Solutions · Formulations & CASE

Formulations: constrained designs, solubility screening and costed candidates

The public mirror of the Formulations & CASE depth studio. Designs and solubility estimates are computed, reformulations are predicted, and each output carries its evidence class and interval band.

Jobs to be done

What formulations teams come here to do

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

  • Design on the simplex — Generate D-optimal (coordinate exchange), simplex-lattice or extreme-vertex designs under per-component bounds — every row sums to 100% (/formulations/depth).
  • Screen solvents by solubility — Hansen parameters from group contribution, ranked by Ra distance and RED (RED < 1 means the solvent is inside the sphere) (/formulations/depth).
  • Compute BOM properties with bands — Mixture density, viscosity and $/kg with an uncertainty band and a live mass-balance check (/formulations/depth).
  • Reformulate to targets — Search alternatives that meet property and cost targets when a component is restricted or expensive — labelled PREDICTED (/formulations/depth).
  • Price before the bench — TEA turns the shortlist into raw-material, equipment, CapEx/OpEx and end-of-life numbers (/studio/tea).
Evidence & limits

What the outputs can honestly support

Screening-grade where true; engine-labelled everywhere.

Screening-grade, and labelled that way.Mixture designs and Hansen parameters are COMPUTED group-contribution estimates (HSP roughly ±2-3 MPa^0.5); BOM mixture properties carry an uncertainty band; reformulation is PREDICTED. The designs decide which experiments are worth running — the bench remains the final word. Matflow does not certify regulatory compliance (VOC, REACH or otherwise); those statements belong to your own regulatory process.
FAQ

Questions before you start

Including the limitations we would rather state now.

How accurate is the Hansen solubility estimate?
It is a group-contribution screening approximation reported with roughly ±2-3 MPa^0.5 uncertainty — a triage tool for choosing experiments, not a replacement for measurement (client/src/pages/docs/content.jsx, domain-packs/formulations).
Are mixture constraints enforced or filtered afterward?
Design rows sum to 1 under the bounds you set, and optimization applies compositional constraints during candidate generation rather than filtering infeasible rows later (client/src/pages/docs/content.jsx, models/optimization_algorithms).
What evidence class is the reformulation output?
PREDICTED. Mixture design and HSP/RED results are COMPUTED; a suggested reformulation is a prediction to test (client/src/pages/docs/content.jsx, domain-packs/formulations).
Does Matflow check VOC or REACH compliance?
No — Matflow does not certify regulatory compliance. Cost, property bands and solubility screening are the verifiable outputs here; any compliance statement should come from your own regulatory process.
Can I bring measured solubilities?
Yes. Upload measured data as a dataset and use it alongside the screening estimates; predictions train on your data and carry conformal intervals.

Design the next plate, not the same corner

Build a constrained mixture design, rank solvents by RED, and export a costed shortlist — free tier, no card.