The methodology described here could also be employed to remove the spectrum of a buffer solution from IR specra. It least squares fits the spectrum with a combination of a reference spectrum, the substrate (buffer) spectrum, and a polynomial, and then subtracts the spectrum of the substrate + polynomial. The process is relatively insensitive to the profile of the reference spectrum especially at low polymer orders.

Note, that subtraction applies to spectra in Absorbance, not in Transmission.

Article A multivariate statistical investigation of background subtr...

Hugh J Byrne The paper you suggested was developed for solid samples in Raman spectroscopy. While it has some similarities to my question with regards to FTIR, I am looking for methods that were developed and tested on FTIR. But, thank you for your time and answer anyways :)

Ardeshir Moeinian - it doesn't matter - I regularly use it for subtraction of water from Raman spectra of aqueous samples and it can also be used for FTIR spectra, as long as you convert them to Absorbance from Transmisison.

Th algorithm doesn't know whether the samples are solid or liquid, or whether the spectrum is from FTIR or Raman.

Hugh J Byrne-Thank you for the suggestion. I tried the algorithm for FTIR Absorbance spectra and unfortunately it doesn't work. The reason is that the calculation of a baseline is not straight forward. FTIR spectra of aqueous solutions contain not just positive but also negative peaks due to absorption by water being stronger than for example proteins. Perhaps this algorithm can be adopted for FTIR but at the moment it is not compatible.

FTIT spectra should not have negative peaks.

Hi Ardi,

also some suggestions from my side. However, maybe you tried some already:

1. Would it be possible to use the buffer solution as background spectrum? You should end up with already quite "clean" spectra.

2. Could you use the baseline correction method from your spectrometer control software? A certain manufracturer offers a quite smooth option to simply click "aqueos solution" and let the method do its magic.

3. Could you use deuterated solvents to shift the strong absorbance features of water further away from your protein bands?

4. Could you use third party software that lets you try to multiply and subtract the sample spectrum with a buffer spectrum by hand, to get a feeling what is happening in your spectra?

5. Could you use more advanced chemometric tools to compensate the water features?

6. Why transmission? ATR typically is much simpler to handle in terms of sample application and solvent absorbance.

Hope some of my suggestions are useful for your issue

Cheers

Julian

Hi Julian Haas;

Thank you very much for your answer!

Unfortunately it is not possible to use to the buffer solution as background according to the FTIR-setup's manufacturer. We can only use water as background measurement. Afterwards, the spectra of buffer and protein solution are measured with reference to water.

The main issue: Because the concentration of water in buffer and protein solutions is less than pure water (of course!) the spectra of buffer and protein solution show negative absorption at wavenumbers where water has strong absorption peaks e.g. 1640 cm-1. This makes the calculation of a baseline difficult. Also, because the concentration of water in the buffer solution and the solution containing proteins is different, the spectra show a different absorption at wavenumbers where water absorbs in the MIR range, which is around the Amide I band of proteins. Although the difference is not a lot, it does make it difficult to compare the buffer and protein spectrum in a quantitative manner.

We can't use deutrated solvents because the goal is to study the proteins in water.

At the moment I am using Matlab to calculate the difference spectrum (protein - buffer spectrum) and I am looking for algorithms that would do the trick.

For the calculation of Difference spectrum, the most common method mentioned in literature, is to use a scaling factor when calculating the Difference spectrum such that the resulting spectrum has a 0 baseline between 1750-2000 cm-1. This wavenumber range is used as reference because any absorption here should be only due to water. When a factor is used, the resulting spectrum does have a 0 baseline between 1750-2000cm-1 but it also contains negative peaks where the buffer solution shows absorption peaks.

I use a Transmission measurement cell with an optical path length of 8 µm for the measurements. According to multiple manufacturers transmission cells with such optical path lengths are optimal for aqueous solutions with low protein concentrations (less than 10 mg/ml). The optical path length is similar to multibounce ATR-FTIR and should give very similar results. The advantage of transmission measurement, is that the optical path length stays always the same and can be measured very precisely.

I hope everything's going great!

I think it is necessary to add a word of caution: In contrast to Raman spectra, the additivity of absorbance requires an additional level of approximation. In other words, even when you have a (thermodynamically) ideal mixture, the spectrum of the mixture will only for weak peaks be a linear mixture of the pure component spectra. We recently proved this for mixtures of Benzene and Toluene experimentally (the paper will appear soon in Applied Spectroscopy). The theoretical proof is simple and based on the fact that light polarizes matter (actually one could say that the existence of Raman spectroscopy proves that absorbances cannot be additive). See Article Beyond Beer's Law: Spectral Mixing Rules

Hi Ardeshir Moeinian ,

Don't know if you're still working on this issue.

I've done a lot of protein structure analysis in water, buffer etc. In my case it was insulin with concentrations from as low as 3 mg/ml or 0.5 mg/ml for concentration determinations. We used ATR-spectroscopy - and even with your comment on why you use transmission cells, it might be worth having a look at fiberoptical ATR probes. In comparison to multibounce ATR elements, you can submerge the whole ATR element into the sample and collect reproducible spectra.

A comment on Julian Haas suggestions - Using spectroscopy software, that offers some manual tools for the spectral subtractions, gives very valuable information. Most software (I use OPUS from Bruker) also offer uni- and multivariate data anlysis tools based on least square algorithms.

Happy new year!

Sven Delbeck If you used Beer's approximation for calibration (or chemometrics based on Beer's approximation), then I hope you used molar concentrations.