You are on page 1of 10

EDITOR’S CHOICE

www.pss-b.com

Crystal Phase Distribution and Ferroelectricity in Ultrathin


HfO2–ZrO2 Bilayers
Martin E. McBriarty,* Vijay K. Narasimhan, Stephen L. Weeks, Ashish Pal,
Huazhi Fang, Trevor A. Petach, Apurva Mehta, Ryan C. Davis, Sergey V. Barabash,
and Karl A. Littau

but requires a ferroelectric memory ele-


(Hf,Zr)O2 ultrathin films are used as ferroelectric layers in emerging digital logic ment.[2] To make these devices reliable
and nonvolatile memory devices. The ferroelectric properties of (Hf,Zr)O2 can be and cost effective, ferroelectric materials
improved by interface engineering, such as the formation of nanolaminates which are compatible with standard semi-
conductor processing technology are needed.
with distinct HfO2 and ZrO2 layers. Herein, the ferroelectric performance of
HfO2 and ZrO2 are of particular interest
HfO2–ZrO2 ultrathin bilayer devices is shown to depend on the stacking order of as ferroelectric materials for these applica-
HfO2 and ZrO2, which affects the quantity of the noncentrosymmetric ortho- tions.[3,4] These oxides are widely used as
rhombic Pca21 crystal phase. By combining X-ray diffraction with a novel dielectrics in microelectronics thanks to
extended X-ray absorption fine structure (EXAFS) analysis procedure, the their high permittivity, wide band gap,
and ease of fabrication of conformal,
orthorhombic, tetragonal, and monoclinic phase fractions are quantified for
high-quality ultrathin films by atomic layer
bilayers composed of 3 nm HfO2 and 3 nm ZrO2. A significantly larger ortho- deposition (ALD). The functional proper-
rhombic ZrO2 phase fraction is found when ZrO2 has an unconstrained surface ties of these oxides are linked to their crys-
during annealing, whereas the presence of a ZrO2 interface with the substrate tal structure.[3] Pca21 (orthorhombic, oFE)
results in a substantial tetragonal ZrO2 phase fraction and a 2.4 smaller and the recently discovered R3 (rhombohe-
remanent polarization. HfO2 is found to be less susceptible than ZrO2 to crystal dral, rFE) crystal structures are noncentro-
symmetric and demonstrate ferroelectric
phase templating. The methods presented herein enable mechanistic studies of
behavior.[5–10] However, the ferroelectric
ferroelectric wake-up, fatigue, and processing effects in (Hf,Zr)O2 films, accel- crystal phases are unstable in pure bulk
erating the development of electronic devices that rely on ferroelectric oxides. HfO2 and ZrO2 at room temperature,
requiring dopants, defects, or interface
engineering to produce functioning devi-
ces.[3,8,10–12] Other technologically interest-
1. Introduction ing polymorphs, including the P42/nmc (tetragonal, t) and Fm3̄m
(cubic, c) phases, are also unstable in pure bulk HfO2 and ZrO2
Ferroelectric materials are key components of emerging electronic relative to the P21/c (monoclinic, m) phase at standard tempera-
devices. Negative capacitance field-effect transistors (NC-FETs) ture and pressure. Thus, the desirable properties of (Hf,Zr)O2
rely on a ferroelectric gate insulator in combination with a standard films must be unlocked by altering the composition, thickness,
dielectric to enable <60 mV decade1 subthreshold slopes and processing conditions, and choice of substrate or capping layer to
lower operating voltages, yielding smaller logic devices operating favor a particular crystal structure.
HfxZr1-xO2 films with x  0.5 show good ferroelectric perfor-
at a lower power.[1] Ferroelectric random access memory (FeRAM)
mance, with a maximum remanent polarization (2Pr) of about
offers faster switching than conventional nonvolatile memory
34 μC cm2 for a 9 nm film.[13] ZrO2/HfO2 nanolaminate struc-
tures yield even larger remanent polarization than (Hf,Zr)O2
Dr. M. E. McBriarty, Dr. V. K. Narasimhan, Dr. S. L. Weeks, Dr. A. Pal, solid solutions of the same overall composition.[14,15] The
Dr. H. Fang, Dr. S. V. Barabash, Dr. K. A. Littau presence of solid–solid interfaces between amorphous HfO2
Technology Group
Intermolecular, Inc. or ZrO2 films can either arrest[16] or template[17] crystallization
3011 N. 1st Street, San Jose, CA 95134, USA during annealing, depending on the structure of the adjacent
E-mail: martin.mcbriarty@intermolecular.com material. Thus, the interfaces between HfO2 or ZrO2 and a
Dr. T. A. Petach, Dr. A. Mehta, Dr. R. C. Davis substrate, electrode, or annealing gas (i.e., an unconstrained
Stanford Synchrotron Radiation Lightsource surface) affect the final distribution of crystal phases, and there-
SLAC National Accelerator Center fore the ultimate ferroelectric performance, in a nanolaminate.
Menlo Park, CA 94025, USA
Ultrathin ALD HfO2 films are amorphous as deposited on tech-
The ORCID identification number(s) for the author(s) of this article nologically relevant substrates at deposition temperatures at or
can be found under https://doi.org/10.1002/pssb.201900285. below 300  C.[14,17,18] ZrO2 is more prone to crystallization than
DOI: 10.1002/pssb.201900285 HfO2, and thin ZrO2 films readily crystallize to the t phase.[16,17]

Phys. Status Solidi B 2020, 257, 1900285 1900285 (1 of 10) © 2019 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.pss-b.com

Developing robust devices based on (Hf,Zr)O2 ferroelectrics Martin E. McBriarty earned his
demands an understanding of the links between the crystal degrees in Materials Science and
structure, processing/annealing conditions, and changes to the Engineering (B.S., University of
material during electrical cycling.[15,19,20] In previous studies on Florida, 2008; Ph.D., Northwestern
(Hf,Zr)O2, a predictor of ferroelectric performance has been the University, 2014). His doctoral
suppression of the m crystal phase as indicated by grazing inci- research focused on heterogeneous
dence X-ray diffraction (GIXRD),[12,21,22] but this measurement catalysis, and in 2011 he won a
does not uniquely reveal the presence or abundance of the oFE Fulbright fellowship to study at the
phase attributed to the ferroelectric performance of these devices. Fritz Haber Institute. In 2014, he
Although the m phase has a distinctive double diffraction peak cor- began postdoctoral research on
responding to the {111} and {1̄11} families of diffracting planes, interfacial geochemistry at Pacific Northwest National
the t and oFE phases have similar diffraction patterns which can Laboratory. In 2018, he joined Intermolecular, Inc. as a
only be distinguished when films are sufficiently thick or when the scientist, where he investigates atomic layer deposition
substrate composition is well controlled.[10,13,23] Real devices may processes for electronic devices.
use films thinner than 10 nm deposited on crystalline electrodes
such as TiN, which interfere further with peak discrimination and Vijay K. Narasimhan is a senior staff
phase identification. Although attempts have been made to decom- research engineer at Intermolecular,
pose the diffraction features of the oFE and t phases of doped Inc., where he develops deposition,
HfO2[10] and (Hf,Zr)O2[24] films, crystal phase distributions have characterization, and integration
not been measured for ferroelectric (Hf,Zr)O2 films thinner than processes for emerging electronic
10 nm. Even under conditions in which these crystal phases can be materials, including ferroelectrics,
distinguished by GIXRD, the contributions of each material in a chalcogenides, and III–V
HfO2/ZrO2 nanolaminate are probably indistinguishable. semiconductors. Previously, he was an
Despite their similar diffraction patterns, the t and oFE international Fulbright Science and
structures of HfO2 and ZrO2 have distinct cation coordination Technology Fellow at Stanford
structures. The oFE phase has a similar local structure to the m University, where he completed a Ph.D. in Materials
phase, with seven metal–oxygen bonds with different lengths Science and Engineering in 2015. His doctoral research
within the first coordination shell. In contrast, the t local structure focused on the optical and electrical properties of
consists of two distorted tetrahedra, with four short bonds and nanostructured surfaces. He earned an M.Phil in Micro-
four long bonds in the first coordination shell. Methods sensitive and Nanotechnology Enterprise from the University of
to local structure instead of periodicity, such as extended X-ray Cambridge and a B.A.Sc. in Computer Engineering from
absorption fine structure (EXAFS) spectroscopy, can better resolve the University of Ottawa in Canada.
these crystal phases. An EXAFS spectrum encodes the coordina- Karl A. Littau is a physical chemist
tion structure(s) of one element in a sample as an oscillation in the with expertise in thin-film deposition,
X-ray absorption caused by photoelectron scattering from neigh- ALD, CVD, surface and gas phase
boring atoms. EXAFS is typically analyzed using shell-by-shell fit- chemistry, and solid state physics. He
ting, in which the atoms surrounding an absorbing atom are holds a B.S. in Chemistry from UC
modeled as coordination shells described by an average bond Berkeley and a Ph.D. in Physical
length, coordination number, and disorder parameter. Such an Chemistry from Stanford University.
approach works well for structures with high local symmetry, He has held positions at AT&T Bell
but statistical limitations hinder the accurate fitting of EXAFS Laboratories, Applied Materials, Inc.,
spectra which encode a complex local structure or spectral contri- and Xerox’s Palo Alto Research
butions from multiple coexisting phases.[25] Previous studies used Center, Stanford University, and is currently CTO at
EXAFS linear combination fitting to discriminate phases of HfO2 Intermolecular Inc. where he has been directing research
based on photoelectron scattering contributions from Hf–Hf inter- and development of new materials and methods for
actions,[26] but these fits mostly neglected the useful structural advanced IC devices and emerging technologies.
information in the first Hf–O shell and used basis spectra which
did not account for thermal disorder, inspiring the development of
more accurate approaches to (Hf,Zr)O2 phase identification. and oFE phases of ZrO2 and HfO2, providing a new route for
Here, a novel EXAFS analysis method is shown to distinguish accurate nondestructive characterization and ferroelectric prop-
the t and oFE phases of 3 nm ZrO2 / 3 nm HfO2 bilayers grown erty prediction for these technologically important materials.
by ALD, and the measured crystal phase distribution is linked to
ferroelectric performance. EXAFS “fingerprints” for different
crystal structures were simulated for known phases, validated 2. Results
by comparison with experimental spectra, and then added in lin-
ear combination to fit experimental data encoding unknown crys- 2.1. Thin Film Characterization
tal phase distributions. GIXRD results were used to further
distinguish the m-HfO2 and oFE-HfO2 phases. This combined Bilayer stacks of ZrO2 and HfO2 were grown by ALD on Si(100)
approach reliably determines the phase fractions of the m, t, substrates with native oxide. The target thickness for each bilayer

Phys. Status Solidi B 2020, 257, 1900285 1900285 (2 of 10) © 2019 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.pss-b.com

component was 30 Å, and growth proceeded as substrate/HfO2/ for each crystal phase from each material cannot be resolved
ZrO2 (sample SHZ) and substrate/ZrO2/HfO2 (sample SZH). by XRD alone.
Samples selected for further study were annealed at 500  C
for 10 min in N2.[15] The thickness, density, and interface mixing 2.2. X-Ray Absorption Spectroscopy
of the HfO2–ZrO2 bilayers were determined by X-ray reflectivity
(XRR). XRR data and fits are shown in Figure S1 and Table S1, 2.2.1. X-ray Absorption Near Edge Structure
Supporting Information. The measured densities agree with
bulk values. Film thicknesses for sample SZH were 29.7 Å For some oxides, phase fractions can be measured by analyzing
(ZrO2) and 29.0 Å (HfO2) and for sample SHZ were 27.3 Å features in the X-ray absorption near edge structure (XANES). Zr
(HfO2) and 31.5 Å (ZrO2), all close to the target thicknesses of K-edge and Hf L3-edge XANES spectra for samples SHZ and
30 Å. The interface roughness values of about 3 Å indicate mini- SZH, as well as reference spectra, are shown in Figure S4,
mal intermixing between the ZrO2 and HfO2 layers. Supporting Information. Each HfO2 XANES spectrum features
GIXRD patterns for the as-deposited and annealed bilayer a strong white line, but they are difficult to distinguish from
stacks are shown in Figure S2, Supporting Information. The m-HfO2, regardless of whether the samples had been annealed.
as-deposited films show no evidence of crystallinity (an artifact ZrO2 XANES are more clearly distinguishable, but linear combi-
at 2θ ¼ 44.5 is present for all samples). The annealed films show nations of m-ZrO2 and t-ZrO2 standard spectra did not match the
several peaks which can correspond to either the t or oFE phase of salient features of annealed SHZ or SZH. To the authors’ knowl-
ZrO2 and HfO2 or some combination thereof. There is a small edge, no XANES spectra for the oFE phases of HfO2 or ZrO2 have
contribution of the m phase which is apparent as the 1̄11m and been published, and the accurate simulation of XANES spectra is
111m shoulders surrounding the 111o/011t peak. XRD peaks beyond the scope of this work. Further analysis focuses on the
were fit over 20 < 2θ < 42 as Gaussian curves atop a decaying EXAFS region of the X-ray absorption spectra.
exponential background. Five peaks were fit, labeled in Figure 1.
The widths of the 1̄11m, 111m, and 111o/011t peaks were kept 2.2.2. EXAFS Simulation
equal. The 111m/1̄11m intensity ratio was fixed to 0.744, which
is the theoretical ratio based on the calculated structure, Lorentz, EXAFS linear combination fitting (LCF) is used to determine the
and polarization factors (see Table S2 and S4, Supporting phase fraction distribution of the 3 nm ZrO2 and HfO2 films
Information). Temperature factors are assumed to be the same comprising the bilayers of interest. LCF assumes that an
for the analyzed reflections. unknown experimental spectrum consists of a weighted sum
XRD fit results are shown in Table S2, Supporting of basis spectra, where each spectrum in the basis set corre-
Information. Using the Scherrer equation with a shape factor sponds to a unique crystal phase. The basis set of EXAFS spectra
of 0.9, a characteristic crystallite size of about 6 nm is estimated, may be simulated starting from experimental or theoretical crys-
which matches the total bilayer thickness and supports previous tal structures. The details of EXAFS simulation are covered else-
findings of crystalline continuity between HfO2 and ZrO2.[15,17] where,[27–29] but a key point is that physically reasonable EXAFS
The monoclinic phase fraction is higher for sample SZH. simulations need to include accurate disorder parameters (σ2)
Because of the strong overlap between the 111o and 011t peaks and amplitude reduction factors (S02). Disorder damps the ampli-
of both ZrO2 and HfO2, as well as the overlap between each tudes of photoelectron scattering paths in proportion to the width
monoclinic peak ([1]11m) of both materials, the contributions of the time-averaged path length distribution, which includes

Figure 1. Fits (aqua dashed lines) to the GIXRD patterns in the vicinity of 2θ ¼ 30 for samples SHZ (red dots) and SZH (blue dots). Both samples were
annealed at 500  C in N2 for 10 min.

Phys. Status Solidi B 2020, 257, 1900285 1900285 (3 of 10) © 2019 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.pss-b.com

both thermal and configurational contributions. Thermal disor- which is typical for nanophase materials below 10 nm.[31]
der can be simulated directly and accurately using first-principles Thus, to parameterize disorder for nanoscale ZrO2, simulated
molecular dynamics[25,30]; however, these simulations can be oFE-ZrO2 spectra were fit against the Zr EXAFS spectrum for
prohibitively expensive. Configurational disorder due to strain sample SHZ. Keeping S02 ¼ 0.92, the resulting disorder para-
fields or interface effects can further dampen the EXAFS of nano- meterization was σ2 ¼ 0.0016*R þ 0.0002 Å2. LCFs to sample
materials.[31] S02 accounts for many-body electronic relaxation SHZ with other ZrO2 phases using the nanoscale disorder para-
effects, but since it is difficult to calculate ab initio, it is typically meterization yielded a 91%  2% oFE-ZrO2 phase fraction
treated as an EXAFS fit parameter.[27] (see Section 2.2.3); therefore, the systematic error on the
In this work, σ2 is parameterized as a linear function of the EXAFS-derived ZrO2 phase fractions is assumed to be 9%.
photoelectron scattering path length, with S02 as a scale factor Disorder parameters for HfO2 were determined by fitting sim-
(see Section 5 for details). A key assumption is that the same ulated EXAFS spectra to an experimental spectrum for a 9.6 nm
disorder parameterization and S02 may be used for the m, t, m-HfO2 film. This sample was grown by ALD on TiN and then
and oFE crystal phases for the same material, crystallite size, capped by ALD TiN at 430  C, yielding a m-HfO2 fraction of
and measurement temperature. Thus, separate disorder param- 90%  1% and a (t þ oFE)-HfO2 fraction of 10%  1% according
eterizations must be established for nanoscale ZrO2 and to XRD (Figure S3, Supporting Information). The consequences
nanoscale HfO2 by comparison with experimental standard of this minority phase fraction on the m-HfO2 EXAFS spectrum
spectra. depend on its identity as t-, oFE-, or some phase combination
ZrO2 thermal disorder parameters were determined in two thereof. The local structure of oFE is similar to that of m, so,
steps. First, simulated m-ZrO2 spectra were fit to an experimental a minority oFE phase fraction would barely affect the EXAFS
spectrum for μm-scale m-ZrO2,[32] yielding σ2 ¼ 0.0012*R þ spectrum; however, the introduction of a minority t phase would
0.0003 Å2 for bulk m-ZrO2. The fitted S02 value (0.92) is higher dampen the EXAFS spectrum, resulting in larger σ2 or smaller
than reported values of around 0.75,[33] although successful S02 than would be obtained for a pure m-HfO2 film. Indeed, the
EXAFS fits to ZrO2 have also been obtained with S02 ¼ 1.[34] fitted S02 ¼ 0.80 is somewhat smaller than prior values,[35] and
Using this initial parameterization, LCFs to sample SHZ with σ2 ¼ 0.0013*R  0.0011 Å2. However, the smaller S02 value does
the lowest R-factor included contributions from oFE-ZrO2 not definitively establish a t-HfO2 minority phase; therefore, a
(73%–74%) and the EXAFS spectrum from the nominally systematic error of 10% is included in all EXAFS-derived phase
amorphous as-deposited SHZ or SZH samples (26%–27%) fraction measurements for HfO2.
(Figure S5, Supporting Information). Because the amorphous Figure 2 shows simulated and experimental reference EXAFS
phase fraction has a small EXAFS amplitude, the primary spectra for the m, oFE, and t phases of HfO2. The simulated
effect of including a significant amorphous component in the m-HfO2 spectrum closely matches the experimental spectrum
LCF is to dampen the amplitude of the fitted spectrum. to which it was calibrated. Discrepancies at k < 2 Å1 may be
However, it is likely that the smaller amplitude of the experi- attributed to artifacts in the experimental spline fit, and
mental spectrum actually indicates configurational disorder, χ(k) ¼ 0 below k ¼ 1.0 Å1 because of the positive ΔE0 offset

Figure 2. k3-weighted EXAFS spectra (a) and phase-uncorrected Fourier transforms over 3.5 < k < 11.0 Å1 (b) for m (top), oFE (center), and t (bottom)
phases of HfO2. Simulated spectra with the nanoscale disorder parameterization are shown as black solid lines; the red dashed lines correspond to
experimental spectra for the 9.6 nm >90% m-HfO2 film.

Phys. Status Solidi B 2020, 257, 1900285 1900285 (4 of 10) © 2019 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.pss-b.com

in the simulation (þ4 eV). The differences between m and oFE bonds in the first coordination shell.[36] This differs from the
spectra are subtle. While the nodes and antinodes of the EXAFS broad, higher magnitude first coordination shell feature for
occur at similar k values, there are differences in the peak ampli- t-HfO2 (Figure 2b); the larger path length difference between
tudes, particularly at higher k values, where the amplitude enve- Hf–O subshells (2.06 and 2.40 Å) results in more constructive inter-
lope is larger for oFE. The t-HfO2 EXAFS shows more structure ference in the sum of their scattering contributions. The experi-
at low k, but at high k the amplitude is much larger than that mental t-ZrO2 spectrum[33] shows a somewhat different first-
for the m or oFE phase. The differences are also clear in the shell structure than the simulated spectrum, with a larger peak
real-space EXAFS Fourier transforms (FTs), especially in at 1.5 Å but a smaller feature at 2.0–2.2 Å. This could be attributed
the Hf–Hf shell (2.9 < R < 4.0 Å) as previously observed.[26] to a significantly larger degree of disorder for the longer Zr–O sub-
The larger magnitude of this shell for t-HfO2 indicates its greater shell[34] which is not captured in the linear parameterization of the
local order. Note that the EXAFS FTs do not directly represent disorder as a function of the photoelectron path length. Low static
radial distributions of atoms. Features in the FTs are offset by disorder in the Zr–Zr shell leads to a significantly larger magnitude
a few tenths of an Å negative of the real coordination distances of the EXAFS FT relative to the m and oFE phases.
due to photoelectron wave phase shifts, and destructive interfer-
ence between photoelectron scattering contributions may
dampen features which would otherwise have larger magnitudes 2.2.3. EXAFS Fitting
in a radial distribution function.[25]
Similar trends are observed between phases of ZrO2. Simulated EXAFS LCF results are given in Table 1, and fits for the HfO2 and
spectra with the nanoscale disorder parameterization and a bind- ZrO2 components are shown in Figure 4 and 5, respectively.
ing energy offset ΔE0 ¼ 8 eV are shown in Figure 3 along These results were selected as the set of best fits, i.e., their
with spectra simulated using the bulk disorder parameterization, R-factors are within 1.25 the minimum (absolute best fit)
which agree well with experimental spectra for bulk m-ZrO2 and R-factor from all permutations of LCFs using the m, t, and/or
t-ZrO2. The m and oFE phases have an overall similar structure, oFE crystal phases and with all included phases contributing
although broader features around 7.8 < k < 10.0 Å1 distinguish phase fractions of at least 5%. For every experimental spectrum,
m-ZrO2 from oFE-ZrO2. The t-ZrO2 spectrum is characterized by the fits reveal a significant oFE phase fraction. The ZrO2 compo-
a large amplitude at high k. In the EXAFS FTs, the m and oFE nent of sample SHZ is nearly completely oFE; good fits can be
phases are characterized by two relatively narrow peaks signify- achieved with about 9% of either m- or t-ZrO2. There are also two
ing the Zr–O and Zr–Zr coordination shells. In contrast, the fits of similar quality for the HfO2 component of sample SHZ,
t-ZrO2 EXAFS FT has a small magnitude in the region corre- which both include a minority t-HfO2 phase. While the fit to this
sponding to Zr–O bonding and a large magnitude for the spectrum with the lowest R-factor includes all three phases, there
Zr–Zr shell. The small Zr–O peak magnitude is due to the is another fit slightly higher in R-factor which only includes t- and
destructive interference of photoelectron scattering contribu- oFE-HfO2. These two fits highlight the strong similarity in the
tions[25] from four short (2.10 Å) and four long (2.35 Å) Zr–O Hf L3 EXAFS spectra of the m and oFE phases. While fits were

Figure 3. k3-weighted EXAFS spectra (a) and phase-uncorrected Fourier transforms over 3.5 < k < 11.0 Å1 (b) for m (top), oFE (center), and t (bottom)
phases of ZrO2. Simulated spectra with the nanoscale disorder parameterization are shown as black solid lines. Simulated spectra with the bulk disorder
parameterization (blue dotted lines) are shown along with experimental spectra (red dashed lines) for bulk m-ZrO2[32] and t-ZrO2.[33]

Phys. Status Solidi B 2020, 257, 1900285 1900285 (5 of 10) © 2019 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.pss-b.com

Table 1. Phase Fraction Results from EXAFS LCFs.

Element Sample R-factor Monoclinic Orthorhombic (FE) Tetragonal


f a)
ΔE0 [eV]b)
f ΔE0 [eV] f ΔE0 [eV]

Hf SZH 0.0379 0.27  0.11 4 0.64  0.11 6 0.09  0.11 12


SHZ 0.0772 0.19  0.11 4 0.67  0.12 6 0.15  0.10 8
0.0848 – – 0.84  0.10 6 0.16  0.10 8
Zr SZH 0.0722 – – 0.59  0.09 6 0.41  0.09 6
SHZ 0.0544 – – 0.92  0.09 8 0.09  0.09 0
0.0610 0.09  0.09 4 0.91  0.09 10 – –

a)
Phase fraction errors include the error from the nonlinear least-squares fit added in the quadrature with systematic errors determined from the disorder parameterization for
each material; b)Best-fit binding energy offset.

Figure 4. Experimental Hf L3 EXAFS spectra (a) and Fourier transforms (b), as well as best fits corresponding to those in Table 1, for the HfO2 component
of bilayer samples SZH and SHZ.

Phys. Status Solidi B 2020, 257, 1900285 1900285 (6 of 10) © 2019 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.pss-b.com

Figure 5. Experimental Zr K EXAFS spectra (a) and Fourier transforms (b), as well as best fits corresponding to those in Table 1, for the ZrO2 component
of component of bilayer samples SZH and SHZ.

performed only over 3.5 < k < 11.0 Å1, the fits and data gener- 3. Discussion
ally agree at k values higher than the fitted range.
3.1. Structural Analysis

2.2.4. Ferroelectric Performance Phase fraction ranges, given in Table 2, were determined as the
upper and lower limits of the uncertainty bounds of each phase
HfO2–ZrO2 bilayers were grown using the same procedures as fraction in Table 1. The oFE-ZrO2 phase is clearly dominant in
samples SHZ and SZH on a conductive TiN layer (samples sample SHZ, whereas there is a significant phase fraction of
STHZ and STZH, respectively) and annealed at 500  C in N2 t-ZrO2 in sample SZH. All other measurements reveal a
for 10 min, followed by deposition of TiN top contacts at small-to-nonexistent t phase fraction. The conservative formula-
250  C. Electric polarization curves for these samples, following tion of uncertainty used here leads to large phase fraction ranges
mild wake-up cycling, are shown in Figure 6. Both samples show for m- and oFE-HfO2 in sample SHZ, as expected because the
clear ferroelectric behavior: sample STHZ has a remanent polar- EXAFS signals from these two phases are nearly degenerate.
ization 2Pr of 24.5 μC cm2, whereas sample STZH’s 2Pr is Statistically significant differences in the HfO2 phase fraction
10.2 μC cm2. between samples are therefore unresolvable by EXAFS alone,

Phys. Status Solidi B 2020, 257, 1900285 1900285 (7 of 10) © 2019 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.pss-b.com

20
SZH. To understand these phenomena, prior work is surveyed
to relate interfacial phenomena and ferroelectricity in ultrathin
(Hf,Zr)O2 films prepared under conditions in which interfaces
dominate the crystallization behavior (i.e., few-nm films proc-
10 essed at or below 500  C).
Polarization [ µC cm-2 ]

In an early study of HfO2–ZrO2 bilayers, Kim et al.[17] found


that ZrO2 deposited by ZrCl4/H2O ALD onto SiOx at 300  C
0 immediately crystallized into the tetragonal phase, whereas
HfO2 deposited by HfCl4/H2O remained amorphous. Zhao
et al.’s[37] systematic study of ultrathin film crystallization with
annealing temperature also proved that ZrO2 crystallizes at sig-
-10 nificantly lower temperatures than HfO2. Figure S2, Supporting
STZH
STHZ Information, shows nearly identical diffraction patterns with no
detectable crystallinity for both bilayer samples as deposited at
-20 260  C, indicating an amorphous initial condition for both
bilayer configurations. Thus, the propensity of ZrO2 to crystallize
-4 -3 -2 -1 0 1 2 3 4
more readily than HfO2 is expected to play a significant role in
Electric Field [MV cm-1] the crystal phase distribution after annealing.
Figure 6. Ferroelectric polarization curves for HfO2/ZrO2 bilayers with In their detailed study on HfO2–ZrO2 bilayers, Lu et al.[14]
different stacking orders. found that increasing the thickness of a ZrO2 capping layer from
1.5 to 6 nm atop a 6 nm HfO2 layer on Pt led to an increasing total
bilayer phase fraction of t and/or oFE versus m. Similarly, this
although more careful choices of reference spectra can improve work shows a much larger fraction of m-HfO2 when HfO2
the precision of future measurements by reducing systematic has an unconstrained surface (sample SZH) versus an interface
error. with underlying Si (sample SHZ). Although Lu et al.[14] sug-
To better distinguish oFE-HfO2 from m-HfO2, quantitative gested that the confinement of HfO2 enhances the oFE phase
XRD analysis is constrained using the more clearly resolved fraction in HfO2, their electron microscopy analysis neglected
t-HfO2 and m-, oFE-, and t-ZrO2 phase fractions from EXAFS. the ZrO2 phase, which is implicated in the enhanced ferroelectric
The analysis is discussed in detail in the Section S3, performance of the substrate/HfO2/ZrO2 stack. The larger
Supporting Information, and results are shown in square brack- t-ZrO2 phase fraction for sample SZH has no apparent templat-
ets in Table 2. This analysis yields a much smaller uncertainty ing effect on the t-HfO2 phase fraction, which is small for both
range on the m-HfO2 phase fraction and subsequently refines samples.
the oFE-HfO2 phase fraction ranges. Considering the aggregate By correlating Hf0.5Zr0.5O2 crystallite size and strain with fer-
phase fractions of ZrO2 and HfO2, sample SHZ has a signifi- roelectric performance, Park et al.[38] found that a higher density
cantly larger oFE phase fraction than sample SZH, which in turn of randomly oriented crystallites provides appropriate strain to
correlates with the improved ferroelectric performance of sample form the oFE phase. Such a condition can be expected from
STHZ (Figure 6). This matches the expectation that the noncen- nucleation at a free surface, as is the case for ZrO2 in sample
trosymmetric oFE phase is primarily responsible for ferroelectric SHZ, which is nearly completely in the oFE phase after anneal-
performance.[13] ing. Replacing the free interface of ZrO2 (sample SHZ) by an
interface with native SiOx (sample SZH) results in greater sus-
3.2. Interfacial Effects on Crystallization ceptibility to form the more stable t phase.
A consequence of the different deposition orders of samples
Table 2 shows that the surface ZrO2 layer is almost completely SHZ and SZH is that the bottom layer was held at the ALD tem-
(82%–100%) composed of the oFE phase for sample SHZ, perature (260  C) for a longer time (3200 s for SZH, 5400 s for
whereas there is 32%–50% t-ZrO2 in sample SZH, where SHZ) than the top layer. Under these conditions, few-nm seed
ZrO2 contacts the underlying Si. The oFE phase fraction of nuclei (likely t-ZrO2) might form, which would influence
HfO2 is less dramatically affected by deposition order, although crystallization during subsequent annealing.[39–41] At a higher
there is a somewhat larger minority m-HfO2 phase for sample ALD temperature (300  C), the formation of a crystalline ZrO2

Table 2. Summary of Crystal Phase Distributions.

Sample HfO2 ZrO2 Bilayer, totalb)


fma) foFE ft fm foFE ft fm foFE ft
c)
SZH 16–38 [28–32] 54–75 [50–70] 0–20 0–9 50–68 32–50 8–23 [14–20] 52–71 [50–69] 16–35
SHZ 0–30 [5–12] 54–94 [65–88] 5–26 0–18 82–100 0–18 0–24 [2–15] 67–97 [73–93] 2–22

a)
Phase fraction ranges are given as percentage values for the monoclinic ( fm), ferroelectric orthorhombic ( foFE), and tetragonal ( ft) phases of HfO2 and ZrO2; b)Total bilayer
phase fractions are given as sums of HfO2 and ZrO2 ranges weighted by the thickness of each layer; c)Values in square brackets are from XRD-constrained analysis.

Phys. Status Solidi B 2020, 257, 1900285 1900285 (8 of 10) © 2019 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.pss-b.com

“seed layer” was shown to improve 2Pr of a 10 nm Hf0.43Zr0.57O2 of (2θ)/2  0.5 and averaged). GIXRD measurements were performed on
film, indicating a higher oFE crystal phase fraction.[42] However, a Bruker D8 Discover diffractometer using a monochromated Cu X-ray
the EXAFS spectra of the as-deposited 3 nm HfO2 and ZrO2 tube (Cu Kα, λ ¼ 1.5418 Å).The incident beam angle was fixed at 0.7 ,
and XRD patterns were collected over the range 20 < 2θ < 110 in
films in the present work are approximately identical between
0.05 steps using a position-sensitive detector. XRD peaks were fit using
samples SHZ and SZH (Figure S5, Supporting Information), the lmfit package[47] in Python 3.6.
yielding no evidence of a difference in nucleation during depo- GIXAS measurements were performed at SSRL beamline 11-2.
sition. In summary, this analysis suggests that the interfacial Monochromatic X-ray energies were selected by a Si (220) double-crystal
nucleation of ZrO2 crystallites during annealing is key to the final monochromator calibrated to 9560.7 eV (Hf L3 edge) and 17997.6 eV
crystal phase distribution, and therefore the ferroelectric perfor- (Zr K edge) using the maximum of the first inflection point of the first deriv-
mance, of ultrathin HfO2–ZrO2 bilayers grown by ALD at 260  C. ative of spectra collected using Hf and Zr metal foils. Harmonic rejection
and collimation of the beam was done upstream of the monochromator
with a Rh-coated silicon mirror set with a 21 keV cut-off. The monochro-
mated beam was focused with a toroidal silicon mirror and in-hutch slits
4. Conclusions to achieve a vertical focus of 100 μm and a nominal spot width of
1 mm. Fluorescence signal was collected with a monolithic 100-element
A larger Pca21 orthorhombic phase fraction is linked to a Canberra germanium detector mounted at a 90 angle to the incident beam.
2.4 larger remanent polarization for a HfO2–ZrO2 bilayer with A Huber two-circle goniometer independently controlled the grazing
an unconstrained ZrO2 surface during annealing. Crystal phase incidence angle, which was selected to be sufficiently high to minimize
fractions were measured using a combined XRD and EXAFS self-absorption effects. The samples were azimuthally spun during measure-
analysis method which distinguishes signals for different crystal ments to smear out sharp diffraction peaks from the Si substrate.
EXAFS Analysis: EXAFS spectra were analyzed by LCF using a basis set
phases that are nearly degenerate in either measurement alone.
of simulated “fingerprint” spectra corresponding to single-phase (m, oFE,
By providing an accurate measurement of the ferroelectric phase or t) ZrO2 or HfO2 (Figure 2 and 3). These spectra were simulated from
fraction in ultrathin (Hf,Zr)O2 films, this analysis resolves a cru- static average atomic positions (structures described in Section S2,
cial problem in the evidence-directed development of devices that Supporting Information) using FEFF8-lite software[28,29] by summing
rely on ferroelectric materials. the EXAFS contributions of photoelectron scattering paths up to a cutoff
EXAFS measurements are element-selective and can pene- radius of 6.0 Å. The disorder σ2 for each photoelectron scattering path was
parameterized as a linear function of the effective photoelectron scattering
trate several μm of material; therefore, they can be used to probe
path length, Reff (σ2 ¼ aReff þ b). The best-fit coefficients a [Å] and b [Å2]
whole devices without interference from materials such as were determined by simulating an ensemble of EXAFS spectra for a crystal
TiN.[13] Measurements may be performed quickly over a limited phase (χs(k)) with different a and b values, then fitting each spectrum to an
k range,[30] enabling fast screening of ultrathin films or even in experimental EXAFS spectrum of a sample with the same crystal phase
situ mechanistic studies of devices under processing[43] or oper- (χe(k)) The best-fit coefficients were determined by minimizing an R-factor
ating[44] conditions. The “fingerprinting” method described between experimental and simulated k3-weighted EXAFS spectra over
herein enables the interpretation of complex or mixed-phase 3.5 < k < 11.0 Å1 (Equation (1)). The amplitude reduction parameter
S02 was simultaneously fitted as an amplitude scaling factor for the
materials which were previously inscrutable by EXAFS analysis simulated spectrum.
and can be extended to other spectroscopies which encode the
P
unique structural signatures of different crystal phases.[45,46] k ðk
3
χ ðkÞ  k3 S20 χ s ðkÞÞ2
R¼ Pe 3 (1)
k ðk χ e ðkÞÞ
2

5. Experimental Section Spectra for the m, oFE, and t crystal structures were simulated using the
best-fit a, b, and S02 parameters, and binding energy offsets ΔE0 were
Sample Preparation: To deposit 3 nm of HfO2 [ZrO2], 24 [41] ALD applied in increments of 2 eV to generate a basis set for LCF. Treating
cycles of Hf[N(CH3)2]4 (TDMA-Hf, Air Liquide) [(tris(dimethylamino)cyclo- ΔE0 as a variable accommodates uncertainty in defining k ¼ 0 Å1 during
pentadienyl)-zirconium (ZyALD), Air Liquide] and 4% ozone / 96% O2 the initial spline fit to the raw X-ray absorption spectrum μ(E). The under-
were run with a substrate temperature of 260  C. After ALD on lying assumption of the LCF method is that the structures used to generate
300 mm Si(100) wafers with a native oxide overlayer, the wafers were the basis set of spectra provide accurate distributions of bond lengths and
cleaved into 45 mm squares, of which some were set aside (the “as-depos- coordination numbers, so these are not treated as fit parameters (as is
ited” condition) and some were annealed at 500  C for 10 min in N2 in an typically done in shell-by-shell EXAFS analysis). However, due to the strong
Allwin21 rapid thermal processing tool. Two additional bilayer samples correlation between ΔE0 and bond length,[48] ΔE0 acts as an ad hoc param-
were prepared for electrical testing on a Si wafer with 300 nm thermal eter for isotropic distortions[25] such as strain or the 1% overestimation
oxide coated with 5 nm TiN by physical vapor deposition (PVD). Using of the simulated HfO2 lattice parameter due to PBE-GGA approximation.
P
the same ALD recipes as earlier, 3 nm HfO2 was deposited atop 3 nm LCFs were performed using Athena software[49] by fitting i f i k3 χ s,i ðkÞ
ZrO2 (sample STZH) or vice versa (sample STHZ), then annealed at to k χ(k) measured for films with an unknown crystal phase composition,
3

500  C for 10 min in N2. TiN top contacts (100 nm thick) were then depos- yielding the phase fractions fi for 2–3 simulated basis spectra χ s,i ðkÞ
ited by PVD at 250  C. The circular contacts (0.305 mm diameter; included in each fit. Fits were constrained such that 0 ≤ fi ≤ 1 and
0.073 mm2 area) were defined by a shadow mask. Σfi ¼ 1, and fit quality was assessed by an R-factor evaluated over the fitting
Electrical Characterization: Polarization curves were collected using a region of 3.5 < k < 11.0 Å1. Additional fits which included EXAFS spectra
Radiance Precision II ferroelectric tester and a Cascade probe station. from the amorphous as-deposited films were attempted during the initial
Polarization–voltage scans were collected with a triangular waveform disorder calibration, but the final fits assume a completely crystallized film.
(0.25 kHz) after applying a mild wake-up stress of 2.5 V at 1 kHz for 1 s.
X-Ray Measurements: XRR was measured at beamline 2-1 at the
Stanford Synchrotron Radiation Lightsource (SSRL) at SLAC National Supporting Information
Accelerator Laboratory using an 8.00 keV X-ray beam (43 μm wide).
Prior to analysis, the specular XRR measurement was corrected by sub- Supporting Information is available from the Wiley Online Library or from
tracting the background scattering (measured with an incident beam angle the author.

Phys. Status Solidi B 2020, 257, 1900285 1900285 (9 of 10) © 2019 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.pss-b.com

Acknowledgements [20] M. Pešić, F. P. G. Fengler, L. Larcher, A. Padovani, T. Schenk,


E. D. Grimley, X. Sang, J. M. LeBeau, S. Slesazeck, U. Schroeder,
The authors thank Prof. Oscar Juan Dura and Prof. Georges Calas for pro- T. Mikolajick, Adv. Funct. Mater. 2016, 26, 4601.
viding ZrO2 EXAFS reference spectra. Use of the Stanford Synchrotron [21] É. O’Connor, M. Halter, F. Eltes, M. Sousa, A. Kellock, S. Abel,
Radiation Lightsource, SLAC National Accelerator Laboratory, was sup-
J. Fompeyrine, APL Mater. 2018, 6, 121103.
ported by the US Department of Energy, Office of Science, Office of
[22] V. Mukundan, K. Beckmann, K. Tapily, S. Consiglio, R. Clark,
Basic Energy Sciences under Contract No. DE-AC02-76SF00515.
G. Leusink, N. Cady, A. C. Diebold, MRS Adv. 2019, 4, 545.
[23] T. S. Böscke, J. Müller, D. Bräuhaus, U. Schröder, U. Böttger, Appl.
Phys. Lett. 2011, 99, 102903.
Conflict of Interest [24] M. H. Park, H. J. Kim, Y. J. Kim, W. Lee, T. Moon, C. S. Hwang, Appl.
The authors declare no conflict of interest. Phys. Lett. 2013, 102, 242905.
[25] M. E. McBriarty, S. Kerisit, E. J. Bylaska, S. Shaw, K. Morris, E. S. Ilton,
Environ. Sci. Technol. 2018, 52, 6282.
[26] P. S. Lysaght, J. C. Woicik, M. Alper Sahiner, B. H. Lee, R. Jammy,
Keywords Appl. Phys. Lett. 2007, 91, 2.
extended X-ray absorption fine structure analyses, ferroelectrics, hafnium [27] J. J. Rehr, R. C. Albers, Rev. Mod. Phys. 2000, 72, 621.
oxide, nonvolatile memories, zirconium oxide [28] A. L. Ankudinov, B. Ravel, J. J. Rehr, S. D. Conradson, Phys. Rev. B
1998, 58, 7565.
Received: May 20, 2019 [29] J. J. Rehr, J. J. Kas, M. P. Prange, A. P. Sorini, Y. Takimoto, F. Vila,
Revised: August 8, 2019 C. R. Phys. 2009, 10, 548.
Published online: September 11, 2019 [30] M. E. McBriarty, J. A. Soltis, S. Kerisit, O. Qafoku, M. E. Bowden,
E. J. Bylaska, J. J. De Yoreo, E. S. Ilton, Environ. Sci. Technol. 2017,
51, 4970.
[1] S. Salahuddin, S. Datta, Nano Lett. 2008, 8, 405. [31] J. Rockenberger, L. Tröger, A. Kornowski, T. Vossmeyer, A. Eychmüller,
[2] J. S. Meena, S. M. Sze, U. Chand, T.-Y. Tseng, Nanoscale Res. Lett. J. Feldhaus, H. Weller, J. Phys. Chem. B 1997, 101, 2691.
2014, 9, 526. [32] O. J. Dura, R. Boada, M. A. López de la Torre, G. Aquilanti,
[3] M. H. Park, Y. H. Lee, H. J. Kim, Y. J. Kim, T. Moon, K. Do Kim, A. Rivera-Calzada, C. Leon, J. Chaboy, Phys. Rev. B 2013, 87, 174109.
J. Müller, A. Kersch, U. Schroeder, T. Mikolajick, C. S. Hwang, [33] O. Dargaud, L. Cormier, N. Menguy, L. Galoisy, G. Calas, S. Papin,
Adv. Mater. 2015, 27, 1811. G. Querel, L. Olivi, J. Non-Cryst. Solids 2010, 356, 2928.
[4] M. Hoffmann, M. Pešić, K. Chatterjee, A. I. Khan, S. Salahuddin, [34] L. M. Acuña, D. G. Lamas, R. O. Fuentes, I. O. Fábregas, M. C. A.
S. Slesazeck, U. Schroeder, T. Mikolajick, Adv. Funct. Mater. 2016, Fantini, A. F. Craievich, R. J. Prado, J. Appl. Crystallogr. 2010, 43, 227.
26, 8643. [35] R. Viennet, H. Roussel, L. Rapenne, J. L. Deschanvres, H. Renevier,
[5] J. Müller, U. Schröder, T. S. Böscke, I. Müller, U. Böttger, L. Wilde, V. Jousseaume, E. Jalaguier, M. G. Proietti, Phys. Rev. Mater. 2018, 2,
J. Sundqvist, M. Lemberger, P. Kücher, T. Mikolajick, L. Frey, J. Appl. 055002.
Phys. 2011, 110, 114113. [36] C. J. Howard, R. J. Hill, B. E. Reichert, Acta Crystallogr., Sect. B 1988,
[6] X. Sang, E. D. Grimley, T. Schenk, U. Schroeder, J. M. LeBeau, Appl. 44, 116.
Phys. Lett. 2015, 106, 162905. [37] C. Zhao, G. Roebben, M. Heyns, O. Van der Biest, Key Eng. Mater.
[7] T. Shimizu, K. Katayama, H. Funakubo, J. Ceram. Soc. Jpn. 2018, 2002, 206, 1285.
126, 269. [38] M. H. Park, H. J. Kim, Y. J. Kim, T. Moon, C. S. Hwang, Appl. Phys.
[8] Y. Wei, P. Nukala, M. Salverda, S. Matzen, H. J. Zhao, J. Momand, Lett. 2014, 104, 72901.
A. S. Everhardt, G. Agnus, G. R. Blake, P. Lecoeur, B. J. Kooi, [39] M. H. Park, Y. H. Lee, H. J. Kim, Y. J. Kim, T. Moon, K. Do Kim,
J. Íñiguez, B. Dkhil, B. Noheda, Nat. Mater. 2018, 17, 1095. S. D. Hyun, T. Mikolajick, U. Schroeder, C. S. Hwang, Nanoscale
[9] S. V. Barabash, J. Comput. Electron. 2017, 16, 1227. 2018, 10, 716.
[10] M. H. Park, T. Schenk, C. M. Fancher, E. D. Grimley, C. Zhou, [40] M. H. Park, Y. H. Lee, T. Mikolajick, U. Schroeder, C. S. Hwang, Adv.
C. Richter, J. M. LeBeau, J. L. Jones, T. Mikolajick, U. Schroeder, Electron. Mater. 2019, 5, 1800522.
J. Mater. Chem. C 2017, 5, 4677. [41] Y. H. Lee, S. D. Hyun, H. J. Kim, J. S. Kim, C. Yoo, T. Moon, K. Do Kim,
[11] T. Mittmann, M. Materano, P. D. Lomenzo, M. H. Park, I. Stolichnov, H. W. Park, Y. Bin Lee, B. S. Kim, J. Roh, M. H. Park, C. S. Hwang, Adv.
M. Cavalieri, C. Zhou, C.-C. Chung, J. L. Jones, T. Szyjka, M. Müller, Electron. Mater. 2019, 5, 1800436.
A. Kersch, T. Mikolajick, U. Schroeder, Adv. Mater. Interfaces 2019, 6, [42] T. Onaya, T. Nabatame, N. Sawamoto, A. Ohi, N. Ikeda, T. Chikyow,
1900042. A. Ogura, Appl. Phys. Express 2017, 10, 81501.
[12] A. Pal, V. K. Narasimhan, S. Weeks, K. Littau, D. Pramanik, T. Chiang, [43] M. H. Park, C.-C. Chung, T. Schenk, C. Richter, K. Opsomer,
Appl. Phys. Lett. 2017, 110, 3. C. Detavernier, C. Adelmann, J. L. Jones, T. Mikolajick,
[13] J. Müller, T. S. Böscke, U. Schröder, S. Mueller, D. Bräuhaus, U. Schroeder, Adv. Electron. Mater. 2018, 4, 1800091.
U. Böttger, L. Frey, T. Mikolajick, Nano Lett. 2012, 12, 4318. [44] I. V Ciuchi, C. C. Chung, C. M. Fancher, J. Guerrier, J. S. Forrester,
[14] Y. W. Lu, J. Shieh, F. Y. Tsai, Acta Mater. 2016, 115, 68. J. L. Jones, L. Mitoseriu, C. Galassi, J. Eur. Ceram. Soc. 2017, 37, 4631.
[15] S. L. Weeks, A. Pal, V. K. Narasimhan, K. A. Littau, T. Chiang, ACS [45] B. Zhou, H. Shi, X. D. Zhang, Q. Su, Z. Y. Jiang, J. Phys. D: Appl. Phys.
Appl. Mater. Interfaces 2017, 9, 13440. 2014, 47, 115502.
[16] D. M. Hausmann, R. G. Gordon, J. Cryst. Growth 2003, 249, 251. [46] C.-T. Ko, P.-S. Yang, Y.-Y. Han, W.-C. Wang, J.-J. Huang, Y.-H. Lee,
[17] H. Kim, P. C. McIntyre, K. C. Saraswat, J. Mater. Res. 2004, 19, 643. Y.-J. Tsai, J. Shieh, M.-J. Chen, Nanotechnology 2015, 26, 265702.
[18] D. Y. Cho, T. J. Park, K. D. Na, J. H. Kim, C. S. Hwang, Phys. Rev. B [47] M. Newville, T. Stensitzki, D. B. Allen, A. Ingargiola, 2016, http://doi.
2008, 78, 1. org/10.5281/zenodo.11813.
[19] T. Schenk, U. Schroeder, M. Pešić, M. Popovici, Y. V Pershin, [48] G. G. Li, F. Bridges, C. H. Booth, Phys. Rev. B 1995, 52, 6332.
T. Mikolajick, ACS Appl. Mater. Interfaces 2014, 6, 19744. [49] B. Ravel, M. Newville, J. Synchrotron Radiat. 2005, 12, 537.

Phys. Status Solidi B 2020, 257, 1900285 1900285 (10 of 10) © 2019 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

You might also like