Zotero Catalogue

One nanny isn’t cutting it anymore. Some parents are spending upward of $250,000 on teams for their children. Luxury services offer potty-training, baby chefs and bike-riding lessons.

A collaborative AI workspace, built on your company context. Build and orchestrate agents right alongside your team's projects, meetings, and connected apps.

A 48-year-old woman with a stimulating electrode implanted in the right thalamic nucleus ventralis posterolateralis developed compulsive self-stimulation associated with erotic sensations and changes in autonomic and neurologic function. Stimulation effects were evaluated by neuropsychologic testing, endocrine studies, positron emission tomographic measurements of regional cerebral metabolic rate for glucose, EEG and evoked potentials. During stimulation, vital signs and pupillary diameter increased and a left hemiparesis and left hemisensory loss developed. Verbal functions deteriorated and visuospatial processing improved. Plasma growth hormone concentrations decreased, and adrenocorticotrophic hormone and cortisol levels rose. With stimulation, glucose metabolism increased in both thalami and both hemispheres, reversing baseline right-sided hypometabolism and right-left asymmetries. EEG and both somatosensory and brain-stem auditory evoked potentials remained unchanged during stimulation, while visual evoked potentials revealed evidence of anterior visual pathway dysfunction in the left eye. This case establishes the potential for addiction to deep brain stimulation and demonstrates that widespread behavioral and physiological changes, with concomitant alteration in the regional cerebral metabolic rate for glucose, may accompany unilateral thalamic stimulation.

Data centers of cloud providers hold millions of processor cores, exabytes of storage, and petabytes of network bandwidth. Research shows that in 2019, data centers consumed more than 2% of global electricity production, where 50% of consumption targeted for cooling infrastructures. While the most effective solution for thermal distribution is liquid cooling, technical challenges and complexities make it expensive. We suggest using living spiders as cooling devices for data centers. A prior work shows that spider silk has high thermal conductivity, close to that of copper: the second-best metallic conductor. Spiders not only generate spider silk but maintain it. Recruiting spiders for the job requires no more than inserting bugs to the data center for the spiders to catch. This solution is effective, self-sustaining, and environment-friendly, but requires solving a number of non-trivial technical and zoological challenges on the way to make it practical.

Frontier LLMs can perform multi-step reasoning over content-free filler tokens like dots or counting sequences, producing correct answers with no visible chain-of-thought (CoT). This is a limit case for behavioral oversight, where surface tokens carry no information about the underlying reasoning. But hidden from the output is not the same as hidden from us. On four task families (fact retrieval, parallel numeric composition, string manipulation, and in-context computation), two open-weights frontier models (DeepSeek V3, Kimi K2) compute over filler tokens in a structured, legible way: attention routes the question through the filler region to the answer, logit-lens readouts show retrieved facts emerging early and their composition crystallizing in late layers, and KV-cache transplants at filler positions causally swap outputs between examples. We introduce an unsupervised decoding pipeline that takes only hidden states as input and recovers intermediate values with 80-95% accuracy (best LLM judge) across both models and all four tasks, without ground-truth labels or training. Hidden computation that defeats behavioral CoT monitoring is, on these tasks, directly readable from the residual stream, suggesting monitorability is a property of the model's full computational trace, not just its surface tokens.

Lessons learned from building multiplayer world models. Built a video tokenizer with spatial attention and a dynamics model with action spaces.

28 August – 1 September 2023 University of Pavia (Italy) From August 28th to September 1st, 2023, Pavia (Italy) hosted the first in-person edition of the “Advanced Quantum ESPRESSO Scho…

Generalized gradient approximations (GGA's) for the exchange-correlation energy improve upon the local spin density (LSD) description of atoms, molecules, and solids. We present a simple derivation of a simple GGA, in which all parameters (other than those in LSD) are fundamental constants. Only general features of the detailed construction underlying the Perdew-Wang 1991 (PW91) GGA are invoked. Improvements over PW91 include an accurate description of the linear response of the uniform electron gas, correct behavior under uniform scaling, and a smoother potential.

An algorithm is presented for carrying out decomposition of electronic charge density into atomic contributions. As suggested by Bader [R. Bader, Atoms in Molecules: A Quantum Theory, Oxford University Press, New York, 1990], space is divided up into atomic regions where the dividing surfaces are at a minimum in the charge density, i.e. the gradient of the charge density is zero along the surface normal. Instead of explicitly finding and representing the dividing surfaces, which is a challenging task, our algorithm assigns each point on a regular (x,y,z) grid to one of the regions by following a steepest ascent path on the grid. The computational work required to analyze a given charge density grid is approximately 50 arithmetic operations per grid point. The work scales linearly with the number of grid points and is essentially independent of the number of atoms in the system. The algorithm is robust and insensitive to the topology of molecular bonding. In addition to two test problems involving a water molecule and NaCl crystal, the algorithm has been used to estimate the electrical activity of a cluster of boron atoms in a silicon crystal. The highly stable three-atom boron cluster, B3I is found to have a charge of −1.5e, which suggests approximately 50% reduction in electrical activity as compared with three substitutional boron atoms.

The molecular structure hypothesis--that a molecule is a collection of atoms linked by a network of bonds-- provides the principal means of ordering and classifying observations in chemistry. However this hypothesis is not related directly to the physics which governs the motions of atomic nuclei and electrons. It is the purpose of this important new book to show that a theory can be developed to establish the molecular structure hypothesis, demonstrating that the atoms in a molecule are real, with properties predicted and defined by the laws of quantum mechanics, and that the structure their presence imparts to a molecule is indeed a consequence of the underlying physics. As a result, the classification based upon the concept of atoms in molecules is freed from its empirical constraints and the full predictive power of quantum mechanics can be incorporated into the resulting theory--a theory of atoms in molecules. Eminently accessible and readable, the book will interest all scientists involved with experiment and observation at the atomic level, in addition to theoreticians. , The molecular structure hypothesis--that a molecule is a collection of atoms linked by a network of bonds-- provides the principal means of ordering and classifying observations in chemistry. However this hypothesis is not related directly to the physics which governs the motions of atomic nuclei and electrons. It is the purpose of this important new book to show that a theory can be developed to establish the molecular structure hypothesis, demonstrating that the atoms in a molecule are real, with properties predicted and defined by the laws of quantum mechanics, and that the structure their presence imparts to a molecule is indeed a consequence of the underlying physics. As a result, the classification based upon the concept of atoms in molecules is freed from its empirical constraints and the full predictive power of quantum mechanics can be incorporated into the resulting theory--a theory of atoms in molecules. Eminently accessible and readable, the book will interest all scientists involved with experiment and observation at the atomic level, in addition to theoreticians.

Creating fast and accurate force fields is a long-standing challenge in computational chemistry and materials science. Recently, several equivariant message passing neural networks (MPNNs) have been shown to outperform models built using other approaches in terms of accuracy. However, most MPNNs suffer from high computational cost and poor scalability. We propose that these limitations arise because MPNNs only pass two-body messages leading to a direct relationship between the number of layers and the expressivity of the network. In this work, we introduce MACE, a new equivariant MPNN model that uses higher body order messages. In particular, we show that using four-body messages reduces the required number of message passing iterations to just two, resulting in a fast and highly parallelizable model, reaching or exceeding state-of-the-art accuracy on the rMD17, 3BPA, and AcAc benchmark tasks. We also demonstrate that using higher order messages leads to an improved steepness of the learning curves.

Many important applications of electronic structure methods involve molecules or solid surfaces in a solvent medium. Since explicit treatment of the solvent in such methods is usually not practical, calculations often employ continuum solvation models to approximate the effect of the solvent. Previous solvation models either involve a parametrization based on atomic radii, which limits the class of applicable solutes, or based on solute electron density, which is more general but less accurate, especially for charged systems. We develop an accurate and general solvation model that includes a cavity that is a nonlocal functional of both solute electron density and potential, local dielectric response on this nonlocally determined cavity, and nonlocal approximations to the cavity-formation and dispersion energies. The dependence of the cavity on the solute potential enables an explicit treatment of the solvent charge asymmetry. With four parameters per solvent, this “CANDLE” model simultaneously reproduces solvation energies of large datasets of neutral molecules, cations, and anions with a mean absolute error of 1.8 kcal/mol in water and 3.0 kcal/mol in acetonitrile.

The solvation model proposed by Fattebert and Gygi [Journal of Computational Chemistry 23, 662 (2002)] and Scherlis et al. [Journal of Chemical Physics 124, 074103 (2006)] is reformulated, overcoming some of the numerical limitations encountered and extending its range of applicability. We first recast the problem in terms of induced polarization charges that act as a direct mapping of the self-consistent continuum dielectric; this allows to define a functional form for the dielectric that is well behaved both in the high-density region of the nuclear charges and in the low-density region where the electronic wavefunctions decay into the solvent. Second, we outline an iterative procedure to solve the Poisson equation for the quantum fragment embedded in the solvent that does not require multi-grid algorithms, is trivially parallel, and can be applied to any Bravais crystallographic system. Last, we capture some of the non-electrostatic or cavitation terms via a combined use of the quantum volume and quantum surface [Physical Review Letters 94, 145501 (2005)] of the solute. The resulting self-consistent continuum solvation (SCCS) model provides a very effective and compact fit of computational and experimental data, whereby the static dielectric constant of the solvent and one parameter allow to fit the electrostatic energy provided by the PCM model with a mean absolute error of 0.3 kcal/mol on a set of 240 neutral solutes. Two parameters allow to fit experimental solvation energies on the same set with a mean absolute error of 1.3 kcal/mol. A detailed analysis of these results, broken down along different classes of chemical compounds, shows that several classes of organic compounds display very high accuracy, with solvation energies in error of 0.3-0.4 kcal/mol, whereby larger discrepancies are mostly limited to self-dissociating species and strong hydrogen-bond forming compounds.

Quantum ESPRESSO is an integrated suite of open-source computer codes for quantum simulations of materials using state-of-the-art electronic-structure techniques, based on density-functional theory, density-functional perturbation theory, and many-body perturbation theory, within the plane-wave pseudopotential and projector-augmented-wave approaches. Quantum ESPRESSO owes its popularity to the wide variety of properties and processes it allows to simulate, to its performance on an increasingly broad array of hardware architectures, and to a community of researchers that rely on its capabilities as a core open-source development platform to implement their ideas. In this paper we describe recent extensions and improvements, covering new methodologies and property calculators, improved parallelization, code modularization, and extended interoperability both within the distribution and with external software.

Soft floating robots (SFRs) represent a shift from rigid machines, offering gravity-defying, compliant, and tactile embodiments for indoor cohabitation. However, their development remains fragmented across isolated prototypes, lacking a coherent design vocabulary. Without a systematic understanding of their interactional capabilities, designers struggle to leverage SFRs’ unique affordances, and these systems often remain limited to novelty applications that are difficult to integrate into everyday life. To address this, we propose a design space for interaction with SFRs. Informed by an exploratory study with 12 experts from HCI, Design, and Robotics, we identify ten design dimensions spanning physical, interactive, and behavioral properties, along with a range of application scenarios. We further present proof-of-concept design examples to demonstrate how this design space can support diverse interaction possibilities. This work contributes a structured framework for understanding and designing interactions with SFRs, supporting their integration into everyday indoor environments.

Machine learning interatomic potentials (MLIPs), also known as machine learning force fields (MLFFs), offer scalable means of simulating complex systems and processes at \textit{ab initio} level accuracy. One such process is the critical yet still poorly understood formation of the solid electrolyte interphase (SEI) at the anode of a Li-ion battery (LIB) during the first charge cycle, where electrochemical reduction of the electrolyte leads to the generation of decomposition products. MLIPs are uniquely poised to atomistically describe these electrochemical processes, as they are not as affected by the same limitations in bonding and electron transfer as classical force fields. Nonetheless, training MLIPs to run accurate dynamics of a condensed phase with two different oxidation states, such as in electrochemistry, is challenging for many architectures. In this work, we show that by using MPNICE, a message passing MLIP architecture with iterative charge equilibration, we are able to accurately (within 1 kcal/mol) train models along two potential energy surfaces (reduced and unreduced) for LIB-relevant electrolyte systems. Importantly, we demonstrate strategies for sampling and training to examples of anion radicals of these species, which often are not centered on any atom (off-center radicals, or OCRs). We additionally discuss well known limitations of global charge equilibration (Qeq) algorithms in erroneously de-localizing charge, and test methods to alleviate the impact on resulting dynamics. Simulations using these models reveal new insights into electrolyte reduction and considerations for the realistic simulation of electron transfer processes in the condensed phase.

Oxygen evolution reaction (OER) is of crucial importance to sustainable energy and environmental engineering, and layered double hydroxides (LDHs) are among the most active catalysts for OER in alkaline conditions, but the reaction mechanism for OER on LDHs remains controversial. Distinctive types of reaction mechanisms have been proposed for the O-O coupling in OER, yet they compose a coupled reaction network with competing kinetics dependent on applied potentials. Herein, we combine grand-canonical methods and micro-kinetic modeling to unravel that the nature of dominant mechanism for OER on LDHs transitions among distinctive types as a function of applied potential, and this arises from the interplay among applied potential and competing kinetics in the coupled reaction network. The theory-predicted overpotentials, Tafel slopes, and findings are in agreement with the observations of experiments including isotope labelling. Thus, we establish a computational methodology to identify and elucidate the potential-dependent mechanisms for electrochemical reactions.

Electrocatalysis plays a central role in clean energy conversion, enabling a number of sustainable processes for future technologies. This review discusses design strategies for state-of-the-art heterogeneous electrocatalysts and associated materials for several different electrochemical transformations involving water, hydrogen, and oxygen, using theory as a means to rationalize catalyst performance. By examining the common principles that govern catalysis for different electrochemical reactions, we describe a systematic framework that helps to understand trends in catalyzing these reactions, serving as a guide to new catalyst development, while highlighting key gaps that need to be addressed. We conclude by extending this framework to emerging clean energy reactions including hydrogen peroxide production, carbon dioxide reduction and nitrogen reduction, where the development of improved catalysts could allow for the sustainable production of a broad range of fuels and chemicals.

Efficient electrocatalysts for the oxygen evolution reaction (OER) are paramount to the development of electrochemical devices for clean energy and fuel conversion. However, the structural complexity of heterogeneous electrocatalysts makes it a great challenge to elucidate the surface catalytic sites and OER mechanisms. Here, we report that catalytic single-site Co in a well-defined brookite TiO2 nanorod (210) surface (Co-TiO2) presents turnover frequencies that are among the highest for Co-based heterogeneous catalysts reported to date, reaching 6.6 ± 1.2 and 181.4 ± 28 s−1 at 300 and 400 mV overpotentials, respectively. Based on grand canonical quantum mechanics calculations and the single-site Co atomic structure validated by in situ and ex situ spectroscopic probes, we have established a full description of the catalytic reaction kinetics for Co-TiO2 as a function of applied potential, revealing an adsorbate evolution mechanism for the OER. The computationally predicted Tafel slope and turnover frequencies exhibit exceedingly good agreement with experiment.

Efficient electrocatalysts for the oxygen evolution reaction (OER) are paramount to the development of electrochemical devices for clean energy and fuel conversion. However, the structural complexity of heterogeneous electrocatalysts makes it a great challenge to elucidate the surface catalytic sites and OER mechanisms. Here, we report that catalytic single-site Co in a well-defined brookite TiO2 nanorod (210) surface (Co-TiO2) presents turnover frequencies that are among the highest for Co-based heterogeneous catalysts reported to date, reaching 6.6 ± 1.2 and 181.4 ± 28 s−1 at 300 and 400 mV overpotentials, respectively. Based on grand canonical quantum mechanics calculations and the single-site Co atomic structure validated by in situ and ex situ spectroscopic probes, we have established a full description of the catalytic reaction kinetics for Co-TiO2 as a function of applied potential, revealing an adsorbate evolution mechanism for the OER. The computationally predicted Tafel slope and turnover frequencies exhibit exceedingly good agreement with experiment.

The design of advanced electrocatalysts is often hindered by uncertainties in identifying and controlling the active surfaces and catalytic centers within heterogeneous materials. Here we present the synthesis of single-site Co catalysts, substitutionally doped into surface-controlled TiO2 anatase nanocrystals, aimed at enhancing the oxygen evolution reaction (OER). Grand canonical quantum mechanics calculations reveal that the kinetics of the OER, following an adsorbate evolution mechanism, is markedly influenced by the coordination environment of Co. The simulations suggest significantly higher turnover frequencies when Co is doped into the (001) surface of TiO2 compared to the (101) surface. Consistent with the computational findings, experimental results show that Co-doped TiO2 (Co-TiO2) nanoplates with selectively exposed {001} surfaces exhibit enhanced current densities and turnover frequencies compared to Co-TiO2 nanobipyramids with {101} surfaces. This study highlights the synergy between theoretical calculations and precision synthesis in the development of more effective catalysts.

Titanium dioxide (TiO2) presents a long-standing challenge for approximate Kohn-Sham density functional theory (KS-DFT), as well as to its Hubbard-corrected extension, DFT+U. We find that a previously proposed extension of first-principles DFT+U to incorporate a Hund's 𝐽 correction, termed DFT+U+J, in combination with parameters calculated using a recently proposed linear-response theory, predicts fundamental band gaps that are accurate to well within the experimental uncertainty in rutile and anatase TiO2. Our approach builds upon established findings that Hubbard correction of both the titanium 3⁢𝑑 and oxygen 2⁢𝑝 subspaces in TiO2, symbolically giving DFT+U𝑑,𝑝, is necessary to achieve acceptable band gaps using DFT+U. This requirement remains when the first-principles Hund's 𝐽 is included. We also find that the calculated gap depends on the correlated subspace definition even when using subspace-specific first-principles 𝑈 and 𝐽 parameters. Using the simplest reasonable correlated subspace definition and underlying functional, the local density approximation, we show that high accuracy results from using a relatively uncomplicated form of the DFT+U+J functional. For closed-shell systems such as TiO2, we describe how various DFT+U+J functionals reduce to DFT+U with suitably modified parameters, so that reliable band gaps can be calculated for rutile and anatase with no modifications to a conventional DFT+U code.

Transition metal oxides have attracted much attention as photo(electrochemical)-catalysts but practical applications are typically hampered by their low and anisotropic charge mobility. A deep understanding of excess charge carrier transport in these materials requires a dynamical treatment of nuclear motion that goes well beyond standard approaches. Here we introduce DeepPolaron, a machine learning framework boosting the accessible time scale of first principles molecular dynamics of adiabatic polaron transport by three orders of magnitude at a virtually negligible loss in accuracy. We apply our method to excess electron and hole transport in titanium dioxide rutile and anatase. We find that the excess electron in rutile relaxes to a polaron predominantly localized on a single Ti atom with hopping occurring only along the [001] direction, associated with an activation energy of 39 meV and a room temperature mobility of $4.4 \times 10^{-2}$ cm$^2$/Vs in good agreement with experiment. In contrast the hole polaron in anatase is localized on a single O atom, and due to poor O 2p orbital overlap with first nearest neighbors charge transport occurs primarily to second nearest neighbors, with a large activation energy of 139 meV resulting in a small room temperature mobility of $1.4 \times 10^{-3}$ cm$^2$/Vs. This work provides a finite temperature first-principles characterization of small polaron transport in rutile and anatase, with a methodology that is directly transferable to other small polaron forming materials and interfacial charge-transfer processes.

Using time-dependent density functional theory (TD-DFT) and nonadiabatic molecular dynamics (NAMD) simulations, we elucidate how hydrogen bonding and water dynamics regulate hole transfer at the anatase TiO2(101)/water interface. Compared to low-density water (LW), moderate-density water (MW) enhances hydrogen bonding between surface- and nonsurface-adsorbed water, restricting interfacial water mobility. This suppresses nonadiabatic coupling and slows the hole transfer. Conversely, higher-density water (HW) near the vacuum destabilizes hydrogen bonding networks, freeing interfacial water and amplifying thermal motion. Enhanced disorder strengthens nonadiabatic coupling, accelerating the hole transfer. Temperature-dependent simulations show that thermal energy overcomes hydrogen bonding constraints: elevated temperatures intensify water dynamics and nonadiabatic coupling, accelerating hole transfer. These results establish hydrogen bonding and thermal fluctuations as key regulators of charge dynamics at the semiconductor/liquid interfaces.

Photocatalytic CO2 reduction to value-added chemicals is limited by inefficient charge transfer and sluggish multielectron kinetics. Here, we develop a TiO2 (P25) based ternary system incorporating Pt NPs and 1T-dominant MoSe2 NSs as dual cocatalysts (i.e., Pt/TiO2-MoSe2) to direct charge flow and reaction pathways. This Pt/TM architecture accelerates charge separation and provides highly active sites for CO2 activation. The optimized Pt1.5%-TiO2-MoSe2 exhibits a CH4 evolution rate of 17.81 μmol g−1 under 5-sun illumination with ≈ 98% selectivity in the gas phase, achieving a 65-fold enhancement over TiO2 (P25). Under multi-sun irradiation, increased photon flux boosts activity, indicating the critical role of charge carrier density, and facilitates rapid CH4 desorption, thereby overcoming the activity-selectivity trade-off. The coexistence of static interfacial charge redistribution and dynamic photoinduced electron transfer, validated by experiment (XPS, XAS) and Density Functional Theory (DFT) quantum mechanics (QM) calculations, ensures the retention of the active 1T-MoSe2 phase and establishes an efficient TiO2 → MoSe2 → Pt charge-funneling highway. In situ DRIFTS, combined with simulated infrared spectra (DFT), identifies a *CHO-dominated pathway for the conversion (*CO→ *CHO → *CHOH → *CH2OH → *CH2 → *CH3 → CH4), and isotopic labeling confirms the CO2-to-CH4 formation. DFT results further support the cascade charge transfer and reduced energy barriers enabled by the dual cocatalyst system.

TiO 2 is a widely used photocatalyst in science and technology and its interface with water is important in fields ranging from geochemistry to biomedicine. , TiO 2 is a widely used photocatalyst in science and technology and its interface with water is important in fields ranging from geochemistry to biomedicine. Yet, it is still unclear whether water adsorbs in molecular or dissociated form on TiO 2 even for the case of well-defined crystalline surfaces. To address this issue, we simulated the TiO 2 –water interface using molecular dynamics with an ab initio -based deep neural network potential. Our simulations show a dynamical equilibrium of molecular and dissociative adsorption of water on TiO 2 . Water dissociates through a solvent-assisted concerted proton transfer to form a pair of short-lived hydroxyl groups on the TiO 2 surface. Molecular adsorption of water is Δ F = 8.0 ± 0.9 kJ mol −1 lower in free energy than the dissociative adsorption, giving rise to a 5.6 ± 0.5% equilibrium water dissociation fraction at room temperature. Due to the relevance of surface hydroxyl groups to the surface chemistry of TiO 2 , our model might be key to understanding phenomena ranging from surface functionalization to photocatalytic mechanisms.

Water adsorption and dissociation processes on pristine low-index TiO2 interfaces are important but poorly understood outside the well-studied anatase (101) and rutile (110). To understand these, we construct three sets of machine learning potentials that are simultaneously applicable to various TiO2 surfaces, based on three density-functional-theory approximations. Here we show the water dissociation free energies on seven pristine TiO2 surfaces, and predict that anatase (100), anatase (110), rutile (001), and rutile (011) favor water dissociation, anatase (101) and rutile (100) have mostly molecular adsorption, while the simulations of rutile (110) sensitively depend on the slab thickness and molecular adsorption is preferred with thick slabs. Moreover, using an automated algorithm, we reveal that these surfaces follow different types of atomistic mechanisms for proton transfer and water dissociation: one-step, two-step, or both. These mechanisms can be rationalized based on the arrangements of water molecules on the different surfaces. Our finding thus demonstrates that the different pristine TiO2 surfaces react with water in distinct ways, and cannot be represented using just the low-energy anatase (101) and rutile (110) surfaces.

The NequIP framework is redesigned for scalable distributed training and PyTorch 2.0 compilation. AOT Inductor inference and optimized Allegro kernels accelerate molecular dynamics by factors of 5–18 on practical system sizes. , Machine learning interatomic potentials, particularly those based on deep equivariant neural networks, have demonstrated state-of-the-art accuracy and computational efficiency in atomistic modeling tasks like molecular dynamics and high-throughput screening. The size of datasets and demands of downstream workflows are growing rapidly, making robust and scalable software essential. This work presents a major overhaul of the NequIP framework focusing on multi-node parallelism, computational performance, and extensibility. The redesigned framework supports distributed training on large datasets and removes barriers preventing full utilization of the PyTorch 2.0 compiler at train time. We demonstrate this acceleration in a case study by training Allegro models on the SPICE 2 dataset of organic molecular systems. For inference, we introduce the first end-to-end infrastructure that uses the PyTorch Ahead-of-Time Inductor compiler for machine learning interatomic potentials. Additionally, we implement a custom kernel for the Allegro model's most expensive operation, the tensor product. Together, these advancements speed up molecular dynamics calculations on system sizes of practical relevance by up to factors of 5 to 18.

The accurate description of chemical processes often requires the use of computationally demanding methods like density-functional theory (DFT), making long simulations of large systems unfeasible. In this Letter we introduce a new kind of neural-network representation of DFT potential-energy surfaces, which provides the energy and forces as a function of all atomic positions in systems of arbitrary size and is several orders of magnitude faster than DFT. The high accuracy of the method is demonstrated for bulk silicon and compared with empirical potentials and DFT. The method is general and can be applied to all types of periodic and nonperiodic systems.

ALTHOUGH the possibility of water photolysis has been investigated by many workers, a useful method has only now been developed. Because water is transparent to visible light it cannot be decomposed directly, but only by radiation with wavelengths shorter than 190 nm (ref. 1).

How to use a research internship to prepare for grad school.

Inclusive, student-centered research environments that prioritize mentorship and development need new models and new guidelines.

This post demystifies some of the expectations and interactions important to success in undergraduate research.