M11po

A new density functional, M11po, that makes a more accurate class of chemistry methods ("rung 3.5") easy to run in PySCF, a free, open-source chemistry package. Published in The Journal of Chemical Physics.

What it is

Chemists use Density Functional Theory (DFT) to predict how molecules behave (their energies, structures and reactions) without running an experiment. How accurate DFT is depends on one ingredient, the functional, and new functionals keep being proposed.

Functionals are often ranked on a ladder, where each higher rung uses more information about the electrons. “Rung 3.5” functionals sit between two of the standard rungs. They handle static correlation and electron delocalization in a better balance, two effects that simpler functionals struggle to get right together. Until now, though, using them has meant making complex changes to a chemistry program’s code.

M11po is a rung-3.5 functional built to avoid that. It runs in standard Gaussian-basis codes like PySCF without complex modifications, and performs comparably to M11plus, an existing rung-3.5 functional.

What I did

I’m the first author of the paper, working under the supervision of Dr. Benjamin Janesko at Texas Christian University, with Yinan Shu, Dayou Zhang and Donald G. Truhlar.

Implement rung-3.5 functionals, including M11po, in Python on top of PySCF
Benchmark them on GMTKN55, a collection of 55 datasets (nicknamed the “chemical zoo”) covering thermochemistry, reaction barrier heights and non-covalent interactions

How it works

Code: pyscf-r35dft, an extension of PySCF’s dft module
Semilocal part: a fork of the libxc library, libxc-r35dft
Benchmark: GMTKN55, scored by mean absolute error against existing functionals
License: Apache-2.0

Status

Published: Morgan, Janesko, Shu, Zhang & Truhlar, “Convenient implementation of rung-3.5 density functionals in standard Gaussian-basis codes,” The Journal of Chemical Physics 165, 084110 (2026). The code is open source on GitHub.