Three issues arise with the choice of a Voigt versus a Gauss + Lorentz. First, in principle, the Voigt function is a convolution of Gauss and Lorentz, not a sum. This is a critical issue to appreciate. Secondly, the Voigt profile is a non-analytical function and uses approximations when formulated in code routines. This is a minor issue that none-the-less should not be forgotten. Finally, certain key parameters (half-width, area) for a Gauss or Lorentz peak are well-defined "fit-constants", while those from a Voigt profile are not. This is a practical consideration that should be taken in to consideration relative to the needs of the results from the fit.

Whether you _should_ use a Voigt or a Gauss + Lorentz sum is dependent on the first issue - what principle applies for your system. Whether you _do_ use a Voigt instead of a Gauss + Lorentz sum to fit a peak profile may be a matter of practical implications with regard to the third issue - the approximation of a convolution by a sum can be so close where the need to have well-defined peak parameters becomes more important.

To finish about the principle ... The peak profile an instrument measures is a convolution of the intrinsic line-width of the event and the instrument broadening, not a sum. Instrument broadening is always inherently Gaussian (unless it is skewed from a normal distribution). When your intrinsic line profile is Lorentz, you _should_ use the Voigt profile to fit the peak. When it is Gaussian, you should use a straight Gaussian profile, and the broader base in the peak profile arises from something other than a Lorentz component. In either case, clear knowledge of the line broadening from your instrument and/or the expected inherent line widths from the events (Raman scattering in your case) can help prevent you from straying too far to over-estimate one or the other contribution to the final profile.

It is best to use IGOR pro

Fit with "Voigt" (this is a convolution of Lorentz and Gauss), use PseudoVoigt for a linear mixture.

indeed you can with Origin 8 with Voigt function in the nonlinear fit tool

There is a function called voigt profile

In addition to answers that you have already got, I'd like to recomend you to find the programme PeakFit. It is very useful for fitting different peaks, for example in Raman and Photoluminescense spectroscopy.

There is also easy to use program Fityk, which can fit Gaussian, Lorentzian, Voigt functions. For me it looks more flexible than Origin.

Yes. I use Origin 8 in menu "Analysis" option "Peak and Baseline" has option Gauss and Lorentzian which will create a new worksheet with date, also depends on the number of peaks. Maybe make baseline applying the function.I hope I helped.

Some peple do not trust the voigt function.

GO to Google and do a search on Voigt lineshape. If you use MatLab you can download a GUI program, IGOR will also work.

Yes, the multi-peak fitter is a good tool... as others have said voigt or pseudo-voigt should fit your needs... though you should be able to define an arbitrary function as well if you need to, and fit that...

I could suggest to use another program to fit with a pseudo voigt fuction, the program is called fullprof.

Be careful in using Origin function and take into account that (in Origin 7.5 I use) Voigt and PseudoVoigt2 presuppose different widths of G and L contributions, whereas in PseudoVoigt1, these are taken equal.

When you must use separate functions, i.e. Gauss for the Rayleigh peak, Lorentz for the central peak and damped harmonic oscillators for Raman peaks, I understand the full version of Origin can fit them separately. I trick the student version by making a custom function that is the sum of all three, and under "settings" tab in the fitting function I select "advanced" and choose the number of replicas I need. The first replica should only be the Gauss, the second only the Lorentz, and the remainder only the DHO's. It is cumbersome but effective if you have no alternate program.

Hi Tylan,

To fit Raman spectra I suggest you use dedicated programs, like PeakFit or FiTyk, mainly because of the subtraction baseline and work in different regions of the spectrum. But, if you prefer you can use Origin. Make a summation of Lorentzians and Gaussians in a non-linear fit option.

I have successfully been using Voigt functions to simultaneously fit multiple Raman and LIBS peaks (up to 20) using a custom-written Win32 C program that reads SPE and SPC files. As for any peak-fit, the first-guess positions of the peaks must be initially specified. My program does this interactively and displays the data and fitted peaks. The resulting Gaussian and Lorentzian widths depend quite critically on effective baseline removal. I use the SNIPS baseline removal method which is effective, simple and fast..

PseudoVoigt function is Lorentian and Gaussian function togeter

I found that using Excel program does the best for deconvolution of spectra using any function that you like

Hi Tylan,

in Origin you can also create your own function in a quite easy way. Therefore you can define a sum of a lorentzian function and a gaussian function to fit your curves.

Regards

According to my experience, fit peaking module of Origin 8.0 works very good and I am very satisfied with it.

Regarding profile function, I could also recommend Voigt. It is convolution of Gaussian and Lorentzian, and the best model for natural band shape. There are also apparently similar functions, such as pseudo-Voigt, (Gauss+Lorentz) etc. but they have not physical meaning, at least in analysis of vibrational spectral bands.

Use voigt function in the origin 6 for fitting your curves. It is already present in origin 6 and higher version.

Peakfit for windows can fit Raman peaks with voigt function by simply selecting the same.

For origin 7.5 and 8, I suppose a user defined function will be a good option. However, I agree to all those friends, who said that voigt and pseudo-voigt will serve the purpose.

IF you use Voigt profile, i think it will solve your problem. Voigt is a convolution of Lorenzian and Gaussian profiles. It is readily available in origin 8.0 and above pro ver. I have used origin to deconvolute raman peaks of my samples and i assure you that it works.

go through this link to see more about Voigt profile:

http://en.wikipedia.org/wiki/Voigt_profile

Thanks to all for the answers. My problem is coming when I try to use the "Fit Multi-Peaks" option. When using this tool Origin gives you only two choices, Gaussian or Lorenztian but no Voigt or any other profile. I can see that I can fit a single curve using Voigt etc. but I am unable to deconvolute contributions to the overall curve from separate peaks (representing two separate populations of the system). I'm beginning to agree with some above answers that Origin is perhaps not the best software for this. I started using a trial version of PeakFit and it seems to be working quite well as far as fitting peaks with appropriate functions is concerned. The drawback with this program seems to be the time required to asses each spectrum, although I just started familiarizing myself with the program so still not sure. I've been in touch with someone at my institute who developed a Matlab GUI for doing this so hopefully this will be a useful tool.

Referring to Patrick's answer here is a link to an article discussing the basics of curve fitting: https://www.thermo.com/eThermo/CMA/PDFs/Product/productPDF_57565.PDF.

I am relatively new to spectroscopic techniques so I am still approaching everything from a basic level of understanding.

Here is a short piece of literature in which the authors are doing something similar. They use Voigt profiles with Lorentzian imposed to 60%.

Attached is a similar project by a separate author in which they use pure Lorenztian fits to model essentially the same thing. Is this an appropriate model? I think in my case pure Lorenztian mimics physical reality relatively well for some samples but others don't make sense when analyzed this way. If anyone has any other good advise on this or knows of any literature that could be of more help I would definitely appreciate it.

Thanks

to Patrick El-Khoury: the isotropic component in Raman spectra offer the information about the vibrational dephasing time, and the anisotropic component of Raman spectra offer the vibrational and reorientational information.

Dear Tylan,

Looking through the answers to your question (including mine) I’d like to ask you to specify your point. The thing is that data fits is my job and a greater part of my scientific activities, but – on the other hand – I perfectly realize that an approach one is choosing depends on ones aim. For example, my aim in fitting spectroscopic (Raman and Rayleigh) data is time-correlation functions. As I see from your comments, you deal with inelastic Rayleigh line plus vibrational Raman lines (possibly overlapped), isn’t it? So, my question is of what is your final target? If a simple decomposition of the spectra, you may use any software colleagues suggest. If you mean any specific further moves, the line decomposition softs would vary.

Should you be interested in time-correlation functions (or generally in further Fourier transforming decomposed line profiles) please look at my review “Novel approaches in spectroscopy of interparticle interactions. Vibrational line profiles and anomalous non-coincidence effects” published in a book entitled Novel Approaches to the Structure and Dynamics of Liquids: Experiments, Theories and Simulations, 2004, pp.193-227, it is in my papers on the Researchgate site. In it and in references therein, some examples taken from my papers dealing with very tricky cases, like multicomponent Rayleigh plus the so-called Boson peak plus numerous Raman lines, etc., are described. Anyway, an approach presented in this review might be interesting, from the point of view of both the spectra treatment and theories behind various line profiles, from overgaussian to Gaussian to Voight to Lorentzian to overlorentzian. I use a bit specific function to model the spectrum, it’s an analytical Fourier transform of a certain flexible time-corelation function. It allows for modeling any symmetric profile from abovementioned ones. Everything is in the review.

If you are more utilitarian and are planning to obtain the best fit due to an unknown reason (e.g., esthetical, it’s also good and understood), even Origin (after proper dividing the spectrum by several parts) may be useful, but it’s another story…

Summarizing, I would say that should I (or maybe others) know of what is your purpose, I would be more specific in my suggestions.

Don’t hesitate to ask any question by [email protected].

All best!

My end aim is to quantify the populations of two similar but separate complexes that give overlapping peaks in the Raman spectrum. I'm not dealing with Rayleigh scattering, these are all Raman bands. I'm doing something similar to the papers I posted above. In those, two bands are observed, one from free TFSI anions and the other from Li coordinated TFSI anions. The band from Li-TFSI is shifted by about 7-8 cm-1 and overlaps the free TFSI peak. How do you know what model is appropriate when fitting the two bands so you can deduce the contributions from each species and ultimately integrate underneath the fitted curves to calculate how many TFSI anions are coordinated to Li and how many are free in the solution? That is ultimately what I'm getting at. Hope this is more clear.

Thanks again for your responses

I have overlooked citations you made, it appears that you are at the beginning of the road we have already passed. My graduated student is now studying vibrational dynamics in solutions of LiTFSI and EMIM TSFI in various solvents, and a small piece of data has already been published, see a paper attached. It is a continuation of another paper I attach as well (it will be in a next reply since it seems to me that two documents cannot be attached together).

With any questions feel free to write to [email protected].

This is more early paper.

To Tylan Watkins: previously I also employed Raman spectra to study the ion pairs in liquid water, my solution is Mg(NO3)2/water. And I succeed to employ Lorentz profile to decompose the concentration dependent Raman spectra, finally I obtained the concentration dependent of the number of various ion pairs. My suggestion is that you could use Lorentz function to fit your data, and maybe you could think of the solvent shared ion pairs, not only the contacted ion pairs.

Excellent topic, question and discussion. My interest in this topic is related to real time Bragg XRD Microscopy where the rocking curve profiles have multitudes of shapes and these shapes and combinations thereof may be used to interpret the defect types. I have not yet read all your wonderful comments fully. I will and then post my remarks. Thanks for sharing everyone!

Response from: Thangam Parameswaran

A Voigt function is a combination of Lorentzian and Gaussian functions. I have used Voigt functions to fit peaks in my research but I have not used the Software Originlab. However I looked up their website which indicates Voigt as one of the fit functions which you may use. I am attaching this info. I hope it helps.

You don't have to use the default fit functions. You can always create a user defined function according to your needs. This can of course be a superposition of Gaussian and Lorentzian if that makes sense.

Tylan! Please post your data (profile) so that we may get an idea as to the possible convolution of shapes that may apply in your case.

Here is a very helpful link that I have referred to routinely:

Paul Barne, Simon Jacque, Martin Vickers; X-ray diffraction peak shapes – Gaussian, Lorentzian (Cauchi), Pearson VII, Voigt, Pseudo Voigt

a. http://pd.chem.ucl.ac.uk/pdnn/peaks/peakcon.htm

b. http://pd.chem.ucl.ac.uk/pdnn/pdindex.htm#peaks

Here are some typical profiles and the use of "shape fitting" in Bragg XRD Microscopy:

http://www.flickr.com/photos/85210325@N04/8561945111/

http://www.flickr.com/photos/85210325@N04/8550145504/

http://www.flickr.com/photos/85210325@N04/7805008376/

[email protected]

Sviatoslav Kirillov! "“Novel approaches in spectroscopy of interparticle interactions. Vibrational line profiles and anomalous non-coincidence effects” published in a book entitled Novel Approaches to the Structure and Dynamics of Liquids: Experiments, Theories and Simulations, 2004, pp.193-227, it is in my papers on the Researchgate site."

Please send me all your published work regarding the use of peak shape to deconvolute parameters from profiles as soon as convenient. I'm looking through your impressive profile and am very much interested in learning more about this subject. Generally, I'm interested in peak positions, FWHM, Integrated area/volume, asymmetry, shape, distortion at tails/shoulders from predominant shape, etc.

[email protected]

"combination of Gaussian and Lorentzian functions" - Pseudo-Voigt and Other Functions - http://pd.chem.ucl.ac.uk/pdnn/peaks/others.htm

Hope this helps.

Dear Tylan and all:

My apologies. A review on "Novel approaches..." has not been attached yet. Yesterday I've uploaded it in my Publication list.

Dear Ravi,

With this review I uploaded three key papers on the subject, all published in Chem Phys Lett: (i) Fitting the stretched exponential model...; (ii) Markovian frequency modulation...; (iii) Time-correlation functions from band-shape fits...

In other papers (not all of them are mentioned in my ResearchGate publication list) I just use this technique.

I am much flattered with your interest to this method. Long ago, I have thought over an opportunity of introducing it in X-ray analysis, especially, in Rietveld's procedures. Unfortunately, being busy with other obligations (I am less active in vibrational spectroscopy than previously), I was unable even to approach the problem. Maybe you are the one to do this job. If yes, I wish you any success. Don't hesitate to ask any question you may have.

All best,

Slava

Slava!

Please forward PDF files when available to my email. Thanx!

[email protected]

Ravi, I posted a raw data file in my data sets (titled "Raman data"). I am not at liberty to give specifics but you can clearly see the type of line shape convolution I am getting. I have been using a Matlab GUI to fit these peaks with Lorenztian/Gaussian combinations; I was unable to figure out how to do this properly with Origin. Generally about 50% L and 50% G fits pretty well. My biggest question now is how do I know how well I am representing a physical reality in terms of number of states based on integration under these fitted peaks. It seems like I can deviate a bit by changing FWHM, peak positions, etc. and get equally good overall fits and yet obtain slightly different values for integrated areas. If you have any further suggestions based on this data let me know.

Thanks

To Tylan Watkins, you could try Igor Pro software, there is Voigt function to be empolyed easily to fit spectra.

One note of caution to keep in mind with regard to Voigt versus Lorentz + Gauss peak shapes. The exact area under a Voigt peak has to be determined numerically. The areas under Lorentz and Gauss peaks can be determined analytically from their peak parameters.

The voigt function is a combination of the two line shapes and is readily available in Origin.

@T Watkins: I second the suggestion from K Lin for Igor Pro. It is likely a comparable if not a better alternative to Origin. It is a heads-and-shoulders better alternative to MatLab. I also second the suggestion from T Vaczi that you set the parameters for the component peaks in the fit based on physically-relevant reasons rather than that you allow them to fall out based on blind faith in a computer program.

After working with both Origin and MatLab, I wrote a Win32 application in C that interactively fits and plots Lorentzians, Gaussians or Voigtians to peaks in SPE or SPC files. I have successfully used this program to fit groups of up to 30 overlapping peaks in both LIBS and Raman spectra (see attached example). If the Voigtian is inverted properly, the result is both a gaussian width and a Lorentzian width for each peak. The relative sizes of these two widths indicate how much of the two types are present. I have found in practice that the relative amounts of Gaussians and Lorentzians is quite variable, even in a single spectrum, and a Voigtian is therefore preferable. Baseline removal is critical and I use the very simple SNIPS method, which many spectroscopists presently seem unaware of.

Yes Origin 8 to fit those... Instead of using convolution, i prefer pseudo voigt to extract the contribution G and L.

Here's the example: (it's not use assymetric function, so it didn't fit well)

It is 3:15am local time. I have a fulltime (9-5) job in an unrelated field. I do this for fun now! I should've participated and learnt from such discussion 3 decades ago. Better late than never. Thanks for all the free-flow info! Really appreciate all participants' contributions.

I shall present my paradigm as well which is exactly similar yet quite different with similar solutions. I'm sure our quest will be embellished with your wisdom.

I'd like to congratulate Tylan for asking a basic yet a highly relevant question and would like to extend the utility of this discussion further in to spectroscopic analyses.

Maybe even a preplanned "Google Hangout"?!

to be continued.........

Sir Chandrasekhara Venkata Rāman, FRS (7 November 1888 – 21 November 1970) was an Indian physicist whose work was influential in the growth of science in India. He was the recipient of the Nobel Prize for Physics in 1930 for the discovery that when light traverses a transparent material, some of the light that is deflected changes in wavelength. This phenomenon is now called Raman scattering and is the result of the Raman effect.

https://en.wikipedia.org/wiki/C._V._Raman

Dude's attired like my Grandpa (I'll post a photo later)!

My interest in the principles of spectroscopy is for Bragg XRD Microscopy. Several comments caught my attention as we are faced with these challenges in day-to-day 2D real time XRD data analyses.

1. Tamás! " There is certainly a Voigt function that can be used for fitting in Origin8." I'd like to learn this aspect well. Any "short-cuts" to learn?

I have access to Origin but have not yet learnt all the features well. Getting better everyday!

2. Barry! "Baseline removal is critical" - Yes! Identifying correctly and "fitting the baseline" is crucial for analyses.

3. "(e.g. the instrument profile)" This is significant because all measurements are relative to this profile. Must be calibrated with reference standards precisely.

My example is for GaAs (004) crystallographic reflection using XRD (X-ray Diffraction).

1. Theoretical profile (Instrument?) based on optics used:

http://www.flickr.com/photos/85210325@N04/9319044423/

2. Experimentally observed data superimposed on theoretical profile and the deviation quantified:

http://www.flickr.com/photos/85210325@N04/9321837294/in/photostream/

A related detailed discussion on LinkedIn:

http://www.linkedin.com/groupItem?view=&gid=2683600&type=member&item=257747275&qid=a74d4276-69c6-44c8-b947-986a0da76aac&trk=group_most_recent_rich-mc-rr-ttl&goback=%2Egde_2683600_member_261112437%2Egmr_2683600

Please join us and share.

http://www.flickr.com/photos/85210325@N04/9321837294/in/photostream/

I'd like to determine the correct profile and extract parameters using Origin. I have the ability to export the data 1D and 2D!

What about Pearson VII? Anyone have experience with that shape?

A quick purist's note on the precision of terminology to use here. Convolution / Deconvolution are entirely different mathematical operations than Peak or Baseline Fitting. The two operations are also used to separate two entirely different physical effects that contribute to the shape of a peak in spectroscopy (or diffraction).

In a perfect-world approach, you would take two steps to analyze any experimental spectrum. You would first deconvolve the (Gaussian) transmission function of your instrument, and then you would fit physically-relevant (Lorentz + Gaussian) peaks AND baseline functions to the result. The fact that you might fit a Gaussian-broadened component peak as a way to model the transmission function broadening by an instrument should never be promoted as deconvolution, and we never deconvolve a baseline from a spectrum, we fit the baseline.

Jeffrey,

Thanks for the clarification. My intent was to indicate that there may be some additional signal in the "base line" such as a low frequency component. In general, I see your "purist" point regarding the use of "deconvolute" instead of "fit" for the baseline. The peak shape would have to be fit prior in order to deconvolute the profiles (topographs) from diffractograms containing both K Alpha 1 & 2 components.

Many of the shapes we are encountering with Bragg Peak profiles for various materials require other than the pseudo-Voigt "fit". We notice that the Pearson VII (split fit - asymmetric) works better in many cases. Have you any experience with the Pearson VII profile? Would you have any suggestions in that regard?

Here is an attempt (visual) at qualitatively fitting shapes to the theoretical X-ray RCP (rocking curve profile) for (004) GaAs using a Bruker simulation for Bartel (220) beam conditioning optics: http://www.flickr.com/photos/85210325@N04/9417942940/

My objective is to follow-through with the "numerical fit" using Origin. There are several "pointers" already in this discussion for me to follow to fruition. Surely, I continue to need help. Therefore, I present my results for you to review and suggest any improvements.

Other Bruker simulated RCP's for different beam conditioning options. (i.e., instrumental factor):

http://www.flickr.com/photos/85210325@N04/9321836998/

Comparing theoretical & experimentally observed RCP's for GaAs (004):

http://www.flickr.com/photos/85210325@N04/9321837294/

2D Real Time High Dynamic Range & High Resolution XRD Microscopy!

http://www.flickr.com/photos/85210325@N04/9417942940/

@R Avanth: "The peak shape would have to be fit prior in order to deconvolve the profiles ..." I still wonder that you are using the word deconvolve incorrectly and to some unwarranted degree of ambiguity. I recognize that literature abounds with the statements equivalent to "the one peak was deconvolved in to N components" when what is truly meant and should identically be stated instead is "the one peak was fit with N components".

But this word play is an aside to the topic of this thread.

I have no experience with the Pearson peak shape or with fitting curves to XRCP data. My background is peak fitting in chemical spectroscopy. The principles that we both must rigorously apply overlap exactly, even as the applications that we both undertake differ tremendously. I am glad to discuss off-line whether my insights from this point might be of help to you.

Here is the example I was discussing in terms of separating (deconvoluting) the K Alpha 1 & K Alpha 2 topographs from the convoluted diffraction image that is a result of both these wavelength components being present in the incident beam impinging on the Quartz sample wafer:

http://www.flickr.com/photos/85210325@N04/7852409924/

After testing, Gaussian & Pearson VII (m=5) are "near perfect" fit (symmetric) for the shape of these RCP's for the Quartz sample.

I would think the "diffraction topograph" would be the convolution of the incident beam (consisting of both K Alpha 1 & 2 wavelength components) and the wafer Nano structure at the VOXEL of interest. In XRD topography we are able to observe both the real space (topograph) and reciprocal space (Omega/2-Theta) simultaneously. I was referencing deconvoluting the individual rocking curve topographs from these convoluted diffraction images. Would that be an acceptable use of the term? Since we are in a sense integrating the signal arriving at each pixel I'd think the mathematical representation would be a convolution. Wouldn't it?

Since we are making measurements in the reciprocal space in XRD some of these convolutions (in real space) turn out to be simple additions in reciprocal space, especially for Gaussian distributions.

"the one peak was deconvolved in to N components" is not the context I used the term. We accomplish that using FFT's.

This end of the theory was never my forte either. I will learn and improve. Thanks!

For any "off-line" comments, my email is:

[email protected]

@R Ananth: "Here is the example I was discussing ..."

Generally, deconvolution is done with a van Cittert iteration or a Fourier transform method, peak fitting uses chi-squared optimization, and profile extraction pulls out a single layer from an NxMxP matrix. Additions of components in reciprocal space are convolutions of the components in real space. When you subtract out a component in reciprocal space and then invert the result, you are doing a deconvolution operation in real space.

Your use of the term "deconvolution" is correct then. Thank you ... I have learned something here about XRD data analysis.

I continue to learn everyday! Thanks!

Tylan Strike Watkins! I'm looking for the answer to your question as well for fitting the same in XRD profiles using Origin 8.0. If I figure it out before you do, I'll share right here.

Can I do this for 2D data? I do know how to import 2D data into Origin Matrix. Then what? How do I access the "nonlinear fit tool" for 2D data analyses? YouTube tutorial?

I could try 1D to get started. I can import 1D data as well.

If back to the question, what is the purpose to deconvolute the single peak back to gauss-lorentz peak function? What is the physical meaning on that? e.g.: X-ray diffraction, gaussian broadening - instrument, lorentzian - crystallite size etc.

If you just wanna find a peaks inside the peak, you must know an tell the software where is the peaks lie on, and let the software do the job.

Here u the axample of peak that content multiple peaks and sloopy baseline, it also can deconvolute (Voigt) to Gausian-Lorentzian if u need so, just look at the detail of it in Origin 8.

FWHM = 0.5346 * wL + sqrt(0.2166 * wL * wL + wG * wG)

wG = Gaussian FWHM, wL = Lorentzian FWHM

Yes

A nice fit algorithm for a wieghted sum of Gaussian and Lorentzian functions is implemented in GRAMS (by Thermo Scientific), a software developed (originally) by Galactic. These fits are just a small portion of a powerfull suit of programs for spectra collection and manipulation. The suit is sometiomes called as GRAMS/AI, and has been previously known as "GRAMS/386", later "GRAMS/32" .

Links to official sites:

http://www.thermoscientific.com/ecomm/servlet/productscatalog?storeId=11152&categoryId=81850

http://www.adeptscience.co.uk/products/lab/grams32/

http://www.asdi.com/products/grams-suite

http://3d2f.com/programs/41-586-grams-ai-download.shtml

!!!!!!!! Not sure whether you want to do sum or convolution of functions, but many mistakes are made with this in chemical/polymer spectroscopy literature.

Please read as tutorial Vibrational Spectroscopy 39 (2005) 266–269

Michael! "what is the purpose to deconvolute the single peak back to gauss-lorentz peak function? What is the physical meaning on that? e.g.: X-ray diffraction, gaussian broadening - instrument, lorentzian - crystallite size etc." - Perhaps, because the actual shape of the Bragg profile may give us clues in to the material Nano structure!

Please review the following discussion and provide some erudite input if possible:

What are the possible Nano structural causes of asymmetry in the Bragg X-ray Rocking Curve Profiles for GaAs (CZ) & SiC (CVD) mono-crystal wafers?

Are there any publications concerning these topics?

1. Sub-grain/Mosaic Structure

2. Stacking Faults

3. Twinning

4. Dislocations

5. Other

https://www.researchgate.net/post/What_are_the_possible_Nano_structural_causes_of_asymmetry_in_the_Bragg_X-ray_Rocking_Curve_Profiles_for_GaAs_CZ_SiC_CVD_mono-crystal_wafers2

@R Ananth: FWIW, Igor Pro can handle analysis of 2-D data (images). The developers (http://www.wavemetrics.com) and users group (http://www.igorexchange.com) are exceptionally responsive and extremely well-versed in all aspects of data analysis. I recommend it highly over Origin for doing anything serious in the analysis of experimental data.

it's easy using a powerfull origin 8 or 9 software. You can find more details about the fitting pics in the youtub "fitting pics with origin". And it's realized unsig the both functions. See my paper "microcrystalline cellulose 2014."

Hi, I cannot give you an answer to whether or not this particular program can do that, but very relevant is whether the physics is correct to allow for Gaussian+Lorentzian fit. In generally a Voigt profile (product of Lorentz and Gauss) is the physically justified form, leading to correct quantitative results

This is called Voigt profile. You can find the function in the data analysis of Origin. It is used when the lorentzian width (for instance a Raman peak) is of the same order of the resolution (for instance a Gassian laser profile).

Even if you need a sum of Gauss and Lorenz: Origin provides a Tool for creating user-defined fitting functions. (See "Tools ->Fitting Function Builder"). An easy way to create a wanted fitting function is to edit the Gauss and Lorenz Function at first, using "Tools->Fitting Function Organizer", and to copy the codes from the Fields “Function”, “Parameter Settings” and “Parameter Initialization” in any text editor. After that, you can paste them into your user-defined function, using "Tools->Fitting Function Builder". Of course, you need to decide which parameters need to be the same for Gauss and Lorenz and which to be different, and to rename them according to that.

E.Shishonok

I remember that in Origin is there the Voigt option in "multiple peak fit".

This is what you need. The Voigt function is a sum of two functions you mention.

Dear Tylan,

I see your point and can support the idea of a half-Gaussian-half-Lorentzian fit for an absorption spectrum in liquids. The problem is easy to solve and you can use any code. Origin allows you to define your own function, as Andjela suggested.

You will find the proper function in the formula [S6] of the supplementary material to my paper J. Chem. Phys. 118, 5916-5931. The supplementary material can be downloaded at the URL http://keszei.chem.elte.hu/papers/JCP118supplement.pdf .

(Oter related material - including the paper - is available at the URL http://keszei.chem.elte.hu/papers/JCP118.htm .

Are talking about a function which is a combination of Lorentzian and Gaussian???? If so, you can use the function named Voigt available in Origin. This function is basically a combination of above two functions.

Good Luck

I think this question was already answered some time ago. It is called Voigt profile. This kind of profiles happens when the resolution of the system (gaussian profile of the laser) is of the same order of the phonon linewidth. It is an intrinsic function of Origin.

Three issues arise with the choice of a Voigt versus a Gauss + Lorentz. First, in principle, the Voigt function is a convolution of Gauss and Lorentz, not a sum. This is a critical issue to appreciate. Secondly, the Voigt profile is a non-analytical function and uses approximations when formulated in code routines. This is a minor issue that none-the-less should not be forgotten. Finally, certain key parameters (half-width, area) for a Gauss or Lorentz peak are well-defined "fit-constants", while those from a Voigt profile are not. This is a practical consideration that should be taken in to consideration relative to the needs of the results from the fit.

Whether you _should_ use a Voigt or a Gauss + Lorentz sum is dependent on the first issue - what principle applies for your system. Whether you _do_ use a Voigt instead of a Gauss + Lorentz sum to fit a peak profile may be a matter of practical implications with regard to the third issue - the approximation of a convolution by a sum can be so close where the need to have well-defined peak parameters becomes more important.

To finish about the principle ... The peak profile an instrument measures is a convolution of the intrinsic line-width of the event and the instrument broadening, not a sum. Instrument broadening is always inherently Gaussian (unless it is skewed from a normal distribution). When your intrinsic line profile is Lorentz, you _should_ use the Voigt profile to fit the peak. When it is Gaussian, you should use a straight Gaussian profile, and the broader base in the peak profile arises from something other than a Lorentz component. In either case, clear knowledge of the line broadening from your instrument and/or the expected inherent line widths from the events (Raman scattering in your case) can help prevent you from straying too far to over-estimate one or the other contribution to the final profile.

I agree with everybody but Ii will do a silly question: are you fitting your spectra before or after removing background? Because removing the polynomial bckg usually would let you fit to loreentzian function easily with Origin... I do particularly agree with Dr Kirilov,s considerattions but I suspect you are not removing bckg and you look only peak positions,, areas' ntensities and fwhm perhaps, if thats the case remove bkgnd fit lorentzian

I agree with everybody but Ii will do a silly question: are you fitting your spectra before or after removing background? Because removing the polynomial bckg usually would let you fit to loreentzian function easily with Origin... I do particularly agree with Dr Kirilov,s considerattions but I suspect you are not removing bckg and you look only peak positions,, areas' ntensities and fwhm perhaps, if thats the case remove bkgnd fit lorentzian

@N Ahmed: I am entirely confused by your post. It is a set of questions rather than an answer. It seems to be entirely unrelated to the original post. Finally, it has rather confusing grammar.

Perhaps you might post your questions as your own topic (rather than hijacking this thread) and give us sufficient background to allow us to construct good answers.

Dear Jaffrey Weimer,

I am working in the field of LIBS and my interest is to find out the electron density from the Stark broadened observed line profiles which appear in the emission of a Laser produced plasma plume. We may use Lorentzian function to find pout the FWHM. However, we were suggested to use Voigt function, a convolution of Lorentzian and Gaussian,  to extract the FWHM and then calculate the electron number density. I will appreciate if you suggest  which relation is more appropriate to be used to calculate the electron number density from the FWHM:

1. FWHM=  WL : Lorentzian contribution

2. FWHM=  WL - WG :

3.FWHM=  WL + WG

4. FWHM = 0.5346xWL + SQR( 0.2166xWL2 + WG2)

The relation which connects FWHM with the electron number density is:

Ne= (FWHM/2Ws)x1016 , where ws is the Stark broadening impact parameter.

Best regards,

Nisar

There is one big drawback if you use origin or the commercially available software packages. They actually do not offer you the most physically reasonable function which is that of the classical damped harmonic oscillator model (CDHO). In fact, while everybody uses Lorentz-profiles it seems that only few people are aware of the fact that the Lorentz-profile has been derived from the CDHO by applying three different approximations and those hold only very good for comparable weak oscillators.

Another drawback of such software is that the Lorentz-profile was originally introduced for only one oscillator. It seems that the extension to more than one is straightforward, but if you do it in the correct way, then you can not only obtain the oscillator parameters like damping, strength and position, but also calculate the changes of the index of refraction function (without having to invoke the Kramers-Kronig relations!) in the infrared or UV-Vis spectral range.

Read the full story here:

Article Quantitative Evaluation of Infrared Absorbance Spectra - Lor...

I think there is Voigt fitting options in newer version of origin does any tell me whether or not this function is related to Voigt equivalent circuit could it be used for fitting impedance curves?