[FDTD]Unwanted field distributions in FDTD simulation of a cylindrical waveguide

Hello,

I've managed to develop a FDTD code, in cylindrical coordinates, for studying mode propagation in a cylindrical waveguide. I have to cite that there is no PML region attached yet, so the waveguide is treated as a cylindrical cavity.
At z=0 I explicitly set a certain waveguide field distribution, for the Electric field components that exist in that mode. For example, if I want to excite TE_21 mode, I manually set Er and Efi field components transverse distributions.
Naturally, only these two component distributions have to exist and to propagate (with the matching magnetic ones), but Ez field distribution appears. The axial distributions of Efi and Er components are naturally sinusoidals (including the reflections from the metal boundary condition), but the Ez is not (actually is a real mess!). The transverse Ez component distribution has a certain form ( like TM_11 Ez transverse distribution), BUT the amplitude of Ez is about 35-40dB lower than the amplitude of the two others components. I already know that, if someone excites with a hard source an FDTD domain, unwanted field components may appear. But, why these components (in my case Ez component), has a certain distribution? For example, CST shows only noise, not a specific spatial form.

Thanks for your reply in advance!