On-chip Detection for Microfluidic Devices

(1)

Literature Thesis

On-chip Detection for Microfluidic devices

- Review of printing techniques and materials, geometries, obstacles,

and the possibility of multichannel detection -

By

Sharene Veelders

October 2019

Student number Responsible Teacher

11009268 Prof. dr. P.J. Schoenmakers

Research Institute Daily Supervisor

Van ’t Hoff Institute for Molecular Sciences drs. P. Breuer

Research Group Second Reviewer

(2)

2

Abstract

Highly complex samples can be separated (e.g. LC, EC) in two or three dimensions with a high peak capacity within printed microfluidic devices. These devices can be printed using different 3D-printing techniques with various polymer materials that have different chemical and physical properties. Unfortunately, the detection is challenging, especially when spatial separations in multiple channels are performed. Therefore, this literature study will review the different detection possibilities for (3D-printed) microfluidic devices. Firstly, optical detection techniques, such as, absorbance and fluorescence, and their obstacles are examined (i.e. light integration, short pathlengths, low volumes, separation of excitation and emitted light). Optical fibres, lenses, bubble, Z-or U-cells, droplet based systems and 90⁰ measurement angles are commonly-used techniques for optically transparent microfluidic devices and are discussed in this review. Furthermore, (off-line) laser-induced fluorescence imaging or optical fibres with integrated lenses are great strategies for multichannel detection. Secondly, ElectroChemical (EC) amperometric detection will be discussed, for which the separation and sealing of the electrodes is crucial inside the microchannel(s) to prevent short-circuits and leakages. Furthermore, the polymeric material needs to prevent joule heating and be compatible with high potentials. Thirdly, confocal and fibre-based Raman spectroscopy will be reviewed for optically transparent microfluidic devices. Here, a 90⁰ measurement angle and adding gold, or silver coated colloids to the sample to perform Surfaced Enhanced Raman Spectroscopy (SERS), provides a better sensitivity. In addition, optical fibres can be applied for multichannel Raman detection with additional filters in the xyz-space. Raman images can also be recorded in the xy-plane with a confocal Raman microscope with a CCD camera. Finally, Mass Spectrometry (MS) could be coupled with microfluidic devices for highly sensitive identification measurements. MS with ElectroSpray Ionisation (ESI) could be used, for which the (electrode) coupling is preferably done by integrating emitters with fused silica capillaries to (multiple) microchannel(s). However, Matrix Assisted Laser Desorption/Ionisation (MALDI) is also possible by fully automated multichannel(s) droplet spotting onto a MALDI plate, for which the solvent properties need to be taken into account. Unfortunately, detection techniques after spatial separations in the xyz-space have not been discussed in literature yet, thus the proposed strategies in this review for the xyz-space need to be performed in practice in the future to determine their selectivity and sensitivity within microfluidic devices.

(3)

3

Abbreviations

ABS Acrylonitrile Butadiene Styrene

CCD Charge-Coupled Devices

CE Counter Electrode

CNC Computer Numerical Control

COPs Cyclic Olefin Polymers

DMD Digital Mirror Device

EC ElectroChemical

EOF Electro-Osmotic Flow

ESI ElectroSpray Ionisation

FDM Fused Deposition Modelling

IR InfraRed

LC Liquid Chromatography

LCW Liquid Core Waveguides

LED Light-Emitting Diode

LIF Laser Induced Fluorescence

LOC Lab On Chip

LOD Limit Of Detection

LOM Laminated Object Manufacturing

MALDI Matrix Assisted Laser Desorption/Ionisation

MS Mass Spectrometry

PAD Pulsed Amperometric Detection

PC PolyCarbonate

PDMS PolyDiMethylSiloxane

PEG-DA Poly(Ethylene Glycol) and DiAcrylate

PFA PerFluoroAlkoxyalkane

PLA Poly(Lactic Acid)

PMMA Poly(Methyl MethAcrylate)

PMT PhotoMultiplier Tubes

PP PolyPropylene

PS PolyStyrene

RE Reference Electrode

RI Refractive Index

RRS Resonance Raman Spectroscopy

SERS Surfaced Enhanced Raman Spectroscopy

SLA StereoLithography

SLS Selective Laser Sintering

STAMP Separation Technology for A Million Peaks

S/N Signal to Noise ratio

TIR Total Internal Reflection

TOF Time-Of-Flight

UV Ultra Violet

WE Working Electrode

(4)

4

1 Introduction

In analytical chemistry, scientist are challenged with highly complex samples (e.g. environmental, food and biological), which require more specific and progressive separation methods [1]. Liquid Chromatography (LC) separations in two dimensions for these complex samples achieve a higher peak capacity in a shorter time than one dimensional separations [2]. Most optimal, these two separation dimensions are orthogonal to each other and this is regularly performed on a column based scale [1]. Adding a third dimension for spatial 3D-LC separations can theoretically yield peak capacities approaching the one million target, which can potentially be able to separate ten thousand to a hundred thousand peaks. The Separation Technology for A Million Peaks (STAMP) project aims for the realization of this concept by creating devices with 3D-printing technologies. These 3D-printed devices are called microfluidics, Lab On Chip (LOC), or Micro Total Analysis System (µTAS) and integrate all the different functions of the normal scaled laboratory into a small chip [3,4].

These microfluidic devices can be used to perform different type of lab functions such as sample preparation and pre-treatment, different types of separations, various reaction processes and detection [4]. The concept of using these devices was invented by Manz et al. in the early 1990s and since then this concept has been improved and applied extensively in analytical science and other research fields. This is mainly due to the many advantages microfluidics have over large scaled processes[4]. Since microfluidics and LOC are low volume based techniques, there is minimal solvent consumption, hence results in less waste, lower costs and it is better for the environment and the safety of the employees [4,5]. Furthermore, due to its compactness it provides shorter analysis times, enhanced accuracy, better process control, and it is a portable device.

However, there are also some disadvantages of these microfluidic systems regarding the detection. Since the volumes are greatly reduced, the amount of analyte that is available decreases, which has an effect on the sensitivity of the measurement [6]. Therefore, the detection of choice is required to have enough sensitivity and selectivity in order to make an accurate measurement of the analyte(s). Another aspect that needs to be considered is the material of the microfluidic device. The early microfluidic devices were made out of glass and silica substrates which are highly compatible with many detection techniques. Unfortunately, manufacturing these devices requires complex etching processes and can usually only create planar devices [7]. Using polymer substrates (e.g. epoxies and acrylates) is much more efficient, because polymers are easily printed and manipulated in the desired shape and have a great potential for disposable devices [8]. However, the polymer needs to be physically and chemically supportive for the requirements and parameters of each detection technique. Moreover, the polymer should not have any influence on the separation mechanism and needs to be compatible with the utilized solvents and pressures.

The various detection techniques become even more challenging when two-dimensional or spatial three-dimensional separations are performed. The reason for this is that these (3D-printed) microfluidic systems contain multiple channels in the xyz-space, which are difficult to couple with some detection systems [2]. Therefore, this literature thesis examined the offline and on-chip detection possibilities for (3D-printed) microfluidics in the first, second and third dimension. The obstacles and potentials of optical, Raman, electrochemical and mass spectrometry detection methods were discussed and evaluated. Furthermore, the choice of material was taken into account for each detection technique. At the end of this review, advices were given of the possible detection strategies, that are compatible with (multichannel) 3D-printed microfluidic devices.

(6)

6

2 3D printing techniques and material characteristics

There are different 3D-printing techniques available for printing microfluidic devices with low diameter microchannels. However, not every 3D-printing technique is beneficial for creating suitable microfluidic devices [7]. 3D-printing techniques such as; StereoLithography (SLA), Laminated Object Manufacturing (LOM), Fused Deposition Modelling (FDM) and Selective Laser Sintering (SLS) can be used for the printing of microfluidic devices. However, each technique uses different (polymeric) materials, which can have an influence on the physical properties of the device. Therefore, some techniques need additional steps to form stable closed channels within the microfluidic device. In the upcoming sections, the four printing techniques and the chemical and physical properties of the printing materials will be discussed.

2.1 Fused Deposition Modelling

Fused Deposition Modelling (FDM) is a 3D-printing technique that is based on melting polymer filaments with a hot metal cylinder and piping it with a tip into the desired shape, see Figure 1 [9]. The polymer filament is printed onto a mechanically moveable print bed at a certain distance, which also determines the resolution and the thickness of the printed part. The flow rate of the piping is crucial for the thickness of the printed part as a faster flow rate causes a thinner layer. During the printing the print bed is lowered, and additional layers can be applied onto the printed part. The advantage of this printing technique is that polymer filaments are based on cheap biocompatible materials such as PolyLactic Acid (PLA), PolyStyrene (PS), or Acrylonitrile Butadiene Styrene (ABS) [7]. Furthermore, this technique is able to use printable metal filaments to create metal printed parts. Unfortunately, FDM printed parts are easily affected by pressure and stress due to the filament structure of the layers which are often not fully fused with each other [7,10]. This effect can be minimized when additional ironing is applied to ensure more adhesion between the different layers.

(7)

7

2.2 Selective Laser Sintering

Selective Laser Sintering uses a powder based material which is heated up with a focused laser in the shape of the defined design, see Figure 2 [7]. This is established on a mechanically moveable print bed which lowers deeper into the powder when another layer is applied. A roller will then dispense another layer of powder over the previous layer. It is also possible to use metal alloy powder or mix multiple polymer powders with each other. An advantage of using powdered material instead of a liquid resin or polymer filament is that it is chemically more pure and can be compressed together.

Figure 2: Selective Laser Sintering (SLS). Adapted from [7].

2.3 StereoLithography

Unlike SLS or FDM, StereoLithography (SLA) uses a liquid resin which is photopolymerizable [11]. The resin usually contains epoxides or acrylate-epoxy hybrids monomers and oligomers, and photoinitiators or Ultra Violet (UV) absorbers. The resin is activated when it is exposed to laser, or focused Light Emitting Diode (LED), of a certain wavelength and hence causes polymerisation of the monomers [12]. The light source will travel a cross-sectional pattern through mirrors, also known as a Digital Mirror Device (DMD), and is then reflected at the surface. Due to the mechanically moveable platform, different layers can be printed on top of each other. The SLA printing technique can be performed in either the batch configuration (free surface approach), see Figure 3 A, which prints layers on the top of the printed part, or in the bat configuration (constrained surface approach), see Figure 3 B, which prints layers at the bottom of the printed part [7,12] The advantage of using the bat configuration is that the new layers are not exposed to oxygen (which encourages the curing process), because the printed part stays inside the resin reservoir and the layer thickness is not dependent on the depth of the reservoir.

(8)

8

Figure 3: Stereolithography (SLA) in the free surface (A) and bat configuration (B). Adapted from [13].

There are different resin polymers available and each have different chemical properties, which results in different absorption wavelengths [7]. This means that when different resins are exposed to the same parameters, it can result in different structures of the printed part which can be physically unfavourable. Also, the speed of the printing and the curing time between the layers is crucial to avoid fragility of the microfluidic device. The materials that are often used with the SLA are Poly(Methyl MethAcrylate) (PMMA), PolyDiMethylSiloxane (PDMS), or Poly(Ethylene Glycol) DiAcrylate (PEG-DA) [14].

2.4 Laminated Object Manufacturing

The last technique that can be used for the production of microfluidic devices is Laminated Object Manufacturing (LOM), see Figure 4 [7]. With this technique, layers of polymer, or metal plates are laser-cut and linked together via chemical bonding or specific types of coatings. However, this technique focusses more on the printing of metals than of polymers and is a less refined technique. Therefore, it requires more actions to smooth the layers.

(9)

9

2.5 Properties Of 3D-printing Materials

As mentioned before, different (polymeric) materials can be used to print, or mould, microfluidic devices. Each type of (polymeric) material has different chemical and physical properties which are described in Table 1.

Table 1: Material, printing technique and their chemical and physical properties with matching reference

Material Printing Technique Properties Ref PolyDiMethylSilox ane (PDMS) Porous matrix of Si-O backbones SLA Thermally curable + Optical transparent + Good elasticity + Gas permeable + Biocompatible + Low autofluorescence + Can be coupled with (ESI-) MS + Apolar

+ Good electrical insulation

+ Can generate a stable Electro-Osmotic Flow (EOF) - UV blockers prevent fluorescence in low wavelengths - Channel deformation possible

- Not organic solvent and acid and base compatible - Lots of sample absorption

- Not pressure resistant

- Electrodes can cause PDMS to crack if elastomers are flexed

Refractive index = 1.47 Strong adhesion with glass

[10,15– 19] Poly(Methyl MethAcrylate) (PMMA) SLA Vacuum hot embossing Injection moulding Amorphous + Optical transparent

+ Rigid and good mechanical stability (resistant high pressures)

+ Low water absorption

+ Compatible with organic solvents and acids and bases + Can be coupled with (ESI-) MS

+ Good electrical insulation + Can generate a stable EOF - Not biological inert - Not gas permeable Refractive index = 1.491 Tensile strength = 61.5 MPa

Glass transition temperature = 105 ⁰C

[3,10,15– 17,20,21] Poly(Lactic Acid) (PLA) FDM Injection moulding + Biodegradable + Gas permeable + Pressure resistant

+/- Semi-transparent PLA available, fluid flow visible - Thermally unstable

- Not gas permeable

[10,13,22 –24]

(10)

10 Refractive index = 1.89

Tensile strength = 56.6 MPa Pressure resistant to ≈2000kPa

PolyStyrene (PS) FDM

Amorphous

+ Rigid and good mechanical stability (resistant against high pressures)

+ Low water absorption

+ Compatible with organic solvents and acids and bases - Not gas permeable

[15,20] Cyclic Olefin Polymers (COPs) Laser ablation and micromillin g

+ Highly resistant to polar organic solvents + Biological inert

+ Low water absorption (lower than PMMA)

+ Optically transparent in UV-VIS (higher than PMMA, PC and PS)

+ Good electrical insulation

+ Good reproducibility with MS (better than PDMS and PMMA)

+ Cheap

+ Low background Fluorescent signal

- Extremely hydrophobic  attacked by non-polar organic solvents

- Relatively high autofluorescence (higher than PDMS)

[16,25]

PolyCarbonate (PC) FDM

Amorphous

+ Optical transparent in visible light + Rigid polymer

+ Resistant for acids

+ Can be coupled with (ESI-) MS - Limited resistance against alcohols

- Low stability for organic solvents and bases - Not gas permeable

[3,13,16,1 7,20,26] SU-8 (negative photoresist) Epoxy Photolithog raphy

+ Easily structured with standard lithography techniques + Stable to high temperatures

+ Optically transparent + Stable to organic solvents + Can be coupled with (ESI-) MS + biocompatible

- High optical propagation loss in UV - High water permeability

- Expensive

[3,16,17,2 0]

Silica and glass Standard

photolithog raphy

+ Resistant to organic solvents + Reusable microfluidics + Pressure resistant

+ Optically transparent (UV-VIS) + Electrically insulating

(11)

11 + Stable against high temperatures

+ Good for droplet formation + Generate a stable EOF

+ Can be coupled with (ESI-) MS

- (Often) not printable with 3D-printing techniques - Expensive

- Not gas permeable

PolyPropylene (PP) Injection

moulding / FDM Semi-crystalline

+ Robust and chemically inert + Flexible and Cheap

+ Highly transparent in the THz frequency range + Can be coupled with ESI

+ Biocompatible

+ Great chemical resistance - Not gas permeable

Refractive index = 1.495

Glass transition temperature = -20 ⁰C

[3,7,13,17 ,22,27,28]

Silicon Crystalline + Pressure resistant

+ Can be coupled with (ESI-) MS + Thermally stable

- Not compatible with high electrical fields

[3,17]

PlasCLEAR SLA

UV curable

+ Optically transparent (available) + Water resistant

- Not resistant to organic solvents - Low pressure stability

- Thermally unstable - Not Biodegradable

[10,29,30]

Microfluidics SLA

UV curable

+ Partly optically transparent + Water resistant

- Not resistant against organic solvents - Low pressure stability

- Thermally unstable

[29]

Quartz + Optically transparent

+ Can generate a stable EOF - Low conductivity

- Expensive

(12)

12

3 Optical detection

Optical light based detection techniques, such as absorbance and fluorescence, are frequently used due to their flexibility, reproducibility, simplicity and comprehensive library [33]. Furthermore, optical detection techniques are relatively cheap and can be coupled to photon sensitive detectors (e.g. PhotoMultiplier Tubes (PMTs) and Charge-Coupled Devices (CCD)) [6]. The upcoming sections will describe absorbance and fluorescence detection in single- and multichannel separation together with their obstacles and examples in practice will be specified.

3.1 Absorbance

When a sample is irradiated with photons from a light source of a certain energy and wavelength, certain molecules within the sample can absorb these photons [33]. This causes a decrease in the light intensity, which can be measured with absorbance detection. The requirement for absorption in the UV-Vis region (200 – 800 nm) is that the analyte of interest contains a chromophore, meaning it should contain π bonds with or without conjugation [34]. The absorbance is based on the Lambert-Beer law, see equation 1, and thus depends on the concentration, path length and extinction coefficient which is a property dependent on the analyte of interest [33].

𝐴 = 𝑙 𝑐 𝜀

Equation 1: Lambert Beer equation (A = absorption, l = path length, c = concentration, ε = extinction coefficient)

A typical absorbance spectrometer consists of a light source, a monochromator (to select the desired wavelength), a flow cell and a detector that measures the incoming light [35]. Under normal operating conditions, an optical path length of 10 mm is used for the cuvette, which requires a relatively large sample volume [33]. However, the optical path length (l in equation 1) from the microchannels is significantly smaller, causing a lower volume within the microchannels, which means there are less molecules available that can absorb the incoming photons. This makes measuring the decrease in the light intensity more challenging, which decreases the sensitivity and resolution of the measurements. In addition, absorption at wavelengths below 190 nm gives a relatively poor signal-to-noise ratio (S/N) and light can easily interfere with air, rough edges, solvents and other impurities present in the microchannels [34]. Furthermore, the choice of material of the microfluidic device is crucial, because unlike glass and quartz, polymer substrates are not always optically transparent for UV and visible light, which can make them unfavourable [8]. Fortunately, there are some systems and methods that provided solutions for these obstacles which will be discussed in the next subsections.

3.1.1 Light Source and Liquid-Core Waveguides

The first important step for optical detection within microfluidic devices is the generation of light from a LED source or a laser [36,37]. LED light sources are cheap, small, have long lifetimes and their intensity is stable, resulting in less noise in the spectra [38]. However, lasers are usually preferred over LED light sources because of their specific wavelength, coherence and high intensity. Unfortunately, incorporating the laser with the microfluidic device is more expensive, can cause degradation of the device and the sample and is material dependent [36,37].

(13)

13 The light then needs to be transported through a microfluidic device, with minimal stray light contribution. This can be achieved by using Liquid-Core Waveguides (LCWs), or optical fibres, which are both based on Total Internal Reflection (TIR) [33,39,40]. LCWs consist of a tube with a fluid inside that has a higher Refractive Index (RI) than the tube itself, creating the ability for light to travel through the fluid in the tube [39]. Consequently, it is crucial to choose the correct material or coating for the tubing. Glass (silica), quartz and printable polymers have a relatively high RI value, which means that only non-aqueous (organic) solvents with a higher RI can be used as a fluid to maintain the TIR effect. However, since most printable polymers are not compatible with organic solvents, see section 2.5, this becomes problematic. DuPont overcame this issue by creating a polymer material for the tubing with a RI lower than water (around 1.30), which was the fluoropolymer: Teflon AF 2,2-bistrifluoromethyl-4,5-di-fluoro-1,3,-dioxol [33,39]. This fluoropolymer can be applied in two different ways, namely as a coating inside the tube or as the tubing itself. Additional advantages of this material are that it is highly flexible (coiling up is possible), can be smoothened (prevents scattering), is hydrophobic, transparent in region 200-2000 nm, has a low UV transparency and is gas permeable due to its porous structure [39]. Compared to other materials this fluoropolymer has more light loss, but still the recovery is acceptable when inexpensive light sources are used. These optical fibres can also be incorporated in combination with mirrors, which can reflect the light and also increase the pathlengths [39,41].

3.1.2 Optical- fibres and lenses and their applications

Even though optical fibres seem useful, there are some drawbacks that need to be considered. Specifically, incorrect coupling of the optical fibres with the source, channels, filters etc. can cause a loss in excitation energy [6,42]. Furthermore, optical fibres have a relatively weak focus property, causing unnecessary energy dispersion of the excitation light [40]. This can be improved utilizing optical lenses to focus the light into the channel(s). These lenses can be made of printable polymeric materials (e.g. PDMS), which also avoids a change in RI and thus improves the accuracy of the measurement angles.

An example of a polymeric lens fabrication technique is the off-line thermal curing technique of PDMS based lenses. This principle was applied by Sung et al. who dropped a controlled volume (≈50µL) of liquid PDMS onto a heated surface from a standard height (≈20 mm) [18]. Due to interactions with the heated surface, the PDMS droplet stops flowing and spreading, see Figure 5, and because of its viscosity a plano-convex lens shape originated. This plano-convex shape, its radius of curvature and the width of the microchannels are crucial parameters that contribute to the focussing properties [6,42]. A lens with a smaller radius (≈ 70 µm), decreases the width of the excitation light and thus allows for a better focus with a higher intensity. However, a lens with a bigger radius (≈ 160 µm) would barely improve the focussing and intensity. At a surface temperature of 200°C, the droplet would cure almost immediately with a well-defined curvature, without thermal degradation [18]. If there is a desire to create a lens with a broader diameter, a larger volume can be pipped onto the surface. However, with larger volumes (≈ 200 µL) the curvature of the lens decreases, making it flatter, and it increases small deviations in each lens, which decreases the repeatability.

Figure 5: PDMS lens formation when a droplet of a controlled volume is dropped onto a heated surface from a constant drop height. Adapted from: [18].

(14)

14 An example of incorporating within optical fibres is shown in Figure 6 [42]. Camou et al. created a PDMS device with two parallel optical fibres on both sides of the fluidic microchannel. These fibres had to be as close to the microchannel and the PDMS optical lenses as possible to avoid stray and background light. However, with very low concentrations an increase in the path length will be needed for sufficient absorbance measurements.

Figure 6: Microfluidic device created by Camou et al that incorporates optical fibres and optical lenses for fluorescence detection. Adapted from: [42].

Another example was performed by Jindal et al., who utilized it in combination with LC separation on a quartz microfluidic device, see Figure 7 [43]. However, it is also possible to perform this on optically transparent, organic solvent- and pressure resistant 3D-printed devices, see section 2.5. Furthermore, the stationary phase needs to be optically transparent as well (e.g. silica or PDMS based). After the separation, on chip absorbance detection was performed by focussing a light source with a lens in an optical fibre onto the microchannel. The transmitted signal is then collected by a second optical fibre, at the bottom of the microchannel. This set-up provided a LOD in the micromolar range, but still contained some noise due to the fluctuations from the light source, which could be improved by adding filters between the optical fibres.

Figure 7: absorbance detection by using optical fibres in combination with LC separation within a microfluidic device. Adapted from: [43].

3.1.3 Droplet-Based Systems

Several absorbance detection techniques are based on based systems. However, with droplet-based systems (micro-volumes) there are several parameters that need to be taken into account [33]. Firstly, when working with organic solvents the vapour pressure and the boiling points should be considered, since organic solvents easily evaporate into the exposed air. Furthermore, when solvents are extremely hydrophobic, it is harder to pipette droplets onto a plate.

(15)

15 In 2003 an offline droplet absorption system was created that was based on the surface tension and hydrophobicity of the solvent to keep the droplet in a fixed place between two surfaces [33,44]. The absorption of the droplets (0.5-2 µL) was measured with a nanodrop microvolume UV-Vis spectrometer, see Figure 8. The droplet was exposed to the light source through an optical fibre and the absorbance signal was measured with the receiving optical fibre. The optical path length was determined by the distance between the two stainless-steel surfaces, which were mechanically adjustable. Even though this is a promising technique for droplet detection with the use of optical fibres, it would be too much of a challenge and time consuming to couple this online with a microfluidic device. Therefore, other droplet-based systems need to be considered.

Figure 8: Droplet-based system for absorption detection [44].

Lui et al. also created an absorbance detection method for droplets already in 1996, which was able to measure absorption spectra under a continuous flow (≈0.8 mL/min), see Figure 9 [45]. The sample liquid arrives in the system chamber (plastic tube) through liquid delivery tubes and will create a droplet at the end of the steel rod. As the flow continues, the droplet becomes bigger and will eventually detach from the rod and end up in the waste collector area. Light travels from the source-through the droplet towards the detector, which is guided by optical fibres. Using this technique, there is a risk that not all the incoming light will reach the detector due to reflection and refraction, especially when solid particles are present in the sample. The position of the fibre relative to the droplet (described as h and d, in Figure 9) is significant because it determines the pathlength of the absorption measurement [45]. Ideally, d has to be large to obtain a high absorbance signal, however there is a risk that not all the light from the centre of the drop is collected. The obtained spectra will show maxima and minima, due to the size inconsistency of the droplet. Therefore, absorbance spectra that were obtained with this technique under constant detection needed to be pre-processed to obtain a proper spectrum. The LOD that was provided by this set-up was in the micromolar range with an S/N of 3.

(16)

16 Due to the liquid delivery tubes, this can be a promising absorbance detection technique that can be coupled to a microchannel of a microfluidic device. When multiple channels need to be coupled, it would require multiple delivery tubes and optical fibres. However, when gradient based LC separations are performed, the properties of the organic solvents need to be considered. Since different solvents have different viscosities, the pressure and flow properties will change. This means that there will be a change in the consistency of the droplet formation, which needs to be taken into account.

3.1.4 Bubble, Z and U-cells

On capillary/column detection is a common detection approach that provides minimum band broadening. Unfortunately, the sensitivity is still relatively low due to the short path length in microchannels [35]. To overcome this problem Z-cells, U-cells and bubble-shaped cells were created, which increase the path length and increase the S/N [33]. Unfortunately, bubble cells are expensive and as the diameter increases it causes more dispersion in the electrical field strength during electrophoresis (causing longer retention times, causing a loss in efficiency and resolution). Therefore, it is more beneficial to use Z and U cells, which can be formed by bending a small part of the capillary into a Z or U shape, or by using 3D-printing techniques, see Figure 10 A and B [46,47]. The light from the source and the signal towards the detector can be guided with optical fibres as described in section 3.1.1.

Figure 10: A = Z-cell for optical detection. B = U-cell for optical detection. Adapted from: [46,47].

An example of a set-up in which bubble, Z or U cells can be incorporated is shown in Figure 11, which is in combination with capillary electrophoresis. The basic set-up for a functioning microfluidic electrophoresis device consists of two short intersection microchannels that serve as sample injection and waste channels, see Figure 11 [21]. The long microchannel transports the buffer solution for the EOF flow that is needed for the separation. The material of the chip needs to be optically transparent and electrically conductive, see section 2.5. In the detection area, in- and outcoming light can be guided by optical fibres and the path length is preferably increased with bubble cells [8]. Compared to the other absorption detection methods described in the previous sections, this method provides a simple online detection that can easily be incorporated within 3D-printed devices with multiple channels.

Figure 11: Basic set-up of a functioning microfluidic device for electrophoresis and LC separations that can include electrochemical, absorbance and fluorescence detection [21].

(17)

17

3.2 Fluorescence

When a molecule in a sample absorbs a photon of a certain wavelength and thus a certain energy, it can move from the ground state (S0) and enter the excited state (S1, S2 etc.) [48]. When the molecule is in the excited state it can undergo numerous events when returning to the ground state. The energy can be transferred into the surrounding as heat, or to other molecules present in the solution or environment (quenching). The energy can also go back to the ground state by emitting a photon, which is measurable as fluorescence. This fluorescence signal can be measured with a fluorometer and typically gives high resolution spectra due to its sensitivity. However, the requirement for fluorescence is that the molecule of interest contains a fluorophore, which is a chromophore that can emit a photon. Unfortunately, most molecules do not contain a fluorophore and for this reason these need to be chemically attached to the molecule. Usually this is done by introducing a fluorescent dye to the sample in the sample channel [21,48]. Each individual fluorophore can be characterized by its emission wavelength, intensity and fluorescence lifetime. However, it has to be considered that fluorophores can have an influence on the separation mechanism.

A regular fluorometer consists of a light source, such as a laser or LED lamp, and an excitation monochromator to filter the unwanted wavelengths [48]. Usually when the fluorescence spectrum is recorded, the excitation wavelength is fixed (typically the maximum of the absorbance spectrum). A cuvette holder is utilized to hold the sample and after absorption, the emitted fluorescent signal is filtered by an emission monochromator and eventually detected. All the light is generally guided by lenses and mirrors throughout the system. Considering a microfluidic device, the volumes and the path lengths are significantly smaller, making the detection more of a challenge [49]. However, fluorescence is a sensitive and promising technique and therefore more microfluidic devices with integrated optics are being developed, as will be discussed in the upcoming sections.

3.2.1 Laser Induced Fluorescence

Laser-Induced Fluorescence (LIF) is a highly sensitive fluorescence detection method that can easily be incorporated in microfluidic devices [40]. For LIF, a laser is used for the excitation of the molecules and an optical LIF set-up limits the background noise, minimizes refraction and reflection and can distinguish the excitation light from the fluorescence signal. There are different types of LIF-microchannel conformations, which have already been applied extensively in practice, see Figure 12. Figure 12 A shows a confocal set-up, which gives a relatively high sensitive detection, but is difficult to miniaturize. Figure 12 B and C both have incoming excitation light at a 45⁰ angle but these approaches show a high background contribution due to refraction and reflection. Figure 12 D, E and F show set-ups with a 90⁰ angle between the excitation light and the fluorescence signal, which improved the LOD significantly to nanomolar range. The most important benefit of the 90⁰ angle is that the excitation light is not interfering with the fluorescent signal, thus lowers the background signal [50]. Adding an interference filter can also reduce the background signal from the scattering of the excitation light. Furthermore, additional optical fibres and lenses can be added to transport and focus the excitation light onto the microchannel [40].

However, although not mentioned by the authors, combining these exact conformations within microfluidic devices may cause some complications. This is due to the RI values of the different types of materials and solvents, which can influence the angle of incidence. Moreover, if the material is not (completely) optically transparent, or smoothened by the printer it will cause more scattering, making exact angles difficult to realise.

(18)

18

Figure 12: Different conformation set-ups for laser induced fluorescence measurements in microfluidic devices [40]. E = Excitation source, F = Fluorescence Signal, L = Lens, O = optical fibre

An example of a microfluidic device with integrated optical waveguides and LIF detection with optical fibres was created by Hübner et al. in 2002 [50]. The optical waveguides were formed by three layers on top of a ‘silica wafer’, namely a buffer, a core and a cladding layer. Firstly, the relatively thick (≈ 12-15 µm) buffer layer was added to prevent light interaction with the ‘silica wafer’. Secondly, a germanosilicaoxynitride coated core layer (≈6 µm) with a higher RI (≈1-1.5%) than the buffer and cladding layer was placed on top of the buffer layer to guide the light based on the TIR principle (described in section 3.1.1), see Figure 13 A. The core layer was eventually structured and secured with photolithography and reactive ion etching, see Figure 13 B and C. The last pure silica glass cladding layer was added on top of the core layer to secure the optical waveguides within the microfluidic device, see Figure 13 D.

(19)

19 Once all the layers were added on top of each other, the microfluidic channels had to be etched (reactive ion etching) and structured (photolithography) through the three layers onto the silica wafer [50]. Next, the optical fibres were integrated monolithically within the channels to perfectly align them (linearly), see Figure 14, to avoid stray light [51,52]. The advantage of this basic set-up is that both absorbance and fluorescence detection is possible [50]. When absorbance detection is desired, the light source is focussed onto the channel through coupling of the waveguides with an optical fibre. The absorbance signal is collected at the other side of the waveguide with another optical fibre, which is connected to a photomultiplier detector. When fluorescence detection is desired the fibres have to be rearranged to create an optimal 90⁰ angle (fibre 1 in Figure 14). Another advantage of this set-up is, when all the fibres are permanently attached to the chip, is its stability and transportability. In addition, due to its simple design, it is possible to create such devices with 3D-printing techniques like SLA. Furthermore, SLA can print grooves inside the chip to hold optical fibres, which are then pressed (manually) into the grooves and aligned with the channels [28,53].

Figure 14: Fluorescence detection set-up with integrated waveguides and detection possible through optical fibres connected to a laser and a PMT [50].

3.3 Multichannel Optical Detection

As discussed in the introduction, two and three-dimensional liquid chromatography separations are especially useful for complex samples and it is possible to perform this on 3D-printed microfluidics. However, integrating optical detection techniques for multichannel microfluidic devices is a challenge and causes several issues. But there are possibilities to avoid these obstacles, as will be discussed in this section.

3.3.1 Optical Cross-Talk

High background interferences (from outside and within the system) are common with microfluidic fluorescence detection [26]. To avoid background interferences from outside the microfluidic device, the device can be isolated in the dark. Unfortunately, it is more difficult to prevent background interferences from within the system. This is due to the polymeric material of the device, which can reflect and refract some of the excitation and fluorescent light throughout the device and between (adjacent) channels. This interferes with the intensities of the different signals and is called ‘optical cross-talk’. Furthermore, this background noise increases when the distance between the microchannels is smaller. This means that optical cross-talk would be inevitable for three-dimensional microfluidic LC separations in the xyz-space.

(20)

20 This type of background noise is random and not consistent. [26]. However, this can be avoided optically separating the microchannels from each other. This diminishes the optical-cross talk effect and avoids complex fluorescence spectra. Firstly, air gap channels can be incorporated in between the microchannels, but this was proven to be sufficient only when the distance between the channels was relatively large. Secondly and more effectively, black silicone rubber absorbers can be incorporated between the channels, which will absorb the scattering light.

3.3.2 Multichannel optical detection in practice

Lynch et al. created UV detection multichannel device that was able to measure the absorbance of 12 channels simultaneously, see Figure 15 [2]. Here, the UV light source was filtered with a monochromator, focussed on the entrance and then split into multiple light sources with optical fibres. There were 13 optical fibres in total, 12 for the microchannels/capillaries and 1 as a reference. The light from the optical fibres is focussed onto the microchannels/capillaries with a lens and the absorbance is measured with a photodiode sensor. Additional filters could be applied to amplify the obtained signal. The photodiode sensor was placed on a printed circuit board and the electronic system was connected by the authors.

Figure 15: Multichannel absorbance detection created by Lynch et al. which uses multiple optical fibres focussed onto individual microchannels. Adopted form: [2].

It is important that the optical fibres are positioned precisely and symmetrically to reduce stray light and background noise. Furthermore, it is crucial that the electronics on the printed circuit board are connected precisely for minimal noise contribution. In addition, a metal cover for the printed circuit board that is grounded with the microfluidic device can also improve noise reduction. This set-up was applied for two-dimensional LC separation and tested with 100 µg/mL lysozyme and detected at 220 nm. The noise level per channel was similar to commercial absorbance detectors. Moreover, a linear relationship between the absorbance and the concentration (according to Lambert Beer law, see equation 1) could be determined with an R2 _{value of 0.9996 between 3-1000 µg/mL lysozyme.}

(21)

21 LIF imaging detection is also possible with multichannel separations in the xy-plane, as was performed by Das et al., who used it for protein separation based on isoelectric focussing and gel electrophoresis [54,55]. Laser light was focussed with a confocal fluorescence system, guided by mirrors and expanded by a beam expander. The column of light is focused by a cylindrical lens to form a line across the multiple channels, see Figure 16. The fluorescence signal that is produced is filtered by a band-pass filter and detected by a CCD camera.

Figure 16: LIF multichannel detection with a confocal fluorescence set-up [55].

However, despite that the authors claim that multichannel detection was possible, it was not performed on highly complex samples [54,55]. Considering complex mixtures with multiple fluorescent analytes in different channels, makes it more difficult to determine from which channel a signal is originating. In addition, the widths of the fluorescent peaks will easily overlap [56]. Therefore, it is more efficient to integrate optical fibres into the microchannels for a more accurate measurement. Fortunately, Amirkhanian et al. were able to combine this principle with multi-capillary electrophoresis system for DNA analysis in the xy-plane [38]. In this system, a LED light source was connected to multiple optical fibres, similar as Figure 15, which were focussed onto the capillaries with micro-ball lenses. It is possible to measure the fluorescence signal with a 180° angle between the excitation source, as is often applied with multiple parallel capillaries in Capillary Array Electrophoresis (CAE) [57]. Unfortunately, a 180° angle does give a lot of scattering and background signal, which can be removed by additional spatial- and spectroscopic filters or simply be avoided by using a slightly different angle. Overall, this subsection showed that there are various possibilities to perform optical detection of multiple channels in the xy-plane. However, no literature for optical detection in the xyz-space was found. Applying fluorescence imaging techniques with a CCD camera would provide fast measurements in the xy-plane of multiple channels. When spatial separations in the xyz-space are performed, it becomes more challenging, but possible strategies could be off-line droplet-based imaging systems (similar technique as will be discussed in section 6.2). However, for accurate measurements with complex samples utilizing optical fibres is highly recommended. Particularly, because of theirbroad measurement angle possibilities. However, this becomes more difficult when individual channels within the core of microchannel cube in the xyz-space need to be detected (especially due to optical cross-talk). Therefore, integrating multiple optical fibres within an optically transparent device (requires printed grooves) at individual and optically separated microchannels would be an imaginable strategy to provide sensitive and selective results.

(22)

22

4 Electrochemical detection

Other detection methods that can be miniaturized are different modes of ElectroChemical detection (EC) and have been applied extensively in combination with microfluidic electrophoresis due to its sensitivity and selectivity [5]. EC is more sensitive than optical detection methods (e.g. absorbance), because it is not dependent on the path length and thus not influenced by the size of the microchannels. Furthermore, the background current is reduced when working with microfluidics, which increases the S/N to a nano-molar range. Moreover, analytes can be detected without chemical or other derivatizations, but it does require a redox reaction.

When applying EC detection, three types of electrodes need to be incorporated into the microfluidic device [58]. The first one is the Working Electrode (WE) which measures the redox reactions inside the microchannels. The second one is the Reference Electrode (RE) and the third one is influenced by the potentiostat which is the Counter Electrode (CE), giving a balance current and to set the potential for the WE. The CE and WE are in direct contact with the solutions within the microchannels and the RE is in electrical contact with a salt bridge. The current difference that is measured between the WE and RE relates to the electric potential and by making use of the Nernst equation, see equation 2, the concentrations of the analytes can be calculated. Since both EC detection and electrophoresis separation need electrodes to provide the required currents, it is crucial that these electrodes are separated from each other. If this is not done correctly, it will cause additional noise in the spectra and it can cause damage to the potentiostat [59].

𝐸 = 𝐸0+ 𝑅𝑇 𝑛𝐹 ln (

𝐶𝑂

𝐶𝑅

)

Equation 2: Nernst equation. E = electric potential, R = gas constant, T = temperature, n = number of electrons that are transferred, E⁰ = standard potential, CO = concentration oxidiser, CR = concentration redactor. [58] The material of the electrodes is also a crucial parameter for the selectivity for different types of analytes [59]. Typically, the electrodes are metal based such as gold (Au), platinum (Pt) and palladium (Pd), which are selective for thiols, oxygen species and carbohydrates. Their detection is possible due to the interaction with the empty d-orbitals of the metals, which forms EC active compounds [60]. Silver/SilverChloride (Ag/AgCl) electrodes are frequently used as a RE. Furthermore, it is possible to use carbon-based electrodes that are selective for organic analytes (especially amines and aromatics). Integrating these electrodes within the microfluidic device can be done in various ways. Traditionally, the electrodes were placed onto a glass slide instead of directly onto the polymer (such as PDMS), because it caused the polymer to crack [19,58,61]. Usually the electrodes are then sealed for protection and to prevent any leakage. However, present-day 3D-printers (such as SLA) can print grooves within the microfluidic devices for the integration and removal of electrodes. When 3D-printing techniques are used, the choice of (polymer) material is crucial. The material needs to be resistant against high potentials, joule heating and preferably needs to be optically transparent while maintaining the respectable flow properties [5]. PMMA and PDMS are polymers that are cheap, have a good electrical conductivity and can generate a stable EOF [62]. Unfortunately, bonding/sealing all the PMMA substrates together to eventually form the network of channels is quite difficult, which results in a low reproducibility and consistency. During the sealing/bonding process, the use of multiple materials needs to be avoided since this can have a negative influence on the separation and can cause dispersion.

(23)

23

4.1 Amperometric Detection

Amperometry is an EC detection mode that is most common on a microfluidic scale [5]. This type of detection applies a constant potential between the WE and the CE, which encourages redox reactions of the analytes and then measures the current as a function of time [5,58]. This current can be related to the concentration of the analytes. The advantage of amperometry is that there is minimal background current contribution and the selectivity can be improved by tuning the potential. In addition, it is relatively inexpensive and the system can be isolated from the electrophoretic high voltage [63].

As was mentioned before, amperometric detection is often used in combination with (capillary/microfluidic) electrophoresis (basic set-up is described in section 3.1.4, in Figure 11) and the detector should not be in contact with voltage used for the separation [5]. There are several solutions for this problem such as the position of the WE within the microchannel incorporating salt bridges within the microchannels; or use a decoupler that separates the electrophoretic current from the electrochemical cell [58].

Separating the WE can be done in various ways: End-channel detection (on and off-chip), Off-channel detection and On-channel detection, see Figure 17 [5,59]. When the WE is placed at the end (≈5-20 µm of the end) of the microchannel, see Figure 17 A and B, the separation potential has enough time to decrease before it reaches the WE. Unfortunately, this method generates a high background current, has a lower S/N and a low separation efficiency due to the distance between the electrode and the channel exit. Another WE position is inside the channel, see Figure 17 C [5,59,64]. This shortens the distance between the exit of the channel and the electrode, resulting in less band broadening. The WE is than biased by the voltage of the separation, but it is still possible because the potentiostat is not grounded to the earth [59]. The last WE position is off the channel, which requires a (metal based, e.g. Pd) decoupler to separate the separation and detection voltages from each other, see Figure 17 D [5,59]. After the decoupling, the analytes are pushed towards the WE along with the remaining EOF. Compared to the other configurations, off-channel detection provides a sensitive signal and significantly reduces the band broadening of the peaks.

Figure 17: Scheme of the different WE positions within a microfluidic device. A = channel detection on chip. B = End-channel detection off-chip. C = In-End-channel detection. D = Off-End-channel detection with a decoupler [5].

(24)

24 Typically, a constant potential is applied, but this can cause accumulation of carbonaceous material onto the metal based electrodes, which causes an unstable signal [65]. To avoid this from happening, alternative Pulsed Amperometric Detection (PAD) can be applied. García et al. utilized PAD with a PDMS microfluidic device with electrophoresis-based separation of carbohydrates, sulphurs, amino acids, glycols and alcohols, see Figure 18. In this case, first a high positive potential is applied which cleans the surface of the electrode (≈0.05 seconds), secondly a negative potential is applied to reactivate the electrodes (≈0.025 seconds), lastly a moderate potential is applied to detect the analytes (≈0.15 seconds). The currents could be measured with a PAD potentiostat or with regular EC detector, but this will contribute more baseline noise during PAD. The electrodes were positioned in end-channel detection mode with a gold WE inside an individual electrode channel and a platinum RE. Furthermore, a 1-MΩ resistor was added to avoid joule heating. The PAD measurements were successfully performed with different conditions and it was proven that it was possible to measure micro-molar concentrations.

Figure 18: PDMS microfluidic chip for electrophoresis based separations and PAD detection [65].

Amperometric detection is also suitable in combination with microfluidic LC separations as was performed by McEnery et al. in 2000 [66]. The microfluidic chip was created with photolithographic techniques onto a silicon wafer with n-octyltriethoxysilane chemically modified microchannels to perform reversed phase LC. In their design, both UV detection and EC detection were possible. However, due to the small microchannels, a smaller sample injection volume was required, making UV detection less suitable, as is in agreement with section 3.1. The electrodes for the EC detection were based on an Au WE and CE and an Ag/AgCl RE, which were integrated onto a glass plate. The WE was positioned upstream and the CE was placed close to the exit of the channel, see Figure 19. Fortunately, this set-up was able to provide a good LC separation and a respectable linear relationship between the EC signal and the concentration of the analytes (R = 0.989).

(25)

25

4.2 Potentiometric and Conductivity Detection

Potentiometry is another mode of EC detection, which can detect low ion concentrations by using ion-selective microelectrodes [34]. When the ion enters the lipophilic membrane of the detector, it will cause a potential difference which will produce a signal. Using the Nernst equation, see equation 2, the ion concentration can be related to the produced signal. Unfortunately, when comparing this method to the amperometric detection, potentiometric detection requires more preparation, more handling procedures and the sensor lifetime is relatively short. Furthermore, it is difficult to detect multiple analytes, which makes it unfavourable for spatial separations of complex samples [5]. The last mode of EC detection is conductivity detection which uses an applied potential on a conductive mobile phase with conductive analytes, or on a conductive mobile phase with non-conductive analytes (or a combination of both) [34]. When the ion of interest passes an electrode gap, a change in potential is measured by two electrodes that are in connection with the electrolyte solution. This signal can be related to the concentration of the ion of interest. This type of detection is especially used in combination with capillary electrophoresis with LOD values between 10-6_{and 10}-7 M. However, compared to amperometric detection, conductivity detection is also not frequently used in microfluidic electrophoresis, but only in capillary electrophoresis.

4.3 Multichannel EC Detection

A concept that has been under investigation for multichannel amperometric EC detection is the InterDigitated Ultramicroelectrode Array (IDUA), which uses two electrodes with several electrode fingers to measure redox reactions, see Figure 20 [67]. This electrode configuration can provide a higher S/N, more stable signal and can perform quantitative analysis (LOD in µM range). The crucial geometrical parameters of this device are the ratio of the space between the electrodes and the width of the electrodes fingers, electrode height, number of ‘electrode fingers’ and electrode material. The material of the electrodes was usually metal based (Au or Pt), since silicon based electrodes often cause short-circuit when in contact with organic solvents or buffers [67]. Furthermore, a large number of high, narrow electrode fingers that are positioned close to each other provide a higher S/N, but with a higher background signal.

Figure 20: Configuration of IDUA for multichannel detection [67].

The IDUA configuration with gold electrodes was applied as a biosensor on a PMMA microfluidic device by Nohnoot et al. in 2013 [68,69]. In this case, five microchannels were aligned parallel with the electrode fingers, but coupling more microchannels was also possible. It was crucial that the IDUA and the PMMA device were sealed together correctly (thermally or with UV light with additional pressure) to avoid short circuiting and leakage. After a constant current was reached, the measurements could be performed. It was proven that this technique provided reproducible results with relative standard deviations below 5% and a LOD in the µM range.

(26)

26 Another concept that has been investigated more recently, was performed by Moreira et al., who created a multichannel PDMS/glass device with a fully-integrated single working electrode for amperometric EC detection [70]. The fully-integrated single working electrode was shared by an array of five microchannels, which was possible because electron transfer only occurs at the surface of the electrode. The authors designed two different lay-outs for the electrodes onto the microfluidic device, see Figure 21. In the first design, see Figure 21 A, each channel was connected independently to a WE, while the RE and the CE were connected to a single outlet reservoir. In the second design, see Figure 21 B, each channel is individually connected to all three (WE, RE and CE) metal-based electrodes. Furthermore, it was important that there are no leakages between the microchannels and the connected electrodes. Therefore, the solvent compatibility and conductivity of the material of the device needs to be taken into account, and the electrodes need to be sealed within the device (usually sealed with SiO2 films). The sealing process also determines the active area of the electrodes, which is crucial since it determines the amount of current that is obtained.

Figure 21: Altered form: A = single outlet reservoir with RE and CE and each channel connected independently to a WE. B = five independent channels, each connected to the RE, WE and CE. (RE = reference electrode, WE = working electrode, CE =

auxiliary electrode) [70].

The authors used these designs for quantitative amperometric measurements and were able to measure a calibration curve with a standard solution with a known concentration of the analyte of interest (LOD in µM range) [70]. Comparing the results of two designs from Figure 21, it is better to use a microfluidic system without a reservoir to avoid back-flow interferences, which negatively influence the separation. Furthermore, the design without a reservoir has a better resistance against leakage and deformations of the microchannels, especially when pressure is applied.

In short, this subsection showed that configurations, such as IDUAs and a fully-integrated single working electrode, can be applied for amperometric EC detection in the xy-plane within microfluidic devices. Unfortunately, EC detection in the xyz-space has not been mentioned in literature thus far. However, using the information from this section, it can be stated that amperometric EC detection proved to be most suitable for detection in the xyz-space. Furthermore, integrating the (usually metal-based) electrodes within the channels requires proper electrode positions and connections with the potentiostat(s). In addition, the detection electrodes need to be separated from the electrodes that provide the electrophoresis-based separations. The material of the microfluidic devices needs to generate a stable EOF and be compatible with high potentials. Thus, if these requirements can be met and the electrodes are sufficiently connected, EC detection in the xyz-space should be possible.

(27)

27

5 Vibrational spectroscopy

5.1 Raman

Raman spectroscopy is form of vibrational spectroscopy that has been applied frequently for qualitative and quantitative analysis [72]. The sample is irradiated with laser-light with a certain wavelength and thus a certain frequency. This causes the molecule to be a promoted to a virtual excited state, which does not depend on the incoming wavelength. When the molecule returns to the ground state it will cause a scattering signal. When the frequency of the scattering light is exactly the same as the incoming light, it causes Rayleigh scattering. When the frequency of the scattering light is different from the incoming light, which is only a small fraction, a Raman signal is formed. When the scattering light frequency is lower, a Stokes signal is formed and when it is higher, an anti-Stokes signal is formed.

Unlike InfraRed (IR) spectroscopy, the molecule does not require a change in dipole moment, but only a change in polarizability [72]. Also the Raman scattering of water is low, hence is favourable for aqueous samples and it needs little to no sample preparation [72,73]. In addition, the selectivity and the sensitivity of normal Raman spectroscopy can be enhanced with Surfaced Enhances Raman Spectroscopy (SERS), which is not possible with infrared spectroscopy and therefore only Raman detection techniques will be discussed in this section [72].

A regular Raman spectrometer contains a light source, filters (a band pass filter to clean up the laser, a monochromator, and a notch filter), cooled photomultiplier tubes and a detector (a photodiode array, CCD) [72]. Usually, the light source is focussed by (glass) lenses and travels through the spectrometer via (dichroic) mirrors. Unlike optical detection, Raman spectroscopy is still suitable when working with microfluidic devices and can still provide enough sensitivity even when dealing with small volumes and small path lengths [73]. The microfluidic devices can be printed with optical transparent polymers, but these structures are less rigid and it causes more background noise. However, it is possible to incorporate a Raman transparent window into the microfluidic device to avoid this background noise.

5.1.1 Confocal Raman Spectroscopy

Several parameters need to be considered when incorporating Raman spectroscopy into a microfluidic device [73]. The first parameter is the focussing of the excitation light onto the small volumes inside the chip with minimal background scattering. The second parameter is the high power of the laser, since it can damage the sample and the microfluidic device. The last parameter that needs to be borne in mind when working liquid samples is that some droplets tend to stick onto the microchannels walls under static conditions [73]. These droplets can create a permanent Raman background signal that is called the memory effect. To avoid this effect, the device needs to be disposable, the walls need to be cleaned (without damaging the device with the solvent) or a continuous flow within the device needs to applied. Furthermore, a continuous flow decreases the heat production from the laser and it gives an average profile of the spectra which is better for the reproducibility [73,74].

Focussing the laser beam onto the microchannel is possible with an objective lens within the Raman-microscopy set-up, see Figure 22 [73]. When working with microfluidics, it is important that lower magnification objective lenses are used, because these have a longer focal distance that can penetrate through several layers into the sample, which increases the S/N. The signal that is backscattered from the sample can be collected with the same lens and is filtered with a Notch filter (Rayleigh scattering removal). Through the pinhole a small fraction of the signal will eventually enter the spectrometer.

(28)

28

Figure 22: Raman-microscopy set-up for microfluidic devices [73].

When working with droplet-based systems (in or outside of the microfluidic device) the shape of the droplet is a crucial parameter, because it affects the backscattering light through the lens into the spectrometer, see Figure 23. One of the reasons for this effect is that there is a difference in RI between the different materials and the solvent(s). To ensure minimal distortion, the laser excitation laser light should enter the device through a flat, smooth and optically transparent surface. When open microfluidic devices are used, the droplet should be as flat as possible and, therefore, the width of the microchannel should be considered.

Figure 23: effect of the droplet size onto the backscattering light [73].

5.1.2 Fibre Probe-Based Raman Spectroscopy

The disadvantage of confocal Raman spectroscopy is that the microscopy set-up collects a spectrum through the entire microfluidic device, which can cause high background signals [75]. To avoid this, a fibre probe-based Raman spectroscopy detection method was invented by Ashok et al., see Figure 24, who measured the urea concentration in urine. Two separate fibres are inserted directly into the microchannel for excitation and detection, which is relatively similar to the optical fibres described in section 3.1.1. It is important that the probes are connected successfully without any leakage and are aligned sufficiently. Using two separate fibres allows for a more flexible set-up, which makes excitation and detection from different angles possible. Similar to fluorescence detection, a 90° angle results in the lowest (fluorescent) background signal and is consequently used in this example. In addition, filters, such as a bandpass filter and a long pass filter, can be incorporated within the probes. Unfortunately, this method obtains a two times lower detection limit compared to confocal Raman spectroscopy. However, still a minimum detection limit of urea 0.15 M with an acquisition time of 5 seconds was reached, which made it suitable for measuring the physiological level of urea in urine. Furthermore, the resolution of the spectra was not influenced by the flow rate of the system and can thus be used in online LC separation measurements.

On-chip Detection for Microfluidic Devices

Literature Thesis