Solutions · Polymers

Polymers: repeat-unit QSPR, blend miscibility and copolymer Tg

The public mirror of the Polymers studio. Every QSPR output is a screening-grade prediction calibrated on a 23-polymer reference set — never a measurement — and simulation results carry the engine that ran them.

Jobs to be done

What polymers teams come here to do

Each job maps to a module that exists — the links go straight into the platform.

Estimate properties from repeat-unit SMILES — Tg (K), density and Hildebrand solubility parameter from RDKit fragment counts, calibrated on a 23-polymer reference set (/studio/polymers).
Judge blend miscibility — Flory-Huggins chi from the solubility-parameter difference, a RED-style distance and a MISCIBLE / UCST-LIKE / IMMISCIBLE verdict (/studio/polymers).
Design copolymers by composition — Fox-equation Tg and linear solubility-parameter mixing across composition (/studio/polymers).
Cross-check with molecular dynamics — NVT/NPT decks, radial distribution functions, Tg cooling and amorphous cell building, with engine-labelled results (/studio/md).
Relax and compare with potentials — Screen structures with the machine-learned potentials actually installed on the deployment; unavailable ones are shown as unavailable (/studio/mlip).

Modules

The real pieces behind this workflow

No roadmap items and no imagined features: every card opens a shipped module route.

See all modules

QSPR studio

Polymers (QSPR)

Repeat-unit property estimates, Flory-Huggins blend screening and Fox-equation copolymer Tg in one page.

Simulation

Molecular Dynamics

Server-side GROMACS, OpenMM and LAMMPS decks with trajectory analysis — nothing is simulated in the browser.

Simulation

MLIP Potentials

Structure relaxation, MD and equation-of-state fits with the potentials the registry reports as installed.

Modeling

Prediction

Train the explainable ensemble on your measured polymer data, with conformal intervals and an extrapolation flag.

Optimization

Multi-objective search

Trade Tg, cost and other objectives against each other on a Pareto front rather than a single score.

Evidence & limits

What the outputs can honestly support

Screening-grade where true; engine-labelled everywhere.

Screening-grade QSPR — state it, do not hide it.Tg, density and solubility outputs are PREDICTED screening approximations from repeat-unit fragment counts, calibrated on a 23-polymer reference set (Van Krevelen / Bicerano style) — never measurements. The blend mixing rule is COMPUTED; the MISCIBLE / UCST-LIKE / IMMISCIBLE verdict is PREDICTED. MD and MLIP results are labelled by the engine that actually ran; an unavailable potential is reported as unavailable, not substituted silently.

Polymers domain pack →Molecular dynamics →Machine-learned potentials →Atomistic simulations →

FAQ

Questions before you start

Including the limitations we would rather state now.

Are the Tg and solubility predictions experimental quality?

No. They are screening-grade QSPR approximations from repeat-unit fragment counts, calibrated on a 23-polymer reference set. Use them to prioritize candidates, then measure (client/src/pages/docs/content.jsx, domain-packs/polymers).

What does the blend verdict mean?

It is a PREDICTED classification from a Flory-Huggins chi estimate — useful for ranking candidates, not a measured phase diagram. Confirm critical blends experimentally.

Can Matflow model the copolymer Tg curve?

Yes — Fox-equation Tg and linear solubility-parameter mixing versus composition (client/src/pages/docs/content.jsx, domain-packs/polymers).

Is there a real simulation path?

Yes — the MD studio builds server-side GROMACS/OpenMM/LAMMPS decks and the MLIP studio relaxes structures with installed potentials; each result names the engine and evidence class (docs/md/overview, mlip/overview).

Where do I start?

Paste or draw a repeat unit on /studio/polymers, then move promising candidates to blend screening, prediction or MD.

Screen the polymer space before the reactor time

Estimate Tg and solubility from a repeat unit, rank blends by chi, and forward the shortlist to simulation — free tier, no card.