Near-infrared spectroscopy sits in a slightly awkward wavelength space between ultraviolet-visible and infrared absorbance spectroscopy, both of which are commonplace in the analytical laboratory. Though often neglected as an analytical technique, near-infrared spectroscopy has enormous potential for a variety of applications, and is accessible now more than ever thanks to the availability of compact, high-value NIR spectrometers. In this note we provide a brief introduction to near-infrared spectroscopy, from its principles to typical applications. We’ll also demonstrate the suitability of our breakthrough Flame-NIR spectrometer for handheld and portable applications requiring NIR and chemometric analysis.
What is NIR Spectroscopy?
While absorption of ultraviolet and visible electromagnetic waves (200-700 nm) excites electrons to higher lying states, (mid-)infrared radiation spanning 2.5-25 µm (400-4000 cm-1 in frequency) stimulates vibrational motion in molecules. At first sight, near-infrared radiation seems to be stuck in the middle: not energetic enough to cause electronic excitation and too high in frequency for fundamental vibrational motions. However, vibrational motions involving hydrogen bound to a heavy atom — usually carbon, oxygen or nitrogen — can also be excited via so-called overtones or combination bands, a one-step jump skipping some of the rungs in the ladder of vibrational states. These higher-energy transitions fall in the near-infrared range of the electromagnetic spectrum, with wavelengths of 800-2500 nm (4000-12500 cm-1 in frequency).
For many years, spectroscopists did not see any additional value in observing a smaller set of vibrations with weaker absorption bands (more than 100x weaker than fundamental vibrations), particularly as they also tended to be non-specific, broad and overlapping. It took not only a new mindset and a fresh look at spectral analysis, but the abandonment of Lambert-Beer’s trusted law and the addition of today’s computing power to realize the potential of near-infrared spectroscopy. In the absence of distinct, easy-to-assign individual vibrational absorption bands, a multi-wavelength approach was needed to unlock the information contained in the overlapping bands of the NIR.
Chemometrics has been that key, enabling near-infrared spectroscopy to gain acceptance as a fast, powerful and convenient analytical technique. The same factors that were once considered drawbacks have been turned to advantages:
1) Detection of any bond to hydrogen in the molecule allows the technique to be quite universal
2) Deep penetration into the sample due to the weak absorbance carries the benefit of minimal sample preparation.
Near-infrared spectroscopy is now proven as a reliable method for noninvasive and nondestructive analysis of solids and powders, enabling applications in online process analysis, mineral testing, food, agriculture and medicine.
The Peculiarities of NIR Measurements
Just as the analysis of NIR spectra is conducted a little bit differently than in the UV-Visible or mid-infrared, so is the measurement. This is not necessarily an issue, as the benefits are clear, particularly for opaque samples and powders, as would be found in a pharmaceutical tablet. These samples yield too much scattering when studied with visible spectroscopy and too much absorbance when probed using infrared light. Near-infrared spectroscopy hits the sweet spot where the two effects balance and yield a useful measurement, with the additional bonus being that no sample preparation is needed prior to measurement.
NIR measurements are most often performed using a typical diffuse reflection setup, with a white standard such as the WS-1 used as a reference. Illumination can be integrated with light delivery (as in the DR probe or the Vivo), or routed from a tungsten halogen light source to the sample via a reflection probe.
NIR measurements also can be made in transmission. A third option is “interactance” mode, which is a combination of reflection and transmission. In interactance mode the illumination and detection spots on the sample are slightly separated on the same surface to capture light that has scattered deeper into the material and returned to the surface. This increases the effective pathlength and leads to stronger absorbance signals. In fact, due to multiple scatter events, the pathlength in interactance mode is actually substantially longer than the physical distance between illumination and collection.
Much as absorbance uses the Beer-Lambert law, reflectance is often analyzed on a comparable logarithmic scale by calculating log(1/R), with R being the reflectance – the ratio of the intensity reflected off the sample to the intensity reflected off a white standard. To carry the analogy even further, this measure is usually assumed to scale linearly with the concentration of the analyte.
Analysis through Chemometrics
The list of sample properties that can be measured with near-infrared spectroscopy is long. What might be surprising is that it goes well beyond typical chemical concentrations such as the percentage of moisture, sugar, fat for food items, hydrocarbons and additives in fuels, or active ingredients in pharmaceuticals. Chemometrics also enables assessment of physical characteristics such as density, firmness, temperature or particle size, and even complex properties like octane number or fruit ripeness.
The “measurement” of such advanced parameters does not rely on assignments of single absorption bands, but rather on a multi-wavelength correlation between the spectrum and the actual physical or chemical lab analysis, drawn out and revealed using the tools of chemometrics. Based on multivariate statistics, chemometrics borrows heavily from linear algebra techniques such a principle component analysis to allow qualitative and quantitative predictions of concentration, identity and quality.
In a typical quantitative chemometric analysis, a number of spectra are recorded for a variety of samples spanning a wide range of values for the physical or chemical property of interest, each of which is tagged to the measured value as determined in a subsequent laboratory analysis. The spectra are pre-processed to eliminate setup variations and correlated to the lab results, yielding a model that allows the user to predict the property of interest from the spectrum alone. This prediction is first tested with additional known samples before it is used on true unknown samples. To account for long-term changes in the instrument or the seasonality of natural samples, the model needs to undergo continued tests and maintenance with additional validation samples from time to time.
Chemometric Analysis of Food Quality
Almost as long as the list of sample properties that can be predicted with near infrared spectroscopy is the list of applications for this technique. Both nondestructive and noninvasive, the method has found numerous uses in the food and agricultural sector, including quality control, ripeness measurement, soil testing and precision agriculture. Nutrient information for grains was probably the first large-scale application of near-infrared spectroscopy, but the technique is now widely used in agricultural products, from determination of protein levels in soybeans to assessing the presence of gluten in grains.
Process analysis applications in industry are not far behind in popularity. Thanks to minimal sample preparation, near-infrared spectroscopy lends itself to inline analysis in the process flow. This enables continuous quality control, whether to monitor grinding, blending or partitioning steps in the pharmaceutical industry, or for fuel characterization in petroleum refining.
The ability of near-infrared spectroscopy to identify polymer types easily is demonstrated in the following figure, showing scaled diffuse reflectance spectra for five common polymers, acrylic glass (PMMA), polypropylene (PP), polystyrene (PS), polyurethane (PU), and polyethylene (PE). The spectra are very distinct and can be used to sort polymer pieces by their identity in an automatic recycling process.
Yet another important application of near-infrared spectroscopy is moisture analysis. Moisture content is one of the key parameters to monitor and control in any process chain, from the food and beverage industry to pharmaceuticals, soil tests and flour milling.
Other applications include:
- cellulose content in the pulp and paper industry
- starch, sugar and protein percentage in food production
- starch, sugar and proteins in the conversion of biomass to biofuels
- fat and oil content in the meat, dairy and seafood sector
Flame-NIR: A New Option for NIR Chemometrics
Although NIR spectrometers have decreased considerably in size thanks to the use of cooled InGaAs array detectors, their use in portable systems is still often limited by size, power consumption and cost. The new Flame-NIR spectrometer overcomes many of these limitations through the use of an uncooled NIR detector array driven by a proprietary low-noise electronic design. The elimination of detector cooling reduces both system cost and power consumption, which when combined with a shorter optical bench yields a spectrometer with a much smaller footprint and greater portability.
The Flame-NIR is based on the same optical bench design of the Flame product line and all its advantages, including interchangeable slits and low unit to unit variability. Measuring the range 970-1630 nm with ~10 nm optical resolution, the Flame-NIR is an ideal choice for many applications including moisture measurement, grain and feed quality, measurement of fats and oils and pharmaceutical ingredients blending.
To put this new bench to the test, we performed a quick feasibility study on its ability to distinguish between 22 variants of raw and partially processed grain types, from oats and wheat to rice and popcorn. A DynaCup stage was used to hold and rotate the samples for spatial averaging, with tungsten halogen light delivery and collection via a DR probe.
Despite a mere 12 measurements for each sample type, basic principal component analysis was able to clearly distinguish all but a few of the grains (primarily different variants of oats and long grain rice).
At first glance, it might seem surprising that a handheld, uncooled NIR spectrometer could yield the quality of data required for materials identification via chemometric analysis. But consider a few factors:
- NIR spectra of foods and many other materials tend to have broad, smoothly varying spectral features, comparable or greater than the Flame-NIR’s resolution of ~10 nm
- Flame-NIR has a surprisingly high signal to noise ratio – 6000:1!
- As a reflection measurement system with a tungsten halogen light source, signal is in ample supply, allowing integration times to remain short
- Many materials identification applications can tolerate regular re-referencing, minimizing the impact of baseline drift
By using a larger set of data for each grain and exploring other multivariate analysis techniques, it should be possible to further isolate different varieties of the same grain.
NIRQuest: For More Demanding Applications
The Flame-NIR may not be the right choice for every NIR application, however. The NIRQuest series offers a much wider range of options for detector and grating, as well as the option for interchangeable slits and an internal shutter (and thus wavelength range, resolution and sensitivity). The NIRQuest bench also employs thermoelectric cooling of the detector, resulting in a rock-solid baseline and lower noise at long integration times. This reduces the frequency of re-referencing, improves reproducibility and allows higher sensitivity for low-light applications.
Ignored a mere 20 years ago, NIR spectroscopy in combination with chemometrics has emerged as a rapid, convenient and cost-effective tool to measure a large variety of analytes in some very different sample types. From agriculture and food production to the pharmaceutical and petroleum industries, the applications are countless. And with its smaller footprint, low power consumption and affordable price, the Flame-NIR will enable this technique to be more widely deployed than ever.