As a consultant, I have gotten calls where the potential clients states, “I’ve just purchased a NIR instrument; could you tell me what to do with it?” Assuming they do not hang up when I do tell them what they can do with it, I explain the problem(s) associated with “putting the cart before the horse.” But, first, we will examine the plethora of kinds and sizes of NIR spectrometers.
For decades, the vast majority of spectrometers were somewhat sensitive (needed to remain safely on a bench), and were somewhat slow (first prisms, later gratings). For example, to generate a “high resolution”mid-range infrared (MIR) spectrum, the typical instrument was set to the 22-minute scan rate (yes, 22 minutes, not a misprint). Higher energy wavelength ranges, such as the deuterium (UV) or tungsten (visible), took less time to scan, yet they were far from rapid. Even the NIR instruments of the 60’s and 70’s were slow grating or filter units. [And, keep in mind, no one was using computers, except very primitive ones for NIRS, so “computation” often meant a ruler and chart paper.]
At no point were there large-scale attempts to place spectrometers in- line at production facilities. One exception was MIR: simple units were available (likely filter instruments) to measure the water content of a paper web, in real time. The units could be connected with devices to add or lower pressure in rollers to relax or squeeze water from the paper rolling through (at a pretty high speed). This was the rare exception, not the rule.
Subscribe to our e-Newsletters
Stay up to date with the latest news, articles, and events. Plus, get special offers
from American Pharmaceutical Review – all delivered right to your inbox! Sign up now!
Today, when we speak of mid-range infrared instruments, we almost always imply a Fourier-Transform instrument (FT-IR). For several good reasons, it has become the “go-to” design (i.e., IR sources are weak, absorbances are strong, and, in general, detectors are noisy, so an interferometer is the best way to go). When we think of UV/Vis (Ultraviolet/Visible) spectrometers, we think (since the early 1980’s) of diode-arrays, again for good reasons. Raman bespeaks LASERs, so there are fewer choices of mechanics beyond the hand-held versus installed arrangements. The emitted light can, of course, be collected in various manners, but there are few choices.
Near-infrared is a strange beast: it has strong sources, weak absorbances, and extremely fast and quiet detectors. The versatility of the spectral range allows for a plethora of hardware variations. The sources are either tungsten/halogen or LASERs (or LASER diodes), so that is somewhat common to all units. The detectors are, essentially, “just” semi-conductors: lead sulfide (PbS), indium gallium arsenide (InGaAs), germanium (Ge), and so forth. These are chosen for their applications. The wavelength range, speed, sensitivity, etc. all help the designer choose the most relevant material (as well as price constrictions). In just about any case (single detector array), the method or detection is the same: a potential is placed across the detector so that when a photon impacts, an electron moves from the (-) to (+) end, causing a micro-current.
The place where the vast number of variations (hence, choices) happen is in wavelength selection. As if the peaks in NIR overlapping so badly with low absorptivity’s wasn’t bad enough, the spectra are strongly affected by hydrogen bonding (and weakly by van der Waal’s repulsions) with contributions of other phenomena. Consequently, instrument designers are always looking to “an edge” when measuring a particular material.
If a material is coarsely shaped (corn, wheat, tablets), cells were designed to average a large cross-section of the material. Highly scattered light (reflected or transmitted)? Multiple detectors (originally one at 45-degrees to the “normal” then 2 and 4 used) or integrating spheres may be used. Some engineers/applications people worried that sorting the wavelengths prior to shining the radiation on the sample might not have the sensitivity needed, since so little light is left to be reflected or transmitted. Their solution was to shine all the wavelengths onto the sample, THEN perform a separation after the sample reacts with the light. So, we have pre-and post-sample wavelength section with large numbers of adherents on each team.
We also have several ways to capture the wavelengths of interest (before or after sample interaction). Gratings still are quite popular with lab-based (or moving on a cart) instruments. There are decades of experience with them; they are rugged; they are easily reproduced holographically (think DVDs and CDs). A moving grating spreads the wavelengths and they strike the sample in order. The reflected or transmitted light is collected on discreet detectors or via an integrating sphere, where the light is reflected onto a single detector from the walls of the spherical sphere.
Other wavelength “massagers” include FT interferometers, polarization interferometers, and MEMS (Microelectromechanical systems), quite small systems initially developed as telecommunications devices. They send an interference pattern of all or most of the light to the sample and the reflected (transmitted) signal is deconvolved by, for example a Fourier Transform. Since these devices can be made from benchtop size (conventional FT) to hand held (EMS), the one chosen is determined by the application.
Another approach is an acousto-optic tunable filter (AOTF). An acousto-optic modulator consists of a piezoelectric transducer which creates sound waves in a material like glass or quartz. A light beam is diffracted into several orders. Quickly changing the frequency of the impinging sound waves changes the distance between standing waves within the crystal, causing the light to be separated as per Bragg’s law (like an accordion acting as a grating). It is light, fast, and relatively inexpensive.
The diode-array instruments, often bench-top types have been around for years, as well. In these, the full spectrum is shone on the sample, light collected, wavelengths separated by a grating or linear variable filter (LVF) and impinges on a series of diodes (small detectors in an array).
Virtually all modern NIR spectrometers are well-built and reliable. What the consumer/analyst needs to do is determine 1) what the analysis will be (complex or simple), 2) where it will be made (lab or process stream), and 3) whether the data gathered will be for R&D or release purposes. So, choose a dependable vendor and make them prove that their units will do the analysis you need.