From bench to plant — one reaction, twenty-five minutes.
Instead of telling you what ScaleChem does, here's a real scale-up problem walked through the suite end-to-end. Real numbers, real recommendations, real time on the clock.
Define the reaction once.
You start in Reactions. Paste in your stoichiometry — diethyl malonate, benzyl bromide, K₂CO₃, DMF — and PubChem auto-fills molecular weight, density, and boiling point. ScaleChem builds the table. Every downstream tool reads from this same definition; you'll never re-key it.
| Compound | Role | Eq | MW | Mol | Mass | Vol |
|---|---|---|---|---|---|---|
| Diethyl malonate | Substrate | 1.0 | 160.17 | 0.500 | 80.1 g | 76.0 mL |
| Benzyl bromide | Reagent | 1.0 | 171.04 | 0.500 | 85.5 g | 59.5 mL |
| K₂CO₃ | Base | 1.2 | 138.21 | 0.600 | 82.9 g | — |
| DMF | Solvent | — | 73.09 | — | 816 g | 865 mL |
Save this as a template. Next month's scale-up reads it back in one click.
Heat transfer at 50L — the cooling margin compresses.
ScaleChem pulls your reaction enthalpy (ΔH ≈ −85 kJ/mol from literature; you can override), computes peak heat generation rate from your addition profile, and balances it against your jacket cooling capacity at every scale.
At 1L the cooling margin is 9.5× — wildly comfortable. At 50L with the same 60-minute addition, it drops to 4.6×. Still nominally safe, but ScaleChem flags it as tight: any cooling hiccup or jacket fluid degradation pushes you below the 3× safety threshold during the peak window.
The same calculation in a hand-built Excel sheet — assuming you already have one — takes a couple of hours per scale, and requires you to find Sieder-Tate, look up your jacket fluid's Cp and viscosity, and re-derive the heat balance if anything changes. ScaleChem ran all three scales above in 90 seconds.
Mixing — verify regime similarity.
Heat transfer caught one risk. Now verify the mixing regime doesn't shift between scales. ScaleChem computes Reynolds, Power number, mixing time, Damköhler, and tip speed for each setup. Lab uses a magnetic stir bar; pilot uses a 200 mm pitched-blade turbine at 200 RPM.
| Group | Lab (1L, stir bar) | Pilot (50L, PBT) | Δ |
|---|---|---|---|
| Reynolds (Re) | 9,200 | 137,000 | 15× ↑ both turbulent ✓ |
| Mixing time θ₉₅ | 5.2 s | 4.5 s | comparable ✓ |
| Tip speed | 0.94 m/s | 2.09 m/s | 2.2× — within shear-OK band |
| Damköhler (Da) | 0.04 | 0.05 | fast-rxn regime ✓ |
| Power per volume | 0.18 kW/m³ | 0.12 kW/m³ | −33% — still adequate for solid base |
Yield prediction — where the loss comes from.
With heat and mixing analyses in hand, ScaleChem predicts your pilot yield with a per-mechanism breakdown. Lab observed 91%. The model auto-pulls your dimensionless and HT numbers — no re-keying — and projects 50L performance:
Equipment recommendation — concrete, with chiller spec.
All four prior analyses feed into the Equipment Recommender. It synthesizes them into specific gear: reactor type, impeller, jacket fluid, chiller capacity, and addition strategy. Justifications cite your actual numbers, not generic rules-of-thumb.
Process safety — which standards apply, and what could go wrong.
Before the pilot run, you need two answers: which regulations apply (so you don't ship a process under-permitted), and what scenarios should the operating procedure cover. ScaleChem's Process Safety tool answers both from your reagent inventory and node breakdown — no separate consulting engagement.
Then the heuristic generator walks each node — charge, base addition, alkylation, quench, extraction, evaporation — and produces a HAZOP-style worksheet. For this chemistry it surfaces about 40 scenarios on the calibrated 5×5 risk matrix. The headlines:
The tool is heuristic — it produces the draft. Your multidisciplinary PHA team validates every row before sign-off (per OSHA 1910.119(e)(4) team-composition requirements). What ScaleChem saves is the 2-3 weeks of pre-meeting prep, not the team review itself.
The applicability engine no longer just "checks OSHA and EPA." Reagent inventories are evaluated against the canonical published threshold lists, line-by-line, with the entry counts shown below. The verdict is heuristic input to your formal applicability determination — never a substitute for it — but it catches the obvious calls before your EHS even opens the spreadsheet.
- OSHA 1910.119 Appendix A — 133 entries (PSM Highly Hazardous Chemicals)
- EPA 40 CFR §68.130 Table 1 — 77 entries (RMP toxic substances)
- EPA 40 CFR §68.130 Table 3 — 62 entries (RMP flammable substances)
- 10,000 lb aggregate flammable trigger — auto-summed across your inventory
- CFATS 6 CFR Part 27 Appendix A — partial subset (security threshold check)
- CalARP 19 CCR §5130.6 Table 3 — full 272 entries (California — stricter than federal)
- NV CAPP NAC 459.9533 — full 213 entries (Nevada)
- NJ TCPA N.J.A.C. 7:31-6.3 — 8 NJ-specific overrides
- DE EHS — Acute Toxicity Concentration methodology (not a fixed list)
- Contra Costa County ISO — adds to CalARP for that jurisdiction
- R&D exemption — PSM 1910.119(a)(2)(iii) and RMP §68.115(b)(5). Lab activities under qualified supervision; OSHA interprets narrowly.
- Atmospheric tank exclusion — PSM 1910.119(a)(1)(ii)(B) for flammable liquids
- Hydrocarbon fuel exclusion — PSM 1910.119(a)(1)(ii)(A)
- Aqueous-vs-anhydrous handling — per OSHA enforcement: "anhydrous" listings don't cover aqueous solutions
- ~180-entry chemical library — OSHA & EPA listed substances, common solvents, organometallics, oxidizers, reactive intermediates. Hazard data + threshold quantities auto-fill.
- Calibrated 5×5 risk matrix — severity vs. likelihood with consequence-based anchoring
- Heuristic scenario generator — 30–60 scenarios per process, chemistry-aware (e.g., DMF + carbonate-base thermal hazard auto-flagged)
- AI-assist available as Pro add-on for novel chemistry
Cost per kilogram at production — and the levers.
Pilot is just the bridge. Where does this land at 1500 kg/year production? ScaleChem's TEA models a 200L production line: 24 batches/year, 64 kg product per batch, full direct + overhead build-up. With DMF recycled at 70% (industry standard for this chemistry):
The walkthrough above assumes you already have the conditions: DMF, 80 °C, 4 h, BnBr added over 60 min, 91% yield. But what if you don't yet? On a 1,000-candidate design space (5 catalysts × 5 solvents × 10 temps × 4 pressures), the ScaleChem optimizer reaches 99% of the true optimum after running just ~40 experiments — about 4% of the design. A first-generation Bayesian optimizer plateaus at ~83% yield and never reaches the optimum, even after 500 picks. Same data, same batch size, same starting state.
Multi-restart maximum-likelihood fit with per-dimension lengthscales (ARD). The model figures out which variables drive your reaction. No hyperparameter tuning, no defaults to guess at.
Kriging Believer batch selection: when you ask for 5 picks, you get 5 that complement each other — not 5 clustered around the same predicted optimum.
Constraints (e.g. temp ≤ 80 °C), multi-objective with a goal toggle, replicate handling, outlier detection with leave-one-out residuals, cross-validation diagnostic, target-reached stopping signals, team workspaces with role-based permissions.
Twenty-five minutes from question to defended pilot plan.
Real engineering. Real numbers. Calibrated to the chemistry you actually run. Here's how that compares to the alternatives:
A successful first pilot batch is worth $50k–$500k depending on the chemistry. ScaleChem pays for itself the first time it catches a problem — like the 60-min addition that would have run hot at 50L.
Worked example uses real published thermodynamic and engineering data for the malonate alkylation chemistry. Numbers shown are first-principles estimates calibrated against literature; ScaleChem's actual outputs reflect your reaction's specific conditions, properties, and equipment. All scale-up recommendations should be reviewed by a qualified process safety engineer before pilot or production runs.