Full Example of Raman Noodle Suite
To walk you through the full Raman Noodle suite of tools and show the robustness of the tools applied to Raman and Infrared spectra we built tools that can take generic, in this case open source Infrared spectra from the NIST WebBook database and perform analysis.
- Regarding the use case of experimental data we have included in the repository with an example data set of (open source) experimental Raman spectra data collected from a supercritical water gasification reactor - a Juypter notebook example, raw data in Microsoft Excel format, and individual output images can be found in the
Raman-noodles/developmentfolder linked here. In the same folder has been included the old method of baseline subtraction by the group in Matlab script format and formating the file inRaman-noodles/development/MatlabCodelinked here
The following steps outlines how to go through the tools we've created and how they link together. The user interface for these individual tools is in linked Juypter notebooks that are linked along with the step they are associated with.
- NOTE: shoyu is a cleaver pun on the Raman Noodles theme of our software - (for "show you") - and in case you don't know, shoyu is the Japanese name for soy sauce - Link to Wikipedia Soy Sauce Page
The shoyu Juypter notebook is the starting point. It will walk you through (using markdown and inline comments) the set up the environment for using the NIST wrapper and will download IR data and combine component spectra to create a theoretical mixture of components.
- If you have your own data (or want to just focus on our experimental Raman spectra data) you can move right to step 2
Now that data has been defined and imported the next step is to wrangle the data. To do this the following is done:
1. Import from other data source (example excel for experimental data)
2. Unit check - Ensure sure units are the same throughout
3. Baseline subtraction - Used PeakUtils built in baseline function to perform polynomial fit baseline subtraction on NIST database spectra.
- Future work includes making this tool more robust
- Experimental data from the supercritical water gasification reactor is already baseline subtracted, as such, this was used for only the NIST IR spectra data
An example output to the user for the NIST Water IR spectra can be seen below:
4. Curve fit and parameter description - The LMFit package was utilized to identify 5 descriptors per peak in a mixture’s Raman signal including: location of the peak, peak height, and peak width. As a mixture has more peaks (components from decomposition) the amount of descriptors increases. For examples here was with a Cauchy–Lorentz distribution for the example in the Juypter notebook, but can also be easily changed for a Gaussian distribution and Voigt distribution
The Cauchy-Lorentz distribution follows the following equation:
An example of this used for carbon can be seen below:
These steps are shown in the spectrafit_example juypter notebook.
Now the data has been uniformly formatted and 5 identifiers have been defined for each peak in the spectra. The final step is to identify the peaks. This is done currently by using the x-position (wavelength, cm^-1) of literature values for the component, and then comparing to what our spectrafit code has determined. Once components in a ‘testing’ dataset mixture are identified the next step is to take a ‘training’ dataset location and share with the user the Euclidean distance between the two datasets. For this software if a peak location is more then ±10 cm-1 from the literature values the confidence that the peak represents the compound is zero. This range was set from experimental considerations.
Below is an image example that is output to the user showing the peak identified in their spectra of CO2, and the literature values of where the peak should be.
Now you are all set to sit at the table and consume all the Raman or Instant Raman (IR) -pun intended - spectra data your heart (or stomach) desires! 🍜