Strongly correlated physics arises from electron-electron scattering within partially filled orbitals. Organic molecules in open-shell configurations are therefore good candidates to exhibit many-body effects. We focus on electron transport in a two-terminal single-molecule junction setup, in which the molecular bridge consists of an organic radical with a molecular orbital hosting a single unpaired electron [singly occupied molecular orbital (SOMO)]. We perform beyond state-of-the-art numerical simulations combining an ab initio description of the chemical environment with quantum field-theoretical techniques that account for many-body effects. The key observation is that the SOMO resonance is prone to splitting, and we identify a giant electronic scattering rate as the driving many-body mechanism, akin to that of the Mott metal-to-insulator transition. By comparing linear and cyclic radicals, we show that the spatial distribution of the SOMO and its projection on the molecular backbone have dramatic consequences for the transport properties of the junction. We argue that the phenomenon and the underlying microscopic mechanism apply to a broad family of open-shell molecular systems, and can explain puzzling experimental observations such as suppressed conductance in radical junctions.

Modern computers perform pre-defined operations using static memory components, whereas biological systems learn through inherently dynamic, time-dependent processes in synapses and neurons. The biological learning process also relies on global signals-neuromodulators-that influence many synapses at once, depending on their dynamic, internal state. In this study, using optical radiation as a global neuromodulatory signal, nanoscale SrTiO3 (STO) memristors that can act as solid-state synapses are investigated. It is observed that the memristor's photo-conductance exhibits a long-term decay process after photoexcitation (10s of seconds) with an activation energy of 0.33 eV. Based on density functional theory calculations, this long-term photoresponse is attributed to the generation and migration of oxygen vacancies at the Pt-SrTiO3 interface. Additionally, the photo-conductance decay can be accurately controlled through an electrical bias signal and the magnitude of the memristor's photoresponse depends on its electrical conductance state, following a well-defined square root relation. These properties, in combination with the device's low power operation (< 1pJ per optical pulse) and small measurement variability, may pave the way for space- and energy-efficient implementations of complex biological learning processes in electro-optical hardware.

Using an in-house Schroedinger-Poisson (SP) solver, we investigate the creation of a single hole spin qubit inside a triple-gate triangular silicon fin field effect transistor (Si FinFET) quantum device similar to experimental structures. The gate induced formation of the required quantum dot (QD) is monitored based on the Luttinger-Kohn 6x6 kp method accounting for magnetic fields and strain to determine the qubit ground state. Strain arises from the inhomogeneous contraction of the different FinFET components when they are cooled down to cryogenic temperatures. It leads to a renormalization of the qubit energy levels, thus impacting both the heavy-hole (HH) and light-hole (LH) populations as well as their mixing. The dot length, band mixing, g-factor, and Larmor/Rabi frequencies of the considered device are extracted. In particular, we show that these metrics exhibit strong strain-dependent variations of their magnitude, thus underlying the importance of including realistic thermal contraction scenarios when modeling hole spin qubits.

Equivariant Graph Neural Networks (eGNNs) trained on density-functional theory (DFT) data can potentially perform electronic structure prediction at unprecedented scales, enabling investigation of the electronic properties of materials with extended defects, interfaces, or exhibiting disordered phases. However, as interactions between atomic orbitals typically extend over 10+ angstroms, the graph representations required for this task tend to be densely connected, and the memory requirements to perform training and inference on these large structures can exceed the limits of modern GPUs. Here we present a distributed eGNN implementation which leverages direct GPU communication and introduce a partitioning strategy of the input graph to reduce the number of embedding exchanges between GPUs. Our implementation shows strong scaling up to 128 GPUs, and weak scaling up to 512 GPUs with 87

Valence change memory (VCM) cells based on SrTiO_3 (STO), a perovskite oxide, are a promising type of emerging memory device. While the operational principle of most VCM cells relies on the growth and dissolution of one or multiple conductive filaments, those based on STO are known to exhibit a distinctive, 'interface-type' switching, which is associated with the modulation of the Schottky barrier at their active electrode. Still, a detailed picture of the processes that lead to interface-type switching is not available. In this work, we use a fully atomistic and ab initio model to study the resistive switching of a Pt-STO-Ti stack. We identify that the termination of the crystalline STO plays a decisive role in the switching mechanism, depending on the relative band alignment between the material and the Pt electrode. In particular, we show that the accumulation of oxygen vacancies at the Pt side can be at the origin of resistive switching in TiO_2-terminated devices by lowering the conduction band minimum of the STO layer, thus facilitating transmission through the Schottky barrier. Moreover, we investigate the possibility of filamentary switching in STO and reveal that it is most likely to occur at the Pt electrode of the SrO-terminated cells.

Designing nanoscale electronic devices such as the currently manufactured nanoribbon field-effect transistors (NRFETs) requires advanced modeling tools capturing all relevant quantum mechanical effects. State-of-the-art approaches combine the non-equilibrium Green's function (NEGF) formalism and density functional theory (DFT). However, as device dimensions do not exceed a few nanometers anymore, electrons are confined in ultra-small volumes, giving rise to strong electron-electron interactions. To account for these critical effects, DFT+NEGF solvers should be extended with the GW approximation, which massively increases their computational intensity. Here, we present the first implementation of the NEGF+GW scheme capable of handling NRFET geometries with dimensions comparable to experiments. This package, called QuaTrEx, makes use of a novel spatial domain decomposition scheme, can treat devices made of up to 84,480 atoms, scales very well on the Alps and Frontier supercomputers (>80% weak scaling efficiency), and sustains an exascale FP64 performance on 42,240 atoms (1.15 Eflop/s).

Advancements in memristive devices have given rise to a new generation of specialized hardware for bio-inspired computing. However, most of these implementations draw only partial inspiration from the architecture and functionalities of the mammalian brain. Moreover, the use of memristive hardware is typically restricted to specific elements within the learning algorithm, leaving computationally expensive operations to be executed in software. Here we demonstrate reinforcement learning through an actor–critic temporal difference algorithm implemented on analogue memristors, mirroring the principles of reward-based learning in a neural network architecture similar to the one found in biology. Memristors are used as multipurpose elements within the learning algorithm: they act as synaptic weights that are trained online, they calculate the weight updates associated with the temporal difference error directly in hardware and they determine the actions to navigate the environment. Owing to these features, weight training can take place entirely in memory, eliminating data movement. We test our framework on two navigation tasks—the T-maze and the Morris water maze—using analogue memristors based on the valence change memory effect. Our approach represents the first step towards fully in-memory and online neuromorphic computing engines based on bio-inspired learning schemes. A framework based on actor–critic temporal difference learning and employing a biologically plausible network architecture that mimics reward-based learning on memristors and enables full in-memory training for navigation tasks is discussed.

This paper introduces a novel integration method of localized metallic back-gates into fully-depleted silicon-on-insulator (FDSOI) multi-gate FETs, enabling robust front-to-back electrostatic coupling from room temperature to cryogenic conditions, without the need for substrate implantation. The fabrication process, termed the Nanomole process, utilizes nanometric vapor-phase etching of the buried oxide or silicon substrate with vapor-HF and XeF2 gases. This is followed by atomic layer deposition (ALD) of a dielectric material and Pt, with precise patterning achieved through inductively coupled plasma etching. Detailed analysis of the process demonstrates controllable etch rates based on device geometry, providing calibrated guidelines for scalable manufacturing. Symmetric mid-k dual-gating is reported in devices featuring a Si-film thickness of 24 nm, with a top and bottom gate oxide equivalent thickness (EOT) of 6.5 nm. Electrical characterization of multi-gate FDSOI SETs, operated as FETs, confirms effective threshold voltage tuning through dual-gate operation, with consistent performance from room temperature to millikelvin regimes. Additionally, quantum mechanical simulations based on the effective mass approximation at 4 K offer insights into the electrostatic behavior of dual-gated SOI quantum dot devices in both planar and nanowire geometries. This scalable and versatile technological solution opens new possibilities for advanced quantum devices, such as charge and spin qubits, by enabling in situ control over volume inversion, electron valley splitting, and spin-orbit interaction.

We present an ab initio method to calculate the clamped Pockels tensor of ferroelectric materials from density-functional theory, the modern theory of polarization exploiting the electric-enthalpy functional, and automated first- and second-order finite-difference derivatives of the polarizations and the Hellmann-Feynman forces. Thanks to the functional-independent capabilities of our approach, we can determine the Pockels tensor of tetragonal barium titanate (BTO) beyond the local density approximation (LDA), with arbitrary exchange-correlation (XC) functionals, for example, PBEsol. The latter, together with RRKJ ultra-soft pseudo-potentials (PP) and a supercell exhibiting local titanium off-centering, enables us to stabilize the negative optical phonon modes encountered in tetragonal BTO when LDA and norm-conserving PP are combined. As a result, the correct value range of r_51, the largest experimental Pockels coefficient of BTO, is recovered. We also reveal that r_51 increases with decreasing titanium off-centering for this material. The lessons learned from the structural, dielectric, and vibrational investigations of BTO will be essential to design next-generation electro-optical modulators based on the Pockels effect.

The industrial-scale growth of dielectrics on top of a 2D material transistor channel without deterioration of its transport characteristics remains challenging today. Here, we investigate the origin of the performance degradation issue by constructing several atomistic interface models between a WS2 monolayer and an amorphous Al2O3 or HfO2 thin film. We then computed their properties using first-principles methods. We show that, while it is in principle possible to achieve a van der Waals interface between these materials, surface defects (e.g., undercoordinated metal atoms at the surface) are detrimental since they create localized states close to the bottom of the conduction band of WS2. Even in their absence, the inhomogeneity of the surface topology creates a nonuniform potential that is felt by charge carriers in WS2. While surface defects can potentially be kept under control with an appropriate oxide choice, the surface inhomogeneity appears to act as a bottleneck, limiting the performance of WS2 as a transistor channel and, in general, for all 2D materials.

Graph neural networks (GNNs) have shown promise in learning the ground-state electronic properties of materials, subverting ab initio density functional theory (DFT) calculations when the underlying lattices can be represented as small and/or repeatable unit cells (i.e., molecules and periodic crystals). Realistic systems are, however, non-ideal and generally characterized by higher structural complexity. As such, they require large (10+ angstrom) unit cells and thousands of atoms to be accurately described. At these scales, DFT becomes computationally prohibitive, making GNNs especially attractive. In this work, we present a strictly local equivariant GNN capable of learning the electronic Hamiltonian (H) of realistically extended materials. It incorporates an augmented partitioning approach that enables training on arbitrarily large structures while preserving local atomic environments beyond boundaries. We demonstrate its capabilities by predicting the electronic Hamiltonian of various systems with up to 3,000 nodes (atoms), 500,000+ edges, similar to 28 million orbital interactions (nonzero entries of H), and <= 0.53% error in the eigenvalue spectra. Our work expands the applicability of current electronic property prediction methods to some of the most challenging cases encountered in computational materials science, namely systems with disorder, interfaces, and defects.

The inversion of structured sparse matrices is a key but computationally and memory-intensive operation in many scientific applications. There are cases, however, where only particular entries of the full inverse are required. This has motivated the development of so-called selected-inversion algorithms, capable of computing only specific elements of the full inverse. Currently, most of them are either shared-memory codes or limited to CPU implementations. Here, we introduce Serinv, a scalable library providing distributed, GPU-based algorithms for the selected inversion and Cholesky decomposition of positive-definite, block-tridiagonal arrowhead matrices. This matrix class is highly relevant in statistical climate modeling and materials science applications. The performance of Serinv is demonstrated on synthetic and real datasets from statistical air temperature prediction models. In our numerical tests, Serinv achieves 32.3 two orders of magnitude speedup over the sparse direct solvers PARDISO and MUMPS on 16 GPUs.

Memristive devices have drawn significant interest due to their use in novel paradigms such as neuromorphic computing. Neuromorphic systems are developed by implementing artificial neurons and synapses on a hardware level. Hence, memristors with multipurpose and reconfigurable neuromorphic functionalities could be highly beneficial in the design process. In this study, we experimentally verify that both neuronal and synaptic functions can be implemented on a single memristor. By controlling the device current at two different levels, the memristor operates in either a volatile or a nonvolatile retention regime. These two operation regimes are essential to mimic neuronal or synaptic behavior. Towards this end, we use an alloyed filamentary memristor (AgSn/SiO2/Pt) composed of ions with differing mobilities enabling both integrate and fire (IF) operation in the volatile regime and synaptic weights in the nonvolatile regime. By only changing the current compliance, these devices switch reliably between the aforementioned retention regimes. Additionally, our proposed training method significantly improves switching variability in the volatile regime. We show how the mean set voltage statistically reduce from 1.2 to 0.2 V; and the standard deviation of the set voltages reduced from 0.52 to 0.03 V.

Transition metal dichalcogenides (TMDCs) are promising candidates for future nano-transistor channels due to their outstanding intrinsic transport properties. However, their electron mobility is highly sensitive to the surrounding dielectric, often falling well below theoretical expectations. In this work, we explore how a stacked Al2O3 dielectric affects electron mobility in monolayer WS2 using first-principles quantum transport simulations. We identify that fluctuations in the electrostatic potential, arising from the disordered structure of Al2O3, significantly degrade mobility, especially when WS2 interfaces with under-coordinated aluminum atoms. Our calculated mobilities (≃1–30 cm2/(V ⋅ s)) align with experimental observations and remain far from the ideal limit (≃300 cm2/(V ⋅ s)). We further demonstrate that encapsulating WS2 with hexagonal boron nitride (hBN) or employing a crystalline oxide can recover high mobility values. However, these strategies introduce trade-offs in electrostatic control and fabrication complexity, underlining the need for careful dielectric engineering in TMDC-based devices.

The inversion of structured sparse matrices is a fundamental yet computationally and memory-intensive task in many scientific applications, such as Bayesian statistical modeling and material science. In certain cases, only particular entries of the full inverse are required. This has motivated the development of so-called selected inversion algorithms (SIA), capable of computing only specific elements of the full inverse. Currently, most SIA implementations are restricted to shared-/distributed-memory CPU architectures or to single GPUs. Here, we introduce novel numerical methods to perform the parallel selected inversion and Cholesky decomposition of positive-definite, block-tridiagonal with arrowhead matrices. A distributed memory, GPU-accelerated implementation of our approach is presented and integrated into the structured solver library Serinv. We demonstrate its performance on synthetic and real datasets from statistical air temperature prediction models and achieve CPU (GPU) speedups of up to 2.6x (71.4x) over the SIA of the PARDISO library and up to 14x (380.9x) over the MUMPS library, when scaling to 16 processes.

We propose a spin-charge qubit based on a bilayer graphene and WSe2 van der Waals heterostructure that together form a quantum dot and demonstrate its functionality from first-principles simulations. Electron and hole confinement as well as electrically controllable spin-orbit coupling (SOC) are modeled by self-consistently solving the Schrödinger and Poisson equations with material parameters extracted from density functional theory as inputs. In both electron and hole quantum dots, we find a two orders of magnitude enhancement of SOC (1.8 meV) compared to intrinsic graphene, in the layer directly adjacent to WSe2. Time-dependent investigations of the quantum device reveal rapid qubit gate operation in the order of picoseconds. Our simulations indicate that bilayer graphene and WSe2 heterostructures provide a promising platform for the processing of quantum information.

Filamentary Resistive Random Access Memory (RRAM) devices have so far been scaled down to a 2.1 nm [1] feature size, 2 nm thickness [2], and integrated with advanced transistor nodes for active cell selection [5]. Evaluating their ultimate integration density now involves understanding the scales at which switching can be reliably achieved in matrix-like structures, and the associated failure mechanisms. Previous theoretical studies explored ultrascaled RRAM based on static electronic properties and fixed filament geometry [2], or continuum/regular domains [4, 6]. While useful, these approaches do not simultaneously capture the underlying structural irregularity of the switching material, and the finite size/number of atoms involved in the switching process, both of which heavily influence the kinetics of atomic relocation and the resulting current flow [3]. Here we leverage a custom atomistic simulation tool designed to capture such effects to investigate the failure mechanisms which emerge in high-density passive RRAM arrays, as well as insight on how they can be mitigated.

Comprehensive device simulations reveal the potential of ultra-scaled field-effect transistors based on two-dimensional channel materials.

Biological neural networks do not only include long-term memory and weight multiplication capabilities, as commonly assumed in artificial neural networks, but also more complex functions such as short-term memory, short-term plasticity, and meta-plasticity - all collocated within each synapse. Here, we demonstrate memristive nano-devices based on SrTiO3 that inherently emulate all these synaptic functions. These memristors operate in a non-filamentary, low conductance regime, which enables stable and energy efficient operation. They can act as multi-functional hardware synapses in a class of bio-inspired deep neural networks (DNN) that make use of both long- and short-term synaptic dynamics and are capable of meta-learning or "learning-to-learn". The resulting bio-inspired DNN is then trained to play the video game Atari Pong, a complex reinforcement learning task in a dynamic environment. Our analysis shows that the energy consumption of the DNN with multi-functional memristive synapses decreases by about two orders of magnitude as compared to a pure GPU implementation. Based on this finding, we infer that memristive devices with a better emulation of the synaptic functionalities do not only broaden the applicability of neuromorphic computing, but could also improve the performance and energy costs of certain artificial intelligence applications.