Introduction

Recently, mobile communication systems have continued to demand a high data rate and great mobility1,2,3. These trends are increasing the need for filters with higher frequencies and wider bandwidths4. Emerging 5G, Wi-Fi and 4G LTE communication standards are driving up frequencies from the traditional bands (below 2.6 GHz) to as high as 5 GHz to accommodate wider system bandwidths. Microelectromechanical system (MEMS) filters, such as surface acoustic wave (SAW) and bulk acoustic wave (BAW) filters, are promising candidates for operation at new radio frequency (RF) band sections. However, the SAW filter technology used for current mobile communication systems can hardly achieve frequencies above 3 GHz due to the degraded filter performance and several fabrication problems caused by the very thin and delicate IDT electrode patterns5. BAW filters can provide low insertion loss (IL), good selectivity, and high power handling and are preferred in higher frequency applications6,7,8.

Two types of technologies, including film bulk acoustic wave resonators (FBARs) and solidly mounted resonators (SMRs), have been used to manufacture thin film BAW filters1,8. SMR devices require a Bragg reflector, preferably patterned with thin-film layers with alternating low and high acoustic impedances. This technology is deemed superior to FBARs in terms of mechanical robustness and power handling. However, the disadvantage of SMRs is an increase in the number of fabrication process steps that should be controlled in terms of film thickness and film properties. For the FBAR, an air cavity is created between the bottom electrode and the carrier wafer. The air/electrode interface can trap more acoustic waves between the electrodes than the Bragg quarter-wavelength acoustic mirror, thus obtaining better effective electromechanical coupling (Keff2). Furthermore, the FBAR provides a slightly higher Q due to the lack of additional reflector layers in which the acoustic wave can be attenuated or can escape.

Aluminum nitride (AlN) is utilized in various generations of telecommunications technology due to its high longitudinal sound velocity v (11,354 m/s), low temperature coefficient of frequency (TCF, −25 ppm/°C), and low acoustic and dielectric losses9,10,11. Using an optimized FBAR design and a careful choice of materials and nonpiezoelectric layer thickness, the Keff2 of the FBAR can be boosted to values of ~7%12. However, new materials with higher electromechanical coupling render potential choices, resulting in relaxed device designs and fabrication margins. Al1-xScxN (AlScN) is one of those materials, which can be deposited through the same means as AlN, that shows an increase in the piezoelectric coefficient. For example, the piezoelectric coefficient d33 of Al0.57Sc0.43N is up to five times larger than that of pure AlN13,14,15. Moreira et al.12 presented an Al0.85Sc0.15N-based FBAR with a resonant frequency of 2.15 GHz and a Keff2 of 12.07%. Wang et al.16 demonstrated a switchable Al0.7Sc0.3N-based FBAR utilizing the newly discovered ferroelectric behavior of Al0.7Sc0.3N thin films.

In this work, we report an Al0.8Sc0.2N-based FBAR for the design of acoustic filters. By taking advantage of the high-quality Al0.8Sc0.2N film, the FBAR presents a large Keff2 of 14.5% and an excellent figure of merit (FOM) of 62. Furthermore, the Mason model is used to extract the important intrinsic material parameters of the Al0.8Sc0.2N film. Subsequently, high- and low-temperature probe stations are employed to study the temperature characteristics of the Al0.8Sc0.2N-based FBAR. Finally, filters using the Al0.8Sc0.2N film exhibit a center frequency of 4.24 GHz, a −3 dB bandwidth of 215 MHz, a small IL of 1.89 dB, and a rejection above 32 dB. The proposed Al0.8Sc0.2N-based FBAR filters show potential for 5G wideband applications.

Design and fabrication

The designed piezoelectric film bulk acoustic resonator, illustrated in Fig. 1, is composed of six thin films grown on a Si substrate (725 μm thick). Two additional Mo layers of 120 nm and 37 nm outside the cavity are designed on the top and bottom electrodes, respectively, to reduce electrode resistance. As shown in Fig. 1(a) and (b), the top and bottom electrode layers are patterned to quadrilaterals, and electrical connection strips are used to conduct electrical signals to the pads (S). When an RF signal is applied between the two electrodes of the resonator, longitudinal bulk acoustic waves are excited in the piezoelectric film. Figure 1(d) shows the scanning electron microscopy (SEM) view. Its cross-sectional stack at the slice line is exhibited in Fig. 1(c). On the basis of this design, the proposed AlScN-based FBAR is fabricated utilizing a complementary metal-oxide-semiconductor (CMOS)-compatible microfabrication process involving a combination of seven photolithography masks17,18,19. The detailed fabrication process is demonstrated in Fig. 2. The fabrication process starts with etching Si to form the isolation walls, which are used to define the cavity. Chemical mechanical polishing (CMP) is used to provide a flat surface for the deposition of the bottom Mo electrode and piezoelectric AlScN film20. In particular, an isolation wall instead of a swimming pool is used to reduce the difficulty of CMP due to the smaller polished area. Next, a 25 nm-thick AlN seed layer and 100 nm-thick bottom electrode Mo are deposited. Moreover, an additional 120 nm-thick Mo layer is formed on the top surface of the bottom electrode at the position where the opening for the bottom electrode electrical contact will be formed. The presence of the additional Mo layer ensures that the bottom electrode will always retain sufficient thickness after etching the openings in the piezoelectric layer. Then, a 500 nm-thick piezoelectric Al0.8Sc0.2N film is deposited and etched by inductively coupled plasma (ICP) etching to open the bottom electrode. A 100 nm-thick Mo layer is deposited and patterned as the top electrode. Subsequently, a mass loading layer of 37 nm-thick Mo film is deposited and etched above the structure. Finally, the release window is opened, and the whole device is released by a VHF isotropic etcher.

Fig. 1: Al0.8Sc0.2N-based FBAR.
figure 1

a Explosive view of the layered structure of the designed resonator. b The practically enlarged section view of the resonator. c Optical microscope of the fabricated resonator. d SEM image of the cross-sectional view of the fabricated resonator.

Full size image
Fig. 2: Main process steps for the fabrication of AlScN-based FBARs.
figure 2

a Forming isolation walls by etching the Si substrate. b SiO2 deposition and chemical mechanical polishing (CMP). c Bottom Mo dual deposition and pattern. d Al0.8Sc0.2N film deposition and etched by inductively coupled plasma (ICP). e Top Mo dual deposition and pattern. f Opening the release window. g Releasing the cavity by VHF.

Full size image

Results and discussion

The cross-sectional transmission electron microscopy (TEM) images are shown in Fig. 3(a) and (b). The Al0.8Sc0.2N film exhibits a good (002) c-axis preferential crystal orientation, which is preferred for achieving a high Q value for the FBAR21,22. Figure 3(c) shows the electron diffraction pattern of the Al0.8Sc0.2N film in region A of Fig. 3(a). The distance between the two diffracted spots closest to the central spot is 8.345 1/nm, corresponding to a lattice plane (002) spacing of 0.240 nm, which is slightly smaller than that of AlN (0.249 nm). In addition, X-ray diffraction (XRD) 2θ/ω scans of the Al0.8Sc0.2N films include only diffraction maxima attributed to Al0.8Sc0.2N (002), confirming that the Al0.8Sc0.2N layer has a single out-of-plane crystallographic orientation.

Fig. 3: Al0.8Sc0.2N film characterization.
figure 3

a Cross-sectional TEM image of the fabricated FBAR. b High-resolution TEM image of the Al0.8Sc0.2N and Mo interface. c Electron diffraction pattern of Al0.8Sc0.2N in region A. d XRD pattern of the Al0.8Sc0.2N film on a linear scale.

Full size image

The frequency response of Al0.8Sc0.2N-based FBARs is tested in air using a Cascade Microtech GSG probe station (Cascade, America) in conjunction with a network analyzer (N5222B, Agilent Technology). The measured impedance response of the resonator is shown in Fig. 4(a). The Keff2 of 14.5%, Qs of 150 and Qp of 318 are evaluated by23

$$K_{eff}^2 = \frac{{\pi ^2}}{4}\frac{{f_s}}{{f_p}}\frac{{f_p - f_s}}{{f_p}}$$
(1)
$$Q_{s,p} = \frac{{f_{s,p}}}{{\Delta f_{s,p} - 3dB}}$$
(2)

where fs and fp are the series and parallel resonant frequencies, respectively. The high Keff2 value of the FBAR is directly related to the enhanced piezoelectric performance with increased Sc content24. The measured result is modeled with the modified Butterworth-Van Dyke (MBVD) model with the circuit shown in the insets in Fig. 4(a). The model consists of a static capacitor C0, electrode resistance Rs, dielectric loss R0 and the motional branches, including a motional resistor Rm, a motional inductor Lm, and a motional capacitor Cm25. The three motional elements in the MBVD circuit can be defined as

$$C_m = C_0\left( {\left( {\frac{{f_P}}{{f_S}}} \right)^2 - 1} \right)$$
(3)
$$L_m = \frac{1}{{{{{\mathrm{\omega }}}}_s^2C_m}}$$
(4)
$$R_m = \frac{1}{{{{{\mathrm{\omega }}}}_sC_mQ_s}}$$
(5)

where ωs is the angular resonant frequency (ωs = 2πfs). The MBVD model fits the measured data with high accuracy, and using Eq. 6, Qm is calculated to be 424.7. The fitted components are shown in Table 1. Remarkably, due to the thin electrode, the electrode resistance Rs is 1.92 Ω, which is larger than the other two losses. Qs is strongly influenced by ohmic loss Rs, while at fp, the loss is negligible, resulting in Qs being almost half the value of Qp. Figure 4(b) shows the comparison of the performances of the fabricated FBAR in this work and previous studies12,15,16. The resonator presents the advantage of having a high frequency of 4.3 GHz and an excellent FOM of 62, which is essential for fabricating a 5G wideband filter.

$$Q_m = \frac{1}{{R_m}}\sqrt {\frac{{L_m}}{{C_m}}}$$
(6)
Fig. 4: Experimental results of Al0.8Sc0.2N-based FBAR.
figure 4

a Measured and MBVD fitted plot for Al0.8Sc0.2N-based FBAR. b Comparison with different studies regarding the performance of AlScN FBAR12,15,16.

Full size image
Table 1 MBVD model parameters.
Full size table

To further assess the Al0.8Sc0.2N properties, the Mason model is used to fit the measured data to extract important intrinsic material parameters. Figure 5(a) shows the Mason model of the FBAR derived from the one-dimensional acoustic wave equation23. In this model, Z is the characteristic acoustic impedance, k (k = 2πf/v) is the wavenumber, d is the thickness of each layer, and \(N = k_3d_3/(2\pi fk_t^2C_0Z_3)\), where \(k_t^2\) is the intrinsic electromechanical coupling factor of the Al0.8Sc0.2N film. The bulk material constants for the Mo electrodes and the AlN seed layer are used in this extraction procedure. The layer thicknesses are measured in advance by SEM, as reported in Section II. The acoustic impedance Z and the longitudinal acoustic velocity va are adjusted to fit the fp of the impedance curve. The intrinsic electromechanical coupling factor \(K_t^2\) is then adjusted to fit the fs of the curve. The fitted impedance of the Mason model and the measured impedance curve are shown in Fig. 5(b). With the equivalent Mason circuit, we can achieve a very consistent fitting result with the measured impedance. Furthermore, the unknown parameters (ρ, e33, \(C_{33}^E\)) can be obtained by26,27.

$$\rho = \frac{Z}{{v_a}}$$
(7)
$$c_{33}^D = \rho v_a^2$$
(8)
$$e_{33} = \sqrt {K_t^2c_{33}^D\varepsilon _r\varepsilon _0}$$
(9)
$$C_{33}^E = c_{33}^D - \frac{{e_{33}^2}}{{\varepsilon _r\varepsilon _0}}$$
(10)

where ρ is the mass density, \(c_{33}^D\) is the stiffened elastic constant, e33 is the piezoelectric stress constant, \(C_{33}^E\) is the elastic constant, and \(\varepsilon _{{{\mathrm{r}}}}\) is the relative dielectric permittivity. The \(\varepsilon _{{{\mathrm{r}}}}\) of the Al0.8Sc0.2N film is separately extracted from the static FBAR capacitance C0, applying the Formula C0 = \(\varepsilon _{{{\mathrm{r}}}}\varepsilon _0A/d\). Here, A is the active FBAR area, and d is the Al0.8Sc0.2N film thickness. Table 2 lists the extracted film parameters of Al0.8Sc0.2N, highlighting a larger e33 of 2.08 C/m2 than that of AlN (1.55 C/m2)28.

Fig. 5: Comparison between simulated value of Mason model and measured result.
figure 5

a Equivalent Mason circuit model of an FBAR structure. b The fitted impedance curve of the Mason model and the measured impedance curve.

Full size image
Table 2 Extracted film parameters
Full size table

By measuring the frequency responses at different temperatures, the temperature coefficients fs and fp of the proposed Al0.8Sc0.2N-based FBAR are found to be −19.2 ppm/°C and −21.1 ppm/°C, respectively, as shown in Fig. 6. TCF is usually an intrinsic property of materials that is dependent on the growth conditions. The TCF of the proposed device is comparable with previously reported values29,30. The difference in TCF between fs and fp is caused by a slight shift in \(K_t^2\) since the change in temperature may induce thermal stress in the thin film.

Fig. 6: Temperature characteristics of the fabricated Al0.8Sc0.2N-based FBAR.
figure 6

a Frequency responses of the fabricated Al0.8Sc0.2N-based FBAR at temperatures from −20 °C to 80 °C. b Measured TCF of the proposed FBAR.

Full size image

Figure 7(a) and (b) show the SEM image and schematic circuit of the fabricated filter. It consists of 8 elements, including 4 series and 4 shunt resonators. To achieve passband transmit characteristics, a 37 nm Mo mass loading layer is added to the shunt resonator to make its resonant frequency lower than the series resonator. Table 3 summarizes the relevant measured and extracted parameters of the series and shunt resonators used in the filter. Moreover, additional Mo layers are constructed on the interconnect line between each adjacent resonator to improve the performance of the filter. A plot of the measured filter response is shown in Fig. 7(c) and (d). The filter shows a center frequency of 4.24 GHz, a −3 dB bandwidth of 215 MHz, a small IL of 1.881 dB, and a rejection >32 dB. The measured return loss at the center frequency is below −12 dB, indicating a good match to 50 Ω without any helper circuits.

Fig. 7: Experimental results of the fabricated FBAR filter.
figure 7

a SEM image for the fabricated FBAR filter. b Schematic circuit design of the FBAR filter. c Measured transmission response of the fabricated FBAR filter. d Measured return loss of the fabricated FBAR filter.

Full size image
Table 3 Measured and extracted FBAR parameters.
Full size table

Conclusion

In this paper, an Al0.8Sc0.2N-based FBAR is investigated for the design of acoustic MEMS filters. The proposed FBAR devices are yielded by a seven-mask layer wafer process, including sputtering deposition, ICP etching, and cavity release. Characterization of the microstructural and crystal structure properties of the sputtered Al0.8Sc0.2N thin film is performed. The results show that the Al0.8Sc0.2N layer has a good c-axis preferential crystal orientation. For the FBAR devices based on the Al0.8Sc0.2N film with a large e33 (2.08 C/m2), a large Keff2 of 14.5% and an improved FOM of 62 are simultaneously obtained. Furthermore, the temperature characteristics of the Al0.8Sc0.2N-based FBAR are studied. The TCFs of fs and fp of the proposed resonator are found to be −19.2 ppm/°C and −21.1 ppm/°C, respectively, indicating excellent temperature stability. The fabricated filter based on the above resonators has a center frequency of 4.24 GHz, a −3 dB bandwidth of 215 MHz, a small IL of 1.881 dB, and a rejection >32 dB, showing the strong potential for 5G RF front-end applications.