NBODY - astrophysics for star clusters, galactic nuclei, black holes

List of International Teams

• Morgan MacLeod, NASA Einstein Fellow, Institute for Advanced Study, morganmacleod@gmail.com
• Aaron Geller, Northwestern University CIERA & The Adler Planetarium, a-geller@northwestern.edu
• Sambaran Banerjee, Argelander Inst. For Astronomy, Univ. Bonn, sambaran@astro.uni-bonn.de
• Peter Berczik, Main Astronomical Observatory Acad. Of Sciences, Kiev, berczik@mao.kiev.ua
• Michael Fellhauer, Faculty Member, Univ. Concepcion. mfellhauer@astro-udec.cl
• Michele Trenti, Senior Lecturer and ARC Future Fellow, Univ. of Melbourne, michele.trenti@unimelb.edu.au
• Holger Baumgardt, Faculty Member, Univ. of Queensland, h.baumgardt@uq.edu.au
• Peter H. Johansson, Associate Professor of Astrophysics, Dept. of Physics, Univ. of Helsinki, peter.johansson@helsinki.fi
• Douglas C. Heggie and Phil Breen, Dept. of Maths. And Stats, Univ. of Edinburgh, d.c.heggie@ed.ac.uk, Phil.Breen@ed.ac.uk
• Marina Ryabova, Denis Ryabov, Southern Federal Univ., Rostov-on-Don, mryabova@sfedu.ru
• Enrico Vesperini, Faculty Member, Indiana Univ., evesperi@indiana.edu
• Giacomo Fragione, Racah Institute of Physics, Hebrew University of Jerusalem Israel, giacomo.fragione@mail.huji.ac.il
• Mirek Giersz, Nicolaus Copernicus Astronomical Center, Warsaw, mig@camk.edu.pl
• Ross Church, Research Fellow, Dept. Astronomy & Theoretical Physics, ross@astro.lu.se
• Roberto Capuzzo-Dolcetta, Professor, Univ. of Rome La Sapienza, roberto.capuzzodolcetta@uniroma1.it
• Maxwell Cai, Postdoc, Sterrewacht Leiden, cai@strw.leidenuniv.nl
• M.B.N. Kouwenhoven, Faculty Member, Xian Jiaotong Liverpool Univ. at Suzhou, t.kouwenhoven@xjtlu.edu.cn
• Nadine Neumaier, Max-Planck Institute for Astronomy, neumayer@mpia.de , team members at MPIA Alessandra Mastrobuono-Battisti and Arianna Picotti; and team of Glenn van de Ven with Anna Sippel
• Michela Mapelli, Astronomical Observatory of Padova, michela.mapelli@oapd.inaf.it
• Jarrod Hurley, Swinburne Univ. Melbourne, Australia, jhurley@swin.edu.au
• Christian Boily, Univ. of Strasbourg, France, christian.boily@astro.unistra.fr
• Ladislav Subr, Charles Univ. Prague, Czech Rep., subr@sirrah.troja.mff.cuni.cz

Dr. Thorsten Naab of the Max-Planck Institute for Astrophysics in Garching is a co-I on this project. Dr. Markus Rampp of the Max-Planck Computing and Data Facility will be a team member on the Max-Planck side as an expert for high performance and accelerated computational astrophysics, Markus Rampp

Objectives

• Direct NBODY code using individual, hierarchically blocked time steps in combination with high order (4th, 6th, 8th) time integrators
• Few Body Regularization Techniques for persistent binaries and hierarchical systems (binary or multiple stars in astrophysical terms)
• Ahmad-Cohen neighbour scheme using different time steps for irregular neighbour force (intermediate range) and regular force (long range)
• Highly detailed astrophysical prescription of stellar evolution for single and binary stars, including the development of their mass, radius, luminosity, temperature, and in case of binaries mass transfer, common envelope and coalescence; also including formation of compact remnants of stellar evolution, such as white dwarfs, neutron stars, stellar mass black holes and binaries containing these objects.
• Mass loss and time varying tidal field of the simulated star cluster in a galactic environment.
• CUDA/GPU acceleration for the regular forces, OpenMP or AVX/SSE acceleration for the neighbour forces

We find the following barriers for current and future use, in particular for a significant extension of the current user basis; in spite of these barriers we still have a currently large user basis, and it is expected that removal of some or all of these barriers will have a significant effect on the future user basis:

• the management of input and output stems from the time of Fortran 77 and should be upgrade by modern tools using python and web based applets on the surface and a suite of service programs beneath to provide user-selected astrophysical input models. The software Mcluster is already a step into this direction (see https://github.com/ahwkuepper/mcluster ). The author, Andreas Kuepper, does not support it anymore, but has allowed us to use parts or all of the code if useful.
• The internal data structure uses Fortran common blocks mainly, which is prohibitive for parallelisation (MPI, OpenMP, GPU). In the current parallel versions some data copy operations are used to cure this provisionally, but a generic upgrade of the data structure using structured more object oriented arrays (as they can now also be used in modern Fortran) is important to stay on top at the acceleration and parallelization The current structure of the code is highly non-modular. Dynamics and stellar evolution of single and binary stars are coupled on the smallest time scales. It is extremely difficult to replace one package by the other e.g. for stellar evolution or few-body integration. While this strong coupling has physical reasons, an effort should be made to improve the situation. A positive example is the AMUSE package, which has a very nice modular approach. NBODY6++ has been implemented as a python script within AMUSE (Maxwell Cai, Leiden, personal communication), but many functionalities of NBODY6++ cannot be fully used in this context yet.
• The final point to mention here in the objectives is not a barrier but an asset: The current code is based on 40 years of sustained software development, initialized and inspired by Sverre Aarseth, but most notably with contributions from many people over the decades such as Seppo Mikkola (regularizations), Jarrod Hurley and Chris Tout (stellar evolution), Keigo Nitadori (GPU implementation), and our team (parallelization with GPU), just to mention a few. All the teams listed in the end of Sect. 1., all the scientific results produced would not have been possible without the effort of the original developers (mostly astrophysicists) and many more people. It is the most prominent goal of this project to protect and sustain this previous work for the future, by transferring it into a more modern software and hardware environment. This is a fundamentally different approach to writing a new modern software, which other colleagues may want to do; our approach will cause difficulties and compromises, sacrifices, but we have shown in our past work that significant progress can be made while keeping the main code characteristsic intact. All software developed will be open-source and provided through github and/or local services at the Max-Planck Computing and Data Facility in Garching, the Computer Network and Information Center in Beijing, and the Astronomisches Rechen-institut at Heidelberg University.