Magnetic Pendulum Strange Attractor

This chaotic strange attractor represent the final resting positions for a magnetic pendulum suspended over some magnets (shown as black dots). It kind of looks like mixed paint. The 2D animation shows what happens as you decrease the damping factor. The 3D animation was shaded by path length.

Mathematica 4.2 version: 1/24/05; C++ version: 10/27/07; 3D rendered in POV-Ray 3.1

(* runtime: 25 seconds, increase n for higher resolution *)
n = 40; h = 0.25; g = 0.2; mu = 0.07; zlist = {Sqrt[3] + I, -Sqrt[3] + I, -2I};
image = Table[z2 = z[25] /. NDSolve[{z''[t] == Plus @@ ((zlist - z[t])/(h^2 + Abs[zlist - z[t]]^2)^1.5) - g z[t] - mu z'[t], z[0] == x + I y, z'[0] == 0}, z, {t, 0, 25}, MaxSteps -> 200000][[1]]; r = Abs[z2 - zlist]; i = Position[r, Min[r]][[1, 1]]; Hue[i/3], {y, -5.0, 5.0, 10.0/n}, {x, -5.0,5.0, 10.0/n}];
Show[Graphics[RasterArray[image]], AspectRatio -> 1];

The picture on the left shows another version with five magnets. See also my physics pendulums.

Here is some Mathematica code to numerically solve this using the Beeman integration scheme with the predictor-corrector modification:
(* runtime: 45 seconds, increase n for higher resolution *)
n = 40; tmax = 25; dt = 0.1; h = 0.25; g = 0.2; mu = 0.07; zlist = {Sqrt[3] + I, -Sqrt[3] + I, -2I};
image = Table[z = x + I y; v = a = a1 = 0; Do[z += v dt + (4a - a1)dt^2/6; vpredict = v + (3a - a1)dt/2; a2 = Plus @@ ((zlist - z)/(h^2 + Abs[zlist - z]^2)^1.5) - g z - mu vpredict; v += (2a2 + 5a - a1)dt/6; a1 =a; a = a2, {t, 0, tmax, dt}]; r = Abs[z - zlist]; Hue[Position[r, Min[r]][[1, 1]]/3], {y, -5.0, 5.0, 10.0/n}, {x, -5.0, 5.0, 10.0/n}];
Show[Graphics[RasterArray[image]], AspectRatio -> 1];

Links

• Magnetic Pendulum – C++ program by Ingo Berg, uses Beeman integration scheme, here is a nice big picture and here is an animation
• Mathematica code – by Urijah Kaplan
• Butterfly Effect – systems that are highly sensitive to initial conditions
• homemade magnetic pendulum – by David Williamson

Driven Cavity

Here is the flow inside a square box where the flow across the top of the box is moving to the right at Reynolds number 1000. This program uses the finite volume method to solve the Navier-Stokes equations assuming steady, incompressible, viscous, laminar flow. This was calculated on a 300×300 non-uniform grid and took 3 hours to run on my laptop. The animation shows the motion along the streamlines/pathlines.

Driven Cavity – calculated in Fortran 90, plotted in Mathematica 4.2, 11/2/05

The following Mathematica code is not completely accurate at the boundaries, but it gives the basic idea:

(* runtime: 20 minutes *)
n = 50; dx = 1.0/(n - 1); RE = 1000; relax = 0.4; residu = residv = residp = 1;
u = Table[If[j == 1 || i == 2 || i == n, 0, 1], {i, 1, n}, {j, 1, n}];
v = p = pstar = ap = an = as = aw = ae = sc = sp = du = dv = Table[0, {n}, {n}];
lisolv[i0_, j0_, phi0_] := Module[{phi = phi0, p = q =Table[0, {n}]}, Do[q[[j0 - 1]] = phi[[i, j0 - 1]]; Do[p[[j]] = an[[i, j]]/(ap[[i, j]] -as[[i, j]]p[[j - 1]]); q[[j]] = (as[[i, j]]q[[j - 1]] + ae[[i, j]] phi[[i + 1, j]] + aw[[i, j]]phi[[i - 1, j]] + sc[[i, j]])/(ap[[i, j]] - as[[i, j]]p[[j - 1]]), {j, j0, n - 1}]; Do[phi[[i,j]] = p[[j]]phi[[i, j + 1]] + q[[j]], {j, n - 1, j0, -1}], {i, i0, n - 1}]; phi];
calc[i0_, j0_] := Module[{}, {cn, cs, ce, cw} = 0.5 {v[[i,j + 1]] + v[[i0, j0 + 1]], v[[i, j]] + v[[i0, j0]], u[[i + 1, j]] + u[[i0 + 1, j0]], u[[i, j]] + u[[i0, j0]]}dx; cp = Max[0, cn - cs + ce - cw]; {an[[i, j]], as[[i, j]], ae[[i, j]], aw[[i, j]]} = Map[Max[0, 1 - 0.5 Abs[#]] &, RE{cn, cs, ce, cw}]/RE + Map[Max[0, #] &, {-cn, cs, -ce, cw}]; sp[[i, j]] = -cp; ap[[i, j]] = an[[i, j]] + as[[i, j]] + ae[[i, j]] + aw[[i, j]] - sp[[i, j]]];
While[Max[residu, residv, residp] > 0.001, residu = residv = residp = 0; Do[calc[i - 1, j]; sc[[i, j]] = cp u[[i, j]] + dx(p[[i - 1, j]] - p[[i, j]]); du[[i, j]] = relax dx/ap[[i, j]]; residu += Abs[an[[i, j]]u[[i, j + 1]] + as[[i, j]]u[[i, j - 1]] + ae[[i, j]]u[[i + 1, j]] + aw[[i, j]]u[[i - 1, j]] - ap[[i, j]]u[[i, j]] + sc[[i, j]]], {i, 3, n - 1}, {j, 2, n - 1}]; ap /= relax; sc += (1 - relax) ap u; u = lisolv[3, 2, u]; Do[calc[i, j - 1]; sc[[i, j]] = cp v[[i, j]] + dx(p[[i, j - 1]] - p[[i, j]]); dv[[i, j]] = relax dx/ap[[i, j]];residv += Abs[an[[i, j]]v[[i, j + 1]] + as[[i, j]]v[[i, j - 1]] + ae[[i, j]]v[[i + 1, j]] + aw[[i, j]]v[[i - 1, j]] - ap[[i, j]]v[[i, j]] + sc[[i, j]]], {i, 2, n - 1}, {j, 3, n - 1}]; ap /= relax; sc += (1 - relax)ap v; v = lisolv[2, 3, v];Do[an[[i, j]] = dx dv[[i, j + 1]]; as[[i, j]] = dx dv[[i, j]]; ae[[i, j]] = dx du[[i + 1, j]]; aw[[i, j]] = dx du[[i, j]]; sp[[i, j]] = 0; sc[[i, j]] = -((v[[i, j + 1]] - v[[i, j]])dx + (u[[i + 1, j]] - u[[i, j]]) dx); residp += Abs[sc[[i, j]]]; ap[[i, j]] = an[[i, j]] + as[[i, j]] + ae[[i, j]] + aw[[i, j]] - sp[[i, j]]; pstar[[i, j]] = 0, {i, 2, n - 1}, {j, 2, n - 1}]; Do[pstar = lisolv[2, 2, pstar], {2}]; Do[If[i != 2, u[[i, j]] += du[[i, j]](pstar[[i - 1, j]] - pstar[[i, j]])]; If[j != 2, v[[i, j]] += dv[[i, j]](pstar[[i, j - 1]] - pstar[[i, j]])]; p[[i, j]] += 0.3(pstar[[i, j]] - pstar[[n - 1, n - 1]]), {i, 2, n - 1}, {j, 2, n - 1}]];

Here is some Mathematica code to plot streamlines. The blue lines are rotating clockwise and the red lines are rotating counter-clockwise. The plot on the left agrees fairly well with Ercan Erturk’s results.
psi = Table[0, {n - 1}, {n - 1}]; Do[psi[[i, j]] = psi[[i, j - 1]] + dx(u[[i + 1, j - 1]] + u[[i + 1, j]])/2, {j, 2, n - 1}, {i, 1, n - 1}];
psi = ListInterpolation[psi, {{0, 1}, {0, 1}}];
ContourPlot[psi[x, y], {x, 0, 1}, {y, 0, 1}, PlotPoints -> 50, PlotRange -> All, ContourShading -> False,Contours -> {-0.08, -0.077, -0.07, -0.06, -0.045, -0.025, -0.01, -0.0025, 0, -8*^-6, 2*^-6, 3*^-5, 8*^-5, 1*^-4, 3*^-4, 6*^-4, 9*^-4}, ContourStyle -> Table[{Hue[2(1 - x)/3]}, {x, 0, 1, 1/16}]]

Here is some Mathematica code to plot vorticity contours. The plot on the left agrees fairly well with Ghia’s results.
omega = ListInterpolation[Table[(v[[i, j]] - v[[i - 1, j]])/dx - (u[[i, j]] - u[[i, j - 1]])/dx, {i, 2, n - 1}, {j, 2, n - 1}], {{0, 1}, {0, 1}}];
ContourPlot[omega[x, y], {x, 0, 1}, {y, 0, 1},PlotPoints -> 50, PlotRange -> All, ContourShading -> False, Contours -> Range[-14, 7],ContourStyle -> Table[{Hue[2(1 - x)/3]}, {x, 0, 1, 1/21}]]

Computational Fluid Dynamics (CFD) Links

• CFD Gallery – I especially like the simulations of turbulence
• Turbulent Boundary Layer – large simulation by Said Elghobashi
• More CFD Movies – nice simulations of fire
• Colorful Fluid Mixing Gallery
• Gallery of Fluid Dynamics – excellent web site on fluid dynamics by Mark Cramer
• Gallery of Fluid Motion – from the American Institute of Physics
• CFD code – simple Fortran codes by Abdusamad Salih
• FlowLab – educational resource by Fluent

Driven Cavity – Mathematica 4.2, 11/16/06

Here is the same driven cavity using the finite difference method. This code uses the Alternating Direction Implicit (ADI) method to solve the vorticity transport equation assuming unsteady, incompressible, viscous flow.
(* runtime: 1 minute *)
<< LinearAlgebra`Tridiagonal`; n = 20; RE = 1000; dx = 1.0/(n - 1); dt = 0.25; cxy = dt/(4dx); sxy = dt/(2RE dx^2); lambda = 1.5; free = Range[2, n - 1];
omega = v = psi = Table[0, {n}, {n}]; u = Table[If[j == n, 1, 0], {n}, {j, 1, n}];
Do[omega[[All, 1]] = 2psi[[All, 2]]/dx^2; omega[[All, n]] = 2 (psi[[All, n - 1]] + dx)/dx^2; omega[[1, All]] = 2 psi[[2, All]]/dx^2; omega[[n, All]] = 2 psi[[n - 1, All]]/dx^2; omega[[free, free]] = Transpose[Partition[TridiagonalSolve[Drop[Flatten[Table[If[i == n - 1, 0, -(cxy u[[i, j]] + sxy)], {j, 2, n - 1}, {i, 2, n - 1}]], -1], Flatten[Table[1 + 2sxy, {j, 2, n - 1}, {i, 2, n - 1}]], Drop[Flatten[Table[If[i == n - 1,0, cxy u[[i + 1, j]] - sxy], {j, 2,n - 1}, {i, 2, n - 1}]], -1], Flatten[Table[(cxy v[[i, j - 1]] + sxy)omega[[i, j - 1]] + (1 - 2sxy)omega[[i, j]] + (-cxy v[[i, j + 1]] + sxy)omega[[i, j + 1]] + If[i == 2, (cxy u[[i - 1, j]] + sxy)omega[[i - 1, j]],0] - If[i == n - 1, (cxy u[[i + 1, j]] - sxy) omega[[i + 1, j]], 0], {j, 2, n - 1}, {i, 2, n - 1}]]], n - 2]]; omega[[free, free]] = Partition[TridiagonalSolve[Drop[Flatten[Table[If[j == n - 1, 0, -(cxy v[[i, j]] + sxy)], {i, 2, n - 1}, {j, 2, n - 1}]], -1], Flatten[Table[1 + 2sxy, {i, 2, n - 1}, {j, 2, n - 1}]], Drop[Flatten[Table[If[j == n - 1, 0, cxy v[[i, j + 1]] - sxy], {i, 2, n - 1}, {j, 2, n - 1}]], -1], Flatten[Table[(cxy u[[i - 1, j]] + sxy)omega[[i - 1, j]] + (1 - 2sxy)omega[[i, j]] + (-cxy u[[i + 1, j]] +sxy)omega[[i + 1, j]] + If[j == 2, (cxy v[[i, j - 1]] + sxy)omega[[i, j - 1]], 0] - If[j == n - 1, (cxy v[[i, j + 1]] - sxy) omega[[i, j + 1]], 0], {i, 2, n - 1}, {j, 2, n - 1}]]], n - 2]; resid = 1; While[resid > 0.001, resid = 0; Do[resid += Abs[psi[[i, j]] - (psi[[i,j]] = lambda(psi[[i, j - 1]] + psi[[i - 1, j]] + psi[[i + 1, j]] + psi[[i, j + 1]] - dx^2omega[[i, j]])/4 + (1 - lambda)psi[[i, j]])], {i, 2, n - 1}, {j, 2, n - 1}]]; Do[u[[i, j]] = (psi[[i, j + 1]] - psi[[i, j - 1]])/(2dx); v[[i, j]] = -(psi[[i + 1, j]] - psi[[i - 1, j]])/(2dx), {i, 2, n - 1}, {j, 2, n - 1}]; ListContourPlot[Transpose[psi], PlotRange -> All, ContourShading -> False], {100}];

Link: Driven Cavity – simple Fortran program by Zheming Zheng that uses this method

Lorenz Attractor

The Lorenz Attractor is a famous chaotic strange attractor. The equations were originally developed to model atmospheric convection.
(* runtime: 1 second *)
sigma = 3; rho = 26.5; beta = 1;
soln = {x[t], y[t], z[t]} /. NDSolve[{x'[t] == sigma (y[t] - x[t]), y'[t] == rho x[t] - x[t]z[t] - y[t], z'[t] == x[t] y[t] - beta z[t], x[0] == 0, y[0] == 1, z[0] == 1}, {x[t], y[t], z[t]}, {t, 0, 100}, MaxSteps -> 10000][[1]];
ParametricPlot3D[soln, {t, 0, 100}, PlotPoints -> 10000, Compiled -> False];

Here is some Mathematica code to numerically solve this using the 4th order Runge-Kutta method:
(* runtime: 0.2 second *)
sigma = 3; rho = 26.5; beta = 1; dt = 0.01; p = {0, 1, 1}; v[{x_, y_, z_}] = {sigma(y - x), rho x - x z - y, x y - beta z};
Show[Graphics3D[Line[Table[k1 = dt v[p]; k2 = dt v[p + k1/2]; k3 = dt v[p +k2/2]; k4 = dt v[p + k3]; p += (k1 + 2 k2 + 2 k3 + k4)/6, {1000}]]]];

Links

• Chaoscope – beautiful strange attractor software by Nicolas Desprez
• Lorenz System Fattened Trajectory – interesting manifold by Michael Henderson
• Lorenz and Modular Flows – nice renderings by Jos Leys
• Mathematica code – by Mark McClure
• Java applet – by Patrick Worfolk