Publications

The importance of ammonia synthesis under mild conditions is increasing due to growing interest in ammonia for large-scale applications of renewable energy storage and utilization. Being one of the most investigated reactions in heterogeneous catalysis, multi-dimensional literature data are available for this reaction as a base to explore new catalysts. Machine learning (ML) can be applied to develop models using existing literature data. However, ML models developed only from literature data may not be able to efficiently predict or suggest new catalyst formulations without additional experimental data. Herein, we present an active learning (AL) framework for accelerating the discovery of novel ammonia synthesis catalysts initiated by literature data to explore a pre-determined search space based on domain knowledge efficiently. This framework generates and selects features for the ML model to capture the effects of catalyst preparation variables, kinetics, thermodynamics, support, and interactions between Ru, promoter, and the support for data mined from literature. Experimental results showed that the AL framework could discover novel catalysts that exceeded the activity of many state-of-the-art catalysts. AL reduced the number of experiments necessary to reach the best catalyst in the search space by 50%, even when no training data related to the best catalyst exists. Furthermore, AL gave insight into the properties of the catalysts that contribute to higher ammonia synthesis activity.

In order to achieve the net-zero carbon emission target in 2050 by reducing the use of fossil fuels and their environmental impact, ammonia is increasingly considered as a key energy carrier for the future energy transition. Industrial-scale NH₃ synthesis via the Haber–Bosch process revolutionized global agriculture and industry. However, this process runs under harsh reaction conditions and is therefore very energy-intensive and releases vast quantities of CO₂; this leaves ample room for improvement by developing highly efficient catalysts that would allow the process to take place under mild reaction conditions. Since the end of the 20th century, ammonia synthesis has been extensively investigated by employing Ruthenium (Ru)-based catalysts (as second-generation catalysts in this process) because of their high activity and outstanding thermal stability. However, their practical applications are hindered by the extremely low abundance and high cost of the precious Ru metal and the poor understanding of the nature of the Ru catalytic mechanism. Single-atom (SA) systems have emerged as a promising class of catalysts that could enable the highly active and selective conversion of N₂ to NH₃. Notably, reducing the Ru particle size to the sub-nanometer and SA scale could dramatically enhance the catalytic efficiency in NH₃ synthesis by increasing the number of active sites over the surface of catalysts and adjusting the structure sensitivity. This review briefly introduces the fundamental mechanism of NH₃ synthesis over Ru-based catalysts and the general parameters for assessing their catalytic activity. Moreover, we discuss in-depth mechanistic studies and recent advances in Ru SA catalysts as an efficient platform for ammonia synthesis by thermocatalysis, photocatalysis, and electrocatalysis. We also review density functional theory calculations investigating the NH₃ yield of the reported catalysts. Finally, we discuss economical aspects and future perspectives that may provide solutions for improving NH₃ production. The present review aims to provide a comprehensive understanding of the advancement in the catalytic ammonia synthesis field and serves as a guide for the future design of highly efficient SA-embedded catalysts.

Without effective management, the steady increase of waste plastics threatens environmental well-being and ecological balance. Plastic up/recycling is a promising solution but has many challenges. In this work, catalytic cracking of polypropylene glycol (PPG) was investigated at varying reaction temperatures of 350–550 °C under nitrogen and steam, using H-ZSM-5 zeolites with different silica-to-alumina (SiO₂/Al₂O₃) ratios of 23:1 and 50:1. The catalysts were assessed through physisorption, chemisorption, solid-state magic-angle spinning nuclear magnetic resonance spectrometry, thermogravimetry, electron microscopy, and positron annihilation lifetime spectroscopy. Extra framework aluminum and Lewis-to-Bronsted acid site ratios were found to play a significant role in the selectivity towards propionaldehyde, where values ∼ 80 % could be reached. In addition, a possible pathway for the PPG cracking reaction was proposed, which may lead to a better understanding of PPG and waste plastic decomposition via zeolite-based catalysts.

During co-pyrolysis of biomass with plastic waste, bio-oil yields (BOY) could be either induced or reduced significantly via synergistic effects (SE). However, investigating/ interpreting the SE and BOY in multidimensional domains is complicated and limited. This work applied XGBoost machine-learning and Shapley additive explanation (SHAP) to develop interpretable/ explainable models for predicting BOY and SE from co-pyrolysis of biomass and plastic waste using 26 input features. Imbalanced training datasets were improved by synthetic minority over-sampling technique. The prediction accuracy of XGBoost models was nearly 0.90 R² for BOY while greater than 0.85 R² for SE. By SHAP, individual impact and interaction of input features on the XGBoost models can be achieved. Although reaction temperature and biomass-to-plastic ratio were the top two important features, overall contributions of feedstock characteristics were more than 60 % in the system of co-pyrolysis. The finding provides a better understanding of co-pyrolysis and a way of further improvements.

Bio-oils produced from torrefaction or pyrolysis of biomass constitute an under-utilized product that generally requires complicated processing. The high acidity and water content pose storage and transportation issues, and the complex nature of the organic species makes utilization as a chemical feedstock troublesome. Most methodologies to separate and upgrade bio-oils involve multiple distinct steps and solvent-based extractions, complicating the process and adding cost. In this work, we demonstrate a simple one-step solution using ammonia to separate the aqueous and organic phases of a model bio-oil. This process produces an aqueous phase that contains ammonium species that could be utilized as a fertilizer and an organic phase that can be used as an additive for transportation fuels or could be burned to produce electricity.

Among renewable and sustainable energy resources, biomass plays a vital role. Agricultural residues/wastes, energy crops, and lignocellulosic biomass could potentially be major feedstocks for biorefineries. In Thailand, one of the most interesting energy crops is hybrid giant Juncao grass (GJG) or Pennisetum purpureum x Pennisetum typhoideum. GJG can be easily grown and has relatively high yields under tropical climates. Herein, conversion of GJG to biofuels via hydrothermal liquefaction (HTL) was investigated using batch reactors under varying reaction temperatures of 250–350 °C and biomass-to-deionized water concentrations of 15–25 wt% at a fixed residence time of 30 min. Changes in temperature and GJG-to-deionized water concentration were found to markedly affect the yields and distribution of products from HTL of GJG. Yields of the liquid product, or bio-oil, can be up to 50 wt% at 350 °C and 25 wt% GJG-to-deionized water concentration. The yields of solid char and gas products fluctuated within 10–25 wt% and 30–45 wt%, respectively. Higher heating values of the resulting bio-oil and char were remarkably better than those of the raw material. An energy recovery of over 50% from the bio-oil, as well as about 35% from the char, can be obtained. By gas chromatograph-mass spectrometry and nuclear magnetic resonance, the bio-oil obtained was found to be a complex chemical mixture, consisting mainly of phenols, nitrogenous compounds, aliphatic compounds, ketones, carboxylic acids, and aldehydes. The finding is useful in future utilization of GJG via HTL for biofuel and/or biochemical production.

The CuxRh_3–x(BTC)₂ catalyst (abbreviated CuRhBTC, BTC^3– = benzene tricarboxylate) provides excellent dispersion of active metal sites coupled with well-defined, robust structures for propylene hydrogenation reactions. This material therefore serves as a unique prototype for understanding catalytic activity in metal organic frameworks (MOFs). The mechanism of gas-phase hydrogenation at the bimetallic metal nodes of a MOF has been investigated in detail for the first time using in situ spectroscopy and diffraction experiments combined with density functional theory (DFT) calculations. The reaction occurs via a cooperative process in which the metal and linker sites play complementary roles; specifically, H₂ is dissociated at a Rh²⁺ site with a missing Rh–O bond, while protonation of the decoordinated carboxylate linker stabilizes the active sites and promotes H2 dissociation. In situ X-ray diffraction experiments show that the crystalline structure of the MOF is retained under reaction conditions at 20–100 °C. In situ Raman spectroscopy and diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) experiments demonstrate that propylene adsorbs at both Rh²⁺ and Cu²⁺ sites via π bonding. Cu²⁺ is catalytically inactive, but at Rh²⁺ sites, a propyl intermediate is observed when H2 is introduced into the propylene feed. Furthermore, the appearance of the O–H stretch of COOH at ∼3690 cm^–1 in the DRIFT spectra is characteristic of defects consisting of missing Rh–O bonds. These experimental results are in general agreement with a reaction mechanism proposed by DFT, in which the decoordinated carboxylate linker is protonated, and the active Rh²⁺ site remains available for readsorption of reactants in the subsequent catalytic cycle.

In this review, we highlight how design of experiments and machine learning can be utilized in catalysis to help optimize reaction conditions, catalysts, and predict new catalyst formulations. An overview of how the techniques work is presented, and the advantages and disadvantages of the techniques are discussed. We showcase the ability to extract meaningful knowledge utilizing small experimental data sets and the recent advancements in the use of machine learning in catalysis. We conclude the review by presenting a potential method to combine the benefits of both machine learning and design of experiments to help accelerate catalyst discovery and optimization.

Torrefaction is a treatment process for converting biomass to high-quality solid fuels. The investigation and interpretation of this process on highly dimensional, non-linear relationships as large datasets are limited. In this work, machine learning (ML) in combination with collaborative game theory (Shapley additive explanation, SHAP) was applied to develop an interpretable model in predicting solid yields (SY) and higher heating values (HHV) of solid products from biomass torrefaction using 18 independent input features from operating conditions, feedstock characteristics and torrefaction reactor properties. Three novel ML algorithms were evaluated, based on 10-fold cross-validation, with 5 different sets of input features. A gradient tree boosting (GTB) model was found to have the highest prediction accuracy R² of 0.93 with root mean square error (RMSE) of 0.06 for SY while about 0.91 R² with 0.79 RMSE for HHV. With the powerful SHAP algorithm, a new framework was proposed to interpret/explain the GTB model performance and highlight the highly influential features for the system of biomass torrefaction in both local and global points of view. Interactions for any pair of the features on the GTB model can be achieved. This application of ML with SHAP is a useful tool for researchers on biomass conversion.