Active sparse Bayesian committee machine potential for isothermal-isobaric molecular dynamics simulations.

Recent advancements in machine learning potentials (MLPs) have significantly impacted the fields of chemistry, physics, and biology by enabling large-scale first-principles simulations. Among different machine learning approaches, kernel-based MLPs distinguish themselves through their ability to handle small datasets, quantify uncertainties, and minimize over-fitting. Nevertheless, their extensive computational requirements present considerable challenges. To alleviate these, sparsification methods have been developed, aiming to reduce computational scaling without compromising accuracy. In the context of isothermal and isobaric ML molecular dynamics (MD) simulations, achieving precise pressure estimation is crucial for reproducing reliable system behavior under constant pressure. Despite progress, sparse kernel MLPs struggle with precise pressure prediction. Here, we introduce a virial kernel function that significantly enhances the pressure estimation accuracy of MLPs. Additionally, we propose the active sparse Bayesian committee machine (BCM) potential, an on-the-fly MLP architecture that aggregates local sparse Gaussian process regression (SGPR) MLPs. The sparse BCM potential overcomes the steep computational scaling with the kernel size, and a predefined restriction on the size of kernel allows for fast and efficient on-the-fly training. Our advancements facilitate accurate and computationally efficient machine learning-enhanced MD (MLMD) simulations across diverse systems, including ice-liquid coexisting phases, Li10Ge(PS6)2 lithium solid electrolyte, and high-pressure liquid boron nitride.

Does pre-speech auditory modulation reflect processes related to feedback monitoring or speech movement planning?

Previous studies have revealed that auditory processing is modulated during the planning phase immediately prior to speech onset. To date, the functional relevance of this pre-speech auditory modulation (PSAM) remains unknown. Here, we investigated whether PSAM reflects neuronal processes that are associated with preparing auditory cortex for optimized feedback monitoring as reflected in online speech corrections. Combining electroencephalographic PSAM data from a previous data set with new acoustic measures of the same participants' speech, we asked whether individual speakers' extent of PSAM is correlated with the implementation of within-vowel articulatory adjustments during /b/-vowel-/d/ word productions. Online articulatory adjustments were quantified as the extent of change in inter-trial formant variability from vowel onset to vowel midpoint (a phenomenon known as centering). This approach allowed us to also consider inter-trial variability in formant production and its possible relation to PSAM at vowel onset and midpoint separately. Results showed that inter-trial formant variability was significantly smaller at vowel midpoint than at vowel onset. PSAM was not significantly correlated with this amount of change in variability as an index of within-vowel adjustments. Surprisingly, PSAM was negatively correlated with inter-trial formant variability not only in the middle but also at the very onset of the vowels. Thus, speakers with more PSAM produced formants that were already less variable at vowel onset. Findings suggest that PSAM may reflect processes that influence speech acoustics as early as vowel onset and, thus, that are directly involved in motor command preparation (feedforward control) rather than output monitoring (feedback control).

Single-atom catalysts supported on a hybrid structure of boron nitride/graphene for efficient nitrogen fixation via synergistic interfacial interactions.

Hexagonal boron nitride (BN) shows significant chemical stability and promising thermal nitrogen reduction reaction (NRR) activity but suffers from low conductivity in electrolysis with a wide band gap. To overcome this problem, two-dimensional (2D) BN and graphene (G) are designed as a heterostructure, namely BN/G. According to density functional theory (DFT), the higher conductivity of G narrows the band gap of BN by inducing some electronic states near the Fermi energy level (Ef). Once transition metals (TMs) are anchored in the BN/G structure as single atom catalysts (SACs), the NRR activity improves as the inert BN basal layer activates with moderate *NH2 binding energy and further the band gap is reduced to zero. V (vanadium) and W (tungsten) SACs exhibit the best performance with limiting potentials of -0.22 and -0.41 V, respectively. This study helps in understanding the improvement of the NRR activity of BN, providing physical insights into the adsorbate-TM interaction.

Neurophysiological evidence of sensory prediction errors driving speech sensorimotor adaptation.

The human sensorimotor system has a remarkable ability to quickly and efficiently learn movements from sensory experience. A prominent example is sensorimotor adaptation, learning that characterizes the sensorimotor system's response to persistent sensory errors by adjusting future movements to compensate for those errors. Despite being essential for maintaining and fine-tuning motor control, mechanisms underlying sensorimotor adaptation remain unclear. A component of sensorimotor adaptation is implicit (i.e., the learner is unaware of the learning process) which has been suggested to result from sensory prediction errors-the discrepancies between predicted sensory consequences of motor commands and actual sensory feedback. However, to date no direct neurophysiological evidence that sensory prediction errors drive adaptation has been demonstrated. Here, we examined prediction errors via magnetoencephalography (MEG) imaging of the auditory cortex during sensorimotor adaptation of speech to altered auditory feedback, an entirely implicit adaptation task. Specifically, we measured how speaking-induced suppression (SIS)--a neural representation of auditory prediction errors--changed over the trials of the adaptation experiment. SIS refers to the suppression of auditory cortical response to speech onset (in particular, the M100 response) to self-produced speech when compared to the response to passive listening to identical playback of that speech. SIS was reduced (reflecting larger prediction errors) during the early learning phase compared to the initial unaltered feedback phase. Furthermore, reduction in SIS positively correlated with behavioral adaptation extents, suggesting that larger prediction errors were associated with more learning. In contrast, such a reduction in SIS was not found in a control experiment in which participants heard unaltered feedback and thus did not adapt. In addition, in some participants who reached a plateau in the late learning phase, SIS increased (reflecting smaller prediction errors), demonstrating that prediction errors were minimal when there was no further adaptation. Together, these findings provide the first neurophysiological evidence for the hypothesis that prediction errors drive human sensorimotor adaptation.

Mechanisms of sensorimotor adaptation in a hierarchical state feedback control model of speech.

Upon perceiving sensory errors during movements, the human sensorimotor system updates future movements to compensate for the errors, a phenomenon called sensorimotor adaptation. One component of this adaptation is thought to be driven by sensory prediction errors-discrepancies between predicted and actual sensory feedback. However, the mechanisms by which prediction errors drive adaptation remain unclear. Here, auditory prediction error-based mechanisms involved in speech auditory-motor adaptation were examined via the feedback aware control of tasks in speech (FACTS) model. Consistent with theoretical perspectives in both non-speech and speech motor control, the hierarchical architecture of FACTS relies on both the higher-level task (vocal tract constrictions) as well as lower-level articulatory state representations. Importantly, FACTS also computes sensory prediction errors as a part of its state feedback control mechanism, a well-established framework in the field of motor control. We explored potential adaptation mechanisms and found that adaptive behavior was present only when prediction errors updated the articulatory-to-task state transformation. In contrast, designs in which prediction errors updated forward sensory prediction models alone did not generate adaptation. Thus, FACTS demonstrated that 1) prediction errors can drive adaptation through task-level updates, and 2) adaptation is likely driven by updates to task-level control rather than (only) to forward predictive models. Additionally, simulating adaptation with FACTS generated a number of important hypotheses regarding previously reported phenomena such as identifying the source(s) of incomplete adaptation and driving factor(s) for changes in the second formant frequency during adaptation to the first formant perturbation. The proposed model design paves the way for a hierarchical state feedback control framework to be examined in the context of sensorimotor adaptation in both speech and non-speech effector systems.

High-Performing Atomic Electrocatalyst for Chlorine Evolution Reaction.

Electrocatalysts facilitating chlorine evolution reaction (ClER) play a vital role in chlor-alkali industries. Owing to a huge amount of chlorine consumed worldwide, inexpensive high-performing catalysts for Cl2 production are highly demanded. Here, a superb ClER catalyst fabricated through uniform dispersion of Pt single atoms (SAs) in C2 N2 moieties of N-doped graphene (denoted as Pt-1) is presented, which demonstrates near 100% exclusive ClER selectivity, long-term durability, extraordinary Cl2 production rate (3500 mmol h-1 gPt -1 ), and >140 000-fold increased mass activity over industrial electrodes in acidic medium. Excitingly, at the typical chlor-alkali industries' operating temperature (80 °C), Pt-1 supported on carbon paper electrode requires a near thermoneutral ultralow overpotential of 5 mV at 1 mA cm-2 current density to initiate the ClER, consistent with the predicted density functional theory (DFT) calculations. Altogether these results show the promising electrocatalyst of Pt-1 toward ClER.

Individual sensorimotor adaptation characteristics are independent across orofacial speech movements and limb reaching movements.

Sensorimotor adaptation is critical for human motor control but shows considerable interindividual variability. Efforts are underway to identify factors accounting for individual differences in specific adaptation tasks. However, a fundamental question has remained unaddressed: Is an individual's capability for adaptation effector system specific or does it reflect a generalized adaptation ability? We therefore tested the same participants in analogous adaptation paradigms focusing on distinct sensorimotor systems: speaking with perturbed auditory feedback and reaching with perturbed visual feedback. Each task was completed once with the perturbation introduced gradually (ramped up over 60 trials) and, on a different day, once with the perturbation introduced suddenly. Consistent with studies of each system separately, visuomotor reach adaptation was more complete than auditory-motor speech adaptation (80% vs. 29% of the perturbation). Adaptation was not significantly correlated between the speech and reach tasks. Moreover, considered within tasks, 1) adaptation extent was correlated between the gradual and sudden conditions for reaching but not for speaking, 2) adaptation extent was correlated with additional measures of performance (e.g., trial duration, within-trial corrections) only for reaching and not for speaking, and 3) fitting individual participant adaptation profiles with exponential rather than linear functions offered a larger benefit [lower root mean square error (RMSE)] for the reach task than for the speech task. Combined, results suggest that the ability for sensorimotor adaptation relies on neural plasticity mechanisms that are effector system specific rather than generalized. This finding has important implications for ongoing efforts seeking to identify cognitive, behavioral, and neurochemical predictors of individual sensorimotor adaptation.NEW & NOTEWORTHY This study provides the first detailed demonstration that individual sensorimotor adaptation characteristics are independent across articulatory speech movements and limb reaching movements. Thus, individual sensorimotor learning abilities are effector system specific rather than generalized. Findings regarding one effector system do not necessarily apply to other systems, different underlying mechanisms may be involved, and implications for clinical rehabilitation or performance training also cannot be generalized.

Fast atomic structure optimization with on-the-fly sparse Gaussian process potentials.

We apply on-the-fly machine learning potentials (MLPs) using the sparse Gaussian process regression (SGPR) algorithm for fast optimization of atomic structures. Great acceleration is achieved even in the context of a single local optimization. Although for finding the exact local minimum, due to limited accuracy of MLPs, switching to another algorithm may be needed. For random gold clusters, the forces are reduced to ∼0.1 eV Å-1within less than ten first-principles (FP) calculations. Because of highly transferable MLPs, this algorithm is specially suitable for global optimization methods such as random or evolutionary structure searching or basin hopping. This is demonstrated by sequential optimization of random gold clusters for which, after only a few optimizations, FP calculations were rarely needed.

Unveiling the Role of Charge Transfer in Enhanced Electrochemical Nitrogen Fixation at Single-Atom Catalysts on BX Sheets (X = As, P, Sb).

To tune single-atom catalysts (SACs) for effective nitrogen reduction reaction (NRR), we investigate various transition metals implanted on boron-arsenide (BAs), boron-phosphide (BP), and boron-antimony (BSb) using density functional theory (DFT). Interestingly, W-BAs shows high catalytic activity and excellent selectivity with an insignificant barrier of only 0.05 eV along the distal pathway and a surmountable kinetic barrier of 0.34 eV. The W-BSb and Mo-BSb exhibit high performances with limiting potentials of -0.19 and -0.34 V. The Bader-charge descriptor reveals that the charge transfers from substrate to *NNH in the first protonation step and from *NH3 to substrate in the last protonation step, circumventing a big hurdle in NRR by achieving negative free energy change of *NH2 to *NH3. Furthermore, machine learning (ML) descriptors are introduced to reduce computational cost. Our rational design meets the three critical prerequisites of chemisorbing N2 molecules, stabilizing *NNH, and destabilizing *NH2 adsorbates for high-efficiency NRR.

Sparse Gaussian Process Regression-Based Machine Learned First-Principles Force-Fields for Saturated, Olefinic, and Aromatic Hydrocarbons.

Universal machine learning (ML) interatomic potentials (IAPs) for saturated, olefinic, and aromatic hydrocarbons are generated by using the Sparse Gaussian process regression algorithm. The universal potentials are obtained by combining the potentials for the previously trained alkane/polyene systems and the potentials generated with the presently trained cyclic/aromatic hydrocarbon systems, along with the newly trained cross-terms between the two systems. The ML-IAPs have been trained using the PBE + D3 level of density functional theory for the on-the-fly adaptive sampling of various hydrocarbon molecules and these clusters composed of small molecules. We tested the ML-IAPs and found that they correctly predicted the structures and energies of the ß-carotene monomer and dimer. Also, the simulations of liquid ethylene reproduced the molecular volume and the simulations of toluene crystals reproduced higher stability of the α-phase over the ß-phase. These ab initio-level force-fields could eventually evolve toward universal organic/polymeric/biomolecular systems.

Perovskite solar cells with atomically coherent interlayers on SnO2 electrodes.

In perovskite solar cells, the interfaces between the perovskite and charge-transporting layers contain high concentrations of defects (about 100 times that within the perovskite layer), specifically, deep-level defects, which substantially reduce the power conversion efficiency of the devices1-3. Recent efforts to reduce these interfacial defects have focused mainly on surface passivation4-6. However, passivating the perovskite surface that interfaces with the electron-transporting layer is difficult, because the surface-treatment agents on the electron-transporting layer may dissolve while coating the perovskite thin film. Alternatively, interfacial defects may not be a concern if a coherent interface could be formed between the electron-transporting and perovskite layers. Here we report the formation of an interlayer between a SnO2 electron-transporting layer and a halide perovskite light-absorbing layer, achieved by coupling Cl-bonded SnO2 with a Cl-containing perovskite precursor. This interlayer has atomically coherent features, which enhance charge extraction and transport from the perovskite layer, and fewer interfacial defects. The existence of such a coherent interlayer allowed us to fabricate perovskite solar cells with a power conversion efficiency of 25.8 per cent (certified 25.5 per cent)under standard illumination. Furthermore, unencapsulated devices maintained about 90 per cent of their initial efficiency even after continuous light exposure for 500 hours. Our findings provide guidelines for designing defect-minimizing interfaces between metal halide perovskites and electron-transporting layers.

Machine Learning of First-Principles Force-Fields for Alkane and Polyene Hydrocarbons.

Machine learning (ML) interatomic potentials (ML-IAPs) are generated for alkane and polyene hydrocarbons using on-the-fly adaptive sampling and a sparse Gaussian process regression (SGPR) algorithm. The ML model is generated based on the PBE+D3 level of density functional theory (DFT) with molecular dynamics (MD) for small alkane and polyene molecules. Intermolecular interactions are also trained with clusters and condensed phases of small molecules. It shows excellent transferability to long alkanes and closely describes the ab inito potential energy surface for polyenes. Simulation of liquid ethane also shows reasonable agreement with experimental reports. This is a promising initiative toward a universal ab initio quality force-field for hydrocarbons and organic molecules.

Universal Machine Learning Interatomic Potentials: Surveying Solid Electrolytes.

We apply ab initio molecular dynamics (AIMD) with on-the-fly machine learning (ML) of interatomic potentials using the sparse Gaussian process regression (SGPR) algorithm for a survey of Li diffusivity in hundreds of ternary crystals as potential electrolytes for all-solid-state batteries. We show that models generated for these crystals can be easily combined for creating more general and transferable models which can potentially be used for simulating new materials without further training. As examples, universal potentials are created for Li-P-S and Li-Sb-S systems by combining the expert models of the crystals which contained the same set of elements. We also show that combinatorial models of different ternary crystals can be directly applied for modeling composite quaternary ones (e.g., Li-Ge-P-S). This hierarchical approach paves the way for modeling large-scale complexity by a combinatorial approach.

Two Liquid-Liquid Phase Transitions in Confined Water Nanofilms.

The stories behind supercooled bulk and confined water can be different. Bulk water has a metastable liquid-liquid phase transition at deeply supercooled conditions, but the existence of such a phenomenon in confined water is in question. Herein we show simulation results of first-order phase transitions between high- and low-density liquid (HDL and LDL) in confined water in both positive and negative pressures. A mid-density state between these two local states appears, which lets the transition show the hysteresis loop with transiently stable intermediate states. On the basis of Landau theory that we have adapted for mixing of moieties with high- and low-density states, we explain the phase transitions with the order parameter-dependent free energy change which is governed by second- to higher-order interactions among those moieties.

Supra-Binary Polarization in a Ferroelectric Nanowire.

The prediction and observation of supra-binary polarization in a ferroelectric nanowire (FNW) covered with a semicylindrical gate that provides an anisotropic electric field in the FNW are reported. There are gate-voltage-driven transitions between four polarization states in the FNW's cross-section, dubbed vertical-up, vertical-down, radial-in, and radial-out. They are determined by the interplay between the spatial depolarization energy and the free energy induced by an anisotropic external electric field, in clear distinction from the conventional film-based binary ferroelectricity. When the FNW is mounted on a biased graphene nanoribbon (GNR), these transitions induce exotic current-voltage hysteresis in the FNW-GNR transistor. This discovery suggests new operating mechanisms of ferroelectric devices. In particular, it enables intrinsic quaternary-digit information manipulation in parallel to the bit manipulation employed in conventional data storage.

Na/Al Codoped Layered Cathode with Defects as Bifunctional Electrocatalyst for High-Performance Li-Ion Battery and Oxygen Evolution Reaction.

The rational design of bifunctional electrocatalyst through simple synthesis with high activity remains a challenging task. Herein, Na/Al codoped Li-excess Li-Ru-Ni-O layered electrodes are demonstrated with defects/dislocations as an efficient bifunctional electrocatalyst toward lithium-ion battery (LIB) and oxygen evolution reaction (OER). Toward LIB cathode, specific capacity of 173 mAh g-1 (0.2C-rate), cyclability (>95.0%), high Columbic efficiency (99.2%), and energy efficiency (90.7%) are achieved. The codoped electrocatalyst has exhibited OER activity at a low onset potential (270 mV@10 mA cm-2 ), with a Tafel slope 69.3 mV dec-1 , and long-term stability over 36 h superior to the undoped and many other OER electrocatalysts including the benchmark IrO2 . The concurrent doping resides in the crystal lattice (where Na shows the pillaring effect to improve facile Li diffusion), Al improves the stabilization of the layered structure, and defective structures provide abundant active sites to accelerate OER reactions.

Speech auditory-motor adaptation to formant-shifted feedback lacks an explicit component: Reduced adaptation in adults who stutter reflects limitations in implicit sensorimotor learning.

The neural mechanisms underlying stuttering remain poorly understood. A large body of work has focused on sensorimotor integration difficulties in individuals who stutter, including recently the capacity for sensorimotor learning. Typically, sensorimotor learning is assessed with adaptation paradigms in which one or more sensory feedback modalities are experimentally perturbed in real time. Our own previous work on speech with perturbed auditory feedback revealed substantial auditory-motor learning limitations in both children and adults who stutter (AWS). It remains unknown, however, which subprocesses of sensorimotor learning are impaired. Indeed, new insights from research on upper limb motor control indicate that sensorimotor learning involves at least two distinct components: (a) an explicit component that includes intentional strategy use and presumably is driven by target error and (b) an implicit component that updates an internal model without awareness of the learner and presumably is driven by sensory prediction error. Here, we attempted to dissociate these components for speech auditory-motor learning in AWS versus adults who do not stutter (AWNS). Our formant-shift auditory-motor adaptation results replicated previous findings that such sensorimotor learning is limited in AWS. Novel findings are that neither control nor stuttering participants reported any awareness of changing their productions in response to the auditory perturbation and that neither group showed systematic drift in auditory target judgments made throughout the adaptation task. These results indicate that speech auditory-motor adaptation to formant-shifted feedback relies exclusively on implicit learning processes. Thus, limited adaptation in AWS reflects poor implicit sensorimotor learning. Speech auditory-motor adaptation to formant-shifted feedback lacks an explicit component: Reduced adaptation in adults who stutter reflects limitations in implicit sensorimotor learning.

Facile room-temperature self-assembly of extended cation-free guanine-quartet network on Mo-doped Au(111) surface.

Guanine-quadruplex, consisting of several stacked guanine-quartets (GQs), has emerged as an important category of novel molecular targets with applications from nanoelectronic devices to anticancer drugs. Incorporation of metal cations into a GQ structure is utilized to form stable G-quadruplexes, while formation of a cation-free GQ network has been challenging. Here we report the room temperature (RT) molecular self-assembly of extended pristine GQ networks on an Au(111) surface. An implanted molybdenum atom within the Au(111) surface is used to nucleate and stabilize the cation-free GQ network. Additionally, decoration of the Au(111) surface with 7-armchair graphene nanoribbons (7-AGNRs) enhances the GQ domain size by suppressing the influence of the disordered phase nucleated from Au step edges. Scanning tunneling microscopy/spectroscopy (STM/STS) and density functional theory (DFT) calculations confirm the formation of GQ networks and unravel the nucleation and growth mechanism. Our work, utilizing a hetero-atom doped substrate, provides a facile approach to enhance the stability and domain size of the GQ self-assembly, which would be applicable for other molecular structures.

Discrete constriction locations describe a comprehensive range of vocal tract shapes in the Maeda model.

The Maeda model was used to generate a large set of vocoid-producing vocal tract configurations. The resulting dataset (a) produced a comprehensive range of formant frequencies and (b) displayed discrete tongue body constriction locations (palatal, velar/uvular, and lower pharyngeal). The discrete parameterization of constriction location across the vowel space suggests this is likely a fundamental characteristic of the human vocal tract, and not limited to any specific set of vowel contrasts. These findings suggest that in addition to established articulatory-acoustic constraints, fundamental biomechanical constraints of the vocal tract may also explain such discreteness.

Anharmonicity-Driven Rashba Cohelical Excitons Break Quantum Efficiency Limitation.

Closed-shell light-emitting diodes (LEDs) suffer from the internal quantum efficiency (IQE) limitation imposed by optically inactive triplet excitons. Here, an unrevealed emission mechanism of lead halide perovskites (LHPs) APbX3 (A = Cs/CN2 H5 ; X = Cl/Br/I) that circumvents the efficiency limit of closed-shell LEDs is explored. Though efficient emission is prohibited by optically inactive J = 0 in inversion symmetric LHPs, the anharmonicity arising from stereochemistry of Pb and resonant orbital-bonding network along the imaginary A+… X- (T1u ) transverse optical (TO) modes, breaks inversion symmetry, introducing disorder and Rashba-Dresselhaus spin-orbit coupling (RD-SOC). This results in bright cohelical and dark antihelical excitons. Many-body theory and first-principles calculations show that the optically active cohelical exciton is the lowest excited state in organic/inorganic LHPs. Thus, RD-SOC can drive to achieve the ideal 50% IQE by utilizing anharmonicity, much over the 25% IQE limitation for closed-shell LEDs.