Full text of DOI:10.1557/JMR.2000.0207

Microstructure and chemistry of nonstoichiometric
Ba,Sr)TiO thin films deposited by metalorganic chemical
(
3
vapor deposition
Igor Levin
Ceramics Division, National Institute of Standards and Technology, Gaithersburg, Maryland 20899
Richard D. Leapman
Bioengineering and Physical Science Program, Office of Research Services, National Institutes of
Health, Bethesda, Maryland 20892
Debra L. Kaiser
Ceramics Division, National Institute of Standards and Technology, Gaithersburg, Maryland 20899
(Received 27 December 1999; accepted 3 April 2000)
The microstructure and chemistry of (Ba,Sr)TiO thin films deposited on Pt/SiO /Si
3
2
substrates by metalorganic chemical vapor deposition were studied using high-
resolution transmission electron microscopy and quantitative spectrum imaging in
electron energy loss spectroscopy. The grain boundaries in all films with overall Ti
content ranging from 50.7% to 53.4% exhibit a significant increase in Ti/Ba ratio as
compared to the grain interiors. The results suggest that the deviations of Ti/(Ba + Sr)
ratio from the stoichiometric value of unity are accommodated by the creation of
Ba/Sr vacancies, which segregate to the grain boundary regions. The films with Ti
contents equal to or greater than 52% additionally contained an amorphous Ti-rich
phase at some grain boundaries and multiple grain junctions; the amount of this phase
increases with increasing overall Ti content. The analysis indicates that the amorphous
phase can only partially account for the significant drop in dielectric permittivity
accompanying increases in the Ti/(Ba + Sr) ratio.
Recently, much attention has been focused on the fab-
rication of polycrystalline thin films of barium strontium
no such increase was detected for the specimen with
50.7% Ti. Their analysis of the energy loss near edge
structure (ELNES) indicated a broadening of the Ti L2,3
ELNES features for the 53.4% Ti specimen; this was
attributed to the distortion of the oxygen octahedra due to
partial accommodation of excess Ti in the lattice. Levin
1–7
titanate (BST). This material can offer a high dielec-
tric permittivity at a relatively large film thickness in
6–8
dynamic random-access memories, and has properties
7
of interest for tunable microwave devices. Deposition of
1
2
the BST films with an excess of Ti has been found to be
necessary to achieve acceptable leakage currents and di-
et al., working on very similar specimens, demon-
strated an increase in the Ti/Ba ratio at all of the grain
boundaries in a film with 53.4% Ti. They found that all
of the grain boundaries in this film were Ba deficient,
while some of these additionally exhibited an increase in
Ti content. The latter regions were associated with an
amorphous-like phase frequently observed at the grain
boundaries and triple junctions in the 53.4% specimen.
The amorphous phase was found to contain at least Ba,
Ti, and O (no measurements of Sr content were con-
ducted). The present contribution summarizes the results
of the combined HRTEM and quantitative EELS analy-
sis of the BST films with different Ti/(Ba + Sr) ratios.
9
electric lifetimes, and to obtain reproducible, stable, and
smooth film morphology. However, deviation of the Ti/
(Ba + Sr) ratio from the stoichiometric value of unity has
an adverse effect on dielectric permittivity of the BST
10
films. For example, the dielectric constant of 40-nm-
thick films decreases by about 60% when the Ti content
10
is increased from 51 to 53.4 at.%. This effect is even
more pronounced for larger film thicknesses. Despite this
dramatic effect of stoichiometry on the dielectric prop-
erties, the exact mechanism for accommodation of excess
Ti in these films still has not been determined. Recently,
structural imaging in a high-resolution transmission elec-
tron microscope (HRTEM) and high-spatial resolution
electron energy loss spectroscopy (EELS) were applied
to study the microstructure and chemistry of BST films
The (Ba,Sr)TiO thin films were deposited on Pt
3
(100 nm)/SiO /Si substrates by high-composition preci-
2
sion, liquid delivery MOCVD at a substrate temperature
2
of 640 °C. Films with three different Ti contents, here-
deposited on Pt/SiO /Si substrates by metalorganic
after referred to as the 50.7%, 52% and 53.4% speci-
mens, were studied. The average compositions of the
50.7% and the 53.4% BST films, as determined by wave-
length dispersive x-ray fluorescence (XRF) spectroscopy
2
11,12
chemical vapor deposition (MOCVD).
al. reported a systematic increase in the Ti/O ratio at
Stemmer et
11
the grain boundaries of a film containing 53.4% Ti, while
J. Mater. Res., Vol. 15, No. 7, Jul 2000
© 2000 Materials Research Society
1433
http://journals.cambridge.org
Downloaded: 29 Mar 2015
IP address: 129.15.14.53

Rapid Communications
1
3,14
using well-characterized standards,
are summarized
position over the area sampled by spectrum-imaging is
equal to the average composition of the entire film as
measured by XRF; the maps of Ti/Ba ratio for the vari-
ous films were subsequently normalized to the corre-
sponding average values of Ti/Ba. The larger standard
deviations reported for the EELS measurements reflect in
part higher statistical errors associated with plural scat-
tering and diffraction effects, as well as the background
subtraction.
in Table I. The film thickness for all specimens was
about 38 nm, as measured by XRF and subsequently con-
firmed by cross-sectional transmission electron micros-
copy (TEM).
Details of the TEM specimen preparation have been
12
12
given elsewhere. Structural imaging was performed in
a JEOL 3010UHR (300 kV) high-resolution transmis-
sion electron microscope. A VG Microscopes HB501
dedicated scanning transmission electron microscope
equipped with a Gatan model 766 DigiPEELS (parallel
electron energy loss spectrometer) was used for chemical
analysis. For this analysis, the specimen was cooled to
liquid nitrogen temperature to prevent contamination.
Spectrum-images (64 × 64 pixels) were recorded by
scanning a focused probe of about 1.5-nm diameter over
an area of about 80 × 80 nm. The thickness of the ana-
lyzed area was less than that of the BST film (38 nm).
Spectra in the energy loss range from 400 to 800 eV
TEM imaging of all three films revealed similar mi-
crostructure composed of columnar grains with a mean
grain diameter of about 15 nm, independent of composi-
tion. All films are fiber textured with the 〈001〉 direction
parallel to the surface normal. No amorphous phase was
detected in the 50.7% specimen, while disordered re-
gions with amorphous-like contrast were observed occa-
sionally at some of the multiple grain junctions in the
52% specimen (Fig. 1). In contrast, such amorphous-like
regions were frequently observed at the grain boundaries
15
12
(
which includes the Ti–L , O–K, and Ba–M edges)
and grain junctions in the 53.4% specimen.
2,3
4,5
were acquired at each point (pixel). Specimen drift dur-
ing the data acquisition was corrected by using a cross-
correlation procedure implemented in the GATAN
Digital Micrograph v.3. software. The spectrum-images
were subsequently processed to remove the background,
and maps of both the Ti–L2,3 and the Ba–M4,5 intensity
distributions were obtained by integrating the corre-
sponding edges over an energy window of 50 eV. The
coefficient of variance due to counting statistics for the
Ti/Ba ratio was about 0.01. Spectrum-imaging of the
Sr–L2,3 edge (1950 eV) was unsuccessful due to the ex-
ceedingly large acquisition times required to achieve an
acceptable signal-to-noise ratio.
Spectrum-imaging of the 50.7% specimen (Fig. 2)
showed an increase in the Ti/Ba ratio at the grain
boundaries, similar to our previous findings on the
12
53.4% film. Compared to the grain interiors, all of the
boundaries appear to be Ba deficient with no detectable
increase in Ti content. The spatial resolution of the pre-
sent analysis (determined primarily by the beam broad-
ening) is about 5 nm, as can be deduced from the
full width at half-maximum of the peaks in the line pro-
files of Ti/Ba ratio taken across the grain boundaries
[Figs. 3(a) and 3(b)]. The effective thickness of a grain
boundary varies from about 1 nm for the core (Fig. 1) to
some larger value if space charge regions are included.
Assuming a cylindrical shape for the analyzed volume,
the numbers measured at the grain boundaries are related
to the actual Ti/Ba ratios, (Ti/Ba)GB, by a factor of ␲d/4t,
where d is the diameter of the analyzed volume (∼ 5 nm)
and t is the grain boundary thickness. According to our
measurements, the product (Ti/Ba)GB × t is about 6 (with
t in nm). No significant difference in the amount of seg-
regation to the grain boundaries (excluding those with an
amorphous phase) was observed between 50.7% and
53.4% specimens. For the amorphous pockets with a size
EELS data were quantified using single crystals of
BaTiO and SrTiO as standards. The EELS spectra from
3
3
the standards and from the BST film with 53.4% Ti were
recorded during the same microscope session under simi-
lar experimental conditions. Sr and Ti were quantified by
using the L2,3 edge, and Ba by using the M4,5 edge. Both
the Ti/Ba and Ti/Sr ratios in the 53.4% film measured
by EELS from 20 areas, each about 80 × 80 nm in size
(
Table II), agree well with those measured by XRF.
Based on this result, we assumed that the average com-
a
TABLE I. Chemical compositions of two BST films as measured
by XRF.
TABLE II. Cation ratios as measured by XRF and EELS.
53.4% Specimen
5
0.7% Specimen
XRF
Cation ratio
XRF
EELS
XRF
Element
53.4% Specimen
50.7% Specimen
a
Ti/Ba
Ti/Sr
Ba/Sr
Ti/(Ba + Sr)
1.61 ± 0.02
4.0 ± 0.03
2.49 ± 0.03
1.15 ± 0.01
1.59 ± 0.08
3.94 ± 0.25
2.48 ± 0.19
1.13 ± 0.07
1.43 ± 0.01
3.66 ± 0.03
2.55 ± 0.03
1.03 ± 0.01
a
Ti
Ba
Sr
53.44 ± 0.22
50.67 ± 0.22
35.45 ± 0.31
13.88 ± 0.1
b
b
33.21 ± 0.31
13.35 ± 0.10
a
a
Compositions are given in at.% of the total metal fraction. Uncertainties
correspond to a single standard deviation.
Uncertainties correspond to a single standard deviation.
Calculated from Ti/Ba and Ti/Sr ratios.
b
1
434
J. Mater. Res., Vol. 15, No. 7, Jul 2000
Downloaded: 29 Mar 2015
http://journals.cambridge.org
IP address: 129.15.14.53

Rapid Communications
of 2–3 nm, the Ti/Ba ratio in the pocket is estimated to
be between 4 and 5 from the measured ratios of Ti/Ba
present EELS measurements is not sufficient to ascertain
a dependence of the Ba/Sr ratio in the grain interiors on
the overall composition.
[Fig. 3(c)] obtained with a 5-nm probe.
The average Ti/Ba ratios measured in the grain inte-
For the 50.7% and 53.4% specimens, the average Ti/
Ba ratio in the grain interiors is less than the overall
Ti/Ba ratio (Table II). In addition, the Ba content was
lower in the grain boundaries than in the grain interiors,
whereas the Ti content remained approximately the same
[Figs. 2(a) and 2(b)]. These combined results suggest
that the [Ti/(Ba + Sr)] ജ 1 ratios in the specimens are
accommodated by the creation of AסBa,Sr cation va-
cancies which segregate to the grain boundary regions.
An inhomogeneous distribution of the Ba/Sr ratio in the
films can be attributed to differences in the energies
of formation of the Ba and Sr vacancies. Segregation of
the Ba/Sr vacancies to the grain boundaries creates nega-
tive space-charge regions which are expected to be com-
pensated by the positively charged grain-boundary
riors are 1.29 ± 0.04 and 1.36 ± 0.05 for the 50.7% and
3.4% specimens, respectively. The uncertainties corre-
spond to a single standard deviation as determined from
2 grains. Because the BST films studied have an aver-
age Ti/(Ba + Sr) ratio greater than unity, it is reasonable
to assume that the grain interiors in these films are either
stoichiometric or Ti rich. The Ti/Ba ratios inside the
grains for both specimens were found to be reproducibly
5
1
(
1
three maps for each specimen) smaller than the values of
.39 (50.7%) and 1.4 (53.4%) calculated for the case of
Ti/(Ba + Sr) ס 1 using mean values of Ba/Sr ratio meas-
ured by XRF (Table II). The difference is statistically
significant for the film with 50.7%. These estimates sug-
gest an inhomogeneous distribution of the Ba/Sr ratio in
the specimen, with smaller values at the grain boundaries
than in the grain interiors; however, the precision of the
FIG. 1. HRTEM image of the BST film with 52% Ti (planar view). A
region with amorphous-like contrast is indicated by the arrows.
FIG. 3. Line profiles of Ti/Ba ratio across the grain boundaries in (a) the
50.7% and (b) the 53.4% specimens. The line profile for the 53.4%
FIG. 2. Experimental maps of (a) Ti–L2,3 and (b) Ba–M4,5 absorbtion
edge intensity distributions and (c) the calculated map of Ti/Ba ratio
for the 50.7% film. Lighter contrast in the maps corresponds to a
higher intensity of the corresponding absorbtion edges (or Ti/Ba ratio).
specimen was taken across a grain boundary free of amorphous phase. (c)
Line profile across an amorphous region at a grain boundary junction in
the 53.4% specimen. The grain boundaries are indicated by arrows. The
dip observed in (c) has no significance.
J. Mater. Res., Vol. 15, No. 7, Jul 2000
Downloaded: 29 Mar 2015
1435
http://journals.cambridge.org
IP address: 129.15.14.53

Rapid Communications
1
6,17
cores.
This might be achieved either by an excess of
must be some additional factors contributing to the drop
in permittivity. One such factor could be a dependence of
the Ba/Sr ratio in the grain interiors on the overall Ti/
(Ba + Sr) ratio in the film; more precise measurements
of the Sr content (i.e., Ti/Sr ratio) in the grain interiors
are necessary to clarify this issue.
Ti ions or by oxygen vacancies present at concentrations
1
6
higher than those of the Ba/Sr vacancies; in both
cases an increased Ti/O ratio at the grain boundary core
is expected. Surprisingly, measurements of the Ti/O ratio
by Stemmer et al. for a 50.7% film performed at much
higher spatial resolution (a probe size of less than 0.4 nm
has been quoted), did not reveal any visible change in
Ti/O ratio at the grain boundaries. At present, there is no
explanation for this discrepancy.
These results demonstrate that the BST films with Ti
contents ranging from 50.7% to 53.4% consist of at least
two phases with distinct chemical composition: grain in-
teriors and grain boundaries. Assuming a hexagonal
prism shape for grains with an average diameter of
22
11
ACKNOWLEDGMENTS
The specimens for this study were kindly provided by
P.C. van Buskirk, S. Bilodeau, and R. Carl of Advanced
Technology Materials, Inc. The discussions with S.K.
Streiffer (Argonne National Laboratory) are acknowl-
edged. The use of brand or trade names does not imply
endorsement of the product by NIST.
1
5 nm, the volume fraction of grain boundaries with a
18
thickness t ס 1 nm is estimated to be 18%. The grain
boundaries in the present films are connected mostly in
parallel to the grain interiors, and therefore their effect
on the dielectric constant is expected to be roughly pro-
portional to the volume fraction. Because the grain size is
independent of the average composition, the grain
boundaries alone cannot account for the strong compo-
sitional dependence of dielectric constant, even though
the thickness of the space charge regions can vary with
changes in the overall Ti/(Ba + Sr) ratio. As the Ti/
REFERENCES
1. T. Kawahara, M. Yamamuka, T. Makita, J. Naka, A. Yuuki,
N. Mikami, and K. Ono, Jpn. J. Appl. Phys. 33, 5129 (1994).
2
. P.C. van Buskirk, J.F. Roeder, and S. Bilodeau, Integr. Ferro-
electr. 10, 9 (1995).
3
. M. Yoshida, H. Yamaguchi, T. Sakuma, and Y. Miyasaka,
J. Electrochem. Soc. 142, 244 (1995).
4. T. Horikawa, N. Mikami, T. Makita, J. Tanimura, M. Kataoka,
K. Sato, and M. Nunoshita, Jpn. J. Appl. Phys. 32, 4126 (1993).
5
6
. S. Yamamichi, J. Yabuta, T. Sakuma, Y. Miyasaka, Appl. Phys.
Lett. 64, 1644 (1994).
. C.S. Hwang, S.O. Park, H-J. Cho, C.S. Kang, S.I. Lee, and M.Y.
Lee, Appl. Phys. Lett. 67, 2819 (1995).
(Ba + Sr) ratio is increased, a third, amorphous phase,
containing Ba, Ti, O (and probably Sr) evolves at some
of the grain boundaries and multiple grain junctions.
Analysis of structural images suggests that the volume
fraction of the amorphous phase increases from about 1%
to 5–6% with an increase in the Ti content from 52% to
7. J.S. Horwitz, W. Chang, A.C. Carter, J.M. Pond, S.W. Kirchoefer,
D.B. Chrisey, J. Levy, and C. Hubert, Integrated Ferroelectrics 22,
7
99 (1998).
8
9
. A.I. Kingon, S.K. Streiffer, C. Basceri, and S.R. Summerfelt,
MRS Bull. 21(7), 46 (1996).
53.4%. No significant effects are expected for such small
. C. Basceri, S.E. Lash, C.B. Parker, S.K. Streiffer, A.I. Kingon,
M. Grossman, S. Hoffman, M. Schumacher, R. Waser, S. Bilo-
deau, R. Carl, P.C. van Buskirk, and S.R. Summerfelt, in Ferro-
electric Thin Films VI, edited by R.E. Treece, R.E. Jones, C.M.
Foster, S.B. Desu, and I.K. Yoo (Mater. Res. Soc. Symp. Proc.
amounts of amorphous phase connected in parallel; how-
ever, this phase connected in series could strongly affect
the dielectric constant. Because a large fraction of the
amorphous phase is present in the form of pockets (see
cross-sectional HRTEM images in Ref. 12), its contribu-
tion to the dielectric constant is expected to be between
the two extreme cases. A possible effect of the amor-
phous phase can be estimated by assuming that the ef-
4
93, Warrendale, PA, 1998), pp. 9–14.
1
1
1
0. S.K. Streiffer, C. Basceri, C.B. Parker, S.E. Lash, and A.I. Kin-
gon, J. Appl. Phys. 86, 4565 (1999).
1. S. Stemmer, S.K. Streiffer, N.D. Browning, and A.I. Kingon,
Appl. Phys. Lett. 74, 2432 (1999).
2. I. Levin, R.D. Leapman, D.L. Kaiser, P.C. van Buskirk, S. Bilo-
deau, and R. Carl, Appl. Phys. Lett. 75, 1299 (1999).
fective dielectric constant, ⑀ , of the BST films is
eff
1
represented by ⑀ ס ⁄
(
2
(⑀ + ⑀ ) [⑀ ס f ⑀ + (1 − f )⑀
eff
p
s
p
a a
a
c
13. P.C. van Buskirk, J.F. Roeder, and S. Bilodeau, Integr. Ferro-
electr. 10, 9 (1995).
−1
−1
−1
⑀ ) ס f (⑀ ) + (1 − f )(⑀ ) , where ⑀ and ⑀ are di-
s a a a c a c
electric permittivities of the amorphous and the crystal-
14. S.M. Bilodeau, R. Carl, P.C. van Buskirk, J.F. Roeder, C. Basceri,
S.E. Lash, C.B. Parker, S.K. Streiffer, and A.I. Kingon, J. Korean
Phys. Soc. 32, S1591 (1998).
line phases, respectively, and f is a volume fraction of
a
the amorphous phase], where ⑀ and ⑀ are dielectric
p
s
1
1
5. R.D. Leapman and D.E. Newbury, Anal. Chem. 65, 2409 (1993).
6. Y-M. Chiang and T. Takagi, J. Am. Ceram. Soc. 73, 3278 (1990).
permittivities calculated for the crystalline and the amor-
phous components connected in parallel and in sequence,
17. S.B. Desu and D.A. Payne, J. Am. Ceram. Soc. 73, 3391 (1990).
19,20
1
8. R.T. DeHoff and F.N. Rhines, Quantitative Microscopy (Mc-
respectively. Then, an amorphous phase with ⑀ ס 15
and a volume fraction of 5% would reduce the effective
Graw-Hill, New York, 1968).
10
19. B.S. Chiou and M.C. Lin, Thin Solid Films 248, 247 (1994).
0. Z.Q. Shi, Q.X. Jia, and W.A. Anderson, J. Vac. Sci. Technol. A,
0, 733 (1992).
dielectric constant from ␧ ס 280 for the 50.7% speci-
men (no amorphous phase) to ⑀ ס 208 for the 53.4%
specimen. This value is still significantly larger than the
2
1
21. H. Funakubo, Y. Takeshima, D. Nagano, K. Shinozaki, and
10
ס 160 measured for the 53.4% specimen; thus, there
N. Mizutani, J. Mater. Res. 13, 3512 (1998).
⑀
1
436
J. Mater. Res., Vol. 15, No. 7, Jul 2000
Downloaded: 29 Mar 2015
http://journals.cambridge.org
IP address: 129.15.14.53