2D array design and fabrication with pitch-shifting interposer at frequencies from 4 MHz up to 10 MHz

Abstract

High element density and strict constraints of the element’s size have significantly limited the design and fabrication of 2D ultrasonic arrays, especially fully sampled 2D arrays. Recently, 3D printing technology has been one of the most rapidly developing fields. Along with the great progress of 3D printing technology, complex and detailed three-dimensional structures have become readily available with a short iteration cycle, which allows us to reduce the complexity of routing and helps to ameliorate assembly problems in 2D ultrasound array fabrication. In this work, we designed and fabricated 2D ultrasound arrays for an array of applications with a pitch-shifting interposer, which allowed us to fit different array designs with the same circuit design and significantly reduce the requirements in routing and connection for 2D array fabrication at frequencies from 4 MHz to 10 MHz. Results demonstrated that this design would make 2D arrays more available and affordable.

I. INTRODUCTION

Tremendous development has been achieved in matrix ultrasound transducer array fabrication and in its applications in the past few decades [1–4]. 2D ultrasound arrays, which enable real-time volumetric imaging, have been designed, fabricated, and applied in transesophageal echocardiography (TEE) [5, 6] and intracardiac echocardiography (ICE) [7, 8], and shown promising potential in overcoming the limitations of 2D ultrasound imaging with 1D ultrasound arrays. Two-dimensional arrays have also been applied in elastography since a 2D array can generate highly focused beams with the focal point being adjustable in three-dimensions [9]. 1.75D ultrasound arrays [2, 4], whose elements can be accessed individually in elevation are specified for providing electronic focusing in elevation within the whole depth range of the field of view and are capable of providing a highly focused image slice, thereby significantly reducing the influence of aliasing from nearby slices, which increases contrast.

Despite all these advantages, difficulties of fabrication impede the further development of matrix ultrasound array technology. Conventional fabrication process for 1D arrays is difficult to apply for a 2D array due to its high element density. Fiering et al [10], presented a customized flexible printed circuit with a complex fabrication process that was implemented to build a 2D array. In the face of such challenges, multiple solutions have been provided to relieve this bottleneck in the past few decades. Application-specific integrated circuits (ASICs) with pre-amplifier, multiplexer or micro-beamformer for sub-array processing were designed for fully-sampled matrix arrays with the purpose to reduce the channel number to a more reasonable value for most ultrasound systems [2, 7, 11–15]. Such ASIC based matrix arrays have been commonly fabricated directly on the surface of the IC substrate itself, which benefits the signal-to-noise ratio in addition to relieving the 2D routing bottleneck. Thus, in such pitch-matched designs, the ASICs’ pads and related circuitry need to fit into the area of an element. With the help of semiconductor technology, great research has been done in academia and industry and presents different ASICs with promising performance. Even though these designs represent significant work for trading off design parameters to fit with array requirements and improve performance, the limited area still constrains the performance and becomes one of the main challenges for the ASIC design. The complexity of these tradeoffs is further compounded by the fact that in ultrasound it is necessary to utilize both low voltage and high voltage ASIC fabrication processes.

Sparse array and row-column array architectures are alternative approaches to solving the problem of excessive elements in a matrix array. As opposed to fully-sampled matrix arrays, sparse arrays are under-sampled dense arrays [16–19]. Keeping the aperture size the same while reducing the total number of active elements leads to the problem of grating lobes. Significant research has been done to improve this approach by applying various signal theory optimization techniques [19, 20]. In addition, research on non-rectangular sparse 2D arrays provides another potential solution for the challenges of 2D array density [18].

Recent work in the design and implementation of row-column arrays demonstrates another promising solution to 2D array design issues. Recently, a forward-looking miniature endoscope designed for 3D imaging was presented by Katherine Latham et al [21]. This device was based on a 30-MHz row-column array with 64 × 64 orthogonal elements. Utilizing a previously reported novel imaging technique [22], this array presented a clear 3D image of two crossed, depth offset wires with good resolution. However, the novel imaging technique requires different pulse polarity on transmit and receiving which presents further challenges in circuit and system design.

In previous work, we have presented matrix arrays fabricated with 3D printed interposers [2, 9, 23]. The interposers with multiple straight channels were filled with conductive epoxy which acts as a backing layer as well as provides electrical interconnection between elements and locally integrated interface circuits. This design is applicable for both PCB and ASIC substrates. In recent years, 3D printed technology has improved significantly. Stereolithography (SLA), a further type of 3D printing technology that is based on laser scanning and liquid resin solidification [24], has shown excellent printing resolution and demonstrated versatility in research and industrial work. Large 3D structures comprised of nearly 10 million 150 μm pitch periodic cells, were presented by Angkur Jyoti Dipanka Shaikeea et al [25]. To fabricate this design, a moving projection SLA system with the micrometer-scale resolution was realized. 3D printing technology improvements are not limited to better resolution. A novel STA technique which supported multi-material projection was reported by Zhenpeng Xu et al [26]. This approach enabled dissolvable supports in complex architectures, which makes the printing of overhanging, fly-like micro-architectures possible. Progress in 3D printing for high-quality and low-cost micro-scaler resolution has already directly benefited a range of fields including biotechnology [27], optical devices [28], and electronic device [29, 30] to name a few.

In this work, we present a novel 3D printed interposer with pitch-shifting micro-architectures to address the challenging requirements in scaling circuits for matrix arrays with high frequency. In previous work, 3D printed interposers with straight channels that were filled with conductive epoxy were used to provide a connection for the matrix array and acted as a backing layer at the same time. The interposers here are the interface between transducer elements and the circuits, and decouple the bonding in between. Thus, instead of being limited by the two existing interconnect surfaces, the routing problem from the electronics to the elements can be solved with one additional degree of freedom that is introduced by the interposer. The complexity of this architecture is reduced significantly when we take advantage of new 3D printing technology that enables complex 3D architectures with microscale. Additionally, integrating such micro-fabricated pitch-shifting interposers into array fabrication facilitates the use of a single and universal circuit substrate for multiple different transducer matrix designs with different geometries and scales. In conventional designs, the matrix geometry is limited by the ASICs or the PCB. A new array design necessitates redesigns of the IC design as well, which is extremely time-consuming and costly. 3D printing processes are much more affordable than IC design flow in many aspects.

To demonstrate the benefits of the pitch-shifting interposer for universal substrate integration, two matrix arrays with different center frequencies, 4.5 MHz and 10 MHz, were fabricated based on the same circuit substrate with different interposers. The two different interposers matched the universal PCB substrate to the two different arrays which had different element pitches due to the different wavelengths used. The two fabricated arrays further had different intended applications, for neural stimulation and imaging respectively, which again was facilitated by the use of the pitch-shifting interposer. The performance of these fabricated arrays has been evaluated and the results will be presented herein.

II. MATERIAL AND METHODS

A. Printed Circuit Board

High-density interconnect printed circuit boards (HDI-PCB) were used to route the elements to the system cable. These industry-standard substrates have a 2+N+2 architecture, trace and space of 3 mils (75 μm) and are capable of being fabricated by multiple manufacturers. Compared with conventional PCBs, HDI-PCBs are smaller in size and provide different via types instead of through holes only, which are advantageous in reducing routing complexity and increasing the density of the pad matrix. The PCB used for array connection has a 16 × 16 pad matrix with 635 μm pad pitch in the central area, which will be used for array assembly in the following process.

B. Interposer and Array Design

There were two types of arrays fabricated with the HDI-PCB. They shared similar geometries but different pitch, center frequencies, and applications.

1). 4MHz Neural Stimulation Array:

This array was aimed at nerve stimulation. We chose bulk hard PZT material (DL-48, DeL Piezo Specialties, LLC, West Palm Beach, FL, USA) to achieve a higher-pressure acoustic field.

The transducer aperture was a 16 × 16 element matrix [Fig. 1.a] with a 770 μm element pitch. Arrays with 2λ pitch demonstrate grating lobes during steering, yet a larger aperture benefits acoustic field pressure and provides increased penetration depth. The interposer architecture can be decomposed into three parts [Fig. 1.b]. The interposer was designed with an isometric magnification structure and is mainly formed with parallel channels where the pitch increased from 635 μm (at the PCB side) to 770 μm (at the transducer side). The top and bottom were straight channel regions each with a 16 × 16 matrix with matching pitch to the pad matrix on the PCB and the aperture. The thickness of the walls between the channels was 235 μm and 150 μm for the PCB side matrix and the transducer side matrix respectively. The wall thickness to pitch ratio was optimized differently for the PCB vs. the transducer matrix. On the transducer side, a thinner wall was preferred for a more active area of the interposer backing for better acoustic absorption. The thicker walls on the circuit side reduce the risk of short circuits between channels during the assembly process. In the center region, the channels were scaled isometrically and the distance between the channels changed from 635 μm to 770 μm.

Fig. 1.

Conceptual drawings for 16 by 16 2D arrays and interposers. (The matching layer is not applicable for the stimulation array.)

2). 10MHz Imaging Array:

This array was designed as a high-frequency 2D imaging array. A 1–3 composite made with soft PZT material (DL-53, DeL Piezo Specialties, LLC, West Palm Beach, FL, USA) and epoxy as kerf filler (EPO-TEK 301, Epoxy Technology, Billerica, MA, USA) were used as the active material of this array and two matching layers were implemented.

The array aperture had a 16 × 16 matrix with a center frequency of 10 MHz as a trade-off between the field of view and image quality, the element pitch of the array was designed to be 2λ (300 μm). Larger arrays with the same field of view at λ pitch could be implemented by increasing the number of imaging channels and this would improve image performance by reducing side-lobes. The interposer architecture was similar to the stimulation array. In the pitch-shifting area, the pitch reduced from 635 μm to 300 μm. The wall thickness was 235 μm on the circuit side and 60 μm on the transducer side.

For both interposers, we left relatively long lengths for the straight regions. This design was intended to provide more margin for the following steps and the straight channels connected to the 10 MHz elements acted as the backing layer. A thicker backing provides improved attenuation of acoustic energy on the back of the array which is especially important in imaging for reducing ringdown and thereby improving axial resolution.

C. Interposer Fabrication

Since the two interposers differed significantly in terms of overall dimensions and minimum feature size (which is the channel wall at the transducer side) different printers were utilized for the fabrication process.

The stimulation array interposer [Fig. 2. a] requires a large printing volume and has a relatively modest requirement for printing resolution. The 3D Systems HD 3500 plus (3D Systems, Rock Hill, SC) with a minimum resolution of 32 μm was selected in this case. This printer was also frequently used in our previous work [2, 23, 31].

Fig. 2.

Interposer for the stimulation array (a, b, c) and the imaging array (d, e, f). a, d: interposers before e-solder filling; b, c, e, f: interposers filled with e-solder;

The imaging array interposer [Fig. 2. d], placed high demands on the print resolution, and therefore a DLP 3D printer (Kudo, Dublin, CA, USA) with a minimum resolution of 15 μm was used to fabricate this design. After printing, the interposer was immersed in alcohol and placed in an ultrasonic cleaner for 15 to 20 minutes. Due to the porous structure of the interposer, this step is crucial to clean out any remaining resin in the channels and avoid the channels being blocked. The interposers are not fully cured after printing and therefore, after cleaning they were further cured in an oven at 46°C for one hour. Every interposer was checked again to ensure clean channel yield after curing.

The filling process is the same for both interposers. Dams for containing the silver epoxy filling material were fabricated to match the outer dimension of the interposers and were glued to a supporting glass substrate for easy handling. The dams were filled with the silver epoxy and the interposers were carefully lowered into the dam-encapsulated epoxy. An important consideration in this regard is that the dams need to fit snugly with the interposer outer dimension with minimal gaps between the two. If the gap is not large enough, the interposers cannot be lowered due to friction against the walls, while too large a gap would cause leakage in the filling process and lead to the failure of the interposer fabrication due to insufficient filling of the channels and wasted material.

The fabricated dam/glass substrate structures were filled with conducting silver epoxy paste (E-Solder 3022, Von Roll Isola, New Haven, CT, USA) and centrifuged at 1000 rpm for 5 minutes. Centrifuging is important for consolidating the silver epoxy particles to increase the acoustic attenuation of the backing material. After this process, the paste surface was flat and any air bubbles created during the mixing process were removed. Uniformity of the backing material with no air bubbles is important to maintain high-quality acoustic attenuation without any significant echoes back to the transducer composite. The interposers were pushed into the dam-confined silver epoxy slowly and with evenly applied high pressure to ensure that the paste fills every channel. Yield in the fabrication of 2D arrays is multiplicative, and every step must have a target yield of 100% such that the final yield of the overall process is high. Therefore, we optimized the fabrication process of the interposers to achieve the highest yield possible and further validated the electrical functioning of every channel in the completed structures prior to continuing the overall transducer fabrication process.

After the interposer was pushed all the way to the glass, the dam was cut, and the interposer was released. Filled interposers were left in a dry box for 24 hours for curing. Extra silver paste on the surface of each active side of the interposers was desired since the paste shrinks during curing and can lead to dents and holes in the channels. Filling with a given interposer can only be done one time as an attempt to rework an incompletely filled channel invariably leads to an air gap that isolates the previously filled paste and new paste in the channels, and this reduces the yield of the conductive channels significantly. After curing, the silver paste in the channels hardens and forms a conductive pathway. The cured interposers were then lapped at the top and bottom surface to remove the extra e-solder and create a flat surface on each side [Fig. 2. b, c, e, f]. Surface co-planarity is critical for both the subsequent transducer fabrication steps as well as the assembly to the electronic substrate. We therefore incorporated surface measurements into the post-fabrication testing criteria with a goal of co-planarity below +/−10μm. The last step in the fabrication process was dicing on the surface to create isolated assembly pins with air gaps between them [Fig. 3. a]. We further diced the surface to create fiducial marks which are critical for precise alignment [Fig. 2. c] for the subsequent steps in the overall acoustic module fabrication process.

Fig. 3.

Schematic of Acoustic Stack Fabrication, the most left and right pillars are connected to GND: a: Creating electrodes on the circuit side of the interposer; b: Assembling Piezo material (with first matching layer for imaging array); c: Creating separated elements and filling kerf; d: Sputtering gold; e: Attaching second matching layer; f: Finished acoustic stack

D. Acoustic Stack Fabrication and Assembly

The fabrication process for the two acoustic stacks was similar except that the imaging array had the first matching layer attached before assembly. The first matching layer was a mixture of Insulcast 501 (Insulcast, Montgomeryville, PA) and 2–3 μm silver powder. The mixture was cast on the composite while placed in a dam and centrifuged. After curing, it was lapped to the design thickness of 47 μm.

Once the piezo material was ready, it was bonded on the transducer side of the interposer using a thin layer of E-solder [Fig. 3. b]. Interposers with attached materials are shown in [Fig. 4. a, d]. The E-solder paste was applied to the interposer surface first, then the material was aligned and pressed down on the paste. The extra paste squeezed out and surrounded the piezo material. This is by design, as the channels connected to the ground were designed to surround the central channels. The extra paste connects with those channels and is used for ground connection in subsequent fabrication steps. The assembly was left in a dry box overnight for curing. Once cured, the piezo material was diced to create independent elements. The marks created in previous steps were used for the alignment of the dicing saw to the interposer grid [Fig. 3. c]. This process was applied for both directions of the matrix and EPO-TEK 301 was used as kerf filler to support the elements [Fig. 4. b, e]. It is critical to avoid dicing along the edge of the material, as in this case a short circuit would be formed between the elements on the edges and ground [Fig. 4. e]. We, therefore poured additional kerf filler in these kerfs to build a smooth bridge from the element surface to the E-solder that covers the ground. A matching layer detaching issue may happen at this step. As shown in [Fig 4. e], the columns show gold color in locations where the matching layer was delaminated. There could be more columns affected since not every detached matching layer is removed during the dicing process. This issue has the potential to influence the bandwidth and sensitivity of the array, which is observed in the testing results. After the curing process of EPO-TEK 301, the array was sputtered with a Cr/Au (50/100 nm) electrode on the surface of the matrix [Fig. 3. d]. By doing so, the top electrodes of all the elements and the extra E-solder connected to the ground were connected. The “bridge” mentioned earlier is necessary to make this connection possible and its shape directly influences the integrity of this connection. The above-described process steps fully describe the fabrication of the pushing array. For the imaging array, an additional step is added to the fabrication process which is a lamination of the second matching layer [Fig. 3. e].

Fig. 4.

Array fabrication process of the stimulation array (a, b, c) and the imaging array (d, e, f). a, d: piezo material assembled; b, e: dicing along the marks on the interposer to create individual elements; c, f: array assembled on HDI-PCB. The columns exhibiting gold color in e were suffering from matching layer detaching issue, which was the consequence of loss connection at the interface between the matching layer and the gold layer or the interface between the gold layer and the piezo material. Second matching layer was assembled in f.

Finished acoustic stacks were next assembled to the PCBs. The PCB was designed at USC using the Altera CAD tool and fabricated by an external vendor (MKT Electronic Co., Ltd., Guangdong, China). The “stamping” process introduced in our previous work [23] was used for assembly. The circuit side of the interposer was first stamped on a thin layer of E-solder and then the acoustic stack was aligned and assembled on the matrix area of the HDI-PCB. After curing, the array fabrication process was completed [Fig. 4. c, f].

E. Array Performance Testing

To evaluate the array’s performance, each of the two fabricated array types was tested using pulse-echo (P/E) testing, hydrophone measurement (for pushing array), and phantom imaging (imaging array). A vantage 256 ultrasound system (Verasonics Inc, Kirkland, WA, USA) was used to drive the transducers.

1). P/E Testing:

A 75 mm thick quartz cube was used as the target. The arrays were securely mounted on a combined stage which allowed multi-degree freedom movement to adjust the orientation of the array surface. This setup allowed us to avoid the influence of surface tilting on signal uniformity. A similar approach was also implemented in another test. The ultrasound system drove one element at a time, and the pulse frequency was the designed frequency of the transducers (4.5 MHz for the stimulation array and 10 MHz for the imaging array).

2). Hydrophone Measurement:

The pressure of the acoustic field at the focal point was the most important parameter to evaluate the stimulation array. Thus, a hydrophone measurement was implemented to examine the emitted field and record the highest pressure the array could achieve. A needletype PVDF hydrophone (Precision Acoustics Ltd., Dorchester, UK) was mounted on a 3-D stepper motor (SGSP33–200, OptoSigma Corporation, Santa Ana, CA, USA) and scanned an 8 mm × 8 mm plane that was 20 mm away from the surface of the array. This distance was chosen based on the location of the target in the experiment. The drive frequency was 4.5 MHz, with 20 full cycles, and the output voltage was limited to 10 V_Peak to avoid potential damage to the hydrophone.

3). Phantom Imaging:

A target with five micro-scale steel wires was used as the phantom to evaluate the imaging array. The wires were arranged in a trapezoidal pattern with a vertical spacing of 1.5 mm and horizontal spacing of 0.5 mm [Fig. 5]. The frames of the target provided a 6 mm × 6 mm imaging window that matches the geometry of the imaging array. Coherent plane-wave compounding was implemented as the beamforming technique. Each imaging cycle generated 21 angled plane waves in the XZ plane and YZ plane respectively. Images in the XZ plane, YZ plane, and XY plane and a volume image were reconstructed. Image reconstruction was processed by the Verasonics system in real-time.

Fig. 5.

Wire target with 5 micro-scale stainless steel strings. The strings were mounted on the stepwise structure of the frame tightly.

III. RESULTS

A. Stimulation Array Characterization results

The system scanned each element and acquired its respective P/E results. The data was arranged and presented in a matrix according to the element in the matrix. This table visualizes the overall yield of the transducers and the clustering of non-functioning channels. The yield of the stimulation array was 94% [Fig. 6. a]. The main reason for the non-functional elements was blocked interposer channels. The HD 3500 plus printer uses wax to support the structure. Due to the interposers’ porous structure, complete wax removal was difficult to achieve, with the potential for any residual wax to end up blocking the interposer channels. The blocked channels could not provide a continuous conducting path due to the resulting failure of the e-solder extrusion process at that element. Printers which use resin are prone to the presence of residual resin in the channels, however, the removal process for the resin is usually easier and has a higher yield.

Fig. 6.

Array performance maps and acoustic field. a. Sensitivity map of the stimulation array; b. acoustic field measured by hydrophone, focusing at 20 mm in the axial direction; c. Sensitivity map of the imaging array; d. Bandwidth map of the imaging array; e. The absolute response map of the imaging array.

The intended application for the forcing array was neural stimulation. Previous experimental data indicated that an acoustic field pressure higher than 1.7 MPa was required for successful neural stimulation [32]. From the hydrophone measurement [Fig. 6. b], 60 mV voltage was received at the focal point with a 17 V driving voltage. And the pressure of the acoustic field was calculated to be 1.4 MPa. The driving voltage was limited by design during this measurement to protect the hydrophone. In the actual neural stimulation experiment, the Vantage system with HIFU configuration was used to provide up to 90 V_peak output voltage. Therefore, given linear gain in the acoustic field pressure with drive voltage, the required pressure can be easily achieved.

B. Imaging Array Characterization results and images

A similar process was undertaken for the imaging array’s P/E results, and the resulting measured overall yield was 88% [Fig. 6. c]. Most of the disconnection happened at the edges, which was usually a consequence of the overlapping of the electrodes on the edges. Another factor that influenced the assembly process was the area of the array. During the assembly process, the array surface was waxed on a holder. This array’s aperture size was less than 4 of the stimulation arrays, which made it more difficult to mount the array flatly on the holder. The central left area presented lower bandwidth compared with other elements [Fig. 6. d], which was due to matching layer detachment. The average bandwidth of the working elements is 41%, which is below the simulation predicted 62%. Besides the detaching issues, overlapping would be the main reason for the mismatch of the bandwidth.

The crosstalk level between the nearby channels was measured based on pulse-echo results. Between the channels that are right next to each other, the crosstalk level is at −20 dB. And between the channels on the diagonal direction, the crosstalk level was reduced to −40 dB. The crosstalk level between the channels with further distance was very close to what it is between the channels along the diagonal direction.

The array was next used to image the wire targets. Imaging views for XZ, YZ, and XY planes and a volume image were acquired in real-time [Fig. 7]. XZ and YZ planes correspond to the cross-section that crossed the y-axis and x-axis respectively, and the XY plane was located at Z = 60 wavelengths. In the displayed volume image [Fig. 7 d], the matrix formed by a group of black points at the top represents the array aperture and provides aid for recognizing the position of the target relative to the transducer. Two sets of images were recorded [Fig. 7 a–d and Fig 6 e–h]. The target was rotated by 45° in the second image set, which can be seen by the orientation of the wires in XY images [Fig. 7. a. e] and 3D models [Fig. 7. d. h] in volume images. The axial and lateral resolutions of the first wire target, from top to bottom, were 430 μm and 640 μm. For third wire target, the axial and lateral resolutions were 450 μm and 750 μm. And at the fifth wire target, the axial and lateral resolutions were 525 μm and 825 μm.

Fig. 7.

Images of the wires target. a-d were acquired with the wires parallel to the x direction, e-h were acquired after the target was rotated by 45 degree. a, e: XY plane; b, f: XZ plane; c, g: YZ plane; d, h: volume images.

IV. DISCUSSION AND CONCLUSIONS

Two 2D ultrasound arrays with different operating frequencies, element pitch and target application were fabricated and tested in this work. By introducing a pitch-shifting interposer into the array design, these two arrays were able to be fabricated with the same circuit. Arrays were then tested for their performance. The stimulation array’s measured performance demonstrated adequate pushing force to satisfy the application requirements. Volume images were acquired from the imaging array, and in this case, the quality of the results was limited by the bandwidth and sensitivity of the elements. For future work, reduction of the element pitch to 1λ or smaller should benefit in imaging performance.

The arrays presented in this work were examples to show the promising potential of implementing pitch-shifting interposer architectures and 3D printing technology in ultrasound array fabrication. Pitch-shifting interposers were designed following the same idea in our previous work: filling the channels with a conductive epoxy to create conducting channels that act as a backing layer [2, 9, 23, 31]. In addition, with the introduction of a shape transformation along the Z-axis, the interposers are no longer simply for connection and backing, and now also play a role in shaping the array geometry and pitch requirements. The pitch-shifting architecture is a straightforward modification of the original idea of interposers: by applying isometric scaling of the pitch between channels and the diameter of the channels, this design methodology allows us to adapt the fixed pad matrix on the circuit board to any number of different pitch array matrices. The transformation that is inherent in pitch-shifting also introduces new and potentially far-reaching topological possibilities provided by interposers. Specifically, the deformation achieved in such interposers is not limited to pitch scaling alone and can be further used for reshaping the geometry of the aperture. For example, previously described spiral sparse-array geometries [18] could be implemented using this technique. In general, any re-mapping of the transmission and receive element location that was previously constrained by the ASIC pad array definition, can now be realized as long as the geometry of the circuit electrodes provides a one-to-one mapping with the target element matrix, and the structure is supported by the 3D printer used to create the interposer.

Fabrication of filled interposers as described here is a (sometimes messy) hands-on process that is limited to prototyping and therefore not readily amenable to volume production. Since the details of the conductive particle size and viscosity of E-solder are not provided by the manufacturer, the minimum channel width we expect to reach is estimated with our observation and previous work experience. According to our previous work, an interposer with 70 μm×150 μm channels was filled successfully [33]. Thus, a 70 μm×70 μm could be fillable and it could be possible for smaller channels but the process will be challenging and require special tooling. For the quick-turn realization of novel element mapping geometries, this prototyping capability is invaluable to rapidly test and validate design concepts. In the future, this process could be adapted to volume production through the use of injection-molding and industrial epoxy filling extrusion machines.

In such a volume process it would be critical to identify important yield issues including bubbles formed in the epoxy fill which reduce the effectiveness of the acoustic backing. In addition, the yield of electrical connections is critically dependent on issues such as pinholes formed in the walls of the channels which can potentially lead to shorting of neighboring elements. Finally, coplanarity of the top and bottom surface of the interposers is a particularly important yield criteria because it has significant effects on the assembly of both the top side transducer array as well as the bottom side electronic substrate connections. All these issues must be taken into consideration if a viable production process is to be implemented.

The main limitation we are facing to reduce the element pitch of the interposer is the resolution of the 3D printer. The interposer requires high resolution, relatively large printing area and thickness at the same time, which is challenging to achieve. In our previous work [33] we utilized special-purpose high resolution printing technology with resolution down to the micron scale. We anticipate that in the future, these advanced printing technologies will become more widely available, thereby facilitating improved resolution at higher frequency with the interposer architecture.

We expect that future exciting improvements in 3D printing technology will bring more possibilities to interposer design. With new printers that support multiple material types, acoustic backing fill and lapping steps could also be simplified, which will further improve the yield and quality of the interposer.

In summary, with the use of advanced 3D printing technologies, we have implemented novel 3D printed pitch-shifting interposers which enable an adaptive 2D array fabrication process in which the circuit pad topology does not limit the geometry of the transducer matrix. With the refinement of the fabrication process and improvements in 3D printing technology, interposers with continuous deformation of the channel matrix will bring new possibilities for array design and solve circuit design challenges for dense 2D array implementation at low cost and with quick-turn prototyping capabilities

TABLE I.

Stimulation Array Parameters

Parameter	Value
Piezo Materials	Hard PZT, bulk
kt	0.6
Matching Layer	NA
Backing	E-solder
Center Frequency	4.5 MHz
-6dB Bandwidth(measured, average)	11%
Pitch	770 μm
Kerf	150 μm
Number of Elements	256
Yield	94%

TABLE II.

Imaging Array Parameters

Parameter	Value
Piezo Materials	Soft PZT, 1–3 composite
kt	0.6
First Matching Layer	2–3 μ m silver epoxy
1ST ML Sound Speed	1961 m/s
1ST ML Acoustic Impedance	7.84 MRayl
1ST ML Thickness	47 μ m
Second Matching layer	ABS
2ND ML Sound Speed	1850 m/s
2ND ML Acoustic Impedance	2.2 MRayl
2ND ML Thickness	44 μ m
Backing	E-solder
Center Frequency	10 MHz
-6dB Bandwidth(measured, average)	41%
Pitch	300 μm
Kerf	60 μm
Number of Elements	256
Yield	88%

Biographies

Haochen Kang received the B.Eng. degree in electrical engineering from the Huazhong University of Science and Technology, Wuhan, China, in 2015, the M.S. degree in electrical engineering from the University of Southern California, and the Ph.D. degree in biomedical engineering from the University of Southern California, Los Angeles in 2020. where he is currently pursuing the Ph.D. degree in biomedical engineering. His research is focused on the design and application of 2-D ultrasound systems.

Yizhe Sun is a graduate student pursuing his Ph.D. degree in Biomedical Engineering at the Viterbi School of Engineering at the University of Southern California (USC). He received his B.S. degree in Aircraft Engineering and Design at Nanjing University of Aeronautics and Astronautics in 2017. From Jan-July 2018, he worked as a Research Assistant in Hong Kong University of Science and Technology. He is proficient in Mechanical Engineering and Microelectromechanical Systems (MEMS). Currently, he is a Ph.D. student and lab manager at the Ultrasound Transducer Research Center working to develop the Ultrasound Array and Single Element Transducers.

He plans to continue his work in academia include the design, and fabrication of high-frequency ultrasonic transducers and arrays.

Robert Wodnicki received the B.Eng. and M.Eng. degrees in electrical engineering from McGill University, Montreal, in 1992 and 1996, respectively, and the PhD degree in biomedical engineering from the University of Southern California, Los Angeles in 2020. The subject of his dissertation was “Highly integrated 2D ultrasonic arrays and electronics for modular large apertures” under Prof. Qifa Zhou. From 1995 to 2014, he was employed as an Application Specific Integrated Circuit (ASIC) Designer for GE Global Research, Niskayuna, NY, USA. In 2018 he received the Best Student Paper Award at the International Ultrasonics Symposium in Kobe, Japan, for his paper on ASICs and modular acoustic arrays. He is currently employed as a senior research associate in the department of Ophthalmology, at USC Keck in Los Angeles. His research is focused on the implementation of 2-D ultrasound systems using highly integrated ASIC electronics interfaced to single-crystal transducer arrays.

Qingqing He (Graduate Student member, IEEE) received the B.Eng. degree in Mechanical Engineering from Central South University of Forestry and Technology, Changsha, China, in 2018, and the M.S. degree in Mechanical Engineering from the University of Southern California, Los Angeles, CA, USA, in 2021. She is currently pursing the Joint Ph.D. degree with the department of Mechanical Engineering in University of California, San Diego, and San Diego State University, CA, USA.

Her research interests mainly focus on the biomimetic 3D printing, micro fabrication, 3D-printed wearable sensing devices, self-healing material.

Yushun Zeng (Graduate Student member, IEEE) received the B.Eng. degree in Bioengineering from Nanjing Tech University, Nanjing, China, in 2019, and the M.S. degree in Biomedical Engineering from the University of Southern California, Los Angeles, CA, USA, in 2021. He is currently pursuing the Ph.D. degree with the Department of Biomedical Engineering, University of Southern California, Los Angeles, CA, USA.

His research interests mainly focus on the high-frequency ultrasound transducer/array, 3D-printed ultrasonic device, acoustic tweezer.

Gengxi Lu received his bachelor’s degree in Physics from Nanjing University, China, in 2017. He is currently a Ph.D. candidate at the Department of Biomedical Engineering, University of Southern California, Los Angeles, CA, USA. His research interests include ultrasound neuromodulation, ultrasound imaging and elastography, 3D printing, and the development of ultrasound transducers.

Jung-Yeol Yeom has been a professor at the School of Biomedical Engineering of Korea University, Seoul, South Korea since 2015. He received his bachelor’s degree in Nuclear Engineering from Seoul National University, South Korea in 2001, and his master’s and Ph.D. degree in Quantum Engineering and Systems Science from the University of Tokyo, Japan, in 2003 and 2006, respectively, on electronics and detectors for positron emission tomography. Dr. Yeom’s lab focuses on detectors and instrumentations for both medical (such as nuclear medicine, ultrasound and multi-modal imaging) and non-medical (radiation monitoring and homeland security) applications.

Dr. Yang Yang’s research focuses on the machine, materials, and structures development for bioinspired 3D printing. He received his B.S. in Physics from Wuhan University in 2009. He completed the joint Ph.D. at Wuhan University and University of California, Los Angeles (UCLA) in Physics and Bioengineering in 2015. Prior to joining SDSU, Dr. Yang is a Postdoc at The University of Southern California (USC) in the Department of Industrial and Systems Engineering and Center for Advanced Manufacturing. He has received the ‘2022 SME Sandra L. Bouckley Outstanding Young Manufacturing Engineer Award’ and has authored over 50 peer-reviewed publications such as ‘Science Advances’, ‘Advanced Materials’, ‘Energy &Environmental Science’, ‘Research’ and he is the reviewer of several journals such as ‘Advanced Materials’, ‘Small’, and ‘Additive Manufacturing’. His work has been supported by NSF and SDSU seed grants.

Qifa Zhou received his Ph.D. degree from the Department of Electronic Materials and Engineering at Xi’an Jiaotong University, China in 1993. He is currently a professor of Biomedical Engineering and Ophthalmology at the University of Southern California.

Dr. Zhou is a fellow of the Institute of Electrical and Electronics Engineers (IEEE), the International Society for Optics and Photonics (SPIE), and the American Institute for Medical and Biological Engineering (AIMBE). He is also a member of the Technical Program Committee of the IEEE International Ultrasonics Symposium, and is an Associate Editor of the IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control. His research focuses on the development of piezoelectric high-frequency ultrasonic transducers/array for biomedical ultrasound and photoacoustic imaging, including intravascular imaging, elastography and ophthalmic imaging. He is also actively exploring ultrasonic mediated therapeutic technology including trans-sclera drug delivery, as well as ultrasound for retinal and brain stimulation. He has published more than 280 peer-reviewed articles in journals and edited two books.