As published in the Atheneum Global Partners Global Quarterly Newsletter, Dec 2015
Trends in Semiconductor Manufacturing Process Equipment
1. What crucial trends are driving the industry?
Many 150 and 200 mm wafer fabs are using equipment that is 20 years old, or more. Devices being manufactured on these tools were not even conceived when these process tools were developed, and it was always thought that the lifetime of the tools would be on the order of ten years.
While many factories have changed owners and many others have closed, the factories running these older toolsets tend to be at capacity. The equipment, while it may have been resold and often rebuilt in the process, is still being used to perform in many cases much more difficult tasks then when designed. But the use of fully depreciated equipment allows production at costs far below newer 300 mm factories, even if they are not capable of the smaller geometries that the 300 mm tools can run.
Both types of factories have their purposes. The large microprocessor and memory factories churn out high volumes of devices and continue to drive down the costs of state-of-the-art products through productivity enhancements and new designs that pack more functionality onto a given chip area. 200 mm tools can’t compete without the geometry capability and economy of scale that 300 mm tools bring to these markets. And these newest factories will remain at 300 mm for the foreseeable future; in essence, 450 mm may ‘never’ happen. However, while many continue to concentrate on the flagship micro and memory products with the smallest geometries, there are many more supporting devices surrounding these that are required to build most electronic products. And at the same time as 150 and 200 mm factories are running at capacity and looking to add incremental wafer starts though utilization improvements, many 300 mm factories are in cost-cutting mode and reducing output.
2. What market segments will experience the most growth and why?
As the mobile device markets have eclipsed the computer as the largest consumer of chips, and the number of all types of wireless products has grown prolifically, the product mix has further tilted to the semiconductors produced on 150 and 200 mm tools. In particular, to supply the growing wireless device market, III-IV device manufacture, mostly GaAs, has been driven to nearly full capacity and the GaAs device makers are cash rich and acquiring additional equipment as fast as the aftermarket can supply it. And to date, there is no supply of GaAs wafers larger than 150 mm, so it will require more creative methods to provide increased wafer size, such as epitaxial layers on silicon, but until that is readily available, the III-IV market will stay at 150 mm.
In addition, GaAs processing requires some types of etching and deposition not normally used in silicon processing. GaAs devices are often using MEMS technologies to build resonators and filters needed for the frequencies used in wireless communication, and in an effort to provide the best functionality with the smallest device area, are marrying silicon devices with GaAs devices though a sandwich of the two devices plus an intermediate through-hole device to provide the necessary connections.
These wireless device manufacturers are even building new factories and resurrecting old ones to meet current demand. This is mostly driven by the fact that the geometry demands of these devices are not difficult to meet, and since the die sizes are relatively small, particle constraints are not as strict.
Since the devices geometries for most wireless devices are relatively simpler to manufacture, many factories already are capable of making them, and with the addition of a few families of toolsets, they can be upgraded to be able to manage the unique processes required. They are using largely older, fully depreciated equipment, so the cost to manufacture is lower. And since any expansions or factory builds have to provide similar wafer costing, they will also be using predominantly older, used equipment.
3. What are the key challenges?
As noted earlier, this equipment was designed and much of it built more than 20 years ago. There are issues that need to be addressed to maintain productivity and sustainability of that older equipment. A semiconductor manufacturing tool used for etching or deposition of thin films is really a frame with plasma chambers attached. There are robots needed to move wafers in and out of the chambers, pumping systems to provide vacuum to evacuate load locks and remove process gasses and effluents, and power delivery systems needed to energize the gasses in the plasma chamber to perform the etching or deposition required. All of this is coordinated and controlled by an on-board computer system.
What has become difficult is maintaining these older tools in a manner that affords uptimes at or better than they achieved when they were new. There are several issues with this, and several solutions have been applied.
The original tool manufacturers are really not interested in these tools remaining in production. The larger OEM’s get some 60% of revenues from new equipment sales; and they concentrate their efforts in order to maintain market share in competition for the volume tool buys by new 300 mm fab startups. They do still supply spare parts for the older tools, and often at premium prices. But they have largely stopped the progression of upgrades, and yet wish to maintain as much control as possible over the end user’s ability to get parts and support as possible. OEM’s often engage in strong-arm tactics to achieve these ends, that would be deemed illegal if applied to a more mainstream industry such as automobiles, under restraint-of-trade laws.
This leaves a significant gap in which the aftermarket rebuilders are flourishing. These companies purchase used equipment in nearly any state of disrepair and rebuild it, returning it to an operating condition through purchase of off-the-shelf items and obtaining proprietary parts from the OEM or purchasing them from the used market. In cases where neither of these routes are available, they may resort to reverse-engineering to obtain a reliable supply of parts.
Rebuilders normally sell the their equipment for a small fraction of the price charged by the OEM for new equipment purchases, which fits the pricing models of the 150/200 mm factories for purchasing additional equipment. For example, a rebuilder may sell a complete four-chamber tool, with installation and warranty, for less money than the OEM charges for one of their new process chambers that might even go onto the same equipment.
The OEM’s have been dragged back into the rebuild market, sometimes at the request of their larger end-user customers. Some OEM’s have a small section of their field service group devoted to rebuilding. In other cases they have hired one of the rebuilders – in other words, their former competitor – to do this work, as their internal resources are normally fully devoted to the new tool sales.
Another area where the rebuilders are taking the lead is in upgrades to the various subsystems. They have had to devote time to finding credible replacements for aging systems that the OEM’s have been able to maintain a tight channel from their original product manufacturers where the (OPM’s) have maintained exclusivity by enforcing purchasing rules that have kept rebuilders from purchasing the exact units needed. In the cases where they can be purchased, they may be at a high mark-up giving prices equivalent to OEM-to-end user pricing. This ensures that the OPM units sell for a price that may be much higher than comparable products of a different type.
Select, targeted upgrades are key to maintaining the useful lifetime of the tool, and to increasing the capability to run beyond the originally-designed process regime, commonly known as the Best Known Method (BKM). And they are often needed simply to offset the scarcity of parts which have gone obsolete, and become either unobtainable or only at very high prices with unreliable supply chains. As processes have been forced out toward the extremes of performance, they tend to lose the ability to be run repeatably across a toolset. This ability is critical to device manufacturers in order to enable high-volume production of their next generation products. And this is also necessary as the parts of the original systems may have been on a last-time buy status for some time, and there are serious challenges to maintaining a reliable supply of devices that are no longer manufactured.
Data collection and automatic fault interdiction have also become an integral part in maintaining the functionality and uptime of older equipment running toward the edges of process performance. They have begun to replace the simple tool alarms that are based on single parameter outside of a fairly arbitrary standard, for example, a warning limit of 5% of gas flow variation from setpoint, to a recipe with limits that can be learned by the fault detection system to determine normal variation and expected trending, and alarm on any situation that does not fit this normal model. While even these systems can’t reliably predict the outcome of a process in the chamber, they are an important step above the standard tool alarms, of which as many are in place to help ensure that the tool is not in the process of damaging itself as they are to ensure that the product is not getting scrapped by out-of-tolerance processing.
In summary, the bulk of semiconductor manufacturing is done using older equipment, maintained by a combination of OEM and aftermarket companies, and this equipment has been pushed far beyond its expected lifetime. The inventiveness required to keep this equipment able to maintain processing integrity with good uptime is immense, and results vary a lot across the industry. It is largely to the credit of wafer fab staffs and small support companies comprised of ex-fab and ex-OEM equipment technicians that this industry is able to run as well as it does in the face of ever-increasing demands.