Simulation and Optimisation of the JLabAES DC Photoinjector - PowerPoint PPT Presentation

Title: Simulation and Optimisation of the JLabAES DC Photoinjector


1
Simulation and Optimisation of the JLab/AES DC
Photo-injector

• Fay Hannon

2
Overview

• Background
• specification
• Simulation 135pC scenario
• By hand
• Multivariate optimisation
• Simulation 1nC
• Conclusions

3
Background

• Designed by Advanced Energy Systems
• Many iterations (7 cell /- 3rd harmonic)
• Now a demo of 1nC/135pC
• Not completely modelled at AES in present form
• Possible upgrade route

4
Injector Layout
Minimise the drift between gun and accelerating
cavities. Single cell cavities.
750MHz
5
Specification
For 7 cell design
6
Simulation ASTRA model

• A Space-charge TRacking Algorithm
• For rotationally symmetric components on-axis
  field maps give good approximation. Radial E and
  B fields generated from derivative.
• Can be fully 3D
• Field maps created
• using SUPERFISH/
• POISSON

7
Layout Field Maps
8
Simulation technique

• By hand aim for spec.
• Piecewise methodology, starting with gun and
  solenoid
• Parameter scans
• Start with 135pC (easier) case FEL laser
  properties for initial distribution
• Thermal emittance excluded rough modelling
• Difficult to meet the target energy

9
135pC Simulation
10
Results 135pC
JLab laser parameters assumed for initial
distribution
11
Multivariate Optimisation

• Ivan Bazarov _at_ Cornell
• 100mA injector with DC gun showed improved
  design.
• Evolutionary algorithm
• Can optimise a number of beam parameters
  simultaneously
• Start with 2 objectives minimise longitudinal
  and transverse emittance
• After a number of generations solutions converge
  on an optimisation front
• Pick solution that is best suited to the
  application

12
PISA

• Platform and programming language independent
  Interface for Search Algorithms
• Institute of Technology, Zurich
• Evaluate candidate solutions -gt
• select promising candidates -gt
• generate new candidates by variation
• - Text based interface for search algorithms
• - Consists of 2 parts variator and selector

13
Selector algorithm specific operations. Assignin
g fitness functions. Creating new mating pools
from the better solutions. Various selector
modules can be chosen eg. Simulated annealing,
genetic algorithms. Variator Problem
specific. Objectives, decision variables and
constraints. Crossing and mutation operators that
act on the candidates chosen by
selector. Variator includes a parallel
implementation of ASTRA. Variator and Selector
talk to each other via a state machine and
common text files
14
Initial Population
Decision variables Field strengths phases etc
Evaluate ASTRA parallel processing J-Lab HPC
5128 node cluster (24hr limit)
Constraints (spec) Objectives fitness evaluated
Solutions
Mating pool created from best solutions.
Crossing mutation New Population Generation
15
100mA Optimisation Problem
Decision variables (10) Laser spot size Solenoid
B 4 cavities max gradient and phase
Constraints (7) Bunch length lt 3mm KE gt7
MeV Longitudinal emittance lt 55 keV mm Transverse
size gt 1.0 mm Energy spread lt 70keV ltx.xgt,ltz dEgt
lt 0
Objectives (2) Minimise transverse
emittance Minimise longitudinal emittance
Selector Module Strength pareto evolutionary
algorithm 2 (SPEA2)
16
Problems noted

• Tail overtakes the head.

17
Decision Variables
18
Evolution
128 population 200 Generations 1k
Macro-particles Tracked to 5m
19
Results 135pC 100 generations

• Population size 128, 200 generations, 1k
  macro-particles for ASTRA.

20
Results 135pC 200 generations
Doesnt meet longitudinal emittance spec of 15keV
mm Other constraints met
21
Results 135pC XYrms 1 2.5mm
Improvement spec not met. Thermal
emittance/halo increases.
22
Try to re-arrange

• Conventionally 3rd harmonic cavities are placed
  last in an injector purely to make the
  longitudinal phase space linear

23
Best Solution Comparison
24
Best Solution Comparison
Absence of a dedicated buncher cavity makes
longitudinal spec. difficult to meet. First cell
and 3rd harmonic doing the bunching.
25
Energy trade off
7MeV was goal of 7 cell design. Energy can be
reduced for demo.
Lower energy can get close to longitudinal spec
26
Energy Trade Off
27
What if bunch length a variable?
128 population 200 Generations 1k
Macro-particles Tracked to 5m
Trms lt 10ps, Meet specification at 7MeV
28
1nC Simulation
29
1nC (7MeV constraint)
Unsuccessful in achieving both goals (lt45keV mm,
lt10um)
30
1nC Case
All possible solutions (meeting spec) are below
5.5MeV
31
Solution (4.6MeV)
32
Shorter laser pulse
Not typical optimal front because not enough
generations
33
Conclusions Future plans

• Specification can be met in both cases when the
  target energy is lowered to 5MeV
• Spec and 7MeV can be met if laser pulse is 10ps
  duration
• Multivariate optimisation is an excellent tool
  for injector design (operating points and
  dependencies)
• More cells greater flexibility
• Currently under construction at JLab
• Gun laser ready. Diagnostic beamline in design
• More detailed simulation (thermal emittance,
  realistic distributions, macro particles)