IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control ,

vol. 59, no. 2,

February

2012

231

Detailed Investigations of Polymer/Metal Multilayer Matching Layer and Backing Absorber Structures for Wideband Ultrasonic Transducers Minoru Toda and Mitch Thompson Abstract—Detailed investigations of multilayer front and back matching layers and a novel backing absorber have been conducted, the detailed theory for which was presented in a previous paper. To design useful structures using the simple proposed equations, the material parameters of the constituent layers must be identified. Therefore, polyimide (for the matching layer) and adhesive-backed copper tape (for the absorber) were characterized by bonding them to polyvinylidene fluoride–trifluoroethylene P(VDF-TrFE) copolymer ultrasonic transducers and then applying a parameter-fitting algorithm to the resulting impedance data. A double matching layer was designed using an 11-μm PVDF (inner) and 23-μm copper (outer) multilayer construction in the first matching section followed by a 75-μm polyimide layer as a typical quarter-wavelength material in the second (outermost) matching section. This structure was bonded to 330-μm PZT with air backing and the reflection waveform from a short pulse was captured. The FFT frequency response showed a 3.1-MHz bandwidth centered at 6.4 MHz, which agreed with the Mason’s model analysis. The use of multiple layers of copper tape as a backing absorber was also investigated. At 3 MHz, the measured impedance was 4 MRayl, attenuation was 220 dB/cm, and velocity was 890 m/s, which agreed with the design theory. The 4-MRayl copper-tape structure was bonded to a back matching structure made from one layer of polyimide and one layer of brass (multilayer matching), and the effectiveness of the backing absorber made of 10 layers of copper tape on a 3-MHz transducer was confirmed.

I. Introduction

T

he acoustic impedance of piezoelectric materials commonly used in ultrasonic medical imaging transducers is much higher than that of the typical water or biological tissue propagation medium. This impedance difference produces a reflection at the front and back boundaries of the transducer material, creating a sharp resonance and prolonged ringing after pulse excitation. To generate and receive short acoustic pulses with less ringing, the transducer bandwidth must be wide. Therefore, to reduce the multiple internal reflections and to dampen the resonance, front impedance matching layers are required [1]–[8]. Although well-known equations can be used to calculate the exact front matching layer acoustic impedance required for optimum performance, in practice it is often difficult to find

Manuscript received June 23, 2011; accepted November 14, 2011. The authors are with Measurement Specialties Inc., Wayne, PA (e-mail: [email protected]). Digital Object Identifier 10.1109/TUFFC.2012.2183 0885–3010/$25.00

production materials having exactly the calculated acoustic impedance. In such a case, when the exact material is not available, a backing absorber may be necessary to reduce resonance caused by reflections from the back surface [9], [10]. For these reasons, there is a need for effective front and back acoustic matching layers. It is also essential to have simple and reliable design equations and low-cost production processes for these impedance structures. We therefore proposed a new structural design using multiple polymer-metal layers for matching and backing absorber applications in a previous paper [11]. In that work, a design for matching structures was proposed in which a thin polymer layer was at the higher-impedance side of the impedance conversion structure and a thin metal layer was on the lower-impedance side. This is the opposite of the typical approach in which the impedances of the layers decreases monotonically moving away from the transducer material; furthermore, the individual layer thickness in the new concept are typically 1/15 to 1/30 of a wavelength. Using a similar principle, a design having iterative layers of high-acoustic-loss adhesive and thin metal was proposed in [11] for the backing absorber application, although detailed experimental proof was not presented. To verify the design approach and to investigate more carefully, we have fabricated examples of structures with typical, useful acoustic properties and compared the measured results with theoretical analyses. One goal of this paper is to carefully determine the acoustic parameters of each of the materials used in these structures and to validate the proposed design equations by comparing their behavior with measured results. Another goal is to test the effectiveness of front and back matching and backing absorber structures at frequencies typical of medical ultrasonic transducers. A well-accepted measurement method for acoustic absorption and velocity is to determine the time of flight and magnitude declination through a known thickness of the test material. One particular method is to immerse a test material in water and measure propagation time and attenuation of a reflected acoustic pulse [12], [13]. This method produces useful results with a thickness of roughly ~1 cm or more and requires a surface area large enough to accommodate the acoustic beam profile, and therefore this approach is not useful for measuring properties of polymer films less than 1 mm in thickness. Another method is to directly bond two transducers to opposite surfaces of a

© 2012 IEEE

232

IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control ,

bulk material and to measure the propagation characteristics [14], but this is also not useful for thin layers. For thin film characterization, one typical method is acoustic resonance spectroscopy, in which the impedance and resonance behavior of a piezoelectric resonator such as quartz is influenced by a deposited thin film [15], [16]. This method is suitable for very thin films ranging in thickness from micrometers to nanometers, and is also useful for identifying the elastic constant and loss of thicker films. In this paper, the acoustic resonance spectroscopy method was applied to determine acoustic properties of 20- to 100-μm polymer films and thin copper tape layers bonded to polyvinylidene fluorude-trifluoroethylene copolymer P(VDF-TrFE) as a piezoelectric resonator. The inherently low impedance of the polymer piezoelectric resonator is easily influenced by additional layers, and the velocity and loss terms of these additional layers were successfully determined by comparison with theoretical curves obtained from a multilayer Mason model which included loss terms in the complex impedance of the constitutive layers. Thus, key parameters required for the design of new multilayer matching and backing absorber layers were determined. The same curve-fitting approach was used to identify the complex parameters of the piezoelectric polymer itself decades ago by Ohigashi [17] and Brown et al. [18], [19]. The basic principle of this non-traditional positioning of materials in an impedance conversion structure, in which the low-impedance layer is closer to the piezoelectric layer and the high-impedance layer is facing the propagation medium, was previously proposed by one of the authors. One proposal involved air ultrasonic transducers, in which a thin air layer was trapped against PZT by a thin suspended polymer film which was in contact with the propagation medium. This structure provided good impedance matching with wideband performance at 40 kHz [20]. Another proposed application was for a P(VDF-TrFE) copolymer transducer using composite matching layers of polyethylene (inner layer, low impedance) and polyester (outer layer, high impedance). This multilayer matching structure was designed so the dynamic resonant impedance, the force divided by vibration velocity at the surface of propagation medium, is minimized as a result of the maximum displacement at the resonance peak. This impedance matching provided higher acoustic output at 1.3 MHz with narrowband continuous excitation into water [21], different from the wideband matching tuned for short-pulse excitation proposed and investigated in this paper. This paper applies the same basic principle to wideband, short-pulse medical imaging transducers, producing designs suitable for low-cost production using commercially available materials, reproducible structures, and a simple design process. II. Measurement of Acoustic Parameters by the Resonance Method Parameters of a piezoelectric material in thin plate form can be evaluated through features of its impedance

vol. 59, no. 2,

February

2012

curve. The ratio of the real and imaginary parts of the input electrical impedance, Zin, is expressed by the impedance angle:

θ = arctan[imag(Z in)/ real(Z in)].

The angle θ is generally close to −90° for any insulator, including piezoelectric materials having capacitive impedance. The vibration which results from the application of an electric field to a piezoelectric material induces a current which adds a resistive component to the impedance and influences θ, causing a deviation from −90° to a less negative angle. When the frequency is swept through resonance, θ shows a peak at the resonant condition. The peak frequency is a function of the thickness, tp, density, ρp, and acoustic velocity, Vs, of the piezoelectric material, and the height of the θ peak and its width are a function of the mechanical loss factor Qp−1 and the electromechanical coupling constant kT [19], [22]. The basic parameters of a piezoelectric plate can be determined through comparison of an experimental θ curve with a theoretical curve by parameter fitting, wherein the material parameters used to create a theoretical curve are altered until the best agreement with the observed curve is obtained. The parameters under a best-fit condition should be those of the piezoelectric materials being used for measurement. When an unknown film is bonded to the surface of a piezoelectric layer whose material parameters have been carefully measured beforehand, the peak frequency, height, and width of the impedance phase angle curve are influenced by density, ρm, acoustic velocity, Vm, and mechanical loss, Qm−1, of the unknown film (the subscript m indicates the added layer). After the addition of the test film, the new impedance curve yields the material parameters of the film layer after comparison to the best-fit theoretical curve. Fig. 1 shows an experimental θ curve (noisy) and its theoretical matching curve (smoother) from a 228-μmthick piezoelectric P(VDF-TrFE) copolymer transducer with thin sputtered electrodes of 200-Å aluminum, which have negligible effect. The peak height measured from the lowest level is 33° and the half-maximum width is 0.28 MHz. The theoretical curve was obtained through Mason’s model, in which the loss terms of dielectric constant and elastic stiffness of the P(VDF-TrFE) film were expressed as complex numbers, as was done by Ohigashi [16] and Brown et al. [17], [18]. The impedance was measured by an impedance spectrum analyzer (HP 4195A, Agilent Technologies Inc., Santa Clara, CA). The parameters of the theoretical curve were adjusted and the resulting best-fit parameters are a nearly unique combination; therefore, these parameters are thought to be those of the piezoelectric copolymer layer as shown in Table I. Material parameters of an additional layer bonded to the copolymer surface were investigated using the following method. When one layer of copper tape was bonded to one surface of the P(VDF-TrFE) copolymer film, the

toda and thompson: detailed investigations of multilayer structures for wideband transducers

Fig. 1. Experimental curve and theoretical impedance angle curve using best fit parameters for 228-μm P(VDF-TrFE) copolymer.

fundamental mode peak was at 2.6 MHz with a second, higher-order, peak appearing at 6.0 MHz which was much stronger than the fundamental peak. Mason model analysis is consistent with this high-frequency copper tape resonance effect which tends to suppress the fundamental peak. This spurious resonance of copper tape disappears when many layers of tape are stacked (see details later). Because the parameters of P(VDF-TrFE) copolymer were determined at 4.8 MHz, and 6.0 MHz is closer to 4.8 MHz than to the 2.6 MHz resonance, the 6-MHz resonance was used to determine the parameters of copper. Fig. 2 shows experimental and theoretical curves of the P(VDF-TrFE) copolymer layer with one layer of copper tape (3M electrical tape with 25-μm adhesive and 38-μm copper, 3M Co., Maplewood, MN) bonded to the surface. The parameters of P(VDF-TrFE) copolymer in the theoretical curve should be close to those already determined by parameter fitting at 4.8 MHz as shown in Table I. The best-fit parameters of the adhesive are also shown in Table I (Q = 3.5 at 6 MHz) along with the parameters used for copper. In Fig. 2, there is a small peak with a height of 1.5°

233

Fig. 2. Experimental and theoretical curves after bonding copper tape on one side of a P(VDF-TrFE) copolymer transducer. The parameters of the theoretical curve were adjusted to best-fit conditions.

at 4.9 MHz. This may be caused by the small lead wire attachment region where the adhesive of the copper tape did not cover the surface and the exposed area should have still the same resonance at 4.9 MHz as shown in Fig. 1. Fig. 3 shows the θ curve for a 104-μm P(VDF-TrFE) copolymer layer (~half of the thickness used in Fig. 1) with 50 μm of polyimide on both surfaces bonded using Epotek 301 epoxy (~2 μm for each bond surface; Epoxy Technology Inc., Billerica, MA). The θ curve measured before adding polyimide layers showed a peak at 11.9 MHz with a resonance shape very similar to that shown in Fig. 1. The best-fit parameters obtained for polyimide films are shown in Table I. The velocity and dielectric constant are a little different for the two copolymer films; these differences result from element-to-element differences and from the frequency dependence of the parameters measured at different resonant frequencies. The frequency-dependent parameters are well known [23]. These parameters are used for the design of matching layers and absorbers, as shown in the next section.

TABLE I. Material Properties. ρ (kg/m3)

Vs (m/s)

d33 (pC/N)

P(VDF-TrFE) (Fig. 1) P(VDF-TrFE) (Fig. 3) PVDF PZT5H Modfied PZT4 Adhesive of Cu tape

1880 1880 1780 7352 7600 1080

2235 2475 2100 4930 4750 1250

15.0 13.1 12.1 190 115

Polyimide Copper Brass

1454 8960 8100

2100 5010 3830

Mechanical Q 17.2 18.1 10.9 48 250 3.5 (6 MHz) 5.4 (3 MHz) 6.0 (0.63 MHz) 17 1000 1000

Dielectric ε′

Dielectric tan δ

t (μm)

5.10 4.4 5.6 2150 870

0.132 0.14 0.29 0.01 0.0042

228 104 112 330 714 25

50

234

IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control ,

vol. 59, no. 2,

February

2012

Fig. 4. Thickness of each layer: (a) designed structure and (b) actual structure. Fig. 3. Experimental (slightly noisy) and theoretical (smooth) θ curves for P(VDF-TrFE) with 50-μm polyimide on both sides. The parameters in the theoretical model were adjusted to get the best-fit condition.



III. Design and Testing of Ultrasonic Transducer Front Impedance Matching A. Multilayer Matching Structure Design Method Design examples for PZT front impedance matching layers are shown in this section. The first example is for 10 mm × 5 mm × 330 μm PZT5H, for which detailed material parameters were obtained through curve fitting and are shown in Table I. The PZT acoustic impedance is Zp = 36.2 MRayl and the resonance frequency is 7.5 MHz. From conventional quarter-wavelength, double matching layer design [24], [25] for radiation medium impedance ZR = 1.5 MRayl, the best acoustic impedance of the first matching layer (close to PZT) is Z1 = Zp4/7ZR3/7 = 9.25 MRayl, and the ideal second matching layer is Z2 = Zp1/7ZR6/7 = 2.36 MRayl. This combination should work well and give wideband performance even without using a backing absorber (air backing). The high impedance of the first matching layer is generally difficult to obtain from commercially available materials; therefore, a mixture of high-density solid powder and polymer binder is typically used to make a composite material with the correct properties. With such materials, it is generally difficult to control thickness and uniformity, and they are not generally feasible for cost-effective mass production. To create the high-impedance layer using the new approach, we can synthesize the equivalent acoustic impedance of ZMK = 9.25 MRayl with thin polymer directly bonded to PZT and a thin metal layer on top of it. The half-wavelength resonance of PZT is 7.47 MHz and the matching layer center frequency was chosen to be the same as described in [11], in which the following relations were used:

Z MK =

MK = 9.25 × 10 6 (1)

f 0 = (1/2π) K /M = 7.47 × 10 6 (2) M = ρ mt m,

K = ρ pV s2/t p. (3)

Here, ZMK is the effective acoustic impedance of the combined polymer-metal layers, M is the unit area mass for the metal layer, K = c33/tp is the unit area spring constant of the polymer layer, ρp is density, Vs = c 33/ρ p is the acoustic velocity used to get (3), and tp is the thickness of the polymer. Note that the spring constant is defined as applied force to the layer divided by displacement. From (1) and (2), K = 4.34 × 1014 was obtained, and the parameters for PVDF shown in Table I are ρp = 1780 kg/ m3 and Vs = 2100 m/s; using these values, a desired thickness of tp = 18.1 μm was obtained from (3). For the metal layer, M = 0.20 kg/m2 and tm = M/ρm = 22.3 μm using ρm = 8960 kg/m3 for copper. The second matching layer was chosen to be homogeneous polyimide with a quarter-wavelength thickness. Thus, the designed structure was 18 μm PVDF/22 μm copper/70.3 μm polyimide as shown in Fig. 4(a). Note that PVDF was chosen here because of its availability. Lower impedance materials such as polyimide can be used to provide the same K and the same end performance by adjusting the layer thickness. The design shown here is an example which uses commonly available metal and polymer films. Higher-frequency designs (N times higher) are possible using thinner layers for all materials (1/N thickness for all layers), including the piezoelectric layer. At much higher frequencies, greater attention must be paid to other aspects of the construction. For example, when the layers become much thinner, the epoxy bonding lines may impact performance and must therefore be considered. In some cases, bonding can be eliminated by using other well-known and well-controlled fabrication and deposition techniques including, for example, spin coating, solution casting, spray coating, chemical vapor deposition, or plasma polymerization. Metal layers may be deposited by direct plating, sputtering, evaporation, etc.

toda and thompson: detailed investigations of multilayer structures for wideband transducers

235

B. Experimental Investigation of Matching Layer The designed structure was fabricated using commercially available materials. The 18.1-μm PVDF film thickness is not generally available, but piezoelectric PVDF of 10 to 11 μm (nominal 9 μm) is available and although thinner than required, the bonding epoxy will increase its effective thickness because the acoustic impedance of epoxy and PVDF are not very much different compared with that of metal. The thickness of copper (22 μm) can be obtained by chemically etching readily available 35-μm material. The second matching layer of ~70.3 μm polyimide can be made by laminating commercially available films of 50 μm and 25 μm with a very thin epoxy layer. The actual fabricated matching structures are shown in Fig. 4(b). The 3-layer matching structure was first laminated with Epotek 301 epoxy, and a total thickness measurement after lamination yielded 3 to 4 μm of epoxy at each interface. Then the PVDF side of this matching layer was bonded to a 330-μm PZT5H plate. Thus, both surfaces of 11-μm PVDF are bonded by a 3- to 4-μm epoxy layer and therefore the effective polymer thickness is close to 18 μm. Fig. 5(a) shows the received waveform from an echo off of a stainless steel reflector block through 9 cm of water after short-pulse excitation from a model 5072PR Panametrics pulser/receiver (Waltham, MA). The fast Fourier transform (FFT) of the waveform is shown in Fig. 5(b) (dotted line). The theoretical curves in Fig. 5(b) were obtained through calculation using Mason’s model and all the thicknesses of the PDVF, copper, and polyimide layers were modified in the model from −20% to +20% in 5% increments about the initially designed thickness combination (0%). The 0% curve shown by the heavy line is closest to the observed FFT curve shown by dots. However, it suggests that the bandwidth can be broadened using thicknesses that are 10% to 15% thinner than the designed thickness. This broader bandwidth is clearly the result of the reduction in the height of the 8.5-MHz peak at −15% and −20% thicknesses. A Mason’s model analysis confirmed that the widest bandwidth is produced with layers that are a little thinner than those predicted by the mass and spring model. This suggests that it is more accurate to do analyses using Mason’s model (a KLM model would yield the same result), but the mass/spring model is a very convenient starting point for detailed design calculations [11]. In Fig. 5(b), all of the theoretical curves were obtained by Mason’s model with material parameters taken from the mass-spring model. The theoretical curve labeled 0% is the closest to the experimental data but the theoretical curve is a little broader than the experimental curve, particularly at higher frequencies. The reason is unknown but may be due to a non-uniform stress strain distribution, or because the actual structure is a little different from the ideal structure design as a result of the bonding layers, or perhaps resulting from some error in the thickness of each layer.

Fig. 5. (a) Received reflection waveform from short pulse excitation. The structure is 330-μm PZT5H with air backing and front double matching of 11-μm PVDF/22-μm copper/75-μm polyimide. (b) The dotted line is the fast Fourier transform of the waveform shown in (a). The frequency response of the 0% theoretical line uses parameters obtained from the mass-spring model.

IV. Investigation of Multilayer Absorber A. Design Parameters and Resulting Impedance of a Multilayer Absorber at ~3 MHz As investigated in the previous paper [11], the acoustic behavior of thin iterated layers of adhesive/metal is essentially the same as wave propagation in many sections of sequentially connected mass-springs. According to the mass-spring model described in [11], the properties of the multilayer absorber can be expressed by its propagation constant, β, acoustic velocity, VMK, and attenuation constant, α, which are given by

236

IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control ,



β = ω/[ω 0(t m + t p)] (m −1) (4)



V MK = (t m + t p)ω/β = (t m + t p) K ′/M (m/s) (5)



α = (ω/2ω 0)(K ′′/K ′)/[t m + t p ] (m −1) (6)



Q −1 = K ′′/K ′. (7)

The polymer layer is a soft adhesive which acts as a spring in the absorber model and M and K ≅ K′ are given in (3), where K′ is the real part and K′′ is the imaginary part of K (the complex nature of K comes from mechanical loss of the adhesive elastic stiffness c33) and it is related to Q of the polymer as shown in (7). The effective acoustic impedance of the absorber is given by (1). This is consistent with the typical acoustic impedance Za of a regular material, Za = [density] × [velocity], where [density] = M/(tp + tm) and [velocity] = (t m + t p) K ′/M yield ZMK = MK.

vol. 59, no. 2,

February

2012

Fig. 6. Theoretical impedance of a multilayer copper tape absorber.

B. Effective Acoustic Impedance of the Multilayer Absorber The effective acoustic impedance of the multiple layer absorber made from stacked copper tape was calculated using a one-dimensional model. The effective acoustic impedance seen from the adhesive side of the structure (the side attached to the PZT) is shown in Fig. 6, where the outermost copper layer is facing air and acoustic waves are reflected back from the copper-air boundary, and the frequency responses of the impedances of 5, 10, and 15 layers are shown. The curves in Fig. 6 show larger peaks and valleys at lower frequencies for structures with fewer layers, whereas at higher frequencies near 4 MHz, the peaks and valleys disappear, the impedance converges to a monotonic value ~4 MRayl, and curves become smooth more quickly for structures with more layers. The constant ratio of copper to adhesive thicknesses, regardless of the number of layers, accounts for the constant effective acoustic impedance. The reason for the decreased amplitude of the peaks and valleys at higher frequencies is that the attenuation constant α is higher at a high frequency and lower at low frequencies and the reflection at the backside air boundary returns to the input surface and influences the impedance. In this case, 37-μm copper and 25-μm adhesive (Q = 5) were assumed and the parameters are shown in Table I. The equivalent acoustic impedance of a multilayer copper tape absorber was experimentally investigated using the following method. At first, the phase angle θ of the input impedance of a PZT plate was calculated assuming various acoustic impedance materials with infinite thickness added to one surface of the plate. It was found that the peak value of the impedance angle decreases with the increasing acoustic impedance of the added material. The measured and calculated results of modified PZT4 are shown in Fig. 7. The calculation was done using Mason’s model with the parameters of modified PZT 4 as shown

Fig. 7. Theoretical impedance angle of modified PZT4 when one surface is loaded by a material with acoustic impedance in the range of 0 to 10 MRayl. The curve with many irregular peaks is the impedance angle observed without surface loading. The dotted line is from 7 layers of copper tape bonded to one surface.

in Table I. The resultant angle curves are shown in Fig. 7 with acoustic impedances ranging from 0 to 10 MRayl loaded at one surface. When an actual multilayer absorber is bonded to a PZT plate, the measured impedance angle peak decreases. This observed peak is compared with a peak height calculated using a best-fit technique, producing the effective acoustic impedance of the multilayer absorber. The impedance angle of modified PZT4 (714 μm thick) was measured with an HP 4195A impedance/spectrum analyzer, and the result is shown in Fig. 7 with theoretical curves for comparison. The experimental curve with many irregular peaks and valleys is typically observed

toda and thompson: detailed investigations of multilayer structures for wideband transducers

for the unloaded condition of this material and its basic curve, neglecting the irregular peaks, should agree with the 0 MRayl theoretical curve. Because this material has very low mechanical loss (Q ~ 250), transverse waves are excited by planer stress resulting from the d31 term and many reflections traverse the plate in planer directions creating random peaks in the spectra. The pattern is further complicated by higher-order resonances of the reflected transverse waves from the edges of the plate. The plate used in this experiment was a side area cut out from a disk (crescent shape), roughly 13 × 5 mm. Exactly rectangular plates and other shapes were tried, but the irregular peaks were generally stronger for exactly rectangular plates; therefore non-straight shapes were chosen. These spurious resonances were not accounted for in the calculations; therefore, the theoretical curves are smooth. When 7 stacked layers of copper tape (25-μm adhesive and 33-μm copper) were bonded to the PZT, the impedance of the PZT was remeasured; the phase angle is shown by dots in Fig. 7. The curve agrees with the theoretical curves for an impedance of 4 MRayl. Therefore, this copper tape stack has an acoustic impedance of approximately 4 MRayl. This measured impedance angle curve was down shifted by 0.05 MHz to provide a better comparison. The reason for the 0.05 MHz shift is not clear, but may be due to inaccurate adhesive thickness measurements. For this structure, when the impedance was calculated from (1)–(3) with the parameters shown in Table I, ZMK = MK = 4.47 MRayl was obtained, which shows fair agreement. Design equations (1) and (3) predict that higher acoustic impedance of the multilayer absorber can be realized by using thinner adhesive layers. To obtain a thinner adhesive layer, a convenient commercial spray adhesive (Elmer’s Products Inc., Columbus, OH) was used. Fig. 8 shows the measured impedance angle of a 330-μm-thick 5 × 10 mm PZT5H plate with a multilayer absorber structure designed for higher impedance. The larger impedance angle curve extending from a base level of −84° to +70° peak is for unloaded PZT. The peak does not have random peaks and valleys as shown in Fig. 8 because of low Qp material (high dampening of the higher-order modes). The curve with the smaller peak is for the same PZT with a multilayer absorber bonded to one surface. This absorber was made from 10 sheets of 1-cm2, 38-μm-thick stainless steel bonded together using a spray adhesive. All of the sheets were sprayed with the spray nozzle at a distance of ~15 cm while moving the nozzle quickly to obtain a coating that was uniform to the eye. The sheets were stacked and put under 9 tons of force at 90°C. From the total finished thickness, each adhesive layer was determined to be 2.5 μm thick. For the purpose of theoretical estimation, we felt it was appropriate to assume that the spray adhesive and the copper tape adhesive were acoustically similar and that modified-PZT4 and PZT5H were very similar. In Fig. 8, the impedance angle of a PZT5H plate with the multilayer absorber on one surface shows a peak at 7.4 MHz with a

237

Fig. 8. Impedance angles of 330-μm PZT5H. The broader and narrower curves correspond to with and without a multilayer absorber. The adhesive layers are 2.5 μm thick (applied by spraying) laminated between 10 layers of 38-μm stainless steel.

−40° height. The peak position of 7.4 MHz is 0.5 MHz higher than the peak at the unloaded condition. Applying these results to the impedance angles of modified PZT4 shown in Fig. 7, the acoustic impedance of this stainless steel/spray adhesive absorber is estimated to be roughly ~10 MRayl (neglected 0.5 MHz shift). The effective impedance of the stainless steel/spray adhesive absorber was calculated to be 14.7 MRayl using (1)–(3), and using the parameters of the adhesive of copper tape in Table I along with the density of stainless steel (7910 kg/m3). Therefore, it seems clearly possible to obtain a very high acoustic impedance using this type of structure, an impedance similar to that of 50-μm-diameter tungsten power in a 40% volume ratio with epoxy, the highest value obtained by Grewe et al. [12]. C. Experimental Investigation of the Velocity and Attenuation of a Multilayer Absorber Commercially available copper tape (copper thickness ranging from 33 to 38 μm; 3M #1181, 3M Co.) was investigated for propagation constant, attenuation, and acoustic impedance for use as an ultrasonic absorber. The copper tape was laminated into a 50-layer stack and adhered to the front surface of a 28-mm-diameter ultrasonic NDT transducer (V102, 1.0 MHz, Panametrics). Another NDT transducer with a 16 mm diameter (V103 0.5/1.0 MHz, Panametrics) was coupled through ultrasonic gel to the other surface of the stack. The larger NDT transducer was excited by pulses from a Panametrics 5072PR pulser/ receiver, the smaller transducer was used as receiver, the transmitted signals were amplified by the receiver amplifier, and the signal was recorded. Another 10 layers were then added to the stack and the signal was recorded again; then further additional layers were added to achieve 70, 80, and 90 total layers, which were sequentially measured.

238

IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control ,

vol. 59, no. 2,

February

2012

Fig. 9. Received signal through multiple layers of copper tape (25 μm of adhesive and 33 μm of copper in each layer). Stacks containing from 50 to 90 layers are shown with amplitude and delay time.

Fig. 10. Received signal through 5, 10, 15, and 20 layers of copper tape (25 μm adhesive and 37 μm copper). Lowest trace for 5 layers, upper most trace for 20 layers.

All results are shown together in Fig. 9. From the decreasing peak voltage and the signal delay, propagation velocity and attenuation constants were computed. The frequencies calculated from the initial cycle periods were 668, 674, 574, 635, and 603 kHz for the different absorber structures and the average frequency was 630 kHz. This frequency comes from the overall frequency response of the two NDT transducers, one of which was excited by a sharp pulse with a broad frequency spectrum. From the variation of the zero-crossing point of the initial cycles in Fig. 9, the propagation velocity was determined to be 847 m/s and the theoretical value is 870 m/s, which shows fair agreement. If we apply this absorber to a 3.2-MHz PZT transducer, the attenuation constant should be larger in proportion to frequency as shown in (6) assuming K′′/K′ = 1/Q constant. From Fig. 9, the amplitude ratio of 90 layers to 50 layers is 0.42 with a 40-layer difference (58 μm of copper tape) and exp {−40 × 58 × 10−6α} = 0.42 gives α = 374/m. This creates an attenuation of 32.4 dB/cm at 630 kHz, equating to 165 dB/cm at 3.2 MHz if K′′/K′ is constant. Note that the measurement of attenuation using this method eliminates the coupling loss at the transducer surface. The theoretical value of the attenuation constant is α = 646/m using f = 630  kHz, f0 = (1/2π) K /M = 2.40  MHz and K′′/K′ = 1/Q = 1/3.5. The theoretical α is 1.73 times the observed α value. The discrepancy is likely caused by using the Q value previously measured at 6 MHz; the actual Q value at 630 kHz in this experiment is apparently higher than the value of 3.5 shown in Table I, and was deduced to be Q = 6.0 at 630 kHz. Another example of absorption by copper tape layers measured at a higher frequency was investigated. Copper tape stacks of multiple layer thicknesses were bonded to an NDE transducer (V180, 29 mm diameter, 3.0 to 3.5 MHz, Panametrics), which was used as a transmitter

in pulse mode excited by a Panametrics 5072PR pulser. The signal was received by a modified PZT4 disk (28 mm diameter × 0.714 mm, 3.2 MHz without matching layers or a backing absorber), to which the copper layer at the top of the stack was coupled using ultrasonic gel. The received signals through 5, 10, 15, and 20 layers of copper tape are shown in Fig. 10. The reason for it exhibits longer ringing than a conventional NDE transducer pair is the use of the undamped PZT plate as a receiver (sharp resonance). The initial peak at t = 0 μs is direct feedthrough caused by the non-shielded PZT plate. The propagation velocity was determined as follows. All waveforms were triggered by the pulser. The starting time of the delayed receive waveforms (the zero-crossing times as shown by the arrow in Fig. 10) and the times of the sequential positive and negative peaks of the waveforms were recorded as a function of the number of layers as shown in Fig. 11, where the bottom trace (triangle plot) is the starting time of the received signal, the line above it is for positive peaks of first cycle of four different layer numbers, the line further above is the time for the first negative peak, and the other lines are for subsequent positive and negative peaks plotted in a similar manner. By this method, the errors resulting from the electronic delay, transducer delay, and trigger timing were eliminated, which are commonly included in all of the timing data. The error is −0.1 μs, as shown by the intersection of the ordinate and the extrapolated line which should be zero if there are no errors. Also, the first half-cycle in Fig. 10 has lower-frequency component of ~1.2 MHz and successive ringing has higher-frequency component of ~3.2 MHz; the velocities at these different frequencies can be separately determined, but actually the difference was very small as indicated by the roughly parallel lines for all times. From the average gradient of each plot in Fig. 11 the velocity was determined to be 928 m/s using a 0.93 mm

toda and thompson: detailed investigations of multilayer structures for wideband transducers

239

a 29% fill ratio [12]. Because the data from the multilayer absorber is at 3 MHz, to compare with the 5-MHz data in [12], the attenuation must be multiplied by the frequency ratio 1.67 because α ∝ ω from (6), and the resulting attenuation is expected to be much higher than that of the tungsten- and PZT-filled absorbers. The calculated attenuation of 10 layers of 2.5-μm spray adhesive and 38-μm stainless steel used to create the data shown in Fig. 8 is 392 dB/cm at 6.8 MHz, assuming the parameters of adhesive on copper tape shown in Table I and Q = 3.5 (measured at 6.0 MHz). D. Application of a Multilayer Absorber to an Ultrasonic Transducer

Fig. 11. The observed time of waveform of Fig. 10: Zero-crossing points of first rise (triangle), first positive peaks (circle), first negative peaks (×), second positive peaks (+), following negative peaks (*), and third positive peaks (dot in circle). The gradient corresponds to propagation velocity.

difference between 5 sheets and 20 sheets (25 μm of adhesive and 37 μm of copper for each layer) and a 1.003 μs averaged delay. This value is close to theoretical value of 889 m/s calculated using (5). The attenuation was determined by the amplitude of the second cycles (3.0 MHz) of Fig. 10 because the initial cycle has somewhat lower frequency content. The amplitude ratio is 0.0925 for a thickness of 15 layers with total 0.93 mm. From exp{−α × 0.93 × 10−3} = 0.0925, α = 2559/m and the attenuation at 3.0 MHz is 222 dB/cm. This is roughly consistent with, but higher than, the previously predicted α value at 3.2 MHz, which was based on the measured 630 kHz value. The theoretical value in this case is α = 3940/m or 342 dB/cm (assuming Q = 3.5) at 3 MHz, 1.54 times the observed α value at 3 MHz. The consistently lower observed α values both at 630 kHz and 3 MHz suggests that the Q of 3.5 measured with a similar structure at 6 MHz does not accurately predict α at lower frequencies. The observed α at 3.0 MHz corresponds to a Q of 5.4 using (6) and (7). We attribute the disagreement between the loss terms of these structures to variations in material properties, irregularities in adhesive thickness, and a range of metal foil thicknesses. Clearly, the multilayer metal-adhesive iterative structure has enough attenuation to function well as an acoustic absorber at these frequencies. Grewe et al. investigated the attenuation behavior of various sizes of tungsten or PZT particles embedded in various polymers at 5 MHz and found the attenuation to be 40 to 60 dB/cm when particle size is 1 to 5 μm. However when the larger particles (52-μm tungsten) are used, it produces 247 and 178 dB/cm at 36% and 40% volume percentage, and 75-μm PZT particles had 178 dB/cm with

In the general case, a backing absorber is required when the performance of the front matching structure is nonideal, which is often the case because of limitations in available materials or difficulty in the proper design and fabrication of composite materials. In such cases, resonance is formed by reflections from the front and back boundaries. If reflections from the front boundary are zero as a result of the use of an ideal front matching structure, performance is improved because resonance cannot be formed solely from reflections from the back boundary. The multilayer absorber was tested on a specific transducer using 714-μm thick modified PZT-4 (centered at 3.3 MHz). As an example of an insufficient front matching for this PZT, a quarter-wavelength 170-μm PVDF layer was tested, which was composed of 115-μm and 55-μm layers bonded together by 5-μm epoxy for a 3.1-MHz design in which piezoelectricity was not used. The backing was air. In this design, PVDF has acoustic impedance of 3.9 MRayl which is lower than the required value of 7.4 MRayl obtained from conventional matching layer impedance (geometrically averaged values of PZT and water). An acoustic wave was launched into water from this transducer using a sharp-pulse excitation (Panametrics 5072 RR) and a reflection signal from a stainless steel block was recorded, as shown by the larger signal with more ringing in Fig. 12. Next, 10 layers of copper tape were used as backing absorber, but the performance was not impressive and the bandwidth did not become broad enough to give a sufficiently short pulse length. This is because the front matching of the PVDF layer is not satisfactory and the impedance of the absorber was not high enough to suppress the reflection from the back boundary of PZT. Therefore, to provide a better impedance match at the back surface, a back matching layer was designed and inserted between the multilayer absorber and PZT. A signal was then recorded in similar fashion and is shown as the smaller signal with less ringing in Fig. 12. The design details of the back matching structure are as follows: The equivalent acoustic impedance of iterative copper tape (25 μm of adhesive and 37 μm of copper) is given by (1) using the value of MK = 4.72 MRayl from the adhesive K = 6.75 × 1013 N/m2 and from the copper M = 0.33 kg/m2. The back matching structure is 50-μm poly-

240

IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control ,

vol. 59, no. 2,

February

2012

responds to the matched copper tape absorber and the narrower bandwidth to air backing. For both cases, PVDF was used as a single front matching layer. To design the absorber for higher frequency operation, all of the layers must be reduced in thickness by an amount inversely proportional to the frequency increase, and different film and metal deposition processes may be required.

V. Discussions on Process and Performance

Fig. 12. Waveforms of transducers of modified PZT4 with front PVDF quarter-wavelength matching for both. One is with air backing and the other is with a back matching structure inserted between PZT and the multilayer absorber. The time scale of the waveform with air backing is shifted by 30 μs to the right for clarity.

Fig. 13. Comparison of frequency response curves of a modified PZT4 transducer with air backing (narrower) and an absorber backing (broader) having a matching layer of polyimide and brass inserted between PZT and the absorber structure. Front matching is quarter-wavelength PVDF for both cases.

imide against the PZT and 50-μm brass on the absorber side. This back matching layer has an equivalent acoustic impedance of MK = 7.2 MRayl and a center frequency of (1/2π) K /M = 2.83 MHz, which converts acoustic impedance of the multilayer absorber to 11 MRayl at the back of the PZT. The difference of the two waveforms in Fig. 12 (time scale of the larger signal was shifted right by 30 μs for clarity) shows the effect of absorber with back matching structure. The frequency response curves after FFT are shown in Fig. 13, where a wider bandwidth response cor-

Thin copper electroplated onto polyimide film is a commercially available structure commonly used in the manufacture of flexible printed circuit boards, and is a material clearly suitable for low-cost mass production. If a non-commercial thickness of copper is required for a particular ultrasound application, suitable thicknesses can be produced either by thinning the copper through a standard chemical etching process, or by adding copper using a standard copper plating process. Lamination of various layers in these stacks is possible using epoxy, which can also be used to adjust the effective thickness of designed polymer layers because most polymer films and epoxy layers have similar acoustic properties and their combined total thickness behaves as single polymer layer. In these cases, the thickness of the epoxy must be controlled to get the desired end thickness but accurate control requires a careful process. Many epoxies used for general bonding purposes have a high viscosity and typical laminated layer thicknesses range from 30 to 50 μm. If pressure is applied during the lamination process, 20-μm epoxy layers are easily achieved; 10-μm layers are difficult but possible depending on the area of lamination (smaller areas tend to become thinner because epoxy outflow has a shorter distance to travel). Therefore, thin bond layers require the application of thin layers of low viscosity epoxy and the application of pressure. Applying heat can also further lower the viscosity for the initial part of the lamination period and help make the bond lines thinner. A typical low viscosity epoxy is Epotek 301; by using this material on small lamination areas, it is possible to make 2- to 5-μm bond layers. Transducer bandwidth depends on the acoustic impedance and thickness of component films used to construct matching and absorber structures, and although accurate designs are possible, the actual impedance and bandwidth vary because the thickness tolerances of these layers is typically ±10%. From our experience, non-uniformity in the epoxy layer thicknesses in the matching layer structure does not strongly influence the performance, although we speculate that it may alter bandwidth. Interestingly, nonuniform adhesive thickness in the absorber is expected to actually increase the bandwidth and attenuation of the structure. These trends are useful topics for future investigation, especially regarding the consistency of behavior in a high-volume mass-production process.

toda and thompson: detailed investigations of multilayer structures for wideband transducers

As shown in the basic design theory in the previous paper [11], an arbitrary matching layer acoustic impedance can be designed by the combination of metal and polymer layers with specific thicknesses. These structures have the advantages of lower cost and fabrication complexity. However, as these commercially available polymer and metal films are designed for other functions entirely, material thicknesses may be limited. In higher-frequency applications requiring very thin layers, performance may be influenced by bonding layers (e.g., epoxy) and other artifacts. Therefore, in this work readily available materials were used to operate in the 3 to 7 MHz range to demonstrate the applicability and effectiveness of the metal-polymer layer principle. Overall, however, production designs can achieve lower costs and improved performance using combinations of metal and polymer layers with specifically optimized thicknesses. VI. Conclusion Experimental investigations of polymer/metal impedance converters and adhesive/metal absorbers showed that these structures could be used in a practical manner as ultrasonic impedance matching and absorptive backing layers. Simple equations were presented for the design of these structures and for calculating their effective acoustic impedances and center frequencies. These equations, developed for polymer/metal structures, can be used in a manner similar to the more common quarter-wavelength matching layer equations to calculate the impedance conversion ratio. The calculated acoustic impedances and propagation velocities agreed well with experimental observation. The measured attenuation constants were consistent with the theory, assuming that the loss term K′′/K′ = Q−1 increases with frequency. Importantly, the observed attenuation values for a multilayer copper tape absorber and for a layered spray adhesive/metal sheet structure are comparable or higher than the most attenuative tungsten–polymer mixture. The impedance of the absorber can be designed by choosing the thickness ratio of adhesive to metal. References [1] J. H. Goll and B. A. Auld, “Multilayer impedance matching schemes for broadbanding of water loaded piezoelectric transducers and high Q electric resonators,” IEEE Trans. Sonics Ultrason., vol. SU-22, no. 1, pp. 52–53, 1975. [2] H. P. Beerman, “Optimizing matching layers for a three-section broad-band piezelectric PZT-5A transducer operating into water,” IEEE Trans. Sonics Ultrason., vol. SU28, no. 1, pp. 52–53, 1981. [3] H. W. Persson and C. H. Hertz, “Acoustic impedance matching of medical ultrasound transducers,” Ultrasonics, vol. 23, no. 2, pp. 83–89, 1985. [4] Q. C. Xu, C. Madhavan, T. T. Srinivasan, S. Yoshikawa, and R. E. Newnham, “Composite transducer with multiple piezoelectric matching layers,” in 1988 IEEE Ultrasonics Symp. Proc., pp. 507–512. [5] J. S. Kenney and W. D. Hunt, “Acoustic matching network synthesis using discrete space Fourier transforms,” in 1990 IEEE Ultrasonics Symp. Proc., pp. 581–586.

241

[6] B. Hadimioglu and B. T. Khuri-Yakub, “Polymer films as acoustic matching layers,” in 1990 IEEE Ultrasonics Symp. Proc., pp. 1337–1340. [7] S. Thiagarajan, R. W. Martin, A. Proctor, I. Jayawadena, and F. Silverstein, “Dual layer matching (20 MHz) piezoelectric transducers with glass and parylene,” IEEE Trans. Ultrason. Ferroelectr. Freq. Control, vol. 44, no. 5, pp. 1172–1174, Sep. 1997. [8] S. Rhee, T. A. Ritter, K. K. Shung, H. Wang, and W. Cao, “Materials for acoustic matching in ultrasound transducers,” in 2001 IEEE Ultrasonics Symp. Proc., pp. 1051–1054. [9] A. Lutsch, “Solid mixtures with specified impedances and high attenuation for ultrasonic waves,” J. Acoust. Soc. Am., vol. 34, no. 1, pp. 131–132, 1962. [10] G. Kossoff, “The effect of backing and matching on the performance of piezoelectric ceramic transducers,” IEEE Trans. Sonics Ultrason., vol. SU-13, no. 1, pp. 20–30, Mar. 1966. [11] M. Toda and M. Thompson, “Novel multi-layer polymer-metal structures for use in ultrasonic transducer impedance matching and backing absorber applications,” IEEE Trans. Ultrason. Ferroelectr. Freq. Control, vol. 57, no. 12, pp. 2818–2827, Dec. 2010. [12] M. G. Grewe, T. R. Guraja, T. R. Shrout, and R. E. Newnham, “Acoustic properties of particle/plolymer composites for ultrasonic backing applications,” IEEE Trans. Ultrason. Ferroelectr. Freq. Control, vol. 37, no. 6, pp. 506–514, Nov. 1990. [13] H. Wang, T. Ritter, W. Cao, and K. K. Shung, “High frequency properties of passive materials for ultrasonic transducers,” IEEE Trans. Ultrason. Ferroelectr. Freq. Control, vol. 48, no. 1, pp. 78–84, Jan. 2001. [14] S. Lees, R. S. Gilmore, and P. R. Krantz, “Acoustic properties of tungsten–vinyl composites,” IEEE Trans. Sonics Ultrason., vol. SU20, no. 1, pp. 1–2, Jan. 1966. [15] H. Ogi, K. Satoh, T. Asada, and H. Masahiko, “Complete mode identification for resonance ultrasound spectroscopy,” J. Acoust. Soc. Am., vol. 112, no. 6, pp. 2553–2557, Dec. 2002. [16] N. I. Polzikova, G. D. Mansfeld, S. G. Alekseev, I. M. Kotelyanskii, and F. O. Sergeev, “Acoustic resonance spectroscopy of nanoceramics,” in 2008 IEEE Ultrasonics Symp. Proc., pp. 2169–2172. [17] H. Ohigashi, “Electromechanical properties of polarized polyvinylidene fluoride films as studied by piezoelectric resonance method,” Jpn. J. Appl. Phys., vol. 47, no. 3, pp. 949–955, 1976. [18] L. F. Brown and J. L. Mason, “Disposable piezoelectric ultrasonic transducers for non-destructive testing applications,” IEEE Trans. Ultrason. Ferroelectr. Freq. Control, vol. 43, no. 4, pp. 560–568, 1996. [19] L. F. Brown and D. C. Carlson, “Ultrasound transducer model for piezoelectric polymer films,” IEEE Trans. Ultrason. Ferroelectr. Freq. Control, vol. 36, no. 3, pp. 313–318, 1989. [20] M. Toda, “New type of matching layer for air-coupled ultrasonic transducers,” IEEE Trans. Ultrason. Ferroelectr. Freq. Control, vol. 49, no. 7, pp. 972–979, Jul. 2002. [21] M. Toda, “Narrowband impedance matching layer for high efficiency thickness mode ultrasonic transducers,” IEEE Trans. Ultrason. Ferroelectr. Freq. Control, vol. 49, no. 3, pp. 299–306, Mar. 2002. [22] H. Ohigashi, K. Koga, M. Suzuki, T. Nakanishi, K. Kimura, and N. Hashimoto, “Piezoelectric and ferroelectric properties of P(VDFTrFE) copolymers and their application to ultrasonic transducers,” Ferroelectrics, vol. 60, no. 1, pp. 263–276, 1984. [23] K. Kimura and H. Ohigshi, “Generation of very high frequency ultrasonic waves using thin films of vinylidene fluoride-trifluoroethylene copolymer,” J. Appl. Phys., vol. 61, no. 10, pp. 4849–4754, 1987. [24] C. S. Desilets, J. D. Fraser, and G. S. Kino, “The design of efficient broad-band piezoelectric transducers,” IEEE Trans. Sonics Ultrason., vol. SU-25, pp. 115–125, May 1978. [25] J. W. Hunt, M. Arditi, and F. S. Foster, “Ultrasound transducers for pulse echo medical imaging,” IEEE Trans. Biomed. Eng., vol. BME30, no. 8, pp. 453–481, Aug. 1983. Minoru Toda was born in Kyoto, Japan. He received the B.S.E.E. degree from Shizuoka University in 1960 and the Ph.D. degree in physics from Tokyo University in 1968. He joined RCA Research Laboratories, Tokyo, Japan, in 1962 as a member of the technical staff and became a manager of the device research group in 1970, where he has done research in microwave oscillation and propagation in semiconductors, microwave application of magnetic thin

242

IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control ,

films and magnetic semiconductors, and piezoelectric devices. In 1982, he was transferred to RCA David Sarnoff Research Center, Princeton, NJ, where he did research in surface acoustic waves, semiconductor lasers, and quantum structures. In 1990, he joined Atochem Sensors, which became a division of Measurement Specialties Inc., where he has developed a number of new piezoelectric devices for actuators, sensors, medical ultrasonics, and airborne ultrasonic transducers. Now he holds the position of principal innovator. He was awarded four RCA Laboratories Outstanding Achievement Awards. He has authored more than 50 technical papers and he was granted 57 US patents. He is a member of IEEE and the Acoustical Society of America.

vol. 59, no. 2,

February

2012

Mitch Thompson (M’99) received the B.S. degree in meteorology and the M.S. degree in mechanical engineering from The Pennsylvania State University in 1977 and 1983, respectively, and the Ph.D. degree in mechanical engineering from Drexel University, Philadelphia, PA, in 2002. In 1983, he joined IBM, where he worked on advanced photolithography equipment development. He joined Pennwalt Corporation in 1986, and until 2008 was focused on the development and commercialization of piezoelectric polymer sensor technology. He held engineering and management positions in the design and manufacturing of sensors and transducers and is currently the Chief Technology Officer for Measurement Specialties Inc., Hampton, VA.