Polymer microstructure and consequently the polymer properties are strongly dependent on polymerization type and process conditions. In particular for free-radical polymerizations (FRP) the impact of the process conditions on, e.g., the molar mass distribution (MMD), the degree and type of branching, as well as the amount of macromonomer formed is well known. Because of the high complexity of polymerizations, tailoring polymers on demand is challenging. While kinetic Monte Carlo (kMC) simulations of polymerizations based on a full kinetic scheme provide access to detailed microstructural information, e.g., MMDs, copolymer composition and the branching behavior in FRP, large computational resources are required. Further, kMC simulations are not suited for reverse engineering. In contrast, modeling polymerizations with Machine Learning (ML) - based models is less demanding and proceeds in much shorter time, once the ML models were trained with large data sets. In addition, ML-based methods are attractive for reverse engineering in order to propose reaction conditions for the synthesis of polymer with targeted properties. Due to several contradicting objectives, multi-objective optimization (MOO) is established, e.g., with monomer conversion, reaction time, and MMD similarity as objectives. Application of the ML methods to reverse engineering of vinyl acetate and butyl acrylate homopolymerizations is reported. The ML models were trained and tested with data sets generated using kMC simulations.

PEEK is a high-performance thermoplastic for demanding load-bearing applications, where reliable prediction of mechanical behavior is essential. As a semi-crystalline polymer, the mechanical performance of PEEK depends on the crystallinity and morphology, which rely strongly on cooling and flow conditions during molding. Accurate prediction of the resulting mechanical response requires: 1) a description of structure development during processing, and 2) an adequate structure-property relation. This study focuses on the latter and aims to develop a micromechanical model that links morphological details, such as crystallinity and crystalline orientation distribution, directly to the mechanical performance of PEEK.

The micromechanical approach considers the amorphous and crystalline phases separately, connecting them in two-phase layered domains. Their response is coupled via a hybrid interaction law. This model, called the composite inclusion model, has described micro-mechanical relationships in other semi-crystalline materials [1] and is now applied to PEEK.

The amorphous phase follows a phenomenological model, the Eindhoven Glassy Polymer (EGP) model. The crystalline phase is modeled with crystal plasticity, governed by crystallographic slip. Model parameters are identified from uniaxial compression test on fully amorphous PEEK, for the amorphous phase, and on semi-crystalline material, for the crystalline phase.

The identified model parameters enable the model to capture the mechanical behavior of PEEK over a range of strain rates and temperatures. The versatility of the model is demonstrated by comparing experiments and simulations at different levels of crystallinity and for different thermal histories.

We employed a multi-scale computational approach to investigate how temperature affects the dynamics and interfacial width of unentangled cis-1,4 Polybutadiene (cPB) chains in the vicinity of alumina interface. First, we developed an accurate interfacial model by parameterizing classical potentials from intuitively sampled Density Functional Theory (DFT) data, ensuring a chemically precise description of the polymer/alumina interactions. [1] This model provided the foundation for extensive atomistic molecular dynamics (MD) simulations spanning several μs and covering a broad temperature range, both above and below the glass transition (Tg). Using these simulations, we analyzed the temperature-dependent behavior of polymer chains near the interface. Our results reveal that the interface induces strong structural and dynamical heterogeneity within a 2–3 nm region. Polymer mobility decreases near the surface, leading to an elevated local Tg. Additionally, we identified a slow Arrhenius processes (SAP), where an abrupt change in activation energy occurs at a critical temperature approximately 80 K above Tg, marking a transition in polymer dynamics at the interface. Finally, we quantified the interfacial width based on structural and dynamical heterogeneity, observing that it peaks around the bulk Tg and decreases at both lower and higher temperatures. [2]

The properties of high free-volume amorphous polymers are of great interest as potential material for the fabrication of membranes for gas separation due to their large permeabilities and good selectivity for various penetrant pairs, due to their large sorption capacity [1].
This study investigates gas (CO₂ and N₂) sorption in amorphous glassy polymers using cryogenic analysis, interpreted by two main model approaches: BET theory and the Non-Equilibrium Lattice Fluid (NELF) model. A volumetric apparatus was used to conduct sorption isotherms on commercial polymers, including Topas, Matrimid, Torlon, PTMSP, Polyphenylene oxide, to determine gas solubility. By employing N₂ and CO₂ at low temperature and high pressure (up to 100 bar), we assessed the applicability of BET theory [2] to these materials and obtained insights into their structure and gas interactions, with potential implications for gas separation membranes [3].
The results (Fig. 1) demonstrated that those amorphous polymers exhibit an appreciable gas uptake, in spite of the absence of a clear intrinsic porosity. The data obtained provide valuable indications on the dilation of the polymer upon sorption and the available free-volume, essential indications for the determination of gas permeability and selectivity.
Moreover, cryogenic isotherms have been examined by using the NELF model, appropriate to represent the solubility of low molecular weight penetrant in homogeneous amorphous glassy polymers [4]. The model accurately represents the experimental data at different operating conditions, both for high or small free-volume polymers, proving to be a useful tool for the quantification of polymer free-volume.

The kinetics of chemical ageing in most commercial plastics remain long-debated, since the nature of dense polymeric solids inhibits the in-situ experimental investigation of complex degradation paths, while traditional computational techniques fail to reach the timescales associated with the slowly progressing degradative reactions. We present a computational framework in which any chemical path is described as successive elementary transitions of the polymeric system between local minima on the energy landscape. For each elementary transition, the transition state can be identified, allowing the estimation of the free-energy barrier and, thereby, of the transition rate constant by means of transition state theory. The use of an appropriately trained reactive forcefield, for the efficient description of the energy landscape, allows the large-scale sampling of potential reaction paths. The result is a network of states populated by the stationary states visited by the system along the chemical paths.

Studying the autoxidation of glassy polystyrene using the developed framework, we extract from the created network-of-states the energetics and rates of the reactions propagating polymers oxidation, i.e., peroxy-radical and hydroperoxide formation,[1] which to the best of our knowledge had not been achieved in the past. The free-energy barriers of both reactions are spanning over the same orders of magnitude, challenging the common assumption that hydroperoxide formation is the rate-determining step of performance-polymers oxidation. The broad distribution of rate constants for both reactions is indicative of the effect of the dense environment on their kinetics and highlights the importance of studying solid-state reactions in-situ.