Biomass-derived porous carbons have been considered one of the most effective adsorbents for CO₂ capture, due to their porous structure and high specific surface area. In this study, we successfully synthesized porous carbon from celery biomass and examined the effect of external adsorption parameters including time, temperature, and pressure on CO₂ uptake in experimental and molecular dynamics (MD) simulations. Furthermore, the influence of carbon’s surface chemistry (carboxyl and hydroxyl functionalities) and nitrogen type on CO₂ capture were investigated utilizing MD simulations. The results showed that pyridinic nitrogen has a greater tendency to adsorb CO₂ than graphitic. It was found that the simultaneous presence of these two types of nitrogen has a greater effect on the CO₂ sorption than the individual presence of each in the structure. It was also revealed that the addition of carboxyl groups (O=C–OH) to the carbon matrix enhances CO₂ capture by about 10%. Additionally, by increasing the simulation time and the size of the simulation box, the average absolute relative error for simulation results of optimal structure declined to 16%, which is an acceptable value and makes the simulation process reliable to predict adsorption capacity under various conditions.

As it is now established that global warming and climate change are a reality, international investments are pouring in and rightfully so for climate change mitigation. Carbon capture and separation (CCS) is therefore gaining paramount importance as it is considered one of the powerful solutions for global warming. Sorption on porous materials is a promising alternative to traditional carbon dioxide (CO₂) capture technologies. Owing to their sustainable availability, economic viability, and important recyclability, natural products‐derived porous carbons have emerged as favorable and competitive materials for CO₂ sorption. Furthermore, the fabrication of high‐quality value‐added functional porous carbon‐based materials using renewable precursors and waste materials is an environmentally friendly approach. This review provides crucial insights and analyses to enhance the understanding of the application of porous carbons in CO₂ capture. Various methods for the synthesis of porous carbon, their structural characterization, and parameters that influence their sorption properties are discussed. The review also delves into the utilization of molecular dynamics (MD), Monte Carlo (MC), density functional theory (DFT), and machine learning techniques for simulating adsorption and validating experimental results. Lastly, the review provides future outlook and research directions for progressing the use of natural products‐derived porous carbons for CO₂ capture.

The emergence of artificial intelligence (AI) and, more particularly, machine learning (ML), has had a significant impact on engineering and the fundamental sciences, resulting in advances in various fields. The use of ML has significantly enhanced data processing and analysis, eliciting the development of new and improved technologies. Specifically, ML is projected to play an increasingly significant role in helping researchers better understand and predict the behavior of porous media. Furthermore, ML models will be able to make use of sizable datasets, such as subsurface data and experiments, to produce accurate predictions and simulations of porous media systems. This capability could help optimize the design of porous materials for specific applications and improve the effectiveness of industrial processes. To this end, this review paper attempts to provide an overview of the present status quo in this context, i.e., the interface of ML and porous media in six different applications, namely, heat exchanger and storage, energy storage and combustion, electrochemical devices, hydrocarbon reservoirs, carbon capture and sequestration, and groundwater, stressing the advances made in the application of ML to porous media and offering insights into the challenges and opportunities for future research. Each section also entails a supplementary database of the literature as a spreadsheet, which includes the details of ML models, datasets, key findings, etc., and mentions relevant available online datasets that can be used to train ML models. Future research trends include employing hybrid models by combining ML models with physics-based models of porous media to improve predictions concerning accuracy and interpretability.

We studied the performance of fine-tuning open-source LLMs for a range of different chemical questions. We benchmark their performances against “traditional” machine learning models and find that, in most cases, the fine-tuning approach is superior.