High coupling materials for thin film bulk acoustic wave resonators

Radio frequency (RF) filters based on bulk acoustic wave resonances in piezoelectric thin films have become indispensable components in mobile communications. The currently used material, AlN, exhibits many excellent properties for this purpose. However, its bandwidth is often a limiting factor. In addition, no tuning is possible with AlN. Ferroelectrics would offer both larger coupling to achieve larger bandwidths, and tunability. However, their acoustic properties are not well known, especially in the thin film case. The goal of this thesis is to investigate the potential and identify the limitations of ferroelectric thin films for thickness mode resonators in the 0.5 - 2 GHz range. The Pb(Zrx,Ti1-x)O3 (PZT) solid solution system is the main candidate, since it is known for its large piezoelectric constants and its growth is already well studied. As a main test vehicle, free standing thin film bulk acoustic resonator (TFBAR) structures with Pt/PZT/Pt/SiO2 membranes were successfully fabricated using silicon micro-machining techniques. The main drawback of ferroelectrics is the damping of acoustic waves by domain wall motion both in the RF electric field and in the pressure wave. For this reason films with varying orientations and compositions were investigated. From the device structures the electro-mechanical coupling constants kt2, the quality factors (Q-factors) and several materials parameters have been obtained. High coupling constants have been found for sol-gel Pb(Zr0.53,Ti0.47)O3 films with a {100} texture, kt2 is found to be 0.4 for a 1 µm thick film and 0.8 for a 3.8 µm thick film. However, the Q-factors of these films are low, 18 for the first film and 3 for the second film. The increase of kt2 and the decrease of the Q-factor with frequency indicates that the domains present in these films contribute to these characteristic parameters. It was generally observed that high coupling constant are associated to low Q-factors. This became evident when comparing films with 53/47 composition, where both tetragonal and rhombohedral phases are present, to tetragonal films as well as when comparing {100} textures with (111) textures. Both for the 53/47 composition and for the {100} texture, ferroelastic domain walls are thought to play a bigger role than for tetragonal compositions and (111) textures. The highest figure of merit (FOM) of about 15 was found when combining the composition leading to a high coupling constant (53/47) and the orientation leading to lower losses (111). However the losses even in this film are too high for RF-filter applications. On the other hand, films with low Q-factors but high coupling could prove very useful as transducers for ultrasonic imaging applications, where low Q-factors are desired. The stiffness coefficients of the studied PZT films were shown to be higher than expected from ceramics data. Most likely the stiffness of ceramics always contains domain contributions leading to softening. In contrast, in textured films the variety of domain orientation is very much reduced. In order to reduce losses due the presence of ferroelastic domains three different potential solutions were explored. The first idea was to manipulate the domain populations of the films deposited on silicon by using heat and vacuum treatments. Silicon substrates are known from previous works to be unfavourable for high c-domain fractions. It was discovered that an anneal in vacuum at 550 °C lead to a significant reduction of c-domains in tetragonal 30/70 PZT ({100}). On the contrary, if the sample was subjected to a compressive stress during cooling, the c-domain fraction could be increased. Analysis of the film stress versus temperature curves revealed a trend consistent with theoretical predictions, i.e. a phase boundary between the c/a/c/a and the a1/a2/a1/a2 domain patterns between room temperature and the Curie temperature θC. However, even though this method reveals interesting results, it can not be exploited as a method to achieve a sufficient c-domain population. The second idea explored was the implementation of a high thermal expansion material as a substrate. PZT films deposited on MgO are known to be compressive due to the difference in thermal expansion of the two materials. The compressive stress leads to highly c-axis oriented PZT films. Devices using MgO substrates were fabricated, however difficulties in the micro-machining of the MgO substrate inhibited a complete liberation of the membrane. Nevertheless, preliminary measurements indicate these devices could lead to both high coupling and high Q-factors, suggesting that further detailed study of this method is worthwhile. As a third method for avoiding ferroelastic domains, the uniaxial ferroelectric potassium lithium niobate (KLN) was explored. The unique ferroelectric axis in this material means that only 180° domain walls are present, which can theoretically be removed by poling. This material has been deposited in thin film form using pulsed laser deposition (PLD). KLN thin films with a {001} texture were deposited successfully on Pt/Si substrates. The films were piezoelectric with a d33,f value of around 10 pm/V and a dielectric constant of 250. This is the first time that piezoelectric properties were measured on KLN thin films. A columnar structure has been observed, however the small grain size and the rough surface currently make it difficult to apply this material to TFBAR's.

Related material