Modeling of Collimateti Titanium Nitride Physical Vapor Deposition using a Combined Specular-Diffuse Formulation
- PDF / 305,256 Bytes
- 6 Pages / 414.72 x 648 pts Page_size
- 48 Downloads / 207 Views
3 Arizona State University, Tempe, AZ 85287-6006
Re-emissions from surfaces receiving a ballistically transported molecular flux have, in the most general case, a spatial distribution with a combination of specular and diffuse components. The addition of specular transmission in the EVOLVE model allows combined specular-diffuse re-emission to be explicitly modeled. Previous modeling of the step-coverage of sputtered titanium nitride (TiN) films in contact structures produced good predictions with a sub-unity sticking coefficient, 0.6 in value, using a strictly diffuse formulation for material re-emission. New data of collimated sputtered TiN with contact aspect ratios exceeding 4.0 was modeled using the EVOLVE specular-diffuse formulation. The result was more accurate model predictions of experimental data and replacement of the ad hoc empirical parameter adjustment in previous modeling with a more detailed physical description.
Introduction Titanium nitride (TiN) films serve a critical role as a diffusion barrier in the contact and via structures of integrated circuits. The present method of depositing TiN is by physical vapor deposition using a sputtering process. Conventional sputtering systems produce films of inadequate conformality in high aspect ratio contacts. Sputtered Ti/TiN coverage can be improved significantly by the use of a collimating device to direct more of the sputtered flux into the contact. Previous work in modeling the conformality of Ti/TiN PVD in contacts produced accurate predictions of the data [7] with a sub-unity sticking coefficient of 0.6, but required the empirical adjustment of a model parameter (the "beaming coefficient") which depended on the aspect ratio of the collimator. Recent work by Lin and co-authors [6] incorporated a detailed Monte Carlo simulation of the flux distribution exiting the collimator. This addition to the model eliminated the dependence of the over-cosine "beaming coefficient" on collimator aspect ratio. Lin and co-workers used a sub-unity sticking coefficient of 0.7 and a constant value of 2.0 for the over-cosine beaming coefficient. The result was accurate model predictions for both collimated and uncollimated PVD Ti/TiN data. The usual assumption of unity sticking coefficient for a PVD process is not consistent with observations of Ti/TiN deposition inside re-entrant features that we have observed in 575 Mat. Res. Soc. Symp. Proc. Vol. 355 ©1995 Materials Research Society
our data. However, recent work by Bang et al. [1] with PVD Ti (without TiN) indicated no deposition on the underside of a deeply re-entrant "overhang" test structure, implying unity sticking coefficient. These inconsistencies undoubtably arise from the simplifications made in both the models and intrepretations of the data. The sticking coefficient may differ between Ti and TiN and may also depend upon un-modeled variation in operating parameters of the sputtering system. Indeed, the sticking coefficient may be varying over the time of the process, in particular in the transition from s
Data Loading...
