Статьи на перевод PVDF_P(VDF-TrFE) / Piezoelectric Composites for Sensor and Actuator Applications

Abstract—In the last 25 years, piezoelectric ceramicpolymer composites have been conceptualized, prototyped, fabricated, and implemented in an array of applications encompassing medical imaging and military missions, among others. A detailed snapshot of the materials used, and a detailed account of the major innovative methods developed in making various piezoelectric ceramic-polymer composites are presented. The salient aspects of processing of such composites are summarized, and structure-processing-property relations are described using connectivity as the unifying central concept. Computer-aided design (CAD)-based fabrication methods, which result in composites whose structural complexity surpass that of composites obtained with traditional methods, are described to introduce the reader to novel concepts in processing of piezocomposites. A brief survey of some recent advances made in modeling of (0-3), (1-3), and (2-2) composites also is provided.

I. Introduction

Today’s transducer technology demands the use of advanced functional materials. Such functional materials are required to perform several tasks that produce a

useful correlation or feedback mechanism in a transducer system. Specifically, in the case of piezoelectric or electrostrictive materials, one desires to harness sensing and actuation functions concomitantly for a given transducer design.

A piezoelectric/electrostrictive sensor converts a mechanical variable (displacement or force) into a measurable electrical quantity by means of piezoelectric/electrostrictive e ect. Alternatively, the actuator converts electrical signals into the useful displacement or force. Because piezoelectrics and electrostrictors inherently possess both direct (sensor) and converse (actuator) effects, they are considered to be functional materials in the sense indicated above.

The growth of the transducer market has been rapid, and the predictions indicate a steady growth rate in the first decade of the 21st century. The sensor market alone increased to $5 billion in 1990, and projections of a $13 billion worldwide market already have been made for this decade [1]. Piezoelectric/electrostrictive sensors and actuators make up a significant portion of the transducer market with applications in automotive, active vibration damping, and medical imaging industries.

Manuscript received January 14, 2004; accepted July 28, 2004. The authors are with the Department of Ceramic and Ma-

terials Engineering and Malcolm McLaren Center for Ceramic Research, Rutgers University, Piscataway, NJ 08854 (e-mail: safari@rci.rutgers.edu).

The design and fabrication of composite materials enable one to optimize electrical, magnetic, and mechanical forces for special applications. Thereby, properties that cannot otherwise be achieved by the end members alone can be synthesized by the use of several connectivity patterns. It follows from this line of reasoning that the drive for the rapid development of piezoelectric composite materials stems primarily from the need to combine desirable material properties to maximize the input and output response of electromechanical transducer systems. For instance, in an electromechanical transducer, one may wish to maximize the piezoelectric sensitivity, minimize the density to obtain a good acoustic matching with water, and make the transducer mechanically flexible to conform to a curved surface. Because these properties are mutually exclusive, there is no single-phase material that simultaneously satisfies these requirements. Thus, in many applications, one might meet conflicting design requirements by combining the most useful properties of two or more phases that do not otherwise appear together in nature.

In this paper, we focus on the development of piezoelectric ceramic-polymer composites and provide a detailed account following the time line of its historical evolution in the context of connectivity. As a prelude to the main discussion presented herein, the most popular piezoelectric/electrostrictive materials also are touched upon briefly. In what follows, particular emphasis is placed on the processing of such composites, and the most salient processing methods are described in detail along with the properties of the resultant composites. Novel computeraided design (CAD)-based fabrication methods also are introduced as their applications to the processing of piezocomposites provides access to unprecedented structural complexity in such composites that is otherwise not realizable with traditional methods. A brief survey of some recent advances made in modeling of (0-3), (1-3), and (2-2) composites also are provided.

II.Piezoelectric-Ferroelectric Materials

A. Piezoelectric Ceramics and Single Crystals

The largest class of piezoelectric ceramics consists of mixed oxides containing corner-sharing octahedra of O2− ions. The largest structure type of this kind is the perovskite family [2]. Piezoelectric ceramics having the perovskite structure include, but are not limited to, barium titanate (BaTiO3), lead titanate (PbTiO3 or PT), lead

zirconate titanate (Pb(ZrxTi1-x)O3, or PZT), lead lan-

thanum zirconate titanate (Pb1-xLax(ZryT1-y)1-x/4O3, or PLZT), and lead magnesium niobate (Pb(Mg1/3Nb2/3)O3,

or PMN).

After the discovery of ferroelectricity in BaTiO3 ceramics in the 1940s [2] in various parts of the world independently, a large array of ceramic ferroelectric materials have been developed [2]–[4]. Especially, PZT, a binary solid solution of lead zirconate (PbZrO3, denoted as PZ, an antiferroelectric), and lead titanate (PT, a ferroelectric) has found widespread use because of its outstanding piezoelectric properties [2]–[4].

A number of single crystal materials also exhibit piezoelectricity. A list of single crystal piezoelectric materials includes quartz (SiO2), lithium niobate (LiNbO3) and lithium tantalate (LiTaO3), ammonium dihydrogen sulfate, lithium sulfate monohydrate, and Rochelle salt. These materials dominate certain applications, such as frequency-stabilized oscillators in watches and radars, and surface acoustic wave devices in television filters and analog signal correlators [2]–[4]. There have been attempts to grow single crystals of PZT; however, the resulting crystals were small and of poor quality for testing [5], [6]. Also, there has been a lot of research on ternary relaxor-PT type of materials [5]–[8]. In, especially, automotive and aerospace applications, there is a growing need for actuators and sensors that allow broad temperature range operation. As a consequence, there has been a renewed interest in systems with high Curie temperature with special emphasis on single crystals. In that regard, materials such as lead metaniobate (PbNb2O6) [9], [10], or relaxor systems such as Pb(Sc1/2Nb1/2)O3- PbTiO3 [11]–[15], Pb(In1/2Nb1/2)O3-PbTiO3 [16], and Pb(Yb1/2Nb1/2)O3-PbTiO3 [16]–[19] have been investigated extensively. Recently, there has been a lot of interest in high-Curie temperature ferroelectric materials such as (1-2x)BiScO3-xPbTiO3 [20]–[24]. In Table I, a brief comparison of the electromechanical properties of some important ferroelectric materials is provided, which clearly shows that single crystal ferroelectrics exhibit far superior properties than their polycrystalline counterparts. Major strides in the implementation of single crystal piezoelectric

Fig. 1. Schematic of PVDF phases and polarization directions. (Courtesy of T. F. McNulty)

materials in sensor and actuator applications, in which the cost associated with single crystals can be justified, have taken place in the past 3 years. However, several problems such as compositional homogeneity and reproducibility remain to be solved.

B. Piezoelectric Polymers

The piezoelectric behavior of polymers was first reported in 1969 [25], [26]. The behavior results from the crystalline regions formed in these polymers during solidification from the melt. The most widely known piezoelectric polymers are polyvinylidene fluoride, also known as PVDF, polyvinylidene fluoride-trifluoroethylene copolymer [25], or P(VDF-TrFE) [26], [27], and odd number nylons, such as Nylon-11 [26].

The PVDF consists of a carbon-based chain with alternating hydrogen and fluorine units attached to the carbon chain. The chemical formula is [CH2CF2]n [3]. Of the four distinct phases that PVDF can assume, the β-phase is the most common phase that exhibits a spontaneous polarization, and thus piezoelectricity. Stretching the film prior to poling forms this phase. The unit cells of the four phases of PVDF are shown schematically in Fig. 1. As can be seen from Fig. 1, all phases but the α-phase exhibit a net dipole moment. The γ-phase quickly converts to β-phase upon stretching. The αp-phase is formed by poling α-phase at very high electric fields [3].

TABLE II

Electromechanical and Dielectric Properties of Some

Piezoelectric Polymers.

The electromechanical properties of piezoelectric polymers are significantly lower than those of ceramics. The d33 values for PVDF and P(VDF-TrFE) are approximately 33 (×10−12 C/N), and dielectric constant K33 is in the range 6–12. They both have a coupling coe cient (k) of 0.20, and a Curie point (T0) of approximately 100◦C. For Nylon11, d33 is about 2 × 10−12 C/N, and k is approximately 0.11 [26], [27]. They have acoustic impedance close to that of the body and high values for the voltage coe cients, and they are flexible and conformable to any shape. Their limitations for use as transducers include: a low dielectric constant (K) and coupling coe cient (kt ), high electrical losses, and the di culty in poling thick samples. The polymers also have a low Curie point, and the degradation of the polymer occurs at low temperatures (70–100◦C) [28]– [31]. In spite of the drawbacks, the discovery of a relaxortype behavior with a high electric field induced strain in certain polyurethane elastomers have generated a lot of interest for further research in this area [32].

Copolymers containing PVDF and trifluoroethylene (TrFE) have received particular attention in recent years. The P(VDF-TrFE) forms the β-phase upon cooling, thus stretching of the film is unnecessary. Because the material is ferroelectric immediately upon cooling from the melt, complex shapes can be manufactured using common polymer processing techniques [32]. A summary of the electromechanical properties of some piezoelectric polymers is given in Table II.

C. Piezoelectric Ceramic/Polymer Composites

As mentioned above, a number of single crystal, ceramic, and polymer materials exhibit piezoelectric behavior. In addition to the monolithic materials, composites of piezoelectric ceramics with polymers also have been formed. Table III summarizes the advantages and disadvantages of each type of material [33]. Ceramics are less expensive and easier to fabricate than polymers or single crystals. They also have relatively high dielectric constants and good electromechanical coupling. However, they have high acoustic impedance, and, therefore, are a poor acoustic match to water, the media through which it is typically transmitting or receiving a signal. Also, because they are sti and brittle, monolithic ceramics cannot be formed onto curved surfaces, limiting design flexibility in the transducer. They also have a high degree of noise associated with their resonant modes. Piezoelectric polymers

TABLE III

Comparison of Properties Pertinent to Transducer

Applications of Piezoelectric Ceramic, Polymer and

Piezoelectric Ceramic/Polymer Composites.1

1Courtesy of T. R. Gururaja.

2(+) = favorable property, (−) = unfavorable property.

are acoustically well matched to water, are very flexible, and have few spurious modes. However, applications for these polymers are limited because of their low electromechanical coupling, low dielectric constant, and high cost of fabrication. Piezoelectric ceramic/polymer composites have shown superior properties when compared to singlephase materials. As shown in Table III, they combine high coupling, low impedance, few spurious modes, and an intermediate dielectric constant. In addition, they are flexible and moderately priced. A more detailed account of the processing and electromechanical properties of ceramicpolymer piezoelectric composites will be provided in the following sections.

III.Piezoelectric Composites

A.Basic Principles

An electromechanical transducer is a device that converts electrical energy into mechanical vibration and vice versa by using piezoelectricity or electrostriction [2]. The direct piezoelectric e ect enables a transducer to function as a passive sound receiver, or as a pickup by the conversion of acoustic energy into an electrical signal. Applications in which such electromechanical transducers are used include hydrophones for under water low frequency noise detection and microphones. However, the converse piezoelectric e ect permits a transducer to act as an active sound transmitter or loudspeaker.

A transducer also can perform both active and passive functions nearly simultaneously in the so-called pulse-echo mode. When so operated, a transducer element not only emits an acoustic wave into a given medium, but also senses the faint echoes reflected back to it. Echoes are produced when a sound wave strikes a boundary between two substances possessing di erent characteristic acoustic impedances. The strength of the echo is proportional to the acoustic impedance mismatch between the two interface materials.

A hydrophone is a passive listening device used to detect relatively low frequency (<40 kHz) noise under water. The state of stress is considered to be e ectively hy-

drostatic as the wavelengths of sounds in this frequency range are much larger than the transducer dimensions [34]. The voltage produced under hydrostatic pressure is used to measure the sensitivity of a hydrophone. In this regard, a useful parameter in evaluating piezoelectric materials for use in hydrophones is the voltage coe cient gh, which relates electric field to the applied hydrostatic strain. Another parameter of paramount importance is the hydrostatic strain coe cient dh, which describes polarization resulting from a change in stress. The gh coe cient is related to dh by the relation gh = dh/(ε0 K) where K is the relative permittivity, ε0 is the permittivity of free space (8.854×10−12 F/m), and dh = d33 +2d31 with d33 > 0 and d31 < 0. Here, d33 and d31 are the longitudinal and transverse piezoelectric coe cients, respectively. A useful figure of merit (FOM) for hydrophone materials is the product of the voltage coe cient and the hydrostatic strain coe - cient dhgh in m2/N [35]. Recently, a new FOM has been proposed to take into account the dielectric losses of the composite, which is given by (dh gh/ tan δ) [36]. Other desirable properties for hydrophone materials include but are not limited to: low density for good acoustic impedance matching with water; little or no variation of dh and gh with pressure, temperature, and/or frequency; high physical compliance and flexibility to conform the transducer to curved surfaces; and good mechanical shock resistance.

Piezoelectric transducers are widely used in biomedical imaging applications that span the 1–30 MHz frequency range. This indispensable diagnostic tool of modern medicine is approximately a $1 billion market worldwide [37]–[42]. The popularity of this diagnostic tool is due to its ability to produce real-time, high-resolution, threedimensional images of internal soft body tissue without the use of potentially hazardous ionizing radiation such as X-rays. A more detailed discussion of the well-established biomedical imaging applications can be found elsewhere [28], [29], [43].

The piezoelectric longitudinal charge coe cient (d33) characterizes a transducer material’s ultrasonic beam transmission capability. However, its echo receiving sensitivity is directly related to its longitudinal piezoelectric voltage coe cient (g33 = d33/ε0K). Large values for both of these coe cients are highly desirable to maximize the FOM [28], [29]. Of these two piezoelectric coe - cients, g33 may be considered to be slightly more critical because a larger g33 implies an ability to detect low intensity ultrasound [39]–[42]. This is of utmost importance as it limits the possibility—however minimal—for the ultrasonic beam to cause damage to body tissue. The dielectric constant (K) of a material plays an important role for both low-frequency hydrophone and high-frequency medical imaging applications. For instance, a relative permittivity of about 100 provides a large voltage coe cient and eases the electrical impedance matching (tuning) between the transducer and the system instrumentation [42]. The dielectric loss factor (tan δ) also should be minimized; otherwise a substantial fraction of the generated acoustic energy would be dissipated [37]–[42].

The power transduction capability of a piezoelectric material is better described by the thickness-coupling coe - cient (kt ), which is defined as the ratio of mechanical energy obtained in a thickness mode transducer to the electrical energy supplied or vice versa [29], [44]. The transducer geometry typically used in biomedical imaging is a thin disk, or plate whose fundamental resonance vibration is in the planar mode. When clamped laterally and at high frequencies, however, the thickness resonance mode dominates [29], [44]. For maximum e ciency, a thickness mode transducer should have a minimal planar mode-coupling coe cient (kp) so that the (kt /kp) ratio is as large as possible.

The amount of mechanical loss due to internal friction within a transducer material (a.k.a. acoustic viscosity) is quantified by the so-called mechanical quality factor (Qm) in a reciprocal manner, i.e., the lower the losses, the higher the Qm. Although a high Qm is desirable to keep the dissipation of acoustic energy at a minimum [42], but low Qm is needed to limit ringing to enable one to generate short acoustic pulse lengths—the primary requirement for good axial resolution in imaging [29]. A Qm 2–10 was found to be the best compromise to limit ringing without the use of external damping layers [29].

The transducer’s acoustic impedance should be closely matched to that of body tissue (1.5 MRayls) for strong acoustic coupling and minimization of the reflection from the transducer/skin interface (low insertion loss). Otherwise, matching layers would be needed to realize acoustic coupling with the body [45].

A given transducer material ideally should be compliant to eliminate air pockets at the transducer/skin interface, and to be easily shaped for focusing purposes. Its processing also should be adaptable to mass production. Given these rather contradictory requirements, the composite approach has proven to be the best concept to fabricate piezoelectric transducers for a wide range of applications.

B. Connectivity Patterns

The field patterns within the composite, which dictate electromechanical properties, are governed by the arrangement of the phases comprising the composite. The concept of connectivity, first developed by Skinner et al. [46] and Newnham [47], is a convenient way to describe the manner in which the individual phases are self-connected (continuous). There are 10 connectivity patterns for a two-phase (diphasic) system, in which each phase could be continuous in zero, one, two, or three dimensions as illustrated in Fig. 2. The internationally accepted nomenclature to describe such composites is (0-0), (0-1), (0-2), (0-3), (1-1), (1-2), (2-2), (1-3), (2-3), and (3-3). The first digit within the parenthesis refers to the number of dimensions of connectivity for the piezoelectrically active phase, and the second digit is used for the electromechanically inactive polymer phase. Based on this connectivity concept, an array of piezoelectric ceramic/polymer composites have been devel-

Fig. 2. Connectivity families for diphasic composites. Note that the total number of connectivity patterns arising from the 10 families depicted is 16 due to permutations involved in families {1-0}, {0-2}, {0-3}, {2-1}, {2-3}, {1-3}.

oped as shown in Fig. 3 [46], [47]. All of these composites were shown to exhibit improved piezoelectric properties compared to single-phase piezoelectric ceramics. A comparison of the dh.gh figure of merit of these composites is provided in Fig. 4. And in what follows, the details pertinent to the processing and properties of a select group of composites will be presented and discussed [46], [47].

The connectivity concept was extended by Pilgrim et al. [48] to include the property type, tensor rank of the property, size scale, and orientation. As discussed in the preceding paragraph, there are 10 possible connectivity patterns for a diphasic composite. However, in considering the connectivity, the fact that two composites of the same phases and connectivity could di er was not included. For instance, a composite formed from piezoelectric rods embedded in a polymer matrix is expected to have a di erent piezoelectric response than a piezoelectric monolith with polymer-filled channels. Within the framework of the 10 connectivity patterns, such a composite would possess (3- 1) connectivity. In the recently accepted usage, however, the former composite would be classified as (1-3), and the latter would be classified as (3-1). In this convention, the active phase is written first, and by accepting this new convention, Pilgrim et al. [48] suggested six additional connectivity patterns, thus raising the connectivity patterns

observed in a diphasic composite to 16, i.e., (0-1), (0-2), (0-3), (1-2), (1-3), (1-3). With this extended nomenclature, the 16 connectivity patterns are: (0-0), (1-0), (0-1), (0-2), (2-0), (0-3), (3-0), (1-1), (1-2), (2-1), (1-3), (3-1), (2-2), (1-3), (3-2), (3-3). It should be noted, however, that the inclusion of these 6 additional connectivity patterns result in 10 connectivity pattern families, in which a given family is denoted by curly brackets, as for instance, {3-1}. This family, of course, includes the patterns (1-3) and (3- 1), with the active phase’s connectivity written first [48]. In other words, the extended nomenclature renders connectivity patterns in connectivity families, allowing one to properly include all the permutations involved.

C. Methods for Making Piezoelectric Composites

A variety of traditional ceramic processing techniques, including align and fill [49]–[51], dice and fill [52], [53], injection molding [54]–[57], lost mold [58]–[67], tape lamination [68]–[72], dielectrophoresis [73], [74], relic processing [69]–[71], laser or ultrasonic cutting [75]–[77], jet machining [61]–[65], reticulation [78], [79], and co-extrusion [80], [81] have been used for making piezocomposites with different architectures.

Among these the dice and fill, injection molding and lost mold techniques are three of the popular methods for making composites for medical imaging applications.

1.Dice and Fill Method: The dice and fill method involves making a series of parallel cuts on a sintered piezoelectric block whereby a (2-2) connectivity pattern is cre-

ated. If the cut sintered piezoelectric block (2-2 connectivity) is rotated by 90◦ then further cut, one obtains (1-

3)connectivity. The cut ceramic block, which is still attached to the ceramic base, is backfilled with an epoxy, then the base ceramic support is removed by polishing. Using the grain size of the sintered block to be cut has a strong impact on the fineness one can achieve in the dicing process—the smaller the grain size, the better the machinability. The fine-grain-size approach, (1-3) composites with rods about 50 µm in width, separated by kerfs of 25 µm have been shown to be feasible [82]. The simplicity of this method, in conjunction with readily available CAD-based wafer dicing systems, has made it very popular for fabricating (2-2) and (1-3) piezocomposites and arrays for transducer applications in general and medical imaging applications in particular [38], [49]–[51], [83].

2.Lost Mold Method: The first implementation of the lost mold technique dates back to work of Rittenmeyer et al. [84]. Siemens Inc. (Munich, Germany) has also been a pioneer of the lost mold technique for fabricating piezoelectric composites by using a plastic mold manufactured by the LIGA (lithography, galvano-forming and plastic molding) process [59]. In the lost mold process, PZT slurry is filled into the plastic mold containing the desired structure as a negative. After drying, the mold is burned out and the structure sintered to more than >98% of the X-ray density. By this process, fabrication of hexagonal rods 50 µm in diameter, 400 µm in height, and spaced 50 µm apart

have been successfully demonstrated [61]–[65]. The use of well-dispersed fine particle slurries have also allowed the realization of sintered honeycomb structures with 10-µm wall thickness. Composites comprising rods with various sizes, shape, and spacing could be straightforwardly made with the lost mold method, making it a very desirable process. The limitations of this method, however, include the inability to rapidly prototype samples. Although the LIGA process is very accurate, it is also expensive and time consuming [58], [64].

3.Injection Molding: By injection molding using standard equipment, fine scale (2-2) and (1-3) ceramic structures can be made with relative ease. The production of preforms with (2-2) sintered sheet composites as fine as 25 µm, and (1-3) PZT rods as fine as 30–40 µm in diameter have been reported from this technique [56]. This method also can be used to make composites with a variety of rod sizes, shapes, and spacing. Rapid throughput, low material waste, flexibility with respect to the transducer design, and a low cost per part are the major advantages of injection molding. The limitation of this processing method is the time and cost associated with making the mold. However, significant strides have been made to circumvent such problems with the lost mold method [54]–[57]. Materials Systems Incorporated (MSI, Concord, MA) has adapted injection molding to produce net shape piezoceramics (see Fig. 5), greatly facilitating the manufacturing of large volumes of complex ceramic parts for underwater transducer applications, among others [85].

4.Tape Casting: Tape casting has been implemented for fabrication of (2-2) composites. High-frequency ultrasound transducers (>20 MHz) recently have been fabricated using stacked tapes of PZT and polymers [86]. To-

Fig. 4. Comparison of figure of merit (dh gh) for various piezoelectric composites.

day, very thin tapes (down to a several microns) can be cast, allowing manufacture of transducers with extremely high frequency for high-resolution medical imaging. In addition to (2-2) composite, (2-0-2) connectivity also was fabricated using PZT tapes and polymer mixed with di erent types of ceramic powders [87]. Other piezocomposites such as (1-3) can be made using stacked tapes and dicing. In order to generate (1-3) composites, a (2-2) structure is initially made using piezoelectric and polymer tapes, followed by dicing the composite perpendicular to the tapes. Multilayer composites with (2-2) and (1-3) connectivity patterns also have been fabricated using piezoelectric and

Fig. 5. Tube array composites fabricated by injection molding. (Courtesy of Materials System Inc.)

conductive tapes [88]. Better signal-to-noise ratios can be achieved in such multilayered ultrasonic transducers [89].

5. Microfabrication by Coextrusion: Recently, Van Hoy et al. [90] and Halloran [91] developed a microfabrication by coextursion (MFCX) method that entails forcing a thermoplastic ceramic extrusion compound through a die with a given reduction ratio. Objects with complex shapes are fabricated by assembling an extrusion feedrod from a shaped ceramic compound with space-filling fugitive compound as shown in Fig. 6(a). After each reduction state, extrudates are assembled into a feedrod and extruded again, reducing the size and multiplying the number of shaped objects. In this process, the size of the feature of interest can be reduced (after N steps of consecutive extrusions) to:

N −1

SN = So Ri ,

i=1

where SN is the final size, So is the initial size, and R is the reduction ratio of step (1) in the sequences of extrusion processes [90], [91].

In Fig. 6(b), a PMN-PT array fabricated by MFCX is shown, which shows a final spatial resolution of 60 µm. The process has the potential to be used to fabricate objects in the size range of 10 µm [90], [91]. Therefore, the size of

Fig. 6. Top, Stepwise reduction in size by MFCX, Bottom, crosssectional view of the extrudates after each reduction step, and the resultant sintered microstructure. The rectangle delineated with a dashed line is the repeat unit of the structure. (Courtesy of J. Halloran)

the features produced by MFCX is comparable to those produced using the lost-mold technique, synchrotron radiation lithography, and approaches the resolution realized by micromolding using photolithography [90], [91].

6. Solid Freeform Fabrication: The electromechanical properties of the piezocomposites, as discussed in the preceding sections, depend chiefly upon the connectivity, shape, and spatial distribution of the ceramic phase in the composite. Although the traditional ceramic processing techniques described thus far have proven e ective for making composites with simple connectivity patterns, none of them permit the fabrication of a composite possessing complex internal hierarchy and symmetry. The advent of a new manufacturing/prototyping technique to fabricate polymer, ceramic, and metal components with very high-design flexibility for novel structures, and fast prototyping is, therefore, needed. Solid freeform fabrication (SFF), an emerging technology that provides an integrated way of manufacturing three-dimensional components from CAD files, arose from the concept of additive processing— processes by which an arbitrary three-dimensional structure is made by cumulative deposition of material, without using any hard tooling, dies, molds, or machining operations [92]–[97].

In the mid-1990s, several SFF methods were developed as techniques to fabricate polymer, metal, or ceramic structures on a fixtureless platform directly from a CAD file. Some of these techniques are designed to produce large parts with a fast output rate and modest surface finish. Some SFF techniques, however, target the market in which a very high resolution and a good surface finish are very critical. Some of the SFF or rapid prototyping methods that have found commercial success include stereolithography (SLA, 3-D Systems Inc., Valencia, CA) [96], [98], [99], Fused Deposition Modeling (FDM, Stratasys Inc.) [92], [94], Selective Laser Sintering (SLS, DTM Corp.) [96], Laminated Object Manufacturing (LOM, Helisys Inc.) [100], [101], 3-D Printing (3-DP, Soligen Inc.) [102], Robocasting (Sandia National Labs) [102], [103], and Sanders Prototyping (SP, Sanders Prototyping Inc.) [104]. Most of these techniques are designed to manufacture net shape polymer parts for form/fit applications and design verification; however, some also are capable of manufacturing metal or ceramic parts such as the Rutgers fused deposition of ceramics (FDC) and fused deposition of multimaterials (FDMM) process [82].

All SFF techniques begin with a common approach. First, a CAD data description of the desired component is prepared. Second, a surface file (a.k.a. stl file) is created from the CAD file, which is later input to the manufacturing system. Third, the stl file is converted into crosssectional slices, or a slice file, in which each slice can be uniquely defined about its build strategy by varying the tool path. The slices collectively define the shape of the part. And fourth, the information pertaining to each slice then is transmitted in a layer-by-layer fashion to the SFF machine. As shown in Fig. 7, as exemplified by the FDC

Fig. 7. Schematic showing the layer-by-layer fabrication in FDC.

process, the fixtureless platform moves down one step in the Z-direction by an amount equal to the slice thickness after each layer is built. The subsequent layer is built on top of the preceding one in a sequential manner, and this process is repeated until the whole part is completed. The sequential or layered approach of manufacturing a threedimensional object of arbitrary shape is indeed the very quintessential feature of SFF. In the FDMM process, the same approach is pursued as in the FDC process; however, a multihead assembly is used instead of a singlehead assembly. The multihead assembly enables one to build three-dimensional structures comprising two materials or more in a sequential manner. Typical binary material combinations include ceramic-polymer, metal-polymer, ceramic metal, and ceramic-ceramic. Although the FDMM process o ers great flexibility in the design and fabrication of multimaterial performs, codensifications of materials due to di erential shrinkage remains a major challenge. However, the proper materials selection as well as the finetuning of the solids loading in the filament feedstock can circumvent such challenges [82].

Cesarano and Calvert [105] developed the so-called Robocasting process, another near-net-shape processing method and part of the greater SFF family of processes. In Robocasting, as depicted schematically in Fig. 8, the object is built by sequentially stacking thin layers of a given material until complicated, three-dimensional shapes are produced. Most SFF processes revolve around threedimensional printing into a powder bed (for porous parts) or the sequential layering of ceramic loaded polymers or waxes. Green parts made with these techniques typically have 40–55 vol.% polymeric binder. Therefore, very long

Fig. 8. Schematic of the Robocasting process. The pictures on the right depict typical structure obtained by the same process. (Courtesy of D. Dimos)

Fig. 9. Three-dimensional PZT lattices with (3-3) connectivity built by Robocasting. (Courtesy of D. Dimos)

and complicated burnout heat treatments are necessary to produce a dense ceramic. In the case of Robocasting, computer-controlled, layer-wise extrusion of colloidal slurries is carried out. Particle sizes are typically 0.5 to 5 µm, and the slurry solids content ranges between 50 and 60 volume percent solids, depending on the particle size distribution and morphology. Most importantly, the liquid carrier is usually water, with some minor additions of dispersants and organic binders to form pastes with a plastic consistency suitable for extrusion. It recently was shown that Robocasting can be used to create complex piezoelectric lattices, which later can be filled with an inactive material to make composites, as shown in Fig. 9.

The SFF techniques provide many advantages for the manufacturing of advanced functional components. Above all, they enable one to reduce lead times and costs in the development of new ceramic parts. In a typical conventional process, a part normally takes weeks or months to fabricate because of the time spent in fabrication of the mold, tools, machining operations, etc., to name a few; but the same structure can be prototyped in a matter of days using SFF. This technology also provides the ability to do rapid, iterative designing for producing components with optimum shape and desired properties. And, SFF can be

used to fabricate novel and complex structures that are not possible to manufacture by any other techniques because of the CAD design flexibility and additive manufacturing approach [82].

IV. Examples of Piezoelectric Composites

A. Composites with (0-3) Connectivity

In a composite with (0-3) connectivity, a threedimensionally connected polymer phase is loaded with ceramic powder (Fig. 2), which itself is zero dimensional. The ease of fabrication in a variety of forms—including large flexible thin sheets, extruded bars and fibers, and molded shapes—is the major advantage associated with (0-3) composites. Moreover, it also is amenable for mass production and for applications involving conformal surfaces. The properties of (0-3) composites are very sensitive to the choice of piezoelectric and polymer phases, as well as the fabrication method used.

The first attempts to produce (0-3) composites could be traced back to Kitayama and Sugawara [106], Pauer [107], and Harrison [108], who used PZT as a filler material and polyurethane as a matrix. These early (0-3) composites had very low d33 values. Several types of flexible piezoelectric composites consisting of PbTiO3 powder and chloroprene rubber were developed by Banno and co-workers [109]–[111], who obtained FOMs in the range from 1000×10−15 to 5000×10−15 m2/N. They also studied PbTiO3 (0-3) composites having various particle sizes (3.3, 7.3, and 31.8 µm). Smaller size particles produced lower dh values, but they showed little or no pressure dependence.

Giniewicz et al. [112] developed (0-3) ceramic powderpolymer composites using 0.5PbTiO3-0.5Bi(Fe0.98,Mn0.02) O3 [PT-BF] as the active filler and Eccogel (Ernest F. Fullam, Inc., Latham, NY) as the matrix. The composites prepared thereof exhibited virtually no pressure dependence with dhgh as high as 2700 × 10−15 m2/N.

Flexible lead titanate/Eccogel (0-3) composites prepared with highly crystalline, coprecipitated ( 3 µm) PT powder exhibited a much larger hydrostatic FOM (4170 × 10−15 m2/N) than reported for mixed oxide PT composites [113]. Chemical processing also was used to prepare PT-BF (Mn doped) powder by the coprecipitation of an aqueous citrate-based solution. Composites fabricated by die pressing a mixture of the PT-BF powder, and Eccogel polymer exhibited an excellent FOM of 4200 × 10−15 m2/N [114].

Despite the simplicity of (0-3) composite processing— a highly desirable feature—it is di cult to achieve a completely uniform distribution of powder and polymer phases, particularly at higher filler loadings as depicted in Fig. 10(a). Furthermore, low dielectric breakdown strength due to the presence of voids within the composite is a limiting factor, rendering poling to full saturation virtually impossible. As a consequence, a colloid processing method to provide microstructural homogeneity and ensure sup-

Fig. 10. Microstructure of a (0-3) composite (a) with high solid loading showing particle-particle contact that causes deviations from zero dimensional connectivity for the piezoelectric particles in the composite by conventional method. (b) The same composite processed by the colloidal method showing uniform dispersion at the same solids loading.

pression of void formation in (0-3) composites was developed to surmount such di culties [114], [115]. In this technique, a piezo-ceramic powder surface is first coated with a polymer. Second, the polymer driven out of the solution by the addition of a nonsolvent induces precipitation of polymer-powder coacervates. These coacervates then are filtrated and die pressed. The resultant (0-3) composites exhibit a uniform microstructure as shown in Fig. 10(b). Colloidally processed composites based on coprecipitated PT-BF powder and Eccogel polymer have the d33 65 × 10−12 C/N, and dhgh 6000 × 10−15 m2/N, making these composites the best in the (0-3) family of composites. Furthermore, it also was found that composites processed via the colloidal route withstood poling fields up to 150 kV/cm as compared to <120 kV/cm for identical composites (composition and solids loading) processed conventionally [114], [115].

One of the main problems in achieving (0-3) composites with high piezoelectric properties is the di culty involved

in poling the composites. In a (0-3) composite, the electric field, which acts on an individual spherical piezoelectric grain, is mostly controlled by the dielectric constant of the polymer phase. Because most polymers have a lower dielectric constant compared to piezoelectric ceramic materials, the applied electric field lines, therefore, will condense in the lower dielectric constant phase. One way to resolve this di culty in poling is to introduce a third conductive phase between the piezoelectric particles. Sa-Gong et al. [116], [117] prepared such composites by adding carbon, germanium, or silicon to PZT. Another approach to ease poling is to raise the resistivity of the ceramic filler material. This type of composite was prepared for PbTiO3 powder doped with UO2 to increase resistivity, allowing a very large (up to 130 kV/cm) poling field to be applied [113], [118].

Han [114] studied the e ect of the particle size on dielectric and piezoelectric properties of 0-3 composites in which composites of PT powder with di erent particle sizes and PVDF copolymer consisting of 90% vinylidenefluoride, and 10% hexafluoropropylene were prepared. It was found that d33 decreased with the particle size of the filler material. It also was noteworthy that the d33 value decreased from 50 to 33 × 10−12 C/N as the particle size of the PT ceramic changed from 3.5 to 0.5 µm. Unlike the d33 behavior, the dielectric constant was found to be almost independent of the particle size of the ceramic filler, except for slightly lower values in the composites with very fine particle sizes [114]. Lee et al. [119], [120] investigated the e ect of the particle size on the dielectric and piezoelectric properties of (0-3) composites. In this work, PT powder of di erent particle sizes was mixed with Eccogel polymer, and composites were prepared by conventional die pressing methods. They also observed that the dielectric constant and dissipation factor are not a ected by particle size. However, d33 decreased dramatically with particle size.

To investigate the e ect of the polymer on resistivity and dielectric properties, Han [114] prepared PT composites with Eccogel polymer, PVDF copolymer, and ethylene-propylenediene monomer (EPDM) polymer. The PVDF copolymer and EPDM polymer had moderate resistivities (6 × 1014 and 1016 ohm-cm, respectively), but the resistivity of the epoxy was lower than that of the PVDF copolymer (1012 ohm-cm). It was found that, although higher poling conditions could be applied to the PVDF copolymer and EPDM composites, the highest d33 value was obtained from the epoxy composites. The higher electrical conductivity of the polymer matrix may have created more electric flux paths between the ceramic particles. This in turn increased the electric field acting on the ceramic filler and made poling of the ceramic easier. The dielectric constant of the composite with EPDM polymer was slightly lower than that of the other two types of composites, but the dissipation factor of the composites with epoxy was twice that of the PVDF copolymer and EPDM polymer. The epoxy gave the highest dhgh figure of merit (5600 × 10−15 m2/N), and the EPDM gave the lowest figure of merit (600×10−15 m2/N). With the consideration of