Slide 1
L. Wang
Work with T.O. Raubenheimer, J. WuLCLS-II Accelerator Physics
meeting Jan 15th, 2013
Intensive MOGA Optimizations of LCLSII configurationsLCLS-II
Accel. Phys. , L. Wang, SLACMany thanks to J Wu for useful
discussion and help on the Litrack set-up, thanks F. Zhou for his
help on IMPACT simulation data, M. Woodley and Y. Nosochkov for
providing the R56 data.1ProblemLayout
Gun to L0, using IMPACT (thanks to F. Zhou)Particle Tracking
start from L0 to beginning of the Undulator using LiTrackNominal
R56 @BC1= 46 mm (thanks M. Woodley, Y. Nosochkov) Nominal R56 @BC2=
29 mmNominal R56 @Bypass Dogleg= -0.115 mmNominal R56 @HBEND=
-0.383/0.479mm (HXR/SXR)Wake field is included
BC24.5 GeVBSY10/13.5 GeVTCAV3 BC1250/335 MeVL1SwirescannerL1X4
wire-scannersL2-linacL3-linacDL1135 MeVL0gunLCLS-II Accel. Phys. ,
L. Wang, SLACAll surface material are SS (except
Undulator)WakeLCLS-II Accel. Phys. , L. Wang, SLACRegionLength
(m)radius (m)Pipe
shapewakecommentsL014.9S-bandL128.8S-bandL2313.8S-bandL3532.7S-bandLX0.6X-bandBypass
extract550.01circularRW,SSBypass
extract9770.025circularRW,SSBSY1920.01circularRW,SSLTU5100.02circularRW,SSHXR
undulator1150.005flatRW, AlMOGA(Multi-Objective Genetic
Algorithm)Variables:Phase and Voltage of L1,LX,L2,L3R56@BC1R56@BC2
Objectives & constrainsEnergy spreadJitters(RF Voltage and
phase jitter, Charge jitter, Laser timing jitter) Energy at
BC1(250MeV/335MeV)Energy at BC2(4.5GeV) Energy at the beginning of
undulator (13.5GeV/10GeV)Peak current at the beginning of undulator
(3kA/4kA)second order curvature
Optimization MethodLCLS-II Accel. Phys. , L. Wang,
SLAC4Objectives
The energy spread is normalized by 0.03%LCLS-II Accel. Phys. ,
L. Wang, SLACFor instance (I/I)0=12%, (E/E)0=0.1%, ()0=100fs
Various JittersSimplified Jitter model in MOGAL1 Phase: [-30,
-10]LX phase: [-162, -156]L2 phase: [-36, -31]L3 phase : [-10,0]
(based on study at 250pC)LCLS-II Accel. Phys. , L. Wang, SLACThe
range of RF phaseA larger range may be need for more
studies!Converge slow; larger RF power6LCLS-II Accel. Phys. , L.
Wang, SLACBenchmarked with design report
Only About 10 variables, but with many constrainsThe bunch
profile (also the peak current) is not easy to be satisfied. Many
efforts have been taken to make the optimization converge fast and
well.Wake field and residual R56 after L3Wake field couples with
bunch profile and add energy chirp in the phase space, therefore it
can change the bunch profile/peak current ;the non-zero R56 also
rotate the phase space, which causes the variation of the bunch
profile/peak current after L3.CPULCLS-II Accel. Phys. , L. Wang,
SLACComplexity250pC HXR250pC SXR150pC HXR40pC HXR20pC HXR20pC
SXRLCLS-II Accel. Phys. , L. Wang, SLACresultsMOGA optimization of
250pC HXR(BC1@335MeV)LCLS-II Accel. Phys. , L. Wang, SLAC
Converge after 100 generations
Objectives @250pC HXRThe slope in z-profile shows the dI/dz near
bunch center. ( thanks J. Wu) It doesn't mean anything for Gaussian
shape bunch
20pC150pCLCLS-II Accel. Phys. , L. Wang, SLACHTTotal
JittersParameters @250pC HXR---Solutions
Small jitter zoneRed zoneSmall energy spread zoneGreen
ZoneLCLS-II Accel. Phys. , L. Wang, SLACLarge range with LX
phase