Search NASA⌕ Search

SEARCH · Search NASA

Results for “Basis function”

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.

At least 1,477 records · Page 82
All sourcesNASA NTRSDOE OSTI
DOE OSTI2016

LONG TERM OPERATION ISSUES FOR ELECTRICAL CABLE SYSTEMS IN NUCLEAR POWER PLANTS

Nuclear power plants contain hundreds of kilometers of electrical cables including cables used for power, for instrumentation, and for control. It is essential that safety-related cable systems continue to perform following a design-basis event. Wholesale replacement of electrical cables in existing plants facing licensing period renewal is both impractical and cost-prohibitive. It is therefore important to understand the long term aging of cable materials to have confidence that aged cables will perform when needed. It is equally important in support of cable aging management to develop methods to evaluate the health of installed cables and inform selective cable replacement decisions. The most common insulation materials for electrical cables in nuclear power plants are cross-linked polyethylene and ethylene-propylene rubber. The mechanical properties of these materials degrade over time in the presence of environmental stresses including heat, gamma irradiation, and moisture. Mechanical degradation of cable insulation beyond a certain threshold is unacceptable because it can lead to insulation cracking, exposure of energized conductors, arcing and burning or loss of the ability of the cable system to function during a design-basis accident. While thermal-, radiation-, and moisture-related degradation of polymer insulation materials has been extensively studied over the last few decades, questions remain regarding the long-term performance of cable materials in nuclear plant-specific environments. Identified knowledge gaps include an understanding of the temperature-dependence of activation energies for thermal damage and an understanding of the synergistic effects of radiation and thermal stress on polymer degradation. Many of the outstanding questions in the aging behavior of cable materials relate to the necessity of predicting long-term field degradation using accelerated aging results from the laboratory. Materials degrade faster under more extreme conditions, but extension of behavior to long term degradation under more mild conditions, such as those experienced by most installed cables in nuclear power plants, is complicated by the fact that different degradation mechanisms may be involved in extreme and mild scenarios. The discrepancy in predicted results from short term, more extreme exposure and actual results from longer term, more mild exposures can be counter intuitive. For instance, due to the attenuation of oxidation penetration in material samples rapidly aged through exposure to high temperatures, the bulk of the samples may be artificially protected from thermal aging. In another example, simultaneous exposure of cable insulation material to heat and radiation may actually lead to less damage at higher temperatures than may be observed at lower temperatures. The Light Water Reactor Sustainability program of the United States (US) Department of Energy Office (DOE) of Nuclear Energy is funding research to increase the predictive understanding of electrical cable material aging and degradation in existing nuclear power plants in support of continued safe operation of plants beyond their initial license periods. This research includes the evaluation and development of methods to assess installed cable condition.

Fifield, Dr Leonard S↗
NASA NTRS1988

Identification of visual evoked response parameters sensitive to pilot mental state

Systems analysis techniques were developed and demonstrated for modeling the electroencephalographic (EEG) steady state visual evoked response (ssVER), for use in EEG data compression and as an indicator of mental workload. The study focused on steady state frequency domain stimulation and response analysis, implemented with a sum-of-sines (SOS) stimulus generator and an off-line describing function response analyzer. Three major tasks were conducted: (1) VER related systems identification material was reviewed; (2) Software for experiment control and data analysis was developed and implemented; and (3) ssVER identification and modeling was demonstrated, via a mental loading experiment. It was found that a systems approach to ssVER functional modeling can serve as the basis for eventual development of a mental workload indicator. The review showed how transient visual evoked response (tVER) and ssVER research are related at the functional level, the software development showed how systems techniques can be used for ssVER characterization, and the pilot experiment showed how a simple model can be used to capture the basic dynamic response of the ssVER, under varying loads.

Zacharias, G. L.↗
DOE OSTI2025

Free-energy orbital-free density functional theory: recent developments, perspective, and outlook

By summarizing the constraint-based development of orbital-free free-energy density functional approximations, we provide a perspective on progress over the last 15 years, the limitations of existing functionals, and the challenges awaiting resolution. We outline the chronology of the development of noninteracting and exchange-correlation free-energy orbital-free functionals and summarize the theoretical basis of existing local density approximation, second-order approximation, generalized gradient approximation (GGA), and meta-GGAs. We discuss limitations and challenges such as problems with thermodynamic derivatives, free-energy nonadditivity and the closely related issue of all-electron versus valence-only local pseudo-potential performance.

74 ATOMIC AND MOLECULAR PHYSICS↗
DOE OSTI2020

Polarizabilities of neutral atoms and atomic ions with a noble gas electron configuration

Atomic polarizabilities play an important role in the development of force fields for molecular simulations, as well as for the development of qualitative concepts of atomic and molecular behavior. Coupled cluster theory at the coupled cluster singles doubles triples level with very large correlation-consistent basis sets with extended diffuse functions has been used to predict the polarizabilities of the atomic neutrals, mono-cations and mono-anions with a noble gas configuration. Additional corrections for scalar relativistic and spin–orbit effects were also included for the electron configurations of Kr, Xe, and Rn. The results are in excellent agreement with experiment or with other high level calculations where available. The current results for most of these species represent the best available values for the polarizabilities. The results show that the polarizability of H - is very difficult to calculate without extremely diffuse functions. The polarizability of H - is the largest value, 34.05 Å 3 , calculated for all species in the current study. The polarizabilities of the remaining halogen anions are also the best available values. The polarizabilities of the halogen anions (excluding F - ) and H - have a linear correlation with the electron affinity of the neutral atom. Spin– orbit effects, even for closed shell species, cannot be ignored for quantitative accuracy, and the inclusion of spin–orbit effects for Fr + , Rn, and At - increases the polarizability by 4%, 6%, and 15%, respectively.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2023

Accurate Prediction of Adiabatic Ionization Potentials of Organic Molecules using Quantum Chemistry Assisted Machine Learning

In previous work (Dandu et al., J. Phys. Chem. A, 2022, 126, 4528–4536), we were successful in predicting accurate atomization energies of organic molecules using machine learning (ML) models, obtaining an accuracy as low as 0.1 kcal/mol compared to the G4MP2 method. In this work, we extend the use of these ML models to adiabatic ionization potentials on data sets of energies generated using quantum chemical calculations. Atomic specific corrections that were found to improve atomization energies from quantum chemical calculations have also been used in this study to improve ionization potentials. Here, the quantum chemical calculations were performed on 3405 molecules containing eight or fewer non-hydrogen atoms derived from the QM9 data set, using the B3LYP functional with the 6–31G(2df,p) basis set for optimization. Low-fidelity IPs for these structures were obtained using two density functional methods: B3LYP/6–31+G(2df,p) and ωB97XD/6–311+G(3df,2p). Highly accurate G4MP2 calculations were performed on these optimized structures to obtain high-fidelity IPs to use in ML models based on the low-fidelity IPs. Our best performing ML methods gave IPs of organic molecules within a mean absolute deviation of 0.035 eV from the G4MP2 IPs for the whole data set. This work demonstrates that ML predictions assisted by quantum chemical calculations can be used to successfully predict IPs of organic molecules for use in high throughput screening.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
NASA NTRS1987

An algorithm to design finite field multipliers using a self-dual normal basis

Finite field multiplication is central in the implementation of some error-correcting coders. Massey and Omura have presented a revolutionary design for multiplication in a finite field. In their design, a normal base is utilized to represent the elements of the field. The concept of using a self-dual normal basis to design the Massey-Omura finite field multiplier is presented. Presented first is an algorithm to locate a self-dual normal basis for GF(2 sup m) for odd m. Then a method to construct the product function for designing the Massey-Omura multiplier is developed. It is shown that the construction of the product function base on a self-dual basis is simpler than that based on an arbitrary normal base.

Wang, C. C.↗
DOE OSTI2011

A Greenhouse-Gas Information System: Monitoring and Validating Emissions Reporting and Mitigation

Current GHG-mitigating regimes, whether internationally agreed or self-imposed, rely on the aggregation of self-reported data, with limited checks for consistency and accuracy, for monitoring. As nations commit to more stringent GHG emissions-mitigation actions and as economic rewards or penalties are attached to emission levels, self-reported data will require independent confirmation that they are accurate and reliable, if they are to provide the basis for critical choices and actions that may be required. Supporting emissions-mitigation efforts and agreements, as well as monitoring energy- and fossil-fuel intensive national and global activities would be best achieved by a process of: (1) monitoring of emissions and emission-mitigation actions, based, in part, on, (2) (self-) reporting of pertinent bottom-up inventory data, (3) verification that reported data derive from and are consistent with agreed-upon processes and procedures, and (4) validation that reported emissions and emissions-mitigation action data are correct, based on independent measurements (top-down) derived from a suite of sensors in space, air, land, and, possibly, sea, used to deduce and attribute anthropogenic emissions. These data would be assessed and used to deduce and attribute measured GHG concentrations to anthropogenic emissions, attributed geographically and, to the extent possible, by economic sector. The validation element is needed to provide independent assurance that emissions are in accord with reported values, and should be considered as an important addition to the accepted MRV process, leading to a MRV&V process. This study and report focus on attributes of a greenhouse-gas information system (GHGIS) needed to support MRV&V needs. These needs set the function of such a system apart from scientific/research monitoring of GHGs and carbon-cycle systems, and include (not exclusively): the need for a GHGIS that is operational, as required for decision-support; the need for a system that meets specifications derived from imposed requirements; the need for rigorous calibration, verification, and validation (CV&V) standards, processes, and records for all measurement and modeling/data-inversion data; the need to develop and adopt an uncertainty-quantification (UQ) regimen for all measurement and modeling data; and the requirement that GHGIS products can be subjected to third-party questioning and scientific scrutiny. This report examines and assesses presently available capabilities that could contribute to a future GHGIS. These capabilities include sensors and measurement technologies; data analysis and data uncertainty quantification (UQ) practices and methods; and model-based data-inversion practices, methods, and their associated UQ. The report further examines the need for traceable calibration, verification, and validation processes and attached metadata; differences between present science-/research-oriented needs and those that would be required for an operational GHGIS; the development, operation, and maintenance of a GHGIS missions-operations center (GMOC); and the complex systems engineering and integration that would be required to develop, operate, and evolve a future GHGIS. Present monitoring systems would be heavily relied on in any GHGIS implementation at the outset and would likely continue to provide valuable future contributions to GHGIS. However, present monitoring systems were developed to serve science/research purposes. This study concludes that no component or capability presently available is at the level of technological maturity and readiness required for implementation in an operational GHGIS today. However, purpose-designed and -built components could be developed and implemented in support of a future GHGIS. The study concludes that it is possible to develop and provide a capability-driven prototype GHGIS, as part of a Phase-1 effort, within three years from project-funding start, that would make use of and integrate existing sensing and system capabilities. As part of a Phase-2 effort, a requirements-driven, operational GHGIS could be developed, within ten years from project funding start. That schedule is driven by the development and long lead-times for some system components. The two efforts would be focused on different deliverables but could commence concurrently, to save time, if that was deemed desirable. We note that, developing and supporting an operational GHGIS will require a new approach and management, sustained funding and other support, as well as technical advances and development of purpose-built components that meet the requisite specifications. A functioning GHGIS will provide the basis for reasoned choices on how best to respond to rising GHG levels, especially when proposed U.S. actions are compared with or conditioned on the actions of other nations.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2014

Functional form for plasma velocity in a rapidly rotating tokamak discharge

A recently developed technique using charge exchange spectroscopy determines the ion poloidal rotation in tokamak plasmas from the poloidal variation in the toroidal angular rotation speed. The basis for this technique is the functional form for the plasma velocity calculated from the equilibrium equations. The initial development of this technique utilized the functional form determined for conditions where the ion toroidal rotation speed is much smaller than the ion thermal speed. There are cases, however, where the toroidal rotation can be comparable to the ion thermal speed, especially for high atomic number impurities. The present paper extends the previous analysis to this high rotation speed case and demonstrates how to extract the poloidal rotation speed from measurements of the toroidal angular rotation speed at two points on a flux surface.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
DOE OSTI2024

Structure-guided functional suppression of AML-associated DNMT3A hotspot mutations

DNA methyltransferases DNMT3A- and DNMT3B-mediated DNA methylation critically regulate epigenomic and transcriptomic patterning during development. The hotspot DNMT3A mutations at the site of Arg822 (R882) promote polymerization, leading to aberrant DNA methylation that may contribute to the pathogenesis of acute myeloid leukemia (AML). However, the molecular basis underlying the mutation-induced functional misregulation of DNMT3A remains unclear. Here, we report the crystal structures of the DNMT3A methyltransferase domain, revealing a molecular basis for its oligomerization behavior distinct to DNMT3B, and the enhanced intermolecular contacts caused by the R882H or R882C mutation. Our biochemical, cellular, and genomic DNA methylation analyses demonstrate that introducing the DNMT3B-converting mutations inhibits the R882H-/R882C-triggered DNMT3A polymerization and enhances substrate access, thereby eliminating the dominant-negative effect of the DNMT3A R882 mutations in cells. Together, this study provides mechanistic insights into DNMT3A R882 mutations-triggered aberrant oligomerization and DNA hypomethylation in AML, with important implications in cancer therapy.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2025

2025 ASMS Investigation of the Collision-Induced Dissociation Mechanism of Protonated TODGA with IRIS

Title (20 words): Investigation of the Collision-Induced Dissociation Mechanism of Protonated TODGA with IRIS Introduction (120 words): One of the challenges facing wide-spread adoption of nuclear power is the development of efficient separation processes for used nuclear fuel. The molecules in separation processes are subjected to an extreme environment due to the high radiation fields from the used fuel and highly acidic media used for fuel dissolution, which results in significant molecular degradation, leading to reduced process efficiency. These degradation products must be identified and studied so mitigation strategies can be developed to maintain process efficiency. However, complex systems can have many degradation products, complicating identification. Untargeted analysis tools could be used to understand radiation chemistry in complex systems. However, this would necessitate improved understanding of the gas-phase fragmentation mechanisms of fuel cycle molecules like tetraoctyldiglycolamide (TODGA). Methods (120 words): The gas-phase fragmentation of protonated TODGA was investigated using collision-induced dissociation (CID), resonance ejection, and infrared ion spectroscopy (IRIS). CID and resonance ejection experiments were conducted using a Bruker Daltonics (Bremen, Gemany) SolariX XR fourier transform ion cyclotron resonance (FT-ICR) mass spectrometer. IRIS spectra of protonated TODGA and its two CID fragmentation products were measured using a modified Bruker amaZon Speed ETD 3D quadrupole ion trap mass spectrometer coupled to the Free Electron Lasers for Infrared eXperiments (FELIX) free electron laser. Measured spectra were compared with density functional theory (DFT) calculations using the Gaussian 16, Revision C.02 software package with the ?B97X-D functional and def2-TZVPP basis sets. Candidate structures were generated using the CREST 3.0 conformational sampling software tool. Preliminary Data (300 words): Collision-induced dissociation of protonated TODGA ([C36H73N2O3]+, m/z=581.562) results two fragment ions, one at m/z=340.285 assigned as [C20H38NO3]+ and the other at m/z=312.290, assigned as [C19H38NO2]+. Based on the assigned formula and the structure of protonated TODGA, the fragment at m/z=340.285 is likely formed from elimination of neutral dioctylamine. Comparison of the IRIS spectrum of m/z=340.285 with DFT predictions suggests it contains a ring structure, and is assigned as N-octyl-N-(6-oxo-1,4-dioxan-2-ylidene)octan-1-aminium. Based on this structure and the structure of protonated TODGA, we hypothesize this fragment formed from elimination of neutral dioctylamine followed by a ring closure mechanism. Comparison of the IRIS spectrum of the fragment at m/z=312.290 with DFT predictions also indicated the presence of a ring structure, assigned as N-(1,3-dioxolan-4-ylidene)-N-octyloctan-1-aminium. This product could be formed from elimination of carbon monoxide from the ring of m/z=340.285 as a sequential fragmentation or formed directly from protonated TODGA via elimination of neutral N,N-dioctylformamide followed by a ring closure. Resonance ejection experiments where m/z=340 was continuously ejected from the IRC cell showed no decrease in intensity of m/z=312.290 across several collision energies, suggesting that the later, direct formation mechanism, dominates. The location of the ionizing proton in protonated TODGA is important for modeling the fragmentation mechanisms. DFT calculations suggested that the position of bands involving the coupled vibrations of the amide C—N and C=O bonds in TODGA are the most sensitive to proton location. Evaluation of the IRIS spectrum of protonated TODGA suggests that the ionizing proton is located between the two amid oxygens. This protonation location was calculated to lie approximately 30 kJ/mol lower in energy than the next lowest energy location, with the proton located solely on one of the amide oxygens. Novel aspect (20 words): Infrared ion spectroscopy combined with resonance ejection experiments and density functional theory to probe the collision-induced dissociation mechanism of tetraoctyldiglycolamide.

37 - INORGANIC, ORGANIC, PHYSICAL AND ANALYTICAL C↗
DOE OSTI2017

How Well Can the M06 Suite of Functionals Describe the Electron Densities of Ne, Ne 6+ , and Ne 8+ ?

The development of better approximations to the exact exchange-correlation functional is essential to the accuracy of density functionals. A recent study indicated that functionals with few parameters offer more accurate electron densities than recently developed many-parameter functionals for light closed-shell atomic systems. In this study, we calculated electron densities, their gradients, and Laplacians of Ne, Ne 6+ , and Ne 8+ using 19 electronic structure methods, and we compared them to the CCSD reference results. Two basis sets, namely, aug-cc-pωCV5Z and aug-cc-pV5Z, are utilized in the calculations. We found that the choice of basis set has a significant impact on the errors and rankings of some of the selected methods. The errors of electron densities, their gradients, and Laplacians calculated with the aug-cc-pV5Z basis set are substantially reduced, especially for Minnesota density functionals, as compared to the results using the aug-cc-pωCV5Z basis set (a larger basis set utilized in earlier work (Medvedev et al. Science2017, 355, 49–52)). The rankings of the M06 suite of functionals among the 19 methods are greatly improved with the aug-cc-pV5Z basis set. Furthermore, the performances of the HSE06, BMK, MN12-L, and MN12-SX functionals are also improved with the aug-cc-pV5Z basis set. The M06 suite of functionals is capable of providing accurate electron densities, gradients, and Laplacians using the aug-cc-pV5Z basis set, and thus it is suitable for a wide range of applications in chemistry and physics.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2017

Thermal out-of-time-order correlators, KMS relations, and spectral functions

We describe general features of thermal correlation functions in quantum systems, with specific focus on the fluctuation-dissipation type relations implied by the KMS condition. These end up relating correlation functions with different time ordering and thus should naturally be viewed in the larger context of out-of-time-ordered (OTO) observables. In particular, eschewing the standard formulation of KMS relations where thermal periodicity is combined with time-reversal to stay within the purview of Schwinger-Keldysh functional integrals, we show that there is a natural way to phrase them directly in terms of OTO correlators. We use these observations to construct a natural causal basis for thermal n-point functions in terms of fully nested commutators. We provide several general results which can be inferred from cyclic orbits of permutations, and exemplify the abstract results using a quantum oscillator as an explicit example.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
DOE OSTI2025

Method-independent cusps for atomic orbitals in quantum Monte Carlo

Here, we present an approach for augmenting Gaussian atomic orbitals with correct nuclear cusps. Like the atomic orbital basis set itself and unlike previous cusp corrections, this approach is independent of the many-body method used to prepare wave functions for quantum Monte Carlo. Once the basis set and molecular geometry are specified, the cusp-corrected atomic orbitals are uniquely specified, regardless of which density functionals, quantum chemistry methods, or subsequent variational Monte Carlo optimizations are employed. We analyze the statistical improvement offered by these cusps in a number of molecules and find them to offer similar advantages as molecular-orbital-based approaches while remaining independent of the choice of many-body method.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2014

Effects of thermal cycling parameters on residual stresses in alumina scales of CoNiCrAlY and NiCoCrAlY bond coats

In this paper, furnace cycling experiments were performed on free-standing high-velocity oxygen-fuel bond coat samples to investigate the effect of material composition, surface texture, and cycling conditions on the average stresses in the formed oxide scales after cooling. The oxide scale thicknesses were determined by SEM image analyses and information about the stresses were acquired by photo-stimulated luminescence-spectroscopy. Additionally, the scale thickness dependent stress fields were calculated in finite-element analyses including approximation functions for the surface roughness derived on the basis of profilometry data. The evolution of the average residual stress as a function of oxide scale thickness was subject to stochastic fluctuations predominantly caused by local scale spallations. In comparison to the supplemental modeling results, thermal stresses due to mismatch of thermal expansion coefficients are identified as the main contribution to the residual stresses. Finally, the theoretical results emphasize that analyses of spectroscopic data acquired for average stress investigations of alumina scales rely on detailed information about microstructural features.

36 MATERIALS SCIENCE↗
NASA NTRS1982

Integrated Application of Active Controls (IAAC) technology to an advanced subsonic transport project. ACT/Control/Guidance System study, volume 1

The active control technology (ACT) control/guidance system task of the integrated application of active controls (IAAC) technology project within the NASA energy efficient transport program was documented. The air traffic environment of navigation and air traffic control systems and procedures were extrapolated. An approach to listing flight functions which will be performed by systems and crew of an ACT configured airplane of the 1990s, and a determination of function criticalities to safety of flight, are the basis of candidate integrated ACT/Control/Guidance System architecture. The system mechanizes five active control functions: pitch augmented stability, angle of attack limiting, lateral/directional augmented stability, gust load alleviation, and maneuver load control. The scope and requirements of a program for simulating the integrated ACT avionics and flight deck system, with pilot in the loop, are defined, system and crew interface elements are simulated, and mechanization is recommended. Relationships between system design and crew roles and procedures are evaluated.

Source record↗
DOE OSTI2025

Multi-fidelity learning for interatomic potentials: low-level forces and high-level energies are all you need

The promise of machine learning interatomic potentials (MLIPs) has led to an abundance of public quantum mechanical (QM) training datasets. The quality of an MLIP is directly limited by the accuracy of the energies and atomic forces in the training dataset. Unfortunately, most of these datasets are computed with relatively low-accuracy QM methods, e.g. density functional theory with a moderate basis set. Due to the increased computational cost of more accurate QM methods, e.g. coupled-cluster theory with a complete basis set (CBS) extrapolation, most high-accuracy datasets are much smaller and often do not contain atomic forces. The lack of high-accuracy atomic forces is quite troubling, as training with force data greatly improves the stability and quality of the MLIP compared to training to energy alone. Because most datasets are computed with a unique level of theory, traditional single-fidelity (SF) learning is not capable of leveraging the vast amounts of published QM data. In this study, we apply multi-fidelity learning (MFL) to train an MLIP to multiple QM datasets of different levels of accuracy, i.e. levels of fidelity. Specifically, we perform three test cases to demonstrate that MFL with both low-level forces and high-level energies yields an extremely accurate MLIP—far more accurate than a SF MLIP trained solely to high-level energies and almost as accurate as a SF MLIP trained directly to high-level energies and forces. Therefore, MFL greatly alleviates the need for generating large and expensive datasets containing high-accuracy atomic forces and allows for more effective training to existing high-accuracy energy-only datasets. Indeed, low-accuracy atomic forces and high-accuracy energies are all that are needed to achieve a high-accuracy MLIP with MFL.

36 MATERIALS SCIENCE↗
DOE OSTI2019

Toward Fast and Reliable Potential Energy Surfaces for Metallic Pt Clusters by Hierarchical Delta Neural Networks

Data-driven machine learning force fields (MLF) are more and more popular in atomistic simulations, and exploit machine learning methods to predict energies and forces for unknown structures based on the knowledge learned from an existing reference database. The latter usually comes from density functional theory calculations. One main drawback of MLFs is that physical laws are not incorporated in the machine learning models and instead, MLFs are designed to be very flexible to simulate complex quantum chemistry potential energy surface (PES). In general, MLFs have poor transferability, and hence a very large trainset is required to span all the target feature space to get a reliable MLF. This procedure becomes more troublesome when the PES is complicated, with a large number of degrees of freedom, in which building a large database is inevitable and very expensive, especially when accurate but costly exchange-correlation functionals have to be used. In this manuscript, we exploit a high dimensional neural network potential (HDNNP) on Pt clusters of size 6 to 20 as one example. Furthermore, our standard level of energy calculation is DFT GGA (PBE) using a plane wave basis set. We introduce an approximate but fast level with the PBE functional and a minimal atomic orbital basis set, then a more accurate but expensive level, using a hybrid functional or non-local vdw functional and a plane wave basis set, is reliably predicted by learning the difference with HDNNP. Findings show that such a differential approach (named ΔHDNNP) can deliver very accurate predictions (error < 10 meV/atom) in reference to converged basis set energies as well as more accurate but expensive xc functional. The overall speedup can be as large as 900 for 20 atom Pt cluster. More importantly, ΔHDNNP shows much better transferability due to the intrinsic smoothness of delta potential energy surface, and accordingly one can use much smaller trainset data to obtain better accuracy than the conventional HDNNP. A multi-layer ΔHDNNP is thus proposed to obtain very accurate predictions versus expensive non-local vdW functional calculations in which the required trainset is further reduced. The approach can be easily generalized to any other machine learning methods and opens a path to study the structure and dynamics of Pt clusters and nanoparticles.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗