Capacitive micromachined ultrasonic transducers: Acoustic challenges and proposed solutions

Abstract

Capacitive Micromachined Ultrasonic Transducer (CMUT) have been subject to research by several research groups during the last two decades. Despite many potential advantages over traditional piezoelectric ultrasound transducers, the CMUT technology has not yet made a proper commercial breakthrough. One issue which we believe need more investigation is the control of the acoustic crosstalk, both through the silicon substrate and through the adjacent fluid medium.

The work presented in this thesis concerns the modeling of immersed CMUT arrays and the investigation of two different acoustic crosstalk effects which may harm the transducer response. The CMUT array is modeled with an analytical model describing the motion of the single acoustically isolated CMUT cell as a combination of free acoustic vibration modes. Several CMUT cells may be connected to form larger elements, and the vibration modes of adjacent CMUTs are coupled through the fluid outside, by an acoustic impedance matrix. In addition, the model may also include the motion of the silicon substrate and the in uence from the source impedance of the electronics.

The first crosstalk effect which we focus on in this work, is the generation of surface acoustic waves (SAWs) along the surface of the silicon substrate supporting the CMUT array. If the SAWs are not damped in any way, they may couple to waves in the fluid at certain steering angles, and cause a drop in transmission eciency at these angles. In the presented simulation we show that the silicon substrate must be limited in thickness in order for the backing material to be able to damp the SAWs. If the CMUT array is mounted on top of a stack of silicon substrates containing transmit and receive circuitry, the thickness and composition of both the bonding materials and the silicon substrates must be taken into account. We have compared three different bonding techniques, surface liquid interdi usion (SLID) bonding, anisotropic conductive adhesive (ACA) and direct fusion bonding, and we show that fusion bonding is the technique which is best suited for high frequency CMUT arrays with several IC chips underneath. The penetration depth of the SAWs is frequency dependent, and we show that a high frequency CMUT array with center frequency around 30 MHz, stacked on top of three circuit layers, should have a total silicon thickness below 100 m. If the acoustic backing material instead is placed between the CMUT array and the rst circuit chip, other bonding techniques than fusion bonding may also be used, and the thicknesses of the IC chips are no longer of importance. However, the transmission of the electric signal through or around the backing layer might be a challenge.

The other crosstalk effect which has been investigated in this thesis, is the inter-element coupling at the CMUT- fluid interface. We refer to this effect as dispersive guided modes, and show that the excitation of local resonances in the interface region may affect the overall transmission from the array at frequencies well within the 100 % bandwidth of the transducer. These waves are not damped by the acoustic backing material. Many CMUT designers choose to have larger distances between CMUT cells in neighbor elements than between CMUTs within an element. We denote this as double periodicity. We have compared the e ect of the dispersive guided modes in arrays with the same distance between all the CMUT cells, regardless if they are in the same or in adjacent elements (denoted as single periodicity), and arrays with double periodicity. Simulations show that the response from arrays with double periodicity is affected by the crosstalk even at broadside radiation, whereas the effect becomes apparent at o -axis beam steering for arrays with single periodicity.

Through simulations, we have shown that it is possible to mechanically damp the dispersive guided modes substantially by introducing a lossy coating material of a few micrometer thickness. The PDMS material RTV516 from GE Silicones seems to have material properties which are well suited for such damping. We have also shown that introducing a low electric source impedance in the transmit electronics may reduce the e ect of the local CMUT- uid resonances on the detected signal.