We use both transmission and ATR on our samples : transmission for solvent extracts and ATR for solid samples (surface or sections, pigment powders, more rarely liquids). Dealing mainly with non pure analytes (non separated mixes, composites...), we decided not to use the ATR correction as a general rule. The main reason was to keep band ratios as "we"re used to seeing them" : as you mention, this is one of the drawbacks with ATR, which we noticed.

To work around this, we also decided to build our own spectral libraries using both transmission and ATR sampling, whenever possible. When working ATR, we try and use mainly ATR libraries, and reversely for transmission.

We still use ATR correction as a way to use transmission libraries when dealing with unknown spectra acquired using ATR, but that's the only case we do this. For instance, it might be useful when doing FTIR imaging by ATR Imaging once you want to identify phases.

As we're only using FTIR for qualitative identification or imaging purposes, we never worried about the subject that much to be honest.

Thanks for your opinions! I guess it might be interesting to see about what we are talking. Here is an example: PMMA using a ZnSe or Diamond crystal under an angle of incidence of 45°, s-polarized. The black spectrum is how it would and the red spectrum how it should look like (if there would be no optical effects).

The effects increase with increasing oscillator strength, which means they also increase with concentration.

We are using the software (OPUS) provided with the Bruker FTIR-spectrometer (Vertex). The ATR correction has not proved itself to be accurate even when the parameters such as the angle of incidence, the dielectric constant of the ATR crystal, and average high-frequency dielectric constant of the sample are specified. The problem concerns the last one since usually it is not known precisely. Moreover, varying it in the small range dramatically changes a "corrected" spectrum. Therefore, this ATR correction is dubious and thus not popular. Nevertheless, the ATR spectra are still qualitatively correct and their comparison within a defined set of samples yields reasonable interpretation.

Dear Dmytro, many thanks! I can imagine that knowing the high-frequency dielectric constant is critical, in particular if there are regions where through dispersion the index of refraction becomes so high that the ATR condition is no longer valid. Changing the high-frequency dielectric constant and, with it, the high-frequency index o refraction means that a larger or smaller spectral region has been recorded outside the ATR condition - this must strongly influence the correction scheme if the ATR condition plays a role in this scheme which it obviously does. The challenge is that the high frequency dielectric constant does not play any (obvious) role in the spectrum as long as it is recorded well above the critical angle for the non-absorbing region. If it would be possible to change the angle of incidence gradually around this critical angle, it should be possible to determine the high frequency dielectric constant accurately based on the recorded spectra...

Thomas Mayerhöfer, thank you for the feedback and the piece of advice (sorry, I have noticed your reply only now).

I can fully relate to the issue of the ATR condition loss. We often encounter it using diamond as an ATR crystal as well as having the incidence angle fixed to 45°. Otherwise, we are used to determine high-frequency dielectric constants by means of spectroscopic ellipsometry, so getting the value is not an issue. However, it will not help if the spectrum is recorded with locally (spectrum-wise) failed ATR conditions.

Dear Dmytro Solonenko many thanks again! Actually, I have rearranged a program that I used for correcting transflection spectra lately - it is partly based on dispersion analysis and does not care if the ATR condition is violated.

Working on it, I realized that modern ATR accessories do not work with polarized light and the correction scheme requires to know the polarization state as Rs and Rp can be very different.

I assume that the correction schemes of the vendors also need to know the polarization state, but simply assume naturally polarized light, even if we all know that light coming from the optics of a conventional FTIR spectrometer is never naturally polarized...

So it seems one has to perform a kind of determination of the polarization state first before a useful correction is possible, in particular because it seems that at least in some cases polarizers cannot even be added due to space limitations (and, maybe, such a retrofit does not make sense because the polarization state is changed in the accessory...).

One more comment on the advanced correction procedure... actually I mistakenly thought that an advanced correction procedure should actually calculate the optical constants... instead it is much less than that, but actually more than just a multiplication with a normalized wavenumber, so from this perspective it is certainly advanced. On the other hand, why not go the extra step and also safe the spectra for which the index of refraction is too high in certain spectral ranges to perform a correct ATR experiment? Probably to safe the bucks for the polarizer and/or the optimization of the design of the accessory instead...

Dear Thomas Mayerhöfer, thank you for opening this interesting conversation (which I found a little too late).

I use FTIR imaging to identify microplastics. Our standard method is to measure in transmission mode, however, most spectra databases comprise only of ATR spectra.

I am quite new to this field and was surprised how big the differences between transmission (absorption) and ATR spectra are. This issue is seldom explained in the literature on microplastics. We additionally face difficulties due to Mie scattering on particle surfaces, and as a consequence, microplastic spectra often show great differences compared to the schoolbook ATR spectra.

We apply a machine-learning approach to analyze the FTIR images because we have a large amount of data that cannot be handled manually anyway. We use only absorbance spectra to train the algorithm so no ATR correction is necessary.

Still, I am wondering about a very substantial thing which is probably trivial for experts. That is, how ATR correction exactly looks like mathematically. Of course I tried to google it, but couldn't find any satisfying information. The information on ATR correction given by FTIR machine vendors is rather a black box as it seems...?

Dear Jana Weisser , many thanks for your contribution! Unfortunately, some of the subtleties of IR spectroscopy have been lost over the years and others have never been explored. It is the usual way with seemingly standard techniques, and the instrument manufacturers are usually not very interested to promote pitfalls of the methods... Concerning the "Advanced ATR Correction" I could myself not find the full formalism how it works, but from what I understand -lg (R/R0) is divided by an penetration depth or an averaged effective penetration depth (averaged under the assumption of having a spectrometer spill out natural polarized light). Since this requires the knowledge of the function of the index of refraction, I think that for the correction first an approximate index of absorption is calculated, and, after a Kramers-Kronig analysis, an approximate index of refraction is obtained with which -lg (R/R0) is corrected by dividing it by the penetration depth or the averaged effective penetration (depth). This certainly only works as long as long as the index of refraction is for all wavenumbers lower than a certain threshold which marks the limit of the internal reflection depending on the type of ATR crystal and the angle of incidence. In other words, the critical angle, which is a function of wavenumber must always be smaller than the actual angle of incidence of the accessory...

Dear dr. Thomas Mayerhöfer

Good discussion

Best regards

Following

https://www.essentialftir.com/advancedATR.html

https://assets.thermofisher.com/TFS-Assets/CAD/Application-Notes/D10241~.pdf

https://www.labcognition.com/onlinehelp/en/atr_correction.htm

Dear dr. Jana Weisser

Best contribution

Best Regards

Finally I also came in contact with the Advanced ATR Correction. The problem we had was to measure the spectra of ideal liquid mixtures to see if the additivity of absorbance holds in this ideal case.

Certainly, we must make certain that what we see is not an artefact from the type of measurement. The system we looked into was the most famous (quasi-)ideal system Benzene-Toluene and the results before correction were not useful at all (see figure 1).

I was surprised how good the correction worked in this case - we validated the results for the function of the absorption index of the neat substances with literature values (cf. figure 2). One reason for the good correction may be that the refractive indices of Benzene and Toluene in the transparcency region that you need for the correction are virtually identical, which means that this also holds for the mixtures.

Overall, this means that for quantitative application of ATR spectroscopy you definitely need to correct ATR spectra, except if you use very weak peaks (which must not be in the vicinity of strong peaks). By the way, there is still some peak shift after correction, but this is due to the fact that Beer's approximation does not hold even for ideal mixtures. The whole story can be found under Article Beyond Beer’s Law: Quasi-Ideal Binary Liquid Mixtures