Question about simulating asymmetrical biradical spectra

[jchen3]

Hi All,

I have a specific question about simulation of room temperature isotropic EPR of a biradical system with two spins strongly exchange coupled.
The problem is, the spectra will be quite asymmetrical compared to single radical spectra.
An example is shown below where I found online.
https://www.google.com/search?newwindow ... mK9nhJqfM:
Compared to single radical, the biradical lost central symmetry.
What I've tried, is to increase the line width, and even adding modulation amplitude to the system, then the spectra looked more asymmetrical. The problem is, once I did this, the hyperfine structures disappear (rounded to one peak) due to that, and there is no way to balance the hyperfine structures and asymmetrical looking.
So I just wonder if there is other option available to address this problem?
Thanks in advance.

Best,
Ju

[Stefan Stoll]

You might be seeing some motional effects. In order to simulate them, you will need the principal values of all hyperfine tensors, plus a rotational correlation time (which might be anisotropic if the biradical is an elongated molecule like the one in your link.)

[jchen3]

You might be seeing some motional effects. In order to simulate them, you will need the principal values of all hyperfine tensors, plus a rotational correlation time (which might be anisotropic if the biradical is an elongated molecule like the one in your link.)

Hi Stefan,

Thanks for your help! Besides, I have some new questions regarding your reply.
1. I quickly played around with "rotational correlation time" function with Sys.tcorr, and I found "pepper" is not working with that, so does that mean only "chili" will work on this specific issue for an exchange coupled system?
2. When I switched with "chili", the computing time increased significantly, and crashed with error in "liouvhamiltonian", and "out of memory". How can I avoid this with correct hyperfine setting?
3. I found this previously in the forum, that when my hyperfine setting is, for example [x x x 0 0 0;0 0 0 y y y], the spectrum is a flat line, while I could make it work by defining A tensor [x x' x' 0 0 0;0 0 0 y y' y'], I just wonder the meaning of this asymmetrical A tensor structure of an isotropic EPR?
An example of my spin system is like this:

Code: Select all

        Sys.S=[0.5000 0.5000]
        Sys.g=[2.0040 2.0041]
        Sys.A=[18,0 ; 18,0 ; 14.5,0 ; 14.5,0]
     Sys.Nucs='N,N,N,N'
       Sys.lw=0.1000
        Sys.J=-7790000
   

Thank you again!

[katarkon]

You may try to reduce the basis (Opt.LLKM) to decrease the memory consumption. Your spin system is too large for default basis. You may check it with Opt.Verbosity=1.
Also You may write custom simulation function using resfields and arbitrary line broadening. Or use pepper with separate output, postconvolution of broadened lines and acumulation with the same efffect.
At last, for large J values (J>>A) the spectrum may be simulated as monoradical with halved HFC constants via garlic function. Modified exchangefunction (viewtopic.php?f=8&t=465) is fine too.

[jchen3]

You may try to reduce the basis (Opt.LLKM) to decrease the memory consumption. Your spin system is too large for default basis. You may check it with Opt.Verbosity=1.
Also You may write custom simulation function using resfields and arbitrary line broadening. Or use pepper with separate output, postconvolution of broadened lines and acumulation with the same efffect.
At last, for large J values (J>>A) the spectrum may be simulated as monoradical with halved HFC constants via garlic function. Modified exchangefunction (viewtopic.php?f=8&t=465) is fine too.

Hi Katarkon, it is very helpful. Indeed I switched to single radical simulation with garlic, and the computing time is just much better, but not with a larger set of hyperfine. As you mentioned, whenever I try to induce something more, the memory crashed out. I will definitely give it a try on tuning basis. Thank you.
Best,
Ju

[Stefan Stoll]

The theory underlying chili is unfortunately very memory-demanding, since it runs in Liouville space instead of Hilbert space. Each additional nitrogen nucleus increases matrix sizes by a factor of 3^2 = 9. For two electrons and four 14N, the matrices are 100k x 100k.

You might also want to check out the option Opt.PostConvNuclei in chili.

To save time, also use '14N' instead of 'N' - The 15N spectrum is negligible since 15N has <0.5% natural abundance.

[jchen3]

The theory underlying chili is unfortunately very memory-demanding, since it runs in Liouville space instead of Hilbert space. Each additional nitrogen nucleus increases matrix sizes by a factor of 3^2 = 9. For two electrons and four 14N, the matrices are 100k x 100k.

You might also want to check out the option Opt.PostConvNuclei in chili.

To save time, also use '14N' instead of 'N' - The 15N spectrum is negligible since 15N has <0.5% natural abundance.

Thank you Stefan, I will definitely take that into consideration. I absolutely overestimated the computing power of a single PC My advisor told me how he lower the size of matrix 28 years back, and I just learnt that I still need to do something like that!