Institute for Systems Research

This reactor is based on a novel segmented showerhead design in which gas precursor composition can be individually controlled in the gas fed to each segment. Because the exhaust gas is recirculated up through the showerhead though the individual segments, the gas flow pattern created eliminates convective mass transfer between the segment regions. The effect of this design is a CVD system in which across-wafer composition gradients can be accurately predicted and controlled.

This paper discusses the development of a simulator for a three-segment prototype that has recently been constructed as a modification to an Ulvac ERA1000 CVD cluster tool. A preliminary set of experiments has been performed to evaluate the performance of the prototype in depositing tungsten films for a range of wafer/showerhead spacing and segment gas compositions. We discuss the simulation approach taken to developing the simulator for this system focusing on a one-dimensional simulation of transport through the segments and exhaust mixing region, a model valid in the limit of close showerhead/wafer spacing. The use of simulation in the prototype system design, interpreting experimental data, and its ultimate use in controlling the CVD process to achieve true programmable CVD operation all will be discussed. Further information can be found at the project website, http://www.isr.umd.edu/Labs/CACSE/research/progrxr

By monitoring the resonant frequency of exhaust process gases, the depletion of WF6 resulting from the reduction by H2 was readily observed in the 0.5 Torr process for wafer temperatures ranging from 300 to 350 C. Despite WF6 depletion rates as low as 3-5%, in-situ wafer-state metrology was achieved with an error less than 6% over 17 processed wafers.

This in-situ metrology capability combined with accurate sensor response modeling suggests an effective approach for acoustic process sensing in order to achieve run-to-run process control of the deposited tungsten film thickness.

With process heating lamp power held constant, wafertemperature variations of up to 160 degrees K were observed by varying feed gasH_2/N_2 ratio. Heat transfer between the wafer and susceptor wasstudied by shifting the instrumented wafer off the susceptor axis,exposing a portion of the wafer backside to the chamber floor. Modelpredictions and experimental observations both demonstrated that the gasvelocity field had little influence on the observed wafer and predictedgas temperatures.

This paper presents two integrated models for understanding cluster tool behavior. The first model is a network model that evaluates the total lot processing time for a given sequence of activities. By including a manufacturing process model (in the form of a response surface model, or RSM), the model calculates the total lot processing time as a function of the process parameter values and other operation times. This model allows one to quantify the sensitivity of total lot processing time with respect to process parameters and times.

In addition, we present an integrated simulation model that includes a process model. For a given scheduling rule that the cluster tool uses to sequence wafer movements, one can use the simulation to evaluate the impact of process changes including changes to product characteristics and changes to process parameter values. In addition, one can construct an integrated network model to quantify the sensitivity of total lot processing time with respect to process times and process parameters in a specific scenario.

The examples presented here illustrate the types of insights that one can gain from using such methods. Namely, the total lot processing time is a function not simply of each operation's process time, but specifically of the chosen process parameter values. Modifying the process parameter values may have significant impacts on the manufacturing system performance, a consequence of importance which is not readily obvious to a process engineer when tuning a process (though in some cases, reducing process times may not change the total lot processing time much).

Additionally, since the cluster tool's maximum throughput depends upon the process parameters, the tradeoffs between process performance and throughput should be considered when evaluating potential process changes and their manufacturing impact.