I am not sure if this was to be expected, however...
If we could predict everything, there would be no need to do any experiments of any kind (including experiments on computational models).
By playing with different combinations for Ra in the node and the MYSA sections, I find that Ra = 1e9 in either of them makes node fire without stimulus, just as I have seen in my "node-only" model.
If you want to verify that your model is a correct implementation of McIntyre's model node of Ranvier, use the Vector class's record() method to capture the time course of each model's node.v(0.5), and compare the recorded values. The models differ in implementation, so results won't be absolutely identical, but it would be adequate to plot the two time courses and examine visually for significant differences over the first 10-20 ms.
So the node model seems to need the connection to a neighboured section with finite cytoplasmatic
resistivity in order to be in a steady state at the resting voltage.
The node model probably needs an electrical load in order to even have a resting potential. Notice that all of the non-node sections have not only pas but also membrane capacitance--both of which contribute to the stability of the intact axon model.
Take a close look at the spike train produced by the isolated node. Examine the first 15 ms closely, especially the interspike intervals. During the depolarizing phase of each ISI the net ionic membrane current i_ion must be inward (obviously). Not only is i_ion inward, but the more depolarized that v becomes, the faster it depolarizes, so the inward i_ion must be growing larger. After all, the node has voltage-dependent sodium channels that are activated by depolarization, and those account for acceleration of the depolarizing phase of the ISIs.
e_pas is -80 mV, so if you attach a section that has the pas mechanism to a node of Ranvier, that passive section will produce an outward current that tends to counteract the net inward current generated by the node. Furthermore, the more depolarized node.v becomes, the larger the passive section's i_pas will become, reducing the slope of the depolarizing phase of the ISIs and slowing the firing rate. If g_pas is big enough, i_pas will be too large for the node's inward current to drive the node to spike.
The passive section's membrane capacitance also reduces the rate of change of v (remember that dv/dt = (1/c)*net current), and this will have a flattening effect on the ISIs and slow the rate of spiking. If its membrane capacitance is big enough, dv/dt will be so slow that voltage-gated potassium channels have time to open and prevent the sodium channels from generating a spike.
You could explore the role of internodal cm and pas in the McIntyre model by a couple of simple computational experiments.
To make this easy, first create a set of sections that contains only internodal sections.
// make a set that contains all non-node sections
internodal = new SectionList()
forsec "node" internodal.remove()
1. What happens if cm of all internodal sections is negligibly small?
// reduce cm of all non-nodal segments by a factor of 1e6
forsec internodal for (x,0) cm(x)*=1e-6
Now run a simulation; you might want to increase Tstop to 1500 or 2000 ms, just to be sure of stability.
2. What happens if g_pas of all internodal sections is negligibly small?
// restore cm to its original values
forsec internodal for (x,0) cm(x)*=1e6
// reduce g_pas by a factor of 1e6
forsec internodal for (x,0) g_pas(x)*=1e-6
// now run a simulation
You may want to increase Tstop to 1500 or 2000 ms.
3. What happens if both cm and g_pas of all internodal sections is negligibly small?
is there a trick to interrupt the output of topology(), or pipe it into a file? On my terminal (I am under Windows), the output is so long and large that I cannot scroll back to its beginning...
In recent versions of Windows, can't length of history be specified as one of the "preference" settings for a cmd window? What about NEURON's own bash terminal?