A bare bones implementation of a model for well-detached eclipsing binaries. The model is essentially a reimplementation of the light curve generator used in Irwin et al. (2011) on LSPM J1112+7626, which was a modified version of the Fortran subroutine "light" by Southworth et al. (2004 - 2009) and Etzel (1981).

The light curve generator has been rewritten in C, using the analytic method of Mandel & Agol (2002) for the quadratic limb darkening law. The implementation aims for nearly full machine precision, although does not quite achieve this at the present time. The treatment of ellipsoidal variation is based on Binnendijk (1974) and Etzel (1981), and the reflection effect on Milne (1926), Russell (1939), and Etzel (1981).

Most of the additions made in the 2011 paper have been carried over, including the spot model and the corrections for the "classical" light travel time across the system (in the solar system, the effect is called the Romer delay). The model can also compute light ratios and radial velocities.

A Python wrapper is provided, and is intended to be the primary way to use the model, but it can also be called directly from C and via a thin wrapper from Fortran. An experimental IDL wrapper is provided in the subdirectory "idl", but I don't use IDL so it has had very little testing.

An updated version of the Monte Carlo program (in C) from the 2011 paper is also provided in "ebmc". It has a number of dependencies that may make it tricky to get working and is therefore not included in the default set of Makefile targets. This is very bare bones, not general, messy, and has no user interface to speak of. It is not intended to be used as-is, but may serve as a place to borrow ideas from. It has accumulated a large number of changes since the version used for the paper, and therefore does not exactly reproduce the published results.

The file API.txt describes how to use the light curve generator, and gives the meanings of the parameters. Some practical notes on fitting real data intended to assist less experienced researchers are also included in the file HINTS.txt. Note that these documents are not an introductory text on eclipsing binary models, and it is assumed the reader is already familiar with this class of model.

Only a C99 compiler should be needed.

Python >= 2.6 Numpy >= 1.4 (advised; earlier versions may work)

Docstrings are embedded in the module, and should (hopefully) be sufficient to figure out how to use the Python API in conjunction with the examples (*.py) and the C API documentation. The "eb_" or "EB_" prefix used to avoid namespace pollution in C is not needed in the Python bindings and is therefore omitted.

My C subroutine library from "lib" on github. The Makefiles currently assume it's located in a directory "lib" at the same level as the "eb" directory (i.e. ../lib if this README is in the current directory).

MPFIT, C version http://www.physics.wisc.edu/~craigm/idl/cmpfit.html

The original program used levmar instead of MPFIT to provide the Levenberg-Marquardt minimizer. This is still supported (for now) as an alternative to MPFIT by defining USE_LEVMAR (see make.inc). http://users.ics.forth.gr/~lourakis/levmar/

PGPLOT >= 5.1.1 (optional, plots disabled if unavailable) http://www.astro.caltech.edu/~tjp/pgplot/

The colour Postscript ("cps") driver is assumed to be available. Everything else is optional. Successfully linking to PGPLOT can be tricky, particularly on Mac OS X. Some suggestions are included as comments in make.inc. To disable the plots, override PGPLOT_INC to remove the -DHAVE_PGPLOT.

The standard Makefile builds the light curve generator and Python module by default when you type "make". To build the IDL wrapper, type "make idl" (there is no need to type "make" first).

To build the fitter / Monte Carlo program, change to the ebmc directory and run "make". You may need to alter the top level make.inc file or override some of the variables so it can find the dependencies or to disable some features.

Automatic dependency generation for header files is supported, and recommended if you plan to update using git or make modifications. Run "make depend" before "make" to get this.

The default compiler options should produce a compatible binary. On x86-64, use of the FMA instruction set may be worthwhile (available on Intel Haswell / 4000 series Core CPUs or newer, AMD Piledriver or newer, including Zen CPUs such as Ryzen). If the compiled binary will only be run on CPUs with this feature available, adding -mfma to OPT in make.inc should improve performance slightly by using an optimized version of the elliptic integrals implemented using SIMD intrinsics.

The author gratefully acknowledges the work of Paul B. Etzel, Daniel Popper, John Southworth, and Alvaro Gimenez. These authors are thanked for their contributions to this research area, and for making their software available, upon which this model is based. Guillermo Torres is thanked for his extensive contributions to the development, testing and documentation of the present software.