Thermal error reduction is of major importance to the performance of any high-tech machine. Topology optimization can be a tool that is able to design geometries that reduce the thermal error, beyond human creativity. In this research, topology optimization is used to reduce the thermal error on a transient-thermal-mechanical problem. The thermal error of the specific problem was reduced by 30%. Multiphyiscal topology optimization, as applied to the problem, comes with many challenges, under which is the manufacturability of its final design. A new penalization method is proposed, that ensures that the final design contains no intermediate densities, which are common to multiphysical topology optimization, even with SIMP or RAMP penalization. The new penalization method aims at creating intermediate material that is adverse to the optimization problem, causing it to avoid intermediate material. On a test case it was shown capable of creating a well performing black and white design, where grey penalization failed to maintain the performance of the design. Furthermore, an interpretation of the adjoint in the sensitivity analysis is given.