Process variability is a fact of life in semiconductor manufacture. Whether it be tool-to-tool, chamber-to-chamber, or as the result of the ongoing use of a tool in production, detecting and compensating for process variability is essential to maintaining product yield and ramping production volumes.

For processes that depend on the interaction between gases to make good product, measuring changes in process chemistry provides key insights into the real-time and long-term performance of the tool. Comparing the compositional “fingerprints” of chambers running identical processes can speed develop- ment of recipes that match performance of multiple chambers and tools.

Fingerprinting capability provided by Verionix sensors helps quantify the differences in tool design, manu- facture and condition of seemingly identical tools. By flagging when a process transitions from good to bad, fingerprinting enables detection of random or long-term component failure, required maintenance or changes in gas purity or supply used in the process.
 

Problem

  • Chamber A and Chamber B on two separate CVD tools are configured to perform the identical process step, but a higher percentage of the thin films created by Chamber B are separating from the wafer surface in later process steps, impacting product yield.

Process Environment and Setup

  • Verionix Vx-2000 series sensors mounted on the forelines of Chambers A and B are configured to detect compositional changes in multiple gas species as process operates.
  • Gas compositional data is collected and compared in real-time to "fingerprints" of process conditions that are known to produce good product.
  • Sensors signal respective tool controllers if process chemistries deviate from established known good conditions.

Critical Verionix Data

  • Sensor in Chamber B detects an intermittent downward drift of the Gas A signal level, indicating instability in composition relative to the same species in Chamber A.

Results

  • A clogged gas vaporizing component with intermittent behavior is identified as the cause of failure in Chamber B, resulting in its removal from service before additional wafer scrap can occur.
  • Repairs to Chamber B are validated using the Verionix sensor prior to its return to service.
  • Performance of processes in Chamber A and Chamber B are more closely matched.

Home  |  Privacy Policy  |  Legal Statement