Machine learning (ML) has recently gained rapidly in importance. Driven by new algorithms, increasing computing power, and the availability of vast amounts of data, ML has established itself as a useful tool across many disciplines. In thermodynamics ML can also be employed, for example to predict unknown properties— often yielding results that surpass those of established physical approaches. 

My work focuses on the intersection of physical models describing thermodynamic properties and machine learning methods. I aim to determine, among other things, whether and how ML techniques can reveal previously unknown patterns in data, and how physical modeling and ML can mutually benefit from one another through the development of novel hybrid models.

Teaching Assistance

• Thermodynamics I (WS 2023/24)
• Thermodynamics of mixtures (WS 2024/25)
• VLE-Laboratory (SS 2025)
• Heat Transfer(WS 2025/26)

Office Hours

By individual appointment

Publications

• M. Hoffmann, T. Specht, Q. Göttl, J. Burger, S. Mandt, H. Hasse, F. Jirasek: Thermodynamically consistent machine learning model for excess Gibbs energy, Nat. Commun. 17, 3485 (2026) .
• M. Hoffmann, T. Specht, N. Hayer, H. Hasse, F. Jirasek: MLPROP – An Interactive Web Interface for Thermophysical Property Prediction with Machine Learning, Chem. Ing. Tech. 97, 1052-1056 (2025) . [DOI]
• M. Hoffmann, H. Hasse, F. Jirasek: GRAPPA—A hybrid graph neural network for predicting pure component vapor pressures, Chem. Eng. J. Adv. 22, 1000750 (2025). [DOI]
• M. Hoffmann, N. Hayer, M. Kohns, F. Jirasek, H. Hasse: Prediction of pair interactions in mixtures by matrix completion, Phys. Chem. Chem. Phys. 26, 19390-19397 (2024). [DOI]
• M. Heinen, M. Hoffmann, F. Diewald, S. Seckler, K. Langenbach, J. Vrabec: Droplet coalescence by molecular dynamics and phase-field modeling,
  Physics of Fluids 34, 042006 (2022). [DOI]

Presentations

• M. Hoffmann, T. Specht, H. Hasse, F. Jirasek: A Thermodynamically Consistent Machine Learning Model for the Gibbs Excess Energy of Mixtures, Thermodynamik-Kolloquium, Hamburg, 22.-24.09.2025.
• M. Hoffmann, T. Specht, S. Mandt, H. Hasse, F. Jirasek: Thermodynamically Consistent Prediction of Vapor-Liquid Equilibria Using Machine Learning, Thermodynamik-Kolloquium, Stuttgart, 25.-27.09.2024.
• M. Hoffmann, D. Gond, H. Hasse, F. Jirasek: Prediction of Pure Component Vapor Pressures with Graph Neural Networks, 22nd Symposium on Thermophysical Properties, Boulder (CO), USA, 24.-28.06.2024.
• M. Hoffmann, N. Hayer, M. Kohns, F. Jirasek, H. Hasse: Prediction of Unlike Interaction Parameters for Molecular Simulations by Matrix Completion, Thermodynamik-Kolloquium, Hannover, 25.-27.09.2023.

Posters

• M. Hoffmann, T. Specht, N. Hayer, H.Hasse, F. Jirasek: MLPROP – An Open Interactive Web Interface for Thermophysical Property Prediction with Machine Learning, Thermodynamik-Kolloquium, Hamburg, 22.-24.09.2025.
• M. Hoffmann, H.Hasse, F. Jirasek: GRAPPA - Graph neural network for predicting pure component vapor pressures, International Conference on Properties and Phase Equilibria for Product and Process Design, Bad Gögging, Germany, 11.-15.05.2025.
• M. Hoffmann, H.Hasse, F. Jirasek: Graph Neural Networks with Molecular Interaction Pooling for Pure-Component Vapor Pressure Prediction, European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases, Vilnius, Litauen, 07.-11.09.2024.