About Near-infrared Spectroscopy
Near-infrared Spectroscopy (NIRS) is molecular spectroscopy. It saves work and money by avoiding costly destructive analysis of the sample. In exchange for saving "front-end" laboratory analysis, the work of NIRS is at the "back-end". It is statistical and computer-based. The result is that NIRS:
- Offers analysis with the same accuracy and reproducibility as reference (conventional chemical) methods
- Involves simple installation with no drainage or fume exhaust
- Requires little or no sample preparation
- Offers wide flexibility in sample presentation
- Permits analysis of large samples that reduces sampling error
- Is simple and completely safe to operate
- Analyzes samples rapidly without destroying them (non-invasive), making it suitable for analysis of live material such as in plant breeding or medicine
- Analyzes numerous constituents simultaneously
- Describes both compositional and functional properties
- Analyzes samples in the laboratory, on-site, or in the field
- Requires no chemicals and involves no disposal costs
Theory and Operation
Near-infrared spectroscopy is a 45-year old technology developed by Karl H. Norris at the US Department of Agriculture in the 1960's. At that time, computers were emerging with sufficient power to analyze large amounts of spectral information from agricultural samples in the near-infrared (NIR) region of the electromagnetic spectrum, in combination with compositional data.
The region, 750 to 2500 nm, includes wavelengths longer than visible (Vis) light and shorter than infra-red (IR) radiation. It is a spectral region that has been ignored by classical spectroscopists and by analytical chemists in favour of colorimetry, mid-infrared spectroscopy, and techniques based on other spectral regions. This is because NIR absorbance bands are broad, overlapping, and matrix-dependent.
The absorption of NIR radiation by organic molecules is due to overtone and combination bands primarily of O-H, C-H, N-H and C=O groups whose fundamental molecular stretching and bending absorb in the mid-IR region. These overtones are anharmonic, i.e., they do not behave in a simple fashion, making NIR spectra complex and not directly interpretable as in some other spectral regions.
The primary advantage of the NIR region is that absorbances are lower than in neighbouring regions and generally obey the Beer/Lambert law, i.e., absorbance increases linearly with concentration. This is because NIR absorptions are generally 10-100 times weaker in intensity than the fundamental mid-IR absorption bands. The weakness of the absorptions is a benefit, providing direct analysis of samples without dilution or the requirement of short optical pathlengths or dispersion in non-absorbing matrices used in traditional sampling techniques in UV/Vis and mid-IR spectroscopies.
Despite the intuitive disadvantage of broad and overlapping absorption bands, modern chemometric techniques can extract meaningful information from the NIR spectra. The information about samples in the NIR spectra could not easily be accessed until the advent of sufficiently powerful computers that allowed the development of statistical relationships between the spectral data and constituents or parameters (e.g., functional properties) determined by conventional techniques. These statistical relationships between the spectral data and data from reference analyses are called calibrations.
The single step in NIR analysis requiring the most planning preparation is the assembly of the samples, often called the training set, to be used for the development of calibrations. A crucial step in achieving success is ensuring that the samples have been analyzed as accurately and precisely as conventional techniques allow. These analyses are termed reference analyses.
Once calibrations are obtained, they are entered into the NIR spectrophotometer. Following the scanning of unknown samples, requiring milliseconds to < 2 min per sample, numerous constituents or parameters of interest can be predicted. NIRS is a rapid, cost-effective, non-destructive, accurate and efficient analytical method.
Advantages
Unlike most conventional analytical methods, NIRS is non-destructive, requires little or no sample preparation, does not use chemicals, or generate chemical wastes requiring disposal. The technique is totally safe to operate, rapid, can be portable, and simultaneously determines numerous constituents or parameters. NIRS instrumentation is simple to operate by non-chemists, and operates without fume hoods, drains, or other installations.
Drawbacks
A primary drawback of NIRS is that it is not a stand-alone technology. Its accuracy is dependent upon the accuracy and precision of the reference method, although its predictions can be more reproducible than the reference method. Separate calibrations are required for each constituent or functionality parameter. To ensure that calibrations remain reliable, the accuracy of calibrations should be monitored by periodical analysis of some of the samples being predicted by the reference method. It may be necessary to update calibrations several times during the initial phases of use to incorporate samples representing new variance not encountered previously, until the calibrations become highly robust.
Calibration Procedure
Calibrations are statistical operations performed on spectral data to obtain an optimal statistical relationship between the spectral data and the reference data. The calibrations are the basis of predicting future, unknown samples in a rapid manner. In common with conventional spectrophotometric analysis, the calibration provides a simple linear regression relationship between spectral data and concentration of a constituent.
There are several comprehensive software packages available for the development of calibrations. These use common mathematical procedures, including multiple linear regression (MLR) and principal component analysis/partial least squares regression (PCA/PLS). More advanced statistical techniques include genetic algorithm and artificial neural networks, which mimic aspects of biological behaviour functions.
Sample Requirements for Development of Calibrations
The sample requirements for developing an NIR calibration are:
- minimum of 50 samples, but preferably 200 or more samples that have been accurately analyzed for the constituents or parameters of interest by conventional (reference) methods
- samples should represent the full range of concentration of the constituents of interest anticipated in future samples to be predicted by the calibration and, ideally, highest values should be at least twice and preferably ten times the lowest values
- samples should represent the range of physical and chemical composition anticipated in future samples to be predicted by the calibration
- samples should represent the range of environmental conditions, such as temperature
- samples should be chemically and physically unchanged between the time the constituents are scanned by NIRS and analyzed by conventional methods
- samples should be physically uniformly mixed so that the aliquots analyzed by conventional methods and those scanned by NIRS are truly representative of each other. Ideally, the actual sample analyzed by the conventional methods and scanned by NIRS should be the same. In practice, this may not be feasible.