Intel/GT Opportunity
Scholar’s Program 2005-6
Research Agenda
Year: 2005-06
Mentor: Cleon Davis
Scholars: Preston Burden, Daniel Cepeda
Date Submitted:
Research Area:
Semiconductor Process Control
Proposed Research
Tasks:
1.
a. Programming: Additions to
Object-Oriented Neural Network Simulator (ObOrNNS) (Preston Burden)
The Java-based Object-Oriented Neural Network Simulator (ObOrNNS)
is a software package developed by the Intelligent Semiconductor
Manufacturing group at the Georgia Institute of Technology. This program is
capable of constructing, training, and exercising multilayer perceptron
neural networks for various semiconductor manufacturing applications.
ObOrNNS implements the well-known error back-propagation training
algorithm. In addition to standard empirical modeling, ObOrNNS has a
semi-empirical (or “hybrid”) neural network modeling feature, which is used
to derive phenomenological models based on known process chemistry and
physics. ObOrNNS also contains an optimization routine based on genetic
algorithms for use in recipe synthesis. As a Java-based application, the
program is available to all platforms that support the Java Runtime
Environment.
A. Parallel
neural-genetic implementation
i. Reference
Search
a) Neural
networks
b) Genetic
algorithms
c) Summary
ii. Java
Programming
B. Update
ObOrNNS Graphical User Interface
C. Neural
network feedback control implementation
i. Reference
search
a) Neural
network control
b) Summary
ii. Java
Programming
Timeline:
November: Background research and literature
review
December: Exercise neural networks in
ObOrNNS
January: Review java code and start
programming
February: Update ObOrNNS graphical
user-interface
March: Implement neural network controller,
paper write-up/poster development
April: Poster development and presentation
2.
VFM Processing and Analysis
(Daniel Cepeda)
Variable frequency Microwave (VFM) curing can be performed in
a MircoCureä
2100. The controllable parameters for the MicroCureä
2100 are center frequency, bandwidth, sweep rate, power level, and ramp
rate. Samples may be processed at a fixed frequency, at a variable
frequency with a specific center frequency and bandwidth, with a varying
bandwidth, and/or varying sweep rate. The unique feature of this system is
the capability of frequency stepping. The system can step through 4096
frequencies over a 1.15 GHz bandwidth every 0.1 s. This frequency stepping
provides a time-averaged uniform energy distribution throughout the cavity,
which eliminates the non-uniform temperature distribution that occurs in
single-frequency microwave furnaces.
This VFM furnace also has a feedback control system that
regulates the temperature of the sample being processed. The control system
can automatically adjust the power levels to maintain the sample at the
desired temperature, which allows good control of ramp rates and final hold
temperatures of the samples to be processed. The VFM furnace has provisions
to maintain an inert atmosphere during processing of samples. The
processing cavity can be pumped down using a mechanical pump and back-filled
using nitrogen gas for processing in an oxygen-free environment. Another
important characteristic of a VFM furnace is the ability to place metal
inside the microwave cavity because charge build up and arcing due to the
presence of the field is eliminated.
This project is a continuation from last year points a
A.
Background Research (Completed last year)
i. Terms
to define
1) Semiconductors
2) Semiconductor
packaging
3) Semiconductor
dielectric
4) Polymer
5) Polyimide
6) Cure
7) Imidization
8) Emissivity
ii. Polymer
dielectrics
1) What
are they used for?
2) Who
manufactures them?
3) Examples
a. Benzocyclobutene
b. Polyimide
PI 2611
iii. Extensive
Literature Review: Microwave processing of polymers
1) Fixed
Frequency
2) VFM
iv. Order
Sensors for VFM
1) Thermocouple
2) Fiber
optic probes
v. Design
of Experiments/Statistical Design
1) Central
composite circumscribed (CCC)
2) Central
composite inscribed (CCI)
3) 2^n
Factorial
4) Latin
Hypercube Sampling
5) D-Optimal
B.
Do a designed experiment (Statistical Design)
i. Choose
a design (CCC, CSI, 2^n, Latin hypercube, D-optimal)
ii. Implement
Design
1) Determine
VFM inputs that will be used in design
2) Use
RS-Discover for design
C.
VFM processing
D.
Data Acquisition
i. Degree
of cure/percent imidization :FTIR
ii. Optical
properties and film thickness measurements
1) In
plane index of refraction : Metricon Prism coupler
2) Through
Plane index of refraction : Metricon Prism coupler
3) Thickness
:Metricon Prism coupler
iii. Electrical
properties
1) Relative
permittivity
2) Dielectric
loss
3) Parallel
plate capacitor structure needed to measure
4) 
Actual dielectric
measurements made with a LCR 
meter and using the
following equations:
iv. Mechanical
properties
1) Residual
stress
a. -Calculate
radius of curvature for wafer with and 
without films and use the
following equation
b. E/(1-ν)
= biaxial elastic modulus, h is substrate thickness in meters, t is film
thickness, R is the differential radius of curvature (1/R = 1/R2-1/R1,
where R1 is radius of curvature of substrate and R2 is radius of
curvature of the film)
2) Young’s
Modulus
a. Instron
tensile tester (estimated from the stress strain curves obtained from
tensile testing of thin, free-standing films)
b. Tensile
strength
c. Elongation
to break (ETB)
d. TS
and ETB are considered to be the maximum values the polymer film withstood
prior to failure
3) Thermal
Stability
a. Thermo-gravimetric
analysis (TGA)–Seiko TG/DTA 320
4) Thermo
mechanical analysis (TMA):
a. TMA
Model 2940 by TA Instruments
v. Physical
properties
1) Moisture
uptake—Quartz Crystal Nanobalance (QCN)
E.
Input data in to ObOrNNS
i. Neural
network modeling
ii. Genetic
algorithm optimization
iii. Sensitivity
analysis
iv. Verify
simulation results—VFM processing
Timeline:
November: Data
acquisition: Fourier transforms infrared spectroscopy and Metricon prism
coupler measurements
December: VFM
processing and Metricon prism coupler measurements
January: Data
acquisition: Electrical properties
February: Data
acquisition: Physical and mechanical properties
March: Input data in
ObOrNNS and paper write-up/poster development
April: Poster
development and presentation
