A nanobubble is a nanobubble regardless of how it was made. What differs between the three industrial methods is what the machine costs to run, how large it can be built before the physics stops cooperating, and which part of it wears out first. Those are procurement questions, and they are answerable without deciding in advance which method wins.
One disclosure up front, because it bears on how to read the next four sections: Kairospace builds hydrodynamic equipment. That is a reason to check the citations rather than a reason to skip them, and where the published comparisons are indirect or narrow, this post says so instead of collecting the conclusion.
How do you actually make a nanobubble?
Three mechanisms make nanobubbles at industrial scale: hydrodynamic cavitation, acoustic cavitation, and gas forced through a porous membrane. The first two drop the local pressure below the vapor pressure of water so cavities form and collapse. The third pushes gas through fine pores and lets the flowing liquid shear the bubbles off the surface.
Hydrodynamic cavitation works off the Bernoulli relation. Push water through a constriction — a venturi throat, a multi-hole orifice plate, a vortex diode — and velocity rises while static pressure falls. If the static pressure drops below the vapor pressure of the liquid, cavities nucleate; downstream, where the channel opens and pressure recovers, they collapse. The collapse is violent and local, and the intense shear around it fragments injected gas into micro- and nanobubbles. The energy comes from a pump, which is equipment most plants already own and already know how to maintain.
Acoustic cavitation delivers the same pressure swing through sound instead of geometry. An ultrasonic transducer radiates alternating compression and rarefaction cycles into the liquid; during the rarefaction half-cycle, the local pressure falls far enough for cavities to nucleate, and the following compression collapses them. Control is the strong suit here — frequency, amplitude, and duty cycle are all set electrically, so the cavitation regime can be tuned continuously in a way a fixed orifice cannot.
Membrane generation does not use cavitation at all. Pressurized gas is forced through a porous element, typically ceramic, into liquid flowing across the face of it. The gas pressure has to exceed the capillary pressure holding it out of the pores — the bubble point — and once through, the shear of the crossflow detaches each bubble before it can grow. Bubble size tracks the pore size, the surface chemistry of the membrane, and the injection pressure rather than any collapse event. The classroom section on generation mechanisms and stability carries the underlying literature.
Which method is most energy-efficient at scale?
No published figure answers this for gas dissolution. In peer-reviewed work on pollutant degradation, hydrodynamic cavitation returned a cavitational yield of 4.44 × 10⁻⁶ mg/J against 2.098 × 10⁻⁷ mg/J for acoustic cavitation. Those are degradation-chemistry yields; neither figure measures gas dissolution, and no like-for-like comparison across the three methods has been published.
Dollars per cubic meter of a chemical conversion, not energy per kilogram of oxygen dissolved.
Source: Sun & Liu (2023)
Take that caveat seriously, because it is the whole reason the paragraph above names no winner. Cavitational yield measures how much pollutant a process destroys per joule; the biodiesel comparison in the same literature measures dollars per cubic meter of a chemical conversion. Both are downstream of cavitation and both are published research rather than our own measurement. What they establish is how efficiently each method turns input energy into cavitation events. They do not establish energy per kilogram of oxygen dissolved, and no one should quote them as if they did.
There is a structural argument that does not depend on those numbers, and it is only worth anything if both sides are counted from the same place. Start both at the wall. An acoustic system takes electricity through a power supply, a piezoelectric stack and a horn before any of it is a pressure wave in the liquid. A hydrodynamic system takes electricity through a motor and a pump before any of it is flow. Every one of those steps loses something, on both sides, and a comparison that begins the hydrodynamic count at "pump work" has quietly given one method its conversion losses for free.
Nor is the hydrodynamic input the energy the plant already spends moving the water. A cavitating constriction earns its cavities by imposing an additional pressure drop across the line, and the pump work needed to recover that drop is the real cost of the device — it rides on the existing duty rather than being covered by it, which is also why the constriction has to be sized to a design point rather than dropped into any convenient pipe. What that extra head costs per unit of gas dissolved is the number to ask any vendor for, ours included, and it is specific to a flow rate and a line pressure rather than a property of the mechanism.
Membranes sit apart from that comparison because their energy goes somewhere else entirely: into compressing the gas above the bubble point and into the crossflow that sweeps the surface. At low gas rates that can be very economical. What no one has published is a clean, like-for-like energy-per-unit-of-oxygen comparison across all three at industrial flow, and pretending otherwise is how this field acquired its credibility problem.
What limits throughput?
Throughput limits differ by mechanism, and none of the three scale the same way. Hydrodynamic cavitation scales with pipe flow, so capacity follows the pump. Acoustic cavitation is limited by how far the sound field reaches from the emitter. Membranes scale by adding area, which means adding modules and the footprint to hold them.
Hydrodynamic scaling is the least constrained, and it is not unconstrained. A cavitating constriction has a design point — a pressure ratio across the throat that puts the device in the regime it was sized for. Run well below it and cavitation stops; run well above it and the cavities merge into a single vapor pocket that damps the collapses instead of concentrating them.
The device scales with the pipe, but it has to be resized with the pipe, not simply fed more water.
Acoustic scaling has the hardest ceiling, and it is a physical one. Sound attenuates as it travels, so the active zone is a shell around the radiating surface rather than the whole vessel. Doubling the tank does not double the treated volume; it leaves most of the new volume outside the field. Larger duty is met by adding transducers and manifolds, so capital and complexity climb roughly linearly with flow, which is exactly the wrong shape for a large continuous stream.
Membranes scale linearly by construction, which is a virtue and a bill. Gas flux per unit of area is bounded by pore density and by the crossflow available to clear the surface, so more gas means more area, more housings, more manifolding, and more of the plant floor. The scaling is predictable, which matters for design, and it is not cheap, which matters for the capital request.
What fails first in the field?
Field failures cluster by mechanism rather than by brand. Membranes foul and blind, and the pores that set bubble size are the first thing lost. Cavitation hardware erodes where the cavities collapse against metal. Acoustic systems lose the emitting surface itself, which is a consumable by design.
Membrane fouling is the most predictable failure in the set, and the most often left out of a quotation. Anything the water carries — scale from hard water, biofilm, iron, fine solids — accumulates on and in the pores. As pores blind, gas flux falls, the pressure needed to maintain the bubble point rises, and the size distribution drifts upward because the surviving pores are the larger ones. Recovery is a cleaning regime, chemical or backpulse, and eventually replacement. Ask any membrane vendor for the replacement interval on your water, not on clean water.
Cavitation erosion is the hydrodynamic equivalent, and it comes from the same event that does the useful work. A cavity collapsing against a metal surface drives a microjet at it, and repeated often enough that pits the surface. Geometry, material choice, and keeping the device at its design point are the controls; a machine run hard off-design erodes its own throat and drifts out of specification as it does. Small orifices add a second, duller failure: they plug. Upstream straining is not an accessory.
Acoustic systems fail at the horn tip or the transducer bond. The radiating surface is being asked to survive the exact collapse energy it is generating, at close range, continuously — so tip erosion is a scheduled replacement rather than a fault, and the bonded stack behind it degrades with thermal cycling. On the other hand, an ultrasonic system has no small passages to plug, which is a genuine advantage on dirty water.
Two things every method shares.
- Water chemistry. It limits bubble life after generation regardless of what made the bubble, because high ionic strength compresses the electrical double layer that keeps a population from coalescing — the stability physics is a separate post.
- No comparative reliability data. There is no public, like-for-like field-reliability dataset across the three methods: the failure modes above are well characterized, while comparative service lives on real water are not published by anyone, ourselves included.
That is a gap in the field, not a detail. Treat any vendor's uptime figure, ours included, as something a pilot on your own water is meant to test.