Search NASA⌕ Search

SEARCH · Search NASA

Results for “open source codes”

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 721 records · Page 40
All sourcesNASA NTRSDOE OSTI
DOE OSTI2024

A graphics processing unit accelerated sparse direct solver and preconditioner with block low rank compression

We present the GPU implementation efforts and challenges of the sparse solver package STRUMPACK. The code is made publicly available on github with a permissive BSD license. STRUMPACK implements an approximate multifrontal solver, a sparse LU factorization which makes use of compression methods to accelerate time to solution and reduce memory usage. Multiple compression schemes based on rank-structured and hierarchical matrix approximations are supported, including hierarchically semi-separable, hierarchically off-diagonal butterfly, and block low rank. Here, in this paper, we present the GPU implementation of the block low rank (BLR) compression method within a multifrontal solver. Our GPU implementation relies on highly optimized vendor libraries such as cuBLAS and cuSOLVER for NVIDIA GPUs, rocBLAS and rocSOLVER for AMD GPUs and the Intel oneAPI Math Kernel Library (oneMKL) for Intel GPUs. Additionally, we rely on external open source libraries such as SLATE (Software for Linear Algebra Targeting Exascale), MAGMA (Matrix Algebra on GPU and Multi-core Architectures), and KBLAS (KAUST BLAS). SLATE is used as a GPU-capable ScaLAPACK replacement. From MAGMA we use variable sized batched dense linear algebra operations such as GEMM, TRSM and LU with partial pivoting. KBLAS provides efficient (batched) low rank matrix compression for NVIDIA GPUs using an adaptive randomized sampling scheme. The resulting sparse solver and preconditioner runs on NVIDIA, AMD and Intel GPUs. Interfaces are available from PETSc, Trilinos and MFEM, or the solver can be used directly in user code. We report results for a range of benchmark applications, using the Perlmutter system from NERSC, Frontier from ORNL, and Aurora from ALCF. For a high frequency wave equation on a regular mesh, using 32 Perlmutter compute nodes, the factorization phase of the exact GPU solver is about 6.5× faster compared to the CPU-only solver. The BLR-enabled GPU solver is about 13.8× faster than the CPU exact solver. For a collection of SuiteSparse matrices, the STRUMPACK exact factorization on a single GPU is on average 1.9× faster than NVIDIA’s cuDSS solver.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2021

Common workflows for computing material properties using different quantum engines

The prediction of material properties based on density-functional theory has become routinely common, thanks, in part, to the steady increase in the number and robustness of available simulation packages. This plurality of codes and methods is both a boon and a burden. While providing great opportunities for cross-verification, these packages adopt different methods, algorithms, and paradigms, making it challenging to choose, master, and efficiently use them. We demonstrate how developing common interfaces for workflows that automatically compute material properties greatly simplifies interoperability and cross-verification. We introduce design rules for reusable, code-agnostic, workflow interfaces to compute well-defined material properties, which we implement for eleven quantum engines and use to compute various material properties. Each implementation encodes carefully selected simulation parameters and workflow logic, making the implementer’s expertise of the quantum engine directly available to non-experts. All workflows are made available as open-source and full reproducibility of the workflows is guaranteed through the use of the AiiDA infrastructure.

36 MATERIALS SCIENCE↗
DOE OSTI2023

Few-shot leaf segmentation

In this work, we use few-shot learning to segment the body and vein architecture of Populus trichocarpa leaves from high-resolution scans obtained in the UC Davis common garden. Leaf and vein segmentation are formulated as separate tasks, in which convolutional neural networks (CNNs) are used to iteratively expand partial segmentations until reaching stopping criteria. Our leaf and vein segmentation approaches use just 50 and 8 manually traced images for training, respectively, and are applied to a set of 2,634 top and bottom leaf scans. We show that both methods achieve high segmentation accuracy, in some cases exceeding even human-level segmentation. The leaf and vein segmentations are subsequently used to extract 68 morphological traits using traditional open-source image processing tools, which are validated using real-world physical measurements. For a biological perspective, we perform a genome-wide association study using the "vein density" trait to discover novel genetic architectures associated with multiple physiological processes relating to leaf development and function. In addition to sharing all of the few-shot learning code, we are releasing all images, manual segmentations, model predictions, 68 extracted leaf phenotypes, and a new set of SNPs called against the v4 P. trichocarpa genome for 1,419 genotypes.

Lagergren, John [Oak Ridge National Lab. (ORNL), O↗
DOE OSTI2023

Supporting information for Few-Shot Learning Enables Population-Scale Analysis of Leaf Traits in Populus trichocarpa

In this work, we use few-shot learning to segment the body and vein architecture of P. trichocarpa leaves from high-resolution scans obtained in the UC Davis common garden. Leaf and vein segmentation are formulated as separate tasks, in which convolutional neural networks (CNNs) are used to iteratively expand partial segmentations until reaching stopping criteria. Our leaf and vein segmentation approaches use just 50 and 8 manually traced images for training, respectively, and are applied to a set of 2,634 top and bottom leaf scans. We show that both methods achieve high segmentation accuracy, in some cases exceeding even human-level segmentation. The leaf and vein segmentations are subsequently used to extract 68 morphological traits using traditional open-source image processing tools, which are validated using real-world physical measurements. For a biological perspective, we perform a genome-wide association study using the vein density trait to discover novel genetic architectures associated with multiple physiological processes relating to leaf development and function. In addition to sharing all of the few-shot learning code (see https://github.com/jlager/few-shot-leaf-segmentation), we are releasing all images, manual segmentations, model predictions, 68 extracted leaf phenotypes, and a new set of SNPs called against the v4 P. trichocarpa genome for 1,419 genotypes. The data folder includes all images, ground truth segmentations, predicted segmentations, and extracted leaf traits. All images encode the sample ID in the file name by indicating the treatment, block, row, position, and leaf side, respectively. For example, the file, C_1_1_2_bot.jpeg, indicates the control treatment, block 1, row 1, position 2, and the bottom side of the leaf. Tabulated results include position IDs as well as the corresponding genotype IDs. The images folder includes the 2,906 high-resolution leaf scans taken in the field. The leaf_masks folder includes 50 ground truth segmentations used for training the leaf tracing algorithm. The leaf_preds folder includes the 2,906 predicted segmentations from the leaf tracing algorithm. The vein_masks folder includes 8 ground truth segmentations used for training the vein growing algorithm. The vein_preds folder includes the 1,453 predicted segmentations from the vein growing algorithm. The vein_probs folder includes the 1,453 predicted probability maps from the vein growing algorithm before thresholding. The genomes folder includes the set of SNPs called against the v4 P. trichocarpa genome for 1,419 genotypes with a README file detailing the steps taken. The results folder includes: (i) raw values of the 68 predicted leaf traits in digital_traits.tsv, (ii) manually measured values of petiole length and width in manual_traits.tsv, (iii) thin plate spline (TPS) adjusted values of the vein density trait in vein_density_tps_adj.tsv, (iv) best linear unbiased prediction (BLUP) adjusted values of the vein density trait in vein_density_blups.tsv, and (v) GWAS results for the vein density trait, including chromosome positions and corresponding P values, in gwas_results.csv.

09 BIOMASS FUELS↗
NASA NTRS2010

Support for Systematic Code Reviews with the SCRUB Tool

SCRUB is a code review tool that supports both large, team-based software development efforts (e.g., for mission software) as well as individual tasks. The tool was developed at JPL to support a new, streamlined code review process that combines human-generated review reports with program-generated review reports from a customizable range of state-of-the-art source code analyzers. The leading commercial tools include Codesonar, Coverity, and Klocwork, each of which can achieve a reasonably low rate of false-positives in the warnings that they generate. The time required to analyze code with these tools can vary greatly. In each case, however, the tools produce results that would be difficult to realize with human code inspections alone. There is little overlap in the results produced by the different analyzers, and each analyzer used generally increases the effectiveness of the overall effort. The SCRUB tool allows all reports to be accessed through a single, uniform interface (see figure) that facilitates brows ing code and reports. Improvements over existing software include significant simplification, and leveraging of a range of commercial, static source code analyzers in a single, uniform framework. The tool runs as a small stand-alone application, avoiding the security problems related to tools based on Web browsers. A developer or reviewer, for instance, must have already obtained access rights to a code base before that code can be browsed and reviewed with the SCRUB tool. The tool cannot open any files or folders to which the user does not already have access. This means that the tool does not need to enforce or administer any additional security policies. The analysis results presented through the SCRUB tool s user interface are always computed off-line, given that, especially for larger projects, this computation can take longer than appropriate for interactive tool use. The recommended code review process that is supported by the SCRUB tool consists of three phases: Code Review, Developer Response, and Closeout Resolution. In the Code Review phase, all tool-based analysis reports are generated, and specific comments from expert code reviewers are entered into the SCRUB tool. In the second phase, Developer Response, the developer is asked to respond to each comment and tool-report that was produced, either agreeing or disagreeing to provide a fix that addresses the issue that was raised. In the third phase, Closeout Resolution, all disagreements are discussed in a meeting of all parties involved, and a resolution is made for all disagreements. The first two phases generally take one week each, and the third phase is concluded in a single closeout meeting.

Holzmann, Gerald J.↗
NASA NTRSDate not supplied

fmdtools Tutorial: Intro to Resilience Modelling, Simulation, and Visualization in Python With fmdtools

This workshop will cover the basics of using the fmdtools package for the simulation of hazardous scenarios for resilience simulation. The fmdtools simulation package is an open-source python toolkit for simulating the dynamic response of a system to internal and externally-driven hazardous scenarios, including faults and environmental conditions, that can be used to analyze the risks related to these hazards. Prior to the development of fmdtools, researchers had to either adapt an (often limited) propriety toolkit or develop their own design/simulation/analysis codes to develop their models of hazardous events, a significant technical burden to both (1) leveraging resilience modeling methodologies and (2) extending these methodologies with their own contributions. The fmdtools package provides a number of model constructs and simulation and analysis methods to enable the designer to focus to solely on their modeling case-study while still enabling a significant degree of model expressiveness and adaptability via Python-based model definition. This tutorial will present the setup of fmdtools and a high-level overview of its use, as well as some simple examples for understanding how to leverage its modelling, simulation, and analysis capabilities. Familiarity with jupyter notebook and basic python will be assumed.

Daniel Hulse↗
DOE OSTI2021

Design and physics basis for the upcoming DIII-D SAS-VW campaign to quantify tungsten leakage and transport in a new slot divertor geometry

A set of experiments are planned to exploit the high SOL collisionality enabled by a tightly baffled slot divertor geometry to suppress tungsten leakage in DIII-D. A toroidal row of graphite tiles from the Small Angle Slot (SAS) divertor is being coated with 10–15 μ m of tungsten. New spectroscopic viewing chords with in-vacuo optics will measure the W gross erosion source from the divertor surface with high spatial and temporal resolution. In parallel, the bottom of the SAS divertor is changed from a flat to a ‘V’ shape. New SOLPS-ITER/DIVIMP simulations conducted with drifts using the planned ‘V’ shape predict a substantial reduction in W sourcing and SOL accumulation in either B × ∇B direction relative to either the old SAS divertor shape or the open, lower divertor. Dedicated studies are planned to carefully characterize the level of W sourcing, leakage, and scrape-off-layer (SOL) accumulation in DIII-D over a wide range of plasma scenarios. Various actuators will be assessed for their efficacy in further reducing high-Z impurity sources and leakage from the slot divertor geometry. Finally, this coupled code-experiment validation effort will be used to stress-test physics models and build confidence in extrapolations to advanced, high-Z divertor geometries for next-step devices.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
NASA NTRS1993

A capaciflector-based robotic system

This report presents preliminary results obtained from a research grant entitled 'Capaciflector-based Control and Imaging,' with Grant Number NAG 5-780, for the period between February 1, 1993 and August 1, 1993. This report deals with the development of a robotic system which is used to evaluate the feasibility of servicing and repairing the Hubble Space Telescope (HST) using robots. In particular, the task of opening the HST's toolbox is considered in this report. First the main components of the robotic system will be introduced. Then each component will be described in detail from low-level to high-level. Finally tasks that have been accomplished during the reporting period for the development of the robotic system will be presented. Listings of source codes for the accomplished tasks are given at the end of the report.

Nguyen, Charles C.↗
NASA NTRS2018

T-infinity: The Dependency Inversion Principle for Rapid and Sustainable Multidisciplinary Software Development

The CFD Vision 2030 Study recommends that, “NASA should develop and maintain an integrated simulation and software development infrastructure to enable rapid CFD technology maturation.... [S]oftware standards and interfaces must be emphasized and supported whenever possible, and open source models for noncritical technology components should be adopted.” The current paper presents an approach to an open source development architecture, named T-infinity, for accelerated research in CFD leveraging the Dependency Inversion Principle to realize plugins that communicate through collections of functions without exposing internal data structures. Steady state flow visualization, mesh adaptation, fluid-structure interaction, and overset domain capabilities are demonstrated through compositions of plugins via standardized abstract interfaces without the need for source code dependencies between disciplines. Plugins interact through abstract interfaces thereby avoiding N 2 direct code-to-code data structure coupling where N is the number of codes. This plugin architecture enhances sustainable development by controlling the interaction between components to limit software complexity growth. The use of T-infinity abstract interfaces enables multidisciplinary application developers to leverage legacy applications alongside newly-developed capabilities. While rein, a description of interface details is deferred until the are more thoroughly tested and can be closed to modification.

O'Connell, Matthew D.↗
DOE OSTI2020

Continued Validation Studies using the MOOSE Framework for Plasma Simulation with Electromagnetics

Resources and tools for the modeling and simulation of low-temperature plasma (LTP) discharges are increasingly vital to progress in the field in order to properly characterize and study complex source designs and plasma chemistries beyond the scope of traditional diagnostics. Open-source software provides powerful platforms for this work and can enable community-driven LTP R&D. Within the Multiphysics Object-Oriented Simulation Environment (MOOSE) open-source framework [1], capabilities have been demonstrated in the areas of plasma fluids (Zapdos [2]), plasma chemistry (CRANE [3]), and general electromagnetic wave theory (Electromagnetic Library for Kinetics & fluids [ELK] [4]). ELK has since been coupled to Zapdos/CRANE to enable fully coupled electromagnetic plasma simulations. This talk will detail the continued validation efforts and discuss Zapdos-ELK-CRANE code coupling with various-low temperature plasma sources. The impact of fully coupled electromagnetics on process parameters (e.g., temperature and electron/ion energy) versus an electrostatic description will also be discussed. [1] Permann et al., SoftwareX, 11 (2020) 100430. [2] Lindsay et al., J. Phys. D: Appl. Phys. 49 (2016) 235204. [3] Keniley et al., Talk, GEC Session DT3.00002 (2019). [4] Icenhour et al., Talk, GEC Session RR1.00006 (2019). Acknowledgements: INL Graduate Fellowship Program, NSF SI2 Grant 1740300, U.S. Dept. of Energy Office of Science Graduate Student Research Program

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
NASA NTRS2018

The CEOS Data Cube Portal: A User-Friendly, Open Source Software Solution for the Distribution, Exploration, Analysis, and Visualization of Analysis Ready Data

There is an urgent need to increase the capacity of developing countries to take part in the study and monitoring of their environments through remote sensing and space-based Earth observation technologies. The Open Data Cube (ODC) provides a mechanism for efficient storage and a powerful framework for processing and analyzing satellite data. While this is ideal for scientific research, the expansive feature space can also be daunting for end-users and decision-makers who simply require a solution which provides easy exploration, analysis, and visualization of Analysis Ready Data (ARD). Utilizing innovative web-design and a modular architecture, the Committee on Earth Observation Satellites (CEOS) has created a web-based user interface (UI) which harnesses the power of the ODC yet provides a simple and familiar user experience: the CEOS Data Cube (CDC). This paper presents an overview of the CDC architecture and the salient features of the UI. In order to provide adaptability, flexibility, scalability, and robustness, we leverage widely-adopted and well-supported technologies such as the Django web framework and the AWS Cloud platform. The fully-customizable source code of the UI is available at our public repository. Interested parties can download the source and build their own UIs. The UI empowers users by providing features that assist with streamlining data preparation, data processing, data visualization, and sub-setting ARD products in order to achieve a wide variety of Earth imaging objectives through an easy to use web interface.

User Interface↗
DOE OSTI2023

Theoretical framework for new magnetic materials for quantum computing and information storage. Final report for the Award No. DE-SC0018910

The focus of this grant was on molecular magnetic materials for information storage and quantum computing. We have been developing robust, first-principle methods for computing relevant electronic and magnetic properties of molecular building blocks (SMMs) of novel magnetic materials and quantum computers. These tools enable theoretical modeling of SMMs’ behavior, facilitating the interpretation of experimental studies and aiding the design of novel magnetic materials. Our strategy is based on the spin-flip (SF) approach, which extends the hierarchy of black-box single-reference methods to strongly correlated systems. Specifically, we developed general scalable algorithms and computer codes for calculating molecular properties, with an emphasis on spin-related properties, such as zero-field splittings, hyperfine couplings, and g-tensors. While our primary focus was on SF wave functions and SF-TDDFT, the underlying theory and computer codes were formulated using reduced density matrices, such that these tools are applicable to a broader class of methods. To extend the scope of applicability of wave-function-based SF methods to larger systems, we developed reduced-scaling approaches for the equation-of-motion coupled-cluster (EOM-CC) methods and continue developing libtensor (our open-source general tensor contraction library for many-body methods). We carried out extensive benchmarks and also carried out several applications.

36 MATERIALS SCIENCE↗
DOE OSTI2020

Enabling Catalog Simulations of Transient and Variable Sources Based on LSST Cadence Strategies

The American Astronomical Society. All rights reserved.. The Large Synoptic Survey Telescope (LSST) project will conduct a 10 year multi-band survey starting in 2022. Observing strategies for this survey are being actively investigated, and the science capabilities can be best forecasted on the basis of simulated strategies from the LSST Operations Simulator (OpSim). This paper describes the way in which OpSim simulates a stochastic realization of the sequence of LSST pointings over the survey duration, and is based on a model of the observatory (including telescope) and historical data of observational conditions. OpSim outputs contain a record of each simulated pointing of the survey along with a complete characterization of the pointing in terms of observing conditions, and some useful quantities derived from the characteristics of the pointing. Thus, each record can be efficiently used to derive the properties of observations of all astrophysical sources found in that pointing. However, in order to obtain the time series of observations (light curves) of a set of sources, it is often more convenient to compute all observations of an astrophysical source, and iterate over sources. In this document, we describe the open source python package OpSimSummary, which allows for a convenient reordering. The objectives of this package are to provide users with an Application Programming Interface for accessing all such observations and summarizing this information in the form of intermediate data products usable by third party software such as SNANA, thereby also bridging the gap between official LSST products and preexisting simulation codes.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2025

Preliminary Study on Fine-Grained Power and Energy Measurements on Grace Hopper GH200 with Open-Source Performance Tools

The increasing adoption of tightly integrated, heterogeneous architectures, combined with the slowdown of Moore’s law, has made application power and energy-driven optimizations critical to efficiently use high-performance computing systems. This paper introduces a newly developed open-source toolkit that seamlessly integrates the Linux real-time hardware monitoring program hwmon with the Performance Application Programming Interface and the Score-P performance measurement system, thereby enabling fine-grained power and energy measurements for high-performance computing applications. Our primary target platform is the Wombat test bed, which is a system based on the NVIDIA GH200 superchip. The toolkit can capture transient power peaks with high temporal resolution (50 ms) and, thanks to Score-P integration, can map power metrics to specific code regions, thereby providing actionable information on power-intensive operations and inefficiencies. The toolkit also provides a holistic view of both the power and the energy consumption of the entire GH200 superchip by covering all major components: the Grace CPU, the Hopper GPU, and the I/O subsystem. Experiments that use Locally Self-consistent Multiple Scattering, which is an application for first-principles calculations of materials developed at Oak Ridge National Laboratory, have demonstrated the tool’s ability to identify transient power spikes and uncover opportunities for energy-aware optimizations. Additionally, we introduce a Python-based utility for converting Open Trace Format 2 traces to Parquet format, thus enabling advanced data analysis for numerical integration methods applied to power data for accurate energy profiling.

Hernandez Mendoza, Oscar [ORNL] (ORCID:00000002538↗
DOE OSTI2023

Technical Justifications for Liquid Hydrogen Exposure Distances

The previous separation distances in the National Fire Protection Association (NFPA) Hydrogen Technologies Code (NFPA 2, 2020 Edition) for bulk liquid hydrogen systems lack a well-documented basis and can be onerous. This report describes the technical justifications for revisions of the bulk liquid hydrogen storage setback distances in NFPA 2, 2023 Edition. Distances are calculated based on a leak area that is 5% of the nominal pipe flow area. Models from the open source HyRAM+ toolkit are used to justify the leak size as well as calculate consequence-based separation distances from that leak size. Validation and verification of the numerical models is provided, as well as justification for the harm criteria used for the determination of the setback distances for each exposure type. This report also reviews mitigations that could result in setback distance reduction. The resulting updates to the liquid hydrogen separation distances are well-documented, retrievable, repeatable, revisable, independently verified, and use experimental results to verify the models.

08 HYDROGEN↗
NASA NTRS2003

XML: James Webb Space Telescope Database Issues, Lessons, and Status

This paper will present the current concept using extensible Markup Language (XML) as the underlying structure for the James Webb Space Telescope (JWST) database. The purpose of using XML is to provide a JWST database, independent of any portion of the ground system, yet still compatible with the various systems using a variety of different structures. The testing of the JWST Flight Software (FSW) started in 2002, yet the launch is scheduled for 2011 with a planned 5-year mission and a 5-year follow on option. The initial database and ground system elements, including the commands, telemetry, and ground system tools will be used for 19 years, plus post mission activities. During the Integration and Test (I&T) phases of the JWST development, 24 distinct laboratories, each geographically dispersed, will have local database tools with an XML database. Each of these laboratories database tools will be used for the exporting and importing of data both locally and to a central database system, inputting data to the database certification process, and providing various reports. A centralized certified database repository will be maintained by the Space Telescope Science Institute (STScI), in Baltimore, Maryland, USA. One of the challenges for the database is to be flexible enough to allow for the upgrade, addition or changing of individual items without effecting the entire ground system. Also, using XML should allow for the altering of the import and export formats needed by the various elements, tracking the verification/validation of each database item, allow many organizations to provide database inputs, and the merging of the many existing database processes into one central database structure throughout the JWST program. Many National Aeronautics and Space Administration (NASA) projects have attempted to take advantage of open source and commercial technology. Often this causes a greater reliance on the use of Commercial-Off-The-Shelf (COTS), which is often limiting. In our review of the database requirements and the COTS software available, only very expensive COTS software will meet 90% of requirements. Even with the high projected initial cost of COTS, the development and support for custom code over the 19-year mission period was forecasted to be higher than the total licensing costs. A group did look at reusing existing database tools and formats. If the JWST database was already in a mature state, the reuse made sense, but with the database still needing to handing the addition of different types of command and telemetry structures, defining new spacecraft systems, accept input and export to systems which has not been defined yet, XML provided the flexibility desired. It remains to be determined whether the XML database will reduce the over all cost for the JWST mission.

Detter, Ryan↗
NASA NTRS2021

Aircraft Flaps Modeling in OpenMDAO

The goal of this project was to develop a model for a single subsystem in the aerodynamics discipline, in this case the flaps of an aircraft. Flaps are high-lift devices used by planes to allow for quicker takeoffs, and slower landings. A computer-based aircraft model of the flaps of an aircraft was developed using the Python based open-source framework OpenMDAO. OpenMDAO is used to develop multi-disciplinary aircraft models using gradient-based optimization; design optimization (MDO) is concerned with solving design problems involving numerical models of complex engineering systems. There were 4 components in the model; each has input values, output variables, and equations to calculate said outputs. The variables and equations are sourced from NASA Fortran code from the 1970s, in a project called the General Aviation Synthesis Program (GASP). These variables and equations which create the model are being converted to Python for ease of use. The flaps model developed will be integrated into a larger model of a conventional aircraft’s flight phases. All subsystems of the model will first be built using the parameters of a Boeing 737 MAX-8, to validate its functionality and accuracy. Then, the aircraft model will be used for hybrid-electric research; running optimizations to improve efficiency, minimize fuel burn, and advance hybrid-electric technology in the aerospace field.

Computer-based aircraft modeling↗
DOE OSTI2024

Synergizing human expertise and AI efficiency with language model for microscopy operation and automated experiment design

With the advent of large language models (LLMs), in both the open source and proprietary domains, attention is turning to how to exploit such artificial intelligence (AI) systems in assisting complex scientific tasks, such as material synthesis, characterization, analysis and discovery. Here, we explore the utility of LLMs, particularly ChatGPT4, in combination with application program interfaces (APIs) in tasks of experimental design, programming workflows, and data analysis in scanning probe microscopy, using both in-house developed APIs and APIs given by a commercial vendor for instrument control. We find that the LLM can be especially useful in converting ideations of experimental workflows to executable code on microscope APIs. Beyond code generation, we find that the GPT4 is capable of analyzing microscopy images in a generic sense. At the same time, we find that GPT4 suffers from an inability to extend beyond basic analyses for more in-depth technical experimental design. We argue that an LLM specifically fine-tuned for individual scientific domains can potentially be a better language interface for converting scientific ideations from human experts to executable workflows. Such a synergy between human expertise and LLM efficiency in experimentation can open new doors for accelerating scientific research, enabling effective experimental protocols sharing in the scientific community.

97 MATHEMATICS AND COMPUTING↗