AVS 51st International Symposium
    Manufacturing Science and Technology Tuesday Sessions
       Session MS-TuP

Paper MS-TuP1
Reactive Preclean H@sub 2@/He Plasma Prior Copper Deposition, Investigation on the First Wafer Effect and Multivariable Advanced Process Control

Tuesday, November 16, 2004, 4:00 pm, Room Exhibit Hall B

Session: Poster Session
Presenter: R. Petri, STMicroelectronics, France
Authors: R. Petri, STMicroelectronics, France
L. Bucelle, STMicroelectronics, France
Correspondent: Click to Email
As a general concern in the semiconductor industry, the process stability is crucial and the methodology to ensure this stability becomes more and more complex. By years, the submicron technology constrains the manufacturing to introduce thorough process controls: initially guaranteed by specification limits only, it was improved by the introduction of control limits and Statistical Process Control. Recently, the industry is moving to Advanced Process Control based on a multivariable control of machine parameters. By this way, those parameters can eventually be correlated with product yield. On another hand, process integration becomes extremely sensitive to the process environment and therefore any uncontrolled modification of the reactor surface, such as reactor wall state, wall temperature, in-situ plasma clean, could generate a process drift and may induce severe yield lost or reliability issues. As a consequence, the study and the eradication of the first wafer effect, due to its impact on the manufacturing performances, has generated a lot of focus and publications. This paper is an application of the Advanced Control Process to ensure the stability of the reactive H@sub 2@/He plasma preclean process used in dual damascene structure prior copper deposition. It shows that tiny variations of the wafer processing context are easily detected by appropriate APC model. It appears that plasma potential is very dependant on process controlled parameters but also varies significantly with substrat nature as well as reactor wall temperature. We show that any process drift from the standard expected behavior is detected with the developed model. The nature of the deviation (i.e. signature) is also given. Therefore, in a manufacturing environment, APC is an extremely versatile methodology which can detect in line process drift, but also allows more fundamental investigations, like first wafer effect understanding and eradication, much easier, faster and cost effective.