Ceramic Heat Dissipation—Less Is More

Feb. 23, 2010
Ceramic Heat Dissipation—Less is more

Optimisation blocks
Total thermal resistance
Rubalit versus Alunit
Heat spreading

LEDs suffer heat problems, which understandably can limit their success as a light source. Much attention is given to the heat sink, but less to the layers and barriers between LEDs and the heat dissipating surface.

A change of concept and material allows for significant gains in thermal management and reliability, as well as a simplified system. Using ceramics as heat sink, circuit carrier, and part of the product design needs some fresh thinking and the willingness to overcome traditional patterns. A simulation process based on Computational Fluid Dynamics supports thermal optimisation and technical product design. So let’s explain the theoretical approach, the proof of concept, and what and how improvement with ceramic heat sinks can be achieved.

LEDs are known to be efficient and are loved for being tiny. But they are only really tiny as long as heat management is not involved. Incandescent light sources work with temperatures up to 2,500°C; LEDs are much colder and many people stumble upon the fact that heat is such an issue. Being relatively cold, LEDs still do produce heat which is not yet a problem. But they are based on semiconductors which, roughly speaking, simply allow temperatures below 100°C.

According to the law of energy conservation, the thermal energy must be transferred to the surrounding area. The LED can only use a small temperature gap between 100°C of the hot spot and 25°C ambience temperature; offering just 75 Kelvin. Consequently a larger surface and powerful thermal management are needed.


Let’s take a look at the LED in Figure 1. Here Group 1 is the LED itself and mainly remains untouchable. Its centre is a die and a heat slug, a copper part, which connects the die with the bottom of the LED. Thermally, the ideal solution is direct bonding of the die to the heat sink itself. Due to mass production this concept is commercially unrealistic. We consider the LED as a standardised “catalogue” product which can not be modified. It is a black box.

Group 2 is the heat sink, transmitting energy from a heat source to a heat drain. This is usually the surrounding air either with free or forced convection. The less aesthetic the material, the higher is the need to hide it. The more you hide it the less efficient is the cooling. Alternatively, pleasing and worthy materials can be used, directly exposed to the air and being part of the visible product design.

In-between group one and two is Group 3 providing mechanical connection, electrical isolation, and thermal transmittance. That seems contradictory since most materials with good thermal conductivity conduct electricity as well. Vice versa, almost every electrical isolation material translates into a thermal barrier. The best compromise is soldering the LED to a PCB which is glued on the metal heat sink. The original function of a PCB as a circuit board can be kept. Although PCBs exist with various thermal conductivities, they remain an obstacle to thermal transfer.


The thermal resistance of LEDs (die to heat slug pad) and heat sinks is available from the manufacturer. But there is little focus on group 3 and its significant influence on the total thermal performance. Adding all thermal resistances but the LED (group 1), the total thermal resistance Rtt (Fig. 2) is born. The Rtt allows a real comparison of heat.


It is common to optimise only the heat sink. Hundreds of designs are available, essentially of aluminium. But for further improvement it is necessary to advance or even eliminate the third group. Electrical isolation has to come from the heat sink itself by the use of other materials. Our conclusion is ceramic. Ceramics, e.g. Rubalit (Al2O3) or Alunit (AlN), combine two crucial characteristics: they are electrically isolating and thermally conductive.

Rubalit has a lower, Alunit a slightly higher conductivity than aluminium. On the other hand Rubalit is less expensive than Alunit (Fig. 3). Their thermal expansion coefficient is adapted to semiconductors, they are rigid, corrosion-resistant, and RoHS compliant.  Completely inert, they are the last part of a system to die. The simplified construction (without glues, insulation layers, etc.) combined with a direct and permanent bond between the high-power LED and the ceramic heat sink create ideal operating conditions for the entire assembly. Put simply, what isn’t there won’t wear out and materials that expand in proportion to each other won’t separate. The result is excellent long-term stability, secure thermal management, and exceptional reliability. A patent has been filed and the concept has been named CeramCool. 


The ceramic heat sink CeramCool is an effective combination of circuit board and heat sink for the reliable cooling of thermally sensitive components and circuits. It enables the direct and permanent connection of components. Also, ceramic is electrically insulating per se and can provide bonding surfaces by using metallisation pads. Customer-specific conductor track structures, even three-dimensional ones, can be provided if required. For power electronic applications direct copper bonding is possible. The heat sink becomes a module substrate that can be densely populated with LEDs and other components. It quickly dissipates the generated heat without creating any barriers.


The idea to use ceramics was first cross-checked in several simulation models. To predict thermal behaviour of various designs a method based on Computational Fluid Dynamics (CFD) was developed. Equally, an optimised ceramic heat sink for 4W cooling was developed. Manufacturing requirements where taken into account. The optimised geometry allows operation of a 4W LED at a maximum temperature below 60°C which was validated against physical tests. The design is square in shape (38mm x 38mm x 24mm) and comprises longer, thinner fins with a larger spacing. The identical geometry in aluminium with a PCB mounted LED showed significant higher temperatures. Depending on the thermal conductivity of the PCB (from λ = 4 W/mK to λ = 1,5 W/mK), the temperature raised between 6 to 28K.

Already a 6K reduction at the hot-spot implies significantly less stress for the LED. The total thermal resistance of the Rubalit assembly is at least 13% better than aluminium with identical shape. Using Alunit the minimum improvement of CeramCool reaches 31%. These good results are outperformed largely for both ceramics if the heat drop of 28K is taken into account.


The concept is flexible and can be used for different targets. It’s your choice whether you run an LED on its optimum temperature assuring high life time and high lumen per Watt or you accept higher temperatures reducing life time and efficiency. A temperature spread from 50°C to 110°C is common. If more lumina are needed, the 4W-heat sink can be equipped with 5W or 6W LEDs. Splitting the power into several 1W LEDs helps to get better heat spreading (Fig. 4). The results are 65°C with 5W and 70°C with 6W.


With the chip permanently and reliably bonded to the electrically insulating CeramCool, the heat sink takes more heat and becomes hotter. It takes the burden off the LED and does exactly what it is made to do, namely, cool the critical components. The reduced die temperature allows a downsized surface and a smaller heat sink. Its higher temperature makes it possible.


Air cooling reaches its limits at very high power densities. This is where liquid cooling is best suited. One example is CeramCool water cooling which benefits from the inertness of ceramics. No corrosion can cause trouble. The concept follows the same goal as for air cooled heat sinks: shortest (thermal) distance between heat source and heat drain. With ceramic it is feasible that the cooling water is only 2mm away from the LED heat slug. No other concept can do this in combination with the durable nature of ceramics. Multilateral electrical circuits can be printed directly on the ceramic without creating thermal barriers.


Since most of the applications where CeramCool is used are customer-specific solutions, it is essential that the performance can be proved before the first expensive prototypes are made. Intensive studies were made to build up simulation models. These simulation models have been verified against various tests and showed reliable correlations to test results. Based on this knowledge, new concepts or variations are easily evaluated.


The problem of retrofit lamps is mainly one of isolation. Any retrofit lamp has to be class II construction because you cannot guarantee an electrical earth. This means that any exposed metal part has to be isolated from mains wiring by double or reinforced insulation. Often retrofits with metal heat sinks do not comply, as it requires larger distances (like 6mm in air) or double layers of insulation which stop the heat sink from working well. The integrated electronic driver in a GU10 LED is so restricted for space that this is a very complex product. With a ceramic heat sink, even if the driver fails completely mains electricity is not conducted by the heat sink and the product is safe.

The new CeramCool GU10 LED spot works with any LED. Socket and reflector are made from a single material: a high-performance ceramic. Thus it has simple class II construction with safe insulation. A high voltage 4W LED only reaches a maximum temperature below 60°C, so both lifetime and light output are increased. In all CeramCool heat sinks the substrate becomes the heat sink. Here it acts as the lamp, or even the luminaire. The simplified design delivers extremely high reliability. In addition, the mount and reflector of GU10 LED spots are usually made of different materials. With this solution, far fewer materials are used and ceramics are exploited for their electrical insulation, good EMC, and high mechanical and chemical stability.


Ceramics can greatly improve both new and existing LED systems. Using a ceramic CeramCool submount, the PCB between LEDs and metallic heat sink can be replaced with ease, considerably reducing the total thermal resistance Rtt of the system. This offers important advantages such as very good thermal conductivity while delivering perfect electrical insulation and high temperature stability. Whether submount or complete circuit board—ceramic is absolutely corrosion resistant, which eliminates galvanic corrosion especially outside.


High power applications, especially for outdoor usage, gain as well from the features of CeramCool. A family of round heat sinks, which will meet the demands of different power levels, is under development. The concept combines cost efficient production with high flexibility of usage. It is going to be a “semi customised” product family.


To fully exploit the optimisation potential we have to look at metallisation possibilities as well. Ceramic can be coated directly with proven thick-layer technologies with its high adhesion force (WNi(Au), Ag, AgPd, Au, DCB, AMB) or thin film processes with its smooth surfaces (allowing precise light angles). A finish for better soldering can be obtained using electroless nickel or gold (immersion or cathodic deposition). The possibility of metallisation makes the whole surface of the heat sink useable as a circuit carrier which can be firmly packed with LEDs and drivers on customised circuit layouts, while providing reliable electrical insulation. The process can be simplified by bonding the chip directly onto the specially designed metallic surface.


Just a quick glance on how the assembling can be done. One potential assembler is BMK, a well-known German centre for electronic services. They do prototyping as well as series production. The heat sinks are mechanically set in a production framework and a solder cream is automatically applied, which is pressed through a template. In the next step LEDs and other components are mounted onto the heat sinks followed by subsequent processing in reflow ovens. A durable bond is established after cooling. The soldering points and position of the components are visually inspected, followed by a 100% functional test. Only now is the product complete and ready to be sent to the customer.

Sponsored Recommendations


To join the conversation, and become an exclusive member of Electronic Design, create an account today!