McGroundstate: Instance Submission and Solution Server for Ising Spin Glass Instances

This page offers a service to compute the energetic groundstates of Ising Spin Glass problems, in particular those on
  • two- or three-dimensional regular lattices (free or periodic boundaries).
For other types, the Spin Glass Server might currently be a more convenient option.

This service is run jointly by Jonas Charfreitag, Michael Jünger, Sven Mallach, and Petra Mutzel.

Instances can be uploaded using the interface below. If a submitted instance is valid and of reasonable size, and if you provide a valid email-address, the instance will be passed to our solvers and either an optimal solution or, if a time limit is exceeded, intermediate results will be sent to the address specified. The format of the results files is described here. Please note that this service comes without any guarantee or warranty. If you have questions, please contact mcsparse [ at ]

Submit your instance ...

Each instance has to be either an .sg or a .gsg file as defined here.
For submission, such files can also be bundled into the top-level (no directories) of a tar.gz or zip archive.
By submitting an instance, you express your consent that it may be freely used by the project members, and you confirm that you use this service solely for non-profit scientific or private intent.
email address

If you use our service in connection with a scientific publication, please refer to McGroundstate in the text and include the following BibTeX citation:

author = {Jonas Charfreitag and Michael J{\"u}nger and Sven Mallach and Petra Mutzel},
title = {{M}c{S}parse: {E}xact Solutions of Sparse Maximum Cut and Sparse Unconstrained Binary Quadratic Optimization Problems},
editor = {Cynthia A. Phillips and Bettina Speckmann},
booktitle = {2022 Proceedings of the Symposium on Algorithm Engineering and Experiments ({ALENEX})},
year = {2022},
pages = {54--66},
doi = {10.1137/1.9781611977042.5}

For general MaxCut or Binary Quadratic Programming instances, please consider using McSparse or, especially for dense problems, BiqMac, run by our partners Franz Rendl, Giovanni Rinaldi, and Angelika Wiegele, as well as the further solvers BiqBin and BiqCrunch.

This work is supported by the transdisciplinary research network "TRA 1 Modelling" at the University of Bonn as part of the Excellence Strategy of the federal and state governments, and by the Hausdorff Center for Mathematics (HCM) funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) under Germany’s Excellence Strategy EXC-2047/1 390685813. In particular, we are grateful to the HCM for financing compute infrastructure employed to provide this service to the scientific community.