Sensitivity and Directivity Analysis of Piezoelectric Ultrasonic Cantilever-Based MEMS Hydrophone for Underwater Applications

In this paper, we report on the characterization of the sensitivity and the directionality of a novel ultrasonic hydrophone fabricated by micro-electro-mechanical systems (MEMS) process, using aluminum nitride (AlN) thin film as piezoelectric functional layer and exploiting a stress-driven design. Hydrophone structure and fabrication consist of four piezoelectric cantilevers in cross configuration, whose first resonant frequency mode in water is designed between 20 kHz and 200 kHz. The MEMS fabricated structures exploit 1 μm and 2 μm thick piezoelectric AlN thin film embedded between two molybdenum electrodes grown by DC magnetron sputtering on silicon (Si) wafer. The 200 nm thick molybdenum electrodes thin layers add a stress-gradient through cantilever thickness, leading to an out-of-plane cantilever bending. A water resistant parylene conformal coating of 1 μm was deposited on each cantilever for waterproof operation. AlN upward bent cantilevers show maximum sensitivity up to −163 dB. The cross configuration of four stress-driven piezoelectric cantilevers, combined with an opportune algorithm for processing all data sensors, permits a finer directionality response of this hydrophone.


Introduction
Hydrophones are underwater acoustic receivers that play an increasingly important role in submarine resources exploration, marine military, underwater noise monitoring, and sonar systems [1]. Hydrophone design and fabrication are based on micro-electro-mechanical systems (MEMS) technology, combining solid-state physics, mechanics, acoustics, and electronics [2] to detect underwater sounds. A piezoelectric-based MEMS hydrophone is an electroacoustic transducer converting mechanical excitation due to acoustic pressure into electrical signal. Usually, the readout mechanism of micro-electro-mechanical system (MEMS) devices exploits piezoelectric or piezoresistive material as sensing element [3][4][5]. MEMS-based hydrophones have been demonstrated to be able to trace the exact location as well as direction of underwater sound sources [6]. In 1996, a MEMS-based hydrophone was reported as the first directional underwater acoustic sensor [7]. Since then, MEMS ultrasonic transducers, in which thin film of piezoelectric material is used as sensing elements, have been widely studied and used for numerous underwater applications [8,9]. A bio mimetic approach has been successfully pursued to design cantilever and membrane base hydrophone. Two T-shape vector hydrophone using MEMS technology inspired by fish lateral line, based on piezoresistive cantilever 3 of 15 This work is divided in three sections. First, the design and analysis of the micro cantilevers model by FEM is reported; second, the MEMS fabrication process is described; and finally, the sensitivity and directionality of the hydrophone is characterized.

Device Design and Fabrication
Piezoelectricity is a coupling mechanism relating the mechanical and electrical properties of a material. An electrical charge is produced when the piezoelectric material is mechanically deformed and vice versa. The piezoelectric constitutive equations, also known as "coupled equations" are given below [52] in the stress-charge form: where S is the strain tensor, s E is the elasticity matrix, T is the stress tensor, e is piezoelectric coupling matrix, D is the tensor of electric displacement, ε is the electrical permittivity, and E is the electric field. In order to explore nitride-based MEMS directional hydrophones and to maximize the receiving sensitivity over the desired frequency range, a stress-driven structure of the device was designed and simulated by using the finite element method (COMSOL Multiphysics). For designing the transducer, acoustics-structure interaction and piezoelectric effect was simulated in water environment (see Figure 1a). The cantilever was fixed at one end while all the other faces were unconstrained, allowing the bending of the device. The mesh was composed of 202,168 elements, using free quad and free tetrahedral finite elements. Two different structures have been simulated, exploiting a piezoelectric aluminum nitride (AlN) functional layer of 1 µm and 2 µm thickness, having two flexural stiffnesses of approximately 3 × 10 −11 N m 2 and 9 × 10 −11 N m 2 , respectively. Higher AlN thickness provides higher flexural stiffness making the cantilever less responsive and brittle. The following piezoelectric coefficient for AlN has been used: e 13 = −0.58 C m −2 and e 33 = 1.55 C m −2 [53], where 200 nm thick top and bottom molybdenum (Mo) electrodes were implemented. Finally, in order to avoid short circuits in water, a parylene conformal coating of 1 µm thickness was applied. Table 1 shows the mechanical properties of materials and their thicknesses, used to design and simulate the micro cantilever's response. In a first set of simulations, different lengths of cantilevers, ranging between 100 µm to 500 µm and a constant width of each cantilever equal to 70 µm, have been investigated. Out of plane upward curvature was fixed at 5 × 10 −4 µm −1 (one of the curvature values experimentally found) for both cantilevers and a constant acoustics force per surface unit equals to 5 kN/m 2 was applied. FEM simulations in Figure 1b demonstrate two different fluid structure interaction mechanisms: for a fixed upward curvature at a constant acoustic intensity, there is an increase of signal with length while higher signals are generated by increasing the flexural stiffness of the cantilever. In fact, in "out-of-plane" bent piezoelectric cantilevers, the signal is strongly dependent on the geometrical features of the beam. The acoustic wave perturbation in the water, distributing its drag force on the whole beam surface, is intense enough to further bend the cantilever and generate a signal from the sound traveling through water. At the same acoustic intensity, the drag force acting on each sensor is directly dependent from the apparent cantilever area (e.g., by increasing the length), as a consequence [44]. A further control in the hydrophone response to the acoustic excitation can be introduced by changing the flexural stiffness of the cantilever-based hydrophone, achieved by exploiting different layer thicknesses. Provided the same apparent cantilever area and a fixed mechanical excitation, thinner cantilever distributes the acoustic deformation all along the cantilever beam, experiencing a lower stress on the hinge with a consequent lower piezoelectric signal. In contrast, thicker cantilevers will experience a higher external mechanical stress localized at the hinge; therefore, a higher piezoelectric voltage and a better sensitivity is obtained. Both mechanisms are observed experimentally.
The designed 3-D model of micro cantilevers with different lengths (between 100 µm to 1000 µm) have been analyzed using the eigenfrequency study to set the first resonant frequency mode in water between 20 kHz to 200 kHz, the desired acoustic range of the device ( Figure 1c). As expected, Figure 1c shows that the first resonance frequency of micro cantilevers decreases as their length increases. It is noteworthy, due to damping effect, that the first resonance mode of frequency of each micro cantilever in water decreases as compared to that in air for both thicknesses [54]. Both cantilevers resonances fall in the ultrasonic frequency range. whole beam surface, is intense enough to further bend the cantilever and generate a signal from the sound traveling through water. At the same acoustic intensity, the drag force acting on each sensor is directly dependent from the apparent cantilever area (e.g., by increasing the length), as a consequence [44]. A further control in the hydrophone response to the acoustic excitation can be introduced by changing the flexural stiffness of the cantilever-based hydrophone, achieved by exploiting different layer thicknesses. Provided the same apparent cantilever area and a fixed mechanical excitation, thinner cantilever distributes the acoustic deformation all along the cantilever beam, experiencing a lower stress on the hinge with a consequent lower piezoelectric signal. In contrast, thicker cantilevers will experience a higher external mechanical stress localized at the hinge; therefore, a higher piezoelectric voltage and a better sensitivity is obtained. Both mechanisms are observed experimentally. The designed 3-D model of micro cantilevers with different lengths (between 100 µm to 1000 µm) have been analyzed using the eigenfrequency study to set the first resonant frequency mode in water between 20 kHz to 200 kHz, the desired acoustic range of the device ( Figure 1c). As expected, Figure 1c shows that the first resonance frequency of micro cantilevers decreases as their length increases. It is noteworthy, due to damping effect, that the first resonance mode of frequency of each micro cantilever in water decreases as compared to that in air for both thicknesses [54]. Both cantilevers resonances fall in the ultrasonic frequency range.  These cantilevers were fabricated by a standard MEMS process, previously applied to the piezoelectric aluminum nitride-based design grown on a silicon substrate [44]. A thin layer of molybdenum (200 nm) was deposited by DC magnetron sputtering. Mask aligner (SUSS MA8/BA8) exposure was used after spin coating of AZ5214E photoresist with 2 µm thickness for bottom electrode definition ( Figure 2a). Then, aluminum nitride (1 µm and 2 µm) and molybdenum (200 nm) were deposited by DC magnetron sputtering. TI35E photoresist with thickness of 4 µm was spin coated for piezoelectric functional layer and top electrode definition (Figure 2b). MIF AZ826 was applied for development of all photoresists. Inductively coupled plasma (ICP) etching with Boron Trichloride (BCl3, 45 sccm) and Nitrogen (N2, 20 sccm) for Mo layer, while Boron Trichloride (BCl3, 100 sccm) and Argon (Ar, 25 sccm) for AlN layer were used to remove exposed layer. Both recipes had RF powers applied to platen equals to 250 W and to coil equals to 600 W, respectively. For cantilever release, silicon etching was performed at 700 sscm of SF6, coil power of 2600 W and pressure of 100 mTorr at temperature 18 • C. Two possible curvatures (Figure 2c), upwards bending (tensile stress) or downwards bending (compressive stress), can be observed. J. Mar. Sci. Eng. 2020.

of 16
These cantilevers were fabricated by a standard MEMS process, previously applied to the piezoelectric aluminum nitride-based design grown on a silicon substrate [44]. A thin layer of molybdenum (200 nm) was deposited by DC magnetron sputtering. Mask aligner (SUSS MA8/BA8) exposure was used after spin coating of AZ5214E photoresist with 2 µm thickness for bottom electrode definition ( Figure 2a). Then, aluminum nitride (1 µm and 2 µm) and molybdenum (200 nm) were deposited by DC magnetron sputtering. TI35E photoresist with thickness of 4 µm was spin coated for piezoelectric functional layer and top electrode definition ( Figure 2b). MIF AZ826 was applied for development of all photoresists. Inductively coupled plasma (ICP) etching with Boron Trichloride (BCl3, 45 sccm) and Nitrogen (N2, 20 sccm) for Mo layer, while Boron Trichloride (BCl3, 100 sccm) and Argon (Ar, 25 sccm) for AlN layer were used to remove exposed layer. Both recipes had RF powers applied to platen equals to 250 W and to coil equals to 600 W, respectively. For cantilever release, silicon etching was performed at 700 sscm of SF6, coil power of 2600 W and pressure of 100 mTorr at temperature 18 °C. Two possible curvatures (Figure 2c), upwards bending (tensile stress) or downwards bending (compressive stress), can be observed.  Figure 3 show selected cantilevers with different lengths and bending, in dependence of increased thickness of AlN. Figure 3a,c shows cantilevers bending in downward direction (AlN thickness equals to 1 µm) while Figure 3b,d shows cantilevers with bending in upward direction (AlN thickness equals to 2 µm). This is in good agreement with Stoney equation for cantilever [44], where the curvature C is expressed:

SEM images in
where R is radius of curvature t is the thickness of the cantilever, E is the Young modulus, and δ is the modulus of the surface stress. The intrinsic stress of AlN accumulated during the sputtering process and residual stress gradient inside the multilayer structure causes the bending in upward direction for 2 µm thick AlN, having an average tensile surface stress of 438 N/m. In contrast, 1 µm thick AlN cantilevers show a slightly downward direction due to compressive residual stress gradient with an average surface stress of 53 N/m [2]. This is explained because a change on AlN thickness leads to a shift in neutral axis position, due to a change in the whole beam stress due to different growth time and temperature. The mismatch of atomic sizes between different layered  Figure 3 show selected cantilevers with different lengths and bending, in dependence of increased thickness of AlN. Figure 3a,c shows cantilevers bending in downward direction (AlN thickness equals to 1 µm) while Figure 3b,d shows cantilevers with bending in upward direction (AlN thickness equals to 2 µm). This is in good agreement with Stoney equation for cantilever [44], where the curvature C is expressed:

SEM images in
where R is radius of curvature t is the thickness of the cantilever, E is the Young modulus, and δ is the modulus of the surface stress. The intrinsic stress of AlN accumulated during the sputtering process and residual stress gradient inside the multilayer structure causes the bending in upward direction for 2 µm thick AlN, having an average tensile surface stress of 438 N/m. In contrast, 1 µm thick AlN cantilevers show a slightly downward direction due to compressive residual stress gradient with an average surface stress of 53 N/m [2]. This is explained because a change on AlN thickness leads to a shift in neutral axis position, due to a change in the whole beam stress due to different growth time and temperature. The mismatch of atomic sizes between different layered materials or lattice-mismatch developed in the thermal cycling of the chip during material deposition and micro-fabrication is the origin in the stress gradient [45]. If needed, a careful setup of growth conditions would allow to neutralize the bending or to place it out of the AlN thickness position in order to further bends the cantilevers and improve the device signals [2]. In Figure 3e, SEM image of face-to-face, cross-configuration of four cantilevers, having 300 µm length has been shown while Figure 3f shows tip of fabricated micro cantilever with conformal parylene coating. materials or lattice-mismatch developed in the thermal cycling of the chip during material deposition and micro-fabrication is the origin in the stress gradient [45]. If needed, a careful setup of growth conditions would allow to neutralize the bending or to place it out of the AlN thickness position in order to further bends the cantilevers and improve the device signals [2]. In Figure 3e, SEM image of face-to-face, cross-configuration of four cantilevers, having 300 µm length has been shown while Figure 3f shows tip of fabricated micro cantilever with conformal parylene coating.  Electrical characterization was realized by LCR meter (Keysight-E4980AL) at 1 kHz frequency and excitation amplitude of two volts, to measure the transducer capacitance as shown in Figure 4a. The capacitance of cantilevers increases with length and decreases with thickness of aluminum nitride, as expected by theory. Figure 4b shows laser doppler vibrometer (Polytec Vibrometer MSA500) measurements at five volts excitation for different cantilever lengths, ranging between 100 µm and 300 µm, with aluminum nitride thickness of 1 µm and 2 µm. The figure shows that the measured first resonance mode frequency is in agreement with the simulated frequency of the cantilevers. The deviation between simulations and measurements was present only for one length for the 2 µm thick aluminum nitride layer as shown in Figure 4b, most likely due to a small difference between the nominal geometrical dimensions and the fabrication results. Electrical characterization was realized by LCR meter (Keysight-E4980AL) at 1 kHz frequency and excitation amplitude of two volts, to measure the transducer capacitance as shown in Figure 4a. The capacitance of cantilevers increases with length and decreases with thickness of aluminum nitride, as expected by theory. Figure 4b shows laser doppler vibrometer (Polytec Vibrometer MSA500) measurements at five volts excitation for different cantilever lengths, ranging between 100 µm and 300 µm, with aluminum nitride thickness of 1 µm and 2 µm. The figure shows that the measured first resonance mode frequency is in agreement with the simulated frequency of the cantilevers. The deviation between simulations and measurements was present only for one length for the 2 µm thick aluminum nitride layer as shown in Figure 4b, most likely due to a small difference between the nominal geometrical dimensions and the fabrication results.

Underwater Characterization
Characterization of hydrophone mostly refers to the sensitivity and directionality. Measurement set up was composed by of emission transducer (200 kHz), reference hydrophone (Onda HNP-1000), oscilloscope (Tektronix, MSO2000B), MEMS transducer, function generator (DPR 300, control with PC) and a rotary stage. A pulsed acoustic wave centered at 200 kHz was produced by an emission transducer driven by a pulse generator at 5 kHz pulse repetition frequency (PRF) and 475 volts as supply voltage to emission transducer. Measurements parameter for underwater characterization of hydrophone have been described deeply in the previous work [44]. Peak output voltages from each cantilever were measured by oscilloscope. It is clear from the Figure 5a that the average output signal response of each cantilever is increasing with length. The highest values of voltage response were achieved with a thickness of 2 µm of AlN. In fact, cantilevers having upward bending increases the acoustic-structure interaction super linearly. Sensitivity measurements were performed by normalizing the signal response to a reference hydrophone with MEMS transducer. The sensitivity of the hydrophone is expressed [44,55]

Underwater Characterization
Characterization of hydrophone mostly refers to the sensitivity and directionality. Measurement set up was composed by of emission transducer (200 kHz), reference hydrophone (Onda HNP-1000), oscilloscope (Tektronix, MSO2000B), MEMS transducer, function generator (DPR 300, control with PC) and a rotary stage. A pulsed acoustic wave centered at 200 kHz was produced by an emission transducer driven by a pulse generator at 5 kHz pulse repetition frequency (PRF) and 475 volts as supply voltage to emission transducer. Measurements parameter for underwater characterization of hydrophone have been described deeply in the previous work [44]. Peak output voltages from each cantilever were measured by oscilloscope. It is clear from the Figure 5a that the average output signal response of each cantilever is increasing with length. The highest values of voltage response were achieved with a thickness of 2 µm of AlN. In fact, cantilevers having upward bending increases the acoustic-structure interaction super linearly. Sensitivity measurements were performed by normalizing the signal response to a reference hydrophone with MEMS transducer. The sensitivity of the hydrophone is expressed [44,55] as where S MEMS is the sensitivity of the tested hydrophone and S re f (−253 dB re. 1 V/µPa) [56] is the sensitivity of the reference hydrophone (HNP-1000 Broadband Needle Type and Hydrophone Preamplifier, 20 dB gain, Onda Corporation). V MEMS and V re f are the peak voltage response from MEMS and reference hydrophone, respectively, k denotes the number of waves and d is the distance from surface of water to reference hydrophone and to the MEMS transducer under test. Figure 5b shows sensitivity measurements of MEMS fabricated hydrophone at 5 kHz PRF and 475 volts supply voltage. It shows that sensitivity increases with length of the cantilevers and, at the same length increases with thickness. Highest sensitivity up to −163 dB was achieved by Mo/AlN/Mo hydrophone with 300 µm length and 2 µm thick AlN functional layer upward cantilever. where is the sensitivity of the tested hydrophone and (−253 dB re. 1 V/µPa) [56] is the sensitivity of the reference hydrophone (HNP-1000 Broadband Needle Type and Hydrophone Preamplifier, 20 dB gain, Onda Corporation). and are the peak voltage response from MEMS and reference hydrophone, respectively, k denotes the number of waves and d is the distance from surface of water to reference hydrophone and to the MEMS transducer under test. Figure 5b shows sensitivity measurements of MEMS fabricated hydrophone at 5 kHz PRF and 475 volts supply voltage. It shows that sensitivity increases with length of the cantilevers and, at the same length increases with thickness. Highest sensitivity up to −163 dB was achieved by Mo/AlN/Mo hydrophone with 300 µm length and 2 µm thick AlN functional layer upward cantilever.  Table 2 shows the comparison of MEMS hydrophone with other cantilever-based hydrophones based on different transduction mechanisms and including commercial ones. It shows that MEMS hydrophone has comparable sensitivity with other devices. Moreover, it has the advantage of miniaturization and due to cross-configuration; it presents directional response, which allows finding the acoustic source direction.   Table 2 shows the comparison of MEMS hydrophone with other cantilever-based hydrophones based on different transduction mechanisms and including commercial ones. It shows that MEMS hydrophone has comparable sensitivity with other devices. Moreover, it has the advantage of miniaturization and due to cross-configuration; it presents directional response, which allows finding the acoustic source direction.

Directionality
MEMS hydrophone was located on a rotary stage and the directivity pattern of the MEMS hydrophone was measured by collecting the signal response for every 10 • rotation angles with respect to acoustic source. In the present method, all cantilevers have been excited to operate in four main read-out modes: Cardioid, Omni, Dipole, and Quadrant mode. Figure 6 shows directional cardioid shaped response from each single cantilever of the cross configurations. In details, Figure 6a,b show aluminum nitride based cantilevers (2 µm and 1 µm AlN thickness, respectively). The maximum cardioid response was measured for the upward bent cantilever, with 2 µm thick AlN (Figure 6a). Conversely, the 1 µm thick AlN (Figure 6b) shows a lower response because of slightly downward bending. Each single cantilever in cross configuration show directivity pattern with a sensing directionality aperture close to 160 • . Figure 7 shows different Omni directivity pattern signal combinations for the two different cross-configurations, obtained by summing up all signal response of each cantilever. Through Omni directivity configuration, hydrophone can detect sounds uniformly in all the direction. By comparison between Figures 6 and 7, cantilevers from a cross configuration made by upwards cantilevers and a thicker piezoelectric aluminum nitride layer have a higher voltage response and sensitivity, therefore being more sensitive to all the direction as compared to the cross configuration of thinner slightly downward cantilevers. From now on, alternative directionality patterns have been investigated exploiting the most sensitive cantilever cross configuration (i.e., 2 µm thick AlN).

Directionality
MEMS hydrophone was located on a rotary stage and the directivity pattern of the MEMS hydrophone was measured by collecting the signal response for every 10° rotation angles with respect to acoustic source. In the present method, all cantilevers have been excited to operate in four main read-out modes: Cardioid, Omni, Dipole, and Quadrant mode. Figure 6 shows directional cardioid shaped response from each single cantilever of the cross configurations. In details, Figure 6a,b show aluminum nitride based cantilevers (2 µm and 1 µm AlN thickness, respectively). The maximum cardioid response was measured for the upward bent cantilever, with 2 µm thick AlN (Figure 6a). Conversely, the 1 µm thick AlN (Figure 6b) shows a lower response because of slightly downward bending. Each single cantilever in cross configuration show directivity pattern with a sensing directionality aperture close to 160°. Figure 7 shows different Omni directivity pattern signal combinations for the two different cross-configurations, obtained by summing up all signal response of each cantilever. Through Omni directivity configuration, hydrophone can detect sounds uniformly in all the direction. By comparison between Figures 6 and  7, cantilevers from a cross configuration made by upwards cantilevers and a thicker piezoelectric aluminum nitride layer have a higher voltage response and sensitivity, therefore being more sensitive to all the direction as compared to the cross configuration of thinner slightly downward cantilevers. From now on, alternative directionality patterns have been investigated exploiting the most sensitive cantilever cross configuration (i.e., 2 µm thick AlN).    Figure 8a,b shows a dipole beam pattern configuration, consisting of subtracting the sum of signals coming from facing cantilevers on one axis with the sum of cantilevers signals coming from the respective orthogonal axis, for the upward bent 2 µm thick aluminum nitride based cantilevers with length of 300 µm. We can define x-axis direction dipole and y-axis direction dipole by commuting the minuend and subtrahend. The regions, where the dipolar signals are positive, are approximately 40° wide along both x-axis and y-axis direction. The red circle defines the regions where the voltage signal changes between positive and negative value; for each dipole mode, only the positive signal will be exploited for discriminating directionality. Noteworthy, dipole directivity can discriminate axial directions but not the sound source side position (90° or 270° for x-axis in Figure 8a and 0° or 180° for y-axis in Figure 8b). Finally, quadrant modes (their mathematical expression are in each legend of Figure 8c-f) give a pattern with a positive signal in 90° wide aperture range placed in each quadrant (0° to 90°, 90° to 180°, 180° to 270°, and 270° to 360°, respectively, as shown in Figure 8c-f. Therefore, a directional quadrant mode needs to be defined for each quadrant.  Focusing on the direction ranges where signals are positive, a suitable combination of the two dipole and four quadrant configurations allows to easily refine the direction of incoming sounds. As an example, the contemporary combination of the positive range (from +50 • to 140 • and from 240 • to 330 • ) of the dipolar configuration in Figure 8a with the maximum signal range (from 0 • to 90 • ) of the configuration in Figure 8c gives the intersection range which allows to identify the direction of sound coming approximately between 50 • and 90 • with aperture close to 40 • , as shown in Figure 9a. Similarly, if both dipole axis have a value in modulus very close to 0 mV, the quadrant, among the four different quadrant modes, giving the maximum positive value identifies the inter-cardinal direction the acoustic signal is coming from. In fact, as a further example, the combination of a zero signal of the configurations in Figure 8a,b with the maximum positive value of the configuration in Figure 8c identifies the direction of sound coming at approximately 45 • , as shown in Figure 9b. The same strategy can be adopted for the other cardinal directions, exploiting the opportune dipole/quadrant combination. direction the acoustic signal is coming from. In fact, as a further example, the combination of a zero signal of the configurations in Figure 8a,b with the maximum positive value of the configuration in Figure 8c identifies the direction of sound coming at approximately 45°, as shown in Figure 9b. The same strategy can be adopted for the other cardinal directions, exploiting the opportune dipole/quadrant combination.

Conclusions
In this work, design, fabrication, and characterization of a stress-driven cantilever based directional MEMS piezoelectric ultrasonic hydrophone have been described. Experimental results showed that hydrophone in cross configuration has high sensitivity in ultrasonic frequency range and it is able to detect underwater acoustic source directions. Laser doppler vibrometer measurement of 100 µm to 300 µm long cantilevers showed that measured resonance frequencies were in agreement with modeled resonance by FEM. A study on cross configurations of different "out-of-plane" bent cantilever lengths and different aluminum nitride thickness shows sensitivity improvement with length and upward vertical displacement. Highest sensitivity of −163 dB was achieved by 2 µm thick upward aluminum nitride for 300 µm long cantilever. These results show that hydrophone with upward bent cantilevers improves voltage response, has a higher sensitivity and better directivity pattern compared to other cantilevers in different cross configuration. Cardioid directionality pattern shows that each single cantilever is capable to identify acoustic source direction with a very large uncertainty up to 160°. Combination of signals from cross configuration cantilevers allowed to define

Conclusions
In this work, design, fabrication, and characterization of a stress-driven cantilever based directional MEMS piezoelectric ultrasonic hydrophone have been described. Experimental results showed that hydrophone in cross configuration has high sensitivity in ultrasonic frequency range and it is able to detect underwater acoustic source directions. Laser doppler vibrometer measurement of 100 µm to 300 µm long cantilevers showed that measured resonance frequencies were in agreement with modeled resonance by FEM. A study on cross configurations of different "out-of-plane" bent cantilever lengths and different aluminum nitride thickness shows sensitivity improvement with length and upward vertical displacement. Highest sensitivity of −163 dB was achieved by 2 µm thick upward aluminum nitride for 300 µm long cantilever. These results show that hydrophone with upward bent cantilevers improves voltage response, has a higher sensitivity and better directivity pattern compared to other cantilevers in different cross configuration. Cardioid directionality pattern shows that each single cantilever is capable to identify acoustic source direction with a very large uncertainty up to 160 • . Combination of signals from cross configuration cantilevers allowed to define Omni, Dipole, and Quadrant directionality pattern. Acoustic sounds in all the direction can be determined through Omni directionality pattern, which is obtained by summing up all the output voltages response coming from each cantilever in the cross configuration. By exploiting maximum positive amplitude values of Dipole and Quadrant beam pattern and their appropriate combination, underwater acoustics signal can be distinguished for their incoming directions with a lower uncertainty (40 • ) with a virtual negligible uncertainty at inter-cardinal directions. These results suggest stress-driven piezoelectric cantilever based ultrasonic hydrophones have a potential high impact on a wide range of technological underwater applications.