From: “Important Instrumentation and Methods for the Detection of Chemicals in the Field”. Chapter 8 on FTIR Gas Measurement by the real time detection chapter of the AIHA ISBN-13: 978-1-935082-41-5
Chapter 8 Field-Portable Fourier Transform Infrared (FTIR) Spectroscopy for Gas and Vapor Analysis
Antii Heikkila, MSc
Introduction Fourier Transform Infrared (FTIR) spectrometers are widely used in various analytical applications ranging from qualitative identification of solid and liquid samples, to qualitative and quantitative analysis of trace gases and vapors in complex matrices such as contaminated air or process gas streams. The FTIR measurement technique relies on an optical modulator known as an interferometer. The first interferometer design by A.A. Michelson near the turn of the 20th century(1,2) predates infrared spectroscopy. Michelson’s interferometer was built to prove or disprove the existence of ether in space, and Michelson and Morley used interferometer measurements to show that the speed of light was uniform in all directions. This exposed a major flaw in the ether theory of light, and paved the way for future work to describe the theory of relativity by Einstein.(3) Michelson was awarded the 1907 Nobel Prize in physics “for his optical precision instruments and the spectroscopic and metrological investigations carried out with their aid.”(4) While Rayleigh made an early suggestion to Michelson that the interferometer could be used to measure the wavelengths of light(5), the prevalent method at that time (and for the next half century) to measure an optical spectrum was based on wavelength dispersion using a prism or a grating. In 1949, Fellgett performed an approximate Fourier transformation of interferometer data to obtain absorbance
information across a range of wavelengths (an absorbance spectrum).(3) Fellgett recognized that during collection of an interferogram, data from all spectral wavelengths are obtained simultaneously, whereas collection of a spectrum using a prism or series of IR filters allows measurement of only a single wavelength band at any given instant. The ability to collect a wide range of wavelength information simultaneously reduces the time required for measurement of a spectrum with an FTIR instrument compared to a dispersion or filter instrument, and this improved performance is known as “Fellgett’s advantage.” The FTIR spectrometer was limited by the relatively slow computation methods of the Fourier transformation with available computers and algorithms until Cooley and Tukey described the fast Fourier transform in 1965.(6) Instrumentation for FTIR analysis became commercially available soon afterwards. The introduction of cube corner mirrors into FTIR spectrometers by Mattson in 1983(7) represented a major improvement, and allowed requirements for precise alignment of moving optical components to be relaxed, making FTIR suitable for industrial and field measurement applications.
IR Theory and Instrumentation The majority of infrared (IR) gas analyzers operate in the mid-IR region (about 2 to 15 µm wavelengths). Spectra are usually plotted as absorbance or per cent transmission on the y axis against wavenumber on the x axis. The wavenumber 47
Important Instrumentation and Methods for Detection of Chemicals in the Field
is related to IR wavelength as the reciprocal of wavelength (expressed in units of cm), and thus has units of cm-1. The wavenumber describes the number of complete waves for a particular wavelength across a distance of 1 cm, and wavenumbers are directly related to energy levels. The mid-IR wavelengths between 2 and 15 µm correspond to wavenumbers 5000 cm-1 to 670 cm-1. Instruments operating at higher wavenumbers (shorter wavelengths) are termed near-IR analyzers and are used mainly for analysis of liquids in various process applications. The lower wavenumbers (longer wavelengths) of far IR are mainly used in research applications with solid samples. Absorption of mid-IR energy occurs at specific wavelengths depending on the molecular structure of a vibrating molecule. While light energy at these wavelengths is not sufficient to cause electronic transitions, this region does encompass the resonant frequencies of molecular bonds. When the energy of an incoming photon matches the energy difference between two vibrational levels of a molecule the molecule can absorb the photon and transfer from a lower to a higher vibrational energy level. This results in a reduction of IR light arriving at a detector. A dipole moment in a molecule (separate regions of differing electric charge) that will change as the molecule vibrates is required for IR absorption. The probability of the molecule absorbing incoming IR light increases with the magnitude of dipole moment, and vibrations in homonuclear molecules with zero dipole moment do not absorb IR light. This means that diatomic molecules composed of only a single element such as nitrogen (N2) and oxygen (O2) will not absorb IR energy. Figure 8.1 shows a gas phase IR spectrum for a volatile organic compound, with characteristic IR absorption features due to the presence of different functional groups. Three general types of IR absorption instruments are commonly encountered for field use. These are the nondispersive IR (NDIR), dispersive or filter IR (FIR), and FTIR instruments. The NDIR instruments are typically used to detect and quantify only a single pre-selected gas or vapor analyte, while dispersive or FIR instruments are more capable. The field-portable FTIR group includes instrumentation dedicated to the analysis of gases and vapors, as well as other types that may analyze solid and liquid materials. The primary differences between dispersive and FTIR instruments will be described below.
Single-Analyte IR Instruments The simplest IR absorption detector uses only a single wavelength (or narrow wavelength band) and the different 48
Figure 8.1 – Infrared spectrum for 200 ppm cis-1,2-dichloroethylene vapor with absorbance features due to various functional groups as noted. A 980 cm multipass gas cell was used (1.0 atm, 51°C). The spectrum was recorded using a Gasmet FTIR instrument operating with 7.72 cm-1 spectral resolution.
wavelengths of IR light produced from a source are not dispersed. Dispersion is not needed if absorbance at only a single dedicated wavelength is to be measured, and either a monochromatic light source is used or a filter is used to limit the transmission of light to a desired frequency. Single-gas analyzers for either carbon monoxide (CO) or carbon dioxide (CO2) are common examples of the simple NDIR detection approach. For an NDIR instrument used to measure CO, all absorption at a specified wavelength (near 4.8 µm in this example) must be assumed to be due to the target analyte, although this assumption may give rise to cross-sensitivity interference if an unexpected component of the gas phase sample absorbs IR energy at or very near the same region of the spectrum. For example, nitrous oxide (N2O, which absorbs strongly near 4.5 µm) could be an interferent to NDIR measurement of CO.
Dispersive and Filter IR Instruments A more complicated IR detection approach that may provide more information will take into account more than a single absorbance wavelength. If a full-spectrum IR scan is required, it is necessary for a non-FTIR instrument to separately measure the different wavelength values in the spectrum. This may be accomplished by dispersing the IR light with a pivoting reflective grating that causes different wavelengths to strike a detector at different times. Another approach is to sequentially place different filters in the optical path to collect absorbance data over time for a number of discrete IR wavelengths. The Miran Sapphire 205B instrument operates as a FIR instrument, employing seven fixed wavelength filters to
Chapter 8 — Field-Portable Fourier Transform Infrared (FTIR) Spectroscopy for Gas and Vapor Analysis
measure mid-IR absorbance at specific wavelengths between about 1.8 and 4.7 µm, and a variable filter for wavelengths between 7.7 and 14.1 µm. Absorbance information may be obtained over a period of several minutes for the portions of a mid-IR spectrum corresponding to the available filters (if desired) by sequentially scanning with the various filters placed in the optical path. While this approach provides significant capability beyond that of a simple NDIR instrument, the need to sequentially scan makes this approach slow relative to FTIR spectroscopy where absorbance data are simultaneously collected across a range of wavelengths. For detection of a pre-determined gas the FIR instrument may be operated analogously to the NDIR detector using only absorbance at a relevant wavelength, although it is subject to the same types of interferences as a simple NDIR detector when it is used in this way.
possible to calculate from it the absorbance peaks due to the sample present between the interferometer and the detector. For many years the main factor preventing FTIR analysis from becoming more commonplace was the computation time of the Fourier transformation, but following the invention of a fast Fourier transform algorithm discussed previously, and with advances in digital signal processing it has become practical to build small, low-cost FTIR spectrometers capable of rapidly measuring a complete mid-IR spectrum with moderate or high resolution and very good signal-to-noise ratio. Coupled with advanced algorithms for identification and quantification of the chemical species represented in the spectrum, modern gas and vapor FTIR analyzers may identify multiple unknowns, and quantify components of complex gas phase mixtures containing dozens of volatile organic compounds.
Fourier Transform IR Instruments In contrast to NDIR and dispersive or FIR instruments, with an FTIR instrument the signal for all relevant wavelengths arrives at the detector nearly simultaneously, and IR absorbance information for the different wavelengths is separated through computation. This allows for nearly instantaneous scanning of a wide wavelength range, and greater sensitivity due to the Fellgett advantage. In 1989, Levine et al.(8) discussed the relative advantages and disadvantages of FTIR compared to FIR for monitoring gases and vapors of concern to the industrial hygienist. At that time, field-portable FIR instruments were available, but FTIR was still mostly confined to the laboratory. Levine et al. predicted that “FTIR systems will get smaller and cheaper and will compete with the FIR systems for transportable air monitoring applications.” In the roughly twenty intervening years this has occurred. The remainder of this chapter discusses FTIR analysis, mostly regarding field analysis of gas phase compounds. The IR radiation from a light source in an FTIR spectrometer is collected into a collimated beam, which is then split in two branches with different optical path lengths. The length of one or both of these branches is varied with a moving mirror, and the two branches are brought together so that depending on the position of the moving mirror the beams undergo either constructive or destructive interference. The resulting beam with constantly varying intensity is passed from the interferometer through a sample compartment where absorbance occurs, and then to a detector, where the signal is recorded as a function of time. The result is an interferogram, and through Fourier transform analysis it is
Fourier Transform IR Gas and Vapor Analyzer Components IR Source For portable mid-IR instruments a common type of light source is a black body radiator relying on thermal emission of IR light from a hot filament (e.g. kanthal™, a Ni-Cr-Al alloy) or small ceramic rod (for example, a 5 mm diameter x 50 mm silicon carbide ceramic “globar” rod). Both of these simple light sources demonstrate continuous emission over the midIR wavelengths, with low cost and long operational lifespan (years).
Interferometer The interferometer is the heart of an FTIR spectrometer, and the traditional design by Michelson is shown in Figure 8.2, in simplified form. Light emitted from the IR source is directed to a thin reflective metal disk (beamsplitter, typically positioned at a 45 degree angle) which is sandwiched between two windows transparent to IR wavelengths. In the example shown in Figure 8.2 a portion of the radiation (ideally 50%) passes through the beamsplitter towards m2, while the rest is reflected towards m1. Both branches of the beam are then reflected back to the beamsplitter, where the two beams converge and a part of the IR radiation exits the interferometer into the sample compartment while another part is directed back at the light source and is lost. A small amount of light is lost in the optics, meaning that at most only 50% of the original IR radiation actually exits the interferometer for transmission through the sample.
49
Important Instrumentation and Methods for Detection of Chemicals in the Field
Figure 8.2 – The Michelson Interferometer. LS = lightsource, BS = beamsplitter, m1 = moving mirror, m2 = stationary mirror, b1 = infrared beam with varying path length, b2 = infrared beam with fixed path length.
The modulation of the IR beam is achieved by moving one of the mirrors (m1 in Figure 8.2) so that the distance travelled by the beam b1 becomes longer than that of the beam b2 directed towards the stationary mirror m2. When the moving mirror is displaced by one quarter of the wavelength of IR light at a specific wavelength, the distance travelled by the beam b1 becomes one half of the wavelength longer than that of b2 and as a result the peaks and troughs of the two waves line up and the resulting beam has zero intensity as shown where optical density (OPD) ≈ λ/2 in Figure 8.3. When the mirror m1 is displaced further, the peaks of the two waves will be closer to each other until perfect alignment and maximum intensity for the beam exiting the interferometer is reached when the mirror m1 is displaced by a distance of one half of the wavelength and the optical path difference of beams b1 and b2 is equal to the wavelength. As the mirror moves continuously at a steady rate, the modulated output beam intensity will change as a cosine wave over time if there is only one wavelength present in the IR beam. As the beam contains all mid-IR wavelengths and no attempt is made to disperse or filter them, the resulting interference pattern will be more complex than in the example above, but the details of the resulting interferogram hold the information required to calculate the IR spectrum for the sample of interest. In order to record the interferogram with the IR detector, a means for determining the position of the moving mirror is needed. In most commercial FTIR instruments the mirror position is measured using a Helium-Neon gas laser which emits a monochromatic and very stable radiation at 632.08 nm in the visible range. This laser beam is made to pass through the interferometer co-axially with the IR beam, with a dedicated laser detector inside the interferometer. The interferogram recorded for the laser is a cosine wave and the wavelength of the laser beam is constant, allowing the position of the moving mirror to be tracked using the laser signal. This internal calibration of the moving mechanism 50
Figure 8.3 – Interference patterns for two waves (e.g., b1 and b2 of Figure 8.2) of constant and equal frequency, offset by various distances related to λ (wavelength), with corresponding sum waves (b1 + b2) immediately below each distinct interference pattern. The sum waves b1 + b2 represent the beam exiting the interferometer; OPD = optical path difference. When OPD ≈ zero, the sum wave is at its strongest, and when OPD ≈ λλ/2, it is at its weakest.
is one of the strengths of FTIR. With a suitably rugged and rapidly scanning mirror design an FTIR instrument can be reliably used in a wide temperature range of 0 to 40°C, while carried in a backpack, while mounted directly on a smokestack, or inside a moving off-road vehicle. In all of these situations, the wavenumber scale of the instrument remains stable and calibration is maintained remarkably well. Interferometer designs used today differ from Michelson’s original approach. One of the shortcomings of the original design was reliance on plane mirrors, which must remain perfectly perpendicular to the beams. A second limitation was the need to move one mirror over relatively long distances while keeping all other parts stationary. The first shortcoming is most often corrected by using cube corner mirrors(7), which consist of three plane mirrors at 90° angles to each other (see Figure 8.4). Such mirrors reflect light back to the original direction regardless of the mirror orientation, which makes it much easier to develop robust and simple moving mechanisms for the interferometer. The second
Figure 8.4 – The geometry of the inner mirrored surfaces of the corner of a cube is the basis for a retroreflector that returns the incoming beam back to its original direction after three reflections, regardless of the orientation of the cube corner itself. The original positions where the beam struck the mirror surfaces prior to tilting the top of the cube corner slightly to the left are depicted by the black dots in the panel to the right.
Chapter 8 — Field-Portable Fourier Transform Infrared (FTIR) Spectroscopy for Gas and Vapor Analysis
Figure 8.5 – Modern interferometer designs used in FTIR instrumentation: left, Michelson interferometer design with cube corner mirrors (used by various manufacturers); right, Gasmet Technologies GICCOR interferometer.
shortcoming has been addressed by a variety of advanced interferometer designs, where the motion of the mirror is either rotary, or both mirrors m1 and m2 are made to move in unison. Figure 8.5 illustrates two different modern interferometer designs used for FTIR instrumentation.
Sample Cell Compared to solid and liquid samples, IR measurements of air samples must account for the lower density of IR absorbing molecules in the gas phase, and the fact that the primary constituents of air, nitrogen and oxygen, do not absorb IR light. For this reason the optical density or absorbance per unit length of optical beam is always much lower in gas and vapor samples than in solids and liquids. This requires long path length designs, for example where the beam is reflected multiple times within a gas sample cell using mirrors. A number of approaches have been used to increase optical path length in a compact sample cell, including the Herriott cell design, but the best known folded-path gas sample cell design is the White cell.(9) Three spherical mirrors are used in this design, as shown in Figure 8.6. The beam from the interferometer is focused into an image at the entrance window of the cell, which is located in an opening on mirror m2. The beam is collected at the opposite wall of the cell by a spherical mirror m1, which refocuses the beam onto the mirror m2 between the entrance and exit windows. From mirror m2 the beam diverges onto the third mirror m3 from where the beam returns to a different spot on the mirror m2 than in the first pass. Different optical paths can be achieved by selecting the radius of curvature and its origin for the three mirrors, and depending on these specifics the beam will make a certain number of passes back and forth before focusing on the exit window of mirror m2, from where the beam passes to the detector. Optical paths up to 10 m are
Figure 8.6 – Schematic drawing of a White cell design with folded beam path to provide an interface for FTIR analysis of gas phase samples. Focusing mirrors outside of the cell guide the beam from the interferometer to the cell and from the cell to the detector.
routinely achieved in a relatively low sample cell volume of 0.5 L, while physically bigger cells can provide path lengths >100 m. Depending on the gas or vapor to be detected and its IR absorptivity, concentrations down to 100 ppb or less can be measured with a 10 m optical path length, and the use of a folded path is critical to attain this performance in a fieldportable system. Sample cell mirrors are typically coated with gold as it is a material with excellent IR reflectivity and is resistant to most chemicals. The windows in such cells are selected so that they do not absorb the wavelengths of light being measured, and they must also be resistant to the sampled gases and vapors. This precludes the use of hygroscopic materials otherwise in widespread use in many laboratory FTIR instruments, with zinc selenide (ZnSe) and barium fluoride (BaF2) available as possible choices. Light at wavenumbers below about 900 cm-1 is absorbed by BaF2, limiting its usefulness in some applications, e.g. for measurement of chlorinated organic compounds such as carbon tetrachloride. A wider
51
Important Instrumentation and Methods for Detection of Chemicals in the Field
wavenumber area is available when ZnSe is used, but due to its high index of refraction it suffers from signal loss due to reflectance. However when an antireflection coating is used, ZnSe is a durable and versatile material that is often used for gas cell windows. An open path approach is an alternative to a closed sample cell for detecting gas and vapor analytes and measuring their concentrations by FTIR. In an open path system an IR beam is transmitted through the ambient atmosphere. A bistatic open path FTIR instrument employs an IR source kept at a distance from the detector and interferometer, with the path length (in the simplest case) determined by the distance between the source and detector. A monostatic open path FTIR instrument uses a mirror placed at a distance from the source to reflect the IR beam back to a detector which is located with the source and interferometer. In addition to measuring gases and vapors with FTIR spectroscopy using a multipass transmission cell or open path design as explained above, it is possible to use an FTIR instrument to measure IR absorption from solid and liquid samples with the attenuated total reflection (ATR) technique. As the density of IR absorbing molecules in a solid or liquid is much higher than in the gas phase, it is not necessary to use a long path length design (as with an instrument designed for gas phase measurements). The ATR sample interface occurs at the surface of a crystal material (e.g., a diamond material which is transparent to IR radiation). The geometry and relatively high refractive index of a reflectance element is chosen so that the IR beam transmitted into a crystal undergoes internal reflection and does not directly enter the external sample (see Figure 8.7). However, a small portion of
the IR wavefront extends into the sample at the point where the beam is reflected, and this indirect interaction with the sample is enough to cause a measurable attenuation of the IR beam, hence the name “attenuated total reflection.” The use of an ATR sample interface allows for direct placement of a liquid or solid material to be tested above the reflectance element with no special sample preparation requirements. This allows an operator to potentially identify sample components present in solids, liquids or pastes at relatively high (i.e., percent) concentrations by automated comparison of FTIR spectra obtained with those from a reference library.
Detector Detectors for IR fall into two main categories, thermal detectors and semiconductor detectors, which in turn may be further divided to photoconductive and photovoltaic detectors. Radiation in the IR wavelengths changes the temperature of a thermal detector and causes a measurable change in either voltage (thermocouple type) or resistance (bolometer type) across a detector element. The most commonly-encountered thermal detector is the bolometer type, which is typically known by the type of material used, such as deuterated triglycine sulfate (DTGS) or deuterated l-alanine-doped triglycine sulfate (DLATGS). Thermal detectors have a good linearity of response across a wide wavelength range, but their overall peak detectivity is lower than that of semiconductor detectors. Also, a measurable response time is required for a detector which relies on the temperature change in a bulk material. For this reason the semiconductor detectors are favored in gas and vapor analyzers capable of rapidly scanning the IR spectrum (as fast as several scans per second).
Figure 8.7 – Close contact between a solid or liquid material and the protective coating that covers the internal reflective element of an ATR-FTIR instrument allows absorption (attenuation) of IR light that evanesces into the sample.
52
Chapter 8 — Field-Portable Fourier Transform Infrared (FTIR) Spectroscopy for Gas and Vapor Analysis
Semiconductor detectors employ a junction between two pieces of semiconductor material chosen so that the band gap or the energy needed to jump an electron in the semiconductor from the valence band (bound state) to the conduction band (free state) corresponds to the energy of the IR photons being measured. When IR light falls on such a semiconductor, some of the electrons are raised to the conduction band and the result is a small electrical current which can be measured. The selection of the semiconductor material determines the wavelength band where the detector may be used, and a common selection is mercury cadmium telluride (MCT), which is a mixture of HgTe and CdTe. The bandpass frequency where the detector starts to respond to photons can be determined by the mixing ratio of HgTe and CdTe in the detector material. Detectors with a higher bandpass frequency are insensitive to long wavelength radiation but have higher peak detectivity. This produces a trade-off between being able to measure long wavelengths (e.g. a wideband detector) or better sensitivity at short wavelengths (narrowband detector). See Figure 8.8 for a comparison of detectivity as a function of wavelength for semiconductor detectors. In a semiconductor detector random thermal motion of electrons occasionally causes electrons to shift into the conduction band, producing measurable thermal noise in the detector signal. For this reason it can be beneficial to cool semiconductor detectors to minimize this source of noise. For instruments used in laboratories the cooling medium is typically liquid nitrogen (-196°C). The use of cryogen is impractical for a portable field instrument, however thermoelectric detector cooling down to -35°C may be used in a field-portable instrument within the constraints of available power and size limitations.
Fourier Transform IR Data Processing After the interferogram has been recorded with the spectrometer, it must be converted into an absorption spectrum with the Fourier transform algorithm and associated signal processing steps (truncation, zero filling, filtering). The resulting spectrum must also be interpreted in order to perform either qualitative analysis (analyte identification) or quantitative analysis (determination of analyte concentration). This interpretation is typically performed using a computer integrated within the analyzer, a handheld computing device, or with a laptop computer. It is also possible to split the computation into two steps, so that the interferogram is converted into an IR spectrum with dedicated electronics inside the analyzer while the spectral interpretation is carried out in an external computer. The
Figure 8.8 – Detector response as a function of wavelength for various IR detectors. The detectors in question are all thermoelectrically cooled MCT semiconductor devices differing in the mixing ratio of Mercury, Cadmium, and Telluride. The quantity D* is specific detectivity, a figure of merit inversely proportional to noise-equivalent power normalized to detector surface area and bandwidth. Figure courtesy of Teledyne Judson Technologies.
following paragraphs deal solely with the analysis of the IR spectrum, and although the Fourier transformation is an important part of FTIR data processing, it is not discussed in detail in this chapter and the interested reader is invited to consult Griffiths and de Haseth(10) for more information.
Qualitative Analysis Identification of a gas or vapor from its IR spectrum involves comparing the spectrum produced from an unknown sample with a library of spectra for known analytes using one of several available library search routines. The sample spectrum is typically corrected for baseline slope and shift before conducting the library search, and the spectrum may be screened so that the use of those wavelength regions with very strong water and CO2 absorbance is avoided. In its simplest form the library search consists of calculating the correlation between a sample spectrum and each of the library spectra after absorption peak heights are normalized. The numerical correlation may be used as a quality control parameter, as higher correlation values correspond to more certain analyte identification. When a sample contains only a single IR-active component for which a corresponding library entry is available the library spectrum with the highest correlation is the most likely identification for the unknown analyte. However, in the case of analytes with spectra that are not included in the library, or for samples containing mixtures this approach would fail, and the calculated correlation to
53
Important Instrumentation and Methods for Detection of Chemicals in the Field
the closest reference spectrum found in the search would be lower than in the case of a successful search. This basic library search routine may be completed several different ways, (e.g. using the correlation of first derivatives instead of absorbance spectra), by calculation of Euclidean distance of a sample spectrum against library spectra, or by calculation of Mahalanobis distance between a sample spectrum and library spectra. The specifics related to these approaches are beyond the scope of this chapter, but in each case the result is a search for a single library spectrum that best fits the sample, leaving the identification of unknown mixtures problematic. Some spectral search routines enable subtraction of the best fitting library spectra from the sample spectrum, so that the process can be repeated for the next unknown in the mixture and so on. Another approach to identifying the unknown components of mixtures makes use of multicomponent analysis originally developed for quantitative analysis (explained below). A multicomponent library search starts by analyzing the unknown spectrum with a small subset of the search library, and more spectra from the library are added to the analysis and removed from the analysis automatically so that the best fitting group of library spectra is found. This type of analysis can identify several unknown analytes from a mixture in one run at the cost of somewhat longer searches, typically tens of seconds with a 5,000 compound search library, compared to a fraction of a second for a simple correlation search.
Quantitative Analysis The use of an FTIR spectrometer as a quantitative gas and vapor analyzer relies on the integration of spectrometry with chemometric or multivariate data handling approaches. These techniques extract information from a measured spectrum that can be linked directly or indirectly to the concentrations of IR-absorbing gases and vapors in the sample that produced the spectrum. The term “multivariate analysis” highlights the fact that the absorbance information at multiple wavelengths is obtained in a single measurement and the complete spectrum is used as the input data, which is in contrast with narrow waveband NDIR and dispersive or IR methods. Beer’s law Beer’s law (also known as the Beer-Lambert law) is defined by Equation 8-1:
54
Ai (ν) ˜ = ai (ν) ˜ b ci (8-1)
is the absorbance caused by gas or vapor i at where wavenumber , and ˜ is the molar absorption coefficient or absorptivity of the same gas at the same wavenumber. The constant b is the optical path length through the sample cell and ci is the concentration of gas or vapor i. If the absorbance (peak height) is not so large as to cause nonlinearity and the absorbance is all due to a single gas or vapor, the concentration of the gas or vapor in question can be calculated very easily. In practice strong absorbance bands of different analytes may overlap to varying degrees, and a multicomponent method of assessing the absorbance due to each gas or vapor, and the concentration of each is thus required. In the Beer’s law equation, absorbance A is taken from the measured spectrum and concentration c is unknown, while path length b is a constant determined by the hardware used. The undefined term in the equation above is the molar absorption coefficient, and this link is provided by a set of reference spectra with known concentration(s) of one or more gases/vapors in each spectrum. This set of reference spectra is called a reference library or training set depending on the type of analysis being performed. Classical Least Squares Analysis Classical least squares (CLS) analysis is based on the assumption that the absorbance A of a gas and vapor mixture is the sum of individual component absorbance values, which in turn are linked to concentrations by Beer’s law. This assumption holds true for dilute gas and vapor mixtures (for instance, occupational hygiene-relevant concentrations of volatile organic compounds in air), while for solid and liquid mixtures with strong interaction between IR absorbing molecules the assumption does not necessarily apply. Beer’s law can be written for multiple gases and vapors (denoted by index n) absorbing at multiple wavelengths (denoted by the index m) in the following way, with the parameters a and b being combined in a single parameter k:
A1 = k11C1 + k12C2 + … + k1nCn + noise A2 = k21C1 + k22C2 + … + k2nCn + noise
(8-2)
Am = km1C1 + km2C2 + … + kmnCn + noise The absorbance values A1 to Am in Equation 8-2 often comprise the complete sample spectrum or at least large sections of it (the areas with very strong water vapor and carbon dioxide peaks would likely be excluded), and the
Chapter 8 — Field-Portable Fourier Transform Infrared (FTIR) Spectroscopy for Gas and Vapor Analysis
Partial least squares analysis
values for k parameters are obtained from the library of reference spectra. If m = n so that there are as many A values as there are concentrations to be determined, a single solution would exist for this group of equations if not for the noise which is invariably present in FTIR spectra. For this reason the number of absorbance values (data points) from the sample spectrum always exceeds the number of concentrations to be determined and the best solution is found by a CLS regression. As the number of A, k, and C parameters is likely to be large, it is more convenient to express the above equations in matrix form:
A = K • C + residual
(8-3)
In Equation 8-3, A is a matrix vector containing the absorbance values of the sample spectrum, K is a matrix (n by m), and C is a matrix vector containing the unknown concentrations. The K matrix is computed by fitting the reference spectra against the sample spectrum so that the residual term is minimized. Once the best values for K are found, the above equation is easily solved for the C values. The above equation explains why the CLS analysis is sometimes referred to as the K-matrix method. An attractive feature of CLS analysis is that each row of the K matrix represents the spectrum of a pure gas or vapor in the library. If spectra for new gases or vapors are added to the library, there will simply be another column in the K matrix, which does not affect other components of the matrix. This means that a CLS method is easy to expand with the addition of more gases and vapors, as long as there are more absorbance data points (rows in A) than there are unknown gases (columns in K and C). In practice, the CLS method can be applied to 25 or more gases and vapors simultaneously when an entire spectrum is recorded from e.g. 900 cm-1 to 4000 cm-1, and the underlying assumptions about additive absorbance do not cause significant analysis errors. One of the main limitations is that all gases and vapors absorbing in the wavelengths A must be included in the method. If an unknown gas or vapor is present, then the CLS regression will fail, but this results in the spectrum of the unknown being shown in the residual. This can be used as a starting point for identification of the gas or vapor in question. Alternatively, the wavelengths A used for analysis can be selected so that only the gases of interest and a limited number of overlapping or interfering gases are likely to be included in the model.
The partial least squares (PLS) analysis model is mathematically more complex than the CLS analysis model, and it is capable of interpreting samples where the absorbance of a mixture is not the sum of component absorptions. The PLS method was developed to counter the limitations of CLS analysis, and it is in widespread use in the analysis of liquid sample spectra recorded with near IR spectrometers. The PLS method may also be applied to the analysis of gaseous mixtures with absorbance measured in the mid-IR spectrum. The PLS equations are omitted from this short review, but the key concept is that both the A and C matrices are analyzed together to give score and loading vectors, from which the concentrations of the unknowns are finally calculated. In order to achieve this, a larger number of reference spectra, usually called a training set in this context, are needed compared to a CLS approach. These spectra are frequently from mixtures of various gases and vapors and they can be real sample spectra containing different gas-phase analytes in concentrations determined by independent analysis. The PLS models are more complex to build than CLS models, and adding another component into a PLS model requires more extensive changes to the library of reference spectra than for CLS models. On the other hand, a correctly parametrized PLS model is a robust analytical tool for complex mixtures as long as the training set covers the range of variation in the sample spectra which will be analyzed with the PLS model.
Capabilities and Limitations of Fourier Transform IR Instrumentation A correctly-selected FTIR instrument can be an ideal tool for analysis of complex gas and vapor mixtures, or unknown solid and liquid samples. Compared with non-IR techniques such as gas chromatography (Chapter 9), FTIR has an advantage in that a gas or vapor sample may be drawn directly into a transmission cell, or a solid or liquid may be placed on a diamond interface directly above an ATR internal reflectance element with no need for further sample handling or preparation. When compared with a filter IR gas and vapor analyzer that often measures absorbance for a pre-determined analyte using only a single wavelength band, the capability to rapidly measure a complete IR spectrum with medium to high resolution offers several advantages. Molecules with very similar spectra (isomers and molecules with identical functional groups) can be distinguished from each other, interference from water vapor can be 55
Important Instrumentation and Methods for Detection of Chemicals in the Field
effectively minimized, and overlapping spectral peaks can be distinguished as the FTIR spectrum contains enough data to use sophisticated multivariate analysis methods. Lastly, the Fellgett advantage provides both improved speed and sensitivity compared to a filter IR method when a scan of a number of wavelengths is needed. An FTIR instrument does have some limitations. Gas phase analytes exist that cannot be measured with any IR technique, namely the elements that occur as diatomic gases: O2, N2, hydrogen (H2), and chlorine (Cl2) for example. Also, for solid and liquid samples, metals and inorganic salts are not detectable. While FTIR can differentiate absorbance bands that overlap to some degree, it cannot do so if one of the components present in a mixture has sufficient absorbance strength to block nearly all light in the wavenumber area where the measurement of another analyte is to be made. This means that in some cases a ppm-level component can be masked by a percent-level component, especially if the two are chemically similar. A further limitation deals with the chemometric models used. In a CLS-based model the library should contain all of the gases and vapors expected to be present in the sample spectrum, and in a PLS-based model the training set should be based on samples that are sufficiently similar to the sample being measured. A recorded FTIR spectrum can be post-analyzed in many cases by modifying the method, but this may not be practical for field analysis unless the FTIR instrument contains multiple preloaded methods which can be rapidly changed by the user.
Example Instrumentation and Case Studies Demonstrating the use of Field-Portable Fourier Transform IR Analyzers Attenuated Total Reflectance (ATR)-Fourier Transform IR Instruments Norman, et al. described an ATR-FTIR instrument designed for rapid field identification of chemical compounds amenable to IR analysis that are present in solids, powders, pastes, gels, or liquids.(11) The currently available version of the instrument described (Smiths Detection HazMatID™, Figure 8.9) weighs about 10 kg. This instrument incorporates an internal path length of 24 cm and operates in the mid-IR range (4000 to 650 cm-1). A low power diode laser is incorporated in the design to measure the location of the moving mirror within the interferometer, along with an electrothermally-cooled DTGS detector providing resolution of 4 cm-1. Analysis times 56
Figure 8.9 – HazMatID™, ATR-FTIR instrument used to identify unknown components of solid and liquid samples placed on the diamond internal reflectance element interface (1). The sample press (2) is used to compress solid materials to the surface of the internal reflectance element in order to maximize the potential for the wavefront of the internally reflected IR radiation to interact with the sample.
of 2 min are routine, and wireless communication capability allows transmission of spectra and results to a computer at a distance from the spectrometer. In the years since Norman et al. described an early configuration of this instrument several additional manufacturers have produced ATR-FTIR instruments that are compatible with field use that were introduced in 2012, including a handheld version weighing < 2 kg. While the ATR sample interface is primarily used for the FTIR analysis of solid and liquid samples Bryant et al.(12) used an ATR-FTIR spectrometer to demonstrate an FTIR analysis approach for organic vapors that didn’t depend on a White cell sample interface. The exterior surface of the instrument’s diamond sample interface was coated with a thin layer of liquid polymer sorbent to concentrate airborne nerve agent simulants as contaminated air was moved past the coated element. Spectral interpretation showed that the airborne simulant compounds could be identified by their unique FTIR spectra when absorbed into the polymer coating, and that ATR-FTIRabsorbance was correlated with the airborne analyte concentration.
Fourier Transform IR Instrumentation for Gas and Vapor Analysis The relevant characteristics for the DX4040™ FTIR instrument (the most current version of the instruments used in the following applications(13-15)) are described below. The 12 kg person-portable instrument features a silicon carbide (globar) IR source, He-Ne laser source to determine mirror position,
Chapter 8 — Field-Portable Fourier Transform Infrared (FTIR) Spectroscopy for Gas and Vapor Analysis
and moisture resistant ZnSe beamsplitter and windows. The interferometer scans at a rate of 10 scans/s with a medium resolution of 8 cm-1. This choice of scan speed and resolution is made to reduce sensitivity to vibration, shock and temperature change as the mirror movement is very short and the movement is rapid. When the IR spectrum is used for quantitative analysis, the relatively small number of data points in the spectrum (1024) is not a disadvantage as the number of gases or vapors to be analyzed is much smaller (no more than 25 simultaneously) and it is desirable to keep spectrum file size at manageable levels (about 5 kilobytes). The White type gas and vapor sample cell provides a 9.8 m optical path, enabling quantitative detection of sub-ppm volatile organic compound concentrations with a cell volume of 0.45 L and 2 L/min sample flow provided by a built-in sample pump. This provides a short response time, typically ≤ 1 min. The materials in contact with a gas phase sample are either polytetrafluoroethylene (PTFE) or Tygon® outside the sample cell, and the sample cell body and mirrors are coated with a corrosion resistant gold layer so that reactive gases and vapors may be analyzed. The IR detector is a thermoelectrically-cooled MCT type with response from 900 cm-1 to higher wavenumbers, giving good sensitivity over a fairly large wavelength range. The IR signal is digitized to give the interferogram and the interferogram is converted into an IR spectrum using dedicated digital signal processing circuits within the analyzer. The IR spectra obtained may be transferred to a rugged handheld computer through a wireless link. The handheld computer serves as the user interface to the FTIR instrument
and it may perform CLS analysis for up to 25 components. Several different predefined methods are stored on the computer and the operator can quickly switch between methods if the samples to be measured are too diverse to be analyzed with a single method. Spectra are also stored on the handheld device and can be transferred to a computer running advanced analytical software for identification of unknowns and analysis of complex gas mixtures. The analyzer can be carried with a shoulder strap or backpackstyle harness and the interferometer can operate in any orientation, also when carried by a moving person. Biomedical Screening A number of examples may be cited where portable FTIR instruments were used to complete quantitative field measurements for gas and vapor analytes. In 2004, Laakso et al. described the use of a portable FTIR instrument to screen exhaled breath of patients seen at hospital emergency rooms in Finland.(13) The Gasmet FTIR instrument used in this study was fully portable, with a weight of 18 kg, and was capable of operation on battery power for up to 8 h. It employed a 200 mL multipass White cell interface, and an electrothermallycooled MCT detector. This work demonstrated the capability for a portable FTIR instrument to rapidly provide data that could be used either clinically, or for biomedical research. Elevated carbon monoxide CO was detected in the exhaled breath of most smokers, providing a method to systematically differentiate patients who smoke from those who do not (Figure 8.11) based on empirical noninvasive measurements that may be obtained very quickly. The quantitative data provided by these researchers also showed that relatively high methane (CH4) concentration in exhaled breath correlated with increasing age. In addition to the ethanol concentrations measured in the exhaled breath of numerous patients who had consumed alcohol (which can be correlated to blood concentration), methyl ethyl ketone was also detected in the breath of a patient who had apparently consumed alcohol that had been denatured with this compound. Measurement of Anthropogenic Influence on Atmospheric Gases
Figure 8.10 – Gasmet DX4040™ person-portable FTIR instrument in use for identification of unknown gas-phase analytes in the field.
In 2008, Guérin et al. studied the impact of reservoir construction in tropical regions on the net emission of nitrous oxide (N2O) compared to natural “pre-flooding” N2O emission rates. This information was desired as N2O is known to have greenhouse gas properties. These researchers used a plastic chamber floating on the surface of reservoirs located in a tropical region to collect gases emitted from the water. 57
Important Instrumentation and Methods for Detection of Chemicals in the Field
two open path monostatic FTIR instruments to sequentially scan eight retroreflectors positioned around an area where sulfur hexafluoride (SF6) had been released.(16) Approximately sixteen open path measurements were obtained every five minutes, and the spectral data were processed using tomographic reconstruction software to create twenty five sequential two-dimensional concentration maps.
Conclusion Figure 8.11 – Relative frequency polygons for exhaled carbon monoxide concentrations of nonsmokers (shaded area, n = 363) and smokers (no shading, n = 207). The y-axis values represent the proportion of patients in each 0.5-ppm (nonsmokers) or 4-ppm (smokers) interval. From Laakso et al.,(13) copyright Oxford Journals, used with permission.
On-site analysis of the chamber gases to determine N2O and CH4 concentrations was completed using a portable FTIR instrument similar to that described by Laakso et al. immediately above. Measurements at the air/soil interface in the same region completed with the same instrument (by using a stainless steel gas collection chamber in contact with soil) allowed the researchers to estimate the change in N2O emission with construction of the reservoirs, relative to emission from undisturbed soil in the same region.(14) Building Air Quality Teye and Hautala used a similar field-portable FTIR instrument with a gas cell interface to measure airborne ammonia (NH3), CH4, CO2, and water vapor concentrations in a dairy building. The primary purpose of this part of their study was to quantitatively measure NH3 emissions as “high ammonia concentrations in animal buildings are known to affect the welfare of animals, workers, and the life span of the building.”(15) Concentration Mapping Where dangerous chemicals are routinely handled with the possibility for unintentional release to the open atmosphere, concentration mapping could provide important information to emergency responders and civil authorities. However, for emergency situations where this type of instrumentation was not already set up, little would be known initially about potentially dangerous atmosphere conditions and technicians could be exposed to unknown hazards in order to set up such a system. In order to demonstrate the potential to obtain 2-dimensional concentration plots for airborne gases or vapors amenable to FTIR detection, Todd simultaneously operated 58
Person-portable FTIR instruments have become widely available in the recent past, and have seen use in many applications. These portable FTIR spectroscopy instruments are available in configurations to identify unknown chemicals in liquid and solid samples (ATR-FTIR instruments) and to identify unknowns and measure organic and inorganic gas and vapor concentrations (FTIR instruments employing a gas cell interface). The primary strengths of FTIR are the ability to adapt to a wide range of measurement tasks by simply changing the spectral library and analysis settings on a computer, and the ability to operate without carrier gases or calibration reference materials. When an FTIR instrument is operated for identification of unknown gases or vapors spectral library matching may provide the identity of the most prominent component in the sample (but not necessarily its concentration). When used for quantitative measurements it is possible to accurately determine concentrations for multiple gases and vapors in the ppm concentration range. The two modes of operation can be combined in a single instrument with the quantitative analysis relying on a chemometric model that includes data for substances of interest, against which the sample is to be compared. The recorded IR spectra are saved together with the analysis results, which makes it possible to verify the analysis results later and to recover data from measurements where the sample contains gases or vapors that were not initially included in the instrument library. Due to continued development efforts handheld ATR-FTIR instruments are now available to support military chemical detection and identification missions, and civilian hazardous material incident responders. In comparison to well-known filter IR instruments designed to measure airborne gas and vapor hazards FTIR instrumentation has greatly expanded the capabilities for rapid identification of airborne hazards. Both types of FTIR instrumentation may provide high quality information in the field to exposure assessment professionals, improving our ability to protect workers and the general public.
Chapter 8 — Field-Portable Fourier Transform Infrared (FTIR) Spectroscopy for Gas and Vapor Analysis
References 1. Michelson, A.A.: On the application of interferencemethods to spectroscopic measurements, I. Philos. Mag. 31:338–46 (1891). 2. Michelson, A.A.: On the application of interferencemethods to spectroscopic measurements, II. Philos. Mag. 34:280–99 (1892). 3. Ferraro, J.R.: History of Fourier transform-infrared spectroscopy, Spectroscopy 14:28–40 (1999). 4. Michelson, A.A.: Recent advances in spectroscopy. Nobel lecture, December 12, 1907. 5. Rayleigh, L.: On the interference bands of approximately homogeneous light: in a letter to Prof. A. Michelson. Philos. Mag. 34:407–11 (1892).
attending emergency departments. J. Anal. Toxicol. 28:111–17 (2004). 14. Guerin, F., G. Abril, A. Tremblay, and R. Delmas: Nitrous oxide emissions from tropical hydroelectric reservoirs. Geophys. Res. Lett. 35:L06404, doi:10.1029/2007GL033057 (2008). 15. Teye, F.K. and M. Hautala: A comparative assessment of four methods for estimating ammonia emissions at microclimatic locations in a dairy building. Int. J. Biometeorol. 54:63-74 (2010). 16. Todd, L.A.: Mapping the air in real-time to visualize the flow of gases and vapors: occupational and environmental applications. Appl. Occup. Environ. Hyg. 15:106–13 (2000).
6. Cooley, J.W. and J.W. Tukey: An algorithm for the machine calculation of complex Fourier series. Math. Comput. 19:297–301 (1965). 7. White, R.L., P.J. Coffey, and D.R. Mattson: High performance FTIR using a corner cube interferometer. Am. Lab. 15:90–93 (1983). 8. Levine, S.P., Y. Li-Shi, C. Strang, and X. Hong-Kui: Advantages and disadvantages in the use of Fourier transform infrared (FTIR) and filter infrared (FIR) spectrometers for monitoring airborne gases and vapors of industrial hygiene concern. Appl. Ind. Hyg. 4:180–87 (1989).
From: “Important Instrumentation and Methods for the Detection of Chemicals in the Field”. Chapter 8 on FTIR Gas Measurement by the real time detection chapter of the AIHA ISBN-13: 978-1-935082-41-5
9. White, J.U.: Long optical paths of large aperture. J. Opt. Soc. Am. 32:285–88 (1942). 10. Griffiths, P.R. and J.A. de Haseth: Fourier Transform Infrared Spectrometry, 2nd edition. Hoboken, NJ: John Wiley and Sons, Inc., 1986. 11. Norman, M.L., A.M. Gagnon, J.A. Reffner, D.W. Schiering, and J.D. Allen: An FT-IR point sensor for identifying chemical WMD and hazardous materials. SPIE Proc. Ser. 5269:143–49 (2004). 12. Bryant, C.K., P.T. LaPuma, G.L. Hook, and E.J. Houser: Chemical agent identification by field-based attenuated total reflectance infrared detection and solid-phase microextraction. Anal. Chem. 79:2334–40 (2007). 13. Laakso, O., M. Haapala, T. Kuitunen, and J.-J. Himberg: Screening of exhaled breath by low-resolution multicomponent FT-IR spectrometry in patients
ETA Process Instrumentation 119 Foster Street, Bldg #6 Peabody, MA 01960 Tel: (978) 532-1330 Fax: (978) 532-7325 w ww . e t a p i i . c o m
[email protected]
59
