Oxygen Transfer Rate — OTR = k_L·a (C* − C_L)

Start with the number that makes this whole subject hard: oxygen is barely soluble in water. Air-saturated water holds about 9.1 mg/L at 20 °C and about 7.0 mg/L at 35 °C, and a fermentation medium holds less still, because dissolved salts and sugars salt the oxygen out. A dense aerobic culture consumes that entire dissolved inventory in a matter of seconds. Everything difficult about aerobic fermentation follows from that one fact: oxygen cannot be stored in the broth, so it must be delivered continuously, and the rate of delivery is what caps the process.

The delivery rate is written $OTR = k_L a (C^{*} - C_L)$ — a coefficient multiplied by a driving force. The driving force is the shortfall between the saturation concentration the gas phase would support and what the dissolved-oxygen probe actually reads, and it is at most a few milligrams per litre. The coefficient does the rest of the work.

$k_L a$ is a lumped, geometry-specific quantity, and this is the most important thing on the page. It is written as a single symbol because $k_L$ — the liquid-film mass-transfer coefficient — and $a$ — the gas–liquid interfacial area per unit volume — cannot be measured separately in a sparged, stirred vessel, so they are measured together and quoted together. What it depends on is a long list, and every item on it is a property of the equipment or the fluid rather than of oxygen: vessel diameter, impeller type and diameter and speed, number of impellers, baffle arrangement, sparger design and position, superficial gas velocity, liquid height, broth viscosity and rheology, ionic strength, and the presence of antifoam — which suppresses transfer sharply, because it works by promoting bubble coalescence and coalescing bubbles have less surface area.

Change any one of those and $k_L a$ changes. Which means it does not transfer between vessels, and that is the central difficulty of bioreactor scale-up. Geometric similarity does not preserve it. Consider the standard scale-up rules: constant power per unit volume, constant impeller tip speed, constant mixing time, constant $k_L a$. Each gives a different impeller speed for the same larger vessel, and they are mutually incompatible — holding power per volume constant while scaling up by a factor of a thousand in volume raises the tip speed substantially, which raises shear; holding tip speed constant drops the power per volume, which drops $k_L a$. You must choose which to preserve, and the choice depends on whether your organism is more sensitive to oxygen limitation or to shear. A $k_L a$ of 100 h⁻¹ measured in a 5 L bench fermenter tells you very little about the 5000 L vessel the process is going into. Measure it in the vessel you will actually run, in the broth you will actually run — by dynamic gassing-out, by sulphite oxidation, or best of all from an exhaust-gas oxygen balance on a live culture, which captures the real viscosity and the real antifoam.

For scale: a shake flask manages perhaps 5 to 20 h⁻¹, a bench stirred fermenter 50 to 500 h⁻¹, a well-designed production vessel a few hundred. Figures above roughly 1000 h⁻¹ generally mean oxygen enrichment or elevated head pressure rather than better mixing.

There are only four levers on the transfer rate and it is worth knowing all of them. Raise $k_L a$ with more stirrer power or more gas — both cost energy, and stirrer power is the dominant operating cost of a large aerobic fermenter. Raise $C^{*}$ by enriching the sparge gas with oxygen, or by holding back-pressure on the vessel; both work through Henry's law, which makes $C^{*}$ proportional to the oxygen partial pressure. Lower $C_L$, which is free but nearly exhausted, since most aerobes need the dissolved oxygen held above a critical value around 5 to 10 percent of saturation before their own respiration becomes oxygen-limited. And the fifth option, which is not on the list because it is not a transfer lever at all: reduce the demand, by feeding substrate more slowly. That is a large part of why high-cell-density processes are fed-batch.

What this looks like on a real fermentation trace is unmistakable. Demand rises with biomass, which rises exponentially. Supply is fixed by the vessel. So the dissolved oxygen sits near saturation for most of the run and then falls off a cliff within one or two doublings of the end. That moment is when the process stops being biology and starts being mass transfer, and it is the reason oxygen — not substrate — usually sets the maximum cell density a bioreactor can reach.