Experimental Study an‎d Molecular Simulation of Zeolitic Nanostructures to Investigate Carbon Dioxide Adsorption an‎d Mass Transfer Mechanisms Integrating Artificial Intelligence

4/19/2027 12:00:00 AM

In this research, a hierarchical core–shell zeolite 13X@ZIF-8 structure was synthesized for the first time using a two-step in situ growth strategy, in which diethanolamine was employed as a shell-growth directing agent on the zeolite core. Crystallographic an‎d microscopic characterizations confirmed the uniform formation of a nanoscale ZIF-8 shell over the Zeolite 13X core, leading to the development of a hierarchically porous composite while preserving the crystallographic integrity of both constituent phases. Carbon dioxide adsorption experiments were conducted over a temperature range of 25–65 °C an‎d pressures between 1 an‎d 9 bar. The results showed a maximum CO2 adsorption capacity of 6.1 mmol g⁻1 at 1 bar, along with excellent cyclic stability, with more than 94.5 % of the initial adsorption capacity retained after 40 adsorption–desorption cycles. To investigate the governing mass-transfer mechanisms, adsorption kinetics were analyzed by fitting experimental data to several kinetic models. Among them, the Avrami model exhibited the best agreement, confirming the non-linear an‎d multi-stage nature of the adsorption process. Furthermore, by employing dual- an‎d series-resistance frameworks, the individual contributions of film diffusion, Knudsen diffusion, intraparticle diffusion, an‎d surface diffusion were quantitatively distinguished. The results revealed that the effective diffusion coefficient decreases as pressure increases from 1 to 9 bar, with surface diffusion identified as the dominant transport mechanism. Based on Buckingham’s π-theorem, a novel dimensionless correlation was developed to estimate the effective diffusion coefficient in core–shell adsorbents, simultaneously accounting for shell thickness, particle geometry, material structure, an‎d operating conditions. In parallel, the hierarchical zeolite 13X@ZIF-8 structure was modeled for the first time using Materials Studio, an‎d the simulated crystallographic features showed strong agreement with experimental observations. Gran‎d Canonical Monte Carlo (GCMC) simulations of CO2 adsorption were performed at fixed temperatures of 25–55 °C an‎d pressures up to 10 bar to validate experimental adsorption trends an‎d to provide molecular-level insight into adsorption isotherms, isosteric heats of adsorption, an‎d preferential adsorption sites. In this work, molecular simulations were employed exclusively as a validation an‎d mechanistic interpretation tool rather than as a data source for predictive modeling. Beyond molecular-scale analysis, a data-driven framework based on machine learning was developed to predict CO2 adsorption capacity in core–shell zeolite@MOF systems under diverse physicochemical conditions. A comprehensive dataset comprising 670 experimental data points, collected from this study an‎d published experimental literature on Zeolite@MOF core–shell adsorbents, was constructed. The dataset included Si/Al ratios (1–30), temperatures (273–338 K), pressures (0–10 bar), BET surface areas (211–1504 m2 g⁻¹), pore volumes (0.07–0.77 cm3 g⁻1), an‎d TGA mass loss up to 300 °C (3–23 %). Multilayer perceptron (MLP), radial basis function (RBF), an‎d support vector machine (SVM) models were trained an‎d optimized, all demonstrating excellent agreement with experimental results, with coefficients of determination exceeding 0.97. Among them, the RBF neural network achieved the highest predictive accuracy (R² = 0.9988, MSE = 0.003), effectively capturing the complex non-linear interactions between structural an‎d operational descriptors. Interpretability analysis of the machine-learning models indicated that BET surface area, operating pressure, an‎d temperature are the most influential parameters governing CO2 adsorption performance, while framework composition an‎d metal molecular weight exert secondary but synergistic effects through core–shell interfacial interactions. Intermediate Si/Al ratios an‎d metals with moderate atomic weights, such as zinc an‎d copper, were found to balance hydro‎philicity an‎d Lewis acidity, thereby enhancing adsorption efficiency. Overall, the integrated experimental, molecular-scale validation, an‎d data-driven framework developed in this thesis provides a physically grounded an‎d scalable strategy for the rational design an‎d optimization of high-performance hierarchical adsorbents for CO2 capture.

زئوليت , كربن دي اكسيد , شبيه سازي مولكولي , ماشين لرنينگ , ساختار هسته-پوسته

zeolite , Molecular simulation , Carbon dioxide , Machine learning , core-shell structure

fatemeh bahmanzadegan jahromi

ahad ghaemi