Infrared Spectroscopy for Quantitation

Pharmaceutical manufacturing relies on quantitative FTIR to verify polymorphic form and blend uniformity of active pharmaceutical ingredients directly in tablet formulations, using near-line or at-line ATR probes integrated into process analytical technology (PAT) systems mandated by regulatory quality-by-design frameworks. Petrochemical refineries use dedicated mid-IR process analyzers with fixed-path flow cells to quantify aromatic content, oxygenate additives (MTBE, ethanol) in gasoline blends, and sulfur-bearing compounds in real time on production lines, replacing slower wet-chemical or chromatographic assays. Polymer and packaging industries quantify additive loadings — antioxidants, UV stabilizers, plasticizers — by ATR-FTIR exactly as in the worked example above, because the technique requires no sample dissolution and takes under a minute per measurement. Atmospheric and environmental chemistry uses long-path gas-cell FTIR (open-path or extractive) to quantify greenhouse gases, industrial stack emissions, and trace pollutants at parts-per-billion sensitivity, exploiting the same Beer's law framework applied to rotational-vibrational line intensities in the gas phase. Forensic and art-conservation laboratories use micro-ATR-FTIR mapping to identify and semi-quantify binder materials, varnishes, and degradation products in paint cross-sections nondestructively. In biomedical research, FTIR imaging of tissue sections quantifies relative lipid, protein, and nucleic acid content pixel-by-pixel, producing chemically resolved histological maps without staining. Across all these applications the underlying discipline is identical: control path length (or effective path length), keep absorbance in the linear regime, correct baseline and interference, and calibrate against traceable standards — precisely the framework built here from the interferometer through Beer's law to multivariate resolution of overlapping bands.

Ask the tutor: How would the calibration and uncertainty analysis change if I used peak height instead of integrated peak area for a partially overlapping IR band?

A molecule's infrared spectrum begins with its normal modes of vibration. For a nonlinear molecule with $N$ atoms there are $3N-6$ vibrational degrees of freedom (3N−5 for linear molecules), each approximated near equilibrium as a harmonic oscillator with a characteristic wavenumber

where $k$ is the bond force constant (N/m), $\mu$ is the reduced mass of the vibrating atoms, and $c$ is the speed of light, giving $\tilde\nu$ in cm⁻¹ when SI units are used and the result divided by $c$ in cm/s. A stiff, light bond — C≡O triple bond stretch near 2143 cm⁻¹, C–H stretch near 2900–3100 cm⁻¹ — vibrates at high wavenumber; a heavy, floppy bond — C–Cl stretch near 700 cm⁻¹ — vibrates at low wavenumber. This is why the region above 1500 cm⁻¹ is dominated by characteristic stretches of light atoms and multiple bonds, while the crowded "fingerprint region" below 1500 cm⁻¹ contains coupled skeletal vibrations unique to each molecule. For a vibration to absorb infrared radiation at all, it must change the molecular dipole moment during the motion — the selection rule $\partial\mu/\partial Q \neq 0$, where $Q$ is the normal coordinate. Symmetric stretches in centrosymmetric molecules (like the symmetric stretch of CO₂) are infrared-inactive precisely because the dipole moment does not change; this is why CO₂'s IR spectrum shows the asymmetric stretch (2349 cm⁻¹) and bend (667 cm⁻¹) but not the symmetric stretch. The intensity of an allowed transition is proportional to the square of the transition dipole moment, which is why polar functional groups (C=O, O–H, N–H) give strong bands while nonpolar ones (C–C, C=C in a symmetric environment) give weak ones. This intensity, expressed per mole of absorber, is exactly the molar absorptivity $\varepsilon$ that appears in Beer's law. The instrument that records this spectrum in a modern laboratory is almost always a Fourier-transform spectrometer built around a Michelson interferometer. Broadband infrared light from a Globar or Nernst glower source (see Radiation Sources for Spectroscopy) is split by a beamsplitter into two paths: one reflects off a fixed mirror, the other off a mirror that moves along the beam axis with mirror displacement $x$. The two beams recombine and interfere; the detector records total intensity as a function of $x$, called the interferogram $I(x)$. For a single monochromatic wavenumber $\tilde\nu_0$, the interferogram is a cosine of period $1/\tilde\nu_0$; for a real, broadband source and sample, it is the superposition of all these cosines, sharply peaked (the "centerburst") at zero path difference and decaying into wiggles at larger $x$. The Fourier transform relationship

recovers the single-beam spectral intensity $B(\tilde\nu)$ from the measured interferogram, computed numerically by a fast Fourier transform after the mirror scan is digitized. This one relationship is why the technique is called Fourier-transform infrared spectroscopy: the spectrum is never measured directly, only reconstructed from an interference pattern.

Two structural advantages follow directly from this design and dominate why FTIR instruments replaced scanning dispersive monochromators for quantitative work. The Fellgett (multiplex) advantage: every wavenumber is measured simultaneously throughout the entire mirror scan, rather than one narrow slit-width slice at a time as in a scanning monochromator, so for the same total measurement time an FTIR instrument accumulates far more signal per wavenumber, improving signal-to-noise ratio roughly by $\sqrt{M}$ where $M$ is the number of resolution elements in the spectrum (many thousands in a typical mid-IR scan). The Jacquinot (throughput) advantage: the interferometer has no dispersing slit to restrict the beam, so it accepts a larger solid angle of radiation for the same resolution, further improving SNR. Both advantages compound with signal averaging: because each scan takes only a fraction of a second to a few seconds, dozens to hundreds of scans can be coadded, and since interferogram noise averages down as $1/\sqrt{n}$ for $n$ coadded scans, spectral SNR improves accordingly — a key practical lever for quantitative precision. Once the single-beam spectrum of the sample, $B_{\text{sample}}(\tilde\nu)$, and of an empty reference path (or blank), $B_{\text{ref}}(\tilde\nu)$, are both recorded, the instrument computes transmittance and absorbance exactly as in any Beer's law measurement:

This ratio cancels the source spectral profile, detector response, and beamsplitter efficiency — all of which are highly wavenumber-dependent — leaving only the sample's genuine absorption. This is the reason a background scan (empty cell, clean ATR crystal, or open beam) must always be taken close in time to the sample scan: any drift in source intensity or atmospheric water/CO₂ between the two scans directly corrupts the ratio.

The resolution of the instrument — how finely $\tilde\nu$ is sampled and how narrow the instrumental line shape is — is set by the maximum mirror displacement $x_{\max}$ in the scan, with resolution approximately $\Delta\tilde\nu \approx 1/(2x_{\max})$. A typical routine mid-IR spectrum is recorded at 4 cm⁻¹ resolution; quantitative work on sharp gas-phase lines may demand 0.5 cm⁻¹ or finer. Resolution matters for quantitation for a specific reason developed in the derivations below: if the natural bandwidth of the absorption is comparable to or narrower than the instrumental resolution, Beer's law appears to fail — absorbance no longer varies linearly with concentration — an effect indistinguishable in symptom from stray light but arising instead from convolution of a sharp true absorption line with a finite instrumental slit function.

Quantitative IR rests on the same law as any absorption spectroscopy, applied to a vibrational transition rather than an electronic one:

where: $A(\tilde\nu)$ is absorbance at wavenumber $\tilde\nu$ (dimensionless), $\varepsilon(\tilde\nu)$ is the molar absorptivity at that wavenumber (L mol⁻¹ cm⁻¹), $b$ is the path length (cm), and $c$ is the analyte concentration (mol/L). Reading: absorbance at a chosen band is directly proportional to concentration, provided path length and molar absorptivity are fixed and true, exactly the logic used for UV-Vis, but now indexed to a specific vibrational mode rather than an electronic transition. For a mixture, provided bands do not overlap or after resolving overlap by multivariate methods, absorbance is additive across all absorbing species at a given wavenumber:

where: the sum runs over every species $i$ contributing at that wavenumber. Reading: this additivity is what allows multicomponent IR analysis — measuring absorbance at several wavenumbers where different species dominate, or performing a full-spectrum least-squares fit, recovers each $c_i$ simultaneously. In transmission cells, the physical path length $b$ is fixed by a spacer (commonly Teflon or lead, 15–200 μm for liquids, longer for gases) and can be measured independently by counting interference fringes in the empty cell:

where: $N$ is the number of interference fringes counted between two wavenumber limits $\tilde\nu_1$ and $\tilde\nu_2$ (cm⁻¹) in the empty-cell spectrum, and the factor of 2 accounts for the double pass through the cell (light reflecting internally off the cell windows). Reading: the cell path length is not simply "measured with calipers" — it is determined optically and far more reproducibly, because the interference fringe spacing directly encodes the optical path. In attenuated total reflectance (ATR), no physical cell exists; the beam totally internally reflects at the interface between a high-index crystal (ZnSe, Ge, or diamond) and the sample, and an evanescent wave penetrates a short distance into the sample. Beer's law still holds but with an effective path length $b_{\text{eff}}$ that replaces $b$:

where: $d_p$ is the depth of penetration (the distance at which the evanescent field falls to $1/e$ of its interface value), $\lambda$ is the wavelength in vacuum, $n_1$ and $n_2$ are the refractive indices of the ATR crystal and sample respectively, and $\theta$ is the angle of incidence at the interface (must exceed the critical angle for total internal reflection). Reading: ATR absorbance is proportional to concentration exactly as in transmission, but the effective path length is only a few microns and depends on wavelength, so ATR spectra are systematically distorted relative to transmission spectra unless corrected, particularly favoring lower-wavenumber (longer-$\lambda$) bands, which penetrate deeper.

Example 1 — Foundational: transmission-cell Beer's law for a carbonyl band. A 0.0500 M solution of ethyl acetate in carbon tetrachloride is placed in a NaCl transmission cell whose path length was determined by fringe-counting to be $b = 0.0512$ cm. The measured absorbance at the carbonyl stretch, 1740 cm⁻¹, is $A = 0.601$ (baseline-corrected against pure CCl₄ solvent). Find the molar absorptivity, and then predict the absorbance of an unknown solution measured in the same cell that gives $A = 0.288$. From Beer's law,

(three significant figures, consistent with the input data). For the unknown, solve for concentration using the same $b$ and $\varepsilon$:

As a limiting-case check: if the unknown had exactly half the standard's absorbance-to-concentration product it should read half the standard's concentration ($0.0250$ M would give $A = 0.301$); since $0.288$ is slightly less than $0.301$, the recovered concentration, $0.0239$ M, is correspondingly slightly below $0.0250$ M — internally consistent with strict proportionality.

Example 2 — Applied: quantitation of an additive in polyethylene film by ATR-FTIR. An analyst builds an ATR calibration for an antioxidant additive in polyethylene film using a diamond ATR accessory (single reflection, fixed contact pressure) and measuring the baseline-corrected peak area of the additive's characteristic band near 1595 cm⁻¹ (aromatic ring stretch), relative to the polymer's own CH₂ bending band at 1465 cm⁻¹ as an internal standard to correct for sample-to-crystal contact variability. Five standards of known additive mass fraction $w$ (in wt %) give the following area ratios $R = A_{1595}/A_{1465}$:

Linear regression of $R$ on $w$ gives slope $m = 0.0392\ \%^{-1}$, intercept $b_0 = -0.0021$, and correlation coefficient $r = 0.9989$ (values obtained by standard least-squares formulas; residual standard deviation $s_{y/x} = 0.0037$). A film sample gives $R_{\text{unk}} = 0.152$ (mean of triplicate scans). Solve for $w_{\text{unk}}$:

The standard error in $w_{\text{unk}}$ from the calibration line, using the standard propagation formula for the $x$-value predicted from a $y$-calibration with $n=5$ standards, mean $\bar w = 4.338\%$, and $S_{xx} = \sum(w_i-\bar w)^2 = 33.65\ \%^2$, and $M=3$ replicate unknown readings:

so the additive content is reported as $w = 3.93 \pm 0.07\%$ (95% confidence would multiply this standard error by the appropriate $t$-value, roughly $t_{0.05,3}=3.18$ for the three degrees of freedom associated with a 5-point calibration used to predict a triplicate unknown, giving $\pm 0.22\%$ at 95% confidence). Using an internal-standard ratio rather than raw peak area is what makes ATR quantitation reproducible here: pressing pressure and crystal contact vary shot-to-shot, changing both peak areas by nearly the same multiplicative factor, which the ratio cancels.

Example 3 — Exam/competition level: resolving overlapping bands by classical least squares, and diagnosing a resolution artifact. A mixture of acetone and ethanol in CCl₄ is scanned in a 0.100 mm transmission cell. At two wavenumbers chosen because each analyte dominates but both contribute, the following molar absorptivities (L mol⁻¹cm⁻¹) are known from single-component standards:

The mixture's absorbances are $A_{1715} = 0.734$ and $A_{1050} = 0.518$, with $b = 0.0100$ cm. Set up the linear system from the additivity relation $A = b\sum_i \varepsilon_i c_i$:

Divide both sides by 0.0100:

Solve by elimination: multiply the first equation by $62/4 = 15.5$ to match the ethanol coefficient... instead use Cramer's rule directly. The determinant of the coefficient matrix:

So the mixture is $0.220$ M acetone and $0.793$ M ethanol (three significant figures, consistent with the four-significant-figure absorbance inputs). Now the diagnostic twist: suppose the analyst had instead recorded this spectrum at coarse 16 cm⁻¹ resolution, and the acetone C=O band's true natural linewidth (full width at half maximum) is only about 10 cm⁻¹. Convolving a Lorentzian absorption line of FWHM 10 cm⁻¹ with an instrumental slit function of comparable or greater width systematically reduces the observed peak absorbance relative to the true peak absorbance (the instrument averages the sharp peak together with less-absorbing wings and shoulders), an effect that grows nonlinearly with concentration because the true peak saturates (approaches its true absorbance ceiling) before the convolution-broadened, apparently-lower peak does. The practical fingerprint of this artifact is a calibration curve that bends downward (concave, apparent negative deviation from Beer's law) at high concentration, even though the chemistry is perfectly ideal — the fix is simply to rescan at finer resolution (here, 4 cm⁻¹ or better) so that the instrumental slit width is small relative to the 10 cm⁻¹ natural linewidth, restoring the linear regime confirmed by the classical least-squares solution above.

The mathematical machinery that makes modern quantitative FTIR possible — recovering a spectrum from an interference pattern — was built by Jean-Baptiste Joseph Fourier and Albert A. Michelson, working a century apart on entirely different problems, then fused by twentieth-century instrument builders and computational chemists into the routine tool used today.