Search for: All records

Modifying solution viscosity is a key functional application of polymers, yet the interplay of molecular chemistry, polymer architecture, and intermolecular interactions makes tailoring precise rheological responses challenging. We introduce a computational framework coupling topology-aware generative machine learning, Gaussian process modeling, and multiparticle collision dynamics to design polymers yielding prescribed shear-rate-dependent viscosity profiles. Targeting thirty rheological profiles of varying difficulty, Bayesian optimization identifies polymers that satisfy all low- and most medium-difficulty targets by modifying topology and solvophobicity, with other variables fixed. In these regimes, we find and explain design degeneracy, where distinct polymers produce near-identical rheological profiles. However, satisfying high-difficulty targets requires extrapolation beyond the initial constrained design space; this is rationally guided by physical scaling theories. This integrated framework establishes a data-driven yet mechanistic route to rational polymer design. 

Single-chain nanoparticles (SCNPs) are a class of materials formed by the intramolecular cross-linking and collapse of single polymer chains. Because their morphology dictates suitability for specific applications, such as nanoscale reactors and drug delivery vehicles, understanding how to control or tailor morphologies is of interest. Here, we investigate how the morphology of SCNPs depends on both precursor chain attributes, such as linker fraction and backbone stiffness, and an imposed shear flow. Using coarse-grained molecular dynamics simulations, we generate an ensemble of structures from 10,800 unique SCNPs, some formed under quiescent conditions and some in shear flow--the latter of which has not been previously studied. We then characterize morphologies by analysis of a three-dimensional embedding space obtained through unsupervised learning of the simulated structures. This reveals how SCNP morphology depends on dimensionless parameters, related to precursor-chain attributes and shear rate, and offers insight into their relative influence. Interestingly, we find that shear rate has comparable influence to the degree of polymerization and the blockiness of reactive sites. Furthermore, shear, which can be externally controlled independent of precursor chain synthesis, can have persistent effects on morphology, such as enhancing compaction of SCNPs based on stiff chains. This work provides guidelines for designing SCNPs with targeted characteristics based on five dimensionless variables and illustrates the utility of machine learning in analyzing SCNPs formed across a range of conditions. 

Dataset Description This dataset contains 6710 structural configurations and solvophobicity values for topologically and chemically diverse coarse-grained polymer chains. Additionally, 480 polymers include shear-rate dependent viscosity profiles at 2 wt% polymer concentration.The data is provided as serialized objects using the pickle Python module.All files were generated using Python version 3.10. Data There are three pickle files containing serialized Python objects. Key files include: data_aug10.pickle  Contains the coarse-grained polymer dataset with 6710 entries.  Each entry includes: Polymer graph Squared radius of gyration (at lambda = 0). Solvophobicity (lambda). Bead count (N). Chain virial number (Xi). topo_param_visc.pickle   Shear-rate-dependent viscosity profiles of 480 polymer systems. target_curves.pickle  Contains 30 target viscosity profiles used for active learning. Usage To load the dataset stored in data_aug10.pickle, use the following code: import pickle with open("data_aug10.pickle", "rb") as handle:    (        (x_train, y_train, c_train, l_train, graph_train),        (x_valid, y_valid, c_valid, l_valid, graph_valid),        (x_test, y_test, c_test, l_test, graph_test),        NAMES,        SCALER,        SCALER_y,        le    ) = pickle.load(handle) x: node features for each polymer graph y: labels (e.g., predicted properties) c: topological class indices l: topological descriptors graph: NetworkX graphs representing polymer topology NAMES: list of topological class names SCALER: fitted scaler for topological descriptors (l) SCALER_y: fitted scaler for property labels (y) le: label encoder for topological class indices   To load the dataset stored in topo_param_visc.pickle, use the following code: import pickle with open("poly_data_ml.pickle", "rb") as handle:    desc_all, ps_all, curve_all, shear_rate, graph_all = pickle.load(handle) desc_all: topological descriptors for each polymer graph ps_all: fitted Carreau–Yasuda model parameters curve_all: fitted viscosity curves shear_rate: shear rates corresponding to each viscosity curve graph_all: polymer graphs represented as NetworkX objects   First 30: seed dataset Next 150: 5 iterations (30 each) from class-balanced space-filling Following 150: space-filling without class balancing Final 150: active learning samples    To load the dataset stored in target_curves.pickle, use the following code: import pickle with open("target_curves.pickle", "rb") as handle:    data = pickle.load(f) curves = data['curves']params = data['params']shear_rate = data["xx"]   curves: target viscosity curves used as design objectives params: Carreau–Yasuda model parameters fitted to the target curves shear_rate: shear rate values associated with the target curves     Help, Suggestions, Corrections?If you need help, have suggestions, identify issues, or have corrections, please send your comments to Shengli Jiang at sj0161@princeton.edu GitHubAdditional data and code relevant for this study is additionally accessible at https://github.com/webbtheosim/cg-topo-solv 

This dataset compiles the results of our study of nuclear quantum effects (NQEs) and equilibrium isotope effects (EIEs) of 92 chemically diverse, organic molecular liquids. It contains the average macroscopic properties (density, molar volume, thermal expansion coefficient, isothermal compressibility, dielectric constant, heat of vaporization) and their associated standard errors computed with four independent classical and path-integral molecular dynamics (PIMD) simulations of each system and their deuterated counterparts. All simulations use the Topology Automated Force-Field Interactions (TAFFI) framework to describe the potential energy surface. In addition, the computed nuclear quantum effects and equilibrium isotope effects resulting from comparison of these simulations are included. 

Phase separation in multicomponent mixtures is of significant interest in both fundamental research and technology. Although the thermodynamic principles governing phase equilibria are straightforward, practical determination of equilibrium phases and constituent compositions for multicomponent systems is often laborious and computationally intensive. Here, we present a machine-learning workflow that simplifies and accelerates phase-coexistence calculations. We specifically analyze capabilities of neural networks to predict the number, composition, and relative abundance of equilibrium phases of systems described by Flory-Huggins theory. We find that incorporating physics-informed material constraints into the neural network architecture enhances the prediction of equilibrium compositions compared to standard neural networks with minor errors along the boundaries of the stable region. However, introducing additional physics-informed losses does not lead to significant further improvement. These errors can be virtually eliminated by using machine-learning predictions as a warm-start for a subsequent optimization routine. This work provides a promising pathway to efficiently characterize multicomponent phase coexistence. 

Active learning and design-build-test-learn strategies are increasingly employed to accelerate materials discovery and characterization. Many data-driven materials design campaigns target solutions within constrained domains such as synthesizability, stability, solubility, recyclability, and toxicity. Lack of knowledge about these constraints can hinder design efficiency by producing samples that fail to meet required thresholds. Acquiring this knowledge during the design campaign is inefficient, and effective classification of common materials constraints transcends specific design objectives. However, there is no consensus on the most data-efficient algorithm for classifying whether a material satisfies a constraint. To address this gap, we comprehensively compare the performance of 100 strategies designed to classify chemical and materials behavior. Performance is assessed across 31 classification tasks sourced from the literature in chemical and materials science. From these results, we recommend best practices for building data-efficient classifiers, showing the neural network- and random forest-based active learning algorithms are most efficient across tasks. We also show that classification task complexity can be quantified based on task metafeatures, most notably the noise-to-signal ratio. Overall, this work provides a comprehensive survey of data-efficient classification strategies, identifies attributes of top-performing strategies, and suggests avenues for further study. 

Phase separation in multicomponent mixtures is of significant interest in both fundamental research and technology. Although the thermodynamic principles governing phase equilibria are straightforward, practical determination of equilibrium phases and constituent compositions for multicomponent systems is often laborious and computationally intensive. Here, we present a machine-learning workflow that simplifies and accelerates phase-coexistence calculations. We specifically analyze capabilities of neural networks to predict the number, composition, and relative abundance of equilibrium phases of systems described by Flory-Huggins theory. We find that incorporating physics-informed material constraints into the neural network architecture enhances the prediction of equilibrium compositions compared to standard neural networks with minor errors along the boundaries of the stable region. However, introducing additional physics-informed losses does not lead to significant further improvement. These errors can be virtually eliminated by using machine-learning predictions as a warm-start for a subsequent optimization routine. This work provides a promising pathway to efficiently characterize multicomponent phase coexistence. 

We introduce a lattice framework that incorporates elements of Flory–Huggins solution theory and the q-state Potts model to study the phase behavior of polymer solutions and single-chain conformational characteristics. Without empirically introducing temperature-dependent interaction parameters, standard Flory–Huggins theory describes systems that are either homogeneous across temperatures or exhibit upper critical solution temperatures. The proposed Flory–Huggins–Potts framework extends these capabilities by predicting lower critical solution temperatures, miscibility loops, and hourglass-shaped spinodal curves. We particularly show that including orientation-dependent interactions, specifically between monomer segments and solvent particles, is alone sufficient to observe such phase behavior. Signatures of emergent phase behavior are found in single-chain Monte Carlo simulations, which display heating- and cooling-induced coil–globule transitions linked to energy fluctuations. The framework also capably describes a range of experimental systems. Importantly, and by contrast to many prior theoretical approaches, the framework does not employ any temperature- or composition-dependent parameters. This work provides new insights regarding the microscopic physics that underpin complex thermoresponsive behavior in polymers. 

Abstract 2D hybrid organic–inorganic perovskites are potentially promising materials as passivation layers that can enhance the efficiency and stability of perovskite photovoltaics. The ability to suppress ion transport is proposed as a stabilization mechanism, yet an effective characterization of relevant modes of halide diffusion in 2D perovskites is nascent. In light of this knowledge gap, molecular dynamics simulations with enhanced sampling and experimental validation to systematically characterize how ligand chemistry in seven (R‐NH₃)₂PbI₄systems impacts halide diffusion, particularly in the out‐of‐plane direction is combined. It is found that increasing stiffness and length of ligands generally inhibits ion transport, while increasing ligand polarization generally enhances it. Structural and energetic analyses of the migration pathways provide quantitative explanations for these trends, which reflect aspects of the disorder of the organic layer. Overall, this mechanistic analysis greatly enhances the current understanding of halide migration in 2D hybrid organic–inorganic perovskites and yields insights that can inform the design of future passivation materials.