Solving curing-protocol-dependent shape errors in PDMS replication

PolyDiMethylSiloxane (PDMS) is an elastomer increasingly used to produce soft objects by replication, in a variety of fields including soft electronics, microfluidics, tribology, biomechanics and soft robotics. While PDMS replication is usually considered faithful at all scales, down to nanoscales, detailed quantitative comparisons between the geometric features of the mold and the replicated object are still required to further ground this commonly accepted view. Here, we show that the top surface of centimetric parallelepipedic PDMS blocks, molded on a rigid plate, deviates from its expected flatness, the amplitude of the deviation being dependent on the crosslinking protocol. As a practical solution, we identify a suitable two-steps protocol which eliminates those replication errors. Using finite element simulations, we show that the effect originates from a thermal contraction when the sample cools from the curing temperature down to the operating temperature. This phenomenon actually applies at any length scale, and finely depends on the sample's aspect ratio and boundary conditions. Our results should help mitigating replication errors in all applications where a well-defined sample geometry is required.