-
Differentiable Modeling and Optimization of Battery Electrolyte Mixtures Using Geometric Deep Learning
Authors:
Shang Zhu,
Bharath Ramsundar,
Emil Annevelink,
Hongyi Lin,
Adarsh Dave,
Pin-Wen Guan,
Kevin Gering,
Venkatasubramanian Viswanathan
Abstract:
Electrolytes play a critical role in designing next-generation battery systems, by allowing efficient ion transfer, preventing charge transfer, and stabilizing electrode-electrolyte interfaces. In this work, we develop a differentiable geometric deep learning (GDL) model for chemical mixtures, DiffMix, which is applied in guiding robotic experimentation and optimization towards fast-charging batte…
▽ More
Electrolytes play a critical role in designing next-generation battery systems, by allowing efficient ion transfer, preventing charge transfer, and stabilizing electrode-electrolyte interfaces. In this work, we develop a differentiable geometric deep learning (GDL) model for chemical mixtures, DiffMix, which is applied in guiding robotic experimentation and optimization towards fast-charging battery electrolytes. In particular, we extend mixture thermodynamic and transport laws by creating GDL-learnable physical coefficients. We evaluate our model with mixture thermodynamics and ion transport properties, where we show improved prediction accuracy and model robustness of DiffMix than its purely data-driven variants. Furthermore, with a robotic experimentation setup, Clio, we improve ionic conductivity of electrolytes by over 18.8% within 10 experimental steps, via differentiable optimization built on DiffMix gradients. By combining GDL, mixture physics laws, and robotic experimentation, DiffMix expands the predictive modeling methods for chemical mixtures and enables efficient optimization in large chemical spaces.
△ Less
Submitted 1 November, 2023; v1 submitted 3 October, 2023;
originally announced October 2023.
-
Physics-informed State-space Neural Networks for Transport Phenomena
Authors:
Akshay J. Dave,
Richard B. Vilim
Abstract:
This work introduces Physics-informed State-space neural network Models (PSMs), a novel solution to achieving real-time optimization, flexibility, and fault tolerance in autonomous systems, particularly in transport-dominated systems such as chemical, biomedical, and power plants. Traditional data-driven methods fall short due to a lack of physical constraints like mass conservation; PSMs address…
▽ More
This work introduces Physics-informed State-space neural network Models (PSMs), a novel solution to achieving real-time optimization, flexibility, and fault tolerance in autonomous systems, particularly in transport-dominated systems such as chemical, biomedical, and power plants. Traditional data-driven methods fall short due to a lack of physical constraints like mass conservation; PSMs address this issue by training deep neural networks with sensor data and physics-informing using components' Partial Differential Equations (PDEs), resulting in a physics-constrained, end-to-end differentiable forward dynamics model. Through two in silico experiments -- a heated channel and a cooling system loop -- we demonstrate that PSMs offer a more accurate approach than a purely data-driven model. In the former experiment, PSMs demonstrated significantly lower average root-mean-square errors across test datasets compared to a purely data-driven neural network, with reductions of 44 %, 48 %, and 94 % in predicting pressure, velocity, and temperature, respectively.
Beyond accuracy, PSMs demonstrate a compelling multitask capability, making them highly versatile. In this work, we showcase two: supervisory control of a nonlinear system through a sequentially updated state-space representation and the proposal of a diagnostic algorithm using residuals from each of the PDEs. The former demonstrates PSMs' ability to handle constant and time-dependent constraints, while the latter illustrates their value in system diagnostics and fault detection. We further posit that PSMs could serve as a foundation for Digital Twins, constantly updated digital representations of physical systems.
△ Less
Submitted 18 December, 2023; v1 submitted 21 September, 2023;
originally announced September 2023.
-
Dielectric breakdown and sub-wavelength patterning of monolayer hexagonal boron nitride using femtosecond pulses
Authors:
Sabeeh Irfan Ahmad,
Emmanuel Sarpong,
Arpit Dave,
Hsin-Yu Yao,
Joel M. Solomon,
**g-Kai Jiang,
Chih-Wei Luo,
Wen-Hao Chang,
Tsing-Hua Her
Abstract:
Hexagonal boron nitride (hBN) has emerged as a promising two-dimensional (2D) material for many applications in photonics. Although its linear and nonlinear optical properties have been extensively studied, its interaction with high-intensity laser pulses, which is important for high-harmonic generation, fabricating quantum emitters, and maskless patterning of hBN, has not been investigated. Here…
▽ More
Hexagonal boron nitride (hBN) has emerged as a promising two-dimensional (2D) material for many applications in photonics. Although its linear and nonlinear optical properties have been extensively studied, its interaction with high-intensity laser pulses, which is important for high-harmonic generation, fabricating quantum emitters, and maskless patterning of hBN, has not been investigated. Here we report the first study of dielectric breakdown in hBN monolayers induced by single femtosecond laser pulses. We show that hBN has the highest breakdown threshold among all existing 2D materials. This enables us to observe clearly for the first time a linear dependence of breakdown threshold on the bandgap energy for 2D materials, demonstrating such a linear dependency is a universal scaling law independent of the dimensionality. We also observe counter-intuitively that hBN, which has a larger bandgap and mechanical strength than quartz, has a lower breakdown threshold. This implies carrier generation in hBN is much more efficient. Furthermore, we demonstrate the clean removal of hBN without damage to the surrounding hBN film or the substrate, indicating that hBN is optically very robust. The ablated features are shown to possess very small edge roughness, which is attributed to its ultrahigh fracture toughness. Finally, we demonstrate femtosecond laser patterning of hBN with sub-wavelength resolution, including an isolated stripe width of 200 nm. Our work advances the knowledge of light-hBN interaction in the strong field regime and firmly establishes femtosecond lasers as novel and promising tools for one-step deterministic patterning of hBN monolayers.
△ Less
Submitted 7 June, 2023;
originally announced June 2023.
-
Ultrafast Multi-Shot Ablation and Defect Generation in Monolayer Transition Metal Dichalcogenides
Authors:
Joel M. Solomon,
Sabeeh Irfan Ahmad,
Arpit Dave,
Li-Syuan Lu,
Yu-Chen Wu,
Wen-Hao Chang,
Chih-Wei Luo,
Tsing-Hua Her
Abstract:
Transition metal dichalcogenides are known to possess large optical nonlinearities and driving these materials at high intensities is desirable for many applications. Understanding their optical responses under repetitive intense excitation is essential to improve the performance limit of these strong-field devices and to achieve efficient laser patterning. Here, we report the incubation study of…
▽ More
Transition metal dichalcogenides are known to possess large optical nonlinearities and driving these materials at high intensities is desirable for many applications. Understanding their optical responses under repetitive intense excitation is essential to improve the performance limit of these strong-field devices and to achieve efficient laser patterning. Here, we report the incubation study of monolayer MoS${}_{2}$ and WS${}_{2}$ induced by 160 fs, 800 nm pulses in air to examine how their ablation threshold scales with the number of admitted laser pulses. Both materials were shown to outperform graphene and most bulk materials; specifically, MoS${}_{2}$ is as resistant to radiation degradation as the best of the bulk thin films with a record fast saturation. Our modeling provides convincing evidence that the small reduction in threshold and fast saturation of MoS${}_{2}$ originates in its excellent bonding integrity against radiation-induced softening. Sub-ablation damages, in the forms of vacancies, lattice disorder, and nano-voids, were revealed by transmission electron microscopy, photoluminescence, Raman, and second harmonic generation studies, which were attributed to the observed incubation. For the first time, a sub-ablation damage threshold is identified for monolayer MoS${}_{2}$ to be 78% of single-shot ablation threshold, below which MoS${}_{2}$ remains intact for many laser pulses. Our results firmly establish MoS${}_{2}$ as a robust material for strong-field devices and for high-throughput laser patterning.
△ Less
Submitted 20 December, 2021;
originally announced December 2021.
-
Ultrafast Laser Ablation, Intrinsic Threshold, and Nanopatterning of Monolayer Molybdenum Disulfide
Authors:
Joel M. Solomon,
Sabeeh Irfan Ahmad,
Arpit Dave,
Li-Syuan Lu,
Fatemeh HadavandMirzaee,
Shih-Chu Lin,
Sih-Hua Chen,
Chih-Wei Luo,
Wen-Hao Chang,
Tsing-Hua Her
Abstract:
Laser direct writing is an attractive method for patterning 2D materials without contamination. Literature shows that the femtosecond ablation threshold of graphene across substrates varies by an order of magnitude. Some attribute it to the thermal coupling to the substrates, but it remains by and large an open question. For the first time the effect of substrates on femtosecond ablation of 2D mat…
▽ More
Laser direct writing is an attractive method for patterning 2D materials without contamination. Literature shows that the femtosecond ablation threshold of graphene across substrates varies by an order of magnitude. Some attribute it to the thermal coupling to the substrates, but it remains by and large an open question. For the first time the effect of substrates on femtosecond ablation of 2D materials is studied using MoS$_{2}$ as an example. We show unambiguously that femtosecond ablation of MoS$_{2}$ is an adiabatic process with negligible heat transfer to the substrates. The observed threshold variation is due to the etalon effect which was not identified before for the laser ablation of 2D materials. Subsequently, an intrinsic ablation threshold is proposed as a true threshold parameter for 2D materials. Additionally, we demonstrate for the first time femtosecond laser patterning of monolayer MoS$_{2}$ with sub-micron resolution and mm/s speed. Moreover, engineered substrates are shown to enhance the ablation efficiency, enabling patterning with low-power femtosecond oscillators. Finally, a zero-thickness approximation is introduced to predict the field enhancement with simple analytical expressions. Our work clarifies the role of substrates on ablation and firmly establishes femtosecond laser ablation as a viable route to pattern 2D materials.
△ Less
Submitted 1 November, 2021;
originally announced November 2021.
-
Empirical Models for Multidimensional Regression of Fission Systems
Authors:
Akshay J. Dave,
Jiankai Yu,
Jarod Wilson,
Bren Phillips,
Kaichao Sun,
Benoit Forget
Abstract:
The development of next-generation autonomous control of fission systems, such as nuclear power plants, will require leveraging advancements in machine learning. For fission systems, accurate prediction of nuclear transport is important to quantify the safety margin and optimize performance. The state-of-the-art approach to this problem is costly Monte Carlo (MC) simulations to approximate solutio…
▽ More
The development of next-generation autonomous control of fission systems, such as nuclear power plants, will require leveraging advancements in machine learning. For fission systems, accurate prediction of nuclear transport is important to quantify the safety margin and optimize performance. The state-of-the-art approach to this problem is costly Monte Carlo (MC) simulations to approximate solutions of the neutron transport equation. Such an approach is feasible for offline calculations e.g., for design or licensing, but is precluded from use as a model-based controller. In this work, we explore the use of Artificial Neural Networks (ANN), Gradient Boosting Regression (GBR), Gaussian Process Regression (GPR) and Support Vector Regression (SVR) to generate empirical models. The empirical model can then be deployed, e.g., in a model predictive controller. Two fission systems are explored: the subcritical MIT Graphite Exponential Pile (MGEP), and the critical MIT Research Reactor (MITR).
Findings from this work establish guidelines for develo** empirical models for multidimensional regression of neutron transport. An assessment of the accuracy and precision finds that the SVR, followed closely by ANN, performs the best. For both MGEP and MITR, the optimized SVR model exhibited a domain-averaged, test, mean absolute percentage error of 0.17 %. A spatial distribution of performance metrics indicates that physical regions of poor performance coincide with locations of largest neutron flux perturbation -- this outcome is mitigated by ANN and SVR. Even at local maxima, ANN and SVR bias is within experimental uncertainty bounds. A comparison of the performance vs. training dataset size found that SVR is more data-efficient than ANN. Both ANN and SVR achieve a greater than 7 order reduction in evaluation time vs. a MC simulation.
△ Less
Submitted 30 May, 2021;
originally announced May 2021.
-
Deep Surrogate Models for Multi-dimensional Regression of Reactor Power
Authors:
Akshay J. Dave,
Jarod Wilson,
Kaichao Sun
Abstract:
There is renewed interest in develo** small modular reactors and micro-reactors. Innovation is necessary in both construction and operation methods of these reactors to be financially attractive. For operation, an area of interest is the development of fully autonomous reactor control. Significant efforts are necessary to demonstrate an autonomous control framework for a nuclear system, while ad…
▽ More
There is renewed interest in develo** small modular reactors and micro-reactors. Innovation is necessary in both construction and operation methods of these reactors to be financially attractive. For operation, an area of interest is the development of fully autonomous reactor control. Significant efforts are necessary to demonstrate an autonomous control framework for a nuclear system, while adhering to established safety criteria. Our group has proposed and received support for demonstration of an autonomous framework on a subcritical system: the MIT Graphite Exponential Pile. In order to have a fast response (on the order of miliseconds), we must extract specific capabilities of general-purpose system codes to a surrogate model. Thus, we have adopted current state-of-the-art neural network libraries to build surrogate models.
This work focuses on establishing the capability of neural networks to provide an accurate and precise multi-dimensional regression of a nuclear reactor's power distribution. We assess using a neural network surrogate against a previously validated model: an MCNP5 model of the MIT reactor. The results indicate that neural networks are an appropriate choice for surrogate models to implement in an autonomous reactor control framework. The MAPE across all test datasets was < 1.16 % with a corresponding standard deviation of < 0.77 %. The error is low, considering that the node-wise fission power can vary from 7 kW to 30 kW across the core.
△ Less
Submitted 13 July, 2020; v1 submitted 10 July, 2020;
originally announced July 2020.
-
Inference of Gas-liquid Flowrate using Neural Networks
Authors:
Akshay J. Dave,
Annalisa Manera
Abstract:
The metering of gas-liquid flows is difficult due to the non-linear relationship between flow regimes and fluid properties, flow orientation, channel geometry, etc. In fact, a majority of commercial multiphase flow meters have a low accuracy, limited range of operation or require a physical separation of the phases. We introduce the inference of gas-liquid flowrates using a neural network model th…
▽ More
The metering of gas-liquid flows is difficult due to the non-linear relationship between flow regimes and fluid properties, flow orientation, channel geometry, etc. In fact, a majority of commercial multiphase flow meters have a low accuracy, limited range of operation or require a physical separation of the phases. We introduce the inference of gas-liquid flowrates using a neural network model that is trained by wire-mesh sensor (WMS) experimental data. The WMS is an experimental tool that records high-resolution high-frequency 3D void fraction distributions in gas-liquid flows. The experimental database utilized spans over two orders of superficial velocity magnitude and multiple flow regimes for a vertical small-diameter pipe. Our findings indicate that a single network can provide accurate and precise inference with below a 7.5% MAP error across all flow regimes. The best performing networks have a combination of a 3D-Convolution head, and an LSTM tail. The finding indicates that the spatiotemporal features observed in gas-liquid flows can be systematically decomposed and used for inferring phase-wise flowrate. Our method does not involve any complex pre-processing of the void fraction matrices, resulting in an evaluation time that is negligible when contrasted to the input time-span. The efficiency of the model manifests in a response time two orders of magnitude lower than the current state-of-the-art.
△ Less
Submitted 25 May, 2020; v1 submitted 15 March, 2020;
originally announced March 2020.
-
Autonomous discovery of battery electrolytes with robotic experimentation and machine-learning
Authors:
Adarsh Dave,
Jared Mitchell,
Kirthevasan Kandasamy,
Sven Burke,
Biswajit Paria,
Barnabas Poczos,
Jay Whitacre,
Venkatasubramanian Viswanathan
Abstract:
Innovations in batteries take years to formulate and commercialize, requiring extensive experimentation during the design and optimization phases. We approached the design and selection of a battery electrolyte through a black-box optimization algorithm directly integrated into a robotic test-stand. We report here the discovery of a novel battery electrolyte by this experiment completely guided by…
▽ More
Innovations in batteries take years to formulate and commercialize, requiring extensive experimentation during the design and optimization phases. We approached the design and selection of a battery electrolyte through a black-box optimization algorithm directly integrated into a robotic test-stand. We report here the discovery of a novel battery electrolyte by this experiment completely guided by the machine-learning software without human intervention. Motivated by the recent trend toward super-concentrated aqueous electrolytes for high-performance batteries, we utilize Dragonfly - a Bayesian machine-learning software package - to search mixtures of commonly used lithium and sodium salts for super-concentrated aqueous electrolytes with wide electrochemical stability windows. Dragonfly autonomously managed the robotic test-stand, recommending electrolyte designs to test and receiving experimental feedback in real time. In 40 hours of continuous experimentation over a four-dimensional design space with millions of potential candidates, Dragonfly discovered a novel, mixed-anion aqueous sodium electrolyte with a wider electrochemical stability window than state-of-the-art sodium electrolyte. A human-guided design process may have missed this optimal electrolyte. This result demonstrates the possibility of integrating robotics with machine-learning to rapidly and autonomously discover novel battery materials.
△ Less
Submitted 22 October, 2019;
originally announced January 2020.
-
Benchmarking conductivity predictions of the Advanced Electrolyte Model (AEM) for aqueous systems
Authors:
Adarsh Dave,
Kevin L. Gering,
Jared M. Mitchell,
Jay Whitacre,
Venkatasubramanian Viswanathan
Abstract:
High-concentration aqueous electrolytes have shown promise as candidates for a safer, lower-cost battery system. Ionic conductivity is a key property required in high performing electrolytes; the Advanced Electrolyte Model (AEM) has previously shown great accuracy in predicting ionic conductivity in highly-concentrated non-aqueous electrolytes. This work provides extensive experimental data for mi…
▽ More
High-concentration aqueous electrolytes have shown promise as candidates for a safer, lower-cost battery system. Ionic conductivity is a key property required in high performing electrolytes; the Advanced Electrolyte Model (AEM) has previously shown great accuracy in predicting ionic conductivity in highly-concentrated non-aqueous electrolytes. This work provides extensive experimental data for mixed and highly concentrated aqueous electrolyte systems, rapidly generated via a robotic electrolyte testing apparatus. These data demonstrate exceptional accuracy from AEM in predicting conductivity in aqueous systems, with the accuracy being maintained even in highly-concentrated and mixed-salt regimes.
△ Less
Submitted 14 June, 2019;
originally announced June 2019.
-
IITMSAT Communications System : A LeanSat Design Approach
Authors:
Akshay Gulati,
Shubham Chavan,
Joseph Samuel,
Sampoornam Srinivasan,
Pradeep Shekhar,
Akshat Dave,
Aditya Sant,
Sourbh Bhadane,
Mayug Maniparambil,
Vishnu Prasad Sivasankarakurup,
Dhanalakshmi Durairaj,
David Koilpillai,
Harishankar Ramachandran
Abstract:
IITMSAT is a student-built nano satellite mission of Indian Institute of Technology Madras, Chennai, India. The objective is to study the precipitation of high energy electrons and protons from Van-Allen radiation belts to lower altitude of 600-900 km due to resonance interaction with low frequency EM waves. The unique communications system design of IITMSAT evolves from the challenging downlink d…
▽ More
IITMSAT is a student-built nano satellite mission of Indian Institute of Technology Madras, Chennai, India. The objective is to study the precipitation of high energy electrons and protons from Van-Allen radiation belts to lower altitude of 600-900 km due to resonance interaction with low frequency EM waves. The unique communications system design of IITMSAT evolves from the challenging downlink data requirement of 1 MB per day in the UHF band posed by the mission and the satellite's payload, SPEED (Space based Proton and Electron Energy Detector). To ensure continuous downlink data stream in the short Low earth Orbit passes, a robust physical layer protocol was designed to counter time-varying aspects of a Space-Earth telecom link. For the on-board communications system, two types of design alternatives exist for each module. The first option is a custom design wherein a module is developed from scratch using discrete components.The other option is an integrated design wherein an electronics COTS module can be directly plugged into the subsystem. This module is evaluated by carrying out vibration and thermal tests. If an integrated module is low-cost and meets the design requirements, it is preferred over a custom design. In order to carry out performance tests under simulated link conditions, an RF attenuation test setup was designed that can work at extreme temperatures. Burn-In tests for 72 hours at ambient and extreme temperatures were carried out. Integrated tests indicate all IITMSAT design requirements have been met. Hence a robust communications system has been validated. The time taken for development of on-board telecom and GS was less than a year and was achieved at a low cost which agrees to a LeanSat approach.
△ Less
Submitted 3 November, 2017;
originally announced November 2017.
