CITA's expertise and computing infrastructure continue to drive a strong analysis and simulation effort. Current developments include large scale parallel N-body, hydrodynamics, and magnetohydrodynamics. CITA has hired a permanent parallel programmer to assist the utilization of its high performance computing infrastructute. Current hardware includes a 536 2.4 GHz CPU / 284 GB intel cluster, a 32-processor 64 GB shared memory Compaq alphaserver GS320, 8 quad es40 alphaservers, and intel and alpha linux clusters comprising upward of 100 processors. Recent highlights include detailed simulations of the intergalactic medium to model the Sunyaev-Zeldovich effect (Pen and Zhang; Bond, Wadsley and Ruetalo); construction of a dedicated out-of-core machine for cosmological gas and clustering simulations (Pen and Trac); hydrodynamical simulations of cluster formation (Loken and collaborators); hydrodynamical simulations of galaxy interactions (Dubinski and collaborators); simulations of two-phase protoplanetary disks (Humble and collaborators); a new numerical method to solve the equations of MHD (Pen and Arras); and precision calculations of the internal modes of rotating stars (Arras, Pen, and Wu).
Jungyeon Cho carried out numerical studies of MHD turbulence. Cho and Lazarian (Wisconsin) performed numerical simulations of electron MHD turbulence (a.k.a. whistler turbulence), which can be viewed as small scale Alfvenic turbulence. They found that anisotropic structure of whistler turbulence is more pronounced than that of Alfvenic turbulence. Cho (2005) numerically studied relativistic force-free MHD turbulence and found that the scaling relations of the turbulence is very similar to those of non-relativistic Alfvenic turbulence.
Iliev has performed, in collaboration with Paul Shapiro (University of Texas at Austin) and Alejandro Raga (UNAM, Mexico City), the first realistic radiation-hydrodynamical simulations of the photoevaporation of cosmological minihalos by global I-fronts during the epoch of reionization. They have studied this complex process in detail, deriving many important quantities like the evaporation timescale, the total number of ionizing photons required for completion, along with some observable consequences. For the collapsed fraction in minihalos expected during reionization, they found that this process can add up to 1 photon per total atom in the universe to the requirements for completing reionization, potentially doubling the minimum number of photons required to reionize the universe.
Precise modeling of radiative feedback processes with codes which couple radiative transfer directly to hydrodynamics is very important in order to better understand reionization and other radiative feedback-related phenomena. Ilian Iliev, in collaboration with Garrelt Mellema (ASTRON, Netherlands), Marcelo Alvarez (Austin) and Paul Shapiro, has developed a new method for transferring ionizing radiation which is explicitly photon-conserving, fast and efficient. It can accurately follow ionization fronts without adoption of small time-steps, making it very appropriate for direct coupling with gasdynamic and N-body codes. This work is currently continuing through coupling this new radiative transfer method with gasdynamics and applying it to simulations of EOR and other astrophysical problems in which radiative transfer is a key process.
M. Liebendoerfer, M. Rampp (Max-Planck Institute for Astrophysics, Garching, MPA), H.-T. Janka (MPA), and A. Mezzacappa (Oak Ridge National Laboratory, ORNL) have compared in detail their supernova simulations implementing an implicit general relativistic solution of the Boltzmann transport equation or a variable Eddington factor method for the neutrino transport and found satisfactory agreement in spherical symmetry. Machine-readable data from these reference simulations have been published to gauge neutrino transport approximations.
In collaboration with K. Langanke (Univ. of Aarhus), G. Martinez-Pinedo (Institut d'Estudis Espacials de Catalunya, Barcelona, IEEC), J. M. Sampaio (Univ. of Aarhus), D. J. Dean (ORNL), W. R. Hix (University of Tennessee, Knoxville, UTK), O. E. B. Messer (UTK), A. Mezzacappa (ORNL), H.-T. Janka (MPA), and M. Rampp (MPA), Liebendoerfer investigated the electron capture rates on heavy nuclei. They found that Pauli blocking does not occur to the extent assumed in previous core collapse simulations. Simulations with the improved rates showed that these ignored reactions actually dominate throughout core collapse.
D. Richmond (Univ. of Victoria) and M. Liebendoerfer have published an extension of the adaptive grid in the hydrodynamics code AGILE that would allow to dynamically excise the singularity from a supernova simulation when a black hole forms at the center of the event.
C. Froehlich (Univ. of Basel), P. Hauser (Univ. of Basel), M. Liebendoerfer, G. Martinez-Pinedo (IEEC), and F.-K. Thielemann (Univ. of Basel) et al. investigated the impact of weak interactions in the vicinity of the mass cut in nucleosynthesis calculations for supernova explosions. Previous calculations assumed an unchanged progenitor composition. It turns out that the large neutrino fluxes drive the electron fraction to higher values, sometimes even above 0.5 and the correspondingly enhanced abundances of 45Sc, 49Ti, and 64Zn agree better with observation.
J. L. Fisker (Univ. of Notre Dame), E. Brown (Michigan State Univ.), M. Liebendoerfer, F.-K. Thielemann (Univ. of Basel), and M. Wiescher (Univ. of Notre Dame) investigated the rp-process reaction flow on an accreting neutron star and the resulting ashes of a type 1 X-ray burst based on a specialized version of the general relativistic hydrodynamics code AGILE.
M. Liebendoerfer has created a new and concise implementation of cubic domain decomposition with MPI for distributed memory computations in the three-dimensional MHD code developed by Ue-Li Pen, Phil Arras, and ShingKwong Wong. In order to achieve accurate internal energies in a fast and cool flow and at the same time correct energy dissipation in turbulence, a combination of the entropy and total energy equation is solved in a velocity splitting scheme inspired by Trac et al. 2004.
M. Liebendoerfer, U.-L. Pen, and C. Thompson performed preliminary three-dimensional simulations of stellar core collapse with magnetic fields on the McKenzie cluster. The simulations are based on the realistic Lattimer-Swesty equation of state and a parameterization of the relevant neutrino physics. More realistic initial data and improved boundary conditions will make the simulations more definitive.
There was further work refining the widely used Markov Chain Monte Carlo code CosmoMC for parameter estimation, developing new efficient proposal distributions and methods for handling 'fast' and 'slow' parameters. He organised a productive cross-disciplinary workshop with Radford Neal (Toronto): some of Neal's ideas were incorporated into CosmoMC. With Bond and Contaldi the code was applied to parameter forcasting for WMAP and Planck.
Antony Lewis continued studying CMB lensing, improving on a public simulation code and testing the CosmoMC parameter estimation pipeline on lensed CMB skies. He identified poor accuracy in the widely used first-order series expansion in the lensing potential, sufficient to bias parameter constraints with Planck. This lead on to work in late 2004 with Anthony Challinor (Cambridge) on a new improved calculation.
E. Thommes has been developing a hybrid code to study "planet-planet-disk" dynamics. It combines an existing N-body integrator (SyMBA, Duncan, Levison and Lee 1998) with a co-evolving one-dimensional viscous disk, coupling the two parts via azimuthally averaged prescriptions for planet-disk torques. This code reproduces a number of key features of planet-disk interaction found in full hydrodynamic simulations (e.g. migration and gap formation), but due to its simplicity it runs much faster, making possible simulations spanning the entire lifetime (up to 10 Myrs) of a protoplanetary gas disk.
During the early stages of planet formation, dust grains in a protoplanetary gas disk settle to the midplane and grow. Dust is thought to be quickly swept onto the central star as it approaches 1m in size (at ~1AU). Humble and collaborators model how various populations of dust grains settle and migrate using a 3D two-phase hydrodynamic (gas and dust) self-gravitating global disk code. Rapid migration and settling of dust can be seen in some areas of the disks, as can large enhancements in dust surface density. Investigations are continuing into whether (and when, where) the dust layer becomes thin and dense enough to undergo gravitational instability and directly form long lived ~1km planetesimals.
[ Back to CITA Research 2004 | CITA Annual Report 2004 ]