Near-infrared spectroscopy. “N-I-R.”  Let’s illuminate the subject a bit, shall we?

Spectroscopy is a branch of science interested in the interaction of light with matter. Near-infrared (NIR) spectroscopy happens when the light used to do the measurements falls within a certain energy or frequency range; typically, 12000 – 4000 cm-1 (or about 700 – 2500 nm in terms of wavelength).

This idea isn’t new. The first observations of NIR light were made by Herschel in 1800, and Coblentz was considered its pioneer in the early 1900’s. However, this small but mighty portion of the electromagnetic spectrum didn’t debut commercially until the 1970’s, coinciding with advancements in PC computing power that radically simplified it’s application.

Why do people use NIR?

Everything you’ve come to love in your life: people, places, baked goods… they are all made up of molecules. Those molecules are made of atoms, and those atoms are moving and grooving (i.e. the bond lengths and bond angles aren’t static, but rather wagging and scissoring and bending and stretching). We can use NIR to measure that molecular dance party, or more technically, molecular vibrations. Those vibrational modes can tell us stuff about the sample that most QC departments like, think: sample identity or composition. 

When the molecules of a sample are hit with NIR light, the light is either absorbed or scattered. When the light is absorbed, we see a peak in our NIR spectrum. When light scatters due to the physical properties of the sample (e.g. particle size, particle morphology, bulk density), the overall slope of the spectrum is impacted. Chemical bonds that absorb NIR well are: oxygen-hydrogen (O-H), carbon-hydrogen (C-H), carbon-oxygen (C-O), nitrogen-hydrogen (N-H), and sulfur-hydrogen (S-H). While NIR isn’t the magic bullet for every analysis, we see samples that are dominated by these types of bonds in many industries, from pharmaceuticals to pet food.

The series of peaks and valleys that appear in an NIR spectrum is the summation of molecular vibrations of the sample. Consider the spectra of 5 different solutions in the figure below, where Absorbance is shown as a function of wavelength (nm). The red spectrum is pure methanol, the green spectrum is pure water. Peaks resulting from the -OH vibrations of both the alcohol and water, as well as the -CH vibration of the alcohol, are labeled.

Since the energy of the -OH (and -CH) bonds of water and methanol (or donuts) differ, they produce unique spectra, as illustrated in the figure above. Combine that with the fact that spectra (of water, methanol or donuts) is repeatable, and we’re going somewhere. That means, if I scan water with my NIR analyzer 10x, I will get (more or less) the same spectrum, and that spectrum is unique from the other stuff I want to analyze (here, methanol). For those reasons, NIR would be a good tool for identification purposes.

Take another look at the figure above. The peak intensities corresponding to each molecular vibration reflect the relative composition of the molecules (water and alcohol) contributing those bonds to the solution. More specifically, you can see the peak corresponding to the water -OH vibration around 1450 nm increase as the relative proportion of water increases in solution. In the same way, the peak corresponding to the -CH combination vibration around 2250 nm increase as the proportion of methanol increases in the solution. That’s Beer’s Law working for us, where Absorbance at a given wavelength is proportional to the pathlength * molar absorptivity * concentration of the absorbing analyte. Note to the academics: if you dig into the theoretical research, you will see that Beer’s Law applies strictly to ideal solutions. Oh, and if you hadn’t heard, almost nothing is ideal in real life. However, we have mathematical ways to address non-linearity, and in the end, most methods work well. And by work well, we mean produce satisfactory standard errors of prediction. 

But before you get too excited about the infinite possibilities of NIR, let me give you some fine print. Our mortal bodies typically can’t run marathons without training first… and an NIR can’t run a qualitative or quantification application without being trained, either.

More on that, here.

Additional resources:

For a good historical, theoretical and applications overview, see the Handbook of Near-Infrared Analysis, edited by Donald A. Burns and Emil W. Ciurczak