The one approach is to use CLS with known spectra and analyze the rest or use MCR with known compounds and decompose the rest of the spectra to the possible components. Keep in mind that Raman spectroscopy works OK for concentration down to 1-3 %, depending on the material. In environmental samples, it may be everything, including soil, plastic, etc. Organic content shows very distinguished spectra, and the specific Raman band may be assigned to functional groups. The inorganic part is more complicated as the Raman cross section usually small and spectra not very bright. It also depends on the laser excitation. If you are using a 785 nm laser, the chance to collect the spectrum from SiO2 will be minimal if the concentration is not high. For transition metal oxides, 633 nm is the best choice. Environmental samples usually show high fluorescence. Anyway, I would start from MRC or CLS. The neural network requires many spectra and good statistics. It is a challenge in the case of Raman spectroscopy. Any questions, let me know at [email protected].

Radek.... Focusing on a specific method or a technique to detect or determine the content of a samples is difficult task. The subject depends on many factors: 1) the level at which a component is present; 2) the similarity on the composition between the various contents, 3) the nature of the samples and how much is available for the analyst; Raman spectroscopy as a method depends on the vibration al behavior of the atoms and groups in a compound. The introduction of computational models and the ability of electronic deconvolution of spectra from the various components helped the researchers to determine certain components. Other techniques must be employed to detect the behavior of the various components. If Raman spectroscopy can be applied to certain purpose it is definitely unable to detect or determine other components.

The ability to unmix those compounds depends on the orthogonality (or similarity) of the basis spectra. If they are highly similar, then the chance of accurately unmixing them would be low. Is it possible to categorize those compounds into sub-groups? For instance, in skin sample we determine the basis spectra based on lipids, proteins, nucleus, pigments, etc. Reference: Article Raman active components of skin cancer

I agree with Mahmood that other techniques should be incorporated to find the true composition. You can also make a phantom with known concentration and detect with Raman.

As an addition to the great replies, I can suggest Raman libraries/databases. I haven't worked with environmental samples. But if you have some initial guesses on the ingredients of the sample, you can narrow the search. I have worked with a commercial database software, which also had the option to import external libraries.

So, let's say your sample contains organic molecules mainly. You purchase (or create by measuring many samples) an organic molecules database. Then, the sofware lists the most similar components based on their correlation coefficients. You can couple this information with the literature to validate.

Remember, the databases can never give you the quantitiy. It just says "X may be in your sample". Quantification is something you need to do in your experimental setup as Sergey Mamedov suggested. PLS (partial least squeares regression) is another option to his suggesitons, too.

it is not an easy task to analyse the environmental samples by Raman microscope although it can detect and quantify the analyts below 0.01% w/w in my experience.

it can be good to detect organic compounds but definitely not feadible for some inorganic materials, e.g. silicon.

Having said that, it is still worth applying Raman microscope to study the environmental samples, particularly regional organic compound contamination in our environment, due to its advantage, e.g. free sammple preparation, etc..

To achieve that task, I would suggest you to start with a ·clean· soil spiked with some common organic compound contamination. By this you could learn how to perform mapping or imaging for better or optimised configuration and thus informative data.

From my almost 20 years experience of applying Raman microscope, it is highly likely that you cannot get what you need if your configuration is not optimised for each analysis.

Unlike other ordinary techniques, analysis by Raman microscope is case dependent. You have to optimize configuration in data collection for each analysis. This would greatly help yiur second step, i.e. data processing and analysis, anothet challenging task when you use Raman microscope.