Understanding Low CSR Emittance Growth

[Zamank]

Hi Michael,

I have been working on this lattice ffor a while now. Its quite simple: Start with a gaussian distribution. Accelerate it through x-band cavities (giving it a energy chirp). Compress with a 4-bend chicane.

I start with a 1um emittance and end with ~2um emittance, though my analytical calculations say that my emittance growth due to CSR should be much much larger. Any ideas of could be the issue (with my lte or my understanding)?

I thought that because the chicane is distorting the bunch distribution (a high peak ~1kA with a super long tail) that the CSR could not resolve the bunch distribution of the peak. So, I increased the bins, did not help.

Any help would be greatly appreciated!

[Zamank]

Any ideas???

[michael_borland]

I played with your files, changing the number of bins and the number of particles by an order of magnitude. I also turned off smoothing. I didn't get significant changes in the emittance growth, so I consider this to be a reliable result. I'm not sure why your analytical estimates give a different answer.

--Michael

[Zamank]

Hmmm, interesting...thanks for your help Michael.

I was wondering does turning off smoothing increase accuracy of results?

Also, I have a feeling that the long tailed distribution this *.lte file creates is the cause of the low CSR. I am looking into it and working it into my analytical calculations.

[michael_borland]

Turning off smoothing makes results more accurate provided there are sufficient particles to control noise. The smoothing filter kills the high frequencies. It is better to use the HIGH_FREQUENCY_CUTOFF parameters, which gives you better control.

--Michael

[Zamank]

Hi Michael,

So, Ive been looking at this issue for a little while and think I may have some idea:

The strong compression causes a long tail towards the end of the compression cycle (Bends 3 & 4). So, when the user defines the BINS for the CSRSBEND element Elegant calculates the BinSize according to the entire bunch length; for the last two bends the bunch length is very long. The long tails of the last two bends requires an even smaller BinSize (i.e. larger BINS number) to resolve the peaks of the long tail distribution.

BUT, I understand that increasing the BINS number too much will create a messy high frequency linear density and even more messy linear density derivative in the CSR calculation.

So, to test things out I ran a scan of BINS for the 4th bend in my system (attached are the results) with CSR, ISR turned off everywhere but CSR ONLY turned on in the 4th bend. I am a little confused, I was hoping that there would be some sort of taper or convergence as I increased the BINS but as you can see the Sdelta continues to increase even after 2e6 BINS.

Do you have any opinions or ideas on helping me understand this a bit better? Or any corrections to my thought process thus far? Any help would be greatly appreciated!

Thanks in advanced!

[Zamank]

Any thoughts Michael???

[michael_borland]

Sorry for the delayed reply. It's a busy time here.

You need to be sure to increase the number of particles in proportion to the number of bins, otherwise the noise increases and that alone could cause the energy spread to increase.

--Michael

[Zamank]

Thanks for the reply Michael! Greatly appreciated...

...follow up question: Is there anyway I could cut the tail of the particle distribution, say at the entrance of the 3rd bend??

Thanks again

[michael_borland]

Yes, you should be able to do it with the CLEAN element.

--Michael