Settings of new 'poisson' method in 'ion-effects'

Moderators: cyao, michael_borland

Post Reply
Siwei_Wang
Posts: 21
Joined: 27 Jun 2017, 07:28

Settings of new 'poisson' method in 'ion-effects'

Post by Siwei_Wang » 05 Jun 2023, 10:58

Hi there,

Recently I noticed that the newest elegant 2023.2 has the new ‘poisson’ method in ‘ion-effects’. I did several tests with different settings of ion_span and ion_poisson_bins. When I set the ion_span to 10 times the beam sigma and ion_poisson_bins to 128, the simulation goes much faster than ‘bilorentzian’, and the growth rate seems to be between ‘gaussian’ and ‘bilorentzian’ (closer to ‘gaussian’ maybe). When I tried to increase the ion_poisson_bins to higher numbers, say 512, the execution time becomes longer. Does my setting sound reasonable, and what would be the optimum setting for the ‘poisson’ method?

Also, I noticed on elegant manual of ‘ion-effects’, it says ‘The fitting-based methods, i.e., bigaussian, bilorentzian, trigaussian, and trilorentzian, typically show instability when it is not expected and may well have noise challenges that have not been resolved.’ Could it be the reason that I see larger instability when using ‘bilorentzian’ method than ‘poisson’ or ‘gaussian’?

Many thanks,
Siwei

jcalvey
Posts: 19
Joined: 08 Jan 2021, 12:20

Re: Settings of new 'poisson' method in 'ion-effects'

Post by jcalvey » 05 Jun 2023, 17:37

Hi Siwei,

For simulations of the APS gas injection studies with the Poisson method, I've been using 128x1024 bins (since higher resolution is desirable in the vertical plane). But this is probably overkill; simulations with 64x512 bins seem to give the same answer. I'm still playing around with this parameter. Ultimately it's best to test this for your own case; try increasing and reducing the number of bins to see if your answer changes.

The fitting based methods are functional, but have many numerical parameters, and can be unstable if the parameters aren't set right. So that could explain your result. The Poisson method is easier to use and, as you noticed, faster. I've also found that it gives results that are more consistent with the gas injection measurements than either the Gaussian or fitting based methods.

Hope this helps,
-Joe

Siwei_Wang
Posts: 21
Joined: 27 Jun 2017, 07:28

Re: Settings of new 'poisson' method in 'ion-effects'

Post by Siwei_Wang » 06 Jun 2023, 02:30

Hi Joe,

Many thanks for your reply. I'll try more cases with different number of bins. It's a good thing that 'poisson' method is closer to measurement result, it will save more execution time for ion simulation for future studies.

Regards,
Siwei

Siwei_Wang
Posts: 21
Joined: 27 Jun 2017, 07:28

Re: Settings of new 'poisson' method in 'ion-effects'

Post by Siwei_Wang » 15 Jun 2023, 04:48

Dear Joe,

I did some benchmarking of different 'poisson' setup parameters with my previous result using 'bilorentzian'.
It seems with different filling patterns, sometimes the ion effects from 'poisson' method is more stable and sometimes 'bilorentzian' is more stable. But I observed some interesting phenomenon when I tried to benchmark a hybrid filling pattern (1 train with 100 half gap). When using 'poisson' method in this case, there is unexpected emittance growth in the horizontal direction:
horampoiss1.jpg
horemitpoiss1.jpg
Did you meet similar phenomenon with the 'poisson' method? Basically my emittance is about 120pm for horizontal direction and 8pm for vertical, ideally there would only be instability in vertical direction, which is also what I observed with 'bilorentzian' method before. The 'poisson' method setting I used is 128 poisson bin number for both horizontal and vertical, with ion_span 10 times the nominal beam size.

Many thanks,
Siwei

jcalvey
Posts: 19
Joined: 08 Jan 2021, 12:20

Re: Settings of new 'poisson' method in 'ion-effects'

Post by jcalvey » 23 Jun 2023, 15:27

Hi Siwei,

I have not seen a horizontal instability in my Poisson simulations. My suggestion is to increase the number of ion macroparticles generated per bunch. If you have only a few high charge macroparticles, that could cause a numerical instability.

-Joe

Siwei_Wang
Posts: 21
Joined: 27 Jun 2017, 07:28

Re: Settings of new 'poisson' method in 'ion-effects'

Post by Siwei_Wang » 28 Jun 2023, 07:56

Hi Joe,

Thanks for your reply. My previous result was already using 100 macro_particle, and 128 poisson bins for both horizontal and vertical. That filling pattern was a hybrid filling pattern with 1 train, 100 half bucket gap. Following your advice, I tried another hybrid filling pattern case ( 2 train, 100 half bucket gap), with three different settings: 1) 128 poisson bins, 100 macro particle; 2) 512 poisson bins, 100 macro particle; 3) 128 poisson bins, 1000 macro particle. It seems the three different settings gave me similar results:
horemit.jpg
I still see the horizontal emittance growth in these cases. So could it be that when I see the horizontal emittance growth, it could be viewed as a criterion that this case is unstable? I only observed this phenomenon in some hybrid filling patterns. For standard filling patterns (several trains with gaps), I didn't observe the horizontal emittance growth, even if for some cases there were instability in the vertical direction. So hopefully the horizontal emittance growth could be a way to determine the case is unstable. How do you think about this?

Many thanks,
Siwei

jcalvey
Posts: 19
Joined: 08 Jan 2021, 12:20

Re: Settings of new 'poisson' method in 'ion-effects'

Post by jcalvey » 05 Jul 2023, 17:07

Siwei,

Can you post or email me the input files you are using?

-Joe

Post Reply