AI Zeolite Design Technology Landscape — PatSnap Eureka
AI-Accelerated Zeolite Design: The 2026 Innovation Landscape
From NLP synthesis extraction to reactive neural network potentials and generative OSDA design — six years of rapid maturation are reshaping how zeolitic frameworks are discovered, simulated, and synthesized. Explore the patent and literature signals defining this field.
Four Interconnected Sub-Domains Driving AI-Zeolite Innovation
AI-accelerated zeolite design encompasses the application of machine learning, neural network potentials (NNPs), natural language processing (NLP), high-throughput simulation, and generative algorithms to the discovery, synthesis planning, and property prediction of zeolitic frameworks. Zeolites are crystalline microporous aluminosilicates with critical roles in catalysis, separation, and adsorption across major industrial sectors.
These sub-domains are deeply interconnected: NLP populates training databases, ML potentials enable rapid energy evaluation, and the resulting predictions feed OSDA design and synthesizability ranking pipelines. The convergence is transforming a field historically governed by costly trial-and-error experimentation into one driven by data and computation.
According to WIPO trend data, AI-applied materials science filings have grown substantially across the past five years — and zeolite-specific AI applications represent one of the most technically mature sub-sectors within this broader wave. The International Zeolite Association currently recognizes approximately 250 experimentally realized topologies, yet computational databases contain millions of hypothetical frameworks — a gap that AI synthesizability prediction is now beginning to close.
For R&D teams seeking to map this space, PatSnap's materials intelligence platform provides access to the full patent and literature landscape across all four sub-domains described below.
Key Metrics Across the AI-Zeolite Design Landscape
Quantitative signals extracted from patent and literature records in the PatSnap Eureka dataset, spanning OSDA design, simulation scale, and geographic distribution of innovation.
OSDA ML Design: Candidate Pipeline Scale
From training dataset to validated candidates — Rice University's neural network compressed months of molecular dynamics into a ranked shortlist of 469 competitive OSDAs from 4,781 candidates.
Geographic Distribution of AI-Zeolite Innovation
Innovation is distributed across academic and industrial institutions in the US, UK, Spain, China, South Korea, France, Czech Republic, and Japan — no single entity dominates.
Technology Cluster Maturity by Sub-Domain
NNP simulation and OSDA ML design are the most active clusters in the 2021–2022 peak period; synthesizability prediction and NLP extraction are the fastest-growing emerging directions.
Landmark Performance Benchmarks from the Dataset
Selected quantitative achievements from key studies in the AI-zeolite literature, demonstrating the scale of computational acceleration now achievable.
Four Technical Approaches Reshaping Zeolite Discovery
Each cluster addresses a distinct bottleneck in the traditional zeolite design pipeline — from energy evaluation to synthesis planning and experimental accessibility prediction.
Neural Network Potentials for Framework Simulation
NNPs trained on DFT-computed energetics reproduce quantum-mechanical accuracy at computational costs orders of magnitude lower than ab initio methods. Charles University (2022) screened more than 20,000 additional hypothetical frameworks using the SchNet architecture with active-learning DFT database extension. Shanghai Qi Zhi Institution frames NNPs as "the core enabling technology for the next era of zeolite research." Shanghai Key Laboratory explored over 1 million minima on the global potential energy surface of the SiAlPO system.
20,000+ frameworks screened · Charles University 2022Data-Driven OSDA Design
OSDAs are organic molecules that template zeolite framework formation during hydrothermal synthesis. Rice University (2019) trained a neural network on 4,781 candidate OSDAs and identified 469 OSDAs with stabilization energies competitive with known templates for zeolite beta — replacing expensive molecular dynamics evaluation. ITQ Valencia (2021) extended this to bi-selective OSDA design for intergrowth frameworks (CHA/AFX, MTT/TON, BEC/ISV) by screening hundreds of thousands of zeolite-OSDA pairs. The chemicals intelligence capabilities within PatSnap map this innovation cluster in detail.
469 competitive OSDAs · Rice University 2019High-Throughput Screening & Synthesizability Ranking
University of Southampton (2021) achieved a 100× acceleration in in silico porous solid discovery using AI-guided simulation selection across energy-structure-function maps. University College London (2022) combined ML with high-throughput simulation to identify zeolite structures outperforming the leading commercial material for xylene separation. Sorbonne Université's "Zeolite Sorting Hat" (2022) established data-driven synthesizability ranking as a distinct pipeline layer — addressing the gap between millions of hypothetical frameworks and the ~250 IZA-recognized topologies. The EPA's growing interest in industrial separation efficiency further strengthens the business case for this cluster.
100× acceleration · University of Southampton 2021NLP-Assisted Synthesis Knowledge Extraction
MIT (2019) used NLP techniques and text markup parsing tools to automatically extract synthesis information from zeolite journal articles, achieving a cross-validated RMSE of 0.98 T/1000 ų for framework density prediction from synthesis conditions. University of Minnesota (2023) applied Geometric Harmonics, Neural Networks, and Gaussian Process Regression to capture crystallization composition–microstructure relationships for faujasite-type zeolites — achieving the highest reported Si/Al ratio of 3.5 for direct, OSDA-free FAU synthesis. Zhejiang University (2022) reviewed theoretical simulation of zeolite-template interaction as a guide for targeted synthesis, encompassing template design, novel compositions, and new topology generation.
Si/Al 3.5 record · University of Minnesota 2023Where AI-Zeolite Design Is Being Applied
From xylene separation to SCR catalysis and CO₂ capture — the application landscape spans multiple high-value industrial sectors, each with distinct AI-acceleration opportunities.
Petrochemical Separation & Refining
Xylene isomer separation is a high-value industrial process where AI-driven zeolite discovery is directly applied. University College London (2022) identified zeolite candidates superior to commercially deployed materials for xylene separation. Department of Bioengineering (2020) extended AI-guided zeolite design to light alkene/alkane separations critical for polyolefin production via computationally designed OSDAs for ethylene–ethane separation.
Automotive Emissions Control (NH₃-SCR)
The AFX zeolite mining project at Tokyo Institute of Technology (2021), involving automotive industry partners, demonstrates zeolite discovery guided by computational topology screening for selective catalytic reduction catalysts. Johnson Matthey's active patent portfolio on AEI zeolite synthesis (EP 2021, GB 2018, EP 2023) addresses the same domain — AEI and AFX topologies are leading candidates for Cu-zeolite SCR catalysts used in diesel aftertreatment.
Key Institutions & Patent Assignees in This Dataset
Innovation is distributed across academic and industrial institutions in 8+ countries. The IP protection layer for AI-driven zeolite design is still emerging — the overwhelming majority of AI-specific innovation appears in academic literature rather than granted patents.
| Institution / Assignee | Country | Primary Contribution | Record Type | Status |
|---|---|---|---|---|
| Johnson Matthey | UK | AEI zeolite synthesis for SCR catalysis — 4 active patent records | Patent | Active · EP, GB |
| ITQ Valencia (CSIC) | Spain | OSDA ML design, phase selectivity control, intergrowth frameworks — most prolific AI-OSDA institution | Literature | 3+ publications 2021–2022 |
| MIT | USA | NLP-based synthesis data extraction; regression model RMSE 0.98 T/1000 ų | Literature | 2019 landmark study |
| Charles University | Czech Republic | Reactive NNPs on DFT databases; screened 20,000+ hypothetical frameworks | Literature | 2022 |
| University of Southampton | UK | 100× acceleration in porous solid discovery via AI-guided simulation selection | Literature | 2021 |
| Rice University | USA | Neural network on 4,781 OSDA candidates; 469 competitive OSDAs for zeolite beta | Literature | 2019 |
| Tosoh Corporation | Japan | AEI zeolite production method | Patent | Active · EP 2021 |
| Metacl Co., Ltd. | South Korea | Generative AI-based plant design automation platform; filed December 2025 | Patent | Pending · KR 2025 |
Map freedom-to-operate against Johnson Matthey's AEI patent cluster
Four active patent records across EP and GB jurisdictions define a key defensive position in SCR catalyst zeolite synthesis.
Five Directions Gaining Momentum in 2022–2025
Based on the most recent records in this dataset, these directions signal where AI-zeolite design is headed — from plant-level generative AI to OSDA-free synthesis at record Si/Al ratios.
Generative AI for Plant-Level Process Design
A KR pending patent filed December 2025 (Metacl Co., Ltd.) describes an AI system performing deep learning on existing plant design data to create derivation models for automated process design via generative AI and metaverse environments. While not zeolite-specific, it signals industrial translation of generative AI into chemical plant design — the next step beyond material discovery. IP analytics platforms can track this emerging industrial IP cluster in real time.
KR Patent · Metacl Co., Ltd. · Dec 2025Synthesizability Prediction as a Standalone Capability
The Zeolite Sorting Hat framework (Sorbonne Université, 2022) establishes data-driven synthesizability ranking as a distinct layer in the design pipeline, enabling the field to close the gap between the millions of hypothetical frameworks in computational databases and the approximately 250 experimentally realized IZA-recognized topologies. This addresses the critical challenge of translating computational predictions into experimentally accessible targets.
Sorbonne Université · 2022 · Synthesizability RankingActive Learning for NNP Training
Charles University's reactive NNP work (2022) specifically employs active learning to iteratively extend the DFT training database to cover high-energy transition states, not just equilibrium configurations. This enables ML potentials to handle reactive processes such as nucleation and condensation — extending beyond static stability screening. According to Nature publications in this space, active learning is becoming the standard approach for extending ML potential coverage to reactive regimes.
Charles University · 2022 · Active Learning NNPsIntergrowth Design & OSDA-Free Crystal Engineering
ITQ's bi-selective OSDA work (2021) and Johnson Matthey's mixed-template AEI/ITE synthesis patent (EP, 2023) both target intergrown and compositionally complex frameworks. Separately, the University of Minnesota's 2023 demonstration of ML-guided FAU synthesis without an OSDA at a record Si/Al ratio of 3.5 addresses both cost and sustainability drivers simultaneously — pointing toward a high-value research frontier. PatSnap customers in materials R&D are already applying these signals to guide synthesis programs.
Minnesota 2023 · Si/Al 3.5 · OSDA-free FAUWhat the IP and Research Signals Mean for R&D Teams
The IP gap is real and exploitable. In this dataset, AI-accelerated zeolite design innovation is heavily concentrated in academic literature with minimal patent coverage. R&D teams and IP strategists entering this space face a relatively open patent landscape for AI-driven synthesis prediction, OSDA generative design, and NNP-based screening workflows — though this window will narrow as industrial adoption accelerates.
OSDA design is the highest near-term leverage point. Multiple independent groups — Rice University, ITQ Valencia, Department of Bioengineering — have demonstrated that ML/NN-based OSDA design can replace months of molecular dynamics computation. Productizing this capability — whether as software, a synthesis service, or a proprietary database of designed OSDAs — represents a clear commercialization pathway.
Johnson Matthey's AEI patent cluster defines a key defensive position. With four active patent records across EP and GB jurisdictions, Johnson Matthey has established a substantial IP position around AEI zeolite synthesis for SCR catalysis. Competitors targeting Cu-AEI or Cu-AFX SCR catalysts should carefully map freedom-to-operate against this portfolio using PatSnap's patent analytics tools.
Sustainability pressures will accelerate AI adoption. Multiple records in this dataset (BASF SE sustainability perspective, 2022; solvent-free and green synthesis reviews from China University of Petroleum, Jiaxing University) document that conventional zeolite manufacturing is energy-intensive, water-intensive, and CO₂-heavy. AI-accelerated design that reduces synthesis trial-and-error, enables OSDA-free routes, and optimizes process conditions directly addresses regulatory and ESG pressures — strengthening the business case for AI investment across the value chain. The OECD's industrial sustainability frameworks increasingly incentivize exactly this kind of computational-first materials development.
AI-Accelerated Zeolite Design — key questions answered
The field divides into four principal technical sub-domains: machine learning potentials and atomistic simulation (training interatomic potentials on DFT data), data-driven OSDA design (using ML and molecular dynamics to identify or generate OSDAs that selectively template target zeolite topologies), high-throughput screening and synthesizability ranking (deploying ML classifiers to pre-screen millions of hypothetical zeolite structures), and NLP-driven synthesis knowledge extraction (applying text mining and NLP to scientific literature to extract synthesis rules and build training datasets automatically).
Charles University (2022) trained NNPs with the SchNet architecture on an iteratively extended, active-learning DFT database and screened more than 20,000 additional hypothetical frameworks beyond those in existing databases, retaining DFT accuracy for thermodynamic stabilities, vibrational properties, and phase transformations.
In this dataset, AI-accelerated zeolite design innovation is heavily concentrated in academic literature with minimal patent coverage. R&D teams and IP strategists entering this space face a relatively open patent landscape for AI-driven synthesis prediction, OSDA generative design, and NNP-based screening workflows — though this window will narrow as industrial adoption accelerates. Johnson Matthey is the most active assignee in this dataset with four active patent records (EP, GB, BR jurisdictions), focused on AEI framework synthesis.
MIT (2019) used NLP techniques and text markup parsing tools to automatically extract synthesis information from zeolite journal articles. A regression model predicts framework density from synthesis conditions with a cross-validated RMSE of 0.98 T/1000 ų, establishing that literature mining could create training sets for zeolite synthesis regression models.
The University of Southampton (2021) reported a 100× acceleration in the in silico design of porous materials using AI-guided intelligent selection of simulations, demonstrated across the energy-structure-function landscape.
The University of Minnesota (2023) used ML algorithms including Geometric Harmonics, Neural Networks, and Gaussian Process Regression to capture crystallization composition–microstructure relationships for faujasite-type zeolites, enabling identification of synthesis conditions that achieve the highest reported Si/Al ratio of 3.5 for direct, OSDA-free FAU synthesis.
Still have questions? Let PatSnap Eureka answer them for you.
Ask Eureka About Zeolite PatentsAccelerate Your AI-Zeolite R&D with Eureka's Patent Intelligence
Join 18,000+ innovators already using PatSnap Eureka to map technology landscapes, identify IP gaps, and accelerate their R&D programs.
References
- Machine Learning Potential Era of Zeolite Simulation — Shanghai Qi Zhi Institution, 2022
- Thermodynamic Rules for Zeolite Formation from Machine Learning Based Global Optimization — Shanghai Key Laboratory of Molecular Catalysis and Innovative Materials, 2020
- Machine Learning Accelerated High-Throughput Screening of Zeolites for the Selective Adsorption of Xylene Isomers — University College London, 2022
- A Machine Learning Approach to Zeolite Synthesis Enabled by Automatic Literature Data Extraction — MIT, 2019
- Accurate Large-Scale Simulations of Siliceous Zeolites by Neural Network Potentials — Charles University, 2022
- Data-Driven Design of Bi-Selective OSDAs for Intergrowth Zeolites — ITQ, Universitat Politècnica de València / CSIC, 2021
- Targeted Synthesis of Zeolites from Calculated Interaction between Zeolite Structure and Organic Template — Zhejiang University, 2022
- Simplifying Computational Workflows with MAZE — University of California, Davis, 2021
- Machine-Enabled Inverse Design of Inorganic Solid Materials: Promises and Challenges — KAIST, 2020
- Accelerating Computational Discovery of Porous Solids through Improved Navigation of Energy-Structure-Function Maps — University of Southampton, 2021
- Machine-Learning Approach to the Design of OSDAs for Zeolite Beta — Rice University, 2019
- A Priori Control of Zeolite Phase Competition and Intergrowth with High-Throughput Simulations — ITQ, Universitat Politècnica de València / CSIC, 2021
- Ranking the Synthesizability of Hypothetical Zeolites with the Sorting Hat — Sorbonne Université, 2022
- Design of OSDAs to Guide the Synthesis of Zeolites for the Separation of Ethylene–Ethane Mixtures — Department of Bioengineering, 2020
- Machine Learning-Assisted Crystal Engineering of a Zeolite — University of Minnesota, 2023
- Artificial Intelligence for Crystal Growth and Characterization — University of Michigan, 2022
- AFX Zeolite for Use as a Support of NH3-SCR Catalyst — Tokyo Institute of Technology, 2021
- AEI Zeolite Synthesis — Johnson Matthey Public Limited Company, EP 2021
- AEI Zeolite Synthesis — Johnson Matthey Public Limited Company, GB 2018
- Synthesis of AEI Zeolite — Johnson Matthey Public Limited Company, EP 2023
- Aluminosilicate AEI Zeolite Preparation — Johnson Matthey Public Limited Company, EP 2021
- Method for Producing AEI Zeolite — Tosoh Corporation, EP 2021
- Platform for Generative AI-Based Plant Design Automation — Metacl Co., Ltd., KR 2025
- Toward Sustainability in Zeolite Manufacturing: An Industry Perspective — BASF SE, 2022
- Active Precursor Promoting Nucleation/Growth of MWW Zeolite — China University of Petroleum, Beijing, 2023
- WIPO — World Intellectual Property Organization: AI in Materials Science Patent Trends
- Nature — Active Learning for Machine Learning Potentials in Reactive Chemistry
- OECD — Industrial Sustainability Frameworks and AI-Driven Materials Development
All data and statistics on this page are sourced from the references above and from PatSnap's proprietary innovation intelligence platform. This landscape is derived from a limited set of patent and literature records retrieved across targeted searches and represents a snapshot of innovation signals within this dataset only.
PatSnap Eureka searches patents and research to answer instantly.