Condition-aware design from molecular transport to thin-film modules—ranked for mixed-gas performance, durability, and manufacturability.
Start a ProjectCD ComputaBio connects molecular descriptors, polymer or porous-solid structure, sorption and diffusion data, thin-film morphology, and process conditions to screen gas separation membranes. The service covers dense polymers, facilitated-transport systems, inorganic and microporous membranes, and mixed-matrix membranes while keeping each material class within its valid transport physics.
Projects may begin with chemical structures, repeat units, filler structures, isotherms, pure- or mixed-gas permeation, microscopy, thermal/mechanical data, fabrication records, or module test data. Missing conditions and inconsistent units are audited before modeling; predictions are reported with applicability-domain and uncertainty flags.
Curate comparable permeability, diffusivity, solubility, selectivity, and permeance records by gas pair, temperature, pressure, humidity, film thickness, aging time, and test mode. Multi-task QSPR or graph models rank candidates only inside documented domains.
Use sorption–diffusion models, molecular simulation, free-volume analysis, or adsorption/diffusion calculations as appropriate to separate solubility and mobility contributions and test competitive uptake under mixed feeds.
Quantify pressure-dependent plasticization, physical aging, condensable-induced swelling, competitive sorption loss, pinholes, nonselective gaps, carrier deactivation, and filler–polymer debonding. Define stress conditions for validation.
Evaluate selective-layer thickness, crosslink density, filler chemistry/loading, dispersion, interfacial compatibility, support resistance, gutter layers, coating solvent, and defect tolerance against practical permeance targets.
Convert intrinsic transport into flat-sheet, hollow-fiber, or spiral-wound scenarios using area, pressure ratio, stage cut, recycle, pressure drop, concentration polarization, purity, recovery, and compression-duty constraints.
Select the next pure- and mixed-gas tests that most reduce uncertainty, including time-dependent conditioning, cycling, humidity and contaminant challenges, thickness controls, and statistically useful replicate plans.
Each stage preserves the link between material identity, membrane form, measurement protocol, model version, and final process assumption.
| Stage | Key Activities | Decision Output |
|---|---|---|
| 1. Separation Scoping | Define gas pair/mixture, contaminants, temperature, feed and permeate pressures, target purity/recovery, membrane form, and economic boundary. | Operating envelope and acceptance criteria |
| 2. Evidence & Data Audit | Normalize permeability/permeance, thickness, protocol, conditioning, aging time, and uncertainty; identify pure/mixed-gas gaps. | Traceable dataset and gap register |
| 3. Candidate Generation | Enumerate polymers, functional groups, crosslinks, fillers, interfaces, layer stacks, or porous structures within synthesis and coating rules. | Feasible material and architecture space |
| 4. Transport Modeling | Combine interpretable ML with sorption–diffusion, molecular simulation, or process models; propagate uncertainty across pressure and composition. | Condition-specific ranking and mechanisms |
| 5. Robustness & Manufacture | Stress-test aging, plasticization, humidity, condensables, interface defects, film thinning, support resistance, and module configuration. | Process window and failure-risk map |
| 6. Validation & Down-Selection | Prioritize mixed-gas, long-duration, cycling, morphology, and scale-relevant tests with explicit go/no-go thresholds. | Shortlist and executable validation plan |
Unit-normalized structures, compositions, test conditions, film forms, performance values, provenance, quality flags, and missing-data map.
Predicted permeability/permeance and mixed-gas selectivity with confidence intervals, applicability-domain flags, and key descriptors.
Solubility–diffusivity decomposition, competitive sorption interpretation, free-volume or pore-path evidence, and sensitivity analysis.
Plasticization pressure, aging trajectory, contaminant sensitivity, defect/interface risks, and recommended mitigation levers.
Selective-layer thickness, support and coating constraints, filler-loading/dispersion range, module assumptions, purity and recovery scenarios.
Shortlist, ranked experiments, controls, duration and cycling plan, measurement endpoints, decision gates, and machine-readable results.
CO₂/N₂ separation under humid flue gas, low CO₂ partial pressure, SOx/NOx traces, thin-film aging, and compression constraints.
CO₂/CH₄ and H₂S removal under high pressure with water and heavy hydrocarbons, plasticization risk, methane recovery, and recycle.
H₂/CO₂, H₂/CH₄, or H₂/N₂ service across syngas and refinery conditions with temperature, pressure ratio, and impurity tolerance.
O₂/N₂ enrichment where modest selectivity must be balanced against ultrathin-layer permeance, defect control, and module productivity.
He/CH₄ or He/N₂ concentration from dilute feeds using cascade configuration, pressure ratio, recovery, and leak-tightness constraints.
CO₂/CH₄ upgrading or vapor/gas separations with water, condensables, swelling, competitive sorption, cleaning, and intermittent operation.
Published studies show that experimentally curated permeability data can support interpretable machine-learning discovery of polymer membranes, while automated workflows can connect candidate generation with physics-based validation. These results support a hybrid workflow; they do not remove the need to validate mixed-gas transport, thin films, aging, plasticization, and process-level performance for the intended feed.
1 Barnett, J. W.; et al. Designing exceptional gas-separation polymer membranes using machine learning. Science Advances 2020, 6, eaaz4301. https://doi.org/10.1126/sciadv.aaz4301. Distributed under Open Access license CC BY-NC 4.0.
2 Yang, J.; et al. Machine learning enables interpretable discovery of innovative polymers for gas separation membranes. Science Advances 2022, 8, eabn9545. https://doi.org/10.1126/sciadv.abn9545. Distributed under Open Access license CC BY-NC 4.0.
3 Giro, R.; et al. AI powered, automated discovery of polymer membranes for carbon capture. npj Computational Materials 2023, 9, 133. https://doi.org/10.1038/s41524-023-01088-3. Distributed under Open Access license CC BY 4.0, with modification.
Our workflow keeps repeat unit or framework, membrane history, feed condition, selective-layer morphology, transport model, uncertainty, failure hypothesis, and module assumption connected. Predictions are candidate-selection tools, not guarantees of field performance. High-value designs are paired with mixed-gas, time-dependent, and scale-relevant validation. To discuss your feed, membrane chemistry, thin-film process, or existing dataset, please Contact Us or submit the Online Inquiry below.
Submit your project details below, and our team will respond within 24 hours.
Talk to our technical team about your project!
I Want To Talk