a wafer metrology consortium under MAGNET

WP2 - Sampling Optimization

WP2 - Sampling Optimization

The challenge of the metrology companies is to maintain roughly the same per-wafer measurement time for 450mm wafers as for 300mm wafers. Since the only reason for 450mm conversion is reducing manufacturing cost, if wafer measurement time will increase with the wafer size – the ‘cost reduction’ challenge will not be achieved.

One of the obvious solutions that come to mind is simply trying to speed-up measurement time. This solution however almost always requires sacrificing other parameters of the measurement, such as repeatability and accuracy, in exchange for speed. However, previous trends in this industry showed (as is obvious) that in order to have good metrology, repeatability and accuracy must improve, not worsen.
One can think of additional methods for maintaining the current time for measurement; Optimization of the sampling plan (explained below) is the most appealing one, as no additional investment in hardware is needed.

Somewhere along the process chain, some of the wafers may be delivered temporarily to a metrology unit. This unit measures the wafer and sends the data to the host. Usually, this unit has no function in wafer processing. It is there merely to ensure that the process is working correctly.
Today the sampling plans (e.g., how many measurements represent a wafer and how many wafers represent a lot) are fixed, and are set off-line according to the worst case process step sensitivity.
By exploiting the fact that the metrology units already collect data by themselves, sometimes a great deal more than is needed by the host, they may have the ability on their own to decide how frequently sampling should be performed, by introducing a sampling optimizer module inside the metrology unit itself.



The goal of this work group is to demonstrate, at the end of the 3 years, is at least one algorithm that will work as a statistical engine within a metrology tool and will enable an in-tool smart sampling mechanism which will improve overall throughput.