|
Printable PDF Version
September 3, 2003
Giant Dielectric Effect in CaCu3Ti4O12 and its Cd Analogue
C.C. Homes1, T. Vogt1, S.M. Shapiro1, S. Wakimoto1,2, M.A. Subramanian3, and A.P. Ramirez4
1Department of Physics, Brookhaven National Laboratory, Upton, NY;
2Department of Physics, Massachusetts Institute of Technology, Cambridge, MA;
3DuPont Central Research & Development Experimental Station, Wilmington, DE;
4Materials Integration Science Laboratory, K764, Los Alamos National Laboratory, Los Alamos, NM
The perovskite-related material CaCu3Ti4O12 has a very high
static dielectric constant ε0~104 at room temperature, which drops
to about 100 at low temperature. Substituting Cd for Ca reduces the
room temperature value by over an order of magnitude. The origin of the large ε0
is not fully understood, but may be due to an internal barrier layer
capacitance (IBLC) effect. Infrared measurements on the Ca and Cd
compounds show dramatic changes in the nature of the normal modes at
low temperature, suggesting that increasing electronic localization
may lead to a breakdown of the IBLC effect.
Materials with large dielectric constants are highly sought after
in the microelectronics industry for use in memory devices; the
static dielectric constant ε0 acts as a scaling factor and
ultimately determines the level of miniaturization. One of the most
commonly used dielectric materials is silicon nitride with ε0~7;
materials with ε0>7 are generally referred to as high-dielectric
constant materials. It was recently noted that the cubic perovskite-related
material CaCu3Ti4O12 shown in Figure 1 had a tremendously high
dielectric constant at room temperature, ε0~104, which showed little
temperature dependence until about 100 K, below which it decreased
by nearly two orders of magnitude to ε0~100, shown in
Figure 2a. The
substitution of Cd for Ca results in a large reduction of ε0, shown
in Figure 2b. It was initially suggested that this large value for
ε0 was a purely extrinsic effect due to Maxwell-Wagner-type
depletion layers forming between the sample and the metal contacts
used in the capacitive measurements of ε0 in the kHz range. This
effect has indeed been observed in a number of different materials.
However, the addition of a thin a buffer layer of a dielectric with
well-known electrical properties (aluminum oxide) has demonstrated
that the large value for ε0 persists in these materials, and is
therefore a property of the bulk. The dramatic drop of ε0 at low
temperature is curious because it is not accompanied by any change
in the long-range crystallographic structure when probed by
high-resolution x-ray and neutron powder diffraction. In
ferroelectric materials, large changes in ε0 are usually accompanied
by a structural distortion and soft-mode condensation.

Infrared (IR) spectroscopy is a powerful tool to study the
lattice vibrations in insulators, in particular soft-mode
condensation. The large change in ε0 and the absence of any
structural distortion suggested that the normal modes should be
examined. The temperature dependence of the reflectance was
determined over a wide frequency range using the spectrometer at
NSLS beamline U10A and the complex optical properties determined
from a Kramers-Kronig analysis. The real part of the dielectric
function ε1 is shown in Figures 3a and 3b for CaCu3Ti4O12 and
CdCu3Ti4O12, respectively, at a variety of temperatures in the infrared region.

The dielectric function is dominated by the infrared active lattice modes. (The units of cm-1 are commonly used
in IR spectroscopy, 1 THz = 30 cm-1.) While these experiments are
done in the THz range, the low-frequency value for ε1 agrees with
the values for ε0 (kHz) at low temperature. It is clear from
Figure 3 that ε1 at low frequency is also increasing at low temperature.
Fits to simple Lorentzian oscillators describe these features quite
well and indicate that many of the vibrational modes are gaining
strength at low temperature, a violation of the partial f-sum rule
for oscillators. From this result, it may be inferred that the Born
effective charge on the atoms is changing and that the material is
becoming more ionic. We (along with others) have speculated that the
large value of ε0 is due to an internal barrier layer capacitance (IBLC)
effect due to extensive twinning or grain boundaries.

This view is consistent with impedance spectroscopy that indicates these
materials may be described as semiconducting regions separated by
insulating barriers. The increase in ionicity implies increasing
charge localization at low temperatures. This localization may lead
to an increase in the size of the insulating regions and a
commensurate reduction of ε0 within the IBLC picture, suggesting
that the large value for ε0 is, at least in part, due to “extrinsic”
effects.
BEAMLINE
U10A
FUNDING
Department of Energy
PUBLICATION
C. C. Homes et al., “Charge transfer in the high dielectric
constant materials CaCu3Ti4O12 and CdCu3Ti4O12”, Phys. Rev. B
67, 092106 (2003).
|