Lefschetz Center for Dynamical Systems Seminar
Abstract: Generally speaking animals do not waste too much of any scarce resource, of which energy is one. Similarly, if robots are ever to be practical, reasonable energy efficiency would be useful. In loose terms, we have been trying to understand legged locomotion from an energetic point of view. One approach is to make energy- efficient legged machines, the most striking of which are those with no motors that only walk down gentle slopes. Another approach is to study energy-use of humans. Finally one can make mathematical analyses of simple mechanical models. One idea that comes out of such is this: perhaps the non-holonomic nature of the intermittent contact in walking contributes to the stability of walking.
Joint Materials/Solid Mechanics Seminar Series
Prita Pant, Indian Institute of Technology-Bombay, Department of Metallurgical Engineering and Materials Science, Powai, Mumbai - 400076, INDIA Klaus W. Schwarz, IBM Watson Research Center, P.O. Box 218, Yorktown Heights, NY | |
Abstract:
The yield strength of thin metal films is commonly as much as an order of magnitude greater
than in bulk materials having similar composition and thermomechanical processing histories.
Because of this, the metallizations that are found in all kinds of integrated micro- and nano-fabricated
devices (integrated circuits, MEMS, microfluidics, optics, etc.) often support extremely high stresses
that lead to failures via delamination, fracture, void formation, and other mechanisms. Thus there has
been a great deal of interest in understanding the strengthening mechanisms in thin films. A number
of models have been proposed based on dimensional constraints and classical mechanisms such as
interactions of dislocations with grain boundaries and other dislocations. While many of these models
can generate experimentally-observed stress levels, most depend on artificial assumed dislocation
configurations. We have used 3-D dislocation dynamics to study the plastic behavior of single crystal
thin metal films in relatively "parameter-free" simulations. We find that the strength of the film can
largely be attributed to the large fluctuations in stress that develop in the film. Regions of low stress
both permit weak interactions to stop dislocation motion and increase the probability of strong
interactions by concentrating mobile dislocations. The likelihood of strong interactions is correlated
with the coefficient of variation (COV) of the stress field. The evolution of the COV with plastic
strain (dislocation density) has been explored. Both stresses and hardening rates are comparable to
experimentally-observed values.
Note: There will be a dinner for Prof. Baker after the seminar. Please contact Ms. Pat Capece at x1501 if you wish to
attend.
CCMB Seminar Series Lecture
Hosted by: Daniel Weinreich Refreshments served at 3:45 p.m. |
Abstract: Darwin's theory of evolution can be understood as a simple algorithm, a formal step-by-step procedure or mechanism, that produces adaptation in biological systems. Computer science studies a broad range of algorithms and seeks to understand which algorithms work best for which types of problems. Is the algorithm that Darwin described the only algorithm relevant to natural evolution, and if there are others, can they solve adaptive problems that are not solvable with Darwin's model? It is well known that there have been numerous events of horizontal gene transfer, important cases of genetic encapsulation and symbiogenesis, and occasional 'major evolutionary transitions' where "entities that were capable of independent replication before the transition can replicate only as part of a larger whole after the transition" (Maynard Smith and Szathmary). It has been suggested before that these events present a challenge to Darwinian gradualism because the results of these events are not small genetic changes, but this somewhat misses the important point. What is algorithmically interesting about these phenomena is not that the genetic changes are large rather than small, but that the genetics involve the union of one genetic lineage with another lineage evolved in parallel rather than the sequential modification of a single lineage that Darwin described. Using algorithmic concepts we can understand why it makes a difference to evolve things in parallel and then bring them together, rather than evolve things sequentially in a single lineage. In computational terms the difference couldn't be more fundamental: Darwinian linear incremental improvement is analogous to `stochastic local search', a very basic form of optimisation, but `compositional' mechanisms, as I refer to them, are analogous to `divide and conquer' optimisation, a fundamentally different class of algorithm based on problem decomposition.
Brown Analysis Seminar
PDE Seminar
Department of Mathematics Colloquium
<--- 2007 Index