Consistency between EasySpin and Bruker AnisoFit?

[tj@bc]

Hello all--I was wondering if anyone had to deal with the lack of consistency in the results obtained via EasySpin vs. AnisoFit from Bruker. Even though both simulations look excellent, the numbers do not really match up (after the unit conversions, unless my math is wrong!)--is this not an issue, or is there an overarching principle that I may be missing?

Screencaps of an example are attached--as always, many thanks for your help and patience!

Screen Shot 2022-01-13 at 6.08.05 PM.png (100.13 KiB) Viewed 8030 times

Screen Shot 2022-01-13 at 6.08.19 PM.png (478.72 KiB) Viewed 8030 times

[Matt Krzyaniak]

If both of these sets of results are from fitting, I'm not surprised if they produce different numbers. Unless both routines are using the same minimization algorithms, initial guesses and processing and target, they are likely to end up in different local minima. The fitting of the EPR spectrum of nitroxide, at a single field is actually a bit under-determined, meaning that in the parameter space there are mutliple sets of numbers that provide a good fit. You could try a psuedo Monte-Carlo approach to shed some light on this, start both fit routines from different, random start points, and see where minimize. If each random start ends in the same spot your in a global minimum, but if they are end up in different spot, you have a range of local minima and the parameter space might be too flat, or you stopping criteria might be too high.

In addition it appears that in AnisoFit, you have an anisotropy linewitdh while you are using an isotropic one in EasySpin.

A better test of consistency would be to perform a simulation with the set of numbers in both routines, the spectra should match.

[tj@bc]

Thanks very much for this detailed explanation Matt--I appreciate it very much! It makes a lot of sense too...

If you could help me out with one point though: you mention the inconsistency in the linewidth parameter, that I use an isotropic linewidth on EasySpin. From what I understood (perhaps incorrectly!) via the EasySpin Academy videos was that the linewidth can be varied using lwpp, or HStrain, or a g- and A-strain combo (and that these methods should not be mixed)--is this correct? Based on your response, are you suggesting that I use HStrain or g- A-strains for the linewidth variations? Please let me know if there are more options in this regard!

The reason I stuck with lwpp was because it gave me the lowest RMSD--perhaps this is not always the best rule to follow when picking the "most suitable" simulation?

Again--thanks very much; I really appreciate your input!

[Stefan Stoll]

It might be helpful if you could post your spectrum plus the two fits.

[tj@bc]

Please see the attached spectra--and thanks!

Screen Shot 2022-01-13 at 6.05.29 PM.png (785.74 KiB) Viewed 8013 times

Spin-Fit Result.png (1.47 MiB) Viewed 8013 times

[Stefan Stoll]

It looks these are two different experimental spectra, at least judging from the different noise levels. In that case, one can't compare the results directly.

[tj@bc]

Oh dear--you're right of course: I did flip between two identical spectra on EasySpin, and clearly chose the wrong result. #embarrassed!

I will check with the proper spectrum, and compare the results again--thank you professor Stoll!

Meanwhile, could you please confirm that my understanding of the varying of the linewidth parameter (two posts above) is correct?

Thanks again!!

[tj@bc]

Updated EasySpin result is attached: the results are still a bit off, but I've decided to be at peace with it, lol--thanks again!!