How did you get your query plan to use the new unique nonclustered index instead of the primary key (which is also a unique index on column a)?