1. FASTION
L. Mether, G. Rumolo
ABP-CWG meeting

2. Outline Introduction Physics model
Code implementation, documentation, license
Typical usage
Performance
Future plans

3. FASTION
Code for modeling the fast beam-ion instability (FBII) in electron machines
Developed at CERN, by G. Rumolo, based on HEADTAIL for e-cloud
Used for studying FBII in linear CLIC structures
Recent developments to apply to synchrotrons

4. Physics model Linear transport of bunch train through machine lattice
Linac: 4D transport, energy incremented
Synchrotron: 6D with linear bucket, radiation damping (S. White)

5. Physics model Linear transport of bunch train through machine lattice
Linac: 4D transport, energy incremented
Synchrotron: 6D with linear bucket, radiation damping (S. White)
The machine lattice is represented by interaction points (IP)
In each IP, interaction between ions and bunch train is modeled

6. Ion-beam interaction Residual gas ionization
Represent ions generated over s distance between adjacent IP’s
Generated through scattering ionization, or field ionization, if bunch e-field exceeds threshold value
Multiple gas species with different pressures and ionization cross-sections supported
Give velocity kick
Generate ions
Introduce bunch

7. Ion-beam interaction E-field calculation
Ion and beam fields calculated separately using the same grid
Averaged Green’s function PIC solver, open boundary, FFT convolution
Accurate for rectangular (non-square) grid cells
Dual-grid solver: fine grid around beam, coarse grid for ions over larger area
Velocity kicks applied according to e-field of oppositely charged particles
Calculate kick
Generate ions
Introduce bunch

8. Ion-beam interaction E-field calculation
Ion and beam fields calculated separately using the same grid
Averaged Green’s function PIC solver, open boundary, FFT convolution
Accurate for rectangular (non-square) grid cells
Dual-grid solver: fine grid around beam, coarse grid for ions over larger area
Velocity kicks applied according to e-field of oppositely charged particles
Calculate kick
Generate ions
Introduce bunch
Transport bunch

9. Ion-beam interaction E-field calculation
Ion and beam fields calculated separately using the same grid
Averaged Green’s function PIC solver, open boundary, FFT convolution
Accurate for rectangular (non-square) grid cells
Dual-grid solver: fine grid around beam, coarse grid for ions over larger area
Velocity kicks applied according to e-field of oppositely charged particles
Calculate kick
Ions drift
Generate ions
Introduce bunch
Transport bunch

10. Ion-beam interaction E-field calculation
Ion and beam fields calculated separately using the same grid
Averaged Green’s function PIC solver, open boundary, FFT convolution
Accurate for rectangular (non-square) grid cells
Dual-grid solver: fine grid around beam, coarse grid for ions over larger area
Velocity kicks applied according to e-field of oppositely charged particles
Calculate kick
Ions drift
Generate ions
Introduce bunch
Transport bunch

11. Code implementation
Written in C (developed from HEADTAIL), with a few routines in F77
Tested only on Linux, compiled with gcc compiler
The PIC solver uses FFTW
Repository:
Input: config file with beam and vacuum parameters, twiss file for lattice
Documentation: minimal overview of input & output
License: GPL3
Users: CERN (1), light sources: CESR-TA, APS, ESRF, INDUS

12. FASTION with PyECLOUD + PyHEADTAIL
FASTION is based on HEADTAIL for e-cloud
Recently HEADTAIL has been redesigned  PyHEADTAIL
Also for electron cloud build-up: ECLOUD  PyECLOUD
Decision to incorporate FASTION functionality into PyECLOUD – PyHEADTAIL
Benefit from more modern and flexible design
Shared workload of maintenance and development
Access to features already implemented in PyECLOUD/PyHEADTAIL:
PIC solvers with boundary for complex beam chamber profiles (PyPIC)
Ion self space charge
Magnetic fields
Bunch slices
Chromaticity
Transverse feedback
coupled for e-cloud instability studies

13. FASTION with PyECLOUD + PyHEADTAIL
FASTION simulations with PyECLOUD-PyHEADTAIL have been implemented and tested
Suitable for simulations in synchrotrons
Currently only single residual gas species with scattering ionization
PyECLOUD
Calculate kick
Ions drift
Generate ions
PyHEADTAIL
PyHEADTAIL
Introduce bunch
Transport bunch

14. Benchmark study for CLIC damping ring
Bunch train initialized identically in FASTION and PyEC-PyHT
Machine lattice divided in 677 interaction points ~ 60 cm long
Residual gas: water, A = 18
Pressure 20 nTorr
Track over 1 turn
Bunch train centroids after 1 turn
Instability in vertical plane

15. Benchmark study for CLIC damping ring
Bunch train initialized identically in FASTION and PyEC-PyHT
Machine lattice divided in 677 interaction points ~ 60 cm long
Residual gas: water, A = 18
Pressure 20 nTorr
Track over 1 turn
Centroid of last bunch along turn
Good agreement between FASTION and PyECLOUD-PyHEADTAIL

16. Typical use case
FASTION simulations are used to estimate necessary vacuum and possibly other requirements to avoid FBII
Scans over residual gas composition and pressure, initial seeds
Relevant beam parameters: intensity, beam size, bunch spacing, train length
Several hundred to a few thousand simulations for a full study
Use at CERN
In the past applied to linear CLIC structures: main linac, main transfer line
Currently: CLIC damping ring (DR)
Future studies: CLIC structures after re-baselining, FCC-ee
CLIC Main Linac

17. Computational requirements & performance
Typical parameters for CLIC studies (convergence studies on-going)
312 bunches of 104 MP’s
500 ion MP’s/bunch  on average about 7.5 × 104 ion MP’s
PIC grid size: 250×250
For LINAC high resolution is necessary  dual grid + larger grid size
For DR ideally simulate at least a damping time: τ = 2 ms ~ 1400 turns
lsbatch suitable for LINAC simulations, but not really for DR
Computation time
LINAC
Damping ring
IP’s 1
2000
260
1400 * 260
FASTION
5 s
5 h
20 min
20 days
PyEC-PyHT
7 s -
30 min
30 days

18. Parallelization with PyPARIS
Parallelization layer for PyECLOUD-PyHEADTAIL e-cloud simulations (G. Iadarola)
Here applied to FASTION type simulation with PyECLOUD-PyHEADTAIL
Machine lattice divided over processors
Bunch train tracked through the full part of each process bunch-by-bunch
All bunches collected on master at end of every turn 1 2 3 4 5 6 7

19. Parallelization with PyPARIS
Define speed-up for Nproc processors, in terms of time for one turn Tturn
Theoretical maximum S = Nproc
Estimated maximum, taking into account waiting time at beginning and end of turn
Measured on 12-core machine
(LIU-PS-GPU)
Test case with 64 bunches
With ~8 cores:
5-6 minutes/turn
5-6 days/damping time

20. Summary & Outlook Currently two tools for FASTION simulations: FASTION
Suitable e.g. for CLIC linac: requires field ionization, run-time not too long
PyECLOUD-PyHEADTAIL
Suitable for synchrotrons: synchrotron motion, feedback, parallelization, etc.
Gas ionization should be generalized to multiple species
Could be extended also with field ionization, to cover linac case
Hardware for parallel simulations needed
Available manpower main question
Past few years, a fraction of 1 person
Future? Certainly simulations studies are needed at least

21. References
G. Rumolo and D. Schulte,“Fast ion instability in the CLIC transfer line and main linac”, in Proc. 11th European Particle Accelerator Conf. (EPAC’08), Genoa, Italy, Aug. 2008, pp. 655-–657.
G. Rumolo and D. Schulte, “Update on fast ion instability simulations for the CLIC main linac”, in Proc. 23rd Particle Accelerator Conf. (PAC’09), Vancouver, Canada, May 2009, pp. 4658–4660.
J.-B. Jeanneret, G. Rumolo, and D. Schulte, “Vacuum specifications for the CLIC main linac”, in Proc. 1st Int. Particle Accelerator Conf. (IPAC’10), Kyoto, Japan, 2010, pp. 3401– 3403.
A. Oeftiger and G. Rumolo, “Fast beam-ion instabilities in CLIC main linac: vacuum specifications”, CERN, Geneva, Switzerland, Rep. CERN-OPEN , Nov
L. Mether, G. Iadarola and G. Rumolo, “Numerical modeling of fast beam ion instabilities”, Proc. HB2016, Malmö, Sweden, Jul