arXiv subjects

Markus Müller

Publications and source records attributed to Markus Müller.

At least 127 records · Page 7Linked to original sources

arXiv2018

An exactly solvable model for Dynamic Nuclear polarization

We introduce a solvable model of driven fermions that elucidates the role of the localization transition in driven disordered magnets, as used in the context of dynamic nuclear polarization. Instead of spins, we study a set of non-interacting fermions that are coupled locally to nuclear spins and tend to hyperpolarize them. The induced hyperpolarization is a fingerprint of the driven steady state of the fermions, which undergo an Anderson Localization (AL) transition upon increasing the disorder. Our central result is that the maximal hyperpolarization level is always found close to the localization transition. In the limit of small nuclear moments the maximum is pinned to the transition, and the hyperpolarization is strongly enhanced by multi-fractal correlations in the critical state of the nearly localized driven system, its magnitude reflecting multi-fractal scaling.

cond-mat.str-el↗

arXiv2018

Neural Language Codes for Multilingual Acoustic Models

Multilingual Speech Recognition is one of the most costly AI problems, because each language (7,000+) and even different accents require their own acoustic models to obtain best recognition performance. Even though they all use the same phoneme symbols, each language and accent imposes its own coloring or "twang". Many adaptive approaches have been proposed, but they require further training, additional data and generally are inferior to monolingually trained models. In this paper, we propose a different approach that uses a large multilingual model that is \emph{modulated} by the codes generated by an ancillary network that learns to code useful differences between the "twangs" or human language. We use Meta-Pi networks to have one network (the language code net) gate the activity of neurons in another (the acoustic model nets). Our results show that during recognition multilingual Meta-Pi networks quickly adapt to the proper language coloring without retraining or new data, and perform better than monolingually trained networks. The model was evaluated by training acoustic modeling nets and modulating language code nets jointly and optimize them for best recognition performance.

cs.CL↗

arXiv2018

Overcoming correlation fluctuations in two-photon interference experiments with differently bright and independently blinking remote quantum emitters

As a fundamental building block for quantum computation and communication protocols, the correct verification of the two-photon interference (TPI) contrast between two independent quantum light sources is of utmost importance. Here, we experimentally demonstrate how frequently present blinking dynamics and changes in emitter brightness critically affect the Hong-Ou-Mandel-type (HOM) correlation histograms of remote TPI experiments measured via the commonly utilized setup configuration. We further exploit this qualitative and quantitative explanation of the observed correlation dynamics to establish an alternative interferometer configuration, which is overcoming the discussed temporal fluctuations, giving rise to an error-free determination of the remote TPI visibility. We prove full knowledge of the obtained correlation by reproducing the measured correlation statistics via Monte-Carlo simulations. As exemplary system, we make use of two pairs of remote semiconductor quantum dots, however, the same conclusions apply for TPI experiments with flying qubits from any kind of remote solid state quantum emitters.

cond-mat.mes-hall↗

arXiv2018

Multilingual Adaptation of RNN Based ASR Systems

In this work, we focus on multilingual systems based on recurrent neural networks (RNNs), trained using the Connectionist Temporal Classification (CTC) loss function. Using a multilingual set of acoustic units poses difficulties. To address this issue, we proposed Language Feature Vectors (LFVs) to train language adaptive multilingual systems. Language adaptation, in contrast to speaker adaptation, needs to be applied not only on the feature level, but also to deeper layers of the network. In this work, we therefore extended our previous approach by introducing a novel technique which we call "modulation". Based on this method, we modulated the hidden layers of RNNs using LFVs. We evaluated this approach in both full and low resource conditions, as well as for grapheme and phone based systems. Lower error rates throughout the different conditions could be achieved by the use of the modulation.

eess.AS↗

arXiv2017

Micromotion-enabled improvement of quantum logic gates with trapped ions

The micromotion of ion crystals confined in Paul traps is usually considered an inconvenient nuisance, and is thus typically minimised in high-precision experiments such as high-fidelity quantum gates for quantum information processing. In this work, we introduce a particular scheme where this behavior can be reversed, making micromotion beneficial for quantum information processing. We show that using laser-driven micromotion sidebands, it is possible to engineer state-dependent dipole forces with a reduced effect of off-resonant couplings to the carrier transition. This allows one, in a certain parameter regime, to devise entangling gate schemes based on geometric phase gates with both a higher speed and a lower error, which is attractive in light of current efforts towards fault-tolerant quantum information processing. We discuss the prospects of reaching the parameters required to observe this micromotion-enabled improvement in experiments with current and future trap designs.

quant-ph↗

arXiv2017

A microscopically motivated renormalization scheme for the MBL/ETH transition

We introduce a multi-scale diagonalization scheme to study the transition between the many-body localized and the ergodic phase in disordered quantum chains. The scheme assumes a sharp dichotomy between subsystems that behave as localized and resonant spots that obey the Eigenstate Thermalization Hypothesis (ETH). We establish a set of microscopic principles defining the diagonalization scheme, and use them to numerically study the transition in very large systems. To a large extent the results are in agreement with an analytically tractable mean-field analysis of the scheme: We find that at the critical point the system is almost surely localized in the thermodynamic limit, hosting a set of thermal inclusions whose sizes are power-law distributed. On the localized side the {\em typical} localization length is bounded from above. The bound saturates upon approach to criticality, entailing that a finite ergodic inclusion thermalizes a region of diverging diameter. The dominant thermal inclusions have a fractal structure, implying that averaged correlators decay as stretched exponentials throughout the localized phase. Slightly on the ergodic side thermalization occurs through an avalanche instability of the nearly localized bulk, whereby rare, supercritically large ergodic spots eventually thermalize the entire sample. Their size diverges at the transition, while their density vanishes. The non-local, avalanche-like nature of this instability entails a breakdown of single parameter scaling and puts the delocalization transition outside the realm of standard critical phenomena.

cond-mat.stat-mech↗

arXiv2017

Phonemic and Graphemic Multilingual CTC Based Speech Recognition

Training automatic speech recognition (ASR) systems requires large amounts of data in the target language in order to achieve good performance. Whereas large training corpora are readily available for languages like English, there exists a long tail of languages which do suffer from a lack of resources. One method to handle data sparsity is to use data from additional source languages and build a multilingual system. Recently, ASR systems based on recurrent neural networks (RNNs) trained with connectionist temporal classification (CTC) have gained substantial research interest. In this work, we extended our previous approach towards training CTC-based systems multilingually. Our systems feature a global phone set, based on the joint phone sets of each source language. We evaluated the use of different language combinations as well as the addition of Language Feature Vectors (LFVs). As contrastive experiment, we built systems based on graphemes as well. Systems having a multilingual phone set are known to suffer in performance compared to their monolingual counterparts. With our proposed approach, we could reduce the gap between these mono- and multilingual setups, using either graphemes or phonemes.

eess.AS↗

arXiv2017

Multiple scaling behavior and nonlinear traits in music scores

We present a statistical analysis of music scores from different composers using detrended fluctuation analysis. We find different fluctuation profiles that correspond to distinct auto-correlation structures of the musical pieces. Further, we reveal evidence for the presence of nonlinear auto-correlations by estimating the detrended fluctuation analysis of the magnitude series, a result validated by a corresponding study of appropriate surrogate data. The amount and the character of nonlinear correlations vary from one composer to another. Finally, we performed a simple experiment in order to evaluate the pleasantness of the musical surrogate pieces in comparison with the original music and find that nonlinear correlations could play an important role in the aesthetic perception of a musical piece.

physics.soc-ph↗

arXiv2017

Yeah, Right, Uh-Huh: A Deep Learning Backchannel Predictor

Using supporting backchannel (BC) cues can make human-computer interaction more social. BCs provide a feedback from the listener to the speaker indicating to the speaker that he is still listened to. BCs can be expressed in different ways, depending on the modality of the interaction, for example as gestures or acoustic cues. In this work, we only considered acoustic cues. We are proposing an approach towards detecting BC opportunities based on acoustic input features like power and pitch. While other works in the field rely on the use of a hand-written rule set or specialized features, we made use of artificial neural networks. They are capable of deriving higher order features from input features themselves. In our setup, we first used a fully connected feed-forward network to establish an updated baseline in comparison to our previously proposed setup. We also extended this setup by the use of Long Short-Term Memory (LSTM) networks which have shown to outperform feed-forward based setups on various tasks. Our best system achieved an F1-Score of 0.37 using power and pitch features. Adding linguistic information using word2vec, the score increased to 0.39.

cs.CL↗

arXiv2017

Hyper-entanglement of photons emitted by a quantum dot

Entanglement is a unique quantum mechanical attribute and a fundamental resource of quantum technologies. Entanglement can be achieved in various individual degrees of freedom, nonetheless some systems are able to create simultaneous entanglement in multiple degrees of freedom - hyper-entanglement. A hyper-entangled state of light represents a valuable tool capable of reducing the experimental requirements and resource overheads and it can improve the success rate of quantum information protocols. Here, we report on demonstration of polarization and time-bin hyper-entangled photons emitted from a single quantum dot. We achieved this result by applying resonant and coherent excitation on a quantum dot system with marginal fine structure splitting. Our results yield fidelities to the maximally entangled state of 0.81(6) and 0.87(4) in polarization and time-bin, respectively.

quant-ph↗

arXiv2016

Multiferroic magnetic spirals induced by random magnetic exchanges

Multiferroism can originate from the breaking of inversion symmetry caused by magnetic-spiral order. The usual mechanism for stabilizing a magnetic spiral is competition between magnetic exchange interactions differing by their range and sign, such as nearest-neighbor and next-nearest- neighbor interactions. Since the latter are usually weak the onset temperatures for multiferroism via this mechanism are typically low. By considering a realistic model for YBaCuFeO$_5$ we propose an alternative mechanism for magnetic-spiral order, and hence for multiferroism, that occurs at much higher temperatures. We show using Monte-Carlo simulations and electronic structure calculations based on density functional theory that the Heisenberg model on a geometrically non-frustrated lattice with only nearest-neighbor interactions can have a spiral phase up to high temperature when frustrating bonds are introduced randomly along a single crystallographic direction as caused, e.g., by a particular type of chemical disorder. This long-range correlated pattern of frustration avoids ferroelectrically inactive spin glass order. Finally, we provide an intuitive explanation for this mechanism and discuss its generalization to other materials.

cond-mat.mtrl-sci↗

arXiv2016

Magneto-oscillations of the mobility edge in Coulomb frustrated bosons and fermions

We study the crossover from strong to weak localization of hard-core bosons on a two dimensional honeycomb lattice in a magnetic field, as motivated by recent experiments on structured films. Taking into account long range Coulomb interactions among the bosons, an effective mobility edge in the excitation spectrum of the insulating Bose glass is identified as the (intensive) energy scale at which excitations become nearly delocalized. Within the forward scattering approximation in the bosonic hopping we find the effective mobility edge $ε_c$ to oscillate periodically with the magnetic flux per plaquette, $ϕ$. We find non-analytic cusps in $ε_c(ϕ)$ at integer or half-integer flux. The bosonic magneto-oscillations start with an increase of the mobility edge (and thus of resistance) with applied flux, in contrast to the equivalent fermionic problem. The amplitude of the oscillations is much more substantial in bosons than in fermions. Bosons exhibit a single hump per flux period, while fermion characteristics undergo two humps. Those are identical for non-interacting fermions, but Coulomb correlations are shown to lead to systematic deviations from this statistical period doubling. Our theory reproduces several key features observed in the activated magneto-transport in structured films.

cond-mat.dis-nn↗

arXiv2016

Classical impurities and boundary Majorana zero modes in quantum chains

We study the response of classical impurities in quantum Ising chains. The Z2 degeneracy they entail renders the existence of two decoupled Majorana modes at zero energy an exact property of a finite system at arbitrary values of its bulk parameters. We trace the evolution of these modes across the transition from the disordered phase to the ordered one and analyze the concomitant qualitative changes of local magnetic properties of an isolated impurity. In the disordered phase, the two ground states differ only close to the impurity, and they are related by the action of an explicitly constructed quasi-local operator. In this phase the local transverse spin susceptibility follows a Curie law. The critical response of a boundary impurity is logarithmically divergent and maps to the two-channel Kondo problem, while it saturates for critical bulk impurities, as well as in the ordered phase. The results for the Ising chain translate to the related problem of a resonant level coupled to a 1d p-wave superconductor or a Peierls chain, whereby the magnetic order is mapped to topological order. We find that the topological phase always exhibits a continuous impurity response to local fields as a result of the level repulsion of local levels from the boundary Majorana zero mode. In contrast, the disordered phase generically features a discontinuous magnetization or charging response. This difference constitutes a general thermodynamic fingerprint of topological order in phases with a bulk gap.

cond-mat.str-el↗

arXiv2016

Phase super-resolution with N00N states generated by on demand single-photon sources

Multi-photon entangled states such as `N00N states' have attracted a lot of attention because of their possible application in high-precision phase measurements. So far, N00N states have been generated by spontaneous parametric down-conversion sources and by mixing quantum and classical light on a beam splitter. However, single-photon on demand sources promise to be more efficient and scalable than down-conversion sources which are probabilistic in nature. Here, we demonstrate super-resolving phase measurements based on two-photon N00N states generated by a quantum dot single-photon source utilizing the Hong-Ou-Mandel effect on a beam splitter. The quantum dot was excited through a two-photon resonant excitation scheme. Using a $π$-pulse, this results in the deterministic generation of indistinguishable single photons from both, the biexciton and exciton radiative recombination processes. Launching either type of the photons into both inputs of an interferometer, phase super-resolution, i.e. an interference fringe spacing with doubled rate relative to the single photon rate has been observed for both transitions. Interference visibilities of $V_{N=2,\mathit{XX}}=0.54\pm0.01$ for the biexciton and $V_{N=2,\mathit{X}}=0.46\pm0.02$ for the exciton transition have been obtained, respectively.

quant-ph↗

arXiv2016

Thermalization and many-body localization in systems under dynamic nuclear polarization

We study the role of dipolar interactions in the standard protocol used to achieve dynamic nuclear polarization (DNP). In the so-called spin-temperature regime, where the interactions establish an effective thermodynamic behavior in the out-of-equilibrium stationary state, we provide numerical predictions for the level of hyperpolarization. We show that nuclear spins equilibrate to the effective spin-temperature established among the electron spins of radicals, as expected from the quantum theory of thermalization. Moreover, we present an analytical technique to estimate the spin temperature, and thus, the nuclear hyperpolarization in the steady state, as a function of interaction strength and quenched disorder. This reproduces both our numerical data and experimental results. Our central finding is that the nuclear hyperpolarization increases steadily upon reducing the interaction strength (by diluting the radical density). Interestingly, the highest polarization is reached at a point where the establishment of a spin temperature is just about to break down due to the incipient many-body localization transition in the electron spin system.

cond-mat.str-el↗

arXiv2016

Absence of many-body mobility edges

Localization transitions as a function of temperature require a many-body mobility edge in energy, separating localized from ergodic states. We argue that this scenario is inconsistent because local fluctuations into the ergodic phase within the supposedly localized phase can serve as mobile bubbles that induce global delocalization. Such fluctuations inevitably appear with a low but finite density anywhere in any typical state. We conclude that the only possibility for many-body localization to occur are lattice models that are localized at all energies. Building on a close analogy with a model of assisted two-particle hopping, where interactions induce delocalization, we argue why hot bubbles are mobile and do not localize upon diluting their energy. Numerical tests of our scenario show that previously reported mobility edges cannot be distinguished from finite-size effects.

cond-mat.dis-nn↗

arXiv2015

Cavity-enhanced simultaneous dressing of quantum dot exciton and biexciton states

We demonstrate the simultaneous dressing of both vacuum-to-exciton and exciton-to-biexciton transitions of a single semiconductor quantum dot in a high-Q micropillar cavity, using photoluminescence spectroscopy. Resonant two-photon excitation of the biexciton is achieved by spectrally tuning the quantum dot emission with respect to the cavity mode. The cavity couples to both transitions and amplifies the Rabi-frequency of the likewise resonant cw laser, driving the transitions. We observe strong-field splitting of the emission lines, which depend on the driving Rabi field amplitude and the cavity-laser detuning. A dressed state theory of a driven 4-level atom correctly predicts the distinct spectral transitions observed in the emission spectrum, and a detailed description of the emission spectra is further provided through a polaron master equation approach which accounts for cavity coupling and acoustic phonon interactions of the semiconductor medium.

quant-ph↗

arXiv2015

Quantifying Spatial Correlations of General Quantum Dynamics

Understanding the role of correlations in quantum systems is both a fundamental challenge as well as of high practical relevance for the control of multi-particle quantum systems. Whereas a lot of research has been devoted to study the various types of correlations that can be present in the states of quantum systems, in this work we introduce a general and rigorous method to quantify the amount of correlations in the dynamics of quantum systems. Using a resource-theoretical approach, we introduce a suitable quantifier and characterize the properties of correlated dynamics. Furthermore, we benchmark our method by applying it to the paradigmatic case of two atoms weakly coupled to the electromagnetic radiation field, and illustrate its potential use to detect and assess spatial noise correlations in quantum computing architectures.

quant-ph↗