Search

In recent years, researchers have been exploring ways to generalize Boltzmann machines (BMs) to quantum systems, leading to the development of variations such as fully-visible and restricted quantum Boltzmann machines (QBMs). Due to the non-commuting nature of their Hamiltonians, restricted QBMs face trainability issues, whereas fully-visible QBMs have emerged as a more tractable option, as recent results demonstrate their sample-efficient trainability. These results position fully-visible QBMs as a favorable choice, offering potential improvements over fully-visible BMs without suffering from the trainability issues associated with restricted QBMs. In this work, we show that low-dimensional, fully-visible QBMs can learn to generate distributions typically associated with higher-dimensional systems. We validate our findings through numerical experiments on both artificial datasets and real-world examples from the high energy physics problem of jet event generation. We find that non-commuting terms and Hamiltonian connectivity improve the learning capabilities of QBMs, providing flexible resources suitable for various hardware architectures. Furthermore, we provide strategies and future directions to maximize the learning capacity of fully-visible QBMs., Comment: 13 pages, 7 figures; supplementary material 5 pages, 5 figures

The role of inhomegeneity in determining the properties of correlated electron systems is poorly understood because of the dearth of structural probes of disorder at the nanoscale. Advances in both neutron and x-ray scattering instrumentation now allow comprehensive measurements of diffuse scattering in single crystals over large volumes of reciprocal space, enabling structural correlations to be characterized over a range of length scales from 10~{\AA} to 200~{\AA} or more. When combined with new analysis tools, such as three-dimensional difference pair-distribution functions, these advanced capabilities have produced novel insights into the interplay of structural fluctuations and electronic properties in a broad range of correlated electron materials. This review describes recent investigations that have demonstrated the importance of understanding structural inhomogeneity pertaining to phenomena as diverse as superconductivity, charge-density-wave modulations, metal-insulator transitions, and multipolar interactions., Comment: 50 pages, 10 figures

Quantifying correlation and entanglement between molecular orbitals can elucidate the role of quantum effects in strongly correlated reaction processes. However, accurately storing the wavefunction for a classical computation of those quantities can be prohibitive. Here we use the Quantinuum H1-1 trapped-ion quantum computer to calculate von Neumann entropies which quantify the orbital correlation and entanglement in a strongly correlated molecular system relevant to lithium-ion batteries (vinylene carbonate interacting with an O$_2$ molecule). As shown in previous works, fermionic superselection rules decrease correlations and reduce measurement overheads for constructing orbital reduced density matrices. Taking into account superselection rules we further reduce the number of measurements by finding commuting sets of Pauli operators. Using low overhead noise reduction techniques we calculate von Neumann entropies in excellent agreement with noiseless benchmarks, indicating that correlations and entanglement between molecular orbitals can be accurately estimated from a quantum computation. Our results show that the one-orbital entanglement vanishes unless opposite-spin open shell configurations are present in the wavefunction., Comment: 20 pages, 7 figures

Though the effects of metal dopants on the electrostructural transition of rutile VO$_2$ have been studied for many decades, there is still no consensus explanation for the observed trends. A major challenge has been to separate the impact of a dopant's size from other factors such as its electronic configuration, which stems from the difficulty in directly probing the local bonding environment around a dopant atom. This work addresses the special case of larger dopant ions by combining X-ray total scattering experiments on V$_{0.83}$Mo$_{0.17}$O$_2$ and V$_{0.89}$Nb$_{0.11}$O$_2$ single crystals with multiple Monte Carlo method models to simulate local size effects in the high-temperature tetragonal phase (R). We find that sufficiently long apical metal-oxygen bonds (M$-$O$_{ap}$) induce a strain field in the neighboring chains that locally resembles the metal-metal dimer formation present in the low-temperature distorted structure of VO$_2$ (M1). The dimer mode in the M1 structure is antisymmetric along M$-$O$_{ap}$, however, while the strain-induced pseudodimer motif is symmetric. The implied direct competition between motifs is verified experimentally. This finding provides a new mechanistic parameter toward understanding the phase transition. More generally, the work highlights how local strain fields around dopants can lead to complex distortions that are ordinarily attributed to electronic origins., Comment: Article: 19 pages, 9 figures; Appendix: 4 pages, 2 figures, 3 tables

A fundamental step of any quantum algorithm is the preparation of qubit registers in a suitable initial state. Often qubit registers represent a discretization of continuous variables and the initial state is defined by a multivariate function. We develop protocols for preparing quantum states whose amplitudes encode multivariate functions by linearly combining block-encodings of Fourier and Chebyshev basis functions. Without relying on arithmetic circuits, quantum Fourier transforms, or multivariate quantum signal processing, our algorithms are simpler and more effective than previous proposals. We analyze requirements both asymptotically and pragmatically in terms of near/medium-term resources. Numerically, we prepare bivariate Student's t-distributions, 2D Ricker wavelets and electron wavefunctions in a 3D Coulomb potential, which are initial states with potential applications in finance, physics and chemistry simulations. Finally, we prepare bivariate Gaussian distributions on the Quantinuum H2-1 trapped-ion quantum processor using 24 qubits and up to 237 two-qubit gates., Comment: 22 pages, 11 figures

Dislocation engineering has the potential to open new avenues toward the exploration and modification of the properties of quantum materials. Strontium titanate (SrTiO3, STO) and potassium tantalate (KTaO3, KTO) are incipient ferroelectrics that show metallization and superconductivity at extremely low charge carrier concentrations, and have been the subject of resurgent interest. These materials also exhibit remarkable ambient-temperature ductility, and thus represent exceptional platforms for studies of the effects of deformation-induced dislocation structures on electronic properties. Recent work on plastically deformed STO revealed an enhancement of the superconducting transition temperature and the emergence of local ferroelectricity and magnetism near self-organized dislocation walls. Here we present a comprehensive structural analysis of plastically deformed STO and KTO, employing specially designed strain cells, diffuse neutron and x-ray scattering, Raman scattering, and nuclear magnetic resonance (NMR). Diffuse scattering and NMR provide insight into the dislocation configurations and densities and their dependence on strain. As in the prior work on STO, Raman scattering reveals evidence for local ferroelectric order near dislocation walls in plastically deformed KTO. Our findings provide valuable information about the self-organized defect structures in both materials, and they position KTO as a second model system with which to explore the associated emergent physics.

This technical report describes the design and training of novel speculative decoding draft models, for accelerating the inference speeds of large language models in a production environment. By conditioning draft predictions on both context vectors and sampled tokens, we can train our speculators to efficiently predict high-quality n-grams, which the base model then accepts or rejects. This allows us to effectively predict multiple tokens per inference forward pass, accelerating wall-clock inference speeds of highly optimized base model implementations by a factor of 2-3x. We explore these initial results and describe next steps for further improvements., Comment: Original upload 4/29/24, updated 6/6/24 with additional references to concurrent work

We design and implement quantum circuits for the simulation of the one-dimensional wave equation on the Quantinuum H1-1 quantum computer. The circuit depth of our approach scales as $O(n^{2})$ for $n$ qubits representing the solution on $2^n$ grid points, and leads to infidelities of $O(2^{-4n} t^{2})$ for simulation time $t$ assuming smooth initial conditions. By varying the qubit count we study the interplay between the algorithmic and physical gate errors to identify the optimal working point of minimum total error. Our approach to simulating the wave equation can readily be adapted to other quantum processors and serve as an application-oriented benchmark., Comment: 10 pages, 7 figures, 1 table

We present an approach for the data-driven modeling of nonlinear viscoelastic materials at small strains which is based on physics-augmented neural networks (NNs) and requires only stress and strain paths for training. The model is built on the concept of generalized standard materials and is therefore thermodynamically consistent by construction. It consists of a free energy and a dissipation potential, which can be either expressed by the components of their tensor arguments or by a suitable set of invariants. The two potentials are described by fully/partially input convex neural networks. For training of the NN model by paths of stress and strain, an efficient and flexible training method based on a recurrent cell, particularly a long short-term memory cell, is developed to automatically generate the internal variable(s) during the training process. The proposed method is benchmarked and thoroughly compared with existing approaches. These include a method that obtains the internal variable by integrating the evolution equation over the entire sequence, while the other method uses an an auxiliary feedforward neural network for the internal variable(s). Databases for training are generated by using a conventional nonlinear viscoelastic reference model, where 3D and 2D plane strain data with either ideal or noisy stresses are generated. The coordinate-based and the invariant-based formulation are compared and the advantages of the latter are demonstrated. Afterwards, the invariant-based model is calibrated by applying the three training methods using ideal or noisy stress data. All methods yield good results, but differ in computation time and usability for large data sets. The presented training method based on a recurrent cell turns out to be particularly robust and widely applicable and thus represents a promising approach for the calibration of other types of models as well., Comment: 21 pages, 16 figures

We present a comprehensive magnetic and structural phase diagram for $\mathrm{Y_{1-x}Lu_xBaCo_4O_7}$, established through neutron diffraction and magnetization measurements. Our results outline the evolution of various nuclear structures, encompassing both orthorhombic and monoclinic symmetries, in response to changes in composition and temperature. The phase transition temperature of the orthorhombic phase, $T_{s1}$, decreases with increasing Lu content from 310 K for x = 0.0 to 110 K for x = 1.0. In Lu-rich compositions (0.7 \le x \le 1.0), first-order structural transitions are observed with coexisting and competing orthorhombic Pbn$2_1$ and metastable monoclinic Cc phases. Composition- and temperature-dependent refinements of the magnetic structure reveal an antiferromagnetic arrangement of Co spin pairs, below the magnetic transition temperatures of the Y-rich compositions, in the ab plane within both the triangular and Kagom\'e layers. A gradual suppression of long-range magnetic order is observed with increasing the Lu content, accompanied by the development of short-range magnetic correlations present in all the samples.

The van der Waals layered magnet $\alpha$-RuCl$_3$ offers tantalizing prospects for the realization of Majorana quasiparticles. Efforts to understand this are, however, hampered by inconsistent magnetic and thermal transport properties likely coming from the formation of structural disorder during crystal growth, postgrowth processing, or upon cooling through the first order structural transition. Here, we investigate structural disorder in $\alpha$-RuCl$_3$ using x-ray diffuse scattering and three-dimensional difference pair distribution function (3D-$\Delta$PDF) analysis. We develop a quantitative model that describes disorder in $\alpha$-RuCl$_3$ in terms of rotational twinning and intermixing of the high and low-temperature structural layer stacking. This disorder may be important to consider when investigating the detailed magnetic and electronic properties of this widely studied material., Comment: 6 pages; 3 figures; accepted in Physical Review B

Multi-layer Ti3C2Tx coatings have demonstrated an outstanding wear performance with excellent durability due to beneficial tribo-layers formed. However, the involved formation processes dependent on the tribological conditions and coating thickness are yet to be fully explored. Therefore, we spray-coated Ti3C2Tx multi-layer particles onto stainless steel substrates to create coatings with two different thicknesses and tested their solid lubrication performance with different normal loads (100 and 200 mN) and sliding frequencies (1 and 2.4 Hz) using linear-reciprocating ball-on-disk tribometry. We demonstrate that MXenes' tribological performance depends on their initial state (delaminated few-layer vs. multi-layer particles), coating thickness, applied load and sliding frequency. Specifically, the best behavior is observed for thinner multi-layer coatings tested at the lower frequency. In contrast, coatings made of delaminated few-layer MXene are not as effective as their multi-layer counterparts. Our high-resolution interface characterization by transmission electron microscopy revealed unambiguous differences regarding the uniformity and chemistry of the formed tribo-layers as well as the degree of tribo-induced MXenes' exfoliation. Atomistic insights into the exfoliation process and molecular dynamic simulations quantitatively backed up our experimental results regarding coating thickness and velocity dependency. This ultimately demonstrates that MXenes' tribological performance is governed by the underlying tribo-chemistry and their exfoliation ability during rubbing.

Multi-frame algorithms for single-channel speech enhancement are able to take advantage from short-time correlations within the speech signal. Deep Filtering (DF) was proposed to directly estimate a complex filter in frequency domain to take advantage of these correlations. In this work, we present a real-time speech enhancement demo using DeepFilterNet. DeepFilterNet's efficiency is enabled by exploiting domain knowledge of speech production and psychoacoustic perception. Our model is able to match state-of-the-art speech enhancement benchmarks while achieving a real-time-factor of 0.19 on a single threaded notebook CPU. The framework as well as pretrained weights have been published under an open source license., Comment: Accepted as show and tell demo to interspeech 2023

Multi-frame algorithms for single-channel speech enhancement are able to take advantage from short-time correlations within the speech signal. Deep filtering (DF) recently demonstrated its capabilities for low-latency scenarios like hearing aids with its complex multi-frame (MF) filter. Alternatively, the complex filter can be estimated via an MF minimum variance distortionless response (MVDR), or MF Wiener filter (WF). Previous studies have shown that incorporating algorithm domain knowledge using an MVDR filter might be beneficial compared to the direct filter estimation via DF. In this work, we compare the usage of various multi-frame filters such as DF, MF-MVDR, or MF-WF for HAs. We assess different covariance estimation methods for both MF-MVDR and MF-WF and objectively demonstrate an improved performance compared to direct DF estimation, significantly outperforming related work while improving the runtime performance., Comment: Submitted to Interspeech 2023

A true understanding of the properties of pnictide superconductors require the development of high-quality materials and performing measurements designed to unravel their intrinsic properties and short-range nematic correlations which are often obscured by extrinsic effects such as poor crystallinity, inhomogeneity, domain formation and twinning. In this paper, we report the systematic growth of high-quality Na-substituted BaFe2As2 single crystals and their characterization using pulsed-magnetic fields x-ray diffraction and x-ray diffuse scattering. Analysis of the properties and compositions of the highest quality crystals show that their actual Na stoichiometry is about 50-60% of the nominal content and that the targeted production of crystals with specific compositions is accessible. We derived a reliable equation to estimate the Na stoichiometry based on the measured superconducting TC of these materials. Attempting to force spin reorientation and induce tetragonality, orthorhombic Ba1-xNaxFe2As2 single crystals subjected to out-of-plane magnetic fields up to 31.4T are found to exhibit strong in-plane magnetic anisotropy demonstrated by the insufficiency of such high fields in manipulating the relative population of their twinned domains or in suppressing the orthorhombic order. Broad x-ray diffuse intensity rods observed at temperatures between 30 K and 300 K uncover short-range structural correlations. Local structure modeling together with 3D-{\Delta}PDF mapping of real-space interatomic vectors show that the diffuse scattering arises from in-plane short-range chemical correlations of the Ba and Na atoms coupled with short-range atomic displacements within the same plane due to an effective size difference between the two atomic species.

The mathematical formulation of constitutive models to describe the path-dependent, i.e., inelastic, behavior of materials is a challenging task and has been a focus in mechanics research for several decades. There have been increased efforts to facilitate or automate this task through data-driven techniques, impelled in particular by the recent revival of neural networks (NNs) in computational mechanics. However, it seems questionable to simply not consider fundamental findings of constitutive modeling originating from the last decades research within NN-based approaches. Herein, we propose a comparative study on different feedforward and recurrent neural network architectures to model inelasticity. Within this study, we divide the models into three basic classes: black box NNs, NNs enforcing physics in a weak form, and NNs enforcing physics in a strong form. Thereby, the first class of networks can learn constitutive relations from data while the underlying physics are completely ignored, whereas the latter two are constructed such that they can account for fundamental physics, where special attention is paid to the second law of thermodynamics in this work. Conventional linear and nonlinear viscoelastic as well as elastoplastic models are used for training data generation and, later on, as reference. After training with random walk time sequences containing information on stress, strain, and, for some models, internal variables, the NN-based models are compared to the reference solution, whereby interpolation and extrapolation are considered. Besides the quality of the stress prediction, the related free energy and dissipation rate are analyzed to evaluate the models. Overall, the presented study enables a clear recording of the advantages and disadvantages of different NN architectures to model inelasticity and gives guidance on how to train and apply these models.