Ok so this is one of the main lines.  What does black do here? if he captures the white knight, white will take back with the queen and have a strong center.

I believe cxd5 is much more common than Nxd5.  But I see no reason the usual plan of fianchettoing the dark squared bishop shouldn't work.  I would not take the knight unless I could gain time with the queen recapture and was satisfied with my pawn structure and activity after the pawn recapture.  Giving him a developing move to respond to my moving the knight twice in the opening sounds like it's just giving him time, not to mention the center.  I wonder if d6 and Be6 is playable, with or without Nxd5.

Yeah, I think sooner or later black will play ...Nxd5 because if cxd5, it's a benoni structure with pieces traded, that's good for black, and if Qxd5, remember it's possible to kick it out with ...Be6 (after ...d6) and ideally ...d5. Perhaps ...d6 first is good, because the f6 knight while it's there keeps the attack on e4, preventing white from playing it.

cxd5 is much more common but if white does play Nxd5:

5...Nc6 6. Bg5 Be7 7. Nxe7 Qxe7 gives up the bishop pair but gives black good development. Black will play for ...d5 with ...0-0 and ...Rd8.

5...Nxd5 6. Qxd5 d6 (or ...Nc6) 7. Nf3 Nc6 8. e4 Be6 9. Qd1 Be7 with equality.

5...Nxd5 6. cxd5 (the simplification favours black) and ...g6 and ...Bg7 give black a good version of the Benoni structure.

The basic point is that White doesn't gain enough by occupying d5 with his Queen to overcome what he loses by the early trade and subsequent loss of tempo when the Queen is expelled from d5.

Remember the Benoni has been seriously played and analyzed for half a century, and the world's leading grandmasters have uniformly rejected 5 Nxd5 in favor of 5 cxd5 - with very good results for White in recent years, too.  The reason is that Nxd5 is not a practical attempt at advantage against reasonable play.