|
|
|
|
This page outlines the general directions in which the code is now being developed. Plasma simulations are a research field of their own. A conference, published as an ASP Conference volume 247, Spectroscopic Challenges of Photoionized Plasmas, Ferland & Savin, editors, my review article in 2003 Annual Reviews Astronomy & Astrophysics, 41, 517 (available here), and the graduate text Osterbrock & Ferland Astrophysics of Gaseous Nebulae and Active Galactic Nuclei (the publisher's web site is here) summarize the current state of plasma codes and the physics of ionized gas and what needs to be done next. The goal is to do a complete simulation of what happens in nature. The capabilities of the code have always been limited by available computer power. As machines grow more powerful it becomes possible to increase the physical fidelity of the simulation. The highest priority has been to include physics processes that affect the physical state of the gas - that is, its kinetic temperature, ionization, and chemical state, together with its observed spectrum. The following lists directions in which Cloudy is now being developed.
Next major releaseWe now plan to make new releases at the start of every year. The next planned release is January 2007 (version 07.01). The code is now "clean C", meaning that the source files can be renamed to "*.cpp" and compiled as a C++ code. The code will officially become a C++ code around November of this year. A beta of the next release, the first as a C++ code, will follow in December (version 06.12). We will then discontinue updates of the current gold version, 06.02. The C++ version will not remain backwards compatible with ANSI C. I would be interested in any thoughts or comments on the code's future in C++, especially if this might cause any problems that we have not thought of yet. Please post comments on the discussion board.
Newly completed! Time-dependent ionizationThe code was able to do time-dependent calculations early in its history (Ferland & Truran 1981, ApJ 244, 1022-1032). This was not developed along with the rest of the infrastructure but has now been revived. The effects of a time-variable continuum source are again included, a project done in collaboration with Will Henney and Robin Williams and an outgrowth of Henney et al. (2005, ApJ, 621, 328, on the ADS here). Nick Abel has created a pair of animated gifs showing the time evolution of an H+ region with PDR after its ionizing star is turned off and then turned back on. The first animation shows the cloud recombining and becoming molecular and the second shows the hydrogen ionization front moving across the cloud. Time dependent continuum sources will be included in the next major upgrade, due in early 2007.
The atomic - molecular data baseThis is listed as a major item due to the large amount of ongoing human effort involved. The conditions in a non-equilibrium plasma are determined by the underlying microphysics. This in turn rests on the foundation of basic atomic and molecular cross sections and rates. The data handling needs of just keeping Cloudy current with the atomic database are extensive. Roughly 100 papers appear each year with new cross sections or rates (a web search on the Astrophysics part of ADS will find about a third of these – many more are in the Physics and Chemistry literature). Cloudy now predicts the intensities of well over 106 emission lines. New oscillator and collision strengths appear for roughly 104 transitions per year, and it is necessary to produce temperature fits to the collision strengths. Extensive sets of photoionization cross-sections with autoionization resonances and experimental / theoretical recombination rates are now appearing. The atomic data must be continuously updated just to stand still. Finally, large changes in the atomic rates may affect the stability of the code's solvers or change important predictions. Any change to the code has a finite possibility of introducing error, so all changes must be crosschecked. All of this is labor intensive but does not, in itself, result in publications.
Dynamics & ShocksAll clouds are actually dynamical flows of some sort. In collaboration with Robin Williams, Will Henney, and Jane Arthur, I am incorporating hydrodynamics and time-dependent rates and advection, with the initial goal of simulating the flows seen in star-forming regions. This project is described in The Hydrodynamics of Photoionized Flows, 2002, G.J. Ferland, W.J. Henney, R.J.R. Williams, S.J. Arthur, Rev Mex AA (Serie de Conferencieas), 12, 43-49, available here., and in Self-Consistent Dynamic Models of Steady Ionization Fronts. I. Weak-D and Weak-R Fronts by Henney, Arthur, Williams & Ferland, 2005, ApJ, 621, 328, available from astro-ph here. The goal is to have an essentially zero-free parameter description of the evolution of these systems. Once the dynamics work is complete and fully stable it will be straightforward to extend the treatment to shocks. The current development is being done in a manner to facilitate this extension.
Exact radiative transferNearly all plasma codes now use escape probabilities to do line and continuum transport due to limits imposed by available processor power. Computers are now becoming powerful enough to do the radiative transport with exact methods. Most sources are optically thick in the Lyman continuum and some lines, while the line radiative transfer effects are important in most sources. From the beginning, the intention was to eventually incorporate exact radiative transfer methods in Cloudy, and the code was designed to receive this. Work is now underway to prepare the infrastructure for this extension.
Higher order dimensionalityThe code is currently 1D. Higher order dimensionality mainly affects the transfer of the diffuse fields, which in turn affects predictions at the 10 - 15% level. The exact radiative transfer approach will use ALI with short characteristics, and the extension to 2D or 3D geometries will be straightforward.
Parallel machinesOpenMP and MPI are platform-independent standards for doing multiprocessor calculations on shared-memory machines. Clock speeds of processes have stopped increasing at the Moore's Law rate in the past few years, but the transistor count has continued to increase. The extra power is going into hyperthreading and multi-core processors, power that can only be used by a parallel code. It now takes roughly half a day to do a static H II region / PDR model with the full model of the hydrogen molecule included. It takes roughly 100 iterations to converge a dynamical model of a blister H II region. These would require many weeks on a serial processor. The code will be developed for use on parallel machines while still running on serial.
|
