Passive cooling hits a hard physical ceiling as heat flux rises. Here is where active cooling takes over, and what the fanless options actually do.
TL;DR: Passive cooling is limited by natural convection, which tops out at a heat transfer coefficient of roughly 25 W/m2K in air. Active cooling raises that ceiling by an order of magnitude, but active no longer has to mean a rotating fan. The decision hinges on heat flux and enclosure constraints, not on total device power.
Thermal engineers get asked the wrong question constantly: Do I need active cooling for a 20 W board? Total power tells you almost nothing on its own. A 20 W load spread across a 200 cm2 aluminium chassis is a passive problem. The same 20 W concentrated into a 1 cm2 package inside a sealed plastic housing is an active problem, and no amount of extra fin area will fix it.
What actually separates the two regimes is a single physical quantity: the convective heat transfer coefficient, h, at the surface where heat finally leaves your product and enters the air. Everything else in the thermal stack, from die attach to vapour chamber, is transport. The last few hundred micrometres of stagnant air are where the design usually fails.
Passive cooling moves heat using conduction, natural convection and radiation, with no energy input beyond the heat itself. Heatsinks, heat spreaders, chassis-as-radiator designs, heat pipes and vapour chambers all fall in this category. Heat pipes deserve a caveat: they are outstanding transporters of heat but they cannot reject it. A heat pipe moves your problem to a different surface. Something still has to hand that heat to the air.
Active cooling adds external work to increase heat transfer at the boundary. Historically that has meant a fan or a pump forcing fluid across a surface. More recently it also includes thermoelectric modules that pump heat against a gradient, and solid-state air movers that create flow without rotating machinery.
The distinction that matters is not moving parts versus no moving parts. It is whether you are paying energy to renew the fluid at the hot boundary. That is what buys you a higher h.
Standard reference values for the convective heat transfer coefficient in air, as tabulated in Incropera and DeWitt, Fundamentals of Heat and Mass Transfer, set the boundaries of the discussion:
| Regime | h (W/m2K) |
|---|---|
| Free (natural) convection, air | 2 - 25 |
| Forced convection, air | 25 - 250 |
| Forced convection, liquids | 50 - 20,000 |
| Boiling and condensation | 2,500 - 100,000 |
Two things follow immediately.
First, the passive ceiling is low and it is hard. Natural convection in air is driven by buoyancy alone: warm air near the surface becomes less dense, rises, and is replaced. That circulation is slow, so a thick, sluggish thermal boundary layer sits on the hot surface and dominates the total resistance. You can enlarge the surface area, but you cannot make buoyancy stronger. Realistically, a well-designed passive heatsink in free air handles on the order of 0.05 to 0.1 W/cm2 before surface temperature runs away.
Second, the step from natural to forced convection is roughly a factor of ten. That single order of magnitude is the entire value proposition of active cooling. Everything after it, from better fin geometry to liquid loops, is refinement within the forced-convection regime.
There is also a geometric trap. Fin area does not scale for free. Adding fins increases surface area but also increases flow restriction and reduces the temperature difference driving the flow. In natural convection, fin spacing below roughly 6 to 8 mm starts choking the buoyant plume, so past a point you add mass and cost while gaining almost nothing. Passive designs run out of headroom faster than the surface-area arithmetic suggests.
Four failure modes recur, and it is worth recognising which one you have because they have different fixes.
Sustained throttling. The device passes a short benchmark and fails a long one. Passive systems have large thermal mass, so they absorb transients well and fail at steady state. If your part is fine for ninety seconds and throttled at ten minutes, you are at the passive ceiling.
The heatsink is winning the industrial design argument. When the thermal solution dictates enclosure volume rather than fitting inside it, passive has stopped being cheap. A 40 mm finned extrusion on a 15 mm-thick product is not a solution, it is a redesign request.
Sealed enclosures. IP65 and above, outdoor telecom cabinets, automotive control units, medical devices, industrial equipment near dust or coolant mist. Ambient air cannot enter, so the internal air is a closed volume that stratifies and stagnates. Internal air temperature climbs well above the external ambient, and the heatsink you sized against outdoor ambient is now working against a much smaller delta T. This is the case where designers most often discover the ceiling late.
Hot spots. Modern accelerators concentrate heat at the die in ways that spreaders cannot fully smooth. The average surface temperature looks acceptable while a local junction sits 30 K higher. Passive spreading helps, but the last resistance at the surface is still convective, and averaging does not remove a peak.
Rotating fans and blowers. Mature, inexpensive, and unmatched on static pressure. A thin laptop blower delivers tens of pascals, enough to force air through a dense fin stack and a long duct. That pressure capability is the reason fans remain the default and the reason they are hard to displace in fin-and-duct architectures. The costs are equally well known: audible and tonal noise that ramps with load, vibration coupled into optics and sensors, bearing wear, dust ingestion that degrades performance across the deployment life, and a hard z-height floor set by the rotor and hub. Below roughly 5 mm there is very little left to spin.
Liquid cooling. The right answer above roughly 300 W per package or where heat must be moved a long way. It brings pumps, plumbing, leak risk and service burden, and at board level in a sealed edge product it is usually disqualified on complexity rather than on physics. Our guide to direct-to-chip cooling covers what cold plates can and cannot reach.
Thermoelectric (Peltier) modules. The only option here that can drive a surface below ambient. That capability costs a coefficient of performance below one in practical electronics use, meaning every watt of heat you move costs more than a watt of input, and all of it becomes additional waste heat that still has to be rejected conventionally. Excellent for precision spot cooling of sensors and optics; poor as a system-level strategy. We compare the trade-offs directly in plasma actuators vs thermoelectric coolers.
Solid-state air movers. Piezoelectric MEMS membranes, micro-blowers and electrohydrodynamic devices all create airflow without a rotor. They occupy a genuinely new point in the design space: active heat transfer in packages one to three millimetres thick, with no bearing to wear. Our solid-state cooling guide compares the families in detail.
The electrohydrodynamic approach is worth isolating because it is the one that most directly attacks the boundary layer rather than the duct.
In a surface dielectric barrier discharge (DBD) actuator, two electrodes are separated by a thin dielectric layer, one exposed to the air and one encapsulated. Driven with low-frequency AC at kilovolt amplitudes and sub-milliamp current, the exposed edge weakly ionises a shallow layer of air. Those ions accelerate along the field and transfer momentum to the overwhelmingly neutral surrounding air through collisions, producing a wall jet tangent to the surface. Moreau's review in the Journal of Physics D remains the standard treatment of the mechanism and its scaling; Benard and Moreau later provided the most complete electrical and mechanical characterisation of surface AC DBD actuators, reporting induced velocities of the order of several metres per second from a device well under a millimetre thick.
The relevant point for thermal design is where the flow is created. A fan generates bulk flow somewhere else and pushes it toward the surface, where the boundary layer still has to be overcome. A surface actuator generates its jet inside the boundary layer, at the exact location where the dominant resistance lives. Go, Garimella, Fisher and Mongia demonstrated the consequence directly: superimposing an ionic wind on an existing 0.3 m/s bulk flow roughly doubled the average heat transfer coefficient and produced about 20 K of additional cooling, while drawing under 100 mW at the actuator.
This is also why comparing these devices to fans on airflow alone is misleading. Bulk CFM is not the figure of merit when the device is acting on the boundary layer rather than through a duct. The honest framing is delivered cooling per unit of input power in the specific geometry, not catalogue airflow.
The trade is real and should be stated plainly. Surface DBD actuators deliver a static pressure on the order of ten pascals, well below what a blower produces. They will not force air through a dense fin stack or a long restrictive duct, and where raw pressure is the binding constraint a fan remains the better engineering answer. They also require a kilovolt-class AC supply, and any discharge in air generates some ozone, which is a design constraint addressed with catalytic materials and validated by measurement against the applicable 0.1 ppm limit rather than dismissed. We treat that question at length in ionic wind reliability and ozone, explained.
Where they win is where nothing else fits: short, low-impedance flow paths, direct placement at the heat source, sealed volumes where the job is to stir trapped internal air rather than exchange it with outside, and z-heights that disqualify every rotating option. Our comparison of ionic wind vs fans goes through the crossover case by case.
Passive cooling is not a lesser option; it is a bounded one. Within its envelope it is silent, free, and effectively immortal, and the correct move is always to exhaust the passive design first. But the envelope is set by physics, not by effort: natural convection in air gives you an h of at most about 25 W/m2K, and no additional fin area repeals that.
What has changed recently is what happens after the ceiling. For fifty years, crossing it meant accepting a rotor, a bearing and a duct. Sealed, thin and silent products simply had no active option and were designed down to whatever passive could carry. Solid-state air movers do not beat fans on pressure or bulk flow, and they should not be sold as though they do. What they do is make active cooling available in enclosures that were previously restricted to passive, which is a different and more useful claim.
If your design is sealed, thin, or limited by the boundary layer rather than by the heatsink behind it, that is the case worth evaluating. See the technology or book a technical demo and we will look at your geometry with you. If a fan is the right answer for your architecture, we will tell you that too.
Passive cooling removes heat using only conduction, natural convection and radiation, with no energy input: heatsinks, heat spreaders and vapour chambers. Active cooling spends energy to renew the fluid at the hot surface, raising the convective heat transfer coefficient from roughly 2-25 W/m2K in still air to 25-250 W/m2K under forced convection. That order-of-magnitude difference is the whole reason active cooling exists.
When heat flux at the rejection surface exceeds roughly 0.05-0.1 W/cm2 in free air, when the enclosure is sealed so internal air stagnates and stratifies, when sustained load causes throttling that short benchmarks hide, or when the heatsink required has grown large enough to dictate the product's dimensions. Heat flux and enclosure type predict this far better than total device wattage.
Yes. Thermoelectric modules, piezoelectric MEMS membranes, micro-blowers and electrohydrodynamic devices such as DBD plasma actuators all add energy to enhance heat transfer without a rotor or bearing. Surface DBD actuators are sub-millimetre films that generate a wall jet directly inside the thermal boundary layer. They deliver much lower static pressure than a fan, on the order of ten pascals, so they suit short low-impedance flow paths and sealed volumes rather than dense fin stacks and long ducts.