At 10 % gas holdup, bubbles scatter away roughly 78 % of the unscattered beam within the first centimetre of slurry, so a sparged cell runs dimmer than its window dosimetry claims. Hold holdup below 3 %, stop sparging 60 s before a rate measurement, and size the frit for 1–2 mm bubbles at 0.2–0.5 vvm.
Sparging is the obvious source, not the only one. Dissolved gas comes out of solution as the cell warms: air-saturated water at 25 °C holds 8.3 mg/L of O2, about 260 µmol/L, so a 500 mL cell contains roughly 130 µmol of dissolved oxygen before any reaction starts — 30 to 100 times the hourly H2 rate of a 0.5 g/L titania run. Stirring at 800 rpm with a vortex reaches an impeller tip speed of 2–3 m/s and entrains headspace gas. At high rates the reaction nucleates gas too: visible H2 bubbles on Pt/TiO2 appear above about 1 mmol/h, which is where reported rates start to scatter.
Each source needs a different fix, which is why "de-gas the reactor" as a single instruction rarely works: a sparged cell needs the measurement paused, a sealed cell a purge before sealing, an entraining stirrer a lower speed.
The refractive index step is large — 1.00 inside the bubble against 1.33 in water — so each interface both reflects and refracts. For spheres of radius r at volume fraction φ, the projected area per unit path is 3φ/(4r). At 10 % holdup and a 1 mm diameter, that is 1.5 cm−1, so the unscattered beam falls to e−1.5, about 22 %, after 1 cm.
The measured rate does not fall by 78 %, and assuming it does is the second mistake. Most of that light is forward-scattered within a few degrees and still reaches the catalyst a millimetre deeper, so total fluence through the plane drops only 8–15 %. What breaks is the definition of optical path: light no longer travels a known distance, the window-mounted photodiode no longer represents the slurry, and back-reflection at each bubble adds 2–5 % loss per interface. That is why bubble-heavy runs give rates that look plausible but will not reproduce.
Sticking is the visible half. A 2 mm bubble on a quartz window covers 3.1 mm2; thirty of them block about 5 % of a 20 cm2 window and each leaves a dry patch the slurry cannot wet.
| Bubble diameter | Rise velocity in water | Holdup at 0.5 vvm | Optical signature | Typical source |
|---|---|---|---|---|
| 0.2–0.5 mm | 0.03–0.08 m/s | 15–25 % | Uniform haze, near-isotropic scatter, slow to clear | P4 frit, 10–16 µm pores |
| 1–2 mm | 0.15–0.25 m/s | 5–12 % | Forward scatter, clears in under 10 s | P3 frit, 16–40 µm pores |
| 2–3 mm | 0.25–0.30 m/s | 3–8 % | Discrete dark voids crossing the window | P2 frit, 40–100 µm pores; coarse sparger |
| >5 mm | 0.30–0.35 m/s | 1–3 % | Large dry patches, erratic baseline | Open-pipe sparger, vortex entrainment |
Finer is not better. A 10 µm frit gives excellent mass transfer and a cell that looks like milk, with 20 % holdup and an optical path you cannot state. For kinetics, 1–2 mm is the sweet spot: enough interfacial area for gas–liquid transfer, few enough bubbles to leave the beam legible.
A differential pressure tap across the liquid column is the cheapest instrument. With the taps 10 cm apart, full liquid gives 981 Pa; at 10 % holdup the reading falls to about 883 Pa, so 10 Pa resolution sees 1 % holdup. For a quick check, stop the sparge and log the rate for 120 s: if it climbs 10–20 %, holdup was biasing the number, and the plateau is what you report.
Do the same for window dosimetry: nine points, unsparged then sparged; the ratio is the scattering loss for that frit, flow rate and solids loading, and it does not transfer between reactors.
Antifoam and surfactants are a trap. 10 ppm of a non-ionic surfactant halves bubble size and looks like a fix, but it adsorbs on TiO2, shifts the point of zero charge and can cut the rate by 20–40 %. Change the plumbing, not the chemistry.
Yes, through the denominator. Photons are counted from a window-plane irradiance that overstates what the slurry receives, so apparent quantum yield is underestimated by 10–30 % in a heavily sparged cell. Report irradiance measured under the same sparging condition as the run.
30 min at 100 mL/min of N2 for a 500 mL cell — six volume exchanges — takes dissolved O2 below 0.1 mg/L. Then stop the gas and wait 60 s: a 2 mm bubble crosses a 10 cm column in about 0.4 s, so a minute is generous.
16–40 µm (P3) for most kinetic work, giving 1–2 mm bubbles at 0.2–0.5 vvm. Go finer only if gas–liquid transfer limits the rate and you accept 15–25 % holdup and an undefined optical path.
Only for a fixed frit, flow rate and solids loading, because the correction scales with all three. A clean-water blank does not transfer: the catalyst scatters too, and the two populations do not add linearly.
