Chapter 7 — Transmission Spectroscopy

Transmission spectroscopy is historically the most widely used measurement geometry in infrared spectroscopy. In this configuration, radiation passes through a sample and the transmitted intensity is measured as a function of wavelength or wavenumber. The attenuation of the transmitted beam is commonly expressed in terms of absorbance and interpreted using the Bouguer–Beer–Lambert law.

Although this interpretation is often sufficient for routine applications, it represents an idealized approximation. The actual transmission of light through a finite sample is governed by electromagnetic boundary conditions at the sample interfaces, multiple reflections within the sample, and interference between forward and backward propagating waves. These effects can influence both the intensity and the spectral shape of measured bands.

Understanding transmission spectroscopy therefore requires distinguishing between the ideal Beer–Lambert limit and the wave-optical reality of transmission through finite media.

The Bouguer–Beer–Lambert law is often presented as a fundamental principle of absorption spectroscopy. In reality, it represents a special limiting case of electromagnetic wave propagation in absorbing media. When light propagates in a homogeneous medium described by a complex refractive index \(\hat{n}=n+ik\), Maxwell’s equations predict an exponential decay of the electric-field amplitude — and hence of the intensity — with propagation distance.

The Beer–Lambert relation therefore emerges when this propagation occurs in a single medium without significant reflections at interfaces and without interference effects. In practical spectroscopic measurements, however, samples are typically bounded by interfaces or embedded in layered structures, so that reflections and standing-wave effects modify the local electric field distribution. Under such conditions the measured absorbance can deviate from the simple Beer–Lambert prediction even if the intrinsic material absorption described by \(k(\tilde{\nu})\) remains unchanged.

7.1 Ideal Transmission and the Beer–Lambert Limit

The Bouguer–Beer–Lambert law provides a simple relationship between absorbance \(A\), concentration \(c\), and optical path length \(d\):

\[ A = \varepsilon\, c\, d \tag{7.1} \]

where \(\varepsilon\) is the decadic molar absorption coefficient (here a molar quantity, not the dielectric function \(\varepsilon(\tilde{\nu})\) of Chapter 5). Absorbance itself is defined through the logarithmic relation

\[ A = -\log_{10}(T) = -\log_{10}\frac{I_T}{I_0} \tag{7.2} \] with \(I_0\) representing the incident intensity and \(I_T\) the transmitted intensity.

Implicit in this formulation is a very specific physical model: radiation propagates through a homogeneous absorbing medium and experiences exponential attenuation due solely to absorption. Reflection losses and interference effects are neglected. In other words, the Beer–Lambert law describes exponential attenuation during propagation in a homogeneous absorbing medium (and does not account for reflections at boundaries) rather than the transmission through a finite slab bounded by interfaces.

In real experiments, however, samples are finite objects with surfaces separating materials of different refractive index. At each interface, a fraction of the incident radiation is reflected while the remainder enters the sample. Additional reflections occur at the exit interface, and part of this reflected radiation can re-enter the sample and propagate again through the absorbing medium. Consequently, the light interacting with the sample is composed of multiple forward- and backward-propagating waves.

From the standpoint of electromagnetic theory, the transmission through a slab is therefore described by wave-optical formalisms such as the Airy summation (49) or the Abelès transfer-matrix method, (50) which explicitly account for multiple reflections and interference. The Beer–Lambert relation emerges only as an approximation under conditions where reflection losses and interference can be neglected (Figure 7.1).

The Beer–Lambert limit compared with the transmittance of a finite slab, all for the same intrinsic absorption (k). The ideal curve assumes pure exponential attenuation and a baseline at unity. The coherent slab adds an interference ripple from multiple internal reflections. The incoherent slab — where thickness variations or limited coherence wash the fringes out — is perfectly smooth, yet its baseline is still depressed below unity by reflection at the two surfaces — to ((1-R)/(1+R)) in the non-absorbing limit, once incoherent multiple reflections are included (Section 7.1). Beer–Lambert therefore fails even when there is no interference at all; the depression is a pure boundary effect.
Fig. 7.1 
The Beer–Lambert limit compared with the transmittance of a finite slab, all for the same intrinsic absorption 
\(k\). The ideal curve assumes pure exponential attenuation and a baseline at unity. The coherent slab adds an interference ripple from multiple internal reflections. The incoherent slab — where thickness variations or limited coherence wash the fringes out — is perfectly smooth, yet its baseline is still depressed below unity by reflection at the two surfaces — to \((1-R)/(1+R)\) in the non-absorbing limit, once incoherent multiple reflections are included (Section 7.1). Beer–Lambert therefore fails even when there is no interference at all; the depression is a pure boundary effect.

It is worth emphasizing that interference is not the only way the Beer–Lambert law breaks down. Even when the fringes are completely washed out — an incoherent slab, or a pressed pellet whose thickness varies from place to place — the transmittance is not the simple exponential \(e^{-\alpha d}\) but rather, for a plane-parallel slab at normal incidence with the same medium on both faces, \(T = (1-R)^2 e^{-\alpha d} / (1 – R^2 e^{-2\alpha d})\), where \(R\) is the surface reflectance and \(\alpha\) the Napierian absorption coefficient.

The numerator factor \((1-R)^2\) together with the denominator depress the baseline below unity — to \((1-R)/(1+R)\) where absorption is negligible — with the denominator accounting for the incoherent multiple reflections. Taking \(A = -\log_{10}T\) and dividing by the thickness therefore does not return the intrinsic absorption coefficient: the recovered value carries a reflection-dependent offset and the bands take on a slightly reflection-like character. The Beer–Lambert law already fails, in other words, in the very simplest case of a homogeneous slab with no interference whatsoever. (36)

In practice, this limit can often be approached in spectrophotometric measurements when:

  • reflection losses are small,

  • interference effects are averaged out,

  • and spectra are ratioed against appropriate reference measurements.

When these conditions are satisfied, the Beer–Lambert law provides a useful and widely applied approximation that enables quantitative analysis of concentration or optical constants.

However, when the underlying assumptions are violated, such as in thin films, layered samples, or strongly refractive materials, the Beer–Lambert law may lead to substantial errors in retrieved absorption coefficients or derived quantities. Under such conditions, more complete optical models must be employed.

7.2 Gases

Among all forms of matter, gases most closely approximate the conditions required for the Beer–Lambert limit. In gaseous samples the refractive index differs only slightly from unity, and the density of absorbing molecules is typically low. Reflection losses do occur at the cell windows — at both the air/window and window/gas interfaces — and need not be small; but referencing the filled cell against the empty cell removes them, and residual interference effects are generally negligible.

Furthermore, absorption in gases is usually weak compared with condensed phases. The resulting optical path lengths are therefore relatively long, often several centimeters or more, which reduces the influence of boundary effects relative to the overall attenuation of the beam. The long path is also what removes the interference fringes: with the windows widely separated, any channel-spectrum fringes they produce are so finely spaced in wavenumber that beam divergence and finite resolution average them away. Under these conditions the exponential attenuation predicted by the Beer–Lambert law provides an excellent description of measured spectra.

Gas-phase spectroscopy therefore represents the regime in which transmission measurements most closely correspond to the intuitive picture of light passing through a medium and being attenuated solely by absorption, at least for low pressures.

In practical measurements, the Beer–Lambert relation is routinely used to determine concentrations, line strengths, and absorption cross sections of gaseous species. The accuracy of such determinations relies on careful calibration of path length and pressure, but the underlying assumption of exponential attenuation remains valid in most cases.

Nevertheless, even in gas spectroscopy the Beer–Lambert relation should be understood as an approximation rather than a fundamental law of transmission. In high-precision measurements, additional effects—such as scattering, pressure broadening, or instrumental artifacts—can introduce deviations from ideal behavior. At elevated pressures, collisional broadening, line mixing, and density-dependent local-field and refractive-index effects can all contribute, and should not be collapsed into a purely empirical attenuation law. The local-field contribution in particular — the polarization of the surrounding medium acting back on the absorbing molecule — is sometimes overlooked or absorbed into an effective-collision picture; it is treated in detail, for both gases and condensed matter, by Born and Wolf. (51)

Despite these caveats, gases represent the simplest and most direct realization of the Beer–Lambert limit and thus provide a useful reference point for understanding more complex transmission measurements in liquids and solids.

7.3 Liquids: Thick vs Thin Cuvettes

Transmission measurements of liquids are typically performed using cuvettes with infrared-transparent windows separated by a defined path length. In infrared spectroscopy, the optical path length must usually be much shorter than in UV–visible spectroscopy because many solvents have strong mid-infrared absorptions and fully transparent solvent regions are limited. As a result, thin cuvettes with path lengths of a few micrometers to a few tens of micrometers are commonly required in MIR spectroscopy in order to avoid complete absorption.

From an optical perspective, a liquid sample in a cuvette forms a layered system bounded on both sides by the ambient medium (air),

ambient  ∣  window  ∣  liquid  ∣  window  ∣  ambient.

The outer ambient–window faces reflect as well, but for the usual thick windows these reflections are mutually incoherent and are referenced out against the empty cell; the channel-spectrum fringes therefore arise at the two window–liquid interfaces, where partial reflection occurs from the refractive-index contrast. These reflections cause multiple passes of radiation through the liquid layer, so the transmitted light results from the superposition of forward- and backward-propagating waves inside the liquid.

Two limiting regimes can be distinguished.

Thick liquid layers

For sufficiently thick liquid layers, the fringe spacing in wavenumber becomes so small that finite spectral resolution, beam divergence, and small thickness variations average the interference contributions away.

In this regime the effect of multiple reflections is effectively smoothed out and the measured transmittance approaches the Beer–Lambert limit. Under such conditions the absorbance calculated from

\[ A = -\log_{10}\frac{T}{T_0} \tag{7.3} \]

is approximately proportional to concentration and path length (here \(T_0\) is the transmittance of the pure solvent). Note that this is different from the usual definition of absorbance and works well as long as the sample and reference remain optically similar, so that changes in reflection losses and refractive-index contrast between them are small. (36)

Thin liquid layers

When the liquid layer becomes sufficiently thin, typically the case in MIR spectroscopy, the optical path difference between reflected beams becomes comparable to the wavelength of the radiation. The forward- and backward-propagating waves can then interfere coherently, producing interference fringes in the transmission spectrum (Figure 7.2).

The window material, not just the path length, decides how strong the channel-spectrum fringes are. The same 1,4-thioxane liquid is modeled between two identical windows. (a) Between CaF₂ windows, whose index (≈1.4) lies close to that of the liquid (≈1.48), the window–liquid reflection is weak and the baseline stays almost fringe-free; the two path lengths then differ mainly in how deeply the absorption bands cut. (b) Between ZnSe windows (index ≈2.4) the reflection is large and pronounced fringes ride across the whole baseline, more finely spaced for the thicker layer because the spacing scales as (1/(2nd)). In a much thicker layer these fringes become so finely spaced that beam divergence and finite resolution average them away, recovering the Beer–Lambert baseline.

Fig. 7.2
 The window material, not just the path length, decides how strong the channel-spectrum fringes are. The same 1,4-thioxane liquid is modeled between two identical windows. (a) Between CaF₂ windows, whose index (≈1.4) lies close to that of the liquid (≈1.48), the window–liquid reflection is weak and the baseline stays almost fringe-free; the two path lengths then differ mainly in how deeply the absorption bands cut. (b) Between ZnSe windows (index ≈2.4) the reflection is large and pronounced fringes ride across the whole baseline, more finely spaced for the thicker layer because the spacing scales as \(1/(2nd)\). In a much thicker layer these fringes become so finely spaced that beam divergence and finite resolution average them away, recovering the Beer–Lambert baseline.

The liquid in Figure 7.2 uses the measured 1,4-thioxane optical constants of Myers et al., (52) modeled between cuvette windows of CaF₂ (53) and ZnSe — both with their dispersive refractive indices. Because thioxane is itself a liquid, the thin-film examples in Figures 7.3 and 7.4 instead use poly(methyl methacrylate) (PMMA), a solid polymer that genuinely forms supported films, on the substrate stated there.

These fringes correspond to periodic variations of the electric-field intensity inside the liquid layer. Because molecular absorption depends on the local electric-field intensity, the effective absorption becomes modulated by interference. As a result, the measured spectra can exhibit not only oscillatory baselines but also changes in band intensities and shifts in apparent peak positions.

Even when the interference fringes are weak or removed by baseline correction, the underlying redistribution of the electric field may still influence the spectral shape. Consequently, deviations from Beer–Lambert behavior may persist even after apparent fringe removal.

Several experimental strategies are commonly used to suppress interference effects, but they differ in how they act and in whether a wave-optical correction can still be applied. (47)  A wedged window does not so much eliminate the effect as smear it: the local path length varies across the illuminated area, so the recorded spectrum is an average over a range of thicknesses, and a correction would have to average the wave-optical model over that same range.

Slightly tilting the cuvette introduces non-normal incidence, which is itself straightforward to model; by itself it shifts the fringes rather than removing them, and any real suppression comes from the angular spread of a non-parallel (focused or convergent) beam, which superimposes fringe patterns of slightly different period.

Roughening a window surface is qualitatively different: it introduces diffuse, non-specular reflection that scatters light out of the coherent beam and is not captured by a specular wave-optical model at all, so this case cannot be corrected.

However, it should be noted that suppressing interference in this way can remove the possibility of applying a simple specular wave-optical correction procedure. When interference is preserved, it can in principle be modeled and corrected using optical methods based on Maxwell’s equations. When coherence is deliberately destroyed, such corrections can be very complex.

Consequently, the choice between suppressing interference experimentally or correcting it computationally depends on the objectives of the measurement and the required level of quantitative accuracy.

A more fundamental limitation should also be borne in mind: local-field effects keep absorbance from being strictly additive even in the most favorable case of dilute liquids, as is seen most clearly in the water bands. This is not peculiar to transmission but occurs across the other techniques as well. (36)

7.4 Solids: Pellets and Diluted Samples

Transmission measurements of solid materials are frequently performed using diluted samples embedded in transparent matrices, most commonly potassium bromide (KBr) pellets. In this technique the sample is finely ground and mixed with a non-absorbing matrix material before being pressed into a thin pellet.

From an optical standpoint, this configuration differs fundamentally from thin films or layered systems. For the Beer–Lambert-like approximation to hold, the absorbing particles should be much smaller than the wavelength of infrared radiation and well dispersed within the transparent matrix. In the ideal limit of sufficiently small, well-dispersed grains at low volume fraction, the pellet behaves approximately as a micro-homogeneous, effectively incoherent medium whose optical properties represent an average of the optical constants of the matrix and the inclusions. (40) As a very crude first-order approximation the averaged complex index is the volume-weighted sum \(\hat{n} = \varphi\,\hat{n}_{\text{inclusion}} + (1-\varphi)\,n_{\text{matrix}}\), where \(\varphi\) is the volume fraction of absorbing material — a mixing rule about as simple as the Beer–Lambert approximation itself, neglecting local-field effects and higher-order terms. (40)

A key consequence of this configuration is that interference effects are usually absent. The pellet thickness varies locally, and the optical path through individual particles differs throughout the sample. These variations effectively destroy the phase coherence required for stable interference patterns. The transmitted radiation therefore reflects an incoherent average over many optical paths.

Under such conditions, the Beer–Lambert law can again provide a reasonable approximation, particularly when the concentration of absorbing material is low. However, proper application of the Beer–Lambert relation requires careful reference measurements.

In spectrophotometric practice, the transmittance of the sample pellet is therefore ratioed against the transmittance of a pure matrix pellet (cf. eqn. 7.3):

\[ A = -\log_{10}\frac{T_{\text{sample}}}{T_{\text{matrix}}} \tag{7.4} \]

This ratioing procedure largely removes the influence of reflections and transmission losses associated with the matrix itself and brings the measurement closer to conditions compatible with the Beer–Lambert approximation. The reason the ratio works — rather than the negative logarithm of the sample transmittance alone — is physical, not merely instrumental: dividing by the matrix transmittance cancels the \((1-R)^2\) reflection-loss factor discussed in Section 7.1, which would otherwise leave a concentration-dependent offset in the baseline. (36) When applied correctly, the absorbance obtained in this manner is proportional to the amount of absorbing material present in the pellet.

Even so, the agreement with the Beer–Lambert law remains only approximate. As the volume fraction increases and the strongest bands approach zero transmittance, the absorption coefficient recovered through the Beer–Lambert relation falls below its true value and the bands flatten and take on a slightly reflection-like character — a deviation that appears with concentration even though no interference is present. This apparent saturation is sometimes attributed to “shadowing” of one particle by another, but shadowing does not occur for grains much smaller than the wavelength: a wave simply diffracts around such objects. The effect is instead a consequence of the transmittance approaching zero and the growing influence of reflection. (40)

Despite this advantage, pellet measurements also introduce additional complications. Scattering by particles, variations in particle size, and incomplete dispersion of the sample can all influence the measured spectrum. Furthermore, anisotropic crystals may exhibit orientation effects if the particles are not randomly oriented and very small compared to the wavelength.

Nevertheless, when samples are finely ground and properly diluted, pellet transmission measurements remain one of the most practical and widely used approaches for obtaining infrared spectra of solid materials.

7.5 Thin Films on Transparent Substrates

Thin films represent one of the most common configurations in transmission spectroscopy of solids. In this case the sample forms a layer with a well-defined thickness deposited on, or supported by, an infrared-transparent substrate such as CaF₂, BaF₂, ZnSe, or Si.

Optically, this configuration corresponds to a layered system bounded by the ambient medium (air),

ambient  ∣  film  ∣  substrate  ∣  ambient.

As in the cuvette, in a properly referenced measurement the thick substrate’s far (substrate–ambient) face is largely referenced out, so the coherent fringes arise at the ambient–film and film–substrate interfaces. The interference mechanism is exactly the one described for thin liquid layers in Section 7.3: because the film thickness is comparable to the infrared wavelength, the multiply reflected waves recombine coherently and the spectrum carries interference fringes.

The influence of interference is not limited to the baseline of the spectrum. The interference pattern modifies the electric-field intensity inside the film, which in turn determines the strength of light–matter interaction. Consequently, the apparent intensity of vibrational bands may vary with film thickness and with wavenumber. In addition, the positions of band maxima can shift slightly relative to the intrinsic spectral positions of the vibrational modes.

These effects are often particularly important for very thin films and may decrease or become averaged as the film thickness increases or as absorption becomes sufficiently strong to damp the multiple reflections.

It is therefore important to recognize that transmission spectra of thin films cannot generally be interpreted directly using the Beer–Lambert law. Instead, the optical response of the layered system must be described using wave-optical models that account for reflection, interference, and absorption simultaneously.

7.6 Index Contrast and Standing-Wave Effects

The magnitude of interference in thin-film transmission depends strongly on the refractive-index contrast between the film and its substrate. When the substrate index is close to that of the film, the film–substrate reflection coefficients are small, the multiply reflected waves are weak, and the spectrum approaches simple exponential attenuation — particularly for films more than a few micrometers thick. When the indices differ strongly, the reflections are large, the multiply reflected waves reinforce one another, and the spectrum shows pronounced fringes together with distortions of band intensities and shifts of apparent peak positions (Figure 7.3).

Refractive-index contrast controls interference, and “matching” is never perfect across a spectrum. The same 4 µm PMMA film is transmitted into two substrates. On silicon (high, nearly constant index) the strong film–substrate reflection drives pronounced fringes across the whole baseline. On CaF₂ the reflection is weak and the spectrum is almost fringe-free at high wavenumber — but CaF₂ is itself dispersive, its index changing toward the lattice-phonon region, so the film–substrate contrast grows toward lower wavenumber and around the film’s own strong, dispersive bands, leaving residual fringes there. The intrinsic absorption bands are identical for both substrates; only the interference differs.
Fig. 7.3
 Refractive-index contrast controls interference, and “matching” is never perfect across a spectrum. The same 4 µm PMMA film is transmitted into two substrates. On silicon (high, nearly constant index) the strong film–substrate reflection drives pronounced fringes across the whole baseline. On CaF₂ the reflection is weak and the spectrum is almost fringe-free at high wavenumber — but CaF₂ is itself dispersive, its index changing toward the lattice-phonon region, so the film–substrate contrast grows toward lower wavenumber and around the film’s own strong, dispersive bands, leaving residual fringes there. The intrinsic absorption bands are identical for both substrates; only the interference differs.
Index matching, however, is never perfect across a spectrum. A real matched window such as CaF₂ is itself dispersive: its refractive index changes toward the lattice-phonon region, so the film–substrate contrast — and the residual interference — grows steadily toward low wavenumber. The film contributes a second, local mismatch: near each strong absorption band its own refractive index swings sharply (anomalous dispersion), so the contrast spikes right at the bands, where distortion is least welcome. The CaF₂ curve in Figure 7.3 is computed with this dispersive index (53) for exactly that reason — matching suppresses interference but does not abolish it. CaF₂ remains a common choice because it combines broad transparency with a moderate index, limiting reflection losses, yet some interference always survives.

These effects are understood most fundamentally as a redistribution of the electric field inside the film. The forward- and backward-propagating waves form a standing-wave pattern whose nodes and antinodes are fixed by the optical thickness of the layer and the wavelength of the radiation. Because the light–matter interaction depends on the local field intensity, the effective absorption is governed not only by the intrinsic absorption coefficient but also by where the absorbing material sits within this standing wave, so absorption is enhanced or suppressed according to the phase of the waves in the layer. (39),   (38)

The central consequence is that the apparent absorption of a film can vary with thickness even when its intrinsic optical constants are fixed: band intensities rise and fall as the changing optical thickness shifts the standing wave relative to the absorbing material (Figure 7.4). The same mechanism displaces apparent band positions slightly, because the transmission minima reflect a combination of absorption and interference rather than the intrinsic resonances alone.

The thickness-dependent distortion is set by the substrate, not by the chemistry. Shown is the apparent absorbance per unit thickness, (-log_{10}T / d), for PMMA films of several thicknesses, with the intrinsic Beer–Lambert curve dotted. Were Beer–Lambert exact, every thickness would collapse onto that curve. (a) On a CaF₂ substrate, modeled with its dispersive index, the films nearly collapse onto the intrinsic curve — the standing wave is weak — although the thinnest film still leaves a small residual offset, since even CaF₂ is not perfectly matched at these low wavenumbers. (b) On silicon the strong film–substrate reflection drives a large, thickness-dependent distortion: thin films show larger apparent absorbance per micron and the relative band intensities shift, even though the optical constants are identical to those in (a).

Fig. 7.4 The thickness-dependent distortion is set by the substrate, not by the chemistry. Shown is the apparent absorbance per unit thickness, \(-\log_{10}T / d\), for PMMA films of several thicknesses, with the intrinsic Beer–Lambert curve dotted. Were Beer–Lambert exact, every thickness would collapse onto that curve. (a) On a CaF₂ substrate, modeled with its dispersive index, the films nearly collapse onto the intrinsic curve — the standing wave is weak — although the thinnest film still leaves a small residual offset, since even CaF₂ is not perfectly matched at these low wavenumbers. (b) On silicon the strong film–substrate reflection drives a large, thickness-dependent distortion: thin films show larger apparent absorbance per micron and the relative band intensities shift, even though the optical constants are identical to those in (a).

These distortions persist even when no fringes are visible: removing a fringe baseline does not undo the underlying field modulation, so altered band ratios and small peak shifts can remain. Their strength scales with the interface index contrast, the layer thickness, and the absorption strength — strong absorption damps the multiple reflections inside a band and makes it comparatively robust against standing-wave modulation, but the strong dispersion accompanying such bands can itself distort their apparent positions and shapes; weakly absorbing thin films are the most susceptible to standing-wave modulation.

All of these effects are described exactly by solutions of Maxwell’s equations for layered media, which treat reflection, absorption, and interference simultaneously and predict how a transmission spectrum depends on thickness, refractive index, and absorption coefficient. Recognizing them is essential: ignoring them can lead to incorrect conclusions about band intensities, relative concentrations, or the intrinsic optical constants of the film.

7.7 Practical Limits of Quantification

Transmission spectroscopy is widely used for quantitative analysis because the Beer–Lambert law relates absorbance directly to concentration and path length. As the preceding sections have shown, that relation holds well only under specific optical conditions — small reflection losses, negligible or averaged interference, and good reference measurements — as in gas-phase work and dilute-solution spectrophotometry. Where those conditions fail, in thin liquid layers, thin films, and high-contrast layered systems, multiple reflections and the standing wave reshape the spectrum, so that band intensities become thickness-dependent, apparent peaks shift, band shapes distort, and the apparent dependence on path length, film thickness, or concentration may cease to be linear.

Figure 7.5 illustrates these limits for a thin film on a high-index substrate, where reflection losses and standing-wave modulation together displace the apparent absorbance from the straight Beer–Lambert line.

Limits of quantification. The Beer–Lambert law predicts a band absorbance strictly proportional to path length. For a real film on a high-index substrate the measured peak departs from this straight line: reflection losses and standing-wave modulation together displace the apparent absorbance from it (inset). The modulation damps as the film thickens and absorption suppresses the multiple reflections.
Fig. 7.5
 Limits of quantification. The Beer–Lambert law predicts a band absorbance strictly proportional to path length. For a real film on a high-index substrate the measured peak departs from this straight line: reflection losses and standing-wave modulation together displace the apparent absorbance from it (inset). The modulation damps as the film thickens and absorption suppresses the multiple reflections.

When these effects become important, interpretation of transmission spectra requires optical models that explicitly account for reflection and interference in layered systems. Two complementary strategies exist for the inverse problem of recovering the intrinsic optical constants from a distorted spectrum.

The first is a fast, approximative correction that is reliable only for the weak oscillators of organic and biological materials but can often be automated and applied with limited expert intervention.

The second is dispersion analysis — fitting an oscillator model of the dielectric function to the measured spectrum — which is far more flexible and applies to strong oscillators as well, at the cost of requiring good starting parameters and experienced judgment. Dispersion analysis was named by Spitzer and Kleinman in 1961 but had been applied much earlier, the first instance being Czerny’s 1930 study of rock salt. (33) 

Wave-optical approaches of this kind allow the optical constants of materials to be extracted from measured spectra. These optical constants can then be used in physics-informed, complex-valued chemometric procedures such as CLS, ILS, PCR, and PLSR, in contrast to conventional chemometrics built on the Bouguer–Beer–Lambert approximation. (37)

In practice, the choice between simple Beer–Lambert analysis and more advanced optical modeling depends on the required level of quantitative accuracy and the optical properties of the sample. Recognizing the limits of the Beer–Lambert approximation is therefore essential for the reliable interpretation of transmission spectra.

References

  1. Czerny M. Messungen am steinsalz im ultraroten zur prüfung der dispersionstheorie. Z Phys. 1930;65(9-10):600-631. doi:10.1007/BF01391162
  1. Mayerhöfer TG. Wave Optics in Infrared Spectroscopy: Theory, Simulation, and Modeling. Elsevier; 2024. doi:10.1016/C2023-0-00262-X
  1. Mayerhöfer TG, Popp J. Beer’s law derived from electromagnetic theory. Spectrochim Acta A Mol Biomol Spectrosc. 2019;215:345-347. doi:10.1016/j.saa.2019.02.103
  1. Mayerhöfer TG, Popp J. The electric field standing wave effect in infrared transflection spectroscopy. Spectrochim Acta A Mol Biomol Spectrosc. 2017;191:283-289. doi:10.1016/j.saa.2017.10.033
  1. Mayerhöfer TG, Mutschke H, Popp J. The electric field standing wave effect in infrared transmission spectroscopy. ChemPhysChem. 2018;19(21):2916-2923. doi:10.1002/cphc.201700688
  1. Mayerhöfer TG, Popp J. Beyond Beer’s Law: Spectral mixing rules. Appl Spectrosc. 2020;74(10):1287-1294. doi:10.1177/0003702820942273
  1. Mayerhöfer TG, Pahlow S, Hübner U, Popp J. Removing interference-based effects from the infrared transflectance spectra of thin films on metallic substrates: A fast and wave optics conform solution. Analyst. 2018;143(13):3164-3175. doi:10.1039/c8an00526e
  1. Airy GB. On the phænomena of Newton’s rings when formed between two transparent substances of different refractive powers. Lond Edinb Dubl Philos Mag J Sci. 1833;2(7):20-30. doi:10.1080/14786443308647959
  1. Abelès F. Recherches sur la propagation des ondes électromagnétiques sinusoïdales dans les milieux stratifiés. Ann Phys. 1950;12d(5):596-640. doi:10.1051/anphys/195012050596
  1. Born M, Wolf E. Principles of Optics: Electromagnetic Theory of Propagation, Interference and Diffraction of Light. 7th ed. Cambridge University Press; 1999. doi:10.1017/CBO9781139644181
  1. Myers TL, Tonkyn RG, Danby TO, Taubman MS, Bernacki BE, Birnbaum JC. Accurate measurement of the optical constants n and k for a series of 57 inorganic and organic liquids for optical modeling and detection. Appl Spectrosc. 2018;72(4):535-550. doi:10.1177/0003702817742848
  1. Malitson IH. A redetermination of some optical properties of calcium fluoride. Appl Opt. 1963;2(11):1103-1107. doi:10.1364/AO.2.001103

The Field Guide is a living document. Questions, corrections, and suggestions for this chapter are very welcome — join the discussion in the Coblentz Field Guide forum.