Rapid Thermal Processing

One particular area of great current research interest is Rapid Thermal Processing (RTP), which is a versatile, single-wafer approach to semiconductor wafer processing that is suitable for several applications including annealing, cleaning, chemical vapor deposition, oxidation, and nitridation. RTP provides a rich variety of problems in modeling, simulation, and control. Here we provide a brief introduction to our work in this area during 1995 and 1996.

RTP System Gas Phase Simulation and Model Reduction

A. J. Newman; Advisors: R. A. Adomaitis and P. S. Krishnaprasad
Computer Aided Control Systems Engineering Lab and Intelligent Servosystems Lab

Project Background and Goals

Rapid Thermal Processing (RTP) is a versatile, cold-wall, single-wafer approach to semiconductor wafer processing suitable for several applications including annealing, cleaning, chemical vapor deposition (RTCVD), oxidation, and nitridation. RTCVD systems offer several advantages over conventional CVD systems including improved wafer-to-wafer deposition uniformity and reduced thermal budgets. The main obstacles for full acceptance of RTCVD into manufacturing are temperature reproducibility and uniformity during the thermal cycle when thin films are being deposited on the wafer surface. Our research focuses on the NC State Three Zone RTP System. For this system, the gas phase transport of energy and reactant species is examined in order to determine those factors which contribute substantially to deposition nonuniformity. Ultimately, the goal is to develop low order models for use in real-time simulation and model-based sensing and control. In particular, reduced-order gas phase models may be useful for determining the wafer state from available measurements of the tool state.


Reactor system geometry and operating conditions are based on the 3 zone furnace at NC State University. Modeling and high-fidelity simulation of the fluid phase transport of momentum and enthalpy are carried out with the Fluent computational fluid dynamics (CFD) software package. During the initial transient, "snapshots" of the fluid flow and temperature fields are captured. These snapshots have a certain degree of correlation due to the evolution of the flow fields being described by well defined equations of motion. These coherent structures are extracted via the Proper Orthogonal Decomposition, giving an empirically determined set of basis functions.

Project Results

An RTP simulator describing the gas phase flow and thermal dynamics has been developed. We have shown that the details of the flow and temperature fields can be captured with a time-varying linear combination of relatively few empirically determined spatial modes. In addition, initial estimates of heat transfer from the wafer to the gas phase indicate that this heat transfer is not a significant factor in causing nonuniformities.


Decomposing the spatial structure of the gas phase dynamics into its underlying coherent structures results in a set of trial functions which can be used as a basis in a Galerkin discretization of the original equations of fluid motion. This results in an extremely low order ODE model, suitable for real-time simulation. Furthermore, understanding the nonlinear, dynamic coupling between the modes is crucial to understanding the relationship between the manipulated variables and the spatial variations to be controlled, and to combining sensor measurements in order to estimate the wafer state.

Future Directions

Transport of reacting and deposition product species for depletion studies and residual gas analysis sensor development is currently being pursued. With this complete model of gas phase transport, we intend to explore the relationship between coherent structures, sensor placement, and how one might specify the minimum number of gas phase/tool state sensors required to uniquely determine the wafer state.

The NC State University Three Lamp Zone RTCVD System

[NCSU RTP System]

Temperature Field Snapshots

Optimal Eigenmodes of the System

Page maintained by Andrew Newman. Send comments to
[Mail Icon] newman@isr.umd.edu
Latest Update: December 30, 1997.