Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “predicting intent”

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 73 records · Page 4

Detailed Reaction Kinetics for CFD Modeling of Nuclear Fuel Pellet Coating for High Temperature Gas-Cooled Reactors

The research project was related to the Advanced Fuel Cycle Initiative and was in direct alignment with advancing knowledge in the area of Nuclear Fuel Development related to the use of TRISO fuels for high-temperature reactors. The importance of properly coating nuclear fuel pellets received a renewed interest for the safe production of nuclear power to help meet the energy requirements of the United States. High-temperature gas-cooled nuclear reactors use fuel in the form of coated uranium particles, and it is the coating process that was of importance to this project. The coating process requires four coating layers to retain radioactive fission products from escaping into the environment. The first layer consists of porous carbon and serves as a buffer layer to attenuate the fission and accommodate the fuel kernel swelling. The second (inner) layer is of pyrocarbon and provides protection from fission products and supports the third layer, which is silicon carbide. The final (outer) layer is also pyrocarbon and provides a bonding surface and protective barrier for the entire pellet. The coating procedures for the silicon carbide and the outer pyrocarbon layers require knowledge of the detailed kinetics of the reaction processes in the gas phase and at the surfaces where the particles interact with the reactor walls. The intent of this project was to acquire detailed information on the reaction kinetics for the chemical vapor deposition (CVD) of carbon and silicon carbine on uranium fuel pellets, including the location of transition state structures, evaluation of the associated activation energies, and the use of these activation energies in the prediction of reaction rate constants. After the detailed reaction kinetics were determined, the reactions were implemented and tested in a computational fluid dynamics model, MFIX. The intention was to find a reduced mechanism set to reduce the computational time for a simulation, while still providing accurate results. Furthermore, fast chemistry techniques would be coupled to MFIX to effectively treat the complex chemistry thus improve the computational efforts. Based on the reaction kinetics modeling, it was determined that the detailed set of chemical reactions for the thermal decomposition of a methyltrichlorosilane (MTS)/H2 mixture consisted of 45 species and 114 gas-phase reactions. Further work identified a mechanism consisting of approximately 60 surface reactions for the surface chemistry of SiC chemical vapor deposition. A reduced mechanism for the MTS gas-phase pyrolysis was constructed using the scanning method based on optimization concepts, which consisted of only 28 species and 29 reactions. The benefits of this project are that we have determined gas-phase species produced during and after the various decomposition reactions of MTS. The success of the computational approaches can now be used to predict the complex chemistry associated with the CVD process in producing nuclear fuel. It is expected that the knowledge we acquired can be easily transferred and that it will contribute to further experimental investigations. Furthermore, the computational techniques can now be used for reactor design and optimization for the next generation of nuclear reactors.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗

A Study of Single Pass Ion Effects at the ALS

We report the results of experiments on a 'fast beam-ion instability' at the Advanced Light Source (ALS). This ion instability, which can arise even when the ions are not trapped over multiple beam passages, will likely be important for many future accelerators. In our experiments, we filled the ALS storage ring with helium gas, raising the pressure approximately two orders of magnitude above the nominal pressure. With gaps in the bunch train large enough to avoid conventional (multi-turn) ion trapping, we observed a factor of 2-3 increase in the vertical beam size along with coherent beam oscillations which increased along the bunch train. Ion trapping has long been recognized as a potential limitation in electron storage rings. The ions, generated by beam-gas collisions, become trapped in the negative potential of the beam and accumulate over multiple beam passages. The trapped ions are then observed to cause a number of deleterious effects such as an increasing beam phase space, a broadening and shifting of the beam transverse oscillation frequencies (tunes), collective beam instabilities, and beam lifetime reductions. All of these effects are of concern for the next generation of accelerators, such as the B-factories or damping rings for future linear colliders, which will store high beam currents with closely spaced bunches and ultra-low beam emittances. One of the standard solutions used to prevent ion trapping is to include a gap in the bunch train which is long compared to the bunch spacing. In this case, the ions are first strongly-focused by the passing electron bunches and then over-focused in the gap. With a sufficiently large gap, the ions can be driven to large amplitudes where they form a diffuse halo and do not affect the beam. In this paper, we describe experiments that study a new regime of transient ion instabilities predicted to arise in future electron storage rings, and linacs with bunch trains. These future rings and linacs, which will be operated with higher beam currents, small transverse beam emittances, and long bunch trains, will use ion clearing gaps to prevent conventional ion trapping. But, while the ion clearing gap may suppress the conventional ion instabilities, it will not suppress a transient beam-ion instability where ions generated and trapped during the passage of a single train lead to a fast instability. While both conventional and transient ion instabilities have the same origin, namely ions produced by the beam, they have different manifestations and, more importantly, the new transient instability can arise even after the conventional ion instability is cured. This new instability is called the 'Fast Beam-Ion Instability' (FBII). In many future rings, the FBII is predicted to have very fast growth rates, much faster than the damping rates of existing and proposed transverse feedback systems, and thus is a potential limitation. To study the FBII, we performed experiments at the ALS, a 1.5 GeV electron storage ring. At the nominal ALS pressure of about 0.24 nTorr, the FBII is not evident. To study the instability, we intentionally added helium gas to the storage-ring vacuum system until the residual gas pressure was increased about 80 nTorr. This brought the predicted growth rate of the instability at least an order of magnitude above the growth rate of conventional multibunch instabilities driven by the RF cavities and above the damping rate of the transverse feedback system (TFB) in the ALS and, thereby, established conditions very similar to those in a future storage ring. We then filled the ring with a relatively short train of bunches, suppressing conventional ion instabilities. In the following, we will first briefly describe This paper describes the experiment and results in more detail.

43 PARTICLE ACCELERATORS↗

Verification Validation and Uncertainty Quantification for CGS

The overall conduct of verification, validation and uncertainty quantification (VVUQ) is discussed through the construction of a workflow relevant to computational modeling including the turbulence problem in the coarse grained simulation (CGS) approach. The workflow contained herein is defined at a high level and constitutes an overview of the activity. Nonetheless, the workflow represents an essential activity in predictive simulation and modeling. VVUQ is complex and necessarily hierarchical in nature. The particular characteristics of VVUQ elements depend upon where the VVUQ activity takes place in the overall hierarchy of physics and models. In this chapter, we focus on the differences between and interplay among validation, calibration and UQ, as well as the difference between UQ and sensitivity analysis. The discussion in this chapter is at a relatively high level and attempts to explain the key issues associated with the overall conduct of VVUQ. The intention is that computational physicists can refer to this chapter for guidance regarding how VVUQ analyses fit into their efforts toward conducting predictive calculations.

Rider, William J.↗

Design of an Additively Manufactured Recuperator With 800 °C Inlet Temperature for sCO2 Power Cycle Application

Abstract Recuperators with design temperatures at and above 800 °C can further increase the thermal efficiency of supercritical CO2 power cycles by enabling higher turbine exhaust temperatures. Mar-M247 is a well-suited nickel-based superalloy for high temperature service due to its high creep strength that prevents excessive material thickness being required for pressure containment. Additive manufacturing using a high-speed laser-directed energy deposition (L-DED) process presents a promising solution, with build trials demonstrating the ability to produce nonconventional flow channels for enhanced heat transfer. A design process is presented that includes aerothermal and mechanical evaluation to maximize performance within the constraints of the manufacturing process. A 2-D heat transfer network and pressure drop code allows prediction of flow distribution and its effect on overall thermal performance. Established literature correlations, along with CFD simulation, inform the prediction of heat transfer coefficients and friction factors for the flowpaths and enhancement features in the heat exchanger core. Mechanical evaluation using finite element analysis (FEA) modeling with the intent of the ASME Boiler & Pressure Vessel Code (BPVC) Section VIII, Div. 2 assesses the operational safety of the design. The detailed design features annular finned passages that take advantage of helical flow paths to distribute the flow from separated headers to shared heat transfer surfaces. Performance predictions for the recuperator at a 50 kW scale provide insights into the feasibility of the additively manufactured (AM) process to produce recuperators on a commercial scale that extend existing operating envelopes.

Engineering↗

MO-G-17A-08: Applications of Quantitative PET/CT Imaging of Yttrium-90: A Tool for Improving Radioembolization

Purpose: Yttrium-90 (Y90) PET/CT post-treatment imaging of radioembolization has fostered significant interest from both the interventional radiology and nuclear medicine communities over the past few years. Recent literature has demonstrated high quantitative accuracy of Y90 PET at the activity concentrations common in radioembolization. However, few have explored methods in which this information can be clinically applied toward improving patient care. Methods: IRB approval and informed consent was obtained for over 35 Y90 post-treatment imaging studies, performed under 2 distinct protocols. In protocol 1, Y90 PET/CT provided quantitative post-treatment imaging, which was then converted into 3D maps of absorbed-dose. Both images and absorbed dose maps were used to manage patient care. In protocol 2, tumor absorbed-dose measurements from Y90 PET/CT were compared to known tumoricidal thresholds. If insufficient absorbed dose was delivered to the tumor, the patient would have an additional Y90 infusion the same day, providing truly patient-specific Y90 PET/CT based treatment-planning. Results: Y90 PET/CT allowed for a superior post-radioembolization evaluation of technical success compared with conventional Y90 bremsstrahlung SPECT. Due to the exceptional resolution of PET, a direct comparison between the distribution of radioembolization and pre-treatment planning intentions can be made. Further, quantification of tumor absorbed-dose directly from PET/CT imaging allows for the prediction of treatment efficacy based on a comparison with known tumoricidal thresholds. This immediate evaluation allowed treating physicians to consider additional or alternate therapies before discovering clinical failure weeks later. One protocol 2 patient was found to have a subtumoricidal absorbed dose following radioembolization. This patient received a same-day infusion of additional Y90 with identical catheter placement to the first infusion. A robust treatment response was seen on follow-up imaging. Conclusion: Y90 PET/CT produces quantifiable image sets, which can be transformed into 3D maps of absorbed dose. This information can be utilized to improve patient care. SIRTeX Medical, Surefire Medical.

60 APPLIED LIFE SCIENCES↗

Initial CGE Model Results Summary Exogenous and Endogenous Variables Tests

The following discussion presents initial results of tests of the most recent version of the National Infrastructure Simulation and Analysis Center Dynamic Computable General Equilibrium (CGE) model developed by Los Alamos National Laboratory (LANL). The intent of this is to test and assess the model’s behavioral properties. The test evaluated whether the predicted impacts are reasonable from a qualitative perspective. This issue is whether the predicted change, be it an increase or decrease in other model variables, is consistent with prior economic intuition and expectations about the predicted change. One of the purposes of this effort is to determine whether model changes are needed in order to improve its behavior qualitatively and quantitatively.

97 MATHEMATICS AND COMPUTING↗

Machine Learning Emulation of Spatial Deposition from a Multi-Physics Ensemble of Weather and Atmospheric Transport Models

In the event of an accidental or intentional hazardous material release in the atmosphere, researchers often run physics-based atmospheric transport and dispersion models to predict the extent and variation of the contaminant spread. These predictions are imperfect due to propagated uncertainty from atmospheric model physics (or parameterizations) and weather data initial conditions. Ensembles of simulations can be used to estimate uncertainty, but running large ensembles is often very time consuming and resource intensive, even using large supercomputers. In this paper, we present a machine-learning-based method which can be used to quickly emulate spatial deposition patterns from a multi-physics ensemble of dispersion simulations. We use a hybrid linear and logistic regression method that can predict deposition in more than 100,000 grid cells with as few as fifty training examples. Logistic regression provides probabilistic predictions of the presence or absence of hazardous materials, while linear regression predicts the quantity of hazardous materials. The coefficients of the linear regressions also open avenues of exploration regarding interpretability—the presented model can be used to find which physics schemes are most important over different spatial areas. A single regression prediction is on the order of 10,000 times faster than running a weather and dispersion simulation. However, considering the number of weather and dispersion simulations needed to train the regressions, the speed-up achieved when considering the whole ensemble is about 24 times. Ultimately, this work will allow atmospheric researchers to produce potential contamination scenarios with uncertainty estimates faster than previously possible, aiding public servants and first responders.

97 MATHEMATICS AND COMPUTING↗

Hydrodynamic Predictions for the Next Outburst of T Coronae Borealis: It Will Be the Brightest Classical or Recurrent Nova Ever Observed in X-Rays*

T Coronae Borealis (TCrB) is a recurrent nova with recorded outbursts in 1866 and 1946 and possible outbursts in 1217 and 1787. It is predicted to explode again in 2025 or 2026 based on multiple observational studies. The system consists of a massive ( M wd ≳ 1.35 M ⊙ ) white dwarf (WD) and a red giant (M3–M4 III). We have performed 1D hydrodynamic simulations with NOVA to predict the behavior of the next outburst. These simulations consist of a range of mass accretion rates onto ∼1.35 M ⊙ WDs, designed to bound the conditions necessary to achieve ignition of an explosion after an ≈80 yr interoutburst period. We have used both carbon–oxygen and oxygen–neon initial compositions, in order to include the possible ejecta abundances to be measured in the observations of the next outburst. As the WD in the TCrB system is observed to be massive, theoretical predictions reported here imply that the WD is growing in mass as a consequence of the thermonuclear runaway. Therefore, the secular evolution of the WD may allow it to approach the Chandrasekhar limit and either explode as a Type Ia supernova or undergo accretion-induced collapse, depending on its underlying composition. We have followed the evolution of just the WD, after removing the ejected matter from the surface layers. Our intent is to illuminate the mystery of the unique, second maximum in the two well-observed outbursts and we have found conditions that bracket the predictions.

Cataclysmic variable stars↗

Explaining and predicting human behavior and social dynamics in simulated virtual worlds: reproducibility, generalizability, and robustness of causal discovery methods

Ground Truth program was designed to evaluate social science modeling approaches using simulation test beds with ground truth intentionally and systematically embedded to understand and model complex Human Domain systems and their dynamics Lazer et al. (Science 369:1060–1062, 2020). Our multidisciplinary team of data scientists, statisticians, experts in Artificial Intelligence (AI) and visual analytics had a unique role on the program to investigate accuracy, reproducibility, generalizability, and robustness of the state-of-the-art (SOTA) causal structure learning approaches applied to fully observed and sampled simulated data across virtual worlds. In addition, we analyzed the feasibility of using machine learning models to predict future social behavior with and without causal knowledge explicitly embedded. In this paper, we first present our causal modeling approach to discover the causal structure of four virtual worlds produced by the simulation teams—Urban Life, Financial Governance, Disaster and Geopolitical Conflict. Our approach adapts the state-of-the-art causal discovery (including ensemble models), machine learning, data analytics, and visualization techniques to allow a human-machine team to reverse-engineer the true causal relations from sampled and fully observed data. We next present our reproducibility analysis of two research methods team’s performance using a range of causal discovery models applied to both sampled and fully observed data, and analyze their effectiveness and limitations. We further investigate the generalizability and robustness to sampling of the SOTA causal discovery approaches on additional simulated datasets with known ground truth. Our results reveal the limitations of existing causal modeling approaches when applied to large-scale, noisy, high-dimensional data with unobserved variables and unknown relationships between them. We show that the SOTA causal models explored in our experiments are not designed to take advantage from vasts amounts of data and have difficulty recovering ground truth when latent confounders are present; they do not generalize well across simulation scenarios and are not robust to sampling; they are vulnerable to data and modeling assumptions, and therefore, the results are hard to reproduce. Finally, when we outline lessons learned and provide recommendations to improve models for causal discovery and prediction of human social behavior from observational data, we highlight the importance of learning data to knowledge representations or transformations to improve causal discovery and describe the benefit of causal feature selection for predictive and prescriptive modeling.

97 MATHEMATICS AND COMPUTING↗

Clinical Factors Predicting Late Severe Urinary Toxicity After Postoperative Radiotherapy for Prostate Carcinoma: A Single-Institute Analysis of 742 Patients

Purpose: To investigate the clinical factors independently predictive of long-term severe urinary sequelae after postprostatectomy radiotherapy. Patients and Methods: Between 1993 and 2005, 742 consecutive patients underwent postoperative radiotherapy with either adjuvant (n = 556; median radiation dose, 70.2 Gy) or salvage (n = 186; median radiation dose, 72 Gy) intent. Results: After a median follow-up of 99 months, the 8-year risk of Grade 2 or greater and Grade 3 late urinary toxicity was almost identical (23.9% vs. 23.7% and 12% vs. 10%) in the adjuvant and salvage cohorts, respectively. On univariate analysis, acute toxicity was significantly predictive of late Grade 2 or greater sequelae in both subgroups (p <.0001 in both cases), and hypertension (p = .02) and whole-pelvis radiotherapy (p = .02) correlated significantly in the adjuvant cohort only. The variables predictive of late Grade 3 sequelae were acute Grade 2 or greater toxicity in both groups and whole-pelvis radiotherapy (8-year risk of Grade 3 events, 21% vs. 11%, p = .007), hypertension (8-year risk, 18% vs. 10%, p = .005), age {<=} 62 years at RT (8-year risk, 16% vs. 11%, p = .04) in the adjuvant subset, and radiation dose >72 Gy (8-year risk, 19% vs. 6%, p = .007) and age >71 years (8-year risk, 16% vs. 6%, p = .006) in the salvage subgroup. Multivariate analysis confirmed the independent predictive role of all the covariates indicated as statistically significant on univariate analysis. Conclusions: The risk of late Grade 2 or greater and Grade 3 urinary toxicity was almost identical, regardless of the RT intent. In the salvage cohort, older age and greater radiation doses resulted in a worse toxicity profile, and younger, hypertensive patients experienced a greater rate of severe late sequelae in the adjuvant setting. The causes of this latter correlation and apparently different etiopathogenesis of chronic damage in the two subgroups were unclear and deserve additional investigation.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

A force-level theory of the rheology of entangled rod and chain polymer liquids. I. Tube deformation, microscopic yielding, and the nonlinear elastic limit

In this paper, we employ a first-principles-based, force-level approach to construct the anharmonic tube confinement field for entangled fluids of rigid needles, and also for chains described at the primitive-path (PP) level in two limiting situations where chain stretch is assumed to either be completely equilibrated or unrelaxed. The influence of shear and extensional deformation and polymer orientation is determined in a nonlinear elastic limit where dissipative relaxation processes are intentionally neglected. For needles and PP-level chains, a self-consistent analysis of transverse polymer harmonic dynamical fluctuations predicts that deformation-induced orientation leads to tube weakening or widening. In contrast, for deformed polymers in which chain stretch does not relax, we find tube strengthening or compression. For all three systems, a finite maximum transverse entanglement force localizing the polymers in effective tubes is predicted. The conditions when this entanglement force can be overcome by an externally applied force associated with macroscopic deformation can be crisply defined in the nonlinear elastic limit, and the possibility of a “microscopic absolute yielding” event destroying the tube confinement can be analyzed. For needles and contour-relaxed PP chains, this force imbalance occurs at a stress of order the equilibrium shear modulus and a strain of order unity, corresponding to a mechanically fragile entanglement tube field. However, for unrelaxed stretched chains, tube compression stabilizes transverse polymer confinement, and there appears to be no force imbalance. These results collectively suggest that the crossover from elastic to irreversible viscous response requires chain retraction to initiate disentanglement. Finally, we qualitatively discuss comparisons with existing phenomenological models for nonlinear startup shear, step strain, and creep rheology experiments.

74 ATOMIC AND MOLECULAR PHYSICS↗

Far-field Wireless Energy Harvesting for Increased Safeguards Equipment Battery Life.

Modern unattended safeguards equipment (e.g. seals) incorporates many low-power electronic circuits, which are typically powered by expensive and toxic lithium thionyl chloride (LiSOCL2) batteries. The limited life of these batteries necessitates their periodic replacement. This replacement must be performed before total battery discharge to avoid potential loss of continuity of knowledge. Thus, the effective battery capacity becomes significantly less than the actual usable capacity. Additionally, such maintenance is a radiological hazard to personnel, as well as a monetary burden to a safeguards inspectorate. Energy harvesting, a commercially available technology, could extend the operational life of batterypowered equipment to achieve significant efficiencies for safeguards deployments. Energy harvesting is the scavenging and storage of ambient energy sources, such as solar, thermal, and kinetic for use in lowpower electronic applications. While the amount of scavenged energy per unit time may be small, it most often comes from a source that will not be depleted throughout the deployment of the harvesting device. The best-known energy harvesters are solar panels and wind turbines. Recently, far-field wireless energy harvesting has become a commercially available option. Far-field wireless energy harvesting provides consistent, predictable, and un-tethered power over distances up to 50 feet. This process converts radio frequency (RF) energy, both intentionally emitted and ambient, into usable direct current (DC) power. Incorporating far-field wireless energy harvesting into safeguards equipment can significantly extend the equipment’s battery life and perhaps make it indefinite. Furthermore, additional functionality can be added to safeguards equipment without lowering its operational life expectancy. This paper explores the benefits and drawbacks of integrating far-field wireless energy harvesting into a chosen safeguards seal: the Remotely Monitored Sealing Array (RMSA). Specifically, it examines the performance of a commercially available RF harvesting system from Powercast, as well as commercial and custom antenna solutions.

25 ENERGY STORAGE↗

Stockpile Stewardship at Los Alamos(U)

Stockpile stewardship is the retention of nuclear weapons in the stockpile beyond their original design life. These older weapons have potential changes inconsistent with the original design intent and military specifications. The Stockpile Stewardship Program requires us to develop high-fidelity, physics-based capabilities to predict, assess, certify and design nuclear weapons without conducting a nuclear test. Each year, the Lab Directors are required to provide an assessment of the safety, security, and reliability our stockpile to the President of the United States. This includes assessing whether a need to return to testing exists. This is a talk to provide an overview of Stockpile Stewardship's scientific requirements and how stewardship has changed in the absence of nuclear testing. The talk is adapted from an HQ talk to the War college, and historical unclassified talks on weapon's physics.

45 MILITARY TECHNOLOGY, WEAPONRY, AND NATIONAL DEF↗

Light Water Reactor Sustainability Program Materials Research Pathway Technical Program Plan

The Materials Research (MR) Pathway within the Light Water Reactor Sustainability (LWRS) Program is charged with performing the research and development (R&D) to develop the scientific basis for understanding and predicting long-term environmental degradation behavior of materials in nuclear reactors. Furthermore, it is essential to use the mechanistic understanding of degradation phenomena in materials to develop mitigation, repair, and new material alternatives for existing components. The work will provide data and methods to assess performance of systems, structures, and components (SSCs) essential to safe and sustained reactor operations. The R&D products developed from the LWRS Program will be used by utilities, industry groups, and regulators to inform operational and regulatory requirements for materials in reactor SSCs subjected to long-term operation conditions, providing key input to both regulators and industry. The intent of this research is to help in reducing the operating costs, which may be in the form of offset maintenance costs due to better predictive models for component lifetimes, improved analysis of materials through nondestructive evaluation, reduced costs for repairs, or extended performance of plants through the selection of improved replacement materials. To best achieve this, industry experience and guidance are important as is their role in coordinated or collaborative research projects. The objectives of this report are to describe the motivation and organization of the MR Pathway within the LWRS Program, provide details on the individual research tasks within the MR Pathway, describe the outcomes and deliverables of the MR Pathway, including recent technical highlights and progress, and list the requirements for performing the research.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗

Validating a Model for Welding Induced Residual Stress Using High-Energy X-ray Diffraction

Integrated computational materials engineering (ICME) provides a pathway to advance performance in structures through the use of physically-based models to better understand how manufacturing processes influence product performance. As one particular challenge, consider that residual stresses induced in fabrication are pervasive and directly impact the life of structures. For ICME to be an effective strategy, it is essential that predictive capability be developed in conjunction with critical experiments. In the present paper, simulation results from a multi-physics model for gas metal arc welding are evaluated through x-ray diffraction using synchrotron radiation. A test component was designed with intent to develop significant gradients in residual stress, be representative of real-world engineering application, yet remain tractable for finely spaced strain measurements with positioning equipment available at synchrotron facilities. Finally, the experimental validation lends confidence to model predictions, facilitating the explicit consideration of residual stress distribution in prediction of fatigue life.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Less is more: Sampling chemical space with active learning

We present the development of accurate and transferable machine learning (ML) potentials for predicting molecular energetics is a challenging task. The process of data generation to train such ML potentials is a task neither well understood nor researched in detail. In this work, we present a fully automated approach for the generation of datasets with the intent of training universal ML potentials. It is based on the concept of active learning (AL) via Query by Committee (QBC), which uses the disagreement between an ensemble of ML potentials to infer the reliability of the ensemble’s prediction. QBC allows the presented AL algorithm to automatically sample regions of chemical space where the ML potential fails to accurately predict the potential energy. AL improves the overall fitness of ANAKIN-ME (ANI) deep learning potentials in rigorous test cases by mitigating human biases in deciding what new training data to use. AL also reduces the training set size to a fraction of the data required when using naive random sampling techniques. To provide validation of our AL approach, we develop the COmprehensive Machine-learning Potential (COMP6) benchmark (publicly available on GitHub) which contains a diverse set of organic molecules. Active learning-based ANI potentials outperform the original random sampled ANI-1 potential with only 10% of the data, while the final active learning-based model vastly outperforms ANI-1 on the COMP6 benchmark after training to only 25% of the data. Finally, we show that our proposed AL technique develops a universal ANI potential (ANI-1x) that provides accurate energy and force predictions on the entire COMP6 benchmark. Finally, this universal ML potential achieves a level of accuracy on par with the best ML potentials for single molecules or materials, while remaining applicable to the general class of organic molecules composed of the elements CHNO.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Formation Mechanisms and Defect Engineering of Imine-Based Porous Organic Cages

The syntheses of porous organic cages (POCs) represent an important synthetic puzzle in dynamic covalent chemistry based self-sorting. Improved understanding of the formation mechanisms of POCs can lead to control and rational design of cages with desired functionality. Herein, we explore the formation mechanisms of imine-based POCs using time-resolved electrospray mass spectrometry, and electronic structure calculations at the density functional theory and correlated molecular orbital theory levels. We find that the synthesis of the [4+6] cycloimine cage CC3-R and the [2+3] cycloimine cage CC-pentane both proceed through similar intermediates via a series of consecutive reactions. The proposed reaction mechanisms are supported by electronic structure calculations. Based on our observations from both experiments and calculations, we propose a comprehensive method for designing and predicting new POC species. In addition, the observation of stable incomplete cages during CC3-R synthesis inspired us to design intentionally defective cages. These “missing-linker” type molecular defects are installed into CC3-R via non-solvent induced crystallization. The defective CC3-R materials are found to have enhanced CO2 interaction and improved CO 2 uptake capacity due to the additional functional groups present within the CC3 crystals.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗