A Technology Comparison of Near-Infrared Spectroscopy

Near-infrared spectroscopy (NIRS) is an extensively used analysis technique for the identification of raw materials and quantification of substances in industrial and research applications.

This article aims at clarifying differences and similarities between dispersive and Fourier transform analyzers.

Introduction

In the early 1940s, the first commercial dispersive spectrometers emerged for UV-Vis applications. This technology is based on the dispersion of light in dependence of its wave-lengths. The dispersion of light is most normally illustrated using a prism, even though today, the prism is usually replaced by monochromator gratings. Dispersive spectroscopy refers to the method of choice for analysis in the UV-Vis range, but also for near-infrared (NIR) and photoluminescence applications.

Commercial Fourier transform infrared (FT-IR) spectrometers initially appeared in the 1960s. They were almost solely used in basic research fields as they were expensive. FT-IR spectrometers have become more affordable due to technical improvements. They are chiefly used in the mid-infrared (MIR) range for the identification of chemical substances and only recently did they became popular for applications in the near-infrared (NIR) range ^[2].

Both types of spectrometers were steadily improved since their first commercial appearance. They are both state-of-the-art.

In this article, dispersive and Fourier transform technologies are compared from an experimental point of view. The article will explain the design and the physical principles underlying each spectrometer type and it will also elucidate vital aspects concerning NIR applications. Vital aspects include the spectral resolution as well as wavelength accuracy and precision, wavelength range, the signal-to-noise (S/N) ratio and data acquisition speed.

Technical Overview

Dispersive Spectrometers

Polychromatic light is emitted from a light source and diffracted on a grating. The diffraction depends on the width of the grooves of the monochromator grating, the incident angle and the wavelength. In the reflected beam, the wavelengths composing the light are spatially distinguishable from one another. Monochromatic light irradiates the sample through an exit slit. To enable scanning of different wavelengths, the monochromator grating is mounted on a motor, which changes its angular position. The motor is controlled by a digital encoder for a repeatable and precise variation of the angle. Thus, monochromatic light of different wavelengths sequentially passes through the exit slit in order to record the intensity spectrum (see Figure 1) ^[1].

FT Technique

FT spectrometers contain an interferometer, which is composed of a beam splitter and two mirrors, one fixed while and one movable, whose distance to the beam splitter is variable (see Figure 2). Polychromatic light from the source is divided into two beamlets, one reaching the fixed mirror while the other is reflected towards the moving mirror. The beamlets are combined again at the beam splitter.

The resulting light intensity depends on the phase difference between the two beamlets, which is caused by the offset of the movable mirror. The detected intensity of the polychromatic light in function of the mirror offset is called interferogram. This is transformed mathematically in order to obtain a spectrum using the Fourier transform. Determination of the mirror offset of the moving mirror is made by a laser ^{[1, 2]}. Using interferometry, light is not monochromatized and all wavelengths are measured at the same time.

Figure 1. Illustration of the predispersive Metrohm XDS series and Metrohm DS2500 spectrometer in diffuse reflectance mode.

Figure 2. Illustration of an FT spectrometer with Michelson interferometer in reflectance mode.

Dispersive Spectrometer or FT?

When choosing the suitable technique several parameters have to be considered. The most important ones, i.e. the resolution, the wavelength range, the photometric accuracy and precision, the signal-to-noise (S/N) ratio and the data acquisition speed are discussed here.

Wavelength Range

The wavelength range of FT-NIR systems is limited due to the optics. It commonly ranges from 800 nm to 2500 nm. Dispersive spectrometers enable data acquisition down to 400 nm and even below, thus including the visible spectral range (Vis). This allows applications in which the sample parameters of interest are chiefly situated in the Vis-range, such as concentration and color strength measurements and also the quantification of chromatic complexes, amongst others.

Resolution

It is possible to fix the spectral resolution of a dispersive system at a constant value by fixation of several instrument parameters such as the density of the grating steps, the entrance and exit slit widths, the size of the spectrometer, the quality of the encoder, etc.

In Metrohm NIRS XDS and DS2500 spectrometers, the resolution is fixed at 8.75 nm, which is considered to be more than sufficient for most applications. This becomes clear when the natural bandwidth of the signals in NIR is closely inspected. Interferometers, on the other hand, permit the resolution to be defined by adjusting the maximum offset of the movable mirror (also known as «Connes' advantage»). Thus, more highly resolved spectra can be obtained than with dispersive systems. Spectra are typically recorded with resolutions of 8 cm^-1 or 16 cm¹; higher resolutions are not frequently used. At 2500 nm, these resolutions correspond to ~ 10 nm and ~ 25 nm, respectively ^{[3, 4]}.

Besides those stated above, there are two other reasons for not using higher resolutions. The first is that, harmonics and combination bands of pure substances in the NIR range have natural bandwidths bigger than 10 nm. Thus, a higher spectral resolution is not needed. Spectral bands can get even broader, when measuring mixtures, thus a higher resolution is not needed for reliable results. Only a few minerals, rare earth oxides and gaseous samples have bandwidths smaller than 8 nm.

Table 1. Examples of commonly analyzed substances and their bandwidths at the respective characteristic wavelengths.

The second reason to not use resolutions higher than 8 cm^-1 with FT-NIR systems is that an increase in resolution is constantly accompanied by an increase in spectral noise. To obtain a higher resolution, the maximum offset of the moving mirror and the mirror velocity have to be increased in order to attain spectra within the same time as dispersive spectrometers. However, this increases the noise exponentially.

To lower the noise level, either more spectra have to be co-added or the acquisition time has to be drastically increased ^[5]. The practical benefits of the higher resolution remain largely unexplored but could be beneficial in some applications ^[2].

Photometric Precision and Wavelength Accuracy

FT spectrometers are calibrated by collecting a high-resolution spectrum of water vapor, guaranteeing a precise and accurate recording of the wavelength response. Modern NIR monochromators, such as those used in the Metrohm NIRS XDS and DS series, are controlled by an extremely accurate digital encoder to guarantee maximum repeatability.

The spectrometers are calibrated using external and internal standards comprising of rare earth oxides. This calibration concept (standardization of bandwidths, photometric response and response wavelength) with certified reference standards allows direct transfer of spectra and chemometric models.

Data Acquisition Speed

It is possible to compare the data acquisition speed of both spectrometer types, FT and dispersive. Both technologies can attain two scans within a second ^[6]. Some of the main benefits of dispersive spectrometers are the wide wavelength coverage, the very low noise level, which results in an extremely high S/N ratio. According to Shaw and Mantsch (1999), “There are a wide variety of applications that require this combination of speed and accuracy.” ^[2]

Table 2. Comparison of the specifications of FT-NIR and predispersive spectrometers ^[6].

Signal-to-Noise (S/N) Ratio

In NIR spectroscopy, the most important parameter in the acquisition of spectra is generally not the accuracy, resolution or repeatability, but the signal to noise (S/N) ratio.

In UV-Vis and NIR spectroscopy, the main source of spectral noise does not derive from the extremely sensitive detectors (PbS or InGaAs used for NIR spectrometers are 1000 times less noisy when compared to the detectors used in the MIR ^[6]) but from the light intensity: the noise is directly proportional to the light intensity and inversely proportional to the wavelength.

In FT-NIR instrumentation, photon noise is superimposed in the interferogram and the Fourier transform cannot reassign these individual contributions to the corresponding spectral ranges. This disadvantage of distributed noise in FT-NIR can be problematic, for instance, when information of interest is in low-intensity regions at high noise ^{[7, 8]}. By contrast, dispersive systems sequentially scan all wavelengths, so that each absorbance measurement is independent and the noise is directly linked with it. With dispersive systems, it is possible to alter the optical path for a perfect signal amplitude, regardless of areas with higher noise levels.

The new Metrohm NIRS monochromators are based on the patented XDS (off-axis digital synchronous) and DS technologies, which guarantee an ideal focus of the monochromatic beam, thanks to the precise setting of the angular position by the encoder.

These innovations yield an unrivaled, nearly constant noise level over the whole Vis-NIR spectral range from 400 to 2500 nm, while the noise level of FT systems increases dramatically towards the spectral limits due to the optics (see Figure 3). Additionally, the S/N-ratio of dispersive systems is 2–60 times greater than those of FT systems¹.

Figure 3. Noise spectra acquired with the Metrohm NIRS DS2500 and an FT-NIR instrument in reflection mode using a reference material with an absorbance between 0.25 and 0.4. On the FT system, the scanning speed was adjusted to match the same data acquisition time of the Metrohm NIRS DS2500 (~ 20 s) with FT data acquisition parameters: double-sided bi-directional interferogram; phase resolution twice as high as set spectral resolution; scanning speed: 5 kHz to 10 kHz; Blackman-Harris 3-term apodization window; Mertz phase correction.

¹ To obtain these results, the measuring parameters of FT systems were adapted to ensure that data acquisition times were the same as those of a NIRS DS2500 spectrometer for better comparability. An external standard was used and a band was evaluated at 975 nm. The same FT data acquisition and processing parameters were used as described in Figure 3.

The S/N ratio is critical for quantitative and qualitative applications. To detect non-compliant products, for example, pills that lack active ingredient, it is necessary to have a high-quality signal since the absorption coefficients are small in NIR spectroscopy. The spectral deviation brought about by low constituent concentrations can be confused with noise. In this situation, non-compliance would not be detected. Finally, while noise criteria are less important for identification methods based on spectral correlation, noise nevertheless impacts the number of samples needed to build up a library. It is possible to build more robust spectral libraries with dispersive systems.

Because high noise levels lead to high spectral variation, it is possible to develop more robust qualitative and quantitative models using dispersive systems ^[9]. As the very near infrared range (800 to 1100 nm) is detected at a constantly low noise level, models with improved analytical figures of merit are possible with dispersive Vis-NIR systems ^{[6, 10]}.

Other Technical Considerations

Currently used monochromators realize a rapid scanning of the whole spectrum (~ 0.5 s). The acquisition time can be compared to that observed with interferometry with comparable resolution.

In FT instruments, the sample is illuminated by the entire spectral range of the strong light source used at once. The resulting high beam power density is capable of heating up the samples. This is a disadvantage when dealing with photosensitive samples, which might deteriorate under such conditions, but also when dealing with biological samples whose overheating can bring about accelerated bacterial growth or denaturation of sensitive proteins. With predispersive systems, light is monochromatized before scanning the sample; thus, the sample is exposed to a much lower beam power density, preventing damage.

Interferometry requires the use of a laser for the precise measurement of the position of the movable mirror. This laser can be considered a costly consumable as its service life is limited. FT-NIR systems are generally considered to be less robust than monochromators and are less suitable for online, atline and inline process applications, where strong vibrations can take place, generating potential misalignment of the moving parts and thus leading to distortions ^[6].

Almost all available spectrometers use desiccants to control the humidity of the system to avoid aging because of the optical material used in FT-NIR spectrometers. These desiccants have to be regenerated on a consistent basis, which can lead to downtime of the spectrometers. The robust optics used in the Metrohm NIRS XDS and DS2500 spectrometers do not need desiccants. It is possible to eliminate several maintenance steps.

The use of dispersive systems is straightforward when compared to the use of FT systems, where many parameters can be modified, heavily influencing spectra. Even though the mechanics for the positioning of the moving mirror are somewhat elaborate and its position is determined by a reference laser, phase errors can take place, which have a large impact on the spectra. To get rid of them, several phase corrections can be applied (such as Mertz correction, Power spectrum…). However, this needs an experienced user who is familiar with these methods.

Conclusions

Dispersive spectrometers are frequently used in UV/Vis and Vis-NIR instruments. FT instruments were originally developed for the MIR range to enhance the measurement quality, because MIR spectroscopy users are confronted with strong absorptions, low intensity light sources and noisy detectors. Compared to dispersive techniques (Felgeltt's, Jacquinot's and Connes' advantages), FT instrumentation has indisputable advantages in the MIR range (2500–25000 nm; 4000 to 400 cm^-1). Many manufacturers of spectrometers apply the FT technique for the NIR range as well, but the benefits do not have the same impact in low-wavelength ranges (800 to 2500 nm; 12500 to 4000 cm^-1).

The nature of NIR spectroscopy differs from MIR spectroscopy. For NIR, there are weaker absorptions, strong sources and quieter detectors. Advantages perceived in FT-MIR are not necessarily applicable to FT-NIR. Additionally, gratings are extremely reproducible and amenable for information transfer between instruments.

Many scientific articles determine equivalent performance of dispersive and FT systems for routine analysis ^[11–13]. Some studies demonstrate better predictive capabilities with monochromators, i.e. dispersive systems, and recommend this type of technology for offline use in laboratories ^{[6, 14]}.

The choice of NIR instrument has to be evaluated according to many economical and technical characteristics of the application: spectral range, sampling mode, resolution versus signal-to-noise ratio, versatility of the instrument, robustness of the spectrometer, support and supplier responsiveness, control software and chemometric tools, as well as cost of the instrument and its maintenance and repair ^{[5, 6]}.

Summary

While dispersive spectrometers are mainly used for quantitative applications in UV-Vis (200–800 nm), spectroscopy using Fourier transform (FT) techniques is mostly applied for line analysis and identification in the mid-infrared (MIR) region (3000–50000 nm). For the near-infrared (NIR), the spectral range between the visible (Vis) and MIR range, both FT and dispersive technologies are used. Both spectroscopic technologies are comparable in performance because of the continuous technological advancements. The decision between dispersive and FT-NIR spectrometers cannot be made a priori, but relatively depends strongly on the application.

References

[1] D. Bertrand et al., La spectroscopie infrarouge et ses applications analytiques, Tec&Doc, 2000.

[2] R. A. Shaw and H. H. Mantsch, «Near-IR Spectrometers», in Encyclopedia of Spectroscopy and Spectrometry, Academic Press, 1999, pp. 1451–1461.

[3] T. Meyer et al., «Suppression of mechanical noise and the selection of optimal resolution in FT-NIR spectroscopy», NIR news, 17 (8), pp. 12-14, 2006.

[4] O. Kolomiets et al., «The influence of spectral resolution on the quantitative near infrared spectroscopic determination of an active ingredient in a solid drug formulation», Journal of Near Infrared Spectroscopy, 12 (5), pp. 271–277, 2004.

[5] FOSS, NIR Spectrometer Technology Comparison, White paper from FOSS, 2013.

[6] E. Ciurczak et al., «Examination of NIR Spectrometers: Dispersive vs. Interferometric Types», Amer. Pharm. Rev., 11 (4), pp. 3-5, 2008.

[7] E. Voigtman, «The Multiplex Disadvantage and Excess Low- Frequency Noise», Applied Spectroscopy, 41 (7) pp. 1182- 1184, 1987.

[8] F. Grandmont, «Développement d'un spectromètre imageur à transformée de Fourier pour l'astronomie», Doctoral Dissertation, Université Laval, 2006.

[9] J. B. I. Reeves and C. M. Zapf, «Discriminant analysis of selected food ingredients by near infrared diffuse reflectance spectroscopy», J. Near Infrared Spectroscopy, (5), pp. 209– 221, 1997.

[10] D. Cozzolino et al., «The use of visible and near-infrared reflectance spectroscopy to predict colour on both intact and homogenised pork muscle», LWT - Food Science and Technology, 2003.

[11] P. R. Armstrong et al., «Comparison of Dispersive and Fourier-Transform NIR Instruments for Measuring Grain and Flour Attributes», Applied Engineering in Agriculture, 22 (3), pp. 453-457, 2006.

[12] A. Kazeminy et al., «A comparison of near infrared method development approaches using a drug product on different spectrophotometers and chemometric software algorithms», Journal of Near Infrared Spectroscopy, 17 (5), pp. 233–245, 2009.

[13] A. Mouazen, «Near infrared spectroscopy for agricultural materials an instrument comparison», Journal of Near Infrared Spectroscopy, 13 (2), pp. 87–98, 2005.

[14] P. Chalus et al., «Comparaison de spectromètres proche infrarouge pour la détermination de principes actifs dans des comprimés», Spectra Analyse, 247, pp. 44–49, 2005.

This information has been sourced, reviewed and adapted from materials provided by Metrohm AG.

For more information on this source, please visit Metrohm AG.