Optimization of a Virtual CVD Reactor 2015
UO Semiconductor Device Processing Masters Program
Adapted from (Danielle Amatore and Milo Koretsky
Department of Chemical Engineering
Oregon State University)
Objective: Develop an optimal “recipe” (i.e. choice process parameters) for a Low Pressure Chemical
Vapor Deposition (LPCVD) reactor at as low a cost as possible. This recipe will be released into high
volume manufacturing for the next-generation products at Duck QuackerChips.
Logistics: The virtual CVD is accessible at: http://cbee.oregonstate.edu/education/VirtualCVD/ You need
to click “jump into 3D Fab” on the left and install the client software. Your login to the virtual CVD is
“Su14_alphax” or “Su14_omegax”where x is a number from 1-4. Get your group password from
Benjamín. You are charged virtual dollars for every experiment you do, so think carefully about your
experimental design.
Overview:
Your team’s task is to develop a “recipe” for high volume manufacturing of Silicon Nitride
(Si3N4) using Low Pressure Chemical Vapor Deposition (LPCVD). The growth and measurements will
be made via computer simulation in our VirtualCVD reactor. You should develop a recipe that grows
Si3N4 to a target thickness of 1500 Å uniformly within the wafer and from wafer to wafer. Your final
process must have a uniformity of > 85% for the product to be useful at all. Your grade will depend
on the amount of material you use in the process, how much money you spend optimizing the
process, and how much time it takes for the growth step.
The furnace has a capacity for batches of up to two hundred 300 mm diameter wafers. The wafer
spacing is 8 mm. It has 5 temperature zones that can be set individually. In addition, you can set the flow
rates of ammonia (NH3) and dichlorosilane (DCS, SiCl2H2) feed gases, the reactor pressure and the time.
You will also have access to a (Virtual) ellipsometer, with which you can measure the film thicknesses at
the points on any wafer that you select. You will be charged $5,000 for each run and $10 for each
measurement (in Virtual$, of course).
Computation of Standard Error and Uniformity:
The reactor model takes as input the reactor parameters and produces as output the function hi(x, y),
where hi(x, y) is the thickness in Å at wafer i and position (x, y) mm.
Let n be the number of wafers and let h' be the target thickness in Å. Let C be the set of all (x, y) points on
a wafer. Let Ei be the standard error at wafer i, and E be the overall standard error. Let u be the overall
uniformity. The model uses the following equations to obtain standard errors and uniformity:
You will want to consider how to appropriately estimate the uniformity and overall standard error from
your “virtual” experimental data where you have a finite number of points.
Deliverables
A. Design Strategy Presentation. Due Monday July 27th @ 8 am
You need to come up with a strategy to explore the parameter space. It may be helpful to
consider using design of experiments (DOE) in your approach, or it may not. Your group’s design
strategy needs to be explained in a 1-2 slide powerpoint that will be presented during a process
meeting to the class on July 27th @ 8 am. You will have 5 minutes to make your presentation.
B. Release to Production. Due Friday July 25th @ 8 am
Submit your final process recipe for release to production (this must be done in the virtual fab.).
Report out to the class all important variables which should include your recipe, the final
uniformity, thickness and uniformity profile throughout reactor, and total money spent. Note that
the virtual reactor also records these parameters which I have access to.
D. Experiment Journal (subject to random inspection, otherwise ungraded)
As you perform the virtual experiments, you need to keep track of the run parameters, data
analysis, and conclusions you infer from the analysis (similar to information you would track in
“real” lab experiment). Pay special attention to any changes you make in your overall
experimental strategy.
E. Final Written Report Due Thursday July 30th @ 9 am
Your two page report should include the following items (turn in hard copy in class, one report per
team):
i. Design and measurement strategy
ii. Data analysis method and data
iii. Final process recipe (including achieved uniformity within wafer and throughout reactor)
iv. Final cost and score based on below formula.
v. Lessons learned and recommendation for further work
Your final score will be determined as follows:
(1) 50% - ability to meet or exceed key performance metrics in final submitted recipe
This grade will be calculated according to the following formula:
Performance points (out of 100) = 25 pts*(uniformity/ 90%) + 25 pts*(DCS utilization/75%) + 25
pts*(30 min/time) + 25 pts ($100,000/your money spent)
The maximum points you can earn for any category (i.e. uniformity, utilization, time, money spent) is 35
pts.
Uniformity less than 85 % = 0 pts for uniformity. The score normalization factors are indications of what
your boss thinks should be possible. You may be able to do better or it may not be possible to hit all
targets simultaneously.
(2) 50% is based on the quality of the written report, data presented, and data analysis rigor, as
well as on the ability of all team members to accurately explain project during presentations.