by Jim Connett, on November 27, 2019
When we think about process capability systems in MES applications, we often imagine systems with a series of rules that permit material to run on a tool when a tool is capable (i.e., set up and configured to successfully process the material) and rules that prohibit or block material from processing on a tool when that tool is not capable (e.g., failed the last qualification, has the wrong physical chemistry installed, is at the wrong chamber temperature, etc.). And you would be correct—process capability modules must perform this ‘go / no-go’ evaluation before a lot is permitted to be processed. But did you know that process capabilities can potentially save you thousands of dollars in raw material costs and intangible costs associated with preventive maintenance activities?
Now…the scenario that follows is not applicable to all manufacturing environments. Our hope is that you can imagine a scenario that does occur in your manufacturing environment and extrapolate the important components to your specific use case.
Let’s imagine a chemical vapor deposition (CVD) process in the semiconductor industry. As the name implies, raw material from a source is set free to ‘fall’ on the wafer through the introduction of chemicals/gases into the chamber—the result is a measured amount of this raw material uniformly (hopefully!) deposited onto the surface of the device being produced.
In a vast majority of cases, the CVD process is considered dirty—raw material is not only deposited onto the device being produced but also onto the sides of the chamber wall. At some point, the thermal stress placed on the ever-growing layer of material on the chamber wall may result in a chamber environment where flaking occurs, which, obviously, can impact the yield of the wafer due to die fallout from contamination/particles.
Engineering and Maintenance groups use historical chamber performance to schedule preventive maintenance procedures well before flaking and chamber contamination may occur. Preventive maintenance involves opening the chamber, cleaning the chamber, removing the old raw-material target and installing a new one, downtime for source alignment and qualification, and then the chamber is released back to production. This downtime lowers tool utilization, increases material costs, and adds to labor hours. It also invites the possibility of problems during the qualification process, when the chamber is brought to atmosphere for the maintenance procedure. For those who have done their fair share of kit and target changes on deposition tools, you will be acutely aware of the challenges described above!
But, not all processes are dirty. In fact, some processes can have the opposite effect—rather than depositing another layer of raw material into the chamber, the chemical composition of some recipes may serve to protect the chamber wall from flaking. If you can identify chemistries that offer this cleaning side effect, then you can use process capabilities to save money. A LOT of money.
Most process capability modules in an MES contain a counting component that can decrement a counter each time a lot (or even a wafer) processes on a tool. When the counter reaches zero, the process is no longer capable, and the MES can (or may) log the chamber down for a predetermined maintenance task. If you can identify your dirty and clean recipes, you can set up an algorithm in process capabilities to actually INCREASE (rather than decrease) the wafer count when a clean recipe runs. For example, let’s say a new kit and target has a life of 500 wafers. A dirty recipe might decrement this count by 5 for each wafer processed. But a clean recipe could increase the count by 2 for each wafer processed. In this way, the “race to zero” is non-linear and is based on the types of recipes run during a given time period.
Identifying the dirty and clean recipes, the level of ‘clean’ that occurs, and the maximum target life are all matters of tests and trials by Engineering. Armed with these data, you can take control of the counters, decrement where necessary, increment where possible, and avoid unnecessary and costly maintenance procedures. As an added layer of automation, SYSTEMA’s EDDi (Event-Driven Dispatcher application) can be made aware of the current wafer count of the chamber through inventory controls and can increase the visibility of material whose recipes will serve to clean/seal the chamber.
I’ve witnessed firsthand the benefits of this type of process capability algorithm, which—in rare and specific cases—reduced the number of chamber cleans and downtime maintenance events by a factor of 3! Of course, your results may vary depending on chemical processes, target material, and other factors, but if you could possibly reduce preventive maintenance and downtime events for a chamber by just 20%, it would be worth it!
Never miss a blog post again!