I believe the index is the mechanism used to identity duplicates, so I doubt you can have it both ways.
You could write your own duplicate finder, but on a non-indexed field it would be slow.
Generally I think an index is a small price to pay to allow you to disallow duplicates.