Search

The Quantum-Enhanced Space-Time (QUEST) experiment consists of a pair of co-located Power Recycled Michelson Interferometers, each designed to have a broadband, shot-noise limited displacement sensitivity of $2\times10^{-19}$ $\mathrm{m/\sqrt{Hz}}$ from 1 to 200 MHz. Here we present the first results of QUEST that set new upper limits on correlated length fluctuations from 13 to 80 MHz, constituting the first broadband constraints for a stochastic gravitational wave background at these frequencies. In a coincident observing run of $10^{4}$ s the averaging of the cross-correlation spectra between the two interferometer signals resulted in a strain sensitivity of $3\times10^{-20}$ $\mathrm{1/\sqrt{Hz}}$ at 40 MHz, making QUEST the most sensitive table-top interferometric system to date.

Speckle patterns manifesting from the interaction of coherent X-rays with matter offer a glimpse into the dynamics of nanoscale domains that underpin many emergent phenomena in quantum materials. While the dynamics of the average structure can be followed with time-resolved X-ray diffraction, the ultrafast evolution of local structures in nonequilibrium conditions have thus far eluded detection due to experimental limitations, such as insufficient X-ray coherent flux. Here we demonstrate a nonequilibrium speckle visibility experiment using a split-and-delay setup at an X-ray free-electron laser. Photoinduced electronic domain fluctuations of the magnetic model material Fe$_{3}$O$_{4}$ reveal changes of the trimeron network configuration due to charge dynamics that exhibit liquid-like fluctuations, analogous to a supercooled liquid phase. This suggests that ultrafast dynamics of electronic heterogeneities under optical stimuli are fundamentally different from thermally-driven ones.

In Data Assimilation, observations are fused with simulations to obtain an accurate estimate of the state and parameters for a given physical system. Combining data with a model, however, while accurately estimating uncertainty, is computationally expensive and infeasible to run in real-time for complex systems. Here, we present a novel particle filter methodology, the Deep Latent Space Particle filter or D-LSPF, that uses neural network-based surrogate models to overcome this computational challenge. The D-LSPF enables filtering in the low-dimensional latent space obtained using Wasserstein AEs with modified vision transformer layers for dimensionality reduction and transformers for parameterized latent space time stepping. As we demonstrate on three test cases, including leak localization in multi-phase pipe flow and seabed identification for fully nonlinear water waves, the D-LSPF runs orders of magnitude faster than a high-fidelity particle filter and 3-5 times faster than alternative methods while being up to an order of magnitude more accurate. The D-LSPF thus enables real-time data assimilation with uncertainty quantification for physical systems.

Resonate-and-Fire (RF) neurons are an interesting complementary model for integrator neurons in spiking neural networks (SNNs). Due to their resonating membrane dynamics they can extract frequency patterns within the time domain. While established RF variants suffer from intrinsic shortcomings, the recently proposed balanced resonate-and-fire (BRF) neuron marked a significant methodological advance in terms of task performance, spiking and parameter efficiency, as well as, general stability and robustness, demonstrated for recurrent SNNs in various sequence learning tasks. One of the most intriguing result, however, was an immense improvement in training convergence speed and smoothness, overcoming the typical convergence dilemma in backprop-based SNN training. This paper aims at providing further intuitions about how and why these convergence advantages emerge. We show that BRF neurons, in contrast to well-established ALIF neurons, span a very clean and smooth - almost convex - error landscape. Furthermore, empirical results reveal that the convergence benefits are predominantly coupled with a divergence boundary-aware optimization, a major component of the BRF formulation that addresses the numerical stability of the time-discrete resonator approximation. These results are supported by a formal investigation of the membrane dynamics indicating that the gradient is transferred back through time without loss of magnitude.

To make sense of their surroundings, intelligent systems must transform complex sensory inputs to structured codes that are reduced to task-relevant information such as object category. Biological agents achieve this in a largely autonomous manner, presumably via self-supervised learning. Whereas previous attempts to model the underlying mechanisms were largely discriminative in nature, there is ample evidence that the brain employs a generative model of the world. Here, we propose that eye movements, in combination with the focused nature of primate vision, constitute a generative, self-supervised task of predicting and revealing visual information. We construct a proof-of-principle model starting from the framework of masked image modeling (MIM), a common approach in deep representation learning. To do so, we analyze how core components of MIM such as masking technique and data augmentation influence the formation of category-specific representations. This allows us not only to better understand the principles behind MIM, but to then reassemble a MIM more in line with the focused nature of biological perception. We find that MIM disentangles neurons in latent space without explicit regularization, a property that has been suggested to structure visual representations in primates. Together with previous findings of invariance learning, this highlights an interesting connection of MIM to latent regularization approaches for self-supervised learning. The source code is available under https://github.com/RobinWeiler/FocusMIM

The GQuEST (Gravity from the Quantum Entanglement of Space-Time) experiment uses tabletop-scale Michelson laser interferometers to probe for fluctuations in space-time. We present a practicable interferometer design featuring a novel photon counting readout method that provides unprecedented sensitivity, as it is not subject to the interferometric standard quantum limit. We evaluate the potential of this design to measure space-time fluctuations motivated by recent `geontropic' quantum gravity models. The accelerated accrual of Fisher information offered by the photon counting readout enables GQuEST to detect the predicted quantum gravity phenomena within measurement times at least 100 times shorter than equivalent conventional interferometers. The GQuEST design thus enables a fast and sensitive search for signatures of quantum gravity in a laboratory-scale experiment.

The resonate-and-fire (RF) neuron, introduced over two decades ago, is a simple, efficient, yet biologically plausible spiking neuron model, which can extract frequency patterns within the time domain due to its resonating membrane dynamics. However, previous RF formulations suffer from intrinsic shortcomings that limit effective learning and prevent exploiting the principled advantage of RF neurons. Here, we introduce the balanced RF (BRF) neuron, which alleviates some of the intrinsic limitations of vanilla RF neurons and demonstrates its effectiveness within recurrent spiking neural networks (RSNNs) on various sequence learning tasks. We show that networks of BRF neurons achieve overall higher task performance, produce only a fraction of the spikes, and require significantly fewer parameters as compared to modern RSNNs. Moreover, BRF-RSNN consistently provide much faster and more stable training convergence, even when bridging many hundreds of time steps during backpropagation through time (BPTT). These results underscore that our BRF-RSNN is a strong candidate for future large-scale RSNN architectures, further lines of research in SNN methodology, and more efficient hardware implementations., Comment: Accepted at ICML 2024, https://proceedings.mlr.press/v235/higuchi24a.html

We report on a direct search for scalar field dark matter using data from LIGO's third observing run. We analyse the coupling of size oscillations of the interferometer's beamsplitter and arm test masses that may be caused by scalar field dark matter. Using new efficient search methods to maximise sensitivity for signatures of such oscillations, we set new upper limits for the coupling constants of scalar field dark matter as a function of its mass, which improve upon bounds from previous direct searches by several orders of magnitude in a frequency band from 10 Hz to 180 Hz., Comment: 6 pages (including references), 3 figures

Resonant driving of electronic transitions with coherent laser sources creates quantum coherent superpositions of the involved electronic states. Most time-resolved studies have focused on gases or isolated subsystems embedded in insulating solids, aiming for applications in quantum information. Here, we demonstrate coherent control of orbital wavefunctions in pyrochlore $Tb_{2}Ti_{2}O_{7}$, which forms an interacting spin liquid ground state. We show that resonant excitation with a strong THz pulse creates a coherent superposition of the lowest energy Tb 4f states before the magnetic interactions eventually dephase them. The coherence manifests itself as a macroscopic oscillating magnetic dipole, which is detected by ultrafast resonant x-ray diffraction. The induced quantum coherence demonstrates coherent control of orbital wave functions, a new tool for the ultrafast manipulation and investigation of quantum materials.

We present a formalism to discern the effects of fluctuations of the spacetime metric on electromagnetic radiation. The formalism works via the measurement of electromagnetic field correlations, while allowing a clear assessment of the assumptions involved. As an application of the formalism, we present a model of spacetime fluctuations that appear as random fluctuations of the refractive index of the vacuum in single, and two co-located Michelson interferometers. We compare an interferometric signal predicted using this model to experimental data from the Holometer and aLIGO. We show that if the signal manifests at a frequency at which the interferometers are sensitive, the strength and scale of possible spacetime fluctuations can be constrained. The bounds, thus obtained, on the strength and scale of the spacetime fluctuations, are also shown to be more stringent than the bounds obtained previously using astronomical observation at optical frequencies. The formalism enables us to evaluate proposed experiments such as QUEST for constraining quantum spacetime fluctuations and to design new ones., Comment: 21 pages, 3 figures, 2 tables

Many functionally relevant ferroic phenomena in quantum materials can be manipulated by driving the lattice coherently with optical and terahertz pulses. New physical phenomena and non-equilibrium phases that have no equilibrium counterpart have been discovered following these protocols. The underlying structural dynamics has been mostly studied by recording the average atomic position along dynamical structural coordinates with elastic scattering methods. However, crystal lattice fluctuations, which are known to influence phase transitions in equilibrium, are also expected to determine these dynamics but have rarely been explored. Here, we study the driven dynamics of the quantum paraelectric SrTiO3, in which mid-infrared drives have been shown to induce a metastable ferroelectric state. Crucial in these physics is the competition between the polar instability and antiferrodistortive rotations, which in equilibrium frustrate the formation of long-range ferroelectricity. We make use of high intensity mid-infrared optical pulses to resonantly drive a Ti-O stretching mode at 17 THz, and we measure the resulting change in lattice fluctuations using time-resolved x-ray diffuse scattering at a free electron laser. After a prompt increase, we observe a long-lived quench in R-point antiferrodistortive lattice fluctuations. The enhancement and reduction in lattice fluctuations are explained theoretically by considering fourth-order nonlinear phononic interactions and third-order coupling to the driven optical phonon and to lattice strain, respectively. These observations provide a number of new and testable hypotheses for the physics of light-induced ferroelectricity.

Gravitational waves modulate the apparent frequencies of other periodic signals. Low-frequency gravitational waves could therefore be detected by observing frequency modulations in signals from higher-frequency sources, e.g., those from binary white dwarfs detected with the LISA gravitational-wave detector. We propose a concrete method to extract these modulations by coherently adding the cross-spectra of a large number of well-resolved quasi-monochromatic signals. We apply this method to the case of LISA, and find this method would enable the detection of background gravitational wave strain amplitudes of, e.g., $A\simeq10^{-10}$ at a frequency $F\simeq10^{-8}\,\rm Hz$, given current projections for the number and properties of Galactic binary white dwarfs and the sensitivity of the instrument. We also estimate (to within an order of magnitude) that this method could potentially compete with that of current Pulsar Timing Arrays when using signals from binary neutron stars such as those expected to be observed with proposed detectors like DECIGO. Our results show that gravitational-wave detectors could be sensitive at frequencies outside of their designed bandwidth using the same infrastructure, which has the potential to open up unexplored and otherwise inaccessible parts of the gravitational wave spectrum., Comment: 12 pages, 6 figures, 1 table

In this work, we propose polarimetry experiments to search for low-mass (sub-eV) bosonic field dark matter, including axions and axion-like particles. We show that a polarimetry configuration consisting of a thick birefringent solid inside a Fabry-P\'erot cavity is exceptionally sensitive to scalar field dark matter, which may cause oscillatory variations in the solid's thickness and refractive index. In addition, we show that a reconfiguration of this polarimetry experiment, in which two quarter-wave plates are placed inside the Fabry-P\'erot cavity instead of a thick birefringent solid, is very sensitive to axion-like particles. We investigate the possibility of using cross-correlation of twin polarimeters to increase the sensitivity of the experiment, which in turn could allow us to explore unexplored parts of the parameter space and potentially detect a signal in either dark matter scenario.

The event-driven and sparse nature of communication between spiking neurons in the brain holds great promise for flexible and energy-efficient AI. Recent advances in learning algorithms have demonstrated that recurrent networks of spiking neurons can be effectively trained to achieve competitive performance compared to standard recurrent neural networks. Still, as these learning algorithms use error-backpropagation through time (BPTT), they suffer from high memory requirements, are slow to train, and are incompatible with online learning. This limits the application of these learning algorithms to relatively small networks and to limited temporal sequence lengths. Online approximations to BPTT with lower computational and memory complexity have been proposed (e-prop, OSTL), but in practice also suffer from memory limitations and, as approximations, do not outperform standard BPTT training. Here, we show how a recently developed alternative to BPTT, Forward Propagation Through Time (FPTT) can be applied in spiking neural networks. Different from BPTT, FPTT attempts to minimize an ongoing dynamically regularized risk on the loss. As a result, FPTT can be computed in an online fashion and has fixed complexity with respect to the sequence length. When combined with a novel dynamic spiking neuron model, the Liquid-Time-Constant neuron, we show that SNNs trained with FPTT outperform online BPTT approximations, and approach or exceed offline BPTT accuracy on temporal classification tasks. This approach thus makes it feasible to train SNNs in a memory-friendly online fashion on long sequences and scale up SNNs to novel and complex neural architectures., Comment: 12 pages, 4 figures

Low-mass (sub-eV) scalar field dark matter may induce apparent oscillations of fundamental constants, resulting in corresponding oscillations of the size and the index of refraction of solids. Laser interferometers are highly sensitive to changes in the size and index of refraction of the main beamsplitter. Using cross-correlated data of the Fermilab Holometer instrument, which consists of twin co-located 40-m arm length power-recycled interferometers, we investigate the possible existence of scalar field dark matter candidates in the mass range between 1.6$\cdot$10$^{-12}$ eV and 1.0$\cdot$10$^{-7}$ eV. We set new upper limits for the coupling parameters of scalar field dark matter, improving on limits from previous direct searches by up to three orders of magnitude.

With the use of twin, co-located, 3D interferometers, Cardiff University's Gravity Exploration Institute aims to observe quantum fluctuations of space-time as predicted by some theories of quantum gravity. Our design displacement sensitivity exceeds that of previous similar experiments, which have constrained the magnitudes of the fluctuations in the 1-25 MHz band. The increased sensitivity comes in large part from the comparably higher circulating power we aim to achieve, which reduces the overall shot noise. One complication of higher circulating power is an increase in contrast defect light, which includes higher-order modes. We will use the DC-readout scheme, whose dark-fringe offset must sufficiently dominate the contrast defect in order to detect faint signals. However, too much total output power risks saturating the high-bandwidth photodetectors. Suppressing the higher-order mode content of the contrast defect is a key strategy to realising the high circulating power and eliminating non-signal-carrying power that contributes to shot noise. For this, the inclusion of an output mode cleaner, whose design is described, is required., Comment: 6 pages, 3 figures, 1 table, Contribution to the 2021 Gravitation session of the 55th Rencontres de Moriond

Inspired by more detailed modeling of biological neurons, Spiking neural networks (SNNs) have been investigated both as more biologically plausible and potentially more powerful models of neural computation, and also with the aim of extracting biological neurons' energy efficiency; the performance of such networks however has remained lacking compared to classical artificial neural networks (ANNs). Here, we demonstrate how a novel surrogate gradient combined with recurrent networks of tunable and adaptive spiking neurons yields state-of-the-art for SNNs on challenging benchmarks in the time-domain, like speech and gesture recognition. This also exceeds the performance of standard classical recurrent neural networks (RNNs) and approaches that of the best modern ANNs. As these SNNs exhibit sparse spiking, we show that they theoretically are one to three orders of magnitude more computationally efficient compared to RNNs with comparable performance. Together, this positions SNNs as an attractive solution for AI hardware implementations., Comment: 11 pages