# Transverse ac Driven and Geometric Ratchet Effects for Vortices in Conformal Crystal Pinning Arrays

###### Abstract

A conformal pinning array is created by taking a conformal transformation of a uniform hexagonal lattice to create a new structure in which the six-fold ordering of the original lattice is preserved but which has a spatial gradient in the pinning site density. With a series of conformal arrays it is possible to create asymmetric substrates, and it was previously shown that when an ac drive is applied parallel to the asymmetry direction, a pronounced ratchet effect occurs with a net dc flow of vortices in the same direction as the ac drive. Here we show that when the ac drive is applied perpendicular to the substrate asymmetry direction, it is possible to realize a transverse ratchet effect where a net dc flow of vortices is generated perpendicular to the ac drive. The conformal transverse ratchet effect is distinct from previous versions of transverse ratchets in that it occurs due to the generation of non-Gaussian transverse vortex velocity fluctuations by the plastic motion of vortices, so that the system behaves as a noise correlation ratchet. The transverse ratchet effect is much more pronounced in the conformal arrays than in random gradient arrays, and is absent in square gradient arrays due the different nature of the vortex flow in each geometry. We show that a series of reversals can occur in the transverse ratchet effect due to changes in the vortex flow across the pinning gradient as a function of vortex filling, pinning strength, and ac amplitude. We also consider the case where a dc drive applied perpendicular to the substrate asymmetry direction generates a net flow of vortices perpendicular to the dc drive, producing what is known as a geometric or drift ratchet that again arises due to non-Gaussian dynamically generated fluctuations. The drift ratchet is more efficient than the ac driven ratchet and also exhibits a series of reversals for varied parameters. Our results should be general to a wide class of systems undergoing nonequilibrium dynamics on conformal substrates, such as colloidal particles on optical traps.

###### pacs:

74.25.Wx,74.25.Uv## I Introduction

When a particle interacts with an asymmetric substrate under an ac drive, it is possible to realize a ratchet effect in which a net dc motion of the particle can occur. If the ac drive is in the form of a periodic external force, the system is characterized as a rocking ratchet, while if the particles are undergoing thermal agitation and the substrate is periodically flashed on and off, the system is known as a flashing ratchet 1 ; 2 . It is also possible to generate dc motion of a particle in an asymmetric substrate in the absence of periodic driving when the noise fluctuations experienced by the particle are not white but have an additional time correlation. Such systems are known as correlation ratchets 3 ; 4 ; 5 ; 6 . Another type of ratchet system, the geometric or drift ratchet, arises when dc driven particles move through an assembly of asymmetric obstacles. The interactions of the particles with the obstacles produces an additional drift of particles in the direction perpendicular to the dc drive 7 ; 8 ; 9 ; 10 ; 11 . One system in which a remarkably rich variety of rocking ratchet effects has been realized is vortices in type-II superconductors interacting with some form of asymmetric substrate where an ac driving force is applied along the direction of the substrate asymmetry 12 ; 12N ; 14 ; 15 ; 16 ; 17 ; 18 ; 19 ; 20 ; 21 ; 22 ; 23 ; 24 . Such asymmetric pinning arrays include periodic one-dimensional (1D) asymmetric saw-tooth thickness modulations 12 ; 19 ; 23 ; 25 ; 26 , channels with funnel shapes 12N ; 20 ; 21 ; 27 ; 28 , and arrays of pinning sites in which the individual pinning sites have an intrinsic asymmetry 15 ; 17 ; 18 ; 22 ; 24 ; 29 ; 30 . In the case where the vortex-vortex interactions are weak, such as at low magnetic fields, the net dc flow is in the easy direction of the substrate asymmetry, but when collective vortex interactions come into play it is possible to observe reversals of the ratchet effect in which the net dc vortex motion switches to follow the hard direction of the substrate asymmetry, and there may be multiple ratchet reversals as a function of vortex density and ac driving amplitudes 15 ; 18 ; 19 ; 25 ; 26 ; 33 ; 34 ; 35 ; 36 ; 37 . It is also possible to create transverse vortex ratchets where a net dc flow of vortices occurs that is perpendicular to the applied ac driving force 31 ; 38 ; 36a ; 39 ; 40 . Such transverse ratchets have been realized for vortices interacting with arrays of triangular pins or other pinning site shapes that have an intrinsic asymmetry. When vortices interact with such pinning sites, they are partially deflected in the direction perpendicular to the ac force 39 ; 40 .

Vortex ratchet pinning geometries can be created using individual pinning sites that are symmetric by introducing a periodic spatial gradient in the pinning site density 14 ; 41 ; 42 ; 43 or a gradient in the size of the pinning sites 44 , and then applying an ac driving force along the direction of the gradient. Gradient ratchets have been studied for randomly arranged pinning with a spatial gradient 41 and periodically arranged pinning arrays with gradients 41 ; 42 ; 43 ; 44 . A new type of pinning geometry called a conformal crystal was recently proposed 47 . It is constructed by applying a conformal transformation to a uniform hexagonal array, resulting in a structure where each pinning site has six neighbors but there is a spatial gradient in the pinning site density 45 ; 46 . Simulations indicate that the pinning effectiveness for a fixed number of pinning sites is maximized for the conformal arrays compared to uniform random arrays or a random arrangement of pinning sites with a gradient 47 . The conformal array also produces more effective pinning than uniform periodic pinning arrays except for fields at or very near integer matching fields. This optimal pinning by conformal arrays was subsequently confirmed in experiments 48 ; 49 . Other studies have also shown that non-conformal gradient pinning arrays produce enhanced pinning compared to uniform arrays 50 ; M . Simulations demonstrate that the hexagonal conformal arrays exhibit stronger pinning than arrays constructed by conformally transforming square or quasiperiodic arrays, while square pinning arrays containing 1D spatial density gradients have enhanced pinning compared to the hexagonal conformal arrays at certain magnetic field fillings due to commensuration effects 51 .

Vortices interacting with a series of conformal pinning arrays can exhibit a ratchet effect when an external ac drive is applied along the substrate asymmetry direction, and multiple reversals in the ratchet effect can occur as a function of vortex density, pinning force, and drive amplitude due to changes in the vortex flow patterns, as shown in Ref. 52 . In general, the ratchet effect is most pronounced for the conformal pinning arrays compared to random gradient or square gradient pinning arrays; however, at low magnetic fields the ratchet effect is strongest for the square gradient array due to a 1D channeling of vortices along the symmetry direction of the square array.

For conformal or other gradient pinning arrays, it has not been considered whether a transverse ratchet effect could occur when the ac drive is applied perpendicular to the substrate asymmetry direction. Since the individual pinning sites are symmetric in these arrays, it might be expected that transverse ratchet effects would not occur. Here we show that a pronounced transverse vortex ratchet effect appears in conformal pinning arrays, and that the mechanism responsible for this effect differs from that which produces the transverse ratchet effects found for asymmetrically shaped pins or obstacles. For the conformal array, the transverse ratchet effect results when plastic vortex motion generates strong non-Gaussian velocity fluctuations both parallel and perpendicular to the ac driving direction. Due to the pinning gradient, the fluctuations can be stronger in the low pinning density portions of the sample, creating an effective gradient in the size of the fluctuations and producing a dynamical thermophoresis phenomenon. The effect can also be viewed as a realization of a noise correlation ratchet. Noise correlation ratchets were first proposed by Magnasco 3 and Doering et al. 4 , who introduced different types of noise correlations directly into the fluctuating noise term governing the equation of motion of a particle placed on an asymmetric substrate. In the conformal pinning array system that we consider, there is no added stochastic noise term in the vortex equation of motion, so the correlated velocity fluctuations are dynamically generated by the plastic motion of the vortices. The onset of plastic flow in vortex systems has been demonstrated both in experiments 53 and simulations to generate strong non-Gaussian vortex velocity fluctuations both parallel and perpendicular to the external driving direction 54 ; 55 . We find that transverse ratchet effects do not occur in square gradient arrays since the vortex trajectories are nearly one-dimensional in the direction parallel to the ac drive, so the transverse fluctuations are too weak to induce transverse ratchet motion. We show that it is possible to realize geometric or drift ratchet effects in the conformal and random gradient arrays when a dc drive is applied perpendicular to the asymmetry direction and a net vortex drift arises that is perpendicular to the dc drive. Geometrical ratchets have been studied for the dc flow of particles through periodic arrays of asymmetric objects, and can arise even in the limit of a single particle 7 ; 8 ; 9 ; 10 ; 11 . In our system the pinning sites are symmetric but a geometric ratchet effect occurs due to the dynamically generated fluctuations from the plastic flow of vortices. In the single particle limit where collective plastic flow is lost, the transverse or geometric ratchet effect is also absent. The geometric ratchet we study has similarities to the geometric ratchet effect proposed by Kolton 56 for particles in a two-dimensional (2D) system moving over a periodic 1D asymmetric substrate containing additional random pinning sites that create nonequilibrium fluctuations, which result in a drift of particles perpendicular to the dc drive direction. We also find that there can be a series of reversals in both the ac and dc driven transverse ratchet effects which is unique among transverse and drift ratchet systems that have been previously studied. Finally, our results should be general to a variety of of other systems which can be described as an assembly of interacting particles moving over a conformal pinning array, such as colloids on optical trap arrays 57 .

## Ii Simulation and System

We consider a 2D simulation geometry with periodic boundary conditions in the and -directions. Within the system we place vortices, where the number of vortices is proportional to the applied magnetic field which is aligned in the direction. We introduce pinning sites to the sample, and denote the field at which the number of vortices equals the number of pinning sites as . The dynamics of an individual vortex is governed by the following overdamped equation of motion:

(1) |

Here is the damping constant which is set equal to . The first term on the right is the repulsive vortex-vortex interaction force , where is the location of vortex , is the modified Bessel function, is the penetration depth, , , , is the flux quantum, and is the permittivity. The initial vortex positions before application of an external driving force are obtained by performing simulated annealing from a high temperature state down to The vortex-pinning interaction term , where is the Heaviside step function, is the pinning radius, is the pinning strength, is the location of pinning site , , and . All forces are measured in units of and lengths in units of .

Figure 1(a) shows the pinning geometry with three conformal crystals, each of width , placed sequentially along the direction of the sample. We refer to this geometry as “ConfG.” Each conformal array is generated by performing a conformal transformation of a uniform triangular lattice in the complex plane, , where and are integers and is the lattice constant. The transformation that maps the points from the original lattice to the plane is

(2) |

where is a parameter. This transformation maps a semi-annular section covering the region of the original lattice to a rectangular region. We use , an outer annulus radius of , an inner annulus radius of , and set the lattice constant to . As shown in Fig. 1(a) the easy flow direction is along the negative direction. We also consider a system with a spatial density gradient of randomly placed pinning sites, shown in Fig. 1(b) and referred to as “RandG,” as well as a non-conformal gradient array constructed by introducing a one-dimensional density gradient to a regular square lattice, illustrated in Fig. 1(c) and referred to as “SquareG.”

In previous work examining longitudinal ratchet effects, the ac drive was applied along the -direction or parallel to the substrate asymmetry direction, and the net motion was also measured along the asymmetry direction 52 . In the present work we apply the ac driving along the -direction or perpendicular to the substrate symmetry direction, as indicated in Fig. 1, but still measure the net flow of vortices along the direction, such that a finite dc flow signature indicates the existence of a transverse ratchet effect. The ac driving term is , where is the ac amplitude. To measure the ratchet effect we sum the vortex displacements in the direction perpendicular to the ac drive to obtain , where is the position of vortex at time and is the initial position of the vortex when the ac drive is first applied. We also measure the corresponding using the net displacements along the direction, and find that, due to the lack of substrate asymmetry along the direction, for all ac drives that we consider. We focus on the case with a fixed ac period of 8000 simulation time steps unless otherwise noted, and allow the system to run for 50 ac cycles before beginning the measurement of to avoid any transient effects, as was done in previous studies of longitudinal ratchet effects 52 . We also consider the case in which we apply a finite dc force along the positive -direction, and examine after the equivalent of 50 ac cycles of time has passed to quantify the geometric ratchet effect. Finally, we consider the effect of adding thermal fluctuations using , which has the properties and , where is the Boltzmann constant. Unless otherwise noted, we set .

## Iii Transverse ac Ratchet Effect

In Fig. 2(a) we plot , the net displacement per vortex in the -direction, verus time in ac drive cycles for the ConfG conformal array in Fig. 1(a) at and . For , indicating that there is no transverse ratchet effect. At a finite transverse ratchet effect emerges and the vortices each move an average of in the negative direction during 50 ac drive cycles. The vortices translate along the easy flow direction of the substrate, and the displacement increases linearly with time indicating that transient effects are not present. For , ratcheting still occurs but the effect is reduced, with the vortices moving an average of in the negative direction during 50 ac drive cycles.

Figure 2(b) shows at , , and for the different array types. In the SquareG square gradient array from Fig. 1(c), after 50 ac cycles, while the RandG random gradient array from Fig. 1(b) exhibits ratcheting that is 20 times less effective than in the ConfG array, which is also shown for comparison. To better quantify the ratchet as a function of , in Fig. 3(a) we plot the value of at the end of ac drive cycles for ConfG, RandG, and SquareG arrays at and . In the ConfG array, when , and the ratchet reaches its maximum efficiency near , after which gradually approaches zero with increasing . The RandG array shows similar behavior, but has a much weaker overall ratchet effect and exhibits an efficiency maximum at . In the SquareG array, for all values of .

To explore the role of vortex-vortex interactions in the ratchet effect, in Fig. 3(b) we plot the value of after 50 ac drive cycles versus for the ConfG system in Fig. 3(a) at , , , , and . The overall effectiveness of the ratchet decreases with decreasing , and for the ratchet effect is absent. Whenever , there is a portion of the ac cycle during which all of the vortices are depinned and moving, so the loss of the ratchet effect at low is not caused by the vortices becoming pinned when their density is small. Instead, this result indicates that collective vortex-vortex interactions are important for the transverse ratchet to occur.

In Fig. 4(a) we plot the instantaneous vortex positions, vortex trajectories, and pinning site locations for the ConfG array in Fig. 3(a) at where there is no ratcheting. Here the vortex motion is confined to the low pinning density regions of the sample. For lower values of , the width of the regions of moving vortices decreases. For , all the vortices are able to move, as shown in Fig. 4(b) for where finite ratcheting in the negative -direction occurs. The vortices do not move strictly along the or driving direction, but follow winding trajectories that introduce strong -direction velocity fluctuations. In a SquareG array with the same parameters, no ratcheting occurs. Figure 4(c) shows that at in the SquareG array, the vortex trajectories are strongly one-dimensional and are oriented along the direction with few or no fluctuations along the direction. In Fig. 4(d), at for the SquareG array all the vortices can participate in the flow during some portion of the ac cycle, but again the motion follows nearly straight trajectories along the -direction and there is no ratchet effect. As increases above , the direction meandering of the vortex trajectories in the ConfG array is progressively reduced, and this coincides with the drop in ratchet efficiency shown in Fig. 3(a). The vortex trajectories for the RandG array are similar in appearance to those shown for the ConfG array.

The transverse ratchet in the ConfG and RandG arrays can be understood as a realization of a noise correlation ratchet where the correlated noise is generated by the plastic flow of the vortices. To clarify this, we examine the -component velocity distributions of the individual vortices over a fixed time of 50 ac drive cycles. In Fig. 5(a) we plot for the ConfG system from Fig. 3 with and . At , where there is a strong ratchet effect, differs significantly from the simple Gaussian form , which is plotted as a smooth solid line. For where there is no ratcheting, there is a strong peak in at due to the pinned vortices, and the magnitude of the velocity fluctuations are significantly reduced compared to the case. For where the ratchet effect is present but weak, as shown in Fig. 3(a), the width of is smaller than at the optimal ac drive of .

In Fig. 5(b) we plot for the ConfG, SquareG, and RandG arrays from Fig. 3(a) at . The width of is much narrower for the SquareG array than for the ConfG and RandG arrays due to the strongly 1D nature of the vortex trajectories in the SquareG array, as shown in Fig. 4(d). In the RandG array, for is nearly identical to that of the ConfG array; however, close to the RandG array lacks the pinned vortex peak found in the ConfG array and instead maintains a Gaussian form of . For , the ConfG array has a reduction in compared to the RandG array; this asymmetry in is discussed further in Section IV.B. The Gaussian nature of the dynamically generated fluctuations in the RandG array produces weak time correlations in the velocity, leading to a ratchet effect that is weaker than that found in the ConfG array. In the SquareG array the velocity is strongly peaked at a single value, meaning that there are insufficient velocity fluctuations to generate a ratchet effect.

### iii.1 Transverse Ratchet Reversals

It is also possible to realize a reversal of the transverse ratchet effect where the net flow of vortices is in the positive or hard direction of the substrate asymmetry. Such motion is termed a reversed ratchet effect, and it is marked by net flow in the positive direction due to the orientation of our ratchet potential. In Fig. 6 we plot as a function of time for a ConfG array with and . At , the vortices are translating in the positive direction, corresponding to a reversed ratchet effect, while at there is a weak normal ratchet effect in the negative direction and at there is a stronger normal ratchet effect in the negative direction, showing that a ratchet reversal occurs as a function of vortex density. In Fig. 7 we show versus for the ConfG array at fixed and varied . For in Fig. 7(a), there is a normal ratchet effect in the negative direction with a magnitude that is largest for . There are local maxima in the ratchet effectiveness near and , and the ratchet effect disappears for since the pinning is weak enough that the vortices start to form a uniform triangular lattice at the higher fields. In Fig. 7(b) at , there is a normal ratchet effect that is suppressed for where the vortices are unable to depin. The ratchet effect is larger in magnitude than for the case, and the maximum ratchet effectiveness occurs just above . The strong variations in the ratchet effect as a function of field reflect the occurrence of partial commensuration effects, with the most pronounced ratchet motion appearing near , 2.0, 4.0, and . Interestingly, there is no commensuration effect near ; this is in contrast with previous studies of vortex ordering in uniform triangular lattices, where ordered triangular vortex lattices associated with peaks in the critical depinning force occur at matching fields of , 3.0, and , while there is much weaker matching at and when ordered but nontriangular vortex lattices form 58 . The commensurability effects may also be different for square conformal arrays, as studies of uniform square pinning arrays show that different kinds of vortex configurations are stable at different integer matching fields 58 ; New1 ; New2 . Due to the gradient in the ConfG array, commensuration conditions can occur in only part of the sample at a time, as illustrated in previous simulations, so that the matching effects are not centered at integer ratios of 51 . In Fig. 7(c) we plot versus for the same ConfG system at that is highlighted in Fig. 6. The ratchet effect is absent for , while for the vortices exhibit a reversed ratchet effect and flow in the positive direction. For the vortices flow in the negative direction to produce a normal ratchet effect.

The switch from a reversed to a normal ratchet effect occurs due to changes in the -direction fluctuations of the vortices moving along the pinning gradient. When is weak, vortex motion occurs across the entire pinning gradient and the largest transverse fluctuations of the flowing vortices take place in the least densely pinned portions of the sample, while vortices spend more time pinned in the most densely pinned regions, producing smaller transverse velocity fluctuations as shown in Fig. 4(a). In analogy with the thermophoretic effect, in which particles preferentially drift from hotter to colder portions of a sample 59 , the vortices tend to drift from the low pinning density regions to the high pinning density regions, so that within an individual substrate ratchet plaquette, the vortices move in the negative direction. For the normal transverse ratchet effect, the pinning establishes a vortex density gradient that is maximum on the high pinning density side of each substrate ratchet plaquette, and this vortex density gradient breaks an additional symmetry for the fluctuation-induced vortex drift, preventing vortices from moving in the positive direction in order to pass directly from the lowest pinning density strong velocity fluctuation region to the highest pinning density small velocity fluctuation region. When is strong, the flow in the regions with low pinning density primarily consists of interstitial vortices that are not trapped in pinning sites moving between occupied pinning sites. The resulting winding flow creates smooth velocity fluctuations in the direction. In the regions with high pinning density, the small spacing between pinned vortices forces the interstitial vortices to approach the pinned vortices much more closely, and the resulting vortex-vortex interaction forces depin the pinned vortices, generating enhanced fluctuations in in the high pinning density region. As a result, the effective temperature gradient induced by the velocity fluctuations is reversed compared to the case of low , leading to a reversal of the ratchet flow direction.

In Fig. 8 we show a heightfield of the vortex trajectories in a portion of the ConfG sample, obtained by rasterizing the vortex trajectories onto a fine grid over the course of 50 ac drive cycles and measuring the total number of trails that pass through each grid point. At , , and in Fig. 8(a), the ratchet effect is in the normal negative -direction. Vortices flow through the pinning sites, and the vortex trajectories fluctuate the most in the low pinning density portion of the sample. In the high pinning density area, the channeling of the vortices through successive pinning sites suppresses velocity fluctuations transverse to the driving direction. At , , and , the ratchet effect is reversed and the vortices translate in the positive direction. In the highest pinning density portion of the sample, vortex motion occurs via a combination of purely interstitial vortex flow in the middle of the illustrated region and hopping of vortices from one pinning site to the next at the top and bottom of the illustrated region, producing enhanced velocity fluctuations in the direction. In the regions with low pinning density, the vortices at the pinning sites remain pinned most of the time and the vortex motion consists almost entirely of smooth interstitial flow with reduced velocity fluctuations in the -direction. As is increased, the vortices in the low pinning density region begin to depin more frequently, while the flow in the high pinning density region shifts from partially interstitial to channelling along the pins, shifting the relative magnitude of the -velocity fluctuations so that it is highest in the low pinning density region, and switching the ratchet effect back to the normal negative direction. In Fig. 7(c), where is held fixed at as is increased, more vortices occupy the low pinning density regions of the sample as becomes larger, and the increased strength of the vortex-vortex interactions causes the vortices located at the pinning sites to depin. As a result, the magnitude of the -velocity fluctuations in the low pinning density regions increases as the magnetic field increases, leading to the reversal of the ratchet effect.

The ratchet reversals can also occur at a fixed field when the ac driving amplitude is varied, as shown in Fig. 9(a) where we plot versus for a ConfG array with at . There is no ratchet effect for . A reversed ratchet effect with vortex flow in the positive direction occurs for , followed by a region of normal ratchet motion in the negative direction for . The ratchet flow is in the reversed positive direction again for . In the range , the vortex motion in the high pinning density regions of the sample occurs through a mixed interstitial and channelling flow of the type illustrated in Fig. 8(b), while for , the ac drive is large enough to depin all the vortices, producing strongly disordered flow throughout the sample and resulting in a normal negative ratchet effect.

The mechanism responsible for the appearance of the second reversed, positive ratcheting region for differs from that found at lower ac drive. At high ac drives, over a portion of the drive cycle the driving current is large enough to induce dynamical ordering of the vortices in the low pinning density portions of the sample. This lowers the velocity fluctuations in the low pinning density portions of the sample compared to other areas of the sample where the vortex lattice remains more disordered. As a result, the effective shaking temperature is highest in the high pinning density region of the sample even though all the vortices are flowing. Previous simulations and experiments on dc driven vortices in samples with uniform pinning density show that the velocity fluctuations are suppressed when the system enters a dynamically ordered state due to a decrease in the effective shaking temperature 53 ; 56 ; 60 ; 61 ; 62 . Since the drive required to induce dynamical ordering increases with increasing pinning density, specific portions of the ConfG sample become ordered for certain values of the ac driving amplitude. This produces an effective temperature gradient across each pinning plaquette, with the largest effective temperature on the high pinning density side. In Fig. 10, we show snapshots of Voronoi constructions obtained from the system in Fig. 9(a). Figure 10(a) illustrates the configuration when the driving amplitude reaches its maximum magnitude in the direction for a sample with , where there is a normal negative direction ratchet effect. Topological defects are uniformly distributed throughout the sample. Figure 10(b) shows the same point in the ac drive cycle for a sample with , where there is a reversed positive direction ratchet effect. The more disordered regions are correlated with the regions of high pinning density, where the effective shaking temperature temperature is higher. As increases further, vortices throughout the sample are able to dynamically reorder during the portion of the drive cycle at which attains its maximum value, so the effective temperature gradient becomes spatially flat and the ratchet effect is reduced. This is shown in Fig 9(a) at the highest values of .

In Fig. 9(b) we plot versus in the ConfG array with at . Here the normal negative ratchet effect persists over the larger range before the ratcheting switches into the reversed positive direction due to the dynamical ordering effects. The increase in the region over which there is a normal ratchet effect occurs because when the vortex density is lower, a larger external drive must be applied to induce dynamical ordering in the low pinning density regions. In Fig. 9(c) at , the ratchet reversals are lost and there is a normal ratchet effect over the range . There is no longer a reversed ratchet effect at low due to the lack of interstitial vortices in the high pinning density portion of the array. At magnetic fields that are this low, all of the vortices in the high pinning density areas are trapped in singly or doubly occupied pinning sites, and there are no remaining freely flowing vortices that could knock one of the pinned vortices out of a pinning site and generate fluctuations in the direction velocity. As a result, the high pinning density portion of the sample is effectively frozen and prevents the vortices in the lower pinning density portions of the sample from translating in the direction. It is possible that for ac drives larger than those illustrated in Fig. 9(c), a reversed ratchet effect may appear due to the occurrence of partial dynamical ordering in the sample.

In Fig. 11(a) we plot versus in a ConfG array with and . There is no ratchet effect for since for strong pinning all the vortices remain localized at pinning sites during the entire ac drive cycle. For , shown in Fig. 11(b), the ratchet effect is initially in the normal negative direction for , and then switches to the reversed positive direction for . Another switch to the normal negative direction ratchet occurs at , and the normal ratchet effect gradually diminishes to zero at the highest values of . The onset of the second normal negative ratchet regime is correlated with the appearance of doubly occupied pinning sites in the high pinning density regions. These act to reduce the direction velocity fluctuations of the flowing interstitial vortices, so that the effective temperature in the high pinning density region is smaller than in other portions of the sample. Figure 11(c) shows that at , there is a large regime of normal negative ratchet behavior extending from , followed by a reversed positive direction ratchet flow for .

### iii.2 Thermal Fluctuations

Although thermal fluctuations alone do not produce a ratchet effect in the ConfG array in the absence of an external drive, they can enhance the transverse ratchet effect in some cases. In Fig. 12(b) we plot versus for a ConfG sample with and , where is the magnitude of the fluctuations in the thermal force term added to the vortex equation of motion. At and , all the vortices are pinned during the entire ac drive cycle and ; however, as increases a finite ratchet effect emerges that exhibits a maximum efficiency at before dropping back to zero for higher values of . At , there is a weak ratchet effect at which undergoes more than a tenfold increase in magnitude for increasing , reaching its maximum efficiency near . At there is a robust ratchet effect at which shows a small enhancement in magnitude to a maximum efficiency at before dropping to zero at . For there is a similar trend, with a maximum efficiency at . For higher values of where the vortices spend more time in motion during each drive cycle, the addition of thermal effects generally decreases the ratchet efficiency. These results show that when thermal effects are important, the transverse ratchet effect remains robust and can even be enhanced.

In Fig. 12(a) we examine the effects of changing the ac drive period for the same system in Fig. 12(b) at . All of the results presented so far were obtained with an ac drive period of simulation time steps. For , the vortices are pinned during the entire ac drive cycle so that independent of the value of the ac period. For , 1.0, and , the ratchet is weak at small ac drive periods since each vortex simply moves back and forth within its local potential minimum, so that there is no generation of the plastic flow necessary for the transverse ratchet effect to occur. As the ac drive period increases, the ratchet effectiveness increases linearly. This indicates that low frequency ac drive cycles produce stronger transverse ratchet effects.

### iii.3 Comparison To Longitudinal Gradient Ratchets

In previous work, we applied an ac drive in the -direction, parallel to the substrate asymmetry direction, and demonstrated the existence of a longitudinal ratchet effect in the ConfG array by measuring 52 . In Fig. 13(a) we show the time evolution of during ac drive cycles for a ConfG sample with and for , where a transverse ratchet effect appears, and for , where there is a longitudinal ratchet effect. There are strong oscillations in for the longitudinal ratchet that arise because the ac drive direction is the same as the ratchet motion direction. The longitudinal ratchet effect is approximately 2.5 times larger than the transverse ratchet effect, and we find that in general the longitudinal ratchet is more effective than the transverse ratchet by a similar ratio for other parameters. The longitudinal ratcheting is stronger since it is a rocking ratchet effect. In contrast, in the transverse ratchet effect the ac drive does not directly push the vortices in the direction of the asymmetry but instead generates plastic flow, which creates the transverse velocity fluctuations that permit a correlation ratchet effect to occur. We find a similar ratio of the longitudinal to transverse ratchet effects for the RandG arrays as well (not shown); however, the overall ratchet effect is smaller in each case for the RandG array than for the ConfG array. In Fig. 13(b) we plot versus time for a SquareG array under the same and ac driving conditions. Here there is no transverse ratchet effect, but there is still a longitudinal ratchet effect which is about four times less effective than the longitudinal ratchet effect for the ConfG array. In general we find that the transverse ratchet effect is more sensitive to changes in magnetic field than the longitudinal ratchet effect since commensuration effects strongly influence the magnitude of the dynamical fluctuations responsible for the transverse ratchet effect.

## Iv Drift Ratchet

We next consider the case where instead of an ac drive, we apply a dc drive in the -direction, and we measure the net drift of vortices in the -direction to examine the drift or geometric ratchet effect. In Fig. 14(a) we plot versus time in simulation time steps for the ConfG array at , , and varied . For , although there is flow in the -direction, there is no drift of the vortices in the -direction, while at there is a pronounced -direction drift, with individual vortices moving an average of in the negative direction after simulation time steps. In comparison, for an ac drive of during the same amount of time (equivalent to 50 ac drive cycles), Fig. 3(a) indicates that individual vortices move an average distance of only in the negative direction, showing that the transverse drift ratchet is approximately three times more effective at transporting the vortices than the ac transverse ratchet effect. For , Fig. 14(a) shows that the transverse drift is reduced. In Fig. 14(b) we plot versus time at for the SquareG, RandG, and ConfG arrays. There is no transverse drift in the SquareG array since the vortices move in predominately straight trajectories along the -direction. The RandG array shows a transverse drift that is approximately 10 times smaller than the transverse drift in the ConfG array.

In order to compare to the ac driven results, we measure for the dc driven system after simulation time steps, which corresponds to the same time interval required to complete 50 ac drive cycles in the ac driven system with a period of 8000 simulation time steps. In Fig. 15(a) we plot versus for ConfG, RandG, and SquareG arrays with and . The transverse drift is strongest for the ConfG array and rapidly drops off when . In the ac driven systems, strong ratchet effects can persist for since there is a portion of the ac cycle during which the driving force is smaller than so that plastic flow can occur. In the dc driven case, however, when all the vortices are moving at all times and the plastic flow necessary to produce the correlation ratchet effect is lost. The transverse drift ratchet is smaller in the RandG array but persists over a wider range of values of due to the stronger dispersion in the pinning forces for the random array caused by overlap of pinning sites in some locations, which creates local regions where the effective pinning force is larger than . In the SquareG array, the vortex trajectories are one-dimensional along the direction for all values of , so there is no transverse drift ratchet effect. Figure 15(b) shows for the same samples plotted against at fixed . The transverse drift in the ConfG and RandG arrays is lost for low pinning forces when the vortex flow is elastic, as well as at large pinning forces where all the vortices remain pinned. If we apply the external dc drive in the negative -direction (not shown), we find exactly the same drift ratchet effects, with the vortices still moving in the negative -direction.

The drift ratchet effect we observe is different in nature from transverse drift ratchets studied by other groups 7 ; 8 ; 9 ; 10 ; 11 . In those systems, the particle-particle interactions are not important and the ratchet effect arises when particles are deflected during collisions with obstacles or pinning sites, producing a net drift. In contrast, the transverse drift ratchet effect we observe is produced by transverse nonequilibrium fluctuations generated by particle-particle interactions, leading to the emergence of an effective noise correlation ratchet.

### iv.1 Drift Ratchet Reversals

In previously studied drift ratchets, reversals in the direction of drift were not observed 7 ; 8 ; 9 ; 10 ; 11 ; 56 . In the drift ratchet described here, there can be reversals of the drift direction in both the ConfG and RandG arrays. In Fig. 16 we plot versus for ConfG, RandG, and SquareG arrays with and . The conformal array shows a negative drift ratchet effect with local efficiency maxima at and , while the drift is suppressed for . In comparison, in the RandG array the magnitude of the drift is smaller but there are multiple reversals from a normal negative to a reversed positive drift ratchet effect. The SquareG array shows no transverse drift.

For the ConfG array, drift ratchet reversals generally occur for stronger pinning and fillings of . In Fig. 17 we plot versus time in a ConfG array with and . At there is no transverse drift effect, while at there is a normal negative drift and at there is a reversed positive drift, indicating a reversal in the drift direction as a function of magnetic field. We again find that the magnitude of the transverse drift for the dc driven systems is significantly larger than the ac driven transverse ratchet effect.

In Fig. 18(a) we plot after simulation time steps versus for the ConfG array from Fig. 17 with and . For , , while the drift is in the normal negative direction for and in the reversed positive direction for . There are several local extrema in the drift which correspond to changes in the vortex flow. For , shown in Fig. 18(b), the drift is mostly in the normal negative direction with only a small region of reversed positive direction drift near . There are also several local extrema in the drift near , 2.7, and . For in Fig. 18(c), the drift is always in the normal negative direction and is largest over the range . There are some small fluctuations in the drift near which corresponds to the point at which the pinning is strong enough that almost all of the pinning sites are doubly occupied.

Ratchet reversals can also occur as a function of the dc drive magnitude. In Fig. 19(a) we plot versus for a ConfG array at and . There is a reversed positive direction drift for which is correlated with all the pinning sites in the sample being doubly occupied. The interstitial vortices in the high pinning density regions are close enough to the doubly occupied pins to cause vortices to depin, while in the low pinning density regions the interstitial vortices move around the occupied pinning sites and do not induce any vortex depinning. As a result, the transverse fluctuations are largest in the high pinning density portions of the sample, and the drift motion is in the reversed positive direction. For , the drive is large enough that all the vortices in doubly occupied pinning sites can be depinned, producing drift in the normal negative direction, while for the transverse velocity fluctuations become homogeneous throughout the sample and the ratchet effect is lost. Figure 19(b) shows that at , there are no longer enough interstitial vortices in the high pinning density regions of the sample to easily depin vortices from the doubly occupied sites, so only a normal negative direction drift appears. In Fig. 19(c), at there is a normal negative direction drift ratchet effect only in the range

### iv.2 Transverse Velocity Fluctuations

We next consider the drift ratchet transverse velocity fluctuations measured in the same way as for the ac driven transverse ratchet effect in Section III. In Fig. 20(a) we plot for the ConfG array at and . At , where there is no transverse drift, there is a strong peak in at since a portion of the vortices are permanently pinned. For where there is a strong transverse drift ratchet effect, the magnitude of the peak in is reduced since the vortices are only temporarily rather than permanently pinned, while remains large over a wider range of values. At where the drift effect is diminished, the width of is strongly reduced since the vortices are moving primarily along straighter trajectories aligned with the -direction. In Fig. 20(b) we plot for the ConfG, RandG, and SquareG arrays at , , and . The SquareG array has a strong peak in at since the vortices are moving in 1D paths in the -direction. For the RandG array the velocity fluctuations can be fit to a Gaussian curve as indicated by the smooth solid line. When there is a strong transverse drift ratchet effect, is asymmetric about , as shown in Fig. 20(c) where we plot separately for and in the ConfG array at , , and . There is a clear difference in the velocity distributions for vortices moving in the positive and negative directions. The smooth curve is a Gaussian fit highlighting the non-Gaussian nature of the fluctuations. In Fig. 20(d) we show for positive and negative in the same ConfG array at where there is no net drift. Here the velocity distribution is symmetric and can be fit to a Gaussian tail away from .

## V Ratchet Effects for Colloidal Particles

The results we find should be general to a wide class of systems of interacting particles moving over a gradient substrate array where nonequilibrium transverse fluctuations can be dynamically generated. For example, our results could be applied to charge-stabilized colloids interacting with optical trap arrays 57 ; 63 . Charged colloids can be modeled as overdamped particles interacting via a repulsive Yukawa or screened Coulomb potential , where is the screening length and is proportional to the effective charge on the particle. Compared to superconducting vortices, the colloids have a much shorter range interaction but a sharper repulsion at the shortest distances. The effective charge and the screening length can be tuned readily in the colloidal system by modifying the ion concentration in the solution. If transverse ratchets can be realized in a colloidal system, they could potentially provide a new spatial separation technique in which a mixture of colloidal species driven over a substrate has one species gradually move further in the drift direction than the other. In previous studies of the longitudinal ratchet effect on conformal substrates, colloidal particles exhibited a robust ratchet effect similar to the superconducting vortices 52 .

In Fig. 21(a) we plot versus for colloids driven oven a ConfG array with a transverse ac drive for at a filling fraction of . In Fig. 21(b) we plot the corresponding fraction of six-fold coordinated colloids versus . For the smallest values of , the colloids are weakly interacting and become localized at the pinning sites so that there is no transverse ratchet. At intermediate values of , where , the colloids move in the negative direction and exhibit a normal transverse ratchet effect, while at slightly higher values of there is a small window in which the ratchet effect is in the reversed positive direction and . This reversed ratchet effect is similar to what we observe for large values of in the vortex system, and arises when the colloids dynamically order in the low pinning density portions of the sample but remain partially disordered in the high pinning density portions of the sample. For higher values of , when the colloids form a uniform triangular lattice with , the flow becomes elastic and the ratchet effect is lost. Figure 21(c) shows versus for dc driven colloids in the drift ratchet configuration where the drive is applied along the -direction. Here there is only a normal negative direction transverse drift effect, and the magnitude of is larger than that for the ac driven sample in Fig. 21(a). At high enough values of , the system again forms a triangular solid with and the transverse drift is lost. These results show that the transverse ratchet and drift ratchet along with ratchet reversals can also be realized in colloidal systems.

## Vi Summary

Using numerical simulations we show that vortices on conformal pinning arrays driven by an ac force applied perpendicular to the array asymmetry direction exhibit a novel transverse ratchet effect where there is a net drift of vortices perpendicular to the ac drive. This effect arises when the dynamically generated nonequilibrium fluctuations which have non-Gaussian characteristics combine with the asymmetry of the substrate to create what is known as a noise correlation ratchet. This ratchet effect is distinct from previously observed transverse vortex ratchet effects that are caused by geometric deflection of individual vortex trajectories. In our system, the correlated noise is generated by the plastic flow of vortices, which creates strong non-Gaussian velocity fluctuations in the direction perpendicular to the ac drive. We find that the transverse ratchet effect is absent for square gradient arrays since these produce less meandering of the vortex trajectories and therefore have reduced velocity fluctuations. The effect is present for random gradient arrays but is substantially weaker than that produced by the conformal array. We show that it is possible to realize a series of reversals in the direction of flow of the transverse ratchet due to changes in the spatial flow pattern of the vortices across gradient. These reversals can occur as a function of vortex density, ac driving amplitide, and pinning strength. In general the transverse ratchet effect has a magnitude that is about one-half to one-third the size of the longitudinal ratchet effect observed for the same conformal pinning arrays. We also examine the case where a dc drive applied perpendicular to the asymmetry of the substrate produces what is known as a geometric or drift ratchet, where a net flux of vortices perpendicular to the dc drive occurs. The maximum efficiency of the drift ratchet is about 2.5 times larger than that of the corresponding ac driven transverse ratchet. The drift ratchet is also a realization of a noise correlation ratchet, and we find that it exhibits reversals in the drift direction as a function of field, drive amplitude, and pinning strength. Our results should be general to a wide class of systems of interacting particles undergoing dynamically generated fluctuations when driven over a a conformal array, including colloidal particles in optical trap arrays.

## Vii Acknowledgments

This work was carried out under the auspices of the NNSA of the U.S. DoE at LANL under Contract No. DE-AC52-06NA25396.

## References

- (1) P. Reimann, Phys. Rep. 361, 57 (2002).
- (2) P. Hänggi and F. Marchesoni, Rev. Mod. Phys. 81, 387 (2009).
- (3) M.O. Magnasco, Phys. Rev. Lett. 71, 1477 (1993).
- (4) C.R. Doering, W. Horsthemke, and J. Riordan, Phys. Rev. Lett. 72, 2984 (1994).
- (5) P. Galajda, J. Keymer, P. Chaikin, and R. Austin, J. Bacteriol. 189, 8704 (2007).
- (6) M.B. Wan, C. J. Olson Reichhardt, Z. Nussinov, and C. Reichhardt, Phys. Rev. Lett. 101, 018102 (2008).
- (7) A. Van Oudenaarden and S.G. Boxer, Science 285, 1046 (1999).
- (8) D. Ertas, Phys. Rev. Lett. 80, 1548 (1998); T.A.J. Duke and R.H. Austin, Phys. Rev. Lett. 80, 1552 (1998).
- (9) C. Keller, F. Marquardt, and C. Bruder, Phys. Rev. E 65, 041927 (2002).
- (10) J. Herrmann, M. Karweit, and G. Drazer, Phys. Rev. E 79, 061404 (2009).
- (11) S. Savel’ev, V. Misko, F. Marchesoni, and F. Nori, Phys. Rev. B 71, 214303 (2005).
- (12) C.S. Lee, B. Jankó, I. Derényi, and A.L. Barabási, Nature (London) 400, 337 (1999).
- (13) J.F. Wambaugh, C. Reichhardt, C.J. Olson, F. Marchesoni, and F. Nori, Phys. Rev. Lett. 83, 5106 (1999).
- (14) C. J. Olson, C. Reichhardt, B. Jankó, and F. Nori, Phys. Rev. Lett. 87, 177002 (2001).
- (15) J.E. Villegas, S. Savel’ev, F. Nori, E.M. Gonzalez, J.V. Anguita, R. García, and J.L. Vicent, Science 302, 1188 (2003).
- (16) R. Wördenweber, P. Dymashevski, and V. R. Misko, Phys. Rev. B 69, 184504 (2004).
- (17) J. Van de Vondel, C. C. de Souza Silva, B. Y. Zhu, M. Morelle, and V. V. Moshchalkov, Phys. Rev. Lett. 94, 057003 (2005).
- (18) C.C. de Souza Silva, J. Van de Vondel, M. Morelle, and V.V. Moshchalkov, Nature (London) 440, 651 (2006).
- (19) Q. Lu, C. Reichhardt, and C. Reichhardt, Phys. Rev. B 75, 054502 (2007).
- (20) K. Yu, T.W. Heitmann, C. Song, M.P. DeFeo, B.L.T. Plourde, M.B.S. Hesselberth, and P.H. Kes, Phys. Rev. B 76, 220507(R) (2007).
- (21) B.L.T. Plourde, IEEE Trans. Appl. Supercond. 19, 3698 (2009).
- (22) B.B. Jin, B.Y. Zhu, R. Wördenweber, C.C. de Souza Silva, P.H. Wu, and V.V. Moshchalkov, Phys. Rev. B 81, 174505 (2010).
- (23) V.A. Shklovskij and O.V. Dobrovolskiy, Phys. Rev. B 84, 054515 (2011).
- (24) A. Palau, C. Monton, V. Rouco, X. Obradors, and T. Puig, Phys. Rev. B 85, 012502 (2012).
- (25) V.I. Marconi, Phys. Rev. Lett. 98, 047006 (2007).
- (26) V.A. Shklovskij, V.V. Sosedkin, and O.V. Dobrovolskiy, J. Phys.: Condens. Matter 26, 025703 (2014).
- (27) N.S. Lin, T.W. Heitmann, K. Yu, B.L.T. Plourde, and V.R. Misko, Phys. Rev. B 84, 144511 (2011).
- (28) G. Karapetrov, V. Yefremenko, G. Mihajlović, J.E. Pearson, M. Iavarone, V. Novosad, and S.D. Bader, Phys. Rev. B 86, 054524 (2012).
- (29) D. Perez de Lara, F.J. Castaño, B.G. Ng, H.S. Korner, R.K. Dumas, E.M. Gonzalez, K. Liu, C.A. Ross, I.K. Schuller, and J.L. Vicent, Phys. Rev. B 80, 224510 (2009).
- (30) V. Rouco, A. Palau, C. Monton, N. Del-Valle, C. Navau, A. Sanchez, X. Obradors and T. Puig, New J. Phys. 17, 073022 (2015).
- (31) L. Dinis, E.M. González, J.V. Anguita, J.M.R. Parrondo, and J.L. Vicent, Phys. Rev. B 76, 212507 (2007).
- (32) L. Dinis, E.M. González, J.V. Anguita, J.M.R. Parrondo, and J.L. Vicent, New J. Phys. 9, 366 (2007).
- (33) D. Perez de Lara, A. Alija, E.M. Gonzalez, M. Velez, J.I. Martín, and J.L. Vicent, Phys. Rev. B 82, 174503 (2010).
- (34) D. Perez de Lara, M. Erekhinsky, E.M. Gonzalez, Y.J. Rosen, I.K. Schuller, and J.L. Vicent, Phys. Rev. B 83, 174507 (2011).
- (35) J. Van de Vondel, V.N. Gladilin, A.V. Silhanek, W. Gillijns, J. Tempere, J.T. Devreese, and V.V. Moshchalkov, Phys. Rev. Lett. 106, 137003 (2011).
- (36) C. J. Olson Reichhardt and C. Reichhardt, Physica C 432, 125 (2005).
- (37) E.M. Gonzalez, N.O. Nunez, J.V. Anguita, and J.L. Vicent, Appl. Phys. Lett. 91, 062505 (2007).
- (38) A.V. Silhanek, J. Van de Vondel, V.V. Moshchalkov, A. Leo, V. Metlushko, B. Ilic, V.R. Misko, and F.M. Peeters, Appl. Phys. Lett. 92, 176101 (2008).
- (39) D. Perez de Lara, L. Dinis, E.M. Gonzalez, J.M.R. Parrondo, J.V. Anguita, and J.L. Vicent, J. Phys.: Condens. Matter 21, 254204 (2009).
- (40) L. Dinis, D. Perez de Lara, E.M. Gonzalez, J.V. Anguita, J.M.R. Parrondo, and J.L. Vicent, New J. Phys. 11, 073046 (2009).
- (41) T.C. Wu, R. Cao, T.J. Yang, L. Horng, J.C. Wu, and J. Kolacek, Solid State Commun. 150, 280 (2010).
- (42) T.C. Wu, L. Horng, J.C. Wu, R. Cao, J. Kolacek, and T.J. Yang, J. Appl. Phys. 102, 033918 (2007).
- (43) T.-C. Wu, L. Horng, and J.-C. Wu, J. Appl. Phys. 117, 17A728 (2015).
- (44) W. Gillijns, A.V. Silhanek, V.V. Moshchalkov, C.J. Olson Reichhardt, and C. Reichhardt, Phys. Rev. Lett. 99, 247002 (2007).
- (45) D. Ray, C.J. Olson Reichhardt, B. Jankó, and C. Reichhardt, Phys. Rev. Lett. 110, 267001 (2013).
- (46) P. Pieranski, in Phase Transitions in Soft Condensed Matter, T. Riste and D. Sherrington, Eds. (Plenum, New York, 1989), p. 45; F. Rothen, P. Pieranski, N. Rivier, and A. Joyet, Eur. J. Phys. 14, 227 (1993).
- (47) F. Rothen and P. Pieranski, Phys. Rev. E 53, 2828 (1996).
- (48) Y.L. Wang, M.L. Latimer, Z.L. Xiao, R. Divan, L.E. Ocola, G.W. Crabtree, and W.K. Kwok, Phys. Rev. B 87, 220501(R) (2013).
- (49) S. Guénon, Y.J. Rosen, A.C. Basaran, and I.K. Schuller, Appl. Phys. Lett. 102, 252602 (2013).
- (50) M. Motta, F. Colauto, W.A. Ortiz, J. Fritzsche, J. Cuppens, W. Gillijns, V.V. Moshchalkov, T.H. Johansen, A. Sanchez, and A.V. Silhanek, Appl. Phys. Lett. 102, 212601 (2013).
- (51) V.R. Misko and F. Nori, Phys. Rev. B 85, 184506 (2012).
- (52) D. Ray, C. Reichhardt, and C. J. Olson Reichhardt, Phys. Rev. B 90, 094502 (2014).
- (53) C. Reichhardt, D. Ray, and C. J. Olson Reichhardt, Phys. Rev. B 91, 184502 (2015).
- (54) A.C. Marley, M.J. Higgins, and S. Bhattacharya, Phys. Rev. Lett. 74, 3029 (1995).
- (55) C.J. Olson, C. Reichhardt, and F. Nori, Phys. Rev. Lett. 80, 2197 (1998).
- (56) A.B. Kolton, D. Domínguez, and N. Grønbech-Jensen, Phys. Rev. Lett. 83, 3061 (1999).
- (57) A.B. Kolton, Phys. Rev. B 75, 020201 (2007).
- (58) K. Xiao, Y. Roichman, and D.G. Grier, Phys. Rev. E 84, 011131 (2011).
- (59) C. Reichhardt, C. J. Olson, and F. Nori, Phys. Rev. B 57, 7937 (1998).
- (60) G.R. Berdiyorov, M.V. Milosevic, and F.M. Peeters, Phys. Rev. B 74, 174512 (2006).
- (61) G.R. Berdiyorov, M.V. Milosevic, and F.M. Peeters, New J. Phys. 11, 013025 (2009).
- (62) C. Ludwig, Sitzungber Bayer. Akad. Wiss. Wien Math-Naturwiss. Kl. 20, 539 (1856); S. Duhr and D. Braun, Proc. Natl. Acad. Sci. (U.S.A.) 103, 19678 (2006).
- (63) C.J. Olson, C. Reichhardt, and F. Nori, Phys. Rev. Lett. 81, 3757 (1998).
- (64) A.E. Koshelev and V.M. Vinokur, Phys. Rev. Lett. 73, 3580 (1994).
- (65) A.B. Kolton, R. Exartier, L.F. Cugliandolo, D. Domínguez, and N. Grønbech-Jensen, Phys. Rev. Lett. 89, 227001 (2002).
- (66) J. Mikhael, J. Roth, L. Helden, and C. Bechinger, Nature (London) 454, 501 (2008).