Search

Artificial neural networks are notoriously power- and time-consuming when implemented on conventional von Neumann computing systems. Consequently, recent years have seen an emergence of research in machine learning hardware that strives to bring memory and computing closer together. A popular approach is to realise artificial neural networks in hardware by implementing their synaptic weights using memristive devices. However, various device- and system-level non-idealities usually prevent these physical implementations from achieving high inference accuracy. We suggest applying a well-known concept in computer science -- committee machines -- in the context of memristor-based neural networks. Using simulations and experimental data from three different types of memristive devices, we show that committee machines employing ensemble averaging can successfully increase inference accuracy in physically implemented neural networks that suffer from faulty devices, device-to-device variability, random telegraph noise and line resistance. Importantly, we demonstrate that the accuracy can be improved even without increasing the total number of memristors., Comment: 22 pages, 18 figures, 4 tables

MXenes are a rapidly growing family of 2D materials that exhibit a highly versatile structure and composition, allowing for significant tuning of the material properties. These properties are, however, ultimately limited by the surface terminations, which are typically a mixture of species, including F and O that are inherent to the MXene processing. Other and robust terminations are lacking. Here, we apply high-resolution scanning transmission electron microscopy (STEM), corresponding image simulations and first-principles calculations to investigate the surface terminations on MXenes synthesized from MAX phases through Lewis acidic melts. The results show that atomic Cl terminates the synthesized MXenes, with mere residual presence of other termination species. Furthermore, in situ STEM-electron energy loss spectroscopy (EELS) heating experiments show that the Cl terminations are stable up to 750 {\deg}C. Thus, we present an attractive new termination that widely expands the MXenes functionalization space and enable new applications., Comment: 14 pages, 4 figures, 5 pages supplemental info

Spectator fragments resulting from relativistic heavy ion collisions, consisting of single protons and neutrons along with groups of stable nuclear fragments up to Nitrogen (Z=7), are measured in PHOBOS. These fragments are observed in Au+Au (sqrt(sNN)=19.6 GeV) and Cu+Cu (22.4 GeV) collisions at high pseudorapidity ($\eta$). The dominant multiply-charged fragment is the tightly bound Helium ($\alpha$), with Lithium, Beryllium, and Boron all clearly seen as a function of collision centrality and pseudorapidity. We observe that in Cu+Cu collisions, it becomes much more favorable for the $\alpha$ fragments to be released than Lithium. The yields of fragments approximately scale with the number of spectator nucleons, independent of the colliding ion. The shapes of the pseudorapidity distributions of fragments indicate that the average deflection of the fragments away from the beam direction increases for more central collisions. A detailed comparison of the shapes for $\alpha$ and Lithium fragments indicates that the centrality dependence of the deflections favors a scaling with the number of participants in the collision., Comment: To be submitted to PRC 17 pages, 19 figures

Precise measurements of the proton electromagnetic form factor ratio $R = \mu_p G_E^p/G_M^p$ using the polarization transfer method at Jefferson Lab have revolutionized the understanding of nucleon structure by revealing the strong decrease of $R$ with momentum transfer $Q^2$ for $Q^2 \gtrsim 1$ GeV$^2$, in strong disagreement with previous extractions of $R$ from cross section measurements. In particular, the polarization transfer results have exposed the limits of applicability of the one-photon-exchange approximation and highlighted the role of quark orbital angular momentum in the nucleon structure. The GEp-II experiment in Jefferson Lab's Hall A measured $R$ at four $Q^2$ values in the range 3.5 GeV$^2 \le Q^2 \le 5.6$ GeV$^2$. A possible discrepancy between the originally published GEp-II results and more recent measurements at higher $Q^2$ motivated a new analysis of the GEp-II data. This article presents the final results of the GEp-II experiment, including details of the new analysis, an expanded description of the apparatus and an overview of theoretical progress since the original publication. The key result of the final analysis is a systematic increase in the results for $R$, improving the consistency of the polarization transfer data in the high-$Q^2$ region. This increase is the result of an improved selection of elastic events which largely removes the systematic effect of the inelastic contamination, underestimated by the original analysis., Comment: 30 pages, 17 figures, 4 tables. Final version to appear in Phys. Rev. C

Pseudorapidity distributions of charged particles emitted in $Au+Au$, $Cu+Cu$, $d+Au$, and $p+p$ collisions over a wide energy range have been measured using the PHOBOS detector at RHIC. The centrality dependence of both the charged particle distributions and the multiplicity at midrapidity were measured. Pseudorapidity distributions of charged particles emitted with $|\eta|<5.4$, which account for between 95% and 99% of the total charged-particle emission associated with collision participants, are presented for different collision centralities. Both the midrapidity density, $dN_{ch}/d\eta$, and the total charged-particle multiplicity, $N_{ch}$, are found to factorize into a product of independent functions of collision energy, $\sqrt{s_{_{NN}}}$, and centrality given in terms of the number of nucleons participating in the collision, $N_{part}$. The total charged particle multiplicity, observed in these experiments and those at lower energies, assumes a linear dependence of $(\ln s_{_{NN}})^2$ over the full range of collision energy of $\sqrt{s_{_{NN}}}$=2.7-200 GeV., Comment: 25 pages, 29 figures, 8 tables

The centrality dependence of the midrapidity charged-particle multiplicity density ($|\eta|$$<$1) is presented for Au+Au and Cu+Cu collisions at RHIC over a broad range of collision energies. The multiplicity measured in the Cu+Cu system is found to be similar to that measured in the Au+Au system, for an equivalent N$_{\rm part}$, with the observed factorization in energy and centrality still persistent in the smaller Cu+Cu system. The extent of the similarities observed for bulk particle production is tested by a comparative analysis of the inclusive transverse momentum distributions for Au+Au and Cu+Cu collisions near midrapidity. It is found that, within the uncertainties of the data, the ratio of yields between the various energies for both Au+Au and Cu+Cu systems are similar and constant with centrality, both in the bulk yields as well as a function of p$_{\rm T}$, up to at least 4 GeV/$c$. The effects of multiple nucleon collisions that strongly increase with centrality and energy appear to only play a minor role in bulk and intermediate transverse momentum particle production., Comment: Submitted for publication in Phys. Rev. C, 5 pages, 3 figures

Charged particle pseudorapidity distributions are presented from the PHOBOS experiment at RHIC, measured in Au+Au and Cu+Cu collisions at sqrt{s_NN}=19.6, 22.4, 62.4, 130 and 200 GeV, as a function of collision centrality. The presentation includes the recently analyzed Cu+Cu data at 22.4 GeV. The measurements were made by the same detector setup over a broad range in pseudorapidity, |eta|<5.4, allowing for a reliable systematic study of particle production as a function of energy, centrality and system size. Comparing Cu+Cu and Au+Au results, we find that the total number of produced charged particles and the overall shape (height and width) of the pseudorapidity distributions are determined by the number of nucleon participants, N_part. Detailed comparisons reveal that the matching of the shape of the Cu+Cu and Au+Au pseudorapidity distributions over the full range of eta is better for the same N_part/2A value than for the same N_part value, where A denotes the mass number. In other words, it is the geometry of the nuclear overlap zone, rather than just the number of nucleon participants that drives the detailed shape of the pseudorapidity distribution and its centrality dependence., Comment: 5 pages, 4 figures. Presented at the 20th International Conference on Nucleus-Nucleus Collisions (Quark Matter 2008), Jaipur, Rajasthan, India, 4-10 February 2008

We present first results on event-by-event elliptic flow fluctuations in nucleus-nucleus collisions corrected for effects of non-flow correlations where the magnitude of non-flow correlations has been independently measured in data. Over the measured range in centrality, we see large relative fluctuations of 25-50%. The results are consistent with predictions from both color glass condensate and Glauber type initial condition calculations of the event-by-event participant eccentricity fluctuations., Comment: 4 pages, 3 figures, Quark Matter 2008 Proceedings

We present a measurement of the spin-dependent cross sections for the \vec{^3He}(\vec{e},e')X} reaction in the quasielastic and resonance regions at four-momentum transfer 0.1 < Q^2< 0.9 GeV^2. The spin-structure functions have been extracted and used to evaluate the nuclear Burkhardt--Cottingham and extended GDH sum rules for the first time. Impulse approximation and exact three-body Faddeev calculations are also compared to the data in the quasielastic region., Comment: Final version

We present the first measurements of the pseudorapidity distribution of primary charged particles in Cu+Cu collisions as a function of collision centrality and energy, \sqrtsnn = 22.4, 62.4 and 200 GeV, over a wide range of pseudorapidity, using the PHOBOS detector. Making a global comparison of Cu+Cu and Au+Au results, we find that the total number of produced charged particles and the rough shape (height and width) of the pseudorapidity distributions are determined by the number of nucleon participants. More detailed studies reveal that a more precise matching of the shape of the Cu+Cu and Au+Au pseudorapidity distributions over the full range of pseudorapidity occurs for the same Npart/2A value rather than the same Npart value. In other words, it is the collision geometry rather than just the number of nucleon participants that drives the detailed shape of the pseudorapidity distribution and its centrality dependence at RHIC energies., Comment: Submitted to Physical Review Letters

Transverse momentum spectra of pions, kaons, protons and antiprotons from Au+Au collisions at sqrt(s_(NN)) = 62.4 GeV have been measured by the PHOBOS experiment at the Relativistic Heavy Ion Collider at Brookhaven National Laboratory. The identification of particles relies on three different methods: low momentum particles stopping in the first detector layers; the specific energy loss (dE/dx) in the silicon Spectrometer, and Time-of-Flight measurement. These methods cover the transverse momentum ranges 0.03-0.2, 0.2-1.0 and 0.5-3.0 GeV/c, respectively. Baryons are found to have substantially harder transverse momentum spectra than mesons. The pT region in which the proton to pion ratio reaches unity in central Au+Au collisions at sqrt(s_(NN)) = 62.4 GeV fits into a smooth trend as a function of collision energy. At low transverse momenta, the spectra exhibit a significant deviation from transverse mass scaling, and when the observed particle yields at very low pT are compared to extrapolations from higher pT, no significant excess is found. By comparing our results to Au+Au collisions at sqrt(s_(NN)) = 200 GeV, we conclude that the net proton yield at midrapidity is proportional to the number of participant nucleons in the collision., Comment: 15 pages, 12 figures. Submitted to Phys. Rev. C