How to Read Your kLa Data Without Getting Fooled by Your DO Probe

Your DO probe is not lying to you. It is faithfully reporting the dissolved oxygen concentration at its tip. The problem is that the tip is in one place in the vessel, and the bulk dissolved oxygen is something your probe approximates — not something it measures directly. At bench scale, the approximation is excellent. At pilot and commercial scale, the gap between probe reading and bulk average can be significant enough to cause real process decisions based on misleading data.

Understanding this distinction matters for kLa determination, process control setpoint selection, and interpretation of DO excursions during scale-up runs.

How the Measurement Actually Works

Standard Clark-type (polarographic) and optical DO probes both measure dissolved oxygen concentration at the sensing element. Clark probes use an amperometric cell with oxygen-permeable membrane; optical probes measure luminescence quenching of an oxygen-sensitive dye. Both report the local dissolved oxygen partial pressure at the probe tip, converted to percent air saturation using a calibration factor.

More important than calibration factors for scale-up interpretation is probe response time.

Probe Response Time and Mixing Time: The Critical Interaction

Every DO probe has a T90 response time — the time required for the probe reading to reach 90% of the true value after a step change in DO. For standard Clark probes, T90 is typically 15–90 seconds depending on membrane thickness and diffusion characteristics. For optical (luminescent) probes, T90 is often shorter: 10–40 seconds in typical configurations.

Compare this to mixing time in your vessel:

• 2L bench bioreactor at 800 rpm: mixing time ~4–8 seconds
• 100L pilot vessel at 200 rpm: mixing time ~40–80 seconds
• 500L pilot vessel at 150 rpm: mixing time ~90–150 seconds
• 10,000L commercial vessel at 80 rpm: mixing time ~200–400 seconds

At 2L, mixing time is much faster than probe response time. The probe is always measuring a well-mixed bulk that changed slowly enough to stay within the probe's measurement bandwidth. At 500L, mixing time is comparable to or slower than probe response time. The probe responds to local concentration changes that may not reflect the bulk average — and the bulk average itself may not be spatially uniform.

kLa Determination Methods and What They Actually Measure

Static gassing-out method

The vessel is deoxygenated (nitrogen sparging), then aeration is started under target operating conditions, and DO% is recorded as it rises to saturation. kLa is extracted from the slope of the ln[(DO* − DO)/(DO* − DO₀)] vs time plot, where DO* is the saturation concentration.

The probe response time directly confounds this measurement. The actual kLa you calculate from probe data is:

1/kLa_apparent = 1/kLa_true + T_probe

where T_probe is an effective probe time constant (approximately T90/2.3). For a probe with T90 = 40 seconds and a true kLa of 200 h⁻¹, the apparent kLa from an uncorrected gassing-out measurement is approximately 160 h⁻¹ — a 20% underestimate. If you use this kLa in oxygen supply calculations without probe correction, you're predicting more oxygen supply than the vessel actually delivers.

The correction is straightforward: measure T90 by doing a step-change test (expose the probe to deoxygenated water, then to saturated water, record the time course), then apply the correction factor. Most bioreactor software does not apply this correction automatically.

Dynamic gassing-out method

Nitrogen is briefly sparged to deplete DO during fermentation, then oxygen-enriched air is restored, and the re-oxygenation curve is used to calculate kLa. This allows kLa measurement under actual fermentation conditions — with broth physical properties affecting oxygen transfer. It gives you kLa under more realistic conditions than the static method. Broth viscosity, anti-foam concentrations, and protein content all reduce kLa compared to clean water.

Sulfite oxidation method

Sodium sulfite is added to the vessel; its oxidation rate is measured and kLa calculated from sulfite consumption and oxygen stoichiometry. This avoids probe response time issues entirely (it's a chemical consumption measurement) but cannot be used with live cultures. It is used for empty vessel characterization before fermentation.

Probe Position Effects on DO Readings

Radial gradient: impeller zone vs vessel wall

In vessels with Rushton turbine impellers, the highest oxygen transfer rates occur near the impeller tips where turbulence is greatest. Probe position relative to the impeller zone significantly affects DO readings. A probe mounted adjacent to an impeller in the high-shear zone reads consistently higher than one mounted at the vessel wall, especially during peak OUR periods.

Axial gradient: vertical DO stratification in tall vessels

In vessels with liquid height-to-diameter (H/D) ratios greater than 1.5:1, dissolved oxygen at the top of the vessel differs from the bottom for two reasons. First, gas bubble residence time increases with height — bubbles from the sparger have more contact time as they rise. Second, hydrostatic pressure at the vessel base (approximately 0.01 atm per 10 cm of liquid height) increases the local oxygen saturation concentration and thus the driving force for transfer at the bottom of the vessel.

At commercial scale (vessels above 5,000L, liquid height above 3m), cells in the lower zone experience different dissolved oxygen conditions from cells in the upper zone — even with good bulk mixing. Your mid-vessel probe captures neither extreme accurately.

What This Means for Scale-Up Decisions

When your DO setpoint is 30% and your probe reads 32%, your cells in poorly mixed zones may be at 20%. This is not a failure of the probe or control system — it's a consequence of probe position, mixing heterogeneity, and response time lag in a vessel with real spatial extent. Before assuming your DO control is adequate, verify that your probe is positioned to capture the lowest-DO zone of the vessel and that your probe T90 is characterized for your specific membrane and media conditions.

kLa values from clean-water characterization are optimistic. Fermentation broth has surfactants (anti-foam), viscosity-modifying components, and high protein concentrations — all of which reduce kLa compared to the water-calibrated value. Industry practice suggests applying a 0.6–0.8 correction factor to water-measured kLa for typical aerobic fermentation broths. If your CDMO quotes 200 h⁻¹ from their vessel characterization, budget for 130–160 h⁻¹ in practice.

A Practical Protocol for kLa Characterization at Your Pilot Vessel

Before your next pilot run on a new vessel:

1. Characterize probe T90 in deionized water at your operating temperature. Record T90 and apply correction to all subsequent kLa measurements.
2. Run a static gassing-out kLa measurement at your target operating conditions (agitation, aeration, back-pressure, temperature) in media that matches your fermentation broth salinity and viscosity.
3. Repeat with anti-foam at your typical concentration. Compare to step 2 to quantify the anti-foam kLa reduction.
4. Calculate OUR demand at your target peak cell density using your bench-characterized specific oxygen uptake rate. Compare to kLa × (DO* − DO setpoint) to assess the oxygen supply margin.

This takes 2–3 hours and gives you the most reliable input for scale-up oxygen supply prediction. It is the single highest-value experimental investment you can make before a pilot run on a new vessel.

References

• Tribe LA, Briens CL, Margaritis A. Determination of the volumetric mass transfer coefficient (kLa) using the dynamic gas out-dynamic gas in method: analysis of errors caused by dissolved oxygen probes. Biotechnol Bioeng. 1995;46(4):388–392.
• Garcia-Ochoa F, Gomez E. Bioreactor scale-up and oxygen transfer rate in microbial processes: an overview. Biotechnol Adv. 2009;27(2):153–176.
• Doran PM. Bioprocess Engineering Principles. 2nd ed. Academic Press; 2013. Chapter 9.