The AlN based solidly mounted resonators utilizing the all-metal conductive Bragg reflectors have been demonstrated. The devices with different reflectors of Mo/Ti, W/Ti and AlN/Mo pairs have been fabricated and the frequency responses have been compared. The bottom electrode is incorporated into the conductive Bragg reflector, thereby reducing the parasitic resistance and the damping of the acoustic energy. The experimental results reveal that Mo/Ti and W/Ti reflectors exhibit excellent properties of frequency selection and clamping the acoustic wave. The devices show the distinct resonant phenomenon near 2.4 GHz with the excellent suppression of the shear mode and the sidelobe mode. Compared to the conventional device consisted of AlN/Mo reflector, the devices consisted of the conductive reflectors demonstrate the higher resonance frequency and obviously improved performance.
The solidly mounted resonators, The all-metal Bragg reflectors, AlN films, Q factor
The film bulk acoustic resonator (FBAR) has been attracted great attentions because the promising application in high radio frequency integrated circuits (RF ICs) [1-3]. Furthermore, FBAR has also been proved to be significative for mass loading measurements in air and liquids [4-7]. The typical FBAR basically consists of a piezoelectric film sandwiched between two electrodes and works in 1-10 GHz range. The air interface (the membrane structure) or the Bragg acoustic reflector (the solidly mounted resonator, SMR) is formed under the piezoelectric film to clamp the acoustic wave energy. The Bragg reflector is composed of a sequence of a quarter wavelength thick layers of low and high acoustic impedance. Typically, SiO2, ZnO or AlN is used as a low acoustic impedance material, and W or Mo is used as high acoustic impedance material [8, 9]. In particular, W/SiO2 reflector proposed by R. Aigner et al. is very popular commercially [10, 11]. However, compared to the FBAR with the membrane structure, the limitation of the SMR is the relatively lower quality factor (Q factor). In order to improve Q factor, thermal annealing  has been proposed. Another approach is minimizing the parasitic series resistance, which is expected to be more pronounced at high frequency as the electrodes become thinner .
In this paper, the SMR devices utilizing different reflectors consisted of Mo/Ti and W/Ti pairs have been demonstrated. The bottom electrode is incorporated into the conductive Bragg reflectors, thereby evidently reducing the parasitic resistance and the damping of the acoustic energy derived from the separate bottom electrode. In additional, the conductive Bragg reflectors simplify the fabrication process and provide a lower contact resistance and an efficient heat path.
2. Experimental details
The SMR devices consisted of Mo/Ti and W/Ti reflectors have been fabricated. In comparison, the device consisted of the conventional Bragg reflector of AlN/Mo has also been fabricated. The proposed device configurations have been shown in Fig.1. Tab. I summarize the dimension of all film thickness. The thickness of each layer of Bragg reflector should be one quarter wavelength of the resonance frequency. In operation, the bottom electrode or metal acoustic mirror is grounded to excite the acoustic wave of thickness extension mode in the piezoelectric film. The pattern of the top electrodes is one port G-S-G type in order to adapt the coplanar probes. The SRM devices were fabrication on polished 3-in Si wafers. All the films were deposited using the JGP 800 sputtering system (SKY Techno. Co., Ltd. China). The detailed sputtering parameters of each layer have been provided in Tab. II. In order to obtain the Bragg reflector with the high quality and the low surface roughness, the deposition parameters were adjusted. For the three-period Mo/Ti and W/Ti reflectors, the sheet resistivities of 0.066 Ω /sq and 0.081Ω /sq were measured, respectively. In the case of three-period AlN/Mo reflectors, 200 nm Mo bottom electrode with the measured sheet resistivities of 0.96 Ω /sq was formed. The AlN films were deposited by RF reactive sputtering at the equal condition. The target was a 4-inch-diameter aluminum with 99.995% purity. The chamber was evacuated until the base pressure decrease to less than 3-10-5 Pa. The process parameters of AlN film were optimized to obtain a highly c-axis orientation and good quality. The 150 nm Al top electrode was deposited on the AlN film surface and patterned by the conventional photolithography method.
The crystal characteristic was identified by X-ray rocking curve (BRUKER-AXS) at wavelength of 0.15418 nm (Cu Kα). The cross-sectional morphologies of films were observed by field effect scanning electron microscope (SEM, FEI SIRION 200) under the operating voltage of 10 kV. The resonator scattering parameters of the finished devices were obtained using Cascade 9000TM probe station with ACP40 probes and analyzed using Agilent 8722D network analyzer.
3. Results and discussions
Fig. 2 shows the X-ray rocking curves of AlN (002) peak of the film deposited on the Bragg reflectors. The full width at half maximum (FWHM) of the rocking curve is around 3.73° for both Mo/Ti and W/Ti reflectors, which indicates preferred c-axis orientation and high crystal quality. Although the FWHM of the rocking curve shows only a slight broaden for the AlN/Mo reflector, there is no obvious deterioration in the crystal quality.
Fig. 3, Fig. 4 and Fig. 5 show the surface and cross-section view morphologies of the AlN films deposited on the AlN/Mo, Mo/Ti and W/Ti reflectors, respectively. It can be seen that the AlN films have a dense columnar structure. The interfaces between the AlN films and Bragg reflectors are clearly visible and distinct.
Fig. 6 shows the measured reflection coefficient (S11) as a function of the frequency for the three samples of SMR devices from 1 to 4 GHz. A distinct resonant phenomenon could be observed for all devices, which suggesting that the Bragg reflectors have successfully restrained the acoustic dissipation to the substrate. The frequency response of the SMR with AlN/Mo reflector exhibits a return loss of -9.81 dB at the center of 2.07 GHz. The SMR with Mo/Ti and W/Ti reflectors present the increased return loss of -16.9 dB and -15.5 dB at 2.29 GHz and 2.38 GHz, respectively. There are no obvious resonances owing to the shear mode and the higher-order harmonics. It is noted that the resonance frequencies of the devices utilizing the all-metal Bragg reflector are a little higher than that of AlN/Mo reflector. This can be explained that the bottom electrode of the conventional SMR is on the propagation path of the acoustic wave, which leads to the increase of the actual propagation distance.
Fig.7 shows the admittance curve of the devices. The FBAR response and the performance parameters are usually given by using Butterworth-Van Dyke (BVD) equivalent circuit. The BVD equivalent circuit of a FBAR is shown in Fig. 8, where C0 is the clamped capacitance, Rs is the serious resistance, Lm, Cm and Rm are motional inductance, motional capacitance, and motional resistance of the resonator, respectively. The BVD parameters provide a direct relation between the measured electrical quantities and the physical properties of the resonator. Tab. III summarizes the BVD parameters obtained by means of a least squares fitting routine available in the advanced design system (ADS) software for the SMR with different reflectors. Series resonant frequency fs and parallel resonant frequency fp are determined by
The Q factor and the effective electromechanical coupling coefficient K2eff can be derived by the expression:
As seen from Tab. III, the devices with conductive reflectors have higher Q factor compared to the conventional device. In addition, the motional resistance Rm and the serious resistance Rs of the SMRs with conductive reflectors are very smaller than that of the SMR with Mo/AlN reflector. The Rs is derived from the electrode contacting and Rm is related to the frequency dependent loss of the acoustic energy and the parasitic resistance. It is to be note that approximate 2 times reduction in Rm and 5 times reduction in Rs for the SMRs with conductive reflectors. This is in agreement to the measured sheet resistivities of the all-metal reflector and the bottom electrode. The Rm and Rs reduce because the bottom electrode is incorporated into the conductive Bragg reflectors. Therefore the SMRs with conductive reflectors have higher performance.
The AlN based SMR devices consisted of different reflectors have been fabricated and the frequency responses has been compared. The bottom electrode is incorporated into the conductive Bragg reflector, thereby evidently reducing the parasitic resistance of the electrode and the damping of the acoustic energy. The X-ray rocking curves suggest that the films deposited on the Mo/Ti and W/Ti reflectors have good quality. The distinct resonant phenomenon was observed near 2.3 GHz in the reflection coefficient measurement. The all-metal reflectors show good capabilities of sidelobe suppression and frequency selection. The comparison of the performance parameters of the devices indicates that the devices with conductive reflector have improved Q factor compared to the conventional device.
This work is supported by the Research Project of SDUST Spring Bud (2009AZZ056).