Home
Search
Collections
Journals
About
Contact us
My IOPscience
Strain induced modulation of the correlated transport in epitaxial Sm0.5Nd0.5NiO3 thin films
This content has been downloaded from IOPscience. Please scroll down to see the full text. 2015 J. Phys.: Condens. Matter 27 132201 (http://iopscience.iop.org/0953-8984/27/13/132201) View the table of contents for this issue, or go to the journal homepage for more
Download details: IP Address: 169.230.243.252 This content was downloaded on 23/04/2015 at 08:24
Please note that terms and conditions apply.
Journal of Physics: Condensed Matter J. Phys.: Condens. Matter 27 (2015) 132201 (7pp)
doi:10.1088/0953-8984/27/13/132201
Fast Track Communication
Strain induced modulation of the correlated transport in epitaxial Sm0.5Nd0.5NiO3 thin films L Zhang1 , H J Gardner1 , X G Chen1 , V R Singh1 and X Hong1,2 1 2
Department of Physics and Astronomy, University of Nebraska-Lincoln, NE 68588, USA Nebraska Center for Materials and Nanoscience, University of Nebraska-Lincoln, NE 68588, USA
E-mail: [email protected] Received 6 February 2015, revised 16 February 2015 Accepted for publication 25 February 2015 Published 17 March 2015 Abstract
We report a study of the effect of epitaxial strain on the correlated transport properties of 2–40 nm Sm0.5 Nd0.5 NiO3 (SNNO) films grown on different substrates. The metal–insulator transition (MIT) temperature TMI of the SNNO films increases with increasing tensile strain. While films on (0 0 1) LaAlO3 and (1 1 0) NdGaO3 substrates exhibit a sharp MIT and thermal hysteresis in the cooling–heating cycle, signaling a first-order transition, films on (0 0 1) SrTiO3 show a broad, second-order MIT. Hall effect measurements reveal hole-type charge carriers and thermally activated temperature dependence of the carrier density below TMI . The corresponding activation energy is ∼80 meV for films on LaAlO3 and NdGaO3 , and is suppressed to 25 meV for films on SrTiO3 . The carrier mobility in the metallic state and variable range hopping (VRH) transport at a low temperature point significantly enhanced electron localization in SNNO on STO, which we believe is not simply driven by extrinsic effects such as oxygen vacancies, but rather is an intrinsic characteristic for films subject to tensile strain due to the elongated Ni–O bond and hence enhanced dynamic Jahn–Teller distortion. In ultrathin films above the electrical dead layer thickness (2–3 nm), we observe a more than 100 K increase of TMI for films on LaAlO3 , which has been correlated with a crossover from 3D to 2D transport as revealed from VRH. We attribute the distinct transport characteristics to strain induced modulation of various energy scales associated with the Ni–O–Ni bond angle and Ni–O bond length, which collectively determine the delocalization bandwidth of the system. Keywords: strongly correlated oxides, nickelate thin films, epitaxial strain, metal–insulator transition, lattice distortion, electron localization, dynamic Jahn–Teller distortion (Some figures may appear in colour only in the online journal)
transition [3–5]. The MIT transition temperature TMI increases monotonically in compounds with smaller size rare earth ion R due to higher levels of lattice distortion [3–5]. The latticedriven modulation of the MIT can also be achieved by applying hydrostatic pressure to bulk nickelates [3, 6–8] or, for epitaxial thin films, optical excitation of the phonon modes in the substrates [9]. Electron localization in these materials is also
Strongly correlated oxides such as the perovskite nickelates RNiO3 (R = rare earth) exhibit a rich variety of electronic and magnetic ground states that can be sensitively tuned by the lattice degree of freedom [1–4]. For example, the RNiO3 oxides, except LaNiO3 , are charge-transfer type Mott insulators, exhibiting a temperature-driven metal–insulator transition (MIT) as well as a paramagnetic to antiferromagnetic 0953-8984/15/132201+07$33.00
1
© 2015 IOP Publishing Ltd Printed in the UK
J. Phys.: Condens. Matter 27 (2015) 132201
3
10
2
10
1 µm
SNNO 001
4
10
SNNO 002
5
10
LAO
1
3
10
2
SNNO 001
4
10
NGO 110
10
1
5
3
10
2
10
1 µm
STO 001
4
10
SNNO 001
10
STO
3
10
SNNO 001
1
10
23
24
θ -2θ (deg.)
1
10
20
STO 001
SNNO 002
10
STO 002
Intensity (a.u.)
1 µm
10
SNNO 002
NGO 220
10
5
NGO
LAO 002
(b) 106 LAO 001
(a)
Intensity (a.u.)
entangled with the magnetic order. Compounds with lower distortion level, such as PrNiO3 and NdNiO3 , exhibit a firstorder MIT with TMI coinciding with the antiferromagnetic order temperature TN [7, 10]. For compounds with smaller size R ions and higher orthorhombic distortion, the MIT and magnetic transition become second-order, and the transition temperatures are decoupled. A powerful tool to systematically probe the effect of lattice distortion on the MIT in nickelates is to impose different levels of strain on epitaxial thin films through lattice mismatched substrates. It has been reported that epitaxial strain can effectively tune the TMI in NdNiO3 [11–13] and SmNiO3 [14, 15] thin films by more than 100 K. Sm0.5 Nd0.5 NiO3 (SNNO) is an especially interesting system in the RNiO3 family as it is close to the phase boundary of critical tolerance factor where TMI and TN become decoupled [7, 10], which means in this system the competition among different energy scales to stabilize the electronic and magnetic ground states are highly susceptible to small external perturbations such as structural variation. In addition, the TMI of SNNO is close to room temperature, which has high technological relevance in building oxide-based electronic and spintronic devices. To date, although extensive research on the effect of epitaxial strain has been carried out on the two parent compounds, NdNiO3 [11–13] and SmNiO3 [14–17], studies on high quality epitaxial thin films of Sm0.5 Nd0.5 NiO3 have been very limited and mostly focused on the growth [18, 19]. In this work, we report a comprehensive study of the transport properties of 2–40 nm epitaxial SNNO films grown on (0 0 1) SrTiO3 (STO), (0 0 1) LaAlO3 (LAO), and (1 1 0) NdGaO3 (NGO) substrates. Bulk SNNO has a pseudocubic lattice constant of ∼3.804 Å [20], corresponding to a compressive strain of −0.37% on LAO (aLAO = 3.79 Å), and tensile strains of 1.5% on NGO (aNGO = 3.86 Å) and 2.7% on STO (aSTO = 3.905 Å). While bulk SNNO shows a secondorder MIT, SNNO films grown on LAO and NGO exhibit a sharp MIT with thermal hysteresis observed in the resistance cooling–heating cycle, signaling a first-order transition. The temperature dependence of the carrier density right below TMI can be well described by thermally activated behavior, from which we extracted the activation energies of 79 and 77 meV for the films on LAO and NGO, respectively. In sharp contrast, films grown on STO show the characteristic of a second-order MIT and a softened activation energy of 25 meV. Based on the carrier mobility in the metallic state and the low temperature variable range hopping (VRH) transport, we conclude that such distinct transition characteristics are driven by different electron localization mechanisms rather than simply due to different levels of structural disorders, such as oxygen vacancies, associated with strain relaxation. We have also investigated the effect of the film thickness on the correlated transport properties. While TMI for films on NGO and STO only show weak thickness dependence, there is a more than 100 K increase of TMI in ultrathin films on LAO above the electrical dead layer thickness (2–3 nm), which has been correlated to a dimensionality crossover from 3D to 2D based on VRH at low temperature. Unlike the chemical substitution and hydrostatic pressure experiments, where the key control parameter is the Ni–O–Ni bond angle [3–8], we
30
40
θ -2θ (deg.)
50
Figure 1. (a) AFM images of 4.4 nm SNNO films grown on LAO, NGO, and STO substrates showing clear atomically flat terraces. The RMS roughness of all three films is 1–2 Å. (b) XRD θ –2θ scans for 40 nm SNNO films grown on LAO (upper panel), NGO (middle panel), and STO (lower panel) substrates. Inset: Laue finite size fringes around the (0 0 1) peak (solid line) with the fit (dashed line) for the 40 nm SNNO on STO.
believe strain induced modulation in the Ni–O bond length also plays a critical role in driving the diverse transport behaviors in epitaxial SNNO films on different substrates. Such modulation would change the cooperative oxygen displacement frequency and in turn control the dynamic Jahn–Teller distortion, which competes with the effect associated with the Ni–O–Ni bond angle in limiting the delocalization bandwidth. We fabricated the epitaxial SNNO films using off-axis radio frequency magnetron sputtering. The films were grown at a deposition temperature of 500 ◦ C in 135 mTorr process gas composed of Ar and O2 (ratio of 1 : 2). After growth, the films were cooled down to room temperature in 1 atmosphere of O2 . For each type of substrate, we have fabricated films with six different thicknesses in the range of 2–40 nm. We examined the surface morphology of the samples using atomic force microscopy (AFM). As shown in figure 1(a), films grown on all three types of substrates possess atomically smooth surfaces with root mean square (RMS) roughness of 1–2 Å. The structural properties of the films were characterized using x-ray diffraction (XRD). Figure 1(b) shows the x-ray θ–2θ scans taken on 40 nm SNNO films grown on different substrates. The film thicknesses were determined from x-ray reflectivity measurements and Laue fringes around the (0 0 1) Bragg peaks (figure 1(b) inset). The x-ray spectra reveal (0 0 1) growth in the pseudo-cubic structure for all samples, with no impurity phases observed for the thickest films studied. The c-axis lattice constants are 3.81 Å, 3.76 Å, and 3.79 Å for the 40 nm films on LAO, NGO, and STO, respectively, consistent with strained films. The corresponding rocking curves for the 2
J. Phys.: Condens. Matter 27 (2015) 132201
(0 0 2) peaks have full-width-half-maximums of 0.12◦ , 0.08◦ , and 0.06◦ , reflecting very high crystalline quality. The out-ofplane lattice constant for the 40 nm film on STO is higher than that for the film on NGO, even though it is subject to a higher tensile strain. Reciprocal space mapping measurements on this sample reveal an in-plane lattice constant of 3.85 Å, showing SNNO on STO is partially relaxed at this thickness. For transport measurements, we patterned the SNNO films into standard Hall bar devices using photolithography. The four-point resistance channels have aspect ratios of 1 or 2 with the channel width varying from 5 to 40 µm. Transport measurements were carried out between 10 and 350 K at a cooling/heating rate of ∼4 K min−1 using Quantum Design PPMS combined with Keithley 2400 SourceMeter. For each sample, we examined the I –V relation at different temperatures to verify that the resistance value obtained is current-independent. Nonlinear I –V characteristics has only been observed in the metallic phase at an excitation current below 1 µA, which is common in correlated oxides due to either work function mismatch at the contacts or phase inhomogeneity within the sample [21]. All measurements were taken in the linear I –V regime with low excitation currents below 10 µA to avoid Joule heating, so that the observed modulation of the transport properties is intrinsic to the sample. We also fabricated multiple Hall bar devices (up to 4) at different locations on the same film, and confirmed that the film properties are homogeneous. Figure 2(a) shows the temperature dependence of sheet resistance R (T ) for a 18 nm SNNO film grown on LAO, which is subject to a moderate compressive strain. The resistance of the film exhibits metallic temperature dependence at high temperature, followed by a sharp transition to an insulating behavior below TMI . There is a pronounced thermal hysteresis between the R (T ) curves on heating and cooling, signaling a first-order MIT [10]. This is distinctly different from bulk SNNO, where the MIT is second-order [7, 10]. The transition temperature TMI , defined as the temperature where dR/dT changes sign on cooling, is 195 K. The thermal hysteresis leads to an 11 K shift between the deflection points for R (T ) on cooling and heating. In the insulating phase, the film exhibits two distinct temperature dependences. Below TMI , there is a sharp transition to the insulating phase between 170 and 135 K for R (T ) on cooling. This temperature dependence has previously been modeled as percolative transport due to coexisting metallic and semiconducting phases [11]. To probe the formation of a correlated gap, we examined the temperature dependence of the carrier density extracted from Hall effect measurements (figure 3). The film possesses hole-type charge carriers with a temperature-independent density of n ∼ 4 holes per unit cell (uc) in the metallic phase. Below TMI , n(T ) decreases exponentially with decreasing temperature, well described by the thermally activated model. Fitting to n(T ) = n0 exp[−(Ea /kB T)] reveals an activation energy Ea of 79 meV (figure 3). This energy gap is between the values reported for the parent compounds NdNiO3 (∼20 meV) [11] and SmNiO3 films (∼170 meV) [22], consistent with the picture that a smaller R ion size leads to a reduced electronic
Figure 2. Temperature dependence of sheet resistance R (T ) on cooling (black) and heating (red) for 18 nm SNNO films grown on (a) LAO, (b) NGO, and (c) STO substrates. Insets: ρ versus T −1/4 on cooling with fits (dashed lines) to the 3D VRH model or the parallel conduction model with both thermal activation and VRH behaviors. The dotted lines mark the temperature Tcr at which the VRH hopping energy W ∼ kB T .
bandwidth [3–5]. The conduction shows a weaker temperature dependence as the sample is cooled below ∼135 K, where phonon assisted hopping between impurity states within the energy gap becomes progressively important [23]. Below 30 K, the temperature dependence of resistivity is dominated by 3D VRH type transport as ρ = ρ0 exp (T0 /T )1/4 . We then investigated the SNNO films grown on NGO and STO substrates with the same thickness, which exhibit significantly different transport characteristics even though both systems are subject to tensile strain. For the 18 nm SNNO film grown on NGO, R (T ) shows qualitatively similar behavior as the film on LAO, exhibiting percolative conduction below TMI followed by impurity dominated hopping at low temperature (figure 2(b)). From n(T ) we extracted a similar activation energy of 77 meV (figure 3). Compared with the films on LAO, this film shows a higher TMI of 260 K, suggesting enhanced electron localization, and a reduced 3
10
23
10
22
quantitative VRH analyses of the low temperature ρ(T ) in these three systems. In Mott’s VRH model [23], conductivity σ in a 3D disordered system follows the relation: T0 1/4 , (1) σ = σ0 exp − T
4 holes/uc
0.6 2
µ (cm /Vs)
-3
n (cm )
J. Phys.: Condens. Matter 27 (2015) 132201
10
21
100
0.2 0.0
LAO NGO STO
where √ N (EF ) a 1/2 σ0 T = 3e2 γph 8π kB
0.4
200 250 300 350
T0 =
18.1 . kB N (EF )a 3 (2)
Here a is the localization length of the charge carriers, γph ∼ 1013 s−1 is the optical phonon frequency calculated from the Debye temperature, and N (EF ) is the density of states of impurity levels at the Fermi level. Fitting ρ(T ) on cooling to equation (1) (figure 2 insets), we obtained the 3D VRH parameters as listed in table 1. We first considered the localization length a for the charge carriers, which measures the spatial decay of the electron wavefunction and reflects directly the degree of charge localization. The localization lengths are 20 nm, 14 nm, and 0.9 nm for the films on LAO, NGO, and STO, respectively, clearly indicating that the hopping charges are significantly more localized in the film on STO. We then examined the extracted densities of state for impurity states at the Fermi level N (EF ), which is responsible for the hopping conduction at low temperature. The N (EF ) values are 6.2 × 1018 cm−3 eV−1 , 1.2 × 1019 cm−3 eV−1 , and 6.5 × 1021 cm−3 eV−1 for the films on LAO, NGO, and STO, respectively. The impurity level increases with the level of strain, consistent with the picture of structural defect formation due to strain relaxation. However, the N (EF ) in the film on STO is four orders of magnitude lower than the values previously reported in NdNiO3 thin films, where a first-order MIT was observed [13]. In addition, it has been shown for SmNiO3 thin films on STO that when a substantial fraction of the sample changes in the oxygenation state from Ni3+ to Ni2+ for strain relaxation, there is an expansion of the out-ofplane lattice parameter, opposite to what is expected from a tensile strain [16, 17]. The fact that the c-axis lattice constant for our films on STO is lower than the bulk value also points to a low density of oxygen vacancies in the system. We thus conclude that the structural disorder alone cannot account for the qualitatively different MIT transition characteristic in SNNO on STO, but rather it is intrinsic to the films under tensile strain and likely due to a different driving mechanism for localization. Above 30 K, equation (1) can no longer give an accurate description of ρ(T ) for films on LAO and NGO. It is important to note that in VRH, the averaging hopping energy W and hopping distance R are temperature dependent quantities given by: 1/4 9a 3 and R = . (3) W = 4π R 3 N (EF ) 8π kB T N (EF )
T (K)
200
and
300
T (K) Figure 3. Temperature dependences of carrier density and (inset) carrier mobility for 18 nm SNNO films on LAO (green squares), NGO (red circles), and STO (blue diamonds) substrates. The dashed lines are fits to the thermally activated model. The dotted line marks the carrier density corresponding to a bulk density of 4 holes uc−1 (7.3 × 1022 cm−3 ).
thermal hysteresis of 4 K in the R (T ) cycle. The film grown on STO, on the other hand, is significantly more resistive in the metallic phase and shows an above room temperature TMI of 330 K (figure 2(c)). The thermal hysteresis is fully suppressed, pointing to a second-order MIT as in bulk SNNO. Fitting to n(T ) yields an activation energy Ea of 25 meV (figure 3), which is surprisingly lower than the values for films on LAO and NGO. In addition, a good fit to R (T ) over a wide temperature range can be achieved by incorporating both the thermal activation and VRH mechanisms (figure 2(c) inset) with an Ea of 29 meV. This Ea is close to the value extracted from the Hall data, indicating the robustness of our method in obtaining the activation energy, and confirming that the energy gap in SNNO on STO is indeed suppressed. It is worth noting that this small Ea value is close to the transport gap observed in NdNiO3 thin films, where transport has also been interpreted as through parallel conduction of thermal activation and VRH [11]. As we will discuss later, it is likely that the dual conduction model only applies to systems with small energy gap and strong localization, so that the activation energy and hopping energy can both be comparable to the thermal energy over an extended temperature range. For films on LAO and NGO, which have much larger energy gaps, incorporating the parallel conduction mechanism does not yield a noticeable improvement to the fit, similar as in the case of SmNiO3 films [22]. We find that even though the TMI in SNNO films increases with increasing tensile strain, the magnitude of the transport gap does not follow the same trend. A possible scenario that can account for the suppressed energy gap in films on STO, which is subject to a substantial tensile strain (∼2.7%), is a higher disorder level due to the formation of high density oxygen vacancies for strain relaxation [16, 17]. To understand the electron localization mechanism and clarify the role of disorder in the conduction characteristic, we performed
The conditions for equation (1) to be applied is W/kB T > 1 and R/a > 1. As shown in table 1, the former condition is 4
J. Phys.: Condens. Matter 27 (2015) 132201
Table 1. Thermal activation energy Ea extracted from n(T ) below TMI and 3D VRH parameters based on equation (1) for 18 nm SNNO films on different substrates. Also listed are the crossover temperature Tcr at which the hopping energy W in equation (3) equals to the thermal energy kB T , and the ratio between the hopping distance R and the localization length a at Tcr .
Ea (meV)
T0 (K)
σ0 T 1/2 (−1 cm−1 K1/2 )
a (nm)
N (EF ) (cm−3 eV−1 )
Tcr (K)
R/a(Tcr )
Sample LAO NGO STO
79 77 25
4.4 × 103 6.6 × 103 3.8 × 104
3.6 × 102 4.2 × 102 2.6 × 103
20 14 0.9
6.2 × 1018 1.2 × 1019 6.5 × 1021
20 25 150
1.5 1.5 1.5
unsatisfied at a lower temperature. In figure 2 insets, we mark the crossover temperature Tcr at which W/kB T = 1. It is clear that around Tcr , ρ(T ) starts to deviate from the VRH model, and a more complicated model is required to describe the conduction associated with hopping. The strongly localized nature of the charge carriers in the film on STO is also manifested in the high temperature transport. In the metallic phase, the resistivity of the SNNO films increases with tensile strain, which can be due to either lower carrier density or reduced charge mobility. Figure 3 shows the carrier density of the 18 nm films. Films on all three substrates possess hole-type charge carriers with a similar density of ∼4 holes uc−1 in the metallic phase and thermally activated n(T ) in the insulating phase. We then calculated the mobility of the charge carriers using µ = σ /ne (figure 3 inset). In the metallic phase, films on LAO and NGO have similar carrier mobility of 0.3 cm2 V−1 s−1 and 0.4 cm2 V−1 s−1 , respectively, comparable with the value reported in NdNiO3 thin films (∼0.4 cm2 V−1 s−1 ) [24]. For the film on STO, the mobility has a significantly lower value of ∼0.05 cm2 V−1 s−1 , further confirming that the charges are more localized in this system. There is a clear contradiction between the results obtained on epitaxial thin films and those from chemical substitution and hydrostatic pressure experiments, where the increase of the lattice parameters corresponds to a reduced electronic bandwidth [3–8]. For the films on STO, which are subject to a higher tensile strain, the enhanced TMI is resulting from stronger electron localization, while the energy gap is actually suppressed. To explain this discrepancy, we need to consider the fundamental difference between the structural modifications induced by epitaxial strain and the other two approaches. In the chemical substitution experiments [3– 6], a smaller rare earth ion leads to a larger rotation of the NiO6 octahedral to accommodate the larger mismatch between the equilibrium R–O and Ni–O bond lengths. Such rotation results in a smaller Ni–O–Ni bond angle and reduces the electronic bandwidth. In the hydrostatic pressure experiments [6–8], since the Ni–O bond is more compressible, the pressure reduces the length difference between the R–O and Ni–O bonds, restoring the NiO6 octahedral rotation and resulting in less lattice distortion. In both experiments, the rotation of the NiO6 octahedral, or the Ni–O–Ni bond angle, is adjustable to minimize the energy of the system. In sharp contrast, epitaxial strain not only modifies the Ni– O and R–O bond lengths to accommodate the lattice mismatch, but also imposes the symmetry of the substrate crystal onto the strained film. This means the rotation of the oxygen
octahedral of the SNNO film has to conform to that of the substrate. SrTiO3 has a close to ideal cubic structure, and we expect that the Ni–O–Ni bond angle in strained SNNO on STO should be larger than the bulk value. It is thus subject to two competing mechanisms: the larger Ni–O–Ni bond angle enhances the electronic bandwidth, while the increased bond length enhances electron localization. From the x-ray data, we find the pseudo-cubic unit cell volume for the 40 nm SNNO film is ∼2% larger than the bulk value, significantly higher than the 0.2% volume expansion as nickelates going through the MIT [3]. Therefore, it is natural to expect that the effect of the increased Ni–O bond length dominates the competition, and the metallicity of the film will be suppressed. We then consider how these two mechanisms collectively determine electron localization in the system. In previous studies of the effect of hydrostatic pressure on thermal transport [6, 7] and the isotope effect [25] in the RNiO3 family, it has been proposed that the cooperative oxygen oscillation, and the associated modulations of the Ni–O bond length and local lattice deformation, can lead to dynamic Jahn–Teller effect and phase fluctuation between the localized and itinerant electronic states, which would narrow the electronic bandwidth in analogy to the small polaron scenario [26]: Wb = Wb exp(−λε/¯hωo ),
(4)
where Wb is the electronic bandwidth defined by the Ni–O–Ni bond angle, Wb is the modified bandwidth, ε is analogous to the polaron stabilization energy, ωo−1 is the phase fluctuation period for the cooperative oxygen displacements, and λ = ε/Wb is a measure of the coupling strength for the electron to the local distortion. In this scenario, the softening of the oxygen vibration modes with low frequency ωo would significantly reduce the electronic bandwidth of itinerant electrons. This model can naturally explain the anomalous transport properties of SNNO films on STO, i.e. a suppressed charge-transfer gap with enhanced electron localization. Under substantial tensile strain, the films possess increased Ni–O bond length that reduces the bonding strength and the associated oxygen oscillation frequency. The net effect is the bandwidth Wb in equation (3) can be substantially reduced even in the presence of an enhanced Wb due to an increased Ni–O–Ni bond angle imposed by the STO substrates. Within this framework, we can also explain why the MIT gradually evolves from first-order to second-order with increasing tensile strain. In the conventional picture of nickelates, there is a single structural parameter, the Ni–O– Ni bond angle, that controls both the electronic bandwidth and the superexchange coupling [3]. In epitaxial thin films, 5
J. Phys.: Condens. Matter 27 (2015) 132201
different thicknesses. Figure 4(a) shows the R (T ) for 2, 3, 4.4, and 18 nm SNNO films on LAO. For the 2 nm film, no MIT has been observed over the entire temperature range investigated. The critical thicknesses of such an electrical ‘dead layer’ are 2–3 nm for films on all substrates, similar to the values reported in other correlated oxide thin films [27]. The existence of such a dead layer has been attributed to the modified electronic structures at the interfacial and surface layers [28, 29]. For thick films the TMI is ∼200 K and does not show apparent variation with film thickness. It increases substantially in films thinner than 6 nm, reaching ∼310 K for the 3 nm films. In contrast, the TMI for films on NGO and STO shows very weak thickness dependence, even though the level of lattice relaxation increases with film thickness. This lack of thickness dependence in TMI and transport characteristics for SNNO on STO over a large film thickness range yields strong support to our previous conclusion that the softening of activation energy and enhanced electron localization are not simply driven by structural disorders, such as oxygen vacancies, induced by strain relaxation. Instead, they are the intrinsic characteristics of the system due to the substrate symmetry and tensile strain. For all three systems, films thinner than 8 nm exhibit 1/3 at low 2D VRH transport with ρ = ρ 0 exp T02D /T temperature. In films on LAO, which is subject to compressive strain, the oxygen vibration modes are stiffened and have a minor effect on limiting the electronic bandwidth. The dimensionality crossover may thus lead to a substantial increase of TMI , since enhanced electron localization is expected in 2D systems. For films on NGO and STO, on the other hand, electron localization is dominated by the dynamic local distortion depicted by equation (4), which is short-ranged and less sensitive to the reduced dimensionality. This result further highlights the importance of the competition among different energy scales in strongly correlated systems that can be tuned by strain. It is worth noting that the thermal hysteresis, signaling the first-order transition, persists in the 3 nm film on LAO, whose TMI is comparable with that of the films on STO. This also indicates that the quenched hysteresis in the film on STO is an intrinsic feature of a second-order MIT, rather than due to large thermal fluctuation that suppresses the dynamics of phase transformation in the phase coexistence regime [3]. In conclusion, working with high quality epitaxial SNNO thin films, we have demonstrated that epitaxial strain can not only tune the MIT transition temperature by more than 100 K, but also modify the characteristics of the MIT between firstorder and second-order. We attribute the diverse transport characteristics to strain induced modulation of the Ni–O bond length and Ni–O–Ni bond angle, and emphasize the critical role played by the corporative fluctuations of oxygen displacements in the system in limiting electron delocalization. Our results indicate that tailored electronic and magnetic properties can be achieved in SNNO thin films via controlling the various competing energy scales through strain engineering.
Figure 4. (a) Temperature dependence of sheet resistance for 2, 3, 4.4, and 18 nm SNNO films on LAO. The 2 nm film is insulating over the entire temperature range and does not show thermal hysteresis. (b) Thickness dependence of TMI for SNNO films on LAO, NGO, and STO substrates. The dotted line marks the crossover from 2D VRH to 3D VRH.
the modified Ni–O bond length also contributes to electron localization, which can either complement or compete with the effect associated with the Ni–O–Ni bond angle. For films on LAO, a substrate with a less distorted rhombohedral structure, the shorter Ni–O bond length and a larger superexchange angle collectively lead to higher TN and lower TMI [11]. When the TMI is lower than the virtual TN , the antiferromagnetic coupling between Ni magnetic moments is suppressed by the itinerant electrons, leading to a sudden quench of magnetic order and a first-order MIT and magnetic transition [10, 14]. For comparison, in the hydrostatic pressure experiments, where the Ni–O–Ni bond angle and Ni–O bond length cannot be controlled independently, no first-order transition has been observed in bulk SNNO at pressure up to 15.7 kbar [6, 7]. As SNNO is at the phase boundary of critical tolerance factor where TN and TMI decouple, a small, −0.37% compressive strain on LAO is sufficient to induce a first-order transition. In SmNiO3 , which is more distorted, a first-order MIT has only been induced when the film is subject to a higher compressive strain of −0.9% [14]. Conversely, for films on STO, the stretched Ni–O bond length significantly enhances electron localization temperature so that TMI > TN , leading to a secondorder transition. To further understand the effect of strain relaxation, we investigated the transport properties of SNNO films with
Acknowledgments
We would like to thank X Xu for insightful discussions and technical assistance. LZ and HJG contributed equally to this 6
J. Phys.: Condens. Matter 27 (2015) 132201
work. The sample growth and structural characterizations have been supported by NSF CAREER Grant No. DMR1148783. The device fabrication and transport studies have been supported by the Center for NanoFerroic Devices (CNFD) and the Nanoelectronics Research Initiative (NRI). The data analysis and manuscript preparation have been supported by NSF Grant No. DMR-1409622 and the Nebraska Materials Research Science and Engineering Center (MRSEC) (Grant No. DMR-1420645). This research was performed in part in Central Facilities of the Nebraska Center for Materials and Nanoscience, which is supported by the Nebraska Research Initiative.
[10] Zhou J-S, Goodenough J, Dabrowski B, Klamut P and Bukowski Z 2000 Phys. Rev. Lett. 84 526 [11] Catalan G, Bowman R and Gregg J 2000 Phys. Rev. B 62 7892 [12] Tiwari A, Jin C and Narayan J 2002 Appl. Phys. Lett. 80 4039 [13] Xiang P-H, Zhong N, Duan C-G, Tang X, Hu Z, Yang P, Zhu Z and Chu J 2013 J. Appl. Phys. 114 243713 [14] Catalano S et al 2014 APL Mater. 2 116110 [15] Shukla N, Joshi T, Dasgupta S, Borisov P, Lederman D and Datta S 2014 Appl. Phys. Lett. 105 012108 [16] Conchon F, Boulle A, Guinebreti`ere R, Girardot C, Pignard S, Kreisel J, Weiss F, Dooryh´ee E and Hodeau J-L 2007 Appl. Phys. Lett. 91 192110 [17] Conchon F et al 2008 J. Appl. Phys. 103 123501 [18] Ambrosini A and Hamet J-F 2003 Appl. Phys. Lett. 82 727 [19] Lian X, Chen F, Tan X, Wang L, Xuan X, Gao G, Jin S and Wu W 2013 AIP Adv. 3 062133 [20] Frand G, Bohnke O, Lacorre P, Fourquet J, Carr´e A, Eid B, Th´eobald J and Gire A 1995 J. Solid State Chem. 120 157 [21] Singh V R, Zhang L, Rajapitamahuni A K, Devries N and Hong X 2014 J. Appl. Phys. 116 033914 [22] Ha S D, Aydogdu G H and Ramanathan S 2011 J. Appl. Phys. 110 094102 [23] Paul D and Mitra S 1973 Phys. Rev. Lett. 31 1000 [24] Scherwitzl R, Zubko P, Lezama I G, Ono S, Morpurgo A F, Catalan G and Triscone J M 2010 Adv. Mater. 22 5517 [25] Medarde M, Lacorre P, Conder K, Fauth F and Furrer A 1998 Phys. Rev. Lett. 80 2397 [26] Holstein T 1959 Ann. Phys. 8 343 [27] Hong X, Posadas A and Ahn C H 2005 Appl. Phys. Lett. 86 142501 [28] Dulli H, Dowben P A, Liou SH and Plummer EW 2000 Phys. LRev. B 62 R14629 [29] Dulli H, Plummer E W, Dowben P A, Choi J and Liou S-H 2000 Appl. Phys. Lett. 77 570
References [1] Imada M, Fujimori A and Tokura Y 1998 Rev. Mod. Phys. 70 1039 [2] Goodenough J B and Zhou J S 2001 Transport properties Localized to Itinerant Electronic Transition in Perovskite Oxides, Structure and Bonding vol 98, ed J B Goodenough (Berlin: Springer) p 17 [3] Medarde M L 1997 J. Phys.: Condens. Matter 9 1679 [4] Catalan G 2008 Phase Transit. 81 729 [5] Torrance J, Lacorre P, Nazzal A, Ansaldo E and Niedermayer C 1992 Phys. Rev. B 45 8209 [6] Zhou J-S, Goodenough J, Dabrowski B, Klamut P and Bukowski Z 2000 Phys. Rev. B 61 4401 [7] Zhou J-S, Goodenough J and Dabrowski B 2003 Phys. Rev. B 67 020404 [8] Medarde M, Mesot J, Lacorre P, Rosenkranz S, Fischer P and Gobrecht K 1995 Phys. Rev. B 52 9248 [9] Caviglia A et al 2012 Phys. Rev. Lett. 108 136801
7
Search
Collections
Journals
About
Contact us
My IOPscience
Strain induced modulation of the correlated transport in epitaxial Sm0.5Nd0.5NiO3 thin films
This content has been downloaded from IOPscience. Please scroll down to see the full text. 2015 J. Phys.: Condens. Matter 27 132201 (http://iopscience.iop.org/0953-8984/27/13/132201) View the table of contents for this issue, or go to the journal homepage for more
Download details: IP Address: 169.230.243.252 This content was downloaded on 23/04/2015 at 08:24
Please note that terms and conditions apply.
Journal of Physics: Condensed Matter J. Phys.: Condens. Matter 27 (2015) 132201 (7pp)
doi:10.1088/0953-8984/27/13/132201
Fast Track Communication
Strain induced modulation of the correlated transport in epitaxial Sm0.5Nd0.5NiO3 thin films L Zhang1 , H J Gardner1 , X G Chen1 , V R Singh1 and X Hong1,2 1 2
Department of Physics and Astronomy, University of Nebraska-Lincoln, NE 68588, USA Nebraska Center for Materials and Nanoscience, University of Nebraska-Lincoln, NE 68588, USA
E-mail: [email protected] Received 6 February 2015, revised 16 February 2015 Accepted for publication 25 February 2015 Published 17 March 2015 Abstract
We report a study of the effect of epitaxial strain on the correlated transport properties of 2–40 nm Sm0.5 Nd0.5 NiO3 (SNNO) films grown on different substrates. The metal–insulator transition (MIT) temperature TMI of the SNNO films increases with increasing tensile strain. While films on (0 0 1) LaAlO3 and (1 1 0) NdGaO3 substrates exhibit a sharp MIT and thermal hysteresis in the cooling–heating cycle, signaling a first-order transition, films on (0 0 1) SrTiO3 show a broad, second-order MIT. Hall effect measurements reveal hole-type charge carriers and thermally activated temperature dependence of the carrier density below TMI . The corresponding activation energy is ∼80 meV for films on LaAlO3 and NdGaO3 , and is suppressed to 25 meV for films on SrTiO3 . The carrier mobility in the metallic state and variable range hopping (VRH) transport at a low temperature point significantly enhanced electron localization in SNNO on STO, which we believe is not simply driven by extrinsic effects such as oxygen vacancies, but rather is an intrinsic characteristic for films subject to tensile strain due to the elongated Ni–O bond and hence enhanced dynamic Jahn–Teller distortion. In ultrathin films above the electrical dead layer thickness (2–3 nm), we observe a more than 100 K increase of TMI for films on LaAlO3 , which has been correlated with a crossover from 3D to 2D transport as revealed from VRH. We attribute the distinct transport characteristics to strain induced modulation of various energy scales associated with the Ni–O–Ni bond angle and Ni–O bond length, which collectively determine the delocalization bandwidth of the system. Keywords: strongly correlated oxides, nickelate thin films, epitaxial strain, metal–insulator transition, lattice distortion, electron localization, dynamic Jahn–Teller distortion (Some figures may appear in colour only in the online journal)
transition [3–5]. The MIT transition temperature TMI increases monotonically in compounds with smaller size rare earth ion R due to higher levels of lattice distortion [3–5]. The latticedriven modulation of the MIT can also be achieved by applying hydrostatic pressure to bulk nickelates [3, 6–8] or, for epitaxial thin films, optical excitation of the phonon modes in the substrates [9]. Electron localization in these materials is also
Strongly correlated oxides such as the perovskite nickelates RNiO3 (R = rare earth) exhibit a rich variety of electronic and magnetic ground states that can be sensitively tuned by the lattice degree of freedom [1–4]. For example, the RNiO3 oxides, except LaNiO3 , are charge-transfer type Mott insulators, exhibiting a temperature-driven metal–insulator transition (MIT) as well as a paramagnetic to antiferromagnetic 0953-8984/15/132201+07$33.00
1
© 2015 IOP Publishing Ltd Printed in the UK
J. Phys.: Condens. Matter 27 (2015) 132201
3
10
2
10
1 µm
SNNO 001
4
10
SNNO 002
5
10
LAO
1
3
10
2
SNNO 001
4
10
NGO 110
10
1
5
3
10
2
10
1 µm
STO 001
4
10
SNNO 001
10
STO
3
10
SNNO 001
1
10
23
24
θ -2θ (deg.)
1
10
20
STO 001
SNNO 002
10
STO 002
Intensity (a.u.)
1 µm
10
SNNO 002
NGO 220
10
5
NGO
LAO 002
(b) 106 LAO 001
(a)
Intensity (a.u.)
entangled with the magnetic order. Compounds with lower distortion level, such as PrNiO3 and NdNiO3 , exhibit a firstorder MIT with TMI coinciding with the antiferromagnetic order temperature TN [7, 10]. For compounds with smaller size R ions and higher orthorhombic distortion, the MIT and magnetic transition become second-order, and the transition temperatures are decoupled. A powerful tool to systematically probe the effect of lattice distortion on the MIT in nickelates is to impose different levels of strain on epitaxial thin films through lattice mismatched substrates. It has been reported that epitaxial strain can effectively tune the TMI in NdNiO3 [11–13] and SmNiO3 [14, 15] thin films by more than 100 K. Sm0.5 Nd0.5 NiO3 (SNNO) is an especially interesting system in the RNiO3 family as it is close to the phase boundary of critical tolerance factor where TMI and TN become decoupled [7, 10], which means in this system the competition among different energy scales to stabilize the electronic and magnetic ground states are highly susceptible to small external perturbations such as structural variation. In addition, the TMI of SNNO is close to room temperature, which has high technological relevance in building oxide-based electronic and spintronic devices. To date, although extensive research on the effect of epitaxial strain has been carried out on the two parent compounds, NdNiO3 [11–13] and SmNiO3 [14–17], studies on high quality epitaxial thin films of Sm0.5 Nd0.5 NiO3 have been very limited and mostly focused on the growth [18, 19]. In this work, we report a comprehensive study of the transport properties of 2–40 nm epitaxial SNNO films grown on (0 0 1) SrTiO3 (STO), (0 0 1) LaAlO3 (LAO), and (1 1 0) NdGaO3 (NGO) substrates. Bulk SNNO has a pseudocubic lattice constant of ∼3.804 Å [20], corresponding to a compressive strain of −0.37% on LAO (aLAO = 3.79 Å), and tensile strains of 1.5% on NGO (aNGO = 3.86 Å) and 2.7% on STO (aSTO = 3.905 Å). While bulk SNNO shows a secondorder MIT, SNNO films grown on LAO and NGO exhibit a sharp MIT with thermal hysteresis observed in the resistance cooling–heating cycle, signaling a first-order transition. The temperature dependence of the carrier density right below TMI can be well described by thermally activated behavior, from which we extracted the activation energies of 79 and 77 meV for the films on LAO and NGO, respectively. In sharp contrast, films grown on STO show the characteristic of a second-order MIT and a softened activation energy of 25 meV. Based on the carrier mobility in the metallic state and the low temperature variable range hopping (VRH) transport, we conclude that such distinct transition characteristics are driven by different electron localization mechanisms rather than simply due to different levels of structural disorders, such as oxygen vacancies, associated with strain relaxation. We have also investigated the effect of the film thickness on the correlated transport properties. While TMI for films on NGO and STO only show weak thickness dependence, there is a more than 100 K increase of TMI in ultrathin films on LAO above the electrical dead layer thickness (2–3 nm), which has been correlated to a dimensionality crossover from 3D to 2D based on VRH at low temperature. Unlike the chemical substitution and hydrostatic pressure experiments, where the key control parameter is the Ni–O–Ni bond angle [3–8], we
30
40
θ -2θ (deg.)
50
Figure 1. (a) AFM images of 4.4 nm SNNO films grown on LAO, NGO, and STO substrates showing clear atomically flat terraces. The RMS roughness of all three films is 1–2 Å. (b) XRD θ –2θ scans for 40 nm SNNO films grown on LAO (upper panel), NGO (middle panel), and STO (lower panel) substrates. Inset: Laue finite size fringes around the (0 0 1) peak (solid line) with the fit (dashed line) for the 40 nm SNNO on STO.
believe strain induced modulation in the Ni–O bond length also plays a critical role in driving the diverse transport behaviors in epitaxial SNNO films on different substrates. Such modulation would change the cooperative oxygen displacement frequency and in turn control the dynamic Jahn–Teller distortion, which competes with the effect associated with the Ni–O–Ni bond angle in limiting the delocalization bandwidth. We fabricated the epitaxial SNNO films using off-axis radio frequency magnetron sputtering. The films were grown at a deposition temperature of 500 ◦ C in 135 mTorr process gas composed of Ar and O2 (ratio of 1 : 2). After growth, the films were cooled down to room temperature in 1 atmosphere of O2 . For each type of substrate, we have fabricated films with six different thicknesses in the range of 2–40 nm. We examined the surface morphology of the samples using atomic force microscopy (AFM). As shown in figure 1(a), films grown on all three types of substrates possess atomically smooth surfaces with root mean square (RMS) roughness of 1–2 Å. The structural properties of the films were characterized using x-ray diffraction (XRD). Figure 1(b) shows the x-ray θ–2θ scans taken on 40 nm SNNO films grown on different substrates. The film thicknesses were determined from x-ray reflectivity measurements and Laue fringes around the (0 0 1) Bragg peaks (figure 1(b) inset). The x-ray spectra reveal (0 0 1) growth in the pseudo-cubic structure for all samples, with no impurity phases observed for the thickest films studied. The c-axis lattice constants are 3.81 Å, 3.76 Å, and 3.79 Å for the 40 nm films on LAO, NGO, and STO, respectively, consistent with strained films. The corresponding rocking curves for the 2
J. Phys.: Condens. Matter 27 (2015) 132201
(0 0 2) peaks have full-width-half-maximums of 0.12◦ , 0.08◦ , and 0.06◦ , reflecting very high crystalline quality. The out-ofplane lattice constant for the 40 nm film on STO is higher than that for the film on NGO, even though it is subject to a higher tensile strain. Reciprocal space mapping measurements on this sample reveal an in-plane lattice constant of 3.85 Å, showing SNNO on STO is partially relaxed at this thickness. For transport measurements, we patterned the SNNO films into standard Hall bar devices using photolithography. The four-point resistance channels have aspect ratios of 1 or 2 with the channel width varying from 5 to 40 µm. Transport measurements were carried out between 10 and 350 K at a cooling/heating rate of ∼4 K min−1 using Quantum Design PPMS combined with Keithley 2400 SourceMeter. For each sample, we examined the I –V relation at different temperatures to verify that the resistance value obtained is current-independent. Nonlinear I –V characteristics has only been observed in the metallic phase at an excitation current below 1 µA, which is common in correlated oxides due to either work function mismatch at the contacts or phase inhomogeneity within the sample [21]. All measurements were taken in the linear I –V regime with low excitation currents below 10 µA to avoid Joule heating, so that the observed modulation of the transport properties is intrinsic to the sample. We also fabricated multiple Hall bar devices (up to 4) at different locations on the same film, and confirmed that the film properties are homogeneous. Figure 2(a) shows the temperature dependence of sheet resistance R (T ) for a 18 nm SNNO film grown on LAO, which is subject to a moderate compressive strain. The resistance of the film exhibits metallic temperature dependence at high temperature, followed by a sharp transition to an insulating behavior below TMI . There is a pronounced thermal hysteresis between the R (T ) curves on heating and cooling, signaling a first-order MIT [10]. This is distinctly different from bulk SNNO, where the MIT is second-order [7, 10]. The transition temperature TMI , defined as the temperature where dR/dT changes sign on cooling, is 195 K. The thermal hysteresis leads to an 11 K shift between the deflection points for R (T ) on cooling and heating. In the insulating phase, the film exhibits two distinct temperature dependences. Below TMI , there is a sharp transition to the insulating phase between 170 and 135 K for R (T ) on cooling. This temperature dependence has previously been modeled as percolative transport due to coexisting metallic and semiconducting phases [11]. To probe the formation of a correlated gap, we examined the temperature dependence of the carrier density extracted from Hall effect measurements (figure 3). The film possesses hole-type charge carriers with a temperature-independent density of n ∼ 4 holes per unit cell (uc) in the metallic phase. Below TMI , n(T ) decreases exponentially with decreasing temperature, well described by the thermally activated model. Fitting to n(T ) = n0 exp[−(Ea /kB T)] reveals an activation energy Ea of 79 meV (figure 3). This energy gap is between the values reported for the parent compounds NdNiO3 (∼20 meV) [11] and SmNiO3 films (∼170 meV) [22], consistent with the picture that a smaller R ion size leads to a reduced electronic
Figure 2. Temperature dependence of sheet resistance R (T ) on cooling (black) and heating (red) for 18 nm SNNO films grown on (a) LAO, (b) NGO, and (c) STO substrates. Insets: ρ versus T −1/4 on cooling with fits (dashed lines) to the 3D VRH model or the parallel conduction model with both thermal activation and VRH behaviors. The dotted lines mark the temperature Tcr at which the VRH hopping energy W ∼ kB T .
bandwidth [3–5]. The conduction shows a weaker temperature dependence as the sample is cooled below ∼135 K, where phonon assisted hopping between impurity states within the energy gap becomes progressively important [23]. Below 30 K, the temperature dependence of resistivity is dominated by 3D VRH type transport as ρ = ρ0 exp (T0 /T )1/4 . We then investigated the SNNO films grown on NGO and STO substrates with the same thickness, which exhibit significantly different transport characteristics even though both systems are subject to tensile strain. For the 18 nm SNNO film grown on NGO, R (T ) shows qualitatively similar behavior as the film on LAO, exhibiting percolative conduction below TMI followed by impurity dominated hopping at low temperature (figure 2(b)). From n(T ) we extracted a similar activation energy of 77 meV (figure 3). Compared with the films on LAO, this film shows a higher TMI of 260 K, suggesting enhanced electron localization, and a reduced 3
10
23
10
22
quantitative VRH analyses of the low temperature ρ(T ) in these three systems. In Mott’s VRH model [23], conductivity σ in a 3D disordered system follows the relation: T0 1/4 , (1) σ = σ0 exp − T
4 holes/uc
0.6 2
µ (cm /Vs)
-3
n (cm )
J. Phys.: Condens. Matter 27 (2015) 132201
10
21
100
0.2 0.0
LAO NGO STO
where √ N (EF ) a 1/2 σ0 T = 3e2 γph 8π kB
0.4
200 250 300 350
T0 =
18.1 . kB N (EF )a 3 (2)
Here a is the localization length of the charge carriers, γph ∼ 1013 s−1 is the optical phonon frequency calculated from the Debye temperature, and N (EF ) is the density of states of impurity levels at the Fermi level. Fitting ρ(T ) on cooling to equation (1) (figure 2 insets), we obtained the 3D VRH parameters as listed in table 1. We first considered the localization length a for the charge carriers, which measures the spatial decay of the electron wavefunction and reflects directly the degree of charge localization. The localization lengths are 20 nm, 14 nm, and 0.9 nm for the films on LAO, NGO, and STO, respectively, clearly indicating that the hopping charges are significantly more localized in the film on STO. We then examined the extracted densities of state for impurity states at the Fermi level N (EF ), which is responsible for the hopping conduction at low temperature. The N (EF ) values are 6.2 × 1018 cm−3 eV−1 , 1.2 × 1019 cm−3 eV−1 , and 6.5 × 1021 cm−3 eV−1 for the films on LAO, NGO, and STO, respectively. The impurity level increases with the level of strain, consistent with the picture of structural defect formation due to strain relaxation. However, the N (EF ) in the film on STO is four orders of magnitude lower than the values previously reported in NdNiO3 thin films, where a first-order MIT was observed [13]. In addition, it has been shown for SmNiO3 thin films on STO that when a substantial fraction of the sample changes in the oxygenation state from Ni3+ to Ni2+ for strain relaxation, there is an expansion of the out-ofplane lattice parameter, opposite to what is expected from a tensile strain [16, 17]. The fact that the c-axis lattice constant for our films on STO is lower than the bulk value also points to a low density of oxygen vacancies in the system. We thus conclude that the structural disorder alone cannot account for the qualitatively different MIT transition characteristic in SNNO on STO, but rather it is intrinsic to the films under tensile strain and likely due to a different driving mechanism for localization. Above 30 K, equation (1) can no longer give an accurate description of ρ(T ) for films on LAO and NGO. It is important to note that in VRH, the averaging hopping energy W and hopping distance R are temperature dependent quantities given by: 1/4 9a 3 and R = . (3) W = 4π R 3 N (EF ) 8π kB T N (EF )
T (K)
200
and
300
T (K) Figure 3. Temperature dependences of carrier density and (inset) carrier mobility for 18 nm SNNO films on LAO (green squares), NGO (red circles), and STO (blue diamonds) substrates. The dashed lines are fits to the thermally activated model. The dotted line marks the carrier density corresponding to a bulk density of 4 holes uc−1 (7.3 × 1022 cm−3 ).
thermal hysteresis of 4 K in the R (T ) cycle. The film grown on STO, on the other hand, is significantly more resistive in the metallic phase and shows an above room temperature TMI of 330 K (figure 2(c)). The thermal hysteresis is fully suppressed, pointing to a second-order MIT as in bulk SNNO. Fitting to n(T ) yields an activation energy Ea of 25 meV (figure 3), which is surprisingly lower than the values for films on LAO and NGO. In addition, a good fit to R (T ) over a wide temperature range can be achieved by incorporating both the thermal activation and VRH mechanisms (figure 2(c) inset) with an Ea of 29 meV. This Ea is close to the value extracted from the Hall data, indicating the robustness of our method in obtaining the activation energy, and confirming that the energy gap in SNNO on STO is indeed suppressed. It is worth noting that this small Ea value is close to the transport gap observed in NdNiO3 thin films, where transport has also been interpreted as through parallel conduction of thermal activation and VRH [11]. As we will discuss later, it is likely that the dual conduction model only applies to systems with small energy gap and strong localization, so that the activation energy and hopping energy can both be comparable to the thermal energy over an extended temperature range. For films on LAO and NGO, which have much larger energy gaps, incorporating the parallel conduction mechanism does not yield a noticeable improvement to the fit, similar as in the case of SmNiO3 films [22]. We find that even though the TMI in SNNO films increases with increasing tensile strain, the magnitude of the transport gap does not follow the same trend. A possible scenario that can account for the suppressed energy gap in films on STO, which is subject to a substantial tensile strain (∼2.7%), is a higher disorder level due to the formation of high density oxygen vacancies for strain relaxation [16, 17]. To understand the electron localization mechanism and clarify the role of disorder in the conduction characteristic, we performed
The conditions for equation (1) to be applied is W/kB T > 1 and R/a > 1. As shown in table 1, the former condition is 4
J. Phys.: Condens. Matter 27 (2015) 132201
Table 1. Thermal activation energy Ea extracted from n(T ) below TMI and 3D VRH parameters based on equation (1) for 18 nm SNNO films on different substrates. Also listed are the crossover temperature Tcr at which the hopping energy W in equation (3) equals to the thermal energy kB T , and the ratio between the hopping distance R and the localization length a at Tcr .
Ea (meV)
T0 (K)
σ0 T 1/2 (−1 cm−1 K1/2 )
a (nm)
N (EF ) (cm−3 eV−1 )
Tcr (K)
R/a(Tcr )
Sample LAO NGO STO
79 77 25
4.4 × 103 6.6 × 103 3.8 × 104
3.6 × 102 4.2 × 102 2.6 × 103
20 14 0.9
6.2 × 1018 1.2 × 1019 6.5 × 1021
20 25 150
1.5 1.5 1.5
unsatisfied at a lower temperature. In figure 2 insets, we mark the crossover temperature Tcr at which W/kB T = 1. It is clear that around Tcr , ρ(T ) starts to deviate from the VRH model, and a more complicated model is required to describe the conduction associated with hopping. The strongly localized nature of the charge carriers in the film on STO is also manifested in the high temperature transport. In the metallic phase, the resistivity of the SNNO films increases with tensile strain, which can be due to either lower carrier density or reduced charge mobility. Figure 3 shows the carrier density of the 18 nm films. Films on all three substrates possess hole-type charge carriers with a similar density of ∼4 holes uc−1 in the metallic phase and thermally activated n(T ) in the insulating phase. We then calculated the mobility of the charge carriers using µ = σ /ne (figure 3 inset). In the metallic phase, films on LAO and NGO have similar carrier mobility of 0.3 cm2 V−1 s−1 and 0.4 cm2 V−1 s−1 , respectively, comparable with the value reported in NdNiO3 thin films (∼0.4 cm2 V−1 s−1 ) [24]. For the film on STO, the mobility has a significantly lower value of ∼0.05 cm2 V−1 s−1 , further confirming that the charges are more localized in this system. There is a clear contradiction between the results obtained on epitaxial thin films and those from chemical substitution and hydrostatic pressure experiments, where the increase of the lattice parameters corresponds to a reduced electronic bandwidth [3–8]. For the films on STO, which are subject to a higher tensile strain, the enhanced TMI is resulting from stronger electron localization, while the energy gap is actually suppressed. To explain this discrepancy, we need to consider the fundamental difference between the structural modifications induced by epitaxial strain and the other two approaches. In the chemical substitution experiments [3– 6], a smaller rare earth ion leads to a larger rotation of the NiO6 octahedral to accommodate the larger mismatch between the equilibrium R–O and Ni–O bond lengths. Such rotation results in a smaller Ni–O–Ni bond angle and reduces the electronic bandwidth. In the hydrostatic pressure experiments [6–8], since the Ni–O bond is more compressible, the pressure reduces the length difference between the R–O and Ni–O bonds, restoring the NiO6 octahedral rotation and resulting in less lattice distortion. In both experiments, the rotation of the NiO6 octahedral, or the Ni–O–Ni bond angle, is adjustable to minimize the energy of the system. In sharp contrast, epitaxial strain not only modifies the Ni– O and R–O bond lengths to accommodate the lattice mismatch, but also imposes the symmetry of the substrate crystal onto the strained film. This means the rotation of the oxygen
octahedral of the SNNO film has to conform to that of the substrate. SrTiO3 has a close to ideal cubic structure, and we expect that the Ni–O–Ni bond angle in strained SNNO on STO should be larger than the bulk value. It is thus subject to two competing mechanisms: the larger Ni–O–Ni bond angle enhances the electronic bandwidth, while the increased bond length enhances electron localization. From the x-ray data, we find the pseudo-cubic unit cell volume for the 40 nm SNNO film is ∼2% larger than the bulk value, significantly higher than the 0.2% volume expansion as nickelates going through the MIT [3]. Therefore, it is natural to expect that the effect of the increased Ni–O bond length dominates the competition, and the metallicity of the film will be suppressed. We then consider how these two mechanisms collectively determine electron localization in the system. In previous studies of the effect of hydrostatic pressure on thermal transport [6, 7] and the isotope effect [25] in the RNiO3 family, it has been proposed that the cooperative oxygen oscillation, and the associated modulations of the Ni–O bond length and local lattice deformation, can lead to dynamic Jahn–Teller effect and phase fluctuation between the localized and itinerant electronic states, which would narrow the electronic bandwidth in analogy to the small polaron scenario [26]: Wb = Wb exp(−λε/¯hωo ),
(4)
where Wb is the electronic bandwidth defined by the Ni–O–Ni bond angle, Wb is the modified bandwidth, ε is analogous to the polaron stabilization energy, ωo−1 is the phase fluctuation period for the cooperative oxygen displacements, and λ = ε/Wb is a measure of the coupling strength for the electron to the local distortion. In this scenario, the softening of the oxygen vibration modes with low frequency ωo would significantly reduce the electronic bandwidth of itinerant electrons. This model can naturally explain the anomalous transport properties of SNNO films on STO, i.e. a suppressed charge-transfer gap with enhanced electron localization. Under substantial tensile strain, the films possess increased Ni–O bond length that reduces the bonding strength and the associated oxygen oscillation frequency. The net effect is the bandwidth Wb in equation (3) can be substantially reduced even in the presence of an enhanced Wb due to an increased Ni–O–Ni bond angle imposed by the STO substrates. Within this framework, we can also explain why the MIT gradually evolves from first-order to second-order with increasing tensile strain. In the conventional picture of nickelates, there is a single structural parameter, the Ni–O– Ni bond angle, that controls both the electronic bandwidth and the superexchange coupling [3]. In epitaxial thin films, 5
J. Phys.: Condens. Matter 27 (2015) 132201
different thicknesses. Figure 4(a) shows the R (T ) for 2, 3, 4.4, and 18 nm SNNO films on LAO. For the 2 nm film, no MIT has been observed over the entire temperature range investigated. The critical thicknesses of such an electrical ‘dead layer’ are 2–3 nm for films on all substrates, similar to the values reported in other correlated oxide thin films [27]. The existence of such a dead layer has been attributed to the modified electronic structures at the interfacial and surface layers [28, 29]. For thick films the TMI is ∼200 K and does not show apparent variation with film thickness. It increases substantially in films thinner than 6 nm, reaching ∼310 K for the 3 nm films. In contrast, the TMI for films on NGO and STO shows very weak thickness dependence, even though the level of lattice relaxation increases with film thickness. This lack of thickness dependence in TMI and transport characteristics for SNNO on STO over a large film thickness range yields strong support to our previous conclusion that the softening of activation energy and enhanced electron localization are not simply driven by structural disorders, such as oxygen vacancies, induced by strain relaxation. Instead, they are the intrinsic characteristics of the system due to the substrate symmetry and tensile strain. For all three systems, films thinner than 8 nm exhibit 1/3 at low 2D VRH transport with ρ = ρ 0 exp T02D /T temperature. In films on LAO, which is subject to compressive strain, the oxygen vibration modes are stiffened and have a minor effect on limiting the electronic bandwidth. The dimensionality crossover may thus lead to a substantial increase of TMI , since enhanced electron localization is expected in 2D systems. For films on NGO and STO, on the other hand, electron localization is dominated by the dynamic local distortion depicted by equation (4), which is short-ranged and less sensitive to the reduced dimensionality. This result further highlights the importance of the competition among different energy scales in strongly correlated systems that can be tuned by strain. It is worth noting that the thermal hysteresis, signaling the first-order transition, persists in the 3 nm film on LAO, whose TMI is comparable with that of the films on STO. This also indicates that the quenched hysteresis in the film on STO is an intrinsic feature of a second-order MIT, rather than due to large thermal fluctuation that suppresses the dynamics of phase transformation in the phase coexistence regime [3]. In conclusion, working with high quality epitaxial SNNO thin films, we have demonstrated that epitaxial strain can not only tune the MIT transition temperature by more than 100 K, but also modify the characteristics of the MIT between firstorder and second-order. We attribute the diverse transport characteristics to strain induced modulation of the Ni–O bond length and Ni–O–Ni bond angle, and emphasize the critical role played by the corporative fluctuations of oxygen displacements in the system in limiting electron delocalization. Our results indicate that tailored electronic and magnetic properties can be achieved in SNNO thin films via controlling the various competing energy scales through strain engineering.
Figure 4. (a) Temperature dependence of sheet resistance for 2, 3, 4.4, and 18 nm SNNO films on LAO. The 2 nm film is insulating over the entire temperature range and does not show thermal hysteresis. (b) Thickness dependence of TMI for SNNO films on LAO, NGO, and STO substrates. The dotted line marks the crossover from 2D VRH to 3D VRH.
the modified Ni–O bond length also contributes to electron localization, which can either complement or compete with the effect associated with the Ni–O–Ni bond angle. For films on LAO, a substrate with a less distorted rhombohedral structure, the shorter Ni–O bond length and a larger superexchange angle collectively lead to higher TN and lower TMI [11]. When the TMI is lower than the virtual TN , the antiferromagnetic coupling between Ni magnetic moments is suppressed by the itinerant electrons, leading to a sudden quench of magnetic order and a first-order MIT and magnetic transition [10, 14]. For comparison, in the hydrostatic pressure experiments, where the Ni–O–Ni bond angle and Ni–O bond length cannot be controlled independently, no first-order transition has been observed in bulk SNNO at pressure up to 15.7 kbar [6, 7]. As SNNO is at the phase boundary of critical tolerance factor where TN and TMI decouple, a small, −0.37% compressive strain on LAO is sufficient to induce a first-order transition. In SmNiO3 , which is more distorted, a first-order MIT has only been induced when the film is subject to a higher compressive strain of −0.9% [14]. Conversely, for films on STO, the stretched Ni–O bond length significantly enhances electron localization temperature so that TMI > TN , leading to a secondorder transition. To further understand the effect of strain relaxation, we investigated the transport properties of SNNO films with
Acknowledgments
We would like to thank X Xu for insightful discussions and technical assistance. LZ and HJG contributed equally to this 6
J. Phys.: Condens. Matter 27 (2015) 132201
work. The sample growth and structural characterizations have been supported by NSF CAREER Grant No. DMR1148783. The device fabrication and transport studies have been supported by the Center for NanoFerroic Devices (CNFD) and the Nanoelectronics Research Initiative (NRI). The data analysis and manuscript preparation have been supported by NSF Grant No. DMR-1409622 and the Nebraska Materials Research Science and Engineering Center (MRSEC) (Grant No. DMR-1420645). This research was performed in part in Central Facilities of the Nebraska Center for Materials and Nanoscience, which is supported by the Nebraska Research Initiative.
[10] Zhou J-S, Goodenough J, Dabrowski B, Klamut P and Bukowski Z 2000 Phys. Rev. Lett. 84 526 [11] Catalan G, Bowman R and Gregg J 2000 Phys. Rev. B 62 7892 [12] Tiwari A, Jin C and Narayan J 2002 Appl. Phys. Lett. 80 4039 [13] Xiang P-H, Zhong N, Duan C-G, Tang X, Hu Z, Yang P, Zhu Z and Chu J 2013 J. Appl. Phys. 114 243713 [14] Catalano S et al 2014 APL Mater. 2 116110 [15] Shukla N, Joshi T, Dasgupta S, Borisov P, Lederman D and Datta S 2014 Appl. Phys. Lett. 105 012108 [16] Conchon F, Boulle A, Guinebreti`ere R, Girardot C, Pignard S, Kreisel J, Weiss F, Dooryh´ee E and Hodeau J-L 2007 Appl. Phys. Lett. 91 192110 [17] Conchon F et al 2008 J. Appl. Phys. 103 123501 [18] Ambrosini A and Hamet J-F 2003 Appl. Phys. Lett. 82 727 [19] Lian X, Chen F, Tan X, Wang L, Xuan X, Gao G, Jin S and Wu W 2013 AIP Adv. 3 062133 [20] Frand G, Bohnke O, Lacorre P, Fourquet J, Carr´e A, Eid B, Th´eobald J and Gire A 1995 J. Solid State Chem. 120 157 [21] Singh V R, Zhang L, Rajapitamahuni A K, Devries N and Hong X 2014 J. Appl. Phys. 116 033914 [22] Ha S D, Aydogdu G H and Ramanathan S 2011 J. Appl. Phys. 110 094102 [23] Paul D and Mitra S 1973 Phys. Rev. Lett. 31 1000 [24] Scherwitzl R, Zubko P, Lezama I G, Ono S, Morpurgo A F, Catalan G and Triscone J M 2010 Adv. Mater. 22 5517 [25] Medarde M, Lacorre P, Conder K, Fauth F and Furrer A 1998 Phys. Rev. Lett. 80 2397 [26] Holstein T 1959 Ann. Phys. 8 343 [27] Hong X, Posadas A and Ahn C H 2005 Appl. Phys. Lett. 86 142501 [28] Dulli H, Dowben P A, Liou SH and Plummer EW 2000 Phys. LRev. B 62 R14629 [29] Dulli H, Plummer E W, Dowben P A, Choi J and Liou S-H 2000 Appl. Phys. Lett. 77 570
References [1] Imada M, Fujimori A and Tokura Y 1998 Rev. Mod. Phys. 70 1039 [2] Goodenough J B and Zhou J S 2001 Transport properties Localized to Itinerant Electronic Transition in Perovskite Oxides, Structure and Bonding vol 98, ed J B Goodenough (Berlin: Springer) p 17 [3] Medarde M L 1997 J. Phys.: Condens. Matter 9 1679 [4] Catalan G 2008 Phase Transit. 81 729 [5] Torrance J, Lacorre P, Nazzal A, Ansaldo E and Niedermayer C 1992 Phys. Rev. B 45 8209 [6] Zhou J-S, Goodenough J, Dabrowski B, Klamut P and Bukowski Z 2000 Phys. Rev. B 61 4401 [7] Zhou J-S, Goodenough J and Dabrowski B 2003 Phys. Rev. B 67 020404 [8] Medarde M, Mesot J, Lacorre P, Rosenkranz S, Fischer P and Gobrecht K 1995 Phys. Rev. B 52 9248 [9] Caviglia A et al 2012 Phys. Rev. Lett. 108 136801
7
