DISPATCH simulations

• Åke Nordlund

Room: 122:026 We use the adaptive mesh refinement codes RAMSES and DISPATCH to model protoplanetary disks in realistic star formation environments, covering a range from a 40 pc outer scale to an inner scale of 1/10 of an Earth radius (more than 12 orders of magnitude). The simulations are done in four steps, with the first step following individual star formation in a 40 pc GMC model with 120 AU resolution. In the 2nd step, the neighborhoods of several stars with a final system mass of 1-2 solar masses are followed during the accretion process, with a smallest mesh size of 2 AU, sufficient to follow the development of the large scale structure of the accretion disks and their accretion histories. A selection of these disks are then studied over 100-1000 yr, with mesh sizes down to 0.015 AU. Using the new DISPATCH adaptive mesh refinement code, we then study the dynamics of gas, dust, and ensembles of chondrule size pebbles in the disk, following the accretion of pebbles onto planetary embryos surrounded by proto-atmospheres, with three dimensional radiative energy, and a resolution of 600 km. The unprecedented scale range is made possible by the new DISPATCH code, which is explicitly constructed for the exascale computing era, with the ability to scale to essentially unlimited number of cores with sustained performance per core. This is achieved by using task based scheduling, where tasks do MHD, radiative energy transfer, and evolve the positions of billions of pseudo-particles representing dust or pebbles, using time steps restricted only by local condition.

Techniques for well-behaved force-free regions in relativistic MHD simulations

• Kyle Parfrey

Room: 122:026

Energisation by reconnection within test-particle approach

• Dhrubaditya Mitra

Room: 122:026 We study the role of turbulence in magnetic reconnection, within the framework of magneto-hydrodynamics, using three-dimensional direct numerical simulations. For small turbulent intensity we find that the reconnection rate obeys Sweet-Parker scaling. For large enough turbulent intensity reconnection rate departs significantly from Sweet-Parker behaviour, becomes almost a constant as a function of the Lundquist number. We further study energisation of test- particles in the same setup. We find that the speed of the energised particles obeys a Maxwellian distribution, whose variance also obeys Sweet-Parker scaling for small turbulent intensity but depends weakly on the Lundquist number for large turbulent intensity. Furthermore, the variance is found to increase with the strength of the reconnecting magnetic field.

Radiation emission, radiation reaction and QED processes for PIC codes

• Marija Vranich

Room: 122:026 Particle-in-cell codes have been successfully employed to model particle acceleration in laboratory (e. g. laser-wakefield acceleration) and in space (e.g. in collisionless shocks). With the advent of laser technology, one can experiment with very intense fields, that would otherwise be available only in astrophysical objects (e.g. in pulsars or magnetars). At extreme intensities, new physics becomes relevant for modelling laser-plasma interactions, which brings new challenges for PIC development. For example, radiation can be emitted at high frequencies that are not resolved by the simulation grid. If a fraction of energy this radiation carries is negligible, we can compute the output radiation spectra by post-processing the particle trajectories or by using a real- time diagnostics that does not interfere with the PIC loop itself. However, if this radiation accounts for a large fraction of the particle energy, one needs to correct the particle momentum by including a classical description of radiation reaction (e.g. Landau & Lifshitz equation of motion instead of the Lorentz force). One can expect a correct post-processing account of the emitted radiation only if the particle trajectories themselves are correct and the emissivity calculation includes radiation damping corrections. An even greater computational challenge is modelling a quantum regime of emission - when a particle can emit a single photon that carries a large fraction of its energy. Such a photon is treated as an additional particle species, which can propagate through the simulation box and later decay into an electron- positron pair. The new pairs re-accelerate in the laser field, and they emit new photons. Repeated occurrence of this process can induce a so-called “QED cascade”, that generates an exponentially rising number of particles in the simulation box. Macroparticle merging algorithm is then necessary to keep the simulation load to a manageable level. We have developed a merging scheme that resamples particles in the simulation and preserves the particle distribution function. I will discuss the implementation of the above-mentioned computational developments in OSIRIS and show examples of physical problems where they are essential.

Discussion on MHD numerics

• Åke Nordlund

Room: 122:026

Reception at Nordita East Room: 122:026

Coffee Room: 122:026

PIC simulation of the thermal pressure-driven expansion of a blast shell into a magnetized ambient medium

• Mark Dieckmann

Room: 122:026 A large gradient of the thermal pressure in a collision-less plasma triggers the formation of a rarefaction wave. Rarefaction waves can accelerate ions to speeds of the order of hundreds to thousands of km/s in laser-plasma experiments and the collision of these fast ion beams with an ambient plasma triggers the formation of shocks. Forthcoming experimental campaigns will introduce a background magnetic field into the ambient plasma that yields a beta value of the order unity and study the magnetized shocks. I will present results from recent PIC simulation studies that investigated such shocks and show how a fast magnetosonic shock forms in such a plasma, which is trailed by a tangential discontinuity, and how a radially expanding blast shell interacts with a uniform background magnetic field.

Numerical methods in WarpX and PICSAR, new tools toward exascale AMR-PIC modeling of relativistic plasmas

• Jean-Luc Vay (Lawrence Berkeley National Laboratory)

Room: 122:026

Lunch Room: AlbaNova restaurant

Reconnection in magnetically-dominated plasmas

• Lorenzo Sironi (Columbia University)

Room: 122:026 In astrophysical relativistic outflows, the magnetic energy density might exceed even the plasma rest-mass energy density, a regime that is dramatically different from laboratory plasmas. We explore the physics of relativistic magnetic reconnection with fully-kinetic particle-in-cell (PIC) simulations, with an emphasis on large-scale computational domains with open boundary conditions. Our results have implications for magnetically-dominated jets, pulsar winds and accretion disk coronae.

Discussion on PiC methods

• Mark Dieckmann
• Jean-Luc Vay (Lawrence Berkeley National Laboratory)

Room: 122:026

Coffee Room: 122:026

Boat trip to Birka Room: Departs from Stadshusbron

Coffee Room: 122:026

Magnetic energy dissipation

• Andrei Beloborodov (Columbia University)

Room: 122:026

Radiative PIC simulations and their applications to pulsars

• Benoit Cerutti

Room: 122:026 I will present some of the recent efforts to model particle acceleration and emission of energetic radiation in pulsars, using global radiative PIC simulations. These studies show that the equatorial current sheet forming in the pulsar wind is the main site of particle acceleration. Relativistic reconnection dissipates magnetic energy which is then efficiently channeled into energetic particles and high-energy synchrotron radiation. Synthetic lightcurves, spectra and polarization present robust features reminiscent of observed gamma-ray pulsars. I will also discuss how particle acceleration proceeds in the wind far from the star but before it reaches the nebula.

Lunch Room: AlbaNova restaurant

Simulating the Twisted Life of Magnetars

• Alexander Chen

Room: 122:026

Discussion on radiative PiC

• Andrei Beloborodov (Columbia University)
• Benoit Cerutti

Room: 122:026

Coffee Room: 122:026

Workshop dinner Room: Prinsen restaurant

Coffee Room: 122:026

Vlasiator - enabling large scale hybrid-Vlasov simulations

• Sebastian von Alfthan (CSC)

Room: 122:026 Vlasiator is a 6D hybrid-Vlasov simulation code, that has been developed for simulating Earth's magnetosphere at kinetic scales. Here I will present some of the key techniques used for enabling current 5D simulations on Petascale machines, and also present some of the ongoing development efforts. In particular I will present the semi- Lagrangian solver used for propagating the Vlasov fluid, and describe the sparse representation that enables 5D simulations by reducing the problem size by orders of magnitude. I will also discuss current work on a new heterologous domain decomposition, in which the spatial domain of one mpi task is not identical for the field- and Vlasov solvers. Solving the two problems on separate grids allows load balancing to be performed separately and overcome the previously existing scaling problems. Finally I will also discuss experiences in porting Vlasiator to the Xeon Phi processor (Knights Landing), and share experience in running large scale simulations on a large Xeon Phi based supercomputer (Marconi).

Collisionless shocks in Vlasov-hybrid simulations

• Rami Vainio

Room: 122:026 I will present results from hybrid-Vlasov simulations using Vlasiator, the world's first global magnetospheric hybrid Vlasov code. I will consider both the Earth's bow shock and interplanetary shocks, analysing their similarities and differences in terms of global shock structure and kinetic ion processes at and near the shock fronts.

Lunch Room: AlbaNova restaurant

Radiative Vlasov simulations

• Joonas Nättilä (University of Turku)

Room: 122:026

Discussion on open questions & Nordita program

• Rami Vainio

Room: 122:026

Coffee Room: 122:026

End of Workshop Room: 122:026