accuracy and performance upgrade of pic and hybrid space-charge solvers
DESCRIPTION
Accuracy and Performance Upgrade of PIC and Hybrid Space-Charge Solvers. Leonid Vorobiev Muons, Inc., Batavia, IL 60510, USA. ABSTRACT - PowerPoint PPT PresentationTRANSCRIPT
CERN, Space Charge 2013, April 17 L.G.Vorobiev 1Muons, Inc.
Accuracy and Performance Upgrade of PIC and Hybrid Space-Charge Solvers
Leonid Vorobiev
Muons, Inc., Batavia, IL 60510, USA
SPACE CHARGE 2013
16-19 April, CERN
ABSTRACT
For the high luminosity LHC (HL-LHC), space charge study requires high quality modeling. In this talk the sources of numerical errors in space charge algorithms are analyzed, and their mitigation is suggested for both conventional space charge grid solvers (based on PIC formalism) and hybrid solvers. The implementation of these improvements may significantly increase the accuracy, speed and physical validity of numerical results and contribute to achieve the required beam parameters in the LHC accelerator complex.
CERN, Space Charge 2013, April 17 L.G.Vorobiev 2Muons, Inc.
Overview
General SC: Grid(PIC), HybridComplex boundaries, Image Forces,Wakes, Beam Loading, Halo, Clusters ...
Works fine for many Works fine for many applications at Macro-levelapplications at Macro-level
High-Current Beam Simulation = Tracker & SC Solver
• Symplectic maps
• Integrators, high orderTracking enginesTracking engines =
&
SCSC
Zero/Non-Zero Frozen SCFree space, Pre-assigned Distribution: KV, Gaussian,…,A starting point
EXACT, Not General APPROXIMATE, General
~EXACT
Recent findings: Frank Schmidt, et al. Micro-scale effects ARE persistent.
E.G.: ORBIT code J.Galambos, J.Holmes, D.Olsen, A.Luccio and J.Beebe-Wang, 1999.
CERN, Space Charge 2013, April 17 L.G.Vorobiev 3Muons, Inc.
Some History
Micro-scale unphysical effects: H-detuning (left), fake tuneEvolutions, artificial dipole oscillations,…
May jeopardize the validity of beam modeling
WANTED: Consistency between ~Exact Trackers and approximate SC Solvers
Search for remedies
Fake Tune Evolution PTC-ORBITH-DETUNING PTC-ORBIT
F.SchmidtAPC Seminarat Fermilab,November 2012
CERN, Space Charge 2013, April 17 L.G.Vorobiev 4Muons, Inc.
Outline
1. PIC Formulation. Clouds & Grid Functions. Errors & Accuracy Upgrades. Limitations.
2. Hybrid Space Charge Solvers.
3. SC Templates. Back to PIC - Split operator revisit. Hierarchy of SC Templates
4. Discussion.
5. Conclusion. Credits.
6. References.
7. SC Solvers vs Applications.
CERN, Space Charge 2013, April 17 L.G.Vorobiev 5Muons, Inc.
1. PIC Formulation
The Green’s function in Differential form, for electrostatic approximation, leads to
n
uand/or , :conditionsBoundary
),,( ),,,( ,4
u
zyxzyxuuu
STEPS: A. Build a Grid. Charge density on the grid.B. Potential/Fields on the grid ( FFT-series, Multi-Grid, Iterations
with accel., Boundaries/Symmetries.C. Interpolation between nodes.
Refs.: Birdsall and Langdon [1], Hockney and Eastwood [2]
Poisson equation:
CERN, Space Charge 2013, April 17 L.G.Vorobiev 6Muons, Inc.
1.A Grid Density. Known-Unknown?
IF SC density is known (Gaussian,…exact)
(Multi-) Grid solvers may approach a solution with machine accuracy, exact SC Fields
Refs. www.nr.com [3], Saraniti [4]
IF SC density is approximate, - a starting
point in SC errors
CERN, Space Charge 2013, April 17 L.G.Vorobiev 7Muons, Inc.
1.A Grid Density. Clouds.
Contour lines (left), 3D plot (right). Simplest Cloud. Strong Numerical noise
PIC Clouds:Size, Shape, Ramp Redistribution of SC
over the grid
From: Ref. [5], Vorobiev and Hirata.
CERN, Space Charge 2013, April 17 L.G.Vorobiev 8Muons, Inc.
1.A Grid Density. Num. Noise Movie
Contour lines (left), 3D plot (right). Noise – suppressed. What about Accuracy?
100K Part.256x256
CERN, Space Charge 2013, April 17 L.G.Vorobiev 9Muons, Inc.
1.A SC Density. Grid Dim & Npart
1M, 256x256 4M, 256x256 4M, 512x512
More part.More grid
dimensions
Slow Down
Viciouscircle
CERN, Space Charge 2013, April 17 L.G.Vorobiev 10Muons, Inc.
1.A Grid Density. Num. Noise Movie
Contour lines (left), 3D plot (right). Noise – suppressed. What about Accuracy?
4M Part.512x512
CERN, Space Charge 2013, April 17 L.G.Vorobiev 11Muons, Inc.
1.B Grid Potential and Fields
Num. Noise concealed,but persists in dynamics
y
uE
x
uE
ugrad
y
x
E
2
2
2
2
is y
u
x
u
CERN, Space Charge 2013, April 17 L.G.Vorobiev 12Muons, Inc.
1.B Field Errors Suppression
Field Ex,y errors, depending on SC Clouds
What is ?max,
*,,
,yx
yxyxerryx E
EEE
*
, yxE
CERN, Space Charge 2013, April 17 L.G.Vorobiev 13Muons, Inc.
1.B Field Errors, Snapshots
CERN, Space Charge 2013, April 17 L.G.Vorobiev 14Muons, Inc.
1.B Field Errors. Choice of
Field errors, depending on SC Clouds and
*, yxE
*, yxE
CERN, Space Charge 2013, April 17 L.G.Vorobiev 15Muons, Inc.
Field errors, dependingon SC Cloud shape/size
*0
*0
200
function goal ,2
function goal ,2
4)(1
y
x
Eyy
u
Exx
u
rur
ur
rr
*, yxEmax
,
*,,
,yx
yxyxerryx E
EEE
min. But what is ?
Simplest benchmarking:
1.B Field Errors. Choice of *, yxE
CERN, Space Charge 2013, April 17 L.G.Vorobiev 16Muons, Inc.
1.B Clusters
All this works similarlyfor Symmetries,Boundary constraints,as well as forclustered beams
Contour lines, 3D shape, Potential (top), Fields (bottom)
CERN, Space Charge 2013, April 17 L.G.Vorobiev 17Muons, Inc.
1.C Interpolation.
SC Fields are solved on the Grid
Particles are in between Grid nodes:
Bi-Tri-Piecwise Interpolation - WrongBi-Tri-Linear Interpolation - InsufficientBi-Tri-Cubic Interpolation: Splines - Best
Ref. Vorobiev and Hirata [5]
CERN, Space Charge 2013, April 17 L.G.Vorobiev 18Muons, Inc.
1. Grid Solvers. Limitations
PIC Grid Solvers: solid/clustered beams, OkFilaments, Sheet beams, Halo, Stray particles not OkMesh refinement, • Density errors, irregular Clouds, use more particles• Errors due to interpolation on irregular grid• Slowing down speed…
Aside from PICs:• Particle-Core, Envelope (linear) & non-linear model ?• Integral representation of the Green’s function: OkComplex Boundary 3D, Sheet Beam (Photo-Injector) Ok• Hybrid SC Solvers
CERN, Space Charge 2013, April 17 L.G.Vorobiev 19Muons, Inc.
2. Hybrid SC Solvers. Outline.Hybrid Space Charge Solvers = Green’s function
in Integral form with Tricks
The Best ones for photo-injectors, el.optics, complex boundaries:VERY accurate, SLOW
Refs: Roger Harrington [6], Miklosh Szilagyi [7], Valentin Ivanov [8],In part: Fikera, Quang, Ryan, Kapin, Hess, Krassilnikov, Vorobiev
This Presentation:A. Space charge templates. Instead of integrated GF, use
Library of EXACTLibrary of EXACT SC potential & fields within boundariesSC potential & fields within boundariesB. From SC Templates to PIC: Split Operator RevisitSplit Operator RevisitC. Hierarchy of Templates
CERN, Space Charge 2013, April 17 L.G.Vorobiev 20Muons, Inc.
Disk-Template, LibraryDisk-Template, Library
Ring-Template = SuperpositionRing-Template = Superpositionof positive (BLUE) & negative of positive (BLUE) & negative (RED) derived from Disc-(RED) derived from Disc-TemplateTemplates
3D beam (non-elliptical!)Within conductingBoundary (not shown)3D Templates withinthe same boundary
NO GRID 3D Poisson Equation, INSTEAD: Convolve Template Potentials/Fields
)()(
, )(4)(
xx
xxx
beambeam
beambeam
Uu
u
)()(
, )(4)(
xx
xxx
tmptmp
tmptmp
Uu
u
zdzzSzzyxuzyxu yxtmpbeam~))~(),~(,~,,(),,( ,
3.A Hybrid Solvers, SC Templates
Ref. Vorobiev and York [9]
CERN, Space Charge 2013, April 17 L.G.Vorobiev 21Muons, Inc.
Library of TemplatesPre-Assigned densities
Ring Disc Template
• General • Halo
• Hollow Beam
EXP
ARB
-r_m 0 +r_m
3.A Hybrid Solvers, SC Templates
Private Communications:
A. Friedman, D. Grote, I. Hofmann, M. Reiser, J. Struckmeier and R. York
CERN, Space Charge 2013, April 17 L.G.Vorobiev 22Muons, Inc.
3.A Template Solvers, Machinery
SC Templates - Machinery
Template PotsFree space (green)Cond. pipe (blue)Image Dens (red)
Beam, Cond.Pipe Template Pots Template Fields Template uzz
Beam Pot Beam Field Beam uzz
CERN, Space Charge 2013, April 17 L.G.Vorobiev 23Muons, Inc.
3.B Hybrid Solvers. From Templates Back to PICs
BEAM NOT A RODBEAM NOT A RODE.G.
FNAL Booster, MI,… x84,x588
Acceleration withtransition crossing
Squeeze Head Tail
2.5D – series of 2D solutions, as if
BEAM BEAM == ROD ROD“BEAM=ROD” model - Good for PSR, SNS,… 2D beam, 1 long bunch 80% of circumference)
CERN, Space Charge 2013, April 17 L.G.Vorobiev 24Muons, Inc.
3.B Hybrid Solvers,Split-Operator Revisit
2
2
2
2
2
2
4z
u
y
u
x
u
corry
u
x
u 42
2
2
2
4/)/(),,( 22 zuzyxcorr
Ref: Vorobiev and York, Phys. Rev. ST Accel. Beams 3, 114201 (2000)
2
22
4
1),(
z
u
H
yxTZ
D
-0.2 -0.1 0.0 0.1 0.2
Z [m]
-10000
-6000
-2000
2000
Off-
axis
dE
z/dz
[V
/m2 ]
Sub-3D
CERN, Space Charge 2013, April 17 L.G.Vorobiev 25Muons, Inc.
3.B Hierachy of Space Charge Templates
SC Templates - Flowchart
CERN, Space Charge 2013, April 17 L.G.Vorobiev 26Muons, Inc.
4. Discussion. Existing SC Classes.Project X,Project X, Main Injector, H- Multiturn Injection
Transverse Painting (x,x’),(y,y’), Longitudinal Painting
FNAL BoosterTransition Crossing:RF cavities - narrow
SC issues
Ref. Vorobiev [11-12]
CERN, Space Charge 2013, April 17 L.G.Vorobiev 27Muons, Inc.
Mu2e Exp. In Fermilab:Extinction interval for mu2e detector:NO STRAY PROTONS
Long. dynamics w/o (left), with (cntr) beam loading. Histograms (right)
4. Discussion. Existing SC Classes
Ref. Balbekov and Vorobiev [10]
CERN, Space Charge 2013, April 17 L.G.Vorobiev 28Muons, Inc.
4. Discussion: Done,…, TBD
Accuracy upgrades• SC Density. Clouds. Size. Shape. +• Interpolation. Splines… +
• 3D SC Templates for Ez +
• Revisit Split Operator, Sub-3D +• 2D Templates, Envelope & Images +• Halo TBD
Memory,
Speed• No Grid
• Much lesser part
• Lesser Memory demands due to parameterization
Beam Loading & Wakes, 1++ Kicks/turn? TBD
Existing SC modules: Multi-turn H- injection (Project-X), Beam Loading (mu2e), Long. Dynamics (Transition Xing, Head-Tail…
On the way to full Synchro-Betatron dynamics studies
CERN, Space Charge 2013, April 17 L.G.Vorobiev 29Muons, Inc.
4. Discussion. HL-LHC Computational Needs
HL-LHC high intensity mode of operation
Linacs → Series of Synchrotrons → LHC
Linacs, Multiturn Injection, Acceleration, Trans. Crossing, Circulation
Tune shift, Image forces, Wakes, B. Loading Halo, Losses…
Wanted: SC Solvers – consistent to Symplectic Trackers and Integrators
SC Performance - Overall Accuracy & Speed Upgrade, Reasonable Memory Capacities, Physical Validity
CERN, Space Charge 2013, April 17 L.G.Vorobiev 30Muons, Inc.
5. Conclusion. Credits.
Charles Ankenbrandt, Thomas Roberts, Jeffrey Holmes
Richard York, Pavel Zenkevich
ANDRolland Johnson, Elias Metral and Frank Schmidt
Optional Space Charge Solvers in existing code
Cross-checking between codes, Stress/Regression Tests
Keep development of different codes by different authors – MUST
GOAL: reliable, insightful modeling for HL-LHC
Credits to:
CERN, Space Charge 2013, April 17 L.G.Vorobiev 31Muons, Inc.
6. ReferencesPIC1. C.K.Birdsall and A.B. Langdon (1985). Plasma Physics via Computer Simulation. McGraw-Hill. 2. R.W.Hockney and J.W. Eastwood (1988). Computer Simulation Using Particles. CRC Press.
Multi-Grid Poisson Solvers3.W.H. Press, S.A.Teukolsky, W.T.Vetterling, B.P.Flannery, “Numerical recipes”, www.nr.com .4. M.Saraniti, et al. IEEE Transactions on CAD, Volume: 15 Issue: 2, Page(s): 141-150 (1996).
Symplectic SC solver5. Vorobiev and Hirata, Report KEK 95-12; http://ccdb5fs.kek.jp/cgi-bin/img/allpdf?199524012\
Green’s Function6. R.Harrington, “Field Computation by Moment Methods”, Macmillan, New York (1968)7. M.Szilagyi, “Electron and Ion Optics”, Plenum, New York (1988)8. V.Ivanov “Green’s Function Techniques in Forming Intense Beams”, (1989, 2009), Int. J.Mod.Phys A,
Vol.24, No. 5, 869–878,…In part: Quang, et al. (2006), Kapin (2002), Hess, et al.(2007), Vorobiev (1999).
SC Templates:9. Vorobiev and York: Phys.Rev. STAB 3, 114201 (2000); MSUCL-Report 1117, 1998;PAC’1999, pp. 2781-2783; Springer: http://link.springer.com/chapter/10.1007%2F3-540-47789-6_33PAC’2001 pp. 3075-3078; EPAC 2002, pp. 1679-1681; PAC’2003 pp. 3533-3535;Fermilab Report 08-236, http://lss.fnal.gov/archive/2008/pub/fermilab-pub-08-236-apc.pdf
BL + Transition crossing: 10. Balbekov and Vorobiev, http://beamdocs.fnal.gov/AD-public/DocDB/ShowDocument?docid=407011. Vorobiev, https://indico.fnal.gov/conferenceDisplay.py?confId=4497
http://beamdocs.fnal.gov/AD-public/DocDB/ShowDocument?docid=413412. Vorobiev, http://beamdocs.fnal.gov/AD-public/DocDB/ShowDocument?docid=4258
CERN, Space Charge 2013, April 17 L.G.Vorobiev 32Muons, Inc.
7. SC Solvers vs ApplicationsApplications to
Single-pass systemsSingle-pass systems
• (Photo-) Injectors, Electron Optics• Linacs: LBT, Acceleration• Transport, Final Focusing
Multi-pass systemsMulti-pass systems
• Rings: Multi-Turn Injection+ Acceleration
• Storage Rings, Colliders
?(maybe/no) Symplecticity (maybe/yes)? ?(maybe/yes) Complex Boundary (no)?
?(maybe/yes) Multi-step SC (no/maybe UMd ERing)??(maybe/no) SC Kicks (yes)?
?(maybe) Beam Loading, Wakes (yes)?…
Yes / Maybe / Not Yes / Maybe / Not