banner image
No data available.
Please log in to see this content.
You have no subscription access to this content.
No metrics data to plot.
The attempt to load metrics for this article has failed.
The attempt to plot a graph for these metrics has failed.
The dynamics of network coupled phase oscillators: An ensemble approach
Rent this article for
View: Figures


Image of FIG. 1.
FIG. 1.

In-degree distributions for the Erdös-Renyi and scale-free networks used in this paper. The Erdös-Renyi network’s degree distribution (dot-dashed line) is peaked around 100. Past a minimum degree, the scale-free network takes on a degree distribution (dotted line) of the form , as is more clearly seen in the inset, which is the same plot shown on a log-log scale.

Image of FIG. 2.
FIG. 2.

Eigenspectrum plots for the three networks used in this paper: (a) a directed network with uniform in-degree, (b) an undirected Erdös-Renyi network, and (c) an undirected scale-free network (). In all cases, and . Since the Erdos-Renyi and scale-free graphs are undirected, all eigenvalues in those cases are real.

Image of FIG. 3.
FIG. 3.

Bulk order parameter r vs. time for systems simulated using the theta formulation (solid line) [Eqs. (1) and (2)], as well as our ensemble formulation (dashed line) [Eqs. (15) and (16)], performed on the networks introduced in Sec. I B: (a) uniform in-degree, (b) Erdös-Renyi, and (c) scale-free. Results were generated numerically using a fourth-order Runge-Kutta integration scheme with fixed time step. Each curve represents a single simulation—no curves are averaged. A time step Δt = 0.1 was used for all theta formulation simulations, save for the scale-free, for which Δt = 0.05 was used, while all ensemble formulation simulations used a time step ten times larger than that was used for the corresponding theta formulation simulations. The width of the frequency distribution was set to Δ = 0.1 and the coupling strength to . Theta formulation curves were shifted horizontally to provide best apparent fit.

Image of FIG. 4.
FIG. 4.

Same as Fig. 3 but with .

Image of FIG. 5.
FIG. 5.

Same as Figs. 3 and 4 but with .

Image of FIG. 6.
FIG. 6.

Long-time-averaged values of r vs. k for systems simulated using the theta formulation (solid squares) [Eqs. (1) and (2)] and our ensemble formulation (open circles) [Eqs. (15) and (16)] as well as ρ calculated from the transcendental equation (solid line) [Eq. (26)]. The data presented corresponds to the three networks described in Sec. I B. Black points correspond to the uniform in-degree (upper) and scale-free (lower) networks, while the grey points correspond to the Erdös-Renyi network. Also shown as dashed lines are the critical coupling value k c , which is approximately the same for all three networks, and the values of ρmax [Eq. (27)] for the three networks. The same integration scheme was used as for Figs. 3–5. Simulations were generally run for 300 time units for the theta formulation simulations, with averaging done over the last 50 time units, while the ensemble formulation simulations were run until they converged (generally between 200 and 500 time units). Selected points were re-run at smaller time step size and longer simulation runtime to ensure validity.

Image of FIG. 7.
FIG. 7.

and vs. for two different values of . When , there is no nonzero intersection of the two curves [thus, no nonzero solution to Eq. (28)].

Image of FIG. 8.
FIG. 8.

Bulk order parameter r vs. t for our uniform in-degree network simulated using our ensemble formulation [Eqs. (15) and (16)] (dashed line) plotted with ρ calculated from Eq. (31) (solid line). The normalized L 2 deviation, D, from the manifold given by Eq. (29) is also plotted (dotted line) along with the slope given by Eq. (40) (dash-dotted line).

Image of FIG. 9.
FIG. 9.

(a) Bulk order parameter r plotted vs. time for a theta formulation simulations [Eq. (1)] on our uniform in-degree network using a bimodal distribution (solid line) and for a simulation using Eqs. (50) and (51) (dashed line). A time step of 0.05 was used for both simulations. The parameters of the simulation were k = 40, Δ = 0.5, and ω0 = 0.2. (b) A parametric polar plot (a, ψ) of the same simulations, starting at incoherent initial conditions (r ≪ 1).

Image of FIG. 10.
FIG. 10.

Phase diagram in parameter space showing regions corresponding to different attractor types denoted by I (incoherent steady-state attractor at r = 0), SS (steady-state attractor with r > 0), and LC (limit cycle attractor corresponding to time periodic variation of r). Bifurcations of these attractors occur as the region boundaries are crossed.34 The dashed horizontal lines at correspond to the scans of parameter shown in Fig. 11.

Image of FIG. 11.
FIG. 11.

Long-time behavior of r vs. for systems simulated using theta formulation [Eqs. (1) and (2)], plotted in black, and for identical systems simulated using our simplified bimodal ensemble formulation [Eqs. (50) and (51)], plotted in grey, for four different values of : (a) , (b) , (c) , and (d) . When the theta formulation attractors are apparently steady state (SS) or incoherent (I) (in both cases some noise is always present), we discard the first 1000 time units of our simulations (a time step size of 0.05 time units was used), time average of the results over the next 1000 time units,35 and plot the averages as solid black squares. When the theta formulation results are apparently limit cycles (noisy), the results are plotted as vertical bars indicating the range of r values in the oscillation. Similarly, when the reduced formulation gives oscillation, the vertical range of the grey gives the oscillation range of r. Vertical dashed lines represent the region boundaries of Fig. 10. The coupling strength k was held fixed at k = 40. Simulations were performed on the uniform in-degree network introduced in Sec. I B. Whereas the data in previous figures were obtained by starting the simulations in an incoherent state; simulations in this figure were run twice, once from an incoherent state and again from a coherent initial condition (obtained by pre-running the simulations for large k).

Image of FIG. 12.
FIG. 12.

(Color) Polar plot of for a variety of initial conditions for and . The solid lines represent simulations performed using our reduced ensemble equations [Eqs. (50) and (51)] and are color-coded to indicate which attractor each simulation ended on (blue for synchronized steady-state, red for incoherent). The locations of each attractor and of a saddle point are marked by grey dots. The regions surrounding each attractor are blown up in (b) (SS), and (c) (I), with orbits from theta formulation simulations [Eq. (1)] shown in black with transients removed.

Image of FIG. 13.
FIG. 13.

(Color) Polar plot of for a variety of initial conditions for and . The solid lines represent simulations performed using our reduced ensemble equations [Eqs. (50) and (51)] and are color-coded to indicate which attractor each simulation ended on (blue for synchronized steady-state, red for incoherent). The location of each attractor and of the saddle point is marked by a grey dot. (b) A magnification of the region of interest, with points on the orbit of a theta formulation simulation [Eq. (1)] plotted in black, showing the system starting in the incoherent attractor and escaping to the steady-state attractor. (c) plotted vs. time for the same theta formulation simulation plotted in (b).

Image of FIG. 14.
FIG. 14.

Variance of plotted vs. for fixed values of , and (a) , corresponding to an I attractor and (b) , corresponding to an SS attractor. For each value of , three simulations were run using the same undirected uniform in-degree network with a given value of (generated in the same manner as described in Sec. I B), but using different independent random realizations of the oscillator frequencies. A time step of Δt = 0.05 and a simulation run time of 5000 time units (with averaging done over the last 4000 time units) were used throughout.

Image of FIG. 15.
FIG. 15.

Two of the graphs from Fig. 11 re-plotted with uniform in-degree data replaced with data corresponding to simulations done on the Erdös-Renyi network were introduced in Sec. I B.


Article metrics loading...


Full text loading...

This is a required field
Please enter a valid email address
752b84549af89a08dbdd7fdb8b9568b5 journal.articlezxybnytfddd
Scitation: The dynamics of network coupled phase oscillators: An ensemble approach