indexing with partition question(s)

Hello, We are going to have a long denormalized table soon. It will be filled with records that get marked deleted ( rec_type=d ), and of course other rec_types. But the largest part of the table that I want to eliminate in queries are the deleted records. So I was thinking of fragmenting the table by rec_type. Now, they want to query on rec_type and other fields also. So how would the indexes work ? If I make a composite index with rec_type as the first column, and say ssn and school_token as the second and third columns, will fragmentation elimination occur to eliminate all the deleted records ? I am thinking it won't because the index can't be created to follow the fragmentation scheme since it has other columns in it. Is that right ? Should I just make an index on rec_type to follow the scheme, and hope the optimizer uses that index and then the others ? A little confused about how it works. Thanks, floyd