depth-graded MLS fitting

[Dmitry Kuznetsov]

Hello, David.

I'm a Ph.D. student at FOM DIFFER / Twente University (working together with Robbert).
First, thank you very much for the new version.

I look on MLS more from deposition point of view, and recently, giving to the magnetron targets usual amount of
pre-heating targets layers, I got nicely broadened Bragg peaks on GIXRR graph:
https://drive.google.com/file/d/0B6T92xyewYcpNXFEdU00QTVDX1k/edit?usp=sharing
(scanned for sample with d_spacing 3.70 nm, but is the same shape for all samples in this coating run)

and broadened Soft x-ray peak:
https://drive.google.com/file/d/0B6T92xyewYcpT3g1T2ZVYnNITTA/edit?usp=sharing
(raw data in .txt from ALS)
System is 175-period LaN/B, aiming to gamma=0.40, "effective" period (from GIXRR) around 3.52 nm

I've chosen LaN layer depth_grading according to QCM info as a ln-function (thickness of LaN rises according to
the function total by 1.4 A from 1st to 175-th layer of MLS) in IMD, and then tried to add some fitting parameters.
Even if I add one fitting parameter (random one), a sentence below white window with parameters appears:
"Too many fitting parameters defined (1) - no degrees of freedom available".
This error appeared only once, and noe I'm not able to reproduce it.
If the depth grading is tried to be off, then an error appears:
https://drive.google.com/file/d/0B6T92xyewYcpckYtanlzWnF5RlU/edit?usp=sharing
Then IMD shuts itself down.
This error is reproducible on my PC.

It would be nice if I'm doing smth. wrong here and the error appears for that reason)

Generally, I have a question about fitting of the non-intentionally depth graded MLS - how the fitting parameters for this case
should be properly introduced into IMD, has it been already tested with or without success and so on.
From my opinion, this is significant to take into account, because modeling in IMD of ldeal LaN/B MLS and with depth_graded
LaN thickness total by 1.0 A (assuming ln-function)  shows drop about 5-6% of reflectivity at 6.X nm wavelength of interest.

Thank you for the time spent!
Dmitry Kuznetsov.

[David Windt]

Dear Dmitry,

You have apparently uncovered a bug in IMD: it seems to occur when you (a) configure a structure to have depth grading, (b) add fit parameters, and then (c) remove the depth grading. I will correct this problem for the next release. In the meantime, you can avoid this error by removing all fit parameters BEFORE you remove depth grading. 

For additional help with fitting, please consult chapter 8 of the documentation (IMD.pdf).

Best regards,

David

[Dmitry Kuznetsov]

Thank you for the reply, David!

Nice to be helpful as a bug-investigator )    I will update you with any new observations if there will be any.



Best wishes,
Dmitry Kuznetsov.

[Dmitry Kuznetsov]

Hi, David!

Two more errors:   1)  If there is a depth-graded MLS with some number of periods (175 in my case), and after adding all the fit parameters one will realize that experimental data is for 50-periods MLS (like me in my case) )  then an attempt to set the proper amount of periods yields the same error, the IMD shuts itself down, after you've agreed with the error message:
https://drive.google.com/file/d/0B6T92xyewYcpejh2M0U2eTg0ems/edit?usp=sharing

2)  If for some reason one will type in random period thickness error delta(z) larger that one of the layers, it will be reasonably an error about negative layer thickness, and the calculation aborted - fully agree. But thing is that IMD keeps "remembering" the error and even after you've agreed with the error, and, for instance, just highlight one of the lines with MLS parameters by mouse click, keeps showing the error that calculations will be aborted, etc.:
https://drive.google.com/file/d/0B6T92xyewYcpbkNvcU16eHpMM3c/edit?usp=sharing
      Resembling story if too large delta (z) is included into fitting parameters - but there IMD gets stucked because the wrong (larger than one of the layers) random error is being set and attributed as MLS actual parameter - so, program just shut downs itself after you agree with the error. Trying to set the delta (z) to zero or reloadind the IMD and the file with MLS yield nothing.

Best wishes with bug-elimination!

Dmitry Kuznetsov.

[David Windt]

Hi Dmitry,

I am unable to reproduce the first error you reported. Also, the screenshot you provided suggests to me that the problem may be somehow related to an independent variable. Can you please provide a .imd file that demonstrates this error?

I'll do my best to deal with the negative thickness error reporting problem. Meanwhile, my advice is to avoid generating negative thicknesses. In the meantime, if you do encounter a negative thickness error, try deleting the independent variable/coupled parameter/fit parameter that is causing the problem.

Best regards,

David

[Dmitry Kuznetsov]

Hello, David,

To get the first error for some reason three or more depth_grading fitting parameters must be added (like
top/bottom thicknesses of B and La layers, in my case).
This is the freshly registered same error:
https://drive.google.com/file/d/0B6T92xyewYcpemNzY3QzMm5BLUE/edit?usp=sharing
and the .imd file corresponding to enjoy it =) :
https://drive.google.com/file/d/0B6T92xyewYcpUDNfcGVHaFZBMFk/edit?usp=sharing

Best wishes,
Dmitry S. Kuznetsov

[David Windt]

Thanks, Dmitry. I'll fix this bug for the next release.

[Dmitry Kuznetsov]

Hello, David!

One more observation - if I make up a MLS (for details screenshot on the link),
https://drive.google.com/file/d/0B6T92xyewYcpTGdzZ2dkR2ROUWs/edit?usp=sharing
 "designed to work" around ~67.0 A wavelength,
but in the experimental data the wavelength are given in nm (6.7; 6,8;..), and I import this data, lambda values will be assumed as A (so, on
imported data there will be 6.7 A and so on).
And in this case IMD writes smth. that "there are no degrees of freedom - too much fitting parameters" (actually, it can be only one fitting parameter chosen - but still the same behavior).

Not an issue, maybe a pop-up reminding message makes sense for beginners, and the question from software: In which units are lambda in exper. data?   nm / A   (to choose one).
That is a few-seconds-work to multiply the values by ten in a table and import then into txt, but for chasing maximum user-friendly behavior and automation ... why not    =)