Is the PDMS somehow bonded? Then you could try it with hexane. It leads to swelling of PDMS and often allows a removal.
Maybe check also
DOI: 10.1021/ac0346712Solvent Compatibility of Poly(dimethylsiloxane)-Based Microfluidic Devices Jessamine Ng Lee, Cheolmin Park, and George M. WhitesidesAnalytical Chemistry 2003 75 (23), 6544-6554
for other swelling agents. If the PDMS sticks to your master structure on a Si wafer, you should consider silanization of your master.
If this is not fitting to your problem, please describe your problem more detailed.
If the PDMS is plasma bonded to the native oxide layer on the Si, then you can't get truly "residue free" removal with solvents since the bonding process results in covalent bonds between the PDMS and the surface oxide layer on the Si. If you don't mind removing the native oxide layer, then you can have a truly "residue free" Si surface (within the limits of your process cleanliness, of course). I would suggest a buffered oxide etch after removing the bulk of the PDMS with the solvent process. A new oxide layer will form and allow you to use the wafer as if it were new (just a tiny bit thinner). Best of luck to you!