EDS quantification

[Eric Leroy]

Hi,
I wonder if it is possible to quantify an EDS spectrum or spectrum image inside hyperspy. It would be a nice feature implement.

I also noticed a bug for the importation of rpl eds datacubes in GMS2: the importation works fine but after no quantification is possible. The software always ask about the detector parameters and even after they are entered still continue to ask.

Regards

Eric

[Francisco de la Peña]

Hi Eric,

I wonder if it is possible to quantify an EDS spectrum or spectrum image inside hyperspy. It would be a nice feature implement.

Hyperspy does not have the necessary features to perform absolute quantification of EDS spectra. It would certainly be interesting to implement this feature but, although I am happy to help to anybody who wants to do it, I don't usually work with EDS spectra so I am not the man for the job.

A nice open source package for EDS quantification and simulation already exists: DTSAII. I might be wrong but it seems that it is a java GUI for the EPQ library, that has Python bindings. What neither DTSAII or EPQ seems to have are tools to work with multidimensional data (what happens to be Hyperspy main strenght). Therefore, don't you think that the easiest way forward is to write "wrappers" around EPQ/DTSAII functions to enable EDS quantification in Hyperspy?  

 

I also noticed a bug for the importation of rpl eds datacubes in GMS2: the importation works fine but after no quantification is possible. The software always ask about the detector parameters and even after they are entered still continue to ask.

It seems that Digital Micrograph needs some tags that are not exported in the RPL file in addition to the detector parameters. To solve this problem we need to know which are the missing tags but I don't have a way to find it out myself because I don't have the EDX quantification DM package... Can someone provide us with this information?

Best wishes,

Francisco
 

Regards

Eric

--
You received this message because you are subscribed to the Google Groups "hyperspy-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to hyperspy-user...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.
 
 

[Sergio Lozano-Perez]

Hi,

 

I agree that it’d be nice to have some tools for EDX quantification in Hyperspy but it’ll require some time and effort. As Francisco pointed out, there are already some useful free packages such as DTSA II that do a good job and come with an open-source library of functions. I should, however, point out that DTSA II is mostly focussed on SEM EDX and is not really suitable for thin-foil analysis/simulation. To overcome that problem I’m collaborating with some people in Cadiz (Spain) to re-write the original DTSA code (which was fully featured for TEM work) into a more current platform (at present it only works for PowerMacs). I’ll update you when we have something ready.

 

Regards,

 

Sergio

[Luiz Fernando Zagonel]

Dear all,

I can work in the EDS quantification tags as I have all DM plugins (DM
version 1.83.842). Still, I am not sure how the quantification
software (plugin) uses the input information within the tags. If
somehow they affect the quantification, the results may be incorrect.
Some tags, for instance, are really microscope dependent and do affect
the quantification quite dramatically like the HT.

I think some discussion would be needed to see which tags can be
filled by the RPL import (with some "standard" valued, like takeoff
angle, detector type, etc) and which tags could be added manually
(maybe with the help of an GUI), like HT, etc.

I think that some script to copy all tags from a single spectrum
acquired as usual within DM to another single spectrum resulting from
Hyperspy is a simple and elegant (although limited) solution. Such
script could try to see if the target is a volume and deal with that
appropriately as well. Should we go for that?

With best regards,

Luiz Fernando Zagonel.

[Eric Leroy]

Hi,

concerning the quantification of EDS spectra, in fact it more simpler than SEM EDS quantification. Since that the method employed (Cliff-Lorimer method) is only based on the calculation of the different ratio of the peaks present in the sample, the only thing we need is the possibility to fit the spectrum with gaussian peaks at the different energies of the elements. After we have to solve a system of equations with the different Ca/Cb=kab*Ia/Ia ratios plus a supplemental condition that is sum of the concentrations is 100%. I wonder if something like that is difficult to implement in hyperspy. A second step could the to introduce an absorption correction with the Beer Lambert law to take account of the thickness of the sample. An finally a great feature that is not any microanalysis software would be the determination of the local thickness when two series of X-Ray lines of the same element are present in the spectrum (K and L or L and M) : if you quantify the same spectrum once with for example the K line and once with the L line, you should obtain the same result. Generally it is not and when you introduce the absorption correction you can make two loops with the thickness variing and you can reasonabily consider that when the result is the same for K and L lines you found the local thickness. Then finally you take this thickness and apply it to the quantification of the spectrum. Generally a precision of 10% in the thickness can be obtained. With this method you can have reliable quantification and also the thickness map.
What do you think about this ?
Regards

[Francisco de la Peña]

Eric,

Fitting Gaussian to  EDS spectra is something that can be easily implemented in Hyperspy. It just requires a new EDSModel Model subclass that, similarly to the EELSModel subclass, reads the position of the peaks from a database and adds peaks to the model with reasonable starting parameters. It seems that, due to the relative simplicity of the EDS spectrum, it is possible for an algorithm to make a satisfactory guess of the elements present in the sample, what means that the generation and fit of the model can be fully automatize. This will probably be enough to obtain qualitative distribution maps. From your message I guess that getting quantitative information involves post-processing of the results of the fit and probably requires some extra input (e.g. k-factors).

I think that what makes Hyperspy different from other software solutions to similar problems is that it is not only open source but also open development and that it is written in a language, Python, that is easy to learn, yet powerful. These explains why it has many features that are not available in other programmes: there is a subset of users that are at the same time developers and "promoting" from user to developer is relatively easy (compare to other software solutions). Obviously the users are the ones who better knows their needs and such a development community is a fertile ground for new interesting features. I mention this because it'll be definitely great to implement the EDS quatification features that you mention in Hyperspy but, because I am not an EDS user at present, I cannot lead this development. However, as I mentioned before, I will be very happy to help to anybody with experience in EDS analysis who wants to do it.

Best wishes,

Francisco

2013/2/20 Eric Leroy <ricou...@gmail.com>

[Pierre Burdet]

Hi,

I'm working on 3D EDS with a FIB/SEM. Discussing with Francisco, I found that hyperspy could be a nice platform to developp the tools I need, such as EDS quantification and different spectrum image treatment. With the help of Francisco, I just start adapting hyperspy for EDS (for the moment importing data from INCA oxford-instrument). I'm mainly focused on SEM. I would be happy to help on TEM side when the main tools for SEM are robust enough.

Regards

Pierre