www.neuron.yale.edu

In hoc file I create two hh cells with length 1000um, diameter 10um, and
coonect them by gap junction

objref gap_l, gap_r
gap_r = new Gap()
gap_l = new Gap()
RLG.axon2 gap_r.loc(0.999)
LLG.axon2 gap_l.loc(0.999)
setpointer gap_r.vnb , LLG.axon2.v(0.999)
setpointer gap_l.vnb , RLG.axon2.v(0.999)

and Gap junction model

NEURON {
POINT_PROCESS Gap
POINTER vnb
RANGE r, i
NONSPECIFIC_CURRENT i
}
PARAMETER { r = 1e10 (megohm) }
ASSIGNED {
v (millivolt)
vnb (millivolt)
i (nanoamp)
}
BREAKPOINT { i = (v - vnb)/r }

when I set the resistance of gap junction approach to zero like 0.0001

I found action potential can not propagate during these two cells.

If I just create one hh cell with length 1000um , diameter 10um, action potential can propagate from one edge to other edge.
Can anyone know if my model is correct ?

Think like an experimentalist. Come up with a list of testable hypotheses, and test them.
To help you get started, here are the hypotheses that I would test if I were you.
1. The implemenation of the gap junction is incorrect.
2. The implementation of one axon, or both axons, is incorrect.
3. The gap junction and the axons are implemented correctly, but coupling is too weak.

How would you test hypothesis 1? Make a toy model with two small, passive single
compartment cells, connected by the gap junction. Inject a current into one of the cells
(use an IClamp) and see what happens to membrane potential in both cells. Also see
how changing the coupling resistance affects the simulation results. Are the results
qualitatively correct? Are they _quantitatively_ correct? (easy to tell--you can measure
and calculate the input resistance of each individual cell, and you know the coupling
resistance)

How would you test hypothesis 2?

If I just create one hh cell with length 1000um , diameter 10um, action potential can propagate from one edge to other edge.

Yes, but what is the value of nseg? Is it large enough to give good spatial accuracy?
See this discussion thread:
https://www.neuron.yale.edu/phpBB2/view ... value+nseg
If nseg is too small, spatial accuracy will be poor. Also, it will take a larger stimulus to
trigger a spike (small nseg means large compartments, which means large local
capacitance, so that more stimulus current is needed to depolarize the cell enough
to trigger a spike). Two long axons with nseg = 1 will require much stronger coupling
to make the "downstream" axon fire a spike.

I don't have to tell you how to test hypothesis 3.

--Ted