Thermal Interface Material Dispensing Control on Assemblies

A thermal interface material fills the microscopic gaps between a component and its heat sink, and its purpose is to replace the air that would otherwise occupy those gaps. Air is a poor conductor, so the interface is often the largest single resistance in a thermal path, particularly where a small die has to transfer a high power density into a large heat sink. Applying the material correctly is therefore a process with a thermal consequence, and the two parameters that matter most are the volume that is dispensed and the thickness of the layer that results.

What Thermal Interface Material Does

The material is a soft, thermally conductive compound that conforms to the surfaces it touches. It does not conduct heat as well as metal, but it conducts far better than air, and the surface roughness of a real component and a real heat sink means that a significant fraction of the interface would be air without it.

The target is a thin, uniform layer with no gaps, and the difficulty is that both properties have to be achieved across a surface that is neither flat nor smooth. A layer that is too thin does not fill the roughness, while one that is too thick adds its own thermal resistance in proportion to its thickness. The process window is therefore narrow, and it is defined by the bond line thickness rather than by the mass of material used.

Material Types and Selection

Thermal interface material is supplied as a grease, a paste, a gap filler, a phase change material or a cured pad. The choice depends on the gap, on the assembly method and on whether the product will be reworked, and the thermal conductivity quoted on the datasheet is only one of several relevant properties.

Where the gap is large, a gap filler with high conductivity is needed; where the surfaces are flat and the pressure is controlled, a thin grease gives a lower total resistance. The selection should be made on the resistance of the finished layer, which depends on the thickness and the pressure as much as on the conductivity.

Dispensing Methods

Material is applied by syringe, by a time and pressure dispenser, by a positive displacement pump or by a stencil print. The method determines the repeatability of the volume, and repeatability is what makes the bond line thickness predictable.

Time and pressure dispensing is the simplest and the least repeatable, because the volume depends on the viscosity, which changes with temperature and with the age of the material. Positive displacement pumping is more expensive and far more consistent, and it is the normal choice where the thermal requirement is tight.

Volume and Bond Line Thickness

The volume is chosen to produce the intended bond line thickness after the parts are pressed together, and the relationship depends on the area of the interface and on how much material escapes at the edges. The calculation is a starting point; the actual thickness has to be verified by measuring the assembled part, because the amount of material that escapes at the edges depends on the surface energy of the parts and on the closing speed.

Bond line thickness is normally measured by a section, by an ultrasonic method or by a thermal measurement that infers the layer from the temperature difference. Whichever method is used, the measurement should be made on the production assembly rather than on a flat test coupon, because the real surfaces have their own flatness and their own roughness. The tolerance that applies to the finished layer is the same kind of tolerance that governs any other assembly dimension, as described in the guide to component tolerance and reliability.

Pattern and Coverage

The pattern controls how the material spreads when the parts come together. A dot spreads outward in a circle, a line spreads along its length and a cross or a star pattern spreads in several directions at once. The pattern should be chosen so that the material reaches the corners of the interface before it escapes at the edges.

Coverage is the result. A pattern that is too small leaves the corners dry, which creates a hot spot that no amount of material elsewhere will fix. A pattern that is too large wastes material and produces squeeze out that can contaminate neighbouring components or short a connection, and the excess can also lift a component that is supposed to sit flat.

Void and Air Entrapment

Voids form when air cannot escape as the material spreads. A dispensed dot that is pressed quickly traps a ring of air around its perimeter, and a pattern that closes off the path to the edge traps air in the middle. The result is a layer with a thermal resistance that is higher than the calculation and a hot spot that moves with the position of the void.

Control of voids comes from the pattern, from the closing speed and from the placement force. Slower closing, a pattern with a defined escape path and a controlled pressure all reduce the amount of air that is trapped, and the effect should be verified by inspecting the squeeze out and by an X-ray or ultrasonic check where the requirement is critical.

Curing and Handling

Some materials cure and some remain as a paste, and the two behave differently in the assembly. A curing material has a defined pot life, a defined open time after dispensing and a defined cure schedule, and a part that is assembled outside those windows will not achieve the intended thickness.

Non curing materials stay soft, which is useful for rework but means that the layer depends entirely on the mechanical pressure of the assembly. In both cases the material has a shelf life and a storage condition, and an expired or badly stored material will not dispense consistently whatever the machine settings are. Material that has been frozen must be thawed to room temperature before use, or condensation will introduce water into the layer.

Inspection and Verification

Inspection is difficult because the layer is hidden once the parts are assembled. The practical checks are the dispensed volume, which can be measured on a sample, the squeeze out pattern, which shows how the material spread, and the thermal performance of the finished assembly, which is the property that matters.

Where the requirement is critical, a section or an ultrasonic scan of a sample gives the bond line thickness directly. The measurement should be made on a defined frequency of assemblies and the results recorded, so that a change in the process is visible from the data rather than from a field failure. On a metal backed or ceramic substrate the interface behaves differently again, and the comparison of those constructions is described in the guide to metal core and ceramic boards.

Process Control and Records

The controls are the material, the machine, the pattern and the assembly pressure. Each of them should be recorded for the lot, together with the dispensing parameters and the verification result, so that a thermal problem can be traced back to the process that produced the part.

Records also make the process transferable, because a line that changes shift or operator will reproduce the same result only if the parameters are written down. The discipline is the same one applied to the rest of the thermal design, which is described in the guide to thermal management design.

FAQ

How much thermal interface material should be dispensed? Enough to fill the roughness and the flatness of the interface and no more, which is determined by the target bond line thickness rather than by a fixed weight. The right volume is the one that produces the intended thickness after the assembly is closed.

Does a higher conductivity always give a lower temperature? No. The total resistance is the conductivity divided by the thickness, so a moderately conductive material applied as a thin, uniform layer can outperform a highly conductive material applied too thickly. The layer thickness is the variable that is most often wrong.

Can the layer be inspected after assembly? Only indirectly. The squeeze out and the measured thermal performance are the practical checks, and a section or an ultrasonic scan on a sample gives the thickness directly. Visual inspection of the assembled part tells very little about the interface underneath.