It should defintely be possible to replicate these results given that they used Lumerical FDTD in the paper with our standard material library. It seems they did a lot of analysis that may require some advanced scripting. I would check the material explorer to make sure you have a solid fit given the bandwidth of the simulation.Your sapphire for example:
The refer to the paper’s description.
Two detector boxes were used to measure the spectral response. The
periodic boundary condition was used in the x-direction, while
the perfect matching layers (PML) were used in the z- and ydirections (Figs. 3f and 8a). The time monitors were used to
detect the change in the electric field over time. An ultrafine 1-
nm mesh size was used in all simulations. A refractive index
of nsup = 1.0 was chosen for air to generate an asymmetric
environment while indices for Al and SiO2 were obtained by
dispersion relation provided by Palik. For the given geometric
parameters of BNA array, the angle of incidence and polarization orientation of the electromagnetic field were also swept
to understand their dependence on LSPR peak position and Efield EF. For a given BNA parameter, the effective index of
antenna environment was numerically changed in the range of
1.33 to 1.95 and spectral reflectance was simulated for normal
incidence to understand the applicability of these BNA structures as inexpensive refractive index sensors. The absorbed
power was calculated by P = −0.5ω|E|
2 Im(ε), where ω = 2πf,
f is the frequency of the light, |E|
2 is the electric field intensity,
and Im(ε) is imaginary part of the dielectric function
See setup Diagram
From what I can see they use air as the background, and they have employed a smaller mesh. The resonant wavelength is pretty sensitive to the mesh size. It seems that they are using a diffracting plane wave in y in order to use PML BC in this direction. Finally, you may want to clean up the time signal or increase your sim time. You are looking for a spectrum from the time monitor that looks like this.
Currently, it looks like this
I am not sure where the 2um artifact is coming from and there seems to be some noise around 1100nm. I think you could get better results apodizing the time-signal.