CUDA-for-clusters: a system for efficient execution of CUDA kernels on multi-core clusters

Rapid advancements in multi-core processor architectures coupled with low-cost, low-latency, high-bandwidth interconnects have made clusters of multi-core machines a common computing resource. Unfortunately, writing good parallel programs that efficiently utilize all the resources in such a cluster is still a major challenge. Various programming languages have been proposed as a solution to this problem, but are yet to be adopted widely to run performance-critical code mainly due to the relatively immature software framework and the effort involved in re-writing existing code in the new language. In this paper, we motivate and describe our initial study in exploring CUDA as a programming language for a cluster of multi-cores. We develop CUDA-For-Clusters (CFC), a framework that transparently orchestrates execution of CUDA kernels on a cluster of multi-core machines. The well-structured nature of a CUDA kernel, the growing popularity, support and stability of the CUDA software stack collectively make CUDA a good candidate to be considered as a programming language for a cluster. CFC uses a mixture of source-to-source compiler transformations, a work distribution runtime and a light-weight software distributed shared memory to manage parallel executions. Initial results on running several standard CUDA benchmark programs achieve impressive speedups of up to 7.5X on a cluster with 8 nodes, thereby opening up an interesting direction of research for further investigation.

Influence of binder solvent on carbon-layer structure in electrical-double-layer capacitors

Porous activated-carbons with a large surface-area have been the most common materials for electrical-double-layer capacitors (EDLCs). These carbons having a wide pore distribution ranges from micropores to macropores in conjunction with a random pore connection that facilitates the high specific-capacitance values. Pore distribution plays a central role in controlling the capacitance value of EDLCs, since electrolyte distribution inside the active material mainly depends on the pore distribution. This has a direct influence on the distribution of resistance and capacitance values within the electrode. As a result, preparation of electrodes remains a vital issue in realising high-performance EDLCs. Generally, carbon materials along with some binders are dispersed into a solvent and coated onto the current collectors. This study examines the role of binder solvents used for the carbon-ink preparation on the microstructure of the electrodes and the consequent performance of the EDLCs. It is observed that the physical properties of the binder solvent namely its dielectric constant, viscosity and boiling point have important role in determining the pore-size distribution as well as the microstructure of electrodes which influence their specific capacitance values.

Vertical electric field stimulated neural cell functionality on porous amorphous carbon electrodes

We demonstrate the efficacy of amorphous macroporous carbon substrates as electrodes to support neuronal cell proliferation and differentiation in electric field mediated culture conditions. The electric field was applied perpendicular to carbon substrate electrode, while growing mouse neuroblastoma (N2a) cells in vitro. The placement of the second electrode outside of the cell culture medium allows the investigation of cell response to electric field without the concurrent complexities of submerged electrodes such as potentially toxic electrode reactions, electro-kinetic flows and charge transfer (electrical current) in the cell medium. The macroporous carbon electrodes are uniquely characterized by a higher specific charge storage capacity (0.2 mC/cm(2)) and low impedance (3.3 k Omega at 1 kHz). The optimal window of electric field stimulation for better cell viability and neurite outgrowth is established. When a uniform or a gradient electric field was applied perpendicular to the amorphous carbon substrate, it was found that the N2a cell viability and neurite length were higher at low electric field strengths (<= 2.5 V/cm) compared to that measured without an applied field (0 V/cm). While the cell viability was assessed by two complementary biochemical assays (MTT and LDH), the differentiation was studied by indirect immunostaining. Overall, the results of the present study unambiguously establish the uniform/gradient vertical electric field based culture protocol to either enhance or to restrict neurite outgrowth respectively at lower or higher field strengths, when neuroblastoma cells are cultured on porous glassy carbon electrodes having a desired combination of electrochemical properties. (C) 2013 Elsevier Ltd. All rights reserved.

Patterned growth and differentiation of neural cells on polymer derived carbon substrates with micro/nano structures in vitro

The growth of neuroblastoma (N2a) and Schwann cells has been explored on polymer derived carbon substrates of varying micro and nanoscale geometries: resorcinol-formaldehyde (RE) gel derived carbon films and electrospun nanofibrous (similar to 200 nm diameter) mat and SU-8 (a negative photoresist) derived carbon micro-patterns. MTT assay and complementary lactate dehydrogenase (LDH) assay established cytocompatibility of RE derived carbon films and fibers over a period of 6 days in culture. The role of length scale of surface patterns in eliciting lineage-specific adaptive response along, across and on the interspacing between adjacent micropatterns (i.e., ``on'', ``across'' and ``off'') has been assayed. Textural features were found to affect 3',5'-cyclic AMP sodium salt-induced neurite outgrowth, over a wide range of length scales: from similar to 200 nm (carbon fibers) to similar to 60 mu m (carbon patterns). Despite their innate randomness, carbon nanofibers promoted preferential differentiation of N2a cells into neuronal lineage, similar to ordered micro-patterns. Our results, for the first time, conclusively demonstrate the potential of RE-gel and SU-8 derived carbon substrates as nerve tissue engineering platforms for guided proliferation and differentiation of neural cells in vitro. (C) 2013 Elsevier Ltd. All rights reserved.

New insights into designing metallacarborane based room temperature hydrogen storage media

Metallacarboranes are promising towards realizing room temperature hydrogen storage media because of the presence of both transition metal and carbon atoms. In metallacarborane clusters, the transition metal adsorbs hydrogen molecules and carbon can link these clusters to form metal organic framework, which can serve as a complete storage medium. Using first principles density functional calculations, we chalk out the underlying principles of designing an efficient metallacarborane based hydrogen storage media. The storage capacity of hydrogen depends upon the number of available transition metal d-orbitals, number of carbons, and dopant atoms in the cluster. These factors control the amount of charge transfer from metal to the cluster, thereby affecting the number of adsorbed hydrogen molecules. This correlation between the charge transfer and storage capacity is general in nature, and can be applied to designing efficient hydrogen storage systems. Following this strategy, a search for the best metallacarborane was carried out in which Sc based monocarborane was found to be the most promising H-2 sorbent material with a 9 wt.% of reversible storage at ambient pressure and temperature. (C) 2013 AIP Publishing LLC.

Experimental evidence for `carbon bonding' in the solid state from charge density analysis

The validity of the newly proposed `carbon bonding', an interaction where a carbon atom acts as an electrophilic site towards a variety of nucleophiles, has been investigated in the solid state. X-ray charge density analysis provides experimental evidence for this hitherto unexplored interaction and unravels its nature and strength.

Design of substrate-integrated lead-carbon hybrid ultracapacitor and experimental validation

Suitability of substrate-integrated lead-carbon hybrid ultracapacitors for low-power back-up applications is studied. A practical application that provides 30 W power back-up to low-power medical gadgets for use in grid-power-deficient rural areas is presented. An ultracapacitor bank is designed for this application and the sizing calculations are provided. Experimental validation and results are also discussed.

Nitrogen deposition: how important is it for global terrestrial carbon uptake?

Global carbon budget studies indicate that the terrestrial ecosystems have remained a large sink for carbon despite widespread deforestation activities. CO2 fertilization, N deposition and re-growth of mid-latitude forests are believed to be key drivers for land carbon uptake. In this study, we assess the importance of N deposition by performing idealized near-equilibrium simulations using the Community Land Model 4.0 (CLM4). In our equilibrium simulations, only 12-17% of the deposited nitrogen is assimilated into the ecosystem and the corresponding carbon uptake can be inferred from a C : N ratio of 20 : 1. We calculate the sensitivity of the terrestrial biosphere for CO2 fertilization, climate warming and N deposition as changes in total ecosystem carbon for unit changes in global mean atmospheric CO2 concentration, global mean temperature and Tera grams of nitrogen deposition per year, respectively. Based on these sensitivities, it is estimated that about 242 PgC could have been taken up by land due to the CO2 fertilization effect and an additional 175 PgC taken up as a result of the increased N deposition since the pre-industrial period. Because of climate warming, the terrestrial ecosystem could have lost about 152 PgC during the same period. Therefore, since pre-industrial times terrestrial carbon losses due to warming may have been more or less compensated by effects of increased N deposition, whereas the effect of CO2 fertilization is approximately indicative of the current increase in terrestrial carbon stock. Our simulations also suggest that the sensitivity of carbon storage to increased N deposition decreases beyond current levels, indicating that climate warming effects on carbon storage may overwhelm N deposition effects in the future.

Correspondence between neutron depolarization and higher order magnetic susceptibility to investigate ferromagnetic clusters in phase separated systems

It is a tough task to distinguish a short-range ferromagnetically correlated cluster-glass phase from a canonical spin-glass-like phase in many magnetic oxide systems using conventional magnetometry measurements. As a case study, we investigate the magnetic ground state of La0.85Sr0.15CoO3, which is often debated based on phase separation issues. We report the results of two samples of La0.85Sr0.15CoO3 (S-1 and S-2) prepared under different conditions. Neutron depolarization, higher harmonic ac susceptibility and magnetic relaxation studies were carried out along with conventional magnetometry measurements to differentiate subtle changes at the microscopic level. There is no evidence of ferromagnetic correlation in the sample S-2 attributed to a spin-glass phase, and this is compounded by the lack of existence of a second order component of higher harmonic ac susceptibility and neutron depolarization. A magnetic relaxation experiment at different temperatures complements the spin glass characteristic in S-2. All these signal a sharp variance when we consider the cluster-glass-like phase (phase separated) in S-1, especially when prepared from an improper chemical synthesis process. This shows that the nonlinear ac susceptibility is a viable tool to detect ferromagnetic clusters such as those the neutron depolarization study can reveal.

Concentration Fluctuations and Segmental Dynamics in Weakly Dynamic Asymmetric Blends in the Presence of Surface-Functionalized Multiwall Carbon Nanotubes

Surface-functionalized multiwall carbon nanotubes (MWCNTs) are incorporated in poly(methyl methacrylate)/styrene acrylonitrile (PMMA/SAN) blends and the pretransitional regime is monitored in situ by melt rheology and dielectric spectroscopy. As the blends exhibit weak dynamic asymmetry, the obvious transitions in the melt rheology due to thermal concentration fluctuations are weak. This is further supported by the weak temperature dependence of the correlation length ( approximate to 10-12 angstrom) in the vicinity of demixing. Hence, various rheological techniques in both the temperature and frequency domains are adopted to evaluate the demixing temperature. The spinodal decomposition temperature is manifested in an increase in the miscibility gap in the presence of MWCNTs. Furthermore, MWCNTs lead to a significant slowdown of the segmental dynamics in the blends. Thermally induced phase separation in the PMMA/SAN blends lead to selective localization of MWCNTs in the PMMA phase. This further manifests itself in a significant increase in the melt conductivity.

Electric field induced ultra-high actuation in a bulk carbon nanotube structure

Electric-field induced nonlinear actuation behavior is demonstrated in a bulk nanotube (CNT) structure under ambient conditions. Completely recoverable and non-degradable actuation over several cycles of electric-field is measured in these structures. A symmetric and polarity independent displacement corresponding up to an axial strain of 14% is measured upon application of a low strength electric field of 4.2 kV/m in the axial direction. However, a much lower strain of similar to 1% is measured in the radial (or, transverse) direction. Furthermore, the electric field induced actuation increases by more than a factor of 2 upon impregnating the CNT cellular structure with copper oxide nano-particles. An electrostriction mechanism, based on the electric-field induced polarization of CNT strands, is proposed to account for the reported actuation behavior. (C) 2013 Elsevier Ltd. All rights reserved.

Carbon nanoflake growth from carbon nanotubes by hot filament chemical vapor deposition

We report the growth of carbon nanoflakes (CNFs) on Si substrate by the hot filament chemical vapor deposition without the substrate bias or the catalyst. CNFs were grown using the single wall carbon nanotubes and the multiwall carbon nanotubes as the nucleation center, in the Ar-rich CH4-H-2-Ar precursor gas mixture with 1% CH4, at the chamber pressure and the substrate temperature of 7.5 Ton and 840 degrees C, respectively. In the H-2-rich condition, CNF synthesis failed due to severe etch-removal of carbon nanotubes (CNTs) while it was successful at the optimized Ar-rich condition. Other forms of carbon such as nano-diamond or mesoporous carbon failed to serve as the nucleation centers for the CNF growth. We proposed a mechanism of the CNF synthesis from the CNTs, which involved the initial unzipping of CNTs by atomic hydrogen and subsequent nucleation and growth of CNFs from the unzipped portion of the graphene layers. (C) 2013 Elsevier Ltd. All rights reserved.

Carbon nanotube based multifunctional flame sensor

Carbon nanotubes (CNT) due to its multifunctional characteristics has been presented as a flame sensor by combining both radiation and chemical sensitivity. Chemical functionalization enhances the sensitivity of CNT sensor toward any chemical modifications that are induced by the flame. Response of the sensor is revealed to be dependent on the measurement direction (longitudinal and transverse) as well as the radiation intensity. A nonlinear relation between the sensitivity and its distance from the source is used to calibrate the intensity of the flame. The present method allows a simpler approach for the flame detection by utilizing a calibration scheme to operate at any particular bias current and tune its sensitivity with respect to any working distance at a particular bias current. (C) 2013 Elsevier B.V. All rights reserved.

Chemical vapor detection using nonlinear electrical properties of carbon nanotube bundles

We demonstrate the electrical transport behavior of carbon nanotubes (CNTs) upon exposure to organic analytes (namely ethanol, benzene, acetone and toluene). The resulting nonlinear current-voltage characteristics revealed a power law dependence of the differential conductivity on the applied bias voltage. Moreover, suppression of differential conductivity at zero bias is found to be dependent on different selective analytes. The power law exponent values have been monitored before, during and after exposure to the chemicals, which revealed a reversible change in the number of electron conducting channels. Therefore, the reduction in the number of conductive paths can be attributed to the interaction of the chemical analyte on the CNT surfaces, which causes a decrease in the differential conductivity of the CNT sample. These results demonstrate chemical selectivity of CNTs due to varying electronic interaction with different chemical analytes.