Deep neural network learning of complex binary sorption equilibria from molecular simulation data

Sun, Yangzesheng; DeJaco, Robert F.; Siepmann, J. Ilja

doi:10.1039/c8sc05340e

Cited by 47 publications

(45 citation statements)

References 75 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…A few examples of the use of machine learning in the field of classical simulations for materials science are the deep neural network learning of complex binary sorption equilibria from molecular simulation data, solving vapor–liquid flash problems using artificial neural networks, predicting thermodynamic properties of alkanes, charge assignment, prediction of partition functions, simulation of infrared spectra, predicting the mechanical properties of zeolite frameworks, CO 2 capture using MOFs, prediction of methane adsorption performance of MOFs, chemically intuited, large‐scale screening of MOFs, screening for precombustion carbon capture using MOFs, screening of MOF Membranes for the separation of gas mixtures, and screening of MOFs for use as electronic devices . Using ML to combine the accuracy and flexibility of electronic structure calculations with the speed of classical potentials is a very active research field .…”

Section: Parameterizationmentioning

confidence: 99%

Design, Parameterization, and Implementation of Atomic Force Fields for Adsorption in Nanoporous Materials

Dubbeldam

Walton

Vlugt

et al. 2019

Advcd Theory and Sims

View full text Add to dashboard Cite

Molecular simulations are an excellent tool to study adsorption and diffusion in nanoporous materials. Examples of nanoporous materials are zeolites, carbon nanotubes, clays, metal-organic frameworks (MOFs), covalent organic frameworks (COFs) and zeolitic imidazolate frameworks (ZIFs). The molecular confinement these materials offer has been exploited in adsorption and catalysis for almost 50 years. Molecular simulations have provided understanding of the underlying shape selectivity, and adsorption and diffusion effects. Much of the reliability of the modeling predictions depends on the accuracy and transferability of the force field. However, flexibility and the chemical and structural diversity of MOFs add significant challenges for engineering force fields that are able to reproduce experimentally observed structural and dynamic properties. Recent developments in design, parameterization, and implementation of force fields for MOFs and zeolites are reviewed.contain silicon and oxygen. They are readily available, very stable, and relatively cheap. The material should have the right combination of high adsorption selectivity, combined with adequate capacity for use in fixed-bed devices. Recently, new classes of nanoporous materials have been designed that have stability, high void volumes, and well defined tailorable cavities of uniform size. Example are metal-organic frameworks (MOFs), [1][2][3][4][5][6][7] covalent organic frameworks (COFs), [8,9] and zeolitic imidazolate frameworks (ZIFs). [10] These novel materials posses almost unlimited structural variety because of the many combinations of building blocks that can be imagined. The building blocks self-assembly during synthesis into crystalline materials that, after evacuation of the structure, can find applications in adsorption separations, air purification, gas storage, chemical sensing, and catalysis. [4,11] The judicious selection of building blocks allows the pore volume and functionality to be tailored in a rational manner. Because of the designprinciple underlying the synthesis, it is important to understand structure-properties relations, such as the mechanisms of gas separation as a function of shape and size of the pore system, in order to create "blueprints to MOF-design." Computer simulation is cost-effective, fast, and allows for rapid identification of important structural parameters affecting adsorption.Most of the published works on molecular simulation studying zeolites have used the Kiselev model. [12] In this model, zeolite atoms are fixed and interactions of guest atoms with silicon atoms is accounted for by an effective interaction only with the surrounding oxygen atoms. Interactions between sorbate molecules and the host are modeled by placing Lennard-Jones sites and partial charges on all framework atoms and sorbate molecules. The Kiselev model is attractive because of its simplicity and computational efficiency. This model has shown significant success, both in using the molecular dynamics method to compute, for example, the diffu...

show abstract

Section: Parameterizationmentioning

confidence: 99%

Design, Parameterization, and Implementation of Atomic Force Fields for Adsorption in Nanoporous Materials

Dubbeldam

Walton

Vlugt

et al. 2019

Advcd Theory and Sims

View full text Add to dashboard Cite

show abstract

“…Recently, Sun et al reported a multitask deep NN (SorbNet) for the prediction of binary adsorption isotherms on zeolites. 479 Their idea was to use a model architecture in which the two components have two independent branches in the neural network close to the output and share layers close to the inputs, which are the initial loading, the volume, and the temperature. They then used this model to optimize process conditions for desorptive drying, which highlights that such models can help avoid the need for iteratively running simulations for the optimization of process conditions (we discuss the connection between materials simulation and process engineering in more detail in the next section).…”

Section: Applications Of Supervised Machine Learningmentioning

confidence: 99%

Big-Data Science in Porous Materials: Materials Genomics and Machine Learning

et al. 2020

View full text Add to dashboard Cite

By combining metal nodes with organic linkers we can potentially synthesize millions of possible metal–organic frameworks (MOFs). The fact that we have so many materials opens many exciting avenues but also create new challenges. We simply have too many materials to be processed using conventional, brute force, methods. In this review, we show that having so many materials allows us to use big-data methods as a powerful technique to study these materials and to discover complex correlations. The first part of the review gives an introduction to the principles of big-data science. We show how to select appropriate training sets, survey approaches that are used to represent these materials in feature space, and review different learning architectures, as well as evaluation and interpretation strategies. In the second part, we review how the different approaches of machine learning have been applied to porous materials. In particular, we discuss applications in the field of gas storage and separation, the stability of these materials, their electronic properties, and their synthesis. Given the increasing interest of the scientific community in machine learning, we expect this list to rapidly expand in the coming years.

show abstract

“…where the initial condition can be obtained from the prediction of the pre-trained model at time step 20. In order to illustrate the effectiveness and efficiency of the transfer learning mode, two contrasting examples are introduced, following the work by Sun et al (2019): retrain the first three layers while the last four layers are randomly initialized, and retrain the network completely. Table 6 and Figure 11 show the prediction results of the transfer learning model and the two contrasting examples.…”

Section: Transfer Learning Based On Tgnnmentioning

confidence: 99%

“…Table 6 and Figure 11 show the prediction results of the transfer learning model and the two contrasting examples. Compared with contrasting example 1, whose network parameters of the last four layers are randomly initialized and fixed during the training process, transfer learning model has better accuracy, which may indicate that the pre-trained parameters of the last four layers learned some information of the system during the former training process, and the information is transferable when dealing with similar new systems (Sun et al, 2019). Compared with contrasting example 2, whose parameters are retrained totally, the transfer learning model only exhibits a slight advantage regarding accuracy, while efficiency is noticeably improved since fewer parameters need to be trained.…”

Section: Transfer Learning Based On Tgnnmentioning

confidence: 99%

Deep learning of subsurface flow via theory-guided neural network

Wang

Zhang

Chang

et al. 2020

Journal of Hydrology

198

View full text Add to dashboard Cite

Active researches are currently being performed to incorporate the wealth of scientific knowledge into data-driven approaches (e.g., neural networks) in order to improve the latter's effectiveness. In this study, the Theory-guided Neural Network (TgNN) is proposed for deep learning of subsurface flow. In the TgNN, as supervised learning, the neural network is trained with available observations or simulation data while being simultaneously guided by theory (e.g., governing equations, other physical constraints, engineering controls, and expert knowledge) of the underlying problem. The TgNN can achieve higher accuracy than the ordinary Artificial Neural Network (ANN) because the former provides physically feasible predictions and can be more readily generalized beyond the regimes covered with the training data. Furthermore, the TgNN model is proposed for subsurface flow with heterogeneous model parameters. Several numerical cases of two-dimensional transient saturated flow are introduced to test the performance of the TgNN. In the learning process, the loss function contains data mismatch, as well as PDE constraint, engineering control, and expert knowledge. After obtaining the parameters of the neural network by minimizing the loss function, a TgNN model is built that not only fits the data, but also adheres to physical/engineering constraints. Predicting the future response can be easily realized by the TgNN model. In addition, the TgNN model is tested in more complicated scenarios, such as prediction with changed boundary conditions, learning from noisy data or outliers, transfer learning, and engineering controls. Numerical results demonstrate that the TgNN model achieves much better predictability, reliability, and generalizability than ANN models due to the physical/engineering constraints in the former.

show abstract

Deep neural network learning of complex binary sorption equilibria from molecular simulation data

Abstract: We employed deep neural networks (NNs) as an efficient and intelligent surrogate of molecular simulations for complex sorption equilibria using probabilistic modeling.

Cited by 47 publications

References 75 publications

Design, Parameterization, and Implementation of Atomic Force Fields for Adsorption in Nanoporous Materials

Design, Parameterization, and Implementation of Atomic Force Fields for Adsorption in Nanoporous Materials

Big-Data Science in Porous Materials: Materials Genomics and Machine Learning

Deep learning of subsurface flow via theory-guided neural network

Contact Info

Product

Resources

About