A Micromachined Pb(Mg_1/3Nb_2/3)O₃-PbTiO₃ Single Crystal Composite Circular Array for Intravascular Ultrasound Imaging

Abstract

This paper describes the design, fabrication, and characterization of a micromachined high-frequency Pb(Mg_1/3Nb_2/3)O₃-PbTiO₃ (PMN-PT) single crystal/epoxy 1–3 composite ultrasound circular array. The 1–3 composites were fabricated by deep reactive ion etching (DRIE) of PMN-PT single crystal. The feature size of single crystal pillars was 18 μm in diameter. The kerf between pillars was less than 4 μm. A 50-element circular array transducer (radially outward) with the pitch of 100 μm was wrapped around a needle resulting in an outer diameter of 1.7 mm. The array test showed that the center frequency reached 39±2 MHz and −6-dB fractional bandwidth was 82±6%. The insertion loss was −41 dB, and crosstalk between adjacent elements was −24 dB. A radial outward imaging testing with phantom wires (D = 50 μm) was conducted. The image was in a dynamic range of 30 dB to show a penetration depth of 6 mm by using the synthetic aperture method. The −6-dB beam width was estimated to be 60 μm in the axial direction at 3.1 mm distance away from the probe. The results suggest that the 40 MHz micromachined 1–3 composite circular array is promising for intravascular ultrasound (IVUS) imaging applications.

Introduction

Atherosclerotic cardiovascular disease is considered one of the worldwide leading causes of death, and the disease often manifests without warning [1]. Based on the update of Heart Disease and Stroke Statistics in 2013, there are more than 2200 American deaths every day, which means one death in every 40 s [2]. For 75% of acute coronary syndromes, it was reported that the underlying pathological mechanism is postulated to be the atherosclerotic plaque rupture. The atherosclerotic plaque, a thin layer of atheroma and a large necrosis core underneath, possesses the morphological features, which are acoustically visible [3]. These features are frequently used as the indicators of coronary artery disease and potential risks. Therefore, the evaluation and characterization of the atherosclerotic plaques are one of the most active areas in the research of cardiology and biomedical imaging [4].

High-frequency transducers have been attractive due to the benefit of high resolution [5,6] in the intravascular ultrasound (IVUS) imaging [7], ophthalmology [5,8], dermatology [9], and small animal studies [10]. Within last two decades, IVUS imaging has evolved from an experimental technique to a clinical standard [11]. The extent and severity of disease are frequently reflected by the echoes in ultrasound imaging. With the IVUS, lipid-laden lesions appear hypoechoic; fibromuscular lesions generate low intensity echoes, and fibrous or calcified tissues are echogenic. Therefore, the IVUS image relies on the ultrasound transducers performance, especially the sensitivity and resolution.

Moreover, to overcome the constraints of the fixed depth-of-field and frame rate of single-element transducer-based IVUS imaging system, IVUS arrays are desirable for their beam steering, subaperture firing, and higher frame rates. In 1997, a circular array of 20 MHz was developed by Volcano (Volcano Corporation, San Diego, CA) for IVUS imaging, which was at the upper limit of conventional frequencies [12]. The 3.5 French 20 MHz imaging catheters consist of 64 elements with ∼50 μm pitch, integrated with five application-specific integrated circuit (ASIC) chips. In 2006, the capacitive micromachined ultrasonic transducers (CMUT) arrays were also considered for IVUS, with a forward-looking ring array for the volumetric intravascular image [13]. Recently, much progress has been made in enhancing IVUS image quality [14], mainly with the increase in frequency. The attenuation of an ultrasonic wave propagation, on the other hand, increases with the frequency, and hence, the decay of propagating acoustic energy is a concern for the required imaging depth. At present, the 30–50 MHz ultrasound systems are targeted on the applications with imaging depth in the range of 3–6 mm [5].

To enhance the high-frequency imaging performance, high sensitivity and broad bandwidth transducers are required [15]. The performance of transducers mainly relies on the properties of the piezoelectric materials, including electromechanical coupling coefficient (k), dielectric permittivity (ε), acoustic impedance (Z), etc. Relaxor-PT-based ferroelectric single crystal PMN-PT offers high coupling coefficients; thus, it has been broadly studied for the transducer applications [16]. In recent years, PMN-PT single crystal 1–3 composite has been increasingly applied in medical imaging transducers for its low acoustic impedance and relatively high electromechanical coupling coefficient (k_t) [17,18]. Typically, 1–3 composites are an array of periodic pillars of piezoelectric material. To avoid the lateral vibration near the designed center frequency, the sizes of the pillar and kerf need to be carefully designed [19,20]. For fabrication of 1–3 composites, “dice-and-fill” method has been widely adopted in the industry. However, due to the limitation of the blade thickness (usually higher than 10 μm), this method is not appropriate for fabrication of high frequency (> 25 MHz) 1–3 composites [21]. Other alternative methods include the interdigital bonding technique, stacked plates or lamination techniques, fiber processing, tape casting, and laser machining [22–24]. In recent years, micromachining technique was successfully developed for PMN-PT 1–3 composite fabrication through a frequency range from 25 to 75 MHz [25–27]. Key processes involved in this micromachining method include photolithography, electroplating, deep reactive ion etching (DRIE), epoxy kerf filling, and plate lapping. Based on this technique, single element and array ultrasound transducers were demonstrated for medical imaging and nondestructive testing applications [26–28]. More recently, pioneer works have been presented on linear arrays and annular arrays at high frequency. In 2012, Liu et al. presented an annular array using micromachined PMN-PT 1–3 composites for medical imaging, and in 2016, Cummins et al. took advantage of the micromachining technique to fabricate a 60 MHz PMN-PT 2–2 composites transducer array for ultrasound-guided breast biopsy [26,29]. To date, a high-frequency 1–3 composite (> 20 MHz) circular array has not yet been developed for IVUS imaging with enhanced performance.

In this paper, a 50-element 40 MHz micromachined PMN-PT single crystal 1–3 composite circular array was designed, fabricated, and characterized for IVUS applications. A radial outward imaging with phantom wires was conducted to show a gray image by using the synthetic aperture method.

Materials and Methods

Material Design and Fabrication.

The PMN-PT single crystal (CTS Corporations, Bolingbrook, IL) and EPO-TEK301 (Epoxy Technology, Inc., Billerica, MA) were selected as active material and epoxy filler for 1–3 composites, respectively. The piezoelectric pillar was designed in a cylindrical post (D = 17.5 μm) with kerf width of 3.5 μm. With the thickness of 26 μm, the composite layer was aimed to resonate at 40 MHz. The comsol program (COMSOL, Inc., Burlington, MA) was used to simulate the resonance of the material and compared with experimental results. As shown in Fig. 1, a piece of 1–3 composites was modeled in the simulation. The four-sided wall was set as symmetric boundary condition, and the top and bottom were set free–free boundary condition. The parameters in the simulation were summarized in Table 1 [30]. The simulation reached numerical convergence at the end of computational process.

Fig. 1.

A three-dimensional model of 1–3 composite in comsol simulation

Table 1.

Key parameters of PMN-PT and EPO-TEK 301

PMN-PT
Elastic property	$c_{11}^E$	11.5 × 1010 N/m2	$c_{33}^E$	10.3 × 1010 N/m2
	$c_{12}^E$	10.3 × 1010 N/m2	$c_{13}^E$	10.2 × 1010 N/m2
$c_{44}^E$	6.9 × 1010 N/m2	${\rho }^C$	8060 kg/m3
Dielectric property	${\epsilon }_{33}^S$	680 ${\epsilon }_0$	${\epsilon }_{11}^S$	1434 ${\epsilon }_0$
Piezoelectric property	$e_{33}$	20.3 C/m2	$e_{15}$	10.1 C/m2
$e_{31}$	−3.9 C/m2
EPO-TEK 301
Elastic property	$c_{11}^E$	8.1 × 109 N/m2	$c_{12}^E$	8.1 × 109 N/m2
${\rho }_{11}^p$	1500 kg/m3
Dielectric property	$\epsilon$	4 ${\epsilon }_0$

The fabrication procedure of micromachined 1–3 composites is shown in Fig. 2 [31]. Initially, Cr/Au-coated PMN-PT single crystal plate was patterned with photoresist KPMR (MiicroChem, Westborough, MA) by photolithography as a lattice template for subsequent hard mask electroplating. Thick hard mask (nickel) was then electroplated through the circular or square shape photoresist template [32,33]. The photoresist was then cleaned by an oxygen plasma etching process (March PM-600 Plasma Asher; Nordson March, Concord, CA). With the pattern nickel mask left on the crystal plate, DRIE was performed to etch the crystal where the part was not covered by Ni, forming crystal pillars and kerfs [34,35]. The epoxy resin (EPO-TEK 301; Epoxy Technology, Inc., Billerica, MA) was then filled into the etched kerfs of the crystal plate. After the epoxy cured, the plate was lapped to the desire thickness. The composite was deposited with Cr/Au as electrodes on the both sides. The composite was then poled with a direct current electric field of approximately 10 kV/cm for 10 min at room temperature. Composite's material capacitance and dielectric loss were measured with an impedance analyzer (4294A; Agilent Tech. Inc., Santa Clara, CA), and the dielectric permittivity was estimated. The electrical impedance and phase spectrum were also measured using the same system. Based on the test results, electromechanical coupling coefficient k_t and acoustic impedance Z were calculated according to the IEEE standards [36].

Fig. 2.

The schematic of process for micromachined 1–3 composites

Array Ultrasound Design, Fabrication, and Characterization.

As reported in our previous work [37], the design parameters of an individual element in an array were simulated using the KLM (Krimholtz, Leedom, and Mattaei) model [38]. In the stack design, the 1–3 composites were bonded with the conductive backing material E-Solder 3022 (Von Roll Isola), and passivation of Parylene (Specialty Coating Systems, Inc., Amherst, NH) with 10 μm thickness served as the matching layer.

For the array design, however, there is a basic tradeoff for array systems between the number of connections (i.e., wires and testing system) from the transducers and degrees of the beam forming performance. Because of the lack of high-frequency imaging system for array ultrasound, each element was connected with a single-wire electric instrument for the imaging setup. To console the workload, the element number of the array was set to 50. Thus, the pitch element size was sacrificed (approximately a 100-μm, or 2.5-λ at 40 MHz, element pitch). Based on such design, the 50 elements are evenly distributed over the circumference of a 1.3-mm-diameter needle (Fig. 3). The reasonable images can be obtained with this system by using a customized synthetic aperture principle, which was addressed later in the manuscript. The details of the array design were presented in our previous conference report [37]. The flex circuit was adopted for the interconnections and the element control. The width (after the dicing-cut) of the single element was ∼80 μm. After the element array had been wrapped around a 1.3 mm diameter needle, the diameter of the outer edge of the array was around 1.7 mm. The design parameters were summarized in Table 2.

Fig. 3.

The schematic view of circular array structure

Table 2.

Array design parameters and material properties

Specifications	Thickness	Values
Array aperture	—	1.5 mm by 5 mm
Element number	—	50
Single element size	—	1.5 mm by 80 μm
Pitch size	—	100 μm
Active layer (PMNT 1–3 composite)	26 μm	Impedance (MRayl)	18
		Longitudinal velocity (m/s)	3150
		5600
Density (kg/m3)
Matching layer (Parylene)	10 μm	Impedance (MRayl)	3.16
		Longitudinal velocity (m/s)	2770
Density (kg/m3)	1140
Backing layer (E-solder)	220 μm (∼4 λ)	Impedance (MRayl)	5.5
		Longitudinal velocity (m/s)	2110
		Density (kg/m3)	2590

In the fabrication, the conductive backing material, E-solder 3022, was cast onto a free-standing composite material, which was lapped to the final thickness (26 μm). E-solder 3022 was then cured for 24 h at room temperature. The acoustic backing layer was lapped down to approximately 200 micron-meter, which resistance rate was measured to be 1.2 Ohm/mm. The piezo and backing stack was performed dicing process to form the pitch element (ZH05 series, DISCO Corporation, Japan), and the backing was partly undercut to keep the stack as a whole piece. The element array was then aligned and bonded with the flex-circuit using epoxy (EPO-TEK 301; Epoxy Technology, Inc., Billerica, MA). The flexible circuits consisted of 5-μm thick Au traces and a 25-μm thick polyimide surface, which was designed to minimize the effect on the high-frequency wave propagation. The array was then coated with a layer of 10 μm Parylene as matching and passivation layer. Finally, the array was wrapped carefully, onto an American wire gauge (AWG) 16 needle (diameter = 1.26 mm). To avoid any unexpected movement or detachment, the EPO-TEK 301 was cast into the device; when the epoxy was cured, the device was fixed on the needle.

The pulse-echo tests for individual element were performed using a pulser-receiver (5900PR; Panametrics, Inc., Waltham, MA). A pulse excitation (1-μJ) was applied to the transducer under P/E mode. The frequency response of the transducer was analyzed from the echo waveform, the center frequency and −6 dB fractional bandwidth were determined. Insertion loss was measured by exciting a representative array element with a 40-MHz (f_c) burst and receiving the reflected echo from a polished steel reflector placed at the elevation focus (d). The received signal (V_R) across a 50-Ω load was referenced to the source signal (V_T) delivered to a 50-Ω reference load and expressed in decibels. The insertion loss was estimated using following formula [39]:

$$\text{IL}=20\hspace{0.17em}\text{log}\hspace{0.17em}\frac{V_R}{V_T}+0.6+2.2\times {10}^{-4}\cdot 2d\cdot f_c^2$$ (1)

The crosstalk between adjacent elements was also measured. The electrical and acoustical separation between elements was determined by the crosstalk test. A function generator (Tektronix AFG3101, Beaverton, OR) was used to excite a reference element with burst signal. The responses from the first and the second adjacent elements were measured to compare with the reference value from the excited element.

Phantom Wire Imaging.

The array was then used to image the 50-μm-diameter steel wires phantom. The two wire targets positioned in a gelatin phantom are about 2.5 mm and 5 mm away from the array, respectively (Fig. 4). Given by operating frequency of 40 MHz, there was no high-frequency beam former available to produce real-time imaging. To perform imaging testing with this array, a customized synthetic aperture method was adopted for the image reconstruction. Because of the circular geometry, and the finite acceptance angle (an angular span of ∼40 deg), only a limited part of the array can be used for beam forming along a given direction. Based on calculated beam pattern and the element pitch size, three adjacent elements from both sides of center transmit element were considered sensitive to the scattered echoes [31,40].

Fig. 4.

The relative positions in the wire imaging test setup

By manually connecting each of the 50 array elements to a pulser–receiver, individual time-domain responses were acquired, with the same settings employed for the pulse-echo test. While one element was excited, the element itself and the adjacent 3 adjacent elements from both sides (7 receiving elements in total) were successively connected with the system to record the RF receiving data. The amplitude of each A-line RF signal was compensated based on the pulse-echo results at a different axial distance. Since the pitch size was relatively large, and limited number of receiving element, the lateral resolution was expected to be low with respect to the working frequency based on such synthetic aperture algorithm.

Results and Discussion

Material Characterization.

A PMN-PT crystal 1–3 composite was fabricated using DRIE technique by H.C. Materials (CTS Corporation, Bolingbrook, IL) [41]. The diameter of PMN-PT post was ∼18 μm, and kerf in between was ∼3 μm. The electrode composites showed the height of 26 μm. Based on the geometry of the composites unit cell and material selection, a finite element simulation was conducted to estimate the composites property, which was also used to compare the test results. Top view and side view of the micromachined composite with the backing are shown in Fig. 5.

Fig. 5.

PMN-PT 1–3 composite with conductive backing (a) top view and (b) side view

Electrical impedance and phase spectra of the composite material (0.4 mm by 0.3 mm) were measured at frequencies from 15 to 85 MHz (Fig. 6). The series and parallel resonant frequencies were 39 MHz and 59 MHz, respectively. Compared with the simulation result, it shows a good match on series resonance (∼40 MHz), but less accordant on parallel resonance frequency (59 MHz in the test result and 79 MHz in simulation result). The gap of two results may cause by that the dielectric loss and coupling loss were neglected in the simulation, which the ideal configuration also affect other values, such as the magnitude of the impedance at resonance and peak phase of the resonance. In the testing, some other factors which impact the performance cannot be neglected.

Fig. 6.

Impedance and phase of the PMN-PT 1–3 composite: (a) simulation results using comsol and (b) test results

• (1)The wire connection would introduce dielectric loss.
• (2)The polling process may not fully exploit the piezoelectric performance.
• (3)The instrument may not be calibrated to perfect condition.

The electromechanical coupling coefficient k_t was calculated to be 0.76. Based on the volume fraction of 65%, the density was determined to be 5631 kg/m³. The longitudinal velocity was estimated to be 3150 m/s. The relative permittivity (ε_r) was calculated to be 1851 at 1 kHz, with the dielectric loss of 48 mU. With 1–3 composite model, other material parameters were estimated and summarized in Table 3 [42].

Table 3.

Material properties of 1–3 composite

Parameters	Value
Volume fraction	65%
$\rho$ (kg/m3)	5631
εr	1851
$V_L$ (m/s)	3150
$Z_L$ (MRayl)	18
kt	0.76
Thickness (μm)	26

Array Characterization.

The photograph picture of the prototype circular array is shown in Fig. 7. The individual element test results for the completed array are summarized in Table 4. The mean (± standard deviation) capacitance of a single element was 72 ± 12 pF, dielectric loss was 108 ± 19 mU, and center frequency was 38.7±3 MHz with an average −6 dB fractional bandwidth of 82 ± 10%. A full result of all element performance was summarized in Fig. 8. Figure 9 shows the typical time response of pulse-echo test and corresponding frequency spectra. The measured −20 dB pulse length for a typical element is 108 ns for the array. Both of these measurements were consistent with the initial design of the array.

Fig. 7.

A photograph of the circular array and the aperture of the element array under the microscope (the zoom-in figure)

Table 4.

Measured properties for the circular array

Property	Value
Number	50
Average frequency	38.7 ± 3 MHz
Average bandwidth	82 ± 10%
P/E sensitivity (at 2 mm)	100 mV
Pulse length (−20 dB)	∼110 ns
Electrical impedance (40 MHz)	180 ± 40 Ohm

Fig. 8.

The test result of the circular array: (a) dielectric capacitance and loss values for each element in array and (b) center frequency and bandwidth values for each element in array

Fig. 9.

Measured pulse-echo response and its fast Fourier transform spectra for a representative element in the array

In the insertion loss characterization, the corrections for diffraction in the azimuth direction and attenuation in water were used to obtain an estimation of –41 dB (Fig. 10), which is relatively low compared to other high-frequency arrays [43]. The small array size and unneglectable dielectric loss were the potential causes for the relatively low insertion loss. Improving both the electrical interconnect method and transmission line fabrication process in the next generation of this array will likely help improve sensitivity and insertion loss.

Fig. 10.

Measured insertion loss

The crosstalk measurements indicated acceptable but not ideal element-to-element separation (Fig. 11). The value near the center frequency of the array was measured to be −24 dB and −28 dB between the first and second adjacent elements. Two likely possibilities contributed to this higher-than-desired crosstalk. First, the backing layer is not cut through on which the elements are interconnected. Thus, the acoustic surface and bulk wave may travel within the backing layer underneath, resulting in acoustic or mechanical cross-coupling between elements. The steel needle serves as the substrate also has a contribution to the crosstalk as well. The second source is the crosstalk through the water-coupled narrow kerfs between elements (20–25 μm).

Fig. 11.

Measured cross-talk of the array

Imaging Results.

An image of steel-wires phantom in the water tank was reconstructed by synthetic aperture method (Fig. 12, in a dynamic range of 30 dB). Generally, the wire position from ultrasound image (Fig. 12) shows good consistency with wire position in phantom (schematic in Fig. 4). The misalignment between image and real position is on micron meter level. The error may come from the variation of sound speed in the phantom. In the image reconstruction, the sound speed was simply assumed constant at 1540 m/s. Based on the plot of a single line (Fig. 13), it exhibits full-width half-maximum resolutions of 60 μm in axial and ∼600 μm in lateral. While the axial resolution is an improvement over previously reported high-frequency linear arrays, the lateral resolution is lower than desired value, which may be improved by a finer element pitch and higher element number. In the future work, the probe would be connected to a multiplexer and a high-frequency imaging system to perform the multi-element firing and phase-angle steering configuration to improve the lateral resolution.

Fig. 12.

An image of steel wires reconstructed by synthetic aperture method, with the dynamic range of 30 dB

Fig. 13.

The beam profile of single line at 3 mm in axial direction (a) and circumferential direction (b)

Conclusion

This paper reported the development and characterization of a 40-MHz circular array using micromachined 1–3 composite elements. The composite material demonstrated improved performance compared with single crystal counterpart. The k_t was 0.76, and acoustic impedance was estimated as 18 MRayl. Initial characterization results of the array showed a center frequency near 39 MHz, an 82% bandwidth, and a −20 dB pulse length of ∼110 ns. The primary results showed potential as a good candidate for receiving transducer array to the vasa vasorum assessment. The technique, as known as acoustic angiography, utilizes the ultrabroadband responses of contrast agents, exciting them near 2–5 MHz range, and receiving harmonics at >20 MHz [44,45]. This provides a very high signal to noise, high-resolution contrast imaging, which enable us to acquire the exceptional sensitivity to microvessel structural mapping. The image of steel wires target was acquired using a customized synthetic aperture principles reconstruction. The results from this imaging test demonstrate beam width exceeding 60 μm axially and 600 μm laterally. Those results indicate that this prototyped array would be suitable for clinical IVUS applications.

However, there are several significant limitations of this study. First, according to the operating frequency of 40 MHz, the element pitch of 100 μm (∼2.5 λ) is relatively large compared to a normative design for a transducer array. To reconstruct a decent image, the dimension of the pitch was usually set within 0.5–1 λ, which benefit the image resolution in lateral. Bezanson et al. developed a miniaturized 64-element phased array, for which the element-to-element pitch was ∼1 λ [6]. Their impressive image of the porcine brain suggested that the resolution was advanced by using a small pitch size. In our study, the pitch size was compromised with the aperture (the circumference of the circular array in diameter of 1.7 mm) and element number of 50. To compensate the lateral resolution and grating lobe induced, a customized synthetic aperture method was adopted to make the image reconstruction.

Another drawback of this research lies in the imaging setup. Due to the lack of high-frequency imaging system, the whole process is time-consuming, and the vibration-induced error may affect the image accuracy when each element wire was connected with the electric system. In the future, enhanced imaging quality can be expected with a high-frequency electric system. Finally, the circular array will be integrated into a catheter for interventional imaging. Therefore, the length of the transmission line cable is not neglectable. Electric impedance matching needs to be implemented for the catheter integration.

Nevertheless, the imaging result of the micromachined single crystal composite circular array on the wire phantom is encouraging, and the use of single crystal composites in high-frequency medical imaging will no doubt play more of a role in the intravascular ultrasound in the future.

Funding Data

• National Institutes of Health (R01EB015508 and 1R41EY021392-01).

• The U.S. Army Medical Research and Material Command (PC111309).

• National Science Foundation (ECCS-1542015).