The FitzHugh-Nagumo (FHN) equations [#!FitzH:61!#,#!NaArYo:62!#] are a simplified version of the Hodgkin-Huxley equations [#!HoHu:52!#]. They model nerve axon dynamics and are given by
Travelling wave solutions of the form , where are solutions of the following ODE system:
We aim to find a -homoclinic orbit at a Shil'nikov bifurcation. All the commands given here are in the file fnb.auto. First we obtain a homoclinic orbit using a homotopy technique (see FrDoMo:94), using ISTART=3, for the parameter values .
Among the output we see:
BR PT TY LAB PERIOD L2-NORM ... PAR(17) 1 20 UZ 3 2.92257E+01 2.37916E-01 ... -1.68000E-09and a zero of PAR(17) means that a zero of an artificial parameter has been located and the right-hand end point of the corresponding solution belongs to the plane that is tangent to the stable manifold at the saddle. This point still needs to come closer to the equilibrium, which we can achieve by further increasing the period to 300, while keeping PAR(17) at 0:
BR PT TY LAB PERIOD L2-NORM ... PAR(2) 1 190 UZ 10 3.00000E+02 7.37932E-02 ... 1.79286E-01
Next we stop using the homotopy technique and increase the period even further, to 1000.
BR PT TY LAB PERIOD L2-NORM ... PAR(2) 1 80 UZ 13 1.00000E+03 4.04183E-02 ... 1.79286E-01
A continuation in PAR(2)= and PAR(1)= needs to be performed to arrive at the place where we wish to find a 2-homoclinic orbit: . At the same time we monitor PAR(22) to locate Belyakov points.
BR PT TY LAB PAR(2) L2-NORM ... PAR(1) PAR(22) 1 6 UZ 15 1.31812E-01 3.28710E-02 ... 2.17166E-01 -6.31243E-06 1 23 UZ 19 -8.54548E-08 1.56158E-02 ... 2.74218E-01 -9.88772E-02Hence, there exists a Belyakov point at . At label 19 we have a lower value of than at the Belyakov point, and by inspection of the file d.4 we can observe that the equilibrium has one positive eigenvalue and a complex conjugate pair of eigenvalues with negative real part, and conclude that this orbit is of Shil'nikov type. Before starting the homoclinic branch switching, we calculate the adjoint to obtain a `Lin vector':
BR PT TY LAB PAR(9) L2-NORM ... PAR(3) 1 2 EP 28 -1.00000E+00 1.56158E-02 ... 2.50000E-03Next, we continue in the time (PAR(21)), the gap (PAR(22)) and (PAR(1)), and by setting ISTART=-2 we try to locate a 2-homoclinic orbit:
BR PT TY LAB PAR(21) ... PAR(1) PAR(22) ... 1 175 UZ 46 1.64818E+02 ... 2.74218E-01 4.59218E-16 1 180 UZ 47 1.44759E+02 ... 2.74218E-01 -1.43728E-14 1 184 UZ 48 1.24939E+02 ... 2.74218E-01 1.55506E-13 1 189 UZ 49 1.04615E+02 ... 2.74218E-01 -2.37665E-11 1 193 UZ 50 8.53538E+01 ... 2.74218E-01 1.02165E-11 1 198 UZ 51 6.37899E+01 ... 2.74218E-01 -5.74204E-14Each of these homoclinic orbits differ by about 20 in the value . This is about the time it takes to make one half-turn close to and around the equilibrium, so that orbits differ by the number of half turns around the equilibrium before a big excursion in phase space. Note that the variation of is so small that it does not appear.
A plot of vs. gives insight into how the gap is opened and closed in the continuation process. This is depicted in Figure 27.5.
|