AMReX Astro Microphysics Networks

pynucastro can generate a C++ network that works in the AMReX-Astro Microphysics. This is done through the AmrexAstroCxxNetwork class. A simple C++ 3-alpha network that works with Microphysics can be created via:

import pynucastro as pyna

reaclib_library = pyna.ReacLibLibrary()

mylib = reaclib_library.linking_nuclei(["he4", "c12", "o16"])

net = pyna.AmrexAstroCxxNetwork(libraries=[mylib], rate_params=[r])
net.write_network()

The C++ network uses a set of template files in pynucastro/templates/amrexastro-cxx-microphysics/ which have placeholders of the form <func>(num). Here,

• func is a tag that corresponds to a function that will be called to write the output in this location

• num is the number of indentations for that line of code

In BaseCxxNetwork, there is a dictionary of function tags and the corresponding function called ftags. Any network class that derives from this can add to this list to specialize the output.

When the network is written, each template file is read and output line by line, injecting the code output by a function if the line has a function tag in it.

For the generation of the righthand side of the network itself, we use SymPy to build up an expression for each term and then use SymPy’s cxxcode function to translate it into C++ code. This is managed by the SympyRates class.

This will directly write out the C++ code into a collection of headers and source files. These are:

• actual_network.H

  This header simply stores the number of rates, and separately the number of ReacLib and tabular rates, as well as defines the arrays that hold the rate data.

• actual_rhs.H

  This defines the function to compute the energy release from the network as well as evaluate the rates and construct the “ydot” and Jacobian terms. Function tags are used to compute the screening factors, compute the tabular rates, and build the ydot and Jacobian, and include any plasma neutrino losses.

• derived_rates.H

  This computes any rates that are derived via detailed balance, with a function provided for each rate.

• interp_tools.H

  This contains interpolation functions used by tabulated rates.

• partition_functions.H

  This holds the partition function data (inline) and the functions to interpolate it.

• reaclib_rates.H

  This computes the ReacLib reaction rates, with a function provided for each rate.

• table_rates.H

  This manages TabularRate rates and interpolating the data. The rate data itself is stored inline in the header file.

  Warning

  We do not check if the thermodynamic conditions for evaluating the rate fall outside of the tabulation. Instead we simply extrapolate from the edge of the table. See Tabulated weak nuclear reaction rates for the bounds of the data.

• temperature_table_rates

  This manages TemperatureTabularRate rates and interpolating their data. The rate data itself is stored inline in the header file.

• tfactors.H

  This stores the struct that holds the different temperature factors used in the reaction rates.

Finally, there are a few meta-data files:

• pynucastro.net

  This lists the nuclei in the network and their properties in a format that Microphysics requires.

• Make.package

  This is a stub for the build system that list the files needed to build the network.

• _parameters

  This lists the runtime parameters that affect the network.

Using the network

To use the network, there are 2 options.

• You can use the unit tests in Microphysics/unit_test/, for example, burn_cell will do a single-zone burn.

• You can add the argument standalone_build=True to the write_network() function to output a basic main.cpp and GNUmakefile. This still requires a copy of Microphysics, and expect the environment variable MICROPHYSICS_HOME to point to its root directory. You can then build simply as:

  make
  

  and run as:

  ./main3d.gnu.ex testing.density=1.e6 testing.temperature=1.e9
  

  where the density and temperature are set via runtime parameters as shown. To control the screening, use the SCREEN_METHOD build parameter from Microphysics. For example, to disable screening, we can do:

  make SCREEN_METHOD=null