Wafer to Wafer Control

Wafer-to-wafer control is a run-to-run process control strategy that measures and adjusts process parameters individually for each wafer rather than only at the lot level, specifically to catch and correct within-lot variation — such as slot-position effects in a batch furnace — that lot-to-lot run-to-run control, averaging across every wafer in a lot, simply cannot see. As single-wafer process tools like modern etch, CVD, and PVD systems have become the dominant equipment architecture at advanced nodes, wafer-to-wafer control has shifted from a specialized refinement into the natural, expected control granularity for an increasing share of fab process steps.

Measuring the critical output parameter on every individual wafer, rather than relying on a lot-average measurement taken from only a sampled subset, is what makes wafer-to-wafer control meaningfully different from ordinary lot-to-lot run-to-run control in the first place — it trades reduced measurement throughput for the ability to see and react to variation that a lot average would simply smooth away. A lot can show a perfectly acceptable average measurement while individual wafers inside that same lot vary considerably around that average, and it's exactly that individual-wafer variation — invisible at the lot-average level — that wafer-to-wafer control is specifically designed to detect and correct.

Exponentially weighted moving average or model-based correction algorithms update the process recipe after each individual wafer's measurement comes back, applying a calculated adjustment intended to pull the next wafer's expected output back toward target — a feedback loop operating at wafer-level cadence rather than the considerably slower lot-level cadence that lot-to-lot control operates at. This faster correction cadence is exactly what gives wafer-to-wafer control its responsiveness advantage: a drift that starts partway through a lot gets detected and corrected within that same lot, rather than only showing up in the lot average after the entire lot has already finished processing.

Slot-dependent effects — systematic variation tied to a wafer's specific physical position within a batch furnace or similar multi-wafer processing chamber — are a textbook example of exactly the kind of within-lot variation that wafer-to-wafer control exists specifically to compensate for, since a batch tool's different slot positions genuinely do experience somewhat different process conditions even when nominally running the identical recipe. Without wafer-to-wafer correction accounting for a wafer's known slot position, that position-dependent variation would show up as unexplained within-lot spread; with it, the control algorithm can apply a position-specific correction that directly counteracts the known, repeatable slot-position effect.

Controller state threaded per-chamber across a multi-chamber tool, rather than treated as a single shared value across every chamber, matters because different chambers on the same tool genuinely drift independently from each other, and a wafer-to-wafer correction calculated from one chamber's recent history shouldn't be blindly applied to a different chamber that may be drifting in a completely different direction. This per-chamber state management is exactly why wafer-to-wafer control implementations have to track separate controller history for every individual chamber on a multi-chamber tool rather than maintaining one global correction value — doing otherwise would risk applying a correction calculated for one chamber's specific drift pattern to a different chamber it doesn't actually describe.

Single-wafer process tools — the dominant architecture for modern etch, CVD, and PVD equipment — make wafer-to-wafer control the natural, inherent control granularity for these tools, since each wafer is already processed individually rather than as part of a shared batch, and advanced-node process specifications increasingly require exactly the kind of tight within-lot uniformity that only wafer-level control granularity can actually deliver. This alignment between single-wafer tool architecture and wafer-to-wafer control methodology is exactly why wafer-to-wafer control has moved from a specialized, occasionally-used refinement toward becoming the expected default control strategy across an increasing share of advanced-node process steps — the tool architecture and the control strategy were essentially built for each other.