Why TaC Coating Spalls on AIXTRON Susceptors: Causes & Fixes

If you have been using AIXTRON reactors for some time, you will have already experienced a susceptor failure before expected due to the Tac coating peeling or flaking off. It's very frustrating, particularly when it's all well and good at first. The coating should protect the graphite and maintain the stability of the process, but if there is any problem in coating or operation, it is prone to build up to cause spalling. This guide will explain why this is occurring in real production conditions, and how you can decrease the risk level, with simple explanations and tips that you can immediately use.

Thermal Expansion Mismatch Between CVD TaC Coating and High-Purity Graphite Susceptor

Applicable by CVD, TaC is dense and has a lower thermal expansion than high-purity graphite. Graphite, however, expands more at high temperatures and contracts more at lower temperatures. A susceptor is heated to very high temperature during a growth cycle followed by a cooling. The CVD TaC coating Guide Ring layer is pulled and pushed at the end of each cycle. On the repeated cycles, small cracks start to occur. This is usually only visible after a number of runs in order to get to know the flow of production. One team that was growing GaN noticed small flakes starting to form around the edge of their susceptor around 20 cycles. The edge was beginning to peel off, but the center was still ok. The reason is that the temperature gradient is higher on the edges, resulting in a larger mismatch in the expansion of the edges. Another thing that many people don't consider is coating thickness. The thicker the TaC layer, the more the internal stress. As the graphite beneath expands, the thicker coating is less able to bend along with the graphite, and thus is prone to cracking. Narrow, controlled coatings are generally more durable since they have the ability to absorb strain. One of the first steps in minimizing this issue is to determine the amount of time it takes for the reactor to warm up or cool down. Stress is higher in regions of rapid temperature changes.The stress is higher in areas of rapid temperature changes. Ramp rates can be the difference if they are slower. In addition, coordinate coating with suppliers to ensure the coating process is consistent with the graphite grade. Graphite is not uniform in nature and slight variations in density or grain size can result in varying expansion.

Residual Stress Accumulation in AIXTRON G5/G10 Thermal Cycling Processes

Another hidden cause for the failure of TaC coatings is residual stress, particularly in tools such as AIXTRON G5 and G10, where thermal cycling is a daily routine. Stress is already trapped during the CVD materials process even though the coating may appear smooth and well-bonded at the beginning. This stress does not get dissipated each time the susceptor is subjected to the heat up and cool down cycles. It is postulated that CVD TaC is formed during the coating process at high temperature and then cooled to room temperature. It is by itself very stressful, as the coating and graphite cool off in different ways. After the first cycle a little more strain is introduced in each successive cycle. It's similar to bending a thin metal strip back and forth, there's no damage until repeated many times, then it cracks. This is frequently manifested in production runs on G5 and G10 systems, as delayed failure. A susceptor can perform well for dozens of cycles, and then begin to flake. One operator reported that their parts were visually checked after every run, and that spalling was seen actually happen overnight after about 30 cycles. It was actually the stress increasing over time until it reached the limit of endurance. That's where temperature is a significant factor. In high throughput applications, fast ramp-up and aggressive cooling are frequent occurrences, but also lead to stress accumulation. The outer area of the susceptor is subjected to the highest temperature cycles since it has the highest heating and cooling rates. This unequal activity brings more stress to the surface. The uniformity of the coating is also important. Any areas that are somewhat thicker or denser, will hold on to more stress. These areas lose strength with age. After a crack has been initiated, it will rapidly propagate throughout the coating. The first step to dealing with this is to examine your thermal recipe. Ramp rates can be slowed down a bit to decrease the stress per cycle. This can slightly increase the time per run, but can be considerable time saved by extending susceptor life. Also, see if your process has unneeded temperature changes. It helps if cycles are kept as stable as possible. A good one to take is to use stress-relief techniques with the coating suppliers. This may involve adjusting deposition temperature and/or introducing a controlled cooling process after coating. In addition, some users perform lower stress “conditioning cycles” prior to full production to help condition the coating. Don't focus on failures, look for patterns. When spalling occurs at the same number of cycles, this is a definite indication of residual stress accumulation. This will enable you to anticipate when to change or replace parts before impacting your process.

Interface Contamination: The Hidden Risk Behind TaC Coating Peeling

One of those is the presence of interface contamination, an otherwise nice TaC coating which can ruin its performance. Even if the coating process is conducted properly there is a potential for contamination occurring between the graphite and the TaC layer, reducing the bond from the outset. Coating may appear to be fine upon deposition but it is not very adhesive to the surface. This usually begins at the time of handling and/or preparing graphite. High purity graphite will readily get contaminated by dust and/or oils from hands, machining residue or even the moisture from air. These are not fully cleaned before undergoing CVD coating and they create thin barriers at the coating/substrate interface. The TaC layer is then grown on top of that barrier, instead of directly to the graphite. It can be observed as early peeling in actual production. One team noticed that the flaking of a new batch of susceptors started after a couple cycles, early on. They investigated this and discovered the cleaning process prior to the coating had been reduced to save time. The little amount of residual change produced fine particles that resulted in substandard bond strength area-wise. One other difficult factor is contamination is not necessarily apparent. The surface may appear to be clean but residues may be present that impact bonding. This is where surface preparation is critical. Different batches can have different coating life performances, even from the same batch, due to differences in cleaning procedures. To decrease such risk, pay attention to good cleaning control. This means the correct use of ultrasonic cleaning, high temperature bake-out to remove moisture and not touching the surface of the graphite with your hands. When coating, be sure to use clean handling tools and keep parts in controlled areas prior to coating. Sure, it's an extra contribution, but it's much better than solving much bigger issues down the road. It also facilitates the process of checking the quality of the incoming graphite. Porosity variations will cause the trapping of contaminants within the material that will release it during coating and will reduce the bond between coating and substrate. Cooperation with suppliers who ensure product quality will make a great difference.

Coating Thickness & Deposition Parameters in CVD TaC Coating Stability

The structure of the coating is also influenced by the deposition parameters. These factors influence both the density and the disposition of the grains – temperature, flow of gases, and the rate of deposition. If the process is too fast, the coating may not be as uniform or collect defects. These minuscule weaknesses become weak points later. However, a slower and more stable process tends to produce a denser and more uniform layer that will be more resistant to stress. The uniformity throughout the susceptor is also critical. If there are some areas thicker than others, then the stress will not be evenly distributed. The places with these uneven spots are the first places where spalling starts. This is particularly common in areas of edges or complex shapes, where the gas flow could change during deposition. First, make a realistic thickness target, rather than aiming for the highest possible thickness, for increased stability. Coordinate closely with the coating supplier to coordinate thickness to your process temperature and cycle count. In addition, keep an eye on the consistency of the deposition—over time, the quality of the coating may vary slightly according to a change in temperature or gas flow. Wherever possible, look at cross sections of coated components. This will reveal to you whether there are any defects or if the layer is uniform. If you take care of these details early, you can prevent some of the failures you may experience after multiple production cycles.

Preventive Maintenance & AIXTRON Susceptor Cleaning Best Practices

One of the most effective ways to slow down the failure of TaC coatings is preventative maintenance but, in production lines, it is often just one of many low priorities. The reality is that, when spalling begins, it doesn't normally stop on its own. Proper cleaning and maintenance at the appropriate time can yield much longer lives for susceptors than attempting to correct issues post-facto. Small residues of process by-products are left behind after each run of the susceptors in AIXTRON. These may accumulate over time on the surface of the TaC. Initially, it may simply be a minor discoloration or a washed out look. This accumulation in the time alters the surface heat distribution leading to non-uniform heating. This non-uniformity contributes to stress in the coating and accelerates cracking. For one production line operating with GaN epitaxy, operators realized that when they cleaned their susceptors on a set schedule, they seemingly lasted longer than those of the operators that cleaned them only when performance fell. It wasn't so much the way the cleaning was done, as the consistency. Irregular cleaning caused deposits to harden, making them more difficult to remove and likely to cause damage to the coating as a result of cleaning.