-
Beyond Scaling: Chemical Intuition as Emergent Ability of Universal Machine Learning Interatomic Potentials
Authors:
Shinnosuke Hattori,
Kohei Shimamura,
Aiichiro Nakano,
Rajiv K. Kalia,
Priya Vashishta,
Ken-ichi Nomura
Abstract:
Machine Learning Interatomic Potentials (MLIPs) have successfully demonstrated scaling behavior, i.e. the power-law improvement in training performance, however the emergence of novel capabilities at scale remains unexplored. We have developed Edge-wise Emergent Decomposition (E3D) framework to investigate how an MLIP develops the ability to derive physically meaningful local representations of ch…
▽ More
Machine Learning Interatomic Potentials (MLIPs) have successfully demonstrated scaling behavior, i.e. the power-law improvement in training performance, however the emergence of novel capabilities at scale remains unexplored. We have developed Edge-wise Emergent Decomposition (E3D) framework to investigate how an MLIP develops the ability to derive physically meaningful local representations of chemical bonds without explicit supervision. Employing an E(3)-equivariant network (Allegro) trained on molecular data (SPICE~2), we found that the trained MLIP spontaneously learned representations of bond dissociation energy (BDE) by decomposing the global potential energy landscape. The learned BDE values quantitatively agree with literature and its scalability are found to be robust across diverse training datasets, suggesting the presence of underlying representation that captures chemical reactions faithfully beyond given training information. Our E3D analysis utilizing Shannon's entropy reveals a close interplay between the decomposability of potential energy learning, scalability of learning, and emergent chemical reactivity, thus providing novel insights of scaling limitations and pathways toward more physically interpretable and predictive simulations.
△ Less
Submitted 9 June, 2025;
originally announced June 2025.
-
Allegro-FM: Towards Equivariant Foundation Model for Exascale Molecular Dynamics Simulations
Authors:
Ken-ichi Nomura,
Shinnosuke Hattori,
Satoshi Ohmura,
Ikumi Kanemasu,
Kohei Shimamura,
Nabankur Dasgupta,
Aiichiro Nakano,
Rajiv K. Kalia,
Priya Vashishta
Abstract:
We present a foundation model for exascale molecular dynamics simulations by leveraging an E(3) equivariant network architecture (Allegro) and a set of large-scale organic and inorganic materials datasets merged by Total Energy Alignment (TEA) framework. Thanks to the large-scale training sets, the obtained model (Allegro-FM) is versatile for various materials simulations for diverse downstream ta…
▽ More
We present a foundation model for exascale molecular dynamics simulations by leveraging an E(3) equivariant network architecture (Allegro) and a set of large-scale organic and inorganic materials datasets merged by Total Energy Alignment (TEA) framework. Thanks to the large-scale training sets, the obtained model (Allegro-FM) is versatile for various materials simulations for diverse downstream tasks covering 89 elements in the periodic table. Allegro-FM exhibits excellent agreements with high-level quantum chemistry theories in describing structural, mechanical, and thermodynamic properties, while exhibiting emergent capabilities for structural correlations, reaction kinetics, mechanical strengths, fracture, and solid/liquid dissolution, for which the model has not been trained. Furthermore, we demonstrate the robust predictability and generalizability of Allegro-FM for chemical reactions using the Transition1x that consists of 10k organic reactions and 9.6 million configurations including transition state data, as well as calcium silicate hydrates as a testbed. With its computationally efficient, strictly-local network architecture, Allegro-FM scales up to multi-billion-atom systems with a parallel efficiency of 0.964 on the exaflop/s Aurora supercomputer at Argonne Leadership Computing Facility. The approach presented in this work demonstrates the potential of the foundation model for a novel materials design and discovery based on large-scale atomistic simulations.
△ Less
Submitted 20 February, 2025; v1 submitted 9 February, 2025;
originally announced February 2025.
-
Enhancement of superconductivity coexisting with charge density wave in lattice expanded $\textrm{NbTe}_2$
Authors:
Takaya Shimokawa,
Yukiko Obata,
Hayato Makino,
Kaito Sato,
Kazutoshi Shimamura,
Hiroyuki Okamoto,
Masao Obata,
Tatsuki Oda,
Yasuo Yoshida
Abstract:
We report a significant enhancement of superconducting transition temperature ($\textit{T}_\textrm{c}$) of transition metal dichalcogenide (TMD) superconductor $\textrm{NbTe}_2$ from 0.56 K to 2.8 K. Detailed x-ray structure analysis reveals that our $\textit{T}_\textrm{c}$-enhanced sample has an anisotropic lattice distortion inducing ~1% expansion of the unit cell volume and multi-domain formati…
▽ More
We report a significant enhancement of superconducting transition temperature ($\textit{T}_\textrm{c}$) of transition metal dichalcogenide (TMD) superconductor $\textrm{NbTe}_2$ from 0.56 K to 2.8 K. Detailed x-ray structure analysis reveals that our $\textit{T}_\textrm{c}$-enhanced sample has an anisotropic lattice distortion inducing ~1% expansion of the unit cell volume and multi-domain formation in the $\textit{ab}$ planes. Despite the unit cell expansion, the distorted 1T structure, closely related to the charge density wave (CDW) order in this material, persists. Hall measurements show almost identical behaviors for both samples indicating that electronic structure does not change much due to the unit cell expansion. These results suggest that the CDW still coexists with the enhanced superconductivity unlike the other TMD superconductors.
△ Less
Submitted 2 October, 2024;
originally announced October 2024.
-
Field-induced quantum phase in a frustrated zigzag-square lattice
Authors:
Hironori Yamaguchi,
Kazutoshi Shimamura,
Yasuo Yoshida,
Akira Matsuo,
Koichi Kindo,
Kiichi Nakano,
Satoshi Morota,
Yuko Hosokoshi,
Takanori Kida,
Yoshiki Iwasaki,
Seiya Shimono,
Koji Araki,
Masayuki Hagiwara
Abstract:
This study presents the experimental realization of a spin-1/2 zigzag-square lattice in a verdazyl-based complex, namely ($m$-Py-V-2,6-F$_2$)$[$Cu(hfac)$_2]$. Molecular orbital calculations suggest the presence of five types of frustrated exchange couplings. Our observations reveal an incremental increase in the magnetization curve beyond a critical field, signifying a phase transition from the an…
▽ More
This study presents the experimental realization of a spin-1/2 zigzag-square lattice in a verdazyl-based complex, namely ($m$-Py-V-2,6-F$_2$)$[$Cu(hfac)$_2]$. Molecular orbital calculations suggest the presence of five types of frustrated exchange couplings. Our observations reveal an incremental increase in the magnetization curve beyond a critical field, signifying a phase transition from the antiferromagnetic ordered state to a quantum state characterized by a 1/2 plateau. This intriguing behavior arises from the effective stabilization of a zigzag chain by the external fields. These results provide evidence for field-induced dimensional reduction in a zigzag-square lattice attributed to the effects of frustration.
△ Less
Submitted 25 September, 2023;
originally announced September 2023.
-
Construction of Machine-Learning Interatomic Potential Under Heat Flux Regularization and Its Application to Power Spectrum Analysis for Silver Chalcogenides
Authors:
Kohei Shimamura,
Koura Akihide,
Fuyuki Shimojo
Abstract:
We propose a data-driven approach for constructing machine-learning interatomic potentials (MLIPs) trained under a regularization with the aim of avoiding nonphysical heat flux. Specifically, we introduce a regularization term for the heat flux into the cost function of MLIPs to be minimized. Since the treatment of heat flux using MLIPs with regularization can be decomposed into elemental contribu…
▽ More
We propose a data-driven approach for constructing machine-learning interatomic potentials (MLIPs) trained under a regularization with the aim of avoiding nonphysical heat flux. Specifically, we introduce a regularization term for the heat flux into the cost function of MLIPs to be minimized. Since the treatment of heat flux using MLIPs with regularization can be decomposed into elemental contributions or conducted in frequency space, this approach is expected to be useful for investigating the origin of thermal conductivity obtained from the Green-Kubo formula. However, the strength of regularization needs to be appropriately set because it may reduce not only the nonphysical part but also the intrinsic heat flux one. To this end, we investigated the conditions for constructing MLIPs that can reproduce the power spectra of heat flux associated with the empirical interatomic potential of Ag$_2$Se, which consists of pairwise functions and do not contain a nonphysical heat flux. The appropriate strength could be estimated from the variation of the magnitude of regularization term as well as root mean square errors for total potential energy, atomic force, and virial stress with respect to the strengths, without reference spectrum data. As an application example, we explored the differences in power spectra between superionic and nonsuperionic conducting phases based on the heat flux regularization to MLIPs trained with the first-principles calculation data of Ag$_2$S. Furthermore, our results demonstrate that training with the regularization improves the robustness of MLIPs as well as the reduction of the nonphysical heat flux.
△ Less
Submitted 4 April, 2022;
originally announced April 2022.
-
Precise magnetization measurements down to 500 mK using a miniature 3He cryostat and a closed-cycle 3He gas handling system installed in a SQUID magnetometer without continuous-cooling functionality
Authors:
Kazutoshi Shimamura,
Hiroki Wajima,
Hayato Makino,
Satoshi Abe,
Yoshinori Haga,
Yoshiaki Sato,
Tatsuya Kawae,
Yasuo Yoshida
Abstract:
We have conducted precise magnetization measurements down to 0.5 K with a miniature 3He cryostat and a closed-cycle 3He gas handling system for a commercial superconducting quantum interference device magnetometer [Magnetic Property Measurement System (Quantum Design)]. The gas handling system contains two sorption pumps filled with granular charcoals. We pressurize 3He gas up to ambient pressure…
▽ More
We have conducted precise magnetization measurements down to 0.5 K with a miniature 3He cryostat and a closed-cycle 3He gas handling system for a commercial superconducting quantum interference device magnetometer [Magnetic Property Measurement System (Quantum Design)]. The gas handling system contains two sorption pumps filled with granular charcoals. We pressurize 3He gas up to ambient pressure for the liquification at 3 K and then pump the vaper for cooling. The lowest sample temperature is ~ 0.5 K and it can persist for 34 hours. We demonstrate the performance of the system by observing the Meissner effect of aluminum below the superconducting transition temperature ~ 1 K. We also measured the magnetization curve of the heavy fermion superconductor CeCoIn5 resulting in successful observation of the lower critical field at 0.5 K.
△ Less
Submitted 18 February, 2022;
originally announced February 2022.
-
Micrometer-scale monolayer SnS growth by physical vapor deposition
Authors:
H. Kawamoto,
N. Higashitarumizu,
N. Nagamura,
M. Nakamura,
K. Shimamura,
N. Ohashi,
K. Nagashio
Abstract:
Recently, monolayer SnS, a two-dimensional group IV monochalcogenide, was grown on a mica substrate at the micrometer-size scale by the simple physical vapor deposition (PVD), resulting in the successful demonstration of its in-plane room temperature ferroelectricity. However, the reason behind the monolayer growth remains unclear because it had been considered that the SnS growth inevitably resul…
▽ More
Recently, monolayer SnS, a two-dimensional group IV monochalcogenide, was grown on a mica substrate at the micrometer-size scale by the simple physical vapor deposition (PVD), resulting in the successful demonstration of its in-plane room temperature ferroelectricity. However, the reason behind the monolayer growth remains unclear because it had been considered that the SnS growth inevitably results in a multilayer thickness due to the strong interlayer interaction arising from lone pair electrons. Here, we investigate the PVD growth of monolayer SnS from two different feed powders, highly purified SnS and commercial phase-impure SnS. Contrary to expectations, it is suggested that the mica substrate surface is modified by sulfur evaporated from the Sn2S3 contaminant in the as-purchased powder and the lateral growth of monolayer SnS is facilitated due to the enhanced surface diffusion of SnS precursor molecules, unlike the growth from the highly purified powder. This insight provides a guide to identify further controllable growth conditions.
△ Less
Submitted 2 December, 2020;
originally announced December 2020.
-
Weak Antilocalization and Conductance Fluctuation in a Sub-micrometer-sized Wire of Epitaxial Bi2Se3
Authors:
Sadashige Matsuo,
Tomohiro Koyama,
Kazutoshi Shimamura,
Tomonori Arakawa,
Yoshitaka Nishihara,
Daichi Chiba,
Kensuke Kobayashi,
Teruo Ono,
Cui-Zu Chang,
Ke He,
Xu-Cun Ma,
Qi-Kun Xue
Abstract:
In this study, we address the phase coherent transport in a sub-micrometer-sized Hall bar made of epitaxial Bi2Se3 thin film by probing the weak antilocalization (WAL) and the magnetoresistance fluctuation below 22 K. The WAL effect is well described by the Hikami-Larkin-Nagaoka model, where the temperature dependence of the coherence length indicates that electron conduction occurs quasi-one-dime…
▽ More
In this study, we address the phase coherent transport in a sub-micrometer-sized Hall bar made of epitaxial Bi2Se3 thin film by probing the weak antilocalization (WAL) and the magnetoresistance fluctuation below 22 K. The WAL effect is well described by the Hikami-Larkin-Nagaoka model, where the temperature dependence of the coherence length indicates that electron conduction occurs quasi-one-dimensionally in the narrow Hall bar. The temperature-dependent magnetoresistance fluctuation is analyzed in terms of the universal conductance fluctuation, which gives a coherence length consistent with that derived from the WAL effect.
△ Less
Submitted 28 February, 2012; v1 submitted 16 January, 2012;
originally announced January 2012.
-
Scattering functions of knotted ring polymers
Authors:
Miyuki K. Shimamura,
Kumiko Kamata,
Akihisa Yao,
Tetsuo Deguchi
Abstract:
We discuss the scattering function of a Gaussian random polygon with N nodes under a given topological constraint through simulation. We obtain the Kratky plot of a Gaussian polygon of N=200 having a fixed knot for some different knots such as the trivial, trefoil and figure-eight knots. We find that some characteristic properties of the different Kratky plots are consistent with the distinct va…
▽ More
We discuss the scattering function of a Gaussian random polygon with N nodes under a given topological constraint through simulation. We obtain the Kratky plot of a Gaussian polygon of N=200 having a fixed knot for some different knots such as the trivial, trefoil and figure-eight knots. We find that some characteristic properties of the different Kratky plots are consistent with the distinct values of the mean square radius of gyration for Gaussian polygons with the different knots.
△ Less
Submitted 5 April, 2005;
originally announced April 2005.
-
Geometrical complexity of conformations of ring polymers under topological constraints
Authors:
Miyuki K. Shimamura,
Tetsuo Deguchi
Abstract:
One measure of geometrical complexity of a spatial curve is the number of crossings in a planar projection of the curve.
For $N$-noded ring polymers with a fixed knot type, we evaluate numerically the average of the crossing number over some directions. We find that the average crossing number under the topological constraint are less than that of no topological constraint for large $N$. The de…
▽ More
One measure of geometrical complexity of a spatial curve is the number of crossings in a planar projection of the curve.
For $N$-noded ring polymers with a fixed knot type, we evaluate numerically the average of the crossing number over some directions. We find that the average crossing number under the topological constraint are less than that of no topological constraint for large $N$. The decrease of the geometrical complexity is significant when the thickness of polymers is small.
The simulation with or without a topological constraint also shows that the average crossing number and the average size of ring polymers are independent measures of conformational complexity.
△ Less
Submitted 22 November, 2002;
originally announced November 2002.
-
Knot complexity and the probability of random knotting
Authors:
Miyuki K. Shimamura,
Tetsuo Deguchi
Abstract:
The probability of a random polygon (or a ring polymer) having a knot type $K$ should depend on the complexity of the knot $K$. Through computer simulation using knot invariants, we show that the knotting probability decreases exponentially with respect to knot complexity. Here we assume that some aspects of knot complexity are expressed by the minimal crossing number $C$ and the aspect ratio…
▽ More
The probability of a random polygon (or a ring polymer) having a knot type $K$ should depend on the complexity of the knot $K$. Through computer simulation using knot invariants, we show that the knotting probability decreases exponentially with respect to knot complexity. Here we assume that some aspects of knot complexity are expressed by the minimal crossing number $C$ and the aspect ratio $p$ of the tube length to the diameter of the {\it ideal knot} of $K$, which is a tubular representation of $K$ in its maximally inflated state.
△ Less
Submitted 11 July, 2002; v1 submitted 11 July, 2002;
originally announced July 2002.
-
Finite-size and asymptotic behaviors of the gyration radius of knotted cylindrical polygons
Authors:
Miyuki K. Shimamura,
Tetsuo Deguchi
Abstract:
Several nontrivial properties are shown for the mean square radius of gyration $R_K^2$ of ring polymers with a fixed knot type K.
Through computer simulation, we discuss both finite-size and asymptotic behaviors of the gyration radius under the topological constraint for self-avoiding polygons consisting of N cylindrical segments with radius r. We find that the average size of ring polymers wi…
▽ More
Several nontrivial properties are shown for the mean square radius of gyration $R_K^2$ of ring polymers with a fixed knot type K.
Through computer simulation, we discuss both finite-size and asymptotic behaviors of the gyration radius under the topological constraint for self-avoiding polygons consisting of N cylindrical segments with radius r. We find that the average size of ring polymers with a knot K can be much larger than that of no topological constraint. The effective expansion due to the topological constraint depends strongly on the parameter r which is related to the excluded volume. The topological expansion is particularly significant for the small r case, where the simulation result is associated with that of random polygons with the knot K.
△ Less
Submitted 7 February, 2002;
originally announced February 2002.
-
Anomalous finite-size effects and canonical asymptotic behaviors for the mean-squared gyration radius of Gaussian random knots
Authors:
Miyuki K. Shimamura,
Tetsuo Deguchi
Abstract:
Anomalously strong finite-size effects have been observed for the mean square radius of gyration $R^2_K$ of Gaussian random polygons with a fixed knot $K$ as a function of the number $N$ of polygonal nodes. Through computer simulations with $N < 2000$, we find for several knots that the gyration radius $R^2_K$ can be approximated by a power law: $R^2_K \sim N^{2 ν_{K}^{eff}}$, where the effectiv…
▽ More
Anomalously strong finite-size effects have been observed for the mean square radius of gyration $R^2_K$ of Gaussian random polygons with a fixed knot $K$ as a function of the number $N$ of polygonal nodes. Through computer simulations with $N < 2000$, we find for several knots that the gyration radius $R^2_K$ can be approximated by a power law: $R^2_K \sim N^{2 ν_{K}^{eff}}$, where the effective exponents $ν_{K}^{eff}$ for the knots are larger than 0.5 and less than 0.6. A crossover occurs for the gyration radius of the trivial knot, when $N$ is roughly equal to the characteristic length $N_c$ of random knotting. For the asymptotic behavior of $R^2_K$, however, we find that it is consistent with the standard one with the scaling exponent 0.5. Thus, although the strong finite-size effects of $R^2_K$ remain effective for extremely large values of $N$, they can be matched with the asymptotic behavior of random walks.
△ Less
Submitted 25 September, 2001; v1 submitted 31 August, 2001;
originally announced August 2001.
-
Gyration radius of a circular polymer under a topological constraint with excluded volume
Authors:
Miyuki K. Shimamura,
Tetsuo Deguchi
Abstract:
It is nontrivial whether the average size of a ring polymer should become smaller or larger under a topological constraint.
Making use of some knot invariants, we evaluate numerically the mean square radius of gyration for ring polymers having a fixed knot type, where the ring polymers are given by self-avoiding polygons consisting of freely-jointed hard cylinders. We obtain plots of the gyrati…
▽ More
It is nontrivial whether the average size of a ring polymer should become smaller or larger under a topological constraint.
Making use of some knot invariants, we evaluate numerically the mean square radius of gyration for ring polymers having a fixed knot type, where the ring polymers are given by self-avoiding polygons consisting of freely-jointed hard cylinders. We obtain plots of the gyration radius versus the number of polygonal nodes for the trivial, trefoil and figure-eight knots. We discuss possible asymptotic behaviors of the gyration radius under the topological constraint. In the asymptotic limit, the size of a ring polymer with a given knot is larger than that of no topological constraint when the polymer is thin, and the effective expansion becomes weak when the polymer is thick enough.
△ Less
Submitted 19 April, 2001; v1 submitted 2 April, 2001;
originally announced April 2001.
-
On the Dominance of Trivial Knots among SAPs on a Cubic Lattice
Authors:
Akihisa Yao,
Hiroshi Matsuda,
Hiroshi Tsukahara,
Miyuki K. Shimamura,
Tetsuo Deguchi
Abstract:
The knotting probability is defined by the probability with which an $N$-step self-avoiding polygon (SAP) with a fixed type of knot appears in the configuration space. We evaluate these probabilities for some knot types on a simple cubic lattice. For the trivial knot, we find that the knotting probability decays much slower for the SAP on the cubic lattice than for continuum models of the SAP as…
▽ More
The knotting probability is defined by the probability with which an $N$-step self-avoiding polygon (SAP) with a fixed type of knot appears in the configuration space. We evaluate these probabilities for some knot types on a simple cubic lattice. For the trivial knot, we find that the knotting probability decays much slower for the SAP on the cubic lattice than for continuum models of the SAP as a function of $N$. In particular the characteristic length of the trivial knot that corresponds to a `half-life' of the knotting probability is estimated to be $2.5 \times 10^5$ on the cubic lattice.
△ Less
Submitted 28 August, 2001; v1 submitted 17 March, 2001;
originally announced March 2001.
-
Topological entropy of a stiff ring polymer and its connection to DNA knots
Authors:
Miyuki K. Shimamura,
Tetso Deguchi
Abstract:
We discuss the entropy of a circular polymer under a topological constraint. We call it the {\it topological entropy} of the polymer, in short. A ring polymer does not change its topology (knot type) under any thermal fluctuations. Through numerical simulations using some knot invariants, we show that the topological entropy of a stiff ring polymer with a fixed knot is described by a scaling for…
▽ More
We discuss the entropy of a circular polymer under a topological constraint. We call it the {\it topological entropy} of the polymer, in short. A ring polymer does not change its topology (knot type) under any thermal fluctuations. Through numerical simulations using some knot invariants, we show that the topological entropy of a stiff ring polymer with a fixed knot is described by a scaling formula as a function of the thickness and length of the circular chain. The result is consistent with the viewpoint that for stiff polymers such as DNAs, the length and diameter of the chains should play a central role in their statistical and dynamical properties. Furthermore, we show that the new formula extends a known theoretical formula for DNA knots.
△ Less
Submitted 6 March, 2001;
originally announced March 2001.
-
Characteristic length of random knotting for cylindrical self-avoiding polygons
Authors:
Miyuki K. shimamura,
Tetsuo Deguchi
Abstract:
We discuss the probability of random knotting for a model of self-avoiding polygons whose segments are given by cylinders of unit length with radius $r$. We show numerically that the characteristic length of random knotting is roughly approximated by an exponential function of the chain thickness $r$.
We discuss the probability of random knotting for a model of self-avoiding polygons whose segments are given by cylinders of unit length with radius $r$. We show numerically that the characteristic length of random knotting is roughly approximated by an exponential function of the chain thickness $r$.
△ Less
Submitted 18 August, 2000;
originally announced August 2000.
