Research plans

My main research goals:

  1. Explain LIBS Imaging as simple as clearly as possible

    • "LIBS Imaging for dummies" and this research web page

    • Wikipedia articles in English and Finnish

  2. Give the tools for others to do LIBS Imaging.

    • Reference datasets

    • Analysis codes

    • Device descriptions

    • LIBS analysis workshop course for teaching/learning

  3. Understand and model LIBS spectrum from physical principles

    • From laser beam to heated sample to spark to atomic excitations to emissions to detected spectrum.

    • The math formula from elemental concentrations to a spectrum as we see it on a spectrometer.

    • The reverse math=algorithm to find from a measured spectrum the elemental concentrations and plasma variables. (with gradient descent using single element contributions as variables)

    • An easily usable software tool based on this for anyone to do LIBS imaging analysis with.

Here below are download links to my previous research plans for context. Note that these are not up to date.

English translated version research plan pdf

Finnish version tutkimussuunnitelma pdf

More detailed notes on plans

Most of the goals mentioned in those pdf research plans are still valid but main focus is on the above mentioned three goals.

On how to Understand and model LIBS spectrum from physical principles (3.)

Laid out steps to understand and model LIBS spectrum from physical principles

  1. Need a model to calculate the LIBS spectrum from theory. This has been done already by example at libs-info and by NIST LIBS website, but they are lacking. Using the Atomic Spectral Database peak list to make up an artificial LIBS spectrum. Optimally would be done by modelling the plasma behaviour to get the temperature and density curves, but first we skip this part and use a reasonable estimate.

  2. Peak widening (Doppler and such) probably simply with a convolution / moving window method. Or alternatively fitting in the double-lorentz / voigt profile if needed but probably convolution is easier and good enough.

  3. Optional self absorption model. This isnt too significant with most of our data but would be nice to understand.

  4. Spectrometer light catching amount and sensitivity curve

  5. Black body radiation continuum model. Or we can use continuum removal and skip this but likely it's better if we don't need to use a general continuum removal but can actually include it in the model.

  6. Then we use this model and search for its parameters using gradient descent to make the theoretical spectrum match an experimental one. Parameters are elemental concentrations and plasma properties and spectrometer properties until latter ones are locked in then only elemental concentrations. Cost function should penalize going below zero to avoid fitting elements that arent actually there and possibly focus on logarithmic scale to catch small things and the differing scales.

  7. Result is the elemental concentrations. And the error between the theoretical and experimental spectrum.

When doing this with big LIBS images we can use the neighbouring pixel values as good initial guess for faster computation.

On how do I share my code (2.)

Git repositories. I need to clean my existing ones from any proprietary information and should throw away most of the (useless) legacy code.

CC BY-SA 4.0 Ilkka Laine. Last modified: January 28, 2025. Please contact me by email for any suggestions, comments or improvements.