Proximal minimizations with D-functions and the massively parallel solution of linear network programs

This paper discusses the massively parallel solution of linear network programs. It integrates the general algorithmic framework of proximal minimization with D-functions (PMD) with primal-dual row-action algorithms. Three alternative algorithmic schemes are studied: quadratic proximal point, entropic proximal point, and least 2-norm perturbations. Each is solving a linear network problem by solving a sequence of nonlinear approximations. The nonlinear subproblems decompose for massively parallel computing. The three algorithms are implemented on a Connection Machine CM-2 with up to 32K processing elements, and problems with up to 16 million variables are solved. A comparison of the three algorithms establishes their relative efficiency. Numerical experiments also establish the best internal tactics which can be used when implementing proximal minimization algorithms. Finally, the new algorithms are compared with an implementation of the network simplex algorithm executing on a CRAY Y-MP vector supercomputer.     

Divisibility of the stable Miller-Morita-Mumford classes

We determine the sublattice generated by the Miller-Morita- Mumford classes in the torsion free quotient of the integral cohomology ring of the stable mapping class group. We further decide when the mod reductions vanish.

     

Lateralization and frequency selectivity in normal and impaired hearing

The onset-time difference delta T required to lateralize a 30-ms bifrequency tone burst toward the leading ear was measured as a function of the frequency difference delta F between the tone in the left ear and the tone in the right ear. At center frequencies of 0.5 and 4 kHz, four normal listeners tested at 80 and 100 dB SPL had delta Ts that were relatively constant at subcritical delta Fs, but increased at delta Fs wider than a critical band. At 1 kHz, delta T increased with delta F even at subcritical delta Fs. Ten listeners with cochlear impairments were tested at 100 dB SPL. Seven had normal delta Ts at 4 kHz, despite hearing losses between 50 and 70 dB. At 0.5 and 1 kHz, mildly impaired listeners had nearly normal lateralization functions, whereas more severely imparied listeners had very large delta Ts and no frequency selectivity. These and other findings indicate that listeners even with moderate to severe hearing losses can lateralize normally on the basis of interaural differences in onset envelope, but not on the basis of temporal differences in the fine structure.     

Temporal integration and multiple looks, revisited: weights as a function of time

This study tests the hypothesis that temporal integration for detection of tone bursts with various durations can be explained by optimally combining multiple looks of brief signal segments whose contribution to detection increases over time. Detectability was measured for signals consisting of six consecutive 25-ms, 1-kHz tone pulses presented in a 50-Hz-wide masker or in maskers consisting of seven 50-Hz-wide noises, one critical band apart, with either coherent or incoherent envelopes. The level of each signal pulse varied randomly around masked threshold according to a Gaussian distribution. The slopes of conditional psychometric functions--plotted in terms of d'2 as a function of the squared signal-pulse intensity for pulses in a particular temporal position--yielded estimates of the contribution to detection provided by each pulse. Results for three normal listeners showed a small, but significant, effect of the temporal location of the pulse. Multiple-looks predictions of temporal-integration functions based on the measured weights and on measured psychometric functions were compared to measured temporal-integration functions. For the single-band and incoherent maskers, the predicted temporal-integration slopes were reasonably consistent with those measured, but for the coherent masker it was not. Whereas no current theory can explain the very steep temporal-integration functions obtained in the coherent masker, the present results are not inconsistent with the multiple-looks hypothesis as an explanation for the decrease in threshold with increasing duration of signals presented in random maskers.     

Monte Carlo Code ODIN and Energy Density Analysis in Relativistic Heavy Ion Collisions

The energy density reached in relativistic heavy ion collisions is analyzed based on the Monte Carlo code ODIN. The effect of different parameters in ODIN code on the energy density is investigated. The results show that the actually reached energy density is lower than Bjorken's estimation due to the finite collision time.     

Probing Degradation in Lithium Ion Batteries with On-Chip Electrochemistry Mass Spectrometry**

The rapid uptake of lithium ion batteries (LIBs) for large scale electric vehicle and energy storage applications requires a deeper understanding of the degradation mechanisms. Capacity fade is due to the complex interplay between phase transitions, electrolyte decomposition and transition metal dissolution; many of these poorly understood parasitic reactions evolve gases as a side product. Here we present an on-chip electrochemistry mass spectrometry method that enables ultra-sensitive, fully quantified and time resolved detection of volatile species evolving from an operating LIB. The technique's electrochemical performance and mass transport is described by a finite element model and then experimentally used to demonstrate the variety of new insights into LIB performance. We show the versatility of the technique, including (a) observation of oxygen evolving from a LiNiMnCoO₂ cathode and (b) the solid electrolyte interphase formation reaction on graphite in a variety of electrolytes, enabling the deconvolution of lithium inventory loss (c) the first direct evidence, by virtue of the improved time resolution of our technique, that carbon dioxide reduction to ethylene takes place in a lithium ion battery. The emerging insight will guide and validate battery lifetime models, as well as inform the design of longer lasting batteries.     

Divergent and Multi-Stage Photoisomerization of Four-Coordinated Boron Compounds with a Naphthyl-Pyridyl/Thiazolyl Backbone

Examination of the photoreactivity of a new class of N,C-chelate organoboron compounds, including a series of unsymmetrically substituted boron molecules, B(naph-pyridyl)(Ar¹)(Ar²) and B(naph-thiazolyl)(Ar¹)(Ar²), led to the discovery of new and divergent photothermal isomerization phenomena. These include the clean and regioselective photoisomerization by unsymmetrical boron, forming borepin isomers, some of which further isomerize to the corresponding boratanorcaradiene diastereomer pairs as a result of the generation of two chiral centers. Significantly, the boratanorcaradienes involving a 3-thienyl substituent on boron were found to thermally convert to BN-fluoranthene annulated borapentalene via an unprecedented reversible boratacyclopropane–boratacyclopentene rearrangement. Changing the pyridyl donor to a thiazolyl donor on the boron was found to provide the B(naph-thiazolyl)(Mes)₂ compounds with a distinct new photoisomerization pathway—instead of borepin, forming new blue fluorescent polycyclic azaborinine species. This work illustrates the richness and complexity of boron photochemistry.     

Effect of masker-fringe onset asynchrony on overshoot (L)

The study examines how overshoot is influenced by masker-signal onset asynchrony when the masker contains frequencies above or below the signal frequency. Masked thresholds were measured for a 2-ms tone at 5 kHz. The measurements were made in a reference condition with a narrow center-band (CB) noise masker (4590-5464 Hz), and in conditions with either a low-fringe (1900-4590 Hz) or a high-fringe noise band (5500-11 000 Hz) added to the CB. The signal always came on 2 ms after the onset of the CB. The time interval, between fringe and signal onsets varied from -98 ms (signal onset before fringe onset) to +502 ms (signal onset after fringe onset). Results show that overshoot is largest, 8-11 dB, for a high fringe with onset occurring between 8 ms before the signal onset and 12 ms after it. Thus, pronounced overshoot is observed even when the high fringe is gated on after the signal's offset. Low fringes produce no more than 4 dB of overshoot, much less than high fringes produce. The finding of pronounced overshoot elicited by a high fringe presented shortly after the end of the signal suggests that overshoot is governed, at least in part, by central mechanisms.