Cross‐Linking Matters: Atomistic Simulations of the Thermal Degradation of Polymethylsiloxanes

Abstract Utilizing an advanced machine‐learning interatomic potential (MLIP) for Si─C─O─H, the thermal breakdown of polysiloxanes covering polymers from linear polydimethylsiloxane (PDMS) to highly networked polymethylsilsesquioxane (PMSQ), is investigated. The reactive simulations reveal that increasing cross‐linking density enhances both thermal stability and ceramic yield after pyrolysis. Multiple pathways for forming siloxane oligomers are uncovered–pathways that depend on the number of unbranched units in the polymer. By simulating the release of gaseous species, the mass‐retention profiles observed in experiments are reproduced: PMSQ exhibits the highest mass retention, while PDMS is prone to substantial mass loss. Non‐isothermal simulations produce thermogravimetric studies and show that higher thermal stability and the peak of mass loss rate align with the degree of cross‐linking. Moreover, these simulations enable a kinetic analysis of TGA data, yielding an activation energy for PDMS degradation that is consistent with literature values. Finally, it is shown that PDMS domains embedded within PMSQ matrices act as sacrificial templates, carving out nanoscale voids that give rise to tailored porosity in the final SiCO ceramics. Overall, the atomistic simulations provide insights into reaction mechanisms and validate experimental trends and offer predictive guidance for tuning preceramic polymers toward tailored ceramic yields and microstructures.