What is NIR spectroscopy? How near-infrared light reads food and feed
Short answer: NIR (near-infrared) spectroscopy shines invisible light, roughly 780 to 2500 nm, on a sample and records how much comes back at each wavelength. Bonds between hydrogen and carbon, nitrogen or oxygen absorb some of that light, so the pattern reveals water, protein, fat and carbohydrates. A calibration model turns the pattern into numbers in seconds, without chemicals or sample destruction.
What is near-infrared light?
Visible light runs from violet at about 400 nm to red at about 700 nm. Just past red, from roughly 780 to 2500 nm, is the near-infrared. The eye cannot see it, but a detector can, and food and feed samples interact with it in a useful way: they absorb a little at specific wavelengths and let the rest scatter back or pass through.
Beyond 2500 nm lies the mid-infrared, where the same molecules absorb far more strongly. That sounds better, but it means mid-infrared light barely gets into a sample, which is why mid-infrared methods need thin films or special accessories. NIR's weaker absorption is what lets it read a cup of ground grain or a vial of milk as it is.
How does NIR spectroscopy work?
Think of each chemical bond as a spring holding two atoms. It vibrates at its own natural frequency, and light at that frequency gets absorbed. For the bonds that matter in food, the main vibration sits in the mid-infrared. NIR picks up the fainter echoes: overtones, at roughly two and three times the main frequency, and combination bands where two vibrations add together.
Bonds to hydrogen vibrate fast and are strongly anharmonic, so their overtones are strong enough to measure. That is why NIR reads O-H (water, carbohydrates), N-H (protein) and C-H (fat, carbohydrates) so well, and why it barely sees anything else.
Weak absorption is NIR's advantage. Because overtones absorb faintly, light travels into the sample and back out, so NIR can read whole kernels, powders and liquids with little or no preparation.
How does an NIR instrument read a sample?
Every NIR analyzer does the same three things: shine light, let it meet the sample, and measure what comes out at each wavelength. What differs is the geometry, chosen to suit the sample.

| Mode | Light path | Typical samples |
|---|---|---|
| Diffuse reflectance | into the surface layer and scattered back to a detector on the same side | ground grain, flour, feed, powders, meat |
| Transmittance | straight through the sample to a detector on the far side | milk and other liquids; whole grain at shorter NIR wavelengths |
| Transflectance | through the sample, off a reflector, back through again | oils, pastes, slurries, liquids with a probe |
You will also see "NIT", near-infrared transmittance. It is the same science in transmittance mode, common in whole-grain analyzers that read kernels without grinding.
How does a spectrum become a protein number?
A spectrum is a curve of absorbance against wavelength, with broad overlapping humps. No single peak says "12.4% protein". The number comes from a calibration: a model built by scanning a set of samples that were also tested by the reference method, such as Kjeldahl or Dumas for protein. The usual tool, partial least squares (PLS) regression, learns which combination of wavelengths tracks the reference values and applies it to every new spectrum in a fraction of a second.
So an NIR result is only as good as two things: the reference values it learned from and how well the calibration samples cover the product you measure today. The next part of this path, what NIR measures and where it fails, goes deeper on both.
Fast can be confidently wrong. A calibration built on too few or too similar samples will still show a number for every scan. Speed is only worth having when the model behind it is sound.
NIR vs wet chemistry: what changes in the lab?

| NIR | Wet chemistry reference | |
|---|---|---|
| Time per sample | seconds to about a minute | much longer, often including digestion, extraction or drying steps |
| Chemicals and waste | none | acids, solvents or catalysts, depending on the method |
| Analytes per scan | several at once (for example moisture, protein, fat) | usually one method per analyte |
| Sample after testing | intact, can be re-scanned | consumed |
| Where accuracy comes from | the calibration, which learned from the reference | the method itself |
NIR does not replace the reference lab; it multiplies it. The lab keeps producing the reference values that build and check the calibration, while NIR answers the hundreds of routine questions in between.
Where is NIR used in food and feed?
NIR analysis of food grew out of agricultural research in the 1960s, including early work by Karl Norris and colleagues measuring fat and moisture in meat directly from the spectrum. Today it runs wherever fast composition answers drive a decision:
- Grain receiving: protein and moisture before the truck leaves, which set the grade and the price.
- Feed mills: raw material checks at intake and finished feed checks before bagging, feeding least-cost formulation.
- Dairy: fat, protein and solids in milk and cheese, at the bench or inline.
- Meat: fat, moisture and protein for blending and labelling.
- Oilseeds and flour milling: oil, protein, moisture and ash-related checks across the process.
Start with one decision, not one instrument. Write down the question NIR must answer (accept the truck? adjust the blend?), the analytes it needs and how fast. The right setup follows from that list.
Meet a term you don't know? The free NIR glossary explains 29 NIR and chemometrics terms, such as RPD, SEP, PLS, SNV and calibration transfer, in plain English.
Worked example: a first week with NIR
ILLUSTRATIVE A made-up lab to show the reasoning; not a customer case.
A small feed lab gets its first benchtop NIR, with a vendor calibration for protein and moisture in soybean meal. Day one, the technician scans ten samples that the reference lab already tested. Protein agrees closely on eight and reads high on two, both from a supplier the lab started using last month.
Instead of trusting the vendor calibration as it is, the lab keeps sending a few samples a week from each supplier to the reference lab, logs the difference and adds the new supplier's samples to the calibration once it has enough. Within a few weeks the two "odd" samples are no longer odd. Nothing about the light changed; the model learned the new product.
Quick check
Step 1 of 6: How NIR light works
- 1How NIR light worksYou are here.Current
- 2What it measures, where it failsReliable analytes, limits and sample presentation.Read part 2 →
- 3Inside the instrumentLight source, sample interface, detector.Read part 3 →
- 4Molecular vibrationsOvertones and combination bands in depth.Read part 4 →
- 5Beer-Lambert and its limitsWhy absorbance is not simply proportional in NIR.Read part 5 →
- 6Why NIR needs chemometricsOverlapping bands and the maths that untangles them.Read part 6 →
Frequently asked questions
What is NIR spectroscopy in simple terms?
A way to measure what a sample is made of by shining invisible near-infrared light on it and recording how much light comes back at each wavelength. A calibration turns that pattern into numbers such as moisture or protein.
What wavelength range is near-infrared?
Roughly 780 to 2500 nm, between visible red light and the mid-infrared. Exact boundaries vary slightly between sources.
What does NIR spectroscopy measure in food and feed?
Mainly components with C-H, N-H and O-H bonds present in reasonable amounts: moisture, protein, fat and oil, starch and fibre.
What is the difference between NIR and mid-infrared spectroscopy?
Mid-infrared (usually measured by FTIR) reads the strong fundamental vibrations, so samples must be thin or prepared. NIR reads weaker overtones and combinations, so light penetrates and most food samples can be measured as they are.
Is NIR analysis destructive?
No. The sample is only exposed to light and can be scanned again or sent to the reference lab afterwards.
What is near-infrared transmittance (NIT)?
NIR measured in transmittance mode: light passes through the sample to a detector on the other side. It is common in whole-grain analyzers that read kernels without grinding.
Next questions people ask
Learn NIR from the light up
The course starts with how near-infrared light works and builds to calibration and routine use, in 32 short lessons for food, grain, feed and dairy labs. One payment, lifetime access.
Sources
- Pasquini, C. "Near infrared spectroscopy: a mature analytical technique with new perspectives — a review." Analytica Chimica Acta 1026 (2018): 8–36. doi:10.1016/j.aca.2018.04.004
- Pasquini, C. "Near infrared spectroscopy: fundamentals, practical aspects and analytical applications." Journal of the Brazilian Chemical Society 14, no. 2 (2003): 198–219. doi:10.1590/S0103-50532003000200006
- Ozaki, Y., Huck, C., Tsuchikawa, S., Engelsen, S. B. (eds.). Near-Infrared Spectroscopy: Theory, Spectral Analysis, Instrumentation, and Applications. Springer, 2021.
- Burns, D. A., Ciurczak, E. W. (eds.). Handbook of Near-Infrared Analysis, 3rd ed. CRC Press, 2007.
- Ben-Gera, I., Norris, K. H. "Direct spectrophotometric determination of fat and moisture in meat products." Journal of Food Science 33, no. 1 (1968): 64–67. doi:10.1111/j.1365-2621.1968.tb00885.x