Modeling, optimization and understanding of adsorption process for pollutant removal via machine learning: Recent progress and future perspectives

Coupling machine learning and theoretical models to compare key properties of biochar in adsorption kinetics rate and maximum adsorption capacity for emerging contaminants

Insights into key properties of biochar with a fast adsorption rate and high adsorption capacity are urgent to design biochar as an adsorbent in pollution emergency treatment. Machine learning (ML) incorporating classical theoretical adsorption models was applied to build prediction models for adsorption kinetics rate (i.e., K) and maximum adsorption capacity (i.e., Qm) of emerging contaminants (ECs) on biochar. Results demonstrated that the prediction performance of adaptive boosting algorithm significantly improved after data preprocessing (i.e., log-transformation) in the small unbalanced datasets with R2 of 0.865 and 0.874 for K and Qm, respectively. The surface chemistry, primarily led by ash content of biochar significantly influenced the K, while surface porous structure of biochar showed a dominant role in predicting Qm. An interactive platform was deployed for relevant scientists to predict K and Qm of new biochar for ECs. The research provided practical references for future engineered biochar design for ECs removal.

Deep learning and big data mining for Metal-Organic frameworks with high performance for simultaneous desulfurization and carbon capture

Carbon capture and desulfurization of flue gases are crucial for the achievement of carbon neutrality and sustainable development. In this work, the "one-step" adsorption technology with high-performance metal-organic frameworks (MOFs) was proposed to simultaneously capture the SO2 and CO2. Four machine learning algorithms were used to predict the performance indicators (NCO2+SO2, SCO2+SO2/N2, and TSN) of MOFs, with Multi-Layer Perceptron Regression (MLPR) showing better performance (R2 = 0.93). To address sparse data of MOF chemical descriptors, we introduced the Deep Factorization Machines (DeepFM) model, outperforming MLPR with a higher R2 of 0.95. Then, sensitivity analysis was employed to find that the adsorption heat and porosity were the key factors for SO2 and CO2 capture performance of MOF, while the influence of open alkali metal sites also stood out. Furthermore, we established a kinetic model to batch simulate the breakthrough curves of TOP 1000 MOFs to investigate their dynamic adsorption separation performance for SO2/CO2/N2. The TOP 20 MOFs screened by the dynamic performance highly overlap with those screened by the static performance, with 76 % containing open alkali metal sites. This integrated approach of computational screening, machine learning, and dynamic analysis significantly advances the development of efficient MOF adsorbents for flue gas treatment.

Machine learning vs. statistical model for prediction modeling and experimental validation: Application in groundwater permeable reactive barrier width design

Permeable reactive barrier (PRB) is an effective in-situ technology for groundwater remediation. The important factors in PRB design are the width and reactive material. In this study, the beaded coal mine drainage sludge (BCMDS) was employed as the filling material to adsorb arsenic pollutants in groundwater, aiming to design the width of PRB. The design methods involving traditional continue column experiments and empirical formulas, as well as machine learning (ML) predictions and statistical methods, which are compared with each other. Traditional methods are determined based on breakthrough curves under several conditions. ML method has advantages in predicting the width of mass transfer zone (WMTZ), which simultaneously consider the characteristics of material, pollutant, and environmental conditions, with data collected from articles. After data preprocessing and model optimizing, selected the XGBoost algorithm based on the high accuracy, which shows good prediction for WMTZ (R2 = 0.97, RMSE = 0.15). The experimentally derived WMTZ values were also used to validate the predictions, demonstrating the ML low error rate of 7.04 % and the feasibility. Subsequent statistical analysis of multiple linear regression (MLR) showed the error rate of 39.43 %, interpret superiority of ML due to the complexity of influencing factors and the insufficient precision of math regression. Compared to traditional width design methods, ML can improve design efficiency and save experimental time and manpower. Further expansion of the dataset and optimization of algorithms could enhance the accuracy of ML, overcoming existing limitations and gaining broader applications.

Enhanced mitigation of acidic and basic dyes by ZnO based nano-photocatalysis: current applications and future perspectives

Dye wastewater possess immense toxicity with carcinogenic properties and they persist in environment owing to their stability and resistance to chemical and photochemical changes. The bio degradability of dye-contaminated wastewater is low due to its complex molecular structure. Nano-photocatalysts based on zinc oxide are reported as one of the effective metal oxides for dye remediation due to their photostability, enhanced UV and visible absorption capabilities in an affordable manner. An electron-hole pair forms when electrons in the valence band of ZnO nano-photocatalyst transfer into the conduction band by absorbing UV light. The review article presents a detailed review on ZnO applications for treating acidic and basic dyes along with the dye degradation performance based on operating conditions and photocatalytic kinetic models. Several acidic and basic dyes have been shown to degrade efficiently using ZnO and its nanocomposites. Higher removal percentages for crystal violet was reported at pH 12 by ZnO/Graphene oxide catalyst under 400 nm UV light, whereas acidic dye Rhodamine B at a pH of 5.8 was degraded to 100% by pristine ZnO. The mechanism of action of ZnO nanocatalysts in degrading the dye contamination are reported and the research gaps to make these agents in environmental remediation on real time operations are discussed.

Application of machine learning approach (artificial neural network) and shrinking core model in cobalt (II) and copper (II) leaching process

The leaching laboratory experiment uses the artificial neural network (ANN) to predict and evaluate copper and cobalt recovery. This study aimed to evaluate the efficacy of using the shrinking core model in conjunction with an artificial neural network (ANN) as part of a machine learning strategy to improve the leaching process of cobalt (II) and copper (II). The numerous factors in the leaching process, such as acid concentration, leaching time, temperature, soil-to-solution ratio, and stirring speed, are adjusted using an ANN with several layers, feed-forward, and back-propagation learning methods. These variables are in charge of the high cobalt recovery during the reduced sulfuric acid leaching procedure. The ANN algorithm has 10 hidden layers, 5 input variables describing the leaching parameters, and two neurons as output layers corresponding to copper and cobalt leaching recovery. The optimum conditions were found to be acid concentration of 100 g/L, leaching duration 120 min, temperature 55 °C, soil-to-solution ratio of 1:40 g/mL, and stirring speed 300 rpm. The optimized trained neural networks tested, trained, and validated steps are represented by R2 values of 0.94, 0.99, 0.97, and 0.97, respectively, equating to 97.5% copper recovery and 95.4% cobalt recovery.

Integrating molecular descriptors for enhanced prediction: Shedding light on the potential of pH to model hydrated electron reaction rates for organic compounds

Hydrated electron reaction rate constant (ke-aq) is an important parameter to determine reductive degradation efficiency and to mitigate the ecological risk of organic compounds (OCs). However, OC species morphology and the concentration of hydrated electrons (e-aq) in water vary with pH, complicating OC fate assessment. This study introduced the environmental variable of pH, to develop models for ke-aq for 701 data points using 3 descriptor types: (i) molecular descriptors (MD), (ii) quantum chemical descriptors (QCD), and (iii) the combination of both (MD + QCD). Models were screened using 2 descriptor screening methods (MLR and RF) and 14 machine learning (ML) algorithms. The introduction of QCDs that characterized the electronic structure of OCs greatly improved the performance of models while ensuring the need for fewer descriptors. The optimal model MLR-XGBoost(MD + QCD), which included pH, achieved the most satisfactory prediction: R2tra = 0.988, Q2boot = 0.861, R2test = 0.875 and Q2test = 0.873. The mechanistic interpretation using the SHAP method further revealed that QCDs, polarizability, volume, and pH had a great influence on the reductive degradation of OCs by e-aq. Overall, the electrochemical parameters (QCDs, pH) related to the solvent and solute are of significance and should be considered in any future ML modeling that assesses the fate of OCs in aquatic environment.

Assessment and machine learning prediction of heavy metals fate in mining farmland assisted by Positive Matrix Factorization

The accurate pollutant prediction by Machine Learning (ML) is significant to efficient environmental monitoring and risk assessment. However, application of ML in soil is under studied. In this study, a Positive Matrix Factorization (PMF) assisted prediction method was developed with Support Vector Machine (SVM) and Random Forest (RF) for heavy metals (HMs) prediction in mining farmland. Principal Component Analysis (PCA) and Redundancy Analysis (RDA) were selected to pretreat data. Experiment results illustrated Cd was the main pollutant with heavy risks in the study area and Pb was easy to migrate. The method effects of HMs total concentration predicting were PMF > Simple > PCA > PCA - PMF, and RF predicted better than SVM. Data pretreatment by RDA prior inspection improved the model results. Characteristic HMs Tessier fractions prediction received good effects with average R value as 0.86. Risk classification prediction performed good in Cd, Cu, Ni and Zn, however, Pb showed weak effect by simple model. The best classifier method for Pb was PMF - RF method with relatively good effect (Area under ROC Curve = 0.896). Overall, our study suggested the combination between PMF and ML can assist the prediction of HMs in soil. Spatial weighted attribute of HMs can be provided by PMF.

Sustainable wastewater purification with crab shell-derived biochar: Advanced machine learning modeling & experimental analysis

Detoxifying ecologically persistent dyes is vital for environmental and human well-being. Herein, crabshell waste is transformed into porous carbon (CB900) through pyrolysis, achieving a remarkable removal rate of 90.5% (CR-RR) and adsorption capacity (∼256.36 mg g-1, qCR). Employing XGBoost modeling, with a robust R2 ∼0.996, proved its superiority over others in predicting CR adsorption. PSO-XGB optimization led to an optimal configuration: 0.051 g adsorbent, 460.56 mg L-1 CR concentration, pH 3.16, and a 94.01 min contact time, resulting in 68.39% CR-RR and 822.15 mg g-1 qCR, simultaneously; sensitivity analysis unveiled the pivotal role of pH and adsorbent dose. CB900 exhibited physical, spontaneous, endothermic following both Langmuir and Freundlich isotherms. Remarkably, CB900 effectively eliminated various contaminants, including chromium and sulfasalazine antibiotic. Pilot-scale CB900 production cost via pyrolysis was $8.5/kg, a fraction of commercial powdered activated carbon, underscoring its economic viability and potential as a sustainable solution for the elimination of toxic contaminants from aqueous environments.

Unraveling the roles of microporous and micro-mesoporous structures of carbon supports on iron oxide properties and As (V) removal performance in contaminated water

This study investigates the impact of microporous (SP-C) and micro-mesoporous carbon (DP-C) supports on the dispersion and phase transformation of iron oxides and their arsenic (V) removal efficiency. The research demonstrates that carbon-supported iron oxide sorbents exhibit superior As(V) uptake capacity compared to unsupported Fe2O3, attributed to reduced iron oxide crystallite sizes and As(V) adsorption on carbon supports. Maximum As(V) uptake capacities of 23.8 mg/g and 18.9 mg/g were achieved for Fe/SP-C and Fe/DP-C at 30 wt% and 50 wt% iron loading, respectively. The study reveals a nonlinear relationship between As(V) sorption capacity and iron oxide crystallite size after excluding As(V) adsorption capacity on carbon supports, suggesting the iron oxide phase (Fe3O4) plays a role in determining adsorption capacity. Iron oxide-loaded DP-C sorbents exhibit faster adsorption rates at low As(V) concentrations (5 mg/L) than SP-C sorbents due to their bimodal pore structure. Adsorption behavior varies at higher As(V) concentrations (45 mg/L), with Fe/DP-C reaching maximum capacity more slowly due to limited available adsorptive sites. All adsorbents maintained near-complete As(V) removal efficiency over five cycles. The findings provide insights for designing more efficient adsorbents for As(V) removal from contaminated water sources.

Yesterday, Today, and Tomorrow. Evolution of a Sleeping Beauty: The Freundlich Isotherm

The name of Herbert Freundlich is commonly associated with a power relationship for adsorbed amount of a substance (C_ads) against the concentration in solution (C_sln), such that C_ads = KC_slnⁿ; this isotherm (together with the Langmuir isotherm) is considered to be the model of choice for correlating the experimental adsorption data of micropollutants or contaminants of emerging concern (pesticides, pharmaceuticals, and personal care products), but it also concerns the adsorption of gases on solids. However, Freundlich's 1907 paper was a "sleeping beauty", which only started to attract significant citations from the early 2000s onward; moreover, these citations were too often wrong. In this paper, the main steps in the historical developments of Freundlich isotherm are identified, along with a discussion of several theoretical points: (1) derivation of the Freundlich isotherm from an exponential distribution of energies, leading to a more general equation, based on the Gauss hypergeometric function, of which the power Freundlich equation is an approximation; (2) application of this hypergeometric isotherm to the case of competitive adsorption, when the binding energies are perfectly correlated; and (3) new equations for estimating the Freundlich coefficient K_F from physicochemical properties such as the sticking surface or probability. From new data treatment of two examples from the literature, the influence of several parameters is highlighted, and the application of linear free-energy relationships (LFER) to the Freundlich parameters for different series of compounds is evoked, along with its limitations. We also suggest some ideas that may be worth exploring in the future, such as extending the range of applications of the Freundlich isotherm by means of its hypergeometric version, extending the competitive adsorption isotherm in the case of partial correlation, and exploring the interest of the sticking surfaces or probabilities instead of K_F for LFER analysis.