Browse Prior Art Database

Method for fault rejection in semiconductor manufacturing run-to-run process control

IP.com Disclosure Number: IPCOM000021716D
Publication Date: 2004-Feb-04
Document File: 3 page(s) / 61K

Publishing Venue

The IP.com Prior Art Database

Abstract

Disclosed is a method for fault rejection in semiconductor manufacturing run-to-run process control. Benefits include improved functionality and improved throughput.

This text was extracted from a Microsoft Word document.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 73% of the total text.

Method for fault rejection in semiconductor manufacturing run-to-run process control

Disclosed is a method for fault rejection in semiconductor manufacturing run-to-run process control. Benefits include improved functionality and improved throughput.

Background

         The lithographic processing feedback control signal is conventionally created using measurements from metrology equipment in semiconductor manufacturing. When the measurement contains error, the control signal cannot be correctly created. The equipment must be shut down, and the process cannot be correctly adjusted and controlled.

General description

         The disclosed method uses estimated output data to control the run-to-run lithographic process. When a problem occurs, the control signals are created and processing continues.

Advantages

         The disclosed method provides advantages, including:

•         Improved functionality due to enabling processing to continue when a fault is detected

•         Improved functionality due to creating lithographic process control signals not connected to measurements from metrology equipment

•         Improved throughput due to enabling processing to continue when a fault is detected

Detailed description

         The disclosed method is a model-based procedure to estimate the true processing output and compute feedback control signals (see Figure 1).

         The model can be built in the following form:

                  (2)

         The equation contains the following values:

•         y(k) is the output measured at step k.

•         u(k) is the control signal computed at step k.

•         a and b are constant coefficients.

•         w(k) is a combination of process noise and measurement noise.

•         ŵ(k) is the noise estimate.

•         ŷ(k) is the output estimate.

Step 1. Set up a prior estimate of output at k with previous estimation of ŷ(k-1):

                  (1)

Step 2. Estimate the noise at k with measured value y(k) and previous estimation ŵ(k-1):

                                           (2)

         The value of λ is from 0 to 1. When the value is greater than 0.5, the impact from historical noise is emphasized. When the value is less than 0.5, the impact from the difference of model and measurement is emphasized.

Step 3. Estimate the output at k with measured y(k) and estimated noise ŵ(k):

                  (3)

Step 4. Compute the control signal at k+1:

                  (4)

         The disclosed method compares the preidentified process model...