Thin Film quantitative analysis

[Pointman]

Hi, everyone!

To start with, I'm a beginner at Monte Carlo simulations and EPMA.  I have just recently came by your forum and i was looking for some help.

I am trying to verify some thin film EDS quantitative analysis measurements that were made in our lab and so far I was unsuccessfull at doing so.

I have already figured out that i can use CASINO simulations, to get intensities + I came across the program CalcZAF which gives matrix corrections and it seems that together they provide accurate info on bulk samples, but i don't quite understand how i should work with thin films. I also found tutorials on the forum, but they are quite confusing and i wasn't sure if they are explaining the things that i need to figure out.

So, in short. Lets consider 170 nm. thick TiN film (5,4 g/cm3) on Si substrate, which is being penetrated by 10 kV beam. How should i figure out the quantitative analysis results for this particular sample?

I would greatly appreciate some advise 

[Probeman]

Hi Pointman,
Probeman here.   

You can also use the Penepma 2012 GUI to model thin film geometries from a specified composition and geometry as described here:

http://probesoftware.com/smf/index.php?topic=57.msg208#msg208

And yes, CalcZAF can calculate composition of thin films from intensities, but it does not model substrate effects.  That said, it does a surprisingly good job so long as there are no significant fluorescences from the substrate. There is some discussion of using CalcZAF for matrix correction of particle geometries here (for thin film geometries just use the first geometry option):

http://probesoftware.com/smf/index.php?topic=281.msg1354#msg1354

But you might also try DTSA II which is available here free from NIST:

http://www.cstl.nist.gov/div837/837.02/epq/dtsa2/

It appears Nicholas has recently added improved thin film calculation options in the latest version:

NIST DTSA-II Jupiter 2016-05-04

    Requires Oracle Java 8
    A major upgrade from Iona - many major and minor updates, new features and improvements
    New - "Standard Bundles" make using complex standards much easier
    New - ζ-factor quantification of thin films
    Many subtle improvements when quantifying low-energy x-rays
    Cleaner, more modern icons and appearance.
    Moved to Jython 2.7 for scripting


john

[Pointman]

Thanks, Probeman

Unfortunately Im still having trouble with CalcZAF and  Penepma.

The problem is that Penepma says straigth upfront that simulation takes 10000 s. SERIOUSLY??? Why would it take hours? And the tutorials only talks about the k ratios, but is that with ZAF correction or without it? It is really confusing.

CalcZAF with the thin film correction just returned me identical results. Might be missing something though.

 Right now im giving a try for the third option NIST DTSA-II. Ill tell how it goes

[Pointman]

Ok, It seems that NIST DIST-II have just did the trick. Although i have one more question for the probing community.

I have just done the quantification for the previously described 170 nm thick TiN film. But i got quite huge uncertainties (circled).



Any ideas how to make this huge uncertainty lower?

In the program, i've used pure bulk Ti, N, and Si. as standarts.
 

[John Donovan]

Ok, It seems that NIST DIST-II have just did the trick. Although i have one more question for the probing community.

I have just done the quantification for the previously described 170 nm thick TiN film. But i got quite huge uncertainties (circled).



Any ideas how to make this huge uncertainty lower?

In the program, i've used pure bulk Ti, N, and Si. as standarts.

Hi Pointman,
Funny you should ask!

Nicholas Ritchie just opened a new forum board on using DTSA that he will be moderating here:

http://probesoftware.com/smf/index.php?board=32.0

So if I were you, I would create a new topic on that board and post your question there.  I'll bet the developer (Nicholas) will respond appropriately!   
john

[Probeman]

Thanks, Probeman

Unfortunately Im still having trouble with CalcZAF and  Penepma.

The problem is that Penepma says straigth upfront that simulation takes 10000 s. SERIOUSLY??? Why would it take hours? And the tutorials only talks about the k ratios, but is that with ZAF correction or without it? It is really confusing.

CalcZAF with the thin film correction just returned me identical results. Might be missing something though.

 Right now im giving a try for the third option NIST DTSA-II. Ill tell how it goes

You can specify the Penepma simulation time for whatever precision you want.  The default of 10000 sec is for high precision. But remember, with Penepma you need to run two simulations. One for the unknown intensity and one for the standard intensity.  You then divide them to get the k-ratio, which is the intensity without any matrix correction.

Also, if you use a pure element for the standard, you can easily extract the matrix correction because the matrix correction with a pure element standard is simply the unkconc * stdint/unkint. Where the unkconc is in weight fraction.

[Pointman]

Ok, since DTSA II option is off the table for now. Lets focus on CalcZAF option.

Im not sure if i'm doing the right thing, because i dont seem to be getting anything useful out of it.

For example i use the material editor and define a bulk TiN sample and press "Calculate"




... and it gives me the desired k ratios and ZAF matrix corrections - awesome

Now lets add thin film correction from "Analytical" and then "Particle or thin film". Set the diameter - 0.17 um. Density - 5.4 g/cc and it gives me these results




Who are almost identical to the previous ones. Am i just missing something?
I assumed that it would show, for example a combined 30% of mass portion for the thin film elements - then one could write off the remaining 70% for the substrate. But, obviously it didn't do that. The provided tutorials dont discuss thin film stuff that much as well. On the other hand, it might be that i just dont understand how it all works - some explanation would be more than helpfull.


By the way. I've tried out GMR Film but so far its no use because it constantly gives me this error. It happens when i have finished inputing the data and the calculations start. Any ideas what is wrong?



So the question remains about how one should quantify such a Film

[Probeman]

Hi Pointman,
It looks to me that you are in the default mode which calculates intensities from concentrations. What you want is one of the calculate concentrations from intensities modes.

There are 4 options in the same window as the Calculate button. You want one of the options other than the first one.
John

[Ben Buse]

Hi Pointman,

Are you sure you want concentration or can you just compare simulated kratios to measured kratios - please see reply to the post in DTSA-II

http://probesoftware.com/smf/index.php?topic=728.0

Ben

[Pointman]

Hi, everyone!

Thank you all for giving me good responses. It took me some time to figure things out but I've made it. CalcZAF did the trick and it works. My mistake was that i didn't understand correctly what exactly the thin film correction does - I was expecting quantitative analysis results, which would include the substrate, when actually having the correction enabled, but it does the opposide - duhh...

Anyway, I would like to follow up with an actual research related question:

We are still talking about the same 170 nm thick TiN film, which was measured with a bruker FEI Quanta 200 FEG SEM. The measurements were done at 2, 5, 10, 20 and 30 keV. In the chart, I have subtracted Si substrate and some trace elements, comparing only the mass fraction of titanium and nitrogen. These are the results that i got:



It is clearly visible that when the energy becomes higher it starts to register more and more nitrogen and less Titanium. It looks weird. The measurement of other, thinner TiN film also returned similar results. It looked like it is some kinf of malfunction, but, surprisingly CASINO simulation and CalcZAF calculation of a similar model film,consisting of 40% Ti and 60% of Nitrogen returned very similar results:




And these are the results i got, when using the CalcZAF thin film correction:



And the up-down trend is also visible here, although not nearly as much. It seems that CalcZAF thin film correction did a pretty good job, but what could be the explanation for the other two: the actual measurement and the simulation?

So my question is: what could be the causes for registering much more N and less Ti at higher voltages then there actually is?

Pointman

[Probeman]

Hi Pointman,
How confidant are you that the thickness is actually 170 nm?

Try some different thickness values and see if you get more consistent results with a thicker or thinner layer specified.

One nice thing about STRATAGem is that you input multiple voltage k-ratios and it iterates to find the optimum composition *and* thickness.

You also might try different matrix correction models, for example one of the PAP models which are optimized for light elements.  In addition you should use empirical k-ratios for N Ka by Ti. 

Finally: are these k-ratios from EDS or WDS?  Even if you used WDS there is a very significant interference of Ti Lb on N Ka.  If you used EDS, forget about it.

One last thing:  with WDS there are significant peak shape changes for the N ka emission line due to the fact that it lies right on the Ti L absorption edge.

In short you could not have picked a more difficult system to attempt quantification on, even ignoring the thin film issues.  On this forum are discussions on performing quantification of light elements such as oxygen, carbon, nitrogen and boron.  You should read those posts carefully.
john

[Pointman]

Hi, Probeman

Tp start with it was EDS that the measurements were accquired with. Also, the thickness of the film should be around 170 nm. give or take, because the cross-section of the film was measured with SEM, and i dont have STRATAGem - only CASINO, CalcZAF and Excel for the graphs

Changing matrix correction models could be an interesting option. I will look into it.

[Probeman]

Hi, Probeman

Tp start with it was EDS that the measurements were accquired with. Also, the thickness of the film should be around 170 nm. give or take, because the cross-section of the film was measured with SEM, and i dont have STRATAGem - only CASINO, CalcZAF and Excel for the graphs

Changing matrix correction models could be an interesting option. I will look into it.

Hi Pointman,
By the way, what is your real name- mine is John.

I suspect you will have great difficulty with extracting reasonable net intensities from EDS spectra for TiN compositions due to the extreme overlap of Ti Lb on N Ka.  What I mean is the deconvolution of the peaks will be very problematic.

Frankly I wouldn't even attempt it with EDS (even with WDS and a bulk geometry it will be difficult enough), but give it a try I guess.  I suspect that the variation you are seeing is probably mostly (entirely?) an artifact of the deconvolution method you are using.  By the way, what EDS software are you using to extract the net intensities?
john

[Probeman]

By the way, have you tried measuring a "standard" TiN composition?  It should yield around 77% Ti and 23 % N (by wt).

That would be my first effort...

[Pointman]

My real name is Justinas

The EDS software was the one that came with the Buker SEM. it's Quantax.

By the way, could you give me any link or an explanation of how exactly does the CalcZAF thin film correction works? Im actually using your software and the info from our discussion for an assingment that im doing in my university right now, so it would usefull to understand it better

Also, using Ti L line for quantification actually made things look better in the simulations. Perhaps i will also try it on the actual SEM.

Justinas