Cells consume dissolved oxygen (DO) continuously; the reactor must resupply it at least as fast as the culture consumes it. The rate of resupply is governed by the volumetric oxygen mass-transfer coefficient kLa, estimated with the empirical van't Riet correlation:
k_L a = a · (P/V)^b · v_s^c
P/V = N_p · f_g · ρ · N³ · D⁵ / V (impeller power per volume, W/m³)
v_s = Q_gas / A_tank (superficial gas velocity, m/s)
a=0.026, b=0.40, c=0.50 (stirred-tank, coalescing media)
A "scale-down model" is a small (1–20 L) bench reactor tuned so its kLa — and therefore its oxygen-transfer environment — matches the full production vessel (here fixed at 2000 L), so process-development experiments run at bench scale stay predictive of the plant. Because power per volume falls as vessel size grows (N must drop roughly as D-2/3 at fixed P/V), a much slower impeller at production scale can reproduce the same kLa a fast bench impeller achieves.
- Working volume / impeller speed / aeration rate — set the bench reactor's operating point; geometry (tank and impeller diameter) is derived assuming a standard height:diameter ratio of 1 and impeller:tank ratio of 0.35, with a Rushton-turbine power number (Np=5) and a gassed-power correction factor.
- Production impeller speed — tune by hand, or press Auto-match kLa to binary-search the production RPM that reproduces the bench kLa at the fixed 2000 L production volume and matching vvm.
- Scale-down mismatch — the percent difference between the two kLa values; qualification is typically accepted within ±20%.
Real-world relevance: this is the core exercise behind "scale-down model qualification" in upstream biologics process development — before running expensive DoE campaigns on a 5 L bench reactor, engineers must first prove its mass-transfer (and shear) environment is representative of the GMP production bioreactor the process will ultimately run in.