You may use 2D axi-symmetric if your geometry is symmetrical, otherwise I would recommend you to use full 3D geometry. This is to ensure that the surface tension forces that control bubble break-up and coalescence are captured correctly (if you want to track the interface).

You can see a description and comparison of using 2D and 3D for modelling slug flow in microchannels in my IJMF paper (doi:10.1016/j.ijmultiphaseflow.2009.11.009) and in the paper attached.

Thank you so much Rafael, your paper wil be very helpful. Actually, I have a horizontal pipe connected with vertical tank and there are two oulet legs in the tank's body, one at the top and one at the bottom. So far I am using 2D to simulate the steam-water two phase flow in this system and the B.C is velocity inlet and pressure outlet, with Eulerian-Eulerian model. But the convergence is still hard to get and the net of the mass flow rate between inlet and outlet is (-3.4546632e-6).So, if there is any person who is interested to explain why the net of the mass flow is negative, and which model is suitable to use in multiphase flow by using Fluent.

As suggested by Rafael its better to use 3D if the pipe is not axi-symmetric in order to get better results

you can use 2D geometry to simulate two phase flow with VOF model in Fluent code, you can try the dam-brick example