Observable Error Bounds of the Time-Splitting Scheme for Quantum-Classical Molecular Dynamics
Not provided.
SEARCH · Search NASA
Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.
Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.
Not provided.
Remote access to large-scale scientific datasets, like those generated by combustion simulations or other high-performance computing (HPC) applications, presents a significant challenge. Downloading entire datasets is often impractical due to their size and the bandwidth limitations of typical networks. To address this challenge, we propose a novel approach that enables efficient remote access to large datasets distributed across multiple facilities. Our method enables technologies to download only the data values of a select variable, in a select region of interest, to a user-defined accuracy. For this purpose, we extended the ADIOS IO library to provide read functions with user-defined accuracy, a remote data server that understands multidimensional selections of specific variables, steps and accuracy from an ADIOS dataset, and which uses lossy compression on the remote site to reduce the data to be transferred back to the client. In addition, our extension of the ADIOS library collects metadata from multiple datasets in small files called Campaign Archives, which can be shared among project participants on any HPC, cloud or laptop, and which can easily facilitate the discovery of content and pointers to the data location as well as remote access to the data by local tools as if data was local. This feature called Campaign Management, enables a group of scientists to manage related datasets stored in multiple files, across multiple facilities as if it was in a single file/database. We demonstrate the effectiveness of our approach using a 1.5 TB dataset from the S3D combustion simulation on Frontier at the Oak Ridge Leadership Facility. Even a single variable from this dataset, at 64 GB, is too large to be processed on a standard laptop. We show two different reading patterns for 2D plots and 3D visualization, with careful settings that a scientist studying combustion data would do and show that running the same Python scripts on Frontier directly takes comparable time than running them on the local laptop with remote access to the data on Frontier.
A verification tool receives a finite precision definition for an approximation of an infinite precision numerical function implemented in a processor in the form of a polynomial of bounded functions. The verification tool receives a domain for verifying outputs of segments associated with the infinite precision numerical function. The verification tool splits the domain into at least two segments, wherein each segment is non-overlapping with any other segment and converts, for each segment, a polynomial of bounded functions for the segment to a simplified formula comprising a polynomial, an inequality, and a constant for a selected segment. The verification tool calculates upper bounds of the polynomial for the at least two segments, beginning with the selected segment and reports the segments that violate a bounding condition.
Explore the source record for details and available documents.
In this technical report we investigate efficient methods for numerical simulation of active suspensions. The prototypical system is a suspension of swimming bacteria in a Newtonian fluid. Rheological and other macroscopic properties of such suspensions can differ dramatically from the same properties of the suspending fluid alone or of suspensions of similar but inactive particles. Elongated bacteria, such as E. coli or B. subtilis, swim along their principal axis, propelling themselves with the help of flagella, attached at the anterior of the organism and pushing it forward in the manner of a propeller. They interact hydrodynamically with the surrounding fluid and, because of their asymmetrical shape, have the propensity to align with the local flow. This, along with the dipolar nature of bacteria (the two forces a bacterium exerts on a fluid - one due to self-propulsion and the other opposing drag - have equal magnitude and point in opposite directions), causes nearby bacteria to tend to align, resulting in a intermittent local ordering on the mesoscopic scale, which is between the microscopic scale of an individual bacterium and the macroscopic scale of the suspension (e.g., its container). The local ordering is sometimes called a collective mode or collective swimming. Thanks to self-propulsion, collective modes inject momentum into the fluid in a coherent way. This enhances the local strain rate without changing the macroscopic stress applied at the boundary of the container. The macroscopic effective viscosity of the suspension is defined roughly as the ratio of the applied stress to the bulk strain rate. If local alignment and therefore local strain-rate enhancement, are significant, the effective viscosity can be appreciably lower than that of the corresponding passive suspension or even of the surrounding fluid alone. Indeed, a sevenfold decrease in the effective viscosity was observed in experiments with B. subtilis. More generally, local collective swimming resulting from bacterial alignment can significantly alter other macroscopic properties of the suspension, such as the oxygen diffusivity and mixing rates. In order to understand the unique macroscopic properties of active suspensions the connection between microscopic swimming and alignment dynamics and the mesoscopic pattern formation must be clarified. This is difficult to do analytically in the fully general setting of moderately dense suspensions, because of the large number of bacteria involved (approx. 10 10 cm -3 in experiments) and the complex, time-dependent geometry of the system. Many reduced analytical models of bacterial have been proposed, but all of them require validation. While comparison with experiment is the ultimate test of a model's fidelity, it is difficult to conduct experiments matched to these models assumptions. Numerical simulation of the microscopic dynamics is an acceptable substitute, but it runs into the problem of having to discretize the fluid domain with a fine-grained boundary (the bacteria) and update the discretization as the domain evolves (bacteria move). This leads to a prohibitively high number of degrees of freedom and prohibitively high setup costs per timestep of simulation. In this technical report we propose numerical methods designed to alleviate these two difficulties. We indicate how to (1) construct an optimal discretization in terms of the number of degrees of freedom per digit of accuracy and (2) optimally update the discretization as the simulation evolves. The technical tool here is the derivation of rigorous error bounds on the error in the numerical solution when using our proposed discretization at the initial time as well as after a given elapsed simulation time. These error bounds should guide the construction of practical discretization schemes and update strategies. Our initial construction is carried out by using a theoretically convenient, but practically prohibitive spectral basis, which is a Galerkin basis of functions with global support. At the end of this report we propose localization techniques while maintaining acceptable error bounds. No numerical experiments were conducted as part of this study, but we envision that we may undertake such studies and further development of the method, jointly or individually.
The next-generation radio astronomy instruments are providing a massive increase in sensitivity and coverage, largely through increasing the number of stations in the array and the frequency span sampled. The two primary problems encountered when processing the resultant avalanche of data are the need for abundant storage and the constraints imposed by I/O, as I/O bandwidths drop significantly on cold storage. An example of this is the data deluge expected from the SKA Telescopes of more than 60 PB per day, all to be stored on the buffer filesystem. While compressing the data is an obvious solution, the impacts on the final data products are hard to predict. In this paper, we chose an error-controlled compressor – MGARD – and applied it to simulated SKA-Mid and real pathfinder visibility data, in noise-free and noise-dominated regimes. As the data have an implicit error level in the system temperature, using an error bound in compression provides a natural metric for compression. MGARD ensures the compression incurred errors adhere to the user-prescribed tolerance. To measure the degradation of images reconstructed using the lossy compressed data, we proposed a list of diagnostic measures, exploring the trade-off between these error bounds and the corresponding compression ratios, as well as the impact on science quality derived from the lossy compressed data products through a series of experiments. We studied the global and local impacts on the output images for continuum and spectral line examples. We found relative error bounds of as much as 10%, which provide compression ratios of about 20, have a limited impact on the continuum imaging as the increased noise is less than the image RMS, whereas a 1% error bound (compression ratio of 8) introduces an increase in noise of about an order of magnitude less than the image RMS. For extremely sensitive observations and for very precious data, we would recommend a 0.1% error bound with compression ratios of about 4. These have noise impacts two orders of magnitude less than the image RMS levels. At these levels, the limits are due to instabilities in the deconvolution methods. We compared the results to the alternative compression tool DYSCO, in both the impacts on the images and in the relative flexibility. MGARD provides better compression for similar error bounds and has a host of potentially powerful additional features.
The increase of computer processing speed is significantly outpacing improvements in network and storage bandwidth, leading to the big data challenge in modern science, where scientific applications can quickly generate much more data than that can be transferred and stored. As a result, big scientific data must be reduced by a few orders of magnitude while the accuracy of the reduced data needs to be guaranteed for further scientific explorations. Moreover, scientists are often interested in some specific spatial/temporal regions in their data, where higher accuracy is required. The locations of the regions requiring high accuracy can sometimes be prescribed based on application knowledge, while other times they must be estimated based on general spatial/temporal variation. In this paper, we develop a novel multilevel approach which allows users to impose region-wise compression error bounds. Our method utilizes the byproduct of a multilevel compressor to detect regions where details are rich and we provide the theoretical underpinning for region-wise error control. With spatially varying precision preservation, our approach can achieve significantly higher compression ratios than single-error bounded compression approaches and control errors in the regions of interest.We conduct the evaluations on two climate use cases – one targeting small-scale, node features and the other focusing on long, areal features. For both use cases, the locations of the features were unknown ahead of the compression. By selecting approximately 16% of the data based on multi-scale spatial variations and compressing those regions with smaller error tolerances than the rest, our approach improves the accuracy of post-analysis by approximately 2 × compared to single-error-bounded compression at the same compression ratio. Using the same error bound for the region of interest, our approach can achieve an increase of more than 50% in overall compression ratio.
Practical error analysis is essential for the design, optimization, and evaluation of Noisy Intermediate-Scale Quantum(NISQ) computing. However, bounding errors in quantum programs is a grand challenge, because the effects of quantum errors depend on exponentially large quantum states. In this work, we present Gleipnir, a novel methodology toward practically computing verified error bounds in quantum programs. Gleipnir introduces the (ρ,δ)-diamond norm, an error metric constrained by a quantum predicate consisting of the approximate state ρ and its distance δ to the ideal state ρ. This predicate (ρ,δ) can be computed adaptively using tensor networks based on the Matrix Product States. Gleipnir features a lightweight logic for reasoning about error bounds in noisy quantum programs, based on the (ρ,δ)-diamond norm metric. Furthermore, our experimental results show that Gleipnir is able to efficiently generate tight error bounds for real-world quantum programs with 10 to 100 qubits, and can be used to evaluate the error mitigation performance of quantum compiler transformations.
The Aharonov-Bohm effect is a fundamental issue in physics. It describes the physically important electromagnetic quantities in quantum mechanics. Its experimental verification constitutes a test of the theory of quantum mechanics itself. The remarkable experiments of Tonomura et al. ['Observation of Aharonov-Bohm effect by electron holography', Phys. Rev. Lett 48, 1443 (1982) and 'Evidence for Aharonov-Bohm effect with magnetic field completely shielded from electron wave', Phys. Rev. Lett 56, 792 (1986)] are widely considered as the only experimental evidence of the physical existence of the Aharonov-Bohm effect. Here we give the first rigorous proof that the classical ansatz of Aharonov and Bohm of 1959 ['Significance of electromagnetic potentials in the quantum theory', Phys. Rev. 115, 485 (1959)], that was tested by Tonomura et al., is a good approximation to the exact solution to the Schroedinger equation. This also proves that the electron, that is, represented by the exact solution, is not accelerated, in agreement with the recent experiment of Caprez et al. in 2007 ['Macroscopic test of the Aharonov-Bohm effect', Phys. Rev. Lett. 99, 210401 (2007)], that shows that the results of the Tonomura et al. experiments can not be explained by the action of a force. Under the assumption that the incoming free electron is a Gaussian wave packet, we estimate the exact solution to the Schroedinger equation for all times. We provide a rigorous, quantitative error bound for the difference in norm between the exact solution and the Aharonov-Bohm Ansatz. Our bound is uniform in time. We also prove that on the Gaussian asymptotic state the scattering operator is given by a constant phase shift, up to a quantitative error bound that we provide. Our results show that for intermediate size electron wave packets, smaller than the ones used in the Tonomura et al. experiments, quantum mechanics predicts the results observed by Tonomura et al. with an error bound smaller than 10{sup -99}. It would be quite interesting to perform experiments with electron wave packets of intermediate size. Furthermore, we provide a physical interpretation of our error bound.
In this report we compare the long-time error bounds and spatial resolution of finite difference methods with different spatial discretizations for the Dirac equation with small electromagnetic potentials characterized by $\varepsilon \in (0, 1]$ a dimensionless parameter. We begin with the simple and widely used finite difference time domain (FDTD) methods, and establish rigorous error bounds of them, which are valid up to the time at $O(1/\varepsilon)$. In the error estimates, we pay particular attention to how the errors depend explicitly on the mesh size $h$ and time step $\tau$ as well as the small parameter $\varepsilon$. Based on the results, in order to obtain "correct" numerical solutions up to the time at $O(1/\varepsilon)$, the $\varepsilon$-scalability (or meshing strategy requirement) of the FDTD methods should be taken as $h = O(\varepsilon^{1/2})$ and $\tau = O(\varepsilon^{1/2})$. To improve the spatial resolution capacity, we apply the Fourier spectral method to discretize the Dirac equation in space. Error bounds of the resulting finite difference Fourier pseudospectral (FDFP) methods show that they exhibit uniform spatial errors in the long-time regime, which are optimal in space as suggested by the Shannon's sampling theorem. Extensive numerical results are reported to confirm the error bounds and demonstrate that they are sharp.
Neural networks are increasingly integrated into scientific discovery, where input data reduction and model quantization play a key role in accelerating inference. However, understanding and mitigating the impact of these techniques on output error is critical for ensuring reliable results, particularly in tasks demanding high numerical precision. This paper introduces a comprehensive framework for optimizing neural network inference in scientific computing by combining data reduction and weight quantization while maintaining error-controlled outcomes. We develop theoretical analyses to bound error propagation under these reductions and propose a framework that balances computational performance with error constraints. Evaluation on real-world learning-based combustion simulations and satellite image classification demonstrates that our derived error bounds accurately predict observed errors while enabling significant computational speedup under our framework. This work highlights the potential for further leveraging advancements in modern lossy compression algorithms and hardware accelerators that support lower-precision formats.
This paper is the second part of a two-part paper that documents the numerical results for the partial derivatives model presented in part I. In this paper, we derive the error bounds for the analytical point-wise error expression and verify our bounds with numerical experiments. The point-wise error expressions make available, and bound, the sources that contribute to the total error of the interpolated cross section in terms of the Lagrange interpolation errors and the model form error. MPACT is used to generate two-group homogenized cross sections for Westinghouse's AP1000 Region 4 lattice to evaluate the accuracy of the bounds. Error bounds calculated over a grid are compared to numerical data for uni-variate and multi-variate interpolation. The point-wise error bounds of a typical case matrix - two branches in each state variable - are displayed for bi-variate interpolation in the state variables: moderator density, fuel temperature, and boron concentration. The error bounds are shown to be highly accurate compared to numerical results, and in accordance with the underlying physics. We then discuss and show how the sources of error contribute to the total error, and consider the improvement of each error source. Finally, we mention future work such as propagating our cross section error bounds through a reactivity calculation. (authors)
Reynolds lubrication approximation is used extensively to study flows between moving machine parts, in narrow channels, and in thin films. The solution of Reynolds equation may be thought of as the zeroth order term in an expansion of the solution of the Stokes equations in powers of the aspect ratio {var_epsilon} of the domain. In this paper, we show how to compute the terms in this expansion to arbitrary order on a two-dimensional, x-periodic domain and derive rigorous, a-priori error bounds for the difference between the exact solution and the truncated expansion solution. Unlike previous studies of this sort, the constants in our error bounds are either independent of the function h(x) describing the geometry, or depend on h and its derivatives in an explicit, intuitive way. Specifically, if the expansion is truncated at order 2k, the error is O({var_epsilon}{sup 2k+2}) and h enters into the error bound only through its first and third inverse moments {integral}{sub 0}{sup 1} h(x){sup -m} dx, m = 1,3 and via the max norms {parallel} 1/{ell}! h{sup {ell}-1}{partial_derivative}{sub x}{sup {ell}}h{parallel}{sub {infinity}}, 1 {le} {ell} {le} 2k + 2. We validate our estimates by comparing with finite element solutions and present numerical evidence that suggests that even when h is real analytic and periodic, the expansion solution forms an asymptotic series rather than a convergent series.
A range of core operations and planning problems for the national electrical grid are naturally formulated and solved as stochastic programming problems, which minimize expected costs subject to a range of uncertain outcomes relating to, for example, uncertain demands or generator output. A critical decision issue relating to such stochastic programs is: How many scenarios are required to ensure a specific error bound on the solution cost? Scenarios are the key mechanism used to sample from the uncertainty space, and the number of scenarios drives computational difficultly. We explore this question in the context of a long-term grid generation expansion problem, using a bounding procedure introduced by Mak, Morton, and Wood. We discuss experimental results using problem formulations independently minimizing expected cost and down-side risk. Our results indicate that we can use a surprisingly small number of scenarios to yield tight error bounds in the case of expected cost minimization, which has key practical implications. In contrast, error bounds in the case of risk minimization are significantly larger, suggesting more research is required in this area in order to achieve rigorous solutions for decision makers.
The objective of this work is to develop error-bounded lossy compression methods to preserve topological features in 2D and 3D vector fields. Specifically, we explore the preservation of critical points in piecewise linear vector fields. We define the preservation of critical points as, without any false positive, false negative, or false type change in the decompressed data, (1) keeping each critical point in its original cell and (2) retaining the type of each critical point (e.g., saddle and attracting node). The key to our method is to adapt a vertex-wise error bound for each grid point and to compress input data together with the error bound field using a modified lossy compressor. Our compression algorithm can be also embarrassingly parallelized for large data handling and in situ processing. We benchmark our method by comparing it with existing lossy compressors in terms of false positive/negative/type rates, compression ratio, and various vector field visualizations with several scientific applications.
Extreme-scale simulations and high-resolution instruments have been generating an increasing amount of data, which poses significant challenges to not only data storage during the run, but also post-processing where data will be repeatedly retrieved and analyzed for a long period of time. The challenges in satisfying a wide range of post-hoc analysis needs while minimizing the I/O overhead caused by inappropriate and/or excessive data retrieval should never be left unmanaged. In this paper, we propose a data refactoring, compressing, and retrieval framework capable of 1) fine-grained data refactoring with regard to precision; 2) incrementally retrieving and recomposing the data in terms of various error bounds; and 3) adaptively retrieving data in multi-precision and multi-resolution with respect to different analysis. With the progressive data re-composition and the adaptable retrieval algorithms, our framework significantly reduces the amount of data retrieved when multiple incremental precision are requested and/or the downstream analysis time when coarse resolution is used. Experiments show that the amount of data retrieved under the same progressively requested error bound using our framework is 64% less than that using state-of-the-art single-error-bounded approaches. Parallel experiments with up to 1, 024 cores and ~ 600 GB data in total show that our approach yields 1.36× and 2.52× performance over existing approaches in writing to and reading from persistent storage systems, respectively.
This paper develops an interpretable, non-intrusive reduced-order modeling technique using regularized kernel interpolation. Existing non-intrusive approaches approximate the dynamics of a reduced-order model (ROM) by solving a data-driven least-squares regression problem for low-dimensional matrix operators. Our approach instead leverages regularized kernel interpolation, which yields an optimal approximation of the ROM dynamics from a user-defined reproducing kernel Hilbert space. We show that our kernel-based approach can produce interpretable ROMs whose structure mirrors full-order model structure by embedding judiciously chosen feature maps into the kernel. The approach is flexible and allows a combination of informed structure through feature maps and closure terms via more general nonlinear terms in the kernel. We also derive a computable a posteriori error bound that combines standard error estimates for intrusive projection-based ROMs and kernel interpolants. In conclusion, the approach is demonstrated in several numerical experiments that include comparisons to operator inference using both proper orthogonal decomposition and quadratic manifold dimension reduction.
An a posteriori error-estimation framework is introduced to quantify and reduce modeling errors resulting from approximating complex mesoscale material behavior with a simpler macroscale model. Such errors may be prevalent when modeling welds and additively manufactured structures, where spatial variations and material textures may be present in the microstructure. We consider a case where a <100> fiber texture develops in the longitudinal scanning direction of a weld. Transversely isotropic elastic properties are obtained through homogenization of a microstructural model with this texture and are considered the reference weld properties within the error-estimation framework. Conversely, isotropic elastic properties are considered approximate weld properties since they contain no representation of texture. Errors introduced by using isotropic material properties to represent a weld are assessed through a quantified error bound in the elastic regime. Lastly, an adaptive error reduction scheme is used to determine the optimal spatial variation of the isotropic weld properties to reduce the error bound.