May 01, 2018
As a design engineer for power electronics systems, you require your selected power module to fulfill its electrical function as described in its data sheet and you expect this module to be reliable meaning that it should operate under given conditions, in a defined period of time and within an acceptable failure rate. While solder fatigue and wire lift offs have been the main limiting factors for the lifetime of conventional modules, new technologies for assembly and packaging of semiconductors have emerged and new module generations achieve much longer lifetime. Looking forward, metallized ceramic substrates will remain key components in the modules to guarantee their functionality and reliability.
Understand the failure mechanisms
Metal and ceramic have different coefficient of thermal expansion (CTE), meaning that these two materials do not expand and shrink at the same pace when they are exposed to temperature variations. Already during the joining process, mechanical stress will be captured in the substrates as copper and ceramic are joined at a very high temperature up to typically 800°C for active metal brazing (AMB) or 1,060°C for direct bonded copper (DBC) substrates. Then, during operation, substrates will be exposed to significant temperature swings because the ambient temperature may vary and the junction temperature of the semiconductor devices will change as they switch on and off the load. This will cause alternating compression and tensile stress in the substrates. This stress is not homogeneously distributed over the entire surface but concentrated on the corner and edges of the copper pads and ultimately can cause edge cracks in the ceramic if there were no other factors limiting module reliability.
How to test and assess the lifetime of metallized ceramic substrates?
Thermal cycling is the most common test method used for assessment of the reliability of metallized ceramic substrates. It is an accelerated test method and as such cannot give any information about the lifetime of the substrates in a real application case but provides a good comparison between various substrates. While almost every manufacturer of substrates will use the same test method, you should pay attention to the temperature profile as different test conditions may lead to different acceleration factors and hence significantly influence the test results. In most cases, substrates will be tested between -40°C and +125°C but one can find -55°C up to 150°C as recommended in some norms for military applications.
Various test methods like dye penetration test and measurement of resonance frequency can be applied to reveal the number and length of cracks but they are either destructive or not suitable for any test vehicle. Scanning acoustic microscopy is a well-known non-destructive test method for failure analysis. It employs ultrasound to penetrate solid materials and make visible image of defects such as cracks, delamination and voids. The contrast on the image highlights the location of a defect. By means of some suitable software for image analysis and with a very good and consistent resolution of the image, the progress of degradation can be measured as a percentage of the total metallization area.
Not every form of degradation in the substrate is critical for your application. However, a minimum requirement should be that the heat dissipation from the chips is not impacted. This very much depends on the position of the chips relative to the degradation. A visual assessment of the images is usually not sufficient to determine the end of life of the substrates. It should be combined with a measurement of the thermal impedance to fully evaluate the criticality of any defect, this is a very good tool to compare and qualify new materials or suppliers.
Select the best material combination for your application
Highly thermally conductive aluminum nitride (AlN) is the best choice for the purpose of heat dissipation but unfortunately it shows poor mechanical properties and as a consequence AlN DBC substrates are prone to fail after few temperature cycles. Alumina (Al2O3) DBC substrates are more robust and they can survive almost twice as many cycles as AlN DBC substrates even though their typical thickness is cut by half. Once doped and reinforced with zirconia, Al2O3 ceramics have improved mechanical properties and reliability is increased again by a factor of two in the same test conditions. Silicon nitride (Si3N4) ceramics are even tougher and, once joined with copper by active metal brazing, Si3N4 substrates can survive more than 1000 cycles.
The source of supply and quality of the ceramic has a tremendous influence on the reliability of substrates, too. To secure business continuity, multiple sources of supply are necessary. Rogers has been investigating many sources of supply and only qualified the few that deliver good and consistent quality to guarantee the best possible reliability.
The lifetime of substrates also depends on the thickness ratio between copper and ceramic. Generally speaking, a thick and strong ceramic substrate metallized with a thin copper layer is expected to withstand many more cycles than a thin and fragile ceramic substrate with a thick copper metallization. Those thicknesses are usually designed to achieve the required insulation, heat dissipation and current density. But the best electrical and thermal design is not likely to be the most reliable. A trade-off will be required and your decision depends on the targeted application as the requirements and mission profiles vary.
Design for reliability
The most reliable materials rarely are the cheapest materials. Luckily there are other options to increase the reliability by design and without adding significant costs to the solution. curamik® substrates have a proprietary dimple design for the purpose of thermal stress relief. As dimples reduce the copper thickness on the edge of the copper pads, there is less Cu material applying compressive forces on the ceramic, the thermo-mechanical stress in these critical areas decreases and lifetime is significantly improved.
Another important design consideration is to equally balance the stress on both sides of the ceramic substrate. We recommend having the same copper pattern and same copper thickness on both sides. As a plain copper surface is often required on the back side, copper should be removed on the component side only when necessary to fulfil the requirements regarding electrical isolation. Particular attention should be given to the copper free perimeters on both sides as cracks may easily appear if they differ by more than 500µm.
Also, the design of the corners of copper pads impacts the reliability of the substrates. Two identical designs but with sharp respectively rounded corners will not achieve the same number of temperature cycles. Surprisingly, substrates with sharp corners survive more cycles than the same substrates with rounded corners.
Today, the lifetime of power modules is not limited by the substrate and there are many ways to design robust modules with existing substrate technologies. A good knowledge of your application and its mission profile is necessary to select the best possible material at the lowest costs. And some design tweaks can help to work around some potential reliability issue and make the best usage of the available substrate surface. Rogers Power Electronics Solutions (PES) offers the largest choice of substrates with its curamik brand. But as technology keeps on evolving and in an effort to steadily improve the performance and reliability of our products, our team is consistently looking for new materials and new processes to innovate and tackle the upcoming challenges.
Do you have any design question or require some assistance for the selection of a suitable substrate for your application? Rogers PES’ experts are available to help. Please contact us today.