Hello Everyone !
I have a samples a satinless steel , I performed the LIBS Measurement with hLIBS (SciAps Z-300).
Total of 12 measurements on different locations were done on the same sample piece , After the acquisition of Raw Spectra (Unprocessed) , I did try to normalize the spectra by two techniques :-
Attached is the plot , for Normalization (by one of the Matrix Element i.e Fe) , By which other techniques , could i improve the results ?
Also for building the calibration , is it advisable to average the 12 processed spectra (Baseline correction + Normalization) ,inturn to produce 1 spectra , which would represent my sample as a whole ?
All your suggestions will be truly appreciated and helpful !
Thanks and Regards,
Rahul Patil
You can try Standard Normal Variate (SNV) normalisation. You should subtract the mean of each spectral feature (in this case intensity for each wavelength) and then divide it with standard deviation. In other words, you find the mean and std of all intensities for each wavelength, and then do the aforementioned subtraction and division.
This should bring each spectrum to the mean of zero and std close to one. It should look better.
Averaging is advisable to reduce LIBS reproducibility error. Every LIBS spectrum is actually average over 10 or 20 successive laser shots. Very often authors do 10 shots on one place, then repeat it for lets say 20 other places on the target. Then the final spectrum is the average of all 200 spectra. On these spectra, calibration models are often constructed.
I agree with the answer of Ivan Traparic. LIBS measurements are usually having very high Standard Deviation from the mean value. This makes LIBS lagged, compared with AAS, ICP, ICP_MS. This is one of the main drawbacks of this technique. To overcome this drawback, you can use different areas of the target sample for LIBS measurements as well as averaging the results of each sample zone.
Thanks a lot, the results are indeed better after SVN.
Just one more question, When I am building a Regression Curve , can't we plot all the datapoints, for example, if I had done a measurement at 12 spots on a single sample, can I do a regression for all those 12 intensity values with concentration?
What the points or advantage of Averaging, those 12 spots results , into just 1 resultant spectra ?
Thanks and Regards,
Rahul
It would be better to average all 12 intensity values, because calibration curve would be more accurate. Do this for all the standard samples you have.
Just watch to select good spectral lines for calibration curve, as some of them can be prone to self - absorption. Just bear this in mind (just a remainder, nothing else).
Best regards,
Ivan
@Ivan , how can i identify the lines , which are prone to self-absorbtion ?
First of all, don't use resonant lines as are they are most prone to self - absorption.
Next, if you want to be sure, you can check the line intensity ratio of lines within the same multiplet. If theoretical and experimental intensity ratios agree or are close, then you can say that there is no self-absorption. For this you can use NIST database.
Best regards,
Ivan
- Badges
- Science topic
- Similar topics
- Biological Science
- Physiology
- Physiological Processes
- Stress