The rapid pace of semiconductor evolution places huge demands on the companies making the equipment used to build integrated circuits. With each new silicon node comes a host of new requirements, and it’s easy to imagine that older equipment rapidly becomes obsolete. While that can be the case, well-designed equipment can often be adapted to new tasks.
In order to see how this works, we can look at one example: an etch machine known as the Versys® Kiyo45™. Released 10-15 years ago – an eternity in the semiconductor world – it was designed to perform reactive-ion etching (RIE) (also known simply as plasma etch) for silicon wafers. Its role was to remove silicon quickly enough to meet the tough economic requirements of high-volume chip production. It came onto the scene when 200-mm diameter wafers were still in production, but 300-mm diameter wafers were emerging, so it had the capability of running both wafer sizes.
For years, the machines have been performing well, although, as requirements tightened, newer versions took over for the new, more aggressive silicon nodes. Even so, the older version remained in service, providing a continued supply of chips built on mature technology nodes.
RIE is very effective at removing material, but, at the end of the process, the surface of the wafer has been altered such that a thin top layer is no longer crystalline. With that in mind, a customer came with a new problem: they needed to etch wafers used in the fabrication of high electron mobility transistors, (HEMTs) for new power device applications; and they needed to be able to do so with much less damage than conventional RIE would create. But there was a catch: these wafers weren’t just made of silicon; they were made of silicon but with GaN or AlGaN epitaxy on 200 mm wafers.
There was no equipment at the time that could meet this need. Newer machines were focused exclusively on 300 mm wafers, and non-silicon wafers were considered somewhat exotic; they weren’t part of the mainstream. So, there was no obvious solution for these GaN wafers.
We took a look at the Kiyo45, the last of the etch machines to work with 200-mm wafers, to see if there might be a way to adapt it to these new wafers. The original designers of the equipment would never have foreseen this particular use; the question was, had they designed in enough flexibility to handle this task?
Moving to a new material can mean new etch gases, different timing, different power levels, and perhaps even a change in the sequence of how things work. How broadly those can be changed depends partly on the physical design of the machine: can it withstand different gases, and are all the physical interconnects compatible? It also depends on how the process recipes are implemented. The more that is done by software, and the more points of control that are available, the better chance there is of making the machine do something completely new.
We found that we could implement this new GaN recipe on the older machine by implementing low damage processes that utilized optimized end-point detection resulting in a smooth GaN etched surface.
So, the Kiyo45, having once been the “older” machine, was now performing on new, leading-edge wafers for a wide range of new markets. In addition to power electronics, it also opened up opportunities for image sensors, MEMS, RF circuits, and other new areas. The specific roles it played in these markets were in etching gates, creating trenches and spacers, providing etch-back, opening hard masks, and a variety of other tasks.
It’s an example of ongoing innovation, of making use of what exists first before having to create something completely new. It’s easy to fall into the expectation that older equipment must be on its way out. If looked at with a fresh, creative eye, there may be whole new opportunities that will give old equipment a new role.