Introduction

Correlated transition metal oxides with partially-filled d electrons undergo intriguing metal-insulator electronic phase transition (MIT)1, 2 as a result of electron interactions3. The capability to control those exotic phenomena with external stimuli is a major subject in correlated oxide heterostructure research due to their extreme sensitivity of those materials systems near metal-insulator phase boundary. For example, in these materials, strain and doping can cause small changes in the crystal structures or charge density, and these changes can shift the phase boundaries, and thereby lead to large change in the electrical and optical properties4,5,6. Among the diverse stimuli, oxygen vacancies are inevitable defects in perovskite oxides (ABO3−δ, where δ is the oxygen deficiency) during the synthesis of oxide thin films. The quantity of oxygen vacancies can be stabilized in the range of δ, and thus have a strong influence on the crystal and electronic structures. Because oxygen vacancies have a strong influence on the d-band filling, as well as structural changes, the ability to control the degree of oxygen deficiency may provide the opportunity to tailor the electrical7, optical8 or electrochemical device functionalities9,10,11 of ABO3−δ.

Bulk rare-earth nickelates (RNiO3, where R is a trivalent rare-earth ion) have strongly-correlated unpaired electrons in Ni3+ e g band, and therefore undergo a first-order phase transition to an insulating state at 200 K due to strongly correlated unpaired electrons in Ni3+ e g band under cooling12. Because nickel ions tend to be stabilized as Ni2+ valence state, instead of Ni3+, bulk RNiO3 is prone to oxygen deficiency unless it is synthesized under extremely high partial pressure p(O2) of oxygen7, 13. For this reason, the MIT and optical absorption coefficient of RNiO3 are both sensitively dependent on δ 6, 7, 13,14,15.

Likewise, RNiO3 epitaxial films synthesized using vacuum growth techniques (e.g., pulsed laser deposition16) are also prone to the oxygen deficiency due to the low p(O2) during the process. RNiO3 thin films show inconsistent MIT characteristics, and can have different T MI even at the same lattice mismatch6, 17,18,19,20. For example, NdNiO3 (NNO) films grown on LaAlO3 substrates have T MI ranging from 175 K to 0 K6, 18, 20, 21, which indicates the difficulty of synthesizing RNiO3 that has the desired cation ratio (R/Ni) and the degree of oxygen deficiency (δ). In the previous studies, oxygen deficiency in RNiO3 epitaxial films has been controlled by varying p(O2) during their growth13, 15, 20. However, p(O2) can significantly affect the cation stoichiometry of RNiO3 by increasing the scattering of ablated species in the background gas during pulsed laser deposition; the ablated elemental species experience different scattering conditions in different oxygen pressure as they propagates toward the substrate, influencing film cation stoichiometry, as well as oxygen deficiencies. Consequently, despite several studies that used p(O2) to modulate oxygen deficiency in RNiO3, to the best of our knowledge, no report has been demonstrated on the tuning of T MI in RNiO3 using oxygen deficiency as a single variable while maintaining cation stoichiometry. To investigate the effect of pure oxygen vacancy on T MI in RNiO3 thin films, their effect must be isolated from other stimuli such as cation stoichiometry and strain states.

In this Article, we report systematic control of oxygen deficiency in NNO epitaxial thin films with in-plane tensile strain and investigate the influence of oxygen deficiency on the MIT characteristics. In particular, unlike the previous reports, oxygen deficiency was spontaneously generated in tensile-strained NNO thin films, and was finely adjusted by varying growth temperature T D without modifying other growth parameters; this result allows tuning the degree of oxygen deficiency while maintaining other stimuli. T MI was directly correlated with the unit-cell’s structural expansion, which is a direct measure of oxygen deficiency. Consequently, these results demonstrate the influence of T D on the formation of oxygen vacancies, and also provide insights into the effect of oxygen deficiency on the correlated phase in rare-earth nickelates.

Results

Epitaxial NNO thin films were grown on (001)-oriented (LaAlO3)0.3-(SrAl0.5Ta0.5O3)0.7 (LSAT, a = 3.868 Å) or SrTiO3 (STO, a = 3.905 Å) single-crystal substrate by pulsed laser deposition at 500 ≤ T D ≤ 800 °C under p(O2) = 300 mTorr with a KrF excimer laser fluence of 2.2 J/cm2. Increase in in-plane tensile strain tends to destabilize the Ni3+ valence state, and as a result, increase oxygen deficiency22,23,24; therefore, in-plane tensile strain was strategically applied in pseudocubic NNO films on both cubic substrates (+1.58% for NNO on LSAT and +2.51% for NNO on STO, Fig. 1a) to systematically generate oxygen deficiency. To ensure that NNO layers were fully-strained on both substrates (Fig. 2a,c), the NNO films used were 30 nm thick on LSAT and 15 nm thick on STO.

Figure 1
figure 1

Tensile-strain-induced oxygen deficiency in NNO epitaxial thin films. (a) In-plane lattice constants and lattice mismatch of NNO with LSAT and STO substrates. (b,c) XRD θ–2θ scan of NNO epitaxial thin films on (001) LSAT (b) and (001) STO (c), as a function of growth temperature (T D). Black vertical lines: (002) diffraction peak of substrates (dashed) and pseudocubic NNO epitaxial thin films (solid). (d) Quantitative EDS analysis for cation stoichiometry in NNO thin films grown at various T D; cation atomic ratios of all samples differ by less than the error bar.

Full size image
Figure 2
figure 2

Structural modulation of tensile-strained NNO films as a function of oxygen deficiency. (a) Reciprocal space mapping (RSM) around the (\(\bar{1}03\)) LSAT Bragg peaks for 30-nm-thick NNO thin films grown at various T D. (b) Unit-cell volume of NNO epitaxial thin films grown at various T D on LSAT substrate. (c) RSM around the (\(\bar{1}03\)) STO Bragg peaks for 15-nm-thick NNO thin films grown at various T D. (d) Unit-cell volume of NNO epitaxial thin films grown at various T D on STO substrate.

Full size image

θ–2θ X-ray diffraction (XRD) symmetrical scan of NNO thin films on (001) LSAT (Fig. 1b) and STO (Fig. 1c) substrates grown at 500 ≤ T D ≤ 800 °C show a (002) NNO peak near the (002) substrate peak; this observation indicates that the NNO films grew epitaxially in the c-axis orientation regardless of the growth temperature (T D). In both cases, as T D increased from 500 °C, NNO peak was moved to slightly higher scattering angle until to 600 °C for NNO on LSAT, and until 650 °C on STO, then shifted to the lower scattering angle as T D increased further. The decrease of the scattering angle represents the expansion of the out-of-plane lattice constant of the NNO thin films. To check the possibility of cation non-stoichiometry as an origin of this expansion, all samples grown at different temperature were characterized by energy dispersive spectroscopy (EDS). The cation atomic ratios of all samples were within 2% of 50% (Fig. 1d); this observation indicates that cation non-stoichiometry is unlikely to be the origin of the structural expansion. Instead, oxygen deficiency can be only modulated by T D, because the sticking coefficient and atomic mobility of light-weighted oxygen in the deposited film can be significantly influenced by T D while the ablated elemental species (Nd, Ni) experience similar scattering conditions at the same laser fluence and p(O2). The presence of oxygen vacancies would modify the Ni valence from Ni3+ (r = 0.56 Å) to Ni2+ (r = 0.69 Å) by supplying electrons from the missing oxygen; this reduction process expands the crystal lattice by chemical expansion25,26,27.

To further verify structural modulation as a function of oxygen deficiency (δ), the volume change of the unit-cell was monitored using reciprocal space mapping (RSM) around the \((\bar{1}03)\) Bragg reflection of (001)-oriented NNO thin films grown at different T D on LSAT (Fig. 2a) and STO (Fig. 2c) substrates. qx was the same for all peaks from substrates and NNO films; this consistency means that all NNO thin films are fully-strained on both substrates (Fig. S1). The out-of-plane lattice constants extracted from RSM qz values are consistent with those extracted from XRD θ–2θ scans (Fig. S1); this agreement confirms that the lattice expansion caused by oxygen deficiency is accommodated by constraint-free out-of-plane lattice expansion, whereas in-plane axes are tightly constrained by interaction with the substrate in both cases (Fig. 2a,c).

When the films are strained under tensile strain, the purely strain-induced out-of-plane lattice constant c NNO of the NNO thin film can be accurately calculated as

$${c}_{{\rm{NNO}}}=[(1+\nu ){a}_{0}-2\nu {a}_{{\rm{NNO}}}]/[1-\nu ],$$
(1)

where ν = 0.35 (ref. 28) is Poisson’s ratio, a 0 = 3.806 Å is the unstrained lattice constant, and a NNO is the in-plane lattice constant of NNO (3.868 Å for NNO/LSAT (same as LSAT) or 3.905 Å for NNO/STO (same as STO)). Based on this relationship, estimated c NNO under tensile strain were 3.741 Å on NNO/LSAT and 3.701 Å on NNO/STO. However, the smallest measured c NNO were 3.768 Å on NNO/LSAT grown at 600 °C (i.e., 0.72% greater than calculated) and 3.778 Å on NNO/STO grown at 650 °C (i.e., 2.08% greater than calculated).

The calculations assume pure strain, so the observation that measured c NNO exceeds calculated c NNO indicates that, on both substrates, oxygen vacancies are incorporated into even optimized NNO films with the smallest unit-cell volume. Therefore, the experimental value of c NNO cannot be simply understood by purely strain-induced Poisson-type contraction. Furthermore, the deviation from the pure strain effect (i.e., difference between measured c NNO and calculated c NNO) becomes more pronounced in the case of more tensile-strained NNO films, i.e., NNO on STO. Previous researchers claimed that oxygen vacancies provide a channel that can relieve tensile strain in perovskite oxides grown on substrates with large lattice parameters; this claim suggests that the number of oxygen vacancies generated in NNO films to accommodate strain can increase as tensile strain increases9, 22,23,24, 29. Therefore, the deviation in unit-cell volume from that expected due to the pure strain effect is attributed to the formation of oxygen vacancies, which are increased by tensile strain.

In some perovskite oxides, e.g., CaMnO3, oxygen deficiency formed at high-temperature growth are unstable over time due to oxygen exchange with the environment, which is facilitated by tensile strain; when left exposed to ambient conditions, the films equilibrate and adopt a very small vacancy concentration without capping layer30. However, in our samples, the oxygen vacancies in NNO are not easily diffused out even under ambient conditions: To check time-dependent degree of oxygen deficiency in NNO films, we re-examined the change in the XRD peaks of all NdNiO3−δ samples after 10 months (Fig. S2). Not only did all samples maintain their XRD pattern, but the trend of volcano shapes in out-of-plane scattering angles was also unchanged even after 10 months. Thus, although high degree of oxygen deficiency generated at high temperature are in a metastable state, oxygen exchange with atmosphere appeared to be suppressed in our NdNiO3−δ films because of very limited kinetics of oxygen (or oxygen vacancies) at room temperature. Indeed, previous literature reported that the out-of-plane oxygen vacancy migration barrier of NNO is much larger than that of other perovskite oxide system, such as CaMnO3 31.

To probe how MIT characteristics can be modulated by adjusting δ in NdNiO3−δ films, we measured temperature-dependent in-plane electrical resistivity for the NNO thin films on LSAT (Fig. 3a) and on STO (Fig. 4a) with different T D. All samples showed a sharp MIT, except grown at T D = 750 and 800 °C, which shows degraded MIT characteristic. At room temperature (RT), the resistivity was higher in the optimized NNO on STO (>10−3 Ω cm at T D = 650 °C) than in the optimized NNO on LSAT (<10−3 Ω cm at T D = 600 °C); this difference indirectly confirms that the number of oxygen vacancies created to accommodate tensile strain in films increases as the amount of applied strain increases9, 22,23,24, 29. T MI can be seen clearly in a plot of the first derivative of resistivity with respect to temperature (dR/dT) as a function of growth temperature (T D) (Figs 3b and 4b). Interestingly, the T MI and RT resistivity were quite correlated with unit-cell volume in NNO films grown on LSAT: both quantities were lowest at the minimum of unit-cell volume (Fig. 3c); this result indicates that the electrical properties of this correlated oxide are strongly influenced by the degree of oxygen deficiency (δ). This tendency is also true for NNO/STO samples: the RT resistivity was lowest at the minimum of unit-cell volume, although T MI variation from T D = 500 to 650 °C does not correspond to the unit-cell volume change in NNO/STO.

Figure 3
figure 3

Correlation between MIT characteristics and structural modulation in oxygen-deficient NdNiO3−δ films on LSAT substrates. (a) Temperature (T)-dependent in-plane resistivity for the NNO films on LSAT grown at various T D. (b) The first derivative of resistivity with respect to temperature (dR/dT) as a function of growth temperature (T D) to clarify T MI. (c) Correlation between T MI, room temperature (RT) resistivity and unit-cell volume in oxygen-deficient NdNiO3−δ films. In case of NNO/LSAT grown at 800 °C, temperature dependence of resistivity curves show all insulating behaviors within the measured range (90 to 300 K), which indicates that T MI is outside the plot window (T MI > 300 K).

Full size image
Figure 4
figure 4

Correlation between MIT characteristics and structural modulation in oxygen-deficient NdNiO3−δ films on STO substrates. (a) Temperature (T)-dependent in-plane resistivity for the NNO films on STO grown at various T D. (b) The first derivative of resistivity with respect to temperature (dR/dT) as a function of growth temperature (T D) to clarify T MI. (c) Correlation between T MI, room temperature (RT) resistivity and unit-cell volume in oxygen-deficient NdNiO3−δ films. In case of NNO/STO grown at 750 and 800 °C, temperature dependence of resistivity curves show all insulating behaviors within the measured range (90 to 300 K), which indicates that T MI is outside the plot window (T MI > 300 K). Unlike NNO/LSAT, T MI variation at 500 ≤ T D ≤ 650 °C does not correspond to the unit-cell volume change in NNO/STO.

Full size image

Discussion

The results of this experiment suggest that oxygen vacancies induced in NNO film increases as T D deviates from the optimal T D (600 °C for NNO/LSAT, 650 °C for NNO/STO). Oxygen vacancies increase the unit-cell volume of RNiO3 13, 20, and also increase the RT resistivity of the metallic state6, 7, 13,14,15, 32. The degree of oxygen deficiencies in our NNO films was quantitatively estimated by introducing a simple model in order to analyze the relationship between unit-cell volume and oxygen vacancies in Fig. 5a. According to the empirical model, the unit-cell volume V of perovskite NdNiO3−δ can be expressed as

$${\rm{V}}={A}^{3}{({r}_{B}+{r}_{anion})}^{3}$$
(2)

where r B and r anion are the ionic radii of the B-site cation (nickel site) and anion, respectively, and A is a constant close to 2.

Figure 5
figure 5

Quantitative estimation of oxygen deficiency (δ) and octahedral rotation in oxygen-deficient NdNiO3−δ films. (a) Estimation of the oxygen deficiency (δ) in NdNiO3−δ films using quantitative model. (b) Thermodynamic and kinetic model of inverse temperature (1/T)-dependent log δ for NNO/LSAT and NNO/STO to explain quantitative estimation of δ. (c) Out-of-plane Ni-O-Ni bond angles (θNi-O-Ni) with respect to T D. The Ni-O-Ni bond angle is calculated from θNi-O-Ni = 2 arcsin(c/2dNi-O), where Ni-O bond length (dNi-O) was assumed to be equal to that of bulk NNO (1.942 Å). Note that there exists drastic change of Ni-O-Ni bond angle in T D = 500 to 650 °C of NNO/STO.

Full size image

By assuming oxygen-deficient NdNiO3−δ due to the unchanged cation ratio (Nd/Ni) characterized by EDS (Fig. 1d), the effective cation and anion radii could be defined as

$${r}_{B}=(1-2\delta ){r}_{N{i}^{3+}}+(2\delta ){r}_{N{i}^{2+}}$$
(3)
$${r}_{anion}=(\frac{3-\delta }{3}){r}_{O}+(\frac{\delta }{3}){r}_{V}$$
(4)

where \({r}_{N{i}^{3+}}\), \({r}_{N{i}^{2+}}\), r O , and r V are the ionic radii of Ni3+, Ni2+, oxygen, and oxygen vacancy, respectively27, 33. Consequently, the degree of oxygen deficiency (δ) (Fig. 5a) was roughly estimated to be ~0.05 and ~0.09 for NNO/LSAT grown at 600 °C and NNO/STO grown at 650 °C, respectively, from the measured unit-cell volume (Fig. 2b,d) using this empirical model. This consistent trend on oxygen deficiency in NNO films grown on two different substrates, i.e., the volcano shape as a function of T D , indicates that there should be different mechanisms to generate oxygen vacancies at high temperature (600 °C~800 °C for NNO/LSAT and 650 °C~800 °C for NNO/STO) and at low temperature (500 °C~600 °C for NNO/LSAT and 500 °C~650 °C for NNO/STO).

In case of high-temperature regime, a thermodynamic model of oxygen deficiency can be utilized to explain the mechanism based on given experimental evidences. The formation energy of oxygen vacancies (ΔH) decreases with the tensile strain23 and the degree of oxygen deficiencies (δ) can be expressed as follows:

$$\delta \propto exp(-\frac{{\rm{\Delta }}G}{kT})$$
(5)

where ΔG = ΔH − TΔS (ΔH is the enthalpy of oxygen vacancy formation). If we ignore small entropy effects (ΔG = ΔH), the inverse temperature (1/T)-dependent log δ for NNO/LSAT and NNO/STO are estimated as shown in Fig. 5b. Moreover, as tensile strain increases, i.e., NNO/STO, the formation energy of oxygen vacancies decreases23; this trend can affect oxygen vacancies under various growth temperature (T D), and the change of oxygen deficiency causes systematic variation in T MI as a function of T D. While this thermodynamic model provides a plausible explanation for the increase of oxygen deficiency in the high temperature regime (black and red solid lines in Fig. 5b), it does not account for the increase of oxygen deficiency in the low temperature regime.

In case of low-temperature regime, the insufficient mobility of oxygen in surface migration during the growth might lead to more oxygen-deficient NNO films on both substrates at lower temperature: Since oxygen atoms are unlikely to be supplied sufficiently at low temperature, the amount of oxygen incorporated into the thin film is kinetically limited, resulting in the increase of oxygen deficiency (blue solid lines in Fig. 5b) with decreasing the growth temperature. Considering higher concentration of oxygen vacancies in NNO/STO at low temperature, surface migration of oxygen was more suppressed in NNO/STO than in NNO/LSAT. Indeed, it was previously reported that the oxygen vacancy migration barrier of NNO/STO is larger than that of NNO/LSAT due to the tensile strain-induced suppression of oxygen vacancy migration to the surface31.

Despite the close relationship between unit-cell volume and T MI above, the modulation of T MI by unit-cell volume was inconsistent at 500 ≤ T D ≤ 650 °C in highly tensile-strained NNO films on STO substrates. Beyond the explanation using unit-cell volume caused by oxygen vacancies, this can be explained by the influence of the octahedron distortion by tensile strain. In case of LaNiO3, in-plane tensile strain weakly modifies the out-of-plane Ni-O distance d Ni-O, but the out-of-plane Ni-O-Ni bond angle θ Ni-O-Ni is highly sensitive to the strain34. Since our samples are all fully-strained state (RSM data, see Figs 2 and S1), it is reasonable to assume that in-plane Ni-O-Ni bond angle is fixed. Because d Ni-O (out-of-plane) is almost insensitive to the strain, θ Ni-O-Ni (out-of-plane) can be estimated from the relationship θ Ni-O-Ni = 2arcsin(c/(2dNi-O)) based on an assumption that d Ni-O (out-of-plane) is equal to that of bulk NNO (1.942 Å). By inserting the observed out-of-plane lattice parameters (c), Ni-O-Ni bond angles (θ Ni-O-Ni) was estimated as shown in Fig. 5c. In NNO/STO, Ni-O-Ni bond angles drastically increased up to 156° as T D deviated from the optimal T D~650 °C down to 500 °C; this trend is unlike that of NNO on LSAT. Because Ni-O-Ni bond angle determines the degree of overlap between Ni 3d and O 2p orbitals, orbital overlap increases and, therefore T MI decreases despite the increase in unit-cell volume. Therefore, in highly tensile-strained NNO films on STO, the effect of octahedron distortion on T MI is more influential than that of oxygen vacancies on T MI in the range of T D = 500 to 650 °C.

Conclusion

In summary, we report systematic control of metal-insulator transition (MIT) in tensile-strained NdNiO3−δ (NNO) epitaxial thin films by varying the degree of oxygen deficiency δ naturally by applying in-plane tensile strain, and by adjusting δ at growth temperature T D through thermodynamic and/or kinetic control of defect formation. Unit-cell volume and MIT temperature T MI both increased as δ increased. Our study provides useful approaches to finely modulate δ in correlated oxides, and provides insights on functional defects in rare-earth nickelate system.

Experimental Methods

Epitaxial NdNiO3 (NNO) thin films were grown on (001)-oriented (LaAlO3)0.3-(SrAl0.5Ta0.5O3)0.7 (LSAT) and SrTiO3 (STO) single-crystal substrates (CrysTec Gmbh, Germany) by pulsed laser deposition with a KrF excimer laser (λ = 248 nm, Coherent Compex Pro 102 F). A polycrystalline stoichiometric NdNiO3 ceramic target was used to fabricate the thin film. During deposition, the laser pulse frequency was 10 Hz, laser energy was 2.2 J/cm2, and p(O2) was 300 mTorr. The growth ( = substrate) temperature T D was varied from 500 to 800 °C. After each deposition, the sample was cooled to room temperature for 30 min under p(O2) = 300 mTorr without any in-situ post-annealing.

Structural characteristics of the NNO films were measured using θ–2θ scans and reciprocal space mapping (RSM) with a high-resolution X-ray diffractometer (XRD, Bruker D8 Discover X-ray diffractometer) with Cu Kα1 radiation (λ = 1.5406 Å). Cation stoichiometry of the film was determined in high vacuum by using a field-emission scanning electron microscope (FE-SEM, JEOL JSM-7401F) coupled with an energy dispersive spectroscopy system (EDS, Oxford INCA). In-plane electrical transport was measured using the Van der Pauw method at 90 ≤ T ≤ 300 K during heating.

To check time-dependent change of oxygen deficiency in NNO films, we re-examined the change in the XRD peaks of all NdNiO3−δ samples used for the manuscript after 10 months. All samples were stored in the desiccator at room temperature for 10 months without any capping layer.