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.

The scenario
You ran an SN2 alkylation of diethyl malonate with benzyl bromide at 1L bench scale with 91% yield. Your boss wants 50L pilot batches next month, then 1500 kg/year at production. Will the chemistry survive the scale-up? What size reactor? What does it cost?
90-second walkthrough
video coming soon — scroll for the full version
Step 01 · 2 min

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.

Reactions · diethyl malonate alkylation
1L bench batch
CompoundRoleEqMWMolMassVol
Diethyl malonateSubstrate1.0160.170.50080.1 g76.0 mL
Benzyl bromideReagent1.0171.040.50085.5 g59.5 mL
K₂CO₃Base1.2138.210.60082.9 g
DMFSolvent73.09816 g865 mL
Concentration1.5 M (limiting reagent basis)
ConditionsDMF, 80°C, 4 h, BnBr added over 60 min
Observed yield (lab)91% isolated

Save this as a template. Next month's scale-up reads it back in one click.

Step 02 · 3 min

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.

🜂 Heat Transfer · cooling margin across scales
live
unsafe (margin < 1×) tight (margin < 3×) 9.5× 7.0× 4.6× 1 L (lab) 10 L 50 L (target) 10× cooling margin
Peak heat rate @ 50L
1.0 kW
over 60-min addition
Jacket capacity
4.8 kW
UA=80 W/K, ΔT=60 K
MTSR (worst case)
112°C
below DMF BP (153°C) ✓
!
Recommendation: Extend the BnBr addition to 120 minutes (was 60 min). Peak rate drops from 1.0 kW to 0.52 kW; cooling margin improves to 9.2×. No other change needed. Operator SOP shows max safe rate as a function of bulk temperature — printable for the floor.

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.

Step 03 · 2 min

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.

Dimensionless · lab vs pilot
DMF, 80°C
GroupLab (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
Mixing scales cleanly. Both lab and pilot are deeply turbulent; mixing time stays comparable. The K₂CO₃ suspension may need verification on first pilot run — if you see base settling, increase RPM to 220.
Step 04 · 4 min

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:

Yield · 50L pilot prediction
conf: medium-high
Lab observed
91%
isolated, n=3
Pilot prediction
85%
±4% (95% CI)
Predicted loss
−6 pts
across 4 mechanisms
Loss attribution
Mixing / base contact
−2.0 pts
Side reactions (BnBr hydrolysis)
−1.5 pts
Thermal hot-spots
−1.5 pts
Holdup / workup
−1.0 pt
The biggest controllable lever is mixing/base contact. Using 220 RPM and increasing K₂CO₃ to 1.3 eq could reclaim ~1.5 pts. Adopting the recommended 120-min addition reduces the thermal hot-spot loss by another ~0.8 pts. Combined: 87–88% expected at 50L — within striking distance of lab.
Step 05 · 3 min

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.

Equipment · 50L pilot recommendation
3 alternatives ranked
Reactor50 L jacketed borosilicate glass, dished bottom
ImpellerPitched-blade turbine, D = 200 mm, 4-blade, 45° tilt
Speed200 RPM (Re ≈ 137,000, fully turbulent)
Jacket fluid50/50 ethylene glycol / water
Jacket setpoint20 °C supply temperature
Chiller capacity800 W min @ 20 °C (1.5× peak load)
Dosing pumpperistaltic, 80 mL/min capacity, BnBr-compatible
Addition profilelinear ramp 0 → 60 mL/min over 120 min
Monitoringinternal Pt100 + jacket inlet/outlet RTDs
Two alternatives ranked: (2) glass-lined steel if scaling above 100L is on the roadmap (better U value, harder to clean for early development). (3) continuous flow if you anticipate >500 kg/yr production (better thermal control, smaller footprint, but higher capex).
Step 06 · 4 min

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.

Process Safety · Standards applicability
auto-evaluated
!
Borderline — DMF aggregate flammable trigger. 200L line at 70% DMF recycle still carries ~6,200 lb DMF on-site (storage + work-in-process). Below the 10,000 lb federal PSM threshold, but in the watch zone — re-check on any inventory increase. State of operation matters: Utah follows federal as-is; California (CalARP) is stricter.
Standards summary
Maybe
OSHA 29 CFR 1910.119 (PSM)
Aggregate flammable: 6,200 lb. Confirm with EHS and re-check on scale-up.
No
EPA 40 CFR Part 68 (RMP)
No listed substance above threshold.
Applies
IEC 61882 (HAZOP) + CCPS
Methodology basis. Ranking framework + worksheet structure.
Applies
NFPA 70 + 30 (flammable handling)
Class I area classification + flammable storage.

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:

Process Safety · PHA worksheet (highlights)
heuristic + chemistry-aware
High risk
5
priority for LOPA review
Medium risk
18
verify safeguards before pilot
Low risk
17
document and move on
Top high-risk scenarios surfaced
N2 (base addition) — More T → DMF + K₂CO₃ + heat
If cooling fails or addition is too fast and bulk T exceeds ~75 °C, DMF starts to decompose (analog system documented in OPRD 2019). Recommendation: independent high-T trip below 70 °C; consider NMP substitution if cooling margin is tight.
N3 (alkylation) — As Well As: operator exposure to BnBr
Benzyl bromide is a strong lachrymator and skin sensitizer. Closed-transfer system + LEV at the charge port mandatory. Single fatality not credible at this scale, but lost-time injury is.
N6 (evaporation) — Loss of vacuum + air ingress with hot mass
If vacuum break is air (not N₂) and the mass is still hot with DMF vapor: flammable atmosphere forms. Verify N₂ break valve; high-level alarm on cold trap.
40-scenario worksheet exports as CSV for the team meeting + a printable HTML report. Smart-suggestions panel automatically flagged the DMF + carbonate-base thermal hazard zone before generation, so the PHA team has a heads-up on the chemistry-specific risk before the kickoff.

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.

What's new · Process Safety
Every threshold list a U.S. process chemist needs — encoded.

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.

Federal lists
  • 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)
State-specific rules
  • 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
Exemptions encoded
  • 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
Library & matrix
  • ~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
Honest framing: This is a heuristic determination, not a legal one. Treat the verdict as the starting input to your written PSM/RMP applicability determination signed by site EHS or a qualified consultant — not the determination itself.
Step 07 · 6 min

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):

$ Techno-Economic · 200L line, 1500 kg/yr
cost per kg
Cost per kg
$58.08
all-in, including overhead
Annual production cost
$87.1k
at 1500 kg/yr target
Batches per year
24
2 per month, 64 kg each
Cost levers, biggest first
Benzyl bromide
$14.47
Labor
$12.53
Overhead (25%)
$11.62
Diethyl malonate
$9.41
Workup solvents
$5.50
DMF (30% fresh)
$2.36
Utilities
$1.25
K₂CO₃
$0.94
Biggest single opportunity: benzyl bromide at $14.47/kg = 25% of total cost. A 10% supplier reduction saves $1.45/kg = $2,170/year. RFQ three suppliers before the next quote cycle. DMF recycle was already factored in — pushing it from 70% to 80% saves another $0.79/kg. Most everything else is structural.
What's new · Experiment Optimizer
The upstream tool: find the conditions before you scale them.

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.

Self-tuning model

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.

Smart parallel batching

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.

Real-chemistry features

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.

Brute force
$200,000
1,000 experiments × $200
First-gen BO
> $100,000
500+ experiments, didn't converge
ScaleChem
$8,000
~40 experiments to converge
Methodology: 1,000-candidate synthetic benchmark with categorical catalyst-solvent interactions, a non-monotonic temperature peak at 60 °C, and a log-pressure dependency. Both optimizers given identical seed, batch size 5. "First-generation BO" = Gaussian-process Bayesian optimizer with hardcoded hyperparameters (lengthscale = 0.5, noise = 0.01) and naive top-N batch selection. Real-world performance varies with reaction surface complexity. Full benchmark and methodology in the technical white paper.

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:

Excel + handbooks
3–6 weeks
your time · plus debugging your own correlations
Engineering consultancy
6–12 weeks
$20k–$80k · plus your time on calls

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.