Professional Documents
Culture Documents
DOI: 10.1038/NMAT4493
Lei Zhang1,2, Yuanjun Zhou3, Lu Guo1,2,†, Weiwei Zhao4, Anna Barnes5,6, Hai-Tian Zhang1,2,
Craig Eaton1,2, Yuanxia Zheng1,2, Matthew Brahlek1,2, Hamna F. Haneef5,6, Nikolas J. Podraza5,6,
Moses H. W. Chan4, Venkatraman Gopalan1, Karin M. Rabe3, Roman Engel-Herbert1,2,*
Affiliations:
1
Department of Materials Science and Engineering, The Pennsylvania State University,
University Park, Pennsylvania 16802, USA.
2
Materials Research Institute, The Pennsylvania State University, University Park, Pennsylvania
16802, USA.
3
Department of Physics and Astronomy, Rutgers University, Rutgers, New Jersey 08854 USA.
4
Department of Physics, The Pennsylvania State University, University Park, Pennsylvania
16802, USA.
5
Department of Physics and Astronomy, University of Toledo, Toledo, Ohio 43606 USA.
6
Wright Center for Photovoltaics Innovation and Commercialization, University of Toledo,
Toledo, Ohio 43606 USA.
†
now at the Department of Materials Science and Engineering, University of Wisconsin-
Madison, Madison, Wisconsin 53706, USA.
*To whom correspondence should be addressed, E-mail: rue2@psu.edu.
1
NATURE MATERIALS | www.nature.com/naturematerials 1
A1. X-ray diffraction. On-axis 2θ-ω scans were performed for all films. X-ray intensity
of SrVO3 (CaVO3) recorded in the 2θ interval from 45-49° (42-51°) is shown in Fig. S1a (Fig.
S1c). Pronounced thickness fringes evidenced the high quality of the thin films. The XRD data
were fitted using a simple model by taking into account film thickness as well as film and
substrate lattice parameters1. The results are summarized in Table S1 (Table S2), the
comparisons of experimental data and fit is shown for the 45 nm SrVO3 film (38nm CaVO3 film)
A2. Atomic force microscopy: All surfaces of thin films were atomically smooth,
exhibiting a terrace structure, see Fig. S2. The root mean square (RMS) roughness of 4, 10, 12,
15, 20 and 45 nm thick SrVO3 films were determined to 6.5, 1.8, 1.8, 2.5, 1.5, and 2.0 Å,
respectively. RMS roughness of 4, 9, 14, 18, and 38 nm thick CaVO3 films were determined to
A3. Spectroscopic ellipsometry of SrVO3 films: Three SrVO3 films, nominally 10, 20,
and 45 nm thick, were grown on single-side polished LSAT substrates and studied using ex-situ
over the IR to UV (0.74 to 5.89 eV; M-2000 J. A. Woollam Co.) and IR (0.049 to 0.75 eV,
FTIR-VASE J. A. Woollam Co.) spectral ranges. The complex dielectric function spectra
(=+i) and microstructural parameters (film layer thickness df, surface roughness thickness
model consisting of a semi-infinite LSAT substrate / SrVO3 film / SrVO3 surface roughness layer
/ air ambient. Free parameters corresponded to the thicknesses of the SrVO3 film and the surface
A
, (A)
E E 02
2
where A and E0 are the amplitude and resonance energy, respectively, and E0 must be
A E0
, (B)
E E 2 i E
2
0
where A is the amplitude, Γ is the broadening, and E0 is the resonance energy. Each Tauc-
AE 0 ( E Eg )2 E E
2
2 ( E E02 ) 2 2 E 2
g
E , (C)
E Eg
0
and
2 2
1 P d , (D)
Eg 2 E2
where is the electrical resistivity, is the mean free relaxation time, 0 the vacuum
permittivity and the reduced Planck constant, is used to interpret the intraband response due to
The optical properties of the surface roughness layer are represented using the
8
Bruggeman effective medium approximation consisting of a 0.5 SrVO3 and 0.5 void volume
fractions. This surface roughness layer was initially included in the model, however it was not
detectable during the data analysis, which is consistent with the film surface being atomically
thicknesses for the nominally 10, 20, and 45 nm films are given in Table S3. After thicknesses
The optical conductivity as a function of frequency for SrVO3 is shown as Fig. S3a.
The plasma frequency p was calculated from the real part of using 11, 12:
p2 c
1 ( ) d . (G)
8 0
Systematic errors determining the plasma frequency arise from the choice of the upper
integration limit c, the approximation of optical conductivity contribution below the lowest
frequency measured (282 cm-1), and interband contribution to 1 from transition between the t2g
frequencies below 282 cm-1 were omitted. The choice of c is to include all contribution from
free carriers and to exclude contribution from interband transition. Doubling c to 000 cm-1
increased p by less than 10%, while the contribution from t2g to t2g interband transition
calculated by DFT contributed less than 4% of the total Drude peak. With sample-to-sample
deviations less than 3% the total error in p was estimated to 10%. The plasma frequency of
SrVO3 was estimated to be p = (2.10.2) eV. The sum-rule plasma frequency as a function of
thick, was grown on single-side polished SLAO substrates and studied using ex-situ
i =55°, 65° and 75° angle of incidence using single rotating-compensator spectroscopic
ellipsometers over the IR to UV (0.73 to 5.04 eV; M-2000U J. A. Woollam Co.). The complex
dielectric function spectra (=+i) and microstructural parameters (film layer thickness df)
were first extracted using a least squares regression analysis and an unweighted error function to
fit the experimental ellipsometric spectra collected over the near IR to UV range to an optical
model consisting of a semi-infinite SLAO substrate / CaVO3 film / air ambient. Free parameters
corresponded to the thicknesses of the CaVO3 film and a parameterization of for CaVO3. The
constant additive term to 1 represented by ∞. The Lorentz and Tauc-Lorentz oscillator are the
same, as introduced in Eqn (B)-(D). After thicknesses were obtained using the parametric model,
SrVO3 and CaVO3, we have fabricated Hall bar devices on 45-nm SrVO3 and 38-nm CaVO3 thin
films and measured resistivity along three different directions: <100>, <110> and <010>. The
resistivity at room temperature and residual-resistance ratio (300K/5K) are almost identical and
The deposition of polycrystalline films from correlated metals with a sizeable anisotropy
in electrical resistivity can affect the overall sheet resistance. While in cubic and pseudo-cubic
crystal structures, such as SrVO3, CaVO3, but also SrNbO3, SrMoO3, CaMoO3, SrRuO3 and
LaNiO3, the anisotropy ratio of resistivity :|| is expected to be small. However, the
anisotropies in correlated metals with layered structure, or uniaxial symmetry might be larger,
and should be taken into consideration. MgB213 :|| ~3, Sr2RuO414 :|| ~23, PdCoO215 ~:||
metallic-like behavior of SrVO3, which is also typical for a simple metal described by a free non-
interacting electron gas, i.e. temperature-independent, high carrier concentration and monotonic
resistivity increase with temperature, the electron motion in a correlated metal is also affected by
many-body interactions. Alteration in the carrier dynamics from the expectation of a non-
interacting electron gas can be utilized to identify the type of interaction – such as electron-
Despite the itinerant behavior of carriers being preserved in the Fermi liquid, a strong electron
constant is common and independent of the model used to describe the correlation phenomena.
In this regard Zk is a hallmark for the correlation strength and a metric for the carrier mass
enhancement m * m band
* *
/ Z k compared to the band effective mass mband of non-interacting
carriers, and thus at the heart of the new design scheme for transparent conductors. The
without correlation, e.g. the ratio of the square of the plasma frequencies Z k pExp / pDFT or
2
the ratio of kinetic energies Z k K Exp / K DFT from band structure calculation without correlation
effects KDFT and experimental values of the experimental kinetic energy KExp reduced by band
width normalization. In both cases the experimental values can be extracted from optical
film of thickness t, situated between two dielectric media (air) and forming an abrupt upper and
lower interface was considered. The absorption coefficient of the film is given by
4 k
, (H)
monochromatic radiation the transmission coefficient T, i.e. the ratio of incident intensity at the
upper interface and the transmitted intensity at the lower interface is given by 17:
I 1 R 4 R sin 2
2
T t , (I)
I 0 e R 2 e t 2 R cos 2( )
R
n 1 k 2
2
, (K)
n 12 k 2
2 n t 2k
; tan 1 . (L)
n k2 1
2
The term 2 R cos 2( ) takes into account the interference effect arising from the upper and
lower film interface. The refractive index n and extinction coefficient k are intimately coupled to
1 i 2 n 2 k 2 i 2nk (M)
The thickness dependent transmission at 550 nm wavelength for freestanding conducting films
Ag, ITO, SrVO3 and CaVO3 were calculated and are shown in Fig. S6. Since the choice of a
specific wavelength can change and R quite significantly, T averaged over the visible spectrum
Tave
400nm
, (N)
800nm
400nm
d
B3. Calculation of enhanced surface scattering effect. The electron mean free path
3 2 n 3 ,
1
vF (O)
e
where is the relaxation time, v F the Fermi velocity, the carrier mobility, n the carrier
concentration, and e the elemental charge. Values for the EMFP of selected materials considered
here, including conventional metals, degenerately doped semiconductors, and SrVO3 and CaVO3
19 20
are given in Table S3. Fuchs and Sondheimer attributed the pronounced conductivity
decrease in ultrathin films to enhanced surface scattering, if film thickness is on the order of the
EMFP or below:
31 p 1 1 1 e
FS 0 1
2 1
3
1 pe
5
d ,
(P)
with 0 the bulk conductivity, = t/ the normalized length with t referring to the film
thickness and the EMFP and p the reflection coefficient ( 0 p 1 , p=0 diffuse: random
direction and magnitude of velocity after surface scattering event, p=1 specular: elastic scattering
at the surface with reversal of the normal component). Mayadas and Shatzkes 21 have expanded
the Fuchs Sondheimer model by considering the following three main mechanisms: background
with the EMFP, D the average grain size and R the reflection coefficient. The total
resistivity =1 of SrVO3, CaVO3, Ag and ITO thin films have been calculated as a function of
thickness and the resistivity increase due to enhanced carrier scattering in the films was
calculated 22 using FS MS 0 and the experimental data and fitting for SrVO3, CaVO3
B4. Calculation of the figure of merit for transparent conductors. Two parameters
determine the performance of a transparent conducting film: optical transmission T, and sheet
resistance Rs. Both are connected via the film thickness t to the material specific parameters
optical absorption and electrical conductivity . The best performance is achieved for
simultaneously high electrical conductivity and optical transmission. The figure of merit (FOM)
T 10
TC ( t ) 0 . (S)
t
Here T is the optical transmission and (0+)/t represents the sheet resistance Rs that takes into
account surface scattering effects, as described in Section C3 of the supplement. The exponent of
the transmission in Eqn. (S) balances the relative weights of T and (0+)/t contributing to TC
and determines the transmission for the thickness at maximum TC. Using an exponent of 10
supplement and shown in Fig. S6, the choice of a specific wavelength affects T, which can
10
spectral sensitivity of the human eye) the average transmission [Eqn. (N)] was used instead.
namely conventional metals, represented by Ag, degenerately doped wide band gap
semiconductors, represented by ITO, and the proposed correlated metals, represented by SrVO3
and CaVO3, the FOM has been modified to incorporate effects of enhanced carrier scattering,
relevant for films with thicknesses smaller than the electron mean free path (see C3), and
changes in the transmission from multiple reflection at the top and bottom surface and the
interference effects arising from it for the free standing film (see C2), relevant for film
thicknesses with optical path length on the order of /4. Taking into account both effects caused
Figure S5 shows the impact of the different corrections to TC for the case of Ag and
ITO. In the first case, averaging the transmission over the entire visible spectrum compared to
the transmission at 550 nm resulted in a small TC increase, attributed to the strong interband
transition of Ag around 360 nm. Taking reflection into account the transmission increased from
interference and multiple reflection within the film. The effect of enhanced surface scattering of
electrons is rather strong. Although FOM only scales linearly with this effect, the electron mean
free path (EMFP) in Ag is rather large (52 nm, see Table S5) The maximum TC for a film
thickness around 4 nm is only a small fraction of the EMFP, resulting in a more than one order of
magnitude increase in the conductivity compared to the bulk value, see Fig. S8a 23, 24.
11
doped wide band gap semiconductors, where low electrical conductivity is compensated by low
optical absorption allowing a larger film thickness. In case of ITO, averaging T over the entire
visible spectrum resulted in a reduction of the average T, which is further reduced by taking
reflection into account, showing pronounced interference effects in TC with a local minimum at
a film thickness of 60 nm. Taking surface scattering effects into consideration TC only
marginally decreased at the maximum around 150 nm, but was reduced at the second local
conducting film, or limiting the model to only one wavelength, or ignoring the thickness
Careful consideration of all relevant effects and using a complete set of experimental data, i.e.
the in the entire range of wavelengths considered and thickness dependent electrical
conductivity for the material to benchmark, are mandatory for a meaningful comparison.
of SrVO3 were performed using the generalized gradient approximation (GGA) with the Perdew-
(VASP-5.2.12) 26, 27. The projector augmented wave (PAW) potentials used contained 10 valence
electrons for Sr (4s24p65s2), 13 for V (2s22p63d34s2) and 6 for O (2s22p4) 28, 29. We used 500 eV
as the energy cutoff. An 8x8x8 Monkhorst-Pack k-point mesh was used for the structural
relaxation, and a 31x31x31 Γ centered k-mesh was used for the interband transition calculation.
12
antiferromagnetic and nonmagnetic configurations showed that the relevant features of the band
structures and interband transitions are not sensitive to U or the magnetic ordering. For U=0, the
ground state was found to be non-magnetic. Since SrVO3 is paramagnetic at room temperature 30,
31, 32
, we performed the analysis for the nonmagnetic state with U=0. The c lattice parameter of
the tetragonal five-atom unit cell was relaxed with in-plane lattice constant fixed to the
experimental value 3.868Å, preserving P4/mmm symmetry, yielding a c/a ratio of 0.979.
The unscreened plasma frequency was calculated from the band structure using11, 33
where denote the direction, V denotes the volume of the unit cell, f is the occupancy,
E is the band energy, e is the unit vector, and g is the weight of k point. For SrVO3, the plasma
frequency is determined by the t2g bands that cross the Fermi level. Since the t2g band structure is
relatively simple, the uncertainty in the plasma frequency is determined by the uncertainty in the
t2g bandwidth. We have computed band structures for different U and k-meshes, and the t2g
bandwidth is found to vary by less than 2%. The uncertainty in p can thus be estimated to be
2
The intraband contribution to the dielectric function was obtained empirically from the
Drude model
2 2 ne 2
int ra 1 2
1
, intra 3 2
, and ; (U)
2 2 me
using measured values =2×10-5 Ω×cm, n=2×1022 cm-3, to estimate =0.1eV and
me 4.5m0 . The interband contribution int(1)er , int( 2 )er to the dielectric function was calculated using
13
4 2 e 2
er , 2 w E E vk u ck e q u vk u v k u ck e q ,
1
int
( 2)
lim 2 k ck (V)
V q 0 q c , v ,k
where V is the volume of the unit cell, the index c and v refer to conduction and valence
bands, respectively, and u ck is the cell periodic part of the wavefunction at k-point k. e and e
denote unit vector in Cartesian directions. lim u ck e q u vk gives the α component of the dipole
q 0
The real part of the dielectric function was obtained using the Kramers-Kronig
transformation,
(1)
int er , 1 2
P d ' (W)
0
' 2 2 i
Figure S6 illustrates the steps to calculate the contribution from interband transition to the
imaginary part of the dielectric function. Figure S6a shows the band structure highlighting the 8
bands with interband transitions up to 3.6 eV: the three highest valence bands, 2p1 (highest), 2p2
(second highest) and 2p3 (third highest), the two V 3d eg bands eg1 (lowest) and eg2 (highest), and
the three V 3d t2g bands t2g1 (lowest), t2g2 (middle) and t2g3 (highest). Figures S6b-f show energy
difference E and dipole matrix element M2 of optical transitions between initial and final states
in the Brillouin zone of SrVO3 along high symmetry lines. Red symbols indicate same
occupancy of initial and final state, suppressing the transition, while blue symbols represent a
filled to empty, and thus active optical transition. Sphere, triangle and cube symbols correspond
14
a transition with notable M2 in the visible spectrum is suppressed due to same occupancy of
initial and final state (red symbols) or the magnitude of M2 is small. The combination of a large
M2 and a filled to empty transition in the visible spectrum does not occur. The contribution with
both, E in the visible spectrum and sizable M2 were selected for the isosurface representation in
Figs. 4c-4e. The respective contribution of these transitions and low energy interband transition
between different t2g bands to inter , are shown in Fig. S9g.
( 2)
15
Figure S1. High resolution 2- X-ray diffraction scans of a, SrVO3 films with varying
thicknesses grown on LSAT substrates. b, Experimental and calculated X-ray diffraction
intensities of the 45-nm-thick SrVO3. c. SrVO3 films with varying thicknesses grown on LSAT
substrates. d, Experimental and calculated X-ray diffraction intensities of the 38-nm-thick
CaVO3. Film lattice parameter and thickness given in Table S1 and Table S2 were extracted
from these fits.
16
17
18
Figure S4. Complex dielectric function ( = 1 + i2) spectra of CaVO3 extracted from room
temperature spectroscopic ellipsometry measurements
19
Figure S5. a, Schematic and b, optical photograph of Hall bar devices, here for CaVO3.
Temperature dependent resistivity of c, 45-nm-thick SrVO3 and d, 38-nm-thick CaVO3 films
along the different directions. The scale bar indicates 500 µm.
20
Figure S6. Optical transmission of free standing films, accounting for reflection and
interference, at a wavelength =550nm [Eqns. (H), (I), (K), (L)], and averaged over the range of
the visible spectra [Eqn. (N)]. Data of the complex refractive index of Ag, ITO, SrVO3 and
CaVO3 were taken from Refs. 35, 36 and this work.
21
Figure S7. Thickness dependence of the sheet resistance Rs for a, SrVO3 a, CaVO3 c, ITO, and
d, Ag taking the thickness dependent resistivity into account. The increase in resistivity
normalized to the bulk resistivity 0 is shown below. Data for Ag, ITO, SrVO3 and CaVO3 were
taken from Refs. 37, 38 and this work, respectively.
22
Figure S8. Consideration for the figure of merit (FOM) for transparent conductors. Solid black
line: FOM defined after Haacke 23, calculated for light at wavelength =550 nm. Dashed gray
line: FOM using transmission value taken as average across the visible spectrum [Eqn. (N)].
Dotted blue line: Averaged transmission incorporating multiple reflection and interference
effects from top and bottom interface of a free standing film by calculating the transmission from
Eqn. (H) and averaging over the visible range, see Fig. S4. Thick solid green line. Figure of merit
from Eqn. (S) that includes thickness dependent increase of electrical conductivity due to
enhanced surface scattering for films with thicknesses below the electron mean free path, see
Fig. S5. a, Example of Ag representing ultrathin metal films. b, Example of ITO representing
degenerately doped wide band gap semiconductors.
23
Figure S9. a, Band structure and nomenclature of bands involved in interband transitions of
SrVO3 with energies below 3.6 eV. The superscript is introduced to distinguish the different
bands involved in the transition. b-f, Dipole matrix element M2 and energy difference E
between the lower and upper electronic state along high symmetry lines in the Brillouin zone of
24
25
Nominal
Lattice film ρ
film Rs Tfs Tfilm/Tsub
parameter thickness (300 K)
thickness
(Å) (nm) (Ω cm) (%) (%)
45 nm 3.822 ± 0.001 45.9 ± 0.2 2.90×10-5 6.3 51.0 73.3
26
Nominal
Lattice film ρ
film Rs Tfs Tfilm/Tsub
parameter thickness (300 K)
thickness
(Å) (nm) (Ω cm) (%) (%)
38 nm 3.782 ± 0.002 37.44 ± 0.2 3.86×10-5 10.3 55.7 75.4
27
28
29
30
31
3. Johs B, Woollam JA, Herzinger CM, Hilfiker J, Synowicki R, Bungay CL. Overview of
variable angle spectroscopic ellipsometry (VASE), part II: advanced applications. SPIE
Proceedings; 1999; 1999. p. 29-58.
5. Collins RW, Ferlauto AS. Handbook of Ellipsometry. William Andrew Norwich NY,
2005.
9. Oldham WG. Numerical techniques for the analysis of lossy films. Surf Sci 1969, 16(0):
97-103.
10. Abdallah LS, Zollner S, Lavoie C, Ozcan AS, Raymond M. Optical conductivity of
Ni1−xPtxSi monosilicides (0 < x < 0.3) from spectroscopic ellipsometry. J Vac Sci Technol
B 2014, 32(5): 051210.
12. Basov DN, Averitt RD, Van Der Marel D, Dressel M, Haule K. Electrodynamics of
correlated electron materials. Rev Mod Phys 2011, 83(2): 471.
32
16. Analytis JG, Chu J-H, Erickson AS, Kucharczyk C, Serafin A, Carrington A, et al. Bulk
superconductivity and disorder in single crystals of LaFePO. arXiv preprint
arXiv:08105368 2008.
17. Fan H. Infra-red absorption in semiconductors. Rep Prog Phys 1956, 19(1): 107.
19. Fuchs K. The conductivity of thin metallic films according to the electron theory of
metals. Proc. Cambridge Philos. Soc; 1938: Cambridge Univ Press; 1938. p. 100.
20. Sondheimer EH. The mean free path of electrons in metals. Adv Phys 1952, 1(1): 1-42.
21. Mayadas A, Shatzkes M. Electrical-resistivity model for polycrystalline films: the case of
arbitrary reflection at external surfaces. Phys Rev B 1970, 1(4): 1382.
22. O’Connor B, Haughn C, An K-H, Pipe KP, Shtein M. Transparent and conductive
electrodes based on unpatterned, thin metal films. Appl Phys Lett 2008, 93(22): 223304.
23. Haacke G. New figure of merit for transparent conductors. J Appl Phys 1976, 47(9):
4086-4089.
24. Ellmer K. Past achievements and future challenges in the development of optically
transparent electrodes. Nature Photon 2012, 6(12): 809-817.
25. Perdew JP, Burke K, Ernzerhof M. Generalized gradient approximation made simple.
Phys Rev Lett 1996, 77(18): 3865.
26. Kresse G, Hafner J. Ab initio molecular dynamics for liquid metals. Phys Rev B 1993,
47(1): 558.
27. Kresse G, Furthmüller J. Efficient iterative schemes for ab initio total-energy calculations
using a plane-wave basis set. Phys Rev B 1996, 54(16): 11169.
33
29. Blöchl PE. Projector augmented-wave method. Phys Rev B 1994, 50(24): 17953.
30. Chamberland BL, Danielson PS. Alkaline-earth vanadium (IV) oxides having the AVO3
composition. J Sol Stat Chem 1971, 3(2): 243.
31. Lan YC, Chen XL, He M. Structure, magnetic susceptibility and resistivity properties of
SrVO3. J Alloys Compd 2003, 354(1–2): 95.
32. Onoda M, Ohta H, Nagasawa H. Metallic properties of perovskite oxide SrVO3. Solid
State Commun 1991, 79(4): 281.
33. Millis A. Optical conductivity and correlated electron physics. Strong interactions in low
dimensions. Springer, 2004, pp 195-235.
36. Kim H, Piqué A, Horwitz JS, Mattoussi H, Murata H, Kafafi ZH, et al. Indium tin oxide
thin films for organic light-emitting devices. Appl Phys Lett 1999, 74(23): 3444-3446.
37. OConnor B, Haughn C, An K-H, Pipe KP, Shtein M. Transparent and conductive
electrodes based on unpatterned, thin metal films. Appl Phys Lett 2008, 93(22): 223304-
223304-223303.
38. Bonaccorso F, Sun Z, Hasan T, Ferrari A. Graphene photonics and optoelectronics. Nat
Photonics 2010, 4(9): 611-622.
39. Shigesato Y, Paine DC. Study of the effect of Sn doping on the electronic transport
properties of thin film indium oxide. Appl Phys Lett 1993, 62(11): 1268-1270.
34