Interface Acoustic Wave Devices Made By Indirect Bonding of Lithium Niobate on Silicon

Interface acoustic waves (IAW) propagate along the boundary between two perfectly bonded solids. For a lossless IAW, all displacement fields are evanescent along the normal to the boundary inside both solids, but a variety of leaky IAWs also exist depending on the selected combination of materials. When at least one of the bonded solids is a piezoelectric, the JAW can in principle be excited by an interdigital transducer (IDT) located at the interface. However, the IDT has a finite and non vanishing thickness that must be properly taken into account in actual devices. This difficulty is probably the reason why most studies have remained theoretical and why few experiments have been reported thus far. One possibility which we have discussed at last year's symposium is to bury the IDT inside one of the solids and to subsequently achieve a direct bonding onto the other solid. Another possibility is to deposit a thick layer, for instance silicon oxide, atop an IDT patterned above a piezoelectric plate. This layer must be thick enough so that interface waves are excited rather than Sezawa modes. We discuss the fabrication and characterization of IAW resonators made by indirect bonding of lithium niobate onto silicon. In our fabrication process, IDTs are first patterned over the surface of a Y-cut lithium niobate wafer. A thin layer of SU8 photoresist is then spun over the IDTs and lithium niobate to a final thickness below one micron. The viscosity of the SU8 layer is such that a uniform and flat deposition is achieved. The SU8 covered lithium niobate wafer is then bonded to a silicon wafer using a wafer bonding machine. During the bonding process, we heat the material stack at a temperature of 60 degrees C and we apply a pressure of 10 kg.cm(-2) to the whole contact surface. The stack is subsequently cured and baked to enhance the acoustic properties of the interfacial resist. Measurements of resonators are presented with an emphasis on the dependence of propagation losses with the resist properties. We find in particular that the resist viscosity is about ten times larger than that of crystalline silicon. This is nevertheless quite a smaller value than could be expected for such a polymer.