A Time Dependency CMP Model for Dishing and Erosion in Copper Damascene and STI Process

Transcription

1 1 A Time Dependency CMP Model for Dishing and Erosion in Copper Damascene and STI Process SFR Workshop & Review April 17, 2002 Jianfeng Luo, Runzi (Tiger) Chang and Professor David A. Dornfeld Berkeley, CA 2002 Goal: To build an integrated CMP model for basic mechanical and chemical elements.

2 2 Motivation Need to know what input variables and how they influence the formation of dishing and erosion in copper damascene and STI process.

3 3 ide Erosion and Copper Dishing in Copper Damascene Process Dishing Erosion Copper line thinning is equal to erosion plus dishing. Resistance is a function of line thinning Dishing worsens the copper line topography milar phenomena happen in STI process Source: Hitachi, 2000

4 4 Definition of Input and Output Variables in Damascene Process Th Cu S h 0 h Start point of polishing End point of polishing (L/E): A parameter related to the pad material (Young s modulus E) and pad topography (L) S: Original step height h: Step height during polishing h 0 : Initial deformation of pad asperities P 0 : Down pressure applied on die D: Pattern density of copper line T: Time of Polishing K e and C 1 : Preston s Coefficients for copper (MRR= KeP 0 + C 1 ) K ox and C 2 : Preston s Coefficients for oxide (MRR= K ox P 0 + C 2 )

5 5 First Stage of Material Removal: Linear Time Dependency of Step Height Reduction Cu h h 0 = (L/E)P 0 /(1-D) the deformation of pad asperity Cu h= h 0 Start point of the first stage End point of the first stage h= K e P 0 /(1-D)T when h> h 0. Only higher part of the step is in contact with pad. One single material-- copper is removed. Step height is a linear function of time.

6 6 Second Stage of Material Removal: Exponent Time Dependency of Step Height Reduction Cu h= h 0 Start point of the second stage h End point of the second stage (Dishing is formed before erosion) h= h 0 Exp(-(K e E/L)T) when h< h 0, where h 0 = (L/E)P 0 /(1-D). Both higher and lower part of the step are in contact with pad. Step height is an exponent function of time.

7 7 Third Stage of Material Removal (Over-polishing): Exponent Time Dependency of Dishing and Erosion Er h= h c Cu h= h c h Start point of the third stage End point of the third stage Dishing h= a/b-[a/b- h c ] Exp (-bt)] where a=(-k ox P 0 -C 2 +K e P 0 +C 1 ), b= (E/L)(K ox D+K e (1-D)) and h c is the initial dishing. Erosion Er= (K ox P 0 +C 2 )T+(K ox ED/b[ h c -a/b][1-exp(-bt)] Both dishing and erosion are exponent functions of time.

8 8 Selected mulation Results of the Step Height as a Function of Time 600 Original Step Height S 500nm 500 PD= 0.9 PD= 0.1 Final Dishing Original Copper Thickness Th Down Pressure P nm 12kPa Step Height (nm) First Stage Second Stage Third Stage Preston s Coefficeint for Copper K e and C 1 Preson s Coefficient for ide K ox and C 2 Coefficient E/L Time Constant for Second Stage K e E/L Time Constant for Third Stage (E/L)(K ox D+K e (1-D)). K e = 0.278nm/(Sec.kPa) C 1 = 2.5nm/Sec. K ox = nm/Sec.kPa) C2= kPa.nm Sec. 261Sec. for PD= Sec. for PD= Time (Second) Duration of Stage 1 Duration of Stage 2 Duration of Stage 3 0 Sec. for PD= Sec. for PD= Sec. for PD= Sec. for PD= Sec. for PD= Sec. for PD= 0.9

9 Normalized Remaining Step Height Polishing Time (Second) 9 Experiment Results VS. Model Predictions of the Time Dependency of Step Height Reduction in The First and Second Stages Linear Stage Exponent Stage Experimental data from Stavreva et. al., Microelectronic Engineering, Vol. 33, Experimental (PD= 0.9) Experimental (PD= 0.85) Experimental (PD= 0.8) Experimental (PD= 0.67) Experimental (PD=0.5) Experimental (PD=0.2) Model (PD= 0.9) Model (PD= 0.85) Model (PD= 0.8) Model (PD= 0.67) Model (PD= 0.5) Model (PD= 0.2)

10 10 Experiment Results VS. Model Predictions of the Time Dependency of Dishing in The Third Stage Dishing (nm) Experimental Data 1 Model Experimental Dta 2 Model Experimental Data 3 Model Overpolishing (%) Experimental data from Pan et. al., IEEE IITC, San Francisco, 1999.

11 11 Discussion Dishing and erosion depend on polishing time, pattern density, pad material, topography, and material removal mechanism (Parameters related to Preston s coefficient including abrasive size, slurry chemicals, etc.) This model can be integrated with the comprehensive material removal model developed earlier. Dishing and erosion depend on processes before CMP such as electroplating, CVD, and copper line thickness before polishing. More experimental verification is needed. Model Limitations (1) An important parameter: copper line width is not included in the model yet. (2) Model may be invalid when the copper line width is small in comparison with the pad asperity.

12 Goals Develop comprehensive chemical and mechanical model. Perform experimental and metrological validation, by 9/30/2003.