MSc Chemistry
Analytical sciences
Master Thesis
Photoacoustic Spectroscopy
An overview of the theoretical background and recent developments
of the technique
by
Rico Singer
2630061 (VU) /11773782 (UvA)
November 2019
12 EC
February 2019 - November 2019
Supervisor/Reviewer:
2
ndreviewer:
V
RIJE
U
NIVERSITEIT
A
MSTERDAM
– F
ACULTY
OF
SCIENCE
A
BSTRACT
This literature study was performed at the Faculty of Science of the Vrije Universiteit as part of the masters track ‘Analytical sciences’ at the Universiteit van Amsterdam (UvA) and the Vrije Universiteit (VU). The aim of the study was to create an overview of the current status of Photoacoustic Spectroscopy (PAS) and the recent developments and new applications reported.
PAS is a spectroscopic technique based on the photoacoustic effect discovered by G.A. Bell in 1881. In the photoacoustic effect, pressure waves are creates due to non-radiant relaxation processes. In PAS, these pressure waves are detected by microphones or other detectors.
This study found that the recent developments and applications of PAS are not focussed on merely one aspect of the technique. For instance, Fourier-Transform Infrared photoacoustic spectroscopy (FTIR-PAS) is broadly applied for the characterization of solid samples by measuring the photoacoustic signal over a broad wavelength range. Especially soil characterization studies were found to successfully test the potentials of using PAS. Another area in which PAS was found to be tested is clinical imaging. Compared to other available imaging techniques, Photoacoustic Imaging (PAI) shows great potential. PAI was found able to produce in vivo imaging of the vascular system in mice brains without any surgery required, up to images with nanometre resolution for imaging organelles. However, with an increasing resolution, the penetration depth is reduced.
Furthermore, several improvements of PAS were discussed. First, two different detection systems designed to replace the conventional microphone were treated. Cantilever enhanced PAS (CEPAS) and Quartz Enhanced PAS (QEPAS) were found to significantly increase the detection limit of PAS systems. Especially for CEPAS, several studies were compared focussing on the detection of different trace gases. These systems showed detection limits in the ppb range, and for some trace gases ppt range detection limits were found. Besides improving the limit of detection, studies were found focussing on the development of portable CEPAS and QEPAS systems usable for in-field and on-site measurements. Compared to other trace gas detection systems, CEPAS shows potential to become a competitor to current conventional gas detection systems.
Finally, Erbium-doped fibre amplification (EDFA) was found to be tested for PAS systems (EDFA-PAS). As the photoacoustic signal is linearly related to the power of excitation, increasing the laser power increases the detectability. EDFA is one technique to increase the laser power whilst measuring at very specific wavelengths. These systems were successfully tested to increase the limit of detection compared to systems without EDFA.
To conclude, according to the general overview of the current status of PAS, the technique shows a large potential to become more important for many different applications in the future.
Table of content
Abstract...3 Abbreviations...5 1 Introduction...6 2 Conventional spectroscopy...7 2.1 Absorbance Spectroscopy... 7 2.2 Fluorescence spectroscopy... 7 2.3 Phosphorescence spectroscopy... 8 2.4 Raman Spectroscopy... 82.5 Fourier Transform Infrared spectroscopy...8
2.6 Cavity ring-down spectroscopy... 9
3 The fundamentals of photoacoustic spectroscopy...10
3.1 A brief history... 10
3.2 The photoacoustic effect... 11
3.3 The photoacoustic spectroscopy experiment...12
4 Fourier Transform Infrared – Photoacoustic Spectroscopy (FTIR-PAS)...16
4.1 Introduction... 16
4.2 Soil characterization... 17
4.3 Differentiation of biological samples...20
5 Cantilever enhanced PAS (CEPAS)...22
5.1 Introduction... 22
5.2 Multilayer graphene cantilever... 26
5.3 Tuneable CEPAS systems for multicomponent analysis...27
5.4 Portable CEPAS system... 28
6 Quartz Enhanced PAS (QEPAS)...33
6.1 Introduction... 33
6.2 Portable QEPAS gas sensors... 34
7 Erbium-doped fibre amplified photoacoustic spectroscopy (EDFA-PAS)...36
7.1 Introduction... 36
7.2 Fibre-ring Laser intracavity EDFA PAS...38
7.3 Stimulated emission EDFA QEPAS... 41
7.4 Comparison of the EDFA systems discussed...42
8 Photoacoustic imaging...45
8.1 Introduction... 45
8.2 Competing spectroscopic imaging techniques...45
8.3 Varieties of photoacoustic imaging techniques...47
9 Conclusion and recommendations...54
A
BBREVIATIONS
ADM Acoustic Detection ModuleBTX Benzene, Toluene and o-, m- and p-Xylene
CEAS Cavity-Enhanced Absorption Spectroscopy
CEPAS Cantilever- Enhanced Photoacoustic Spectroscopy
CERPAS Cavity-Enhanced Resonant Photoacoustic Spectroscopy
CM Confocal microscopy
cw continues wave
DAQ Data Acquisition card
DCM (dye Laser) 4-dicyanomethylene-2-methyl-6-p-dimethylaminostyryl-4H-pyran
EDFA Erbium-doped fibre amplifier
FBG Fibre Bragg Grating
FBGA FBG analyser
FC Fibre Collimator
FLI Fibre-Ring Laser Intracavity
FTIR Fourier Transform Infrared spectroscopy
ISO Optical Isolator
IVM Intravital microscopy
L Lens
LIU Laser-induced ultrasound
MDL Minimum Detection Limit
MLG Multilayer graphene
mR Micro Resonator
NEC Noise equivalent concentration
OPO Optical parametric oscillator
OPS Orthogonal polarization spectral imaging
PACT Photoacoustic computed tomography
PAE Photoacoustic endoscopy
PAI Photoacoustic Imaging
PAM Photoacoustic microscopy
PAS Photoacoustic Spectroscopy
PAT Photoacoustic Tomography
PCA Principal Component Analysis
PLSR Partial Least Squares Regression
ppm/ppb/ppt Parts per million / parts per billion / parts per trillion
PZT Piezoelectric Transducer
QEPAS Quartz-Enhanced Photoacoustic spectroscopy
QTF Quartz Tuning Fork
RSI Raman Spectral Imaging
S/N Signal-to-noise ratio
SHG Second harmonic generation (microscopy)
TA Transimpedance Amplifier
THG Third harmonic generation (microscopy)
TPM Two-photon microscopy
TRSD Time-resolved stress detection
1
1
I
NTRODUCTION
Photoacoustic spectroscopy (PAS) is, compared to several other spectroscopic techniques, relative unknown to the general public. This in spite of the fact that the photoacoustic effect, on which this technique is based, was already discovered by A.G. Bell in 1881 [1]. In PAS, analytical measurements are performed by a combination of light and sound. Sample excitation is often achieved by pulsed or continuous wave (cw) lasers and the pressure waves that are formed subsequently are measured by microphones or other sensors [2].
Over the last decades, various studies were performed focussing on the improvements and applications of PAS. Consequently, PAS has potentially changed a lot over the last decades. Therefore, the Faculty of Science of the Vrije Universiteit would like to create an overview of the current status of PAS and the applications and improvements that were reported so far. To create this overview, a literature study was assigned as part of the masters track ‘Analytical Sciences’ of the Universiteit van Amsterdam (UvA) and the Vrije Universiteit (VU).
First, several conventional spectroscopic techniques are described in chapter 2. The goal of this chapter is to create an understanding of the techniques to which PAS is compared in this study by listing the advantages and disadvantages of these techniques. Hereafter, an overview is presented of the history of PAS, the physics of the photoacoustic effect and general PAS measurements in chapter 3. In chapter 4, Fourier-Transform Infrared – Photoacoustic spectroscopy (FTIR-PAS) is treated. This application of PAS is often used for the analysis of solid samples and shows promising results for, amongst others, soil characterization. Hereafter, chapters 5 and 6 discuss Cantilever-Enhanced photoacoustic spectroscopy (CEPAS) and Quartz enhanced photoacoustic spectroscopy (QEPAS). Both of these technique focus on improving the detection system of PAS. However, besides improving the limit of detection, studies are discussed focussing on developing portable PAS systems for in-field or on-site measurements. Chapter 7 treats Erbium-doped fibre amplified photoacoustic spectroscopy (EDFA-PAS). This technique focuses on improving the laser power used in PAS experiments. Finally, chapter 8 is dedicated to Photoacoustic imaging (PAI) which could be used for clinical applications.
2
2
C
ONVENTIONAL
SPECTROSCOPY
2.1 A
BSORBANCES
PECTROSCOPYThe most well-known and basic form of (analytical) spectroscopy is absorbance spectroscopy. In this technique, a sample is excited with a light source of a specific wavelength(s). Depending on the extinction coefficient of the analyte for the chosen wavelength, the pathlength the light travels through the sample and the concentration of the analyte, a certain amount of photon energy will be absorbed by the sample. By measuring the intensity of the light beam after the sample to its starting condition and by using the Lambert-Beer law (see equation 1 ( )), the mentioned sample characteristics can be determined. Here, I is the measured signal intensity, I0 is the intensity of the signal before entering
the sample, ε is the extinction coefficient which is unique per analyte for a specific wavelength, c is the concentration of the analyte, and d is the pathlength the light travels through the sample. Equation
1 ( ) can be re-written to equation 2 ( ). Using this equation, the absorbance (A) is used to determine the concentration of a compound. [3]
(
SEQ Equation
¿
1
) I=I0∙ 10−εcd(
SEQ Equation
¿
2
)A=log
(
I
0I
)
=
εcd
In infrared experiments, the wavenumber is traditionally used instead of the wavelength. Therefore, equation 1 ( ) is altered for infrared experiments giving equation 3 ( ). [3]
(
SEQ Equation
¿
3 )I(ν )=I
0(
ν )∙ exp
ε (ν)cd2.2 F
LUORESCENCE SPECTROSCOPYAfter the analyte absorbed the photons energy and gets into a certain excited state (~ 10-15 s), it could
lose (part) of its energy by emitting a photon with a specific wavelength. After part of the energy is released due to internal conversions and vibrational relaxation, the analyte gets into its first singlet excited state (S1). Hereafter, the remaining energy can be released by emitting a photon. The
wavelength of this photon is dependent on the molecule itself and the surrounding of this molecule. This phenomena, called fluorescence, occurs in about 10-9 – 10-7 s after the excitation took place. On
average the limit of detection (LOD) of fluorescence spectroscopy is a factor 10 – 100 improved compared to absorption spectroscopy. However, not all compounds can be measured using fluorescence spectroscopy as they are by nature non-fluorescent. This can be altered by fluorescent tags, which can be added by chemical derivatization [3].
2.3 P
HOSPHORESCENCE SPECTROSCOPYPhosphorescence spectroscopy can only be applied for a very limited amount of compounds that show phosphorescence. To show phosphorescence, an excited molecule experiences so called intersystem crossing. As a result, the molecule gets into its first triplet excited state (T1). Compared to transition
from S1, which occurs within 10-9 – 10-7 s after excitation, the transition from T1 to the ground state
occurs 10-3 up to 1 second after excitation [3].
2.4 R
AMANS
PECTROSCOPYIn Raman spectroscopy, the vibrational spectrum of molecules is used for sample characterisation. This includes, but is not limited to, crystallization form and the redox state of a sample. Inelastic scattering of photons by the analytes is used to obtain this information. Normally, a scattered photon loses no energy during a scattering event (elastic scattering. However, 1 out of 107 photons lose part of
their energy to a vibrational state of the analyte (inelastic scattering). The energy difference of the photons prior and after inelastic scattering corresponds to a vibrational state of the analyte [3]. However, the Raman signal often has a low intensity due to the low probability of inelastic scattering. Consequently, if an analyte produces fluorescent signals, this will cause severe interference with the Raman signal. This can be solved by using an excitation laser with longer wavelengths (infrared) as these are often too weak to create fluorescence signals [3].
2.5 F
OURIERT
RANSFORMI
NFRARED SPECTROSCOPYFourier Transform Infrared (FTIR) Spectroscopy is an absorption method conventionally used for gases. A typical FTIR setup used for the analysis of gases consists of four parts. (1) an infrared source, which can be a laser or a black body radiator. (2) A spectrometer or a filter, which is required to ensure that the measurement is performed at the intended wavelength. Multiple options are available here, like an interferometer, a dispersive monochromator or optical filters. However, the usage of a spectrometer or filter is nonessential if a tuneable laser is used as an IR source. (3) A sample cell containing the sample gas. (4) The IR detector [4]. Figure 1 shows such a conventional setup for a gas analysis using FTIR. The concentration of a certain gas is determined by measuring the absorption (see equation 2 ( )) at a certain wavelength or wave number. For gases, FTIR techniques show
detection limits in the ppm range [5], although there are instrument manufacturers claiming to measure ppb levels in high-purity gas samples [6].
However, further improvement of the sensitivity of FTIR detectors will have minor effects. This is caused by the effect that most FTIR systems already operate close to the theoretical limit of infrared detectors. Consequently, FTIR systems will not gain much higher sensitivity in the future [4].
Figure 1 the conventional low-resolution Fourier transform infrared (FTIR) spectrometer used for gas measurements. Image taken from [4]
2.6 C
AVITY RING-
DOWN SPECTROSCOPYIn cavity ring-down spectroscopy, the analyte is placed inside a chamber with two highly reflective mirrors on both sides. Consequently, the excitation (laser) beam is trapped inside the cavity and passes through the sample a multiplicity of times compared to traditional absorbance spectroscopy. This way, the effective path length mentioned in the Lambert-Beer law (equation 1 ( )) can be increased up to a kilometre [3]. Although able to detect analytes in the ppb range [7], cavity-ring down spectroscopy requires careful alignment of all the optical parts in the system. This makes the system less robust and difficult to implement on industrial sites and in hand held devices.
3
3
T
HE
FUNDAMENTALS
OF
PHOTOACOUSTIC
SPECTROSCOPY
3.1
A
BRIEF HISTORYIn 1881, A.G. Bell described the phenomenon now called the photoacoustic effect [1]. During his study, Bell discovered that a wide range of materials emit a sound when exposed to radiant energy. Although all tested solids created sound, the loudest sounds were created by dark-coloured fibrous materials. Bell tried to explain this by assuming the heat capacity of dark materials and the air-filled cavities of fibres could cause larger air shocks. Besides solids, liquids and gases also showed photoacoustic properties. Although Bell merely described the phenomenon itself, he noticed how the fact that different substances produce sounds of different intensities under similar conditions. He also proposed that the effect could be used to investigate the absorption of infrared light by substances. Since the human eye is unable to see this light, the photoacoustic effect could help to determine if a compound radiates infrared light. A lamp-black material would be placed in the receiver part of his ‘spectrophone’, which would only produce a sound if the material radiates infrared light [1]. The first system based on the photoacoustic effect used to measure mixtures of gas was developed in 1938 by Viengerov. This system used a microphone as a detector and a blackbody as an infrared light source. However, the interest for photoacoustic spectroscopy (PAS) declined after 1940 until the invention of the laser in the 1960’s. [8]
In 1968, more than 80 years after Bell’s discovery, the first photoacoustic analytical device was developed [9].The rediscovery of PAS is often linked to the development of laser technology, since this enabled sensitive and selective gas analysers based on PAS [10]. The first gas analyser was designed to overcome an issue with conventional absorbance spectrometers, namely the intensity loss caused by scattering [9]. Conventional absorbance spectrometers determine the gas concentration by measuring the difference in intensity between the emitted and received light. However, scattering caused by other components could also cause an intensity difference. This would result in higher gas concentration then truly present. The spectrophone used a laser to heat a gas by absorption, after which the pressure difference is measured. Scattering does not contribute to this pressure rise and could therefore not produce false measurements. The ability to measure without disturbance by scattering
light also led to renewed interest to use the photoacoustic effect to measure liquids and solids. In the 1970’s, photoacoustic was successfully used to obtain spectra of (semi)solid materials including crystalline, amorphous and gel types [11].
3.2 T
HE PHOTOACOUSTIC EFFECTAs in other spectroscopic techniques, Photoacoustic spectroscopy is based on the absorption of electromagnetic radiation by molecules. However, conventional absorption spectroscopy uses the difference in intensity between the excitation radiation (I0) and the measured reflected or
transmitted light (I) to determine the absorption. In PAS the absorption of electromagnetic radiation is determined by measuring pressure fluctuations generated as a consequence of the radiation of the sample. The generation of pressure fluctuations, or sound waves, by the irradiation of a sample is called the photoacoustic effect [2]. The photoacoustic effect is the fundamental principle behind PAS. Both light and sound is used in this technique [12]. Pressure fluctuations are created due to non-radiative relaxation processes after irradiation. These processes, of which collisions between molecules is an example, cause local warming in a sample. This causes thermal expansion, creating the pressure fluctuations which are detectable as acoustic or ultrasonic waves [2]. Figure 2 shows the processes occurring from sample irradiation until signal detection. In conventional spectroscopic techniques, low
absorption coefficients, the opaqueness of a sample, and scattering could cause obstruction of your results. In all these cases, PAS could be advantageous. The main advantage of PAS is its capability to determine the absorption coefficient over several orders of magnitude. Consequently, PAS is able to determine both weak absorptions of gases using relative short pathlengths, all the way to the high absorption of opaque solids [2]. Furthermore, classical absorption techniques require a background correction which is unnecessary in PAS. The detected photoacoustic signal is directly proportional to both the incident radiation and the absorption coefficient of the absorbing material [13]. Since the signal of PAS depends on the power of the excitation radiation, the sensitivity can be tuned by the choice of the radiation source, like a lamp or a laser. However, PAS is an indirect measuring technique. Consequently, quantitative measurements require calibration steps in their method [14].
Figure 2 The physical processes occurring in PAS from sample excitation to signal detection. Image taken from [16]
3.3 T
HE PHOTOACOUSTIC SPECTROSCOPY EXPERIMENTA PAS experiment roughly consists of three major parts. First, the sample is excited. Excitation is followed by the signal generation. Finally the generated signal is detected [2].
3.3.1 Excitation
As mentioned, the pressure fluctuation required for PAS measurements are generated by optical absorption. Periodic heating and cooling is required for these fluctuation to occur. There are two ways to do this, either by modulated or pulsed excitation [2].
3.3.1.1Modulated excitation
In modulated excitation, the intensity of the excitation source fluctuates. The form of this fluctuation is either a square or sine wave. Chopped or modulated lamps, or IR sources, are often used to determine UV/VIS or IR absorption of opaque solids. Depth analysis of these solid samples is possible by tuning the modulation frequency. Modulated continuous wave lasers are the most common excitation sources used for the analysis of gases. Using modulated excitation, one single acoustic frequency is excited. The pressure fluctuations created in modulated excitation result in sound waves that can be detected by microphones or other forms of detectors [2].
3.3.1.2Pulsed excitation
In pulsed excitation, laser pulses in the nanosecond range are often used. Due to the short pulses and relatively slow repetition rate, fast and adiabatic thermal expansions are created. The short pressure fluctuations that result from this are recorded by oscilloscopes, boxcar systems or fast A/D converters. Using pulsed excitation, a broad band of acoustic frequencies is excited. Furthermore, this method allows for depth analysis of solid samples, using time delay [2]. This is explained in sub-chapter 3.3.3.
3.3.2 Signal generation
Signal generation is performed either direct, or indirect. With direct signal generation the signal is generated inside a gas, liquid of solid sample. The signal is hereafter detected inside, or at the surface of a sample. However, there is a difference for measuring gases, or liquids and solids. In indirect signal generation, the heat is first generated inside a sample, after which it is transported to an interface. Indirect signal generation is only performed with liquid or solid samples. Figure 3 shows an overview of the different excitation and signal generation methods.
Figure 3 An overview of the excitation and signal generation possibilities in PAS, and the samples state of matter it is performed on. Image taken from [15]
3.3.2.1Direct signal generation for measuring gases
PAS measurements performed on gases are usually performed inside a cylindrical cell. A laser, either chopped cw or modulated, irradiates the gas inside this cell. The pressure waves that are generated are detected by microphones [2]. An example of the PA cell used during these measurements is shown in Figure 4. The amplitude of the signal is plotted against the wavelength responsible for the generated pressure fluctuation [15]. From the measured amplitude, the absorption coefficient of the sample can be calculated according to equation (1) [2].
(1) p=F W0μa
Here, p is the signal amplitude, F is the cell constant, W0 is the incident radiation power and µa is the
absorption coefficient [2]. The cell constant F is determined by equation 5 ( ) [2].
(
SEQ Equation
¿
5
)F=
G (γ−1) L
ωV
Here, G is a geometric factor, γ is the adiabatic coefficient of the gas sample, L is the length of the PA cell, ω is the modulation frequency calculated by ω=2 πv , and V is the volume of the PA cell [2]. According to equation 1 and 2, the amplitude of the signal increases with a longer PA cell, while it decreases with an increasing cell volume. Furthermore, the signal amplitude decreases with an increasing modulation frequency. However, there is an optimum S/N ratio which can be achieved, since the noise increases with an increasing V/L ratio and modulation frequency [16].
These measurements can either be performed in the non-resonant-, or resonant mode. Measuring in the non-resonant mode means that the modulation frequency used is lower than the first resonant frequency of the cell. Therefore, the generated pressure wave is larger than the cell dimension and it is not possible to generate standing waves in the cell. If the modulation frequency is increased, the first resonant frequency of the cell is reached at some point. This leads to an increase of the observed signal as it affects the cell constant f according to equation 6 ( ) [2].
(
SEQ Equation
¿
6 )F
res=
Q
iC
i(γ −1)
ω
iV
Here, Q is a quality factor amplifying the signal under resonant conditions and C is a factor dependent on the position of the laser beam and microphone. As (3) indicates, Q, C and ω are all dependent on the ith eigenmode of the cell. The amplitude increasement caused by Q can be up to 103 [2].
Figure 4 Example of a PA cell used during PAS measurements performed on gases. Image taken from [2]
3.3.2.2Direct signal generation for measuring liquids and solids
For liquids and solids, direct signal generation is only possible with an pulsed excitation source. Here, the amplitude of the signal (p) can be described as shown in equation 7 ( ). Here, β is the thermal expansion coefficient of the sample, c is the speed of sound inside the sample and Cp is its heat
capacity, E0 is the energy of the incident laser and µa the absorption coefficient of the sample [2].
Depth analysis of solids samples with an pulsed excitation source is possible, as further explained in sub-chapter 3.3.3.
(
SEQ Equation
¿
7
)p∝
β c
2C
pE
0μ
a3.3.2.3Indirect signal generation for measuring liquids and solids
Direct measurements of liquid and solids with modulated excitation is impossible due to acoustic impedance mismatches, which causes strong interference. This phenomena is further explained in sub-chapter 3.3.3. Consequently, measurements on liquids and solids with modulated excitation is only possible by indirect signal generation. A well-known example of indirect PAS measurements with modulated excitation is Fourier Transform Infrared PAS (FTIR-PAS) [2]. Chapter 4 of this study is dedicated to FTIR-PAS measurements and its applications.
For both direct and indirect PAS measurements of solids it is often required to continuously purge the sample inside the PAS chamber with dry helium as water can cause severe interference [17]. However, sometimes it is required to dry the sample before placing it, which is often performed by freeze drying [18–21].
3.3.3 Depth profiling in PAS
If pulsed excitation is used for time resolved PAS, the depth (z) from which a signal originates, is calculated by equation
8
( ). Here c is the speed of sound inside the sample and t the time from the laser pulse till the signal detection [2]. In this equation, the time it takes for the laser to reach and excite the sample is neglected. By measuring the signal for multiple values of t, a depth profile is created.(
SEQ Equation
¿
8 ) z=c ∙tAs equation 7 ( ) showed, the signal amplitude is depending on the absorption coefficient of the sample. This coefficient also determined the optical penetration depth (δ) of a sample as indicated in equation 9 ( ) [2].
(
SEQ Equation
¿
9 ) δ=1/µaHowever, depth profiling at values of z higher than δ is still possible by using laser-induced ultrasound (LIU). Acoustic waves have higher decay lengths compared to the maximum penetration depth of laser signals. If a pressure wave traveling through a medium (acoustic impedance) encounters an area where the speed of sound suddenly changes, it experiences an impedance mismatch [22]. Consequently, part of the acoustic wave is reflected. In LIU, this reflected acoustic wave is measured for depth analysis [23]. Figure 5 shows the signal generation and detection of a LIU experiment, which provides information on “sound velocities, modulus of elasticity, film thickness, interface properties and delamination parameters” [23].
Figure 5 Signal generation and detection in laser-induced ultrasound. (a) – Using an ultrashort laser pulse (picosecond range), a photoacoustic signal is generated at the surface of a sample. (b) – The acoustic wave travels through the medium till it reaches an inhomogeneity (c). Here, part of the wave is reflected due to impedance mismatch. (d) – The reflected acoustic wave reaches the surface and is detected .Image taken from
Depth analysis with modulated excitation is also possible, as often performed with FTIR-PAS. This is explained in chapter 4.
4
4
F
OURIER
T
RANSFORM
I
NFRARED
– P
HOTOACOUSTIC
S
PECTROSCOPY
(FTIR-PAS)
4.1 I
NTRODUCTIONFourier Transform Infrared – Photoacoustic Spectroscopy (FTIR-PAS) was developed and studied in the 1980’s [24]. The technique has application in the analysis of surface compositions, chemical structures and depth profiling of solids.
Performing PAS experiments in the mid infra-red region affords surface composition information due to the well-documented absorption by functional groups in this region. However, the mid infra-red sources often provide weak intensities. Therefore, IR techniques are combined with Fourier Transformation to improve signal/noise ratios [24]. For FTIR-PAS, samples with high surface areas provide higher photoacoustic signals. This is thought to be caused by a larger surface to gas contact area. Consequently, more heat is transferred from the sample to the surrounding air, causing larger pressure changes. This provides another advantage of surface analysis by PAS compared to conventional spectroscopic techniques, as these experience more scattering and signal interference for higher surface areas [24].
Figure 6 shows a typical FTIR-PAS setup used for surface and depth analysis [25]. Inside the FTIR interferometer, infrared light is split into two beams. Half of the light is directed to a fixed mirror and half to a moving mirror. Different modulation frequencies (ω) are created for each wavelength by combining these two beams. After the interferometer the beam is focussed on the sample inside the sample holder. A disadvantage of PAS measurements on solids performed is the interference caused by carbon dioxide (CO2) and water vapor (H2O). To eliminate these interferences, the sample cell is
purged by nitrogen (N2) and Helium (He) gas [17]. Consequently, the IR beam passes through an
IR-window to keep the N2 gas inside the compartment. Different detectors may be used to measure the
Figure 6 Typical FTIR-PAS setup for surface and depth analysis of solid samples. Inside the FTIR interferometer, different modulation frequencies (ω) are created for each wavelength. The beam is focussed on the sample inside the sample holder, where the PAS signal is created. The sample is constantly purged with N2 to
eliminate interference by CO2 and H2O. A detector is installed to measure the PAS signal, in this case a
cantilever detector. Image taken from [25]
FTIR-PAS systems are employed for a wide range of application. In this chapter, its use in soil characterization (sub-chapter 4.2) and differentiation of biological samples (sub-chapter 4.3) are discussed.
4.2 S
OIL CHARACTERIZATIONOne application FTIR-PAS is tested for, is the characterization of soil. Standard techniques for soil analysis are chemical tests which are time-consuming [26]. More advance techniques include X-ray microtomography and X-ray fluorescence. However, these techniques are time-consuming, expensive and require high-quality systems [17]. Over the past two decades, infrared based techniques became more promising for soil analysis as they are non-destructive and require little sample for representative data. Yet, they require time-consuming sample preparative procedures and the low transmissivity of soil causes sensitivity issues.These types of samples can still be accurately measured by PAS based systems as it does not measure the absorption or transmittance of a sample for a certain wavelength. Additionally, it allows to perform depth profiling analysis, which could create more insight in the structure, composition and function of soil. Therefore, a number of studies were performed to test if FTIR-PAS is able to replace traditional soil tests. Most of these studies use commercially available machinery and do not propose any new developments for FTIR-PAS. However, they show how FTIR-PAS may be applied in field studies.
A study by C. Du et al. [26] studied the difference between traditional tests versus FTIR-PAS for the determination of available nitrogen (N), phosphorus (P), potassium (K) and organic matter in different soil samples. Traditional tests performed are extraction by KCl for the available N, extraction by NaHCO3 for the available P, extraction by NH4OAc for the available K and wet oxidation for the
Bruker. The PAS samples were purged with He for 10 seconds prior, and throughout to the measurements. Of 56 soil samples, the mentioned characteristics were first determined by traditional tests. 42 samples were used to perform partial least squares regression (PLSR) on wavenumbers corresponding to a specific characteristic of the FTIR-PAS spectra (4000 – 500 cm-1) obtained of all
soil samples. Once a model was created, the remaining 14 samples were used to test the model. C. Du et al. found that their model predicted the available N, P and organic matter with a Root Mean Square Error (RMSE) of 2.36, 2.93 and 0.99 mg/kg respectively which they describe as excellent predictions. However, with an RMSE of 38.09 mg/kg the prediction for the available K is insufficient. PLSR was also applied by C. Peltre et al. [27] and Z. Xing et al. [28] to predict soil characteristics using FTIR-PAS. C. Peltre et al. compared PLSR results obtained for both FTIR-PAS, NIR and a combination of FTIR-PAS and NIR. These characteristics were the total organic carbon (TOC), available N, clay content and content of the labile fraction of the soil organic carbon (SOC). With the exception of the clay content, FTIR-PAS provided improved results compared to NIR. However, PLSR performed using the combined data from both FTIR-PAS and NIR provided the most reliable results. Z. Xing et al. compared FTIR-PAS with Raman for the determination of soil organic matter (SOM) in four different soil samples all originating from China. According to this study, FTIR-PAS also provides more reliable results compared to Raman. However, the best results were obtained by combining the data of Raman and FTIR-PAS.
Another study by Z. Xing et al. [17] focussed on depth profiling of soil by FTIR-PAS. Rosencwaig and Gersho [11] developed a theory for the thermal diffusion length (µ) of the PAS signal in solids, which is dependent on the experimental angular frequency (ω) (see equation 10 ( )
ADDIN CSL
CITATION{citationItems :[{id:ITEM-1 , itemData:{author :[{dropping-particle: , family :Joseph A. Gardella ,given :Jr. , non-dropping-particle : , parse-names: false , suffix : }, {dropping-particle : ,family : Da-Zhen , given: Jiang , non-dropping-particle : , parse-names : false , suffix : }, {dropping-particle : , family : William P ,given :McKenna , non-dropping-particle: , parse-names :false , suffix : },{dropping-particle : , family :Edward M. , given: Eyring , non-dropping-particle: , parse-names :false , suffix : }], id :ITEM-1, issued :{date-parts:[[0]]}, page :36-49 , title : APPLICATIONS OF FOURIER TRANSFORM INFRARED PHOTOACOUSTIC SPECTROSCOPY (VF-IR/PAS) TO SURFACE AND CORROSION PHENOMENA , type : bill, volume:15 }, uris :[ http://www.mendeley.com/documents/?uuid=9360ec2c-7474-43dd-bd25-0b75ecb24ad9]}], mendeley :{formattedCitation:[29] , plainTextFormattedCitation :[29] , previouslyFormattedCitation :[29]}, properties :{noteIndex : 0 }, schema :https://github.com/citation-style-language/schema/raw/master/csl-citation.json }[29]
)(
SEQ Equation
¿
10
)µ=
(
2 K
ρc ω
)
1 2
Here, µ is the thermal diffusion length of the PAS signal (cm), K is the thermal conductivity of the sample (W·cm-1·K-1), ρ is the density of the sample (kg/cm3), c is the specific heat of the sample
(W·s·kg-1·K-1), ω is the experimental angular frequency, or chopper frequency (rad·s-1).
According to equation
10
( ), the thermal diffusion length is small when using high modulation frequencies. Therefore, surface analysis becomes possible. Furthermore, depth analysis is achieved by using variable values of ω. However, equation10
( ) can be altered as shown below to relate µ to the mirror velocity speed (V) and the wavenumber (ύ), giving equation 13 ( ) [17]µ=
(
2 K
ρc ω
)
1 2 (SEQ Equation
¿
11
)ω=4 πV ´v
µ=
(
2 K
4 ρc πV ´v
)
1 2 (SEQ Equation
¿
12
)D=
K
ρc
(SEQ Equation
¿
13 )µ=
(
D
2 πV ´v
)
1 2Here, V is the mirror velocity (cm · s-1), ύ is the wavenumber (cm-1) and D is the thermal diffusivity
(cm2·s-1). Equation 11 ( ) was obtained from
ADDIN CSL
CITATION{citationItems :[{id:ITEM-1 , itemData:{author :[{dropping-particle: , family :Joseph A. Gardella ,given :Jr. , non-dropping-particle : , parse-names: false , suffix : }, {dropping-particle : ,family : Da-Zhen , given: Jiang , non-dropping-particle : , parse-names : false , suffix : }, {dropping-particle : , family : William P ,given :McKenna , non-dropping-particle: , parse-names :false , suffix : },{dropping-particle : , family :Edward M. , given: Eyring , non-dropping-particle: , parse-names :false , suffix : }], id :ITEM-1, issued :{date-parts:[[0]]}, page :36-49 , title : APPLICATIONS OF FOURIER TRANSFORM INFRARED PHOTOACOUSTIC SPECTROSCOPY (VF-IR/PAS) TO SURFACE AND CORROSION PHENOMENA , type : bill, volume:15 }, uris :[ http://www.mendeley.com/documents/?uuid=9360ec2c-7474-43dd-bd25-0b75ecb24ad9]}], mendeley :{formattedCitation:[29] , plainTextFormattedCitation :[29] , previouslyFormattedCitation :[29]}, properties :{noteIndex : 0 }, schema :https://github.com/citation-style-language/schema/raw/master/csl-citation.json }[29]
, relating the angular frequency (ω) to the mirror moving velocity (V). Equation 12 ( ) wasobtained from
ADDIN CSLCITATION{citationItems :[{id:ITEM-1 , itemData:{author :[{dropping-particle: , family :Bryon Bird , given: R , non-dropping-particle : , parse-names: false , suffix : }, {dropping-particle : , family :Stewart ,given : W , non-dropping-particle: , parse-names :false ,suffix : },{dropping-particle: , family :Lightfoot , given :E, non-dropping-particle: , parse-names :false , suffix : }], id :ITEM-1 , issued :{date-parts:[[1960]]}, title: Transport phenomena , type: book }, uris :[http://www.mendeley.com/documents/?uuid=25c2677f-4bd1-4966-8cf1-9d1428346000 ]}], mendeley : {formattedCitation :[30], plainTextFormattedCitation :[30] , previouslyFormattedCitation :[30] }, properties :{noteIndex :0 }, schema : https://github.com/citation-style-language/schema/raw/master/csl-citation.json }[30]
, showing how the thermal diffusivity (D) is calculated. As equation
13
( ) shows, the diffusion lengths now depends on the velocity of the moving mirror, and the wavenumber used to irradiate thesample. However, a paper by C. Du et al [31] uses
µ=
(
D
πV ´v
)
1
2 ,without an explanation how they
derived it. Nonetheless, smaller wavenumbers, or lower frequencies, will penetrate deeper into the sample. During measurements, depth profiling is achieved by modulating the velocity of the moving mirror (V) (see Figure 6). Three different values of V were tested by Z. Xing et al. ; 0.16, 0.32 and 0.63 cm ·s-1, where deeper signals are recorded at lower V according to equation
13
( ). With thesesettings, differences in SOM up to 4 µm from the surface were studied, whereas depth measurements using Raman often reach a maximum depth of 2 µm [32]. Their measurements were performed with a commercially available FTIR spectrometer from Thermo Scientific (Nicolet 6700) [33] with the MTEC model 300 PA accessory [32]. Within their article it is not specified which laser is used as an excitation source. However, in an FTIR manual written by Thermo Fisher the only available source mentioned suitable for the mentioned spectral range is the ETC. No further specification were given or
found about this source [34]. Figure 7 show the FTIR-PAS results for homogenized black soil (no chemical difference between top layer and deeper layers) recorded at the three mentioned V values. To obtain these results, the average signals was taken for 32 successive scans. As these results show, there is a difference between the spectrum recorded at 0.63 cm · s-1 and the spectra recorded at lower
values of V, which show much more similarity. Especially around 2000 cm-1 there is a clear
dissimilarity. These results were also observed for two other soil samples discussed in their study. This observation is explained by a difference in hydrophobicity between the top layer of soil and the deeper layers. This difference was already proven in other studies, and now confirmed by FTIR-PAS. Therefore, they conclude that measuring at different moving mirror velocities (V) is beneficial for comparing soil samples at a micrometre scale.
Figure 7 FTIR-IR results of black soil recorded by Z. Xing et al. for three different moving mirror velocities (V); 0.16, 0.32 and 0.63 cm · s-1. The least optical sample penetration is obtained for the highest V.
4.3 D
IFFERENTIATION OF BIOLOGICAL SAMPLESAnother field in which FTIR-PAS is widely applied is for the differentiation of two or more biological samples. In this chapter, three studies are discussed using FTIR-PAS to distinguish organisms with comparable characteristics based on multivariable statistical analysis.
In one study, FTIR-PAS was used to differentiate multiple fishes of the same species (Astyanax
altiparanae) living in different streams of the Paraná River in South America. For this, they performed
measurements from 4000 to 400 cm-1 on the inside of the scales of the fishes. 11 absorption peaks
were chosen, and the intensities were processed by multivariate statistical analysis. Beside differentiation based on the habitat conditions, it was proven possible to differentiate on the diet of the fishes as well. The advantages of using FTIR-PAS for these measurements is that is requires very little sample preparation. After removing the scales from the fish, which causes little harm to the fish, the dried in a vacuum for at least 12 hours. Hereafter, they can be analysed [18]. Besides differentiation
within a species, FTIR-PAS has also been used to distinguish different species by analysing their scales [21].
Another study used FTIR-PAS combined with multivariate statistical analysis to distinguish male, worker and queen ants (Ectatomma vizottoi). After the ants were caught, their abdomen was removed and placed in a vacuum dryer for 48 hours. Hereafter, direct FTIR-PAS was applied on the dried samples and the intensities of 13 absorption peaks were used. This study also proved that FTIR-PAS is a suitable technique to distinguish biological samples which are genetically seen comparable [19].
Beside differentiation based on habitat or sex, FTIR-PAS has also been applied to identify contaminated plants. Soybean rust, also known as Asian rust, has an estimated economical effect of billions of dollars. Therefore, an early detection of an infection would be very beneficial. A FTIR-PAS method was developed in which a small sample was taken from the centre of a leaf. This tissue was dried for 48 hours before being analysed. The FTIR-PAS spectra (4000 – 400 cm-1) of infected and
uninfected plants were compared. A clear difference was observed in the region 1600 – 1500 cm-1
5
5
C
ANTILEVER
ENHANCED
PAS (CEPAS)
5.1 I
NTRODUCTIONDue to the wide bandwidth response of microphones, they are sensitive to background noises [35]. Therefore, multiple studies focussed on developing new methods for detecting the pressure waves caused by the photoacoustic effect. One of the possibilities is to replace the microphone by a cantilever system. These systems have shown to be able to measure trace gases in the ppt range.
In gas analysis, low resolution Fourier transform infrared (FTIR) spectrometers were known to be the best analytical instruments available. However, during the development of new FTIR gas analysers the theoretical performance limits were being reached as explained in chapter 2.5. This meant that the sensitivity and selectivity could not be improved much further. Therefore, studies were performed to examine the possibility of PAS to replace FTIR, since it is unrestricted by any law of physics. In PAS, the major limit of sensitivity is the pressure sensor. In the early days of PAS, a microphone was used to measure the pressure difference caused by the photoacoustic effect. Although being the most obvious choice to measure sound, it limited the sensitivity of the technique to a point where it was less sensitive than FTIR [4]. This limitation is caused by the physics of the microphone itself. Figure 8 shows the structure of a (standard) condenser microphone. This type of microphone contains a thin membrane which deforms due to pressure waves. Often mylar films are used as a membrane in condenser microphones. These are metal coated and form a condenser together with the other electrode present. Between the membrane and the electrode, a gas is present. The capacitance (C) of the system, calculated by equation 14 ( ), changes (∆C) as a consequence of the deformation of the membrane. ∆C is proportional to the pressure change caused by the photoacoustic effect (∆p), and can be calculated by equation 15 ( ) as long as the deformation in the membrane (∆h) is negligible compared to the total distance h. Here, ε is the dielectric constant of the gas present between the electrodes, A is the area of the electrode and h is the distance between the membrane and the electrode. To increase the sensitivity of the microphone, A must be increased and h must be decreased. However, eventually this reaches a maximum and the system cannot be further improved. Hence, the microphone had to be replaced by a new measurement method in order to improve the sensitivity of PAS systems.
Figure 8 Structure of a (standard) condenser microphone. h is the distance between the mylar film (membrane) and the metal electrode. Image taken from [4]
(
SEQ Equation
¿
14
)C=
εA
h
(
SEQ Equation
¿
15 )∆ p
∝ ∆C=
−
εA
3 h
2∆ h(∆ h
≪h)
One of the proposed methods to enhance the sensitivity of PAS during the analysis of gases is to replace the microphone by a detection system using a laser source. Kauppinen et al. [4] used the setup shown in Figure 9 during their experiments. Their setup consists of two connected chambers where both chambers are filled with the analysis gas sample. A rectangular silicon cantilever (4 × 2 mm) with a thickness of 5 µm is used. The cantilever is attached on one side to the surrounding box, and only a small gap (30 µm) is present at the three other sides of the cantilever. The movement of the cantilever, caused by the pressure waves caused by the photoacoustic effect, is measured by a 630 nm laser (not further specified in the article) and three photodiodes (D1, D2 and D3 in Figure 9). In this interferometer, a beam-splitter and a mirror are present. These are adjusted in such a way that when the cantilever is in its resting position (no movement) the split beam will reach photodiodes D1 and D2 with half of an interference fringe (total destructive interference – no signal). When the cantilever bends, D1 and D2 have a 90° phase difference. Additionally, D1 and D3 have a phase difference of exactly 180°. Consequently, this system is able to measure the displacement of the cantilever over a complete change of 2π. Using this setup they were able to reach a detection limit of 800 ppt for a methane gas sample, compared to 10 ppm with an FTIR method [4]. However, during this study, only one cantilever type was tested for one specific gas.
Figure 9 The setup used by Kauppinen et al. to measure the photoacoustic effect by using a silicon cantilever. Image taken from [4]
Table 1 shows the LODs for different trace gases measured on CEPAS systems and compares the found numbers with different other spectroscopic techniques. However, as the LODs were determined in different studies, on different (CEPAS) systems, they are hard to compare. Furthermore, most of these LODs are determined in pure gases. But the LOD increases once it is determined in gas mixtures as shown by C.B. Hirschmann et al. [36]. This phenomena is more deeply treated in sub-chapter 5.3. Last, it must be amplified that not all available studies applying CEPAS for the mentioned gases could be read within the time set for this study. As far as possible, the CEPAS measurements were compare to other spectroscopic studies focussing on a certain analyte. However, if no such article was found by this study, it is compared to other techniques. Of course, it is very well possible that studies were performed able to reach lower limits of detection which were not found by this study. Articles claiming to be able to measure a specific gas without mentioning their LOD were not processed within this study. For a complete picture, more research is required where (CE)PAS gas detection results are compared to a wide range of other gas detection systems. However, the results shown in Table 1 give an first overview of the state of CEPAS for gas detection.
As these results in Table 1 show, CEPAS is able to measure most mentioned gases in the ppm, ppb or even sub-ppb range. With these results, it can compete or even surpass most of the other techniques mentioned in Table 1. Therefore, CEPAS shows a high potential to become more important for gas detection systems in the years to come.
Table 1 Comparison of the detection limit of several trace gases measured by CEPAS, compared to other spectroscopic techniques (if available). Abbreviations: NA; not available, WMS; wavelength modulation spectroscopy, Hrc-RGO; hierarchical nanostructure reduced graphene oxide,
Trace gas Analytical technique Detection
limit (ppb) Measuremen t time (s) Sourc e Methane (CH4) CEPAS 0.8 100 [4] Multipass cell + WMS 1.2 650 [37]
Acetic acid (CH3COOH) CEPAS 540 0.3 [13]
Visible-light-enhanced gas sensing of CdSxSe 1-x nanoribons
870 30 [38]
Methanol (CH3OH) CEPAS 800 0.3 [13]
SnO2 inverse opal sensor 1000 NA [39]
Acetone (C3H6O) CEPAS 24800 0.3 [13]
Cavity ringdown spectroscopy
110 9 [40]
Hydrogen sulphide (SH2) CEPAS 33 10 [41]
WMS 224 24 [42]
Acetylene (C2H2) CEPAS 0.08 200 [43]
Ag/ZnO Hrc-RGO 3000 57 [44]
Carbon Monoxide (CO) CEPAS 110 1 [45]
Absorption spectroscopy 1 1 [46]
Sulphur hexafluoride (SF6) CEPAS 1750 NA [47]
Cavity enhanced absorption spectroscopy
18 NA [48]
Formaldehyde (CH2O) CEPAS 0.623 15 [49]
CuO nanocubes 6 NA [50]
Nitrogen dioxide (NO2 ) CEPAS 0.05 1 [51]
Amperometric sensor 20 60 [52]
Benzene (C6H6) CEPAS 0.521
4.32 562 [36]
Absorption spectroscopy GC with PID detector
20 0.2 2400 1140 [53] [54] Toluene (C7H8) CEPAS 0.511 7.42 562 [36] SnO2-ZnO
GC with PID detector
100 0.26 1000 1140 [55] [54] p-Xylene (C8H10) CEPAS 0.591 11.02 562 [36]
GC with PID detector 0.8 1140 [54]
o-Xylene (C8H10) CEPAS 0.611
6.22
562 [36] GC with PID detector 1.7 1140 [54]
m-Xylene (C8H10) CEPAS 0.521
12.52
562 [36] Absorption spectroscopy
GC with PID detector
10 0.8
2400 1140
[53]
1 Determined by measuring the pure compound
5.2 M
ULTILAYER GRAPHENE CANTILEVERA study performed by J. Suchánek et al. [56] developed a cantilever based on multilayer graphene (MLG). Using MLG to construct the cantilever makes it is possible to build ultrathin structures down to a single atomic layer. Consequently, the sensitivity is increased compared to measurements performed with thicker cantilevers. During their study, three detection systems were compared by measuring the detection limit for methanol gas; a condenser microphone, and two cantilever system based on an MLG membrane and an MLG cantilever. The circular MLG membrane had a diameter of 4 mm and a thickness of 100 nm. To lock the membrane in its place, it was sealed in EVA-PET foil attached to a glass window, both with a circular opening at the MLG membrane position. The MLG cantilever was rectangular (6 mm × 3 mm) with a thickness of 50 µm. For their measurements, a tuneable CO2 laser (Edinburgh Instruments WL-8-Gt – no laser power mentioned or found) was used
with an excitation wavenumber of 1033.488 cm-1. To measure the displacement of the cantilever and
membrane system, a He-Ne laser was used in combination with a quadrant detector. The signals were measured for 200 seconds. The detection limit for methanol was determined to be 0.75 ppm, 0.42 ppm and 0.33 ppm with signal-to-noise ratios of 19, 61 and 70 for the condenser microphone, MLG membrane and MLG cantilever respectively.
A second study by J. Suchánek et al. [13] tested the MLG cantilever for multiple trace gasses. Here, they focussed on measuring acetic acid in a mixture with methanol and acetone. They chose acetic acid as a concentration above 80 ppb in the human breath is an indicator of the gastro-oesophageal reflux disease [57]. Their results were also shown in Table 1. Here, the same CO2 laser (excitation)
and He-Ne laser (cantilever displacement) were used as mentioned above. The laser power ranged from 170 mW to 1 W, depending on the excitation wavelength used. The limit of detection was determined at a measurement time of 0.3 s for fast data acquisition. This resulted in LODs of 540, 800 and 24.800 ppb for acetic acid, methanol and acetone respectively. Their system was also tested for a prepared mixture of 6.2 ppm acetic acid, 6.5 ppm methanol and 65 ppm acetone, all present above their individual LOD. To calculate the concentration, five absorption lines were used; 9.24, 9.5, 9.58, 10.2 and 10.24 µm. These values were based on the absorption lines of the mentioned gases between 1082 – 947 cm-1. The calculated values using their MLG cantilever system were 6.0 ppm for acetic
acid (∆ ≈ 3%), 3.4 ppm for methanol (∆ ≈ 48%) and 70 ppm for acetic acid (∆ ≈ 20%). For a second mixture (10.7 ppm acetic acid, 10.6 ppm methanol and 3950 ppm acetone) the calculated values were 2.6 ppm for acetic acid (∆ ≈ 76%), 12.5 ppm for methanol (∆ ≈20%) and 4200 ppm for acetone (∆ ≈ 7%). As no measurement time is mentioned for these specific measurements, a time of 0.3 seconds is assumed. In their paper, J. Suchánek et al. conclude that their system is, with the numbers presented above, suitable for measuring acetic acid at a high background of acetone and is valuable for medical breath analysis. However, as mentioned, a concentration of 80 ppb of acetic acid is an indicator of the gastro-oesophageal reflux disease. With their system, J. Suchánek et al. are not able to measure these
concentrations and is therefore unsuitable as a reliable device for this purpose. Additionally, their method calculates concentration with a large deviation up to 76% of the real concentration. Therefore, adjustments are required before their system is reliable enough for clinical purposes.
5.3 T
UNEABLECEPAS
SYSTEMS FOR MULTICOMPONENT ANALYSISAs already mentioned in sub-chapter 5.2, CEPAS systems have been developed capable of measuring the concentration of several trace gasses by using tuneable excitation lasers.
A study by C. Hirschmann et al. [36] developed a CEPAS system to detect benzene, toluene, p-, m-and o-xylene. Together, these compounds are often abbreviated as BTX within the petrochemical industry. Although important as starting chemicals or as solvents, exposure can have negative effects on the human health and the environment [58–60]. In their study, C. Hirschmann et al. tested the possibility to combine a small commercially available optical parametric oscillator (OPO – dimensions; 125×70×45 mm) with their own CEPAS system. The OPO is tuneable from 3237 to 3296 nm, with a wavelength depended laser power from 88 to 103 mW. Their CEPAS system contained a cylindrical photoacoustic cell (95 mm in length, 4 mm in diameter) and a silicon cantilever of 5×1.2×0.01 mm (L×W×H). To test their system, the detection limit (LOD) of single-compound gas and multi-compound gas was tested for BTX. During all measurements, the BTX (both single- and multi-compound) was mixed with a methane and water vapour blend in nitrogen gas. For the determination of the LOD with single compound gas, two methods were used. First, the photoacoustic signal was measured for all compounds at 3288 nm. This wavelength was chosen as all BTX compounds show absorption at this point while both methane and water show none [36]. The LOD was statistically determined by measuring the photoacoustic signal for 0.951 seconds at three different concentrations and determine three times the standard deviation (3σ). The results of these measurements are found in Table 2. The LOD’s found vary from 9.8 ppb for toluene to 16.0 for o-Xylene. The different LOD’s found are explained by the difference in the absorption coefficient at 3288 nm for the BTX compounds. Second, the LOD’s were determined by a science-based calibration [61] using the photoacoustic signal at 591 spectral point between 3237 – 3296 nm. Each spectral point was measured for 0.951 seconds, giving a total acquisition time of 562 seconds/9.3 minutes. This multivariate data analysis uses both the known absorption spectrum of methane, water and the BTX gas to determine the contribution of a single compound to the measured photoacoustic signal. As these results show (see Table 2), this method improves the LOD’s to sub-ppb levels. Finally, the science-based calibration was applied to simultaneously calculate the LOD’s of the BTX gases whilst measuring the photoacoustic signal between 3237 – 3296 nm. The same data acquisition time was used as mentioned above. The concentration of the compounds varied between 4.96 ppm to 10.05 ppm. As these results show (see Table 2), measuring the LOD’s within a mixture provides higher values compared to single-compound measurements using the science-based calibration approach.
Where this decrease originates from is not explained within their study. However, with the exception of m-Xylene, lower LOD’s were found for the multi-compound measurements when compared to the single-compound measurements using the photoacoustic signal generated at only one wavelength (3288 nm).
Table 2 Results obtained by C. Hirschmann et al. [36] determining the LOD’s of BTX in both a single-compound and multi-compound setup. Two different approaches were used to determine the LOD; (1) measuring at a single wavelength (3288 nm) and calculate the LOD using three times the standard deviation (3σ) and (2) take the signal at 591 spectral point between 3237 – 3296 nm and calculate the LOD by science-based calibration using 3σ.
Compound LOD
single-compound at 3288 nm (3σ, 0.951 s) (ppb) LOD single-compound nm (3σ, 591 spectral points, 0.951 s each) (ppb) LOD multi-compound nm (3σ, 591 spectral points, 0.951 s each) (ppb) Benzene 12.0 0.52 4.3 Toluene 9.8 0.51 7.4 p-Xylene 13.2 0.59 11.0 o-Xylene 16.0 0.61 6.2 m-Xylene 10.1 0.52 12.5
However, to calculate the presented detection limits C. Hirschmann et al. only took the interference by water into account. A study by J. Chin et al [62] showed that vapour from gasoline (a product from the petrochemical industry) consists of many more compounds besides BTX. Each of these compounds could potentially cause interference which is not taken into account with their current method. Therefore, it is unlikely that the LOD values shown will be reliable for real-life gas samples. Furthermore, no data was presented where the actual gas concentration is compared to the concentration calculated by their system. Therefore, it remains unknown if this system is able to produce reliable concentrations even with clean gas samples.
5.4 P
ORTABLECEPAS
SYSTEMAs Table 1 shows, CEPAS systems are capable of delivering high sensitivity for multiple trace gases. For many scientific and industrial processes, the detection of these gases is important. Unfortunately, most developed CEPAS systems are not very compact and certainly not portable. Furthermore, precise alignment procedures are required in order to obtain high performances [10]. The system shown in Figure 9 is an example. Therefore, in many cases small gas detectors such as electro-chemical and metal-oxide gas detectors are still favoured. However, these systems are not capable of detecting trace gases in the sub-ppm level [63]. Consequently, multiple studies focussed on developing portable CEPAS systems.
Most cantilever-Enhanced PAS systems developed require a pump to actively transport the trace gas into the PAS system before it can be analysed. Although these systems are often very sensitive, such
as the systems discussed in chapter 5.3, the requirement of a gas transportation system provides multiple disadvantages. First, a pump requires energy and space within the CEPAS system, which makes it less portable. Furthermore, the transportation of a trace gas often causes dilution, making it hard to determine the exact concentration at a specific spot. Moreover, measuring biological processes by the gas consumption/production is often performed in closed environments by active sampling. As this may affect the diffusion rates [63], there is a need for highly sensitive in situ trace gas detectors. Zhou and Iannuzzi [63] developed a CEPAS system capable of performing in situ trace gas detection. Their system uses a fibre-tip photoacoustic sensor as shown in Figure 10. With a total volume of 63 µL, it is small enough to fit in small rooms like reagent tubes. The photoacoustic signal is generated in the photoacoustic cell. This cell is produced from transparent glass, and is protected by a quartz tube. At one end of the photoacoustic cell (the cell inlet), a carbon fibre filament is placed with a pore-size of 10 µm. This allows the diffusion of trace gas molecules into the cell, while protecting it from contaminations. At the other end of the cell, the cantilever is placed. The cantilever consists of a micro-mirror and a carbon fibre, capable of transferring the photoacoustic signal into a mechanical vibration. Furthermore, the centre of the micro-mirror is transparent. This allows the introduction of the excitation signal via an excitation fibre through the micro-mirror. The photoacoustic signal is measured via a readout fibre, located at 50 µm from the micro-mirror. The displacement of the cantilever is measured interferometrically. For this, a commercially available interferometer (OP1550, Optics11) was used. This interferometer has a tuneable laser source between 1528 – 1563 nm [64]. This allows to scan the Fabry-Perot cavity, or plane-parallel cavity, at different wavelengths and choose the wavelength providing the optimal linearity and sensitivity [10]. Figure 11 explains the readout and excitation setup used by Zhou and Iannuzzi [63]. Although a different sensor is used in this example, the same interferometric principle is used. Thus, the excitation signal enters the photoacoustic cell via the excitation fibre, through the micro-mirror. Here, the trace-gas molecules generate a photoacoustic signal which is translated into a mechanical vibration. The vertical displacement of the cantilever is measured interferometrically by a laser signal through the readout fibre. The noise equivalent concentration (NEC) was determined by the Allan deviation. For a 1 second integration time, a NEC of 24.7 ppb was found. At 230 seconds, the optimal integration time was found, providing a NEC of 3 ppb. However, the gas pressure during the measurements was put at 0.35 bar, as the optimal PAS signal is obtain at this pressure. Therefore, it is expected that measurements performed at regular conditions (1 bar) will provide lower detection limits. Finally the response time of the system was found to be 30 seconds.
Figure 10 . [LEFT-TOP] Photograph of the photoacoustic sensor, compared to a USB connector to illustrate the true size of the sensor. [MIDDLE] Sketch of the complete fibre-tip photoacoustic sensor, showing the different parts. [RIGHT-BOTTOM] Sketch of the micro-mirror cantilever system, showing the position of the excitation fibre, readout fibre and photoacoustic cell. The arrows show the path of laser beams through the different fibres. As the dotted line after the excitation fibre indicates, the centre of the micro-mirror is transparent, enabling the excitation beam to pass through and enter the photoacoustic cell . This image was taken from [63], showing all parts to scale.
Figure 11 Illustration of the excitation and readout setup used by Gruca et al. [10]. Although the photoacoustic sensor shown here is different from the one used by Zhou and Iannuzzi [63], they use the same interferometric principle. One laser, shown in green, is used to create the photoacoustic signal in the sensor. The other laser, shown in salmon, is split into two beams by the optical coupler. One beam is directed to the sensor, where it hits a cantilever and is redirected back into the optical fibre it originated from. In the optical coupler, this signal is combined with the second laser beam heading toward the detector. Depending on the vertical displacement of the cantilever due to the photoacoustic signal, the two beam could constructively or destructively interfere. The amplitude of the displacement can be coupled to a certain concentration by means of calibration [10]. This image was taken from [10]
Zhou and Iannuzzi [63] tested their fibre-tip photoacoustic sensor by measuring the CO2 production of
a yeast fermentation process within a test tube. Due to the limited space within a test tube, conventional (cantilever) PAS systems could only monitor the CO2 production by enclosing the
