Electronic and Mechanical Realizations of One-way Coupling in One and Two Dimensions

Barbara J. Breen, Aaron B. Doud, Jamie R. Grimm, Andrew H. Tanasse, and Stuart J. Tanasse
Physics Department, The University of Portland, Portland, Oregon 97203, USA

John F. Lindner and Katsuo Maxted

Physics Department, The College of Wooster, Wooster, Ohio 44691, USA (Dated: July 20, 2010) One-way or unidirectional coupling is a striking example of how topological considerations the parity of an array of multistable elements combined with periodic boundary conditions can qualitatively regulate dynamics. Here we introduce a simple electronic model of unidirectional coupling in one and two dimensions and experimentally compare it to an improved mechanical model and an ideal mathematical model. Using experiment and computation, we demonstrate the richer phenomenology of one-way coupling in higher dimensions. In media where bidirectional coupling would dissipate all excitations, soliton-like waves propagate in dierent directions with dierent speeds, provided the coupling is suciently strong.
PACS numbers: 05.45.Yv, 82.40.Bj, 89.75.Kd

I.

INTRODUCTION

II.

IDEAL REVERSER ARRAY

One-way coupling has the notable eect of facilitating the propagation of solitary waves or solitons [1] in media that would normally be dissipative. It is a new paradigm for the eects of topology on dynamics [410]. This type of coupling, described in detail below, was rst introduced by In et al. to improve the performance of uxgate magnetometers [4]. Recently, Lindner et al. realized a simple mechanical model of generic one-way coupling that facilitates study of arrays of many elements [11]. Here we describe an even simpler electronic model of one-way coupling that generalizes to two and higher dimensions. We experimentally compare the behavior of the electronic model in one-dimension with an improved version of the mechanical model and demonstrate their qualitatively similar dynamics. In addition, the compact electronic array permits the experimental exploration of two-dimensional one-way coupled arrays. We compare the behavior of the twodimensional arrays with predictions from theory and simulation, where soliton-antisoliton pairs spatially separate domains of degenerate ground states. For dierent initial conditions, solitons propagate at dierent speeds and directions corresponding to topologically distinct modes. Common to our ideal mathematical model and our mechanical and electronic apparati is the logical idea of an reverser, which we explicate below. The paper is structured in the following way. Section II quantitatively describes an idealized model of one-way coupling, including its extension to two dimensions. Section III describes a rened mechanical one-way coupled array. Section IV introduces an electronic one-way coupled array, and compares its dynamics to that of the mechanical apparatus in one-dimension. It also reports the richer dynamics of a two-dimensional one-way electronic apparatus and compares it with the idealized model of Section II. Section V oers conclusions and future work.

A bistable potential V = a2 + b4 connes a torsional oscillator of rotational inertia I and viscosity . Link such oscillators two-ways with coupling k , so the equations of motion are x + x = ax b3 k (x1 x ) k (x+1 x ), (1) I x with periodic boundary conditions 0 = N and 1 = N +1 , for x = 1, 2, . . . , N , where the overdots indicate dierentiation with respect to time t. Delete one of the coupling terms to link the oscillators one-way, so the equations of motion become x + x = ax b3 k (x1 x ) I x
3 = Ax bx kx1 ,

(2)

where A = a + k . Drop the inertial term to focus on the over-damped case, so x = x 3 x 1 x1 , (3)

where the rescaled angle x = x b/A, the rescaled coupling 1 = k/A = k/(a + k ), and the prime indicates differentiation with respect to the rescaled time = tA/ . The nal term 1 x1 in Eq. 3 is the reverser, whose torque is reversely proportional to the previous oscillators deection. For two dimensions, generalize this to x,y = x,y 3 x,y 2 (x,y 1 + x1,y ), (4)

with periodic boundary conditions, where the rescaled coupling 2 = k/(a + 2k ). In D dimensions, D = k/(a + Dk ). Note that each oscillator is inuenced only from below and left (on a standard cartesian plane). Therefore the coupling direction is from bottom-left to top-right.

1:1

3:1

3:5

FIG. 1: (Color online) Angles x (left) and angular velocities x (right) as a function of time in an N = 21 array for coupling 1 = 0.74; cylindrical view emphasizes periodic boundary conditions. Odd number of elements forbids completely alternating ground state. Solitons annihilate in pairs, leaving one soliton to propagate endlessly.

FIG. 2: (Color online) Angles x,y (left) and angular velocities x,y (right) for two-dimensional soliton m:n modes in a 4141 array for coupling 2 = 0.74; toroidal views emphasize doubly periodic boundary conditions. Modes are robust with respect to noise and disorder. Color coding as in Fig. 1.

The phenomenology of the one-dimensional array is well known [8]. From random initial conditions, moving discontinuities referred to as solitons separate regions of opposite degenerate ground state in which oscillators alternate between their two stable equilibria. Soliton speed increases with coupling, but also depends on noise and disorder and on an exponentially attenuated nonlinear interaction. These are soliton-antisoliton pairs in the sense that a trailing soliton reverses the oscillator equilibria left by a leading soliton, and they annihilate in pairs upon close approach. In even arrays, annihilations result in a global ground state of quiescent oscillators. However in odd arrays, one soliton is always left over, propagating endlessly in a frustrated attempt to reach an impossible global equilibrium, as in Fig. 1. Here we extend these results to higher dimensions using computer simulations. We numerically integrate Eq. 4 using a fourth-order Runge-Kutta algorithm implemented in C/C++ with typical time steps of dt = 0.01. We check this with BDF and Adams integrators implemented using Mathematica. In two dimensions, periodic boundary conditions enforce a toroidal topology. Moving discontinuities still separate regions of degenerate ground states, but are now spatially extended from points to lines. Topologically distinct modes move at discretely dierent speeds and directions vmn , where m :n represent a soliton (or domain

wall) that wraps m times in the toroidal direction for n times in the poloidal direction, as in Fig. 2. These modes are robust with respect to noise and disorder. For small coupling, individual oscillators are stuck in their own equilibria and nothing propagates. For large coupling, about half of randomly selected initial conditions result in static dislocations orthogonal to the coupling direction, but all others induce solitons. Almost all these solitons are in the fast symmetric 1:1 mode but about one in a hundred are in the slower asymmetric 1:3 and 3:1 modes at angles of 23 from the coupling direction. For suciently large arrays, about one in 10 000 is in the 3:3 mode, as summarized by Fig. 3. For large coupling 2 , random initial conditions generate additional very slow modes at large angles > 45 that are mirror images of the asymmetric modes. For example, a -1:3 mode is the mirror image of a 1:3 mode, where the negative sign refers to the slope of the wavefront. In practice the more complex modes are only possible in large arrays. Soliton speed vmn increases monotonically with coupling 2 , as in Fig. 4. The two-dimensional one-way array has a maximum soliton speed v11 proportional to the coupling. This maximum speed is only realized for solitons that propagate in the coupling direction v11 . Asymmetrical modes that propagate at angle from v11 propagate more slowly, as in Fig. 5. A typical elastic medium like a drumhead supports wave propagation at one speed. One-way coupling in two dimensions supports wave propagation at multiple speeds in multiple directions. These waves can serve, for example, as the basis of clocks with dierent frequency stan-

(a)
+1:3 1:1 1:3

3:3

1.2 1.0 0.8 0.6 0.4 0.2

1:1 3:5 1:3 1:5 1:7 1:13 1:11 1:7 1:5 1:3 3:5 1:1 0.0 0.2 0.4 0.6 0.8 1.0

+3:1 1:1 static

0.0

(b)
10 10 10 10 10 10
0

3:1 1:1 static

1:1

-1

FIG. 4: (Color online) Simulated soliton speeds vmn as a function of coupling 2 for selected modes m:n of a 41 41 array. Weak coupling isolates individual oscillators at their equilibria. Negative modes with angles > 45 (red squares) are mirror images of asymmetric positive modes with angles < 45 (blue circles).
1.2 1.0 0.8

-2

1:3 & 3:1

-3

-4

0.6

-5

3:3
3 7 11 15 19 23 27 31 35 39 43 47 51 55 59

0.4 0.2 0.0 0 30 45 60 90

FIG. 3: (Color online) (a) Simplest modes of a 15 15 array and their propagation velocity vectors vmn for coupling 2 = 1. (b) Steady state probabilities P for random initial conditions as a function of array size N indicate the relative sizes of the basins of attraction for the modes in (a).

dards, the frequencies determined by the array size and the soliton direction, with the latter selected by choosing specic initial conditions.

FIG. 5: (Color online) Simulated soliton speeds vmn as a function of angle for coupling 2 = 1 for selected modes m:n of a 41 41 array. Angle separates the propagation of the fast 1:1 mode and the m:n mode. Dashed line is v11 cos2 . Inset depicts the corresponding velocity vectors. Colors as in Fig. 4.

III.

MECHANICAL REVERSER ARRAY

A recent paper [11] describes a mechanical apparatus built at The College of Wooster that realizes one-way coupling and uses it to investigate the annihilation of soliton-antisoliton pairs. Here, we improve this device and use it as a benchmark for our electronic apparatus described below in Sec. IV. Like the Wooster apparatus, our device employs bistable oscillators consisting of inverted pendulums balanced by restoring springs. If one oscillator rotates clockwise, a mechanical reverser, described by Fig. 6, rotates the next oscillator counterclockwise. The reverser works

by deecting a water jet o a seesaw connected to the next oscillator, thereby exploiting the fact that the downward force of the water jet is independent of the transverse force that directs it. Unlike the Wooster device, where water weight supplements jet pressure, our device uses jet pressure alone to torque each oscillator, thereby more closely mimicking the 1 x1 reverser terms in the idealized one-dimensional array of Sec. II. To test our array, we record the times to annihilation T as a function of the initial soliton separation n for an N = 16 mechanical array, as reported in Fig. 7. As expected, the larger the initial separations of the solitons, the longer they survive before annihilation. Such studies are dicult computationally because of the very long transient times, but the results agree well with previous

4 rent to ow through the p-channel producing high output voltage, as in Fig. 8. The high and low voltages correspond to the bistable states of the oscillators in the ideal array. Connecting an odd number of such inverters (or NOT gates) in series with periodic boundary conditions forms a ring oscillator. Such oscillators are well known in electrical engineering [12]. For example, they often provide the frequency standard for phase lock loop control systems, with frequencies typically in the megahertz range. Recently, ring oscillators have been connected by diodes to create a chaotic circuit [13]. We modify the basic design to create simple physical models of one-way coupling, in both one and two dimensions A 16V constant voltage across the inverters powers the coupling, a 470F capacitance slows the solitons to one second time scales, and bi-colored light-emitting diodes (LEDs) visualize the currents. We use the open source analog electronic circuit simulator SPICE [14] to vet our designs. (For our larger arrays, we generate the SPICE code algorithmically using Mathematica.) The phenomenology of the electronic arrays resembles that of the ideal and mechanical arrays. For example, by briey connecting two nodes for various times in the even one-dimensional array of Fig. 10, we create a solitonantisoliton pair and recorded its time to annihilation, as reported in Fig. 11. As with the mechanical array, the larger the initial separations of the solitons, the longer they survive before annihilation.

FIG. 6: (Color online) Mechanical reverser mimics the Eq. 3 reverser coupling term. When the previous water jet (dashed) deects the inverted pendulum counterclockwise, the linked arm deects the next water jet clockwise (left), and vice versa (right). Red disks represent xed axles parallel to the array.
200

150

100

50

p
0 0 5

p low out low in

p high out

10

15

high in n p n

FIG. 7: Experimental distribution of 150 mechanical array annihilation times T in seconds for solitons initially separated by N oscillators in an array of length N = 16.

work [11]. Asymmetries in the plot probably reect slight rotational asymmetries in the oscillators.

IV.

ELECTRONIC REVERSER ARRAY

FIG. 8: (Color online) Electronic CMOS inverter mimics the Eq. 3 reverser coupling term. High input voltage causes current to ow through the n-channel discharging the capacitor and producing low output voltage (left); low input voltage causes current to ow through the p-channel charging the capacitor and producing high output voltage (right).

Generalizing the mechanical reverser array to study one-way coupling dynamics in two dimensions may not be practical. Instead, we fabricate an electronic circuit that mimics the mathematical reverser terms in the idealized arrays of Sec. II and the mechanical reverser of Sec. III. The key component is a Complementary Metal Oxide Silicon (CMOS) inverter, which contains both an n-type and a p-type eld eect transistor. High input voltage causes current to ow through the n-channel producing low output voltage, while low input voltage causes cur-

To create an electronic analog of two-dimensional oneway coupling, note that the Eq. 4 coupling term 2 (x,y1 + x1,y ) = 22 x,y1 + x1,y 2 (5)

is proportional to the average of adjacent inverters. Accomplish this by using the Fig. 9(a) passive average cir-

(a)

(b)

60 50 40 30 20 10

FIG. 9: (a) Passive average circuit. (b) Simplied schematic for two-dimensional CMOS inverter array. In addition to periodic boundary conditions, each node connects to a capacitor (to control the time scale) in series with a bi-colored light emitting diode (to visualize the currents caused by passing solitons).

0 0 5 10

15

20

FIG. 11: Experimental distribution of 305 electronic array annihilation times T in seconds for solitons initially separated by N oscillators in an array of length N = 24.

cuit, whose output = IR = V1 /R1 + V2 /R2 = V1 + V2 , V 1/R1 + 1/R2 2 (6)

provided the resistances R1 = R2 . Connect two CMOS inverter outputs to each node, as in Fig. 9(b), with periodic boundary conditions; explicit resistors to augment the line resistance are unnecessary. As in the onedimensional array, to control the time scale and visualize the currents, at each node connect a capacitor and a bi-colored LED to ground. Figure 12 shows a 13 13 electronic array built on prototype boards sustaining a soliton in a 1:1 mode.

FIG. 12: (Color online) Inexpensive electronic 13 13 array of CMOS inverters and capacitors on prototype boards (background) facilitates exploration of one-way coupling in two dimensions. Sequence of a soliton in a 1:1 mode visualized by a square grid of LEDs on a printed circuit board (foreground).

V.

CONCLUSIONS

FIG. 10: (Color online) Inexpensive electronic N = 24 array of CMOS inverters, capacitors, and LEDs on prototype boards facilitates exploration of one-way coupling dynamics. Sequence shows a timer circuit induced soliton-antisoliton pair.

One-way coupling is not merely a mathematical phenomenon; we have realized two radically dierent instantiations (mechanical and electrical) of the same physics; the key feature is coupling that reverses adjacent states. We have also demonstrated that one-way coupling induces rich additional behavior in two-dimensions. Why study one-way coupling? One-way coupling is fascinating and so far largely unexplored nonlinear dynamics. Current research takes one-way coupling from ideality to reality, while future research may take it from

6 reality to practicality. Two-way coupled bistable media quickly dissipate all excitations. By contrast, one-way coupling the same media with cyclic boundary conditions can induce oscillations and wave propagation indenitely. Changing the parity (oddness or evenness) of the array radically changes the dynamics. The system seeks a global lowestenergy alternating state, but oddness forever frustrates this equilibrium, resulting in a kind of perpetual motion! (Energy is conserved because the coupling must be powered.) The solitary waves are models of solition-antisoliton pairs; what the leading one does the trailing one undoes, and they are always created or annihilated in pairs. Indeed, in two-dimensions, one-way coupled solitons are like domains walls separating degenerate vacua in particle physics. Damped dynamical systems typically have short transients followed by indenite steady states. These systems have transients that increase exponentially with the separation of the solition-antisoliton pairs, resulting in the tall peaks of Figs. 7 and 11. Practically speaking, for long N > 100 arrays, transients are all one ever observes. Typical media, like drumheads, oer just one speed of propagation, the speed of sound. By contrast, one-way coupled media oer discretely dierent speeds in dierent direction (although never backwards) for dierent initial conditions. The soliton modes are topologically robust with respect to noise; converting a 1:1 mode to a 1:3 mode requires increasing noise (or heating) to completely randomize the entire array and then decreasing noise (or cooling) to fall into a dierent basin of attraction. Solitons circulate in these array with robust frequencies that increase linearly with coupling and number of elements (with the latter characteristic varying only discretely) so that they might serve as exible frequency standards or clocks. Noise and disorder cause the solitons to slow and speed, approach and annihilate, so homogeneous arrays with a soliton-antisoliton pair started far apart might serve as noise detectors: if the array is later observed to be quiescent, noise must have destroyed its solitons. One-way coupling is an extreme example of wave propagation in anisotropic media, such as light in calcite. Some of its signature features might be present but overlooked in more (but not completely) isotropic media. One-way coupling implies one-way propagation. Natural examples of the latter include wave propagation in the (supersonic) solar wind, light propagation near a black hole, and nerve propagation along axons. In the future, we hope to use electric analogs of oneway coupling in two and higher dimensions to investigate the dynamical eects of noise and disorder (temporal and spatial inhomogeneities). Early computer simulations already suggest additional phenomena in three dimensions. Field Programmable Gate Arrays (FPGAs) might facilitate realization of larger arrays.

Acknowledgments

This research was supported by the Clare Boothe Luce Program of the Henry Luce Foundation and NSF Grant No. DMR-0649112. J. F. L. thanks The College of Wooster for making possible his sabbatical at the University of Portland. We thank Allen R. Hansen and Craig E. Henry for assistance in building our arrays.

[1] T. Dauxois and M. Payrard, Physics of Solitons (Cambridge University Press, 2006) [2] J. M. Carcione Wave Fields in Real Media: Wave Propagation in Anisotropic, Anelastic, Porous and Electromagnetic Media, 2nd Ed. (Elsevier Science, 2007) [3] A. A. Fuki, Y. A. Kravtsov, O. N. Naida Geometrical Optics of Weakly Anisotropic Media (CRC Press, 1998) [4] V. In, A. R. Bulsara, A. Palacios, P. Longhini, A. Kho, J. D. Ne, Phys. Rev. E 68, 045102(R) (2003). [5] A. R. Bulsara, V. In, A. Kho, P. Longhini, A. Palacios, W. Rappel, J. Acebron, S. Baglio, B. Ando, Phys. Rev. E 70, 036103 (2004). [6] J. A. Rogge, D. Aeyels, J. Phys. A 37, 11135 (2004). [7] V. In, A. R. Bulsara, A. Palacios, P. Longhini, A. Kho, Phys. Rev. E 72, 045104(R) (2005). [8] J. F. Lindner, A. R. Bulsara, Phys. Rev. E 74, 020105

(2006). [9] V. In, A. Palacios, A. R. Bulsara, P. Longhini, A. Kho, J. D. Ne, S. Baglio, B. Ando, Phys. Rev. E 73, 066121 (2006). [10] A. R. Bulsara, J. F. Lindner, V. In, A. Kho, S. Baglio, V. Sacco, B. Ando, P. Longhini, A. Palacios, W. Rappel, Phys. Lett. A 353, 4 (2006). [11] J.F. Lindner, K.M. Patton, P.M. Odenthal, J.C. Gallagher, B.J. Breen, Phys. Rev. E 78, 066604 (2008). [12] J. A. McNeill and D. Ricketts, The Designers Guide to Jitter in Ring Oscillators (Springer, 2009). [13] Y. Hosokawa and Y. Nishio, Int. J. Bifurcat. Chaos 14, 2513-2524 (2004). [14] K. Kundert, The Designers Guide to SPICE and Spectre (Springer, 1995).