The thesis of this extended abstract is simple. High productivity comes from high level infrastructures. To measure this, we introduce a methodology that goes beyond the tradition of timing software in serial and tuned parallel modes. We perform a classroom productivity study involving 29 students who have written a homework exercise in a low level language (MPI message passing) and a high level language (Star-P with MATLAB client). Our conclusions indicate what perhaps should be of little surprise: (1) the high level language is always far easier on the students than the low level language. (2) The early versions of the high level language perform inadequately compared to the tuned low level language, but later versions substantially catch up. Asymptotically, the analogy must hold that message passing is to high level language parallel programming as assembler is to high level environments such as MATLAB, Mathematica, Maple, or even Python. We follow the Kepner method that correctly realizes that traditional speedup numbers without some discussion of the human cost of reaching these numbers can fail to reflect the true human productivity cost of high performance computing. Traditional data compares low level message passing with serial computation. With the benefit of a …

This paper accompanies a poster that is being presented atthe SciDAC 2006 meeting in Denver, CO. This project focuses on leveragingscientific visualization and analytics software technology as an enablingtechnology for increasing scientific productivity and insight. Advancesincomputational technology have resultedin an "information big bang,"which in turn has createda significant data understanding challenge. Thischallenge is widely acknowledged to be one of the primary bottlenecks incontemporary science. The vision for our Center is to respond directly tothat challenge by adapting, extending, creating when necessary anddeploying visualization and data understanding technologies for ourscience stakeholders. Using an organizational model as a Visualizationand Analytics Center for Enabling Technologies (VACET), we are wellpositioned to be responsive to the needs of a diverse set of scientificstakeholders in a coordinated fashion using a range of visualization,mathematics, statistics, computer and computational science and datamanagement technologies.

Microfluidic reactors are investigated as a mechanism tocontrol the growth of semiconductor nanocrystals and characterize thestructural evolution of colloidal quantum dots. Due to their shortdiffusion lengths, low thermal masses, and predictable fluid dynamics,microfluidic devices can be used to quickly and reproducibly alterreaction conditions such as concentration, temperature, and reactiontime, while allowing for rapid reagent mixing and productcharacterization. These features are particularly useful for colloidalnanocrystal reactions, which scale poorly and are difficult to controland characterize in bulk fluids. To demonstrate the capabilities ofnanoparticle microreactors, a size series of spherical CdSe nanocrystalswas synthesized at high temperature in a continuous-flow, microfabricatedglass reactor. Nanocrystal diameters are reproducibly controlled bysystematically altering reaction parameters such as the temperature,concentration, and reaction time. Microreactors with finer control overtemperature and reagent mixing were designed to synthesize nanoparticlesof different shapes, such as rods, tetrapods, and hollow shells. The twomajor challenges observed with continuous flow reactors are thedeposition of particles on channel walls and the broad distribution ofresidence times that result from laminar flow. To alleviate theseproblems, I designed and fabricated liquid-liquid segmented flowmicroreactors in which the reaction precursors are encapsulated inflowing droplets suspended in an immiscible carrier fluid. The synthesisof CdSe nanocrystals in such microreactors exhibited reduced depositionand residence time …

This Washington Closure Hanford science and technology (S&T) plan documents the activities associated with providing S&T support to the River Corridor Closure Project for fiscal year 2006.

In the quest for precision cosmology, one must ensure that the cosmology is accurate as well. We discuss figures of merit for determining from observations whether the dark energy is a cosmological constant or dynamical, with special attention to the best determined equation of state value, at the ``pivot'' or decorrelation redshift. We show this is not necessarily the best lever on testing consistency with the cosmological constant, and moreover is subject to bias. The standard parametrization of w(a)=w_0+w_a(1-a) by contrast is quite robust, as tested by extensions to higher order parametrizations and modified gravity. Combination of complementary probes gives strong immunization against inaccurate, but precise, cosmology.