-
Abelian $p$-groups with a $p$-bounded factor or a $p$-bounded subgroup
Authors:
Justyna Kosakowska,
Markus Schmidmeier,
Martin Schreiner
Abstract:
In his 1934 paper, G.Birkhoff poses the problem of classifying pairs $(G,U)$, where $G$ is an abelian group and $U\subset G$ a subgroup, up to automorphisms of $G$. In general, Birkhoff's Problem is not considered feasible. In this note, we fix a prime number $p$ and assume that $G$ is a direct sum of cyclic $p$-groups and $U\subset G$ is a subgroup. Under the assumption that the factor group…
▽ More
In his 1934 paper, G.Birkhoff poses the problem of classifying pairs $(G,U)$, where $G$ is an abelian group and $U\subset G$ a subgroup, up to automorphisms of $G$. In general, Birkhoff's Problem is not considered feasible. In this note, we fix a prime number $p$ and assume that $G$ is a direct sum of cyclic $p$-groups and $U\subset G$ is a subgroup. Under the assumption that the factor group $G/U$ is $p$-bounded, we show that each such pair $(G,U)$ has a unique direct sum decomposition into pairs of type $(\mathbb Z/(p^n),\mathbb Z/(p^n))$ or $(\mathbb Z/(p^n), (p))$. On the other hand, if we assume $U$ is a $p$-bounded subgroup of $G$ and in addition that $G/U$ is a direct sum of cyclic $p$-groups, we show that each such pair $(G,U)$ has a unique direct sum decomposition into pairs of type $(\mathbb Z/(p^n),0)$ or $(\mathbb Z/(p^n), (p^{n-1}))$. We generalize the above results to modules over commutative discrete valuation rings.
△ Less
Submitted 3 December, 2023;
originally announced December 2023.
-
Implicit Transfer Operator Learning: Multiple Time-Resolution Surrogates for Molecular Dynamics
Authors:
Mathias Schreiner,
Ole Winther,
Simon Olsson
Abstract:
Computing properties of molecular systems rely on estimating expectations of the (unnormalized) Boltzmann distribution. Molecular dynamics (MD) is a broadly adopted technique to approximate such quantities. However, stable simulations rely on very small integration time-steps ($10^{-15}\,\mathrm{s}$), whereas convergence of some moments, e.g. binding free energy or rates, might rely on sampling pr…
▽ More
Computing properties of molecular systems rely on estimating expectations of the (unnormalized) Boltzmann distribution. Molecular dynamics (MD) is a broadly adopted technique to approximate such quantities. However, stable simulations rely on very small integration time-steps ($10^{-15}\,\mathrm{s}$), whereas convergence of some moments, e.g. binding free energy or rates, might rely on sampling processes on time-scales as long as $10^{-1}\, \mathrm{s}$, and these simulations must be repeated for every molecular system independently. Here, we present Implict Transfer Operator (ITO) Learning, a framework to learn surrogates of the simulation process with multiple time-resolutions. We implement ITO with denoising diffusion probabilistic models with a new SE(3) equivariant architecture and show the resulting models can generate self-consistent stochastic dynamics across multiple time-scales, even when the system is only partially observed. Finally, we present a coarse-grained CG-SE3-ITO model which can quantitatively model all-atom molecular dynamics using only coarse molecular representations. As such, ITO provides an important step towards multiple time- and space-resolution acceleration of MD. Code is available at \href{https://github.com/olsson-group/ito}{https://github.com/olsson-group/ito}.
△ Less
Submitted 28 October, 2023; v1 submitted 29 May, 2023;
originally announced May 2023.
-
Transition1x -- a Dataset for Building Generalizable Reactive Machine Learning Potentials
Authors:
Mathias Schreiner,
Arghya Bhowmik,
Tejs Vegge,
Jonas Busk,
Ole Winther
Abstract:
Machine Learning (ML) models have, in contrast to their usefulness in molecular dynamics studies, had limited success as surrogate potentials for reaction barrier search. It is due to the scarcity of training data in relevant transition state regions of chemical space. Currently, available datasets for training ML models on small molecular systems almost exclusively contain configurations at or ne…
▽ More
Machine Learning (ML) models have, in contrast to their usefulness in molecular dynamics studies, had limited success as surrogate potentials for reaction barrier search. It is due to the scarcity of training data in relevant transition state regions of chemical space. Currently, available datasets for training ML models on small molecular systems almost exclusively contain configurations at or near equilibrium. In this work, we present the dataset Transition1x containing 9.6 million Density Functional Theory (DFT) calculations of forces and energies of molecular configurations on and around reaction pathways at the wB97x/6-31G(d) level of theory. The data was generated by running Nudged Elastic Band (NEB) calculations with DFT on 10k reactions while saving intermediate calculations. We train state-of-the-art equivariant graph message-passing neural network models on Transition1x and cross-validate on the popular ANI1x and QM9 datasets. We show that ML models cannot learn features in transition-state regions solely by training on hitherto popular benchmark datasets. Transition1x is a new challenging benchmark that will provide an important step towards develo** next-generation ML force fields that also work far away from equilibrium configurations and reactive systems.
△ Less
Submitted 1 September, 2022; v1 submitted 25 July, 2022;
originally announced July 2022.
-
NeuralNEB -- Neural Networks can find Reaction Paths Fast
Authors:
Mathias Schreiner,
Arghya Bhowmik,
Tejs Vegge,
Peter Bjørn Jørgensen,
Ole Winther
Abstract:
Quantum mechanical methods like Density Functional Theory (DFT) are used with great success alongside efficient search algorithms for studying kinetics of reactive systems. However, DFT is prohibitively expensive for large scale exploration. Machine Learning (ML) models have turned out to be excellent emulators of small molecule DFT calculations and could possibly replace DFT in such tasks. For ki…
▽ More
Quantum mechanical methods like Density Functional Theory (DFT) are used with great success alongside efficient search algorithms for studying kinetics of reactive systems. However, DFT is prohibitively expensive for large scale exploration. Machine Learning (ML) models have turned out to be excellent emulators of small molecule DFT calculations and could possibly replace DFT in such tasks. For kinetics, success relies primarily on the models capability to accurately predict the Potential Energy Surface (PES) around transition-states and Minimal Energy Paths (MEPs). Previously this has not been possible due to scarcity of relevant data in the literature. In this paper we train state of the art equivariant Graph Neural Network (GNN)-based models on around 10.000 elementary reactions from the Transition1x dataset. We apply the models as potentials for the Nudged Elastic Band (NEB) algorithm and achieve a Mean Average Error (MAE) of 0.13+/-0.03 eV on barrier energies on unseen reactions. We compare the results against equivalent models trained on QM9 and ANI1x. We also compare with and outperform Density Functional based Tight Binding (DFTB) on both accuracy and computational resource. The implication is that ML models, given relevant data, are now at a level where they can be applied for downstream tasks in quantum chemistry transcending prediction of simple molecular features.
△ Less
Submitted 1 September, 2022; v1 submitted 20 July, 2022;
originally announced July 2022.
-
Octave-spanning 1 emission across the visible spectrum from single crystalline 1,3,5,7-tetrakis-($p$-methoxyphenyl)adamantane
Authors:
Marius J. Müller,
Ferdinand Ziese,
Jürgen Belz,
Franziska Hüppe,
Saravanan Gowrisankar,
Bastian Bernhardt,
Sebastian Schwan,
Doreen Mollenhauer Peter R. Schreiner,
Kerstin Volz,
Simone Sanna,
Sangam Chatterjee
Abstract:
Sustainable efficient light emitter based solely on elements-of-hope are needed to replace current compounds based on less-abundant materials. Functionalized diamondoids are a potential solution for this challenge as they offer efficient, octave-spanning emission across the visible spectrum in their single-crystalline form. Its large quantum efficiency increases towards higher-than-ambient tempera…
▽ More
Sustainable efficient light emitter based solely on elements-of-hope are needed to replace current compounds based on less-abundant materials. Functionalized diamondoids are a potential solution for this challenge as they offer efficient, octave-spanning emission across the visible spectrum in their single-crystalline form. Its large quantum efficiency increases towards higher-than-ambient temperatures to beyond 7 %. The stability beyond 200 °C renders such functionalized diamondoids as sustainable phosphors for LED applications. Detailed structural and theoretical investigations suggest a crucial role of exciton states accompanied by structural modifications (self-trapped excitons) in the emission process.
△ Less
Submitted 14 April, 2022;
originally announced April 2022.
-
Catalogue of new Herbig Ae/Be and classical Be stars. A machine learning approach to Gaia DR2
Authors:
M. Vioque,
R. D. Oudmaijer,
M. Schreiner,
I. Mendigutía,
D. Baines,
N. Mowlavi,
R. Pérez-Martínez
Abstract:
The intermediate-mass pre-main sequence Herbig Ae/Be stars are key to understanding the differences in formation mechanisms between low- and high-mass stars. The study of the general properties of these objects is hampered by the fact that few and mostly serendipitously discovered sources are known. Our goal is to identify new Herbig Ae/Be candidates to create a homogeneous and well defined catalo…
▽ More
The intermediate-mass pre-main sequence Herbig Ae/Be stars are key to understanding the differences in formation mechanisms between low- and high-mass stars. The study of the general properties of these objects is hampered by the fact that few and mostly serendipitously discovered sources are known. Our goal is to identify new Herbig Ae/Be candidates to create a homogeneous and well defined catalogue of these objects. We have applied machine learning techniques to 4,150,983 sources with data from Gaia DR2, 2MASS, WISE, and IPHAS or VPHAS+. Several observables were chosen to identify new Herbig Ae/Be candidates based on our current knowledge of this class, which is characterised by infrared excesses, photometric variabilities, and H$α$ emission lines. Classical techniques are not efficient for identifying new Herbig Ae/Be stars mainly because of their similarity with classical Be stars, with which they share many characteristics. By focusing on disentangling these two types of objects, our algorithm has also identified new classical Be stars. We have obtained a large catalogue of 8470 new pre-main sequence candidates and another catalogue of 693 new classical Be candidates with a completeness of $78.8\pm1.4\%$ and $85.5\pm1.2\%$, respectively. Of the catalogue of pre-main sequence candidates, at least 1361 sources are potentially new Herbig Ae/Be candidates according to their position in the Hertzsprung-Russell diagram. In this study we present the methodology used, evaluate the quality of the catalogues, and perform an analysis of their flaws and biases. For this assessment, we make use of observables that have not been accounted for by the algorithm and hence are selection-independent, such as coordinates and parallax based distances. The catalogue of new Herbig Ae/Be stars that we present here increases the number of known objects of the class by an order of magnitude.
△ Less
Submitted 4 May, 2020;
originally announced May 2020.
