Tibor Szilvási
@SzilvasiGroup
Followers
845
Following
965
Media
21
Statuses
306
Assistant Professor at The University of Alabama Computational Chemistry | Catalysis | Material Design
Tuscaloosa, AL
Joined October 2020
Our paper on how to train transferable water potentials using equivariant neural networks is out in JPCL! @JPhysChem .Great work @tgmaxson! .Thanks for the funding @doescience!.
pubs.acs.org
Machine learning interatomic potentials (MLIPs) have emerged as a technique that promises quantum theory accuracy for reduced cost. It has been proposed [J. Chem. Phys. 2023, 158, 084111] that MLIPs...
0
5
52
For researchers looking to choose an MLIP architecture, we suggest selecting equivariant MLIP architectures if the complexity of the system is a challenge.
0
0
1
Moving forward, we recommend benchmarking efforts shifting focus from marginal accuracy improvements in energy and force errors toward identifying and understanding model failure modes, assessing transferability, and evaluating how their errors affect observable predictions.
1
0
0
- The HEA and Zr–O data sets are identified as challenging tests for future benchmarks and MLIP model architecture developments as they show significant differentiation in error between MLIP architectures.
1
0
1
Key points:. - Our analysis highlights that low errors in energy and force predictions do not guarantee reliable observables. - Equivariant MLIPs offer 1.5–2× improvements over non-equivariant MLIPs in energy and force error for structurally or compositionally complex systems.
1
0
1
We tested five MLIP architectures (MACE, NequIP, Allegro, MTP, and Torch-ANI), focusing not only on traditional metrics (energies, forces, and stresses) but also explicitly validating derived physical observables.
1
0
1
MS25 presents diverse materials-relevant systems including MgO surfaces, liquid water, zeolites, a catalytic Pt surface reaction, high-entropy alloys (HEAs), and disordered Zr-oxides.
1
0
1
Our materials science-focused MLIP benchmark dataset that goes beyond energy and forces is out in @JCIM_JCTC. Congrats, @tgmaxson, @soyemiademola, @XZhangChem, and Benjamin Chen!.
pubs.acs.org
We present MS25, a benchmark data set for evaluating machine learning interatomic potentials (MLIPs) across diverse materials-relevant systems including MgO surfaces, liquid water, zeolites, a...
3
22
68
Our contribution to the Athanassios Z. Panagiotopoulos Festschrift is now online. We show how to model complex aqueous electrolytes with realistic (0.1 M) concentrations at DFT-quality in agreement with experiments.
pubs.acs.org
Understanding the structure and thermodynamics of solvated ions is essential for advancing applications in electrochemistry, water treatment, and energy storage. While ab initio molecular dynamics...
2
3
35
Interested in constant chemical potential simulations at interfaces? We developed a simple method that allows DFT-quality constant chemical potential simulations. Congrats @soyemiademola. Thank you @ENERGY for funding.
arxiv.org
Chemical potential of species in solution is essential for understanding various chemical processes at interfaces. Molecular dynamics (MD) simulations, constrained by fixed compositions, cannot...
0
3
17
New study is out on the depolymerization of dehydrochlorinated polyvinyl chloride using our multiscale simulation framework (MUSIK). Our work explains why depolymerization is incomplete and provide guidance on how to improve by catalyst design.
0
2
13
New paper is out in Journal of Catalysis. We wrote a comprehensive review on the application of machine learning potentials in heterogeneous catalysis. Congrats Jide, Khagendra, Sophia, and Ademola!.
0
4
18
If you are interested in catalysis and machine learning, we wrote a comprehensive review on the application of machine learning potentials in heterogeneous catalysis.
3
14
55
RT @secatsoc: Social Media Chair: Jason Bates (@jasonsbates).Communication Manager: Md Masudur Rahman.Directors: Fanxing Li and John Kuhn.R….
0
1
0
Congrats to all the students!.If you are wondering, I am on the picture because I serve as the President of SECS, not because I started another PhD.
The SECS held another successful Annual Symposium on Feb 10-11! Excellent talks and posters were given by our students. Several students were awarded "Top Oral Presentation" and "Top Poster",.Top Oral Presentation:.Sijie Guo - UTK.Sara Haidar - VT.Suchi Vijayaraghavan - GT
1
0
10
Our framework can match experimentally observed reaction times within an order of magnitude without any parameter estimation thanks to affordable high-level coupled cluster calculations.
0
0
2