)94@0<=4>C:1%099=C7@,94,)94@0<=4>C:1%099=C7@,94,
'.3:7,<7C:88:9='.3:7,<7C:88:9=
0;,<>809>,7%,;0<="' 0;,<>809>:1",>0<4,7='.409.0924900<492
#:@08-0<
9G?09.0:1,>4:9$</0<:9>304070.><4.%<:;0<>40=:1%-"29G?09.0:1,>4:9$</0<:9>304070.><4.%<:;0<>40=:1%-"2
#-#-
$$
%-'.%-'.
#-#-
$$
%"#%'#&07,B:<0<<:070.><4.=%"#%'#&07,B:<0<<:070.><4.=
!0:9,<-0<
)94@0<=4>C:1%099=C7@,94,
%0>0< ,@40=
)94@0<=4>C:1%099=C7@,94,
/,@40=7<=8?;0990/?
:77:A>34=,9/,//4>4:9,7A:<6=,>3>>;=<0;:=4>:<C?;0990/?8=0+;,;0<=
&0.:8809/0/4>,>4:9&0.:8809/0/4>,>4:9
,<-0<!,@40=% 9G?09.0:1,>4:9$</0<:9>304070.><4.%<:;0<>40=:1%-"2
#-
$
%-'.
#-
$
%"#%'#&07,B:<0<<:070.><4.=&0><40@0/1<:83>>;=<0;:=4>:<C?;0990/?
8=0+;,;0<=
:;C<423>(3080<4.,90<,84.':.40>C&0;<49>0/1<:8
:?<9,7:1>3080<4.,90<,84.':.40>C
*:7?80
==?0#:@08-0<;,20=
(34=;,;0<4=;:=>0/,>'.3:7,<7C:88:9=3>>;=<0;:=4>:<C?;0990/?8=0+;,;0<=
:<8:<0491:<8,>4:9;70,=0.:9>,.><0;:=4>:<C;:-:B?;0990/?
9G?09.0:1,>4:9$</0<:9>304070.><4.%<:;0<>40=:1%-"29G?09.0:1,>4:9$</0<:9>304070.><4.%<:;0<>40=:1%-"2
#-#-
$$
%-'.%-'.
#-#-
$$
%"#%'#&07,B:<0<<:070.><4.=%"#%'#&07,B:<0<<:070.><4.=
-=><,.>-=><,.>
(300110.>:1>30=4>0.,>4:9.3084=><C,9/:</0<492:9>30/4070.><4.;<:;0<>40=:1=:74/=:7?>4:9=49>30
B
%-"2
#-
$
%-'.
#-
$
%"#%'#;0<:@=64>0=C=>08A,=0B,8490/49=,8;70=A4>3
≤
B
≤(30<8,7,990,7492><0,>809>=A0<00110.>4@04949/?.4927:92<,920=4>0:</0<49>30
=,8;70=A4>349>34=.:8;:=4>4:9,7<,920(30A077:</0<0/7,<20.3084.,7/:8,49.0<,84.=0B34-4>
<07,B:<-03,@4:<?;>:
B
E1:<34230<@,7?0=:1
B
9:<8,710<<:070.><4.-03,@4:<A,=:-=0<@0/:<
B
≤
<0/?.>4:9=49>30.3084.,7/:8,49=4D03,/9:=4294F.,9>0110.>:9>30A0,6F07//4070.><4.
;<:;0<>40=-?>49/?.0/,><,9=4>4:9>:<07,B:<-03,@4:<1:<
B
E(30/4=:</0<0/%'#<4.3=,8;70=
?9/0<2:,=;:9>,90:?=D0<:F07/<07,B:<>:10<<:070.><4.><,9=4>4:9=4847,<>:>3,><0;:<>0/;<0@4:?=7C1:<
%'#(30F07//0;09/09>;<:;0<>40=:1.:8;:=4>4:9=7C492.7:=0=>>:>30<07,B:<>:10<<:070.><4..<:==:@0<
0B34-4>0/>3034230=>=09=4>4@4>C>:,7>0<,>4:9=49>30.3084.,7:</0<(30;<:;0<>40=:1>34==C=>08,<0
.:9=4=>09>A4>3,<,9/:8=4>0/0=.<4;>4:9:1>30:</0<0/%-βH
β
$
=><?.>?<0A4>3βH
"2
B
#-
B
'.
B
,9/βI#-
:8809>=:8809>=
:;C<423>(3080<4.,90<,84.':.40>C&0;<49>0/1<:8
:?<9,7:1>3080<4.,90<,84.':.40>C
*:7?80==?0#:@08-0<;,20=
(34=5:?<9,7,<>4.704=,@,47,-70,>'.3:7,<7C:88:9=3>>;=<0;:=4>:<C?;0990/?8=0+;,;0<=
Influence of Cation Order on the Dielectric Properties of
Pb(Mg
1/3
Nb
2/3
)O
3
–Pb(Sc
1/2
Nb
1/2
)O
3
(PMN-PSN) Relaxor Ferroelectrics
Leon Farber and Peter Davies*
Department of Materials Science and Engineering, University of Pennsylvania, 3231 Walnut St.,
Philadelphia, Pennsylvania 19104-6272
The effect of the B-site cation chemistry and ordering on the
dielectric properties of solid solutions in the (1ⴚx)Pb(Mg
1/3
Nb
2/3
)O
3
–Pb(Sc
1/2
Nb
1/2
)O
3
(PMN-PSN) perovskite system was
examined in samples with 0.1 < x < 0.9. Thermal annealing
treatments were effective in inducing long-range B-site order
in the samples within this compositional range. The well-
ordered, large chemical domain ceramics exhibit relaxor be-
havior up to x ⴝ ⬃⬃0.5; for higher values of x, normal
ferroelectric behavior was observed. For x < 0.5 reductions in
the chemical domain, size had no significant effect on the
weak-field dielectric properties, but induced a transition to
relaxor behavior for x > ⬃⬃0.6. The disordered PSN-rich
samples undergo a spontaneous zero-field relaxor to ferroelec-
tric transition similar to that reported previously for PSN. The
field-dependent properties of compositions lying closest to the
relaxor to ferroelectric crossover exhibited the highest sensi-
tivity to alterations in the chemical order. The properties of
this system are consistent with a “random site” description of
the 1:1 ordered Pb(ⴕ
1/2

1/2
)O
3
structure with ⴕ ⴝ (Mg
(2ⴚ2x)/3
Nb
(1ⴚx)/3
Sc
x
) and ⴖ ⴝ Nb.
I. Introduction
T
HE relationship between the B-site cation order and ferroelec-
tric properties of the Pb(Mg
1/3
Nb
2/3
)O
3
(PMN) family of
perovskites has been extensively investigated.
1
For several years,
it has been recognized that the relaxor behavior in these systems is
related to localized inhomogenieties in the distribution of the metal
cations on the B sublattice. Recent experimental and theoretical
investigations of Pb(Mg
1/3
Ta
2/3
)O
3
(PMT) and PMN have pro-
vided convincing evidence that the order on the B-sites conforms
to a “random site” model.
2–9
In this model, one of the cation
positions () in the 1:1 ordered Pb(⬘
1/2
⬙
1/2
)O
3
phases is occupied
by a ferroelectrically active cation (niobium or tantalum), whereas
the second contains a random mixture of active and inactive
(magnesium) cations. Using thermal treatments to tailor the degree
of cation order, alterations in the size of the chemically ordered
domains in relaxors based on PMN and PMT were shown to have
relatively little effect on the weak-field dielectric response.
4–8
This is quite different from the behavior of the Pb(Sc
1/2
Ta
1/2
)O
3
(PST) and Pb(Sc
1/2
Nb
1/2
)O
3
(PSN) systems where alterations in
the degree of order produce a transition from relaxor (disordered)
to normal (ordered) ferroelectric behavior.
10
In this paper, we
examine the response of the dielectric properties of a solid solution
of PMN-PSN to alterations in the degree of chemical order.
Support for the random site model for the B-site order in
PMN-type systems has come from new investigations of their
ordered structures
2,3
and from the observation of extensive in-
creases in the size of the chemical domains and degree of order in
samples equilibrated at elevated temperature.
4–8
In a previous
paper,
9
we reported that the chemical order in PMN was enhanced
significantly by the substitution of relatively small (⬃10 mol%)
concentrations of PSN. After slow-cooling or extended annealing,
extensive order was observed in samples across the (1⫺x)PMN–
(x)PSN solid solution system when x ⱖ 0.1. By monitoring the
changes in the order as a function of temperature, the order-
disorder boundary for the PMN–PSN system was established. The
transition temperature for pure PMN was estimated to be ⬃950°C,
and the highest transition, 1360°C, was observed for x ⫽ 0.5. The
changes in stability across the system were consistent with the
“random site” description of the cation order where sites in the
Pb(⬘
1/2

1/2
)O
3
1:1 ordered solid solution have ⬘ ⫽ (Mg
(2⫺2x)/3
Nb
(1⫺x)/3
Sc
x
) and ⬙ ⫽ Nb. Because of their higher ordering
temperatures, the degree of chemical order and chemical domain
size in the x ⱖ 0.1 compositions could be modified extensively by
thermal annealing.
Several studies have been made of the dielectric properties of
the PSN and PMN end members. For PSN (and its tantalate
counterpart, PST) it is well known that the formation of a 1:1
ordered structure is accompanied by a transition from a relaxor to
normal ferroelectric response.
10
Because of its overall 1:1 B-site
stoichiometry, in PSN, complete structural order is accompanied
by complete chemical order with one cation position occupied by
scandium and the other by niobium (or tantalum in PST). In
contrast, for the PMN-type systems, the formation of a well-
ordered 1:1 B-site structure does not produce any significant
change in the weak-field dielectric response and relaxor behavior
is observed regardless of the degree of order or the chemical
domain size.
7,8
The difference can be attributed to the absence of
complete chemical order in the 1:1 random site form of the PMN
systems. Because of the overall 1:2 B-site stoichiometry, 1:1
structural order in PMN is accompanied only by chemical order on
one cation position (), and a random cation distribution is always
retained on the second site (⬘). For the PSN systems, lead
vacancy formation has also been shown to induce additional
variations in the dielectric behavior. In particular, disordered,
vacancy-free compositions (prepared by annealing in a PbO-rich
environment) were observed to undergo a spontaneous zero-field
relaxor-to-normal ferroelectric transition at temperatures below the
dielectric maximum.
11
This transition was suppressed in samples
that contained small vacancy concentrations. Similar observations
of a zero-field relaxor-to-normal ferroelectric transition have also
been reported in the PMN-PT system for 10% PT.
12
Because the two end members exhibit quite different responses
to alterations in their chemical order, we were interested in
examining at what point this system would exhibit a crossover
from PMN-type to PSN-type behavior. We demonstrate that the
weak-field dielectric properties of PMN-rich compositions (x ⬍
0.5) retain their insensitivity to the chemical domain size and
observe a transition to PSN-type behavior for higher values of x.
H. U. Anderson—contributing editor
Manuscript No. 187430. Received October 1, 2001; approved November 13, 2002.
This work was supported by the Office of Naval Research through grants
N00014-98-1-0583 and N00014-01-1-0860, and by the National Science Foundation
through grant DMR 98-09035. The work also made use of the MRSEC shared
experimental facilities supported by the NSF under grant DMR96-32598.
*Member, American Ceramic Society.
J. Am. Ceram. Soc., 86 [11] 1861–66 (2003)
1861
journal
II. Experimental Methods
PMN–PSN ceramics were synthesized from high-purity
(99.9%) oxides via a modified “columbite-type” route.
13
Stoichi
-
ometric quantities of predried oxides of the B-site metals were
mixed in an agate mortar for 10 min and then calcined at 1100°C
for 6 h. The resultant calcine was ball-milled in a polyethylene jar
for 3 h using yttria-stabilized zirconia balls and acetone as a
milling medium. After drying and recalcining, the slurry at 1100°C
for 12 h, stoichiometric amounts of lead oxide were added and
mixed in an agate mortar. This mixture was precalcined at 750°C
for 3 h, ball-milled for 3 h, and then reheated at 900°Cfor3h.All
these treatments were conducted in a closed platinum crucible.
After a final ball-milling for 10 h, pellets 4–5 mm thick and 10 mm
in diameter were isostatically pressed at 600 MPa. Final sintering
of the ceramics was conducted at a temperature of 1225°–1250°C
for 1 h. During the sintering and subsequent high-temperature
annealing treatments, the ceramics were buried in a protective
powder of the same composition.
X-ray diffraction (XRD; Model D-max B, Geigerflex, Rigaku,
Tokyo, Japan) was used to determine the phase purity. The degree
of order was monitored by scaling the integrated intensity of the
(1/2, 1/2, 1/2) supercell reflection, I
o
, to the (001) subcell reflec
-
tion of the perovskite subcell, I
100
. The order parameter ␣ was
calculated from the scaled intensities from ␣
2
⫽ (I
o
/I
100
)
observed
/
(I
o
/I
100
)
max
, where (I
o
/I
100
)
max
is a value calculated assuming
complete 1:1 order within the random site model. In this calcula-
tion, it was assumed that the lead and oxygen ions are not
displaced from their ideal positions in the perovskite unit cell;
however, it is recognized that these displacements do exist and will
induce errors in our calculations of the subcell peak intensities.
The microstructures of the ceramics were characterized using
transmission electron microscopy (TEM; Model 420 EX, Philips,
Eindhoven, Netherlands) operated at 120 kV. The TEM specimens
were prepared by grinding, polishing, and dimpling ceramic thin
sections. The final thinning to perforation was conducted via
argon-ion milling (5.5 kV and 20 mA, PIPS, Gatan, Pleasanton,
CA).
Low-field dielectric measurements were performed using a
high-precision LCR meter (HP 42824A, Agilent Technologies,
Palo Alto, CA). The temperature-dependent measurements were
made using an environmental chamber (Delta 9023, Delta Design,
San Diego, CA). High-field polarization measurements used a
ferroelectric test system (Model RT66, Radiant, Albuquerque,
NM).
III. Results
The extensive degree of chemical order induced by slow-
cooling the (1⫺x)PMN–(x)PSN samples is evident in the dark-
field TEM images in Fig. 1 that were collected using the (3/2, 3/2,
3/2) supercell reflections. Details of the thermal conditions re-
quired to induce the order are provided elsewhere.
9
Although the
microstructure of the x ⫽ 0.1, 0.5, 0.7, and 0.9 specimens all show
evidence for high levels of chemical order (bright regions in Fig.
1), the size of the chemically ordered domains is clearly different
in each sample. The largest ordered regions (⬃200 nm) are
observed in x ⫽ 0.5, which was previously shown to also exhibit
the highest order–disorder transition temperature. The variation in
the domain size across the system also parallels the trend in the
ordering temperatures discussed in the previous paper.
9
The response of the dielectric properties to the alteration in bulk
chemistry, degree of order, and chemical domain size was exam-
ined by collecting weak-field data from the well-ordered samples,
and from specimens exhibiting minimal cation order. Data col-
lected from the ordered ceramics, with the microstructures shown
previously in Fig. 1, are shown in Fig. 2. For x ⱕ 0.5 the
permittivity of the ordered samples exhibits diffuse and frequency-
dependent behavior characteristic of a relaxor type response.
However, for x ⱖ 0.6, the ordered compositions show a normal
ferroelectric behavior and the dielectric properties are similar to
those observed in the pure ordered PSN end member. It is
important to note that the crossover from relaxor-to-normal behav-
ior (between x ⫽ 0.5 and 0.6) shows no correlation to the size of
the chemical domains in the ordered samples. The composition
with the largest chemical domains (x ⫽ 0.5) is a relaxor, whereas
Fig. 1. Dark-field image collected from chemically ordered (1 ⫺
x)PMN–(x) PSN with (a) x ⫽ 0.1, (b) x ⫽ 0.5, (c) x ⫽ 0.7, and (d) x ⫽ 0.9.
Fig. 2. (a) Real part of the permittivity, (b) dielectric loss for ordered
samples of (1 ⫺ x)PMN–(x)PSN with x ⫽ 0.2, 0.5, 0.7, and 0.9.
Frequencies correspond to 100 Hz, 1 kHz, 10kHz, 100 kHz, and 1MHz.
1862 Journal of the American Ceramic Society—Farber and Davies Vol. 86, No. 11
a composition with small domains (x ⫽ 0.9) has a normal
ferroelectric response.
To examine the possible influence of lead-vacancies on the
dielectric properties, which have been shown to induce relaxor-
type behavior in PSN (and PST), the properties of the ordered
samples were remeasured after a low-temperature anneal in a
lead-rich atmosphere. The lead anneal did not produce any
significant change in the dielectric properties. The absence of a
correlation between the type of ferroelectric response, the chemical
domain size, or the presence of lead vacancies suggests that the
dielectric properties of the ordered ceramics in the PMN–PSN
system are controlled by the chemistry of the ordered random-site
structure, specifically by the composition of the ⬘ site. This will
be discussed later.
The dielectric properties of the disordered PMN-PSN ceramics
are shown in Fig. 3. Compositions with x ⬍ 0.5, which exhibit
relaxor behavior in their ordered forms, show very little change
with the reduction in the degree of order. However, for x ⬎ 0.5
(normal ferroelectrics in their ordered forms), the reduction in the
degree of ordering induces a transition to frequency-dependent,
relaxor behavior. This change is accompanied by an increase in the
magnitude of the permittivity and the temperature of the permit-
tivity maximum (e.g., ⬃⌬⌻ ⫽ ⬃35°C for x ⫽ 0.9). The disordered
PSN-rich samples also exhibit a discontinuity in the real part of the
permittivity at temperatures just below the permittivity maximum
(see Figs. 3 and 4). This type of discontinuity has also been
observed in the disordered PSN end member and arises from a
spontaneous zero-field, relaxor-to-normal ferroelectric phase
transformation.
11
Further evidence for the different behavior of the samples
with x ⱕ 0.5 and x ⬎ 0.5 was obtained from measurements of
the thermal hysteresis of the dielectric properties. A typical
example of the data collected during thermal cycling (cooling
followed by heating) for the PSN-rich compositions is shown in
Fig. 5(b) for x ⫽ 0.9. The real part of the permittivity and the
dielectric loss exhibit thermal hysteresis associated with the
spontaneous relaxor-to-normal ferroelectric-phase transition. In
contrast, similar experiments conducted on the PMN-rich com-
positions did not reveal any measurable hysteresis in the
dielectric response for either the disordered or ordered samples
(see Fig. 5(a) for x ⫽ 0.2) and the data collecting on heating and
cooling were identical.
More detailed measurements were made on the x ⫽ 0.5
composition, which lies at the boundary between the relaxor
and normal sides of the system. The very high order/disorder
transition temperature for this composition (T
trans
⫽⬃1360°C)
prevented preparation of ceramics with complete cation disor-
der; however, elevated temperature heat treatments could be
used to prepare samples with differing order parameters (␣).
Although all the samples exhibited relaxor-type behavior, some
hysteresis was observed (Fig. 6). The difference in the proper-
ties on cooling and heating were very small in the sample with
a lower degree of ordering (Fig. 6(a), ␣⫽0.76), but increased
with the degree of order (Fig. 6(b), ␣⫽0.97). Therefore, this
composition at the relaxor to normal boundary combines
features of the PMN-rich and disordered PSN-rich sides of the
system.
The field-dependent properties of the solid solutions were also
probed as a function of temperature. Fig. 7(a) shows the polariza-
tion loops recorded at different temperatures for a PMN-rich
ordered composition with x ⫽ 0.2. The hysteresis loops are typical
of those observed in relaxors and showed little dependence on the
degree of order. The temperature dependence of the remanent
Fig. 3. (a) Real part of the permittivity, (b) dielectric loss for disordered
samples of (1 ⫺ x)PMN–(x)PSN with x ⫽ 0.2, 0.5, 0.7, and 0.9.
Frequencies correspond to 100 Hz, 1 kHz, 10kHz, 100 kHz, and 1MHz.
Fig. 4. Real part of the permittivity for disordered and ordered forms of
(a) x ⫽ 0.2 and (b) x ⫽ 0.9.
November 2003 Influence of Cation Order on Dielectric Properties of PMN-PSN Relaxor Ferroelectrics 1863
