Create an aux track of the same type as the vocal track you are trying to delay (mono, stereo)
On one of the sends on the vocal track, choose an available bus.
On the input of the aux track, select the same bus.
Insert either medium, long or extra long delay in the aux track.
The tempo should sync to your session tempo. (if your audio was not played to a click along with the session tempo, you will have to determine the tempo of the recorded audio; but I will not get into that here, as that would be a tutorial all in itself)
Select the note duration that you want the delay to follow (whole, half, quarter, eighth, sixteenth) with triplet and/or dot modifier if necessary.
That’s it for setting up the delay to tempo sync
And to catch just that one phrase, you would automate the send so that it is only active for the particular phrase that you want to repeat. You can do this by automating the send volume, or by muting the send, depending on the sound you are looking for. Automating by mute might not be as smooth as automating the volume of the send. (I prefer automating the volume).