Dear Dr. Sun Thanks for your explanation. In above convergence test, the dipole polarization is oscillating in-plane of the cancelled monitor. The resultant far-field patterns are significantly different with respect to the size of open monitor box.
As you suggested, I tried out-plane dipole polarization relative to the cancelled monitor, giving rise to the distinguished far-field projection patterns as shown in the attached images, based on different sizes of the open monitor box.
In the practical simulation, the dipole is embedded in the substrate, and the substrate is infinite along x and y, and the open monitor box is in vacuum space. Therefore, the source is outside of the box, leading to a worse error of far-field integration I guess...Do you have any suggestion to compensate for this case?
Bests Jingzhong