Short
According to the logic of interferometry, phase shift should be build-up between two waves by using passive or active methods. Passive methods require no additional power, like changing pathlength of one waveguide, but don't allow to modify after production. Active methods, on the contrary, give oppurtunity to manipulate phase of the waves after production and do it in very high speed(GHZ range). There are two methods mostly in usage, thermal or electrical. Thermal phase shifters can be very compact, but they are relatively slow compared to electrical charge accumulation methods. In our design, we will use passive method.
In order to design a Mach-Zehnder interferometer production-ready, some of the parameters should considered as "constant", like height of waveguides, 220nm in our case, due to already defined silicon wafer thickness by CMOS foundry. Besides, according to mode theory, which solves Maxwell's EM Equations in 2D, after passing a value for width of a waveguide, additional modes are emerging, and it affects the efficiency negatively. Thus, we selected 500nm as waveguide width, which is very close to theoretical limit value. Slab waveguide geometry is selected, due to low bending loss and ease of production. (There is no active component on chip, so it is not needed to use rib waveguide at all)
Waveguide mode has been defined by using effective index method, which calculates 1D slab modes in 2 different orientations and combine the results. To get more realistic data and compare effective index method vs. vectorial analysis, fully vectorial eigen mode solution is also simulated in Lumerical Mode. Wavelength is selected as 1550nm, which is widely used in telecommunication.