arXiv subjects

Holger Fehske

Publications and source records attributed to Holger Fehske.

At least 109 records · Page 6Linked to original sources

arXiv2012

Sparse matrix-vector multiplication on GPGPU clusters: A new storage format and a scalable implementation

Sparse matrix-vector multiplication (spMVM) is the dominant operation in many sparse solvers. We investigate performance properties of spMVM with matrices of various sparsity patterns on the nVidia "Fermi" class of GPGPUs. A new "padded jagged diagonals storage" (pJDS) format is proposed which may substantially reduce the memory overhead intrinsic to the widespread ELLPACK-R scheme. In our test scenarios the pJDS format cuts the overall spMVM memory footprint on the GPGPU by up to 70%, and achieves 95% to 130% of the ELLPACK-R performance. Using a suitable performance model we identify performance bottlenecks on the node level that invalidate some types of matrix structures for efficient multi-GPGPU parallelization. For appropriate sparsity patterns we extend previous work on distributed-memory parallel spMVM to demonstrate a scalable hybrid MPI-GPGPU code, achieving efficient overlap of communication and computation.

cs.DC↗

arXiv2012

Charge carrier localisation in disordered graphene nanoribbons

We study the electronic properties of actual-size graphene nanoribbons subjected to substitutional disorder particularly with regard to the experimentally observed metal-insulator transition. Calculating the local, mean and typical density of states, as well as the time-evolution of the particle density we comment on a possible disorder-induced localisation of charge carriers at and close to the Dirac point within a percolation transition scenario.

cond-mat.str-el↗

arXiv2012

Resonant charge transfer at dielectric surfaces

We report on the theoretical description of secondary electron emission due to resonant charge transfer occurring during the collision of metastable nitrogen molecules with dielectric surfaces. The emission is described as a two step process consisting of electron capture to form an intermediate shape resonance and subsequent electron emission by decay of this ion, either due to its natural life time or its interaction with the surface. The electron capture is modeled using the Keldysh Green's function technique and the negative ion decay is described by a combination of the Keldysh technique and a rate equation approach. We find the resonant capture of electrons to be very efficient and the natural decay to be clearly dominating over the surface-induced decay. Secondary electron emission coefficients are calculated for aluminum oxide, magnesium oxide, silicon oxide, and diamond at several kinetic energies of the projectile. With the exception of magnesium oxide the coefficients turn out to be of the order of 0.1 over the whole range of kinetic energies. This rather large value is a direct consequence of the shape resonance acting as a relay state for electron emission.

physics.plasm-ph↗

arXiv2011

Metal-to-insulator transition and electron-hole puddle formation in disordered graphene nanoribbons

The experimentally observed metal-to-insulator transition in hydrogenated graphene is numerically confirmed for actual sized graphene samples and realistic impurity concentrations. The eigenstates of our tight-binding model with substitutional disorder corroborate the formation of electron-hole-puddles with characteristic length scales comparable to the ones found in experiments. The puddles cause charge inhomogeneities and tend to suppress Anderson localization. Even though, monitoring the charge carrier quantum dynamics and performing a finite-size scaling of the local density of states distribution, we find strong evidence for the existence of localized states in graphene nanoribbons with short-range but also correlated long-range disorder.

cond-mat.str-el↗

arXiv2011

Aharonov-Bohm interference for a hole in a two-dimensional Ising antiferromagnet in a transverse magnetic field

We show that a proper consideration of the contribution of Trugman loops leads to a fairly low effective mass for a hole moving in a square lattice Ising antiferromagnet, if the bare hopping and the exchange energy scales are comparable. This contradicts the general view that because of the absence of spin fluctuations, this effective mass must be extremely large. Moreover, in the presence of a transverse magnetic field, we show that the effective hopping integrals acquire an unusual dependence on the magnetic field, through Aharonov-Bohm interference, in addition to significant retardation effects. The effect of the Aharonov-Bohm interference on the cyclotron frequency (for small magnetic fields) and the Hofstadter butterfly (for large magnetic fields) is analyzed.

cond-mat.str-el↗

arXiv2011

Hybrid-parallel sparse matrix-vector multiplication with explicit communication overlap on current multicore-based systems

We evaluate optimized parallel sparse matrix-vector operations for several representative application areas on widespread multicore-based cluster configurations. First the single-socket baseline performance is analyzed and modeled with respect to basic architectural properties of standard multicore chips. Beyond the single node, the performance of parallel sparse matrix-vector operations is often limited by communication overhead. Starting from the observation that nonblocking MPI is not able to hide communication cost using standard MPI implementations, we demonstrate that explicit overlap of communication and computation can be achieved by using a dedicated communication thread, which may run on a virtual core. Moreover we identify performance benefits of hybrid MPI/OpenMP programming due to improved load balancing even without explicit communication overlap. We compare performance results for pure MPI, the widely used "vector-like" hybrid programming strategies, and explicit overlap on a modern multicore-based cluster and a Cray XE6 system.

cs.DC↗

arXiv2011

Excitonic resonances in the 2D extended Falicov-Kimball model

Using the projector-based renormalization method we investigate the formation of the excitonic insulator phase in the two-dimensional (2D) spinless Falicov-Kimball model with dispersive $f$ electrons and address the existence of excitonic bound states at high temperatures on the semiconductor side of the semimetal-semiconductor transition. To this end we calculate the imaginary part of the dynamical electron-hole pair susceptibility and analyze the wave-vector and energy dependence of excitonic resonances emerging in the band gap. We thereby confirm the existence of the exciton insulator and its exciton environment within a generic two-band lattice model with local Coulomb attraction.

cond-mat.str-el↗

arXiv2011

Auger de-excitation of metastable molecules at metallic surfaces

We study secondary electron emission from metallic surfaces due to Auger de-excitation of diatomic metastable molecules. Our approach is based on an effective model for the two active electrons involved in the process -- a molecular electron described by a linear combination of atomic orbitals when it is bound and a two-center Coulomb wave when it is not and a metal electron described by the eigenfunctions of a step potential -- and employs Keldysh Green's functions. Solving the Dyson equation for the retarded Green's function by exponential resummation we are able to treat time-nonlocal self-energies and to avoid the wide-band approximation.Results are presented for the de-excitation of \NitrogenDominantMetastableState\ on aluminum and tungsten and discussed in view of previous experimental and theoretical investigations. We find quantitative agreement with experimental data for tungsten indicating that the effective model captures the physics of the process quite well. For aluminum we predict secondary electron emission due to Auger de-excitation to be one to two orders of magnitude smaller than the one found for resonant charge-transfer and subsequent auto-detachment.

cond-mat.other↗

arXiv2011

Dynamic charge correlations near the Peierls transition

The quantum phase transition between a repulsive Luttinger liquid and an insulating Peierls state is studied in the framework of the one-dimensional spinless Holstein model. We focus on the adiabatic regime but include the full quantum dynamics of the phonons. Using continuous-time quantum Monte Carlo simulations, we track in particular the dynamic charge structure factor and the single-particle spectrum across the transition. With increasing electron-phonon coupling, the dynamic charge structure factor reveals the emergence of a charge gap, and a clear signature of phonon softening at the zone boundary. The single-particle spectral function evolves continuously across the transition. Hybridization of the charge and phonon modes of the Luttinger liquid description leads to two modes, one of which corresponds to the coherent polaron band. This band acquires a gap upon entering the Peierls phase, whereas the other mode constitutes the incoherent, high-energy spectrum with backfolded shadow bands. Coherent polaronic motion is a direct consequence of quantum lattice fluctuations. In the strong-coupling regime, the spectrum is described by the static, mean-field limit. Importantly, whereas finite electron density in general leads to screening of polaron effects, the latter reappear at half filling due to charge ordering and lattice dimerization.

cond-mat.str-el↗

arXiv2011

Plasma walls beyond the perfect absorber approximation for electrons

Plasma walls accumulate electrons more efficiently than ions leading to wall potentials which are negative with respect to the plasma potential. Theoretically, walls are usually treated as perfect absorber for electrons and ions implying perfect sticking of the particles to the wall and infinitely long desorption times for particles stuck to the wall. For electrons we question the perfect absorber model and calculate, specifically for a planar dielectric wall, the electron sticking coefficient $s_e$ and the electron desorption time $τ_e$. For the uncharged wall we find $s_e\ll 1$ and $τ_e\approx 10^{-4}s$. Thus, in the early stage of the build-up of the wall potential, when the wall is essentially uncharged, the wall is not a perfect absorber for electrons. For the charged wall we find $τ_e^{-1}\approx 0$. Thus, $τ_e$ approaches the perfect absorber value. But $s_e$ is still only of the order of $10^{-1}$. Calculating $s_e$ as a function of the wall potential and combining this expression with the quasi-stationary balance equations for the electron and ion surface densities we find the selfconsistent wall potential, including surface effects, to be 30% of the perfect absorber value.

physics.plasm-ph↗

arXiv2010

Parallel sparse matrix-vector multiplication as a test case for hybrid MPI+OpenMP programming

We evaluate optimized parallel sparse matrix-vector operations for two representative application areas on widespread multicore-based cluster configurations. First the single-socket baseline performance is analyzed and modeled with respect to basic architectural properties of standard multicore chips. Going beyond the single node, parallel sparse matrix-vector operations often suffer from an unfavorable communication to computation ratio. Starting from the observation that nonblocking MPI is not able to hide communication cost using standard MPI implementations, we demonstrate that explicit overlap of communication and computation can be achieved by using a dedicated communication thread, which may run on a virtual core. We compare our approach to pure MPI and the widely used "vector-like" hybrid programming strategy.

cs.PF↗

arXiv2010

Stability of edge states and edge magnetism in graphene nanoribbons

We critically discuss the stability of edge states and edge magnetism in zigzag edge graphene nanoribbons (ZGNRs). We point out that magnetic edge states might not exist in real systems, and show that there are at least three very natural mechanisms - edge reconstruction, edge passivation, and edge closure - which dramatically reduce the effect of edge states in ZGNRs or even totally eliminate them. Even if systems with magnetic edge states could be made, the intrinsic magnetism would not be stable at room temperature. Charge doping and the presence of edge defects further destabilize the intrinsic magnetism of such systems.

cond-mat.mtrl-sci↗

arXiv2010

Phase separation of multicomponent excitonic Bose-Einstein condensates

For the observation of Bose-Einstein condensation, excitons in cuprous oxide are regarded as promising candidates due to their large binding energy and long lifetime. High particle densities may be achieved by entrapment in a stress induced potential. We consider a multi-component gas of interacting para- and orthoexcitons in cuprous oxide confined in a three-dimensional potential trap. Based on the Hartree-Fock-Bogoliubov theory, we calculate density profiles as well as decay luminescence spectra which exhibit signatures of the separation of the Bose-condensed phases.

cond-mat.quant-gas↗

arXiv2010

Momentum average approximation for models with boson-modulated hopping: Role of closed loops in the dynamical generation of a finite quasiparticle mass

We generalize the momentum average approximation to study the properties of single polarons in models with boson affected hopping, where the fermion-boson scattering depends explicitly on both the fermion's and the boson's momentum. As a specific example, we investigate the Edwards fermion-boson model in both one and two dimensions. In one dimension, this allows us to compare our results with exact diagonalization results, to validate the accuracy of our approximation. The generalization to two-dimensional lattices allows us to calculate the polaron's quasiparticle weight and dispersion throughout the Brillouin zone and to demonstrate the importance of Trugman loops in generating a finite effective mass even when the free fermion has an infinite mass.

cond-mat.mtrl-sci↗

arXiv2010

Spectral signatures of the BCS-BEC crossover in the excitonic insulator phase of the extended Falicov-Kimball model

We explore the spontaneous formation of an excitonic insulator state at the semimetal-semiconductor transition of mixed-valence materials in the framework of the spinless Falicov-Kimball model with direct $f$-$f$ electron hopping. Adapting the projector-based renormalization method, we obtain a set of renormalization differential equations for the extended Falicov-Kimball model parameters and finally derive analytical expressions for the order parameter, as well as for the renormalized $c$- and $f$-electron dispersions, momentum distributions, and wave-vector resolved single-particle spectral functions. Our numerical results proved the valence transition picture, related to the appearance of the excitonic insulator phase, in the case of overlapping $c$ and $f$ bands. Thereby the photoemission spectra show significant differences between the weak-to-intermediate and intermediate-to-strong Coulomb attraction regimes, indicating a BCS-BEC transition of the excitonic condensate.

cond-mat.str-el↗

arXiv2010

Polarons and slow quantum phonons

We describe the formation and properties of Holstein polarons in the entire parameter regime. Our presentation focuses on the polaron mass and radius, which we obtain with an improved numerical technique. It is based on the combination of variational exact diagonalization with an improved construction of phonon states, providing results even for the strong coupling adiabatic regime. In particular we can describe the formation of large and heavy adiabatic polarons. A comparison of the polaron mass for the one and three dimensional situation explains how the different properties in the static oscillator limit determine the behavior in the adiabatic regime. The transport properties of large and small polarons are characterized by the f-sum rule and the optical conductivity. Our calculations are approximation-free and have negligible numerical error. This allows us to give a conclusive and impartial description of polaron formation. We finally discuss the implications of our results for situations beyond the Holstein model.

cond-mat.str-el↗

arXiv2010

Distribution of the local density of states as a criterion for Anderson localization: Numerically exact results for various lattices in two and three dimensions

Numerical approaches to Anderson localization face the problem of having to treat large localization lengths while being restricted to finite system sizes. We show that by finite-size scaling of the probability distribution of the local density of states (LDOS) this long-standing problem can be overcome. To this end we reexamine the approach, propose numerical refinements, and apply it to study the dependence of the distribution of the LDOS on the dimensionality and coordination number of the lattice. Particular attention is given to the graphene lattice. We show that the system-size dependence of the LDOS distribution is indeed an unambiguous sign of Anderson localization, irrespective of the dimension and lattice structure. The numerically exact LDOS data obtained by us agree with a log-normal distribution over up to ten orders of magnitude and thereby fulfill a nontrivial symmetry relation previously derived for the non-linear $σ$-model.

cond-mat.str-el↗

arXiv2009

Performance limitations for sparse matrix-vector multiplications on current multicore environments

The increasing importance of multicore processors calls for a reevaluation of established numerical algorithms in view of their ability to profit from this new hardware concept. In order to optimize the existent algorithms, a detailed knowledge of the different performance-limiting factors is mandatory. In this contribution we investigate sparse matrix-vector multiplication, which is the dominant operation in many sparse eigenvalue solvers. Two conceptually different storage schemes and computational kernels have been conceived in the past to target cache-based and vector architectures, respectively. Starting from a series of microbenchmarks we apply the gained insight on optimized sparse MVM implementations, whose serial and OpenMP-parallel performance we review on state-of-the-art multicore systems.

cs.PF↗