3 SAMPLE PRE-TREATMENT FOR IMAGING SPECTROSCOPY
4.3 Fourier transform infrared spectroscopy (FTIR)
While the analytical methods for MP analysis have developed quickly during 2010s, spectroscopic methods have gained popularity. The first MP studies utilized light microscopy alone or followed by chemical analysis methods such as FTIR or morphological analysis such as SEM (Hidalgo-Ruz et al., 2012). Back in the beginning of 2010s, Hidalgo-Ruz et al. (2012) suggested that visual sorting remains as an obligatory step for MP analysis. However, FTIR microscopy with single element detector was reported to be suitable method for quantitation of MPs in 2012 (Harrison et al., 2012). Focal plane array (FPA) or linear array detector based FTIR imaging of microplastics (enriched to filters) was published in 2015 for marine plankton and sediment (Löder et al., 2015) and wastewater (Tagg et al., 2015).
4.3.1 Measurement techniques in FTIR spectroscopy
Infrared spectroscopy can be performed in multiple measurement modes; ATR (attenuated total reflection), transmission and reflection. Each of them is suitable for MP analysis, but for different particle sizes. ATR-FTIR is usually done with a benchtop spectrometer. In MP research, transmission and reflection modes are usually conducted with a spectrometer coupled to a microscope. The benchtop ATR-FTIR method has usually been used for the verification of light microscopy identification (e.g. Álvarez-Hernández et al., 2019; Biginagwa et al., 2016), and is suitable for analysing larger, typically >500 µm particles (Veerasingam et al., 2020).
It is rather time-consuming, as particles have to be measured manually one by one (Primpke et al., 2020a). Thus, many studies have utilized ATR-FTIR for analysing only a small proportion of all particles. However, ATR is relatively affordable and easy to use, which makes it popular and suitable method for analysing larger MPs.
The terms FTIR microscopy, µ-FTIR, or FTIR microspectroscopy refer to the similar instrumentations, in which infrared spectrum is measured via a microscope. While ATR-FTIR is suitable for larger particles, FTIR microscopy enables the measurement of smaller particles, down to approximately 10 µm, limited by the diffraction of light (Rocha-Santos and Duarte, 2017). FTIR microscopes have different types of detectors – single element or point detector measures one spectra at time, whereas FPA detector consists of a matrix of typically 64x64 or 128x128 elements. In between is a linear array detector, which consists of typically 16 elements. Each detector element measures one spectrum, which means that FPA detectors measure up to 128*128 = 16 384 spectra simultaneously. FPA resembles a digital camera, but each pixel contains a spectrum. Compared to single element detector, FPA enables fast measurement of large areas containing representative number of particles. With single element detector, each particle has to be located and measured individually through microscope (Negrete Velasco et al., 2020; Talvitie et al., 2017). Point measurements can be performed in FTIR microscope with ATR accessory, too (Bordos et al., 2019). In principle, point detector can be used for mapping of larger areas. Maps consist of pixels measured one by one, but because mapping is highly time-consuming, only a few studies have utilized it (Harrison et al., 2012; Renner et al., 2020; Vianello et al., 2013). However, pre-selection methods of particles, based on image analysis, can reduce the measurement time (Renner et al., 2020).
Compared to point measurements, FPA-FTIR is very fast. The other advantage of FPA-FTIR is that contradictory to other microscopic methods, it does not require visual pre-selection of particles, which is always more or less subjective and therefore prone to bias (Löder et al., 2015). FPA-FTIR allows automatic measurement of large areas, and visual selection is not necessary. It is, hence, one of the most selective
Figure 3. Energy level diagram of infrared absorption and Raman Stokes scattering.
The output of both techniques is a spectrum, where x axis presents wavelength or Raman shift and y axis absorbance or intensity. Spectrum consists of peaks, which correspond to specific chemical bonds or their combinations. Raman and FTIR spectra have the peaks of the same molecule in the same positions and can be interpreted similarly. Thus, when spectra from MP sample are compared to reference spectra of plastics, they reveals whether particles are MPs or not (Rocha-Santos and Duarte, 2017).
4.3 FOURIER TRANSFORM INFRARED SPECTROSCOPY (FTIR)
While the analytical methods for MP analysis have developed quickly during 2010s, spectroscopic methods have gained popularity. The first MP studies utilized light microscopy alone or followed by chemical analysis methods such as FTIR or morphological analysis such as SEM (Hidalgo-Ruz et al., 2012). Back in the beginning of 2010s, Hidalgo-Ruz et al. (2012) suggested that visual sorting remains as an obligatory step for MP analysis. However, FTIR microscopy with single element detector was reported to be suitable method for quantitation of MPs in 2012 (Harrison et al., 2012). Focal plane array (FPA) or linear array detector based FTIR imaging of microplastics (enriched to filters) was published in 2015 for marine plankton and sediment (Löder et al., 2015) and wastewater (Tagg et al., 2015).
4.3.1 Measurement techniques in FTIR spectroscopy
Infrared spectroscopy can be performed in multiple measurement modes; ATR (attenuated total reflection), transmission and reflection. Each of them is suitable for MP analysis, but for different particle sizes. ATR-FTIR is usually done with a benchtop spectrometer. In MP research, transmission and reflection modes are usually conducted with a spectrometer coupled to a microscope. The benchtop ATR-FTIR method has usually been used for the verification of light microscopy identification (e.g. Álvarez-Hernández et al., 2019; Biginagwa et al., 2016), and is suitable for analysing larger, typically >500 µm particles (Veerasingam et al., 2020).
It is rather time-consuming, as particles have to be measured manually one by one (Primpke et al., 2020a). Thus, many studies have utilized ATR-FTIR for analysing only a small proportion of all particles. However, ATR is relatively affordable and easy to use, which makes it popular and suitable method for analysing larger MPs.
The terms FTIR microscopy, µ-FTIR, or FTIR microspectroscopy refer to the similar instrumentations, in which infrared spectrum is measured via a microscope. While ATR-FTIR is suitable for larger particles, FTIR microscopy enables the measurement of smaller particles, down to approximately 10 µm, limited by the diffraction of light (Rocha-Santos and Duarte, 2017). FTIR microscopes have different types of detectors – single element or point detector measures one spectra at time, whereas FPA detector consists of a matrix of typically 64x64 or 128x128 elements. In between is a linear array detector, which consists of typically 16 elements. Each detector element measures one spectrum, which means that FPA detectors measure up to 128*128 = 16 384 spectra simultaneously. FPA resembles a digital camera, but each pixel contains a spectrum. Compared to single element detector, FPA enables fast measurement of large areas containing representative number of particles. With single element detector, each particle has to be located and measured individually through microscope (Negrete Velasco et al., 2020; Talvitie et al., 2017). Point measurements can be performed in FTIR microscope with ATR accessory, too (Bordos et al., 2019). In principle, point detector can be used for mapping of larger areas. Maps consist of pixels measured one by one, but because mapping is highly time-consuming, only a few studies have utilized it (Harrison et al., 2012; Renner et al., 2020; Vianello et al., 2013). However, pre-selection methods of particles, based on image analysis, can reduce the measurement time (Renner et al., 2020).
Compared to point measurements, FPA-FTIR is very fast. The other advantage of FPA-FTIR is that contradictory to other microscopic methods, it does not require visual pre-selection of particles, which is always more or less subjective and therefore prone to bias (Löder et al., 2015). FPA-FTIR allows automatic measurement of large areas, and visual selection is not necessary. It is, hence, one of the most selective
methods for analysing small (<100–300 µm) MPs. However, FPA-FTIR instruments are expensive and require a trained operator.
All types of FTIR microscopy are rather time-consuming, which reduce their usability for the monitoring of large sample sets. To overcome this disadvantage, quantum-cascade laser (QCL) based infrared spectroscopy has been reported to be as accurate, but faster method for analysing MPs from environmental samples (Primpke et al., 2020c). However, it is currently a new technique and its applicability has not been tested as widely as other FTIR techniques.
4.3.2 FPA-FTIR in practise
FPA-FTIR measurements can be done in transmission or reflection mode (Löder et al., 2015). Before measurement, samples need pre-treatment to separate MPs efficiently from the matrix (Löder et al., 2017). The pre-treatment has significant effect on the selectivity of FPA-FTIR analysis. Because FPA-FTIR aims to quantify MPs down to 10–20 µm, they have to be separated and purified from other materials in the environmental samples. If sample contains high amounts of other particles, MPs can be covered by them and do not produce clear spectra of plastics. Pre-treatment methods are covered in Chapter 3.
Transmission mode is suitable for small and thin particles, because large and thick particles absorb infrared radiation too intensively to produce a representative spectra (Primpke et al., 2020a). This is called total absorption, resulting in detector saturation.
On the other hand, reflection mode is suitable for larger particles. However, in reflection, the particle must be reflective, thus dark-coloured particles can lead to full absorption of the incident light and no spectra is acquired.
The most common FTIR measurement techniques, their limitations and deliverables are summarized in Table 4.
Table 4. Overview of FTIR techniques in MP analysis. *Approximate value – depends on the instrumentation, measurement parameters, and features of the particle, such as chemical composition and thickness (Löder et al., 2015).
Technique Particle size
(µm) Quantitative Visual sorting Sample pre-treatment
Benchtop ATR >500 Semi Yes Simple
Microscopic techniques
Point mode - transmission 10–500* Semi Yes Moderate
Point mode - reflection 10–1000* Semi Yes Moderate
FPA Imaging – transmission 10–500* Semi/Yes No Complex
FPA Imaging - reflection 10–1000* Semi/Yes No Complex
After particles have been separated from the matrix, they are deposited or filtered to a substrate suitable for FTIR. Filters or substrates should not absorb IR in the typical measured range, which is at least 800–3600 cm-1. In transmission, substrates should pass the infrared radiation completely through, and in reflection, substrates should reflect the radiation completely. Various materials have been tested for both modes.
The most suitable materials so far for transmission are zinc selenide windows (Liu et al., 2019; Simon et al., 2018; Vianello et al., 2019), aluminium oxide (Anodisc) filter (Haave et al., 2019; Mintenig et al., 2019) and silicon filter (Käppler et al., 2015).
Reflection measurements have been performed with silver membrane filters (Primpke et al., 2020b), gold-coated polycarbonate filters (Uurasjärvi et al., 2021) and microscope reflection slides (Simon et al., 2018). Generally, sample deposition is easier with filters than non-porous substrates, because filtration is fast compared to pipetting and evaporation of suspension to a substrate.
The measurement parameters which affect the quality of FTIR spectra include number of scans and spectral resolution (Löder et al., 2015). The number of scans means how many spectra are measured from one point for averaging. The higher the number of scans, the better signal to noise ratio is obtained. However, in MP analysis, the imaged areas are typically very large, and increased number of scans leads to long measurement times. Spectral resolution means interval between data points in a spectrum, and it defines smallest features that are possible to distinguish from the
methods for analysing small (<100–300 µm) MPs. However, FPA-FTIR instruments are expensive and require a trained operator.
All types of FTIR microscopy are rather time-consuming, which reduce their usability for the monitoring of large sample sets. To overcome this disadvantage, quantum-cascade laser (QCL) based infrared spectroscopy has been reported to be as accurate, but faster method for analysing MPs from environmental samples (Primpke et al., 2020c). However, it is currently a new technique and its applicability has not been tested as widely as other FTIR techniques.
4.3.2 FPA-FTIR in practise
FPA-FTIR measurements can be done in transmission or reflection mode (Löder et al., 2015). Before measurement, samples need pre-treatment to separate MPs efficiently from the matrix (Löder et al., 2017). The pre-treatment has significant effect on the selectivity of FPA-FTIR analysis. Because FPA-FTIR aims to quantify MPs down to 10–20 µm, they have to be separated and purified from other materials in the environmental samples. If sample contains high amounts of other particles, MPs can be covered by them and do not produce clear spectra of plastics. Pre-treatment methods are covered in Chapter 3.
Transmission mode is suitable for small and thin particles, because large and thick particles absorb infrared radiation too intensively to produce a representative spectra (Primpke et al., 2020a). This is called total absorption, resulting in detector saturation.
On the other hand, reflection mode is suitable for larger particles. However, in reflection, the particle must be reflective, thus dark-coloured particles can lead to full absorption of the incident light and no spectra is acquired.
The most common FTIR measurement techniques, their limitations and deliverables are summarized in Table 4.
Table 4. Overview of FTIR techniques in MP analysis. *Approximate value – depends on the instrumentation, measurement parameters, and features of the particle, such as chemical composition and thickness (Löder et al., 2015).
Technique Particle size
(µm) Quantitative Visual sorting Sample pre-treatment
Benchtop ATR >500 Semi Yes Simple
Microscopic techniques
Point mode - transmission 10–500* Semi Yes Moderate
Point mode - reflection 10–1000* Semi Yes Moderate
FPA Imaging – transmission 10–500* Semi/Yes No Complex
FPA Imaging - reflection 10–1000* Semi/Yes No Complex
After particles have been separated from the matrix, they are deposited or filtered to a substrate suitable for FTIR. Filters or substrates should not absorb IR in the typical measured range, which is at least 800–3600 cm-1. In transmission, substrates should pass the infrared radiation completely through, and in reflection, substrates should reflect the radiation completely. Various materials have been tested for both modes.
The most suitable materials so far for transmission are zinc selenide windows (Liu et al., 2019; Simon et al., 2018; Vianello et al., 2019), aluminium oxide (Anodisc) filter (Haave et al., 2019; Mintenig et al., 2019) and silicon filter (Käppler et al., 2015).
Reflection measurements have been performed with silver membrane filters (Primpke et al., 2020b), gold-coated polycarbonate filters (Uurasjärvi et al., 2021) and microscope reflection slides (Simon et al., 2018). Generally, sample deposition is easier with filters than non-porous substrates, because filtration is fast compared to pipetting and evaporation of suspension to a substrate.
The measurement parameters which affect the quality of FTIR spectra include number of scans and spectral resolution (Löder et al., 2015). The number of scans means how many spectra are measured from one point for averaging. The higher the number of scans, the better signal to noise ratio is obtained. However, in MP analysis, the imaged areas are typically very large, and increased number of scans leads to long measurement times. Spectral resolution means interval between data points in a spectrum, and it defines smallest features that are possible to distinguish from the
shapes of peaks. High spectral resolution lowers signal to noise ratio and requires higher number of scans, which increases measurement time.
The choice of parameters is a compromise between spectral quality and measurement speed. Löder et al. (2015) have reported that spectral quality did not increase significantly after 6 scans and spectral resolution 8 cm-1 is enough for identifying MPs. The maximum spectral range is usually limited by detector and substrate, and MPs are identified better if wide enough range is measured. Therefore, the range does not need to be limited from the maximum. Additionally, FPA-FTIR instruments typically have options for spatial pixel sizes (or binning factors) and objectives.
Depending on the instrumentation, the measurement speed may be increased by using larger pixel size. However, it produces lower spatial resolution and is not suitable for detecting the smallest MPs, close to the diffraction limit in 10 µm. These options define the resolution of a matrix measured, which affects how small particles can be detected (Table 5).
Table 5. The effects of measurement parameters in imaging FTIR, assuming that the imaged area is constant.
Parameter Signal to
noise ratio Limit of detection
(particle size) Speed Data volume
Pixel size/binning and/or objective x x x x
Spectral resolution x - x x
Spectral range - - x x
Number of scans (averaging) x - x x
4.3.3 Data analysis methods
The most common data analysis method is to compare measured spectra and reference spectra of known plastics. The comparison is usually done by calculating Person’s correlation coefficient for two spectra, which gives a number between 0–1, or as percentage 0–100. The higher the correlation coefficient, the higher is the probability for identification. As a rule of thumb, 0.7 (or 70%) is generally set as a limit for reliable recognition. However, MPs from the environment have experienced harsh conditions such as ultraviolet (UV) radiation, moisture, oxygen and mechanical abrasion, which affect their chemical and physical properties
(Veerasingam et al., 2020). Therefore, if data analysis is performed using library of spectra measured from commercial virgin plastics, the limit of recognition cannot be unambiguously set (Cowger et al., 2020b).
The other problem with commercial libraries is that different measurement techniques cause certain features to spectra. In the optimal situation, the reference library should be measured with the same instrument and the same parameters.
Many researchers have collected own reference libraries, but to harmonize the methods, more open source spectra libraries measured with different methods from weathered MPs are needed (Cowger et al., 2020b; Primpke et al., 2018).
Reference libraries and data analysis software can be commercial or custom-made (Cowger et al., 2020b). An inclusive library should contain spectra of at least the most common polymer types and natural polymers, such as proteins and cellulose materials, depending on what is common in the sample matrix (Primpke et al., 2018).
Inclusion of natural polymers prevents the mismatching them as plastics.
Single spectra from benchtop ATR or point mode microscopy are usually compared to references one-by-one. However, FPA-FTIR produces excessive amount of data, up to millions of spectra or tens of gigabytes. FPA-FTIR images are three-dimensional matrixes, in which the first and second dimensions define the location of the spectrum in the map, and third contains the spectral information. To compare these spectral maps with reference spectra, every pixel has to be compared to each reference. The process would be impossibly time-consuming to do manually.
Therefore, automatic data analysis software have been developed for quantifying particles by their materials (Primpke et al., 2017, 2020b; Renner et al., 2019b; Wander et al., 2020). Moreover, spectra can be pre-processed to remove noise, correct baseline or normalize before the data analysis to match the spectra better with library references (Renner et al., 2019a).