Andreas Ipp

Research Group

Lattice Gauge Theory & Machine Learning Group

We are a research group at the Institute for Theoretical Physics of TU Wien (Vienna, Austria), led by Andreas Ipp (Principal Investigator, ORCID 0000-0001-9511-3523). Our research combines fundamental theoretical physics — in particular lattice gauge theories and the physics of heavy-ion collisions — with modern machine learning methods such as symmetry-preserving (equivariant) neural networks and generative diffusion models.

Research topics

• Machine learning for lattice gauge theory: Lattice Gauge Equivariant Convolutional Neural Networks (L-CNNs) with exact gauge symmetry, machine-learned classically perfect fixed-point actions, HMC simulations and classically perfect gradient flows.
• Glasma and heavy-ion collisions: (3+1)D dilute Glasma initial conditions, real-time lattice simulations of the early stages of heavy-ion collisions, jet and heavy-quark momentum broadening.
• Generative models for physics: equivariant diffusion models for non-Abelian lattice gauge theory, neural gradient flow, and physics-driven learning for inverse problems in QCD.

Group members

We regularly have openings for Bachelor, Master and PhD students — see teaching or contact ipp@hep.itp.tuwien.ac.at.

Recent graduates

Former members

Funding

• FWF project PAT3667424 — "Machine Learning with Renormalization Group for Lattice QCD" (2025–2029, PI: A. Ipp)
• FWF project P 34764-N27 — "Impact of glasma on heavy flavor azimuthal correlations and spectra" (2021–2026)
• FWF project P 32446-N27 — "Upscaling Glasma simulations using machine learning" (2019–2024)
• FWF project P 26582-N27 — "Exploring quark-gluon-plasma dynamics with CPIC simulations" (2014–2019)

Selected recent publications

• K. Holland, A. Ipp, D. I. Müller, U. Wenger, Machine-Learned Renormalization-Group-Improved Gauge Actions and Classically Perfect Gradient Flows, Phys. Rev. Lett. 136 (2026) 031901 (PRL, arXiv:2504.15870)
• G. Aarts, D. E. Habibi, A. Ipp, D. I. Müller, T. R. Ranner, L. Wang, W. Wang, Q. Zhu, Generalizable Equivariant Diffusion Models for Non-Abelian Lattice Gauge Theory (2026), arXiv:2601.19552
• G. Aarts, K. Fukushima, T. Hatsuda, A. Ipp, S. Shi, L. Wang, K. Zhou, Physics-driven learning for inverse problems in quantum chromodynamics, Nature Reviews Physics 7 (2025) 154, arXiv:2501.05580
• A. Ipp, M. Leuthner, D. I. Müller, S. Schlichting, K. Schmidt, P. Singh, Energy-momentum tensor of the dilute (3+1)D Glasma, Phys. Rev. D 109 (2024) 094040, arXiv:2401.10320
• K. Holland, A. Ipp, D. I. Müller, U. Wenger, Machine learning a fixed point action for SU(3) gauge theory with a gauge equivariant convolutional neural network, Phys. Rev. D 110 (2024) 074502, arXiv:2401.06481
• M. Favoni, A. Ipp, D. I. Müller, D. Schuh, Lattice gauge equivariant convolutional neural networks, Phys. Rev. Lett. 128 (2022) 032003, arXiv:2012.12901

Full list: Publications

Supervised theses

• Kayran Schmidt — PhD 2026, (3+1)D dilute Glasma initial conditions in simulations of heavy-ion collisions, reposiTUm
• Markus Leuthner — PhD 2025, Aspects of the dilute Glasma, reposiTUm
• Matteo Favoni — PhD 2024, Symmetry-preserving neural networks in lattice field theories, reposiTUm
• Felix Schlemmer — Diplom 2025, Using modern equivariant machine learning architectures as effective Hamiltonians for Monte Carlo simulations of quantum spin systems, reposiTUm
• Thomas R. Ranner — Diplom 2024, Sampling lattice field theories with score-based diffusion models, reposiTUm
• Lukas Einramhof — Diplom 2022, Group equivariant neural networks for a scalar field theory, reposiTUm
• Srinath Bulusu — Diplom 2021, Supervised learning in φ⁴ lattice field theory, reposiTUm
• Patrick Kappl — Diplom 2018, Implementing a 3+1-dimensional simulation for a color glass condensate model in C++, reposiTUm
• Peter Somkuti — Diplom 2011, Modifications of spectra and correlations through yoctosecond photon double pulses from a quark-gluon plasma, reposiTUm

Open-source software

• lge-cnn — Lattice Gauge Equivariant CNNs (L-CNNs), fixed-point actions and HMC simulations
• OpenPixi — simulation framework for the initial stages of heavy-ion collisions

Collaborations

We collaborate with the groups of K. Holland (University of the Pacific) and U. Wenger (University of Bern) on machine-learned fixed-point actions, with S. Schlichting (Bielefeld) and P. Singh on Glasma physics, and we participate in international community initiatives on machine learning and large physics models.

• DM-QFT Collaboration — theoretical physicists working at the interface of generative diffusion models and quantum field theory, from non-perturbative lattice methods to physics-driven learning for QFT.
• EuCAIF
  — European Consortium for AI in Fundamental Physics.
• COST Action AIFORIA (CA25172) — AI for Fundamental Physics: Open Roadmapping and Inclusive Actions.
• SCIANCE — AI in Science initiative (AI in Science Working Group “Astronomy and Fundamental Physics”).

Community & service

• EuCAIF — Fellow of the European Consortium for AI in Fundamental Physics, EuCAIF members.
• COST Action AIFORIA (CA25172) — member of the Management Committee (Austria), cost.eu/actions/CA25172.
• SCIANCE — member of the AI in Science Working Group “Astronomy and Fundamental Physics”, sciance.eu/aiswg-members.