What a micro fan or microcooler actually is at chip scale: miniature fans, MEMS coolers, micro-blowers, synthetic jets and DBD plasma, compared fairly.
Sunon markets its Mighty Mini range as "the smallest (8mm), the thinnest (3mm) and the lowest energy-consuming innovative design for mini fans and blowers" [1]. That is roughly where a rotating fan stops being buildable. In the same market, xMEMS publishes a µCooling part measuring 9.26 x 7.6 x 1.08 mm [2][3][4]. Two air movers, comparable footprints, and one of them is a third of the height of the other because it contains no rotor.
If you have searched for a micro fan, a microfan or a microcooler, that gap is the thing you are actually shopping across. The word covers at least five mechanisms that share a size class and share almost nothing else.
A micro fan is a millimetre-scale air mover mounted on or beside a hot chip, and at that scale the category spans five distinct mechanisms: miniature rotary fans, piezoelectric MEMS coolers, piezoelectric micro-blowers, synthetic jets, and surface plasma actuators with no moving element at all.
A conventional axial fan or blower is the default answer, and for good reason: it is cheap, it is understood, and the supply chain is deep. Sunon's own figures put the practical bottom of that range at an 8 mm frame and 3 mm of height [1].
The floor is not a manufacturing accident. Three things bind at once. The bearing needs physical volume, and shrinking it shortens life rather than preserving it. The hub and motor take a growing share of the swept area as the frame shrinks, so less of the fan is actually fan. And to hold airflow constant as the rotor gets smaller, blade tip speed must rise, which raises tonal noise. Fans also fail in a familiar way: bearing wear and dust ingestion. In a sealed, dusty or continuously loaded box, the intake and the bearing fail before the silicon does. We look at ionic wind against fans in more detail separately.
None of this means small fans are bad. It means that below a few millimetres of z-height, the question stops being which fan and starts being whether a fan.
These replace the rotor with a membrane driven at ultrasonic frequency, rectifying the oscillation into net airflow.
Frore Systems describes AirJet as using "Ultrasonic MEMS membranes pull in air and create high-velocity pulsating jets of air that saturate with heat" [5]. The AirJet Mini G2 is published as 2.65 mm thick, 7 g, removing 7.5 W of heat per chip, generating 1750 Pascals of back pressure at 21 dBA [5]. The earlier AirJet Mini is listed at 5.25 W of heat removal, the same 1750 Pa of back pressure, a maximum of 1 W of power and 21 dBA [6].
xMEMS builds the same idea in silicon. Its XMC-2400 is published at 9.26 x 7.6 x 1.08 mm with bi-directional flow adjustable up to 39 cc/s, an estimated 30 mW of power, an IP58 rating, SMT-reflowable assembly, and operation described as "Inaudible; all mechanical operation is at ultrasonic frequencies" [2][3]. Converge Digest reports the same overall dimensions and describes it as a "fan-on-a-chip design" using "piezoMEMS technology" [4].
Two things deserve saying plainly. First, that back pressure figure is the important one. 1750 Pa is what lets a thin device push air through a dense fin stack, where most thin-form air movers simply choke. Second, MEMS coolers are further along in commercial shipping than DBD plasma is, and it would be dishonest to imply otherwise. Their wear mechanism is fatigue of a moving membrane and its piezoelectric film.
The oldest member of the family. Murata's MZB1001T02 drives an air chamber with, in Murata's words, "the ultrasonic vibrations of the ceramic as the drive source" [7]. Its datasheet gives 20.0 +/-0.15 mm outer dimension, 1.85 +/-0.2 mm thickness excluding the nozzle, flow of at least 0.70 L/min, static pressure of at least 1.42 kPa, a drive voltage of 10 to 20 Vp-p and a resonant frequency of 24.0 to 27.0 kHz [8].
That is a strong pressure head from a 1.85 mm part orderable through ordinary distributors for years. The catch is flow rate: 0.70 L/min is under 12 cc/s, and the device is specified as an air pump, so there is no published thermal rating. In practice these serve gas sampling more often than they cool silicon.
A synthetic jet oscillates a diaphragm over a cavity with an orifice, ejecting a train of vortices without any net mass entering the system. Advanced Thermal Solutions describes it as "a zero-mass-flux jet comprised entirely of the ambient fluid" [9]. Because the jet is unsteady, it disrupts the thermal boundary layer more aggressively than steady flow at the same average velocity. The same article reports that "for an SPL-A of 40 dBA, the synthetic jet solution had a 12% better thermal performance than the fan-sink", and that "for small fans, it was found that the L10 reliability at 60ºC is around 50,000 hours, while for the Nuventix SynJet is at least 300,000 hours at 60ºC" [9]. That reliability comparison is a vendor-supplied figure from 2011, so treat it as directional rather than as a specification.
A dielectric barrier discharge actuator places two electrodes either side of a thin dielectric, one exposed to air and one encapsulated. A high-voltage AC waveform ionises a shallow layer of air at the exposed electrode edge, the ions accelerate along the surface, and they drag neutral air with them into a wall jet of a few metres per second. The discharge is non-thermal, so the gas stays near ambient temperature and adds no meaningful heat to what it cools. The Y-Flow actuator is roughly 200 micrometres thick and has no moving element whatsoever.
The evidence base is narrow but real. Go, Garimella, Fisher and Mongia found that in the presence of a bulk flow, ionic winds "distort the boundary layer, increasing heat transfer from the wall", with experiments "corresponding to local enhancement of the heat transfer coefficient by more than twofold" [10]. Superimposed on an existing 0.3 m/s bulk flow, that work measured roughly 20 K of additional cooling while drawing under 100 mW, specifically 67 mW at 15 µA [10]. A 2021 review in Sādhanā catalogues the wider literature on ionic wind in thermal management, listing "silent operation, quick response, minimum power and compactness" as its favourable characteristics [11].
| Approach | Published thickness | Moving parts | Failure mode | Best at |
|---|---|---|---|---|
| Miniature axial fan or blower | 3 mm thinnest, 8 mm smallest frame [1] | Rotor, bearing, motor | Bearing wear, dust ingestion | Bulk volumetric flow, lowest cost, deepest supply chain |
| Piezoelectric MEMS cooler | 2.65 mm (AirJet Mini G2) [5]; 1.08 mm (xMEMS XMC-2400) [2][3][4] | Vibrating membrane | Membrane and piezo film fatigue | High static pressure in a thin package; 1750 Pa through a fin stack [5][6] |
| Piezoelectric micro-blower | 1.85 mm excluding nozzle [8] | Vibrating ceramic diaphragm | Diaphragm and valve fatigue | Pressure head from a mature, distributor-stocked part |
| Synthetic jet | Varies by module, not published as a single figure | Oscillating diaphragm | Diaphragm fatigue | Unsteady boundary-layer disruption at low acoustic output [9] |
| Surface DBD plasma actuator | Roughly 200 µm | None | Dielectric ageing under electrical stress | Adding local flow where nothing fits and nothing currently moves |
| Thermoelectric (Peltier) | Module dependent | None | Solder fatigue under thermal cycling | Driving a spot below ambient |
| Vapour chamber or heat pipe | Module dependent | None (passive) | Wick dry-out, non-condensable gas | Spreading heat, never rejecting it |
Thermoelectric modules and vapour chambers sit in the last two rows because searchers reach them from the same queries, but they are adjacent tools rather than competitors. A vapour chamber moves heat sideways and rejects none of it, so it still needs an air mover downstream. A Peltier can push a junction below ambient, which nothing else here can do, at a poor coefficient of performance; we cover plasma actuators against thermoelectric coolers separately.
Every option on that table loses somewhere, including ours.
Miniature fans lose in sealed, dusty or continuously loaded enclosures, and below about 3 mm they run out of geometry. MEMS coolers and micro-blowers remove a bounded amount of heat per part, so scaling means adding parts and board area, and their wear mechanism is fatigue of something that moves millions of times an hour. Synthetic jets have no consistent published spec sheet across vendors, which makes them hard to design against.
Surface DBD plasma loses in three clear places. Static pressure is negligible, so it will not force air through a restrictive fin stack, and against 1750 Pa it is not in the same contest. Bulk flow is metres per second, not tens of metres per second, so a genuinely high volumetric flow problem still wants a fan. And electrical-to-fluid conversion efficiency for corona-based ionic wind is poor in absolute terms, which only stops mattering when the honest alternative in that space is no airflow at all. It also needs a kilovolt-class AC supply, generates some ozone depending on waveform, geometry and duty cycle, and its dielectric ages under continuous electrical stress. YPlasma is early stage and the Y-Flow actuator is in development, not shipping in volume the way the MEMS parts above are.
The practical way to choose is to work out which constraint actually binds. If it is volume of air, a fan still wins. If it is pressure through a fin stack in a thin chassis, the MEMS coolers are the strongest published answer. If it is height, and the surface you need to cool has under a millimetre of clearance, a 200 micrometre film is the only thing in the list that fits. That last case is common in edge AI thermal management and in the stagnant air around dense accelerator boards, which is why fanless GPU cooling keeps coming back as a question. For a head-to-head of the vendors shipping today, see our solid-state cooling landscape.
A micro fan is a millimetre-scale air mover placed on or beside a hot chip. The term now covers miniature rotary fans, piezoelectric MEMS coolers, piezoelectric micro-blowers, synthetic jets and surface plasma actuators, which share a size class but work by entirely different physics.
Sunon publishes its Mighty Mini range as the smallest at an 8 mm frame and the thinnest at 3 mm [1]. Below that, the bearing, hub and motor stop fitting usefully, and the blade tip speed needed to hold airflow raises noise instead.
In chip cooling, microcooler usually means a solid-state air mover with no rotor. Published examples include the xMEMS XMC-2400 at 9.26 x 7.6 x 1.08 mm [2][3][4] and the Frore AirJet Mini G2 at 2.65 mm thick [5].
MEMS coolers are further along. Frore and xMEMS publish full product specifications and ship commercially. DBD plasma actuators, including YPlasma's Y-Flow, are earlier stage and in development.
Usually yes. Every air mover here removes heat from a surface, and a spreader or fin stack is what gives it enough surface area to work against. Vapour chambers and heat pipes spread heat but reject none of it, so they pair with an air mover rather than replace one.