
ECE 695A Lecture 37: Radiation Induced Damage – An overview
20 Apr 2013  Online Presentations  Contributor(s): Muhammad Alam
Outline:
 Introduction and short history of radiation damage
 Radiation damage in various types of components
 Sources of radiation
 A basic calculation and simulation approaches
 Conclusions

ECE 695A Lecture 37R: Review Questions
20 Apr 2013  Online Presentations  Contributor(s): Muhammad Alam
Review Questions:
 Why is SOI more radiation hard compared to bulk devices? What do you feel about radiation hardness of FINFET?
 What type of radiation issues could arise for thinbody devices like FINFET?
 What is error correction code? Why does it correct for MBU?
 What is the difference between SEE and SEU?
 What is ‘displacement damage’? What is its unit?
 What is LET? What are the two types of units used to describe LET?
 Explain the origin of the term Blackwall? What is the blackwall thickness.

ECE 695A Lecture 34: Scaling Theory of Design of Experiments
18 Apr 2013  Online Presentations  Contributor(s): Muhammad Alam
Outline:
 Introduction
 Buckingham PI Theorem
 An Illustrative Example
 Recall the scaling theory of HCI, NBTI, and TDDB
 Conclusions

ECE 695A Lecture 34A: Appendix  Variability by Bootstrap Method
18 Apr 2013  Online Presentations  Contributor(s): Muhammad Alam

ECE 695A Lecture 33: Model Selection/Goodness of Fit
18 Apr 2013  Online Presentations  Contributor(s): Muhammad Alam
Outline:
 The problem of matching data with theoretical distribution
 Parameter extractions: Moments, linear regression, maximum likelihood
 Goodness of fit: Residual, Pearson, Cox, Akika
 Conclusion

ECE 695A Lecture 33R: Review Questions
18 Apr 2013  Online Presentations  Contributor(s): Muhammad Alam
Review Questions:
 With higher number of model parameters, you can always get a good fit – why should you minimize the number of parameters
 Least square method is a subset of maximum likelihood approach to data fitting. Is this statement correct?
 What aspect of the distribution function does CoxOakes method emphasize?
 Can MLE be used for any distribution function?
 How would you change the MLE condition if you had 3 independent parameters to estimate?
 Does increase in model parameters increase chances of passing c2 test?
 How does the methods affected by censored data (e.g. , TDDB test yet to finish?)

ECE 695A Lecture 32R: Review Questions
17 Apr 2013  Online Presentations  Contributor(s): Muhammad Alam
Review Questions:
 Why do people use Normal, lognormal, Weibull distributions when they do not know the exact physical distribution?
 What is the problem of using empirical distributions? What are the advantages?
 If you must choose an empirical distribution, what should be your criteria?
(Nos. of parameters, physical principles, etc.)
 Why does everyone suggest the use of CDF for empirical datafitting, rather than PDF? (Obviously one can go from one function to the other)
 There are all sorts of distribution functions (e.g. survivability function) ? If everything is related to everything else, why do we need so many?
 How would you determine the BFRW failure rates? Mean Hazard rate?

ECE 695A Lecture 32: Physical vs. Empirical Distribution
17 Apr 2013  Online Presentations  Contributor(s): Muhammad Alam
Outline:
 Physical Vs. empirical distribution
 Properties of classical distribution function
 Momentbased fitting of data
 Conclusions

ECE 695A Lecture 31R: Review Questions
15 Apr 2013  Online Presentations  Contributor(s): Muhammad Alam
Review Questions:
 What is the difference between parametric estimation vs. nonparametric estimation?
 What principle did Tacho Brahe’s approach assume?
 What is the difference between population and sample? When we collect data for TDDB or NBTI, what type of data are we collecting?
 What problem does Hazen formula avoid regarding Fi=i/n? How is this justified?
 What is the problem of Hazen formula with respect to censored data?
 Can you think of a situation where data censoring may be necessary for NBTI test?
 What is the difference between an outlier vs. a censored data?
 Do I need to know what the physical distribution is before using Hazen or Kaplan formula? Why or why not?

ECE 695A Lecture 31: Collecting and Plotting Data
15 Apr 2013  Online Presentations  Contributor(s): Muhammad Alam
Outline:
 Origin of data, Field Acceleration vs. Statistical Inference
 Nonparametric information
 Preparing data for projection: Hazen formula
 Preparing data for projection: Kaplan formula
 Conclusion

ECE 695A Lecture 31A: Appendix  Bootstrap Method Introduction
15 Apr 2013  Online Presentations  Contributor(s): Muhammad Alam

ECE 695A Lecture 30R: Review Questions
08 Apr 2013  Online Presentations  Contributor(s): Muhammad Alam
Outline:
 What is the difference between extrinsic vs. intrinsic breakdown?
 Does gas dielectric have extrinsic breakdown? Why or why not?
 What does ESD damage and the plasma damage to thin oxides?
 Can you explain the physical meaning of infant mortality ? How does it relate to yield of semiconductor manufacturing?
 Can you reinterpret the Apgar tests in terms of infant mortality?
 What is the difference between the Weibull for thick vs. thin oxides?

ECE 695A Lecture 30: Breakdown in Dielectrics with Defects
08 Apr 2013  Online Presentations  Contributor(s): Muhammad Alam
Outline:
 Introduction
 Theory of preexisting defects: Thin oxides
 Theory of preexisting defects: thick oxides
 Conclusions

ECE 695A Lecture 29R: Review Questions
08 Apr 2013  Online Presentations  Contributor(s): Muhammad Alam
Review Questions:
 Mention a few differences between thick and thin oxide breakdown.
 Is breakdown in thick oxides contact dominated? Can I use AHI theory here?
 How does the Paschen’s cascade initiate?
 What does it mean to have a fractal dimension of 1.7 for 2D breakdown? Why does the number suggest spatial correlation ?
 What is a color center? How does color center help us visualize breakdown in polymers?
 Explain physically the origin of the minimum breakdown voltage for gas dielectric?
 Is gas dielectric breakdown reversible? What about solid dielectric BD?

ECE 695A Lecture 29A: Appendix  Dimension of a Surface
08 Apr 2013  Online Presentations  Contributor(s): Muhammad Alam

ECE 695A Lecture 28: Circuit Implications of Dielectric Breakdown
08 Apr 2013  Online Presentations  Contributor(s): Muhammad Alam
Outline:
 Part 1  Understanding PostBD FET behavior
 BD position determination
 Hard and Soft BD in FETs
 Distinguishing leakage and intrinsic FET parameters shifts
 Part 2  Impact of breakdown on digital circuit operation
 BD in ring oscillator
 BD in SR AM cell
 Timing, BD into soft node
Based on a IRPS Tutorial by Ben Kaczer

ECE 695A Lecture 29: Circuit Implications of Dielectric Breakdown
08 Apr 2013  Online Presentations  Contributor(s): Muhammad Alam
Outline:
 Introduction
 Spatial and temporal dynamics during breakdown
 Breakdown in bulk oxides: puzzle
 Conclusions

ECE 695A Lecture 27: Correlated TDDB in OffState HCI
29 Mar 2013  Online Presentations  Contributor(s): Muhammad Alam

ECE 695A Lecture 27R: Review Questions
29 Mar 2013  Online Presentations  Contributor(s): Muhammad Alam

ECE 695A Lecture 261: Statistics of Soft Breakdown via Methods of Markov Chains
28 Mar 2013  Online Presentations  Contributor(s): Muhammad Alam
Outline:
 Spatial vs. Temporal correlation
 Theory of correlated Dielectric Breakdown
 Excess leakage as a signature of correlated BD
 Conclusions

ECE 695A Lecture 262: Statistics of Soft Breakdown (Breakdown Position correlation)
28 Mar 2013  Online Presentations  Contributor(s): Muhammad Alam
Outline:
 Position and time correlation of BD spot
 How to determine the position of the BD Spot
 Position correlation in BD spots
 Why is localization so weak?
 Conclusions

ECE 695A Lecture 26R: Review Questions
28 Mar 2013  Online Presentations  Contributor(s): Muhammad Alam

ECE 695A Lecture 25R: Review Questions
27 Mar 2013  Online Presentations  Contributor(s): Muhammad Alam
Review Questions:
 Explain why percolation resistance is area independent?
 Why is the physical origin of the distribution of percolation resistance?
 How would the ratio of hard and soft breakdown change with an auxiliary parallel capacitor in constant voltage stress? Explain.
 What is the evidence from HBD that a thermal process may be involved?
 In constant voltage stress, do you expect that BD would be harder for thicker oxide? Or does it not matter?
 How might people have missed SBD even for thin oxides where BD in operating condition is expected?
 Do you find a limitation of the SPICE like model used in the analysis?

ECE 695A Lecture 24: Statistics of Oxide Breakdown  Cell percolation model
21 Mar 2013  Online Presentations  Contributor(s): Muhammad Alam
Outline:
 Observations: Failure times are statistically distributed
 Models of Failure Distribution: Extrinsic vs. percolation
 Percolation theory of multiple Breakdown
 TDDB lifetime projection
 Conclusions

ECE 695A Lecture 24R: Review Questions
21 Mar 2013  Online Presentations  Contributor(s): Muhammad Alam