FCC-ee optics tuning WG

→ Europe/Zurich
6/2-004 (CERN)

6/2-004

CERN

40
Show room on map
Zoom Meeting ID
65516078996
Host
Rogelio Tomas Garcia
Useful links
Join via phone
Zoom URL

Dynamic aperture optimization using ML/AI at SuperKEKB, Taichiro Koga

A New optimization method is developed to define the sextupolar strengths in SuperKEKB. Previously simplex was used. This is a complex problem as SuperKEKB has about 50 sextupole pairs per ring.

Resulting settings with the new settings have larger Momentum Acceptance (measured as the area in the DA vs dp/p plot). Strickingly beta*y shows a bit larger changes versus momentum for the new optimal settings.

Optimizations including collimators are also performed successfully. Beam-beam interactions are included.

CMA-ES:  Covariance Matrix Adaption - Evolution Strategy

Some imperfections were assumed as small sextupoles offsets. Actually the optimization does not target exactly area, but Tousheck lifetime (which is derived from the DA versus dp/p). 

Claude assisted in suggesting the algorithm and in job submission management. Bernardo Abreau also considered CMA-ES algorithm for having a good performance. For FCC-ee Marc uses multiple objectives algorithms. 

Combining CMA-ES with surrogate models improved results.

 
Beta correction by using quadrupoles adjacent to sextupoles, Takashi Asami, Takashi Asami 
Goal is to correct the feed-down from sextupoles with quadrupoles. Assigning 100-200 um misalignments generates 10-20% beta-beating that can be corrected very well.
Measured beta-beating can be modeled with some quadrupolar components next to sextupoles. Hence this approach is demonstrated. Some further considerations for application in the machine are needed. In next operation there is a plan to use dedicated quadrupoles for these corrections.
Oide-san recalls that there were related studies in the past where orbit bumps were used to correct the beta-beating and coupling (with the vertical offset).
 
 
Estimation of quadrupole errors in the SXABE section Takashi Asami, Takashi Asam
 
SXABE sextupoles have a fixed setting to fulfill extraction requirements. They are not used in MA optimization.
Mael's identified errors in SXABE regions agree with results also found with LOCO-like algorithm by the Japanese colleagues. Hence both techniques are in agreement. Similar results are observed in other optics with lower beta*.  For this section in particular the quadrupoles within the -I transformation where not used in correction on purpose, which explains the uncorrected error in phase advance from 180 degrees. in next commissioning these errors will be addressed.
There will be machine studies. Detailed schedule to be distributed later.
 
 
Non-linear chromaticity re-evaluated with new abort section model , Mael Le Garrec
 
Further details on the SXABE errors are presented. Required quadupole corrections are maximum 0.4%. Using these quadrupolar errors helps to improve drastically the agreement of resonance driving terms (amplitude and phase), and amplitude detuning. For non-linear chromaticity the agreement can be largely improved if using all quardupolar errors found around ring (not only the SXABE region).
 
At beta*y=1mm some sextupolar RDTs show still some disagrement, probably form a new source in the IR.Maybe a b3 error of QC2LE.
 
Detailed measurements and corrections of many of these quantities (Q'', Q''', amplitude detuning, RDTs) should be checked in commissioning. Combination of TbT, COD and off-momentum should suffice.
Mael will visit SuperKEKB in February, when the optics squeeze will be commissioned down to 0.9mm, when new sextupoles settings will also be tested, followed by final optics commissioning.
It would be ideal to measure and correct also chromatic coupling (rotatable sextupoles only in LER). As seen in few meetings ago in Kevin's presentation, skew quadrupoles can correct chromatic coupling. It would be ideal to try these corrections in the machine as emittance growth is seen to arise from chromatic coupling.
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
     
 
 
 
 
 
 
 
 
 
 
 
There are minutes attached to this event. Show them.
    • 09:00 → 09:05
      News and introduction 5m
      Speakers: Jacqueline Keintzel (CERN), Rogelio Tomas Garcia (CERN)
    • 09:05 → 09:25
      Dynamic aperture optimization using ML/AI at SuperKEKB 20m
      Speaker: Taichiro Koga
    • 09:25 → 09:45
      Beta correction by using quadrupoles adjacent to sextupoles 20m
      Speaker: Takashi Asami
    • 09:45 → 10:05
      Estimation of quadrupole errors in the SXABE section 20m
      Speaker: Takashi Asami
    • 10:05 → 10:25
      Non-linear chromaticity re-evaluated with new abort section model 20m
      Speaker: Mael Le Garrec (Laboratoire d'Annecy de Physique des Particules (CNRS IN2P3-LAPP))