pH electrodes and dissolved-oxygen sensors

pH and dissolved oxygen are the two variables a bioreactor controls continuously, and both are inferred rather than counted. Each is a physical transduction with its own failure behaviour, and in single-use systems those failures decide how long a run can be trusted.

The pH electrode is two half-cells, and only one of them is the sensor

A glass electrode measures a potential difference between an internal reference in contact with the pH-sensitive glass membrane and an external reference half-cell, usually silver/silver chloride, connected to the sample through a porous junction. The glass membrane develops a potential proportional to the activity difference across it; the reference half-cell is supposed to contribute a constant. Almost all real error lives in that assumption.

The junction carries a liquid-junction potential that changes with the sample’s ionic composition, and it is physically open, so protein, cells and precipitates foul it. Silver ions leaching outward meet sulfide and protein and form deposits that block it further. The result is an offset that grows with exposure — visible as a shifted zero rather than a wrong slope, which is why an electrode that still calibrates on slope can be quietly wrong in broth.

Temperature enters twice, and only one of the two is automatic. The Nernst relation makes the slope proportional to absolute temperature: about 59.2 mV per pH unit at 25 °C and about 61.5 mV at 37 °C. Instruments compensate this from a temperature probe. What they cannot compensate is that buffers and samples have their own temperature-dependent pH — a phosphate buffer’s assigned value differs measurably between 25 °C and 37 °C — so calibrating at bench temperature and measuring at culture temperature introduces a systematic error unless the calibration is performed at, or corrected to, the measurement temperature. Steam sterilisation ages the electrode as well: each cycle hydrates and erodes the glass gel layer and depletes the reference electrolyte, so a reusable probe has a finite number of qualified cycles.

Two ways to measure oxygen, with opposite weaknesses

A Clark-type sensor is amperometric. Oxygen diffuses through a gas-permeable membrane to a polarised cathode, is electrochemically reduced, and the current is proportional to the oxygen partial pressure at the membrane. Because it reduces the oxygen it detects, it consumes it: in stagnant liquid it depletes its own boundary layer and reads low, so the sensor is flow-dependent and its response time is set by diffusion through the membrane and electrolyte.

An optical sensor measures the luminescence lifetime of an immobilised luminophore quenched by oxygen, following Stern–Volmer behaviour and read out as a phase shift in modulated excitation. It consumes no oxygen and is therefore flow-independent, which is the decisive advantage in a poorly mixed or very small vessel. Its weakness is the luminophore: repeated excitation photobleaches it, so calibration drifts with accumulated illumination rather than with elapsed time, and sampling frequency becomes a calibration variable.

The single-use constraint

In a pre-sterilised bag, sensors are installed and irradiated with the vessel and factory-calibrated before it is sealed. Once inoculated, the system is closed by design, and no calibration buffer or gas can reach the sensor. Practice is therefore a one-point offset: an aseptically drawn sample is measured on an at-line blood-gas analyser, and the in-line reading is corrected to it. That fixes an offset at one instant. It does not fix drift, and over a fourteen-day perfusion run — with junction fouling on one side and photobleaching on the other — accumulated drift, not initial accuracy, is what limits how far the measurement can be trusted.