• No results found

An Overview of Wide-field Raman Imaging and Its Applications

N/A
N/A
Protected

Academic year: 2021

Share "An Overview of Wide-field Raman Imaging and Its Applications"

Copied!
90
0
0

Bezig met laden.... (Bekijk nu de volledige tekst)

Hele tekst

(1)

MSc Chemistry

Analytical Science

Literature Thesis

An Overview of Wide-field Raman Imaging and Its

Applications

Supervisor/ Examiner

Dr. Freek Ariese

Second reviewer

Prof. Dr. Govert W. Somsen

Qianru Zuo

11692464 UvA /2629666 VU

12 ECTS

March 2020 - May 2020

(2)

1

Wide-field Raman Imaging (WRI) is a powerful technique for generating detailed chemical images. It combines Raman spectroscopy and Raman imaging techniques and can provide spatial distribution for one or a few specified chemical components in the sample. Features like molecular specificity, non-destructive and non-labeling testing, and the capability of analyzing wet samples make WRI a prospective alternative method to fluorescence, IR, and X-ray imaging in some application fields. Recent advancements in WRI instrumental setups facilitated the application in many areas; the raising applications and market demands, in turn, pushed the development of instruments. In this context, this study provides an overview of WRI with a special focus on recent advances. Firstly, the theoretical background of the Raman effect and several forms of Raman spectroscopy, such as SERS (Surface-Enhanced Raman Spectroscopy), Resonance Raman Spectroscopy (RRS), and Coherent anti-Stokes Raman Spectroscopy (CARS) are provided in Chapter 2. WRI and its instrumental setup are then introduced in Chapter 3. Furthermore, Chapter 4 describes the application of WRI in some presentative areas, including material analysis, biology, biomedicine, pharmaceutical, food, threat detection, and forensic science.

(3)

2

Abbreviation Meaning

2D Two-Dimensional

2D-hBN Two-Dimensional hexagonal Boron Nitride

3D Three-Dimensional

ACG AgCu@Graphene

AN ammonium nitrate

AOI Angle Of Incidence

AOTF Acousto-Optical Tunable Filter APIs Active Pharmaceutical Ingredients

Bg Bacillus globigii

BuPTCD (n-butylimido) perylene

CARS Coherent Anti-Stokes Raman Scattering

CCD Charge-Coupled Device

CHO Chinese Hamster Ovary

CNTs Carbon NanoTubes

DNT 2,4-Dinitrotoluene

DPI Dry Powder Inhaler

FAST Fiber Array Spectral Translator

FCC Face-Centered Cubic

FWHM Full-Width Half Maximum

GC Gas Chromatography

Hb Hemoglobin

HO Heterotopic Ossification

HPDR13 Methacrylic Homopolymer of Disperse Red-13 HPLC High-Performance Liquid Chromatography

IR InfraRed

LB Langmuir-Blodgett

LCTF Liquid-Crystal Tunable Filter

(4)

3

LSDRS Light Sheet-excited Direct Raman Spectroscopy LT-LCTF Lyot-type Liquid Crystal Tunable Filter

LVLWP Linear Variable Long-Wave Pass filter LVSWP Linear Variable Short-Wave Pass filter LVTF Linear-variable tunable filters

LWP Long-Wave-Pass

MCF Multi-Conjugate Filter

MoS2 Molybdenum disulfide

MRI Magnetic Resonance Imaging

MS Mass Spectrometry

NA Numerical Aperture

NIR Near-InfraRed

NPs Nanoparticles

NM Not Mentioned

PETN PentaErythritol TetraNitrate

PS PolyStyrene

QA/QC Quality Assurance / Quality Control

RF Radio Frequency

RRS Resonance Raman Spectroscopy

SEM Scanning Electron Microscope

SERS Surface-Enhanced Raman Spectroscopy

SIL Solid Immersion Lenses

SWP Short-Wave-Pass

SR Spontaneous Raman

TF-TBPF Thin-Film Tunable bandpass filters

TNT 2,4,6-trinitrotoluene

UV UltraViolet

w/w weight by weight

(5)

4 Abstract ... 1 List of Abbreviations ... 2 Introduction ... 6 Fundamental Principles ... 8 2.1 Raman Spectroscopy ... 8

2.1.1 Concept of the Raman Effect ... 8

2.1.2 Theoretical Background ... 9

2.1.3 Instrument Setup ... 17

2.2 Other Forms of Raman Spectroscopy Techniques ... 18

2.2.1 Surface Enhanced Raman Spectroscopy (SERS)... 19

2.2.2 Resonance Raman Spectroscopy (RRS)... 19

2.2.3 Coherent Anti-Stokes Raman Spectroscopy (CARS) ... 20

Wide-field Raman Imaging ... 23

3.1 Raman Imaging ... 23

3.2 System Performance ... 26

3.2.1 Spectral Resolution ... 26

3.2.2 Spatial Resolution ... 27

3.3 Instrumentation of Wide-field Raman Imaging ... 28

3.3.1 Excitation Source ... 28

Contents

(6)

5

3.3.2 Wavelength Selection Device ... 29

3.3.3 Fiber Array-based Raman Imaging ... 42

Applications of Wide-field Raman Imaging ... 45

4.1 Materials and Polymers ... 45

4.2 Pharmaceutical ... 49

4.3 Biology and Biomedicine ... 52

4.3.1 Cells ... 53

4.3.2 Tissue ... 55

4.3.3 Bones ... 56

4.3.4 Small Animals ... 57

4.3.5 Disadvantages in Biomedical Applications ... 57

4.4 Threat Detection ... 58 4.5 Food Safety ... 60 4.6 Forensic science ... 62 4.7 Others... 62 Conclusion ... 64 Reference ... 67 ... 80

(7)

6

Wide-field Raman imaging (WRI) is a type of Raman imaging technology that captures the image of a sample by Raman scattering. It detects the presence of specific chemical components and features their locations in the samples simultaneously. The first implementation of WRI can be traced back to 1975 [1]. In this experiment, a tunable filter, a laser with a fixed wavelength, and a microscope were used to visualize chemical crystals. It was worth noting that they failed to use a tunable dye laser due to the insufficient filter performance at that time. During the past few decades, the developments of lasers, filters, and data processing algorithms have fueled numerous studies on the instrumentations and applications of WRI.

WRI has many promising features thanks to the properties of Raman scattering, such as non-destructive testing, molecular specificity, simple sample preparation, and the capability of analyzing wet samples. It also shows unique characteristics among Raman imaging techniques. In general, there are three types of Raman imaging: Point mapping, line scanning, and wild-field imaging. Point mapping scans the sample point by point with a tightly focused laser beam and reconstructs the image pixel by pixel afterward. Line scanning is another scanning method where a line-shaped laser beam scans the samples. Both scanning methods record the full spectra and enable high spectral resolution; however, the whole procedure is slow and often causes laser-induced damage to the sample. On the contrary, WRI illuminates the sample entirely with a defocused laser source, and the two-Dimensional (2D) image is recorded at once for a specific Raman wavenumber. As a result, wide-field imaging offers

(8)

7

rapid and even real-time Raman imaging, which facilitates the application in many fields such as biology, biomedicine, pharmaceutical, threat detection, food safety, etc.

Although WRI shows great opportunities in the above applications fields, there are few published reviews on its instrumentations and applications so far (accessed on ninth of May 2020). Most reviews are about general Raman imaging in which applications of wide-field imaging, point mapping, and line scanning are mixed instead of specified [2]–[5]. The same situation occurs for their instrumentations.

This literature review is an introduction to WRI, the aim of which is to provide information for readers who want to utilize it in their experiments. Since the best way to apply any technique to a new sample is to look at similar existing applications, this review mainly discusses the instrumental developments and applications of the WRI technique. The fundamental principles of the Raman effect and Raman techniques that were combined with wide-field imaging are provided as background information in Chapter 2. The recent advancements in instrumentation are described and compared in Chapter 3. The application fields highlighted in Chapter 4 are material characterization, biomedical, pharmaceutical, threat detection, and food safety. Moreover, the advantages and limitations are discussed; some other areas are also briefly reviewed. The typical applications, as well as their instrumental setup, are summarized in the Appendix due to the page limitation. Finally, we conclude this paper with the discussion of WRI instruments and applications in Chapter 5.

(9)

8

This chapter describes the fundamental principles of Raman spectroscopy (RS). In the first part of this chapter, the phenomenon of the Raman effect, essential theoretical background, as well as a simple setup used to measure Raman scattered light are discussed. The first part of this chapter only focuses on the conventional Raman spectroscopy that is nonresonant and spontaneous; the second part of this chapter briefly introduces other modes of Raman spectroscopy.

2.1 Raman Spectroscopy

2.1.1 Concept of the Raman Effect

The Raman effect is produced by the inelastic scattering of photons that hit the molecules. When a beam of monochromatic light shines on a substance, some of the light is transmitted, some is absorbed, and other light is scattered. Most of the scattered light, called Rayleigh scattering, has the same wavelength as the incident light. A small fraction (about 1 out of 106

to 108) has a different wavelength to the incident light [6]. The scattered light that is shifted

in wavelength is known as Raman scattering, and the corresponding spectrum is called the Raman spectrum. The Raman effect can also be observed in daily life with the help of filters: when water is exposed to green laser light (532 nm), it looks green because the majority of the scattered light shares the same color as the incident light. However, when a filter which blocks the green light is applied, red Raman scattering is revealed (around 650 nm). The theories about the Raman effect are discussed in the following sections.

(10)

9

2.1.2 Theoretical Background

Classical Theory

In classical theory, light is electromagnetic radiation consisting of oscillating electromagnetic waves. When a molecule is illuminated by light, it is in an electric field E, which is given by Equation 2-1.

𝑬 = 𝐸0cos(2𝜋𝜈𝑖𝑡) Equation 2-1

where E0 represents the magnitude of the electric field, and νi is the frequency of the incident

light. The interaction of the molecule and the incident electromagnetic wave generates an induced molecular dipole moment. This oscillating dipole moment deforms the molecule periodically and emits radiation (scattered light). The induced dipole moment is in a linear relationship with the polarizability of the molecule and the incident electric field, which is given by Equation 2-2.

𝑷 = 𝜶𝑬 Equation 2-2

Where P refers to the induced dipole moment, E refers to the incident electric field, and α refers to the polarizability of a molecule. The polarizability describes how easily the electric density distribution of a molecule can be deformed in an electric field. Equation 2-2 can be expressed as a matrix using the Cartesian coordinate system. The matrix is shown in Equation 2-3. [ 𝑃𝑥 𝑃𝑦 𝑃𝑧 ] = [ 𝛼𝑥𝑥 𝛼𝑥𝑦 𝛼𝑥𝑧 𝛼𝑦𝑥 𝛼𝑦𝑦 𝛼𝑦𝑧 𝛼𝑧𝑥 𝛼𝑧𝑦 𝛼𝑧𝑧 ] [ 𝐸𝑥 𝐸𝑦 𝐸𝑧 ] Equation 2-3

(11)

10

When the molecule vibrates, the polarizability is changed and can be expressed as Taylor expansion in Equation 2-4, which includes the equilibrium polarizability and a small component that is caused by the displacement q from the equilibrium position.

𝛼(𝑞) = 𝛼0+𝜕𝛼

𝜕𝑞∙ 𝑞 + ⋯ Equation 2-4

Where 𝛼(𝑞) refers to the changed polarizability of a molecule with the displacement q, 𝛼0

represents the polarizability in the equilibrium position.

q is related to the corresponding vibration of the molecule and is given by Equation 2-5. 𝑞 = 𝑞0cos(2𝜋𝜈𝑣𝑡) Equation 2-5

Where 𝑞0 is the maximum value of q and 𝜈𝑣 is the vibrational frequency of the molecule.

By replacing q in Equation 2-4 with Equation 2-5, we get Equation 2-6,

𝛼(𝑞) = 𝛼0+ 𝜕𝛼

𝜕𝑞∙ 𝑞0𝑐𝑜𝑠(2𝜋𝜈𝑣𝑡) Equation 2-6

By replacing 𝑬 in Equation 2-2 with Equation 2-1, we have Equation 2-7

𝑃 = 𝛼𝐸0𝑐𝑜𝑠(2𝜋𝜈𝑖𝑡) Equation 2-7

Inserting Equation 2-6 into Equation 2-7, we get Equation 2-8

𝑃 = 𝛼0𝐸0cos(2 𝜋𝜈𝑖𝑡) + 𝜕𝛼

𝜕𝑞∙ 𝑞0cos(2𝜋𝜈𝑣𝑡) 𝐸0cos(2𝜋𝜈𝑖𝑡) Equation 2-8

Equation 2-8 can be reformulated as the following equation (Equation 2-9)

𝑃 = 𝛼0𝐸0cos(2 𝜋𝜈𝑖𝑡) +1

2 𝜕𝛼

𝜕𝑞∙ 𝑞0𝐸0[cos(2𝜋𝑡(𝜈𝑣 + 𝜈𝑖)) cos(2𝜋𝑡(𝜈𝑣− 𝜈𝑖))]

(12)

11

Where the first part describes the induced dipole moment that causes Rayleigh scattering with the same frequency as the incident light; the second part describes the dipole moment that produces Stokes Raman scattering with the frequency of (𝜈𝑣− 𝜈𝑖) and the anti-Stokes Raman scattering with the frequency of (𝜈𝑣+ 𝜈𝑖).

The intensity of the emitted radiation is proportional to the square of the second time derivative of the induced dipole moment. Therefore, the intensity of Rayleigh and Raman scattering can be deduced by Equation 2-9, shown in the following Equations, where 𝐼𝑖 is the intensity of incident light.

𝐼𝑅𝑎𝑦𝑙𝑒𝑖𝑔ℎ∝ 𝜈𝑖4𝛼04𝐼 𝑖 Equation 2-10 𝐼𝑆𝑡𝑜𝑘𝑒𝑠 ∝ (𝜈𝑖 − 𝜈𝑣)4𝛼 04𝐼𝑖 Equation 2-11 𝐼𝐴𝑛𝑡𝑖−𝑆𝑡𝑜𝑘𝑒𝑠∝ (𝜈𝑖 + 𝜈𝑣)4𝛼 0 4𝐼 𝑖 Equation 2-12

According to equation 2-11 and 2-12, two Raman scattering intensities are predicted to be similar in classical theory; however, they are observed to be different in practice. Therefore, the classical theory fails to explain the different intensity between Stokes and anti-Stokes Raman scattering, which is explained by quantum theory.

Quantum theory

In quantum theory, light is treated as photons. Raman scattering is the photons that are inelastically scattered by the molecule. When a molecule absorbs the energy from the photon, it is excited to a virtual state. The virtual state is not a real energy level but an intermediate state, whose energy level is determined by the energy of the incident photon. The molecule then immediately falls back to its ground state in several ways. Here we only discuss three

(13)

12

cases, shown in Figure 2-1: Rayleigh scattering (in green), Stokes Raman scattering (in red), and anti-Stokes Raman scattering (in blue):

i. Rayleigh scattering: The molecule falls back to the initial energy level and emits a photon with the same energy as the incident photon.

ii. Stokes Raman scattering: The molecule falls back to anther vibrational level in the electronic ground state and emits a photon with reduced energy than the incident photon.

iii. Anti-Stokes Raman scattering: The molecule is on a higher vibrational level before the photon strikes it, however, falls to a lower vibrational level after it reaches the virtual state. In this case, a photon with higher energy than the incident photon is emitted.

Figure 2-1 Jablonski diagram and spectrum scheme illustrating Rayleigh, Stokes Raman scattering and anti-Stokes Raman scattering.

The energy levels are drawn in such a way that the lowest energy level is shown at the bottom with a series of increasing energy levels above it. Electronic and vibrational energy levels are drawn as solid line; Virtual states are drawn as dashed lines. S0 and S1 refer to the ground and excited states of a

molecule; Different vibrational energy levels are illustrated with the same initial “v”. The corresponding peaks in the spectrum are drawn under the Jablonski diagram, the position of the peaks is described in wavenumber(cm-1)

(14)

13

Here, we define the frequency of the incident light as 𝜈𝑖 and the energy between the first

vibrational energy level and the ground state as ∆𝐸. Based on Planck's Equation (Equation 2-13): the energy of the incident light is ℎ𝜈𝑖; the energy of the illustrated Rayleigh scattering is also ℎ𝜈𝑖; the corresponding energies of Stokes and anti-Stokes Raman scattering are ℎ𝜈𝑖 − ∆𝐸 and ℎ𝜈𝑖 + ∆𝐸, respectively.

𝐸 = ℎ𝜈 Equation 2-13 Where h is Planck's constant (6.62607015×10−34 J⋅s); v is the frequency in s-1.

As shown in the schematic diagram of the Raman spectra (Figure 2-1). The Raman spectra records Raman shifts in wavenumber to avoid the difference in spectra, which is caused by different excitation light. Equation 2-15 is the reform of Equation 2-14 using wavenumber. Wavenumber can be directly calculated from the wavelength by Equation 2-15, where λ0 is

the wavelength in cm in the vacuum.

𝐸 = ℎ𝑐𝜈̃ Equation 2-14

c is the speed of light in cm/s, and 𝜈̃ is wave number in cm-1.

𝜈̃ = 1/𝜆0 Equation 2-15

In quantum theory, the relative intensity of Stokes and anti-Stokes Raman scattering is determined by the relative abundance of the molecules that produce the two scatterings. Because the molecule population in different states follows Boltzmann's law, according to which most of the molecules are in the lowest vibrational energy level at room temperature. Thus, photons are more likely to be Stokes scattered, producing a stronger Stokes scattering than anti-Stokes Raman scattering.

(15)

14

Selection Rules

Raman selection rules define the types of transitions that are Raman active. From Equation 2-9, it can be concluded that the Raman effect only occurs when (𝜕𝛼

𝜕𝑞)0 ≠ 0. Accordingly, the

essential requirement for Raman scattering is the change in polarizability of the molecule. The change in polarizability mainly results from the vibrational motion of a molecule. Although rotational contribution can be observed in the gas phase, here, we only focus on the vibrational motion of a molecule in order to simplify the theory. Figure 2-2 shows the vibrational motions and corresponding change in polarizability of the molecule types A2, AB,

linear ABA, and non-linear ABA.

Almost all molecules of low symmetry are Raman active. In molecules that consist of several atoms, the number of degrees of freedom for vibrational motion is 3n-6, where n is the number of atoms. For a large molecule, many vibrational transitions may be Raman active, resulting in a large number ofRaman bands. However, the corresponding Raman spectrum is

Figure 2-2 Vibrational motions and corresponding Raman activity of four types of molecules.

The ball and stick models imitate the vibrational motions of different molecules; The variation of polarizability is illustrated by the functions in 2D normal coordinate; (𝜕𝛼𝜕𝑞)

0is the derivative of

(16)

15

relatively simple because the different vibrational motions are close in energy, causing the overlap of Raman bands [7].

Difference between Raman and Infrared

Raman spectroscopy has been regarded as complementary to Infrared (IR) spectroscopy for the similarity they share and the difference between them. Both techniques describe the vibrational structure of molecules. However, Raman and IR spectroscopy reflect different vibrational structures for a molecule due to different selection rules: The Raman scattering requires the change in polarizability, whereas the IR absorption demands for changes in dipole moment. Therefore, the highest intensity in Raman spectra is given by symmetric vibrations, while the highest intensity in IR is given by asymmetric vibrations. Besides, the excitation source is also different in two spectroscopies. In the measuring of IR spectra, the samples only absorb photons that have an exact frequency, which matches the energy difference between two vibrational energy levels. In contrast, samples can be excited to the virtual state by any photons that are of higher energy when measuring Raman scattering.

Fluorescence

Fluorescence is the dominant background in Raman spectroscopy arising from the sample molecules, contaminants in the sample, or from the container/environment of the sample.[8] Fluorescence occurs when the incident energy is high enough to excite a molecule to an upper electronic state. The process of fluorescence is illustrated in the Jablonski diagram in Figure 2-3: A molecule is excited to an upper electronic state by absorbing a photon. The excited molecule then falls to the lowest level of the excited electronic state by non-radiative internal conversion. Finally, the molecule returns back to the electronic ground state by emitting a

(17)

16

photon. The resulted fluorescence emission is of lower frequency than the incident light and usually overlaps with the range of Stokes Raman scattering. The fluorescent background can be orders of magnitude stronger than Raman signals for its much larger cross-section compared to that of Raman.[9]

Many strategies have been developed to reduce fluorescence[10], [11]:

i. Choose laser light with lower frequencies (longer wavelengths). Low-frequency excitation, such as near-infrared excitation, prevents the electronic absorption and therefore reduces fluorescent emission. However, this method sacrifices Raman intensity as we can conclude from Equation 2-11 that Raman intensity decreases with the frequency of the incident light.

ii. Choose deep ultraviolet (UV) excitation. Using the laser with a wavelength below 260 nm can enhance the Raman intensity as well as reduce the fluorescence. As mentioned above, Raman intensity is proportional to the fourth power of excitation frequency. Compared with visible light, deep UV excitation increases the Raman intensity

(18)

17

efficiently. Moreover, it is reported that the use of excitation wavelength below about 260 nm can avoid fluorescence interference [12]. Because few compounds fluoresce below that wavelength and their fluorescence efficiencies are, in general, reduced. iii. Use anti-Stokes Raman scattering. Anti-Stokes Raman scattering is not affected by

fluorescent emission since it has a higher frequency than the excitation light. This method is not practical because of the inherent low intensity of anti-Stokes Raman scattering.

iv. Use time-resolved methods. Fluorescence lifetime (the time that molecule spends in the excited state before emission) is longer than that of Raman scattering. Based on the difference in timescale, pulsed excitation and gated detectors can reduce the fluorescence efficiently.

v. Use shifted excitation. The wavelength of Raman scattering shifts with different excitation light while the fluorescence stays the same. Fluorescence could be filtered out by analyzing two spectra that excited by two lasers close in wavelength.

vi. Other methods to reduce fluorescence, such as quenching the fluorophore with added reagents[13] and use photon blenching [14].

2.1.3 Instrument Setup

The first Raman instrument, setup by C.V. Raman and K.S. Krishan, utilized monochromatic sunlight as an excitation source to irradiate the benzene sample, a filter to filter out elastic scattering light, and the human eye as a detector to observe the Raman scattering. Currently, the Raman spectrometers are generally composed of four parts: an excitation source, filters, a monochromator, and a detector. The laser is the most widely applied excitation source, which provides nearly monochromatic light of high intensity. A filter, usually a narrow

(19)

18

bandpass filter, removes the unwanted background of the laser light. A second filter removes the Rayleigh scattering. A monochromator, such as a diffraction grating, is used to disperse the Raman scattering light into monochromatic light. The detector or camera records the dispersed Raman scattering. The general instrument setup is displayed in Figure 2-4; a more detailed discussion of the four compounds is provided in the next chapter.

2.2 Other Forms of Raman Spectroscopy Techniques

The previous section introduces the concept and theories about the spontaneous and nonresonant Raman scattering. Since the weak intensity of Raman scattering is one of the significant problems for analyzing samples of low concentration, ways to increase the Raman intensity have been extensively explored. This section introduces the techniques that can improve Raman intensity significantly, including surface-enhanced Raman spectroscopy (SERS), resonance Raman spectroscopy (RRS), and coherent anti-Stokes Raman scattering (CARS).

Figure 2-4 A general instrument setup of Raman spectroscopy

The laser beams are purified by the laser transmitting filter, then incident on the sample. The scattered light from the sample passes a laser blocking filter to eliminate the Rayleigh scattering. The remining Raman scattering is then dispersed by the grating groove (monochromator) and recorded by the detector/camera.

(20)

19

2.2.1 Surface Enhanced Raman Spectroscopy (SERS)

Surface-enhanced Raman scattering is a phenomenon that the Raman intensity of a molecule can be increased by a factor of 106 or more when it absorbs on metal colloids or rough metal

surfaces [6]. Moreover, the fluorescence background is also reduced by the interaction between sample and metal nanoparticles [13].

The exact mechanism of SERS is still under discussion. Present theories can be classified into two categories: electromagnetic theory and chemical theory. The two theories are both valid and often combined to interpret SERS. Electromagnetic theory ascribes SERS to the increased incident electromagnetic field caused by the metal surface. When the light is incident on a roughened metal surface, it excites the surface plasmons to a certain frequency. Surface plasmons are a collective oscillation of a group of free electrons at a metal surface. The frequency of the plasmons depends on the metal type (silver, gold, copper, etc.) and the properties of the surface (size, roughness, and colloid diameter). At this frequency, the local electric field intensity is increased by the oscillating electrons. The corresponding Raman intensity is therefore increased along with the local electric field. It is worth noticing that Raman intensity could only be enhanced by oscillations perpendicular to the plane of the metal surface, which could be achieved by roughening the surface. In chemical theory, samples are bonded to the metal surface. The enhancement is caused by the back and forth transfer of the electrons between samples and the metal surface, which increases the effective polarizability.

2.2.2 Resonance Raman Spectroscopy (RRS)

RRS is the technique that records the resonance-enhanced Raman scattering. Resonance Raman scattering occurs when the incident photons have an energy that is close to or

(21)

20

matches the electronic transition energy of the molecule that produces Raman scattering. Resonance enhancement can increase the Raman intensity by a factor up to 106 [15]. RRS is

not only sensitive but also selective, because the enhancement only occurs for specific vibrations that are correlated with the electronic chromophore. Besides, the excitation wavelength provides the electronic information and is often plotted with the resonance Raman band denoted as resonance excitation profiles. Theoretically, resonance enhancement can be applied to any kind of Raman scattering. When it combines with SERS, the Raman intensity could be increase by a factor of 1014. The combination is named as surface-enhanced

resonance Raman spectroscopy (SERRS).

Experimentally, resonance Raman spectroscopy is more challenging to perform than normal Raman for the strict demand of the incident wavelength. A tunable laser is the ideal excitation source. The laser is tuned to the wavelength that matches the energy difference between the first/second vibrational level of the ground state and the first/second vibrational level of the excitation state. Other disadvantages are also associated with the excitation source: The excitation light with the electric transition energy could photochemically damage the samples and intensify the fluorescence background.

2.2.3 Coherent Anti-Stokes Raman Spectroscopy (CARS)

CARS is a non-linear process where multiple photons interact with one molecule at the same time. The term non-linear is to indicate that the scattering efficiency is not linearly related to the intensity of the incident electromagnetic field as it is in spontaneous Raman scattering. The process is illustrated in Figure 2-5. Three photons are involved in this process. One photon with the frequency of ν1 excites the molecule to a virtual state. A second photon (ν2) is of the

(22)

21

presence of ν1 and ν2 at the same time dramatically increases the molecular population on

the excited vibrational state. A third photon ν3 brings the molecule on the excited vibrational

state up to another virtual state. This molecule then returns to the ground state producing the anti-Stokes scattering from this new virtual state. The whole process is called CARS. Experimentally, the process is often simplified by using two different lasers (νpump and νStokes)

instead of three, where ν3 = ν1= νpump andν2= νStokes. The CARS scattering light occurs at the

frequency of νCARS=2νpump- νStokes.[16] CARS would occur when the difference between νpump

and νStokes matches an exact vibrational energy level, which is usually achieved by tuning the

frequency of Stokes beams or using broadband Stokes beams.

The intensity of CARS is proportional to the square of the pump intensity and the Stokes beam intensity: 𝐼𝐶𝐴𝑅𝑆~𝐼𝑝𝑢𝑚𝑝2 I𝑠𝑡𝑜𝑘𝑒𝑠,[17], and as a result, its intensity can be orders of magnitude

larger than that of spontaneous Raman scattering [8]. The intensity can be further increased by a pre-resonance or resonance effect by using an excitation frequency that is close to the electronic transition of the target molecule [6]. Another attractive feature of CARS is that it is an anti-Stokes scattering. Therefore, the fluorescent background is dramatically reduced.

(23)

22

The limitations of CARS are brought up by the nonresonant background, and more importantly, by the undesired CARS signals that are generated from other untargeted molecules in the sample and the sample environment. The sensitivity of the measurement is, therefore, decreased by the above effects.

(24)

23

This chapter introduces the basic concept of WRI as a subtype of Raman imaging, the instrument development, and parameters used to evaluate an imaging system.

3.1 Raman Imaging

Raman imaging is a combination of Raman spectroscopy and imaging techniques. A Raman image can be seen as the display of Raman spectra spatially. A typical data cube (also referred to as hyperspectral data cube) obtained by Raman imaging is shown in Figure 3-1 [4]. The data cube contains four-dimensional data: spectral information (wavelength), Raman intensity, as well as the spatial information in two dimensions (x-axis and y-axis).

Wide-field Raman Imaging

Figure 3-1 Illustration of data cube obtained by Raman imaging (Inspired by Mitsutake et al [4]). (a) Illustration of the data cube containing spectral and spatial information. The data cube contains a set of wavelength specific images represented by the big squares; Each small square in the data cube represents a pixel, whose location is defined by x and y axis. (b) Raman spectrum of a specific pixel (xi,

yj). (c) Graphical illustration of the spatial distribution for a specific wavelength (λ1), which is

(25)

24

Raman imaging is especially useful to measure materials that are spatially heterogeneous in compositions, for it reveals the spatial distribution of chemical components in the samples. Besides, due to the non-destructive and non-invasive nature of the Raman effect, it has been extensively explored in the fields such as material analysis, biology, biomedicine, pharmaceutical, threat detection, just to name a few [4].

Raman imaging techniques can be broadly classified into two types based on the data acquisition methods: scanning (also called mapping) and wide-field (also called global imaging, direct imaging or spectral Raman microscopy [5]). Scanning acquisition often requires the relative motion between samples and laser beams. In contrast, wide-field acquisition illuminates the whole area of the sample without the movement of the laser or the samples. A graphical illustration of different acquisition methods is shown in Figure 3-2 [4].

The scanning techniques can be further categorized into point mapping and line scanning. In point mapping, the spectrum is obtained pixel by pixel (point by point): The laser light is focused on a tiny spot on the sample at a time, and the full spectrum is obtained for this specific point. Afterward, the laser or the sample is moved to a new position resulting in a

Figure 3-2 Illustration of Raman imaging acquisition modes. (Adopted from Mitsutake et al. [4]) The big square refers to the sample surface; Small squares refers to pixels; The spectra are collected from the area coloured in red. (a) Point mapping mode. (b) Line scanning mode. (c) Wide-field imaging mode.

(26)

25

new spectrum. The final image is reconstructed by the collected data on each pixel point. Many advantages make point mapping the most widely used Raman imaging technique. Point mapping is usually performed in a confocal mode, which reduces the fluorescence and secondary scattering background. Besides, point mapping provides spectra with full spectral coverage and high spectral resolution. However, the disadvantages are not negligible. It is time-consuming for point by point measuring, and the image obtained is often limited in resolution. Moreover, laser-induced sample damage is accompanied by the long acquisition time (hours to days).

Line scanning is an extension of point mapping. In line scanning, the spectrum is obtained line by line: the laser light is defocused into line shape and scanned through the area of interests in one direction. The Raman scattered light then passes through a slit and, in most cases, dispersed by the diffraction grating. Eventually, the dispersed light is projected on a 2D Charge-Coupled Device (CCD) camera. The row of the pixels of the CCD camera contains the spectral information from the sample. The column of pixels of the CCD camera indicates the birthplace of a certain spectrum. Hence, the spectrum of a certain point in the illuminated line area is captured by a specific row of CCD. The same as point mapping, the final image is constructed by extracting the intensity of a certain wavelength at each sample point. Compared with point mapping, spectra for each point in a larger surface are recorded at once. Therefore, the line scanning experiment is less time-consuming. Moreover, the resolution of line scanning is also increased in the direction perpendicular to the laser movement. Because in that direction, the imaging resolution is not dependent on the laser width.

In wide-field imaging, the spectrum of the whole sample area is recorded wavelength by wavelength: The laser light is defocused and illuminates the entire sample area. A predefined

(27)

26

wavelength is then recorded for the area illuminated. The image is constructed based on the recorded spectra. Wide-field imaging has many advantages over scanning techniques such as shorter acquisition time and higher spatial resolution. Because the image of the whole sample is acquired at one time, the corresponding acquisition time of wide-field imaging is much shorter than that of scanning techniques when only a few wavelengths are demanded, resulting in highly efficient Raman imaging [18]. Laser-induced damage is also avoided. Besides, the resolution no longer depends on the width and focus of the laser light. Thus diffraction-limited resolution can be obtained by wide-field imaging. The wide-field imaging may not provide sufficient spectral information for characterizing novel materials, where the chemical composition and molecular structure are poorly defined. However, it is practical for the majority of material characterization, where a small number of wavelengths is required (< 30) [3]. Wide-field imaging is especially suitable for rapid and even real-time imaging when few wavelengths are sufficient, and has been widely applied to biomedical, pharmaceutical, threat detection, and material characterization with emphasis on polymer and carbon compounds.

3.2 System Performance

An imaging system is often evaluated or compared with other imaging techniques in terms of resolution. The following sections will introduce spectral and spatial resolution separately.

3.2.1 Spectral Resolution

The spectral resolution defines the smallest wavelength difference of two Raman peaks that can still be distinguished from each other. When two peaks cannot be separated from each other, they appear as one broad peak. Practically, we use the width of a peak at half of the

(28)

27

peak height to represent the spectral resolution for this peak. This parameter is the Full-Width Half Maximum (FWHM).

In point-mapping and line-scanning approaches, it is the entrance slit and size of the spectrometer and the properties of the diffraction grating that affect the spectroscopy spectral resolution. The spectral resolution could be increased by decreasing the entrance size of the spectrometer or increasing the density of grating grooves. These solutions will nevertheless cause a decrease in signal intensity as a trade-off.

In wide-field approach, it is the filtering method that determines the spectral resolution. The spectral resolution of the image is in line with the smallest wavelength difference that can be resolved by the filter applied.

3.2.2 Spatial Resolution

Spatial resolution defines the smallest distance between two points that can still be differentiated. The spatial resolution of WRI is determined by the theoretical limitation and the density of CCD pixels. When the properties of the detector are not taken into consideration, the spatial resolution can be estimated by the following equation:

𝑅 = 𝜆 (2𝑁𝐴)⁄ Equation 3-1 where R refers to the resolution, λ is the imaging wavelength, and NA is the numerical aperture of the optic system. R is also the theoretical limit of the spatial resolution for any optical system. The optical system that achieves the theoretical limit is called diffraction limited. Among the three categories of Raman imaging, a diffraction-limited image can only be obtained by the wide-field approach.

(29)

28

3.3 Instrumentation of Wide-field Raman Imaging

A typical WRI system is made of four main components: the excitation source, a sample illumination system or sampling platform (usually a microscopy), the wavelength selection device, and a multi-channel detector (usually a CCD camera). During the imaging process: the unfocused laser light globally illuminates the sample. The scattered light passes through an optical filter which selects the predefined Raman wavelength. The corresponding image at that wavelength is then recorded by a multi-channel detector.

Nowadays, only two types of instruments are commercially available for WRI. One is Falcon IITM (ChemImage Corporation Pittsburgh, Pennsylvania), which is made specifically for WRI.

Another is inVia™ confocal Raman microscope, which has a wide-field acquisition mode. Since most of the studies had to or preferred to use the instruments that were assembled in their laboratories, this section classifies the instrumentations by the instrument components and highlights the recent advances in excitation source and wavelength selection device.

3.3.1 Excitation Source

The laser has been applied as a universal excitation source for Raman scattering since invented. This section introduces different laser types that are used in WRI. As explained in chapter 2, the wavelength of Raman scattered light is defined by the excitation light and the vibrational energy level of the molecules. Accordingly, the laser source should agree with the filter to detect the desired Raman scattering and filter out the background signal. In order to achieve the wavelength scan in WRI, two combinations of excitation source and wavelength selection device are mostly applied: a tunable laser with a fixed filter, which or a fixed laser

(30)

29

wavelength with a tunable bandpass filter. Another alternative way is to use a fixed laser wavelength with several selected bandpass filters, as proposed in the study of Chen et al. [19].

Tunable laser

The use of a tunable laser was first reported in the study of Puppels et al. [20]. They coupled an argon-ion laser pumped tunable dye laser with a narrowband dielectric transmission filter in a Raman imaging system. The Raman scattering was selected by tuning the laser wavelength to the extent that the targeted Raman scattering was in the range of the narrowband dielectric filter. The transmitted Raman scattering was then recorded by a CCD camera. This device was able to provide an image with a spectral resolution of 10 cm-1. More

recently, Li et al. [21] built another Raman imaging platform with a tunable dye laser, presenting a spectral resolution of 1.5 cm−1 and a spatial resolution close to the diffraction

limit. The laser in their experiment can be tuned continuously over the range of 550 nm to 680 nm. The wavelength selection device was composed of a set of fixed filters combining with a Fabry-Perot etalon. Two long-pass filters were used to reduce the Rayleigh scattering, one narrowband filter worked as the essential filter to select the desired Raman signal, and an air-spaced etalon was utilized to constrict the transmitted Raman scattering.

The combination of tunable laser and narrow bandpass filters with fixed wavelength allows the high image sensitivity owing to the high transmission of the filters. Compared with systems using tunable filters, tunable laser-based systems are of higher sensitivity and provides better spectral resolution.

(31)

30

Monochromators and filters are the two most common wavelength selection device. Monochromator-based wavelength selection device, such as diffraction gratings, selects the desired wavelength by dispersing light rays in a wavelength-dependent spatial dimension, which allows the wavelength selection over a broad spectral range. Hence, it is widely used in Raman spectrometers as well as in point-mapping and line-scanning Raman imaging. However, it is not applicable to image a 2D field directly, thus not applicable for wide-field imaging.

In contrast, when coupled with a CCD camera, filters allow the capture of a 2D image directly. Therefore, they are widely used in WRI. Several kinds of filters are available, including thin-film filters, acousto-optical tunable filter (AOTF), and liquid-crystal tunable filter (LCTF). Tunable filters are primarily designed to select a predefined wavelength at one time, which makes them the ideal wavelength selection device in WRI. Applications of thin-film bandpass filters (refers to as bandpass filters in this thesis), thin-film tunable bandpass filters (TF-TBPF), AOTF, LCTF, and photonic crystal as wavelength selection devices are provided in this section.

Narrow bandpass filters

The most straightforward wavelength selection device in WRI is using narrow bandpass filters. The narrow bandpass filter is a type of bandpass filter that transmits a defined bandpass with narrow bandwidth. It is usually coupled with tunable lasers to achieve wavelength scans. Moreover, a series of discrete bandpass filters with a fixed laser is also appliable in experiments where only a few known wavelengths are required for Raman imaging [22], [23]. Even multi-channel WRI systems can be constructed by multiple bandpass filters. In the study conducted by Wei et al. [23], they used a fixed laser (785 nm) and four narrow bandpass filters

(32)

31

with different center wavelengths to capture the Raman imaging at different wavelength simultaneously. Their instrumental setup is illustrated in Figure 3-3 (A). The imaging system was set up in a backscattering arrangement. First, laser light was expanded and illuminated the sample after passing through the objective lens (OL). The backscattered light was then collected by the same OL, deflected by the mirror, and transmitted by the dichroic mirror (DM). Wavelengths below 785 nm were blocked by a long-pass (LP) filter (cutoff wavelength at 785 nm). Afterward, the filtered light was separated into four identical parts by two 2 × 2 lens arrays (LA 1 and LA 2). Each part passed through a different bandpass (BP) filter and finally reached the CCD detector resulting in narrowband imaging. Note that there were four bandpass filters, and each of them had a different passband. A method to reconstruct the full spectrum was also developed by relating the narrow-band measurements with the full Raman spectrum. In the calibration stage, they related the narrow-band measurement data to the full Raman spectrum of pure urea, potassium formate, PMMA, and trans-stilbene in a Wiener matrix. In the testing stage, the mixture of the previously measured calibration samples is tested by narrowed-band measurement. They were then able to deduce the full spectrum from four recorded wavelengths for any pixel point using Wiener estimation, shown in Figure 3-3 (D) is the reconstructed spectrum from the location “a”, “b”, “c” respectfully. By extracting the target wavelength from the reconstructed spectrum in each pixel, they featured the geological distribution of urea, potassium formate, PMMA in a mixture successfully. The spatial resolution mainly depended on the pixel density, and the spectral resolution was decided by the bandwidth of the narrow band filter.

(33)

32

The narrow bandpass filter is one of the most economical filters in WRI. It guarantees high transmittance (about 90- 100%), narrow bandwidth, and perfect out of band blocking. However, one narrow bandpass filter is selected for a specific Raman scattering at a particular excitation wavelength. Therefore, it is not feasible for universal detection, where several wavelengths are required. Even if the wavelength selection can be achieved by swapping filters, the whole process is limited in speed.

Acousto-optic Tunable Filter (AOTF)

AOTF is an electrically tunable filter based on the anomalous Bragg diffraction of the incident light when it passes anisotropic crystals under the acousto-optic interaction. Firstly, an oscillating radio frequency (RF) signal is applied to the transducer generating a varying acoustic wave that spreads through the anisotropic crystal. The refractive index of the anisotropic crystals then changes periodically by the acoustic wave. Finally, the alternating

refractive index Bragg diffracts the incident light, only transmitting undiffracted zero-order beams. In this way, AOTF can diffract incident polychromatic light according to the RF signal, resulting in specific monochromatic light. AOTF is a fast electrical wavelength selection device that covers a wide range of wavelengths and has been widely used in spectral analysis

Figure 3-3 Schematic of a multi-channel WRI system, the Raman images and reconstructed spectrum (adopted from Wei et al. [23]). (A) Instrumental setup. (B) Multi-channel images of a 1951 USAF resolution test chart recorded simultaneously under 785 nm excitation without the LB and BPs in the setup. (C) WRI taken at the wavelength of 880 nm. (D) Full spectra reconstructed at a, b, c point in (C).

(34)

33

instruments. The application of AOTF in Raman imaging can be traced back to 1992 when Treado et al. [24] utilized a tellurium dioxide AOTF in their experiment. The instrument setup was straight forward: The laser light was defocused by a lens and illuminated the sample globally. The scattered light then passed through an AOTF filter and a holographic Raman filter to select the desired Raman scattered light. Finally, the Raman scattered light hit the CCD camera, providing a high-fidelity image with a spectral resolution of 50 cm-1 (at 630 nm).

AOTFs have been developed and utilized in many studies since it was first reported in the early 1990s [25]. Nowadays, AOTF based systems allow a spectral resolution of about 0.5 nm in the visible range and 2 nm in the NIR range [26], which would correspond to a Raman spectral resolution of about 18 cm-1 and 32 cm-1 respectively.

The advantage of an AOTF is that it enables narrow bandpass and rapid wavelength selection over a broad spectrum range. Besides, it is not affected by heat or intense light. However, a small fraction of the light near the center wavelength will be diffracted by AOTF during the experiment, leading to a blurred image [25]. This phenomenon can be seen in Figure 3-4 (C): the bars and the numbers are blurry in the image captured using AOTF. The price of AOTF spectroscopy instruments is generally high due to the expensive AOTF crystals. At present, the only anisotropic crystal that has been widely used in the AOTF instrument is TeO2, which

has an operating wavelength range of 350 nm to 5500 nm. Accordingly, the main applications are currently limited to the visible and near-infrared bands.

(35)

34

Liquid crystal tunable filters (LCTF)

LCTFs are designed based on the electronically controlled birefringence effect of liquid crystals and were applied in Raman imaging almost at the same time with AOTFs. Generally speaking, AOTFs offer higher transmittance and broader spectral range, while LCTFs provide larger aperture, improved image quality, and diffraction-limited spatial resolution [25],[27],[28]. More specifically, the performances of different types of LCTFs are slightly different. Nowadays, several types of LCTFs are available: Fabry-Perot LCTF, Lyot-type LCTF, Solc LCTF, Evans Element LCTF, among which, Lyot-type, Solc, as well as Evans Split-Element LCTFs are polarization-dependent and known as polarization interference filters. The introduction and application of Fabry-Perot LCTF, Lyot-type LCTF, and Evans Split-Element LCTF are described separately below.

Lyot-type Liquid Crystal Tunable Filter (LT-LCTF)

The LT-LCTF, based on the design of the Lyot filter, is composed of a series of individual filter stages. Each stage is made up of an unchangeable retardance birefringent element and a liquid crystal wave plate sandwiched in linear polarizers. Tuning is achieved by electronically

Figure 3-4 The images of 1951 USAF resolution test chart. Images taken from H. R. Morris et al (1994) [25]. The imaged area contains two elements 5 and 6 with a spacing between two bars of 4.9 and 4.3 µm, respectively. (A) Brightfield image collected with the microscope using a bandpass filter. (B) Image collected by the same microscope using LCTF; (C) Image collected by the same microscope using AOTF.

(36)

35

rotating crystal axes of the liquid crystal, which changes the transmission wavelength of the LCTF correspondingly [29]. LT-LCTF allows rapid tuning (<20 µs), broad spectral coverage (from 500 nm to 900 nm) [30], and narrow bandpass, which makes it suitable for WRI. Examples abound. The first system that utilized the LT-LCTF provided the diffraction-limited image of polystyrene microspheres. The average spectral resolution was 7.6 cm-1 across the

full spectral range ( 564 cm -1 to 4052 cm -1) and a transmittance range from 6.7 to 16.3% [27].

Later in the 1990s, Kline et al. [31] conducted Raman imaging of breast tissue utilizing a commercially available LT-LCTF (CRI) from VariSpec, providing the image with a resolution close to the theoretical limit (diffraction-limited). LT-LCTFs can provide images with excellent spectral and spatial resolution; however, they are often limited by the innately low transmission. Because the Raman scattered light is initially of low intensity, longer exposure time is requested for filters with lower transmittance, which limits the rapid imaging and causes potential photodamage of samples.

Evans split element LCTF

One alternative to LT-LCTFs is the Evans split element LCTFs, which is inspired by the Evans split element filter [32]. In an Evans split-element filter, two stages of Lyot filter are merged into one stage by splitting up the birefringent element of one stage into two parts and placing the split elements on both sides of the birefringent element of another stage. Evans-split element LCTFs improve the transmittance of Lyto filters by decreasing nearly half of the number of the polarizer while preserving the narrow bandpass and broad spectral range [33][34].

In a recent study, Evans split element LCTF provided a spectral resolution of 9.9 cm-1 over a

(37)

36

nm) [35]. Evans split element LCTF outperforms the LT-LCTFs concerning transmittance. However, its performance is profoundly affected by temperature. Evans split element LCTFs suffer from severe peak drifts and decreased transmission under higher temperatures.

Fabry–Perot LCTF

Fabry-Perot LCTF is designed based on the theoretical concept of Fabry-Perot Interferometer: multiple-beam interference occurs when light passes through a cavity, which is in between two reflective parallel surfaces. For a Fabry-Perot LCTF, liquid crystal, whose refractive index can be controlled by the applied voltage, is used as the cavity material. The desired transmission wavelength is then selected by tuning the voltage.

Compared with LT-LCTF and Evans split element LCTF, Fabry-Perot LCFT improves the transmission dramatically. The transmission of a Fabry-Perot LCTF can be modified by the coating design. With proper coating, it can be up to 60- 70% [36]. However, Fabry-Perot LCTF is limited by its spectral bandpass resulting in low spectral resolution, smaller wavelength range leading to a limited free spectral range, and low out-of-band rejection. Moreover, it also suffers from the thermal-induced bandpass drift [3].

Multi-Conjugate Filter (MCF)

A recent advancement of LCTF is the development of MCF. MCF is modified based on the Evans split element LCTF. Compared with other polarization interference filters (Lyot type and Evans split element LCTF), each stage of MCF is increased in finesse by 1.5× to 2×. Overall, MCF allows higher transmittance, good out of band blocking efficiency and broader spectral range. Table 3-1 compares the essential optical properties between MCF and Evans split element filter [3].

(38)

37

Thin-Film Tunable bandpass filters (TF-TBPFs)

Thin-film filters are substances such as optical glass with single or multiple-layered dielectric thin-film coatings. The filtering effect results from the internal interference of the incident light. TF-TBPFs share the advantages of all thin-film filters: high transmission (can be up to 100%), good out of band blocking ( 106 ), a high threshold for laser damage, and affordable

price due to the commercial availability. Compared with other tunable filters, TF-TBPFs also show unique features such as adjustable bandwidth and excellent image properties. However, the spectral range of TF-TBPF is relatively small, and tuning speed is limited by the mechanical process.

Two types of thin-film tunable bandpass filters are currently available: Linear-variable tunable filters (LVTF) and angle-tuned thin-film filters.

Linear-variable tunable filters (LVTF)

LVTFs are filters with nonuniform thin-film layer thickness, which varies continuously along one linear direction. Therefore, the center wavelength is approximately in a linear relationship with the spatial position of the LCTF. Tuning is achieved by moving it along a linear direction. One typical setup of TF-TBPF is made up of two filters: One Linear Variable

(39)

38

Long-Wave Pass filter (LVLWP), and One Linear Variable Short-Wave Pass filter (LVSWP) [37]. This setup is illustrated graphically in Figure 3-5. The LVLWP and LVSWP are moving in the plane that is perpendicular to the incident light for wavelength selection. Bandwidth can also be tuned by the relative movement between LVLWP and LVSWP. Although LVTF offers high transmission and good image properties, it is not widely applied in Raman imaging, which could be due to its broad bandwidth (approximately 35 nm according to [37] ).

Angle-tuned thin-film filters

Angle-tuned thin-film filter is developed by utilizing the phenomenon that changing the angle of incidence (AOI) of a thin-film interference filter would change its center wavelength [38]. As the AOI increases, the center wavelength decreases. However, larger AOI deteriorates passband ripple for s-polarized light and diminishes p-polarized light, reducing the overall transmittance. The polarization splitting can be alleviated by a novel setup that combines a short-wave-pass (SWP) edge filter with a long wave pass edge filter (LWP) [39], shown in Figure 3-6. The LWP is turned to bock the wavelength longer than the center wavelength, and the SWP further blocks the wavelength shorter than the center wavelength. In this system,

(40)

39

the AOI can be adjusted from 0° to 60° without sacrificing filter performance, offering a twelve percent tunable range of any wavelength.

A recent study by Iga et al. [40] demonstrated the applicability of TF-TBPF in SERS imaging. Their filtering system was a TF-TBPF in combination with multiple bandpass filters, achieving a tunable FWHM of 1.5 nm, 2 nm, and 3 nm over the spectral range of 530 nm to 610 nm. The desired wavelength was selected by changing the angle between the incident light and filter surface, which is achieved by rotating the TF-TBPF. The resulting image shows the diffraction-limited spatial resolution and a minimum spectral resolution of 1.5 nm, which would correspond to a Raman spectral resolution of about 40 cm-1.

Figure 3-6 Illustration of TF-TBPF using angle-tuned thin film filters. The desired wavelength is selected by rotation of the two tunable filters

(41)

40

Photonic crystal

Photonic crystals are materials in which dielectric optical nanostructures are arranged periodically. Light is scattered by the periodically varied the dielectric constant when it propagates within a photonic crystal. The interaction between light and the photonic crystal can be constructive or destructive interference, and it is subject to the wavelength, polarization, direction of the incident light as well as the chemical and structural composition of the crystal lattice. Therefore, a certain wavelength will be strongly enhanced at a particular AOI by a specific photonic crystal, which is known as Bragg diffraction (illustrated in Figure 3-7 [41]). The constructively interfering wavelength is given the following function [42]:

𝑛𝜆 = 2𝑑𝑚 sin 𝜃 Equation 3-2

Where n is a positive integer referring to the diffraction order, λ is the wavelength of the light, d is the spacing between two lattice planes, m is the refractive index of the crystal, and θ is the glancing angle. For a particular crystal, d is considered to be constant. Thus, photonic crystals can be used as a tunable filter in Raman imaging: wavelength can be selected by changing the AOI, which is achieved by rotating the photonic crystal stage.

Figure 3-7 Schematic of the Bragg diffraction. Figure adopted from Baskaran [41]. Incident light (solid line) hits the crystal lattices by an angle θ with respect to the crystal plane. The two incident light will interfere constructively when the difference between their travelling distances ( 2𝑑𝑚 sin 𝜃) is equal to a whole number multiple of their wavelength. The blue dot represents the atoms in crystal lattice; The dash line is the imaginary crystal planes that have a spacing of d.

(42)

41

A photonic crystal based wavelength selection device was first utilized in WRI by Hufziger et al. [43] The face-centered cubic (FCC) photonic crystals were self-assembled by polystyrene nanoparticles of about 102.7 nm diameter. The crystal was mounted on a rotatable stage, and the AOI was tuned between 5° and 40°during experiments. The displayed FWHM was about 9 nm, and the diffraction efficiency was about 70% varying with the incidence angles. The Raman spectral resolution present here is about 330 cm-1, for a wavelength range of 490 to

540 nm. It is relatively broad compared with the other Raman imaging system, such as LCTF based systems, which offer a spectral resolution of 10 cm-1.

Photonic crystals are also developed to meet up the shortage of deep UV filters. A recent advance is reported by the same group. They synthesized an FCC silica type photonic crystal by self-assembling 35.5 nm silica nanoparticles in solution. This photonic crystal presented an FWHM of about 1 nm (100 cm-1) over the deep UV range [44].

The advantage of using a photonic crystal is that it offers a high diffraction efficiency (about 70%), and it is highly customizable in respect of spectral range and bandwidth. The spectral range is covered from IR to deep UV by using different photonic crystals. The bandwidth can be minimized by several methods, such as reducing the particle diameters, reducing the crystal thickness, and raising the array order. However, one limitation of the photonic crystal based device is that the spectral range is limited to the altering of incident angles. In practice, the angle of incident light can only be changed over a limited range due to the loss of diffraction efficiency at larger angles. Another disadvantage is that the impurities with charge introduced in synthesis would cause disorder and increases the bandwidth in practice.

(43)

42

Summary and comparison of the wavelength selection devices

Several characteristics of the filters mentioned above are compared in Table 3-2. A more detailed summary of the instrumental setup is provided in Appendix table I1.

Table 3-2 Comparison of different wavelength selection devices. The number of stars refers to how good the filter performance is for one specific property. The color also acts as an indicator (The darker the color, the better the filter). NA means the property does not apply to the filter.

LCTF AOTF LVTF angle-tuned thin-film filter Photonic crystal Bandpass filter

Angular field of view ★★★★ ★★ ★★ ★ ★★★

Aperture ★★★★ ★ ★★★ ★★★★ NA ★★★★

Bandpass ★★★★ ★★★ ★★ ★★★ ★★ ★★★

Free spectra range ★★★★ ★★★★★ ★★ ★★ ★★★ NA

Imaging quality ★★★★ ★★ ★★★ ★★★★ ★★★ ★★★★

Low cost ★★ ★★ ★★★★ ★★★★ ★ ★★★★★

Optical damage threshold ★★ ★★★ ★★★★ ★★★★ NA ★★★★

Out-of-band blocking ★★ ★★ ★★★★ ★★★★ ★★★★ ★★★★

Transmission/diffractive efficiency ★★ ★★★ ★★★★ ★★★★ ★★★★ ★★★★★

Tuning speed ★★★★ ★★★★★ ★★ ★★ ★★ NA

3.3.3 Fiber Array-based Raman Imaging

The microscope is one of the most popular sampling devices of Raman imaging. In WRI, the microscope is used in a conventional way, where a sample is illuminated entirely, scattered light is collected, and the final image is a magnified view of the sample. The utilization of the microscope is also the reason why WRI is referred to as spectral microscopy in some literature [5].

(44)

43

A recent advance of the sampling device is the development of fiber array-based Raman imaging. It was first reported in the study of Myrick et al. [45], named as dimension reduction arrays, and has been highlighted recently, known as fiber array spectral translator (FAST ) [46]–[48]. The scheme of the device is shown in Figure 3-8 [2]. A bunch of fibers is used for sample illumination as well as signal collection. One end of the optical fibers (proximal end), which is close to the sample, is arranged into a 2D plane surface. Fibers at another end (distal end), which is connected to a spectrograph, are lined up with special ordering. In the imaging process, the Raman signal of the sample is first captured by the proximal end of the fibers, then transport to the distal end. Finally, the light from each fiber is dispersed by the spectrograph simultaneously and reaches the CCD detector. The full spectrum from each distal end is therefore recorded. The image is reconstructed based on the spectral contract and the location of the proximal end. The array in this application acts as a

(45)

44

reduction device allowing both spectral and spatial information (two-dimension) to be recorded by a 2D detector concurrently without loss of spectral resolution.

However, the poor signal transmission limits the performance of fiber array: nearly 99% light is lost, as reported by [45]. Besides, the spatial resolution is dependent mainly on the number of fibers in the fiber bundle and the maximum pixels of a CCD detector. Although the resolution can be improved by adding more fibers to the bundle, much more detection rows will be required. For example, at least 900 fibers and 900 detection rows in a CCD camera are required to obtain a 30×30 pixel image. Hence, 90,000 fibers and 90,000 detection rows would be demanded to improve the resolution by a factor of 10, resulting in a 300×300 pixel image. However, a commonly used CCD camera only has about 1,000 detection rows. It is obviously not practical for high-resolution imaging by current devices.

Despite the limitations, advantages of FAST such as rapid measurement, the full spectral range, which also enables imaging in deep UV [49], high spectral resolution make it a promising approach for biomaterialsand stand-off detection of explosives.

(46)

45

This chapter describes various applications of WRI with specialized topics: materials and polymers, pharmaceutical products, biology and biomedicine, threat detection, food safety, and forensic science. The representative applications in each field are summarized in the table A1 to table A5, respectively (see Appendix).

4.1 Materials and Polymers

The Raman effect has been widely used to study 2D nanomaterials such as graphene, and other graphene-like materials, including two-dimensional hexagonal boron nitride (2D-hBN) and molybdenum disulfide (MoS2) [50]. Graphene is composed of hexagonal lattices which

are formed by sp2 hybridized carbon atoms. It is the basic unit of several other carbon

allotropes, such as graphite, charcoal, carbon nanotubes, and fullerene. Researches on graphene are growing extensively due to its optical properties, electrical performance, and high mechanical strength. For nanomaterials like graphene, Raman spectra reveal not only qualitative and quantitative information but also the defects /contaminant, doping, strain, and layer number due to the strong optical resonances.

Compared with Raman spectroscopy and point mapping Raman imaging, WRI allows rapid imaging over a large area with sufficient spectral resolution. Therefore, it is especially useful to locate the target chemicals in a large range of arbitrary substrates, and to real-time image dynamic processes of the material fabrication. Examples abound.

Havener et al. [51] attempted to image graphene via WRI. Their experiments show that WRI can

Referenties

GERELATEERDE DOCUMENTEN

Consultant Barend Löbker van Vortus benadrukte dat voor een optimale productie een juiste combinatie van licht, temperatuur, CO 2 en luchtvochtigheid nodig is.. “Dat vraagt

As the role of the Value Chain in creating shared value was highlighted in the literature review, this research theorized that focusing CSR activities in the Value Chain could be

As a demonstration of the hardness-based necking model, the uniaxial tensile tests of the main grades and the intermediate grades are simulated using a 0.5mm solid element model and

Bijlage B Overzicht betrokken partijen Voor deze verdiepingsfase hebben we samengewerkt met partijen die bij de zorg voor vrouwen met bekkenbodemklachten zijn betrokken:

In the previous section, the individual's role in the forming of Solidarities was explored along the lines of them either directly or indirectly engaging with their environments.

The aim of this study was to investigate the longitudinal performance on the RFFT by repeating the test over an average follow-up period of three and six years in a large cohort

Section 2.3 further elaborates on precision medicine, and describes various analysis methods that can be performed on (time series) data collected in psychopathology research..

For the ∼1-μm-long junction, the Fabry-Pérot interferences-induced oscillations of the critical supercurrent are observed, suggesting the ballistic transport of the surface