Physics-Informed Deep Learning Approach for Reintroducing Atomic Detail in Coarse-Grained Configurations of Multiple Poly(lactic acid) Stereoisomers.

Multiscale modeling of complex molecular systems, such as macromolecules, encompasses methods that combine information from fine and coarse representations of molecules to capture material properties over a wide range of spatiotemporal scales. Being able to exchange information between different levels of resolution is essential for the effective transfer of this information. The inverse problem of reintroducing atomistic degrees of freedom in coarse-grained (CG) molecular configurations is particularly challenging as, from a mathematical point of view, it is an ill-posed problem; the forward mapping from the atomistic to the CG description is typically defined via a deterministic operator ("one-to-one" problem), whereas the reversed mapping from the CG to the atomistic model refers to creating one representative configuration out of many possible ones ("one-to-many" problem). Most of the backmapping methods proposed so far balance accuracy, efficiency, and general applicability. This is particularly important for macromolecular systems with different types of isomers, i.e., molecules that have the same molecular formula and sequence of bonded atoms (constitution) but differ in the three-dimensional configurations of their atoms in space. Here, we introduce a versatile deep learning approach for backmapping multicomponent CG macromolecules with chiral centers, trained to learn structural correlations between polymer configurations at the atomistic level and their corresponding CG descriptions. This method is intended to be simple and flexible while presenting a generic solution for resolution transformation. In addition, the method is aimed to respect the structural features of the molecule, such as local packing, capturing therefore the physical properties of the material. As an illustrative example, we apply the model on linear poly(lactic acid) (PLA) in melt, which is one of the most popular biodegradable polymers. The framework is tested on a number of model systems starting from homopolymer stereoisomers of PLA to copolymers with randomly placed chiral centers. The results demonstrate the efficiency and efficacy of the new approach.

Deep convolutional neural networks for generating atomistic configurations of multi-component macromolecules from coarse-grained models.

Despite the modern advances in the available computational resources, the length and time scales of the physical systems that can be studied in full atomic detail, via molecular simulations, are still limited. To overcome such limitations, coarse-grained (CG) models have been developed to reduce the dimensionality of the physical system under study. However, to study such systems at the atomic level, it is necessary to re-introduce the atomistic details into the CG description. Such an ill-posed mathematical problem is typically treated via numerical algorithms, which need to balance accuracy, efficiency, and general applicability. Here, we introduce an efficient and versatile method for backmapping multi-component CG macromolecules of arbitrary microstructures. By utilizing deep learning algorithms, we train a convolutional neural network to learn structural correlations between polymer configurations at the atomistic and their corresponding CG descriptions, obtained from atomistic simulations. The trained model is then utilized to get predictions of atomistic structures from input CG configurations. As an illustrative example, we apply the convolutional neural network to polybutadiene copolymers of various microstructures, in which each monomer microstructure (i.e., cis-1,4, trans-1,4, and vinyl-1,2) is represented as a different CG particle type. The proposed methodology is transferable over molecular weight and various microstructures. Moreover, starting from a specific single CG configuration with a given microstructure, we show that by modifying its chemistry (i.e., CG particle types), we are able to obtain a set of well equilibrated polymer configurations of different microstructures (chemistry) than the one of the original CG configuration.

Hierarchical modelling of polystyrene melts: from soft blobs to atomistic resolution.

We demonstrate that hierarchical backmapping strategies incorporating generic blob-based models can equilibrate melts of high-molecular-weight polymers, described with chemically specific, atomistic models. The central idea is first to represent polymers by chains of large soft blobs (spheres) and efficiently equilibrate the melt on large scales. Then, the degrees of freedom of more detailed models are reinserted step by step. The procedure terminates when the atomistic description is reached. Reinsertions are feasible computationally because the fine-grained melt must be re-equilibrated only locally. We consider polystyrene (PS) which is sufficiently complex to serve method development because of stereo-chemistry and bulky side groups. Our backmapping strategy bridges mesoscopic and atomistic scales by incorporating a blob-based, a moderately coarse-grained (CG), and a united-atom model of PS. We demonstrate that the generic blob-based model can be parameterised to reproduce the mesoscale properties of a specific polymer - here PS. The moderately CG model captures stereo-chemistry. To perform backmapping we improve and adjust several fine-graining techniques. We prove equilibration of backmapped PS melts by comparing their structural and conformational properties with reference data from smaller systems, equilibrated with less efficient methods.

Structure and dynamics of stereo-regular poly(methyl-methacrylate) melts through atomistic molecular dynamics simulations.

Poly(methyl-methacrylate), PMMA, is a disubstituted vinyl polymer whose properties depend significantly on its tacticity. Here we present a detailed study of the structure, conformation, and dynamics of syndiotactic, atactic, and isotactic PMMA melts at various temperatures (580, 550, 520, and 490 K) via all-atom molecular dynamics simulations. The calculated volumetric properties are close to experimental data. The Tg and chain dimensions of PMMA model systems are found to depend strongly on tacticity in agreement with experimental findings. The backbone bonds in trans (t), diads in tt, and inter-diads in t|t torsional states are the most populated for all PMMA stereo-chemistries and their fractions increase with the number of syndiotactic sequences. Also, the effective torsional barrier heights for the backbone, ester side group, and α-methyl group are larger for syndiotactic PMMA compared to the isotactic one. The structure of the PMMA chains is studied by computing the intra- and inter-chain static structure factors, S(q), and the radial pair distribution functions. In the first peak of S(q), both intra- and inter-chain components contribute, whereas the second and third peaks mainly come from inter- and intra-chain parts, respectively. For all PMMA stereo-isomers a clear tendency of ester-methyl groups to aggregate is observed. The local dynamics are studied by analyzing torsional autocorrelation functions for various dihedral angles. A wide spectrum of correlation times and different activation energies are observed for the motions of different parts of PMMA chains. The stereo-chemistry affects the backbone, ester side group, and α-methyl motions, whereas the ester-methyl rotation remains unaffected. The dynamic heterogeneity of the PMMA chains is also studied in detail for the different stereo-chemistries via the temperature dependence of the stretching exponent. Furthermore, the reorientational dynamics at the chain level and translational dynamics for monomer and chain centers-of-mass are analyzed.

Dynamic heterogeneity in fully miscible blends of polystyrene with oligostyrene.

Binary blends of polystyrene with oligostyrene are perfectly miscible (χ=0) yet dynamically heterogeneous. This is evidenced by independent probing of the dipole relaxation perpendicular to the backbone by dielectric spectroscopy and molecular dynamics. The self-concentration model with a single intramolecular length scale qualitatively describes the slower segmental dynamics. A quantitative comparison based on MD, however, requires a composition-dependent length scale. The pertinent dynamic length scale that best describes the slow segmental dynamics in miscible blends relates to both intra- and intermolecular contributions.

Aggregation and vesiculation of membrane proteins by curvature-mediated interactions.

Membrane remodelling plays an important role in cellular tasks such as endocytosis, vesiculation and protein sorting, and in the biogenesis of organelles such as the endoplasmic reticulum or the Golgi apparatus. It is well established that the remodelling process is aided by specialized proteins that can sense as well as create membrane curvature, and trigger tubulation when added to synthetic liposomes. Because the energy needed for such large-scale changes in membrane geometry significantly exceeds the binding energy between individual proteins and between protein and membrane, cooperative action is essential. It has recently been suggested that curvature-mediated attractive interactions could aid cooperation and complement the effects of specific binding events on membrane remodelling. But it is difficult to experimentally isolate curvature-mediated interactions from direct attractions between proteins. Moreover, approximate theories predict repulsion between isotropically curving proteins. Here we use coarse-grained membrane simulations to show that curvature-inducing model proteins adsorbed on lipid bilayer membranes can experience attractive interactions that arise purely as a result of membrane curvature. We find that once a minimal local bending is realized, the effect robustly drives protein cluster formation and subsequent transformation into vesicles with radii that correlate with the local curvature imprint. Owing to its universal nature, curvature-mediated attraction can operate even between proteins lacking any specific interactions, such as newly synthesized and still immature membrane proteins in the endoplasmic reticulum.

Adsorption of air pollutants onto silver and gold atomic clusters: DFT and PNO-LCCSD-F12 calculations.

We provide a comprehensive investigation of intermolecular interactions between atmospheric gaseous pollutants, including CH4, CO, CO2, NO, NO2, SO2, as well as H2O and Agn (n = 1-22) or Aun (n = 1-20) atomic clusters. The optimized geometries of all the systems investigated in our study were determined using density functional theory (DFT) with M06-2X functional and SDD basis set. The PNO-LCCSD-F12/SDD method was used for more accurate single-point energy calculations. Compared to their isolated states, the structures of the Agn and Aun clusters undergo severe deformations upon adsorption of the gaseous species, which become more significant as the size of the clusters decreases. Considering that, in addition to adsorption energy, we have determined the interaction and deformation energy of all the systems. All our calculations consistently show that among the gaseous species examined, SO2 and NO2 exhibit a higher preference for adsorption on both types of clusters, with a slightly higher preference for the Ag clusters compared to the Au clusters, with the SO2/Ag16 system exhibiting the lowest adsorption energy. The type of intermolecular interactions was investigated through wave function analyses, including natural bond orbital (NBO) and quantum theory of atoms in molecules (QTAIM), showing that NO2 and SO2 are chemisorbed on the Agn and Aun atomic clusters, whereas the other gas molecules exhibit a much weaker interaction with them. The reported data can be used as input parameters for molecular dynamics simulations to study the selectivity of atomic clusters towards specific gases under ambient conditions, as well as to design materials that take advantage of the studied intermolecular interactions.

Multiscale modeling of soft matter: scaling of dynamics.

Many physical phenomena and properties of soft matter systems are characterized by an interplay of interactions and processes that span a wide range of length- and time scales. Computer simulation approaches require models, which cover these scales. These are typically multiscale models that combine and link different levels of resolution. In order to reach mesoscopic time- and length scales, necessary to access material properties, coarse-grained models are developed. They are based on microscopic, atomistic descriptions of systems and represent these systems on a coarser, mesoscopic level. While the connection between length scales can be established immediately, the link between the different time scales that takes into account the faster dynamics of the coarser system cannot be obtained directly. In this perspective paper we discuss methods that link the time scales in structure based multiscale models. Concepts which try to rigorously map dynamics of related models are limited to simple model systems, while the challenge in soft matter systems is the multitude of fluctuating energy barriers of comparable height. More pragmatic methods to match time scales are applied successfully to quantitatively understand and predict dynamics of one-component soft matter systems. However, there are still open questions. We point out that the link between the dynamics on different resolution levels can be affected by slight changes of the system, as for different tacticities. Furthermore, in two-component systems the dynamics of the host polymer and of additives are accelerated very differently.

A novel method for measuring the bending rigidity of model lipid membranes by simulating tethers.

The tensile force along a cylindrical lipid bilayer tube is proportional to the membrane's bending modulus and inversely proportional to the tube radius. We show that this relation, which is experimentally exploited to measure bending rigidities, can be applied with even greater ease in computer simulations. Using a coarse-grained bilayer model we efficiently obtain bending rigidities that compare very well with complementary measurements based on an analysis of thermal undulation modes. We furthermore illustrate that no deviations from simple quadratic continuum theory occur up to a radius of curvature comparable to the bilayer thickness.

Atomistic molecular dynamics simulation of the temperature and pressure dependences of local and terminal relaxations in cis-1,4-polybutadiene.

The dynamics of cis-1,4-polybutadiene (cis-1,4-PB) over a wide range of temperature and pressure conditions is explored by conducting atomistic molecular dynamics (MD) simulations with a united atom model on a 32-chain C128 cis-1,4-PB system. The local or segmental dynamics is analyzed in terms of the dipole moment time autocorrelation function (DACF) of the simulated polymer and its temperature and pressure variations, for temperatures as low as 195 K and pressures as high as 3 kbars. By Fourier transforming the DACF, the dielectric spectrum, epsilon* = epsilon' + i epsilon" = epsilon*omega, is computed and the normalized epsilon"/epsilon(max)" vs omega/omega(max) plot is analyzed on the basis of the time-temperature and time-pressure superposition principles. The relative contribution of thermal energy and volume to the segmental and chain relaxation processes are also calculated and evaluated in terms of the ratio of the activation energy at constant volume to the activation energy at constant pressure, Q(V)/Q(P). Additional results for the temperature and pressure dependences of the Rouse times describing terminal relaxation in the two polymers show that, in the regime of the temperature and pressure conditions covered here, segmental and chain relaxations are influenced similarly by the pressure and temperature variations. This is in contrast to what is measured experimentally [see, e.g., G. Floudas and T. Reisinger, J. Chem. Phys. 111, 5201 (1999); C. M. Roland et al.,J. Polym. Sci. Part B, 41, 3047 (2003)] for other, chemically more complex polymers that pressure has a stronger influence on the dynamics of segmental mode than on the dynamics of the longest normal mode, at least for the regime of temperature and pressure conditions covered in the present MD simulations.