Cooperative multi-physics simulation on self-heating effect of capacitive RF MEMS switch

The self-heating effect of the capacitive RF MEMS switch is caused by increasing incident power of the RF signal, which leads to deformation of the membrane. Thus, the air gap of the switch is changed. Eventually the drift of the actuation voltage of the switch is prompted, which seriously affect the reliability of the switch. Because the failure mechanisms of self-heating effect involves complex multi-physics coupling, the failure mechanisms are analysed and the failure modes are described by proposing the electromagnetic-thermo-mechanic multi-physics cooperative simulation method. Firstly, the dissipation power of the membrane under different incident power is got by constructing the electromagnetic simulation model of the switch in HFSS, which is taken as a heat source. Then the distribution of the surface temperature of the membrane is got by constructing the thermal simulation model of the switch in ePhysics, which is taken as a load. Next, the deformation behavior model of the switch is obtained by constructing the stress simulation model of the switch in ePhysics. At last, according to the change of the air gap caused by deformation, the failure prediction model of the drift of the actuation voltage is obtained. Taking a typical capacitive RF MEMS switch with rectangular membrane geometry for an instance, the distribution along the edge of the length of the surface current density of the membrane is got with this method. And the temperature gradually reduces along the edge of the length, with the highest temperature in the center and the lowest at the anchor. It is found that the maximum deformation point of the membrane appears on the edges of the long side. And the deformation presents a saddle surface. The linear relationship of the drift between the actuation voltage of the switch and the incident power (0-5 W) of the RF signal is fitted by getting maximum deformation value of the membrane under different temperature incident power (0-5 W). The effectiveness of the proposed method is proved by comparing with the measured data of the references.