The basic idea of this method is contour integration, with a new approach for connecting image solutions into continuous image tracks, which is presented in the paper: Light Curve Calculations for Triple Microlensing Systems.

This software contains some basic data structures of the VBBinaryLensing software, and some optimizations (e.g., the Quadrupole test) which were presented in V. Bozza et al., MNRAS 479 (2018) 5157, see also their relevant papers V. Bozza, MNRAS 408 (2010) 2188, and V. Bozza, E. Khalouei and E. Bachelet, arXiv:2011.04780.

The complex root solving algorithm is adopted from Jan Skowron & Andy Gould code (see their paper) and translated to C++ by Tyler M. Heintz and Ava R. Hoag.

After compile with "make" command, the test example can be run as:

./bin/testtriple

Here we show how the closed image tracks are connected:

In pymodule/python_bindings.cpp, please change the path of pybind11.h and stl.h (by "locate pybind11.h" in terminal). Then simply run the following to build the python module "TripleLensing".:

sudo python3 setup.py install
(or python3 setup.py install --user)

Please check the example notebook for the basic usage of this module.

Image topologies change as the source is moving (please see ./doc/top_lkv.mp4 for original movie):

This code is under more testing. We thank the feedbacks from the community and welcome more. One way to test is rerun MCMC optimization (use this code to calculate magnifications) with the published solutions of triple events as initial guess. Here is an example result on OGLE-2012-BLG-0026:

When the mass ratio is small (below ~ 1e-5), the solutions from the lens equation solver are more accurate when the origin of the coordinate system is set to be close to the smallest mass.