RF MEMS switches have been shown to have better performance than their solid state counterparts on account of their low insertion loss and high isolation. Despite their superiority, these switches suffer from several reliability issues which limit their lifetime when compared with p-i-n diodes and GaAs FET switches. One of the major reliability issues is the reduction in lifetime of these switches when switched under hot switching conditions i.e. when a DC voltage or RF signal is applied across the contact while it is switching from an off to on position or vice versa. In this work, contact damage in Ruthenium-on-Ruthenium microcontacts has been investigated under hot switching conditions. Using an AFM based test setup, developed at Northeastern University for the purpose of contact testing, a large number of experiments were performed to observe and understand the mechanisms that lead to microcontact damage and ultimately its failure. The structure used was a clamped-clamped beam structure with a contact bump at its center. A flat topped mating pillar formed the other end of the contact and this pillar was mounted on a piezoactuator whose expansion and contraction, leading to contacts closing and opening, replicated switching cycles. It was observed that material transfer was the primary cause for contact failure in DC hot switching. When the applied hot switching voltage exceeded 2.5 V, the direction of material transfer appeared to be polarity dependent and is always found to be from the anode to the cathode. This gives rise to the formation of a pit at the anode and a mound on the cathode. Prolonged material transfer leads to contact erosion until at one point the contact resistance becomes too high leading to contact failure. It was determined, through models and experiments, that the mechanisms leading to contact erosion operate when the electrodes are separated by either a few A or are barely touching. For leading edge hot switching, i.e. hot switching during the make phase of the contact, the damage mechanism was found to be associated with very low current and was prominent even when a current limiting resistance up to 1 MegO was placed in series with the contact. For both leading and trailing edge hot switching, when a hot switching voltage of 3.5 V is applied and a 50 O resistance is placed in series, the amounts of material transfer observed at a cycling frequency of 500 Hz were found to be almost identical. However, leading and trailing edge hot switching were also found to be different under other conditions such as when a high external resistance of 5 kO is placed in series. Also, for trailing edge hot switching, when contacts are separated extremely slowly, two different mechanisms - one polarity dependent and one polarity independent - were found to exist. These mechanisms were found to operate before the contacts fully come apart, probably when a molten metal bridge is formed between them. By examining microcontacts under a variety of hot switching conditions, ranging from different voltages, different polarities and different approach and separation rates, it was concluded that hot switching damage is an extremely complex phenomenon for microcontacts. It consists of a number of different mechanisms all occurring simultaneously in different degrees depending on the exact hot switching conditions. Even a small hot switching voltage of 0.25 V can cause damage that is significant when compared with pure cold switching i.e. when a voltage is applied only when the contact is fully closed. However, hot switching also gave rise to lower contact resistance compared with cold switching. Under bipolar hot switching, microcontacts were able to last up to more than 100 million cycles while still maintaining a contact resistance of less than 1.
展开▼
