithography and Catalytic Etching for Broadband and

Chapter 3

99

Omnidirectional Elimination of Fresnel Reflection. Langmuir 26, 12855–

12858 (2010).

23. Kuo, C., Shiu, J., Cho, Y. & Chen, P. Fabrication of Large-Area Periodic

Nanopillar Arrays for Nanoimprint Lithography Using Polymer Colloid

Masks. Adv. Mater. 13, 1065–1068 (2003).

24. Schmidt, M. S., Boisen, A. & Jörg, H. Towards Easily Reproducible Nano-

structured SERS substrates. in IEEE Sensors Conference 1763–1767 (2009).

doi:10.1109/ICSENS.2009.5398468



24, OP11–OP18 (2012).

26. Mu, J. et al. Direct laser writing of symmetry-broken nanocorrals and their

applications in SERS spectroscopy. Appl. Phys. B Lasers Opt. 117, 121–125

(2014).




28. Zhou, Q. et al. A Surface-Enhanced Raman Scattering Sensor Integrated with

Battery-Controlled Fluidic Device for Capture and Detection of Trace Small

Molecules. Sci. Rep. 5, 12865 (2015).

29. Cumpston, B. H. et al. Two-photon polymerization initiators for three-

dimensional optical data storage and microfabrication. Nature 398, 51–54

(1999).



1–7 (2013).

31. Obata, K., El-tamer, A., Koch, L., Hinze, U. & Chichkov, B. N. High-aspect

3D two-photon polymerization structuring with widened objective working

range ( WOW-2PP ). Light Sci. Appl. 12, 8–11 (2013).

32. Maher, R. C., Cohen, L. F., Le Ru, E. C. & Etchegoin, P. G. On the

experimental estimation of surface enhanced raman scattering (SERS) cross

sections by vibrational pumping. J. Phys. Chem. B 110, 19469–19478 (2006).

33. Davies, R. A., Chong, N. S. & Ooi, B. G. Chemical Enhancement of the

Surface Enhanced Raman Scattering Signals of Anilines via Their Ortho-


100

Substituents. Opt. Photonics J. 03, 13–23 (2013).

34. Lombardi, J. R., Birke, R. L., Lu, T. & Xu., J. Charge-transfer theory of

surface enhanced Raman spectroscopy : Herzberg-Teller contributions. J.

Chem. Phys. 84, 4174 (1986).

35. Creighton, J. A. Surface raman electromagnetic enhancement factors for

molecules at the surface of small isolated metal spheres: The determination of

adsorbate orientation from sers relative intensities. Surf. Sci. 124, 209–219

(1983).



Chem. C 111, 13794–13803 (2007).



12544–12548 (2005).

38. Luo, Z., Fang, Y. & Yao, J. A New Approach for Non-destructive Detection

of Dye Molecules by Combination of Terahertz Time-domain Spectra and

Raman Spectra. Trends Appl. Sci. Res. 2, 295–303 (2007).

39. Bio-Rad. KnowItAll ® AnalyzeIt MVP. (2019). Available at: http://www.bio-

rad.com/webroot/web/pdf/spectroscopy/global/english/literature/docs/95392

-Bio-Rad_KnowItAll_Software_AnalyzeIt_MVP_PCA_Datasheet.pdf.

40. Shrivastava, A. & Gupta, V. B. Methods for the determination of limit of

detection and limit of quantitation of the analytical methods. Chronicles

young Sci. 2, 21–25 (2011).

41. Pestka, J. J. & Smolinski, A. T. Deoxynivalenol : Toxicology and Potential

Effects on Humans. J. Toxicol. Environ. Health 8, 39–69 (2015).

42. Marasas, W. F. O. Discovery and Occurrence of the Fumonisins : A Historical

Perspective. Environ. Health Perspect. 109, 239–243 (2001).

43. Lerda, D. Mycotoxins Factsheet. (European Union, 2011).

44. Emge, D. et al. Bacillus Spore Classification via Surface-Enhanced Raman

Spectroscopy and Principal Component Analysis. Appl. Spectrosc. 62, 267–

272 (2008).



Chapter 3

101








102

Chapter 4

103

Chapter 4

4 Integrated Confocal Raman Probe Combined with a

Freeform Reflector Embedded Lab-on-Chip

4.1 Introduction to a freeform reflector-based confocal Raman

spectroscopy lab-on-chip system

Miniaturized optical components, such as lenses, filters, optical fibers and mirrors can

be integrated to a microfluidic Raman lab-on-chip device to increase the excitation

and collecting efficiencies.1–4 Diane De Coster, a former PhD student in our research

group has developed a freeform reflector-based confocal Raman spectroscopy lab-on-

chip system for material characterization.4 As discussed in Chapter 1, one of the

objectives in this PhD is to further miniaturize this Raman spectroscopy experimental

setup and to enhance the Raman signal. To start with, this section gives an overview

of the confocal Raman spectroscopy lab-on-chip system developed by Diane De

Coster.

Part of this chapter has been submitted as:

Liu, Q., Barbieri, G., Thienpont, H., Ottevaere, H. (2017). Integrated confocal Raman probe

combined with a free-form reflector based lab-on-chip. In Optical System Alignment,

Tolerancing, and Verification XI (Vol. 10377). [1037706] SPIE. DOI: 10.1117/12.2274936

Liu, Q., De Coster, D., Loterie, D., Van Erps, J. A., Vervaeke, M., Missinne, J., Ottevaere, H.

(2016). Proof-of-concept demonstration of free-form optics enhanced confocal Raman

spectroscopy in combination with optofluidic lab-on-chip. In Micro-Optics 2016 (Vol. 9888).

[UNSP 98880E] SPIE. DOI: 10.1117/12.2227386

De Coster, D., Liu, Q., Vervaeke, M., Van Erps, J.A., Missine, J., Thienpont, H., Ottevaere, H.

(2017). Optofluidic chip for single-beam optical trapping of particles enabling confocal Raman

measurements. IEEE J. Sel. Top. Quantum Electron., 23(2), [5500109].

Chapter 4 Integrated Confocal Raman Probe Combined with a Freeform Reflector Embedded Lab-on-Chip

104

4.1.1 Confocal principle

Confocal microscopes are widely used in biological research for studying cell

structures as they can provide background-free images with high resolution by

rejecting the out-of-focus light. Typically, in a confocal system, a pinhole is placed at

the image plane of the lens, as shown in Figure 4.1. Light emitted from a source in the

focal plane can be focused by the lens to a spot and collected inside the pinhole. The

out-of-focus emissions will be strongly attenuated or rejected by the pinhole.

Therefore, a confocal system can suppress the background signals from the out-of-

focus regions to increase the quality of the wanted signal for a sample under test.

Figure 4.1 Illustration of confocal principle.4

4.1.2 Design of the confocal Raman spectroscopy LoC setup

The confocal Raman spectroscopy setup Diane De Coster has developed consists of a

freeform reflector embedded LoC and external optics for microfluidic measurements4,

as shown in Figure 4.2. The LoC on the bottom of the setup consists of three PMMA

layers. The 200µm thick bottom PMMA layer is embedded with a freeform reflector

coated with a reflective gold layer. The diameter and sag of the reflector are 1.684mm

and 300µm respectively. The 500µm thick middle layer consists of a 500µm width

fluidic channel and a chamber with a diameter slightly bigger than that of the reflector.

The focal point of the reflector is located in the center of the fluidic chamber, and

50µm above the bottom of the channel layer. The 500µm thick top layer is the sealing

layer for the fluidic channel. All the three layers are bonded together by UV curing

adhesive. The LoC enables a very high numerical aperture with aqueous flow filled

inside the chamber (NA=1.28) and allows a high collection efficiency of the Raman

scattering.

In this setup, the freeform reflector in the LoC and the collection lens work together

as the lens, and the core facet of the collection fiber works as the pinhole during the

confocal measurement, respectively. The confocal Raman spectroscopy setup consists

of an excitation and a collection path. In the excitation path, a 785nm wavelength laser

coming out of the single-mode fiber is collimated and reshaped by a set of lenses (L1)

Chapter 4

105

to a light beam of which its diameter is in line with that of the freeform reflector. The

collimated beam is then reflected by a laser line mirror to the LoC, and focused by the

freeform reflector to the focal point inside the fluidic chamber. The Raman scattering

will be induced where the laser light excites the sample under test due to the

vibrational modes of the molecules of the analytes in the flow. In the collection path,

the scattered Raman signal collimated by the freeform reflector passing through the

laser line mirror can be collected into the multi-mode collection fiber by the collection

lens (L2). A notch filter is placed in front of the collection lens to attenuate the

Rayleigh scattering of the laser light. The light bundle of the Raman scattering

originating from out-of-focus sources cannot pass through the fiber facet because of

the confocal principle. In this case, the Raman scattering of the out-of-focus regions,

especially from the polymer layers will be suppressed significantly.

Figure 4.2 (Left) schematic of the confocal Raman spectroscopy setup with a freeform

reflector embedded LoC. (Right) Geometrical parameters of the LoC.4

The shape of the freeform reflector is a derivation of a parabolic mirror by taking into

account the refractive transition between the polymer layer and the fluids inside the

chamber according to Fermat’s principle. The surface profile of the freeform reflector

is calculated by a numerical approach. The calculated shape is then introduced in the

Advanced Systems Analysis Program (ASAP) to perform non-sequential ray tracing

simulations together with the three PMMA layers and the other external optical

components. In the non-sequential ray tracing, the sensing area is defined as an area

with a length of 80µm in the z-direction around the focal point. Each ray from the

excitation is supposed to generate multiple Raman scattering sources homogenously

along its optical path in the sensing area. The positions of the Raman sources for all

excitation rays are selected randomly using a Poisson distribution as the position of

each Raman source should be independent of that of the other Raman sources. Figure


106

4.3 shows a non-sequential ray tracing simulation of excitation and the generation of

Raman scattering. The non-sequential ray tracing simulations demonstrate the

confocal behavior and give us the alignment tolerances of the Raman spectroscopy

system, which are used as input in the implementation phase of the setup.

Figure 4.3 ASAP simulation of the (a) excitation and (b) generation of Raman scattering in

the microfluidic channel.4

4.1.3 Implementation of the confocal Raman spectroscopy LoC setup

Chip fabrication

After designing the freeform reflector and performing non-sequential ray tracing

simulations, the freeform reflector is fabricated in PMMA with in-house ultra-

precision diamond tooling using a Nanotech 350FG CNC machine. The convex

surface of the fabricated reflector is coated with a 200nm thick reflective Au layer by

thermal evaporation coating, performed at the Center for Microsystems Technology

(CMST) of UGent (Prof. J. Missine). In a next step, the channel layer and sealing layer

of the optofluidic chip are fabricated by micro-milling and bonded together with the

reflector embedded bottom layer by a UV curing adhesive. The PMMA in- and outlet

connections for the fluidics are UV cured, the fused silica tubing is used to transfer

the sample flow. The final chip is shown in Figure 4.4.

Figure 4.4 (a) Schematic drawing and (b) photograph of the LoC with integrated freeform

reflector. 4

Chapter 4

107

Proof-of-concept demonstration

A proof-of-concept demonstration setup is built in our lab by aligning the freeform

reflector embedded microfluidic LoC and the external optics including the lenses,

fibers, filters and mirrors, as shown in Figure 4.5. A single-mode fiber with a core size

of 5µm and a NA of 0.12 is used to couple the 785nm wavelength diode laser light

into the setup. Raman signals are captured through a multi-mode fiber with a core

diameter of 200µm, connected to a spectrometer with a slit width of 25µm. The noise-

equivalent concentration (NEC) that corresponds to a SNR of 1 is approximately

20mM for the detection of urea aqueous solution when using an integration time of

15 seconds and a laser power of 190mW.

Figure 4.5 Photographs of the confocal Raman spectroscopy setup. (a) side view; (b) front

view.4

4.1.4 Conclusion

Diane De Coster designed a freeform reflector embedded microfluidic LoC for

confocal Raman measurements. The confocal performance and alignment tolerances

were assessed by non-sequential ray tracing simulations using ASAP®. The shape of

the freeform reflector was calculated via a numerical approach and fabricated by an

ultra-precision diamond tooling approach. The fabricated LoC was integrated with

external optics to implement a proof-of-concept demonstration setup for microfluidic

Raman spectroscopy measurements. The experiments with this original confocal

Raman spectroscopy LoC setup demonstrate that this Raman system allows robust

confocal measurements with a NEC of 20mM.

Although PMMA polymer as material for the LoC is considerably more cost-efficient

compared to the quartz chip used in the literature5,6, the direct fabrication approach


108

with ultra-precision diamond tooling is still relatively expensive in terms of labor and

equipment costs. In addition, the NEC of this setup is a bit higher compared with the

one of the quartz chip. This difference can be a result of a higher excitation power and

a spectrometer with a higher sensitivity they utilized in combination with the quartz

chip.4 Therefore, in the next steps, we work, as part of this PhD, on the optimization

of the external optics to implement an integrated Raman probe to further lower the

detection limit and allow mass fabrication of the LoCs to reduce the cost of a

microfluidic LoC analysis.

4.2 Integrated Confocal Raman probe optimized for microfluidic

lab-on-chip

4.2.1 Concept of Raman probe and state-of-the-art

Many application domains require remote sampling by using compact and portable

devices, and often optical instruments are the best choice. A good example hereof are

endoscopes for the diagnosis of diseases and minimally invasive or non-invasive

surgeries in a clinical setting.7–9 Endoscopes are generally utilized by imaging the

morphological features of organs or cells. Typically, an endoscope consists of an

illumination and an imaging path, similar to a Raman spectroscopy setup that also

consists out of an excitation and a collection path. In fact, Raman probes have already

been investigated and utilized occasionally for the in vivo diagnosis of cancers since

the Raman spectra obtained with label-free and contactless detection approaches

provide much more useful information of the status of the human body.10–19 The

dimensions of the Raman probe for clinical applications are strictly constrained by the

anatomical requirements, which are usually less than 2mm. In addition, Raman probes

for diagnosis and non-invasive surgeries require a flexible probe head and body as to

penetrate through the natural orifices and tracts such as esophagus and intestinal tracts.

Undeniably, fiber optic-based Raman probes can best meet these requirements. Figure

4.6 shows the schematics of a Raman probe configured with optical fibers. The

minimum size of a Raman probe can be achieved by using a single multi-mode optical

fiber with flat-tip to propagate the excitation light and collect the Raman signals.10,11,20

A dichroic filter is used to split the excitation and collection paths in this configuration.

On the other hand, two flat-tipped optical fibers can be employed for the excitation

and collection paths respectively. However, the collection efficiency of any of the

above configurations is very low. A common way to reinforce the collection

efficiency is to surround the excitation fiber with a bunch of fibers so that the

overlapping areas of the excitation path and collection path on the sample can be

Chapter 4

109

increased14,17, as shown in Figure 4.7. Beveled-tip fiber probes are reported to have

significantly higher collection efficiencies compared with the flat-tipped fiber probes

in virtue of the best overlapping between the excitation beam spot and the detection

areas.12 Another way to reinforce the collection efficiency is to place a lens in front of

the probe tip to focus the excitation beam as well as collect the Raman scattering16,17,

as shown in Figure 4.8.

Figure 4.6 Layouts of various fiber optic-based Raman probes10.

Figure 4.7 (A) Schematic of a Raman endoscopy setup for in vivo diagnostics. (B)

Photograph of the Raman probe. (C) The Raman probe used during clinical examination.14


110

Figure 4.8 Cross-sections of a fiber-based Raman probe with a ball lens to enhance the

collection efficiency.16

Besides of the fiber optic-based Raman probe, Raman probes with miniaturized

optical components and optimized layouts have also been reported for clinical

diagnosis and other application domains.18,19,21 The fundamental optical components

are optical fibers and filters, lenses and mirrors, as shown in Figure 4.9 and Figure

4.10. Typically, the dimensions of these types of Raman probe with miniaturized

optical components are a bit larger than the fiber optic-based probes, but they allow

Raman detection with higher excitation and collection efficiencies, as well as higher

external and internal tolerances. In addition, the working distance of a miniaturized

optics-based Raman probe is much longer, which makes remote detection possible.

Fiber optics with micro-lenses and filters have been investigated to achieve the Raman

detection goals with minimal sizes and high efficiencies.21

Figure 4.9 Schematic of a Raman probe design for epithelial tissue detection. EP: Epithelium;

ST: Stroma.19

Chapter 4

111

Figure 4.10 Layout of a handheld Raman probe.21

In our case, we aim for developing a compact Raman probe for a microfluidic LoC

analysis. Although the Raman spectrum can provide “fingerprint” information of a

sample under test, the intensity of the induced Raman scattering is rather low

compared to the excitation light which indicates that the Raman signal could easily be

obscured by the background noise.22 A way to enhance the Raman efficiency is to

optimize both the excitation and the collection path. Our Raman probe design is based

on an existing Raman lab-on-chip made out of polymer material as described in the

previous sections. By combining the Raman probe and LoC we achieve a “confocal

structure” in which the back-ground noise outside the interesting area can be

suppressed significantly. The fiber tip works as a pinhole and the reflector works as a

confocal lens in a confocal system.

Although there are constrains on the dimensions of a Raman probe in virtue of the

functional area of the LoC, the requirement to the Raman probe’s size is not that strict

compared to clinical endoscopic devices. We determine to utilize miniaturized optics

in our design as it allows robust Raman measurements which can be introduced to

microfluidic devices when needing high excitation and collection efficiencies, high

internal and external tolerances, long working distances and large freedoms in

maintenance.

4.2.2 Design considerations of an integrated Raman probe for confocal

Raman lab-on-chip measurements

Different layouts of a miniaturized optics-based Raman probe exist and each one may

be distinguished from each other by the properties of the optical and mechanical

components.18,19,21 In general, they can be classified into two types: the collection-

prioritized configuration and the excitation-prioritized configuration. The type of


112

optical components used in the two configurations are almost the same. The main

difference between them involves the dichroic mirrors and optical path lengths of the

excitation and collection paths. In the collection-prioritized configuration, a long-pass

dichroic mirror is used, and the optical path length of the excitation path is longer than

that of the collection path due to an additional optical path length of the excitation

path between the mirror and dichroic mirror. While in the excitation-prioritized

configuration, a short-pass dichroic mirror is used, and the optical path length of the

excitation path is shorter than that of the collection path. The shorter the optical path

length and the lower the number of optical surfaces or optical components, the lower

the energy losses due to absorption and scattering of the optical components and

media, and the better the alignment tolerances. Therefore, the excitation-prioritized

configuration has a higher excitation efficiency and lower collection efficiency

compared with the ones of the collection-prioritized configuration and vice versa. The

two configurations of the Raman probe design for our microfluidic lab-on-chips are

shown in Figure 4.11. We investigate both configurations to find an optimal layout

for the Raman probe.

Figure 4.11 (a) Collection-prioritized configuration and (b) excitation-prioritized

configuration of the Raman probe for the freeform reflector-based microfluidic chip. The red

and blue rays refer to the excitation and collection path respectively.

The Raman probe performance largely depends on the quality of the selected optical

components and their characteristics. The role of optical fibers is to guide the

excitation light from the laser source to the sensing area (excitation fiber), and to carry

the Raman signal originated in the sensing area towards the spectrometer (collection

Chapter 4

113

fiber). Materials of the fiber core and cladding, profile dimensions of the fiber and the

length will largely influence the propagation performance of the light inside the fiber.

The major drawback of light propagation through an optical fiber is the background

noise generated from the fiber medium23,24. In this regard, optical filters are employed

for noise reduction according to their operation wavelength range25. Long-pass and

notch filters are the most commonly used filters present in the probe design. A band-

pass filter is employed along the excitation path to suppress the noise generated by

the light travelling through the excitation fiber and the side bands of the laser light.

On the other hand, either a notch filter for reducing Rayleigh scattering or a long-pass

filter for suppressing other sources of noise may be employed along the collection

path. For filtering the entire incoming radiation, these optical components need to

work properly over the range of angles covered by the NA of the fiber. The lens placed

in front of the excitation fiber works for collimating the laser light towards the sensing

area while the lens in front of the collection fiber focuses the Raman scattered light

into the collection fiber. In some cases, light collimated by the collimation lens

directly interacts with the sample under test to induce the Raman scattering. However,

detection lenses are always necessary in front of the sample to increase the Raman

scattering and enhance the collection efficiency. Thus, a proper lens selection allows

an efficient delivery and collection of light from the excitation path to the collection

path. To better understand the two layouts and the influences of the optical

components of the Raman probe, we performed non-sequential ray tracing simulations.

Non-sequential ray tracing simulations for the Raman probe

The optimum layout, the characteristics of the best optical components and the

tolerances of the Raman probe are simulated and determined using Breault Advanced

Systems Analysis Program (ASAP) and Matlab. ASAP is suited for performing non-

sequential ray-tracing simulations of the entire optical path: a light beam propagation

through each defined surface, its collimation and flux variations are monitored. Once

the incident light has reached the sensing area, then the ray position, direction and flux

are exported to Matlab. The latter software tool works only for simulating the effect

of the Raman scattering in the sensing area and it creates one Raman scattering source

for any location of the light rays. The generated Raman scattering is then imported

back into ASAP where the non-sequential ray tracing continues. The simulation

mechanism of Raman scattering including the definition of the sensing area, the

distribution and selection of the Raman sources are the same as described in the

previous section. The specifications of the optical components in our probe design are

listed in Table 4.1. We keep the excitation and collection fibers of the initial confocal

Raman LoC system as described in section 4.1 of Chapter 4, and modify the other


114

optical components such as the lenses and filters to further reduce the size of the

external optics towards a probe design.

Table 4.1 List of the optical components modelled in the simulation model and present in the

final probe design.

Optics Geometrical features Main features

Excitation fiber Core-size = 4.9 µm, NA = 0.12 Single mode fiber

Collection fiber Core-size = 200 µm, NA = 0.5 Multimode fiber

Excitation lens (EL) Clear aperture = 5.5 mm, Thickness =

2.17 mm, NA = 0.15

Aspheric lens

Collection lens (CL) Clear aperture = 8 mm, Thickness =

3.69 mm, NA = 0.5

Aspheric lens

Long-pass dichroic

mirror (DM)

Clear aperture = 8 mm, Thickness =

1.05 mm, Tilt = 45°

Average transmissivity>93%

@ (804 - 1600 nm)

Absolute reflectivity>94%

@ (780 - 790 nm)

Short-pass dichroic

mirror

(DM)

Clear aperture = 8 mm, Thickness =

1.05 mm, Tilt = 45°

Average reflectivity>93%

@ (804 - 1600 nm)

Absolute transmissivity >94%

@ (780 - 790 nm)

Mirror (M) Clear aperture = 8.6 mm, Thickness = 2

mm

Average reflectivity>98%

@ (500 - 2000) nm

Bandpass filter

(BPF)

Diameter = 12.5 mm, Thickness = 3 mm OD = 3 @ (300 - 1200 nm),

Average transmissivity>98% @

785 nm

Notch filter (NF) Diameter = 12.5 mm, Thickness = 3 mm OD = 6 @ 785 nm (FWHM = 20

nm)

Long-pass filter

(LPF)

Clear aperture = 10 mm, Thickness = 2

mm

Average transmissivity>96%

@ (815 - 1650 nm)

The simulation starts from a circular aperture, which simulates the core of the

excitation fiber in terms of NA, geometrical dimensions and material. The excitation

wavelength is 785nm and the power of each ray corresponds to 1 unit, an internal

parameter of the software which may be read as the maximum power available. The

two basic types of layout, excitation-prioritized and collection-prioritized

configuration, are simulated respectively. In the excitation-prioritized or short-pass

configuration, the 785nm wavelength excitation laser is collimated by the excitation

lens to a collimated beam and passes through a band-pass filter to narrow its band.

The beam then passes through the short-pass dichroic mirror and the three layers of

the microfluidic chip. Thanks to the freeform reflector embedded on the bottom layer

of the microfluidic chip, the collimated excitation beam is focused inside the fluidic

channel, where the Raman scattering is generated. The Raman sources in the sensing

Chapter 4

115

area are collimated by the freeform reflector and collected along the collection path

to the collection fiber. In the collection-prioritized or long-pass configuration, the

collimated 785nm wavelength laser beam is reflected by the reflective mirror and a

long-pass dichroic mirror successively to the microfluidic chip. The collimated Stokes

Raman scattering with longer wavelengths can pass through the long-pass dichroic

mirror and be collected along the collection path.

By adding a detection plane in the optical path, the ASAP software can calculate the

direction and the flux of each ray interacting with the optical components. Here we

define the excitation efficiency 𝜂𝑒𝑥𝑐𝑖𝑡𝑎𝑡𝑖𝑜𝑛 and collection efficiency 𝜂𝑐𝑜𝑙𝑙𝑒𝑐𝑡𝑖𝑜𝑛 as

suitable parameters for discussing and comparing the simulation results for the two

configurations.

𝜂𝑒𝑥𝑐𝑖𝑡𝑎𝑡𝑖𝑜𝑛 =𝜙𝑐ℎ𝑎𝑛𝑛𝑒𝑙

𝜙𝑖𝑛𝑐𝑖𝑑𝑒𝑛𝑡

× 100 % （4.1）

𝜂𝑐𝑜𝑙𝑙𝑒𝑐𝑡𝑖𝑜𝑛 =𝜙𝑓𝑖𝑏𝑒𝑟

𝜙𝑐ℎ𝑎𝑛𝑛𝑒𝑙

× 100 % （4.2）

Where, 𝜙𝑖𝑛𝑐𝑖𝑑𝑒𝑛𝑡 is the flux of incident light from the excitation fiber, 𝜙𝑐ℎ𝑎𝑛𝑛𝑒𝑙 is the

flux of incident light at the focus in the fluidic channel, 𝜙𝑓𝑖𝑏𝑒𝑟 is the flux of the Raman

scattering collected by the multi-mode fiber.

The total efficiency in the simulation is given by

𝜂𝑡𝑜𝑡𝑎𝑙 = 𝜂𝑒𝑥𝑐𝑖𝑡𝑎𝑡𝑖𝑜𝑛 × 𝜂𝑐𝑜𝑙𝑙𝑒𝑐𝑡𝑖𝑜𝑛 =𝜙𝑓𝑖𝑏𝑒𝑟

𝜙𝑖𝑛𝑐𝑖𝑑𝑒𝑛𝑡

× 100 % （4.3）

Figure 4.12 and Figure 4.13 show the excitation and collection path of the excitation-

prioritized and collection-prioritized configuration using non-sequential ray tracing

simulations, respectively. For the excitation path of the excitation-prioritized

configuration, the collimated excitation laser light passes thought the two surfaces of

the dichroic mirror and enters the chip, as shown in Figure. 4.12(a). For the excitation

path of the collection-prioritized configuration, the collimated excitation laser light is

reflected by the mirror and the reflective surface of the dichroic mirror successively

and then enters the chip, as shown in Figure 4.13(a). For the collection paths in Figure

4.12(b) and Figure 4.13(b), the light beams are slightly deviated because the Raman

sources from the out-of-focus regions are also considered in the non-sequential ray

tracing simulations. However, due to the confocality of the setup, these out-of-focus

Raman sources will be blocked by the apertures of the optical components, especially

by the core of the collection fiber.


116

Figure 4.12 (a) The excitation path and (b) collection path of the excitation-prioritized

configuration using non-sequential ray tracing simulations.

Figure 4.13 (a) The excitation path and (b) collection path of the collection-prioritized

configuration using non-sequential ray tracing simulations.

In the simulation for each configuration, 106 rays with 1-unit flux per ray are initiated

from the excitation fiber. The numerical simulations and the excitation and collection

efficiencies for both layouts are reported in Table 4.2.

Table 4.2 Simulation results of different layouts and the calculated efficiencies.

Configuration 𝜙𝑖𝑛𝑐𝑖𝑑𝑒𝑛𝑡 𝜙𝑐ℎ𝑎𝑛𝑛𝑒𝑙 𝜙𝑓𝑖𝑏𝑒𝑟 𝜂𝑒𝑥𝑐𝑖𝑡𝑎𝑡𝑖𝑜𝑛 𝜂𝑐𝑜𝑙𝑙𝑒𝑐𝑡𝑖𝑜𝑛 𝜂𝑡𝑜𝑡𝑎𝑙

Excitation-prioritized 100,000 88,281 3,241 88.28 % 3.67 % 3.24 %

Collection-prioritized 100,000 67,423 3,892 67.42 % 5.77 % 3.89 %

From the simulation results, the excitation efficiency 𝜂𝑒𝑥𝑐𝑖𝑡𝑎𝑡𝑖𝑜𝑛 of the excitation-

prioritized configuration is 88.28%, which is approximately 21% higher than the

Chapter 4

117

efficiency of the collection-prioritized configuration (67.42%). However, the

collection efficiency 𝜂𝑐𝑜𝑙𝑙𝑒𝑐𝑡𝑖𝑜𝑛 of the excitation-prioritized configuration is 3.67%,

which is around 2% lower than the efficiency of the collection-prioritized

configuration (5.77%). In addition, according to the simulation results, the overall

efficiency 𝜂𝑡𝑜𝑡𝑎𝑙 of the collection-prioritized configuration is 3.89% compared to 3.24%

of the overall efficiency of the excitation-prioritized configuration.

Based on these results, the collection-prioritized configuration is the most performant

and it is the optimum layout for the fiber optic-based Raman probe, as shown in Figure

4.14.

Figure 4.14 Optimal layout of the integrated Raman probe for the microfluidic lab-on-chip,

inside the dashed oval is the reflector based LoC.

Misalignment tolerances

The misalignment tolerances can be determined such that the overall efficiency varies

no more than 20% (-1dB loss) of the nominal value26, being 3.89 % as reported in

Table 4.2. We investigate the internal and external tolerances of the confocal Raman

probe by shifting the positions and angles of the components and detect the flux

change of the Raman signals in the non-sequential ray tracing simulations. The

internal tolerances refer to the misalignment tolerances between the optical

components inside the Raman probe, while the external tolerances refer to the

misalignments between the lab-on-chip and the Raman probe. The internal and


118

external misalignment tolerances of our confocal Raman probe are reported in Table

4.3 and Table 4.4.

Table 4.3 Misalignment tolerances between the optical components within the probe. (d: the

distance between the dichroic mirror DM and the mirror M; L: the length of the probe refers

to the vertical distance between the fiber tip and the dichroic mirror)

Internal tolerances d [mm] L [cm] DM tilt [°] M tilt [°]

Range allowed >4.2 9 – 20.9 ±0.25 ±0.25

Table 4.4 Misalignment tolerances between the chip and the Raman probe.

External tolerances X [µm] Y [µm] WD [mm] Tilt around X [°] Tilt around Y [°]

Range allowed -450 ~ 455 -460 ~ 475 <71 -1.65 ~ 1.53 -1.55 ~ 1.47

The tolerance analysis shown in Table 4.3 and Table 4.4 guarantees that the total

Raman efficiency for the optimum design of the Raman probe decreases less than -

1dB. The internal tolerances take into account the influences of the distance between

the two mirrors (d), their tilt with respect to the optical axis (DM tilt and M tilt) and

the length of the Raman probe. On the other hand, the external tolerances involve the

tolerance of the misalignment of the Raman probe with respect to the lab-on-chip in

the x- or y-axis, the tilts around the x- or y-axis, and the working distance (WD)

between the probe head and the lab-on-chip. As we can see from Table 4.3 and Table

4.4, the most crucial parameters that affect the Raman efficiencies are the tilts, which

is the same situation as for the alignment of the components within the probe.

Therefore, we have to pay more attention to the angular tolerances of the mirrors and

of the LoC during implementation. A proof-of-concept demonstration setup was built

in the lab in the frame of a Thorlabs cage system which is easy to align and maintain.

The technical parameters of the components in the Raman probe are the same as

described in Table 4.1. Figure 4.15 shows our implemented proof-of-concept

demonstration setup in combination with our lab-on-chip. A 785nm wavelength diode

laser (Sacher Lasertechnik TEC-500-0785-300) is used as excitation source, a

spectrometer (AvaSpec-HERO) is used to obtain the Raman spectra. The lateral

dimension of the proof-of-concept demonstration Raman probe is approximately

60mm, which is three times smaller than our previous confocal setup. The minimal

probe dimension of 13.3mm is possible according to our simulation results by

introducing a 3D printed mechanical holder for the optical components instead of the

cage system used.

Chapter 4

119

Figure 4.15 Different points of view of the “collection-prioritized configuration” of the

Raman probe realized, the proof-of-concept demonstration was built in a Thorlabs cage

system. Lab-on-chip support and syringe for filling fluids in the microchannel of the chip are

highlighted by the orange and green rectangles, respectively.

4.2.3 Experimental performance of the Raman probe combined with a LoC

Spectra of different fluidic samples (Figure 4.16) are preliminarily measured by our

fiber optic-based Raman probe with a 5 second integration time and 190mW laser

power on the chip. Ethanol has the highest Raman scattering at 879cm-1 (a) due to C-

C stretching, and methanol has a Raman peak at 1019cm-1 (b) corresponding to C-O

stretching. We can distinguish the peaks of ethanol and methanol from the spectrum

of the mixture of ¼ methanol and ½ ethanol in aqueous solution (c). In each spectrum,

including the spectrum of water (d), there is a small peak approximately at 920cm-1

which corresponds to the background noise from the polymer of the LoC, highly

suppressed because of the confocal approach.


120

Figure 4.16 Spectra of different fluidic samples. Ethanol and methanol can be distinguished

due to their different Raman peaks.

The miniaturized confocal Raman lab-on-chip setup, where the lateral scale of the

external optics was reduced from over 200mm to 13.3mm, replaces the bulky

laboratory equipment used before and may pave the way to a portable system for

medical applications. Hence, the idea of realizing a fully integrated optical system

where also the laser source and spectrometer are integrated in an innovative, low-cost,

easy to design and user-friendly system for analyses becomes more realistic.

4.2.4 Conclusion

Raman spectroscopy is a powerful tool for analytical measurements in many

applications. Traditional Raman spectroscopic analyses require bulky equipment,

considerable time of signal acquisition and manual sampling of substances under test.

In this work, we took a step from bulky and manual consuming laboratory testing

towards lab-on-chip analyses. We miniaturized the Raman spectroscopic system by

combining a freeform reflector-based polymer LoC with a customized Raman probe.

By using the confocal detection principle, we aim to enhance the detection of the

Raman signals from the substance of interest due to the suppression of the background

Raman signal from the polymer of the chip. Next to the LoC we miniaturized the

external optical components, surrounding the reflector embedded optofluidic chip,

and assembled these in a Raman probe. We evaluated the misalignment tolerance of

the internal optics (LoC) and external optics (Raman probe) by non-sequential ray

tracing which shows that off-axis misalignments are around ±400µm and the

maximum working distance of our Raman probe is 71mm. The non-sequential ray-

tracing simulations also show that the most critical misalignments come from the tilts

Chapter 4

121

of the mirror and dichroic mirror. Using this probe, the system could be implemented

as a portable reader unit containing external optics, in which a low-cost, robust and

mass manufacturable microfluidic LoC containing a freeform reflector is inserted, to

enable confocal Raman spectroscopy measurements.

4.3 Mass manufacturing of the LoC

4.3.1 Drawbacks of our previous Raman LoCs and solutions

Due to its confocality, the microfluidic chip we originally designed and produced can

greatly suppress Raman scattering from the material itself, while the freeform

reflector has a high numerical aperture, thus increasing the sensitivity of Raman

detection. But the shortcomings of this chip are also obvious. First of all, the chip uses

two polymer cubes to mount the in- and outlet, so the size of the chip, especially the

overall thickness in the direction orthogonal to the microfluidic channel and reflector

on chip is large. At the same time, this also adds additional production and packaging

processes, which increases the fabrication cost. Secondly, because the chip itself is

produced with ultra-high precision diamond tooling, the fabrication efficiency is

relatively low, and the cost of an individual chip is high. For the fabrication of each

chip, the diamond tooling machine needs to be first calibrated, the PMMA material

needs to be measured and pre-machined, followed by roughing and fine finishing of

the chip. The cost of operation and maintenance of diamond tooling machines is very

high in terms of manual and financial requirements. Thirdly, we found that the

performance of the PMMA-based chip deteriorates over time and the sensitivity of

the lab-on-chip decreases as a result. With a microscope, we observed that there are

some micro-cracks on the surface of the channel of the chip which has been used for

a period of time, as shown in Figure 4.17. These micro-cracks, initiated in PMMA due

to Environmental Stress Cracking (ESC)27,28, have reduced the optical transmittance

and mechanical strength of the chip, resulting in a decrease of the detection efficiency.

In addition, our chip uses fused silica tubing to transfer liquid. Although fused silica

is resistant to many chemicals, the tubing made from this material is hard and brittle.

It is inconvenient during a measurement as we must avoid bending the tubing.

Therefore, we need to find ways to solve these problems.


122

Figure 4.17 Micro-cracks on the surface of the polymer which was in contact with a fluidic

sample for a period of time.

To reduce the overall dimension of the chip, which is relatively large due to the

PMMA cubic connectors, we optimized the structural design of our chip. We maintain

the structure of the bottom layer with integrated freeform reflector but change the

structure of the channel and top layer, as shown in Figure 4.18. The channel layer

consists of two symmetrical polymer parts with a gap in between for inserting the

tubing. The top sealing layer is a complete polymer plate in our optimized design. In

this way, the total thickness of the chip equals to the sum of the thicknesses of the

three polymer layers, which is less than 3mm.

Figure 4.18 Scheme (left) and photograph (right) of our optimized PMMA chip.

For the other disadvantages, including the high cost, low resistance to environmental

stress cracking and the inflexibility of the fused silica tubing, we developed new

designs using TOPAS cyclic olefin copolymers (COC) polymers with better chemical

resistance and optical properties. We also optimized the entire fabrication process and

applied double-sided hot embossing for the mass manufacturing of the chip to reduce

the cost.

4.3.2 TOPAS COC polymers

Topas® is the trade name for Topas Advanced Polymers’ cyclic olefin copolymers

(COC) copolymerized from norbornene and ethylene using a metallocene catalyst29.

The molecular structure of COC is illustrated in Figure 4.19. It has been deployed in

a wide range of application domains from biomedical devices where biocompatibility

Chapter 4

123

and chemical inertness are rigorous, to packaging and electronics fields where there

are high demands on the mechanical and thermal properties.

Figure 4.19 Molecular structure of Topas COC.

The Topas COC family consists of different grades with distinguishing characteristics

for various application domains. The list of leading grades of Topas COC is shown in

Table 4.5.

Table 4.5 Leading grades of Topas COC polymers.30

Grade Property Characteristic

8007S-04 Standard For injection molding, high purity

5013L-10 High Flow General purpose for injection molding

6013M-07 Standard Injection molding extrusion molding common use, high purity,

better surface appearance, heat resistance

6017S-04 Standard Injection molding extrusion molding common use, high purity,

better surface appearance, highest heat resistance

8007F-600 Film Improved moldability

8007F-04 Film High purity

9506F-500 Film Improved Moldability, Low Tg

Each grade of Topas® refers to a COC polymer with different molecular weights and

norbornene/ethylene monomer ratios by which the range of physical properties is

achieved. For instance, as shown in Figure 4.20, the glass transition temperature (Tg)

of COC is correlated with the mole fraction of norbornene monomer as given by31

𝑇𝑔 (𝐶𝑒𝑙𝑐𝑖𝑢𝑠) ≅ 4 ∙ 𝑁𝑜𝑟𝑏𝑜𝑟𝑛𝑒𝑛𝑒 (𝑚𝑜𝑙 %) − 65 （4.4）


124

Figure 4.20 Copolymer composition of COC and the heat resistance displayed by its glass

transition temperature.29

As a matter of fact, according to the naming rules of Topas®, the first two digits of

the grade number represent the molecular weight of the COC polymer, while the last

two digits refer to its Tg. For example, Topas COC 6013 refers to a COC with an

average molecular weight around 60 000 and a Tg close to 130°C, and Topas COC

8007 implies that the COC polymer has an average molecular weight around 80 000

and a Tg close to 70°C. We performed Gel Permeation Chromatography (GPC)

measurements at the Department of Organic and Macromolecular Chemistry of

UGent (Prof. P. Dubruel) to determine the molecular weight32 of several types of

Topas COCs and PMMA. The GPC results are shown in Table 4.6. During the GPC

measurements, a refractive index (RI) detector was used as an indicator and

chloroform was used as a solvent with a flow of 1 mL/min.

Table 4.6 GPC results of different types of Topas COCs and PMMA.

Polymer

Grade

Norbornene

wt%

Tg

(°C)

HDT (°C)

@0.45Mpa

Mn Mw Mp Mz Mz PDI

COC 6017 82 178 170 40635 79651 72403 132800 191568 1.961

COC 6013 76 138 130 43312 79842 79675 120847 166918 1.844

COC 5013 75 134 127 48624 90450 85101 140065 196114 1.860

COC 8007 65 80 75 51501 96797 89250 151089 208993 1.880

PMMA XT 0 105 101 50887 94895 100820 139527 181346 1.865

(Mn: number average molecular weight; Mw: weight average molecular weight; Mp: peak

molecular weight; Mz: Z-average molecular weight; PDI: Polydispersity index = Mw/Mn)

The COC 6013 is a clear grade with a heat distortion temperature (HDT) of 138°C, a

value which cannot be attained by many other amorphous polymers.29 Its physical and

chemical properties such as high transparency, high purity, high chemical resistance

and low water absorption makes it appropriate for microfluidic devices. The

Chapter 4

125

transmission of Topas COC 6013 is shown in Figure 4.21. In addition, the better

finishing surface appearance of 6013 grade compared to other Topas grades makes it

an ideal material for mass manufacturing of optical components where the surface

roughness is crucial. In addition, Topas COC 6013 also delivers low fluorescence

which is extremely attractive when using for Raman measurements. Therefore, we

choose Topas COC 6013 as material for our microfluidic Raman LoCs.

Figure 4.21 Transmission of TOPAS 6013 and PMMA. (Measured with an AvaSpec-

UV/VIS/NIR spectrometer for plates with 2mm thickness)

4.3.3 Design of the freeform reflector-based LoC for mass manufacturing

Since COC polymer is used for the LoC and it has different optical properties

compared to PMMA, we re-designed the shape of the freeform reflector employing

the same numerical approach as mentioned in the previous section (4.1). The geometry

of the freeform reflector is determined by the thicknesses of the reflector and the

bottom layer, as well as by the focal height, as shown in Figure 4.22. We designed

two configurations of a COC-based freeform reflector and denoted them as G2 and

G3, while our previous PMMA-based freeform reflector is denoted as G1. The G2

COC based freeform reflector has similar geometrical and optical features as G1. The

main difference between the G2 and G3 reflectors are their focal height and bottom

thickness. As a result, they have a slightly different radius, NA and collecting

efficiency, as shown in Table 4.7.


126

Figure 4.22 Geometry of the freeform reflector.

Table 4.7 Specifications of the freeform reflectors.

G1:PMMA G2:COC G3:COC

Focal Height (µm) 50 50 100

Bottom Thickness (µm) 200 200 250

Reflector Thickness (µm) 300 300 300

NA 1.28 1.28 1.27

Radius (mm) 0.841 0.853 0.925

Efficiency 42% 42% 39%

The profiles of the PMMA-based freeform reflector and the two types of COC-based

freeform reflector are illustrated in Figure 4.23.

Figure 4.23 Profiles of the PMMA- and COC-based freeform reflectors.

The profile of the freeform reflector obtained by our numerical approach is a set of

data points, which needs to be fitted into a polynomial curve for the design and

fabrication of master molds, needed for double-sided hot embossing in a next step.

Therefore, we investigated the curve fitting of the surface profiles to find the proper

polynomial coefficients for the fabrication. The polynomial results of different

Chapter 4

127

degrees for both the PMMA- and COC-based freeform reflector are illustrated in

Figure 4.24, Figure 4.25 and Figure 4.26, and listed in Table 4.8.

Figure 4.24 Polyfitting results for the PMMA-based G1 chip.

Figure 4.25 Polyfitting results for the COC-based G2 chip.

Figure 4.26 Polyfitting results for the COC-based G3 chip.


128

Table 4.8 Polynomial fitting results of different freeform reflector designs.

(unit in mm) Poly2 Poly4 Poly6 Poly8 Poly10 Poly12 Poly14 Poly16

G1 Mean -1.76E-03 5.44E-04 -1.01E-04 -1.23E-05 1.59E-05 -3.53E-06 -1.50E-06 1.20E-06

STD 2.05E-03 5.58E-04 8.11E-05 2.94E-05 1.69E-05 3.11E-06 2.58E-06 1.16E-06

PV 5.93E-03 2.56E-03 2.27E-04 1.85E-04 5.35E-05 8.92E-06 1.42E-05 5.00E-06

G2 Mean -2.62E-03 7.47E-04 -6.74E-05 -5.34E-05 2.09E-05 4.12E-06 -4.85E-06 1.92E-07

STD 2.98E-03 7.01E-04 6.68E-05 6.57E-05 1.69E-05 8.91E-06 4.54E-06 1.41E-06

PV 7.89E-03 2.91E-03 3.09E-04 3.42E-04 4.89E-05 5.51E-05 1.27E-05 8.92E-06

G3 Mean -2.09E-03 4.24E-04 -3.84E-05 -1.30E-05 5.79E-06 -1.95E-07 -5.63E-07 1.63E-07

STD 2.20E-03 3.99E-04 2.77E-05 1.70E-05 5.19E-06 7.11E-07 6.37E-07 1.28E-07

PV 5.54E-03 1.79E-03 7.81E-05 1.01E-04 1.49E-05 4.80E-06 2.87E-06 3.73E-07

The accuracy of polynomial fitting of the reflector’s surface profile will greatly affect

the optical performance of the LoC system. The total integrated scatter (TIS) is

considered in the design to evaluate the influence of the surface quality of the reflector.

The TIS of an optical surface or thin film coating is given by33

𝑇𝐼𝑆(𝑅𝑞) = 𝑅0[1 − 𝑒−(4𝜋𝑅𝑞 cos 𝜃𝑖

𝜆)2

] （4.5）

Where, 𝑅𝑞 is the RMS roughness of the surface, 𝜃𝑖 is the angle of incidence, 𝑅0 is the

total reflectance, 𝜆 is the wavelength of the incident light. Figure 4.27 shows the

changes of scattering proportions of an optical surface with respect to the incident

angle and surface roughness for 785nm excitation light.

Figure 4.27 TIS of a reflective surface for 785nm wavelength light with different incident

angles.

Chapter 4

129

As stated by the TIS equation, incident light with a shorter wavelength gives rise to a

larger TIS for a surface with the same RMS roughness. In general, in order to

guarantee that the TIS is less than 20% for the 785nm excitation in our design, the

RMS roughness Rq should be less than 30nm. The requirement to the surface

roughness should be fulfilled during both the design and the fabrication phase.

According to the accuracy of the polynomial fitting listed in Table 4.8, a fitting degree

of 14 or higher is necessary.

Besides, we also determined the tolerance of the bottom thickness of the COC-based

chips via non-sequential ray tracing simulations in ASAP. The layout of our

simulation model is the same as the one used for the PMMA-based chip. We just

replace the PMMA-based chip with the COC-based 50µm/100µm focus LoC. The

normalized simulation results are shown in Figure 4.28. The maximum collection

efficiency of the G2 (50µm focal height) design is slightly higher than the one of the

G3 (100µm focal height) design because of its higher NA. However, we can observe

that the G3 design has a higher tolerance of the bottom thickness compared to the G2

design. This is mainly because the collection efficiency decreases sharply at a bottom

thickness of 50µm. This can be explained by the fact that most of the excitation rays

focus in the polymer bottom layer rather than in the microfluidic channel, resulting in

a higher Raman response of the COC polymer that obscures the Raman response of

the sample. For each design, in order to ensure that the normalized flux is more than

80%, the thickness error needs to be less than 20µm. This thickness tolerance is

particularly important in the follow-up fabrication process.

Figure 4.28 Thickness tolerance of the bottom thickness. The nominal bottom thickness of the

50µm focus design (G2) is 200µm, while the nominal bottom thickness of the 100µm focus

design (G3) is 250µm.


130

Base upon the simulations, we decided to investigate both designs and work in a first

step with the G2 chip.

4.3.4 Double-sided hot embossing for the 50µm focus G2 LoC

Design of the COC-based G2 chip

In a first step, we fabricated the COC-based 50µm focus G2 chip by double-sided hot

embossing replication. We have introduced the hot embossing fabrication process for

a microfluidic LoC in Chapter 1. Double-sided hot embossing is the replication

technique that stamps 3D structures on both sides of a polymer sheet by raising the

temperature of the polymer slightly above its glass transition temperature using two

master molds from both sides. Double-sided hot embossing is an ideal fabrication

technique for low-volume mass-manufacturing of optofluidic LoC devices where high

precision and high surface quality are essential34. Unlike the PMMA-based LoC for

which the reflector and the channel layer are fabricated out of different polymer plates,

the reflector, microfluidic channel and chamber can be stamped out of a single COC

plate. Next, we can bond the sealing layer on top of the hot-embossed replica by UV

curing adhesive. Thanks to the high overlay accuracy (2 µm) of the hot embossing

machine, we find that the alignment precision between the freeform reflector and

fluidic chamber of the COC-based G2 chip is higher than the precision of the PMMA-

based chip where the alignment had to be adjusted manually. This has also greatly

reduced the complexity and cost of the chip fabrication. Figure 4.29 and Figure 4.30

illustrate the structure and dimensions of our COC-based G2 chip. The diameter of

the channel/reflector layer is 25.4mm, the size of the sealing layer is 25.4 × 25.4 mm².

Figure 4.29 Drawing of the COC-based 50µm focus G2 chip with integrated freeform

reflector. The channel/reflector layer of the chip is fabricated via double-sided hot embossing.

Chapter 4

131

Figure 4.30 Dimensions of the COC-based 50µm G2 chip.

Fabrication of the brass-based master molds

Before we can replicate the COC-based G2 LoC by double-sided hot embossing, we

fabricated the master molds from brass material with ultra-precision diamond tooling.

Figure 4.31 shows the fabrication procedure of the reflector mold. A brass plate is first

prefabricated with a CNC milling machine and polished by diamond turning. Then

the mold structure for the reflector and rim of the chip are fabricated by ultra-precision

diamond tooling with the Nanotech 350FG CNC machine. The channel mold is

fabricated by micro-milling with the same Nanotech 350FG CNC, as shown in Figure

4.32.

Figure 4.31 Photographs of ultra-precision diamond tooling for the reflector mold fabrication.

(left: before tooling; middle: during tooling; right: after tooling)


132

Figure 4.32 Photograph of micro-milling for the channel mold fabrication.

The drawings and photos of the reflector and channel molds are illustrated in Figure

4.33. As the surface roughness of the replicas is highly dependent on the surface

quality of the molds, we characterized the surface of the reflector and channel molds

by using a non-contact optical profiler (Bruker Contour GT-I) based on a Mirau

interference microscope. We measured an RMS roughness of (20.5 ± 5.3) nm for the

reflector mold, and an RMS roughness of (6.4 ± 1.3) nm for the channel mold. Each

mold is evaluated within 10 evaluation areas of 64µm × 48µm.

Figure 4.33 Drawings and photographs of the reflector mold (left) and channel mold (right).

Chapter 4

133

Chip replication

To fabricate the COC-based 50µm focus G2 chips, the reflector and channel mold are

first fixed on the holders of the machine. Then, a Topas COC 6013 polymer sheet with

a dimension of 30mm × 30mm × 1mm is inserted into the chamber of the embossing

machine between the two molds. By increasing the temperature of the molds to 175°C

for 30 seconds and loading 12000N force under vacuum conditions, the micro-

structures of the fluidic channel as well as the freeform reflector are stamped on the

polymer sheet due to thermal plastic deformation. For demolding, the polymer is

cooled down to 115°C, while maintaining the applied force to avoid sink marks, voids

and shrinkage. These optimized parameters such as temperature, force and waiting

time guarantee that the bottom thickness of the replica is within the tolerance range,

which is crucial for our Raman signal acquisition. Finally, the stamped part is

demolded and removed from the machine. The cycle time for each replica takes

approximately 5 minutes. Figure 4.34 shows a schematic and photo of the double-

sided hot embossing process.

Figure 4.34 (a) Schematic and (b) photograph of double-sided hot embossing for the chip

replication.

Of course, other combinations of these factors are also able to control the bottom

thickness within the tolerance range. For instance, increasing the force while reducing

the waiting time and keeping the temperature constant, or keeping the force

unchanged, but lowering the temperature and extending the waiting time. Adjusting

all three factors is also possible to achieve the same objective. But when doing that,

we must take into account the corresponding possible consequences. For example,

increasing the force may not only deform the polymer sheet, but also the metal molds.


134

A higher temperature may also change the physical and chemical properties of the

polymer.

We noted that the initial thickness of the blank polymer sheet can largely influence

the final bottom thickness of the replica as defined in Figure 4.22. Figure 4.35 shows

the thickness of the replica with respect to the original polymer sheet. Basically, a

thicker polymer sheet can result in a thicker sample thickness. By measuring the

thickness of the polymer sheets and selecting the proper ones, we can fabricate the

LoCs with the correct bottom thicknesses. Figure 4.36 shows the raw COC polymer

sheet cut by milling machine and the fabricated reflector/channel layer using hot

embossing.

Figure 4.35 Influence of the sheet thickness. The raw COC plate is cut into 30mm by 30mm

pieces with a milling machine. The thickness of the COC plate is measured before hot

embossing, and the thickness of the chip is measured after hot embossing. In general, a

thicker COC plate leads to a thicker chip thickness.

Figure 4.36 COC polymer sheet and the hot embossed reflector/channel layer. The size of the

original sheet is 30mm × 30mm × 1mm. The diameter of the reflector/channel layer is

25.4mm.

Chapter 4

135

The surface roughness of the freeform reflector and the channel for the first set of

replicas (20 pieces) is (29 ± 4.9) nm and (24.8 ± 3.0) nm respectively, measured by a

non-contact optical profiler (Bruker Contour GT-I). We then applied a 200nm thick

reflective Au layer on the freeform reflector by sputter coating. In a last step, the

reflector/channel layer is bonded with a sealing layer by UV curing adhesive (Loctite

AA 3301 LC, Viscosity 160 mPa·s), as shown in Figure 4.37.

Figure 4.37 Photographs of the microfluidic lab-on-chip with integrated freeform reflector.

The hot embossed circular reflector/channel layer and the squared sealing layer are bonded by

UV curing adhesive.

Mold degradation and micro-cracks in the replicas

When repeating the hot embossing process to fabricate a second batch of COC-based

50µm focus G2 LoCs, we found that the surface quality of both molds and the replicas

degraded a lot. The RMS roughness of the reflector and the channel mold became

(168.1 ± 23.8) nm and (67.9 ± 16.2) nm respectively, as shown in Figure 4.38. The

RMS roughness of the corresponding replicas are (162.3 ± 21.9) nm and (69.2 ± 13.5)

nm respectively.

Figure 4.38 Surface roughness of the reflector and the channel mold before and after hot

embossing. The RMS roughness of the reflector mold increased from (20.5 ± 5.3) nm to

(168.1 ± 23.8) nm, and the RMS roughness of the channel mold increased from (6.4 ± 1.3) nm

to (67.9 ± 16.2) nm, respectively.


136

The deterioration of the mold’s surface roughness is mainly caused by two reasons.

Firstly, the high temperature during the hot embossing process can easily induce

oxidation and corrosion of the copper on the surface of the mold. Although the

replication is performed under vacuum conditions, the humidity, oxygen and other

gases adsorbed by the COC polymer can be released and adsorbed by the brass surface

during the hot embossing process. According to the datasheet of Topas, COC polymer

absorbs 100ppm humidity, 600ppm oxygen and 1200ppm nitrogen in general

conditions at 25°C. Because of the humidity, the copper surface easily adsorbs or

agglomerates H2O molecules resulting in a thin aqueous film. In this way, the

corrosion or electrochemical corrosion is more likely to occur. The released residuals

of the COC polymer are dissolved in the aqueous film and will greatly accelerate the

oxidation and corrosion of the copper at high temperatures. The higher the

temperature, the faster the reaction. The oxidation and corrosion can also be visually

observed through the discoloration of the brass surface. Secondly, the high pressure

applied on the microstructures during the hot embossing leads to a deformation of the

brass surface, which results in a degradation of the surface quality at macro scale. In

addition, heat can be hardly dissipated from the closed structure between two molds,

locally resulting in a higher temperature, which further aggravates the oxidation and

corrosion of the brass surface.

The degradation of the surface roughness of the molds naturally leads to a decreased

surface quality of the COC replicas. In addition, we even observed some micro-cracks

at the thinnest part (200µm bottom thickness) of the replicas around the reflector.

Some of the micro-cracks spread into the interior region of the freeform reflector,

which inevitably reduce the optical and mechanical properties of the LoCs, and

eventually affect the Raman detection efficiencies. The cracking defects are caused

by the shrinkage of the material during cooling35 because the mechanical strength of

the thinnest parts is much less than the surrounding regions. The cracking defects of

the COC replicas can be reduced by increasing the bottom thickness. Therefore, we

switched to the G3 LoC design by optimizing the structure of our LoC and the

fabrication process to solve problems such as the variation of the bottom thicknesses

and the degradation of the surface quality.

4.3.5 Mass fabrication of the COC-based 100µm focus G3 LoC

Design of the COC-based G3 LoC

The 100µm focus G3 LoC design has a bottom thickness of 250µm, which reinforces

the mechanical strength of the bottom layer and avoids cracking defects compared to

the G2 LoC design (200µm bottom thickness). We introduced fluorinated ethylene

propylene (FEP) tubing (IDEX Health & Science) instead of fused silica capillaries

Chapter 4

137

because of their higher flexibility and extreme resistance to a chemical attack36. A

general view of our COC-based G3 LoC is shown in Figure 4.39(a). Figure 4.39(b)

indicates the dimensions of our G3 LoC. The diameter of the freeform reflector is

1.851mm, the bottom thickness is 250µm. The integrated reflector/channel layer is

bonded with a circular sealing layer. The FEP tubing is inserted into the two holes of

the sealing layer that work as in- and outlet. The smallest volume we can measure

with the G3 LoC is 0.5µL which corresponds to the dimensions of the microfluidic

chamber.

Figure 4.39 (a) General view and (b) cross-section of the COC-based G3 LoC.

Fabrication process

The optimized fabrication process is shown in Figure 4.40. In a first step, we

fabricated two baseplates by CNC milling starting from brass material. A nickel

phosphorus (NiP) alloy layer was coated on top of each baseplate by electrochemical

plating. This NiP coating can improve the resistance of the brass mold to chemical

erosion and oxidation, especially when embossing at high temperatures.37 Next the

rough NiP surfaces on the brass molds were pre-milled before placing on the Nanotech

350FG CNC machine to obtain a good surface quality. In a next step, the reflector

mold was fabricated by ultra-precision diamond tooling and the channel mold by

micro-milling. Figure 4.41 shows the fabricated master molds out of NiP-coated brass.


138

Figure 4.40 Full fabrication flow for the COC-based 100µm focus G3 LoCs.

Figure 4.41 Photographs of the (a) reflector mold and (b) channel mold. The mold surface is

coated with NiP material.

The NiP-coated brass molds are then fixed on the hot embossing machine. Unlike the

fabrication of the COC-based G2 chip where the thickness of the replica was

controlled by the hot embossing settings and the dimensions of the polymer plate, we

designed and fabricated a physical stopper to control the bottom thickness of the

replicas, as shown in Figure 4.42. The physical stopper is equipped with 4 metal shims

and fixed around the reflector mold. When conducting the hot embossing process, we

put a 1mm-thick COC sheet in-between the two molds. By increasing the temperature

in the vacuum chamber and loading force on the molds, the polymer is deformed and

its thickness is decreased. The movement of the two molds will be stopped once both

molds hit the metal shim. In this way the 250µm thickness of the bottom layer can be

Chapter 4

139

precisely controlled, no matter how much the thickness of the raw polymer sheet

varies.

Figure 4.42 (a) General view and (b) dimensions of the physical stopper to control the bottom

thickness of the LoC during hot embossing. The thickness of each shim is 250µm.

In a next step, we applied a 200nm Au layer on the freeform reflector surface using a

sputter coater (JEOL® JFC-2300HR FINE COATER). The surfaces of the

reflector/channel layer as well as the sealing layer were treated with a Piezobrush®

PZ2 plasma pen to increase the surface energy and activate the polymers due to outer

chain scissions in just a few seconds.38–40 As a result, the adhesion of the polymer

surface can be improved during bonding using a thermal treatment or UV curing.

Figure 4.43 shows our final COC-based G3 chip.

Figure 4.43 Photographs of the fabricated COC-based 100µm focus G3 chip.

The RMS surface roughness of our optimized NiP-coated brass reflector and channel

mold are (12.4 ± 2.3) nm and (6.3 ± 0.4) nm respectively. The surface roughness of

the reflector and channel of the hot embossed G3 replica chip are (17.0 ± 3.9) nm and

(8.1 ± 0.4) nm respectively. After several runs, we did no longer observe a degradation

of the NiP-coated brass mold, and only regular cleaning of the molds’ surfaces with

isopropyl alcohol or acetone is needed to remove the residues, paving the way towards

the mass fabrication of our COC-based G3 chips.


140

4.3.6 Bonding and shear strength tests

We performed shear strength tests on the COC 6013 and PMMA samples bonded with

different approaches including UV curing adhesive, laser welding and thermal

bonding to better understand the improvement of plasma treatment on the adhesion of

polymers. Each sample under test consists of two COC or PMMA sheets with a

dimension of 25mm × 25mm × 1mm and an overlapping area of 10mm × 25mm, as

illustrated in Figure 4.44. The contact surface of each polymer sheet is exposed to a

plasma for 30 seconds with a Piezobrush® PZ2 plasma pen. For the UV curing

adhesive, a droplet of 50µL UV glue (Loctite AA 3301 LC) is placed on the

overlapping area and evenly distributed due to capillary forces. After illuminating the

glue with a UV lamp for 30 seconds, the two polymer sheets are bonded. The laser

welding is performed on a LPKF PrecisionWeld 3000 machine that is equipped with

a 1940nm-wavelength Thulium fiber laser with a spot size of 65µm. The power

density for the laser welding is 2.11kW/mm2. The laser spot on the interface of the

sheets melts the thermoplastic and generates physio-chemical bonds at the interface

by subsequent cooling41. The weld pattern is formed with 3 vertical lines (20mm long

each) and 6 horizontal lines (8mm long each), as shown in Figure 4.44(b). The total

weld seam area is 21.24mm2 approximately. The thermal bonding is performed using

the Jenoptik HEX04 hot embossing machine, which we also applied for the mass

fabrication of the chips. During the thermal bonding, we heated the sample up to

140°C and continuously applied a force of 1000N for 5 minutes. Here, a lower

temperature and force, compared to the replication of the chips, are applied to avoid

deformation of the polymer sheets.

Figure 4.44 (a) Scheme of UV/thermal bonding and (b) laser welding.

The bonded samples are placed into a Universal Testing Machine with both ends being

fixed in separate clamps. The shear stress 𝐹𝑝, coplanar with the overlapping plane, is

measured as shown in Figure 4.45. The samples without surface plasma treatment are

also tested as benchmark. The shear stress of the bonded samples with surface plasma

treatment is significantly improved except for thermal bonding. Because the plasma

treatment breaks some of the chemical bonds of the polymers on the surface, we

Chapter 4

141

understand it is easier to reconnect with the bonds of molecules from another polymer

sheet in the weld seam area or from the UV photoresist. While for the thermal bonding,

the heat and loaded force make the two polymer components to join permanently into

a complete assembly by cohesive bonds across the interface,41 the improvement in

adhesion by plasma treatment is not obvious.

Figure 4.45 Shear stresses of the samples bonded with different methods. (LW: laser welding;

UV: UV curing adhesive; TB: Thermal bonding; legend: with and without plasma treatment)

The shear strength of the bonding is given by42

𝜏 =𝐹𝑝

𝐴 （4.6）

Here A is bonding area, and 𝐹𝑝 is the shear stress measured. The calculated shear

strength of plasma treated, and non-plasma treated samples are listed in Table 4.9.

Table 4.9 Shear strength of different bonding approaches for different materials. (unit: MPa)

Method - Material Shear strength without PT Shear strength with PT

LW - COC 8.12 ± 4.71 32.02 ± 3.07

LW - PMMA 21.11 ± 2.21 26.47 ± 10.03

UV - COC 0.80 ± 0.27 1.08 ± 0.27

UV - PMMA 1.14 ± 0.30 3.55 ± 0.52

TB - COC 3.40 ± 0.89 3.45 ± 0.33

(PT: Plasma treatment. LW: laser welding; UV: UV curing adhesive; TB: Thermal bonding)

The laser welded COC sample with prior plasma treatment possesses the highest shear

strength 𝜏 compared to UV curing and thermal bonding. However, the weld seams

may deform the microfluidic channel and could lead to leakage. UV curing adhesive,

on the other hand, is a fast and simple bonding method, but only suitable for low-

speed and low-volume prototyping of LoC devices. Thermal bonding has the highest


142

resistance to the leakage for the COC material which is important for high-pressure

microfluidic detection.43

4.4 Mass manufactured LoCs in combination with the developed

Raman probe

4.4.1 Implementation of the proof-of-concept demonstration for confocal

measurements

We used the proof-of-concept setup of the Raman probe shown in Figure 4.15 in

combination with our hot embossed LoCs for the confocal microfluidic Raman

measurements, as shown in Figure 4.46.

Figure 4.46 Proof-of-concept Raman setup by combining the Raman probe with the mass

fabricated LoC. (BPF: Band-pass filter, M: Mirror, DM: Dichroic mirror, LPF: Long-pass

filter, NF: Notch filter, CL: Collimating lens, MMF: Multi-mode fiber.)

4.4.2 System calibration

In a first step, we perform Raman measurements with our confocal Raman setup to

calibrate the system. The Raman setup is aligned by adjusting the positions of the

optical components and the LoC to the optimal conditions. We compare the

performance of the hot embossed G2 and G3 chips by measuring the spectra of ethanol,

Chapter 4

143

as shown in Figure 4.47. The power of the laser excitation is 190mW, and the

integration time of the spectrometer is 15 seconds for each spectrum. In general, the

average Raman signal obtained by our 100µm focus G3 chip is 3 times larger as the

one obtained by the G2 chip. In addition, the repeatability of the spectra by the G3

chip is also much better than for the G2 chip. This can be explained by the lower

energy loss according to the smaller surface roughness and higher precision of the

bottom thickness of the G3 chip compared to the G2 chip.

Figure 4.47 Raman spectra of ethanol obtained by the mass fabricated G2 and G3 chips. The

dashed lines refer to the upper and lower range of the spectra. Figures are on the same scale.

Since the G3 chip performs twice better than the G2 chip, we continue with our G3

chip and calibrate our Raman system by measuring the Raman spectra of urea

solutions with different concentrations ranging from 2.5mM to 500mM, as shown in

Figure 4.48. The Raman peak at 930cm-1 comes from the background of COC polymer.

An internal standard with a constant concentration of 100mM KNO3 is added to each

urea solution for each measurement to eliminate uncontrollable fluctuations induced

by the setup.


144

Figure 4.48 Raman spectra of urea solutions with different concentrations. An internal

standard of 100mM KNO3 is added to the solution to normalize the spectra.

After subtracting the background of the water spectra, we normalize the Raman peak

of urea at 997cm-1 to the characteristic peak of KNO3 at 1040cm-1 for each urea

solution. Figure 4.49 shows the calibration curve with the changes of the normalized

Raman intensities of urea for different concentrations. The limit of detection (LOD),

or the detection limit, of the system can be obtained according to the calibration

curve:44

𝐿𝑂𝐷 = 3𝑆𝑎/𝑏 （0.32）

Where 𝑆𝑎 is the standard deviation of the response, and b is the slope of the linear

regression fitted calibration curve.

The LOD of our Raman probe in combination with the freeform reflector embedded

LoC is 10.2mM (60 ppm) for the detection of urea according to the linear regression

approach. The noise-equivalent-concentration (NEC) of our setup is 3.4mM for urea

detection, which is almost 6 times better than the NEC obtained with the LoC system

available at the start of this PhD. The LOD of our polymer-based Raman LoC is

comparable with the one obtained by the quartz glass-based Raman-on-chip (7.5mM)

with a much higher excitation power (700mW) 5,6.

Chapter 4

145

Figure 4.49 Calibration curve of our proof-of-concept Raman setup using our mass-fabricated

G3 chips in combination with our Raman probe.

4.4.3 Proof-of-concept for mycotoxin detection

As we mentioned in the previous chapter, mycotoxins including fumonisin45–47 and

deoxynivalenol48,49 are commonly existing in a variety of different crops. Their

detection is of great importance in the field of food safety. Therefore, we prepared

fumonisin and deoxynivalenol solutions with different concentrations and performed

Raman measurements using our Raman setup. The fumonisin B1 is dissolved in a 1:1

water-acetonitrile solution and the concentration of fumonisin B1 is prepared in three

levels: 2 ppb, 1250 ppb and 1500ppb, which are marked with the initial letter of the

toxin name, namely F followed by the numbers 1, 2 and 3, respectively. The

deoxynivalenol is dissolved with methanol and various concentrations (5 ppb, 1000

ppb and 1250 ppb) are prepared, marked with the initial letter of the toxin name,

namely D followed by numbers 1, 2 and 3, respectively. Each mycotoxin solution is

injected into the microfluidic channel of our COC-based G3 chip through the inlet,

and 10 sets of Raman spectra with 15 seconds integration time are obtained under the

785nm wavelength laser with a power of 190mW. The Raman spectra of the solvents

are also measured under the same conditions as reference.

All spectra are imported in MATLAB and smoothed with a 2nd order Savitsky-Golay

filter. We corrected the offset of the spectra by taking into account the mean level of

the spectra within the smooth bands between 1670cm-1 and 2360cm-1. In a next step,

we performed Principal Component Analysis (PCA) to the processed Raman spectra

of fumonisin B1 and deoxynivalenol.


146

Figure 4.50 shows the PCA results of the fumonisin Raman spectra. The spectra form

well separated groups in the PC1-PC2 plane, which accounts for 96% of total variance.

According to Figure 4.50(c), the solvent (MWA) and F1 data are close to each other.

Although the samples are classified, they are not ordered along the PC1 direction

according to their contamination level. PC2 is equal for all groups except for F3,

therefore this difference could be related to the toxin itself.

Figure 4.50 PCA results of the fumonisin B1 Raman spectra. (a) Average spectra of solvent

and fumonisin b1 with different concentration levels. (b) loadings of different principal

components. (c) Scores of PC1 and PC2.

Figure 4.51 shows the PCA results of the deoxynivalenol Raman spectra. The spectra

form well separated groups in the PC1-PC2 plane, which account for 96% of total

variance, as shown in Figure 4.51(c). The group of pure methanol is not close to D1

as expected. The groups are also not ordered according to their contamination level

along either PC1 or PC2. Moreover, two methanol spectra mix up with the D3 group.

Chapter 4

147

Figure 4.51 PCA results of the deoxynivalenol Raman spectra. (a) Average spectra of solvent

and deoxynivalenol with different concentration levels. (b) loadings of different principal

components. (c) Scores of PC1 and PC2.

In conclusion, Raman spectra of mycotoxins obtained by our Raman setup show a

very good repeatability (tight group of replies), which allow to detect sensible

between-sample differences. However, these differences look more related to random

factors than to their contamination. Only in case of Fumonisin B1, a detection of

contamination seems possible (and only along the 2nd PC), but it should be confirmed

by analyzing some more heavily contaminated samples (above 100ppm). When low

contamination levels are analyzed with traditional Raman spectroscopy, even a very

small difference in sample preparation could have an effect comparable to or even

stronger than the toxin itself. Therefore, the combination of SERS with microfluidic

LoC is necessary to increase the sensitivity of our Raman spectroscopy measurements

for mycotoxin detection.

4.5 Conclusion

In the first part of this chapter, we introduced a confocal Raman spectroscopy system

designed and implemented by our research group, prior to the start of this PhD. The

Raman spectroscopy system consists out of a freeform reflector, an integrated lab-on-


148

chip and external optics, which can suppress the background signal from the out-of-

focus regions significantly. We achieved a noise-equivalent concentration of 20mM

for urea detection. The PMMA-based lab-on-chip with embedded freeform reflector,

manufactured directly by ultra-precision diamond tooling, is robust and performant

from a mechanical and optical point of view in terms of misalignment tolerance and

high NA respectively. However, the cost for the direct fabrication of the freeform

reflector is quite high as it is not compatible with mass manufacturing techniques. The

chemical resistance of PMMA material is also relatively low for various types of

analytes and solvents. In addition, the dimensions of the external optics of the Raman

system is still large such that we cannot use it for in-situ analyses.

Therefore, in a first step of this PhD, we integrated the components of the external

optics into a miniaturized Raman probe. We accessed the misalignment tolerances by

means of non-sequential ray tracing simulations as well as experiments. In a next step,

we re-designed the freeform reflector and the lab-on-chip from many aspects for the

purpose of mass fabrication. The new design of the LoC is based on Topas COC

material, which shows a higher chemical resistance and better optical properties

compared to PMMA. The shape of the freeform reflector is calculated via a numerical

approach according to Fermat’s principle. We fabricated NiP-coated brass master

molds, which are resistant to chemical erosion and oxidation during double-sided hot

embossing. The NiP coating on top of the brass structure also improves the strength

of the molds and results in a higher surface quality. We replicated COC-based LoCs,

bonded them with a sealing layer and inserted FEP tubing as in- and outlet.

We integrated the mass manufactured LoC with the miniaturized Raman probe and

implemented a proof-of-concept demonstration setup. We conducted various types of

experiments to get insight into the performance of our Raman setup. We calibrated

our Raman system by measuring the Raman spectra of urea solutions with different

concentrations. A noise equivalent concentration of 3.4mM and a limit of detection of

10.2mM for urea solution is achieved by our LoC system. Finally, we employed our

Raman system for the detection of two types of mycotoxin, fumonisin B1 and

deoxynivalenol, and performed Principle Component Analysis to the Raman spectra

obtained. The PCA results show that the Raman spectra of mycotoxins obtained by

our setup are clustered and allow to detect differences between various samples.

However, the sensitivity of our LoC system using traditional Raman measurements is

too low. Therefore, the combination of microfluidic-lab-on-chips with SERS is

necessary to extend the applicability of Raman spectroscopy.

Chapter 4

149

References

1. Watson, D. A. et al. A Flow Cytometer for the Measurement of Raman

Spectra. Cytom. Part A 73, 119–128 (2008).

2. Ngernsutivorakul, T. et al. Design and Microfabrication of a miniature fiber

optic probe with integrated lenses and mirrors for Raman and fluorescence

measurements. Anal Bioanal Chem 409, 275–285 (2017).

3. Bates, K. E. & Lu, H. Biophysical Perspective Optics-Integrated Microfluidic

Platforms for Biomolecular Analyses. Biophys. J. 110, 1684–1697 (2016).



(2015).



80870J (2011).

6. Dochow, S. et al. Raman-on-chip device and detection fibres with fibre Bragg

grating for analysis of solutions and particles. Lab Chip 13, 1109 (2013).

7. Yamamoto, H. et al. Diagnosis and Treatment of Small-Intestinal Diseases.

Clin. Gastroenterol. Hepatol. 3565, 1010–1016 (2004).

8. Goetz, M. & Wang, T. D. Imaging and Advanced Technology Molecular

Imaging in Gastrointestinal Endoscopy. Imaging Adv. Technol. 138, 828–833

(2010).

9. Kudo, S., Tamura, S., Nakajima, T. & Yamano, H. Diagnosis of colorectal

tumorous lesions by magnifying endoscopy. Gastrointest. Endosc. 44, 0–6

(1996).

10. COONEY, T. F., SKINNER, H. T. & S. M. ANGEL. Comparative Study of

Some Fiber-Optic Remote Raman Probe Designs . Part II : Tests of Single-

Fiber , Lensed , and Flat- and Bevel-Tip Multi-Fiber Probes. Appl. Spectrosc.

50, 849–860 (1996).

11. F.Cooney, T., SKINNER, H. T. & S.M.Angel. Comparative Study of Some

Fiber-Optic Remote Raman Probe Designs. Part I: Model for Liquids and

Transparent Solids. Appl. Spectrosc. 50, 836–848 (1996).


150

12. Shim, M. G., Song, L. W. K., Marcon, N. E. & Wilson, B. C. Rapid

Communication In vivo Near-infrared Raman Spectroscopy : Demonstration

of Feasibility During Clinical Gastrointestinal Endoscopy. Photochem.

Photobiol. 72, 146–150 (2000).

13. Saar, B. G., Johnston, R. S., Freudiger, C. W., Xie, X. S. & Seibel, E. J.

Coherent Raman scanning fiber endoscopy. Opt Lett. 36, 2396–2398 (2012).

14. M.S.Bergholt et al. In Vivo Diagnosis of Esophageal Cancer Using Image-

Guided Raman Endoscopy and Biomolecular Modeling. Technol. Cancer Res.

Treat. 10, 103–112 (2011).

15. Komachi, Y., Katagiri, T., Sato, H. & Tashiro, H. Improvement and analysis

of a micro Raman probe. Appl. Opt. 48, 6–13 (2009).

16. Motz, J. T. et al. Optical fiber probe for biomedical Raman spectroscopy. Appl.

Opt. 43, 542–554 (2004).

17. Short, M. A. et al. Development and preliminary results of an endoscopic

Raman probe for potential in vivo diagnosis of lung cancers. Opt. Lett. 33,

711–713 (2008).

18. Jermyn, M. & Desroches, J. A miniature confocal Raman probe for

endoscopic use. Phys. Med. Biol. 54, 7077–7087 (2009).

19. Mo, J., Zheng, W. & Huang, Z. Fiber-optic Raman probe couples ball lens for

depth-selected Raman measurements of epithelial tissue. Biomed. Opt.

Express 1, 4712–4721 (2010).

20. Prasad, P. N. Nanophotonics. (John Wiley & Sons, Inc., 2004).

21. Cordero, E., Latka, I., Matthäus, C., Schie, I. W. & Popp, J. In-vivo Raman

spectroscopy : from basics to applications. J. Biomed. Opt. 23, 071210 (2018).

22. Sarychev, A. K., Ivanov, A., Lagarkov, A. & Barbillon, G. Light

Concentration by Metal-Dielectric Micro-Resonators for SERS Sensing.

Materials (Basel). 12, 1–39 (2019).

23. M.L.Myrick & S.M.Angel. Elimination of Background in Fiber-Optic Raman

Measurements. Appl. Spectrosc. 44, 565–570 (1990).

24. Ma, J. & Li, Y.-S. Optical-Fiber Raman Probe with Low Background

Interference by Spatial Optimization. Appl. Spectrosc. 48, 1529–1531 (1994).

Chapter 4

151

25. Archibald, D. D., Lin, L. T. & Honigs, D. E. Raman Spectroscopy over

Optical Fibers with the Use of a Near-IR FT Spectrometer. Appl. Spectrosc.

42, 1558–1563 (1988).

26. Yamada, H., Nozawa, M., Kinoshita, M. & Ohashi, K. Vertical-coupling

optical interface for on-chip optical interconnection. Opt. Express 19, 698

(2011).

27. Materials, O. F. Fatigue crack propagation of PMMA in organic sovents. 9,

1896–1898 (1974).

28. Arnold, J. C. The effects of diffusion on environmental stress crack initiation

in PMMA. J. Mater. Sci. 33, 5193–5204 (1998).

29. Topas Advanced Polymers. TOPAS - Cyclic Olefin Copolymers. 4 (2015).

Available at:

http://www.topas.com/sites/default/files/files/TOPAS_Brochure_E_2014_06

(1).pdf.

30. TOPAS® COC Products. Polyplastics Global Website (2020). Available at:

https://www.polyplastics-global.com/en/product/topas.html.

31. O’Neil, C. E., Taylor, S., Ratnayake, K., Pullagurla, S. & Soper, S. A.

Chracterization of Activated Cyclic Olefin Copolymer: Effects of Ethylene /

Norbornene Content on the Physiochemical Properties. Analyst 141, 6521–

6532 (2017).

32. Wu, J. et al. Cereus sinensis Polysaccharide and Its Immunomodulatory

Properties in Human. Mar. Drugs 15, 140 (2017).

33. Schröder, S., Herffurth, T., Duparré, A. & Harvey, J. E. Impact of surface

roughness on scatter losses and the scattering distribution of surfaces and thin

film coatings. in SPIE - The International Society for Optical Engineering

8169 1–8 (2011). doi:10.1117/12.896989

34. Leech, P. Hot Embossing Of Microchannels in Cyclic Olefin Copolymers. in

Mater. Res. Soc. Symp. Proc. OO03-01 (2009). doi:10.1557/PROC-1191-

OO03-01

35. Shin, H. & Park, E. Analysis of Crack Phenomenon for Injection-Molded

Screw Using Moldflow Simulation. J. Appl. Polym. Sci. 113, 2702–2708

(2009).


152

36. Rath, S. K., Palai, A., Rao, S., Chandrasekhar, L. & Patri, M. Effect of

Solvents in Radiation-Induced Grafting of 4-Vinyl Pyridine onto Fluorinated

Ethylene Propylene Copolymer. J. Appl. Polym. Sci. 108, 4065–4071 (2008).

37. Barycka, I., Hołodnik, B. & Misiuk, A. NiP as a New Material for Thick Film

Technology. Electrocompon. Sci. Technol. 7, 221–226 (1981).

38. Kaplan, S. L. & Rose, P. W. Plasma surface treatment of plastics to enhance

adhesion. Int. J. Adhes. Adhes. 11, 109–113 (1991).

39. Hegemann, D., Brunner, H. & Oehr, C. Plasma treatment of polymers for

surface and adhesion improvement. Nucl. Instruments Methods Phys. Res.

Sect. B Beam Interact. with Mater. Atoms 208, 281–286 (2003).

40. Noeske, M., Degenhardt, J., Strudthoff, S. & Lommatzsch, U. Plasma jet

treatment of five polymers at atmospheric pressure: Surface modifications and

the relevance for adhesion. Int. J. Adhes. Adhes. 24, 171–177 (2004).

41. Roy, P. K., Malik, T. & Sinha, T. K. Thermal bonded Nonwoven - An

Overview. TechnicalTextile.Net 1–15 (2011). Available at:

https://www.technicaltextile.net/articles/thermal-bonded-nonwoven-an-

overview-5402.

42. Pelsmaeker, J. De et al. Clear to clear laser welding for joining thermoplastic

polymers: a comparative study based on physicochemical characterization. J.

Mater. Process. Technol. 255, 808–815 (2017).

43. Rosenbergerl, F., Jones, E., Lee, C. & Devoe, D. HIGH PRESSURE

THERMAL BONDING FOR SEALING. Micro Total Anal. Syst. 1, 404–405

(2002).

44. Shrivastava, A. & Gupta, V. B. Methods for the determination of limit of

detection and limit of quantitation of the analytical methods. Chronicles

young Sci. 2, 21–25 (2011).

45. Marasas, W. F. O. Discovery and Occurrence of the Fumonisins : A Historical

Perspective. Environ. Health Perspect. 109, 239–243 (2001).





Chapter 4

153





49. Pestka, J. J. & Smolinski, A. T. Deoxynivalenol : Toxicology and Potential

Effects on Humans. J. Toxicol. Environ. Health 8, 39–69 (2015).

Chapter 5

155

Chapter 5

5 A Tunable Freeform Segmented Reflector in a

Microfluidic System for Conventional Raman

and SERS Spectroscopy

As we described in Chapter 2, Raman spectroscopy is a powerful optical detection

technique and nowadays it is widely used in biological research, pharmaceutics,

chemical sciences and many other fields1–6. When a bundle of photons interacts with

a molecule, the majority of the photons will be elastically scattered by the molecule

as Rayleigh scattering, while a small portion (typically millionths) of the photons will

be absorbed and re-emitted almost simultaneously by the molecule with a frequency

shift due to the molecular vibrational modes. The inelastic scattering of the photons

with energy shift is referred as Raman scattering, which can be captured and detected

by a spectrometer as a Raman spectrum. The Raman spectra contain fingerprint

information of the molecules under test that can be used for material characterization.

Raman spectroscopy is suitable for label-free analysis of solids, liquids and gases

qualitatively or quantitatively. Especially in recent years, the combination of a

microfluidic chip with Raman spectroscopy has largely reduced the size and cost

compared to traditional bulky Raman equipment7–9. One way to miniaturize the

Raman system is to remove the lenses for collimation and focusing but introduce

optical fibers directly into the microfluidic chip for excitation and collection

components10–13. The optical fiber-based Raman-on-chip systems are very compact.

In such systems, the distance between the fiber facets and the fluid should be as small


Liu, Q.; Stenbæk Schmidt, M.; Thienpont, H.; Ottevaere, H. A Tunable Freeform-Segmented

Reflector in a Microfluidic System for Conventional and Surface-Enhanced Raman

Spectroscopy. Sensors 2020, 20, 1250.

Chapter 5 A Tunable Freeform Segmented Reflector in a Microfluidic System for Conventional Raman and SERS

156

as possible, otherwise the excitation and collection efficiencies of the microfluidic

chip will decrease significantly. One way to satisfy this requirement is to keep the

fiber ends in the fluidic channel and be in contact with the sample directly. However,

the analytes can contaminate the fiber facets easily during a measurement, therefore

decreasing the reliability and stability of the Raman analysis. To prevent this, a

compromise is to miniaturize some of the optical components into micro-optics and

to integrate them with the microfluidic chip14,15. These optofluidic chips typically have

higher sensitivity in comparison with the fiber-based microfluidic chips due to the

higher excitation and collection efficiencies. We have introduced freeform reflector-

based PMMA and COC lab-on-chips for Raman spectroscopy with a noise-

equivalent-concentration (NEC) for urea solution of 20mM and 3.4 mM respectively

in Chapter 4. The lab-on-chip has suppressed the background of polymer - out of

which the chip is fabricated - by a factor of 7. Nevertheless, the cost of an individual

chip is very expensive since the freeform reflector embedded on the chip is fabricated

via ultra-precision diamond tooling with considerably high manual and equipment

requirements. In addition, the detection limit of our previous lab-on-chip with

conventional Raman spectroscopy is not low enough to satisfy the requirements for

biological and toxicological applications where the concentrations of the analytes are

typically at ppm or even ppb level. Spectroscopy based on Surface-Enhanced Raman

Spectroscopy (SERS) has been applied increasingly to reach a higher detection

sensitivity by integrating nanoparticles, nanodimers or specific SERS-biotags in the

microfluidic systems for drug and cell detection16–19. With SERS, the Raman

scattering of adsorbates can be enhanced by localized surface plasmon resonances

(LSPR) on rough metal surfaces or nanostructures with noble metallic coating such as

Au or Ag. The SERS enhancement factors are typically 104–106 and can be as high as

1012 for single molecule detection20–22. In this chapter we investigate a freeform

reflector-based tunable Raman spectroscopy setup for microfluidic lab-on-chip

analysis. Although COC is the material needed to achieve a better chemical resistance,

however in a first step to demonstrate the proof-of-concept, we use PMMA to

fabricate our lab-on-chip. The versatile Raman setup we built for the experimental

proof-of-concept demonstration is compatible with our own PMMA-based lab-on-

chip for conventional Raman and surface-enhanced Raman detection, as well as lab-

on-chips that are commercially available.

Chapter 5

157

5.1 Design of the freeform segmented reflector

5.1.1 An overview of the segmented reflector

The lab-on-chip, reported in Chapter 4, consists of three layers, as shown in Figure

5.1(a). The 200µm thick bottom layer is embedded with a freeform reflector with

200nm gold coating to focus the excitation beam and collect the scattered Raman

signal. The middle layer is a 500µm height fluidic layer for the sample flow. The top

PMMA layer is the sealing layer for the fluidic channel including in- and outlets. The

focal point of the freeform reflector is located in the fluidic layer 50µm above the top

of the reflector layer. The fixed freeform reflector on chip enables robust confocal

Raman detection but has no flexibility to introduce SERS, since the SERS affective

area is typically close to the surface of the metallic surface in a range smaller than

100nm. Moreover, the SERS substrates are covered with a metallic coating and

therefore opaque. The incident beam and the Raman scattered light will be blocked in

the reflector-based system when introducing the SERS substrate without any

modification.

Figure 5.1 Illustration of (a) our freeform reflector embedded lab-on-chip (NA=1.28)

discussed in Chapter 4 and (b) the segmented reflector-based Raman system with SERS

microfluidic chip (NA=1.15). The red and blue rays refer to the incident light that interact

with different segments of the reflector. (S1: center segment; S2: middle concave segment;

S3: marginal segment.)

To this end, we bring forward a segmented mirror design that works together with a

microfluidic chip, by which the majority of the excitation light can be utilized and all

the scattered light within a certain field can bypass the SERS substrate and can be

collected. The microfluidic chip consists of three layers, including the polymer-based

bottom and top sealing layers and a channel layer for the fluidic samples, as shown in

Figure 5.1(b). The reflector is composed of three segments, the center segment, the

middle concave segment and the marginal segment at the outer rim of the reflector. In


158

the excitation path, the majority of excitation light will be directly reflected to the

microfluidic channel by the middle segment. The outer part of the beam will firstly be

reflected by the marginal segment and then focused by the center segment to the same

focal point as that of the concave segment. Only a small amount of the excitation

energy is blocked by the SERS substrate. And vice versa, in the collection path, the

center segment and the marginal segment provide the Raman signal collection with

two reflective surfaces and can work together to collimate the Raman scattering close

to the optical axis. At the same time the middle segment can collimate the rest of the

Raman scattered light towards the same direction.

5.1.2 The middle concave segment design

There are two methods to design the middle segment, either by conventional

sequential ray tracing to reduce the various aberrations and obtain an optimal surface,

or by a numerical approach to calculate the surface profile directly. We will not

discuss the ray-tracing-based optimization method in this PhD but focus on the

numerical approach.

Figure 5.2 Surface profile of the concave segment, the parallel rays are reflected through the

bottom layer of the microfluidic chip into the channel layer.

As shown in Figure 5.2, the refractive index of the sample under test in the channel

and the bottom polymer layer are 𝑛1 and 𝑛2, respectively. Additionally, the medium

between the bottom polymer layer and the reflector has a refractive index of 𝑛3.

The excitation ray AB that is perpendicular to the microfluidic chip is reflected by the

concave surface at point 𝐶, enters the bottom layer of the microfluidic chip at point D

and is refracted to the upper surface of the polymer layer at point E. The angle of

incidence, angle of refraction and angle of emergence at the bottom polymer layer are

Chapter 5

159

𝜃1, 𝜃2, 𝜃3 respectively. 𝐹 is the focal point for all the excitation rays. Points 𝑂 and 𝐺

are the intersections of the central ray with the bottom and upper surfaces of the

bottom polymer layer, respectively. 𝑀 is the lowest point of the reflector’s surface.

Suppose the focal height, or the distance between the upper surface of the polymer

layer and the focal point, is ℎ𝑓. The thickness of polymer layer is ℎ𝑝, the distance

between the lower surface of the polymer layer and the reflector is ℎ𝑚.

According to Snell's law

𝑛1 sin 𝜃1 = 𝑛2 sin 𝜃2 = 𝑛3 sin 𝜃3 （5.1）

As stated by Fermat’s principle23, the optical path length of all rays leaving the

polymer layer of plane 𝑥 and focusing to point 𝐹 is a constant, namely

𝑛3 ∙ (𝐵𝐶 + 𝐶𝐷) + 𝑛2 ∙ 𝐷𝐸 + 𝑛1 ∙ 𝐸𝐹

= 2𝑛3 ∙ ℎ𝑚 + 𝑛2 ∙ ℎ𝑝 + 𝑛1 ∙ ℎ𝑓 （5.2）

Let’s construct a Cartesian coordinates system with point 𝑂 as the origin and the

central ray 𝑂𝐹 as the z-axis. Then the point 𝐶 on the surface of the reflector can be

represented by the coordinates (𝑥, 𝑧) . We can also get the following geometric

relationships

−𝑥 = 𝐵𝐶 ∙ tan 𝜃3 + ℎ𝑝 ∙ tan 𝜃2 + ℎ𝑓 ∙ tan 𝜃1 （5.3）

−𝑧 = 𝐵𝐶 （5.4）

𝐶𝐷 = −𝑧/ cos 𝜃3 （5.5）

𝐷𝐸 = ℎ𝑝/ cos 𝜃2 （5.6）

𝐸𝐹 = ℎ𝑓/ cos 𝜃1 （5.7）

When inserting the equations (5.3) - (5.7) into the equation (5.2) we get

𝑧 = [𝑛2ℎ𝑝(1

𝑐𝑜𝑠 𝜃2

− 1) + 𝑛1ℎ𝑓(1

𝑐𝑜𝑠 𝜃1

− 1) − 2𝑛3 ∙ ℎ𝑚]

/[𝑛3 ∙ (1 +1

𝑐𝑜𝑠 𝜃3

)]

（5.8）


160

Also, the abscissa of the surface 𝑥 is given by

𝑥 = 𝑧 ∙ tan 𝜃3 − ℎ𝑝 ∙ tan 𝜃2 − ℎ𝑓 ∙ tan 𝜃1 （5.9）

Because 𝜃2 and 𝜃3 can be represented by 𝜃1 according to equation (5.1), and as

demonstrated by the equation (5.8) and (5.9), we notice that once the ℎ𝑓, ℎ𝑝 and ℎ𝑚

are confirmed, the surface profile of the concave segment is a function of the angle

𝜃1. If we fill the same medium as 𝑛1 in the space between the polymer layer and the

reflector (𝑛1 = 𝑛3), 𝜃1 = 𝜃3. The focal height ℎ𝑓 will change proportionally with

respect to the height of reflector ℎ𝑚 , making the system tunable in the vertical

direction with a stationary spot size. This is crucial as we can place the focal point in

the very middle of the channel to minimize the Raman background from the polymer

or keep it exactly on the nanostructures of a SERS substrate maximizing the SERS

enhancement. Since most of the sample solutions for biological and chemical research

are aqueous24–27, and many other organic solvents such as ethanol or acetone have a

similar refractive index as water, we use water as the medium between the reflector

and polymer layer (𝑛1 = 𝑛3 = 𝑛𝑤𝑎𝑡𝑒𝑟).

5.1.3 The center and marginal segment design

The center and marginal segments are also directly designed by a numerical approach.

Figure 5.3 illustrates the surface profile of all three segments. In this step we use the

same Cartesian coordinates system as mentioned before.

Figure 5.3 Design schematic of all segments: center, concave and marginal: (a) general view,

(b) zoom-in view of the rays in the microfluidic channel and (c) zoom-in view of the rays

interacting with the center segment.

Chapter 5

161

Assume 𝑆𝑖(𝑥𝑠𝑖 , 𝑧𝑠𝑖) is the virtual focal point of a reflected ray by the concave surface,

which refers to the intersection of the extension of a reflected ray and the optical axis.

Note that 𝑆𝑖 is different from the actual focal point 𝐹, which can be obtained by ray

tracing of the reflected rays according to the surface shape of the concave segment.

𝑃𝑖(𝑥𝑖 , 𝑧𝑖) is a random point of the center segment. 𝑆𝑖𝑃𝑖 is the relevant ray that passes

through the same focal point of the concave segment. The angle of inclination of the

center segment at point 𝑃𝑖 is 𝜃𝑖. The angle between ray 𝑆𝑖𝑃𝑖 and the optical axis is 𝛼𝑖,

and the angle between ray 𝑆𝑖𝑃𝑖 and the reflected ray by center segment is 𝛽. 𝛽 is a

constant for all rays. The angle of inclination of the marginal segment is γ. M is the

end point of the concave segment profile.

For the ray S𝑖𝑃𝑖, its slope is given by

tan 𝛼𝑖 = (𝑥𝑖 − 𝑥𝑆𝑖)/(𝑧𝑖 − 𝑧𝑆𝑖) （5.10）

For the surface of center segment at point 𝑃𝑖 , the angle of inclination 𝜃𝑖 is given by

𝜃𝑖 = (𝛽 − 𝛼𝑖)/2 （5.11）

If we describe the center segment by

𝑧 = 𝑓(𝑥) （5.12）

Then the slope of center segment is given by

tan 𝜃𝑖 = (𝑧𝑖+1 − 𝑧𝑖)/(𝑥𝑖+1 − 𝑥𝑖) （5.13）

The numerical calculation of the surface profile of the center segment starts from the

initial point 𝑃0 that the radius of the center segment |𝑥0| is determined by the lateral

dimension of the SERS substrate. The angle value of 𝛽 can firstly be derived from the

inclination of ray 𝑃0𝑀. Then tan 𝛼0, the slope of the ray S0𝑃0, is obtained by equation

(5.10). The derived angle value of 𝛼0 is introduced to equation (5.11) to get the value

of 𝜃0. In a next step, the coordinate of 𝑃1(𝑥1, 𝑧1) is calculated as the intersection point

of ray S1𝑃1 and the curve 𝑃0𝑃1. The overall profile of the center segment is calculated

by iteration of this process from the coordinates of 𝑃1 to 𝑃𝑖 .

The profile of the marginal segment is a straight line with an angle of inclination that

is given by

𝛾 = (𝜋 − 𝛽)/2 （5.14）

Normally, the thickness of the polymer layer for the microfluidic chip should be larger

than 200µm to guarantee mechanical strength. To reduce the fabrication complexity,


162

we introduce a 1mm thick commercial polymer plate in our design, which is identical

to many other commercial microfluidic chips. We accomplished the segmented

freeform design of two configurations with a diameter of the center segment of 5mm

and 10mm respectively, as shown in Figure 5.4. The numerical aperture of both

configurations is 1.15, making their collecting efficiencies close to that of a 100×

magnification objective lens. The overall diameter of the segmented reflector

increases from 30mm to 38mm when the diameter of the center segment increases

from 5mm to 10mm. The sag of the whole reflector also increases with a larger inner

diameter. As a result, the tunability of the system decreases from 4mm to 0.5mm.

Since the 5mm diameter of the center segment is wide enough to cover the SERS

substrate for our application, we utilize the 5mm center segment design in the

following non-sequential ray tracing simulation, fabrication and experiments.

Figure 5.4 Two configurations of reflector and microfluidic chip design, the diameter of the

center segment for the (a) and (b) configurations are 5mm and 10mm, respectively, and the

overall diameters are 30mm and 38mm. NA of both configurations is 1.15. Moreover, the NA

of the center segment is 0.34 and 0.45 respectively. The blue and red rays refer to the incident

light that interact with different segments.

5.2 Non-sequential simulation of the Raman spectroscopy with

segmented reflector

We performed a non-sequential ray tracing simulation for our system using the

Henyey-Greenstein model in OpticStudio (Zemax, Kirkland, WA, USA). This model

was initialized by Henyey and Greenstein in 1941 to describe the angular distribution

of light scattered by interstellar matter28. Nowadays it has been applied as one of the

typical Rayleigh models in various situations, ranging from the scattering of light by

chemical emulsions29, biological tissues30 to atmospheric and interstellar clouds31.

Since Raman scattering has common features as Rayleigh scattering, the intensity of

Chapter 5

163

scattered light is inversely proportional to the fourth power of the wavelength of the

incident light and the size of particles (typically on the order of 1 nm) inducing the

scattering is much smaller than the wavelength of the excitation, we use this model

for our non-sequential simulations.

In the Henyey-Greenstein model28, the scattered light has the following angular

distribution 𝜌(𝜃)

𝜌(𝜃) = 1

4𝜋

1 − 𝑔2

(1 + 𝑔2 − 2𝑔 cos 𝜃)32

（5.15）

Where, 𝜃 is the angle of scattered light with respect to the incident light. 𝑔 is an

asymmetric factor ranging from -1 to 1. According to equation (5.15), the forward

scattering is dominant when g>0. While g<0, backscattering predominates. For an

oriented molecule or crystal, the angular distribution of Raman scattering is

heterogeneous32. But in most cases, as the molecules in the sample are randomly

oriented, so the Raman scattering is nearly isotropic in 4π steradian, we set g=0 to

simplify the simulation.

The other two important parameters of the Henyey-Greenstein model in Zemax

OpticStudio are transmission and mean free path33. The transmission parameter

indicates how much of the incident light is attenuated during scattering. The mean

free path (M) refers to the probability of the ray being scattered. In Zemax

OpticStudio, the integrated probability of a ray traveling a distance 𝑥 within the

medium and undergoing scattering is given by 33

𝑝(𝑥) = 1.0 − 𝑒−𝑥𝑀 （5.16）

The larger the value of M, the lower the probability that the incident light will be

scattered.


164

Figure 5.5 Non-sequential simulation in ZEMAX OpticStudio: (a) Overall layout, and (b) the

microfluidic chip and segmented reflector. The blue rays and green rays refer to the excitation

light and Raman light respectively. (EL: Excitation lens; CL: Collection lens; BP: Band-pass

filter; M: Mirror; LP: Long-pass filter; DM: Dichroic mirror.)

In addition to the microfluidic chip and segmented reflector, the microfluidic system

also contains external optics consisting out of an excitation path and a collection path

(Figure 5.5). In the excitation path, the excitation laser coming out of the facet of the

excitation fiber is collimated into a parallel beam by the excitation lens. A band-pass

filter (BP) is used to suppress the laser intensity at other frequencies. The mirror (M)

and dichroic mirror (DM) then reflect the collimated beam to a beam expander

consisting out of a concave and a convex lens. The expanded laser beam with the same

diameter of the segmented reflector passes through the microfluidic chip and focuses

inside the fluidic channel by the segmented reflector.

In the collection path, the generated Raman scattering is collimated by the reflector

and bypasses the central region of the microfluidic chip. Then the Raman beam is

reduced in diameter by the beam expander, passes through the dichroic mirror and the

long-pass filter, and is collected into the collection multi-mode fiber by the collection

lens.

Chapter 5

165

In the non-sequential simulation, the wavelength of excitation is 785nm, and the

wavelength of scattering is 850nm, which is close to the fingerprint Raman bands of

ethanol, urea and potassium nitrate. These chemicals are also the analytes we use in

the following experiments. For the microfluidic chip, the channel layer with fluidics,

as well as the two polymer layers can be defined as scattering volumes respectively in

accordance with the Henyey-Greenstein model. The thickness of each layer is 1mm.

We perform non-sequential ray tracing simulations to evaluate three characteristics of

our Raman system: confocality, background suppression, and alignment tolerances of

the external optics with respect to the reflector.

Our Raman system is a confocal system, in which the collection fiber works as the

pinhole to attenuate the out-of-focus scattering. In this confocal system, only the

scattering close to the focus will be collected into the multi-mode fiber (MMF).

Scattering from above or below the focal point is regarded as background and will be

suppressed due to the proper choice of the core diameter of the multi-mode fiber.

Figure 5.6 Confocal behavior of the system by simulation. Each subfigure shows the

efficiency changes with respect to the displacement of the point source in (a) x-, (b) y- and (c)

z-axis, respectively. 100µm, 200µm and 400µm MMF are used as collection fiber.


166

To evaluate the confocality, we place a point source in the fluidic channel to

demonstrate the Raman source from a molecule. The light from this point source can

be received by the MMF through the collection path. By displacing the point source

out of the optimal position in different directions and to detect the flux received on

the fiber tip, we can get the confocality curve, as shown in Figure 5.6. MMFs with

diameters of 100µm, 200µm and 400µm are modeled in our simulation. The collection

efficiency is defined as the proportion of Raman scattered light that can be collected

by our system over the entire 4π steradian of all Raman scattered light. If all backward

Raman scattering can be collected, the normalized efficiency should be 50%.

According to the simulation, a maximum normalized efficiency of 23.8% can be

realized by our Raman setup when using a 400µm MMF. After fitting the confocality

curve with a normal distribution, we can obtain the Full Width Half Maxim (FWHM)

of the flux change with respect to the displacement of the point source, as listed in

Table 5.1.

From the simulation result, we notice that a large core diameter of multi-mode fiber

leads to a higher collection efficiency, but a lower confocality with larger FWHM. It

is reasonable that more scattering from both the in-focus and out-of-focus regions are

collected with a larger core size. Besides, two extrema appear on the confocality curve

at the z direction. This is because the parameters of all the optical components for

simulation as well as experiments are from commercial companies, and the

aberrations when they are aligned together are relatively large, resulting in non-

parallel rays towards the reflector, thereby giving rise to separated focal points for

different segments of the reflector. This phenomenon is more obvious when the

diameter of MMF is small. Therefore, we should select a suitable fiber to get a trade-

off between efficiency and confocality for different scenarios.

Table 5.1 Confocality behavior of the setup from simulations.

MMF (µm) Max Efficiency Confocality - FWHM (µm)

x y z

100 13.4 % 41.8 35.2 120.0

200 19.3 % 57.8 57.4 124.6

400 23.8 % 100.0 100.4 155.6

Confocality refers to the system’s ability to suppress the background outside the

region of interest. However, it is very difficult to measure this parameter

experimentally because no perfect point source is available in reality. Therefore, we

use another related parameter, the suppression factor (SF), to represent the confocality

that is both theoretically and experimentally obtainable.

Chapter 5

167

In the simulation of confocal performance, we define the top, bottom and the channel

layer as scattering volume by the Henyey-Greenstein model. The Raman intensity

observed at the collection fiber from these layers are 𝐼𝑡 , 𝐼𝑏 and 𝐼𝑚 respectively with

the identical excitation. The suppression factor (SF) can be calculated via the

following equation in the simulation

𝑆𝐹 = 𝐼𝑚/(𝐼𝑡 + 𝐼𝑏) （5.17）

As our setup includes the segmented reflector, the microfluidic chip as well as the

external optics, the alignment of the different parts can be critical. Especially when

the collection fiber in the Raman probe plays the role of the pinhole in the confocal

system, a small displacement of the MMF can reduce the Raman intensity obtained.

To assess the influences of the misalignment, we displace the MMF in the x, y and z

direction and measure the flux collected by the fiber. Simulation results are compared

with the experimental results in the next section, as shown in Figure 5.13.

The non-sequential ray tracing simulations also demonstrate that the segmented

reflector can collect 20% more Raman scattering compared to a simple concave mirror

in combination with a 4mm by 4mm SERS substrate inside the fluidic chamber. In

addition, ideally, the spot diameter of the excitation light on the focal plane could

reach the diffraction limit as our segmented reflector is designed via a numerical

approach such that all excitation rays can be focused in a common point. However,

this goal is difficult to achieve because of the aberrations of the external optics used.

According to our non-sequential ray tracing simulation with an optimized layout of

all optical components, 75% of the tracing rays fall within a 5µm diameter circle on

the focal plane. The spot size of the excitation light could be further minimized by

using custom-made optical components with less aberrations.

5.3 Fabrication of the segmented reflector and the microfluidic

chip

After the design and simulation of the segmented freeform reflector, we fabricated

this segmented reflector and the microfluidic chip separately with different

approaches. The base material of the segmented reflector is brass. It was pre-

fabricated to the approximate shape of the segmented reflector and coated with a

Nickel Phosphorus (NiP) layer by electroless nickel plating. A NiP layer can increase

the hardness, corrosion and wear resistance for the segmented reflector when

immersing in water34. Then we use ultra-precision diamond turning to create the


168

segmented surface. But the reflectivity of NiP is only 53% for NIR light. Therefore,

we finalize the segmented reflector by applying a 50nm thick Au layer using a sputter

coater. The reflectivity after coating is larger than 95% for the NIR light. The

brass/NiP approach with Au coating makes it easy to clean the surface with isopropyl

alcohol (IPA) or acetone after an experiment. Figure 5.7 shows the NiP surface of our

segmented reflector and the surface after gold coating.

Figure 5.7 NiP surface of segmented reflector by ultra-precision diamond turning (a) and after

gold coating (b). The reflectivity increases from 53% to 95% for NIR light when adding the

Au coating.

The surface roughness and profile of our segmented reflector are measured with a

non-contact profilometer (Contour GT-I, Bruker, Billerica, MA, USA) and a

multisensor coordinate measurement machine (VideoCheck UA400 with Werth Fiber

Probe, Werth Messtechnik, Giessen, Germany), respectively. The RMS roughness of

the gold coated surface is (14.7 ± 1.4) nm when measuring 10 different areas of

200µm200µm. The standard error of the surface profile is less than 2µm, close to

the detection limit of the Werth Fiber Probe used.

For the microfluidic chip, the three layers are fabricated from PMMA polymer by

laser cutting. The width of the channel is 600µm, and the detection chamber in the

center of the middle layer has a diameter of 6mm. Then we use UV curing adhesive

to bond these three layers together.

We implement a proof-of-concept setup consisting out of the external optics, the

segmented reflector and the microfluidic chip, as shown in Figure 5.8. We use a diode

laser (TEC-500-0785-300, Sacher Lasertechnik, Marburg, Germany) with a

maximum laser power of 300mW to emit 785nm wavelength light for excitation, and

a spectrometer (AvaSpec-HERO, Apeldoorn, Netherlands) to obtain the Raman

signals for collection. The spectrometer is equipped with a 1024x58 pixel back

illuminated TE cooled CCD detector and a blaze grating with grooves of 830 lines/mm

for the 788–1020nm wavelength range. The slit width of the spectrometer is 25µm

corresponding to a spectral resolution of 0.7nm. The conventional Raman spectra

showed are obtained with an excitation power of 150mW and an acquisition time of

Chapter 5

169

15s. In order to prevent the damage of the SERS samples, the SERS measurement is

performed with an excitation power of 70mW and an acquisition time of 5s.

Figure 5.8 Proof-of-concept demonstration setup of our Raman spectroscopy system (left): (a)

the external optics, (b) Microfluidic chip and segmented reflector, (c) Raman probe, (d) beam

expander; Microfluidic chip and segmented reflector in detail (right).

5.4 Preparation of the SERS substrates and SERS chip

The SERS substrate is fabricated with the combination of maskless reactive ion

etching and electron beam evaporation35, as shown in Figure 5.9.

Figure 5.9 The fabrication process of SERS substrate: (a–b) Maskless reactive ion etching

forms silicon nanopillars; (c–d) Electron beam evaporation forms Au layers; (e) SEM image

of the nanopillars with coatings35.


170

An undoped single crystal silicon wafer with a diameter of 4 inches is processed with

an Advanced Silicon Etcher (Surface Technology Systems MESC Multiplex ICP,

Newport, UK) to form a large area of aperiodic nano-pillars with a pillar width of 50–

80nm and a height around 600nm. A 200nm thick Au layer is placed on top of the

nano-pillars with electron beam evaporation (SCM 600, Alcatel, Annecy, France).

The processed wafer is then laser cut into 4mm by 4mm pieces for SERS analysis.

The analytical enhancement factor of this SERS substrate35 with Au coated ø80nm

nano-pillars can be as large as 2.4 × 106, in line with the maximum SMEF by FDTD

simulation.

We bond the backside of the SERS substrate on the top PMMA layer by UV curing

adhesive. The SERS substrate is then integrated inside the microfluidic chamber by

bonding the PMMA-based channel layer and bottom layer together with UV curing

adhesive, as shown in Figure 5.10. By tuning the height of the segmented reflector,

we place the focal point right on the surface of SERS substrate to obtain the best

enhancement.

Figure 5.10 (a) Microfluidic chip integrated with SERS substrate to enhance the sensitivity of

the Raman analysis. (b) Cross-section of the SERS microfluidic Chip.

5.5 Experiments with the microfluidic Raman setup

5.5.1 Suppression factor of microfluidic Raman setup

Because our microfluidic chip is made from PMMA material, the Raman scattering

collected by our setup will also detect the background signal from PMMA. We use

ethanol to determine the experimental suppression factor. The initial Raman spectra

of ethanol and PMMA can be measured by commercial Raman spectroscopy (Figure

5.11(a)). Ethanol has a very strong Raman response at around 875cm-1. Assuming its

peak value is 𝐼𝑠0 at this Raman shift. The Raman signal of PMMA is significantly

strong at 807cm-1, and its peak value is 𝐼𝑛0 under the same criterion as for the ethanol

Chapter 5

171

detection. The Raman spectrum obtained by our setup has peak values of 𝐼𝑠 and 𝐼𝑛 at

875cm-1 and 807cm-1, respectively. The experimental suppression factor can be

calculated as

𝑆𝐹 = (𝐼𝑠

𝐼𝑛

) ∗ (𝐼𝑠0

𝐼𝑛0

) （5.18）

The term 𝐼𝑠0/𝐼𝑛0 is added in the equation to normalize the Raman intensity of different

substances.

As shown Figure 5.11(b), the background from PMMA is suppressed significantly

compared to the Raman spectrum of PMMA in our confocal setup. The comparison

of the suppression factors obtained by experiments and simulation is shown in Figure

5.12. Same as in the simulation, we use different MMFs with core diameters of

100µm, 200µm and 400µm. The smaller the core diameter of the MMF, the larger the

suppression factor, therefore we obtain better confocality. However, in this case the

amount of light coupled into the collection fiber decreases. The suppression factor for

the 400µm and 200µm MMF is around 6, and the suppression factor for the 100µm

MMF is over eight in our proof-of-concept setup.

Figure 5.11 (a) Raman spectra of ethanol and PMMA measured by commercial Raman

microscope (Bruker Senterra, Department of Chemistry of Ghent University), and (b) Raman

spectrum of ethanol in the PMMA chip obtained with our setup using a 200µm diameter

MMF fiber. The PMMA background is suppressed with a factor of 6. (figures are on the same

scale.)


172

Figure 5.12 Comparison of the suppression factors obtained by simulation and experiments.

The average and standard deviation of the experimental suppression factor for each MMF is

calculated over 15 measurements.

5.5.2 Alignment tolerances

For analyzing the alignment tolerances we also use ethanol as analyte. By adjusting

the position of the collection fiber in the x-, y- and z-direction as shown in Figure 5.5

and obtaining the corresponding peak values of the Raman spectrum at 875cm-1, we

get the alignment tolerance curves as shown in Figure 5.13. The average and standard

deviation of each point in the curve are obtained over 10 measurements. All of the

peak values are normalized over 4π steradian to compare with the simulation results.

The results from simulations and experiments are in good agreement, which approves

the validity of our non-sequential ray-tracing approach using the Henyey-Greenstein

model. We use the full width at half maximum (FWHM) of the alignment tolerance

curve (50% drop in efficiency) to determine the misalignment tolerance. Both the

simulated and experimental misalignment tolerance in the horizontal direction (x- and

y-displacements) is approximately 150µm, 220µm and 380µm for the 100µm, 200µm

and 400µm MMF, respectively. The misalignment tolerance in the vertical direction

(z-displacement) is 1.9mm, 2.1mm and 2.7mm for the 100µm, 200µm and 400µm

MMF in the simulations, while according to the experiments the tolerance in the z-

direction is 1.5mm, 1.5mm and 2.3mm. Basically, the misalignment tolerance in the

vertical direction is larger than in the horizontal direction according to the confocal

behavior. We also note that the symmetry of the alignment tolerance curve in the

horizontal direction is better than that in the vertical direction. It is due to the

aberrations of the external optics and the resulted splitting of the focal points of the

different segments, which we have discussed in the confocality analysis.

Chapter 5

173

Figure 5.13 Raman efficiency of our setup with respect to misalignments of the MMF along

(a) x-, (b) y-, (c) z-axis, respectively. 100, 200 and 400µm MMFs are used. The solid lines

refer to the simulation results, and the dashed lines refer to the experimental results.

5.5.3 Limit of detection for conventional Raman analysis

We used two kinds of analytes, urea and potassium nitrate (KNO3), to evaluate the

sensitivity, or the detection limit of the setup for conventional Raman. Urea is a

common simple-structure organic compound with chemical formula CO(NH2)2. It is

one of the metabolites of animal cells and is therefore used as an important indicator

in many biological and medical tests. Urea has a strong Raman scattering near

1000cm-1. Potassium nitrate is one kind of inorganic salt widely used in the industry.

The Raman peak of KNO3 is located around 1040cm-1.

In the sensitivity experiments, we set the integration time to 15 seconds, and the power

of excitation on the microfluidic chip to 150mW. The wavelength of the excitation is

785nm as before. We take 10 measurements for each solution to obtain the Raman

spectra.

First, deionized water was injected into the microfluidic channel and its Raman

spectrum was taken as a reference for the following measurements. Then, the Raman

spectra of a series of aqueous urea and KNO3 solutions from 15mM to 300mM were

measured, as shown in Figure 5.14. After subtracting the reference, we obtained the

mean peak value and standard deviation for each solution. When we plot the Raman


174

intensity of both urea and KNO3 as function of concentration, we get the calibration

curve (Figure 5.15(a)). The signal-to-noise ratio (SNR) for each concentration can be

obtained by dividing the average Raman intensity of each solution by the standard

deviation, as shown in Figure 5.15(b). In a next step, the sensitivity of our setup for

urea and KNO3 can be accessed by the noise-equivalent-concentration (NEC). NEC

is the relative concentration in which its Raman intensity of interest equals to the

noise, or SNR=1. According to the SNR curve, the NEC of our setup for urea and

KNO3 are approximately 19mM and 18mM respectively.

Figure 5.14 Raman spectra of aqueous urea (U) and KNO3 (P) solution with a concentration

from 15 to 300mM. The baseline is corrected by subtracting the reference spectrum of water.

Each spectrum is obtained over 10 measurements.

Figure 5.15 (Left) Calibration curves and (right) SNR curves of aqueous urea and potassium

nitrate solutions. The average and standard deviation for each solution is obtained over 10

measurements. According to the SNR curves, the NEC of our setup for urea and KNO3 are

19mM and 18mM, respectively.

Chapter 5

175

5.6 Using the microfluidic chip and segmented reflector in

combination with a SERS substrate

To further improve the limit of detection, we perform a Raman spectroscopy

measurement with our microfluidic chip in combination with the etched SERS

substrate. For the SERS measurements, we still use a 785nm wavelength laser as

excitation source, but reduce the power of the excitation into the chip from 150mW

to 70mW and decrease the integration time from 15s to 5s to avoid damage of the Au

coated nanopillars. The SERS spectra of 0.1 mM (6 ppm) urea solution and 0.1 mM

(10 ppm) KNO3 solution are shown in Figure 5.16(a). By subtracting the background

SERS signal of water we obtain the baseline corrected SERS spectra of urea and

KNO3, as shown in Figure 5.16(b). The SERS measurements show that even with a

lower power and a shorter integration time, the peaks of 0.1mM analytes can be clearly

observed by the SERS chip. The 0.1mM concentration of urea solution we tested with

our SERS chip is 6.7 times better than the result in literature36. The 0.1mM

concentration of KNO3 detected with our SERS chip is a bit larger than the 1ppm limit

of detection reported37. This can be partially explained by a longer integration time

(60s) used. Nevertheless, the ability of our SERS chip for quantitative analysis needs

to be further investigated. We also measured the SERS spectrum of 10µM Rhodamine

B (RhB) solutions with our SERS chip. The Raman peaks of RhB at 612, 1130, 1274,

1351 and 1499cm−1 can be clearly discriminated. From the SERS-based experimental

results, we observe that RhB is more sensitive than urea and KNO3. The main reason

is a higher cross-section of RhB. In addition, the SERS enhancement can also be

largely impacted by the assimilation of molecules with the surface of the SERS

substrate. RhB molecules can easily adsorb on the SERS substrate, thus inducing a

stronger Raman response of RhB sample. The main SERS bands and their relative

assignments of molecular vibrational modes are listed in Table 5.2. However, it must

be noted that the SERS spectra of water and urea present several abnormal Raman

peaks and bands in the regions around 600, 900, 1500 and 2000cm−1. The presence of

these abnormal Raman bands is likely to be related to the impurity of the analytes and

the contamination during the sample preparation as well as the experiments. As a

result, other molecules rather than the analyte with high assimilation and Raman

cross-sections could generate amplified Raman signals by the high enhancement

factor (~106) of our SERS substrate and are as such added to the spectrum of the

analyte.


176

Figure 5.16 (a) SERS spectra of Rhodamine B, urea, potassium nitrate and water sample

obtained with our microfluidic chip in combination with maskless ion etched SERS

substrates, (b) the baseline corrected SERS spectra of Rhodamine B, urea and potassium

nitrate by subtracting water as reference. Each spectrum is obtained over 10 measurements.

Table 5.2 Main SERS bands of urea, potassium nitrate and Rhodamine B and their relative

intensities and assignments. (ν: stretching, δ: deformation; S: strong, M: medium, W: weak)

Analyte Literature (cm−1) SERS (cm−1) Assignment

Urea38

1012 S 1000 S ν (N-C-N)

1474 W 1467 M ν (N-C-N)

1542 W 1540 M δ (NH2)

Potassium nitrate39 1050 S 1040 S ν (NO3)

1343 W 1357 W ν (NO3)

Rhodamine B40,41

619 S 615.9 S ν (Aromatic C-C)

1130 W 1130 W δ (Aromatic C-H)

1199 M 1190 S δ (Aromatic C-H)

1284 S 1274 M δ (C-C)

1360 S 1351 S ν (Aromatic C-C)


1591 W 1581 S ν (C=C)


5.7 Conclusions

We designed a freeform-segmented reflector-based microfluidic system for

conventional Raman and SERS measurements. The surface profile of our freeform-

segmented reflector is calculated by a numerical approach. We performed non-

sequential ray tracing simulations in OpticStudio by using the Henyey-Greenstein

model to define the scattering volumes. The confocality, alignment tolerances of the

external optics with respect to the microfluidic chip system and background

Chapter 5

177

suppression performance of our setup have been simulated. We fabricated the

segmented reflector and microfluidic chip and implemented a proof-of-concept setup.

The experimental results obtained with our setup are in good agreement with the

simulation results, which proves the feasibility of our simulation approach.

The NEC of our setup for aqueous urea and KNO3 solution is approximately 20mM,

comparable to the value obtained with our previously reported lab-on-chip setup.14 To

further increase the sensitivity of our Raman system, we integrated a SERS substrate

into the channel of the microfluidic chip. The SERS microfluidic chip in combination

with our Raman setup is capable to discriminate 0.1mM urea and KNO3 solutions,

which is compatible with literature results.36,37 Our SERS microfluidic chip is also

able to detect the Raman peaks of RhB solution with a concentration of 10µM. The

SERS microfluidic chip has extended potential applications of our setup to the

biological and chemical application domain that require a lower limit of detection.

However, it should be noticed that the Raman enhancement effect is highly correlated

with the number of molecules adsorbed on the nanostructures. As the adsorption of

different molecules varies widely, the Raman enhancement for various analytes are

quite different even when the samples have the same concentration. The uniformity

of the SERS substrate also greatly affects the Raman response. For the moment, the

existence of compounds in an aqueous solution with low concentration can be

determined with our setup. Nevertheless, the sensitivity of our SERS microfluidic chip

can be further improved by optimizing the optical components in the external optics

and improving the uniformity of the SERS substrate. In addition, we didn’t combine

2PP printed SERS substrates with our microfluidic system because of a lower EF

compared to the Silmeco substrate as well as the restriction of the glass baseplate. But

the integration of the microfluidic chip with 2PP printed SERS substrates should be

further investigated by cutting the glass plate to match with the dimensions of the

microfluidic chamber. Finally, the quantitative analysis capabilities of our system for

SERS measurements should in future be investigated using p-mercaptobenzoic acid

(p-MBA) and biological samples.

References

1. Raman, C. V. A new radiation. India J. Phys 2, 387–398 (1928).

2. Tu, Q. & Chang, C. Diagnostic applications of Raman spectroscopy.

Nanomedicine Nanotechnology, Biol. Med. 8, 545–558 (2012).


178

3. Lim, I. I. S. et al. Gold and magnetic oxide/gold core/shell nanoparticles as

bio-functional nanoprobes. Nanotechnology 19, 305102 (2008).




(2013).

5. Matteini, P. et al. Site-Selective Surface-Enhanced Raman Detection of

Proteins. ACS Nano 11, 918–926 (2017).

6. Krafft, C. & Popp, J. The many facets of Raman spectroscopy for biomedical

analysis. Anal. Bioanal. Chem. 407, 699–717 (2015).

7. Habermehl, A. et al. Lab-on-chip, surface-enhanced Raman analysis by

aerosol jet printing and roll-to-roll hot embossing. Sensors 17, 1–11 (2017).

8. Liu, J., Devoe, D. L. & White, I. M. Microfluidic SERS Using a 3-

Dimensional Porous Monolith as a SERS-Active Solid Phase in a

Microchannel. OSA Conf. 1, 6–7 (2010).

9. Uusitalo, S. et al. Performance and flow dynamics studies of polymeric

optofluidic sers sensors. J. Eur. Opt. Soc. 10, (2015).

10. Khijwania, S. K., Tiwari, V. S., Yueh, F. Y. & Singh, J. P. A fiber optic

Raman sensor for hydrocarbon detection. Sensors Actuators, B Chem. 125,

563–568 (2007).

11. Ashok, P. C., Singh, G. P., Tan, K. M. & Dholakia, K. Fiber probe based

microfluidic Raman spectroscopy. Opt. Express 18, 7642–9 (2010).

12. Ashok, P. C., Singh, G. P., Rendall, H. A., Krauss, T. F. & Dholakia, K.

Waveguide confined Raman spectroscopy for microfluidic interrogation. Lab

Chip 11, 1262 (2011).

13. Dochow, S. et al. Quartz microfluidic chip for tumour cell identification by

Raman spectroscopy in combination with optical traps. Anal. Bioanal. Chem.

405, 2743–2746 (2013).

14. Coster, D. et al. Free-Form Optics Enhanced Confocal Raman Spectroscopy

for Optofluidic Lab-on-Chips. IEEE J. Sel. Top. Quantum Electron. 21,

2701108 (2015).

15. Wang, W., Zhao, J., Short, M. & Zeng, H. Real-time in vivo cancer diagnosis

Chapter 5

179

using Raman spectroscopy. J. Biophotonics 8, 527–545 (2015).

16. Zhou, Q. et al. A Surface-Enhanced Raman Scattering Sensor Integrated with

Battery-Controlled Fluidic Device for Capture and Detection of Trace Small

Molecules. Sci. Rep. 5, 12865 (2015).

17. Perozziello, G. et al. Microfluidic device for continuous single cells analysis

via Raman spectroscopy enhanced by integrated plasmonic nanodimers. Opt.

Express 24, 180–190 (2016).

18. Kline, N. D. et al. Optimization of Surface-Enhanced Raman Spectroscopy

Conditions for Implementation into a Micro fluidic Device for Drug Detection.

Anal. Chem. 88, 10513–10522 (2016).




9, 4328–4336 (2015).



Chem. C 111, 13794–13803 (2007).

21. Blackie, E. Quantification of the Enhancement Factor in Surface-Enhanced

Raman Scattering. (Victoria University of Wellington, New Zealand, 2010).

22. Pilot, R., Signorini, R., Durante, C. & Orian, L. A Review on Surface-

Enhanced Raman Scattering. Biosensors 9, 1–100 (2019).

23. Born, M. & Wolf, E. Principles of Optics: electromagnetic theory of

propagation, interference and diffraction of light. (Pergamon Press, 1970).

24. Pilát, Z. et al. Microfluidic Cultivation and Laser Tweezers Raman

Spectroscopy of E. coli under Antibiotic Stress. Sensors 18, 1623 (2018).

25. Cui, P. & Wang, S. Application of micro fluidic chip technology in

pharmaceutical analysis : A review. J. Pharm. Anal. 9, 238–247 (2019).

26. Niciński, K., Krajcz, J., Kudelski, A., W, E. & Trzcińska, J. Detection of

circulating tumor cells in blood by shell-isolated nanoparticle – enhanced

Raman spectroscopy ( SHINERS ) in microfluidic device. Sci. Rep. 9, 9267

(2019).

27. Lai, C. W., Hsiung, S. K., Yeh, C. L., Chiou, A. & Lee, G. Bin. A cell delivery


180

and pre-positioning system utilizing microfluidic devices for dual-beam

optical trap-and-stretch. Sensors Actuators, B Chem. 135, 388–397 (2008).

28. L. G. Henyey, J. L. G. Diffuse radiation in the Galaxy. Astrophys. J. 93, 70–

83 (1941).

29. Michels, R., Foschum, F. & Kienle, A. Optical properties of fat emulsions.

Opt. Express 16, 5907–5926 (2008).

30. Binzoni, T., Leung, T. S., Gandjbakhche, A. H., Rüfenacht, D. & Delpy, D.

T. The use of the Henyey-Greenstein phase function in Monte Carlo

simulations in biomedical optics. Phys. Med. Biol. 51, 313–322 (2006).

31. Berk, A. et al. MODTRAN cloud and multiple scattering upgrades with

application to AVIRIS. Remote Sens. Environ. 65, 367–375 (1998).

32. Nafie, L., Stein, P., Fanconi, B. & Peticolas, W. L. Angular Dependence of

Raman Scattering Intensity. J. Chem. Phys. 1584–1588 (1970).

doi:10.1063/1.1673171

33. Gangadhara, S. Zemax OpticStudio Knowledgebase - Bulk Scattering with

the Rayleigh Model. (2008). Available at:

https://customers.zemax.com/os/resources/learn/knowledgebase/bulk-

scattering-with-the-rayleigh-model. (Accessed: 5th February 2008)

34. Barycka, I., Hołodnik, B. & Misiuk, A. NiP as a New Material for Thick Film

Technology. Electrocompon. Sci. Technol. 7, 221–226 (1981).



24, OP11–OP18 (2012).

36. Li, M. et al. Microfluidic surface-enhanced Raman scattering sensor with

monolithically integrated nanoporous gold disk arrays for rapid and label-free

biomolecular detection. J. Biomed. Opt. 19, 111611 (2014).

37. Tran, C. T. K., Tran, H. T. T., Bui, H. T. T., Dang, T. Q. & Nguyen, L. Q.

Determination of low level nitrate / nitrite contamination using SERS- active

Ag / ITO substrates coupled to a self-designed Raman spectroscopy system.

J. Sci. Adv. Mater. Devices 2, 172–177 (2017).

38. Frost, R. L., Kristof, J., Rintoul, L. & Kloprogge, J. T. Raman spectroscopy

of urea and urea-intercalated kaolinites at 77 K. Spectrochim. Acta Part A 56,

Chapter 5

181

1681–1691 (2000).

39. Frost, R. L. & Erickson, K. L. Vibrational spectroscopic study of the nitrate

containing hydrotalcite Mbobomkulite. Spectrochim. Acta Part A 61, 2919–

2925 (2005).



12544–12548 (2005).





182

Chapter 6

183

Chapter 6

6 A Compact Conical Beam Shaper and Freeform

Segmented Reflector for SERS Analysis

As described in Chapter 2 and Chapter 3, Raman spectroscopy is a powerful tool for

material characterization. However, the intensity of the scattered photon is inherently

very weak compared to the number of incident photons (proportional to 𝜐𝑜4). Applying

a shorter wavelength laser as excitation gives stronger Raman scattering, but

meanwhile may also results in a higher fluorescence and probably a sample

degradation1. Therefore, surface enhanced Raman scattering (SERS) has been utilized

more and more in research to amplify the intensity of the Raman signal by employing

nanotextured-surfaces with metallic coating2–4. Raman spectroscopy and SERS have

a wide variety of applications in biological research and life sciences such as cancer

diagnostics5–7, DNA/RNA identification8–10, toxin and drug detection11–13. Typically,

SERS measurements are performed by focusing the excitation laser beam onto the

nanotextured-surfaces with a high NA objective lens5–13, as shown in Figure 6.1(a).

The objective lens also plays the role to collect the Raman signal scattered from the

molecules adsorbed on the surfaces. However, the objective lenses are always

designed with refractive structures and therefore are sensitive to wavelengths

especially in the infrared or UV region14. The achromatic and apochromatic objectives

might help to minimize chromatic aberrations15–17. Some of the reflective objective

lenses can eliminate the chromatic aberrations but the obstructions in the beam path

reduces the performance of the objective inevitably18,19. Besides, the working

distances of the high NA objectives are typically very short. Applying a freeform


Liu, Q.; Stenbæk Schmidt, M.; Thienpont, H.; Ottevaere, H. A Compact Conical Beam Shaper

and Freeform Segmented Reflector for SERS analysis. Optics Express 2020, 28, 11.

Chapter 6 A Compact Conical Beam Shaper and Freeform Segmented Reflector for SERS Analysis

184

reflector with high NA can substantially reduce the size of the optical system, and

together with confocal Raman spectroscopy, the background can also be suppressed20.

We have investigated a microfluidic lab-on-chip system with a tunable freeform

segmented reflector for conventional Raman and SERS measurements in Chapter 5.

The non-sequential ray tracing simulations in this design demonstrate that the

segmented reflector can collect 20% more Raman scattering compared to a simple

concave mirror as the Raman scattered light can bypass the SERS substrate thanks to

the segmented design. However, some of the excitation light is still blocked by the

SERS substrate. The segmented design mainly increases the collection efficiency

rather than the excitation efficiency.

To this end, we developed a freeform segmented reflector in combination with a

conical beam shaper for SERS measurements. Our freeform segmented reflector has

a high NA and a long working distance in comparison with commercial achromatic

and apochromatic objective lenses. In combination with a conical beam shaper, all

excitation light hits the sample under test. Also, the obstruction of our segmented

reflector is 0% which implies that the Raman signal can bypass the metallic surfaces

of the SERS substrate to achieve a high collecting efficiency. In addition, a larger

working distance makes the sample preparation and handling easier for SERS

measurements. We calculated the surface profile of our segmented reflector and

conical beam shaper by numerical approaches. The new segmented reflector is more

compact with a smaller sag compared with the one obtained by the numerical

approach given in Chapter 5. The performance of our conical beam shaper and

freeform segmented reflector were assessed by non-sequential ray tracing simulations

and compared to the experimental results obtained. The freeform segmented reflector

and conical beam shaper were fabricated by means of ultra-precision diamond tooling

and turned into a proof-of-concept demonstration setup for SERS measurements.

6.1 Design and fabrication of the freeform segmented reflector

and beam shaper

6.1.1 Working principle of the conical beam shaper and segmented reflector

The two main components of our Raman spectroscopic system are the beam shaper

and the freeform segmented reflector, as shown in Figure 6.1(b).

Chapter 6

185

Figure 6.1 (a) Illustration of conventional SERS measurements with an objective lens; (b)

Principle of our freeform segmented reflector and the conical beam shaper; and (c) the profile

of the segmented reflector calculated by a numerical approach. (S1: center segment; S2:

middle segment; S3; outer segment.)

The beam shaper consists of two reflective conical surfaces with a certain

displacement along the axis. The outside of the conical surface is cut laterally to make

an open aperture for the light rays. In the excitation path, a circular light beam that is

parallel to the optical axis coming through the beam shaper will be expanded into a

ring-beam. The ring-beam is then focused on the SERS substrate by the freeform

reflector which consists out of three segments. The central and outer segments work

together to focus the outer part of the ring-beam onto the surface of SERS substrate.

The middle segment works independently to focus the inner part of the beam-ring to

the same focal point of the two other segments. The diameter of the central segment

is consistent with that of the inner conical surface, making sure that all the excitation

rays are reflected to the focal point. In the collection path, the scattered Raman rays

also interact with the different segments of the freeform reflector. The rays close to

the optical axis are reflected by the central segment and the outer segment

successively (red rays in Figure 6.1(b)). The rays with higher angles of emission can

be collimated directly by the middle segment. Therefore, the Raman scattered light is

collimated into a ring-beam. The beam shaper then reduces the diameter of the

collimated ring-beam of Raman scattering to a circular beam with a smaller diameter.

In this case, most of the Raman scattered light can bypass the opaque SERS substrate

and be delivered to the external optics. The obstruction of the segmented reflector in

combination with the conical beam shaper is 0%.

6.1.2 Numerical approaches to calculate the surface profile of the segmented

reflector

The profile of the segmented reflector is shown in Figure 6.1(c). We calculated the

shape by a numerical approach. The central, middle and outer segments are denoted

by S1, S2 and S3 respectively. 𝑆𝑖 stands for the Raman source of each ray, and in our

calculation all rays come from a common source 𝑆0. In the first step, we can simply


186

obtain the profile of the middle segment S2, which is parabolic according to the

position of the focus 𝑆0 and the radius of curvature R:

𝑧(𝑥) =

𝑥2

𝑅(1 + √1 − (1 + 𝑘)𝑥2

𝑅2)

（6.1）

Where, the conic constant k equals to -1.

In a next step, the center part of the parabolic segment S2 (blue dashed curve in Figure

6.1(b)) is replaced by a new center segment S1. Assume 𝑃𝑖 : (𝑥𝑖 , 𝑧𝑖) is a point on

segment S1, and 𝐸𝑖 : (𝑎𝑖 , 𝑏𝑖) is an arbitrary point belonging to segment S3, we have

𝑃𝑖: (𝑥𝑖 , 𝑧𝑖) ∈ 𝑧 = 𝑓(𝑥)

𝐸𝑖 : (𝑎𝑖 , 𝑏𝑖) ∈ 𝑧 = 𝑔(𝑥)

𝑃0: (𝑥0, 𝑧0) is situated at the intersection of S1 and S2. Segment S3 intersects with S2

at 𝐸0. The focus of all segments is 𝑆0: (𝑥𝑠, 𝑧𝑠).

Then the gradient of the ray 𝑆𝑖𝑃𝑖 can be given by

tan 𝛼𝑖 = (𝑥𝑖 − 𝑥𝑠)/(𝑧𝑖 − 𝑧𝑠) （6.2）

Where, 𝛼𝑖 is the emission angle of the Raman scattered light.

The slope of S1 at 𝑃𝑖 is the derivative of 𝑓(𝑥):

𝑚𝑖 = tan 𝜃𝑖 = 𝑓′(𝑥𝑖) = (𝑧𝑖+1 − 𝑧𝑖)/(𝑥𝑖+1 − 𝑥𝑖) （6.3）

Here 𝜃𝑖 is the inclination of S1 at 𝑃𝑖 .

If we denote the angle between the optical axis and the reflected ray from 𝑃𝑖 by 𝛽𝑖,

according to the law of reflection, we can get the following equation

𝜃𝑖 = (𝛽𝑖 − 𝛼𝑖)/2 （6.4）

Since all the reflected rays pass through a vertex 𝐶𝑖: (𝑥𝑐 , 𝑧𝑐), the tangent of 𝛽𝑖 can be

expressed as

Chapter 6

187

tan 𝛽𝑖 = (𝑥𝑖 − 𝑥𝑐)/(𝑧𝑐 − 𝑧𝑖) （6.5）

The inclination of the normal line of S3 at 𝐸𝑖 equals to half of angle 𝛽𝑖 in agreement

with the law of reflection

𝛾𝑖 = 𝛽𝑖/2 （6.6）

Then the slope of segment S3 at 𝐸𝑖 can be given by

𝑘𝑖 = −1/tan (𝛾𝑖) = (𝑏𝑖 − 𝑧)/(𝑎𝑖 − 𝑥) （6.7）

To calculate the surface profile of S1 by the numerical approach, we start from

𝑃0: (𝑥0, 𝑧0), and calculate the value of 𝛼0 according to equation (6.2), and then get the

value of 𝛽0 by equation (6.5). The inclination angle of S1 at 𝑃0 can be then calculated

by equation (6.4). The next point 𝑃1 on S1 with a tiny step away from 𝑃0 can be

retrieved by equation (6.3). By repeating the above process, we are able to obtain the

surface profile of segment S1.

𝑃0: (𝑥0, 𝑧0) → 𝛼0, 𝛽0 → tan 𝜃0 → 𝑃1: (𝑥1, 𝑧1)

𝑃1: (𝑥1, 𝑧1) → 𝛼1, 𝛽1 → tan 𝜃1 → 𝑃2: (𝑥2, 𝑧2)

…

𝑃𝑖 : (𝑥𝑖 , 𝑧𝑖) → 𝛼𝑖, 𝛽𝑖 → tan 𝜃𝑖 → 𝑃𝑖+1: (𝑥𝑖+1, 𝑧𝑖+1)

To calculate the surface profile of outer segment S3, we start from 𝐸0: (𝑎0, 𝑏0), and

obtain the inclination of the normal line 𝛾𝑖 based on equation (6.6). Then the next

point 𝐸1 on S3 besides 𝐸0 can be calculated by equation (6.7). By repeating the above

calculation, we obtain the overall shape of segment S3.

𝐸0: (𝑎0, 𝑏0), 𝛽0 → 𝑘1 → 𝐸1: (𝑎1, 𝑏1)

𝐸1: (𝑎1, 𝑏1), 𝛽1 → 𝑘2 → 𝐸2: (𝑎2, 𝑏2)

…

𝐸𝑖: (𝑎𝑖 , 𝑏𝑖), 𝛽𝑖 → 𝑘𝑖+1 → 𝐸𝑖+1: (𝑎𝑖+1, 𝑏𝑖+1)


188

Once the surface profiles of different segments are calculated, we obtain the 3D

surface of the reflector by rotating a generatrix, consisting out of the three segments,

around the z-axis as shown in Figure 6.1(c).

Taking into account the size of the SERS substrate (Silmeco ApS, København,

Denmark) which is 4mm × 4mm × 0.7mm, we determined the freeform segmented

reflector with the specifications listed in Table 6.1. For benchmarking reasons, we

designed a segmented reflector with the same diameter as determined by the numerical

approach described in Chapter 5, as shown in Figure 6.2.

Table 6.1 Dimensions of the freeform segmented reflector.

(mm) Previous numerical approach New numerical approach

S1 S2 S3 Total S1 S2 S3 Total

Δx 3.00 5.00 2.52 10.6 7.00 3.00 0.55 10.6

Sag 1.39 2.29 3.92 6.21 2.01 2.12 0.83 2.97

The segmented reflector has an outer radius of 10.55mm to match with the mount of

the Thorlabs’ 1-inch cage system used for the proof-of-concept demonstration. The

center segment has a radius of 7mm to cover the surface area of the SERS substrate.

Our segmented reflector has an overall NA of 0.984. The NA of the central segment

is 0.87, which implies that the central and outer segments capture 61.9% of the total

collection efficiency. The working distance of our segmented reflector - the vertical

distance between the focal point and the front edge of segment S3 - is 1mm, which is

also much longer than commercial dry objective lenses with high values of NA. The

new freeform segmented reflector design has a smaller dimension and longer working

distance compared with the one obtained by the numerical approach described in

Chapter 5.

Figure 6.2 Comparison of the profile of the segmented reflector obtained by the numerical

approaches discussed in Chapter 5 (a) and Chapter 6 (b), respectively. The overall radius of

Chapter 6

189

both reflectors is 10.6mm. However, the sag of the new segmented reflector design is

2.97mm, which is much less than 6.21mm obtained by the previous numerical approach.

The beam shaper has an inner radius of 7mm in line with the geometric boundaries of

the center segment , and an outer radius of 12mm to match with the segmented

reflector as well as to the mount of the Thorlabs cage system used for the proof-of-

concept demonstration with a 1.45mm alignment tolerance along the z-axis.

6.1.3 Fabrication of the conical beam shaper and segmented reflector

We fabricated the beam shaper and segmented reflector by ultra-precision diamond

tooling. The two parts of the beam shaper are made from Poly(methyl methacrylate)

(PMMA) material. The reflective surfaces of the two conical components are coated

with a 200nm thick gold layer using sputtering such that a reflectivity up to 99% for

the excitation and Raman scattered light is obtained, as shown in Figure 6.3(a, b). We

bonded the two parts together by UV curing to make the conical beam shaper, as

shown in Figure 6.3(c). The inner component of the beam shaper has a corner radius

of 50µm due to the diamond tool used for the fabrication. However, the energy loss

caused by this radius value is 1.44% according to our simulation, which can be

neglected. The segmented reflector is made from brass and also coated with a 200nm

thick gold layer to increase the reflectivity, as shown in Figure 6.3(d). The RMS

roughness of the conical surfaces and the segmented reflector are around 7nm

measured by an optical non-contact profilometer (Bruker Contour GT-I, Bruker,

Billerica, MA, USA).

Figure 6.3 Two components (a, b) of the conical beam shaper made from PMMA material

with localized gold coating. (c) Integration of (a) and (b) into a beam shaper and (d) the

freeform segmented reflector made from brass. All reflective surfaces are fabricated by ultra-

precision diamond tooling and coated with gold layers by sputtering.

(a) (b) (c) (d)


190

6.2 Optical system design and simulation

6.2.1 Non-sequential ray tracing simulations for the Raman system

In order to assess the performance of our freeform segmented reflector and the beam

shaper, we performed non-sequential ray tracing simulations in ZEMAX OpticStudio.

Figure 6.4(a) shows a scheme of the entire Raman spectroscopy system. In addition

to the reflector and beam shaper, our system also contains a Raman probe and a mirror

as external optics. In the excitation path, the 785nm wavelength laser emitted from

the single-mode fiber passes through the excitation lens to form a collimated Gaussian

beam, after which the bandpass filter blocks the side bands around 785nm. The

Gaussian beam is then reflected by the long-pass dichroic mirror and a mirror

successively and turns into a ring-beam by the conical beam shaper. The ring-beam is

focused on the SERS substrate by our freeform segmented reflector to generate Raman

scattering. For the collection path, we define a 4mm×4mm scattering surface to mimic

the surface of the SERS substrate. When each 785nm wavelength excitation ray

interact with the scattering surface, it will shift the wavelength to 890nm which is

associated to a 1502cm-1 Raman shift and reemit in an arbitrary direction from the

surface. The scattered Raman rays are collimated by our segmented reflector and

beam shaper, and pass through the long-pass dichroic mirror. A collecting lens is used

to focus the Raman beam into the multi-mode fiber (MMF). To further minimize the

influences of the excitation source and anti-Stokes scattering to the detection, we

placed a long-pass filter in front of the collecting lens. The details of our freeform

segmented reflector and the beam shaper in the non-sequential ray tracing simulation

are shown in Figure 6.4(b).

Figure 6.4 Scheme of the Raman spectroscopy system in non-sequential ray tracing

simulation (a), and the details of the segmented reflector, beam shaper and the SERS substrate

Chapter 6

191

(b). The blue and green lines refer to the excitation and Raman rays respectively. Our system

also contains a Raman probe (the red dashed box) and a mirror as external optics.

6.2.2 Confocal behavior of the Raman system

Our Raman spectroscopy system can be regarded as a confocal system, because the

out-of-focus flare which interferes the Raman signal can be blocked by the core of the

multi-mode fiber. We evaluate the confocality by placing a point source with 890nm

wavelength to mimic the Raman source of an excited molecule and moving it away

from the focus along different axes. A circular virtual detector with 200nm diameter

is placed in front of the multi-mode fiber to investigate the flux changes with respect

to the displacement of the point source. The simulation results are shown in Figure

6.5(a). When the point source is moved away from the focus, the Raman intensity

detected decreases. We use the full-width half-maximum (FWHM) of the curves in

Figure 6.5(a) to evaluate the confocality. The FWHMs for x/y and z displacement are

45µm and 68µm respectively. We find that the Raman intensity changes are

symmetric with the displacement of the point source along the x/y axis, but the

symmetry is poorer along the optical axis. This can be explained by the dissimilar

collimating abilities of the different segments. When the point source is below the

focal plane, as is shown in Figure 6.5(b), the middle segment will reflect the rays to a

convergent beam, but the central and outer segments will convert the rays to a

divergent ring-beam respectively. The performance of the segments is opposite

compared to the situation when the point source is above the focus, as is shown in

Figure 6.5(c). This results in a non-symmetrical Raman intensity change in our

simulations. Ideally, the spot size of the excitation light in the focal plane could reach

the diffraction limit as our segmented reflector is designed via a numerical approach

such that all excitation rays hit a common point. However, in practice this goal is

difficult to achieve because of the aberrations of the external optics used. According

to our non-sequential ray tracing simulations, 75% of the tracing rays out of 106 rays

fall within a 2µm diameter circle on the focal plane.


192

Figure 6.5 Normalized Raman intensity change (a) with respect to the displacement of the

point source from focal plane along the z-axis (b, c). The red and blue rays refer to different

sections of the Raman scattered light interacting with different segments that change to

different directions when the Raman source is moved along the z-axis.

6.3 Proof-of-concept demonstration of the Raman system

We designed the structural frame with CAD software and implemented a proof-of-

concept demonstration setup with a Thorlabs’ cage system, as shown in Figure 6.6(a,

b). The segmented reflector is mounted on a cage translation stage with 1/2" z-axis

travel. Both the beam shaper and the multi-mode fiber are fixed on Thorlabs’ Ø1"

actuated translators such that they can be aligned actively. A single-mode fiber with

5µm core size is connected to a 785nm wavelength diode laser (TEC-500-0785-300,

Sacher Lasertechnik, Marburg, Germany), which is the same wavelength as has been

used in our non-sequential ray tracing simulations. The SERS substrate textured with

nanopillars employed for our detection is fabricated through a combination of

maskless reactive ion etching and electron beam evaporation21, and has an

enhancement factor up to 106. We collect the Raman signal with a 200µm core size

multi-mode fiber and deliver the Raman scattering to a spectrometer (AvaSpec-HERO,

Apeldoorn, Netherlands) for analysis.

Chapter 6

193

Figure 6.6 Structural frame CAD of our Raman spectroscopy (a), and the proof-of-concept

demonstration setup of our Raman spectroscopy system (b). (BPF: Band-pass filter; LPF:

Long-pass filter.)

6.3.1 Performance of the beam shaper

First, we evaluated the performance of our conical beam shaper using non-sequential

ray tracing simulations. Ideally, the 785nm wavelength laser coming out of the single-

mode fiber will be collimated into a Gaussian beam with a diameter of 3.9mm, as

shown in Figure 6.7(a, b). After the conical beam shaper, the Gaussian beam will be

reshaped into a ring-beam, partially shown in Figure 6.7(c, d). We used a CMOS

sensor (Basler Aca 1920-155) to measure the beam profile before and after the conical

beam shaper in the laboratory. The sensor size is 11.3mm × 7.1mm, and it has a pixel

size of 5.86µm × 5.86µm. The shapes of the beam before and after the conical beam

shaper using the CMOS sensor are shown in Figure 6.8(a, b). and Figure 6.8(c, d)

respectively. According to the experimental results, we find that the size of the

Gaussian beam is 3.6mm instead of 3.9mm due to small NA matching errors. However,

the shape of the ring-beam after the beam shaper is in good agreement with the

simulation results. The FWHM widths of the ring-beam in simulations and

experiments are 1.512mm and 1.629mm respectively.

(a) (b)


194

Figure 6.7 Beam shape before (a, b) and after (c, d) the conical beam shaper in the

simulations.

Figure 6.8 Beam shape before (a, b) and after (c, d) the conical beam shaper experimentally

obtained by the CMOS sensor.

6.3.2 SERS measurements for the Rhodamine B

Rhodamine B is a commonly used dye for Raman and fluorescence measurements22.

Therefore, we purchased HPLC Rhodamine B and ethanol (Sigma Aldrich) as

analytes for the experiments. We weighted out 1.437g RhB powder and dissolved it

with ethanol into a 100mL stock solution of 30mM concentration. Then it was diluted

in ethanol into 10µM solutions. Next, 5µL of this 10µM Rhodamine B solution was

deposited on the SERS substrate with a micropipette and dried. We placed the SERS

substrate with the deposited Rhodamine B in front of the freeform segmented reflector

in our setup and conducted experiments to obtain the Raman scattering. To obtain the

Raman spectrum of Rhodamine B we measured with 5 seconds integration time under

5mW excitation power. The spectrum is shown in Figure 6.9. We also measured the

identical Rhodamine B sample under the same excitation power and integration time

but with a 60× commercial microscope objective lens (Newport M-60X) instead of

our beam shaper and segmented reflector. This objective lens has a numerical aperture

of 0.85, and its working distance is 0.3mm.

Chapter 6

195

Figure 6.9 SERS spectra of 10µM Rhodamine B measured with our conical beam shaper in

combination with the freeform segmented reflector, and with a 60× commercial microscope

objective lens. Figures are on the same scale.

The main Raman peaks of Rhodamine B and their intensities are listed in Table 6.2.

We also analyzed the molecular vibrational modes associated with these Raman peaks

accordingly. Rhodamine B has strong Raman peaks at 615.9cm-1, 1187cm-1, 1271cm-

1, 1351cm-1, 1502cm-1, and 1643cm-1. The Raman intensity measured with our

segmented reflector system at 1502cm-1, which is the highest among all peaks, is

approximately 30% higher than the one measured with the 60× objective lens, as

expected from simulations.

Table 6.2 Intensities and peak assignments of main Raman bands of Rhodamine B.

Literature22,23

(cm-1)

SERS & Reflector

(cm-1)

SERS & Objective

(cm-1)

Assignment

619 s 615.9 s 612.3 w Aromatic stretching

1198 s 1187 s 1190 s Aromatic C-H bending

1281 s 1271 s 1271 s C-C bridge-bands stretching

1359 s 1351 s 1351 s Aromatic C-C stretching

1508 s 1502 s 1502 s Aromatic C-C stretching

1523 s 1523 s 1523 s Aromatic C-H bending

1595 w 1595 m 1595 m C=C stretching

1647 s 1643 s 1643 m Aromatic C-C stretching

(s: strong; m: medium; w: weak)

6.3.3 Misalignment tolerances of the setup

The misalignment of the collection fiber with respect to the Raman probe is very

crucial because the fiber tip plays the role as pinhole in the confocal system. Therefore,

we investigated the misalignment of the collection fiber both in non-sequential ray


196

tracing simulations and with SERS measurements. In the non-sequential ray tracing

simulations, we detected the total flux of the 890nm wavelength rays corresponding

to a Raman peak at 1502cm-1 from the scattering surface collected by the multi-mode

fiber from different positions along the x-, y- and z-axis. In the experimental approach,

we measured the Raman spectra of Rhodamine B on the SERS substrate by moving

the position of the fiber which is mounted on a translation stage. Afterwards we got

the curves of the Raman intensity of Rhodamine B at 1502cm-1 with respect to the

displacements of the collection fiber from its optimal position, as shown in Figure

6.10. The average and standard deviation of the Raman intensity at each position are

obtained over 10 measurements. We can conclude that the experimental results are in

good agreement with the simulations.

Figure 6.10 Comparison of misalignment tolerances of the collection fiber in simulation and

experiments. Intensities are normalized. The average and standard deviation of each

displacement are obtained over 10 measurements.

Table 6.3 shows the FWHM values of the misalignment tolerance curves correspond

to -3dB energy loss for the collection fiber along the different axes. The FWHM of

the tolerance along the z-axis is approximately 3 times larger than the one along the

(a) (b)

(c)

Chapter 6

197

x- and y-axis both in simulation and with experimental results obtained. Also the

tolerance curves of the z-axis show less symmetry than the one of x and y. This can

be explained by the fact that different segments of the reflector have different

collecting abilities along the z-axis, identical to the confocal performance in the non-

sequential ray tracing simulations mentioned above.

Table 6.3 FWHM tolerances of the collection fiber.

X (mm) Y (mm) Z (mm)

Simulation 0.119 0.117 0.341

Experiments 0.182 0.144 0.441

We also evaluate the misalignment tolerance of the beam shaper with respect to the

segmented reflector by simulation and experiments respectively, as shown in Figure

6.11. The FWHM tolerances such that the overall efficiency varies no more than 50%

are around 0.225mm both in simulation and experiments.

Figure 6.11 Misalignment tolerance of the beam shaper with respect to the segmented

reflector. Intensities are normalized. The average and standard deviation of each displacement

are obtained over 10 measurements. The FWHM tolerances are around 0.225mm both in

simulation and experiments.

According to the non-sequential ray tracing simulations, a 1mm horizontal

misalignment of the two parts of the conical beam shaper with respect to each other

leads to an energy loss less than -1dB (<20%). In practice, we designed and fabricated

grooves for both parts of the conical beam shaper such that they can be aligned with

a precision better than 5µm in the horizontal direction.

6.4 Conclusion

We designed a freeform segmented reflector with a high NA (0.984) and a conical

beam shaper for confocal SERS measurements. The surface profile of the freeform


198

segmented reflector is based upon a parabolic surface and is calculated by a numerical

approach. We fabricated the freeform segmented reflector and the conical beam

shaper by ultra-precision diamond tooling. The reflective surfaces were coated with

gold layers by sputtering and characterized by an optical non-contact profilometer.

We designed the structural scheme by CAD software and implemented a proof-of-

concept demonstration setup in combination with a Raman probe.

We assessed the confocality of the Raman system using the non-sequential ray tracing

simulation by shifting an ideal point source along different axes and measured the

total flux collected by the multi-mode fiber. The simulation result show that the

Raman intensity changes are symmetric with the displacement of the point source

along the x/y-axis, but the symmetry is less good along the optical axis due to

dissimilar collimating abilities of the different segments. We also evaluated the

performance of the conical beam shaper in non-sequential ray tracing by placing a

virtual detector before and after the beam shaper. The simulation results were

compared and in good agreement with the actual beam shape measured with a CMOS

sensor.

We measured the spectra of 10µM Rhodamine B solution in ethanol deposited on the

SERS substrate and analyzed the assignments of different Raman bands with respect

to the vibrational modes of Rhodamine B. In addition, we benchmarked our system

and performed a reference measurement using the same SERS sample by replacing

our segmented reflector and beam shaper with a 60× commercial objective lens.

According to the experimental results, the collecting efficiency of our freeform

segmented reflector is 30% higher than the efficiency of the commercial objective

lens. The use of a freeform reflector can also greatly reduce the number of optical

surfaces thereby resulting in an optical component with a smaller size. Moreover, we

assessed the misalignment tolerances of the conical beam shaper with respect to the

freeform segmented reflector and the misalignment tolerance of the collecting fiber

respectively in non-sequential ray tracing simulations and with the experiments by

comparing the Raman intensity measured. The experimental results were in good

agreement with the simulation results.

Our freeform segmented reflector in combination with the conical beam shaper

performed well for the confocal SERS measurements. The obstruction of our

reflector-based Raman system is 0% such that all the excitation light can be utilized

and 82% of the SERS signal can be collected under 2πsr detection geometry. Because

of its high NA and confocality, the potential applications of the latter system should

be further investigated. There is a trend in SERS analysis towards illuminating a larger

area to reduce the Raman signal fluctuations due to the variation of the nanostructures.

Chapter 6

199

And for biological research the lack of imaging ability might restrict its applicability.

Therefore, future work involves optimization of the segmented reflector for mapping

larger areas to make a trade-off between collection efficiency and noise reduction, and

for visualizing the sample under test by taking into account the field of view (FOV).

References

1. De Luca, A. C., Mazilu, M., Riches, A., Herrington, C. S. & Dholakia, K.

Online fluorescence suppression in modulated raman spectroscopy. Anal.

Chem. 82, 738–745 (2010).

2. Ru, E. C. Le & Etchegoin, P. G. Principles of Surface-Enhanced Raman

Spectroscopy and related plasmonic effects. Elsevier Ltd 1–656 (2009).

doi:10.1016/B978-0-444-52779-0.00003-9

3. Davies, R. A., Chong, N. S. & Ooi, B. G. Chemical Enhancement of the

Surface Enhanced Raman Scattering Signals of Anilines via Their Ortho-

Substituents. Opt. Photonics J. 03, 13–23 (2013).

4. Samanta, A. et al. Ultrasensitive Near-Infrared Raman Reporters for SERS-

Based In Vivo Cancer Detection. Angew. Chemie Int. Ed. 50, 6089–6092

(2011).

5. Lee, S. et al. Rapid and sensitive phenotypic marker detection on breast

cancer cells using surface-enhanced Raman scattering (SERS) imaging.

Biosens. Bioelectron. 51, 238–243 (2014).

6. Chon, H., Lee, S., Son, S. W., Oh, C. H. & Choo, J. Highly sensitive

immunoassay of lung cancer marker carcinoembryonic antigen using surface-

enhanced raman scattering of hollow gold nanospheres. Anal. Chem. 81,

3029–3034 (2009).

7. Sha, M. Y., Xu, H., Natan, M. J. & Cromer, R. Surface-Enhanced Raman

Scattering Tags for Rapid and Homogeneous Detection of Circulating Tumor

Cells in the Presence of Human Whole Blood. J. Am. Chem. Soc. 130, 17214–

17215 (2008).

8. Song, H., Lee, S., Kang, T. & Kim, K. Silver nanoislands based SERS-active

substrates for DNAs detection. Prog. Biomed. Opt. Imaging - Proc. SPIE


200

9340, (2015).

9. Kang, T., Yoo, S. M., Yoon, I., Lee, S. Y. & Kim, B. Patterned Multiplex

Pathogen DNA Detection by Au Particle-on-Wire SERS Sensor. Nano Lett.

10, 1189–1193 (2010).

10. Bell, S. E. J. & Sirimuthu, N. M. S. Surface-Enhanced Raman Spectroscopy

( SERS ) for Sub-Micromolar Detection of DNA / RNA Mononucleotides. J.

Am. Chem. Soc. 128, 15580–15581 (2006).

11. Zhou, Y., Chen, J., Zhang, L. & Yang, L. Multifunctional TiO2-coated Ag

nanowire arrays as recyclable SERS substrates for the detection of organic

pollutants. Eur. J. Inorg. Chem. 2012, 3176–3182 (2012).




(2013).



14. Sasian, J., Gao, W. & Yan, Y. Method to design apochromat and

superachromat objectives. Opt. Eng. 56, 105106 (2017).

15. Abramowitz, M., Spring, K. R., Keller, H. E. & Davidson, M. W. Basic

Principles of microscope objectives. Biotechniques 33, 772–781 (2002).

16. Greisukh, G. I., Ezhov, E. G., Levin, I. A. & Stepanov, S. A. Design of

achromatic and apochromatic plastic micro-objectives. Appl. Opt. 49, 4379–

4384 (2010).

17. Anthony O’Connor. Microscope objectives. (1973).

18. Jones, L. Handbook of optics. (McGRAW-HILL, INC., 1995).

19. Wang, H., Yang, X., Liu, Y., Jiang, B. & Luo, Q. Reflection-mode optical-

resolution photoacoustic microscopy based on a reflective objective. Opt.

Express 21, 659–661 (2013).



(2015).

Chapter 6

201



24, OP11–OP18 (2012).



12544–12548 (2005).




Chapter 7

203

Chapter 7

7 Conclusions and Perspectives

In this chapter we summarize the most important achievements and future challenges

for the microfluidic Raman and SERS setups developed in this PhD dissertation, and

finish with the most promising perspectives.

7.1 Conclusions

7.1.1 3D printed nanostructures by two-photon polymerization for SERS

analysis

We presented nano-pillar arrays ranging from 200nm to 600nm as SERS substrates

for mycotoxin detection. We built a nominal shape and a voxel-based model for

simulating the enhancement of the electric field of the nano-pillar arrays using Finite-

Difference Time-Domain (FDTD) method. We fabricated the nano-pillar arrays by

means of two-photon polymerization. The nanostructures are coated with a 20nm gold

layer by sputtering. We investigated the morphological specifications of the

nanostructures by analyzing the SEM images and AFM data. A new model was built

based on the AFM data obtained from the fabrication results and introduced to the

FDTD simulation. We demonstrated the enhancement behavior by measuring the

Raman spectrum of Rhodamine B solutions. Besides, we determined the limit of

detection of the 200nm pillar array by performing Raman measurements on

Rhodamine B solutions with different concentrations. The experimental enhancement

factor of our 200nm nano-pillar array is close to 104. The detection limit of our 200nm

nano-pillar array is 0.55µM for Rhodamine B solution. Finally, we discriminated

1ppm Deoxynivalenol and 1.25ppm Fumonisin b1 in acetonitrile solutions by our

SERS substrate in combination with principal component analysis. The two-photon

polymerization approach allows fast and flexible prototyping of SERS substrates for

Chapter 7 Conclusions and Perspectives

204

material characterization. The main drawback of the two-photon polymerized SERS

substrate is the low utilization ratio of polymer material as only a very small portion

of the photoresist is polymerized into nanostructures, while most of the remaining

photoresist is washed away and wasted after polymerization. In addition, this serial

polymerization process is slow, resulting in a high cost for the fabrication of an

individual sample. Important to mention is that the SERS enhancement performance

of our 2PP printed substrates varies in terms of the detection region. This can partially

be explained by the nature of the SERS mechanism where the adsorption, orientation,

distribution and other features of the molecule under test can influence the

enhancement. But the variation of the dimensions of the nanostructures due to

fabrication errors is more influencing fluctuations of the SERS enhancement.

However, it should be possible to reduce this dimension variation by optimizing

various aspects during two-photon polymerization, such as the fluctuations of the

femtosecond laser power, the photoresins and the voxel path compiling.

7.1.2 Mass fabricated lab-on-chip with integrated freeform reflector in

combination with a Raman probe for microfluidic analysis

We developed a working flow for the mass fabrication of a microfluidic lab-on-chip

with an integrated freeform reflector to reduce the cost as well as to increase the

repeatability and reproducibility for Raman measurements. The shape of the freeform

reflector is calculated via a numerical approach. The freeform shape and the

microfluidic channels are replicated starting from Topas COC material by double-

sided hot embossing. The master molds for the reflector and channels are fabricated

from brass material with NiP coatings on the surfaces to enhance the resistance to

chemical erosion and oxidation during hot embossing. The NiP-brass-based master

molds also improve the strength of the molds and result in a higher surface quality of

the replicas. We sealed the microfluidic channel by UV curing adhesive and thermal

bonding and used FEP tubing as in/outlet. We integrated the mass fabricated lab-on-

chip with a miniaturized Raman probe and implemented a proof-of-concept setup. Our

setup has a noise equivalent concentration of 3.4mM for urea detection. We also

conducted measurements with our setup for mycotoxin detection. The spectra of

mycotoxins obtained by our Raman setup show a very good repeatability with tight

group of replies after Principal Component Analysis. However, the sensitivity of our

lab-on-chip with Raman probe for conventional Raman analysis is relatively low,

which cannot meet the requirements in many application domains, including

mycotoxin detection in food safety. Therefore, the combination of SERS with our

microfluidic lab-on-chip is crucial to further increase the sensitivity of Raman analysis.

Chapter 7

205

7.1.3 Freeform segmented reflector design for conventional Raman and SERS

analysis

We presented a freeform-segmented reflector microfluidic system for conventional

and SERS Raman analysis. The segmented reflector with a NA of 1.15 is directly

designed by a numerical approach. The polymer-based microfluidic system strongly

suppresses the undesirable background signal because it enables confocal detection of

Raman scattering through the combination of a freeform reflector and a microfluidic

chip. We performed systematic simulations using non-sequential ray tracing with the

Henyey-Greenstein model to assess the Raman scattering behavior of the substance

under test. We fabricated the freeform reflector and the microfluidic chip by means of

ultra-precision diamond turning and laser cutting respectively. We demonstrated the

confocal behavior by measuring the Raman spectrum of ethanol. Besides, we

calibrated the setup by performing Raman measurements of urea and KNO3 solutions

with different concentrations. The detection limit of our microfluidic system is

approximately 20mM according to the experiment. Next, we implemented a SERS

microfluidic chip and could discriminate 100µM urea and KNO3 solutions.

In addition, we developed a new segmented reflector design with an improved

numerical approach. The new segmented design is more compact compared to the

original segmented reflector and has an NA of 0.984. The new segmented reflector is

used for SERS analysis in combination with a conical beam shaper. We performed

systematic simulations in non-sequential ray tracing to assess the detecting abilities

of the substance under test. We implemented a proof-of-concept setup and

demonstrated the confocal behavior by measuring the SERS signal of 10µM

Rhodamine B solution. The experimental results agree well with the simulations

concerning the misalignment tolerances of the beam shaper with respect to the

segmented reflector and the misalignment tolerances of the collecting fiber. In

addition, we conducted a benchmark analysis for SERS measurements by using a 60×

objective lens with a numerical aperture of 0.85. We found that the main Raman

intensity of Rhodamine B at 1503cm-1 obtained by our segmented reflector working

together with the conical beam shaper is approximately 30% higher than the intensity

obtained with the commercial objective lens.

However, the sensitivity of our Raman setup with the segmented reflector is highly

correlated with the number of molecules adsorbed on the nanostructures. As the

adsorption of different molecules varies widely, the Raman enhancement for various

analytes is different even when the samples have the same concentration. The

uniformity of the SERS substrate also greatly affects the Raman response. At this

moment, the existence of some compounds in an aqueous solution with low


206

concentration can be determined with our setup. Nevertheless, the sensitivity of our

SERS microfluidic chip can be further improved by optimizing the optical

components in the external optics and improving the uniformity of the SERS substrate.

In addition, the quantitative analysis capabilities of our system for SERS

measurements should be further investigated.

7.2 Perspectives

7.2.1 Optimization of the nanostructures for SERS analysis and mass

manufacturing of the nanostructure

With this PhD, we have presented the development of 2PP printed nanostructures used

for SERS analysis. In a preliminary study, we have investigated nano-pillar array,

nano-hemisphere arrays and nano-grid with dimensions from 200nm to 600nm with

an interval of 100nm. The FDTD simulations as well as the experimental results show

that the 200nm pillar array-based SERS substrate has the largest enhancement

compared to the other SERS substrates we fabricated. However, only nano-pillar

arrays with an aspect ratio of 1 are simulated by the FDTD method and fabricated by

2PP in this PhD. In future work, a mapping of different aspect ratios with different

values of height, diameter and pitch can be simulated with the FDTD method. The

dimensions can also be reduced to smaller values. In addition, nanostructures with

other shapes, such as nano-pyramids, -square columns, -rings and even complex

patterns with combined nanostructures, can be studied to find optimal designs with

high LSPR.1–5 Last but not least, although we have discussed the EM field

enhancement of different coating materials for a simple nanosphere, the SERS

enhancement of 2PP printed nanostructures with different coatings is also an

interesting topic for future research. By implementing these ideas an optimal SERS

substrate realized through 2PP or other fabrication methods can be obtained. However,

the main challenge to optimize the nanostructures for SERS is the huge computing

capacity and long computing time required by FDTD because of the nature of the

FDTD method, as well as the complex combination of different parameters. The data

processing of the simulated results as well as the theoretical analysis can be

challenging as well. Possible solutions to shorten the FDTD simulations could be

employing a server with a strong computing capacity or using a commercial

computing cloud. Once the optimal nanostructures for SERS are obtained by FDTD

simulations, we will be able to fabricate the SERS substrates and conduct experiments

to further investigate their enhancement performances.

Chapter 7

207

Two-photon polymerization is suitable for prototyping SERS substrates for academic

analysis. However, the cost of an individual SERS substrate fabricated with 2PP is

relatively high because of the long preparation and writing times and the material

requirements. Mass fabrication of the nanostructures is necessary to increase the

fabrication efficiency and reduce the cost. We have investigated soft lithography for

the replication of 2PP printed nanostructures using PDMS as photoresin, as shown in

Figure 7.1. This mass fabrication approach should be further studied to investigate the

quality of the replicas and their performance for SERS analysis.

Figure 7.1 Nanostructure replication with 2PP printed master molds.

7.2.2 Towards a 2PP fabricated optofluidic lab-on-chip for SERS analysis

In this PhD, we have tested our 2PP printed SERS substrates. We also integrated a

commercial SERS substrate (Silmeco) with a mass fabricated LoC and a tunable

freeform segmented reflector for microfluidic analysis. The ongoing technological

developments on lab-on-chip devices, together with a faster 2PP technology with an

increased accuracy and resolution will enable fully integrated optofluidic lab-on-chips

for SERS analysis.6–9 A promising optofluidic lab-on-chip with SERS substrate that

can directly be fabricated using 2PP is shown in Figure 7.2. The upper polymer is an

integrated optical part with a chamber surrounded by a reflector. The nanostructures

are located in the center of the chamber. Reflective layers for the reflector and the

noble metal surface on top of the nanostructures can be coated by thermal evaporation

or sputtering, the inner surface of the reflector will not be coated. A bottom polymer

layer is used for sealing. A fluidic sample passes through the microfluidic channel and

will fill the chamber surrounded by the reflector. Molecules adsorbed on the surface


208

of the SERS substrate can be excited and generate enhanced Raman scattering under

laser excitation. The SERS signal will be collected by the reflector along the collection

path (blue arrows).

Figure 7.2 Towards a monolithic optofluidic lab-on-chip for SERS analysis. The monolithic

part can be directly fabricated by 2PP with a reflector shape and nanostructures in the center.

The reflective coating layer for the reflector and the metal surface for SERS can be added by

local thermal evaporation.

7.2.3 Towards a lab-on-chip with integrated micro-excitation source and

micro-spectrometer.

With this PhD we have integrated a freeform reflector-based microfluidic setup in

combination with a miniaturized Raman probe. This setup can be used for on-site

Raman analysis where portability is one of the main considerations. The last few years

we have seen the development of Micro-Electro-Mechanical Systems (MEMS) for the

miniaturization of spectrometers on chip.10–14 Micro-laser sources with high

throughput has also been investigated and applied in various fields.15–19 Figure 7.3

shows an electrostatically driven tiltable grating-based MEMS chip used for mobile

micro-spectrometers. The size of the micro-spectrometer equipped with this MEMS

chip is 12.2mm × 6mm × 10.8mm.

Chapter 7

209

Figure 7.3 Scanning grating MEMS chip ((a) face up and (b) face down); and rendered

images of the micro-spectrometer ((c) and (d)). The dimensions of the MEMS chip are 9.6mm

× 5.3mm × 0.5mm. The tiltable diffraction grating of this MEMS chip is electrostatically

driven.14

Our system can be further miniaturized by integrating the excitation source and

spectrometer into a single chip device. The integration of the laser source and

spectrometer, together with a microfluidic system will enable fully in vivo and in situ

analyses in various application domains because of the high flexibility, low cost, small

volume consumption as well as energy saving.

These goals are ambitious but not unrealistic within the expertise and complete value

chain provided by B-PHOT from material development, optical modeling and design

to the mass-fabrication for the integration of optical components and photonics

technology.

References




2. Guo, H., Xing, B., Hamlet, L. C., Chica, A. & He, L. Surface-enhanced

Raman scattering detection of silver nanoparticles in environmental and

biological samples. Sci. Total Environ. 554–555, 246–252 (2016).


210

3. Reznickova, A., Novotna, Z., Kasalkova, N. S. & Svorcik, V. Gold

nanoparticles deposited on glass: Physicochemical characterization and

cytocompatibility. Nanoscale Res. Lett. 8, 1–8 (2013).

4. Zhang, Z., Liao, F., Ma, S., Gao, S. & Shao, M. SERS and FDTD simulation

of gold nanoparticles grafted on germanium wafer via galvanic displacement.

Surf. Interface Anal. 47, 398–402 (2015).






1–7 (2013).

7. Li, Y. et al. Lab on a Chip channels and coupled waveguides via two-photon.

Lab Chip 16, 4406–4414 (2016).

8. Accoto, C. et al. Two-Photon Polymerization Lithography and Laser Doppler

Vibrometry of a SU-8-Based Suspended Microchannel Resonator. J.

Microelectromechanical Syst. 24, 1038–1042 (2015).

9. Liu, Y.-J. et al. A simple and direct reading flow meter fabricated by two-

photon polymerization for microfluidic channel. Microfluid Nanoflui 18,

427–431 (2015).

10. Yang, Z. et al. Single-nanowire spectrometers. Science (80-. ). 365, 1017–

1020 (2019).

11. Cadusch, J. J., Meng, J., Craig, B. & Crozier, K. B. Silicon microspectrometer

chip based on nanostructured fishnet photodetectors with tailored

responsivities and machine learning. Optica 6, 1171–1177 (2019).

12. Erfan, M. et al. On-Chip Micro-Electro-Mechanical System Fourier

Transform Infrared ( MEMS FT Spectrometer-Based Gas Sensing. Appl.

Spectrosc. 0, 1–8 (2016).

13. Tormen, M., Lockhart, R., Niedermann, P. & Overstolz, T. MEMS tunable

grating micro-spectrometer. in International Conference on Space Optics —

ICSO 2008 Proc. of S, 1056607 (2008).

14. Pugner, T., Knobbe, J. & Gru, H. Near-Infrared Grating Spectrometer for

Chapter 7

211

Mobile Phone Applications. Appl. Spectrosc. 70, 734–745 (2016).

15. Marell, M. J. H. & Hill, M. T. Progress in Metal-insulator-metal Waveguide

Lasers at Near-infrared Wavelengths. Prog. Electromagn. Res. 1411 (2009).

16. Lim, H. H. & Taira, T. Sub-nanosecond laser induced air-breakdown with

giant-pulse duration tuned Nd : YAG ceramic micro-laser by cavity-length

control. Opt. Express 25, 19891–19899 (2017).

17. Sieno, L. Di et al. Miniaturized pulsed laser source for time-domain diffuse

optics routes to wearable devices diffuse optics routes to wearable devices. J.

Biomed. Opt. 22, 085004 (2017).

18. Wang, X., Wang, C., Wu, X., Zhu, L. & Jing, H. Researching the 915 nm

high-power and high- brightness semiconductor laser single chip coupling

module. J. Semicond. 38, 024006–1 (2017).

19. Wan, Y. et al. 1.3 μm submilliamp threshold quantum dot micro-lasers on Si.

Optica 4, 940–944 (2017).


212

List of Publications

213


Journal papers

1. Liu, Q.; Stenbæk Schmidt, M.; Thienpont, H.; Ottevaere, H. A Tunable

Freeform-Segmented Reflector in a Microfluidic System for Conventional

and Surface-Enhanced Raman Spectroscopy. Sensors 2020, 20, 1250. DOI:

10.3390/s20051250.

2. Liu, Q.; Vanmol, K.; Lycke, S.; Van Erps, J.; Vandenabeele, P.; Thienpont,

H.; Ottevaere, H. SERS using two-photon polymerized nanostructures for

mycotoxin detection. RSC Advanced 2020, 10, 14274 – 14282. DOI:

10.1039/d0ra01909g.

3. Liu, Q.; Stenbæk Schmidt, M.; Thienpont, H.; Ottevaere, H. A Compact

Conical Beam Shaper and Freeform Segmented Reflector for SERS analysis.

Optics Express 2020, 28, 11. DOI: 10.1364/OE.391623.

4. De Coster, D., Liu, Q., Vervaeke, M., Van Erps, J. A., Missine, J., Thienpont,

H., Ottevaere, H. (2017). Optofluidic chip for single-beam optical trapping

of particles enabling confocal Raman measurements. IEEE J. Sel. Top.

Quantum Electron., 23(2), [5500109]. DOI: 10.1109/JSTQE.2016.2584787.

Conference proceedings

1. Liu, Q., Barbieri, G., Thienpont, H., Ottevaere, H. (2017). Integrated

confocal Raman probe combined with a free-form reflector based lab-on-

chip. In Optical System Alignment, Tolerancing, and Verification XI (Vol.

10377). [1037706] SPIE. DOI: 10.1117/12.2274936.

2. Liu, Q., De Coster, D., Loterie, D., Van Erps, J. A., Vervaeke, M., Missinne,

J., Ottevaere, H. (2016). Proof-of-concept demonstration of free-form optics

enhanced confocal Raman spectroscopy in combination with optofluidic lab-

on-chip. In Micro-Optics 2016 (Vol. 9888). [UNSP 98880E] SPIE. DOI:

10.1117/12.2227386


214

3. Ottevaere, H., Liu, Q., De Coster, D., Van Erps, J. A., Vervaeke, M.,

Thienpont, H. (2017). Optofluidic chips for Raman spectroscopy and optical

trapping. In 2017 Conference on Lasers and Electro-Optics (CLEO) IEEE,

Piscataway, NJ, USA.

4. Ottevaere, H., Liu, Q., De Coster, D., Van Erps, J. A., Vervaeke, M.,

Thienpont, H. (2016). Novel microfluidic devices for Raman spectroscopy

and optical trapping. In Current Developments in Lens Design and Optical

Engineering XVII (Vol. 9947). [UNSP 994709] SPIE. DOI:

10.1117/12.2239996.

Appendix 1 List of Tables

215


Table 1.1 General characteristics of some materials for microfluidic LoC fabrication.

.................................................................................................................................. 15

Table 1.2 Raman peaks and relative intensities of different types of polymers. (s:

strong; m: medium; w: weak) ................................................................................... 16 Table 1.3 Chemical resistance and biocompatibility of some typical materials for lab-

on-chip fabrication. ................................................................................................... 17

Table 2.1 The electromagnetic spectrum.2 ................................................................ 33 Table 2.2 Vibrational modes of some chemical bonds and their spectra intensity.8 . 43

Table 2.3 A comparison of cross sections between different spectroscopy. ............. 48 Table 2.4 Features of Raman, IR absorption and fluorescence spectroscopy. .......... 48 Table 2.5 Sample applicability of Raman, IR absorption and fluorescence

spectroscopy.18–20 ...................................................................................................... 49 Table 2.6 Typical lasers used in Raman spectroscopy. ............................................. 50

Table 2.7 Permittivity of some materials and maximum amplitude of the

electromagnetic field of a 100nm diameter nanosphere under 785nm excitation.

(Lumerical FDTD database & Simulation results) ................................................... 55 Table 2.8 Comparison of different types of SERS substrates. .................................. 67

Table 2.9 Number of publications on Raman spectroscopy and Surface Enhanced

Raman spectroscopy (SERS) from 1955 to 2018 categorized according to the fields

of application. (Statistic data extracted from the Web of Science) ............................ 68

Table 3.1 Dimensions of nano-pillar arrays measured with SEM and AFM. (all unit

in nm) ........................................................................................................................ 86

Table 3.2 The intensities and peak assignments of main Raman bands of Rhodamine

B. .............................................................................................................................. 90

Table 3.3 The characteristic peaks and assignments of main Raman bands of FUM

and DON according to PCA. .................................................................................... 95 Table 4.1 List of the optical components modelled in the simulation model and present

in the final probe design. ........................................................................................ 114

Table 4.2 Simulation results of different layouts and the calculated efficiencies. .. 116

Table 4.3 Misalignment tolerances between the optical components within the probe.

(d: the distance between the dichroic mirror DM and the mirror M; L: the length of

the probe refers to the vertical distance between the fiber tip and the dichroic mirror

or the mirror) .......................................................................................................... 118 Table 4.4 Misalignment tolerances between the chip and the Raman probe. ......... 118 Table 4.5 Leading grades of Topas COC polymers. ............................................... 123 Table 4.6 GPC results of different types of Topas COCs and PMMA. .................. 124


216

Table 4.7 Specifications of the freeform reflectors. ................................................ 126 Table 4.8 Polynomial fitting results of different freeform reflector designs. .......... 128 Table 4.9 Shear strength of different bonding approaches for different materials. (unit:

MPa) ....................................................................................................................... 141 Table 5.1 Confocality behavior of the setup from simulations. .............................. 166

Table 5.2 Main SERS bands of urea, potassium nitrate and Rhodamine B and their

relative intensities and assignments. (ν: stretching, δ: deformation; S: strong, M:

medium, W: weak).................................................................................................. 176

Table 6.1 Dimensions of the freeform segmented reflector. ................................... 188

Table 6.2 Intensities and peak assignments of main Raman bands of Rhodamine B.

................................................................................................................................ 195

Table 6.3 FWHM tolerances of the collection fiber ............................................... 197

Appendix 2 List of Figures

217


Figure 1.1 Sum of times cited per year for 945 results on TOPIC (Raman) AND

TOPIC (microfluidic) from 1989-2019 on the Web of Science data base. (Accessed on

Mar. 25, 2020) ............................................................................................................ 2 Figure 1.2 Workflow of this PhD work. ..................................................................... 5

Figure 1.3 (Left)The first apparatus of flow analysis setup for determination of organic

and in-organic compounds by using (Right) silicone rubber-based graphite electrodes.

(C) Valve, (E1) Indicator electrode, (E2) Reference electrode. Presented by E. Pungor

et al.2,3 ......................................................................................................................... 6 Figure 1.4 The first actual FIA system described by J. Ruzicka and E. Hansen.4 (a)

Polymer blocks; (b) Silicone rubber wall; (c) polyethylene tubing. ........................... 6 Figure 1.5 (Left) Block diagram of the first LoC system developed by C.T. Stephen

et al.6 (Right) Photograph of micro capillary column and gas chromatographic system

on a 5-cm-diameter silicon wafer. .............................................................................. 7 Figure 1.6 Cross-section of the Raman detection chamber and liquid nitrogen cold

stage for in situ detection, presented by B. Barry and R. Mathies.11 .......................... 7 Figure 1.7 (a) Layout of the glass microfluidic chip and (b) photograph of the

microfluidic Raman spectroscopy setup, demonstrated by L. Moonkwon et al.15 ...... 8 Figure 1.8 (Left) schematics of the thiolene-based microfluidic lab-on-chip and (right)

photograph of the Raman spectroscopy system for droplet formulation analysis

presented by S. Barnes et al.16 .................................................................................... 9 Figure 1.9 Schematic of the glass-based SERRS microfluidic lab-on-chip

demonstrated by R. Keir et al.17 ................................................................................ 10 Figure 1.10 (a) Schematic and (b) photograph of the optical probe-based PDMS

Raman lab-on-chip device demonstrated by P. Ashok et al.19 .................................. 10 Figure 1.11 Schematic of the PDMS-based WCRS microfluidic chip presented by P.

Ashok et al.20 ............................................................................................................ 11

Figure 1.12 Design of a fiber Raman probe with miniaturized mirrors and lenses for a

microfluidic lab-on-chip system. (a) Non-sequential ray tracing of the microfluidic

system. (b) Schematic drawing of a mold for PDMS probe fabrication. (c) Schematic

(left) and photograph (right) of the PDMS probe. Demonstrated by T.

Ngernsutivorakul.25 ................................................................................................... 12

Figure 1.13 Freeform reflector embedded microfluidic lab-on-chip demonstrated by

D. De. Coster et al.27 ................................................................................................. 12

Figure 1.14 Schematics of different types of on-chip variable lens systems. (A)

Pneumatically tuned lens. (B) Lens geometry controlled by electrowetting induced


218

changes. (C) Hydrodynamic or electrokinetic lens. (D) Environmentally responsive

lens. (ITO) indium tine oxide. The dashed lines refer to the limits of physical tunability.

Presented by K. Bates el al.26 .................................................................................... 13 Figure 1.15 Schematic of a PDMS-based microfluidic channel for the analysis of

cyanide water pollutant with silver colloids presented by K. Yea et al.29 ................. 14

Figure 1.16 Schematics of tip coated multimode fibers (TCMMF) as Raman probe

presented by (left) C. Shi et al.36 and (right) M. Fan et al.,37 respectively. ............... 14 Figure 1.17 Cost and volume comparison for common lab-on-chip fabrication

technologies.68 (POCT: point-of-care testing, µPAD: micro paper-based analytical

device.) ..................................................................................................................... 18 Figure 1.18 Illustrations of the typical fabrication techniques: (a) micro-milling, (b)

UV lithography, and (c) X-ray lithography.69 ........................................................... 18 Figure 1.19 Schematic illustrations of (left) hot embossing and (right) injection

molding.70 ................................................................................................................. 19

Figure 1.20 Main processes of soft lithography for SERS sensor fabrication, presented

by S.Z. Oo et al.71 ..................................................................................................... 19

Figure 1.21 Fabrication process of a microfluidic SERS lab-on-chip by R2R approach,

presented by A. Habermehl el al.33 ........................................................................... 20

Figure 1.22 Fabrication procedure of a 3D microfluidic SERS lab-on-chip by

femtosecond-laser direct writing, demonstrated by S. Bai et al.32 ............................ 21

Figure 1.23 Schematics of (left) femtosecond laser writing and (right) femtosecond

laser assisted wet etching presented by A. Scott et al.65 ........................................... 21 Figure 2.1 The propagation of radiation.2 ................................................................. 33

Figure 2.2 Ball and spring model of a diatomic molecule. ....................................... 35 Figure 2.3 Potential well of a diatomic harmonic oscillator (dashed red) and the Morse

potential (solid blue) that better approximate the vibrational energy of an actual

molecule. (Figure adapted from en.wikipedia.org/wiki/Morse_potential) ................ 36

Figure 2.4 A molecule interacting with an incident photon results in Rayleigh

scattering, Raman Stokes and Anti-Stokes scattering. .............................................. 39 Figure 2.5 Simplified Jablonski diagram of different transition processes occurring

upon the interaction of an incident photon with a molecule. ‘Virtual energy states’ do

not exist in reality, but they are used to describe the Raman scattering. In reality the

electron falls back to the ground instantaneously. .................................................... 39 Figure 2.6 Schematic diagram of different vibrational modes of a -CH2 group. The

yellow ball represents the carbon atom and the blue ball represents the hydrogen atom

(Figure adapted from https://en.wikipedia.org/wiki/Molecular_vibration) .............. 41 Figure 2.7 Scheme of the ring breathing vibration. .................................................. 41 Figure 2.8 Schematics of out of plane bending. The yellow, green and blue balls

represent different types of atoms. ............................................................................ 42


219

Figure 2.9 Stretching vibration of a diatomic molecule and the related fluctuating

dipole moment.3 ........................................................................................................ 44 Figure 2.10 Changes of the dipole moment or the polarizability of the CO2 molecule

during vibrations. At the symmetrical stretching mode v1, there is a change in the

polarizability but no change in the dipole moment, the molecule is Raman active and

IR inactive. At the asymmetrical stretching mode v2, there is a change in dipole

moment, so the molecule is IR active but Raman inactive. At the deformation modes

v3 and v4, there are only changes in dipole moment, so the molecule is Raman inactive.

.................................................................................................................................. 45

Figure 2.11 Complete Raman spectrum of carbon tetrachloride (CCl4) excited by a

488nm wavelength laser.17 ........................................................................................ 46

Figure 2.12 Jablonski diagram showing the origin of Stokes Raman and Anti-Stokes

Raman lines in the Raman spectrum of CCl4. ........................................................... 46 Figure 2.13 Schematic of a dispersive Raman spectroscopy setup.31 ....................... 50

Figure 2.14 (a) An Au nanoparticle acts as an antenna by excitation of a dipolar

localized surface plasmon resonance (LSPR).40 (b) FDTD simulated Electromagnetic

field enhanced of an Au nanoparticle due to the LSPR, excited by 785nm wavelength

radiation. ................................................................................................................... 54

Figure 2.15 The normalized EF of a nanosphere drops fast down when the distance of

the molecule from the surface increases. The legends refer to the radius of the

nanosphere. ............................................................................................................... 56 Figure 2.16 FDTD simulated electric fields of an Au nano-pyramid, cube and

nanospheres in an external EM radiation at 785nm wavelength. .............................. 57

Figure 2.17 (Left) The Scanning Tunneling Microscope (STM) image of a Pt electrode

roughened by controlled a ORC potential, (Right) the AFM image of a Rh electrode

roughened by a controlled ORC current.47 ................................................................ 59

Figure 2.18 (A) SEM image of silver nanowires, (B) Transmission Electron

Microscopy (TEM) image of an individual Ag nanowire, (C) SEM image of silver

triangular nanoplates, (D) TEM image of silver triangular nanoplates, (E) SEM image

of Ag nanoparticles, (F) TEM image of Ag nanoparticles. 52 ................................... 60 Figure 2.19 TEM images of (A) AuNPs and (B) MIP-AuNPs.55 ............................. 61

Figure 2.20 TEM images of Fe3O4@C@Ag nanospheres with different sizes.57 ..... 61

Figure 2.21 Ag island films with mass thicknesses of 5nm, 10nm, 15nm and 20nm on

crystalline silicon substrates. Each image shows an area of 1µm × 1µm.62 .............. 62 Figure 2.22 SEM images of different Ag island films on a stainless steel substrate at

potential CV rates of 50mV/s, 100mV/s, 150mV/s and 200mV/s, respectively. 64 The

scale bar is 1µm. ....................................................................................................... 62 Figure 2.23 (a) Scheme of the nanoshell array fabrication and (b) SEM image of the

nanoshell array. 65 ..................................................................................................... 63


220

Figure 2.24 (a) AFM of the P4VP arrays, and SEM images of AuNP arrays on the

quaternized PS-b-P4VP films after overgrowth time of (b) 0 minutes, (c) 1 minute, (d)

3 minute. And (e) the scheme of AuNP array fabrication. (i) Solvent annealing. (ii)

Quaternization, (iii) Colloid adsorption, (iv) Overgrowth. (Blue: substrate, green:

P4VP, yellow: Polystyrene; red: Au). 67 ................................................................... 63

Figure 2.25 SEM images of the AgFON SERS substrate from (A) side and (B) top-

down. ........................................................................................................................ 64 Figure 2.26 (a) AFM image of the nanotriangle array; (b) Raman spectrum

benzenethiol from the nanotriangle arrays enhancement.72 ...................................... 64

Figure 2.27 SEM image of nanostructures on the fiber tip with (A) CPA, (B, C) SA,

(D) CPA-SR and (E) SA-SR patterns using 1µm diameter PS nanospheres. Scale bars

are 3µm.73 ................................................................................................................. 65 Figure 2.28 (a) Scheme of the fabrication process of polymer-based nanogratings as

SERS substrate; (1) spin coating and E-beam lithography, (2) Atomic Layer

Deposition (ALD), (3) Sputtering. (b) SEM image of the nanogratings. .................. 65 Figure 2.29 Nanopillar arrays fabricated by (a) nano-imprinting lithography76 and (b)

chemical etching.77 ................................................................................................... 66

Figure 2.30 SEM image of 2PP printed (a) nanocorral complex, (b) nanocorral array,

and (c) nanopyramid. ................................................................................................ 66 Figure 2.31 Number of publications on Raman OR Surface Enhanced Raman

spectroscopy (SERS) per year from 1999 to 2019. (Statistic data extracted from the

Web of Science, accessed on April 7, 2020) ............................................................. 67 Figure 3.1 Scheme of the two-photon polymerization system we used for

manufacturing nanostructures. (AOM: acousto-optical modulator). ........................ 81 Figure 3.2. (a) Drawing of a nominal shape model, and (b, c, d) electric field

distribution of the 200nm, 400nm and 600nm pillar arrays using the nominal shape

model and simulated by the FDTD method. ............................................................. 84

Figure 3.3. (a) Drawing of a voxel-based model, and (b, c, d) electric field distribution

of the 200nm, 400nm and 600nm pillar arrays using voxel-based model and simulated

by the FDTD method. ............................................................................................... 85 Figure 3.4. Morphologies of 200nm, 400nm and 600nm nano-pillar arrays. Measured

with SEM (column 1 and 2) and AFM (column 3). .................................................. 85

Figure 3.5 (a) Drawing of a nominal shape model, and (b, c, d) electric field

distribution of the 200nm, 400nm and 600nm pillar arrays using the fabricated model

and simulated by the FDTD method. ........................................................................ 87 Figure 3.6. Electric fields of 400nm pillar array in voxel-based model without (a) and

with fabrication errors (b-i). 50nm and 100nm closer pitch (b, c) increase the electric

field. 50nm height error (d) and 100nm height error (e) of one pillar, and all pillars

with 50nm height error (f) have little impact on the electric field. Diameter errors due


221

to displacement of different adjacent layers (g-i) have big impact to the relocation and

intensity change of electric fields, as these also reduce the pitch values. ................. 88 Figure 3.7. (a) Raman background of the 200nm pillar array; (b) Average Raman

spectra of 10µM RhB obtained with 200nm pillar arrays. Baselines are corrected by

subtracting the background Raman signal. Spectra are obtained under a 785nm

wavelength and 2.5mW laser excitation with 1 second integration time. The dashed

lines refer to the upper and lower range of the spectra. Average spectrum is obtained

over 16 measurements. ............................................................................................. 89

Figure 3.8. Comparison of enhancement factors for the different models obtained by

FDTD simulations and the experimental result. The average and standard deviation of

each enhancement factor are obtained over 16 measurements on a different detection

area of a nano-pillar array. ........................................................................................ 91 Figure 3.9. Comparison of 10µM RhB Raman spectra on different substrates and the

Raman spectrum of pure RhB. .................................................................................. 91

Figure 3.10. SEM images of 200nm, 400nm and 600nm hemisphere arrays (a-c) and

nano-grids with 200nm, 400nm and 600nm spacing (d-f). ....................................... 92

Figure 3.11. Calibration curves of the 200nm nano-pillar array showing the peak

intensities at 629cm-1, 1287cm-1 and 1363cm-1 with respect to the RhB concentrations

(a). Molecules can easily adsorb on the SERS substrate in the low concentration

condition (b), but a thick layer of molecules with high concentration will restrain the

SERS signal (c). The RhB molecules cover both the nanostructures and the

surrounding flat surfaces, for which the interaction of molecules with nanostructures

has changed the color of the RhB from red violet to green under white light

illumination due to a higher reflectivity at the green light band. .............................. 93 Figure 3.12. Raman spectra of 1ppm DON and 1.25ppm FUM obtained with our

200nm pillar array SERS substrate under a 785nm wavelength and 2.5mW laser

excitation with 1 second integration time. The concentrations of mycotoxins are close

to the detection limit of our SERS substrates and therefore it is difficult to recognize

the mycotoxins directly. ............................................................................................ 94 Figure 3.13. (a) Spectra of DON and FUM are clustered by the scores of the second

and third principal components. (b)-(c) Coefficients of principal components 1-3

which represent 85.9% of the variance. .................................................................... 95

Figure 4.1 Illustration of confocal principle.4 ......................................................... 104 Figure 4.2 (Left) schematic of the confocal Raman spectroscopy setup with a freeform

reflector embedded LoC. (Right) Geometrical parameters of the LoC.4 ................ 105 Figure 4.3 ASAP simulation of the (a) excitation and (b) generation of Raman

scattering in the microfluidic channel.4 .................................................................. 106 Figure 4.4 (a) Schematic drawing and (b) photograph of the LoC with integrated

freeform reflector. 4 ................................................................................................ 106


222

Figure 4.5 Photographs of the confocal Raman spectroscopy setup. (a) side view; (b)

front view.4 ............................................................................................................. 107 Figure 4.6 Layouts of various fiber optic-based Raman probes10. .......................... 109 Figure 4.7 (A) Schematic of a Raman endoscopy setup for in vivo diagnostics. (B)

Photograph of the Raman probe. (C) The Raman probe used during clinical

examination.14 ......................................................................................................... 109 Figure 4.8 Cross-sections of a fiber-based Raman probe with a ball lens to enhance

the collection efficiency.16 ...................................................................................... 110

Figure 4.9 Schematic of a Raman probe design for epithelial tissue detection. EP:

Epithelium; ST: Stroma.19 ....................................................................................... 110 Figure 4.10 Layout of a handheld Raman probe.21 ................................................. 111

Figure 4.11 (a) Collection-prioritized configuration and (b) excitation-prioritized

configuration of the Raman probe for the freeform reflector-based microfluidic chip.

The red and blue rays refer to the excitation and collection path respectively. ...... 112

Figure 4.12 (a) The excitation path and (b) collection path of the excitation-prioritized

configuration using non-sequential ray tracing simulations. .................................. 116

Figure 4.13 (a) The excitation path and (b) collection path of the collection-prioritized

configuration using non-sequential ray tracing simulations. .................................. 116

Figure 4.14 Optimal layout of the integrated Raman probe for the microfluidic lab-

on-chip, inside the dashed oval is the reflector based LoC. .................................... 117

Figure 4.15 Different points of view of the “collection-prioritized configuration” of

the Raman probe realized, the proof-of-concept demonstration was built in a Thorlabs

cage system. Lab-on-chip support and syringe for filling fluids in the microchannel of

the chip are highlighted by the orange and green rectangles, respectively. ............ 119 Figure 4.16 Spectra of different fluidic samples. Ethanol and methanol can be

distinguished due to their different Raman peaks. .................................................. 120

Figure 4.17 Micro-cracks on the surface of the polymer which was in contact with a

fluidic sample for a period of time. ......................................................................... 122 Figure 4.18 Scheme (left) and photograph (right) of our optimized PMMA chip. . 122 Figure 4.19 Molecular structure of Topas COC. .................................................... 123 Figure 4.20 Copolymer composition of COC and the heat resistance displayed by its

glass transition temperature.29 ................................................................................. 124

Figure 4.21 Transmission of TOPAS 6013 and PMMA. (Measured with an AvaSpec-

UV/VIS/NIR spectrometer for plates with 2mm thickness) ................................... 125 Figure 4.22 Geometry of the freeform reflector. .................................................... 126 Figure 4.23 Profiles of the PMMA- and COC-based freeform reflectors. .............. 126 Figure 4.24 Polyfitting results for the PMMA-based G1 chip. ............................... 127 Figure 4.25 Polyfitting results for the COC-based G2 chip. ................................... 127

Figure 4.26 Polyfitting results for the COC-based G3 chip. ................................... 127


223

Figure 4.27 TIS of a reflective surface for 785nm wavelength light with different

incident angles. ....................................................................................................... 128 Figure 4.28 Thickness tolerance of the bottom thickness. The nominal bottom

thickness of the 50µm focus design (G2) is 200µm, while the nominal bottom

thickness of the 100µm focus design (G3) is 250µm. ............................................ 129

Figure 4.29 Drawing of the COC-based 50µm focus G2 chip with integrated freeform

reflector. The channel/reflector layer of the chip is fabricated via double-sided hot

embossing. .............................................................................................................. 130

Figure 4.30 Dimensions of the COC-based 50µm G2 chip. ................................... 131

Figure 4.31 Photographs of ultra-precision diamond tooling for the reflector mold

fabrication. (left: before tooling; middle: during tooling; right: after tooling) ....... 131

Figure 4.32 Photograph of micro-milling for the channel mold fabrication. .......... 132 Figure 4.33 Drawings and photographs of the reflector mold (left) and channel mold

(right). ..................................................................................................................... 132

Figure 4.34 (a) Schematic and (b) photograph of double-sided hot embossing for the

chip replication. ...................................................................................................... 133

Figure 4.35 Influence of the sheet thickness. The raw COC plate is cut into 30mm by

30mm pieces with a milling machine. The thickness of the COC plate is measured

before hot embossing, and the thickness of the chip is measured after hot embossing.

In general, a thicker COC plate leads to a thicker chip thickness. .......................... 134

Figure 4.36 COC polymer sheet and the hot embossed reflector/channel layer. The

size of the original sheet is 30mm × 30mm × 1mm. The diameter of the

reflector/channel layer is 25.4mm........................................................................... 134

Figure 4.37 Photographs of the microfluidic lab-on-chip with integrated freeform

reflector. The hot embossed circular reflector/channel layer and the squared sealing

layer are bonded by UV curing adhesive. ............................................................... 135

Figure 4.38 Surface roughness of the reflector and the channel mold before and after

hot embossing. The RMS roughness of the reflector mold increased from (20.5 ± 5.3)

nm to (168.1 ± 23.8) nm, and the RMS roughness of the channel mold increased from

(6.4 ± 1.3) nm to (67.9 ± 16.2) nm, respectively. ................................................... 135 Figure 4.39 (a) General view and (b) cross-section of the COC-based G3 LoC. ... 137

Figure 4.40 Full fabrication flow for the COC-based 100µm focus G3 LoCs. ....... 138

Figure 4.41 Photographs of the (a) reflector mold and (b) channel mold. The mold

surface is coated with NiP material. ....................................................................... 138 Figure 4.42 (a) General view and (b) dimensions of the physical stopper to control the

bottom thickness of the LoC during hot embossing. The thickness of each shim is

250µm. .................................................................................................................... 139 Figure 4.43 Photographs of the fabricated COC-based 100µm focus G3 chip. ...... 139

Figure 4.44 (a) Scheme of UV/thermal bonding and (b) laser welding. ................. 140


224

Figure 4.45 Shear stresses of the samples bonded with different methods. (LW: laser

welding; UV: UV curing adhesive; TB: Thermal bonding; legend: with and without

plasma treatment).................................................................................................... 141 Figure 4.46 Proof-of-concept Raman setup by combining the Raman probe with the

mass fabricated LoC. (BPF: Band-pass filter, M: Mirror, DM: Dichroic mirror, LPF:

Long-pass filter, NF: Notch filter, CL: Collimating lens, MMF: Multi-mode fiber.)

................................................................................................................................ 142 Figure 4.47 Raman spectra of ethanol obtained by the mass fabricated G2 and G3

chips. The dashed lines refer to the upper and lower range of the spectra. Figures are

on the same scale. ................................................................................................... 143 Figure 4.48 Raman spectra of urea solutions with different concentrations. An internal

standard of 100mM KNO3 is added to the solution to normalize the spectra. ........ 144 Figure 4.49 Calibration curve of our proof-of-concept Raman setup using our mass-

fabricated G3 chips in combination with our Raman probe. .................................. 145

Figure 4.50 PCA results of the fumonisin B1 Raman spectra. (a) Average spectra of

solvent and fumonisin b1 with different concentration levels. (b) loadings of different

principal components. (c) Scores of PC1 and PC2. ................................................ 146

Figure 4.51 PCA results of the deoxynivalenol Raman spectra. (a) Average spectra of

solvent and deoxynivalenol with different concentration levels. (b) loadings of

different principal components. (c) Scores of PC1 and PC2. ................................. 147

Figure 5.1 Illustration of (a) our freeform reflector embedded lab-on-chip (NA=1.28)

discussed in Chapter 4 and (b) the segmented reflector-based Raman system with

SERS microfluidic chip (NA=1.15). The red and blue rays refer to the incident light

that interact with different segments of the reflector. (S1: center segment; S2: middle

concave segment; S3: marginal segment.) .............................................................. 157

Figure 5.2 Surface profile of the concave segment, the parallel rays are reflected

through the bottom layer of the microfluidic chip into the channel layer. .............. 158

Figure 5.3 Design schematic of all segments: center, concave and marginal: (a)

general view, (b) zoom-in view of the rays in the microfluidic channel and (c) zoom-

in view of the rays interacting with the center segment. ......................................... 160 Figure 5.4 Two configurations of reflector and microfluidic chip design, the diameter

of the center segment for the (a) and (b) configurations are 5mm and 10mm,

respectively, and the overall diameters are 30mm and 38mm. NA of both

configurations is 1.15. Moreover, the NA of the center segment is 0.34 and 0.45

respectively. The blue and red rays refer to the incident light that interact with

different segments. .................................................................................................. 162 Figure 5.5 Non-sequential simulation in ZEMAX OpticStudio: (a) Overall layout, and

(b) the microfluidic chip and segmented reflector. The blue rays and green rays refer

to the excitation light and Raman light respectively. (EL: Excitation lens; CL:


225

Collection lens; BP: Band-pass filter; M: Mirror; LP: Long-pass filter; DM: Dichroic

mirror.) .................................................................................................................... 164 Figure 5.6 Confocal behavior of the system by simulation. Each subfigure shows the

efficiency changes with respect to the displacement of the point source in (a) x-, (b)

y- and (c) z-axis, respectively. 100µm, 200µm and 400µm MMF are used as collection

fiber. ........................................................................................................................ 165 Figure 5.7 NiP surface of segmented reflector by ultra-precision diamond turning (a)

and after gold coating (b). The reflectivity increases from 53% to 95% for NIR light

when adding the Au coating. .................................................................................. 168

Figure 5.8 Proof-of-concept demonstration setup of our Raman spectroscopy system

(left): (a) the external optics, (b) Microfluidic chip and segmented reflector, (c) Raman

probe, (d) beam expander; Microfluidic chip and segmented reflector in detail (right).

................................................................................................................................ 169 Figure 5.9 The fabrication process of SERS substrate: (a–b) Maskless reactive ion

etching forms silicon nanopillars; (c–d) Electron beam evaporation forms Au layers;

(e) SEM image of the nanopillars with coatings35. ................................................. 169

Figure 5.10 (a) Microfluidic chip integrated with SERS substrate to enhance the

sensitivity of the Raman analysis. (b) Cross-section of the SERS microfluidic Chip.

................................................................................................................................ 170 Figure 5.11 (a) Raman spectra of ethanol and PMMA measured by commercial

Raman microscope (Bruker Senterra, Department of Chemistry of Ghent University),

and (b) Raman spectrum of ethanol in the PMMA chip obtained with our setup using

a 200µm diameter MMF fiber. The PMMA background is suppressed with a factor of

6. (figures are on the same scale.) ........................................................................... 171 Figure 5.12 Comparison of the suppression factors obtained by simulation and

experiments. The average and standard deviation of the experimental suppression

factor for each MMF is calculated over 15 measurements. .................................... 172

Figure 5.13 Raman efficiency of our setup with respect to misalignments of the MMF

along (a) x-, (b) y-, (c) z-axis, respectively. 100, 200 and 400µm MMFs are used. The

solid lines refer to the simulation results, and the dashed lines refer to the experimental

results. ..................................................................................................................... 173

Figure 5.14 Raman spectra of aqueous urea (U) and KNO3 (P) solution with a

concentration from 15 to 300mM. The baseline is corrected by subtracting the

reference spectrum of water. Each spectrum is obtained over 10 measurements. .. 174 Figure 5.15 (a) Calibration curves and (b) SNR curves of aqueous urea and potassium

nitrate solutions. The average and standard deviation for each solution is obtained

over 10 measurements. According to the SNR curves, the NEC of our setup for urea

and KNO3 are 19mM and 18mM, respectively. ...................................................... 174


226

Figure 5.16 (a) SERS spectra of Rhodamine B, urea, potassium nitrate and water

sample obtained with our microfluidic chip in combination with maskless ion etched

SERS substrates, (b) the baseline corrected SERS spectra of Rhodamine B, urea and

potassium nitrate by subtracting water as reference. Each spectrum is obtained over

10 measurements. ................................................................................................... 176

Figure 6.1 (a) Illustration of conventional SERS measurements with an objective lens;

(b) Principle of our freeform segmented reflector and the conical beam shaper; and (c)

the profile of the segmented reflector calculated by a numerical approach. (S1: center

segment; S2: middle segment; S3; outer segment.) ................................................ 185

Figure 6.2 Comparison of the profile of the segmented reflector obtained by the

numerical approaches discussed in Chapter 5 (a) and Chapter 6 (b), respectively. The

overall radius of both reflectors is 10.6mm. However, the sag of the new segmented

reflector design is 2.97mm, which is much less than 6.21mm obtained by the previous

numerical approach. ................................................................................................ 188

Figure 6.3 Two components (a, b) of the conical beam shaper made from PMMA

material with localized gold coating. (c) Integration of (a) and (b) into a beam shaper

and (d) the freeform segmented reflector made from brass. All reflective surfaces are

fabricated by ultra-precision diamond tooling and coated with gold layers by

sputtering. ............................................................................................................... 189 Figure 6.4 Scheme of the Raman spectroscopy system in non-sequential ray tracing

simulation (a), and the details of the segmented reflector, beam shaper and the SERS

substrate (b). The blue and green lines refer to the excitation and Raman rays

respectively. Our system also contains a Raman probe (the red dashed box) and a

mirror as external optics. ........................................................................................ 190 Figure 6.5 Normalized Raman intensity change (a) with respect to the displacement

of the point source from focal plane along the z-axis (b, c). The red and blue rays refer

to different sections of the Raman scattered light interacting with different segments

that change to different directions when the Raman source is moved along the z-axis.

................................................................................................................................ 192 Figure 6.6 Structural frame CAD of our Raman spectroscopy (a), and the proof-of-

concept demonstration setup of our Raman spectroscopy system (b). (BPF: Band-pass

filter; LPF: Long-pass filter.) .................................................................................. 193

Figure 6.7 Beam shape before (a, b) and after (c, d) the conical beam shaper in the

simulations. ............................................................................................................. 194 Figure 6.8 Beam shape before (a, b) and after (c, d) the conical beam shaper

experimentally obtained by the CMOS sensor. ...................................................... 194 Figure 6.9 SERS spectra of 10µM Rhodamine B measured with our conical beam

shaper in combination with the freeform segmented reflector, and with a 60×

commercial microscope objective lens. Figures are on the same scale................... 195


227

Figure 6.10 Comparison of misalignment tolerances of the collection fiber in

simulation and experiments. Intensities are normalized. The average and standard

deviation of each displacement are obtained over 10 measurements. ..................... 196 Figure 6.11 Misalignment tolerance of the beam shaper with respect to the segmented

reflector. Intensities are normalized. The average and standard deviation of each

displacement are obtained over 10 measurements. The FWHM tolerances are around

0.225mm both in simulation and experiments. ....................................................... 197 Figure 7.1 Nanostructure replication with 2PP printed master molds. ................... 207

Figure 7.2 Towards a monolithic optofluidic lab-on-chip for SERS analysis. The

monolithic part can be directly fabricated by 2PP with a reflector shape and

nanostructures in the center. The reflective coating layer for the reflector and the metal

surface for SERS can be added by local thermal evaporation. ............................... 208 Figure 7.3 Scanning grating MEMS chip ((a) face up and (b) face down); and rendered

images of the micro-spectrometer ((c) and (d)). The dimensions of the MEMS chip

are 9.6mm × 5.3mm × 0.5mm. The tiltable diffraction grating of this MEMS chip is

electrostatically driven.14 ........................................................................................ 209

FREEFORM-BASED OPTOFLUIDIC DEVICES TOWARDS …

Documents

Transcript of FREEFORM-BASED OPTOFLUIDIC DEVICES TOWARDS …