Structure & Data Preparation
Standardize structures, remove disorder and duplicates, identify missing atoms, assign provenance and prepare computation-ready libraries.
AI for Materials
AI-Driven MOF and COF Screening
Physics-aware screening and design of porous frameworks for adsorption, separation, catalysis, sensing, energy and environmental applications.
Start Your ProjectPorous framework discovery
Screen the right framework for the real operating window
MOFs and COFs offer vast, modular design spaces, but geometric porosity alone does not establish adsorption selectivity, stability, transport, reactivity or synthesizability. We combine curated structures, molecular simulation and fit-for-purpose machine learning to reduce a large candidate space to a defensible validation set.
Discuss Your Screening TargetStructure-awarePore geometry, chemistry, defects and flexibility
Condition-specificPressure, temperature, humidity and mixtures
Validation-readyRanked candidates with model boundaries
Core screening capabilities
An integrated MOF and COF screening framework
The project scope is assembled from the capabilities needed for the target decision, without requiring separate service modules.
Pore & Topology Analysis
Quantify accessible surface area, pore-limiting diameter, void fraction, topology and chemistry relevant to molecular access.
Adsorption & Separation Screening
Evaluate uptake, working capacity, selectivity and regenerability using GCMC, mixture conditions and ML surrogates.
Reactive Site Prioritization
Compare nodes, linkers, defects and confined sites through electronic descriptors, DFT calculations and reaction-aware models.
Stability & Feasibility Assessment
Assess thermal, mechanical, chemical and hydrolytic risks together with precursor feasibility and literature evidence.
Functional Property Prediction
Evaluate electronic structure, charge transport, optical response and structure–property relationships for functional frameworks.
Transport & Storage Evaluation
Screen molecular or ion-accessible pores, binding, diffusion and conductivity-relevant features under defined conditions.
Candidate Design & Optimization
Explore constrained combinations of nodes, linkers and topologies, then filter candidates for validity, feasibility and performance.
Integrated approach
A gated workflow, not a single black-box score
Fast models are useful only when structures, operating conditions and prediction boundaries are handled explicitly.
✓
Structure preparation and quality gates
Normalize cells, resolve disorder where possible, remove inaccessible structures and document assumptions.
✓
Multiscale evidence generation
Use geometric analysis, GCMC, MD, DFT or learned interatomic potentials according to the endpoint.
✓
Framework-aware learning
Combine pore descriptors, chemistry, topology and three-dimensional graph representations with customer data.
✓
Applicability and uncertainty control
Detect out-of-domain structures, compare model ensembles and reserve high-fidelity calculation for informative candidates.
Method selection
Methods aligned to the decision endpoint
The computational stack is selected by framework chemistry, target property, data density and required fidelity.
| Project need | Typical methods | Key inputs | Decision output |
|---|---|---|---|
| Library triage | Structure cleaning, deduplication, pore analysis, rule filters | CIFs, framework source, guest-removal rules | Valid computation-ready library |
| Adsorption and separation | GCMC, Henry coefficients, IAST where appropriate, ML surrogates | Frameworks, adsorbates, force fields, T/P/composition | Uptake, selectivity, working capacity and ranking |
| Diffusion and flexibility | MD, free-energy methods, MLIPs, flexible-framework analysis | Atomic model, guest loading, force model, conditions | Diffusivity, barriers and flexibility risk |
| Stability and reactivity | DFT, phonon/energy analysis, defect modeling, literature-informed classifiers | Node/linker chemistry, defects, environment | Relative stability, reactive-site and degradation hypotheses |
| Large-space optimization | GNNs, Gaussian processes, active learning, constrained Bayesian optimization | Descriptors, labels, design constraints, uncertainty target | Pareto-ranked candidates and next calculations |
| New framework generation | Building-block assembly, generative models, validity and synthesizability filters | Allowed nodes, linkers, topologies and property profile | Novel candidates with traceable design rationale |
Scientific boundary: adsorption predictions depend on structure quality, charge assignment, force-field suitability and the treatment of defects, water and flexibility. Model validation is defined for the intended chemical and operating domain.
Project workflow
From target profile to an actionable shortlist
Define the decision
Set the application, operating window, hard constraints, comparator and validation endpoint.
Audit structures and data
Review CIF quality, provenance, target labels, simulation consistency and experimental uncertainty.
Generate evidence
Calculate descriptors and selected high-fidelity labels with controlled protocols.
Model and screen
Validate models without chemical leakage, quantify uncertainty and rank the design space.
Recommend validation
Deliver candidates, trade-offs, failure modes and the next simulation or experiment matrix.
Typical inputs
- Application, adsorbate or reaction and operating conditions
- Customer CIFs, candidate list or permitted framework families
- Target metrics, comparator materials and hard exclusions
- Experimental isotherms, PXRD, stability or transport data, if available
- Allowed metals, linkers, solvents, cost and synthesis constraints
- Preferred force fields, charge schemes or internal protocols
Typical deliverables
- Curated framework library with structure-quality annotations
- Documented simulation and modeling protocols
- Property predictions with uncertainty and applicability flags
- Ranked candidates and multi-objective trade-off analysis
- Pore, chemistry and topology drivers of performance
- Failure-risk and sensitivity assessment
- Recommended high-fidelity calculations or validation experiments
- Technical report, figures and agreed machine-readable outputs
Representative engagements
Screening designed around concrete R&D choices
Humid gas-separation shortlist
Goal: identify frameworks balancing mixture selectivity, working capacity and water tolerance. Workflow: structure audit, geometric filtering, mixture-aware simulation, surrogate screening and stability risk flags.
COF ion-transport design space
Goal: prioritize pore chemistries and stacking motifs for ion accessibility and mobility. Workflow: topology descriptors, binding calculations, MD-derived transport evidence and multi-objective ranking.
MOF catalyst site prioritization
Goal: compare nodes, defects and linker environments for a target transformation. Workflow: site enumeration, DFT descriptors, reaction-path calculations on selected systems and interpretable candidate ranking.
FAQ
Frequently asked questions
Can you screen both experimental and hypothetical MOF/COF structures?
Yes. We keep their provenance separate and apply different confidence gates. Hypothetical structures require stronger validity, stability and synthesizability checks before performance ranking is treated as actionable.
What if our framework structures contain solvent, disorder or missing atoms?
We first audit the structures. Guest removal, disorder resolution, charge assignment and missing-atom repair are documented; structures that cannot be prepared defensibly are flagged rather than silently included.
Can models account for humidity, gas mixtures and framework flexibility?
These effects can be included when the project data and computational models support them. Static single-component screening may be used as an early gate, followed by mixture, water-competition or flexible-framework calculations for selected candidates.
Can you combine our experimental data with simulations and public structures?
Yes. Data sources are harmonized with provenance and domain labels. We use validation splits that reduce leakage across related framework families and report performance for the customer-relevant domain.
Do you guarantee that a top-ranked hypothetical framework can be synthesized?
No computational ranking can guarantee synthesis. We can incorporate building-block availability, reported reaction families, stability indicators and literature evidence, then propose a validation sequence for experimental teams.
How is confidential data handled?
Project data can be isolated and used only within the agreed scope. Data access, retention, reusable model outputs and reporting requirements can be defined before work begins.
Online Inquiry
Submit your project details below, and our team will respond within 24 hours.
x
Need help getting the data you need?
Talk to our technical team about your project!
I Want To Talk