Second derivative analysis

Share icon

Thu, 11/27/2014 - 06:36 am

Hi,

I am performing second derivative analysis on a broad peak related to my collagen raman spectra in order to reveal overllaping bands......However, when i take the second derivative of my original spectrum, it seems that a main peak that i was expecting to be the most intense is actually of lower height than others that seem to be less intense. I use linear baseline to get an initial idea of the major and minor peaks. I am bit confused from literature because i am not sure if second derivative analysis is suitable to separate peaks and define their positions, widths and heights or if the relative height of individual peaks is somehow distorted with respect to the original spectrum. Any ideas?

Thank you
Hi Konstantinos

The second derivative gets the rate of change of gradient. So, a broad peak will have a smaller second derivative in the vicinity of the peak centre than a narrow peak.
You said that you are trying to reveal (resolve ?) overlapping bands - if they are very overlapped then I think your best bet is to fit multiple peaks to the single broad feature.

Hope this helps,
Kurt

November 27, 2014 at 06:56 am - Permalink

Hi Kurt,

Thanks very much for your prompt reply......To illustrate this in a more clear way.......I have a broad enveloppe with several overllaping bands.......I performed second derivative analysis and i found these bands.......I can fit the second derivative in order to find the positions as well.....However, it seems that the relative intensities between peaks (at the second derivative) are different than the ones seen in the original spectrum (obviously due to the reason you gave me). Nevertheless, besides the precise positions of the peaks i want to know what are the true intensities of each convoluted peak and it seems that this is not reflected in the second derivative. I was thinking of using the second derivative to find the number and position of the overllaping peaks and then use the ''edit or add peak'' option and fit the original spectrum. This way i am assuming that Igor Pro will perform combinations and will give me the best values for positions, REAL peak heights and REAL widths according to the best fitting. Does that make sense at all?

Thanks again

November 27, 2014 at 07:53 am - Permalink

Konstantinos Chatzipanagis wrote:
Nevertheless, besides the precise positions of the peaks i want to know what are the true intensities of each convoluted peak and it seems that this is not reflected in the second derivative. I was thinking of using the second derivative to find the number and position of the overllaping peaks and then use the ''edit or add peak'' option and fit the original spectrum. This way i am assuming that Igor Pro will perform combinations and will give me the best values for positions, REAL peak heights and REAL widths according to the best fitting.


I think the best way to get the true peak heights of several Raman bands that make up a broad feature is to fit the spectrum itself and not any derivatives. If you have some knowledge of the peak positions and widths you can add these to the fit function when doing a multiple peak fit to this spectral feature. Depending upon how overlapping your bands are, you may also want to add constraints to the peak height (such as must be greater than zero) to help the fit process.

November 27, 2014 at 11:53 pm - Permalink