Adsorption最新文献

This work introduces a robust machine learning framework for accurately predicting CO₂ adsorption capacity (AC) in porous adsorbent materials, offering a significant advancement over conventional experimental and analytical approaches. Specifically, we combine a cascaded forward neural network (CFNN) with advanced local and global explainable artificial intelligence (XAI) techniques to achieve both high predictive accuracy and interpretability. The model is trained on an expanded dataset exceeding 2700 data points, encompassing diverse families, such as metal-organic frameworks (MOFs), zeolites, porous organic polymers (POP), and carbon-based materials (CBM). The CFNN is further optimized through three distinct learning algorithms, namely Levenberg–Marquardt, Bayesian Regularization, and Scaled Conjugate Gradient. Quantitatively, the CFNN-LM model achieved a determination coefficient (R²) of 0.9991 and a root mean square error (RMSE) of 0.0659, outperforming alternative ML frameworks and surpassing the models previously reported in the literature. Moreover, by integrating SHAP and LIME for global and local interpretability, we provide physical insights into the factors influencing adsorption performance under varying conditions. Beyond modeling accuracy, this framework offers tangible applications enabling rapid, low-cost pre-screening of adsorbents for industrial sectors, while also equipping researchers with a transparent and scalable tool for accelerating material discovery. By bridging the gap between computational intelligence and practical deployment, this work contributes a scientifically credible and operationally valuable asset to the evolving landscape of carbon capture solutions.

The worldwide water issue is mostly caused by water pollution, which poses serious risks to human health and environmental sustainability even with tremendous technological breakthroughs. Many inorganics and newly discovered organic contaminants, including heavy metals, synthetic dyes, petroleum derivatives, hazardous compounds, and industrial effluents, currently make up the bulk of waterborne pollutants. Highly effective adsorbents materials must be developed in order to remediate these contaminants effectively. Among many different types of absorbent present in the current scenario, Metal organic framework (MOF) has emerged as a particularly promising class due to their exceptional physicochemical properties. MOF being a porous material, having an exceptionally large surface area of about 1000 to 10,000 m²/g and precises making it to easily remove the water contaminants. These unique properties make MOFs a highly promising material for water treatment applications. By incorporating MOFs into traditional water filtration system, the effectiveness and selectivity of contamination removal might be greatly increased. In this review the detail characteristics and performance of various MOF-based composites, along with their specific applications in water treatment, are discussed. The study also addresses potential changes to enhance stability and scalability, future possibilities, and the main obstacles to the widespread use of the MOFs for effective and sustainable water filtrations system.

Graphical Abstract

The sorption and release of gases in complex porous materials often involves multiple simultaneous processes. It is difficult to distinguish the dynamics of individual processes from experiments that only allow access to the combined overall dynamics of all processes taking place. This works shows for the case of dimethyl sulfoxide (DMSO) vapor interacting with paper as a porous matrix that it is possible to distinguish experimentally the dynamics of the sorption and release processes of two populations of differently sorbed DMSO molecules, even though the temporal evolution of the sorbed concentration shows no evidence of multiple processes. The key to separating the processes lies in the release behavior. DMSO release from paper saturated with DMSO is a superposition of short- and long-term processes, with release rates of 3(times)10(^{-4})s(^{-1}) and 1.1(times)10(^{-6})s(^{-1}), respectively. Such different rates indicate the presence of two types of sorbed populations with distinct release dynamics. This inspires a strategy to separate the sorption dynamics of the two populations as well. A sorption experiment was designed to determine the concentration of DMSO sorbed in the paper at a certain exposure time, and again after the short-term release was complete. This procedure yields the concentration of each sorbing species as a function of time. Successfully separating the process dynamics is important for clarifying the nature of the processes, such as determining activation energies, and for correctly describing the temporal evolution of sorption and release mathematically. This strategy may help reveal the dynamics of simultaneous, mutually obscuring processes in other systems as well.

Graphical Abstract

Designing adsorption processes requires knowledge of the adsorption isotherms. Measuring accurate isotherms is time consuming and inefficient equidistant points are usually chosen. Here, we combine isotherm measurements with Model-Based Design of Experiments to iteratively determine isotherm models with less experimental effort, while maintaining high model accuracy. Our joint approach combining isotherm model discrimination and parameter precision is validated by thermo-gravimetric experiments for the adsorption pairs Lewatit VP OC 1065 with (textrm{CO}_2) and (textrm{H}_2textrm{O}) and BAM-P109 with (textrm{H}_2textrm{O}), covering isotherm Types I, III, and V. Results show that the experimental effort could be reduced between 70–81%. Furthermore, the framework scheduled measurements for Lewatit VP OC 1065/(textrm{H}_2textrm{O}) to discriminate its isotherm Type between II or III, devoid of our bias as experimenters. Overall, our approach demonstrates potential to streamline the identification of adsorption isotherms while enabling more efficient and unbiased model development.

The prediction of adsorption in carbonaceous materials via molecular simulation is a challenging task. Unlike zeolites and MOFs, carbons are amorphous. Homogeneous activated carbon models are commonly used for characterization. Here, we demonstrate that, despite their relative success, homogeneous models are inadequate for predicting the adsorption of SO₂, a contaminant in flue gases that interferes with CO₂ capture processes. To address this issue properly, we have developed a new set of SO₂ isotherms based on a heterogeneous reactive model (rMD). The isotherms were calculated using the Monte Carlo method in the grand canonical ensemble. Two samples of commercial carbons, C141 and WV1050, had their SO₂ adsorption capacities predicted by combining the regular N₂ at 77 K and CO₂ at 273 K characterization with the new set of SO₂ isotherms in the rMD model. By incorporating the heterogeneous model in the characterization process, we have considerably improved the agreement between experimental and simulated isotherms for both carbons. The use of the rMD heterogeneous kernel will enable new adsorption predictions for adsorbates where previous homogeneous models have failed.

The decomposition of NO dimers on metal surfaces is a critical step in nitrogen oxide reduction technologies. In this study, we investigate the Cu(100) surface as a catalytic platform for this reaction using ab initio molecular dynamics (AIMD) simulations, based on density functional theory (DFT) with van der Waals corrections. A detailed analysis of the energy partitioning in the desorbed (hbox {N}_2hbox {O}) molecule reveals that the activation energy associated with N–O bond cleavage is primarily redistributed into the rotational mode, followed by the translational mode, and small fraction of the energy is transferred into vibrational excitation and dissipated into surface modes. This mode-specific energy redistribution suggests that enhancing the excitation of the rotational degree of freedom could improve the rate and efficiency of NO reduction. In addition, we evaluate the adsorption behavior of the reaction products, (hbox {N}_2hbox {O}) and atomic O, on Cu(100) using DFT calculations. (hbox {N}_2hbox {O}) is found to preferentially adopt a bent chemisorbed geometry, while atomic O favors adsorption at the fourfold hollow site. These stable configurations serve as the final state for the NO dimer decomposition pathway. While for the initial state is flat-ONNO dimer, which was discussed comprehensively in our previous work. Together, these results offer fundamental insights into the desorption dynamics, energy transfer mechanisms, and product–surface interactions that govern NO decomposition on copper catalysts, and may inform the rational design of more effective, non-PGM catalytic systems for environmental applications.

Graphical abstract

Water adsorption on Aluminium Fumarate (Al-Fum), a promising MOF for low-temperature heat-driven applications, exhibits complex isotherm shapes that challenge conventional modeling approaches. In this work, we develop a physically grounded dual-site adsorption model that combines Langmuir adsorption on high-energy sites with cooperative water clustering based on association theory. The model reproduces full isotherm profiles across a broad temperature range (20–90 °C) using eight temperature-independent parameters with clear physical significance. The model achieves excellent agreement (adjusted R² = 0.9957), accurately capturing both low-pressure concavity and the S-shaped transition. It also enables the calculation of isosteres and isosteric heats of adsorption, revealing distinct thermodynamic regimes governed by the two adsorption mechanisms. To demonstrate system-level relevance, the model is applied to a typical intermittent adsorption cooling cycle operating at 10/30/60°C. Al-Fum delivers a thermal COP of 0.785 (excluding heat exchangers’ sensible heat) and a cycled mass of 177.5 g kg⁻¹, outperforming benchmark materials. This work provides a robust, physically sound modeling tool for the design and optimization of advanced adsorption-based thermal systems.

Many studies that use an automated sorption balance to determine a water vapor sorption isotherm for wood collect data until the moisture content change is less than or equal to 0.002% min⁻¹ (20 µg g⁻¹ min⁻¹). This stop criterion has been claimed to give errors in equilibrium moisture content (EMC) predictions of less than 0.001 g g⁻¹ but over the past 10 years, studies have shown that the actual errors can be greater than 0.01 g g⁻¹ because the measurements are stopped well before equilibrium is reached. Despite the large errors associated with this stop criterion, it remains popular due to the speed at which isotherms can be measured. This paper utilizes data from a worldwide interlaboratory study on automated sorption balances to develop a correction method for estimating EMC of western larch (Larix occidentalis Nutt.) from the moisture content corresponding to the 20 µg g⁻¹ min⁻¹ criterion. The study uses data from 72 relative humidity absorption steps with hold times of 7–10 days from 21 different laboratories and eight different instrument models. EMC is defined based on the inherent mass stability of automated sorption balances determined in the first part of this interlaboratory study. On average the sorption process is less than 80% complete when the 20 µg g⁻¹ min⁻¹ criterion is reached, resulting in a mean absolute error (MAE) of 0.006 g g⁻¹. The correction equation for estimating EMC reduces the MAE to 0.001 g g⁻¹. The analysis presented in this paper, along with the correction equation, can be considered for certain use cases to reduce systematic errors and shorten measurement times.

Using the Lattice Boltzmann method (LBM) for simulation of fluid dynamics in complex systems such as adsorption with the advection terms of scalar fields (concentration and temperature distribution), different approaches of advection coupling to the fluid motion can be proposed: “Active or Passive Scalers”. In the present study, the usefulness of active or passive scalars in simulation of an adsorption bed using LBM at different operating conditions such as temperature, pressure and feed flow rate were investigated. In the active scalar approach in LBM, the collision operator in the Boltzmann transport equation consists of two terms: the self and cross collision. On the other hand, the collision term for a passive scalar comes from the Chapman relationship. As the cross collision term in active scalar has an inverse relationship with diffusion coefficient, the effect of this term reduces in gas systems such as adsorption with a high diffusion coefficient; thus, the active and passive approaches become similar. It is obvious that in systems with a lower diffusion coefficient (liquid systems), the cross collision term in collision operator in LBM is high; therefore, it is expected that the active approach with more precise results deviates from the passive approach. Results showed that in most cases, the average relative error compared to experimental data was less in active scalar than in passive scalar approach, indicating that the active scalar approach predicts the adsorption behavior with higher accuracy in comparison with the passive approach.

Nowadays, many pollutants, especially polycyclic-aromatic-hydrocarbons (PAHs), are a high threat to humans as well as animals due to their carcinogenic behaviour. Therefore, in the present research work, we studied the adsorption behaviour of three different PAHs, namely anthracene, benzo[a]pyrene, and chrysene, on monolayer beta phosphorous nitride nanosheet (β-PN-sheet) using the density-functional-theory (DFT) method. Besides, low-dimensional material possesses many features, including a large active surface region and the electronic properties can be fine-tuned easily, which are the main requirements for chemical sensors. Initially, the structural stability of the β-PN-sheet is confirmed with the support of phonon-band-maps and formation energy. Furthermore, the electronic properties of β-PN-sheet are investigated using band maps and projected-density-of-states (PDOS) maps. We also studied the influence of compressive strain on the electronic properties as well as on the adsorption properties of the β-PN-sheet. The computed band gap of β-PN-sheet slightly increases from 3.355 eV to 3.537 eV owing to the compressive strain. The adsorption behaviour of PAH pollutants on β-PN-sheet is studied with significant factors, namely adsorption energy, relative band gap changes, and Mulliken population analysis. Furthermore, the adsorption of PAHs on β-PN-sheet gets slightly improved with applied compressive strain, and the adsorption energy falls in the scale of physisorption (−0.292 eV to −0.404 eV). Furthermore, a fast recovery time is obtained while desorbing PAH pollutants from the β-PN-sheet. The sensing response of β-PN-sheet to PAHs gets enhanced by applying compressive strain.