Here is a simulation file which I have set up to reproduce Figure 2a from the publication:
Fig_2a.fsp (735.6 KB)
I used 2D rectangle objects of PEC material to simulate the infinitesimally thick conducting walls. Using a 2D structure is preferable to using a 3D structure that is very thin, since when you use a 3D object, the simulation mesh needs to be fine enough to resolve the thickness of the structure, otherwise the structure may not be included in the simulation.
I extended the structures into PML boundaries, and I used PML absorbing boundaries in the simulation region to simulate an open boundary. I also used a 2D simulation region instead of 3D simulation region since the device is uniform and infinite in the 3rd dimension.
For the source, I used a frequency of 10 GHz to match what is used in the publication. I noticed that in the simulation file that you attached, the source was set up using time-domain settings, so the default source pulse was not used. For FDTD simulations, typically the default source pulse is the most efficient source signal to use, as discussed here:
I also needed to increase the simulation time setting of the FDTD simulation region to ensure that the simulation time was long enough for the full source pulse to be injected.
Using a frequency-domain profile monitor, I recorded the field profile over the XY plane to reproduce the plot in Figure 2a. In order to record only the resonant mode of the device, I used monitor apodization. Monitor apodization is discussed in more detail here:
The final plot I got when I visualized the Ez field component from the monitor is below:
I think the profile matches with the expected result. Please try out the simulation file and let me know if you have further questions.