I am new to phenix.refine. I'm using the graphical interface.
I hav processed my dfiffraction data and used Phaser MR for molecular replacement.
And I goyt a result: a solution pdb file and the associated mtz file.
1. So I use the mtz file from this Phaser MR solution or the original file from my data
processing in phenix.refine?

The original file. I think Phaser outputs anisotropy-corrected data, and it may be at reduced resolution depending on how you run it.

2. I have set parameters for phenix.refine like this:
Is there other settings for the parameter which I should do for a better result?
PDB data: my solution from Phaser MR
XRAY DATA= ????
Simulated Annealing=1000
Update waters
Refine Target weights: Wxc=0.8 Wxu=1.6
Cycles=5
Fix rotamers
Fix bad side chains

Unless you've already run refinement with this dataset and are confident about the weights, I would not set wxc and wxu initially; automatic weight optimization is a better place to start. Otherwise, which refinement settings are most appropriate depends on the resolution and starting R-free. As Frederic points out, rigid-body refinement may be a good idea at this stage. (By the way, you list both "Fix rotamers" and "Fix bad side chains", but these are the same thing - unless I mislabeled something in the GUI...)