Technology
Mid-infrared spectroscopy
Mid-IR light is absorbed at the energies at which molecular bonds vibrate. Because those energies depend on the bond and on its chemical surroundings, a mid-IR spectrum says what is present and how much of it there is — which is why it became the reference technique for chemical composition.
The physical basis
A chemical bond is not rigid. It stretches, bends and twists at frequencies set by the masses of the atoms and the stiffness of the bond between them. When infrared light passes through a sample, a photon whose energy matches one of those vibrational transitions can be absorbed — but only if the vibration changes the molecule's dipole moment. That selection rule is why infrared is so informative about polar functional groups: carbonyls, hydroxyls, amines, nitriles.
The mid-infrared region, roughly 4000 to 400 wavenumbers, contains the fundamental vibrations of nearly every functional group of interest in process chemistry. Two parts of it behave differently. Above about 1500 wavenumbers, absorptions map cleanly onto identifiable groups — a carbonyl stretch near 1700, an O–H stretch above 3000. Below 1500 lies the fingerprint region, where coupled whole-molecule vibrations produce a pattern that is highly specific to the compound but not interpretable group by group.
Beer's law makes the technique quantitative: absorbance is proportional to concentration, to path length and to the molar absorptivity of the species. That linear relationship is what lets a spectrum become a number.
Mid-IR compared with the neighbouring techniques
Versus near-infrared
NIR measures overtones and combination bands, which are weaker and heavily overlapped. That gives NIR deeper penetration and simpler optics, but it almost always needs a larger calibration set to reach the same specificity. Mid-IR reads the fundamentals directly.
Versus Raman
Raman responds to changes in polarisability rather than dipole moment, so the two techniques see complementary vibrations. Raman handles water well and is strong on solid form and polymorph; mid-IR is stronger on polar organics in solution.
Versus chromatography
HPLC and GC separate components physically and resolve what spectroscopy cannot. They remain the reference method — and the source of the reference values an inline spectroscopic model is calibrated against. What they cannot be is continuous.
Why water is the recurring difficulty
Water absorbs strongly across large parts of the mid-infrared. In an aqueous stream the analyte bands sit on a very large background that also shifts with temperature, and an uncorrected temperature shift looks exactly like a concentration change.
Three things make aqueous mid-IR work in practice. The optical path length has to be short enough that the water background does not saturate the detector, and matched to the concentration range being resolved. Temperature has to be controlled or compensated. And the model has to be multivariate, so it uses the whole spectral pattern to separate analyte from background rather than relying on one peak height.
What the measurement geometry decides
- ATR measures the few micrometres of sample in contact with a crystal — excellent for a clean surface, vulnerable to fouling and unrepresentative of a heterogeneous bulk.
- Transmission through a flow cell measures the flowing bulk rather than a surface film, so a fouled window shows up in the self-test instead of quietly biasing the result.
- Path length in transmission has to be matched to absorbance: too long and the band saturates, too short and the signal is lost in noise.
- Fibre-coupled designs are limited by mid-IR transmission in the fibre itself, which is one reason compact instruments avoid fibres entirely.
Common questions
Can mid-IR measure concentration without a reference method?
Not in a real process stream. Beer's law is linear in principle, but in a mixture with overlapping bands, scatter and a temperature-dependent background, converting absorbance into concentration requires a multivariate model trained against a reference method such as HPLC.
What limit of detection can I expect?
It depends on the molar absorptivity of your analyte, the path length you can use, and how much the background varies. Anyone quoting a single number without asking about your chemistry is quoting a number from a different experiment.
Does the sample survive the measurement?
Yes. Infrared absorption is non-destructive, so sample integrity is preserved for downstream testing.
What a real dataset looks like
Absorbance spectra recorded through a reaction, plotted as a family of curves from start (dark) to end (light). Two things are visible without any modelling: the bands that grow as product forms, and the isosbestic behaviour where curves cross, which is the signature of a clean conversion between two species. What the model does is turn that family of curves into a single concentration against time.
Wondering whether mid-IR can see your chemistry?
Send the composition, the concentration range and the conditions. Where it makes sense we test the technology on your own process samples before you commit.