Elegant users forum

Michael,

I have a solenoid causing emittance growth . To solve the decoupling of transverse coordinates and to reverse the emittance growth I intend to use skew quadrupoles in the lattice. I use the optimization to get the optimized TILT and K values for quadrupoles. Do the TILT and K optimization work at the same time? Since the code is not converging to the result I asked for... I used "simplex" and others as well. I attached the file, would you please have a look at it ..Thanks

Serkan

Serkan,

One problem I see with your files is that you set "emit_y=5.84-6", i.e., you left out the "e" in 5.84e-6.

If I just run your system to look at the emittance growth, I see that it is due to chromatic effects in the solenoid and quads (if you set sigma_dp=0, you'll see no emittance growth). Hence, it can't be fixed entirely with rotation or adjustment of quads. I revised your input files (attached) to work better.

By the way, this is a fairly low energy beam, so you might want to use a program that includes space charge. ASTRA is a good choice as it interfaces easily to elegant.

--Michael

Michael,

Thanks for the fast reply and fixing it for me. You are right I have somewhat large energy spread (a positron source). I understand that it won't entirely fix the emittance growth but the thing is, at least I expect from this optimization process to let me approach to my target value a bit more, but I am not getting any closer. The solenoid doubles the initial emittance and whether I do the optimization with tilted quads or not, the emittance is not reduced even a bit, I need at least about 30-40 % improvement. Might that be because we should look for the uncoupled emittance (if there is such a way in elegant)? Or there is another optimization method that I am not aware of? Thank you for your time.


PS: considered very low current so no significant effects of space charge is expected.

Serkan,

The .ele file I attached to my last message will optimize to less than 2x the target values. To go beyond that, you probably need more knobs. By adding the distances between elements as variables, I was able to get the growth down to 1.6x the target. (See attached.) Adding another quad or two might give better results. You can also try increasing n_restarts; sometimes this helps to find a better solution if you are willing to wait.

You are right that there may be an issue with computing the x and y emittances, since the problem is coupled. You can try adding to the simulation a minimization of the off-diagonal sigma matrix terms, i.e., s13, s14, s23, and s24. Again, to make much progress you may need more knobs (e.g., more quads). However, it is possible that you'll get into a problem of diminishing returns as you add more knobs, due to extra chromatic effects.

--Michael