Research on copper-clad ceramic substrate for high-power light-emitting diodes

The power density of packaged LEDs has increased over the past few years, and the lifetime requirements of the modules have also increased. This has brought new requirements for improved substrate thermal conductivity and reliability to exceed standard FR4 or insulated metal substrates. Copper-clad ceramic (DCB) substrates offer lower thermal resistance and have been successfully used in high-power, high-voltage inverters and solid-state relays.

DCB process

DCB substrates are manufactured using a special hot-melt bonding method, where a copper sheet with a thin layer of copper oxide (oxidized during or before heat treatment) is bonded to Al2O3 ceramic and heated at a temperature of 1065°C to 1085°C (Figures 1 and 2).

Figure 1 Eutectic of oxygen and copper oxide

Figure 2 DCB process

The eutectic melt is bonded to the ceramic while the copper sheet remains solid. The excellent wettability of Al2O3 ceramic is based on the following reaction: CuO + Al2O3 = Cu The

following properties of Al2O4 enable DCB to replace traditional materials for multi-chip power modules.

Despite the relatively thick copper layer (0.3mm), the thermal expansion coefficient is still very low (7.2×10-6);
copper has high peel strength (>50N/cm);
due to the efficient heat dissipation of the thick copper sheet and the direct bonding of copper to the ceramic, the thermal resistance of the substrate is very low;
high mechanical and environmental stability.　
The cross-section of the substrate (Figure 3) shows the close contact between the aluminum oxide (24 W/mK) and the aluminum nitride substrate (180W/mK).

Figure 3 Aluminum oxide (left) and aluminum nitride cross section

Motivation

The dependence of expected catastrophic failure rate and junction temperature is a well-known and documented fact and can be predicted by the Arrhenius model. Higher junction temperature leads to lower lumen and thus shorter expected lifetime of the module. The

main way to produce high quality LED modules is to achieve lower junction temperature with better packaging. The lifetime of assembled LED modules and the price-life ratio can be extended by using the right combination of DCB substrate materials. Both AlN and thin Al2O3 (0.25mm) DCB substrates can provide economical and technical solutions to the above challenges.

When we consider a typical 5W high power LED package and a contact area (contact of the metal sheet supporting the substrate) of about 9mm2, it can be easily calculated according to the table 1 that even a standard Al2O3 ceramic substrate is sufficient, thus avoiding the cost increase caused by using special materials such as Si3N4 or AlN. Depending on the geometric conditions, the thermal resistance can be greatly reduced and is about 60% lower than that of conventional IMS substrates (75μm insulation thickness and 2.2W/mK heat transfer).

When looking at the power forecast development (Figure 4) we can see that by 2010, LED powers could reach up to 100W. We have to understand that this is not a completely new packaging problem. The requirements are the same as in traditional power electronics. Therefore, the same comparison results – the same solutions apply.

Figure 4 LED power development forecast

 Figure 5 shows the power density and operating temperature.

Figure 5 Power density and temperature

We refer to the development trend of packaged high-power LEDs from three major LED manufacturers (Figure 6). This pushes designers to design packages that can reduce thermal resistance.

Figure 6 Development trend of LED power and package thermal resistance

Based on these data, it seems that the further development is to reduce the thermal resistance between the junction and the metal sheet. For LEDs with power values ​​greater than 5W, the thermal resistance value of 4K/W can be achieved in the near future.

For die-to-substrate packages, the substrate itself is already the bottleneck of thermal management. This trend will force further improvements to the substrate.

Thermal characteristics of LED packages

Figure 7 shows the heat dissipation path of power LED packages. Let's not talk about heat sinks and focus on the situation where RJ-B=RJ-S+RS-B.

Figure 7 Thermal resistance simulation

For the packaged LEDs, we used a Lumileds Luxeon V (data from a public data sheet) for simulation and examined the thermal distribution results of the layout pattern for optimized heat dissipation.

The material was an aluminum-copper substrate 1 mm Al / 75 μm dielectric / 70 μm Cu (dielectric: 2,2 W / mK). The boundary condition was to fix the heat sink at 20 degrees. For the die-to-substrate simulation, we used a 2x2mm GaAs cube and the software used was IcePack.

Figure 8 Geometric model

Simulation results for packaged LEDs

The thermal resistance of the substrate material RB shows a dependence on the insulation thickness (Figure 9). The lowest static substrate thermal resistance measured for packaged LEDs is 0.3 K/W.

Figure 9 Simulated thermal resistance (including diffusion)

The temperature distribution inside the package shows that most of the thermal energy is distributed in the metal sheets inside the package.

Figure 10 Total thermal resistance from junction to substrate

Therefore, referring to the overall thermal resistance RJ-B, it is shown that the reduction of the thermal resistance of the substrate does not have a significant effect on the LED chip. Although the temperature has definitely decreased, the Rth drop is not very obvious. This is because the thermal resistance of the package itself is too high and even if the thermal resistance of the substrate is reduced, it does not affect the overall result.

Figure 11 Total simulated thermal resistance from junction to substrate

When the thermal resistance requirements of the package are further reduced, the situation of packaged LEDs needs to be re-evaluated.

The simulation results of CoB

compared with packaged LEDs show that the heat distribution on different substrates is significantly different when using the chip-to-substrate method.

Figure 12 CoB simulation results (200μm copper on 0.25mm A1203, dTmax=7.4℃)

Figure 13 CoB simulation results (75μm copper on IMS, dTmax=22.8℃, junction-to-substrate thermal resistance)

From the simulation results, it can be seen that the DCB substrate provides the possibility of obtaining low thermal resistance. The above 2.4 K/W is an ideal solution for the minimum thermal resistance of the interconnect layer. In real situations, the solder layer and/or adhesive layer increase the above value.

Different from the package type category, the die direct soldering substrate method can fix the chip tightly.

Heat dissipation and dynamic response

. Some short-life products such as flash lamps require three times more current than usual to drive the light-emitting diode. The high heat capacity characteristics of the DCB substrate will be beneficial for such products.

Figure 14 LED PWM brightness adjustment method

In addition, the more widely used method for adjusting the brightness of LEDs is pulse width modulation (PWM as shown in the figure). Using this method, the switching of the LED is a high-frequency specified working cycle, and the naked eye only feels that the light is dimming and cannot detect its cycle.

This process implies the need for thermal management. Packaged LEDs generally use heat sink metal sheets, and the die direct solder substrate package must provide sufficient thermal capacity to provide this operating mode. The

heat dissipation efficiency of thick copper sheets can further improve the heat dissipation performance, which can be measured and/or simulated by finite elements. The effect of thicker DCB copper sheets can be clearly seen from the simulation. It shows that the heat dissipation method is concentrically distributed around the chip.

Figure 15 Standard color map

Figure 16 Alumina substrate with thick copper sheet

This heat dissipation method increases the heat dissipation area. Some combinations of aluminum oxide substrates/and thick copper sheets can even compare to the thermal performance of aluminum nitride DBC.

In terms of values, the static thermal resistance decreases when compared with other substrate materials, and the dynamic thermal performance also shows the effect of increasing heat capacity.

Figure 17 Dynamic performance of CoB on DCB and IMS

Reliability considerations – Thermal expansion coefficients

are different from packaged LEDs. Thermo-mechanical compatibility needs to be considered for die-to-substrate packages. The different thermal expansion coefficients on both sides of any rigid interconnect layer (such as the solder layer) will cause stress on the interconnect layer. When the elasticity and rigidity of the material determine the reliability, more stress will inevitably reduce the reliability of the connection.

Due to the increase in the maximum allowed junction temperature, this situation becomes a reliability issue similar to power electronics. With an increase of 40°C, the different thermal expansion coefficients of copper and GaAs (16.5-5.5) will cause a length mismatch of about 440ppm between the chip and the substrate.

This is a well-known problem in the field of high-power electronics. There are three possible solutions:

1. Use matching materials to reduce the difference in thermal expansion coefficients

2. Reduce the overall temperature

3. Use non-rigid contact surface materials.

The thermal expansion coefficient of alumina DCB as a material is about 7.2 ppm/K, depending on its actual structure. Therefore, this material can provide a matching material between pure copper or aluminum heat sinks and semiconductor chips.

Figure 18 Effect of different thermal expansion rates on power

Improved DCB for power LED applications

The current DCB pitch values ​​are limited to 200-250μm. Since some LED chip manufacturers rely on flip chip technology, the direct chip-to-substrate package for DCB still needs further development. The first goal of changing the structural technology is to make the DCB insulation gap in the range of 100μm.

Further research and development is needed for precise geometric alignment of chip bonding.

Figure 19: Mounting marks on the copper surface

Conclusion

DCB substrates provide an attractive solution for future designs in the field of power LEDs. Since current packaged power LEDs have high thermal resistance, improvements in substrates cannot bring significant benefits. However, future LED packaging and multi-chip direct soldering substrate methods can benefit from the performance of DCB substrates.