Eigenmode coefficient increases along the length of a straight waveguide? How does get_eigenmode_coefficients

[peterropac]

I have an eigenmode Gaussian source at the start of a 2D straight waveguide emitting the first eigenmode. After that I have two mode monitors created with sim.add_mode_monitor at different distances after the source, oriented perpendicular to the direction of the waveguide across the width of the waveguide.

I get the eigenmode coefficients of the first eigenmode at both mode monitors, and when I calculate the ratio between the two coefficient I notice that at certain wavelenghts the eigenmode coefficient of the furthermost mode monitor is greater than the eigenmode coefficient that i get from the closer mode monitor. How is this possible? How does sim.get_eigenmode_coefficients calculate coeffecients?

Numerical example:
The center wavelength of the source is 1.55um and fwidth = 0.2. I measure the eigenmode coeffecients at frequncies 0.58064516 (1.722 um), 0.64516129 (1.55 um), 0.70967742 (1.4091um), and the ratios right_eigenmode_coeff/left_eigenmode_coeffs are 0.80500448, 0.65725958, 1.08718659.

[stevengj]

The ratios are complex numbers. I assume you are reporting to the absolute values (or absolute values squared)? Have you checked that the results are converged with resolution (i.e. doubling the resolution doesn't change anything)?

[peterropac]

Sorry, actually the ratios are abs(right_eigenmode_coeff)²/abs(left_eigenmode_coeff)²
I have doubled the resolution (120 points/um) but nothing changed. I did notice that the values change (except for the one at 1.55um) with respect to the fwidth parameter.

[peterropac]

In case anyone faces the same problem:

• It helps to broaden the source. In my case fwidth = 0.3 was good enough.
• it helps to run the simulation with mp.stop_when_fields_decayed and to set decay_by to something like 10^-8. I had it at 10^-4 and not all fields sufficiently decayed. 10^-8 is way less than I thought I would need...

Sorry for cluttering the discussions.