SiO2 on SiC

[mjpavel]

Hello I have recently come across this community for thin film analysis using EPMA and am wondering if the following situation is calculable using BadgerFilm:

- estimating the thickness of an SiO2 layer (ranging from 30nm - 500nm) formed on an SiC substrate using EDS

I have pieced together a pathway using other posts from this website which led me to using multi-voltage analysis from 3kev to 10 kev and assuming a fixed composition for the SiO2 layer and extracting k-ratios from DTSA II which I had previously been using for just quant eds

I tried using DTSA II to simulate a 100nm film of SiO2 on SiC at 3, 5,  7, and 10kev, and then simulating SiO2 and SiC standards for quantifying the film spectra. Using those k-ratios I calculated a thickness in badgerfilm and my results are not promising - using every combination of fitting I converge to ~250nm. My fitted output also does not look great - the k ratios all follow a trend which I can visually see quite well, however the fitting is not good. Is this sort of a core limitation of the technique or is there something else I should look in to?

Thanks!

[Probeman]

Hello I have recently come across this community for thin film analysis using EPMA and am wondering if the following situation is calculable using BadgerFilm:

- estimating the thickness of an SiO2 layer (ranging from 30nm - 500nm) formed on an SiC substrate using EDS

If you only need to know the thickness and not the composition of the film, you might be better off just doing a cross section and seeing the thickness in SE.  Another method would be critical angle x-ray reflection.

I have done a number of MVA EPMA measurements where the both the substrate and the film in question share a major element and it is possible to get a convincing result but one must constrain either the composition or the thickness.  But it depends greatly on the physics details and the accelerating voltages selected.  In these situations I've always only needed to determine the compositions, so we measured thickness using x-ray reflectivity and constrained that in the MVA models to obtain compositions.

Since it appears that you already know compositions for both your substrate and deposited film, you would declare the substrate and thin film composition as fixed and leave the film thickness as unknown.  It should be able to converge.

I know STRATAGem can often handle this situation, so BadgerFilm should be able to also.

[mjpavel]

Gotcha I may try stratagem and see if I have better luck - I also learned today that some of my detector settings did not copy over well into the new version of DTSA II which may have given me weird simulation results so more to play with there.

The sample geometries are frustrating to the point that they must be cross sectioned via ion beam and are not suitable for XRR and we're just killing time on FIB when we are already looking at the samples for quant EDS - would be more efficient if we could knock out a few measurements in one session. Since many of the samples have been FIB cross sectioned I have a sort of "ground truth" layer thickness to start comparing the fitting against..our lab is getting a lot more of these samples going forward and wanted to see if this was viable before really going after it. I appreciate the insight!

[Ben Buse]

Hello I have recently come across this community for thin film analysis using EPMA and am wondering if the following situation is calculable using BadgerFilm:

- estimating the thickness of an SiO2 layer (ranging from 30nm - 500nm) formed on an SiC substrate using EDS

I have pieced together a pathway using other posts from this website which led me to using multi-voltage analysis from 3kev to 10 kev and assuming a fixed composition for the SiO2 layer and extracting k-ratios from DTSA II which I had previously been using for just quant eds

I tried using DTSA II to simulate a 100nm film of SiO2 on SiC at 3, 5,  7, and 10kev, and then simulating SiO2 and SiC standards for quantifying the film spectra. Using those k-ratios I calculated a thickness in badgerfilm and my results are not promising - using every combination of fitting I converge to ~250nm. My fitted output also does not look great - the k ratios all follow a trend which I can visually see quite well, however the fitting is not good. Is this sort of a core limitation of the technique or is there something else I should look in to?

Thanks!

Hi MJ Pavel

That's an interesting experiment

It would be interesting if you made your data & files available here, - so we can see your results

Obviously you are testing whether Monte Carlo agrees with analytical matrix correction.

Also your setup means it's all about how well you can measure oxygen and carbon, with Si being in both layers,

Did you try the different MACs and models within BadgerFilm?

[Ben Buse]

Gotcha I may try stratagem and see if I have better luck - I also learned today that some of my detector settings did not copy over well into the new version of DTSA II which may have given me weird simulation results so more to play with there.

If you are running simulations in DTSA-II and using the Monte Carlo table of data - the detector shouldn't make a difference - that is for generating the simulation spectra.

[mjpavel]

Hello I have recently come across this community for thin film analysis using EPMA and am wondering if the following situation is calculable using BadgerFilm:

- estimating the thickness of an SiO2 layer (ranging from 30nm - 500nm) formed on an SiC substrate using EDS

I have pieced together a pathway using other posts from this website which led me to using multi-voltage analysis from 3kev to 10 kev and assuming a fixed composition for the SiO2 layer and extracting k-ratios from DTSA II which I had previously been using for just quant eds

I tried using DTSA II to simulate a 100nm film of SiO2 on SiC at 3, 5,  7, and 10kev, and then simulating SiO2 and SiC standards for quantifying the film spectra. Using those k-ratios I calculated a thickness in badgerfilm and my results are not promising - using every combination of fitting I converge to ~250nm. My fitted output also does not look great - the k ratios all follow a trend which I can visually see quite well, however the fitting is not good. Is this sort of a core limitation of the technique or is there something else I should look in to?

Thanks!

Hi MJ Pavel

That's an interesting experiment

It would be interesting if you made your data & files available here, - so we can see your results

Obviously you are testing whether Monte Carlo agrees with analytical matrix correction.

Also your setup means it's all about how well you can measure oxygen and carbon, with Si being in both layers,

Did you try the different MACs and models within BadgerFilm?

Hi Ben,

I actually realized I screwed up the badger film calls and when I brought in my K ratios from DTSA, I did not specify that Si, C and O came from SiC and SiO2 - my own inexperience not thinking through what it is actually calculating. Once I specified what standards my k-ratios were based upon, the film thickness came to ~105nm which for me is plenty good enough resolution for layer thickness.

As for the rest of the community, it would be interesting to know if the 5nm comes down to the slight differences in the Monte Carlo sims and then analysis or not. I have only last few weeks gone down the deep dive into what the other MAC models actually mean - I know better understand the sort of origin in error in carbon analysis (MAC of c from graphite vs c from SiC)

Our solution to the microscopy problem was to acquire a real SiC EDS standard so that we may start to minimize some of the PRZ corrections. We had previously been performing our quant with CaCO3 and SiO2 and our analytical totals were +-10% off 100 so we know we've got some work to do