Subscribe to the gold package and get unlimited access to Shamra Academy
Register a new userGPU-Enabled Particle-Particle Particle-Tree Scheme for Simulating Dense Stellar Cluster System
81
0
0.0
(
0
)
Ask ChatGPT about the research
No Arabic abstract
We describe the implementation and performance of the ${rm P^3T}$ (Particle-Particle Particle-Tree) scheme for simulating dense stellar systems. In ${rm P^3T}$, the force experienced by a particle is split into short-range and long-range contributions. Short-range forces are evaluated by direct summation and integrated with the fourth order Hermite predictor-corrector method with the block timesteps. For long-range forces, we use a combination of the Barnes-Hut tree code and the leapfrog integrator. The tree part of our simulation environment is accelerated using graphical processing units (GPU), whereas the direct summation is carried out on the host CPU. Our code gives excellent performance and accuracy for star cluster simulations with a large number of particles even when the core size of the star cluster is small.
rate research
Read More
In a standard theory of the formation of the planets in our Solar System, terrestrial planets and cores of gas giants are formed through accretion of kilometer-sized objects (planetesimals) in a protoplanetary disk. Gravitational $N$-body simulations of a disk system made up of numerous planetesimals are the most direct way to study the accretion process. However, the use of $N$-body simulations has been limited to idealized models (e.g. perfect accretion) and/or narrow spatial ranges in the radial direction, due to the limited number of simulation runs and particles available. We have developed new $N$-body simulation code equipped with a particle-particle particle-tree (${rm P^3T}$) scheme for studying the planetary system formation process: GPLUM. For each particle, GPLUM uses the fourth-order Hermite scheme to calculate gravitational interactions with particles within cut-off radii and the Barnes-Hut tree scheme for particles outside the cut-off radii. In existing implementations, ${rm P^3T}$ schemes use the same cut-off radius for all particles, making a simulation become slower when the mass range of the planetesimal population becomes wider. We have solved this problem by allowing each particle to have an appropriate cut-off radius depending on its mass, its distance from the central star, and the local velocity dispersion of planetesimals. In addition to achieving a significant speed-up, we have also improved the scalability of the code to reach a good strong-scaling performance up to 1024 cores in the case of $N=10^6$. GPLUM is freely available from https://github.com/YotaIshigaki/GPLUM with MIT license.
Numerical methods to improve the treatment of magnetic fields in smoothed field magnetohydrodynamics (SPMHD) are developed and tested. Chapter 2 is a review of SPMHD. In Chapter 3, a mixed hyperbolic/parabolic scheme is developed which cleans divergence error from the magnetic field. Average divergence error is an order of magnitude lower for all test cases considered, and allows for the stable simulation of the gravitational collapse of magnetised molecular cloud cores. The effectiveness of the cleaning may be improved by explicitly increasing the hyperbolic wave speed or by cycling the cleaning equations between timesteps. In the latter, it is possible to achieve DivB=0. Chapter 4 develops a switch to reduce dissipation of the magnetic field from artificial resistivity. Compared to the existing switch in the literature, this leads to sharper shock profiles in shocktube tests, lower overall dissipation of magnetic energy, and importantly, is able to capture magnetic shocks in the highly super-Alfvenic regime. Chapter 5 compares these numerical methods against grid-based MHD methods (using the Flash code) in simulations of the small-scale dynamo amplification of a magnetic field in driven, isothermal, supersonic turbulence. Both codes exponentially amplify the magnetic energy at a constant rate, though SPMHD shows a resolution dependence that arises from the scaling of the numerical dissipation terms. The time-averaged saturated magnetic spectra have similar shape, and both codes have PDFs of magnetic field strength that are log-normal, which become lopsided as the magnetic field saturates. We conclude that SPMHD is able to reliably simulate the small-scale dynamo amplification of magnetic fields. Chapter 6 concludes the thesis and presents some preliminary work demonstrating that SPMHD can activate the magneto-rotational instability in 2D shearing box tests.
We present a novel method for particle splitting in smoothed particle hydrodynamics simulations. Our method utilizes the Voronoi diagram for a given particle set to determine the position of fine daughter particles. We perform several test simulations to compare our method with a conventional splitting method in which the daughter particles are placed isotropically over the local smoothing length. We show that, with our method, the density deviation after splitting is reduced by a factor of about two compared with the conventional method. Splitting would smooth out the anisotropic density structure if the daughters are distributed isotropically, but our scheme allows the daughter particles to trace the original density distribution with length scales of the mean separation of their parent. We apply the particle splitting to simulations of the primordial gas cloud collapse. The thermal evolution is accurately followed to the hydrogen number density of 10^12 /cc. With the effective mass resolution of ~10^-4 Msun after the multi-step particle splitting, the protostellar disk structure is well resolved. We conclude that the method offers an efficient way to simulate the evolution of an interstellar gas and the formation of stars.
Supersonic turbulence is believed to be at the heart of star formation. We have performed smoothed particle magnetohydrodynamics (SPMHD) simulations of the small-scale dynamo amplification of magnetic fields in supersonic turbulence. The calculations use isothermal gas driven at rms velocity of Mach 10 so that conditions are representative of star-forming molecular clouds in the Milky Way. The growth of magnetic energy is followed for 10 orders in magnitude until it reaches saturation, a few percent of the kinetic energy. The results of our dynamo calculations are compared with results from grid-based methods, finding excellent agreement on their statistics and their qualitative behaviour. The simulations utilise the latest algorithmic developments we have developed, in particular, a new divergence cleaning approach to maintain the solenoidal constraint on the magnetic field and a method to reduce the numerical dissipation of the magnetic shock capturing scheme. We demonstrate that our divergence cleaning method may be used to achieve $ abla cdot {bf B}=0$ to machine precision, albeit at significant computational expense.
We describe a simple method for simulating the dynamics of small grains in a dusty gas, relevant to micron-sized grains in the interstellar medium and grains of centimetre size and smaller in protoplanetary discs. The method involves solving one extra diffusion equation for the dust fraction in addition to the usual equations of hydrodynamics. This diffusion approximation for dust is valid when the dust stopping time is smaller than the computational timestep. We present a numerical implementation using Smoothed Particle Hydrodynamics (SPH) that is conservative, accurate and fast. It does not require any implicit timestepping and can be straightforwardly ported into existing 3D codes.
Log in to be able to interact and post comments
comments
Fetching comments
Sign in to be able to follow your search criteria
