AVS 46th International Symposium
    Manufacturing Science and Technology Group Tuesday Sessions
       Session MS-TuA

Invited Paper MS-TuA4
Yield Improvement Through Multizone Uniformity Control of a CMP Process Utilizing a Pre and Post-Measurement Strategy

Tuesday, October 26, 1999, 3:00 pm, Room 611

Session: Interconnect and Integration
Presenter: J. Moyne, University of Michigan
Authors: J. Moyne, University of Michigan
K. Khan, University of Michigan
J. Colt, IBM
J. Chapple-Sokol, IBM
R. Nadeau, IBM
P. Smith, IBM
T. Parikh, SEMATECH
Correspondent: Click to Email
Achieving good uniformity process control in Chemical Mechanical Polishing (CMP) requires a representative uniformity metric and strong models relating this metric to process tunable inputs. Previous efforts in CMP uniformity control have yielded acceptable results utilizing a Center-to-Edge (CTE) first order non-uniformity metric. Closer analysis of post CMP process non-uniformity, however, reveals significant higher order non-uniformity components such as the center "dimple" and outer "doughnut" regions. These non-uniformity characteristics are due in large part to upstream CVD processing. Utilizing a multi-zone approach to uniformity modeling, a more accurate mathematical model of CMP uniformity has been identified. An optimization function has been developed based on minimizing the removal profile slopes as well as the absolute value of the area under the removal curve. The model and function have been utilized to customize a thickness and uniformity multivariate run-to-run software control solution for the process. The controller is based on the Generic Cell Controller structure, which is a proven enabler for run-to-run control for a number of processes including CMP, vapor phase epitaxy, and etch. The control algorithm is a zero'th order adjustable linear approximation two-stage algorithm with EWMA noise filtering. This algorithm, which supports first order linear and non-linear models, has been demonstrated to be effective in CMP CTE and thickness multivariate control. The control solution has been enhanced to utilize both pre and post CMP process metrology along with process models to suggest process recipe modifications on a run-to-run basis. Results indicate improved control of CMP process non-uniformity qualities of interest. Further, the results quantify the significant benefit of utilizing pre-metrology (feedforward) information in addition to traditional post metrology (feedback) in determining control recipe advices.