116

IEEE TRANSACTIONSON SONICS AND ULTRASONICS, VOL.

su-13, NO. 4, OCTOBER 196(

16 6

Transducers Using Forced Transitions Between Ferroelectric and Antiferroelectric States

nere

D e
P AQ

Abstract-The use of resonant piezoelectric composite structures energy by virtue of the geometric change of the primifor acoustic radiation into fluid media is well established. At low tive unit cell. Since the transition occurs also with tcmperfrequencies, desirable for long distance acoustic transmission, resonant structures are extremely large and expensive. Frequency ature change, thermoelect,ric and thermomechanica] lowering can be accomplished by a variety of means involving use of energy conversion are also possible. With respect to the bending structures or mass-loading, but the resulting transducers heat of transition, t,hese processes are directly coupled, are vulnerable to pressure effects due to deep submersion. A possible but the thermal energy required to provide the necessary alternative to present transducer structures is the use of the electric temperature rise limits power conversion to the 1 percent field-forced transistion from antiferroelectric to ferroelectric. A number of Pb(Zr, Sn, Ti)Oa compositions have been developed eEciency range. The electric field-forced transition was demonstrated which experience these transitions a t relatively low electric field (- 7-15 kV/cm) and typically generate volume strains near 0.1 first in PbZrO, by Shirane et al. [3] in a narrow temperpercent. There is therefore not necessarily a requirement for ature range near the Curie point. Sawaguchi showed that acoustic shielding, and since the developed strain is independent of substitution of about 3 atom percent Ti4+ for Zr4+ in frequency, resonant structures are not needed. Construction is therefore simplified and the size and weight of low-frequency PbZrO, introduces a ferroelectric state (rhombohedral) transducers may be reduced. Operating characteristics of the in a narrow temperature range below the Curie point [4]. antiferroelectric transducer are discussed in detail, but no large The electric field-forced antiferroelectric-ferroelectrictranlow-frequency transducer arrays have yet been built. sition in Pb(Zr, Ti)O, compositions is obtained only over

lese tions vert' pecia we , Equ msdi vice the1 11 nc nera Ltic 1

a narrow temperature range because the free energy of the ferroelectric state is slightly higher than that of the HE TRAXSITION from antiferroelectric to fer- antiferroelectric state only over a narrow temperature roelectric can occur spontaneously with temperature range. Jaffe found that this difference in free energy iS change, increase in electric field, or change in stress minimized over a much wider temperature range by configuration. The technically important ferroelectric ce- substantial substitution of Sn4+ for (Zr, Ti)*' [SI, and ramic materials, with the exception of PbNb,Oe, are all of that further reductions are accomplished b y substitution the perovslute structure type and are therefore only of La3' and Sr2+for Pb2+[6]. slightly distorted from cubic. Some of these materials have both ferroelectric and antiferroelectric distorted structures, GENERAL PRINCIPLESTHE ANTIFERROELECTRIC OF and in these cases, transitions to ferroelectric from either TRANSDUCER cubic (paraelectric) or antiferroelectric involve a n inThe equations which define thc action of the antifenocrease in specific volume. The specific volume of the cubic state is intermediate between those of the ferroelectric electric transducer are derived readily from the thermoand antiferroelectric states. I n the ferroelectric state, the dynamic potential generally called the Gibbs free energy. unit cell i elongated paralle1 t o the polar axis; in the The antiferroelectric-ferroelectric (A-F) transition is a s antiferroelectric state, one of the axes perpendicular to first-order transition, and there are therefore disconthe antiparallel displacements is shorter than in the cubic tinuities in the partial derivatives of the Gibbs free energy with respect t o temperature, stress, or electric field. These (reference) phase. It has been demonstrated t h a t hydrostatic and one- partial derivatives are, respectively, entropy, mechanical dimensional compression favor the antiferroelectric state, strain, and dielectric displacement. The following reand transitions a t pressures as low as a few thousand psi lationships are obtained by sett,ing the Gibbs free energy have been obtained with some compositions [I], [a]. The equal for states A and F. They hold for first-order tramipressure-forced transition may be used for conversion of tions, with derivatives taken along the phase boundaries. mechanical to electric energy. In similar fashion, the =- V / V A electric field-forced antiferroelectric,ferroelectric transiAD I tion may be used for conversion of electrical to mechanical

INTRODUCTION

iere
83, A !

T,, 5
For Prop

d
ese
;tra

($)

Manuscript received April 15, 1966. T i work was supported by hs the Sandia Corporation and the Ofc of Naval Research. This fie g p e r w s presented at the 1965 IEEE Ultrasonics Symposium, a oston, Mass. The author is with the Clevite Corporation, Electronic Research Division, Cleveland, Ohio.

and

d w ctric iival iplec Figu d-fo Pur the e n A. :the :tric iilibi tra .) a n Pigm an

66

BERLINCOURT: TRANSDUCERS USING FORCED TRANSITIONS

117

iere

E V

= =

electric field (V/m), dielectric displacement (C/mz),

= volume (m3) = tempernture (OK), = hydrostatic pressure (N/m), and = thermal energy (J/m3, heat of transition).

D
0

AQ

three conditions of mechanical load: 1) zero pressure load, 2 ) linear mechanical load with energy dissipated in the load, and 3) ideal nonlinear mechanical load. With 1) there is no energy conversion, and a volume change AT occurs a t E,, and a change -AV a t E. with electric field decrease. The dielectrically dissipated energy per unit volume W Dis given approximately by

er-

sd,

Lese equations hold strictly only for equilibrium con;ions, but actually there is always hysteresis. It has vertheless been shown that the equations hold well Iecially for transitions occurring with increasing temperIre, pressure, or electric field [2]. Equation (1) describes the action of the ferroelectric tnsducer for electrical t o mechanical energy conversion vice versa. Equations (2) and (3) describe the action thermoelectric and thermomechanical conversion, and 1 not be discussed further here. Equation (I) can be 1 neralized t o include stress systems other than hydrottic pressure as follows:

W D = (E, - E,)P.Y,
where

(9)

P , = spontaneous polarization of the ferroelectric

state, C/mZ,

E = electric field, V/m, and W , = dissipated energy density, J/m3.

I n b), the mechanical energy delivered per unit volume is W, ( ) ) ( A T / V ) A p . The additional electrical energy dissipated per unit volume is (4) AE AD = ( $ ) ( A V / V ) A p [see (S)]. With the ideal nonlinear load, the mechanical energy delivered per unit volume is A p ( A V / V ) = AE AD. With the linear mechanical load, the ratio of energy delivered t o energy dissipated in the antiferroelectric element is

in $1)
41. ,nrer

d
(5)

of he re is
3Y

id

iere S,, SI = strain parallel and perpendicular to applied electric field (m/m) and T,, TI = stress parallel and perpendicular to applied electric field (N/m*)

Under these ideal conditions the efficiency is

1

00-

For increments in stress and electric field with the propriate boundary conditions, (I), (4),and (5) become AE AD = ( A V / V ) Ap, (6) AE A D = A S , AT3, (7) 3 A E A D = A S , AT,. (8)
ese equations state that a compressive stress raises transition electric field to ferroelectric, that a n electric d will likewise raise the transition stress t o antiferroctric, and t h a t the energies involved are equal. This is iivalent t o stating that these energies are perfectIy ipled, k = 100 percent. Figure 1illustrates the ideal equilibrium (no hysteresis) d-forced ferroelectric transition a t ambient (for practipurposes zero) pressure. The transition t o ferroelectric the electric field E , is marlied by an increase in polarizan AP and specific volume AV/V. With decrease in ctric field, the transition back t o the stable antiferroctric phase takes place here also a t E,. Under noniilibrium conditions, there is significant hysteresis and transition fields for the transitions to ferroelectric ,) and t o antiferroelectric (E,,)are related E. < E , < E,. Figure 2 shows schematically the actual operation of : antiferroelectric transducer for electromechanical :rgy conversion. D-E and A V - p plots are sketched for
5

Y.
a nse

al
eY
!IS.

2)

3)

This shows that the key to efficient operation is low hysteresis (E, - E,) and heavy loading (AB large). This will be discussed further with respect t o data on actual ceramic compositions, but it should be noted that the antiferroelectric transducer cannot be highly efficient. Its strength lies rather in simplicity of design and freedom from the requirement of mechanical resonance. It should perhaps be noted that the treatment has to this point been general, not limited to a fluid load. The antiferroelectric transducer is better matched, of course, t o solid loads. One could, for instance, use an antiferroelectric transducer (low duty cycle) for generation of high amplitude acoustic waves a t any frequency below its switching speed (at least 5 MHz with the best materials). It can thus be used as an extremely broadband (direct current to a few megahertz) delay-line transducer or as a generator of elastic stress for elasto-optic modulation, but in both cases with a severely restricted duty cycle. Figure 3 shows schematically the operating range of a slanted loop antiferroelectric transducer. A dc bias field about equal t o the average of E , and E, is maintained and an ac electric field is applied to generate the volume strain. The resulting ac polarization and ac strain are shown schematically. Due t o nonlinearity and hysteresis, the dynamic polarization and strain are distortecl. The distortion in the polarization results in internal heating [equation (9)], while that in strain results only

118

IJCEEI TRANSACTIONS ON SONICS AND ULTRASONICS

OmOBEI

66

J+,

D'

I E I
Et

i i

E-

E BIAS

Fig. 1. Idealized antiferroelectric-ferroelectric electric field-forced transition.

Fig. 3.

Schematic operation of antiferroelectric transducer wit bias electric field.

g. 5.

EZERO

PPRESSURE LOAD

LINEAR MECHANICAL LOAD

IDEAL NONLINEAR MECHANICAL LOAD

. Fig. 4 PbZrOt-PbTi03 temperature-composition phase diagran

Fig. 2. Antiferroelectric transducer operation; schematic load cycle in (b) and (c) assumes that the stored mechanical energy is dissipated in the mechanical or acoustic load. Here W , = electric energy dissipated in the transition cycle (hysteresis), WE = additional electric energy dissipated due to the mechanical load, and W , = W E = energy dissipated in mechanical load for the cycle.

in distortion of the acoustic signal. The latter results in some degradation of output power a t the driving frequency, but does not otherwise affect operation.

EXPERIMENTAL DATA
There are two antiferroelectric states in modified and unmodified PbZrO,, one with orthorhombic symmetry with only very slight distortion from tetragonal, the other with tetragonal symmetry. I n each case, there is very high cell multiplicity, with over 100 perovskite unit cells in one tetragonal multiple cell. The forced transition to ferroelectric occurs generally only from the tetragonal antiferroelectric state. PbZrO, is orthorhombic antiferroelectric a t room temperature. The temperature-compo-

sition phase diagram of Fig. 4 shows that this state r( mains stable t o the Curie point. Substitution of substanti: amounts of Ti4' for Zr"' [4] (Fig. 4), Ba2' for Pb2' [7 or Nb5' for Zr4+ [8] brings about a ferroelectric (rhomb( hedral) phase. Three ferroelectric phases are s h o r n 1 Fig. 4. The tetragonal ferroelectric phase F T is isostru tural with PbTiOs. There is a first-order transition b tween the two rhombohedral phases, with discontinuitit in lattice constants and polarization, but it has n o t yl been possible to detect any superstructure in either sta or a difference in symmetry [ 2 ] ,[9]. Substitution of Sn4+ for (Zr, Ti)4+ and/or La" f l Pb2+ in Pb(Zr, Ti)O, extends the tetragonal antiferr electric state a t the expense of the rhombohedral f e n electric state. These effects are demonstrated in triaxi phase diagrams in Figs. 5 and 6. I n addition, Fig. 5 shol the establishment of a metastable F R ( L T ) - A T pba boundary by poling. With compositions within the met stable region, the forced transition from antiferroelectr
I

I 1

BERLIKCOURT

TRAKSDUCERS USING FORCED TRANSITIONS

. 5.

35 .60 .65 .ra .75 .BO .e5 .go .ss Triaxial phase diagram for the system Pb.wNb.oz(Zr, Sn, Ti).mO3 at 25C showing metastability due to poling. All specimens poled at 25"C, if possible.

KI)

P nC b

.gran

;e

ri

inti: [7 mbc vn n b uitii 't Y

sta

(Zr, Sn, Ti)Oa a t 25C; specimens not poled. Fig. 6 . Triaxial phase diagram for the system Pb.97La.02

+ f fen f eM iax jho Pha met ectl

---Fig.
Hysteresis

loop

for

Yb.os;Pu'b.o,[(Zr.7Sn.3).91Ti.051.9903

& ,

~ , c m

-I*% r ~ [lz

Wr'dA

at 25C illustrating metastable ferroelectric state.

rI

120
IEEE TRANSACTIONS ON SONICS AND ULTRASONICS
.*I-

OCTOBl

966

3 6

.ea

:
a

.20

J2

*.
I

-I0

- 15;
Fig. 8.

tv
I

0
I

100

L
4c

200 TEMPERATURE -.C

300

10 0 200 TEMPERATURE - " C

xx,

Fig. 9. Thermal expansion for Pb.Q~La.oz(Zr.Q~~Ti.~~Sn.026)C 'ig. 1:

Thermd expansion for Pb.QQNb.oz(Zr.~~Ti.o~Sn~~,),~~O~ (virgin element).

to ferroelectric occurs with electric field increase, but the ferroelectric state remains as a metastable phase with removal of the electric field. An example of this type of behavior is shown in Fig. 7. Only compositions which have the tetragonal antiferroelectric state stable over a useful temperature range may be considered for application as antiferroelectric transducers. For the transition t o occur a t a reasonable electric field, the free energy difference should be low. This, in effect, means that desirable compositions are those which are close to the F R ( L T ) - A T phase boundary in phase diagrams such as Figs. 5 and 6, and on the A T side. The differences in primitive cell volumes for the F R C L T ) , Ao, A,, and cubic (paralectric) phases are illustrated by the thermal expansion curves in Figs. 8 and 9. These curves and other related data show that the primitive unit cell size is in the following order:

ti
4)

1 Ao )
$T FR,LT) F,,HT,

volume increasing.

Fig. 10. Dimensional changes vs. hydrostatic stress for Pb.wNI( Z r , ~ ~ T i . ~ o S n . ~ 5 )LowOand high pressure phases are FR( .9~ a. and AT.

5)'

The specific volume difference i l o - F R ( H T ) is as high as 0.96 percent, but this transition cannot be electric fieldforced. The transition which can be field-forced in specific compositions over a wide temperature range is that between A T and F R [ L T ) The specific volume difference . ranges from 0.5 percent to less than 0.1 percent depending on composition. It should be noted that thermal expansion in the ferroelectric range is markedly anisotropic with poled specimens, This is much more pronounced in F R , , , , than F R ( L T ) .
1 The high temperature ferroelectric rhombohedral phase does not occur in Figs. 8 and 9, and i t does not, in fact exist in compositions exhibiting the forced transition over a wide temperature range.

Fig. 10 shows dimensional changes through a pr sure-f orced f erroelectric-antlf erroele ctric transit ion the coniposition whose thermal expansion curve is sho in Fig. 8. The upper curves show S, and Sj for a PO' specimen and the curve a t lower left shows AV/V ZS, S,. Remanent values of S , and S, are due domain orientation, and remanence in AV/V for 1 poled specimen is the result of a slight volume increa probably due to microcracks, which occurred dur poling. The pressure a t which the transition F , , , T ) occurs is more sharply defined in the poled specim and it occurs at a pressure about 8000 psi higher. T should be expected since the elcctric energy per u volume stored in poling (J E d D ) raises the energy difI

nce 1 ausir 2 qui1 nit I 'he ~p A. Fig 3r tl how he F pig. ransj ure lope icrez jsing terce ainec empc netri iolarj Thi ric f ransi <tics ield pig. ions Lave 3ops. hose ields errot ure lectr orcec
'

* P'

, T

366

BERLINCOURT: TRANSDUCERS USIXG FORCED TRANSITIONS

121
30--

IO

20

30 40
hV/cm

50

- L -

400

-lo
ig.

t
I
-5 0 5 BIAS ELECTRIC FIELD. KV/cm
Y)

IS

2 0

25

IZdOl.

11. Electric field-hydrostatic pressure phase diagram at 25C for Pb. d b . d Z r .OT.D Sn.2 6). 9 8 0 3 . si

ice between the ferroelectric and antiferroelectric states, iusing a siniilar increase in mechanical energy (J p d V / V ) quired to force the transition. The electric energy per lit volume stored is about equal to E,PR = 0.19 J/cm3. he additional mechanical energy is about equal t o p AV/V = 0.23 J/cm3. Figure 11 shows an electric field-pressure phase diagram lr the composition whose thermal expansion curve is iown in Fig, 8 and whose dimensional changes through L pressure-forced transition F R I L e F)-ATare shown in ig. 10. These data were obtained by measurement of ansition pressure, determined from permittivity-presire curves, a t different bias electric field levels. The 3pe of p , (pressure for transition to antiferroelectric with creasing pressure) vs. bias field curve is 960 psi/kV/cm. sing this and P R = 27 pC/cm2, (1) gives AV/V = 0.41 :rcent, in nearly exact agreement with the value ob4 Jned in Fig. 10 and the value extrapolated to room mperature in Fig. 8. The upper curve in Fig. 11is asymetric about zero electric field because of the original JJb.w darity of the poled specimen. VR U T ) There are, in general terms, two different types of elecic field-forced antiferroelectric-ferroelectric A T-FR( LT ) ansitions. Neither, of course, has the ideal character'its sketched in Fig. 1. Typical charge density-electric pres- Id hysteresis loops for the two types are shown in L for g. 12. Al antiferroelectric Pb(Zr, Sn, Ti)O, composil horn Ins with Curie points above about 100C appear t o poled ve either relatively square or slanted hysteresis )ps. I n comparison with the square loop materials, ue to ose with slanted loops have lower transition electric P the Ids (B,), lower relative volume differences between the -roelectric and antiferroelectric states, wider temperauring re ranges over which the transition can be forced by -)-AT lctric field, and much less hysteresis. Furthermore, the imen, -ced transition is faster with the slanted loop maThis

Fig. 12. Slanted and square hysteresis loops a t 25C.

v=

unit ..-

terials, taking place in less than one microsecond. All factors other than the magnitude of the volume dlfference thus favor the slanted loop materials for acoustic power generation. Typical strain-field curves for transitions of these two types are shown in Figs. 13 and 14. The strains S , perpendicular to and S , parallel t o the electric field are shown as well as S y = S, 2S1. The strain S , was measured by means of a strain gage cemented to an electroded surface. The volume strain S v was determined from volume displacement of the specimen when placed inside a small oil-filled fused quartz chamber with a calibrated capillary. The strain S, was calculated ( S , = S v - ax,). Several of the advantages of the slanted loop composition for acoustic power generation are immediately evident in Fig. 12 and are demonstrated by comparison of Figs. 13 and 14. The disadvantage of considerably lower transition strain is also demonstrated. The first one-half cycle is unique, and is shown here only for completeness. There is a small remanence in S, and S,, with that in the former opposite in sign and almost exactly twice the magnitude of the latter. The remanence is due to domain orientation in the forced ferroelectric state, and because of the mechanical nature of this orientation, its effects remain in the stable antiferroelectric state. It is necessary to heat the specimen above the Curie point to remove the remanent strains, and only after doing so can the first one-half cycle be repeated. For use in transducers this is not important, and in any case there is no such effect with the volume strain. It is noteworthy and of considerable practical importance that with the slanted loop material, the strain SI is more than an order of magnitude less than Sa. On repetitive cycling, XI even contributes to the volume strain, but only slightly. I n contrast, piezoelectric ceramic transducers presently employed in radiating

IEEE TRANSACTIONS O N SONICS AND ULTRASONICS

OCMBER

*T

- - - -- l o++
01T .

-0.01-

-0.11

Fig. 14. Strain vs. electric field through a forced antiferroelectric02[( & ferroelectric transition cycle for P b .99Nb. Zr ,$Sn.4). QTi,4. P. Fig. 13. Strain vs. electric field through a forced antiferroelectricferroelectric transition cycle for Pb,stLa.04(Zr.42Ti.l$n.r)Oa.

15

transducers generate maximum strain a t or near mechanical resonance, and require acoustic shielding to prevent strain cancellation due t o cross-contraction. This, in addition to the necessity for frequency lowering and impedance matching, leads to high static stress and consequent large changes in the parameters of the piezoelectric ceramic [lo]-[16]. Figure 15 shows volume strain as a function of polarization for the slanted loop composition. This curve essentially combines the data of the hysteresis loop in Fig. 12 and the volume strain-electric field curve in Fig. 13. The curve in Fig. 15 is virtually linear within the 18 pC/cm2). With P 5 18 transition region (to P pC/cm2, there are piezoelectric effects in addition t,o further transition effects. Near Palthere are no additional transition effects, and the piezoelectric effect reaches its saturation value ( g h 17 X m2/C). Figure 16 shows corresponding data for the pressureforced ferroelectric-antifcrroelectric transition. The curve a t the left is reproduced from Fig. 10. The curve in the center shows loss of polarization of the ferroelectric state through the transition, with the indicated initial slope equal to the measured value of d,. The loss of polarization is shown as a function of AV/V on the right. Just as in Fig. 15, the AV/V-P relationship is linear through the transition region. The slope l / V dV/dP in the transition region is 2.1 x lo- m2/C for the square loop coniposition of Fig. 16 and 3.6 X m2/C for the slanted loop material of Fig. 15. The six to one ratio i just about equal to the s atio of transition values of AV/V. It is also about equal r

P.

IO-C/E~~

Fig. 15. Volume strain vs. polarization for Pb.srLa. 04Zr.42Ti.li3n.~ O J .

Fig.

to the ratio of ferroelectric distortions3 for these compositions. A truly surprising feature of the slanted 1001: compositions is their very slight distortion from thr cubic reference phase in combination with values of Pl about the same as for square loop compositions M7itl much higher distortion. It appears either that there m large ionic displacements which do not result in unit eel distortion with the slanted loop materials, or perhapr more likely, there is a much larger contribution t o t h polarization by electron cloud distortion rather thar ionic displacements. Figure 17 shows the dependence of E,, here definec for a slanted loop composition as the electric field ai which P = +P,, on compressive stress. There is no con
3 The ferroelectric distortions are rhombohedral, with tht difference in axial lengths of longest and shortest diagonal of tht cubic reference cell proportional to the deviation of the rhombohed ral angle from 90.

clus par: late lool: stre:

Ef-a
Fro1 valc iden late1 0.12 Sam tran in t is, k
4

stres

3
OEER

1966

BERLIBCOURT : TRASSDTCERS USING FORCED TRANSITIONS

123

I. \

as 06 . % i

' 0.4-

z
I

0.2-

Fig. 16. Volume strain and loss of polarization vs. hydrostatic pressure for Pb.9~Nb.ol(Zr,B~Ti.07Su.?~).~~0~.

tric-

k.

POOP ' the Ps rith are cell aPs the ian led at onthe the ied-

that the piezoelectric constant d,, is very much less in magnitude than d3, in the forced ferroelectric state, and HYDROSTATIC PRESSURE 2d,, is as a result, the hydrostatic effect d, = d,, considerably stronger than in ordinary piezoelectric ceramics with perovskite structure. These effects are fairly clear from the slopes of the saturated portions of the curves in Fig. 13. The apparent transition strains in Fig. 13 may be chosen a t the electric field a t which the polarization saturates, about 30 kV/cm. Neglecting the 'STRESS T~ (PERPENDICULAR m E,I first one-half cycle, they are 0.115 percent (AV/V), 0.10 percent (&), and 0.01 percent (SI). Agreement with calculated values i fairly good, especially for (AVIV),. s This would normally be expected; one can achieve exactly hydrostatic conditions, but one can only approach experimentally a condition of one-dimensional stress. 1 0 0 S IO I5 20 25 30 It should be pointed out t h a t direct use of a change of 103 PSI COMPRESSIM STRESS specific volume for generation of acoustic power in a Fig. 17. Variation of Ef with stress for Pb.g~La.o~(Zr.r~Ti,18Sn.4)OI. fluid medium cannot be extended to low frequencies without quite large configurations because of the general requirement of dimensions not too small compared to a clusive dfierence in the data for hydrostatic stress and parallel stress. There is very little change of E , with wavelength. Assuming an isotropic strain in a sphere, lateral st,ress. These data were obtained from hysteresis the acoustic power in watts is given by loops taken with stress maintained. Changes of E , with C3 P (- A g stress (not shown) are virtually identical, so the difference 1 pa = (4rR2) 1 (12) E,-E, is for practical purposes independent of stress.' 18 / x \ 2 From the slopes of these curves and (l), (4), and ( 5 ) , values of (AV/V),, (JS'~)~, and with the subscript identifying the strains as transition strains, may be calcu- where lated using ALl = Ps. The values are, respectively, 0.12, R = radius, meters 0.12, and 0.028 percent. Comparison with Fig. 13 for the c = velocity of sound (1500 m/s in water) same nominal composition is fairly difficult since the p = density of water = lo3 kg/m3, and transition is not sharp, and there are piezoelectric strains h = wavelength in meters = c / f . in the forced ferroelectric state. The piezoelectric effect, The antiferroelectric transducer does not generate an is, however, fairly weak. It should nevertheless be noted isotropic strain even though all strain components are in the same sense (Fig. 13). Nevertheless (12) may be There is actually a slight decrease of ErE, with increasing applied to estimate acoustic pow-er available from an stress.
i

STRESS T ,

(PARALLEL TO E 3 1

-n- . i

(y

124

IEEE TRANSACTTONS ON SONES AND ULTRASONICS

OCTOBER

IEEE

TABLE I ACOUSTIC POWER DELIVERED INTERNALLY AND DISSIPATED POWER FOR WITH STRAIN ASSUMED ISOTROPICWITH E r E C = 2 kV/cm, AP = 15 pC/cm AND
Output Power/ Unit Area W/cml 14 X lo-* 0.17 2 18.2 52 102

A SPHERICAL ANTIFERROELECTRIC RADIATOR AND AT7/V = 0 . 1 PERCENT; CAVITATION NEQLECTED.

[31

141

R/A
0.02 0.05 0.1
0.2

Radius, cm 100 Hz 1000 Hz

30 75 150 3 7.5 15 30 45
60

Total Output Power, kW 100 H z 1000 Hz 0.14 11.7 560 1.4

Internally Dissipated* Power, kW 100 H z 1000 Ha 51 800 6300

Efficiency percent 0.3 1.5 8.2 29 43 46

-

~51 [GI [71 181 [QI r101

10-3

0.12

0.3 0.4

5.6 206

1320 3500

0.51 8.0 63 510 1770 4200

* With 15 percent

of total enclosed volume driven.

<

antiferroelectric transducer. Table I shows the acoustic power density ( p , / 4 r R 2 , power per unit area) in water for spherical shells with several values of R/X and a volume strain of 0.1 percent. Cavitation is neglected. Actual values of the radius for the listed values of R/X are shown for frequencies of 100 and lo00 Hz. Corresponding values of total output power are given for all reasonable sphere sizes. The table also lists values for internally dissipated power. With the slanted loop composition of Fig. 12, E,-E. is about as low as has yet been achieved, although materials with considerately lower values of E, are available. I n this case, E,-E, 2 kV/cm and 69 15 pC/cm2 for a volume strain of 0.1 percent about a bias of 20 kV/cm. The heat power is thus 3 W/cm3 a t 100 Hz or 30 W/cm3 a t 1000 Hz. The total internally dissipated power values listed in Table I were calculated assuming that only 15 percent of the total enclosed volume is driven. It is immediately evident that the antiferroelectric transducer will require a low duty cycle or forced cooling or both. With a 1 percent duty cycle and noting that the heat capacity is about 3 J/cm3 (3: the average temperature rise would be about 1C in 100 seconds at 100 Hz or in 10 seconds a t 1000 Hz, assuming no heat loss. It should be pointed out that optimum compositions from this standpoint (low E,-E,) have yet to be developed, and there are indications that considerable improvement can be obtained. It is fortuitous that with temperature rise E r E , decreases, with a corresponding decrease in heat power. This can be seen in an extreme example in Rg. 18, where hysteresis loops for a square loop composition are shown a t several temperatures. With respect to temperature rise, therefore, the antiferroelectric transducer tends to be self-limiting. Narrowing of the loop with the slanted loop material is less dramatic but nevertheless definite. I n actual practice, one would probably use a stack of rings electrically connected in parallel and with major strain component (Fig. 13) axial. The stack of rings should, of course, be capped to increase fluid displacement. Lateral relief is not required due to the nature of the transition strain. If a horn is provided in order to increase further the fluid displacement, lateral relief would, however, be required.

-- 30
20

C/cmZ

f
J

25.C

-- 10

- -

Fig. 18. Progressive changes of hysteresis loop character with temperature for Pb.99Nb. 02[( Zr.6Sn. BTi, 51.9803.

Ab wave: vanta catior descr trans

CONCLUSIONS The use of a n electric field-forced antiferroelectric to ferroelectric transition for electromechanical energy conversion is fairly well understood. Its major advantage is very wideband (nonresonant) operation, and its major disadvantage is high internally dissipated power. It is possible that antiferroelectric transducers may find use in low-frequency sonar s y s t e m with low duty cycle or as wideband delay line transducers where high strain amplitude is important.

I II
ting sisto1 first GaAt in th It is lessei terial Lo taxia itself not : prod1 struc
M E J.

ACKNOWLEDGMENT The author wishes t o thank B. Jaffe, who developed

the ceramic compositions described in this paper, and H. H. A. IXrueger, who made many of the measurements.

REFERENCES
[l] D. Berlincourt, H. Jaffe, H. H. A. Krueger, and B. Jaffe, Release of electric energy in PbNb (Zr, Ti, Sn) O1by temperature and by pressure-enforced phase transitions, A p p l . P h p Lett., vol. 3, pp. 90-92, September 1963. [2] D. Berlincourt, H. H. A. Krueger, and B. Jaffe, Stability of phases in modified lead zirconate with variatinn i . nressure, .~~. _ . . _ . . . . n . electric field, temperature, and composition, J. Ph;. Chem: Solids, vol. 25, pp. 659-674, July 1964.
~

Comp G. McGi!

IEEE TRANSACTIONS O N SONICS AND ULTRASONICS, VOL.

SU-13, NO. 4, OCTOBER 1966

125

Phys., vol. 30, pp. 1804-1810, November 1959. [3] G. Shirane, E. Sawaguchi, and Y. Takagi, Dielectric properties [11] B. A. Rotenberg, The effect of pressure on the piezoelectric of lead zirconate, Phys. Rev., vol. 84, pp. 476-481, November properties of barium titanate, S v e Phys.-Solid Stale, vol. 1, oit 1951. pp. 1627-1630, June 1960. [4] E. Sawaguchi, Ferroelectricity versus antiferroelectricity in the solid solutions of PbZr03 and PbTiO, J . Phys. SOC. [12] H. H. A. Krueger and D. Berlincourt, Effects of high static stress on the piezoelectric properties of transducer materials. Japan, vol. 8, pp. 615-629, September-October 1951. .- [5] B. Jaffe, Antifemoelectric ceramics with field-enforced transiJ . Acoust. Soc: Am., vol. 33, fp. 1339-1344, October 1961. tions: a new nonlinear circuit element, Proc. IRE, vol. 49, [13] S.V. Bogdanov, B. M. Vul, and R. Ya. Razbach, The piezoelectric properties of polycrystalline barium titanate a t high pp. 1264-1267, August 1961. pressures, S v e Phys.-Cyst., vol. 6, pp. 58-61, July-August oit [6] , unpublished. 1961 [7] G. Shirane and S. Hoshino, X-ray study of transitions in [14] R.-F, Brown, Effect of two-dimensional mechanical stress on PbZr03 containing Ba or Si, Acta Cryst.,vol. 7, pp. 203-210, the dielectric properties of poled ceramic barium titanate and February 1954. lead zirconate titanate, Can. J . Phys., vol. 39, pp. 741-753, [SI N. N. Krainik, Phase changes in certain solid solutions containing lead zirconate, Sou. Phys.-Tech. Phys., vol. 3, pp. May 1961. [15] R. F. Brown and G. W. McMahon, Material constants of 493-502, March 1958. ferroelectric ceramics a t high pressure, Can. J. Phys., vol. 40, r91 H. Barnett. Evidence for a new Dhase boundarv in the ferropp. 672-674, May 1962. electric lead zirconate-lead titanate system, J. A p p l . Phys., , Properties of transducer ceramics under maintained [16] vol. 33, p. 1606, April 1962. planar stress, J . Acoust. SOC.Am., vol. 38, pp. 570-573, [lo] D. Berlincourt and H. H. A. Krueger, Domain processes in October 1965. lead titanate zirconate and barium titanate ceramics, J . A p p l .

.--

GaAs Transducers
J. BOHM,
MEMBER, IEEE, AND

G.

w.FARNELL, SENIOR

MEMBER, IEEE

th

Absfracf-GaAs ultrasonic transducers to generate longitudinal waves have been fabricated from semi-insulator material. Advantages of an epitaxial layer transducer are discussed and fabrication steps of a prototype inverse epitaxial layer device are described. Loss measurements performed on these 41 and 120 Mc/s transducers are presented.

INTRODUCTION
to nis or

is
se or in

ed

id ts.
ye, erYS.

HE POPULARITY of GaAs material for various semiconductor device applications is well known; tunnel-diodes, varactor diodes, infrared light-emifting diodes, semiconductor lasers, and experimental transistor structures have been fabricated from it, and the first observation of Gunn oscillation was also made with G a h . The great interest was accompanied by advances in the technology and wider availability of the material. It is also known that GaAs is piezoelectric-though to a lesser degree than the more common piezoelectric materials. Low-resistivity layers of .Gahs have been grown epitaxially on substrates of Si and Ge as well as on GaAs itself. While high-resistivity epitaxial GaAs layers are not yet available, it is likely that such layers will be produced in the near future, thereby facilitating the construction of various integrated acoustic devices; thus it

was considered worthwhile t o investigate some of the acoustic properties of high-resistivity GaAs. As described below, transducers have been fabricated from such highresistivity Gails both from bulk material directly and from material containing a n epitaxial interface. Increasing the resistivity of the transducer material increases the conversion eficiency. For this study, ntype semi-insulator GaAs with lo ohm cm resistivity was chosen and it was cut to generate longitudinal waves in the [lll]direction. Table I summarizes the significant properties of G a h for such longitudinal wave propagation. For experiments such as those involving acoustic gain, the high mobility is very promising [1] but the low coupling coefficient is a disadvantage (for CdS, k M 0.2).

of re,
m.

GAASBULK TRANSDUCERS The initial experiments were intended to evaluate the characteristics of the bulk semi-insulator GaAs itself a s an ultrasonic transducer material. I n order to obtain the highest conversion efficiency, the transducers were designed to operate in the fundamental mode, t,hat is with the thickness equal t o one-half the acoustic wavelength. The fabrication process included orientation in the [lll] direction, slicing, lapping, and polishing; selected areas of the slice were gold plated (with Sn content) to provide ohmic contacts. The plating was followed by an alloying process a t 585C, and finally a scribing operation sepaManuscript received February 7 , 1966. J. Bohm is with McGill University and Northern Electric rated the individual transducers. The quality of the Company, Montreal, Canada. ohmic contacts has not been consistent but was improved G. W. Farnell is with the Department of Electrical Engineering, in many cases by concentrated heat treatments. McGill University, Montreal, Canada.