OK, I hoped someone more proficient in mineral analyis would answer here, but since that does not seem to be the case, here are some general Raman points:

first of all, minerals with weakly polarizable constituents are not exactly the strength of Raman, but you can of course find characteristic bands for which there is also a database:

https://rruff.info/about/about/about_software.php

From existence of these bands you can get a rough idea of your composition. Please note, however, that since the polarizability of many minerals is low, you might miss minor and trace components. Second, the peak intensity is not just determined by the amount of the species in the sample but also proportional to the cross section. The most important factor for the latter is how strongly the polarizability of the crystallite changes during the vibration that you excite and this, in turn, does not only depend on the general crystal structure but also on defects and interfaces.

Thus, while Raman is obviously helpful for a qualitative analysis, a quantitative view will be hard.

I hope this was more than you already knew. Do not hesitate to ask if you have a more detailed question but as long as I'm the only one who answers please remember that I don't use Raman on minerals but other samples.

I suggest to get in touch with Prof. Dr. Andrei Buzatu who is an expert on RAMAN spectra and also a member of Researchgate.

H.G.Dill

Hello,

I have used the openly available software CrystalSleuth in the past, as also indicated above by Dr. Weippert. It is easy and convenient to compare/contrast your measured spectra with 'standard' results from a comprehensive integrated RRUFF database (data can be in *.txt file format). One can also edit the spectra to a certain degree and export them in Excel etc. for additional processing.

Best,

Martin

You can try non-negative least squares fitting to unmix the concentration of individual minerals. I used this concept to unmix the components in human skin: Article Raman active components of skin cancer

You would get the relative concentration, but not the real concentration, because the signal intensity is proportional to the Raman scattering cross section of individual minerals.

Hi Ananya,

One thing you could do is use the PCA analysis to its full capability. The PCA analysis should give you scores of every measured spectrum on each principal component axis. Let's call this the training scores. Next, you could load a spectrum (or multiple) from the RRUFF database and project it on the training PCs and get a score for this too. Let's call this test score. Subtract the test score from the training score (euclidean distance). The least score difference is the closest match to your mineral spectrum.

PS: Another easier way would be to write a code for determining peak positions in your spectra and then comparing it with the RRUFF database peak positions. The comparison should also give you a coefficient to determine how close it is. If you have a lot of measured pixels, PCA can be computationally expensive. Also, you need to have at least a small idea about what the spectra most likely would resemble, instead of testing through the entire RUFF database.

I agree with Dr. Xu Feng, Raman intensities are dependent on the cross-sections and without prior knowledge of this, it is very unreliable to estimate the concentrations. Relative concentrations seem to be better, but again, your intensities also depend on various other factors. My best suggestion would be to report concentrations based on some other technique (I don't know what could be used, sorry not an expert on the same).

Best,

Sachin