GSOC 2026: Electrostatics beyond the atomic charges

[ccastilloo]

📚 Package Description and Impact

FFprime is a Python library designed to derive and evaluate the accuracy of atom-centered force-field parameters from quantum-chemical calculations. The package currently supports the computation of atomic charges, dipoles, and quadrupoles using physically motivated partitioning schemes based on the atom-in-molecules theory.

Electrostatic interactions govern many natural and biological processes and are among the most important contributions to inter- and intramolecular interactions. While most standard biomolecular force fields approximate electrostatics using fixed atomic charges (monopoles), this approximation neglects the anisotropy of the molecular electrostatic potential. We recently demonstrated that including higher-order atomic multipoles derived from specific AIM methods provides a systematic improvement while maintaining computational efficiency https://pubs.rsc.org/en/content/articlelanding/2026/ra/d5ra07866k.

This project aims to extend FFprime by implementing efficient and well-tested routines to compute electrostatic potentials and electric fields generated by atom-centered multipoles (charges, dipoles, and quadrupoles), enabling their direct use in early analysis for future force-field development.

👷 What will you do?

The student will design and implement functions in FFprime to compute:

• The electrostatic potential generated by atomic multipoles (monopoles, dipoles, quadrupoles)
• The electric field derived from these multipoles at arbitrary spatial points

The work will also include:

• Integrating the new functionality with all the required utility functions into FFprime
• Writing unit tests to validate correctness against analytical results and reference calculations

🏁 Expected Outcomes

1. New FFprime modules/functions to compute electrostatic potential and electric field from atomic multipoles.
2. Support for monopole, dipole, and quadrupole contributions
3. Well-documented, tested, and maintainable code
4. Example use cases illustrating the improvement over charge-only electrostatics

Required Skills

Skill	Description
Requied Skills	Python, Scientific OOP, Linear algebra & vector calculus
Preferred Skills	Computational chemistry or molecular modeling, Familiarity with molecular simulations and force fields
Project size	Small
Difficulty	Medium

🙋 Mentors

Name	Email	GitHub
Carlos Castillo-Orellana	ccastilloo1995_at_gmail_dot_com	@ccastilloo
Farnaz Heidar-Zadeh	farnaz.heidarzadeh_at_queensu_dot_ca	@FarnazH
Omid Hossainzadeh	ohosseinzade0_at_gmail_dot_com	@ohzde
--

🏋️ Stretch Goal

Build analysis utilities on top of the electrostatic field implementation, such as projecting the field onto bonds or reaction coordinates, comparing fields using cosine similarity or other metrics.

[rohan-019]

Hi @ccastilloo @FarnazH,

Thank you for outlining this extension so clearly — extending FFprime to compute electrostatic potentials and electric fields from AIM-derived multipoles is a very natural and impactful next step.

After reviewing the repository structure (particularly nb.py, utils/potentials.py, and the existing tests), I’m very interested in contributing to this as a GSoC participant this year. The focus on accurate tensor-based implementations and analytical validation strongly aligns with my interest in scientific Python and numerical methods.

I plan to begin with a clean, vectorized implementation of the monopole and dipole contributions, accompanied by analytical unit tests for correctness and stability. Would this be a reasonable starting point for the project?

I’d appreciate any guidance on recommended initial milestones as I prepare my GSoC proposal.

Looking forward to your feedback.

[AkshatSharma5]

Hi @ccastilloo @FarnazH @ohzde ,
I’m Akshat, a developer from IIT Indore with strong experience in Python, scientific computing, and linear algebra, and I’m very interested in contributing to FFprime for GSoC 2026.

I’ve gone through the project description and paper, and I find the multipole-based electrostatics extension technically exciting. I’d love to get involved, explore the codebase, and start contributing wherever helpful.

Looking forward to learning from you and contributing to FFprime.

Thanks!

[imfarhan0105-bits]

Hi @ccastilloo, @FarnazH, @ohzde,

I’m very interested in contributing to the “Electrostatics beyond atomic charges” project for GSoC 2026.

I’m currently pursuing an MSc in Chemistry alongside a BE in Mechanical Engineering. My academic background includes electrostatics, multipole expansions, intermolecular interactions, and quantum chemistry, and I’ve also worked on projects involving machine learning and deep learning. I’m particularly interested in building strong physics-based foundations that can later interface with data-driven approaches in computational chemistry.

Extending FFprime to compute electrostatic potentials and electric fields from atom-centered multipoles strongly aligns with my interest in translating physically rigorous models into reliable scientific software.

I’ve set up the repository locally and started reviewing the current structure for handling atomic parameters. Before drafting a detailed proposal, I would greatly appreciate any guidance on architectural expectations, preferred design direction, or recommended starter tasks to ensure my proposal aligns closely with your long-term vision for FFprime.

I also have a long-term interest in working with QCDevs to explore deeper into the field of computational chemistry and emerging machine learning approaches in chemistry.

Looking forward to contributing.

Best regards,
Farhan

[anandvenugopal-tech]

Hi @ccastilloo, @FarnazH, @ohzde

I’m excited about the Electrostatics Beyond Atomic Charges project. I have completed M.Sc in chemistry from NIT Trichy and worked in computational chemistry project for one year. I am also skilled in machine learning makes this particularly appealing, especially the goal of moving past atom-centered point charges toward richer models like multipoles, off-center charges, and ML-learned electrostatic potentials.

I’ve been exploring approaches to derive multipole moments from QM data, ESP map fitting, and how these improved electrostatics influence binding and conformational behavior. I’m preparing a proposal focused on implementing a new parametrization pipeline and benchmarking against traditional FFs. Any suggestions on priority areas would be very helpful.

Looking forward to contributing.

Thanks,
Anand Venugopal

[opalclouds]

Hi @ccastilloo , @ohzde, @FarnazH !

I’m Shreya Gupta, a chemical engineering graduate from IIT BHU with experience in Python , machine learning and deep earning.

I’m very interested in contributing to FFprime, particularly extending it to compute electrostatic potentials and fields. I’d like to explore integrating lightweight AI/ML approaches to predict multipole contributions or optimize computations for faster force-field evaluations.
Looking forward to contributing!

Thanks,
Shreya