Journals ITB
JETS

Journal of Engineering and Technological Sciences
  1. Home
  2. Archives
  3. Vol 57 (2025) Issue 4
  4. Articles

Deeper Insight into the Rational Design and Synthesis of Zeolites Revealed by Machine Learning: A Mini Review

Published: | DOI: 10.5614/j.eng.technol.sci.2025.57.4.4 | Vol 57, Issue 4 (2025)

Authors

Abstract

Zeolites are widely applied in various fields owing to their outstanding properties. However, our understanding on the nature of zeolite synthesis is not completed yet due to its high dimensional parameters. Machine learning has the ability to unravel fundamental relationships between complex parameters and predict the possible outcomes; thus, it can potentially reveal the nature of zeolite synthesis. This mini review highlights the current use of machine learning to comprehend the black box issue in zeolite synthesis. Conventional syntheses of zeolite were also elaborated to showcase the gap between traditional methods and machine learning approaches in zeolite synthesis. The future prospects of machine learning applications in zeolite synthesis are also discussed. This mini-review may bring crucial insights on the zeolite synthesis process.

Keywords

Rational design; Biochemical engineering; Engineering drawing; Computer science; Process engineering; Artificial intelligence; Chemical engineering; Mathematics; Engineering; Materials science; Nanotechnology

Introduction

Zeolites have shown pivotal roles in research and industrial sectors owing to their extraordinary properties and great applicability. The high thermal stability, large surface area, and molecular shape selectivity of zeolites allow them to be applied as catalysts (Bae et al., 2021; Kadja, Azhari, Mardiana, et al., 2021; Li et al., 2023; J. Wang et al., 2022; H. Zhang et al., 2022), adsorbents (Fischer, 2020; Hewitt et al., 2022; Mai et al., 2022; Mguni et al., 2022; Pérez-Botella et al., 2019; Shobuke et al., 2022), and ion exchangers (Aragaw & Ayalew, 2019; Campanile et al., 2022; Chen et al., 2016). Moreover, zeolite exhibits incredible performance in the refinery industry (Primo & Garcia, 2014; L. Zhang et al., 2022, 2023), biomass conversion (Bornes et al., 2023; Mardiana et al., 2022; Perego et al., 2017), CO2 capture (Karka et al., 2019; Murge et al., 2019; Thakkar et al., 2016), and water purification (Mahmoodi & Saffar-Dastgerdi, 2019; Tankersley et al., 2020).

Zeolites are composed by repeating TO4 (T represents Si or Al atom) tetrahedra aluminosilicate structures, which shape into different channels, cages, and pore morphologies. This unique structure creates well-defined pores in the zeolite structures. According to the International Zeolite Association, 255 zeolite frameworks have been discovered to date (Gandhi & Hasan, 2022). The synthesis of zeolite was first introduced by Richard Barrer, who reacted natural minerals

with an alkaline solution at high temperatures, i.e., 170-270 ℃ (Cundy & Cox, 2003). Generally, zeolites are synthesized using the hydrothermal method at a temperature of 60-200 ℃ for 1 to 20 days (Rahmah et al., 2023).

A large number of studies about zeolite synthesis focuses on property improvement. In this sense, the research about the synthesis of zeolites aims to control the crystal size and morphology as well as pore size controlling to overcome the diffusion limitation, in particular, hierarchical zeolite synthesis (Al-Ani et al., 2020; Graça et al., 2018; Jia et al., 2019; Kadja, Suprianti, et al., 2020; Khan et al., 2019; L. Wang et al., 2015). Currently, the innovations of zeolite synthesis research are shifting to develop synthesis strategies in order to suppress the production cost as well as to decrease environmental pollution. For instance, various strategies have been used for the green synthesis of zeolite, i.e., zeolite synthesis without solvent or known as solvent-free synthesis (Al-Nahari et al., 2023; Kadja, Rukmana, et al., 2021; Q. Wu et al., 2018), Organic Structure Directing Agent (OSDA)-free (Kadja, Kadir, et al., 2020; Tomita et al., 2022), solventfree and OSDA-free (Kadja, Azhari, Mukti, et al., 2021), using natural precursors, e.g., rice husk as silica (Kadja et al., 2017; Mohamed et al., 2015), and synthesis zeolite in low temperature (Kadja et al., 2016).

As aforementioned, zeolite structures, which are built by different TO4 configurations, could create different topologies. In this regard, different pore morphologies and compositions could also be fabricated (Gandhi & Hasan, 2022). Therefore, it has also become one factor that researchers put much effort into synthesizing new zeolite frameworks. Even so, understanding the zeolite synthesis process is still a big question mark. This process is affected by many factors, i.e., various synthesis conditions and numerous choices of source materials. Furthermore, the conversancy of the complex link between zeolite structures and properties still becomes a challenge. Owing to these reasons, researchers began to develop and/or utilize a tool to predict the suitable synthesis condition with desired properties.

4

Zeolite synthesis, machine learning, and zeolite and machine learning publication indexed by Scopus (April 2025).

The theoretical simulation of zeolite began in the 1970s, while the atomic simulation of the zeolite system started in the 1990s (S. Ma & Liu, 2022b). Scientists have predicted several million zeolite structures, but currently, only hundreds of them can be synthesized. Nevertheless, the density functional theory (DFT) calculation needs a high cost and longtime simulation. Thus, new approaches are required to overcome the bottleneck in zeolite synthesis.

Considering that the parameters of zeolite synthesis are complex and still poorly understood, a faster method that can optimize these intricate parameters is expected to ameliorate the understanding of zeolite synthesis (Moliner et al., 2019). Recently, the rapid development of machine learning has become a potential breakthrough in chemical research. As a part of artificial intelligence, machine learning is capable to understand the relationships between numerous variables and accurately predict the plausible outcomes. Unlike traditional computation techniques, machine learning can learn from a large amount of data without explicitly programmed. Furthermore, machine learning has the ability to produce an output extremely fast compared to conventional techniques, such as DFT. To date, machine learning has been widely applied to propose and identify new materials (Z. Yang et al., 2023). In addition, machine learning has been proven as an impressive tools in studying the complex relationship and several application (D. Ma et al., 2022). Hence, machine learning is projected to unveil the black box in zeolite synthesis system.

Figure 1 shows the number of publications related to zeolite synthesis, which has been growing year by year, along with the research on machine learning, which has increased extensively. The data trend shows a growing interest towards the use of machine learning in zeolite research. Herein, this present mini review focused on the potential of machine learning in revealing the black box issue in zeolite synthesis. It started by discussing the black box issue in conventional zeolite synthesis. Subsequently, the critical innovations in zeolite synthesis employing machine learning techniques in recent years were highlighted. At the end of the review, the future potential of machine learning in assisting chemists in the rational design and synthesis of zeolites was pointed out.

Several reviews have been published concerning the involvement of machine learning techniques in zeolite synthesis, as presented in Table 1 (Gandhi & Hasan, 2022; Kwak et al., 2021; S. Ma & Liu, 2022b, 2022a; Moliner et al., 2019; M. Wu et al., 2025). Despite growing interest in machine learning employ in zeolite synthesis, there remains a lack of comprehensive reviews which focusing in the latest advancement of machine learning potential in revealing the black box issue in zeolite synthesis. Thus, this gap highlights the significance of this mini review to provide a focused yet latest advancements discussion of unexplored avenues.

TitleFocusYearRef
Machine learning applied to zeolite synthesis: the
missing link for realizing high-throughput
discovery		Machine learning techniques to rationalize
zeolite synthesis		2019(Moliner et al., 2019)
The role of zeolite framework in zeolite stability
and catalysis from recent simulation		The theoretical insights of zeolite
framework's role in the stability and
functionality of zeolite		2021(S. Ma & Liu, 2022b)
Recent progress on Al distribution over zeolite
frameworks: linking theories and experiments		Al distribution in zeolite framework2021(Kwak et al., 2021)
Machine learning for the design and discovery of
zeolites and porous crystalline materials		The discovery of zeolites and similar
crystalline materials using ML-based
design		2021(Gandhi & Hasan,
2022)
Machine learning potential era of zeolite
simulation		The zeolite stability and the mechanism in
catalytic reaction		2022(S. Ma & Liu, 2022a)
AI-empowered digital design of zeolites: Progress,
challenges, and perspectives		AI-empowered design of zeolites in
properties prediction, simulation, design,
and synthesis zeolite		2025(M. Wu et al., 2025)
Deeper insight into the rational design and
synthesis of zeolites revealed by machine		The latest advancement of machine
learning potential in revealing the black		2025This work

Table 1 Comparison with existing reviews of machine learning in zeolite synthesis.

"Blackbox" in the Conventional Zeolite Synthesis

learning: a mini review

Generally, zeolites are synthesized using a hydrothermal method within a certain time. The hydrothermal method in the zeolite synthesis process consists of several stages, including (Cundy & Cox, 2005):

box issue in zeolite synthesis

  • 1. Reactants consisting of amorphous silica and alumina are mixed with a cation source in an alkaline medium and a liquid phase, e.g., NaOH and KOH.
  • 2. The mixture is heated in a stainless-steel autoclave, usually above 100℃.
  • 3. The mixture undergoes an induction process by remaining amorphous after a while of heating.
  • 4. Crystalline zeolite can be produced, and all amorphous reactants are converted to zeolite.

The mixture of reactants in the zeolite synthesis process can turn into a solid gel or colloidal suspension, depending on the type of reactants and mixing conditions. Zeolite synthesis is a complex reaction with many influencing variables (Yu, 2007), such as composition combination (Si/Al ratio (Hernando et al., 2018), alkalinity (Salwa Mohd Nazir et al., 2019), solubility (White et al., 2011), inorganic cations (Sasidharan & Kumar, 1997)), reactant sources (Keawkumay et al., 2025; Khaleque et al., 2020), OSDA (Jensen et al., 2021), temperature (Hui & Chao, 2006; Sumari et al., 2019), time (Meftah et al., 2017), aging period (Ahlers et al., 2020), stirring (Hanif et al., 2000), and addition of seeds (Jain & Rimer, 2020). The mechanism of zeolite crystallization consists of three main steps. The first step is the supersaturation stage, which drives the crystallization process through the formation of a supersaturated atmosphere. The second step is nucleation, where the reactant molecules undergo rearrangement to become nuclei. The last step is crystal growth.

Apart from the solid pathway mechanism by rearranging the structure, the mechanism of zeolite formation can also be through the settlement pathway or solution-mediated mechanism. In a solution-mediated mechanism, the amorphous phase in the zeolite will dissolve first, then form a core, and crystal growth will occur. Furthermore, the resulting zeolite, often a metastable species, rendering the more reproducible exact synthesis conditions, is urgently needed (Cundy & Cox, 2005; Grand et al., 2016). Besides, it is also prominent to bear in mind that zeolite as a metastable phase makes it probable to be controlled by the kinetic parameters. Notwithstanding, the coincidence of complex parameters, complex crystallization kinetics, and desired properties renders the process of zeolite synthesis still based on trial-and-error approaches, which causes the resulting product with minimal control, such as amorphous even zeolite with different phase and pore size (Moliner et al., 2019). Moreover, synthesis conditions that comprise even ten parameters make it a daunting task because we will encounter a high-dimensional search space (D. Ma et al., 2022).

It was reported previously that a specific zeolite synthesis method could generate a crystallized zeolite with Si/Al <5 due to the presence of alkali. Furthermore, the SDA (Structure Directing Agent) exploitation replacing the inorganic alkali made the zeolite with a high Si/Al ratio possible, even generating pure silica zeolite, known as silicalite-1 (X. Yang et al., 2022; C. Zhang et al., 2022). Besides, the synthesis of zeolite is also affected by the mixing order of reactants (Prodinger & Derewinski, 2020). Oleksiak et al. have investigated the nucleation of FAU and LTA zeolite using a combination of microscopy, scattering, and diffraction techniques (Oleksiak et al., 2016). The result showed that the exterior surface is the energetically preferred site for the nucleation over the interior of the particle based on the effect of confinement.

Kumar et al. reported the investigation of LTA zeolite synthesis using atomic force microscopy (AFM) at 35 and 45℃. In this work, LTA crystallization exhibits the gel-like islands formation from the aluminosilicate molecules in supersaturated conditions. Beyond these three-dimensional islands' evolution and assembly, other pathways could also occur, i.e., the attachment of nearly oriented. Moreover, a layering mechanism with layers spreading and two-dimensional nucleation also could create in lower supersaturation conditions. Hence, this work highlighted the nonclassical crystallization, which may be related to other zeolite types (Kumar et al., 2018).

Furthermore, apart from the synthesis parameters mentioned, zeolite synthesis research is also shifting to the consensus that providing a sustainable and environmentally friendly synthesis complete with enhancing the mechanistic understanding is also urgently needed. Kadja et al. have thoroughly investigated ZSM-5 zeolite synthesis (Kadja, Azhari, Mukti, et al., 2021). This work uses rice husk silica, seed-assisted, without SDA's presence by the solvent-free method. The result showed that crystalline ZSM-5 was generated in 10 h at 180 ℃, yielding up to ≥ 95%. Moreover, the Avrami equation analysis exhibits that nucleation occurred instantaneously and was the rate-determining step with activation energies of 137 kJ.mol-1 . Hence, the progress of revealing zeolite synthesis black box is still elusive due to many parameters that are essential to be examined.

Machine Learning Prospect for Rational Design of Zeolite Synthesis

As aforementioned before, the zeolite synthesis process is influenced by complex parameters, the trial-and-error principal, and the elusive understanding of crystallization mechanism, which has given rise to machine learning as a promising completion. Principally, machine learning prescribes suitable training data to a proper algorithm to understand the relationships over several variables (Butler et al., 2018; Jordan, M. I. & Mitchell, T. M. 2015). The data in machine learning is processed in four steps, i.e., data collection, feature generation and selection, algorithm selection, and validation and prediction(Mai et al., 2022). Due to its ability to predict, machine learning could also minimize the number of experiments (Louie et al., 2021; Xu et al., 2023). Figure 2 exhibits the use of machine learning in zeolite synthesis, including properties and structure prediction, best synthesis condition, OSDA determination, crystallinity and yield analysis, topology classification, even discovery new zeolite topology.

The role of machine learning in zeolite synthesis

Classification and Discovery of Zeolite Topology

The development of machine learning in zeolite areas has significantly impact the zeolite prediction. Using machine learning, Raman has predicted the extra-large pore structure in zeolite (Raman, 2023b). The data comprise 15 features, so it is simplified using t-SNE (t-distributed stochastic neighbor embedding). This work started with collecting a dataset related to the extra-large pore zeolite using 70-80 papers from several publishers. This collection data process used 80- 90% of the total time in this research. The independent variable (input) used was molar gel synthesis composition, while the dependent variable (output) was the number of T atoms in the zeolite pore ring. Zeolite with less than 12 T atoms and zeolite with more than 12 was labelled as 0 and 1, respectively.

Moreover, to obtain more transparent and interpretable results, SHAP (SHapley Additive exPlanations), a game-theoritic approach, have been conducted. The results showed that the zeolite with a very large pore size was produced from a lower F/T (the comparison of fluoride ions and T atoms) composition. The fluoride ion, in this case, plays a role in converting the synthetic material into a more mobile solution. During the zeolite synthesis process, fluoride ions act as mineralizers and OH− ions. In addition, a high carbon and nitrogen (C/N) ratio of OSDA is also required in the formation of zeolite with large pore sizes. It should be noted that the SHAP value is not only influenced by the specified features but also by other features. Hence, SHAP model is promising to interpreting the blackbox in zeolite synthesis and machine learning. Furthermore, by using XGBoost algorithms on the Heroku cloud platform, this work has achieved an accuracy of up to 86.57%. This result can be applied by other researchers regarding gaining their desired extra-large pore zeolite as well as discovering new zeolites with extra-large pore zeolite.

Long before, Yang et al. reported ZSP, the zeolite-structure predictor, for classifying the zeolite structures posit by the framework types (S. Yang et al., 2008). Among 1436 zeolite crystals from the Inorganic Crystal Structure Database (ICSD) that were analyzed, 179 zeolite frameworks were approved, and 96 framework types were depicted in the data set of zeolites. Furthermore, the topological descriptors using the Delaunay tessellation approach consist of four steps, i.e., generation of zeolite unit cell using the computational crystallography toolbox, replication periodically of the unit cell, determination of framework T-atom as active points, and utilization of Qhull algorithm for each sphere of T-atoms tessellation. Moreover, this work uses random forest algorithms, which exhibit an accuracy of 97.5%.

In another case, Kim and Min have exploited Bayesian active learning, the combination of machine learning and Bayesian optimization, to accelerate the discovery of zeolite structures from numerous hypothetical candidates as depicted in Figure 3(a) (Kim & Min, 2021). The procedures of this work start with calculating the mechanical properties of zeolite structures in the IZA database. The data was collected then using machine learning to predict the bulk and shear moduli. Furthermore, the elastic properties of zeolite structures in the predicted crystallography open database (PCOD) were predicted using the regression model. There were 50 regression models obtained and used for the Bayesian optimization. These procedures were repeated 20 times for validation. The result showed the new database that comprises up to 871 labeled zeolite structures and 23 new zeolite structures with the higher shear modulus than IZA database of 5.61-48.16 GPa. Figure 3(b) exhibits the standard deviation value of the predictive model including the unlabeled data which indicating the predictive model is convergence. It should be mentioned that the proposed platform may be used for other new material discoveries.

2

(a) The Bayesian active learning platform. (b) Standard deviation of data prediction for shear and bulk modulus. Reproduced with permission from Ref. (Kim & Min, 2021). Copyright 2021 ACS.

Furthermore, Xing and Blaisten-Barojas have developed a cloud-based computing system in order to permit the user to access the ZSP using a Web browser (Xing & Blaisten-Barojas, 2013). In this work, a public cloud was presented, i.e., Structure-Adaptive-Materials-Prediction (SAMP), which developed for Windows Azure platform. As a result, 41 framework types of zeolites were automatically obtained. Finally, it is worth noting that this work renders an automated system consisting of supercell, visualization, descriptor, and ZSP.

Mechanical and Structural Properties

Some research on machine learning utilization in zeolite synthesis was focused on the mechanical and structural properties of zeolite. Evans and Coudert have predicted the mechanical properties of the zeolite framework, particularly the elastic properties, i.e., elastic bulk (K) and shear moduli (G), using gradient boosting regressor (GBR) (Evans & Coudert, 2017). The descriptors in this research consist of space group, crystal density, unit cell volume, Si-O distributions, surface area, pore volume, pore size, and dimensionality. Furthermore, this work pointed out that the high-density zeolite has lower K and G than that of low-density. The result showed that from 590448 hypothetical zeolites prediction, the obtained accuracy for log (K) and log (G), respectively, was 0.102 ± 0.034 and 0.0947 ± 0.022. In addition, this model provides cheaper large-scale prediction by excluding the relative energy calculation, which has proven not to improve the accuracy of K and G.

Sours and Kulkarni utilize the deep neural network for predicting the structural properties of pure silica zeolites, as depicted in Figure 4(a) (Sours & Kulkarni, 2022). They used 219 data of pure silica zeolite topologies for DFT calculations. In addition, the structural properties analysis using 187 topologies data through machine learning and DFT optimizations for comparison. The result showed that machine learning potentials and DFT gained excellent performance. Notwithstanding, the machine learning model exhibits a faster performance than DFT up to 1000 times. Moreover, the accuracy of deep neural networks is validated by 32 testing topologies.

Still with the intention to predict the zeolite topology structural from the synthesis condition, another group, i.e., Jensen et al., created the pipeline data collected by automatically extracting zeolite synthesis data from 70000 papers. These data are then plotted as depicted in Figure 4(c). The data was collected by using NLP (natural language processing) techniques, the related synthesis data was parsed using HTML and XML, and the compositional ratios were located and extracted using regular expression (regex) (Figure 4(b)). In contrast with the previous group, which focused on pure silica zeolite, this work has successfully applied these pipeline data using random forest algorithms to predict the germanium (Ge) zeolite high and low framework densities with RMSE (root mean squared error (RMSE) of 0.98T/1000Å3 . Moreover, this work also pointed out the fluoride ion and Ge relationship with respect to the hydrothermal stability (Jensen et al., 2019).

Optimization of Synthesis Condition

In light of the plentiful research nowadays about low-cost zeolite synthesis, Ma et al. have exploited machine learning for the realization of distilling seed-assisted zeolite synthesis (D. Ma et al., 2022). This work uses 385 historical data on seed-assisted zeolite synthesis carried out in the laboratory. These data involve nine synthesis parameters and using Na-borosilicate gel, which consists of OTMAC (Octyltrimethylammonium Chloride) and EUO, MWW, MTT, MFI, ERI, SFE, TON, IWF, and *MRE as seeds. Moreover, this work was intended to suppress the production cost of zeolite synthesis by using a seed-assisted method instead of OSDA. There are six machine learning models used in this work, i.e., logistic regression (LR), Decision Tree (DT), Support Vector Machine (SVM), extreme Gradient Boosting (XGBoost), Adaptive Boosting (AdaBoost), and Random Forest (RF).

4

a). Schematic procedure of deep potential model. Reproduced with permission from Ref. (Sours & Kulkarni, 2022). Copyright 2023 ACS. b). Schematic overview of machine learning approach in zeolite synthesis. c). The resulting data by automatic literature data extraction. The color insensity represents the data frequency. Reproduced with permission from Ref. (Jensen et al., 2019). Copyright 2019 ACS.

The accuracies of each model exhibit that the highest accuracy obtained by the AdaBoost model, followed by RF, XGBoost, DT, SVM, and LR, respectively. Moreover, OTMAC, NaOH, and B2O3 gel composition, framework density, and crystallization time have significant importance scores. It is worth noting that the metastable zeolite formation was affected by NaOH and the seed's framework density. The higher framework density of the seed inclines to no metastable formation. In this regard, MFI with higher framework density, i.e., 17.9, exhibited no transformation compared to IWV seed with a framework density of 15.7 which experienced transformation to MWW and EUO aligned with increasing the crystallization time.

Concerning gaining the general guidelines for zeolite classification, which relate to synthesis condition, Liu and coworkers have defined the global potential energy surface (PES) of 12 T atom zeolite systems, i.e., CHA, ATO, ATV, and

ATS framework using machine learning (S. Ma et al., 2020). The result exhibited that zeolite becomes a thermodynamically stable product due to the exploitation of a proper SDA. Moreover, the pH of zeolite synthesis is sensitive. In this sense, the Si-O-Al bonding favored the basic condition, not the acidic or neutral condition.

2

The decision tree resulted from XGBoost analysis. Reproduced with permission from Ref. (Muraoka et al., 2019). Copyright 2019 Nature.

Muraoka et al. used machine learning to connect synthesis and zeolite structure descriptors. In this research, XGBoost and random forest models showed a high accuracy of around 75-80%(Muraoka et al., 2019). Among these two methods, XGBoost was chosen because its hyperparameter is more suitable for this study. Moreover, it can predict the results of the synthesis as well as the possibility of zeolite formation under certain synthesis conditions. The XGBoost analysis revealed that the most influential descriptors of zeolite synthesis are SiO2, Al2O3, MOH (M = Li, Na, K), and H2O (Figure 5)

Recently, Pan et al. introduced the extending SHAP through aggregation approach to all building units. In this sense, they used ZeoSyn, which consists of 23961 zeolite synthesis routes for 921 OSDAs and 223 zeolite frameworks (Pan et al., 2024). The SHAP values then calculated to define the influenced of each parameter on the zeolite forming probability. Using random forest architecture to train the ZeoSyn datasets, the reported accuracy was 0.73.

The Role of OSDA

Besides investigating the synthesis conditions along with the resulting crystal, the possibility of research about OSDA and zeolite topology using machine learning also attracted scientists' attention. Jensen et al. have investigated the relationship between OSDA and zeolite types using weighted holistic invariant molecular (WHIM) as descriptors (Jensen et al., 2021). Furthermore, to simplify the high dimensionality of WHIM, PCA (principal component analysis) is used to sampling the potential and known OSDAs of 5 zeolite types, i.e., LEV, CHA, AEI, LTA, and AFX. This work collected 5663 databases of synthesis routes using natural language processing. Moreover, this work has successfully predicted 408 potential OSDA for CHA zeolite types. Besides, these generated OSDA are also being tested for a different kind of zeolite, i.e., SFW framework, and the molecular simulations exhibited that OSDA generates are compatible with SFW.

Schwalbe-Koda et al. have proposed metrics of binding by controlling the template selectivity in zeolite synthesis (Schwalbe-Koda, Kwon, et al., 2021). This work collected 549 OSDA data from literature with benchmarking OSDA's phase competition with criteria having both strong and weak binding affinity to several zeolite frameworks, respectively. The literature extraction exhibited 1122 pairs of OSDA and zeolite plotted by templating energy to obtain the compatibility between literature data and binding metric. Furthermore, the result showed that OSDAs for several frameworks and zeolite with the design of dual-OSDA exhibit not accurate and limited binding energy metrics, respectively. To be noted, the templating energy was better in explaining the literature, approximately 70%, than plain binding energies. Nevertheless, the templating energy inaccurately predicted the best host for a certain OSDA. Furthermore, this work successfully identified the energetic, electrostatic, and geometric od OSDAs and designed OSDAs with favorable properties for zeolite CHA.

In another work, Schwalbe-Koda et al. continues designing the biselectivity OSDAs for intergrowth zeolite synthesis by combining data mining and computational simulations (Schwalbe-Koda, Corma, et al., 2021). This work started by using some known OSDAs as a reference, designing new OSDA through shape metrics and phase competition. Moreover, the binding energies were used for energy references, i.e., to calculate the second-best pair of zeolites and OSDAs. The results exhibited that this method has the ability for hypothetical intergrowth realization of AEI/SAV and many other intergrowths.

Recently, still from Rafael Gόmez-Bombarelli group, they applied binding energies to the synthesis of up to 100 zeolites through known OSDAs (Schwalbe-Koda et al., 2022). The resulting data has the ability to construct the small- and medium-pore zeolites, which have a higher selectivity than the large-pore zeolites with respect to the opening size. Moreover, this work emphasizes that the presence of inorganic cations in OSDA enables an effect on competing phases. Nonetheless, new computational methods are still necessary to develop the electrostatic effects in zeolite's template. To this end, this method successfully proposed an alternative OSDA for KFI zeolite, i.e., tetraethylammonium.

The latest, Gόmez-Bombarelli and co-workers reported the efficient zeolite intergrowth synthesis of CHA/AEI through designing a unique OSDA via a priori methodology. This method relies on quantifying zeolite from a phase competition point of view via high-throughput simulations (Bello-Jurado et al., 2022). Generally, CHA and AEI zeolite were synthesized using N,N,N-trimethyladamantylammonium (TMAda), and N-ethyl-N-methyl-2,2,6,6 tetramethylpiperidinium, respectively. In this work, they proposed N-ethyl-N-isopropyl-N-methylpropan-2-ammonium and 1-ethyl-1-isopropylpyrrolidin-1-ium for CHA/AEI intergrowths zeolite synthesis. Moreover, the resulting material, i.e., CHA/AEI intergrowth, exhibits an outstanding performance for NH3-SCR (selective reduction of NOx with ammonia) compared to the pure CHA catalyst. It showed that machine learning has the ability to design and synthesize zeolite more efficiently and is applicable to industry.

Crystallinity and Yield Analysis

Moreover, the application of machine learning on zeolite was utilized not only for the zeolite synthesis process and parameter condition but also for crystallization results. Nikiforov et al. reported the prediction of MFI zeolite's crystallinity by analyzing 650 papers using several machine learning algorithms, i.e., random forest, gradient boosting, and decision trees. There are twelve synthesis parameters collected that is number of moles of Na/Si, the number of moles of Al/Si, the number of moles of H2O/Si, aging time, aging temperature, time and temperature of the first step of hydrothermal, time and temperature of the second step of hydrothermal, number of moles of template/Si, and the obtained degree of crystallinity. Among the aforementioned used algorithms, gradient boosting algorithms showed the most outstanding result, with MAE (mean absolute error) and MSE (mean squared error) of 204.4 and 10.3, respectively (Nikiforov et al., 2022).

Conroy et al. have reported the machine learning utilization for zeolite LTA synthesis. This research is expected to increase the zeolite yield and performance. The result exhibits one of the synthesis parameters that affect the LTA synthesis, i.e., reaction time. The gel composition used is SiO2/Al2O3, H2O/Na2O, Na2O/SiO2 of 2, 30, and 2.5, respectively. In the initial period of zeolite synthesis, the resulting product is still amorphous. However, the crystalline LTA zeolite increased in line with the increase in reaction time. This result was in agreement with SEM images. Moreover, the SEM images and quantitative XRD, which were analyzed with the Rietveld method, revealed the optimal crystalline results of 71 and 72% at 3 and 4 hours, respectively. Besides, the qualitative XRD, which was analyzed by comparing the relative peak intensity, revealed the crystallinity of LTA zeolite of 99 wt% at 3- and 4-hour synthesis time. Therefore, it is worth noting that the hybrid XRD analysis, a combination of quantitative and qualitative XRD, plays a role in amorphous and crystalline determination (Conroy et al., 2022). Furthermore, among several algorithms that applied, i.e., ridge regression, linear regression, regression tree, XGBoost, artificial neural networks, and random forest, an artificial neural networks model (ANN) achieved the highest accuracy with R2 = 0.84. In order to validate the hybrid data approach, the ANN model was analyzed using qualitative, quantitative, hybrid 10, 20, and 40 datasets (the number according to the crystalline zeolite content in wt%). The result showed that the qualitative XRD has the highest accuracy among others. Notwithstanding, the qualitative XRD lies a problem in interpreting the successfulness of most zeolite synthesis. Hence, in fact, the "hybrid 10" was the most accurate model. Nevertheless, this work pointed out that machine learning can be successfully utilized in zeolite LTA synthesis.

Table 2 Summarises of machine learning application in zeolite synthesis

AimDataset collectionAlgorithmPlatformHighlighted ResultRef
Classification and Discovery of Ze
Extra-large pore zeolite predictionManually processed
70-80 papers from
several publishers		XGBoostHeroku
cloud		Accuracy up to 86.57%(Raman,
2023a)
Framework type predictorInorganic Crystal
Structure Database
(ICSD)		Random forest-Accuracy of 83%(S. Yang et al.,
2009)
Zeolite structures with
superior mechanical
properties		IZA databaseDT, BTE, SVM, GPR,
LightGBM		Bayesian
active
learning		23 new zeolite structures(Kim & Min,
2021)
Zeolite structure predictorInorganic Crystal
Structure Database
(ICSD)		Random forestWindows
Azure		Accuracy of 0.98(Xing &
Blaisten-
Barojas, 2013)
Mechanical and structural properties•
Mechanical properties predictionStructure file in CIF formatGradient boosting regressor-Accuracy for log (K) and log (G) = 0.102 ± 0.034 and 0.0947 ± 0.022(Evans &
Coudert,
2017)
Structural properties predictionIZA databaseDeep Potential-RMSE of bulk moduli of 8.6 GPa(Sours &
Kulkarni,
2022)
Zeolite topology structuralZeolite journal
articles using NLP		Random forest-RMSE of 0.98T/1000Å(Jensen et al.,
2019)
Optimization of synthesis ofondition
Seed-assisted zeolite synthesisArchived laboratory recordsLR, DT, SVM, XGBoost,
AdaBoost, Random
Forest		-Accuracy up to >0.9(D. Ma et al.,
2022)
Hydrothermal parametersZeoSynRandom Forest-Accuracy of 0.73(Pan et al.,
2024)
Thermodynamic rules for zeolite formationEnhanced stochastic surface walkingGlobal neural network-General guidelines for
zeolite classification
relate to their stability
and synthesis condition		(S. Ma et al.,
2020)
Linking the synthesis
and zeolite structure
descriptor		Experimental records in literaturesXGBoost, Random
Forest		-Accuracy of around 0.75-0.8(Muraoka et
al., 2019)
·The role of OSDA
Relationship between OSDAs and zeolitesScientific literatures
which published
from 1966 to 2020
using NLP		Generative neural work-408 potential OSDA for
CHA zeolite		(Jensen et al.,
2021)
Designing the
biselectivity OSDAs for
intergrowth zeolite
synthesis		A data set of zeolite-
OSDA pairs from
literatures		--Potential OSDAs for disordered framework(Schwalbe-
Koda, Corma,
et al., 2021)
Repurposing templatesLiteratures of
zeolites synthesis
data		Voronoi and monte carlo docking algorithm-Example of alternatives OSDA for zeolites(Schwalbe-
Koda et al.,
2022)
Crystallinity and yield arnalysis
Degree of crystallinity
prediction for MFI
zeolite		Manually processed
650 papers about
MFI zeolites
synthesis		DT, Random Forest, and gradient boosting-MAE of 10.3 and MSE
of 204.4		(Nikiforov et al., 2022)
Increase the yield and performance of LTA zeoliteSynthesis data from literaturesLinear regression,
artificial neural
network, regression
tree, XGBoost, random
forest, ridge regression		-Accuracy up to 0.84(Conroy et al.,
2022)

Summary and Outlook

Zeolite, applied in several chemistry industries, became more and more center of the researcher's attention. Here, we overview the conventional zeolite synthesis, which still requires a deep understanding to explain the crystallization of zeolite structure, both kinetics and thermodynamics, which remains elusive. It turns out to be a laborious task due to numerous influential parameters, such as the molar composition of reactants and synthesis conditions. On the other hand, machine learning with the capability to predict and discover a material is also under the spotlight. Hence, machine learning is a promising way to demonstrate the synthesis process, predict the best synthetic condition, and even discover a new zeolite topology.

In light of the foregoing, the superiority of machine learning has brought the researchers evidence of the machine learning potential in zeolite research, particularly the zeolite synthesis parameters designing, the zeolite structural properties prediction, even the prediction of quantitative output in zeolite synthesis. Table 2 summarizes the application of machine learning in zeolite synthesis. Generally, there are two deviances of machine learning in revealing the zeolite synthesis process, i.e., (i) determining the data collection techniques to obtain an enormous database for training, testing, and validation and (ii) the preference of suitable learning algorithms to accelerate the achievement of the research target. Moreover, it is worth noting that the feature importance of the machine learning algorithm model brought a convenience impact to analyze the large number of parameters that affect the synthesis process. With respect to these points, machine learning will effectively improve the synthesis process.

Ultimately, among several machine learning approaches to zeolite synthesis, some cases require attention to ameliorate the future prospects. First, the algorithm used ought to have the ability to interplay the synthesis process and desired properties as well as zeolite performance. Second, the machine learning approaches require an ability not only to make an effective synthesis process by the mechanism comprehension but also to prescribe the synthesis process on a large scale. In addition, data generation and feature extraction should also involve simple and efficient methods for capturing information from a large amount of literature regarding the biggest data availability are from patents and journals. This is due to the requisite of high-quality datasets to produce an accurate result in machine learning. Besides, it is also worth noting that the large amount of zeolite structures which is absolutely crucial to be consider in order to database generalization. Moreover, the exegesis of the kinetics and thermodynamic mechanism of zeolite synthesis required a proper approach to ensuring the reproducibility and quality of machine learning. In this case, SHAP and LIME (Local Interpretable Model-agnostic Explanations) can be considered for utilization due to the ability to interpret the method before proving it in the simulation and/or experiment.

On a final note, due to the complexity of zeolite and machine learning, the cross-disciplinary teams would create the best result. In this sense, shared databases and/or standard methods could potentially catalyze the process. Nevertheless, it is worth noting that cross validation, external validation, and experimental test still needed to prove the machine learning prediction result before applied in the laboratory and industry. Thus, if all of these cases are fulfilled, the machine learning models may successfully address the realization of zeolite and even develop more zeolite frameworks, which expected to have a beneficial impact on the chemical industry as well as humankind's life.

Acknowledgement

This work is supported by Hibah Riset Unggulan ITB 2025. SM gratefully acknowledges Institut Teknologi Bandung for tuition provided through the GTA scholarship and Kadja Lab led by Dr. Ir. Grandprix T.M. Kadja for monthly stipend.

Compliance with ethics guidelines

The authors declare they have no conflict of interest or financial conflicts to disclose.

This article contains no studies with human or animal subjects performed by authors.

List of Abbreviations

AbbreviationFull Term
AdaBoostAdaptive Boosting
AFMAtomic Force Microscopy
ANNArtificial Neural Networks
BTEBoosting Tree Ensemble
DFTDensity functional theory
AbbreviationFull Term
DTDecision tree
GBRGradient Boosting Regressor
GPRGaussian Process Regressor
LightGBMLight Gradient Boosting Machine
LIMELocal Interpretable Model-agnostic Explanations
LRLogistic Regression
MAEMean Absolute Error
MSEMean Squared Error
OSDAOrganic Structure Directing Agent
PCODPredicted Crystallography Open Database
PESPotential Energy Surface
RFRandom forest
SAMPStructure-Adaptive-Materials-Prediction
SDAStructure directing agent
SHAPSHaply Additive exPlanations
svmSupport Vector Machine
t-SNEt-distributed stochastic neighbor embedding
WHIMWeighted Holistic Invariant Molecular
XGBoosteXtreme Gradient Boosting

Research Intelligence

Data from OpenAlex ↗

Metrics

0.00
FWCIfield-weighted
13th
Percentilevs same year + field
Review
Work type
GOLD
Open Access

Topics

Zeolite Catalysis and Synthesis Primary
0.97
Inorganic Chemistry › Chemistry › Physical Sciences
Machine Learning in Materials Science
0.96

Related Research

Semantic Profile AI-classified research signals

Core Domains
Rational design 0.46
level 2
Biochemical engineering 0.39
level 1
Engineering drawing 0.37
level 1
Secondary Topics
Computer science 0.35 Process engineering 0.34 Artificial intelligence 0.33 Chemical engineering 0.33 Mathematics 0.33 Engineering 0.32 Materials science 0.22 Nanotechnology 0.20

Institution Network

References

  1. Ahlers, C. B., Talbot, J. B., Taufiqurrahmi, N., Rahman Mohamed, A., Bhatia -, S., Arni, S., Sumari, S., Santoso, A., & Tamara, T. (2020). The Effect of Aging and Crystallization Time on the Synthesis and Characteristics of Zeolite-Y from Malang-Quartzite Silica. IOP Conference Series: Materials Science and Engineering, 833(1), 012060. https://doi.org/10.1088/1757-899X/833/1/012060 DOI: 10.1088/1757-899x/833/1/012060
  2. Al-Ani, A., Freitas, C., & Zholobenko, V. (2020). Nanostructured large-pore zeolite: The enhanced accessibility of active sites and its effect on the catalytic performance. Microporous and Mesoporous Materials, 293, 109805. https://doi.org/10.1016/j.micromeso.2019.109805 DOI: 10.1016/j.micromeso.2019.109805
  3. Al-Nahari, S., Laurencin, D., & Alonso, B. (2023). Solvent-free synthesis of zeolites: New insights into the mechanism and non-mechanochemical route. Microporous and Mesoporous Materials, 350(January), 112445. https://doi.org/10.1016/j.micromeso.2023.112445 DOI: 10.1016/j.micromeso.2023.112445
  4. Aragaw, T. A., & Ayalew, A. A. (2019). Removal of water hardness using zeolite synthesized from Ethiopian kaolin by hydrothermal method. Water Practice and Technology, 14(1), 145–159. https://doi.org/10.2166/wpt.2018.116 DOI: 10.2166/wpt.2018.116
  5. Bae, S. Y., Shin, J. S., Kim, Y. S., & Lee, J. M. (2021). Decision tree analysis on the performance of zeolite-based SCR catalysts. IFAC-PapersOnLine, 54(3), 55–60. https://doi.org/10.1016/j.ifacol.2021.08.218 DOI: 10.1016/j.ifacol.2021.08.218
  6. Bello-Jurado, E., Schwalbe-Koda, D., Nero, M., Paris, C., Uusimäki, T., Román-Leshkov, Y., Corma, A., Willhammar, T., Gómez-Bombarelli, R., & Moliner, M. (2022). Tunable CHA/AEI Zeolite Intergrowths with A Priori Biselective Organic Structure-Directing Agents: Controlling Enrichment and Implications for Selective Catalytic Reduction of NOx. Angewandte Chemie - International Edition, 61(28), 1–6. https://doi.org/10.1002/anie.202201837 DOI: 10.1002/anie.202201837
  7. Bornes, C., Santos-Vieira, I. C. M. S., Vieira, R., Mafra, L., Simões, M. M. Q., & Rocha, J. (2023). Challenges and opportunities for zeolites in biomass upgrading: Impediments and future directions. Catalysis Today, 419(March), 114159. https://doi.org/10.1016/j.cattod.2023.114159 DOI: 10.1016/j.cattod.2023.114159
  8. Butler, K. T., Davies, D. W., Cartwright, H., Isayev, O., & Walsh, A. (2018). Machine learning for molecular and materials science. Nature, 559(7715), 547–555. https://doi.org/10.1038/s41586-018-0337-2 DOI: 10.1038/s41586-018-0337-2
  9. Campanile, A., Liguori, B., Ferone, C., Caputo, D., & Aprea, P. (2022). Zeolite-based monoliths for water softening by ion exchange/precipitation process. Scientific Reports, 12(1), 1–10. https://doi.org/10.1038/s41598-022-07679-2 DOI: 10.1038/s41598-022-07679-2
  10. Chen, L., Jansson, J., Skoglundh, M., & Grönbeck, H. (2016). Mechanism for Solid-State Ion Exchange of Cu+ into Zeolites. Journal of Physical Chemistry C, 120(51), 29182–29189. https://doi.org/10.1021/acs.jpcc.6b09553 DOI: 10.1021/acs.jpcc.6b09553
  11. Conroy, B., Nayak, R., Hidalgo, A. L. R., & Millar, G. J. (2022). Evaluation and application of machine learning principles to Zeolite LTA synthesis. Microporous and Mesoporous Materials, 335(January), 111802. https://doi.org/10.1016/j.micromeso.2022.111802 DOI: 10.1016/j.micromeso.2022.111802
  12. Cundy, C. S., & Cox, P. A. (2003). The hydrothermal synthesis of zeolites: History and development from the earliest days to the present time. Chemical Reviews, 103(3), 663–701. https://doi.org/10.1021/cr020060i DOI: 10.1021/cr020060i
  13. Cundy, C. S., & Cox, P. A. (2005). The hydrothermal synthesis of zeolites: Precursors, intermediates and reaction mechanism. Microporous and Mesoporous Materials, 82(1–2), 1–78. https://doi.org/10.1016/j.micromeso.2005.02.016 DOI: 10.1016/j.micromeso.2005.02.016
  14. Evans, J. D., & Coudert, F. X. (2017). Predicting the Mechanical Properties of Zeolite Frameworks by Machine Learning. Chemistry of Materials, 29(18), 7833–7839. https://doi.org/10.1021/acs.chemmater.7b02532 DOI: 10.1021/acs.chemmater.7b02532
  15. Fischer, M. (2020). Simulation-based evaluation of zeolite adsorbents for the removal of emerging contaminants†. Materials Advances, 1(1), 86–98. https://doi.org/10.1039/d0ma00025f DOI: 10.1039/d0ma00025f
  16. Gandhi, A., & Hasan, M. M. F. (2022). Machine learning for the design and discovery of zeolites and porous crystalline materials. Current Opinion in Chemical Engineering, 35, 100739. https://doi.org/10.1016/j.coche.2021.100739 DOI: 10.1016/j.coche.2021.100739
  17. Graça, I., Bacariza, M. C., Fernandes, A., & Chadwick, D. (2018). Desilicated NaY zeolites impregnated with magnesium as catalysts for glucose isomerisation into fructose. Applied Catalysis B: Environmental, 224, 660–670. https://doi.org/10.1016/j.apcatb.2017.11.009 DOI: 10.1016/j.apcatb.2017.11.009
  18. Grand, J., Awala, H., & Mintova, S. (2016). Mechanism of zeolites crystal growth: New findings and open questions. CrystEngComm, 18(5), 650–664. https://doi.org/10.1039/c5ce02286j DOI: 10.1039/c5ce02286j
  19. Hanif, N., Anderson, M. W., Alfredsson, V., & Terasaki, O. (2000). The effect of stirring on the synthesis of intergrowths of zeolite Y polymorphs. Physical Chemistry Chemical Physics, 2(14), 3349–3357. https://doi.org/10.1039/B002314K DOI: 10.1039/b002314k
  20. Hernando, H., Hernández-Giménez, A. M., Ochoa-Hernández, C., Bruijnincx, P. C. A., Houben, K., Baldus, M., Pizarro, P., Coronado, J. M., Fermoso, J., Čejka, J., Weckhuysen, B. M., & Serrano, D. P. (2018). Engineering the acidity and accessibility of the zeolite ZSM-5 for efficient bio-oil upgrading in catalytic pyrolysis of lignocellulose. Green Chemistry, 20(15), 3499–3511. https://doi.org/10.1039/c8gc01722k DOI: 10.1039/c8gc01722k
  21. Hewitt, D., Pope, T., Sarwar, M., Turrina, A., & Slater, B. (2022). Machine learning accelerated high-throughput screening of zeolites for the selective adsorption of xylene isomers. Chemical Science, 532, 13178–13186. https://doi.org/10.1039/d2sc03351h DOI: 10.1039/d2sc03351h
  22. Hui, K. S., & Chao, C. Y. H. (2006). Effects of step-change of synthesis temperature on synthesis of zeolite 4A from coal fly ash. Microporous and Mesoporous Materials, 88(1–3), 145–151. https://doi.org/10.1016/J.MICROMESO.2005.09.005 DOI: 10.1016/j.micromeso.2005.09.005
  23. Jain, R., & Rimer, J. D. (2020). Seed-Assisted zeolite synthesis: The impact of seeding conditions and interzeolite transformations on crystal structure and morphology. Microporous and Mesoporous Materials, 300, 110174. https://doi.org/10.1016/J.MICROMESO.2020.110174 DOI: 10.1016/j.micromeso.2020.110174
  24. Jensen, Z., Kim, E., Kwon, S., Gani, T. Z. H., Román-Leshkov, Y., Moliner, M., Corma, A., & Olivetti, E. (2019). A Machine Learning Approach to Zeolite Synthesis Enabled by Automatic Literature Data Extraction. ACS Central Science, 5, 892–899. https://doi.org/10.1021/acscentsci.9b00193 DOI: 10.1021/acscentsci.9b00193
  25. Jensen, Z., Kwon, S., Schwalbe-Koda, D., Paris, C., Gómez-Bombarelli, R., Román-Leshkov, Y., Corma, A., Moliner, M., & Olivetti, E. A. (2021). Discovering Relationships between OSDAs and Zeolites through Data Mining and Generative Neural Networks. ACS Central Science, 7(5), 858–867. https://doi.org/10.1021/acscentsci.1c00024 DOI: 10.1021/acscentsci.1c00024
  26. Jia, X., Khan, W., Wu, Z., Choi, J., & Yip, A. C. K. (2019). Modern synthesis strategies for hierarchical zeolites: Bottom-up versus top-down strategies. Advanced Powder Technology, 30(3), 467–484. https://doi.org/10.1016/j.apt.2018.12.014 DOI: 10.1016/j.apt.2018.12.014
  27. Jordan, M. I. & Mitchell,T. M. (2015). Machine learning: Trends,perspectives, and prospects. Science, 349(6245), 255–260.
  28. Kadja, G. T. M., Azhari, N. J., Mardiana, S., Khalil, M., Subagjo, & Mahyuddin, M. H. (2021). Accelerated, Mesoporogen-Free Synthesis of Hierarchical Nanorod ZSM-48 Assisted by Hydroxyl Radicals. Industrial and Engineering Chemistry Research, 60(48), 17786–17791. https://doi.org/10.1021/acs.iecr.1c03586 DOI: 10.1021/acs.iecr.1c03586
  29. Kadja, G. T. M., Azhari, N. J., Mukti, R. R., & Khalil, M. (2021). A Mechanistic Investigation of Sustainable Solvent-Free, Seed-Directed Synthesis of ZSM-5 Zeolites in the Absence of an Organic Structure-Directing Agent. ACS Omega, 6(1), 925–933. https://doi.org/10.1021/acsomega.0c05070 DOI: 10.1021/acsomega.0c05070
  30. Kadja, G. T. M., Fabiani, V. A., Aziz, M. H., Fajar, A. T. N., Prasetyo, A., Suendo, V., Ng, E. P., & Mukti, R. R. (2017). The effect of structural properties of natural silica precursors in the mesoporogen-free synthesis of hierarchical ZSM-5 below 100 °C. Advanced Powder Technology, 28(2), 443–452. https://doi.org/10.1016/j.apt.2016.10.017 DOI: 10.1016/j.apt.2016.10.017
  31. Kadja, G. T. M., Kadir, I. R., Fajar, A. T. N., Suendo, V., & Mukti, R. R. (2020). Revisiting the seed-assisted synthesis of zeolites without organic structure-directing agents: Insights from the CHA case. RSC Advances, 10(9), 5304–5315. https://doi.org/10.1039/c9ra10825d DOI: 10.1039/c9ra10825d
  32. Kadja, G. T. M., Mukti, R. R., Liu, Z., Rilyanti, M., Ismunandar, Marsih, I. N., Ogura, M., Wakihara, T., & Okubo, T. (2016). Mesoporogen-free synthesis of hierarchically porous ZSM-5 below 100°C. Microporous and Mesoporous Materials, 226, 344–352. https://doi.org/10.1016/j.micromeso.2016.02.007 DOI: 10.1016/j.micromeso.2016.02.007
  33. Kadja, G. T. M., Rukmana, M. D., Mukti, R. R., Mahyuddin, M. H., Saputro, A. G., & Wungu, T. D. K. (2021). Solvent-free, small organic lactam-assisted synthesis of ZSM-5 zeolites. Materials Letters, 290, 129501. https://doi.org/10.1016/j.matlet.2021.129501 DOI: 10.1016/j.matlet.2021.129501
  34. Kadja, G. T. M., Suprianti, T. R., Ilmi, M. M., Khalil, M., Mukti, R. R., & Subagjo. (2020). Sequential mechanochemical and recrystallization methods for synthesizing hierarchically porous ZSM-5 zeolites. Microporous and Mesoporous Materials, 308, 110550. https://doi.org/10.1016/j.micromeso.2020.110550 DOI: 10.1016/j.micromeso.2020.110550
  35. Karka, S., Kodukula, S., Nandury, S. V., & Pal, U. (2019). Polyethylenimine-Modified Zeolite 13X for CO2 Capture: Adsorption and Kinetic Studies. ACS Omega, 4(15), 16441–16449. https://doi.org/10.1021/acsomega.9b02047 DOI: 10.1021/acsomega.9b02047
  36. Keawkumay, C., Sosa, N., Osakoo, N., Prayoonpokarach, S., Wittayakun, J., Youngjan, S., Boonyoung, P., Khemthong, P., Amedlous, A., & Mintova, S. (2025). Effect of aluminum source and water content in the precursor suspensions used for the synthesis of nanosized zeolite Y on CO₂ adsorption capacity. Microporous and Mesoporous Materials, 386, 113491. https://doi.org/10.1016/J.MICROMESO.2025.113491 DOI: 10.1016/j.micromeso.2025.113491
  37. Khaleque, A., Alam, M. M., Hoque, M., Mondal, S., Haider, J. Bin, Xu, B., Johir, M. A. H., Karmakar, A. K., Zhou, J. L., Ahmed, M. B., & Moni, M. A. (2020). Zeolite synthesis from low-cost materials and environmental applications: A review. Environmental Advances, 2, 100019. https://doi.org/10.1016/J.ENVADV.2020.100019 DOI: 10.1016/j.envadv.2020.100019
  38. Khan, W., Jia, X., Wu, Z., Choi, J., & Yip, A. C. K. (2019). Incorporating hierarchy into conventional zeolites for catalytic biomass conversions: A review. Catalysts 9(2), 127, MDPI AG. https://doi.org/10.3390/catal9020127 DOI: 10.3390/catal9020127
  39. Kim, N., & Min, K. (2021). Accelerated Discovery of Zeolite Structures with Superior Mechanical Properties via Active Learning. Journal of Physical Chemistry Letters, 12(9), 2334–2339. https://doi.org/10.1021/acs.jpclett.1c00339 DOI: 10.1021/acs.jpclett.1c00339
  40. Kumar, M., Choudhary, M. K., & Rimer, J. D. (2018). Transient modes of zeolite surface growth from 3D gel-like islands to 2D single layers. Nature Communications, 9(1), 2129. https://doi.org/10.1038/s41467-018-04296-4 DOI: 10.1038/s41467-018-04296-4
  41. Kwak, S. J., Kim, H. S., Park, N., Park, M. J., & Lee, W. B. (2021). Recent progress on Al distribution over zeolite frameworks: Linking theories and experiments. Korean Journal of Chemical Engineering, 38(6), 1117–1128. https://doi.org/10.1007/s11814-021-0796-2 DOI: 10.1007/s11814-021-0796-2
  42. Li, X., Han, H., Evangelou, N., Wichrowski, N. J., Lu, P., Xu, W., Hwang, S. J., Zhao, W., Song, C., Guo, X., Bhan, A., Kevrekidis, I. G., & Tsapatsis, M. (2023). Machine learning-assisted crystal engineering of a zeolite. Nature Communications, 14(1), 1–12. https://doi.org/10.1038/s41467-023-38738-5 DOI: 10.1038/s41467-023-38738-5
  43. Louie, S. G., Chan, Y. H., da Jornada, F. H., Li, Z., & Qiu, D. Y. (2021). Discovering and understanding materials through computation. Nature Materials, 20(6), 728–735. https://doi.org/10.1038/s41563-021-01015-1 DOI: 10.1038/s41563-021-01015-1
  44. M. I. Jordan, & T. M. Mitchell. (2015). Machine learning: Trends,perspectives, and prospects. Science, 349(6245), 255–260.
  45. Ma, D., Li, X., Liang, J., Wang, Z., & Yang, W. (2022). Distilling seed-assisted zeolite synthesis conditions by machine learning. Microporous and Mesoporous Materials, 339(April), 112029. https://doi.org/10.1016/j.micromeso.2022.112029 DOI: 10.1016/j.micromeso.2022.112029
  46. Ma, S., & Liu, Z. P. (2022a). Machine learning potential era of zeolite simulation. Chemical Science, 13(18), 5055–5068. https://doi.org/10.1039/d2sc01225a DOI: 10.1039/d2sc01225a
  47. Ma, S., & Liu, Z. P. (2022b). The Role of Zeolite Framework in Zeolite Stability and Catalysis from Recent Atomic Simulation. Topics in Catalysis, 65(1–4), 59–68. https://doi.org/10.1007/s11244-021-01473-6 DOI: 10.1007/s11244-021-01473-6
  48. Ma, S., Shang, C., Wang, C. M., & Liu, Z. P. (2020). Thermodynamic rules for zeolite formation from machine learning based global optimization. Chemical Science, 11(37), 10113–10118. https://doi.org/10.1039/d0sc03918g DOI: 10.1039/d0sc03918g
  49. Mahmoodi, N. M., & Saffar-Dastgerdi, M. H. (2019). Zeolite nanoparticle as a superior adsorbent with high capacity: Synthesis, surface modification and pollutant adsorption ability from wastewater. Microchemical Journal, 145(August 2018), 74–83. https://doi.org/10.1016/j.microc.2018.10.018 DOI: 10.1016/j.microc.2018.10.018
  50. Mai, H., Le, T. C., Chen, D., Winkler, D. A., & Caruso, R. A. (2022). Machine Learning in the Development of Adsorbents for Clean Energy Application and Greenhouse Gas Capture. Advanced Science, 9(36), 1–22. https://doi.org/10.1002/advs.202203899 DOI: 10.1002/advs.202203899
  51. Mardiana, S., Azhari, N. J., Ilmi, T., & Kadja, G. T. M. (2022). Hierarchical zeolite for biomass conversion to biofuel: A review. Fuel, 309(October 2021), 122119. https://doi.org/10.1016/j.fuel.2021.122119 DOI: 10.1016/j.fuel.2021.122119
  52. Meftah, M., Oueslati, W., Chorfi, N., & Ben Haj Amara, A. (2017). Effect of the raw material type and the reaction time on the synthesis of halloysite based Zeolite Na-P1. Results in Physics, 7, 1475–1484. https://doi.org/10.1016/J.RINP.2017.04.013 DOI: 10.1016/j.rinp.2017.04.013
  53. Mguni, L. L., Ndhlovu, A., Liu, X., Hildebrandt, D., & Yao, Y. (2022). Insight into Adsorptive Desulfurization by Zeolites: A Machine Learning Exploration. Energy & Fuels, 36(8), 4427–4438. https://doi.org/10.1021/acs.energyfuels.1c03949 DOI: 10.1021/acs.energyfuels.1c03949
  54. Mohamed, R. M., Mkhalid, I. A., & Barakat, M. A. (2015). Rice husk ash as a renewable source for the production of zeolite NaY and its characterization. Arabian Journal of Chemistry, 8(1), 48–53. https://doi.org/10.1016/j.arabjc.2012.12.013 DOI: 10.1016/j.arabjc.2012.12.013
  55. Moliner, M., Román-Leshkov, Y., & Corma, A. (2019). Machine Learning Applied to Zeolite Synthesis: The Missing Link for Realizing High-Throughput Discovery. Accounts of Chemical Research, 52(10), 2971–2980. https://doi.org/10.1021/acs.accounts.9b00399 DOI: 10.1021/acs.accounts.9b00399
  56. Muraoka, K., Sada, Y., Miyazaki, D., Chaikittisilp, W., & Okubo, T. (2019). Linking synthesis and structure descriptors from a large collection of synthetic records of zeolite materials. Nature Communications, 10(1), 1–11. https://doi.org/10.1038/s41467-019-12394-0 DOI: 10.1038/s41467-019-12394-0
  57. Murge, P., Dinda, S., & Roy, S. (2019). Zeolite-Based Sorbent for CO2 Capture: Preparation and Performance Evaluation. Langmuir. https://doi.org/10.1021/acs.langmuir.9b02259 DOI: 10.1021/acs.langmuir.9b02259
  58. Nikiforov, A. I., Babchuk, I. V., Vorobkalo, V. A., Chesnokov, E. A., & Chistov, D. L. (2022). Application of Machine Learning Methods to Predicting the Degree of Crystallinity of MFI Type Zeolites. Petroleum Chemistry, 62(3), 322–328. https://doi.org/10.1134/S0965544122030057 DOI: 10.1134/s0965544122030057
  59. Oleksiak, M. D., Soltis, J. A., Conato, M. T., Penn, R. L., & Rimer, J. D. (2016). Nucleation of FAU and LTA Zeolites from Heterogeneous Aluminosilicate Precursors. Chemistry of Materials, 28(14), 4906–4916. https://doi.org/10.1021/acs.chemmater.6b01000 DOI: 10.1021/acs.chemmater.6b01000
  60. Pan, E., Kwon, S., Jensen, Z., Xie, M., Gómez-Bombarelli, R., Moliner, M., Román-Leshkov, Y., & Olivetti, E. (2024). ZeoSyn: A Comprehensive Zeolite Synthesis Dataset Enabling Machine-Learning Rationalization of Hydrothermal Parameters. ACS Central Science, 10(3), 729–743. https://doi.org/10.1021/acscentsci.3c01615 DOI: 10.1021/acscentsci.3c01615
  61. Perego, C., Bosetti, A., Ricci, M., & Millini, R. (2017). Zeolite Materials for Biomass Conversion to Biofuel. Energy and Fuels, 31(8), 7721–7733. https://doi.org/10.1021/acs.energyfuels.7b01057 DOI: 10.1021/acs.energyfuels.7b01057
  62. Pérez-Botella, E., Palomino, M., Valencia, S., & Rey, F. (2019). Zeolites and other adsorbents. In Kaneko, K., Rodríguez-Reinoso, F. (eds) Nanoporous Materials for Gas Storage. Green Energy and Technology. Springer, SingaporeGreen Energy and Technology. https://doi.org/10.1007/978-981-13-3504-4_7 DOI: 10.1007/978-981-13-3504-4_7
  63. Primo, A., & Garcia, H. (2014). Zeolites as catalysts in oil refining. Chemical Society Reviews, 43(22), 7548–7561. https://doi.org/10.1039/c3cs60394f DOI: 10.1039/c3cs60394f
  64. Prodinger, S., & Derewinski, M. A. (2020). Synthetic zeolites and their characterization. In Nanoporous Materials for Molecule Separation and Conversion. INC. https://doi.org/10.1016/b978-0-12-818487-5.00003-0 DOI: 10.1016/b978-0-12-818487-5.00003-0
  65. Rahmah, W., Novita, T. H., Wenten, I. G., & Kadja, G. T. M. (2023). Perspective and outlook into green and effective approaches for zeolitic membrane preparation. Materials Today Sustainability, 22, 100345. https://doi.org/10.1016/j.mtsust.2023.100345 DOI: 10.1016/j.mtsust.2023.100345
  66. Raman, G. (2023a). Identifying extra-large pore structures in zeolites with a machine learning approach and its deployment into production. Microporous and Mesoporous Materials, 348(October 2022), 112362. https://doi.org/10.1016/j.micromeso.2022.112362 DOI: 10.1016/j.micromeso.2022.112362
  67. Raman, G. (2023b). Microporous and Mesoporous Materials Identifying extra-large pore structures in zeolites with a machine learning approach and its deployment into production. Microporous and Mesoporous Materials, 348(October 2022), 112362. https://doi.org/10.1016/j.micromeso.2022.112362 DOI: 10.1016/j.micromeso.2022.112362
  68. Salwa Mohd Nazir, L., Fong Yeong, Y., & Leng Chew, T. (2019). Effect of Alkalinity Towards the Formation of NaX Zeolite Membranes. Materials Today: Proceedings, 19, 1287–1293. https://doi.org/10.1016/J.MATPR.2019.11.135 DOI: 10.1016/j.matpr.2019.11.135
  69. Sasidharan, M., & Kumar, R. (1997). Effect of various inorganic cations (Li, Na, K and Cs) and silica sources on the synthesis of the silica analogue of zeolite NCL-1 (Si-NCL-1). Microporous Materials, 8(1–2), 43–47. https://doi.org/10.1016/S0927-6513(96)00056-9 DOI: 10.1016/s0927-6513(96
  70. Schwalbe-Koda, D., Corma, A., Román-Leshkov, Y., Moliner, M., & Gómez-Bombarelli, R. (2021). Data-Driven Design of Biselective Templates for Intergrowth Zeolites. Journal of Physical Chemistry Letters, 12(43), 10689–10694. https://doi.org/10.1021/acs.jpclett.1c03132 DOI: 10.1021/acs.jpclett.1c03132
  71. Schwalbe-Koda, D., Kwon, S., Paris, C., Bello-Jurado, E., Jensen, Z., Olivetti, E., Willhammar, T., Corma, A., Román-Leshkov, Y., Moliner, M., & Gómez-Bombarelli, R. (2021). A priori control of zeolite phase competition and intergrowth with high-throughput simulations. Science, 374(6565), 308–315. https://doi.org/10.1126/science.abh3350 DOI: 10.1126/science.abh3350
  72. Schwalbe-Koda, D., Santiago-Reyes, O. A., Corma, A., Román-Leshkov, Y., Moliner, M., & Gómez-Bombarelli, R. (2022). Repurposing Templates for Zeolite Synthesis from Simulations and Data Mining. Chemistry of Materials, 34(12), 5366–5376. https://doi.org/10.1021/acs.chemmater.2c00064 DOI: 10.1021/acs.chemmater.2c00064
  73. Shobuke, H., Matsumoto, T., Hirosawa, F., Miyagawa, M., & Takaba, H. (2022). Estimation of Adsorbed Amounts in Organoclay by Machine Learning. ACS Omega, 8(1), 1146–1153. https://doi.org/10.1021/acsomega.2c06602 DOI: 10.1021/acsomega.2c06602
  74. Sours, T. G., & Kulkarni, A. R. (2022). Predicting Structural Properties of Pure Silica Zeolites Using Deep Neural Network Potentials. Journal of Physical Chemistry C, 127(3), 1455–1463. https://doi.org/10.1021/acs.jpcc.2c08429 DOI: 10.1021/acs.jpcc.2c08429
  75. Sumari, S., Fajaroh, F., Yahmin, Sholihah, N., Santoso, A., & Budianto, A. (2019). Effect of Temperature Synthesis on Structural Behaviours of NaY Zeolite Using Local Sand as A Silica Source. IOP Conference Series: Materials Science and Engineering, 515(1), 012036. https://doi.org/10.1088/1757-899X/515/1/012036 DOI: 10.1088/1757-899x/515/1/012036
  76. Tankersley, K. B., Dunning, N. P., Carr, C., Lentz, D. L., & Scarborough, V. L. (2020). Zeolite water purification at Tikal, an ancient Maya city in Guatemala. Scientific Reports, 10(1), 1–7. https://doi.org/10.1038/s41598-020-75023-7 DOI: 10.1038/s41598-020-75023-7
  77. Thakkar, H., Eastman, S., Hajari, A., Rownaghi, A. A., Knox, J. C., & Rezaei, F. (2016). 3D-Printed Zeolite Monoliths for CO2 Removal from Enclosed Environments. ACS Applied Materials and Interfaces, 8(41), 27753–27761. https://doi.org/10.1021/acsami.6b09647 DOI: 10.1021/acsami.6b09647
  78. Tomita, J., Elangovan, S. P., Itabashi, K., Chokkalingam, A., Fujinuma, H., Hao, Z., Kanno, A., Hayashi, K., Iyoki, K., Wakihara, T., & Okubo, T. (2022). OSDA-free synthesis of zeolite beta: Broadening the methodology for a successful use of the product as a seed. Advanced Powder Technology, 33(9), 103741. https://doi.org/10.1016/j.apt.2022.103741 DOI: 10.1016/j.apt.2022.103741
  79. Wang, J., Guan, Y., Zhang, Q., Zhu, H., Li, X., Li, Y., Dong, Z., Yuan, G., & Cong, Y. (2022). Well-dispersed ultrafine Pt nanoparticles anchored on oxygen-rich surface of V2CTx (MXene) for boosting hydrogen evolution reaction. Applied Surface Science, 582, 152481. https://doi.org/10.1016/J.APSUSC.2022.152481 DOI: 10.1016/j.apsusc.2022.152481
  80. Wang, L., Zhang, J., Yi, X., Zheng, A., Deng, F., Chen, C., Ji, Y., Liu, F., Meng, X., & Xiao, F. S. (2015). Mesoporous ZSM-5 zeolite-supported ru nanoparticles as highly efficient catalysts for upgrading phenolic biomolecules. ACS Catalysis, 5(5), 2727–2734. https://doi.org/10.1021/acscatal.5b00083 DOI: 10.1021/acscatal.5b00083
  81. White, C. E., Provis, J. L., Proffen, T., & Van Deventer, J. S. J. (2011). Quantitative mechanistic modeling of silica solubility and precipitation during the initial period of zeolite synthesis. Journal of Physical Chemistry C, 115(20), 9879–9888. https://doi.org/10.1021/JP2006217/ASSET/IMAGES/MEDIUM/JP-2011-006217_0008.GIF DOI: 10.1021/jp2006217/asset/images/medium/jp-2011-006217_0008.gif
  82. Wu, M., Zhang, S., & Ren, J. (2025). AI-empowered digital design of zeolites: Progress, challenges, and perspectives. APL Materials, 13(2), 020601. https://doi.org/10.1063/5.0253847/3333679 DOI: 10.1063/5.0253847/3333679
  83. Wu, Q., Meng, X., Gao, X., & Xiao, F. S. (2018). Solvent-Free Synthesis of Zeolites: Mechanism and Utility [Research-article]. Accounts of Chemical Research, 51(6), 1396–1403. https://doi.org/10.1021/acs.accounts.8b00057 DOI: 10.1021/acs.accounts.8b00057
  84. Xing, Q., & Blaisten-Barojas, E. (2013). A cloud computing system in windows azure platform for data analysis of crystalline materials. Concurrency and Computation: Practice and Experience, 25(15), 2157–2169. https://doi.org/https://doi.org/10.1002/cpe.2912 DOI: 10.1002/cpe.2912
  85. Xu, P., Ji, X., Li, M., & Lu, W. (2023). Small data machine learning in materials science. Npj Computational Materials, 9(1), 1–15. https://doi.org/10.1038/s41524-023-01000-z DOI: 10.1038/s41524-023-01000-z
  86. Yang, S., Lach-hab, M., Vaisman, I. I., & Blaisten-Barojas, E. (2009). Identifying zeolite frameworks with a machine learning approach. Journal of Physical Chemistry C, 113(52), 21721–21725. https://doi.org/10.1021/jp907017u DOI: 10.1021/jp907017u
  87. Yang, S., Lach-Hab, M., Vaisman, I. I., & Blaisten-Barojas, E. (2008). Machine learning approach for classification of zeolite crystals. Proceedings of the 2008 International Conference on Data Mining, DMIN 2008, January, 702–706.
  88. Yang, X., Dib, E., Lang, Q., Guo, H., Fu, G., Wang, J., Yi, Q., Zhao, H., & Valtchev, V. (2022). Silicalite-1 formation in acidic medium: Synthesis conditions and physicochemical properties. Microporous and Mesoporous Materials, 329(September 2021), 111537. https://doi.org/10.1016/j.micromeso.2021.111537 DOI: 10.1016/j.micromeso.2021.111537
  89. Yang, Z., Chen, B., Chen, H., & Li, H. (2023). A critical review on machine-learning-assisted screening and design of effective sorbents for carbon dioxide (CO2) capture. Frontiers in Energy Research, 10(January), 1–19. https://doi.org/10.3389/fenrg.2022.1043064 DOI: 10.3389/fenrg.2022.1043064
  90. Yu, J. (2007). Chapter 3 - Synthesis of Zeolites. In J. Čejka, H. van Bekkum, A. Corma, & F. B. T.-S. in S. S. and C. Schüth (Eds.), Introduction to Zeolite Science and Practice, Studies in Surface Science and Catalysis, 168, 39–103. Elsevier. https://doi.org/10.1016/S0167-2991(07)80791-9 DOI: 10.1016/s0167-2991(07
  91. Zhang, C., Chu, S., Jiang, J., Zhao, J., Wen, S., Sun, B., & Xu, W. (2022). Minute-Scale Synthesis of Nano Silicalite-1 Zeolites. Frontiers in Chemistry, 10(April), 1–7. https://doi.org/10.3389/fchem.2022.860795 DOI: 10.3389/fchem.2022.860795
  92. Zhang, H., Samsudin, I. bin, Jaenicke, S., & Chuah, G. K. (2022). Zeolites in catalysis: sustainable synthesis and its impact on properties and applications. Catalysis Science and Technology, 12(19), 6024–6039. https://doi.org/10.1039/d2cy01325h DOI: 10.1039/d2cy01325h
  93. Zhang, L., Hu, Q., Qin, Y., Liu, H., Liu, H., Cao, G., Gao, X., Song, L., & Sun, Z. (2023). Optimizing the accessibility of zeolite Y on FCC catalyst to improve heavy oil conversion capacity. Microporous and Mesoporous Materials, 359(May), 112627. https://doi.org/10.1016/j.micromeso.2023.112627 DOI: 10.1016/j.micromeso.2023.112627
  94. Zhang, L., Hu, Q., Qin, Y., Liu, H., Zhao, X., Gao, X., Song, L., & Sun, Z. (2022). Optimized Zeolite Distribution of FCC Catalysts for Promoting Heavy-Oil Catalytic Cracking. Industrial & Engineering Chemistry Research, 61(32), 11628–11635. https://doi.org/10.1021/acs.iecr.1c04656. DOI: 10.1021/acs.iecr.1c04656