Hands-on Workshop on Computational Astrophysics
February 4-9, 2013
Over the last several years we have been working on the development of a generalized tool for speeding-up computational astrophysics code development. Our framework, called the Astrophysics Multipurpose Software Environment (AMUSE), enables researchers, but also advanced students, to quickly write production quality codes for performing simulations. Our environment is publicly available via http://amusecode.org.
In the workshop we aimed to bring experienced researchers together with graduate students and young postdoctoral level researchers to work together with expert software developers on key problems in multi-scale and multi-physics computational astrophysics. These problems would be implemented as scripts in the AMUSE framework. The AMUSE development team was available to assist in the numerical implementations.
The workshop was attended by around 22 participants divided into 6 teams with one experienced researcher, a junior researcher or graduate student and a member of the AMUSE development team. As two teams interests overlapped the participants decided to merge the team into one and work on their problem from 2 different angles.
The main program consisted of a short talk in the morning to assemble all participants in the room. The rest of the day was focused on working on the problems. At the end of each day we would have a short planery session where the progress of each team was reported and discussed.
The first day of the workshop we formed the teams and (re-) defined the problems. What follows is a list of the problems we tried to tackle, and the impressions of the team members:
1) Modeling planetary systems in star clusters
We wanted to simulate a planetary system around each star in a small cluster. We used mercury to simulate each planetary system with ph4 to simulate the cluster. We had no previous experience with AMUSE. After some installation problems and with some help from the AMUSE team we were able to write a working script and could show the cluster evolving at the end of the week.
2) Handling of multiples in a stellar cluster evolution
We worked on making the handling of close encounters and binary formation by the multiples modules module in AMUSE production ready. Sitting close together we could quickly implement some new schemes and test the resulting simulation. The multiple module improved significantly during the workshop.
3) The retention of stellar winds in intermediate age star clusters.
We combined Evtwin with the Fi sph hydrodynamics code
4) Ionizing feedback from massive stars in a molecular cloud
5) Dynamics of self gravitating systems
The work done at the Meeting was pursuing two aims: one scope has been preparing a friendly driver to generate initial conditions for N-body simulations from a given density distribution in spherical symmetry. Capuzzo-Dolcetta prepared the mathematical scheme and D. Punzo prepared a Python program to do this, with the collaboration of I. Pelupessy. The final objective of this part of the work is to provide a subroutine available through the AMUSE environment. Another more applicative aim was to recheck some results already obtained by Capuzzo-Dolcetta and Spera (2013, in prep.) showing that a violent collapse of an N-body system (N=512) initially unsegregated in mass leads to mass segregation on the collpase time scale, which is essentially the crossing time scale when starting from "cold" (zero virial ratio) initial conditions sampled in a homogeneous sphere. We used both the HiGPUs code (6th order Hermite's code by Capuzzo-Dolcetta, Punzo and Spera) and the PH4 code (4th order Hcermite's ode by McMIllan) both operating in the AMUSE framework. We considered two populations of stars of different mass and the presence of a black hole. We noted the high speed and precision of HiGPUs, which was run on a laptop giving rapidly a set of simulation results which allow us to state that the violent mass segregation effect is a real feature, at least in the case studied. This part of work was mainly done by I. Pelupessy.
Finally we want to highlight the many comments we received about the ease of use and potential of the AMUSE framework. Several of the attending graduate students have continued their work with AMUSE. A number of arrangements were made for future collaborations.
The workshop organizers are very grateful to the Lorentz Center team that supported this workshop. Their skill and professionalism showed during all phases of the workshop organisation. As scientific organisers organizing the workshop was a "breeze". Further, the new Snellius location provided an ideal environment for productive interactions within and between the different teams.
Simon Portegies Zwart (Leiden University, Netherlands)
Steve McMillan (Drexel University, Philadelphia, United States)
Arjen van Elteren (Leiden University, Netherlands)