What is jet impingement cooling?

It is a cooling method in which a high-speed fluid jet from a nozzle is directed onto the surface to be cooled, forcing a very thin boundary layer to form at the impact point. Compared with parallel-flow convective cooling, the local heat transfer coefficient near the stagnation point can be several to more than ten times higher, and the method is widely used for cooling semiconductors, turbine blades and glass tempering.

Why is the heat transfer coefficient maximum at the stagnation point?

When the jet hits the plate, the velocity at that point is nearly zero and the pressure is at its maximum. The flow decelerates perpendicular to the plate, which crushes the thermal boundary layer to a very small thickness. A thinner boundary layer means less thermal resistance and higher h, so h peaks at the stagnation point and then decays with increasing radius as the wall jet develops.

How should the standoff ratio H/D be chosen?

In practice H/D = 4 to 8 is commonly used. If H/D is too small the potential core of the nozzle hits the plate directly with insufficient turbulence, and noise and pressure loss grow. If H/D is too large the jet entrains surrounding air, decelerates and loses momentum at the stagnation point, so h drops. The simplified Martin-type correlation also peaks near H/D = 6.

How is this used for electronics cooling and glass tempering?

For CPUs and IGBT power modules, nozzle arrays direct air or liquid jets onto the back side of a heat sink to remove local high heat flux. In physical (air-quench) glass tempering, jets of air are blown onto hot glass to quench the surface quickly, producing compressive residual stress that increases strength. Both rely on the localized cooling capability of impinging jets.

Jet Impingement Cooling Simulator — Free Online Calculator

Parameters

Nozzle exit velocity U

m/s

Nozzle diameter D

Standoff ratio H/D

—

Temperature difference T_w − T_∞

The fluid is fixed to air (ρ=1.2 kg/m³, ν=1.5×10⁻⁵ m²/s, k=0.026 W/(m·K), Pr=0.71).

Results

—

Reynolds number Re

—

Nusselt number Nu

—

Heat transfer coefficient h

—

Heat flux q = h·ΔT

Nozzle, jet and impingement plate

Top: nozzle / Blue arrow: jet / Bottom: plate (red = hot, blue = cool) / Yellow dot: stagnation point

Radial profile h(r/D)

Horizontal axis: dimensionless radius r/D / Vertical: local heat transfer coefficient h (yellow dot: stagnation r/D=0)

Theory & Key Formulas

When a round air jet impinges perpendicularly on a flat plate, the thermal boundary layer is crushed very thin near the stagnation point and the heat transfer is strongly enhanced. This tool uses a simplified Martin-type correlation that is widely cited in heat transfer textbooks.

Reynolds number at the nozzle exit. U is the exit velocity, D the nozzle diameter, ν the kinematic viscosity:

$$Re = \frac{U\,D}{\nu}$$

Representative Nusselt number. Pr is the Prandtl number, H/D is the standoff ratio:

$$Nu = Re^{0.5}\,Pr^{0.42}\,(H/D)^{-0.1}$$

Heat transfer coefficient h and heat flux q. k is the air thermal conductivity, T_w−T_∞ is the wall-to-fluid temperature difference:

$$h = \frac{Nu\,k}{D},\qquad q = h\,(T_w - T_\infty)$$

h is maximum at the stagnation point and decays with increasing r/D (approximated here by an exponential decay).

What is the jet impingement cooling simulator?

🙋

A CPU is tiny but generates a lot of heat. I have heard that sometimes a plain fan just is not enough — is there a more special cooling method?

🎓

One of the answers is jet impingement cooling. Roughly speaking, a jet of air or liquid is shot from a nozzle directly onto the surface you want to cool, and a very thin boundary layer is forced at the impact point. Try increasing the "Nozzle exit velocity U" above — both Reynolds and Nusselt numbers shoot up.

🙋

The radial profile chart really does have its peak right in the middle. But why is the spot where the jet stops the coolest? The flow seems to be standing still there.

🎓

Good question. The velocity is indeed nearly zero at the stagnation point, but in exchange the jet decelerates perpendicular to the wall, and that crushes the thermal boundary layer extremely thin. Remember: thermal resistance ≈ boundary-layer thickness / conductivity. A thinner layer means lower resistance and higher h, so "slow flow but lots of heat transfer" is exactly what happens here.

🙋

Got it. So the closer the nozzle to the plate, the better it cools? When I set H/D to 2 the Nu number looks largest.

🎓

The interesting thing is, getting too close does not monotonically help. If H/D is too small the potential core of the nozzle hits the plate before enough turbulence is generated, and noise and pressure loss go up too. In practice H/D = 4 to 8 is the sweet spot. The (H/D)^(-0.1) factor in the correlation is very weak — Nu only differs by about 1.3x between H/D=2 and 12. Think of it as a broad valley to land in.

🙋

The bottom-right card shows q = 13 kW/m². Is that a lot?

🎓

On a finned heat sink surface, natural convection gives 5 to 10 W/(m²·K), forced air convection around 30 to 80, and jet impingement can reach hundreds to thousands. With the defaults (U=30, D=10 mm) h is 266 and q is 13 kW/m² at ΔT=50 K — many times more than fan-forced cooling. From electronics cooling to glass tempering to internal cooling of gas-turbine blades, this method is used wherever a small region must be cooled hard.

Frequently Asked Questions

For CPUs, GPUs and IGBT power modules, nozzle arrays direct air or dielectric liquid jets onto the back of a heat sink. Air-based impingement reaches several hundred W/(m²·K) and liquid impingement can reach tens of thousands, removing heat densities that simple fans cannot handle. Applications include data centers and automotive power electronics.

Architectural and automotive tempered glass is made by heating a glass sheet near its softening point and then blowing arrays of air jets on both faces to quench it rapidly. The surface contracts first while the interior is still hot, leaving compressive residual stress on the surface and tensile stress in the core, which raises the strength of the glass several-fold. Nozzle pitch and H/D are key process parameters.

As the representative correlation Nu ∝ Re^0.5 shows, Nu grows with the square root of Re. Doubling the velocity only doubles Nu by a factor of √2 ≈ 1.4, while the fan power scales with the cube of velocity. Designers therefore pick an operating point in Re = 10,000 to 80,000 where the required h is met without excessive blower power.

A single nozzle peaks at the stagnation point and decays outward, so it cannot cool a large area uniformly. A nozzle array distributes many jets over the surface, and the design must consider jet-to-jet interference, the crossflow of spent air, and pitch s/D. The trade-off between average heat transfer and uniformity is what nozzle-array design is all about.

Real-world applications

Cooling of high-density electronics: Server CPUs and GPUs, IGBT inverters for electric vehicles, and laser diodes all use jet impingement to lower the local thermal resistance. Combinations with microchannels and direct dielectric-liquid impingement (using fluorinated fluids) are actively researched and increasingly deployed.

Internal cooling of jet engine and gas turbine blades: The leading edges of turbine blades exposed to hot combustion gas use impingement cooling from inside, where compressor air jets strike the inner blade wall. Combined with film cooling and convective cooling, this allows operation at gas temperatures far above the alloy melting point.

Rapid quenching in glass and metal heat treatment: Glass tempering, continuous heat-treatment lines for steel sheet, and cooling of extruded sections all use jet impingement to deliver a uniform, controllable cooling rate. The cooling rate directly determines residual stress, microstructure and strength, so nozzle design has a direct impact on product quality.

Drying in food and printing industries: Cereals, baked goods, paper and freshly printed or coated surfaces are dried with arrays of impinging hot or dry air jets. Heat and mass transfer are simultaneously enhanced, which gives short and uniform drying times for items moving on a conveyor.

Common misconceptions and caveats

The most common misconception is thinking that pushing the velocity up directly raises h in proportion. As the representative correlation Nu ∝ Re^0.5 shows, h scales only with the square root of nozzle velocity. Meanwhile fan or pump power roughly scales with the cube of velocity, so doubling the velocity gives only √2 ≈ 1.4 times higher h but uses about 8 times more power. Move the U slider in the tool and watch how Nu and the heat flux q grow more and more slowly. In design, "the minimum velocity that achieves the required h" is the key to energy efficiency.

A second common pitfall is assuming that the closer the nozzle is to the plate, the better the cooling. The (H/D)^(-0.1) factor in the correlation is very weak — Nu only differs by about 1.3x between H/D=2 and H/D=12. If H/D is too small, the potential core of the nozzle hits the plate without sufficient turbulence, and noise and backflow grow. In practice H/D = 4 to 8 is the sweet spot, and the choice within this "broad valley" trades off against assembly, noise and pressure loss.

Finally, remember that the tool shows the heat transfer coefficient near the stagnation point, not the average over the entire plate. As the h(r/D) plot shows, h decays rapidly outside the stagnation region — at r/D=3 it is already less than half of the peak value. For larger cooled areas you need nozzle arrays or area-averaged Nusselt correlations (full Martin or Goldstein correlations). Use this tool as a quick reference for an upper bound on hotspot cooling under a single nozzle.

Jet Impingement Cooling Simulator — Heat Transfer on the Plate

What is the jet impingement cooling simulator?

Frequently Asked Questions

Real-world applications

Common misconceptions and caveats

Related Tools