Search

It has been shown that unclocked, recurrent networks of Boolean gates in FPGAs can be used for low-SWaP reservoir computing. In such systems, topology and node functionality of the network are randomly initialized. To create a network that solves a task, weights are applied to output nodes and learning is achieved by adjusting those weights with conventional machine learning methods. However, performance is often limited compared to networks where all parameters are learned. Herein, we explore an alternative learning approach for unclocked, recurrent networks in FPGAs. We use evolutionary computation to evolve the Boolean functions of network nodes. In one type of implementation the output nodes are used directly to perform a task and all learning is via evolution of the network's node functions. In a second type of implementation a back-end classifier is used as in traditional reservoir computing. In that case, both evolution of node functions and adjustment of output node weights contribute to learning. We demonstrate the practicality of node function evolution, obtaining an accuracy improvement of ~30% on an image classification task while processing at a rate of over three million samples per second. We additionally demonstrate evolvability of network memory and dynamic output signals.

Superparamagnetic tunnel junctions are important devices for a range of emerging technologies, but most existing compact models capture only their mean switching rates. Capturing qualitatively accurate analog dynamics of these devices will be important as the technology scales up. Here we present results using a one-dimensional overdamped Langevin equation that captures statistical properties of measured time traces, including voltage histograms, drift and diffusion characteristics as measured with Kramers-Moyal coefficients, and dwell times distributions. While common macrospin models are more physically-motivated magnetic models than the Langevin model, we show that for the device measured here, they capture even fewer of the measured experimental behaviors., Comment: 22 pages (13 pages of main text), 21 figures

Superparamagnetic tunnel junctions (SMTJs) are promising sources of randomness for compact and energy efficient implementations of probabilistic computing techniques. Augmenting an SMTJ with electronic circuits, to convert the random telegraph fluctuations of its resistance state to stochastic digital signals, gives a basic building block known as a probabilistic bit or $p$-bit. Though scalable probabilistic computing methods connecting $p$-bits have been proposed, practical implementations are limited by either minimal tunability or energy inefficient microprocessors-in-the-loop. In this work, we experimentally demonstrate the functionality of a scalable analog unit cell, namely a pair of $p$-bits with programmable electrical coupling. This tunable coupling is implemented with operational amplifier circuits that have a time constant of approximately 1us, which is faster than the mean dwell times of the SMTJs over most of the operating range. Programmability enables flexibility, allowing both positive and negative couplings, as well as coupling devices with widely varying device properties. These tunable coupling circuits can achieve the whole range of correlations from $-1$ to $1$, for both devices with similar timescales, and devices whose time scales vary by an order of magnitude. This range of correlation allows such circuits to be used for scalable implementations of simulated annealing with probabilistic computing.

Machine learning applications are computationally demanding and power intensive. Hardware acceleration of these software tools is a natural step being explored using various technologies. A recurrent processing unit (RPU) is fast and power-efficient hardware for machine learning under development at the University of Maryland. It is comprised of a recurrent neural network and a trainable output vector as a hardware implementation of a reservoir computer. The reservoir is currently realized on both Xilinx 7-series and Ultrascale+ ZYNQ SoCs using an autonomous Boolean network for processing and a Python-based software API. The RPU is capable of classifying up to 40M MNIST images per second with the reservoir consuming under 261mW of power. Using an array of 2048 unclocked gates with roughly 100pS transition times, we achieve about 20 TOPS and 75 TOPS/W.

In clouds of suspended particles (grains, droplets, spheres, crystals, etc.), collisions electrify the particles and the clouds, producing large electric potential differences over large scales. This is seen most spectacularly in the atmosphere as lighting in thunderstorms, thundersnow, dust storms, and volcanic ash plumes where multi-million-volt potential differences over scales of kilometers can be produced, but it is a general phenomenon in granular systems as a whole. The electrification process is not well understood, especially for electrification of insulating particles of the same material. To investigate the relative importances of particle properties (material, size, etc.) and collective phenomena (behaviors of systems at large scales not easily predicted from local dynamics) in granular and atmospheric electrification, we used a table-top experiment that mechanically shakes particles inside a cell where we measure the macroscopic electric field between the electrically conducting end plates. The measured electric fields are a result of capacitive coupling and direct charge transfer between the particles and the plates. Using a diverse range of mono-material particle sets (plastics, ceramic, glass, and metals), we found that all our particle materials electrify and show similar dynamics with long time-scale temporal variation and an electric field amplitude that depends on the particle quantity in a complex way. These results suggest that while particle properties do matter like previous investigations have shown, macroscopic electrification of solids is relatively material agnostic and large scale collective phenomena play a major role., Comment: 12 pages, 7 figures, 1 table, submitted to Phys. Rev. Fluids

Measurement of the differential rotation of the Sun's interior is one of the great achievements of helioseismology, providing important constraints for stellar physics. The technique relies on observing and analyzing rotationally-induced splittings of p-modes in the star. Here we demonstrate the first use of the technique in a laboratory setting. We apply it in a spherical cavity with a spinning central core (spherical-Couette flow) to determine the mean azimuthal velocity of the air filling the cavity. We excite a number of acoustic resonances (analogous to p-modes in the Sun) using a speaker and record the response with an array of small microphones on the outer sphere. Many observed acoustic modes show rotationally-induced splittings, which allow us to perform an inversion to determine the air's azimuthal velocity as a function of both radius and latitude. We validate the method by comparing the velocity field obtained through inversion against the velocity profile measured with a calibrated hot film anemometer. This 'modal acoustic velocimetry' technique has great potential for laboratory setups involving rotating fluids in axisymmetric cavities. It will be useful especially in liquid metals where direct optical methods are unsuitable and ultrasonic techniques very challenging at best., Comment: Accepted for publication on The New Journal Of Physics

We present azimuthal velocity profiles measured in a Taylor-Couette apparatus, which has been used as a model of stellar and planetary accretion disks. The apparatus has a cylinder radius ratio of $\eta = 0.716$, an aspect-ratio of $\Gamma = 11.74$, and the plates closing the cylinders in the axial direction are attached to the outer cylinder. We investigate angular momentum transport and Ekman pumping in the Rayleigh-stable regime. The regime is linearly stable and is characterized by radially increasing specific angular momentum. We present several Rayleigh-stable profiles for shear Reynolds numbers $Re_S \sim O(10^5) \,$, both for $\Omega_i > \Omega_o > 0$ (quasi-Keplerian regime) and $\Omega_o > \Omega_i > 0$ (sub-rotating regime) where $\Omega_{i,o}$ is the inner/outer cylinder rotation rate. None of the velocity profiles matches the non-vortical laminar Taylor-Couette profile. The deviation from that profile increased as solid-body rotation is approached at fixed $Re_S$. Flow super-rotation, an angular velocity greater than that of both cylinders, is observed in the sub-rotating regime. The velocity profiles give lower bounds for the torques required to rotate the inner cylinder that were larger than the torques for the case of laminar Taylor-Couette flow. The quasi-Keplerian profiles are composed of a well mixed inner region, having approximately constant angular momentum, connected to an outer region in solid-body rotation with the outer cylinder and attached axial boundaries. These regions suggest that the angular momentum is transported axially to the axial boundaries. Therefore, Taylor-Couette flow with closing plates attached to the outer cylinder is an imperfect model for accretion disk flows, especially with regard to their stability., Comment: 22 pages, 10 figures, 2 tables, under consideration for publication in Journal of Fluid Mechanics (JFM)

In superfluid helium, vorticity is quantized and constrained on line-like phase singularities, called quantum or quantized vortices. By visualizing the motion of sub-micron frozen particles in superfluid $^{4}$He, we directly observe for the first time the helical Kelvin waves excited after quantized vortex reconnections. We compare the data with self-similar solutions of vortex filament models. We report the results in a fluid dynamic video., Comment: Includes 2 videos. Submitted to the Gallery of Fluid Motion of the 65th Annual DFD Meeting of the APS in San Diego, 2012

Spherical Couette flow (flow between concentric rotating spheres) is one of flows under consideration for the laboratory magnetic dynamos. Recent experiments have shown that such flows may excite Coriolis restored inertial modes. The present work aims to better understand the properties of the observed modes and the nature of their excitation. Using numerical solutions describing forced inertial modes of a uniformly rotating fluid inside a spherical shell, we first identify the observed oscillations of the Couette flow with non-axisymmetric, retrograde, equatorially anti-symmetric inertial modes, confirming first attempts using a full sphere model. Although the model has no differential rotation, identification is possible because a large fraction of the fluid in a spherical Couette flow rotates rigidly. From the observed sequence of the excited modes appearing when the inner sphere is slowed down by step, we identify a critical Rossby number associated with a given mode and below which it is excited. The matching between this critical number and the one derived from the phase velocity of the numerically computed modes shows that these modes are excited by an instability likely driven by the critical layer that develops in the shear layer staying along the tangent cylinder of the inner sphere., Comment: 11 pages, 17 figures

Flow between concentric spheres of radius ratio $\eta = r_\mathrm{i}/r_\mathrm{o} = 0.35$ is studied in a 3 m outer diameter experiment. We have measured the torques required to maintain constant boundary speeds as well as localized wall shear stress, velocity, and pressure. At low Ekman number $E = 2.1\times10^{-7}$ and modest Rossby number $0.07 < Ro < 3.4$, the resulting flow is highly turbulent, with a Reynolds number ($Re=Ro/E$) exceeding fifteen million. Several turbulent flow regimes are evident as $Ro$ is varied for fixed $E$. We focus our attention on one flow transition in particular, between $Ro = 1.8$ and $Ro = 2.6$, where the flow shows bistable behavior. For $Ro$ within this range, the flow undergoes intermittent transitions between the states observed alone at adjacent $Ro$ outside the switching range. The two states are clearly distinguished in all measured flow quantities, including a striking reduction in torque demanded from the inner sphere by the state lying at higher $Ro$. The reduced angular momentum transport appears to be associated with the development of a fast zonal circulation near the experiment core. The lower torque state exhibits waves, one of which is similar to an inertial mode known for a full sphere, and another which appears to be a strongly advected Rossby-type wave. These results represent a new laboratory example of the overlapping existence of distinct flow states in high Reynolds number flow. Turbulent multiple stability and the resilience of transport barriers associated with zonal flows are important topics in geophysical and astrophysical contexts.

We present measurements of the angular momentum flux (torque) in Taylor-Couette flow of water between independently rotating cylinders for all regions of the $\(\Omega_1, \Omega_2\)$ parameter space at high Reynolds numbers, where $\Omega_1$ $\(\Omega_2\)$ is the inner (outer) cylinder angular velocity. We find that the Rossby number $Ro = \(\Omega_1 - \Omega_2\)/\Omega_2$ fully determines the state and torque $G$ as compared to $G(Ro = \infty) \equiv \Gi$. The ratio $G/\Gi$ is a linear function of $Ro^{-1}$ in four sections of the parameter space. For flows with radially-increasing angular momentum, our measured torques greatly exceed those of previous experiments [Ji \textit{et al.}, Nature, \textbf{444}, 343 (2006)], but agree with the analysis of Richard and Zahn [Astron. Astrophys., \textbf{347}, 734 (1999)]., Comment: 4 pages, 4 figures, to appear in Physical Review Letters

A new turbulent Taylor-Couette system consisting of two independently rotating cylinders has been constructed. The gap between the cylinders has a height of 0.927 m, an inner radius of 0.200 m, and a variable outer radius (from 0.279 to 0.220 m). The maximum angular rotation rates of the inner and outer cylinder are 20 and 10 Hz, respectively, resulting in Reynolds numbers up to 3.4 x 10^6 with water as working fluid. With this Taylor-Couette system, the parameter space (Re_i, Re_o, {\eta}) extends to (2.0 x 10^6, {\pm}1.4 x 10^6, 0.716-0.909). The system is equipped with bubble injectors, temperature control, skin-friction drag sensors, and several local sensors for studying turbulent single-phase and two-phase flows. Inner cylinder load cells detect skin-friction drag via torque measurements. The clear acrylic outer cylinder allows the dynamics of the liquid flow and the dispersed phase (bubbles, particles, fibers, etc.) inside the gap to be investigated with specialized local sensors and nonintrusive optical imaging techniques. The system allows study of both Taylor-Couette flow in a high-Reynolds-number regime, and the mechanisms behind skin-friction drag alterations due to bubble injection, polymer injection, and surface hydrophobicity and roughness., Comment: 13 pages, 14 figures

We observe deterministic chaos in a simple network of electronic logic gates that are not regulated by a clocking signal. The resulting power spectrum is ultra-wide-band, extending from dc to beyond 2 GHz. The observed behavior is reproduced qualitatively using an autonomously updating Boolean model with signal propagation times that depend on the recent history of the gates and filtering of pulses of short duration, whose presence is confirmed experimentally. Electronic Boolean chaos may find application as an ultra-wide-band source of radio waves, Comment: 10 pages and 4 figure

We discuss an experimental technique to visualize the motion in the bulk of superfluid $^4$He by tracking micron-sized solid hydrogen tracers. The behavior of the tracers is complex since they may be trapped by the quantized vortices while also interacting with the normal fluid via Stokes drag. We discuss the mechanism by which tracers may be trapped by quantized vortices as well as the dependencies on hydrogen volume fraction, temperature, and flow properties. We apply this technique to study the dynamics of a thermal counterflow. Our observations serve as a direct confirmation of the two-fluid model as well as a quantitative test of the normal fluid velocity dependence on the applied heat flux., Comment: 7 pages, 7 figures, to appear in J. Phys. Soc. Japan

When two vortices cross, each of them breaks into two parts and exchanges part of itself for part of the other. This process, called vortex reconnection, occurs in classical as well as superfluids, and in magnetized plasmas and superconductors. We present the first experimental observations of reconnection between quantized vortices in superfluid helium. We do so by imaging micron-sized solid hydrogen particles trapped on quantized vortex cores (Bewley GP, Lathrop DP, Sreenivasan KR, 2006, Nature, 441:588), and by inferring the occurrence of reconnection from the motions of groups of recoiling particles. We show the distance separating particles on the just-reconnected vortex lines grows as a power law in time. The average value of the scaling exponent is approximately 1/2, consistent with the scale-invariant evolution of the vortices.

Matter at low temperatures exhibits unusual properties such as superfluidity, superconductivity, Bose-Einstein condensation, and supersolidity. These states display quantum mechanical behaviours at scales much larger than atomic dimensions. As in many phase transitions, defects can occur during the transition to the low temperature state. The study of these defects yields useful information about the nature of the transitions and of the macroscopic states themselves. When cooled below the lambda temperature (~2.172 K), liquid helium acquires superfluid properties, and the defects in the superfluid take the form of line vortices with quantized circulation. The formation of these vortices has been suggested as a model for cosmological structure formation. Here we visualize these superfluid (or quantized) vortices by suspending micron-sized solid particles of hydrogen in the fluid. The particles at low concentrations arrange themselves with nearly equal spacing along vortex lines. The number density of the vortex lines in a steadily rotating state compares well with Feynman's prediction. For the first time, it is possible to observe superfluid turbulence in which the superfluid vortices interact, and exist in networks with complex branching. Our method makes it possible to characterize the vortices during the transition across the lambda point and in the turbulent state itself., Comment: submitted to Nature

Experiments observing the liquid surface in a vertically oscillating container have indicated that modeling the dynamics of such systems require maps that admit states at infinity. In this paper we investigate the bifurcations in such a map. We show that though such maps in general fall in the category of piecewise smooth maps, the mechanisms of bifurcations are quite different from those in other piecewise smooth maps. We obtain the conditions of occurrence of infinite states, and show that periodic orbits containing such states are superstable. We observe period-adding cascade in this system, and obtain the scaling law of the successive periodic windows., Comment: 10 pages, 6 figures, composed in Latex2e

Differential rotation occurs in conducting flows in accretion disks and planetary cores. In such systems, the magnetorotational instability can arise from coupling Lorentz and centrifugal forces to cause large radial angular momentum fluxes. We present the first experimental observation of the magnetorotational instability. Our system consists of liquid sodium between differentially rotating spheres, with an imposed coaxial magnetic field. We characterize the observed patterns, dynamics and torque increases, and establish that this instability can occur from a hydrodynamic turbulent background., Comment: 4 pages, 5 figures. Final version, accepted for publication in Physical Review Letters

Self-similar Euler singularities may be useful for understanding some aspects of Navier-Stokes turbulence. Here, a causal explanation for intermittency is given, based on the control of the sudden growth of the gradients by the Euler equations. This explanation uses certain Euler solutions as intermediate asymptotics in Navier-Stokes turbulence \cite{barenblatt} -- controlling the dynamics over a limited spatial and temporal domain. These arise from an analysis of similarity equations, previously discussed by Pelz and Green \cite{pelz}, which yield experimentally testable predictions. Three main points are presented here: scalings of suitable characteristic lengths with time from a critical time $l \sim (t_\circ -t)^\alpha$, $\alpha>1$, a discussion of invariant sets of the similarity equations that result, and a discussion of cutoff mechanisms. The value $\alpha = 3/2$ appears to correspond to Kolmogorov scaling for turbulence. Some limited experimental evidence is presented from Eulerian gradient measurements at the Kolmogorov scale showing $1<\alpha<3$ values. Much testing is necessary to ascertain the final usefulness and validity of these ideas, as several conceptual obstacles remain., Comment: 3 eps figures, attempting again to fix margins