arXiv subjects

Bryan M. Wong

Publications and source records attributed to Bryan M. Wong.

At least 19 recordsLinked to original sources

arXiv2024

FT K-means: A High-Performance K-means on GPU with Fault Tolerance

K-means is a widely used algorithm in clustering, however, its efficiency is primarily constrained by the computational cost of distance computing. Existing implementations suffer from suboptimal utilization of computational units and lack resilience against soft errors. To address these challenges, we introduce FT K-means, a high-performance GPU-accelerated implementation of K-means with online fault tolerance. We first present a stepwise optimization strategy that achieves competitive performance compared to NVIDIA's cuML library. We further improve FT K-means with a template-based code generation framework that supports different data types and adapts to different input shapes. A novel warp-level tensor-core error correction scheme is proposed to address the failure of existing fault tolerance methods due to memory asynchronization during copy operations. Our experimental evaluations on NVIDIA T4 GPU and A100 GPU demonstrate that FT K-means without fault tolerance outperforms cuML's K-means implementation, showing a performance increase of 10\%-300\% in scenarios involving irregular data shapes. Moreover, the fault tolerance feature of FT K-means introduces only an overhead of 11\%, maintaining robust performance even with tens of errors injected per second.

cs.DC↗

arXiv2024

Velocity-gauge real-time time-dependent density functional tight-binding for large-scale condensed matter systems

We present a new velocity-gauge real-time, time-dependent density functional tight-binding (VG-rtTDDFTB) implementation in the open-source DFTB+ software package (https://dftbplus.org) for probing electronic excitations in large, condensed matter systems. Our VG-rtTDDFTB approach enables real-time electron dynamics simulations of large, periodic, condensed matter systems containing thousands of atoms with a favorable computational scaling as a function of system size. We provide computational details and benchmark calculations to demonstrate its accuracy and computational parallelizability on a variety of large material systems. As a representative example, we calculate laser-induced electron dynamics in a 512-atom amorphous silicon supercell to highlight the large periodic systems that can be examined with our implementation. Taken together, our VG-rtTDDFTB approach enables new electron dynamics simulations of complex systems that require large periodic supercells, such as crystal defects, complex surfaces, nanowires, and amorphous materials.

cond-mat.mes-hall↗

arXiv2024

Neutralization of low energy Na$^{+}$ scattered from InAs(001)

The neutralization probability of low energy Na$^{+}$ ions scattered from In- and As-rich InAs(001) surfaces is measured by time-of-flight spectroscopy. It is found that the neutralization probability for projectiles scattered from As sites is larger than from In sites for both types of surfaces. A modification of the resonant charge transfer model is proposed in which a freezing contour that follows the atomic structure is combined with molecular dynamics and density functional theory. Together, these approaches show that the neutralization of alkali projectiles scattered from a compound solid material is determined by multiple factors, particularly the surface atomic and electronic structures. This model is applicable to any system in which the surface potential is inhomogeneous, such as compound materials and adsorbate-covered surfaces.

cond-mat.mtrl-sci↗

arXiv2024

Cost-Effective Methodology for Complex Tuning Searches in HPC: Navigating Interdependencies and Dimensionality

Tuning searches are pivotal in High-Performance Computing (HPC), addressing complex optimization challenges in computational applications. The complexity arises not only from finely tuning parameters within routines but also potential interdependencies among them, rendering traditional optimization methods inefficient. Instead of scrutinizing interdependencies among parameters and routines, practitioners often face the dilemma of conducting independent tuning searches for each routine, thereby overlooking interdependence, or pursuing a more resource-intensive joint search for all routines. This decision is driven by the consideration that some interdependence analysis and high-dimensional decomposition techniques in literature may be prohibitively expensive in HPC tuning searches. Our methodology adapts and refines these methods to ensure computational feasibility while maximizing performance gains in real-world scenarios. Our methodology leverages a cost-effective interdependence analysis to decide whether to merge several tuning searches into a joint search or conduct orthogonal searches. Tested on synthetic functions with varying levels of parameter interdependence, our methodology efficiently explores the search space. In comparison to Bayesian-optimization-based full independent or fully joint searches, our methodology suggested an optimized breakdown of independent and merged searches that led to final configurations up to 8% more accurate, reducing the search time by up to 95%. When applied to GPU-offloaded Real-Time Time-Dependent Density Functional Theory (RT-TDDFT), an application in computational materials science that challenges modern HPC autotuners, our methodology achieved an effective tuning search. Its adaptability and efficiency extend beyond RT-TDDFT, making it valuable for related applications in HPC.

cs.DC↗

arXiv2023

Accelerating Quantum Optimal Control of Multi-Qubit Systems with Symmetry-Based Hamiltonian Transformations

We present a novel, computationally efficient approach to accelerate quantum optimal control calculations of large multi-qubit systems used in a variety of quantum computing applications. By leveraging the intrinsic symmetry of finite groups, the Hilbert space can be decomposed and the Hamiltonians block-diagonalized to enable extremely fast quantum optimal control calculations. Our approach reduces the Hamiltonian size of an $n$-qubit system from 2^n by 2^n to O(n by n) or O((2^n / n) by (2^n / n)) under Sn or Dn symmetry, respectively. Most importantly, this approach reduces the computational runtime of qubit optimal control calculations by orders of magnitude while maintaining the same accuracy as the conventional method. As prospective applications, we show that (1) symmetry-protected subspaces can be potential platforms for quantum error suppression and simulation of other quantum Hamiltonians, and (2) Lie-Trotter-Suzuki decomposition approaches can generalize our method to a general variety of multi-qubit systems.

quant-ph↗

arXiv2023

Anatomy of High-Performance GEMM with Online Fault Tolerance on GPUs

General Matrix Multiplication (GEMM) is a crucial algorithm for various applications such as machine learning and scientific computing, and an efficient GEMM implementation is essential for the performance of these systems. While researchers often strive for faster performance by using large compute platforms, the increased scale of these systems can raise concerns about hardware and software reliability. In this paper, we present a design for a high-performance GEMM with algorithm-based fault tolerance for use on GPUs. We describe fault-tolerant designs for GEMM at the thread, warp, and threadblock levels, and also provide a baseline GEMM implementation that is competitive with or faster than the state-of-the-art, proprietary cuBLAS GEMM. We present a kernel fusion strategy to overlap and mitigate the memory latency due to fault tolerance with the original GEMM computation. To support a wide range of input matrix shapes and reduce development costs, we present a template-based approach for automatic code generation for both fault-tolerant and non-fault-tolerant GEMM implementations. We evaluate our work on NVIDIA Tesla T4 and A100 server GPUs. Experimental results demonstrate that our baseline GEMM presents comparable or superior performance compared to the closed-source cuBLAS. The fault-tolerant GEMM incurs only a minimal overhead (8.89\% on average) compared to cuBLAS even with hundreds of errors injected per minute. For irregularly shaped inputs, the code generator-generated kernels show remarkable speedups of $160\% \sim 183.5\%$ and $148.55\% \sim 165.12\%$ for fault-tolerant and non-fault-tolerant GEMMs, outperforming cuBLAS by up to $41.40\%$.

cs.DC↗

arXiv2022

Improved Band Gaps and Structural Properties from Wannier-Fermi-Löwdin Self-Interaction Corrections for Periodic Systems

The accurate prediction of band gaps and structural properties in periodic systems continues to be one of the central goals of electronic structure theory. However, band gaps obtained from popular exchange-correlation functionals (such as LDA and PBE) are severely underestimated partly due to the spurious self-interaction error (SIE) inherent to these functionals. In this work, we present a new formulation and implementation of Wannier function-derived Fermi-Löwdin (WFL) orbitals for correcting the SIE in periodic systems. Since our approach utilizes a variational minimization of the self-interaction energy with respect to the Wannier charge centers, it is computationally more efficient than the HSE hybrid functional and other self-interaction corrections that require a large number of transformation matrix elements. Calculations on several (17 in total) prototypical molecular solids, semiconductors, and wide-bandgap materials show that our WFL self-interaction correction approach gives better band gaps and bulk moduli compared to semilocal functionals, largely due to the partial removal of self-interaction errors.

cond-mat.mtrl-sci↗

arXiv2022

High-Temperature Decomposition of Diisopropyl Methylphosphonate (DIMP) on Alumina: Mechanistic Predictions from Ab Initio Molecular Dynamics

The enhanced degradation of organophosphorous-based chemical warfare agents (CWAs) on metal-oxide surfaces holds immense promise for neutralization efforts; however, the underlying mechanisms in this process remain poorly understood. We utilize large-scale quantum calculations for the first time to probe the high-temperature degradation of diisopropyl methylphosphonate (DIMP), a nerve agent simulant. Our Born-Oppenheimer molecular dynamics (BOMD) calculations show that the $γ$-Al$_2$O$_3$ surface shows immense promise for quickly adsorbing and destroying CWAs. We find that the alumina surface quickly adsorbs DIMP at all temperatures, and subsequent decomposition of DIMP proceeds via a propene elimination. Our BOMD calculations are complemented with metadynamics simulations to produce free energy paths, which show that the activation barrier decreases with temperature and DIMP readily decomposes on $γ$-Al$_2$O$_3$. Our first-principle BOMD and metadynamics simulations provide crucial diagnostics for sarin decomposition models and mechanistic information for examining CWA decomposition reactions on other candidate metal oxide surfaces.

physics.chem-ph↗

arXiv2022

HADOKEN: An Open-Source Software Package for Predicting Electron Confinement Effects in Various Nanowire Geometries and Configurations

We present an open-source software package, HADOKEN (High-level Algorithms to Design, Optimize, and Keep Electrons in Nanowires), for predicting electron confinement/localization effects in nanowires with various geometries, {arbitrary number of concentric shell layers,} doping densities, and external boundary conditions. The HADOKEN code is written in the MATLAB programming environments to aid in its readability and general accessibility to both users and practitioners. We provide several examples and outputs on a variety of different nanowire geometries, boundary conditions, and doping densities to demonstrate the capabilities of the HADOKEN software package. As such, the use of this predictive and versatile tool by both experimentalists and theorists could lead to further advances in both understanding and tailoring electron confinement effects in these nanosystems.

cond-mat.mes-hall↗

arXiv2021

6 nm super-resolution optical transmission and scattering spectroscopic imaging of carbon nanotubes using a nanometer-scale white light source

Optical hyperspectral imaging based on absorption and scattering of photons at the visible and adjacent frequencies denotes one of the most informative and inclusive characterization methods in material research. Unfortunately, restricted by the diffraction limit of light, it is unable to resolve the nanoscale inhomogeneity in light-matter interactions, which is diagnostic of the local modulation in material structure and properties. Moreover, many nanomaterials have highly anisotropic optical properties that are outstandingly appealing yet hard to characterize through conventional optical methods. Therefore, there has been a pressing demand in the diverse fields including electronics, photonics, physics, and materials science to extend the optical hyperspectral imaging into the nanometer length scale. In this work, we report a super-resolution hyperspectral imaging technique that simultaneously measures optical absorption and scattering spectra with the illumination from a tungsten-halogen lamp. We demonstrated sub-5 nm spatial resolution in both visible and near-infrared wavelengths (415 to 980 nm) for the hyperspectral imaging of strained single-walled carbon nanotubes (SWNT) and reconstructed true-color images to reveal the longitudinal and transverse optical transition-induced light absorption and scattering in the SWNTs. This is the first time transverse optical absorption in SWNTs were clearly observed experimentally. The new technique provides rich near-field spectroscopic information that had made it possible to analyze the spatial modulation of band-structure along a single SWNT induced through strain engineering.

physics.optics↗

arXiv2020

Ultrafast Photoinduced Band Splitting and Carrier Dynamics in Chiral Tellurium Nanosheets

Trigonal tellurium (Te) is a chiral semiconductor that lacks both mirror and inversion symmetries, resulting in complex band structures with Weyl crossings and unique spin textures. Detailed time-resolved polarized reflectance spectroscopy is used to investigate its band structure and carrier dynamics. The polarized transient spectra reveal optical transitions between the uppermost spin-split H4 and H5 and the degenerate H6 valence bands (VB) and the lowest degenerate H6 conduction band (CB) as well as a higher energy transition at the L-point. Surprisingly, the degeneracy of the H6 CB (a proposed Weyl node) is lifted and the spin-split VB gap is reduced upon photoexcitation before relaxing to equilibrium as the carriers decay. Using ab initio density functional theory (DFT) calculations we conclude that the dynamic band structure is caused by a photoinduced shear strain in the Te film that breaks the screw symmetry of the crystal. The band-edge anisotropy is also reflected in the hot carrier decay rate, which is a factor of two slower along c-axis than perpendicular to it. The majority of photoexcited carriers near the band-edge are seen to recombine within 30 ps while higher lying transitions observed near 1.2 eV appear to have substantially longer lifetimes, potentially due to contributions of intervalley processes in the recombination rate. These new findings shed light on the strong correlation between photoinduced carriers and electronic structure in anisotropic crystals, which opens a potential pathway for designing novel Te-based devices that take advantage of the topological structures as well as strong spin-related properties.

cond-mat.mtrl-sci↗

arXiv2019

Charge Density Wave Hampers Exciton Condensation in 1T-TiSe$_2$

The Bose-Einstein condensation of excitons continues to garner immense attention as a prototypical example for observing emergent properties from many-body quantum effects. In particular, Titanium Diselenide (TiSe$_2$) is a promising candidate for realizing exciton condensation and was experimentally observed only very recently. Surprisingly, the condensate was experimentally characterized by a soft plasmon mode that only exists near the transition temperature, $T_c$, of the charge density wave (CDW). Here, we characterize and analyze the experimental spectra using linear-response time-dependent density functional theory and find that the soft mode can be attributed to interband electronic transitions. At the CDW state below $T_c$, the periodic lattice distortions hamper the spontaneous formation of the exciton by introducing a CDW gap. The band gap raises the soft mode and merges it into the regular plasmon. Our surprising results contradict previous simplistic analytical models commonly used in the scientific literature. In addition, we find that a finite electronic temperature, $T_e$, introduces a dissipation channel and prevents the condensation above $T_c$. The combined effect of the CDW and $T_e$ explains the fragile temperature-dependence of the exciton condensation. Taken together, our work provides the first \textit{ab initio} atomic-level framework for rationalizing recent experiments and further manipulating exciton condensates in TiSe$_2$.

cond-mat.mtrl-sci↗

arXiv2019

Indirect but Efficient: Laser-Excited Electrons Can Drive Ultrafast Polarization Switching in Ferroelectric Materials

To enhance the efficiency of next-generation ferroelectric (FE) electronic devices, new techniques for controlling ferroelectric polarization switching are required. While most prior studies have attempted to induce polarization switching via the excitation of phonons, these experimental techniques required intricate and expensive terahertz sources and have not been completely successful. Here, we propose a new mechanism for rapidly and efficiently switching the FE polarization via laser-tuning of the underlying dynamical potential energy surface. Using time-dependent density functional calculations, we observe an ultrafast switching of the FE polarization in BaTiO3 within 200 fs. A laser pulse can induce a charge density redistribution that reduces the original FE charge order. This excitation results in both desirable and highly directional ionic forces that are always opposite to the original FE displacements. Our new mechanism enables the reversible switching of the FE polarization with optical pulses that can be produced from existing 800 nm experimental laser sources.

cond-mat.mtrl-sci↗

arXiv2019

Accurate Electron Affinities and Orbital Energies of Anions from a Non-Empirically Tuned Range-Separated Density Functional Theory Approach

The treatment of atomic anions with Kohn-Sham density functional theory (DFT) has long been controversial since the highest occupied molecular orbital (HOMO) energy, $E_{HOMO}$, is often calculated to be positive with most approximate density functionals. We assess the accuracy of orbital energies and electron affinities for all three rows of elements in the periodic table (H-Ar) using a variety of theoretical approaches and customized basis sets. Among all of the theoretical methods studied here, we find that a non-empirically tuned range-separated approach (constructed to satisfy DFT-Koopmans' theorem for the anionic electron system) provides the best accuracy for a variety of basis sets - even for small basis sets where most functionals typically fail. Previous approaches to solve this conundrum of positive $E_{HOMO}$ values have utilized non-self-consistent methods; however electronic properties, such as electronic couplings/gradients (which require a self-consistent potential and energy), become ill-defined with these approaches. In contrast, the non-empirically tuned range-separated procedure used here yields well-defined electronic couplings/gradients and correct $E_{HOMO}$ values since both the potential and resulting electronic energy are computed self-consistently. Orbital energies and electron affinities are further analyzed in the context of the electronic energy as a function of electronic number (including fractional numbers of electrons) to provide a stringent assessment of self-interaction errors for these complex anion systems.

physics.chem-ph↗

arXiv2018

The Diamine Cation Is Not a Chemical Example Where Density Functional Theory Fails

In a recent communication, Weber and co-workers presented a surprising study on charge-localization effects in the N,N'-dimethylpiperazine (DMP+) diamine cation to provide a stringent test of density functional theory (DFT) methods. Within their study, the authors examined various DFT methods and concluded that "all DFT functionals commonly used today, including hybrid functionals with exact exchange, fail to predict a stable charge-localized state." This surprising conclusion is based on the authors' use of a self-interaction correction (namely, complex-valued Perdew-Zunger Self-Interaction Correction (PZ-SIC)) to DFT, which appears to give excellent agreement with experiment and other wavefunction-based benchmarks. Since the publication of this recent communication, the same DMP+ molecule has been cited in numerous subsequent studies as a prototypical example of the importance of self-interaction corrections for accurately calculating other chemical systems. In this correspondence, we have carried out new high-level CCSD(T) analyses on the DMP+ cation to show that DFT actually performs quite well for this system (in contrast to their conclusion that all DFT functionals fail), whereas the PZ-SIC approach used by Weber et al. is the outlier that is inconsistent with the high-level CCSD(T) (coupled-cluster with single and double excitations and perturbative triples) calculations. Our new findings and analysis for this system are briefly discussed in this correspondence.

physics.chem-ph↗

arXiv2018

The Importance of Short- and Long-Range Exchange on Various Excited State Properties of DNA Monomers, Stacked Complexes, and Watson-Crick Pairs

We present a detailed analysis of several time-dependent DFT (TD-DFT) methods, including conventional hybrid functionals and two types of non-empirically tuned range-separated functionals, for predicting a diverse set of electronic excitations in DNA nucleobase monomers and dimers. This large and extensive set of excitations comprises a total of 50 different transitions (for each tested DFT functional) that includes several n $\rightarrow$ π and π $\rightarrow$ π* valence excitations, long-range charge-transfer excitations, and extended Rydberg transitions (complete with benchmark calculations from high-level EOM-CCSD(T) methods). The presence of localized valence excitations as well as extreme long-range charge-transfer excitations in these systems poses a serious challenge for TD-DFT methods that allows us to assess the importance of both short- and long-range exchange contributions for simultaneously predicting all of these various transitions. In particular, we find that functionals that do not have both short- and full long-range exchange components are unable to predict the different types of nucleobase excitations with the same accuracy. Most importantly, the current study highlights the importance of both short-range exchange and a non-empirically tuned contribution of long-range exchange for accurately predicting the diverse excitations in these challenging nucleobase systems.

physics.chem-ph↗

arXiv2018

Unusual Bandgap Oscillations in Template-Directed π-Conjugated Porphyrin Nanotubes

Using large-scale DFT calculations (up to 1,476 atoms and 18,432 orbitals), we present the first detailed analysis on the unusual electronic properties of recently synthesized porphyrin nanotubes. We surprisingly observe extremely large oscillations in the bandgap of these nanostructures as a function of size, in contradiction to typical quantum confinement effects (i.e., the bandgap increases with size in several of these nanotubes). In particular, we find that these intriguing electronic oscillations arise from a size-dependent alternation of aromatic/non-aromatic characteristics in these porphyrin nanotubes. Our analyses of band structures and orbital diagrams indicate that the electronic transitions in these nanostructures are direct-bandgap, optically active "bright" states that can be readily observed in photoelectron spectroscopic experiments. Most importantly due to their unusual bandgap oscillations, we find that both type I and type II donor-acceptor p-n heterojunctions are possible in these template-directed, "bottom-up synthesized" porphyrin nanotubes - a unique property that is not present in conventional carbon nanotubes.

cond-mat.mes-hall↗

arXiv2018

Inconsistencies in the Electronic Properties of Phosphorene Nanotubes: New Insights from Large-Scale DFT Calculations

Contrary to recent reports, we show that the electronic properties of phosphorene nanotubes are surprisingly rich and much more complex than previously assumed. We find that all phosphorene nanotubes exhibit an intricate direct-to-indirect bandgap transition as the nanotube diameter decreases - a unique property not identified in any prior studies (which claimed either direct or indirect band gaps only) that we uncover with large-scale DFT calculations. We address these previous inconsistencies by detailed analyses of orbital interactions, which reveal that the strain associated with decreasing the nanotube diameter causes a transition from a direct to an indirect band gap for all of the phosphorene nanotubes. We show that our findings are completely general, and extensive calculations across several exchange-correlation functionals verify our conclusions. Most importantly, our results and analyses resolve a long-standing question on the electronic properties of phosphorus nanotubes and brings closure to previously conflicting findings in these unique nanostructures.

cond-mat.mes-hall↗