A bubble you can see is a bubble on its way out. Coarse aeration gives gas a few seconds of contact and then hands most of it back to the atmosphere. Below roughly 200 nm the same gas stops behaving like a bubble at all: it stays where it was put, moves with the water, and is still in suspension at the far end of the pipe.
That difference is the whole engineering case, and it rests on three pieces of physics pulling against each other — viscous drag, Laplace pressure, and interfacial charge. Two are textbook. The third is still argued over in the literature, and this post says so rather than rounding an open question up to a fact.
Why do ordinary bubbles rise but nanobubbles don't?
Buoyant force scales with the cube of the radius while Stokes drag scales only with the radius itself, so rise velocity falls off as radius squared. A 200 nm bubble rises about 22 nm/s, roughly 2 mm a day. Brownian displacement moves it around 2 µm a second, a hundred times faster, in every direction at once.
The arithmetic is worth doing. In the creeping-flow regime a sphere reaches terminal velocity when buoyancy balances viscous drag, which makes rise speed proportional to the square of the radius. Put a 200 nm bubble into that expression — radius 100 nm, water at 20 °C — and terminal velocity comes out at roughly 22 nm/s for an immobile interface, the case real surfactant-bearing water gives you. A clean interface would rise about 1.5× faster and change nothing that follows. And 200 nm is the conservative case: our own measured population centers well below it, so the real drift is slower still.
Compare a bubble from a coarse diffuser. A 1 mm bubble clears a meter of tank in a few seconds, and Stokes' law no longer applies there — about seven orders of magnitude in velocity for four in diameter.
Shrinking a bubble does not slow it a little. It removes rise from the list of things that happen to it.
Meanwhile, the same small size that kills buoyancy hands the bubble to thermal motion. Water molecules strike it from every side, and the imbalance at any instant displaces it far more than gravity does — a hundred to one in favor of noise.
Be precise about what that does not say. Buoyancy is not zero, and over long enough times a ballistic drift outruns a random walk. Undisturbed for a week, a population of 200 nm bubbles would creep upward by roughly a centimeter. No working vessel is that undisturbed — convection in a reservoir, or flow in a pipe, swamps it. Vertical drift is not what ends a nanobubble's life. Dissolution is.
If Laplace pressure should dissolve them, why do nanobubbles persist?
Classical theory says a 200 nm bubble should not survive. Laplace pressure adds roughly 14 atm inside it, which by classical diffusion theory drives the gas out in microseconds. They persist for months anyway. The leading explanations are surface charge, adsorbed gas at the interface, and a locally saturated shell around each bubble.
The paradox comes straight out of the Young-Laplace relation. Surface tension pulls a curved gas-liquid interface inward, and the excess pressure it creates is inversely proportional to the radius: at 100 nm, on the order of 14 atm above ambient. Henry's law then requires the water touching that interface to hold gas in proportion, so the bubble wall is locally supersaturated and gas diffuses away down the gradient.
Worse, the process accelerates itself. As gas leaves, the radius shrinks, which raises the Laplace pressure, steepens the gradient, and drives gas out faster. The Epstein-Plesset treatment of that runaway predicts a lifetime measured in microseconds. Millions of them should not still be in the jar months later. They are. The classical model is missing a term.
Three candidate terms lead the literature, and they are not mutually exclusive:
- Surface charge. A charged interface generates an outward electrostatic stress that offsets part of the inward pull of surface tension, lowering the Laplace pressure the gas has to fight.
- Interfacial gas adsorption. Gas molecules and trace amphiphilic species in any real water accumulate at the interface and armor it, lowering effective surface tension and the area available for diffusion.
- A saturation shell. If the liquid around the bubble is already near saturation, the gradient driving gas outward flattens toward zero and diffusion stalls regardless of internal pressure.
None of the three is settled, and anyone saying otherwise is selling certainty rather than physics. The persistence itself is measured repeatably — in published research and in our own water, by more than one method. Kairospace designs around the measurement, not a preferred explanation.
What does zeta potential have to do with stability?
Nanobubbles carry a negative surface charge. Because every bubble carries the same sign, they repel each other electrostatically instead of merging, and coalescence is the fastest route to a bubble large enough to rise. Whether that charged layer also slows gas out of the bubble is one of the open candidate mechanisms, not a settled one.
Zeta potential is the electrical potential at the shear plane — the boundary between the ion layer that travels with the bubble and the bulk water that does not. Gas-water interfaces in ordinary water measure negative, most often attributed to preferential adsorption of hydroxide ions.
The first consequence is anti-coalescence. Merging is not a rival cause of death; it is the road back to the size where none of the physics above applies. Two bubbles combine, the radius climbs, and buoyancy — which goes as radius cubed — catches up with drag. A cascade reaches a size that rises, and the gas is back to seconds of contact time. Like charges hold the population apart and keep the cascade from starting.
The second consequence is the contested half of this section. If the structured ion layer is more than a passive marker of charge, it is also a physical obstacle to the gas leaving. That stays a candidate rather than a demonstrated mechanism: no measurement yet separates a diffusion barrier at the charged layer from the adsorption and saturation-shell explanations above: all three predict the same observable — a bubble that is still there. The classroom page on generation mechanisms and stability carries the theory in more depth.
It also sets a real limit. Anything that compresses the electrical double layer shortens bubble life: high ionic strength, heavily loaded fertigation, and brackish or marine water. Treated water is stable, not immortal. That is a design input, not a footnote.
How do you know they are actually there?
Nanoparticle tracking analysis is the working method: a laser illuminates the suspension, a camera tracks how far individual scatterers wander per frame, and the diffusion rate gives each one a size. Dynamic light scattering is faster but returns a population average that hides small populations. Neither distinguishes a bubble from a particle on its own.
NTA works because the Brownian wandering described above is itself a measurement. Track a scatterer frame by frame, extract its diffusion coefficient, and the Stokes-Einstein relation returns its hydrodynamic diameter. Because it sizes objects one at a time, it also counts them, so a concentration comes out of the same run as a size distribution.
Arizona State University ran that measurement on Kairospace water. Sample AGPO2-T3, captured March 21, 2023 on a Malvern NanoSight across 1,498 frames in three merged runs.
Mode 79 nm · median 106 nm · D90 201 nm — about nine in ten tracked particles measured below roughly 201 nm. The distribution is the sharper result.
That is independent laboratory characterization: not our field data, and not the literature, but a third thing that carries different weight from either.
The report records that dilution was not recorded, so the concentration is an as-measured figure: if the sample was diluted before tracking, the true number is higher. It is a floor, not a point estimate, and we do not round it up.
NTA also counts particles, not bubbles. It tracks scattering centers and sizes them, and nothing in the measurement itself says a given center is gas rather than grit.
Dynamic light scattering is faster and the one to be careful with. It reads the whole sample at once, and scattering scales steeply with diameter, so one large aggregate can dominate the signal and drag the reported mean up — good for tracking whether a population has shifted between two otherwise identical samples, poor at asserting what is in a sample you have never seen.
The second caveat is the harder one, and no light-scattering method resolves it alone. The discipline is procedural: run the untreated feed water as a blank so the particulate background is known, then rerun after pressurizing or degassing, since gas-filled objects compress and dissolve while solids do not. What behaves like gas across that comparison is read as gas. The classroom section on characterization and measurement techniques covers the method set in full.
What does this mean for water in an actual system?
Persistence is what makes the treatment portable. Gas that stays suspended travels the pipe run, so water dosed at the pump is still carrying it at the emitter, the pond, or the basin hundreds of meters away. A microbubble cloud would have coalesced and collected at the high points of the line long before the water arrived.
Every irrigation, aquaculture, and treatment system has the same structural problem: the point of injection is not the point of use. Gas goes in at the pump house; demand is at the root zone, at the fish, at the biofilm — a filter bank downstream. A treatment that does not survive that trip treats the pipe, not the process. Sub-200 nm gas survives it.
It changes what the water does on arrival, too. Suspended gas keeps transferring across whatever interface it meets, so oxygen continues moving into the root zone or across a gill long after the water left the pump. The cavitation that makes the bubbles also disperses what is dissolved alongside them — the physics driving the chemistry.
Which is exactly where to stop. This is a mechanism post, and the mechanism does not replace input chemistry — it changes how well that chemistry is delivered, and how much oxygen reaches the place that needs it. Whether it changes anything on your ground is a question for a controlled pilot, not a physics argument.