The absorption of CO2 by amine solutions is a fundamental step in carbon capture technology. Accurately predicting the pseudo-first-order rate constants (k0) for this process is essential for evaluating absorbent performance. Experimental methods are both expensive and time-intensive, and traditional kinetic models often lack broad applicability. A machine learning (ML) framework was developed to predict the k0 for CO2 absorption in amine solutions. A dataset containing 1,301 samples covering 27 amines under diverse experimental conditions was compiled. An ESP-based feature engineering strategy (Experimental Parameters, Structural Descriptors and Property Features) was used to construct a chemically informative and high-quality input representation. Four ML algorithms-XGBoost (eXtreme Gradient Boosting), RF (Random Forest), CNN (Convolutional Neural Network) and DNN (Deep Neural Network) were trained and compared, with ESP-XGBoost model achieving the best performance (R2 = 0.992). Compared to traditional kinetic models, ML methods showed stronger generalizability. SHAP analysis suggested that regions with high electron density and accessible surface area enhance the CO2 absorption reaction. This interpretable and robust framework offers a practical tool for predicting reaction performance and optimizing solvent design in CO2 capture processes.

The density of CO₂-unloaded and CO₂-loaded aqueous solutions of 2-amino-2-methyl-1-propanol (AMP) + 1,5-diamino-2-methylpentane (DA2MP) and methyldiethanolamine (MDEA) + DA2MP bi-solvent blends were experimentally investigated under varying amine concentrations (2–5 M), temperatures (25–50 °C), and CO₂ loadings (0–1.0 mol CO₂/mol amine). To predict the experimental results, 4 modeling approaches were applied: artificial neural networks (ANN), response surface methodology (RSM), Nwaoha’s nonlinear model, and Nwaoha et al.’s modified nonlinear model. Results showed that higher CO₂ loading increased density, while higher temperature and amine blend concentration reduced density. Predictive modeling demonstrated that the ANN exhibited the highest accuracy, achieving a maximum absolute average deviation (%AAD) of 0.38%, a high coefficient of determination (R²) of 0.9999, and the highest mean squared error (MSE) of 1.6E-5. Meanwhile, the modified nonlinear model had a maximum %AAD of 0.78%, the highest R² of 0.9697, and the highest MSE value of 1.05E-4. The nonlinear model achieved maximum %AAD of 0.9155%, highest R² of 0.9309, and highest MSE value of 1.15E-4. The RSM showed maximum %AAD of 2.1401%, highest R² of 0.9782, and MSE of 4.97E-4. The predictive performance of the models follows this trend: ANN > modified nonlinear model > nonlinear model > RSM. Notably, the modified nonlinear model, despite requiring significantly less computational effort and expertise compared to ANN and RSM, proved to be highly effective for accurately predicting density.

Although catalytic amine regeneration can effectively reduce heat duty in CO2 capture, the insufficient activity and stability of existing solid acid catalysts in high-temperature basic environments impede their practical implementation. Herein, phosphorus-doped H-Beta (P-H beta) catalysts were employed for CO2 desorption for the first time. Phosphorus doping H beta forms P-OH, which undergoes in situ reconstruction into active -PO3H groups, generating abundant strong acid sites (SAS) and enhancing electron transfer efficiency. This synergistic effect significantly enhances the proton-coupled electron transfer (PCET) process, thereby facilitating CO2 desorption. The optimal P-H beta catalyst improves the CO2 desorption rate by 368.0%, lowers the activation energy (Ea) by 33.3%, and reduces the regeneration energy consumption by 49.1%, while maintaining stable performance over 10 cycles. Online fourier transform infrared spectroscopy (FT-IR) analysis reveals the synergistic catalytic mechanism. This work thus provides a novel perspective for designing highly efficient and robust CO2 desorption catalysts by elucidating the clear mechanism of the PCET process.

This study evaluates the equilibrium CO2 loading, CO2 absorption capacity, cyclic capacity, initial CO₂ desorption rate, heat of absorption, heat of vaporization, and overall regeneration energy of a tri-solvent amine blend system comprising 2-amino-2-methyl-1-propanol (AMP), piperazine (PZ), and diethylenetriamine (DETA). The AMP concentration was fixed at 2 mol/L, while PZ and DETA concentrations each were varied between 0.5 and 1.5 mol/L, maintaining a total amine concentration of 4 mol/L (37–40 wt%). Experimental results demonstrate that the AMP–PZ–DETA tri-solvent blend system exhibits significantly higher equilibrium CO2 loading (33.3–60.8%), CO2 absorption capacity (3.8–26.9%), cyclic capacity (18.5–66.2%), initial CO₂ desorption rates (117–137%), while reducing regeneration energy by 53.8–57.8% relative to the benchmark monoethanolamine (MEA) solution, and lower regeneration energy (26.2–30.1%,) than the AMP–DETA bi-solvent blend. Furthermore, both the heat of absorption and heat of vaporization for the tri-solvent system were found to be lower than those of the AMP–DETA blend and MEA. The AMP–PZ–DETA tri-solvent blends also exhibited relative regeneration energy comparable to, and in several cases lower than, that of the optimal concentrations reported for recently investigated tri-solvent amine blend systems. Overall, the results indicate that the AMP–PZ–DETA system offers improved CO2 absorption and regeneration characteristics, while the variation in tri-solvent blend composition highlights its operational flexibility in maintaining low energy requirements without compromising the CO2 absorption and cyclic capacities for post-combustion CO₂ capture applications.

To address high energy consumption of post-combustion CO2 capture, this pilot study assessed a bi-solvent blend of 2-amino-2-methyl-1-propanol (AMP) and 1,5-diamino-2-methylpentane (DA2MP) for improved performance. The AMP concentration was held constant at 19.wt.%, while the DA2MP concentration varied between 13.5 and 40.5 wt%, resulting in total blend concentrations ranging from 32.5 to 59.5 wt%. Key performance indicators including capture efficiency, CO2 loading, absorption rate, overall mass transfer coefficients in the absorption and desorption columns, and regeneration energy were evaluated and compared with a 30 wt% MEA solution and a 19 wt% AMP + 10 wt% PZ bi-solvent blend. Results indicated that that at higher concentrations (39-59.5 wt%), the AMP + DA2MP blend outperformed MEA in CO2 capture efficiency by 18.9-66.1 %. Across all concentrations, the AMP + DA2MP blend also exhibited a higher CO2 absorption capacity ranging from 5.2 % to 92.3 % compared to the AMP + PZ blend solution. Raising the concentration of AMP + DA2MP from 39 to 59.5 wt% led to notable improvements in mass transfer: absorption column gas-phase coefficients increased by 56.2-233.8 %, while desorption column coefficients increased by 21.2-49.9 % relative to MEA. Similarly, compared to AMP + PZ, the AMP + DA2MP blend achieved 6.6-302.1 % higher absorption and 7.4-43.6 % higher desorption mass transfer coefficients. Moreover, the AMP + DA2MP blend required significantly lowered regeneration energy, 15.4-40.1 % less than MEA and 12.5-54.6 % less than AMP + PZ blend. Based on overall performance, the two optimal blend formulations were identified as 19 wt% AMP + 20 wt% DA2MP and 19 wt% AMP + 27 wt% DA2MP.

The development of accurate thermophysical property data is essential for optimizing solvent systems in CO2 capture processes. Although amine-based solvents such as 2-amino-2-methyl-1-propanol (AMP) and diethylenetriamine (DETA) are promising candidates, limited information exists on the density and viscosity behavior of their aqueous blends under CO2-free and varying CO2 loadings. This study investigates the density and viscosity of CO2-free and CO2-loaded aqueous AMP-DETA bi-solvent blends relevant to CO2 capture applications by conducting experiments across varying AMP-DETA blend concentrations (2-5 M), temperatures (25-50 degrees C), and CO2 loadings (0.0-0.91 mol CO2/mol amine). Results showed that both density and viscosity increased with amine concentration and CO2 loading but decreased with higher temperatures. To predict these properties, four modeling approaches were evaluated: artificial neural networks (ANN), a previously published nonlinear model, response surface methodology (RSM), and a novel modified-nonlinear model. ANN demonstrated the highest accuracy (%AAD and RMSE of 0.00; R-2 = 0.9999), and RSM showed good predictive accuracy (%AAD = 0.37%-15.93%; RMSE = 0.48-3.76; R-2 = 0.9697-0.9788). The novel modified-nonlinear model followed, offering strong performance (%AAD = 0.56%-9.49%; RMSE = 0.48-5.66; R-2 = 0.9522-0.9594) with lower computational effort than ANN and RSM. While RSM slightly outperformed the modified model in predicting density, but less effective for viscosity. The nonlinear model has the least predictive accuracy (%AAD = 1.03%-12.32%; RMSE = 0.62-10.35; R-2 = 0.8325-0.9245). Overall, the four models ranked as follows: ANN > novel modified-nonlinear model > RSM > nonlinear model.

The regeneration of CO2-rich amine absorbents remain a significant bottleneck in post-combustion carbon capture due to substantial energy demand associated with conventional thermal stripping, especially for monoethanolamine (MEA)-based systems, which currently represent the industrial benchmark. Solid acid catalysts (SACs) have emerged as promising materials capable of lowering the energy demand, while simultaneously enhancing CO2 desorption kinetics. However, a comprehensive understanding of the influence of SAC physicochemical properties on CO2 desorption mechanisms and performance remains limited. This review addresses this gap by systematically examining the role of key SACs physiochemical properties, including total acidity and strengths, Br empty set nsted/Lewis (B/L) sites ratio, surface area, pore volume, and pore size, in governing the CO2 desorption mechanisms. A rigorous literature survey was conducted using scientific databases, including Scopus and Google Scholar, to identify relevant peer-reviewed studies published within the stipulated timeframe (2017-2025). All six classes of SACs including metal oxides, clay minerals, molecular sieves, carbon-rich materials, organic frameworks, and composites are discussed in terms of their acid functionality and structural attributes. Particular emphasis is given to catalytic mechanisms involving AmineCOO(-) breakdown and AmineH+ deprotonation pathways. Comparative data shows that SACs improved RHD reduction by 10-57 % relative to non-catalytic system. Findings reveal that the catalyst's performance depends on the availability of acid sites and strengths, B/L ratio, mesoporous surface area, and large pore distribution. Future research should prioritize hybrid SACs with improved acidity, porosity, and stability under cyclic conditions for sustainable industrial deployment.

The electrolyte non-random two-liquid (e-NRTL) model was employed to analyze CO2 solubility in liquid mixtures of amines, including 2-methyl piperazine (2-MPZ), sulfolane (SUL), the ionic liquid (IL) 1-butyl-3-methylimidazolium acetate, and methyl diethanolamine (MDEA), as well as in a mixture of 3-aminopropyl triethoxysilane (TESA), 1-(2-aminoethyl) piperazine (AEP), and bis(3-aminopropyl) amine (APA). The gas phase was modelled using the Redlich-Kwong equation of state. The regression analysis yielded average absolute deviations (%AAD) of 0.31% for temperature, 0.03% for pressure, and 0.03% for molar fraction in the {2-MPZ, SUL, IL, MDEA} system, and 0.04%, 0.44%, and 0.36%, respectively, in the {TESA, AEP, APA} system. This modelling methodology quantitatively determines species concentrations in the liquid phase, demonstrating that increasing CO2 loadings enhances the proportions of CO2-related species while diminishing the abundance of unreacted amines. These findings suggest that the e-NRTL framework offers a more accurate and reliable representation of gas-liquid equilibrium compared to previous models.

Amine‐based chemical absorption remains a promisingly powerful technology for carbon dioxide (CO 2 ) capture. However, enhancing the performance of absorbents remains a significant challenge. In this study, three sterically hindered amines (SHAs), including 2‐( tert ‐butylamino)ethanol, 2‐amino‐2‐methyl‐1‐propanol, 2‐amino‐2‐methylpropane‐1,3‐diol were selected as the primary absorbents with piperazine (PZ) employed as an activator. Absorption and desorption experiments were conducted for various PZ‐SHAs systems with a total amine concentration of 30 wt% under conditions of 313 K and 13 kPa CO 2 partial pressure. The CO 2 absorption capacity, initial absorption rate, desorption rate, and desorption energy consumption were systematically evaluated, indicating that the addition of PZ as an activator could effectively improve the trapping capacity of the spatially hindered amine, and keep a high recycling capacity and low regeneration energy. Finally, the CO 2 reaction mechanism under the new system was clarified by 13 C nuclear magnetic resonance (NMR) technique.

This study evaluates a bench-scale process using a blended solvent system of 2-amino-2-methyl-1-propanol (AMP) and 1,5-diamino-2-methylpentane (DA2MP) for CO2 capture. A parametric sensitivity assessment was conducted considering AMP + DA2MP blend concentration and flow rate, flue gas flow rate, and desorption (reboiler) temperature as independent parameters. Dependent parameters included CO2 capture efficiency, mass transfer coefficients in the absorption and desorption columns, CO2-rich and lean amine loadings, cyclic loading, CO2 absorption rate, sensible and vaporization energy requirements, regeneration energy, and specific solvent cost. Parametric sensitivity analysis results indicated that the AMP + DA2MP blend concentration, AMP + DA2MP blend flow rate, and desorption temperature in this order are the most critical factors, enhancing the overall CO2 capture process system, especially the CO2 capture efficiency, regeneration energy and mass transfer coefficients of the absorption and desorption columns. A new multi-objective optimization methodology using a maximization and minimization objectives is proposed, and the results revealed that the optimal range of amine blend concentration is 2 mol/L AMP+(1.5-1.7 mol/L) DA2MP blend, equivalent to 19 wt% AMP+(20.8-23.2 wt %) DA2MP blend, for a total amine blend concentration of 39.8-42.2 wt%. The optimal amine blend flow rate and desorption temperature were determined to be 49.5-53 mL/min and 109-111 degrees C, respectively. All the optimized AMP + DA2MP blend systems have superior absorption-regeneration performance at a lower desorption temperature (109-111 degrees C) than the benchmark 5 mol/L (30 wt%) MEA solution which operated at 50 mL/min amine flow rate and 120 degrees C desorption temperature.

Abstract Deep eutectic solvents (DESs), as an emerging class of green solvents, have demonstrated great potential in gas absorption and separation owing to their favorable physicochemical properties. However, accurate prediction of CO2 solubility in DESs across a wide range of temperatures and pressures remains a major challenge, limiting their optimization in carbon capture applications. In this work, two input representations, Simplified Molecular Input Line Entry System-based structural coding and physicochemical descriptors, were comparatively evaluated for CO2 solubility prediction in DESs. The dataset includes predominantly choline chloride-based DESs, together with selected betaine-based and ammonium salt-based systems, spanning both hydrophilic and limited hydrophobic subclasses. The dataset contains 2648 experimental measurements corresponding to 93 independent hydrogen bond acceptor-hydrogen bond donor (HBA-HBD) systems under different temperatures, pressures, and compositions. Four machine learning algorithms—extreme gradient boosting, random forest, deep neural network, and convolutional neural network—were evaluated using two input representations. All measurements associated with the same HBA-HBD pair were retained within the same data subset. Among the evaluated model–input combinations, the SC-based RF model achieved the highest test-set performance, with an R2 of 0.971 under the adopted random split. This study provides an exploratory comparison of ML strategies for CO2 solubility prediction within the DES chemical space represented by the collected dataset.

Energy-intensive distillation remains the dominant process for propylene/propane separation, creating a strong demand for membrane-based alternatives. Mixed-matrix membranes (MMMs) based on ZIF-8 and PIM-1 offer high potential but are often limited by interfacial voids and inefficient pore connectivity. Here, we combine multi-scale simulations with experiments to clarify how imidazolium-based ionic liquids (ILs) regulate the ZIF-8/PIM-1 interface and influence gas transport. The results show that IL molecular structure, including cation size and flexibility as well as anion volume and polarity, governs interfacial occupation, free-volume distribution, and polymer-chain rigidity. Two distinct interfacial regulation modes are identified: bulky and flexible ILs form geometry-filling interlayers that densify the interface and remove non-selective voids, whereas compact ILs create interaction-mediated anchoring sites that induce localized ordering while preserving continuous microporous pathways. These two modes lead to different separation behaviors. Geometry-filling ILs improve propylene/propane selectivity but reduce permeability due to interfacial densification. In contrast, interaction-mediated anchoring selectively suppresses large non-selective pores while maintaining diffusion efficiency, predicting an approximately 78% increase in propylene/propane selectivity without significant loss of permeability. These findings establish a structure–interfacial organization–transport relationship, providing a molecular-level design guideline for IL-regulated MMMs.

Amine-based water-lean solvents serve as energy-efficient absorbents for carbon dioxide (CO2) capture but partially face the kinetic challenge of low reaction rates. In this study, stopped-flow experiments were conducted to investigate the CO2 absorption kinetics of 3-Aminopropanol (3AP)-N,N-dimethylethanolamine (DMEA) blends. Results reveal that high-concentration DMEA significantly boosts primary-tertiary amine synergy, enabling "low-water, high-rate" absorption. MD simulations elucidated the mechanism: DMEA acts as a competitive proton acceptor, switching the zwitterion proton transfer pathway toward DMEA and weakening 3AP solvation, increasing free active 3AP. Ab initio molecular dynamics combined with density functional theory calculations further confirmed the enhanced synergy in ethanol is driven by preferential zwitterion proton transfer to DMEA. The proposed mechanism was validated by a kinetic model (AARD <6.00%). This work provides deep insights into primary-tertiary amine synergy in water-lean blends, guiding the development of high-efficiency water-lean CO2 absorbents.

This study investigated the initial CO2 absorption efficiency, initial average overall absorption mass transfer coefficient, absorption heat, heat of vaporization, density, and viscosity of the novel aqueous AMP–1,5-diamino-2-methylpentane (DA2MP)–MEA tri-solvent amine blend solution using a bench-scale absorption column process plant. The concentration of AMP was kept constant at 2 mol·L−1, while the DA2MP concentration ranged from 1 to 2 mol·L−1 and MEA concentration from 2 to 3 mol·L−1, while the total tri-solvent amine blend concentration was kept at 6 mol·L−1. Results showed that the AMP–DA2MP–MEA blend possessed some of the lowest densities but higher viscosities than the other studied amine systems. Results revealed that the AMP–DA2MP–MEA blend has a lower absorber column height (18% to 58%) to achieve initial 90% CO2 absorption efficiency compared to other investigated amine solutions. Notably, the AMP–DA2MP–MEA blends achieved the greatest reduction in absorber height (41–58%) relative to the benchmark MEA solution. The AMP–DA2MP–MEA blend also possesses a higher initial absorber mass transfer coefficient compared with the benchmark MEA solution (65.9% to 118.9%), the AMP–DA2MP blend (6.6% to 40.6%), and the AMP–PZ–MEA blend (44.8% to 91%). The AMP–DA2MP–MEA blend has higher heat of vaporization (up to 9.8%) and absorption heat (up to 28.8%) compared with the other amine systems. These findings indicate that the AMP–DA2MP–MEA tri-solvent blend is a strong candidate for advanced CO2 capture applications and requires further investigation.

Non-aqueous absorbent combinations for absorption-based post-combustion carbon capture have been studied over the past two decades. This study is the first meta-analysis to statistically evaluate the performance of all such combinations reported between 2001 and 2024. Numerous R packages were utilized for comprehensive analysis and visualization, while PRISMA guidelines were strictly followed to select 351 pertinent studies through keywords-based preliminary screening. Applying rigorous inclusion exclusion criteria, 63 studies were shortlisted, providing 299 data points on CO2 loading, cyclic capacity, pre- & post-absorption viscosity, and regeneration energy. For systematic analysis, non-aqueous absorbent combinations were divided into five major classes, further divided into 27 categories and 47 subcategories. Both extensively studied and less explored subcategories of absorbent combinations were considered in the heterogeneity evaluation using random-effects and fixed-effects models. Top-performing combinations for CO2 loading include polyamine-secondary amine-glycol (multiple-amine), potassium prolinate-ethanol-glycol (one-amine), and ionic liquid-organic solvent (amine-free). The highest cyclic capacities were observed in polyamine-primary amine-alcohol (multiple-amine) and polyamine-glycol-alcohol (one-amine), while data for amine-free and DES-based absorbent combinations were insufficient. The lowest pre-absorption viscosity was shown by the sub-category of ethanolamine-hydrochloride-DES, while the minimal post-absorption viscosity was exhibited by the sub-category of secondary amine-organic solvent-alcohol combination, both favorable for industrial flow and energy efficiency. The sub-category of polyamine-glycol-alcohol demonstrated the most energy-efficient regeneration. This meta-analysis emphasizes the importance of optimal absorbent selection through detailed categorization, illustrating clear performance variations across sub-categories and their strong potential for sustainable carbon capture by addressing critical research gaps and enhancing experimental consistency.

Energy efficient solvent was formulated from 2-amino-2-methyl-1-propanol (AMP), piperazine (PZ), and triethanolamine (TEA). Since AMP and PZ can precipitate, ternary blend was explored at unloaded and rich CO2 loading (0.6 mol CO2/mol amine). Either too low TEA concentration, too low PZ/AMP molar ratio, or an increase total amine concentration from 5.0 to 5.5 or 6.0 M induced the precipitation at rich CO2 loading. Even though equilibrium CO2 loaded density and viscosity of the blends were higher than that of 5.0 M MEA (especially, viscosity), they are in operational ranges. A proper solvent formulation strategy was to trade-off between an increment of PZ/AMP molar ratio (favors the absorption) and an elevation of TEA concentration (positively affects the regeneration). Respecting 5.0 M MEA, 0.7/2.8/1.5 has 9 % higher capacity, 73 % greater cyclic capacity, 45 % larger mass transfer coefficient, 5 % higher CO2 removal percentage, and 26 % lower regeneration energy.

Machine Learning (ML) models have demonstrated outstanding performance in predicting essential parameters in the carbon capture process system and support a better understanding of the relationships among the parameters. Their effectiveness in accurately processing and analyzing large volumes of data is well-established. However, these models often function as "black boxes," and the reasoning or processes of the models are often unknown. Practitioners often struggle to understand how specific inputs result in particular outputs. This lack of transparency is a barrier to the wider adoption of ML approaches in sectors such as healthcare, finance, heavy industry, and law, where decisions often need to be transparent and justifiable.Therefore, increasing transparency in ML models is essential for enhancing the adoption of the ML approach. One possible solution is to incorporate Explainable Artificial Intelligence (XAI) techniques, which aim to clarify the decision-making processes of the models. For instance, feature importance metrics and attention mechanisms can identify the most critical inputs in decision-making. By shedding light on the inner mechanisms of ML models, practitioners can enhance their understanding and build confidence in using ML technologies. This paper presents a method designed to enhance the transparency of ML models. Our approach uses advanced modelling techniques and model explanation methods, namely the Decision Trees Ensemble (DTE) and Tree-Based Local Interpretable Model-agnostic Explanation (LIMETree), to make predictions more understandable for practitioners. To validate the approach, a new dataset was generated from a ProMax simulation, in which specifications for the contactors are derived from existing carbon dioxide capture units in North America. Using the same method Wang et al. proposed [1], we first developed an accurate correlation model of the relationships of parameters in the carbon capture process system with the help of a DTE, GAN (Generative Adversarial Network) and PFA (Principal Feature Analysis). Then, we applied the LIMETree method to interpret the DTE model's predictions.

Numerous review papers on solid acid catalyst-based energy-efficient CO2 desorption in solvent regeneration have been published; however, a comprehensive statistical analyzes of the catalytic solvent regeneration has not been investigated yet. This study aims to provide a systematic meta-analysis of the most pertinent studies conducted between 2017-2023 on CO2 desorption and heat duty reduction in catalytic solvent regeneration. After the selection criteria narrowed the number of 4,520 publications that initially matched the keywords to 43 research papers for detailed review and data extraction. Meta-analysis was conducted by the random-effects model using various packages of R and R Studio software. Fourteen types of solvents were found to be applied with three groups: single MEA, MEA blends, and solvents other than MEA. The catalysts' activity in enhancing CO2 desorption was in the sequence: carbon rich materials > metal oxides > organic frameworks > molecular sieves > composite materials > clay minerals. Different levels of time, temperature, CO2 loading, and catalysts to solvent ratio were found to be significant in CO2 desorption. The catalyst characteristics including mesoporous surface area, total surface area, pore diameter, microporous surface area, total acidity, and B/L ratio influenced CO2 desorption. For relative heat duty of solvent regeneration, Lewis acid sites and B/L ratio were found to be significant. This study offers the first systematic enquiry and a detailed statistical analysis to support future research on sustainable approaches for solid acid catalysts synthesis to enhance CO2 capture.