Open Access
Add to BibSonomy
Abstract
This paper reports the fabrication and characterization of the first flexible transparent capacitive micromachined ultrasound transducer (CMUT) array for through-illumination photoacoustic tomography. Fabricated based on an adhesive wafer bonding technique and a PDMS backfill approach, the array has a maximum transparency of 67% in visible light range and can be bent to a radius of curvature of less than 5 mm without the structural layers being damaged. With a center frequency of 3.5 MHz, 80% fractional bandwidth, and noise equivalent pressure (NEP) of 62 mPa/$\sqrt {\mathrm {Hz}}$, the array was successfully used in limited-view photoacoustic tomography of a 100 µm wire target, demonstrating lateral and axial resolutions of 293 µm and 382 µm, respectively, with 46 dB signal-to-noise ratio. Additionally, deep tissue photoacoustic tomography was also demonstrated on a blood tube within a chicken tissue using the fabricated CMUT arrays.
© 2022 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement
1. Introduction
Photoacoustic imaging is an emerging, non-invasive imaging technique which has drawn tremendous attention in recent years. The capability of this method in providing optical contrast of the targets deep in tissue can be used for imaging optically-absorbing endogenous and exogenous chromophores, which has applications in cancer imaging, cardiovascular imaging, and more [1]. Basically, this imaging modality combines optical and ultrasound techniques to generate a detailed image of tissue. First, the target is illuminated with short laser pulses, usually in the range of the visible or near-infrared spectrum. Upon absorbing the laser energy, the targets undergo transient thermoelastic expansion, causing them to emit broadband ultrasound waves, which are then detected using ultrasound transducers in order to construct a photoacoustic image. In general, photoacoustic imaging is done in oblique illumination mode, in which the optical path is different from the ultrasonic path (e.g. the sample is illuminated from the side and the ultrasound waves are detected by the transducer located on top of the sample). Developing a through-illumination photoacoustic imaging system in which the optical and ultrasonic waves share the same path can offer numerous advantages such as less attenuation due to minimized optical and ultrasound paths, more uniform illumination of the targets, and improved signal to noise ratio [2]. In this work, we present the fabrication of the first flexible transparent capacitive micromachined ultrasound transducer (CMUT) array. While the transparency of the array enables the laser to pass through it for through-illumination photoacoustic imaging, its flexibility can be utilized for conformal or curved applications such as tomography arrays, which is the focus of this paper. The larger the angular coverage of a tomography array, the finer the resolution of the images. Moreover, reconstruction artifacts such as edge-waves can be mitigated with a curved tomography array.
2. Background
Conventional photoacoustic imaging systems use a piezoelectric transducer array to detect the ultrasound signals. However, most such piezoelectric materials are not optically transparent and require matching and backing layers and electrodes, which are typically opaque. Polyvinylidene fluoride (PVDF) is a piezoelectric polymer which has promise for flexible or curved transparent transducers [3]. However, the long-term objective of the current work is to fabricate flexible transparent 2D arrays for 3D photoacoustic imaging. This is non-trivial owing to requirements of not only transparent flexible transducer materials but also requirements for dense wiring. Recently, row-column arrays were introduced that enable readout from a 2D array using only row and column addressing to perform real-time 3D photoacoustic imaging [4]. These arrays required elements with bias sensitivity. While PVDF is flexible and can be made optically transparent, it is not necessarily bias sensitive as required for future row-column arrays. Capacitive micromachined ultrasound transducers (CMUTs) are an alternative method that can address these limitations. CMUTs were first introduced to improve the operational frequency range of an airborne ultrasound transducer [5]. However, due to their great performance in immersion applications [6], CMUTs were proposed as a promising alternative to their piezoelectric counterparts to overcome their limitations. Basically, a CMUT is composed of a deflecting membrane and a fixed substrate separated by a vacuum gap, which form the electrodes of a variable capacitor. These devices can compete with piezoelectric transducers in terms of fractional bandwidth [7–9] and receive sensitivity [10,11]. Furthermore, the advances in microfabrication technology as well as material science in recent decades offers numerous advantages on the fabrication side of CMUTs such as more design flexibility, and ease of integration with other electronics. More importantly, CMUTs exhibit bias sensitive operation [12], which makes them suitable for future bias-sensitive 2D arrays.
In recent years, researchers have investigated the fabrication of transparent CMUTs to be used in through-illumination photoacoustic systems. Chen et al [13] pioneered through-transducer illumination photoacoustic imaging by fabrication of an infrared transparent CMUT array. This was achieved by thinning the silicon substrate of the device down to 100 $\mu$m and coating it with silicon nitride as an anti-reflection layer, which offered a transparency of 12% in the infrared region. Zhang et al [14] proposed using a glass substrate and indium tin oxide (ITO) as the electrode material in CMUTs to improve the transparency of the device. However, the transparency was still low in shorter wavelengths due to the silicon membrane in these devices. They used a single-element CMUT with an optical fiber attached to its backside, and scanned it over phantoms to obtain a 2D and 3D photoacoustic images, which can be much slower compared to a linear or two-dimensional array. In another work, they [15] further improved the transparency of the CMUT arrays by using BCB and SU-8 as the adhesive material in an adhesive wafer bonding process for device fabrication. The transparency of the device was measured to reach 70%-80% in visible range, although the performance of the array was not evaluated for photoacoustic applications. Li et al [2] fabricated a transparent CMUT element based on adhesive wafer bonding process with BCB as the adhesive material, ITO as the electrodes, and silicon nitride as the membrane. The performance of the fabricated CMUT element, which had a maximum transparency of 82% in the visible light range, was evaluated in a through-element illumination photoacoustic test. The same device structure was used by Kashani et al [11] to fabricate a transparent CMUT linear array. The fabricated array was used to acquire photoacoustic images of gold wires with the laser passing through the transducer. By employing the optical transparency, Kashani et al [16] used a similar array for combined ultrasonic and optical imaging using a camera on the backside of the CMUT array.
Mechanical flexibility is another useful feature that can be added to ultrasound transducers to extend their usage to shape conformal applications [17–20]. A rigid transducer should be pressed on the curved surface of the tissue to bring all of the elements to contact with the patient body, which results in anatomical changes and patient discomfort. These problems can be addressed by having a flexible transducer that can conform to the body curvatures. A flexible array can also be used in fixed-focus array geometries, tomography arrays, and intravascular imaging in which the transducer can be wrapped around the catheter tips. In recent years, some researchers have investigated the possibility of making flexible CMUTs, which, on the other hand, were not transparent. "Sonic paper" [21] was the first flexible CMUT which was fabricated based on a sacrificial release process with polyethylene terephthalate (PET) as the flexible substrate and SU-8 as the membrane material. The cavities in this polymer-based device were not vacuum-sealed, which excludes immersion applications. Moreover, this device was not transparent. Cheng et al [22] fabricated a non-transparent flexible CMUT using a reverse fabrication method based on sacrificial release process with silicon nitride as the membrane material. The CMUTs had a concave bottom electrode, which, in theory, improves the sensitivity of the device by reducing the parasitic capacitance as the membrane deflection complies with the bottom electrode. However, the fabricated device was not characterized experimentally to show their functionality. A similar CMUT array with concave bottom electrode was fabricated by Shi et al [23]. The device was not only flexible but also stretchable as it was made of PDMS as the structural material and Ga-In-Sn liquid-metal as the electrodes. However, these non-transparent devices had a large cavity depth (around 10-12 $\mu$m) with a very low resonance frequency (200 KHz air), which will not be suitable for imaging purposes. Another non-transparent polymer-based flexible MEMS transducer was introduced by Ge and Cretu [24]. The transducers were fabricated using gray-scale lithography on SU-8 as the structural material. The resonance frequency of the fabricated devices, however, were reported to shift after repeated bending cycles. Also, the cavities in these devices were not vacuum-sealed, limiting their usage in immersion. Zhuang et al [25,26] fabricated a non-transparent flexible non-transparent CMUT array by etching through-wafer trenches in a conventional wafer-bonded CMUT to isolate the elements, and refilling these trenches with PDMS to achieve mechanical flexibility for the whole array. This type of structure can address some of the problems with polymer-based devices. However, since the front side of the array was coated with PDMS in this process, the device had a low-quality factor due to the damping introduced by PDMS on the membranes. In addition, when the array was bent, some cracks formed in the silicon membrane in the regions overlapping with the trenches and propagated to the active device areas. Caronti et al [27] proposed a reverse process to fabricate flexible CMUT arrays encapsulated between an Epoxy backing layer and a Silicone elastomer coating on the top. Since the membrane as well as the sidewalls of the CMUTs were made of silicon nitride, the array was susceptible to crack upon bending. The first flexible row-column addressed CMUT array was fabricated by Chen et al [28] by coating the front side of a conventional row-column addressed array with PDMS followed by etching the silicon substrate in the backside. The proposed devices also have reliability issues as they were reported to develop cracks upon bending the array.
Neither of the previously reported flexible CMUTs discussed above were transparent. The first case of combining the transparency and flexibility in a CMUT element was reported by Pang and Chang [29]. This polymer-based CMUT, which consisted of ITO as the electrodes and SU-8 as the structural material, had a radius of curvature of 40 mm and a transparency of more than 80% in visible light range. This flexible transparent single element CMUT, however, is air-coupled which is not suitable for biomedical applications, requiring acoustic coupling to tissue. Besides, this device was reported to have reliability problems such as ITO electrodes being cracked upon membrane vibrations or SU-8 membrane swelling after fabrication, which got worse after extended use.
Transparent optical-based sensors have also been proposed as another mechanism of ultrasound detection in photoacoustic applications. These include Fabry-Perot interferometers [30], micro-ring resonators [31], and fiber and grating based optical sensors of ultrasound [32]. These devices have very high sensitivity with ultra wide bandwidth compared to conventional ultrasound transducers. However, most of these devices use continuous-wave lasers which can be sensitive to temperature drifts and vibrations [33]. Also, these sensors have been developed on planar substrates and adaption to a flexible or curved transparent array would be highly non-trivial since optical interrogation requires careful optical alignment, and any bending may create stresses that will perturb the optical readout. Moreover, these systems can only be used for receiving ultrasound whereas ultrasound transducers are also able to transmit ultrasound, which enables combined photoacoustic and ultrasound imaging.
3. Methodology
Here we propose an architecture for the first reliable flexible transparent CMUT array for through-illumination photoacoustic tomography. The devices are fabricated on a fused silica substrate based on an adhesive wafer bonding technique with BCB as both an adhesive and sidewall material. The top and bottom electrodes are made of ITO, and the membrane is composed of low-stress silicon nitride. The mechanical flexibility of the array is realized by dicing the array into multiple dies and filling the diced regions with PDMS. By having the dicing cuts only in the vertical direction, which is the case in this work, a cylindrical flexible array is realized. It is also possible to have a spherically flexible array by dicing the wafer in both horizontal and vertical directions, although this will require flexible connections within the PDMS to electrically connect the dies together. The proposed device architecture addresses the reliability problems with previous flexible CMUTs in the literature. Since the CMUT cells are placed on a rigid substrate, the array is robust to formation of any cracks in the structural materials upon bending. The fabricated CMUT array is wire bonded to a flexible printed circuit board (FPCB), and successfully used to acquire photoacoustic images of wire targets suspended in vegetable oil. Due to the relatively large gap distance limited by the spin coating of BCB, the fabricated CMUT arrays require high DC voltages to have an optimal sensitivity, which can be a safety concern for clinical or in-vivo applications. This can be addressed with transparent electrical encapsulation schemes, which are left for future work.
4. Fabrication
The proposed flexible transparent CMUT is fabricated based on an adhesive wafer bonding technique with benzocyclobutene (BCB), which is a photosensitive polymer, as both adhesive and sidewall layers. This fabrication process requires two wafers: one silicon wafer (wafer #1) and one fused silica wafer (wafer #2), which are both cleaned for 20 minutes in Piranha solution (3:1 volume mixture of H$_2$SO$_4$:H$_2$O$_2$) before being processed. The process flow, which is schematically shown in Fig. 1, is described as follows:
Fig. 1. Fabrication process of the transparent flexible CMUT array (a) Fused silica wafer coated with ITO and BCB, (b) Patterning BCB to define CMUT cell cavities, (c) adhesive bonding, (d) Silicon handle removal, (e) Sputtering ITO as the top electrode, (f) Patterning top ITO layer, (g) Accessing the bottom ITO electrode using RIE, (h) Cr/Au lift-off, (i) A close look at the transparent CMUT array, (j) Dicing the wafer along pre-defined directions, (k) filling the backside with PDMS, (l) Transparent flexible CMUT array, (m) 2D cross sectional view of the fabricated CMUT array
- 1. On wafer #1, which is the silicon wafer (Silicon Materials Inc., 100 mm diameter, P/B doped, <100> orientation, 10-20 $\Omega$-cm resistivity, 525 $\mu$m), a silicon nitride layer is deposited using low pressure chemical vapor deposition (LPCVD) to achieve a low stress film. The stress as well as the thickness of this layer are of great importance as it will form the membrane of the CMUT cells.
- 2. The silicon nitride layer deposited on the backside of wafer #1 is etched using RIE (Trion Phantom RIE, 45 sccm of CF$_4$ and 5 sccm of O$_2$, 125 W, 150 mTorr) while protecting the front side with AZ1512 photoresist. After removing the backside silicon nitride, the photoresist on the front side is stripped, and the film stress of the silicon nitride film is measured using KLA-Tencor FLX-2320 to be around 50 MPa tensile stress. The wafer is then cleaned with piranha solution to prepare it for the future steps.
- 3. On wafer #2, which is the fused silica wafer (University Wafer, 100 mm diameter, 500 $\mu$m thickness, DSP), a 250 nm layer of indium tin oxide (ITO) that is a transparent and conductive material is sputtered (75 W, 50 sccm of Ar and 0.5 sccm of O$_2$ at 7 mTorrs) as the bottom electrode.
- 4. Wafer #2 is then cleaned in a 5:1:1 bath of H$_2$O:NH$_4$OH:H$_2$O$_2$, also known as RCA1 solution, at 70$^{\circ }$C for 10 minutes to remove any possible organic contaminants, preparing the surface of the wafer for the next step, which is BCB spin coating.
- 5. To ensure the proper adhesion of the BCB to the bottom electrode, first, a thin layer of AP3000 adhesion promoter (Dow Chemicals) is spun coated on wafer #2 at 3000 rpm for 30 seconds, followed by a soft bake at 150$^{\circ }$C for 60 seconds. This step is also performed on wafer #1 to improve the adhesion in the adhesive bonding step later on in the process (step 8).
- 6. Then, BCB (Cyclotene 4022-25, Dow Chemicals) is spread at 500 rpm for 10 seconds and spun coated at 5000 rpm for 30 seconds on wafer #2 followed by a 90-second soft bake at 60$^{\circ }$ (Fig. 1(a)). The thickness of the BCB layer defines the gap distance in the CMUT cells.
- 7. The BCB layer on wafer #2 is then exposed to UV light through the first photomask in order to define the CMUT cavities. After a post-exposure bake at 50$^\circ$C for 60 seconds, the exposed BCB is developed by drop casting developer DS2100 (Dow Chemicals) on the BCB and waiting for 2 minutes. The developer is then spun off the wafer at 2000 rpm for 30 seconds, followed by drying the wafer with N$_2$ gas (Fig. 1(b)).
- 8. At this point, the adhesive bonding is performed to bond wafer #1 and wafer #2 together using a SUSS ELAN CB6L bonder. First, the wafers are placed onto each other separated by spacers to ensure a vacuum-sealed gap. After loading the wafers, the chamber is pumped down to 0.5 mTorr, followed by removing the spacers, which allows the wafers to come into contact with each other while achieving vacuum sealed cavities. A compressive pressure of 0.5 MPa is then applied over the wafers, during which the wafers are kept at 150$^\circ$C for 15 minutes followed by 250$^\circ$ for an hour. To conclude the bonding process, the wafers are let to cool down below 100$^\circ$ and then removed from the chamber (Fig. 1(c)).
- 9. The next step is to remove the silicon handle, which is done by a combination of dry and wet etching in order to speed up the process. The silicon layer is first thinned down to around 150 $\mu$m using ICPRIE (Alcatel AMS110, 400 sccm of SF$_6$, 2000 W source power, 10 W chuck power). The reminder of the silicon is then wet etched inside a 32% KOH bath at 60$^\circ$, with the silicon nitride layer acting as an effective etch stop (Fig. 1(d)).
- 10. Next, a 200 nm layer of ITO is sputtered over the silicon nitride layer as the top electrode (75 W, 50 sccm of Ar and 0.5 sccm of O$_2$ at 7 mTorrs) (Fig. 1(e)).
- 11. The top ITO layer is then patterned using AZ1512 positive photoresist with the second photomask and wet etched using HCl to define the top electrodes for the CMUT elements (Fig. 1(f)).
- 12. To access the bottom electrode, both silicon nitride and BCB layers need to be etched. Therefore, AZ1529, which is a thicker positive photoresist, is used to pattern the silicon nitride layer using the third photomask. Both silicon nitride and BCB layers are etched in a single RIE process (Trion Phantom RIE, 45 sccm of CF$_4$ and 5 sccm of O$_2$, 125 W, 150 mTorr) to reach the bottom ITO electrode (Fig. 1(g)).
- 13. In order to decrease the resistivity of the top ITO electrodes while preserving the transparency of the device, a 20 nm/200 nm layer of Cr/Au is sputtered (55 sccm of Ar, 7 mTorr. Cr: 300 W, Au: 80 W) and patterned in the form of thin strips using lift-off technique. This is accomplished by using AZ5214 photoresist, which is a positive-tone photoresist to achieve negative side-walls, making it suitable for the lift-off process. The Cr/Au pattern also includes bond pads on both top and bottom electrode for each element in order to facilitate the wire bonding to a printed circuit board (PCB) later on (Fig. 1(h)). At this point, the fabrication of a transparent CMUT array is complete.
- 14. To make the array flexible, further steps need to be performed on the wafer. To this end, using a Disco DAD 3240 dicing tool, the wafer is cut with trenches of 300 $\mu$m in width in predefined horizontal and vertical directions, turning the wafer into islands of separate CMUT dies on a dicing tape with each die containing four CMUT elements (Fig. 1(j)).
- 15. The diced wafer is then temporarily bonded to another handle wafer (not shown in the schematic) from the front side using a two-sided thermal release tape. After removing the dicing tape from the back, a PDMS layer (10:1 weight ratio of elastomer base and curing agent, Sylgard 184 Silicone Elastomer) is spun coated at 350 rpm for 2 minutes on the backside of the wafer followed by curing at 50$^\circ$C for 24 hours. During this step, PDMS flows inside the trenches and mechanically connects the diced dies together, making the whole array flexible. (Fig. 1(k)).
- 16. The substrate is then heated up to above 100$^\circ$C for the tape to release the array off the handle wafer, which concludes the fabrication of the proposed flexible transparent CMUT array (Fig. 1(l)).
The cross sectional view of the fabricated CMUT array and its structural layers are schematically shown in Fig. 1(m). The optical microscope images of the CMUT arrays are also presented in Fig. 2, which are annotated with in-plane dimensions. Table 1 summarizes the geometrical properties of the fabricated devices.
Table 1. Geometrical properties of the fabricated flexible transparent CMUT arrays
5. Characterization
5.1 Capacitance-Voltage Measurements
Before dicing the transparent CMUT array (step (i) in Fig. 1), capacitance-voltage measurements were done randomly on some of the CMUT elements to ensure the functionality of the fabricated devices. To this end, the capacitance of the CMUT elements was measured using a Keithley 4200-Semiconductor Characterization System by increasing the bias voltages from 0 up to 200 V with 0.5 V step size, as shown in Fig. 3. As the bias voltage between the top and bottom electrode of the CMUT element increases, the membranes deflect towards the bottom electrode, increasing the capacitance of the element.
5.2 Transparency
Optical transparency is of great important for this device as the laser needs to pass through the transducer for target illumination in the photoacoustic imaging. Therefore, the transparency of the array was measured from UV to the NIR wavelength range using a spectrophotometer (Perkin-Elmer NIR-UV) with a light beam size of 0.5 cm in width, which covers almost eight CMUT elements. Each of the layers in the CMUT structure including ITO, BCB, silicon nitride, and PDMS was also deposited on separate fused silica substrates to measure their optical transparency. Fig. 4 shows the transparency of the array and its constitutive layers as a function of wavelength. The fabricated CMUT array demonstrated a maximum transmission of 67% at a wavelength of 704 nm. A photograph of the fabricated CMUT array showing its transparency in visible light is presented in Fig. 5(a).
Fig. 4. Transparency measurement on the fabricated flexible transparent CMUT array and its constitutive layers
Fig. 5. Photographs of the fabricated flexible transparent CMUT arrays demonstrating: (a) transparency, (b) flexibility, (c) an array wire bonded to a flexible PCB with a cut-out in the back of the array
5.3 Flexibility
As shown in Fig. 5(b), the flexibility of the fabricated CMUT array is quite high compared to the previously reported flexible transparent CMUT [29] which could only be bent to 40 mm in radius. The current fabricated array can be bent to a radius of curvature of less than 5 mm without the PDMS connections being affected. More importantly, since the CMUT structural layers are sitting on a rigid substrate, all layers will remain intact regardless of how much the array is bent, which is a huge advantage over the other flexible CMUT variants that had reliability issues upon bending [25–27,29]. To assess the reliability of the devices upon bending, strips of flexible transparent CMUT arrays were subjected to over 500 cycles of bending to a radius of 2 cm without the PDMS connections breaking. Fig. 5(c) shows a fabricated flexible transparent CMUT array wire bonded to a flexible PCB, which has a cut-out at the back of the array allowing through-illumination in photoacoustic experiments. The flexibility feature of the array can enable a wide variety of applications ranging from wearable devices to tomography arrays. For tomography arrays, array curvature with large angular coverage has been well documented to achieve improved resolution and reduced artifacts compared to planar arrays [34]. In this work, the flexibility of the array is employed to implement a curved cylindrical transducer array.
5.4 Acoustic Sensitivity
As the next step of the characterization, the acoustic receive tests were done on a single CMUT element, which was wire bonded to a rigid PCB. To this end, an external piezoelectric transducer (OLYMPUS, Panametrics-NDT, V384-SM, 3.5 MHz/0.25") driven by a pulser-receiver (OLYMPUS, Panametric-NDT, Model 5800) was used to generate ultrasound pulses within vegetable oil as the acoustic medium to be received by the CMUT element. First, in a setup shown in Fig. 6(a) in left, the external transducer was calibrated using a needle hydrophone (ONDA HNP-0400) to measure its output pressure. This external transducer was then used in another setup shown in Fig. 6(a) on the right to evaluate the performance of the CMUT element in receive. To be consistent, the piezoelectric transducer was placed at the same distance from the CMUT element as it was from the hydrophone. The received signals by the CMUT element, which was biased using a DC power supply (Stanford Research Systems, Model PS310/1250V-25W), were amplified by 15 dB through an amplifier (OLYMPUS, Panametrics-NDT, Pulser/Receiver, Model 5073PR) and recorded with an oscilloscope (Tektronix, MSO 2014) with 20 MHz bandwidth. Fig. 6(b) depicts the temporal received signal by CMUT elements at 200 V bias voltage as well as its frequency response, which indicates a center frequency of 3.5 MHz in immersion with 80% fractional bandwidth. The resonance frequency of the CMUT elements in air was also measured using an impedance analyzer for different bias voltages as illustrated in Fig. 7, which shows a resonance frequency of around 8 MHz in air. One of the main factors determining the center frequency of the CMUT array is the acoustic medium it is coupled to. The higher the acoustic impedance of the medium, the higher is the damping acting on the CMUT membranes. Therefore, the resonance frequency of the CMUT array is lower in oil compared to air. More damping broadens the bandwidth of the CMUT at the expense of lower detection sensitivity (low quality factor).
Fig. 6. Acoustic Receive Test; (a) Experimental setup for acoustic receive test; Left: Piezoelectric transducer calibration, Right: Receive test on a CMUT element, (b) Received signal by the CMUT element in time and frequency domain at a bias voltage of 200 V, (c) Noise equivalent pressure as a function of applied bias voltage
Having the output pressure of the piezoelectric transducer from the calibration test, the receive sensitivity of the fabricated CMUT elements can be evaluated by calculating the noise equivalent pressure (NEP) factor. This is done by dividing the peak-to-peak output pressure of the piezoelectric transducer to the signal-to-noise ratio (SNR) of the received signals by the CMUT element to the square root of the CMUT bandwidth. The signal-to-noise ratio (SNR) of the received signals is calculated by dividing the peak-to-peak pressure of the received signals to the standard deviation of the noise. Fig. 6(c) shows the noise equivalent pressure (NEP) of the fabricated CMUT elements as a function of DC bias voltage. The NEP reaches 62 mPa/$\sqrt {\mathrm {Hz}}$ at a bias voltage of 275 V. This is comparable to the values reported in the literature, which range from 2.6 to 83 mPa/$\sqrt {\mathrm {Hz}}$ [35–37] and greatly depends on the element size.
5.5 Photoacoustic Tomography
The performance of the fabricated flexible transparent CMUT arrays was evaluated by performing limited-view photoacoustic tomography in through-illumination mode. To this end, as shown in Fig. 5(c), the fabricated CMUT array was wire-bonded to a flexible PCB, which was then mounted inside of a glass beaker of 10 cm in diameter, which is filled with vegetable oil as the acoustic medium. The flexible PCB includes a cut-out behind the array so that the laser can pass through the transducer for illuminating the target. Fig. 8(a) shows the experimental setup for photoacoustic tomography. This setup consisted of three main parts: the laser, Verasonics programmable ultrasound system, and a custom-designed interface board. The target was first illuminated by laser pulses of 532 nm in wavelength with 10 Hz repetition rate (Continuum Surelite). The photoacoustic signals generated by the target were then received by the CMUT array. To be able to acquire the photoacoustic data using the Verasonics system, the custom-made board was used to interface the CMUT array to the Verasonics system using micro-coaxial cables. This interface board has 256 channels with each channel consisting of a bias-tee and a pre-amplifier, as shown in Fig. 8(a). The photoacoustic signals received by the CMUT elements passed through a bias-tee, where DC and AC components were separated, and got amplified by MAXIM 4805A op-amps before being fed to the Verasonics system. The bias voltages of the CMUT cells were supplied by a DC power supply through the bias-tees. Finally, the photoacoustic data acquired by the Verasonics system was beam-formed using a delay and sum method to reconstruct the photoacoustic image of the target. Since a limited view tomography experiment was conducted in this experiment, the envelope detection was performed in only one direction after implementing the back-projection algorithm. However, when a larger angular extent is used for imaging, it may be appropriate to use a multi-view Hilbert transform approach [38,39] to achieve unipolar reconstructions while also ensuring high resolution.
5.5.1 Wire Phantom
To characterize the spatial resolution of the fabricated CMUT arrays, the photoacoustic imaging is performed on an aluminum wire target of 100 $\mu$m in diameter. Fig. 8(b) shows the photoacoustic image of this wire target acquired by 64 CMUT elements at a bias voltage of 150 V and a laser fluence of 18.9 mJ/cm$^2$. The image was obtained by adding the photoacoustic images of the wire target taken at five different locations. The reconstructed image has a signal-to-noise ratio (SNR) of 46 dB and shows no obvious artifacts due to optical absorption of the array in through-illumination. The raw photoacoustic signal acquired through a single CMUT channel with its smoothed frequency response is also given in Fig. 8(c). The large spike at time zero is most probably of an electrical artifact due to large Q-switch voltages since this spike was observed on even elements which had no laser excitation passing through them. Owing to the transparency of the array, very little acoustic signal is expected to be generated from the absorption by the array. As shown in Fig. 9(a), the lateral resolution of the array is obtained by measuring the full width at half maximum (FWHM) of the photoacoustic image in the lateral direction for the target located at the depth of 24 mm. The theoretical in-plane spatial resolution for a full tomography array is half the wavelength, which is 214 $\mu$m for the current devices. With the photoacoustic tomography done with a partial-view array and the wire not being a perfect point target, the obtained spatial resolution is less than the full tomography case. The axial resolution of the CMUT was measured to be 382 $\mu$m, which, as shown in Fig. 9(b), was obtained by taking the full width at half maximum of the envelope curve applied to the photoacoustic signal received by a CMUT element. This is in good agreement with the theoretical axial resolution of 378 $\mu$m, which is calculated based on 0.88$\lambda /BW$ with $\lambda$ and BW being the center wavelength and bandwidth of the transducer, respectively [40,41]. Since there is no acoustic lens on the array, the elevational resolution is limited by far-field diffraction and is estimated as 3 mm at the center of the field of view based on $\lambda R/H$, in which $\lambda$ is the center wavelength, $R$ is the radius of curvature of the array, and $H$ is the element height.
5.5.2 Tissue Phantom
To further evaluate the performance of the fabricated CMUT array in biomedical applications, ex vivo photoacoustic tomography is performed on a chicken breast sample as a tissue mimicking phantom. A tube with an internal diameter of 1.68 mm is inserted into the chicken breast phantom, which is then filled with bovine blood in order to resemble a blood vessel. To perform the photoacoustic imaging, the same setup shown in Fig. 8(a) was used with the wire target replaced with the chicken breast phantom, as shown in Fig. 10(a). The photoacoustic image of the blood vessel acquired by the CMUT array is given in Fig. 10(b), which demonstrates a signal-to-noise ratio of 18 dB.
Fig. 8. Photoacoustic imaging; (a) Experimental setup for photoacoustic imaging, (b) Reconstructed photoacoustic image of wire targets inside vegetable oil acquired by a 64-element flexible transparent CMUT array, (c) Raw photoacoustic signal acquired through a single CMUT channel with its frequency response
Fig. 9. In-plane resolution measurement based on the full width at half maximum of the photoacoustic data obtained from the target located at the depth of 24 mm; (a) Lateral resolution, (b) Axial resolution
Fig. 10. Ex vivo photoacoustic tomography; (a) Chicken breast phantom with an embedded tube filled with bovine blood, (b) photoacoustic image of the blood tube acquired by the fabricated CMUT array
5.6 Discussion and future works
The adhesive BCB layer that also defines the gap height was thicker compared to our previous work [11], which resulted in a slightly lower device sensitivity. This was done in an attempt to avoid dielectric breakdown issues experienced when both high bias voltages and energetic laser pulses were used together in through-illumination photoacoustic applications. Having said that, the damage threshold of the CMUT array in through-illumination will vary greatly depending on not only the laser parameters but also the CMUT architecture and applied bias voltages. This need to be further investigated in future works. In current devices, staying below ANSI limit, which is 20 mJ/cm$^2$ [13], was safe for the CMUTs, however, laser-induced dielectric breakdown was observed when going above these ANSI limits and using bias voltages greater than about 200 V.
The sensitivity of the transducer could be further improved by implementing chip-level amplifiers as close as possible to the CMUTs to boost the received photoacoustic signals and minimize parasitic capacitance from the cables. Current devices exhibited minimal dielectric charging as observed over 20 cycles of C-V testing in the pre-collapse regime. The lack of charging is likely due to the slightly larger gap, which would require higher snap-down voltages. Future works should also investigate the break-down-robust and charging-free architectures with thinner gaps, which could further improve the sensitivity of the device.
The acoustic and photoacoustic experiments were conducted using vegetable oil as a non-conductive acoustic medium, which has acoustic properties similar to water. Vegetable oil was selected rather than water since with the small distance between the CMUT electrodes, water will be locally ionized in high DC voltages, which results in electrical shorting between the top and bottom electrodes. Parylene coating was investigated as a potential solution in this work, however, it was susceptible to burns from laser pulses with intensity above 10 mJ/cm$^2$. In future works, transparent electrical encapsulation methods should be explored to avoid this issue.
The transparency of the device can also be further improved by adding anti-reflection coating or optical matching layers in between the device layers to minimize the reflection of the laser in through-illumination. As the laser passes through the CMUT array with multiple thin films and various geometries, light diffraction can happen. This can be an issue in optical resolution photoacoustic systems such as photoacoustic microscopy in which the laser needs to be focused on the sample. However, in an acoustic resolution photoacoustic imaging system, which is the case in this work, the laser does not need to be focused on the sample as the focusing is done using the ultrasound transducer array.
Maximum benefit of the transparent transducer could eventually be achieved with bias-sensitive two-dimensional arrays enabled by row-column addressing [12]. Such top-orthogonal-to-bottom electrode (TOBE) arrays can achieve readout from every element using strategic bias-encoded acquisition. Fabrication of such arrays could enable wide field, three-dimensional, highly-sensitive photoacoustic imaging. Wearable applications could also be enabled by the device architecture proposed in this work. Such wearable devices could enable longitudinal monitoring of venous blood oxygenation, and other oxygenation monitoring tasks using multi-wavelength photoacoustic imaging.
6. Conclusion
This work presents the fabrication and characterization of the first flexible transparent CMUT arrays for through-illumination photoacoustic tomography systems. The proposed device is fabricated based on an adhesive wafer bonding technique with BCB as the adhesive, ITO as the electrodes, silicon nitride as the membrane, and PDMS as the flexible backing layer, resulting in a transparency of 67% for the array in the visible range. The flexibility of the CMUT array, which can reach a radius-of-curvature of less than 5 mm, is achieved by connecting diced CMUT dies with PDMS. Therefore, the structural layers remain intact regardless of the bending curvature, making the array very reliable upon bending. The receive acoustic tests done on a CMUT element indicated a center frequency of 3.5 MHz with 80% fractional bandwidth, and a noise equivalent pressure of 62 mPa/$\sqrt {\mathrm {Hz}}$ at 275 V bias voltage. Wire bonded to a flexible PCB, a 64-element array is successfully used to acquire through-illumination photoacoustic images of a 100 $\mu$m wire target, showing an SNR of 46 dB with a lateral and axial resolution of 293 $\mu$m and 382 $\mu$m, respectively. Ex vivo photoacoustic tomography was also performed on a chicken breast phantom using the fabricated array in through-illumination. The proposed device architecture has a great potential to be used in a variety of photoacoustic applications such as photoacoustic tomography or wearable devices.
Funding
Natural Sciences and Engineering Research Council of Canada (RGPIN-2018-05788); Alberta Innovates (AICEC 202102269).
Acknowledgments
The authors are grateful for resources supplied by CMC Microsystems Canada. The authors would also like to thank nanoFAB facility staff at the University of Alberta for their support and advice.
Disclosures
The authors had no conflict of interests in this work. Roger Zemp is a founder and shareholder of illumiSonics Inc. and CliniSonix Inc., which, however, did not support this work.
Data availability
Data underlying the results presented in this paper are not publicly available at this time but may be obtained from the authors upon reasonable request.
References
1. M. Xu and L. V. Wang, “Photoacoustic imaging in biomedicine,” Rev. Sci. Instrum. 77(4), 041101 (2006). [CrossRef]
2. Z. Li, A. K. Ilkhechi, and R. Zemp, “Transparent capacitive micromachined ultrasonic transducers (cmuts) for photoacoustic applications,” Opt. Express 27(9), 13204–13218 (2019). [CrossRef]
3. C. Fang, H. Hu, and J. Zou, “A focused optically transparent pvdf transducer for photoacoustic microscopy,” IEEE Sensors J. 20(5), 2313–2319 (2020). [CrossRef]
4. C. Ceroici, K. Latham, R. Chee, B. Greenlay, Q. Barber, J. A. Brown, and R. Zemp, “3d photoacoustic imaging using hadamard-bias encoding with a crossed electrode relaxor array,” Opt. Lett. 43(14), 3425–3428 (2018). [CrossRef]
5. M. Haller and B. Khuri-Yakub, “A surface micromachined electrostatic ultrasonic air transducer,” in 1994 Proceedings of IEEE Ultrasonics Symposium, vol. 2 (1994), pp. 1241–1244.
6. H. Soh, I. Ladabaum, A. Atalar, C. Quate, and B. Khuri-Yakub, “Silicon micromachined ultrasonic immersion transducers,” Appl. Phys. Lett. 69(24), 3674–3676 (1996). [CrossRef]
7. D. M. Mills and L. S. Smith, “Real-time in-vivo imaging with capacitive micromachined ultrasound transducer (cmut) linear arrays,” in IEEE Symposium on Ultrasonics, 2003, vol. 1 (IEEE, 2003), pp. 568–571.
8. I. Wygant, “A comparison of cmuts and piezoelectric transducer elements for 2d medical imaging based on conventional simulation models,” in 2011 IEEE International Ultrasonics Symposium, (IEEE, 2011), pp. 100–103.
9. M. Engholm, H. Bouzari, T. L. Christiansen, C. Beers, J. P. Bagge, L. N. Moesner, S. E. Diederichsen, M. B. Stuart, J. A. Jensen, and E. V. Thomsen, “Probe development of cmut and pzt row–column-addressed 2-d arrays,” Sens. Actuators, A 273, 121–133 (2018). [CrossRef]
10. T. Takezaki, M. Kawano, S. Machida, and D. Ryuzaki, “Through-transmission scanning acoustic tomography using capacitive micromachined ultrasound transducer,” in 2018 IEEE International Symposium on the Physical and Failure Analysis of Integrated Circuits (IPFA), (IEEE, 2018), pp. 1–4.
11. A. K. Ilkhechi, C. Ceroici, Z. Li, and R. Zemp, “Transparent capacitive micromachined ultrasonic transducer (cmut) arrays for real-time photoacoustic applications,” Opt. Express 28(9), 13750–13760 (2020). [CrossRef]
12. R. K. Chee, A. Sampaleanu, D. Rishi, and R. J. Zemp, “Top orthogonal to bottom electrode (tobe) 2-d cmut arrays for 3-d photoacoustic imaging,” IEEE Trans. Ultrason., Ferroelect., Freq. Contr. 61(8), 1393–1395 (2014). [CrossRef]
13. J. Chen, M. Wang, J.-C. Cheng, Y.-H. Wang, P.-C. Li, and X. Cheng, “A photoacoustic imager with light illumination through an infrared-transparent silicon cmut array,” IEEE Trans. Ultrason., Ferroelect., Freq. Contr. 59(4), 766–775 (2012). [CrossRef]
14. X. Zhang, X. Wu, O. J. Adelegan, F. Y. Yamaner, and Ö. Oralkan, “Backward-mode photoacoustic imaging using illumination through a cmut with improved transparency,” IEEE Trans. Ultrason., Ferroelect., Freq. Contr. 65(1), 85–94 (2018). [CrossRef]
15. X. Zhang, O. Adelegan, F. Y. Yamaner, and Ö. Oralkan, “An optically transparent capacitive micromachined ultrasonic transducer (cmut) fabricated using su-8 or bcb adhesive wafer bonding,” in 2017 IEEE International Ultrasonics Symposium (IUS), (2017), pp. 1–4.
16. A. K. Ilkhechi, C. Ceroici, E. Dew, and R. Zemp, “Transparent capacitive micromachined ultrasound transducer lineararrays for combined realtime optical and ultrasonic imaging,” Opt. Lett. 46(7), 1542–1545 (2021). [CrossRef]
17. J. Chang, Z. Chen, Y. Huang, Y. Li, X. Zeng, and C. Lu, “Flexible ultrasonic array for breast-cancer diagnosis based on a self-shape–estimation algorithm,” Ultrasonics 108, 106199 (2020). [CrossRef]
18. Y. Yang, H. Tian, B. Yan, H. Sun, C. Wu, Y. Shu, L.-G. Wang, and T.-L. Ren, “A flexible piezoelectric micromachined ultrasound transducer,” RSC Adv. 3(47), 24900–24905 (2013). [CrossRef]
19. T. F. de Oliveira, C. N. Pai, M. Y. Matuda, J. C. Adamowski, and F. Buiochi, “Development of a 2.25 mhz flexible array ultrasonic transducer,” Res. Biomed. Eng. 35(1), 27–37 (2019). [CrossRef]
20. H. Hu, X. Zhu, C. Wang, L. Zhang, X. Li, S. Lee, Z. Huang, R. Chen, Z. Chen, C. Wang, Y. Gu, Y. Chen, Y. Lei, T. Zhang, N. Kim, Y. Guo, Y. Teng, W. Zhou, Y. Li, A. Nomoto, S. Sternini, Q. Zhou, M. Pharr, F. L. Di Scalea, and S. Xu, “Stretchable ultrasonic transducer arrays for three-dimensional imaging on complex surfaces,” Sci. Adv. 4(3), eaar3979 (2018). [CrossRef]
21. M.-W. Chang, H.-C. Deng, D.-C. Pang, and M.-Y. Chen, “6f-6 a novel method for fabricating sonic paper,” in 2007 IEEE Ultrasonics Symposium Proceedings, (2007), pp. 527–530.
22. C.-H. Cheng, C. Chao, X. Shi, and W. W. F. Leung, “A flexible capacitive micromachined ultrasonic transducer (cmut) array with increased effective capacitance from concave bottom electrodes for ultrasonic imaging applications,” in 2009 IEEE International Ultrasonics Symposium, (2009), pp. 996–999.
23. X. Shi, C.-H. Cheng, and J. Peng, “A novel stretchable cmut array using liquid-metal electrodes on a pdms substrate,” in 2012 7th IEEE International Conference on Nano/Micro Engineered and Molecular Systems (NEMS), (2012), pp. 104–107.
24. C. Ge and E. Cretu, “A sacrificial-layer-free fabrication technology for mems transducer on flexible substrate,” Sens. Actuators, A 286, 202–210 (2019). [CrossRef]
25. X. Zhuang, D.-S. Lin, Ö. Oralkan, and B. T. Khuri-Yakub, “Flexible transducer arrays with through-wafer electrical interconnects based on trench refilling with pdms,” in 2007 IEEE 20th International Conference on Micro Electro Mechanical Systems (MEMS), (2007), pp. 73–76.
26. X. Zhuang, D.-S. Lin, Ö. Oralkan, and B. T. Khuri-Yakub, “Fabrication of flexible transducer arrays with through-wafer electrical interconnects based on trench refilling with pdms,” J. Microelectromech. Syst. 17(2), 446–452 (2008). [CrossRef]
27. A. Caronti, A. Coppa, A. Savoia, C. Longo, P. Gatta, B. Mauti, A. Corbo, B. Calabrese, G. Bollino, A. Paz, G. Caliano, and M. Pappalardo, “Curvilinear capacitive micromachined ultrasonic transducer (cmut) array fabricated using a reverse process,” 2008 IEEE Ultrasonics Symposium, (2008), pp. 2092–2095.
28. A. I.-H. Chen, L. L. P. Wong, S. Na, Z. Li, M. Macecek, and J. T. W. Yeow, “Fabrication of a curved row-column addressed capacitive micromachined ultrasonic transducer array,” J. Microelectromech. Syst. 25(4), 675–682 (2016). [CrossRef]
29. D.-C. Pang and C.-M. Chang, “Development of a novel transparent flexible capacitive micromachined ultrasonic transducer,” Sensors 17(6), 1443 (2017). [CrossRef]
30. S. V. Thathachary and S. Ashkenazi, “Toward a highly sensitive polymer waveguide fiber fabry–pérot ultrasound detector,” J. Biomed. Opt. 23(10), 1 (2018). [CrossRef]
31. H. Li, B. Dong, Z. Zhang, H. F. Zhang, and C. Sun, “A transparent broadband ultrasonic detector based on an optical micro-ring resonator for photoacoustic microscopy,” Sci. Rep. 4(1), 4496 (2015). [CrossRef]
32. R. Shnaiderman, G. Wissmeyer, O. Ülgen, Q. Mustafa, A. Chmyrov, and V. Ntziachristos, “A submicrometre silicon-on-insulator resonator for ultrasound detection,” Nature 585(7825), 372–378 (2020). [CrossRef]
33. D. Ren, Y. Sun, J. Shi, and R. Chen, “A review of transparent sensors for photoacoustic imaging applications,” in Photonics, vol. 8 (Multidisciplinary Digital Publishing Institute, 2021), p. 324.
34. A. B. E. Attia, G. Balasundaram, M. Moothanchery, U. Dinish, R. Bi, V. Ntziachristos, and M. Olivo, “A review of clinical photoacoustic imaging: Current and future trends,” Photoacoustics 16, 100144 (2019). [CrossRef]
35. A. Sampaleanu, P. Zhang, A. Kshirsagar, W. Moussa, and R. J. Zemp, “Top-orthogonal-to-bottom-electrode (tobe) cmut arrays for 3-d ultrasound imaging,” IEEE Trans. Ultrason., Ferroelect., Freq. Contr. 61(2), 266–276 (2014). [CrossRef]
36. S. Vaithilingam, T.-J. Ma, Y. Furukawa, I. Wygant, X. Zhuang, A. De La Zerda, Ö. Oralkan, A. Kamaya, S. s. Gambhir, R. B. Jeffrey, and B. T. Khuri-yakub, “Three-dimensional photoacoustic imaging using a two-dimensional cmut array,” IEEE Trans. Ultrason., Ferroelect., Freq. Contr. 56(11), 2411–2419 (2009). [CrossRef]
37. G. Gurun, P. Hasler, and F. L. Degertekin, “Front-end receiver electronics for high-frequency monolithic cmut-on-cmos imaging arrays,” IEEE Trans. Ultrason., Ferroelect., Freq. Contr. 58(8), 1658–1668 (2011). [CrossRef]
38. G. Li, L. Li, L. Zhu, J. Xia, and L. V. Wang, “Multiview hilbert transformation for full-view photoacoustic computed tomography using a linear array,” J. Biomed. Opt. 20(6), 1 (2015). [CrossRef]
39. L. Li, L. Zhu, Y. Shen, and L. V. Wang, “Multiview hilbert transformation in full-ring transducer array-based photoacoustic computed tomography,” J. Biomed. Opt. 22(7), 076017 (2017). [CrossRef]
40. J. Yao and L. V. Wang, “Photoacoustic microscopy,” Laser Photonics Rev. 7(5), 758–778 (2013). [CrossRef]
41. M. R. Sobhani, K. Latham, J. Brown, and R. J. Zemp, “Bias-sensitive transparent single-element ultrasound transducers using hot-pressed pmn-pt,” OSA Continuum 4(10), 2606–2614 (2021). [CrossRef]
