# API Scale-Up: The 10 Physics-Based Mistakes That Derail Every Project

# Why a Reaction that Works Perfectly at 1 L Can Fail at 10,000 L

Scaling an Active Pharmaceutical Ingredient (API) manufacturing process from a 1 L laboratory flask to a 10,000 L commercial reactor is frequently misunderstood as a simple exercise in multiplying raw material charges by 10,000.

It is not.

While the fundamental reaction chemistry and molecular thermodynamics remain identical, the underlying transport phenomena—hydrodynamics, heat transfer, mass transfer, and characteristic mixing timescales—change by orders of magnitude.

At the 1-liter laboratory bench scale, an R&D chemist charges reagents in seconds, achieves instantaneous micro-mixing, maintains isothermal control via oversized thermal jackets, and directly inspects phase boundaries. At a 10,000 L commercial scale, the identical process involves metric tons of boiling liquid, 1 to 2 hour slow reagent dosing protocols, severe surface-area-to-volume constraints, substantial thermal inertia, and macromixing recirculation times measured in minutes.

The foundational principle of chemical scale-up is:

Chemistry dictates what is thermodynamically and kinetically possible. Transport phenomena determine what actually happens inside an industrial reactor.

Below is a rigorous, physics-based analysis of the 10 fundamental engineering mistakes that derail API scale-up projects, complete with quantitative governing equations, characteristic timescale ratios, and actionable engineering solutions.

# 1. Mistake #1 — Scaling Everything by Geometric Volume (Linear Volume Scaling)

The most elementary scale-up failure is assuming that all process and equipment dimensions scale linearly with batch volume:

1 L ──► 10,000 L (Volume Factor = 10,000x)

While total molar and mass quantities scale directly with batch volume, geometric, hydrodynamic, and thermodynamic parameters scale with fractional power laws:

• Volume (V) scales with the cube of characteristic length: V is proportional to L³
• Characteristic Length (L) scales with volume to the 1/3 power: L is proportional to V^(1/3)
• Wetted Heat Transfer Area (A) scales with length squared: A is proportional to L² is proportional to V^(2/3)
• Specific Surface Area Ratio (A/V) scales inversely with scale: A/V is proportional to 1/L is proportional to V^(-1/3)

# Quantitative Comparison: 1 L Flask vs. 10,000 L Commercial Reactor

Because the available heat-transfer area per unit volume drops from 60 m⁻¹ down to 2.7 m⁻¹, direct geometric scaling inevitably leads to thermal runaway, impurity spikes, or extended cycle times unless an explicit scale-up governing criterion is chosen.

# 2. Mistake #2 — Assuming Mixing Intensity and Agitator Power Are Equivalent

A common operational error is attempting to maintain the same rotational speed (RPM) across scales.

In agitated vessels, power consumption (P) in the turbulent regime (Re > 10,000) is governed by:

P = Np x rho x N³ x D⁵

Where:

• P = Agitator shaft power input (Watts)
• Np = Impeller Power Number (dimensionless, e.g. 0.35 for hydrofoils, 1.5 for pitched blades, 5.0 for Rushton turbines)
• rho = Liquid density (kg/m³)
• N = Rotational speed (revolutions per second, s⁻¹)
• D = Impeller diameter (m)

Because power scales with the 5th power of impeller diameter (D⁵) and the 3rd power of speed (N³), holding RPM constant during scale-up causes power demand to explode by:

Power Scale Factor = (D_plant / D_lab)⁵ = (21.5)⁵ = 4,586,000x

Conversely, if an engineer scales down speed to maintain identical power per unit volume (P/V = constant):

N_plant = N_lab x (D_lab / D_plant)^(2/3) = N_lab x (1 / 21.5)^(2/3) = 0.129 x N_lab

While P/V remains constant, the impeller tip speed (v_tip = pi x D x N) increases by 2.78x, which can shear fragile crystal aggregates, while blend time increases by 2.78x, degrading rapid reaction selectivity!

# 3. Mistake #3 — Ignoring Characteristic Mixing Times & Damköhler Numbers

At 1 L scale, bulk liquid circulation occurs within 1 to 2 seconds. In a 10,000 L reactor, bulk circulation (macromixing) requires 30 to 90 seconds, while molecular dispersion (micromixing) can take tens of milliseconds.

# The Damköhler Number Regime Analysis

To determine whether a reaction will suffer upon scale-up, engineers evaluate the Damköhler Number for Mixing (Da_I):

Da_I = tau_mix / tau_rxn = Characteristic Mixing Time / Characteristic Reaction Half-Life

• Homogeneous Kinetic Regime (Da_I << 0.1): Chemical reaction is much slower than mixing (tau_rxn > 10 min). The vessel remains chemically uniform; scale-up is seamless.
• Diffusion / Mixing-Controlled Regime (Da_I >> 1.0): Chemical reaction occurs faster than mixing (tau_rxn < 5 sec, e.g. acid-base neutralizations, acylations, Grignard couplings).

Reagent Dosing Point (High Local Excess B)
──► Fast Desired Path: A + B ──► Product (C) [k1 = Fast]
──► Slow Side Path: C + B ──► Over-Alkylated Impurity (D) [k2]

In a 10 KL reactor under mixing control, the incoming reagent 'B' forms a localized "hot plume." Because it cannot disperse before reacting, it attacks already-formed Product 'C', creating unacceptably high levels of over-alkylated or bis-adduct impurities that were completely absent in the 1 L lab flask.

# 4. Mistake #4 — Scaling Heat Transfer Without Evaluating A/V Shrinkage

Exothermic chemical reactions generate heat strictly proportional to liquid volume (V):

Q_generated = r_rxn x (-delta-H_rxn) x V

However, jacket heat dissipation is strictly governed by wetted surface area (A):

Q_removed = U x A x delta-T_lm

As demonstrated in Section 1, the ratio A/V is proportional to 1/V^(1/3).

# The Heat Dissipation Mismatch:

• At 1 L, a reaction releasing 100 W generates 100 kW/m³, but has 60 m²/m³ of heat transfer area. A mild 2 °C jacket differential removes all heat effortlessly.
• At 10,000 L, the same chemistry generates 1,000 kW, but the reactor possesses only 27 m² of jacket area (2.7 m²/m³).
• To remove 1,000 kW across 27 m² with U = 350 W/m²·K:
  delta-T_lm = Q / (U x A) = 1,000,000 W / (350 x 27) = 105.8 °C

Because a 105.8 °C cooling delta is physically impossible without cryogenic refrigeration, the commercial batch temperature will uncontrollably ramp upward, triggering thermal runaway or decomposition.

# 5. Mistake #5 — Ignoring Reagent Dosing Profiles & Subsurface Dip Pipes

In the laboratory, adding 100 mL of reagent over 5 minutes is standard. At 10,000 L scale, adding 1,000 L of reagent requires 1.5 to 3.0 hours due to feed pump limits and exothermic heat-removal constraints.

# The Problem of Surface Dripping:

In many plant reactors, reagents are dropped from the top nozzle onto the turbulent liquid surface:

1. Solvent vapors in the headspace react with droplets before they reach the bulk liquid.
2. Vapors boil off locally due to un-dissipated heat of mixing.
3. Top surface flow patterns carry the reagent into the stagnant baffle shadow before reaching the high-shear impeller zone.

# 6. Mistake #6 — Overlooking Gas-Liquid Mass Transfer (kLa & Hatta Number)

Gas-liquid catalytic reactions (e.g. Hydrogenation, Carbonylation, Aerobic Oxidation) depend on transferring gas molecules across the interfacial boundary:

Rate = kLa x (C_star - C_liquid)

Where:

• kLa = Volumetric mass transfer coefficient (s⁻¹ or h⁻¹)
• C_star = Saturated gas solubility at reactor pressure (mol/m³)
• C_liquid = Dissolved gas concentration in bulk solvent (mol/m³)

# The Hatta Number (Ha) Diagnostic:

To identify whether mass transfer or chemical kinetics governs your reactor:

Ha = sqrt( k_rxn x D_AB x C_catalyst ) / k_L

• Ha < 0.3 (Slow Kinetic Regime): Liquid is saturated with gas; scale-up based on temperature and catalyst loading.
• Ha > 3.0 (Fast Mass Transfer Controlled): Reaction occurs entirely in the liquid film surrounding the gas bubble. Increasing catalyst loading in a 10 KL plant will NOT increase reaction rate; it will only cause catalyst poisoning and impurity accumulation.

To maintain kLa >= 0.08 s⁻¹ at 10 KL scale, standard pitched blade turbines must be replaced with Hollow-Shaft Gas-Induction Impellers (e.g. Ekato GASJET) that continuously re-suck headspace gas and disperse microscopic bubbles directly into the batch.

# 7. Mistake #7 — Assuming Constant Overall Heat Transfer Coefficient (U)

Engineers often copy a clean laboratory U-value (e.g. 400 W/m²·K) into commercial reactor sizing calculations.

In reality, overall heat transfer coefficient (U) is a composite series resistance:

1/U = (1 / h_process) + R_fouling,process + (x_wall / k_wall) + R_fouling,jacket + (1 / h_jacket)

Neglecting the glass-lining thermal barrier in a 10 KL Glass-Lined reactor reduces effective heat transfer capacity by 60%, doubling batch cooling durations.

# 8. Mistake #8 — Ignoring Rheological Shifts & Non-Newtonian Viscosity Surges

During API concentration, salt formation, or antisolvent addition, reaction slurries frequently transition from Newtonian fluids (mu = 1.0 cP) to shear-thinning or pseudoplastic non-Newtonian regimes (mu_apparent > 500 – 5,000 cP).

# The Viscosity Collapse Mechanism:

1. As viscosity surges, the Impeller Reynolds Number (Re = rho x N x D² / mu) plunges from turbulent (Re > 50,000) into transitional or laminar flow (Re < 500).
2. Agitator pumping collapses; fluid movement becomes confined to a rotating "cavern" immediately surrounding the impeller blades.
3. Near the vessel wall and jacket baffles, the fluid becomes completely stagnant.
4. Process film coefficient (h_process) drops by 80%, causing localized overheating or precipitation on the cold jacket wall.

# 9. Mistake #9 — Overlooking Solid Suspension Physics (Njs) & Secondary Nucleation

In slurry reactions and crystallization, keeping particles fully off the vessel dish bottom requires exceeding the Zwietering Just-Suspended Agitator Speed (Njs):

Njs = S x (nu)^0.1 x [ (g x delta-r\rho) / rho_L ]^0.45 x (d_p)^0.2 x [ X^0.13 / D^0.85 ]

Where:

• S = Dimensionless vessel geometry parameter
• nu = Kinematic viscosity (m²/s)
• delta-rho = Solid-liquid density difference (rho_solid - rho_liquid)
• d_p = Mean crystal particle diameter (m)
• X = Solid weight percentage (%)
• D = Agitator diameter (m)

# The Scale-Up Crystallization Trap:

• If the commercial agitator operates below Njs, dense API crystals settle into the bottom valve nozzle, forming an un-mixable cake that blocks transfer lines.
• If the agitator runs too fast to compensate, the excessive impeller tip speed (v_tip > 5.0 m/s) creates catastrophic mechanical shear, shattering delicate needle crystals into ultra-fine dust (secondary nucleation), resulting in unfilterable filter cakes in downstream plant centrifuges.

# 10. Mistake #10 — Sub-Optimizing the Reactor While Blinding Downstream Unit Operations

A reactor does not operate in isolation. An API process is an interconnected, multi-unit manufacturing train:

[10 KL Reactor] ──► [Overhead Condenser] ──► [Receivers] ──► [Centrifuge / ANFD] ──► [Vacuum Dryer]

# Typical Downstream Plant Bottlenecks:

1. Condenser Flooding: Scaling reactor boil-up to 800 kg/h without verifying that the overhead shell-and-tube condenser can handle the vapor velocity leads to solvent carryover into the vacuum header.
2. Specific Cake Resistance in ANFDs: A 1 L Buchner funnel filtration taking 2 minutes can translate to 18 hours on a 2 m² Agitated Nutsche Filter Dryer (ANFD) if crystal size distribution (CSD) shifts from 150 microns down to 20 microns.
3. Dryer Thermal Degradation: Large wet cakes in vacuum dryers suffer poor core heat conduction, extending drying cycles to 24–48 hours and degrading thermolabile active ingredients.

# 11. Master Scale-Up Parameter Decision Matrix

When scaling an agitated batch process, no single criterion can satisfy all physical constraints simultaneously. The table below guides engineers in choosing the governing parameter:

# 12. Step-by-Step 5-Stage Scale-Up Protocol for API Projects

To de-risk technology transfer and commercial campaigns, top chemical engineering teams execute this structured 5-stage protocol:

Stage 1: Reaction Calorimetry & Kinetics (RC1e, DSC, ReactIR)
──► Stage 2: Hydrodynamic & Scale-Down Modeling (Visimix, CFD)
──► Stage 3: Intermediate Pilot Scale Validation (100 L to 500 L)
──► Stage 4: Utility, Headspace & Downstream Bottleneck Audit
──► Stage 5: Plant Commissioning & Real-Time PAT Monitoring (FBRM, DCS)

1. Stage 1 — Calorimetric Characterization: Measure heat of reaction (delta-H_rxn), adiabatic temperature rise (delta-T_ad), and time-to-maximum-rate (TMR_ad) using Reaction Calorimetry (Mettler Toledo RC1e) and Accelerating Rate Calorimetry (ARC).
2. Stage 2 — Hydrodynamic Modeling: Use software simulation (e.g. Visimix, DynoChem) to model shear rates, blend times, and heat transfer coefficients across prospective plant vessels.
3. Stage 3 — Pilot Demonstration: Execute at 50 L to 500 L scale to confirm impurity profiles and crystal habit under realistic dosing rates.
4. Stage 4 — Downstream Audit: Size the process condenser, vacuum pump volumetric throughput, ANFD filtration area, and wastewater treatability before committing commercial production suites.
5. Stage 5 — Real-Time Process Monitoring: Implement Process Analytical Technology (PAT) tools such as inline FT-IR and Focused Beam Reflectance Measurement (FBRM) to monitor chord-length distributions and reaction conversion in real time.

# Core Engineering Takeaway

Successful scale-up is never a matter of luck or recipe multiplication.

A successful laboratory experiment proves that the chemistry is sound. A successful commercial scale-up proves that the chemical kinetics, fluid mechanics, and thermodynamics have been harmoniously balanced at industrial scale.

By identifying the rate-limiting physical phenomenon before entering the plant—whether it is micromixing, heat dissipation, solid suspension, or gas-liquid mass transfer—process engineers can guarantee high yields, strict polymorph control, and robust process safety on the very first commercial batch.

# Applicable Engineering Standards & Codes Used

The engineering methodologies, design correlations, and safety criteria detailed in this article adhere to the following international standards and industry codes:

• US FDA 21 CFR Part 211: Current Good Manufacturing Practice for Finished Pharmaceuticals
• EU GMP Annex 1: Manufacture of Sterile Medicinal Products
• ISPE Baseline Pharmaceutical Engineering Guide (Vol 1-7: Bulk Active Pharmaceutical Ingredients, Water & Steam Systems): ISPE Baseline Pharmaceutical Engineering Guide (Vol 1-7: Bulk Active Pharmaceutical Ingredients, Water & Steam Systems)
• WHO Technical Report Series No. 961: Supplementary Guidelines on Good Manufacturing Practices for Heating, Ventilation and Air Conditioning