Performance Assessment of Universal Machine Learning Interatomic Potentials: Challenges and Directions for Materials' Surfaces.

Machine learning interatomic potentials (MLIPs) are one of the main techniques in the materials science toolbox, able to bridge ab initio accuracy with the computational efficiency of classical force fields. This allows simulations ranging from atoms, molecules, and biosystems, to solid and bulk materials, surfaces, nanomaterials, and their interfaces and complex interactions. A recent class of advanced MLIPs, which use equivariant representations and deep graph neural networks, is known as universal models. These models are proposed as foundation models suitable for any system, covering most elements from the periodic table. Current universal MLIPs (UIPs) have been trained with the largest consistent data set available nowadays. However, these are composed mostly of bulk materials' DFT calculations. In this article, we assess the universality of all openly available UIPs, namely MACE, CHGNet, and M3GNet, in a representative task of generalization: calculation of surface energies. We find that the out-of-the-box foundation models have significant shortcomings in this task, with errors correlated to the total energy of surface simulations, having an out-of-domain distance from the training data set. Our results show that while UIPs are an efficient starting point for fine-tuning specialized models, we envision the potential of increasing the coverage of the materials space toward universal training data sets for MLIPs.

Electrochemical and theoretical investigation on the behavior of the Co2+ ion in three eutectic solvents.

Deep eutectic solvents (DESs) have many advantages, making them a promising alternative in replacing ionic liquids and organic solvents. Besides, DESs have received much prominence due to their diverse applications: Electrodeposition of metals, organic synthesis, gas adsorption, and biodiesel production. Therefore, this work analyzed the effect of the temperature increase (298 K-353 K) on the behavior of the Co2+ ions in three eutectic solvents through electrochemical techniques and computational simulations. From the electrochemical analysis realized, the increase in temperature caused a reduction in specific mass and an increase in the diffusion coefficient. Besides, the activation energy values were of 15.3, 29.9, and 55.2 kJ mol-1 for 1ChCl:2 EG, 1ChCl:2U, and 1ChCl:2G, respectively. The computational simulations indicate that the increased temperature effect caused the replacement of HBD molecules by anions chloride around Co2+ ions for the SDW1 and SDW3 systems between the temperatures of 298 K-353 K, except for the SDW2 system that the replaced occurred in the interval of 313 K-353 K. Besides, the increase of temperature occasioned the increase of strength for Co-Cl interaction and weakened the interactions between the Co2+ ions with the oxygen of HBD molecules.

Dynamic Electroporation Model Evaluation on Rabbit Tissues.

Biological electroporation is a process of opening pores in the cell membrane when exposed to intense electric fields. This work provides results for validation of a dynamic model of electroporation on biological tissues. Computational simulations were carried out and results for the electrical current through the tissue and increase of the tissue temperature were compared to experimental results. Two calculation methods were used: Equivalent Circuit Method and Finite Element Method. With Equivalent Circuit Method the dielectric dispersion present in biological tissues was included. Liver, kidney and heart of rabbit were used in the experiments. Voltage pulse protocols and voltage ramps were applied using stainless steel needles electrodes. There is good agreement between the simulated and experimental results with mean errors below 15%, with the simulated results within the experimental standard deviation. Only for the protocol with fundamental frequency of 50 kHz, the simulation performed by the Finite Element Method using a commercial software did not correctly represent the current, with errors reaching 50%. The justification for the error found is due to the dielectric dispersion that was not included in this simulator.

Simulating hybrid energy systems based on complementary renewable resources.

The complementarity between energy resources can influence the performance of hybrid generation and storage systems, and can also decisively influence their design. Renewable resources may have intermittent characteristics that make the study of the influence of complementarity on the performance of hybrid systems quite difficult. The establishment of a performance limit of hybrid systems based on renewable resources and the study of the effects of complementarity considering this limit can provide interesting results. This performance limit can be established with an idealization of the mathematical functions describing the energy availability of the explored renewable resources. This article presents a method for analyzing the performance of hybrid systems based on complementary resources. The method allows to evaluate the influence of different levels of complementarity between the exploited resources on the cost of energy and capacity shortage. Utilizing idealized energy availability, the result sets a performance limit. •A method to evaluate the impact of complementarity on the performance and reliability of hybrid systems.•The energy availabilities of the renewable sources are idealized and allow the characterization of a limit of performance.•Different levels of complementarity can be related with design parameters of hybrid energy and storage systems.

New Approaches to the Computer Simulation of Amorphous Alloys: A Review.

In this work we review our new methods to computer generate amorphous atomic topologies of several binary alloys: SiH, SiN, CN; binary systems based on group IV elements like SiC; the GeSe2 chalcogenide; aluminum-based systems: AlN and AlSi, and the CuZr amorphous alloy. We use an ab initio approach based on density functionals and computationally thermally-randomized periodically-continued cells with at least 108 atoms. The computational thermal process to generate the amorphous alloys is the undermelt-quench approach, or one of its variants, that consists in linearly heating the samples to just below their melting (or liquidus) temperatures, and then linearly cooling them afterwards. These processes are carried out from initial crystalline conditions using short and long time steps. We find that a step four-times the default time step is adequate for most of the simulations. Radial distribution functions (partial and total) are calculated and compared whenever possible with experimental results, and the agreement is very good. For some materials we report studies of the effect of the topological disorder on their electronic and vibrational densities of states and on their optical properties.