Current status theory and simulations ALEGRO WG3 J. Vieira - - PowerPoint PPT Presentation

Jorge Vieira | ALEGRO workshop, Oxford | March 26, 2018

INSTITUTO DE PLASMAS E FUSÃO NUCLEAR

Current status theory and simulations ALEGRO WG3

golp golp

J. Vieira

• GoLP / Instituto de Plasmas e Fusão Nuclear

Instituto Superior Técnico, Lisbon Portugal

web.ist.utl.pt/jorge.vieira epp.tecnico.ulisboa.pt || golp.tecnico.ulisboa.pt

Courtesy W. An et al.

Jorge Vieira | ALEGRO workshop, Oxford | March 26, 2018

Acknowledgements

Contributions from H. Vicenti (CEA), T. Mehrling, (LBNL); R.A. Fonseca (ISCTE, IST), J. England (SLAC); U. Niedermayer (TUD), B. Cowan (TECH-X), W. An (UCLA), D. Gordon (NRL)

)

Ricardo Fonseca | ALEGRO 2018

Physics Models for Plasma Based Collider Simulation

• Enough high energy events per second to measure something

– TeV energy: long plasmas, high peak power lasers – Average power: long time scale behavior (hydro/transport) – Emittance: source emittance, emittance growth

• Ability to generate/manipulate short positron bunches

– Novel wakefield structures – Perturbative or nonlinear QED processes

• Polarization resolution is important to physics case

– Question of polarized sources and polarization preservation

25/03/18 1

Desired conditions at final focus determine the physics issues:

• D. Gordon et al.

Ricardo Fonseca | ALEGRO 2018

Physics Models for Plasma Based Collider Simulation

Injection Stage Laser Module Reasonable guess at major and ancillary components points to physics: Polarization Preservation Acceleration Stages Final Focus Coupling Elements Beam Sources Plasma Sources

Spin-resolved dynamics: Dirac Equation BMT Equation Propagation coupled to rate equations; Flow models (cooling); Material damage. Ionization physics, QED models, Cathode models Plasma lenses, MHD, material damage. Chemical kinetics Hydrodynamics & Transport Beamstrahlung Usual Maxwell- Lorentz; Emittance Growth

• D. Gordon et al.

Ricardo Fonseca | ALEGRO 2018

State-of-the-art of current modeling tools for Laser/Plasma Wakefield Accelera;on

• I. Context
• II. Challenges to overcome in LWFA/PWFA simula;ons
• III. Methods for realis;c ;me/ressource-to-solu;on
• H. Vincen;*, T. Mehrling* and J-L Vay (with contribu;ons from: R. Lehe, C. BenedeL, X. Davoine, W. Ann, T. Grismayer, M. Vranic and K. Lotov) - *speakers

« 3D standard full PIC modelling is extremely challenging »

Mi;gate numerical ar;facts Reduce ;me/length scale dispari;es Reduce dimensionality Require kine>c « Par>cle-In-Cell (PIC) » simula>ons

• Wavebreaking
• Non-linear regimes
• Large ;me/length scale

dispari;es

• 3D geometry very costly
• Numerical ar;facts

>109 macropar;cles/109 cells Driver 1 µm / Acc length cm to 100 m High spa;o-temporal resolu;on required

Dispersion-free Maxwell solvers Lorentz invariant par>cle pushers Quasi-sta>c/ Envelope Adapta>ve Mesh Refinement R-Z geometry With Azimuthal decomposi>on Boosted frame

• IV. Methods to account for physics beyond Vlasov
• Field/collisional ioniza>on
• Radia>on (sub-cell)
• QED Monte-Carlo modules
• Spin polariza>on

Merging/SpliSng Fluid Galilean solvers/ Bumped dispersion Laser envelope

driver: driver:

LWFA PWFA

Ricardo Fonseca | ALEGRO 2018

QuickPIC

QuickPIC simulation of LWFA with a beam load

QuickPIC is a 3D parallel Quasi-Static PIC code, which is developed based on the framework UPIC.

QuickPIC simulation of two-bunch electron-driven PWFA experiment at FACET shown on the Nature cover.

QuickPIC simulation of positron-driven PWFA.

https://github.com/UCLA-Plasma-Simulation-Group/QuickPIC-OpenSource

http://picksc.idre.ucla.edu

Recent Progress MPI+OpenMP with Tiles OO Design Using Fortran 2003 Vectorization on KNL (3.61 ns/particle/step) Open Source on Github Future Plans

Dynamic load balancing

Adaptive 2d and 3d time steps Adaptive particle loading

Adaptive mesh refinement

• W. An et al.

Algorithmic needs for LWFA: quasi-static PIC algorithm QuickPIC

Ricardo Fonseca | ALEGRO 2018

Towards Exascale Computations

• Outstanding progress in computational power since

1950s

• Present systems can reach performances of 0.1 EFlop/s
• Energy cost for calculations has gone down by 14 orders
• f magnitude
• Continuous evolution of architectures and computing

paradigms

• Exascale simulations are within reach
• Present simulations can already track > 1013 particles for

millions of time steps

• Increasing quality and quantitative fidelity of simulations
• Continuously evolve algorithms and codes to effjciently

use new generations of computing hardware

• Community efgort among experts in large scale plasma

simulation

• This evolution presents a formidable challenge for

computational physicists

• Useful to have an ecosystem of codes where ideas are

shared.

• The community needs sustainable support for exascale

software development

7 Derouillat et al, CPC 222 351 (2018)

Dynamic load balancing of a LWFA simulation

• R. A. Fonseca et al.

Ricardo Fonseca | ALEGRO 2018

WG7 (DLA): Optical System Design Combining Analytical Calculation, FDTD, AVM, and NLSE

• Design Study of Integrated Multi-Stage DLA Network
• Realistic Component Parameters
• Adjoint Variable (AVM or “Inverse Design”) Based Structure Optimizations

couplers splitters phase shifters

• T. Hughes, et al. “On-Chip Laser

Power Delivery System for Dielectric Laser Accelerators,” submitted (2018)

• J. England et al.

Ricardo Fonseca | ALEGRO 2018

WG7 (DLA): Particle Tracking Using Semi- Analytic 6D DLA Tracker

One kick per grating cell (numerically lightweight) Symplectic code à No artificial emittance increase à Natural emittance increase properly determined Kicks by resonant Fourier coefficient (one complex number per grating cell) Transverse kick by Panofsky-Wenzel theorem Can be applied to laterally coupled structures Subrelativistic & Relativistic structures Tilted grating structures Alternating phase / Spatial Harmonic focusing

• U. Niedermayer et al. Phys. Rev. Accel. Beams 20, 111302 (2017)
• J. England et al.

Ricardo Fonseca | ALEGRO 2018

Key computational question for DLA collider

• Will beam remain stable over km of propagation?
• Or are beam-driven wakefields a show-stopper?
• Need algorithms/codes for propagation with wakefields
• Wakefields for short (~1 mm–1 cm) structures can be computed with standard

PIC (mostly—see below)

• But even meter scale (~million structure periods) intractable
• Need reduced model for wakefields to propagate with tracking
• Wakefield physics not even fully understood
• Optical bunches have frequency content deep in UV
• What is material response? Need iteration between fundamental modeling and

experiment

• B. Cowan et al.

Jorge Vieira | ALEGRO workshop, Oxford | March 26, 2018

Conclusions

Plasma based acceleration Full PIC codes are the most predictive but are very CPU expensive but . Reduced codes are less

• accurate. Combination and cross check between different models is crucial.

Large efforts devoted to improving the accuracy of the algorithms (e.g. accurate predictions of beam emittance growth requires numerical Cherenkov free solvers). Improving agreement with experiments/understanding the origin of disagreements is critical to improve current designs and requires combined efforts of numerical simulations, theory and experimental diagnostics (e.g. effects of higher order laser modes/accurate description of laser profiles in simulations). Including additional physics beyond electromagnetism is important to tackle open questions (spin tracking, ionisation, disruption, hydro codes, …). Investigating new types of plasma waves (with different topologies, geometries) required to address current challenges (e.g. positron acceleration) and to open new avenues of research DLA Current challenge: staging several accelerating structures is critical to boost energy gains Accurate and systematic numerical modelling of 3D beam dynamics is an on-going effort

Jorge Vieira | ALEGRO workshop, Oxford | March 26, 2018

Agenda