Role of pO₂ and film microstructure on the memristive properties of La₂NiO_4+δ/LaNiO_3−δ bilayers

Abstract

LaNiO₃/La₂NiO₄ bilayers deposited at varying pO₂ conditions resulted in remarkable differences in film microstructure and cell parameters, directly affecting the electrical behaviour of Pt/LaNiO₃/La₂NiO₄/Pt devices. The devices deposited at low pO₂ showed the largest memristance. We propose this is due to the formation of a p-type Schottky contact between LaNiO₃ and La₂NiO₄, where the extent of its carrier depletion width can be modulated by the electric-field induced drift of interstitial oxygen ions acting as mobile acceptor dopants in La₂NiO₄.

---

Introduction

Mixed ionic electronic conducting materials (MIEC) based on transition metal oxides are key elements in many different applications, such as ceramic gas separation membranes, gas sensing or as cathode materials for solid oxide fuel cells.^1,2 Recently, a new application has come into play for the use as MIEC oxides in resistive switching devices, more particularly, in those based on the valence change mechanism (VCMs). The combined electronic conductivity and oxide ion mobility in these materials are key aspects in the functioning of these memristive devices. The drift of oxygen ions induced by the application of an external electric field triggers internal redox reactions, locally changing the structure and composition of the material and therefore also its electrical properties. Typically, VCMs are based on the formation of conducting filaments across an insulating oxide such as SrTiO₃, TiO₂, Ta₂O₅ or HfO₂. This insulator is sandwiched between two metal electrodes; the conducting filament is produced by the ion migration of mobile donors such as oxygen vacancies () or cation interstitials under high electric fields and can be considered as a soft dielectric breakdown.³ In addition, interface-type switching has been demonstrated for some MIEC oxides, such as Pr_1−xCa_xMnO₃ (PCMO), where the mobile defects are oxygen vacancies.⁴ However, very few studies are devoted to other types of MIEC oxides where the main mobile species are oxygen interstitials (), as is the case in La₂NiO₄.^5,6 The additional interstitial oxygen ions (usually expressed as δ) can be stabilized in the form of over-stoichiometric La₂NiO_4+δ, partially oxidizing Ni²⁺ into Ni³⁺ and resulting into a larger concentration of electronic (hole) carriers. The p-type semiconducting properties and high oxygen diffusion of La₂NiO_4+δ make it an interesting material for the fabrication of VCM-based resistive switching devices.

La₂NiO₄ has a layered Ruddelsden–Popper-type structure consisting of alternate La–Ni–O perovskite and La–O rock salt layers along the c-axis of the structure.⁷ The oxygen diffusion has been proven to be highly anisotropic in this layered structure.^8–11 Indeed, the interstitial oxygen anions located in the rock salt layers diffuse preferentially along the a–b plane of the structure through an interstitialcy mechanism.⁸ Similarly, the electronic conductivity takes place preferentially along the perovskite layers, leading to about three orders of magnitude higher electrical conductivity along the a–b plane compared to the c-axis direction at room temperature.^12,13 In this sense the control over the microstructure of the material, and more particularly the preferential orientation of the crystal domains, is essential to understand the memristive properties of this material when used in VCM-type devices.

For this study LaNiO₃, a perovskite-structure oxide material showing metallic conductivity,¹⁴ has been selected as template electrode due its chemical affinity and crystallographic match with La₂NiO₄. To the best of our knowledge the electrical and memristive behavior of the LaNiO₃/La₂NiO₄ bilayer interface remains yet to be explored.

This work aims at studying the microstructure and electrical properties of LaNiO₃/LaNiO₄ epitaxial bilayers deposited under different oxygen partial pressure (pO₂) during film growth in order to tune the oxygen content and therefore the contact properties and memristive behaviour of these complex-oxide heterojunctions.

This study demonstrates that not only the oxygen stoichiometry of the films, but also the overall crystal domain orientations are largely affected by the deposition conditions, leading to substantial differences in the memristive properties of these MIEC-based devices. The results presented here show that oxygen drift across the material, as well as oxygen transport across the interface, are key aspects of the memristive response, and could have implications both as memristive synapses for neuromorphic computing in low-energy consumption electronics, as well as for a wider type of electrochemical devices making use of this type of materials, such as solid oxide fuel cells and electrolysers.

Experimental methodology

LaNiO₃/La₂NiO₄ epitaxial bilayers were grown by Pulsed laser deposition by using a pulsed UV excimer laser (Lambda Physik Compex Pro 201, λ = 248 nm) in a PLD-Workstation from Surface-Tec, GmbH, and corresponding ceramic targets prepared by solid-state reaction. SrTiO₃(001) single crystals (10 × 10 mm², from Crystec, GmbH) were used as substrates. LaNiO₃ was chosen as bottom electrode material, providing a crystal matching for the subsequent epitaxial growth of La₂NiO₄. The deposition temperatures were 600 °C and 700 °C for LaNiO₃ and La₂NiO₄, respectively, as previously determined for optimal crystal quality of single layers deposited on SrTiO₃ substrates. The values of the oxygen pressure during deposition were 12 mTorr and 70 mTorr. The pO₂ range was limited by the PLD system as it can only operate in high vacuum conditions. Other conditions for the sequential deposition of LaNiO₃ and La₂NiO₄ were the same: target-to-substrate distance of 55 mm, and laser fluence of 1.2 J cm⁻² (spot size 7.3 mm²), and pulse repetition rate of 2 Hz. Given the different ablation rates of the two targets and the different pO₂ conditions the number of laser pulses was adjusted to obtain an approximate thickness of 30 nm for both LaNiO₃ and La₂NiO₄ layers. Three sets of bilayers were prepared combining the growth of LaNiO₃ and subsequent growth of La₂NiO₄ at either “High” pO₂ (70 mTorr) or “Low” pO₂ (12 mTorr), resulting in three bilayers with different pressure combinations, namely H/L, L/L and L/H samples. The attempted growth of a H/H bilayer was discarded from the study as it showed insufficient crystal quality.

The deposition protocols (shown in Fig. S1†) were designed to avoid as much as possible some parasitic effects. For instance, the difference in the optimal growth temperature between LaNiO₃ and La₂NiO₄ (600 °C and 700 °C, respectively) made it very important to minimize the time interval between subsequent depositions, since LaNiO₃ is prone to vary its oxygen stoichiometry and eventually decompose at high temperatures and low pO₂, when heating up from 600 °C to 700 °C. Therefore, this was an important parameter to control. For the bilayer grown under constant low pO₂ (L/L), La₂NiO₄ was deposited as soon as the sample reached 700 °C after a heating ramp of 10 °C min⁻¹ (10 min). For the growth of the H/L and L/H bilayers, the pO₂ was only changed after having heated up the sample from 600 °C to 700 °C, and then waiting 1 min for the pO₂ to stabilize before growing the second layer in different pO₂. The samples were all cooled down under the pO₂ used for the growth of the top La₂NiO₄ layer. For a proper comparison single layers of LaNiO₃ and La₂NiO₄ were also grown at the two different pO₂ conditions.

The XRD measurements (2θ/ω and ϕ scans) were carried out using a Malvern-Panalytical X'Pert-PRO MRD system with 4-angle goniometer, and Cu Kα1 radiation (λ = 1.5406 Å, 2xGe(220) monochromator).

Microstructural characterization of the film was carried out by high-resolution transmission electron microscopy (HRTEM) using a FEI Tecnai G2 F20 microscope with field emission gun working at 200 kV. Cross-section samples were prepared by FIB on regions with top Pt electrodes (after the electrical measurements).

During film deposition one part of the sample was covered in situ by a metal mask after the growth of the LaNiO₃ preventing the growth of the subsequent La₂NiO₄ layer. In this way, after deposition, there is one part of the sample which has exposed LaNiO₃ layer. For the electrical characterisation of the bilayers an array of 50–200 μm diameter Pt electrodes were deposited by e-beam evaporation through a hard mask. Some electrodes were located directly on the exposed part of the LaNiO₃ bottom electrode layer, while the rest were on the La₂NiO₄ top layer. Electrodes of 200 μm in diameter were selected to form the devices, with a distance between bottom and top electrodes between 1400 μm and 1800 μm depending on the sample. Electrical measurements were carried out with a Keithley 2604B SourceMeter in a dark box, using two tips to apply a voltage and simultaneously measure the resulting current. The voltage was applied to the Pt/LaNiO₃ electrode while the La₂NiO₄/Pt electrode remained grounded. First, the initial resistance state (IRS) of the device was measured at low voltage. Then, the device was cycled by triangular voltage sweeps (0 V → +V_max → −V_max → 0 V), while progressively increasing |V_max| by 1 V every five cycles, in order to trigger the memristive behavior, until |V_max| = 15 V was reached. The sweep rate was fixed to 10 V s⁻¹. The current was measured after each cycle (at a voltage of 500 mV) to determine the resistance state. No electroforming step was required to trigger the memristance, and thus these devices can be considered forming-free.

Results and discussion

Sample preparation and characterisation

Structural characterization. The XRD measurements performed on the bilayers indicate a high degree of orientation of both LaNiO₃ and La₂NiO₄ layers. Only 00L peaks of both compounds were observed in the 2θ scans as depicted in Fig. 1a. The oscillations in the background of the XRD scans confirm the high crystal quality of the layers. From the position of the peaks it is possible to determine the corresponding out-of-plane parameters, as shown in Table 1. As far as the LaNiO₃ layers are concerned, the c-parameter is 3.83 Å for the layer deposited at high pO₂ of 70 mTorr (H/L sample), while for the LaNiO₃ deposited at low pO₂ we obtain c = 3.86 and 3.87 Å, for L/L and L/H samples, respectively. It seems that the c-axis parameter of LaNiO₃ layer mainly depends on the pO₂ during its growth and less so on the pO₂ present during the subsequent growth of La₂NiO₄ (when grown at a different pO₂).

Fig. 1 (a) High resolution X-ray diffraction 2θ/ω scans in the vicinity of 002 SrTiO₃ substrate reflection of the three bilayer samples combination of the growth at low and high pO₂. The peaks labelled as 006 La₂NiO₄ and 002 LaNiO₃ correspond to c-axis oriented domains. Note that the weak 006 La₂NiO₄ peak in the L/H sample is due to the dominant a-axis orientation of this layer. (The peak with * corresponds to a double-crystal diffraction and is an indication of high crystal coherence of the whole bilayer) (b) ϕ-scans of the 101 La₂NiO₄ reflections indicating either dominant out-of-plane c-orientation or a-orientation of the La₂NiO₄ film, for the extreme H/L and L/H samples, respectively.

Table 1 Values of the c parameter of LaNiO₃ and La₂NiO₄ as well as observed orientation of La₂NiO₄ domains depending on the bilayer deposition conditions (varying the oxygen partial pressure)

Sample	pO2 (mTorr)	LaNiO3c-parameter (Å)	La2NiO4c-parameter (Å)	La2NiO4 orientation
Bilayer (LaNiO3/La2NiO4)	70/10 (H/L)	3.832	12.664	Pure c⊥
	10/10 (L/L)	3.862	12.637	Mixed c⊥ and a⊥
	10/70 (L/H)	3.870	—	Pure a⊥
Single layer	10 (L)	3.877	12.632	Pure c⊥
	70 (H)	3.830	12.686	Pure c⊥

---

The LaNiO₃c-parameter values are consistent with the results obtained for the single layers, as shown also in Table 1, (the corresponding XRD patterns are in the supplementary part: Fig. S2 and S3†) in the sense that the lattice parameters measured for films grown in low pO₂ are expanded (about 1%) with respect to those obtained at high pO₂. This can be explained by a larger oxygen vacancy concentration, as expected for LaNiO_3−δ. The crystal quality of the LaNiO₃ layer is preserved at both pO₂.

However, when comparing the cell parameter values of LaNiO₃ in the single layer deposited at low pO₂ with those obtained in the bilayers at the same conditions, a slight reduction of the cell parameter between 0.2–0.4% is observed. This reduction has been observed regardless of the pO₂ used during the growth of the subsequent La₂NiO₄ layer. This means that LaNiO₃ deposited at low pO₂ oxidizes during the deposition of La₂NiO₄ (either at low or high pO₂), while it maintains a constant oxygen content when deposited at high pO₂. This is consistent with the observed oxygen transport previously reported in LaNiO₃/La₂NiO₄ bilayers at elevated temperatures,¹⁵ although, a possible effect of the impinging energy of the growing species to scavenge oxygen vacancies from the underlying LaNiO₃ cannot be discarded.^16,17 Therefore, the LaNiO₃ layers deposited at low pO₂ (L/L and L/H bilayers) could potentially develop an oxygen vacancy gradient at the interface with La₂NiO₄. A different situation occurs for the La₂NiO₄ layers. At low pO₂ a clear c-axis orientation is observed for La₂NiO₄ (for the H/L and L/L bilayers), with an intense 006 reflection for both films. The c-axis parameter for the L/L samples is about 12.64 Å, similar to that obtained for La₂NiO₄ single layers directly grown on the SrTiO₃ substrate (Fig. S3† and Table 1) for the same pO₂ conditions.

This could be another indication of a certain oxygen ion transport happening between the LaNiO₃ and La₂NiO₄ layers. However, in this case it could not be ruled out that part of the changes in La₂NiO₄ cell parameter are due to a different strain when growing on LaNiO₃ instead of directly on SrTiO₃ substrates.

When growing La₂NiO₄ at high pO₂, in the L/H bilayer, no 006 reflection is observed. Therefore, this La₂NiO₄ film does not present domains with c-axis orientation, contrary to single La₂NiO₄ films which showed c-axis orientation when directly grown on SrTiO₃ substrates (Fig. S2†). This type of scan does not preclude the existence of other orientations of La₂NiO₄, i.e. a-axis orientation, which would overlap with the intense substrate peaks. In order to clarify whether La₂NiO₄ grows with a different orientation Φ-scans of asymmetric reflections, such as 101 La₂NiO₄, were performed at a sample stage tilt angle χ = 16.7°, which is the position expected for the a-oriented domains, as depicted in Fig. 1b. Indeed, the scan over the L/H sample shows the presence of broad 101 peaks with 4-fold symmetry. This indicates the existence of a-axis orientation with perfect in-plane alignment with the substrate [100]/[010] directions. For comparison, the corresponding Φ-scan of the 101 reflection for the c-axis oriented domains in the H/L sample, in this case at χ = 72.1°, was also measured indicating again a four-fold symmetry with a large degree of in-plane arrangement. Therefore, the growth of La₂NiO₄ at high pO₂ promotes the growth of a-axis domains, while c-axis orientation is obtained at low pO₂.

This situation was different to the direct growth of La₂NiO₄ on SrTiO₃(001) substrates where a high degree of c-axis orientation is achieved at both pO₂ values (Fig. S3†). Therefore, it is likely that the exposure of underlying LaNiO₃ to the high pO₂ before the subsequent growth of La₂NiO₄ affects the LaNiO₃ topmost surface to inhibit the perfect cube-on-cube arrangement of perovskite blocks of c-axis oriented La₂NiO₄ on LaNiO₃. In agreement with the La–Ni–O thermodynamic phase diagram,² it is possible that some secondary phases, resulting from LaNiO₃ decomposition into La₂NiO₄ + NiO are also present at the interface, which could alter the subsequent growth of La₂NiO₄.

Bilayer microstructure. In order to address the microstructure of the bilayers cross-sectional samples were prepared from the same set of bilayers.

Fig. 2 shows the cross-sections oriented along [100]-zone axis at the interface between bottom LaNiO₃ and top La₂NiO₄ layers for the three different deposition conditions: H/L (top), L/L (middle), and L/H (bottom). The left column shows HRTEM images of the bilayers where clear differences in microstructure are revealed in top and bottom layers. All the bilayers show highly oriented LaNiO₃ with a large density of vertical planar defects, which very likely correspond to Ruddlesden–Popper type defects, a sort of antiphase boundary defect with a displacement of half primitive perovskite cell along the 〈111〉 direction, as previously shown in.^18,19

Fig. 2 HRTEM images (left column) of cross sections of the different bilayers from top to bottom: H/L, L/L and L/H LaNiO₃/La₂NiO₄ bilayers, as well as corresponding FFT patterns of the La₂NiO₄ layer (second column). The third column shows colour-filtered images for the selected spots in the FFT corresponding to cubic perovskite (red), c_⊥-Ruddlesden–Popper (RP) (green), a_⊥-RP (blue). The column on the right shows a schematic of the film's microstructure and domain orientation (the c-axis is indicated with blue arrows) for the three types of samples, showing the formation of boundaries between a_⊥- and c_⊥-domains: perpendicular (marked in red colour) and parallel (green) to the vertical axis. It also shows the high concentration or low concentration of oxygen interstitials () in the initial state depending on the deposition conditions.

On the other hand the nano-structural differences of the La₂NiO₄ films of the bilayers are more pronounced. The H/L sample shows a sharp interface with highly c-axis oriented La₂NiO₄. The corresponding FFT of the top La₂NiO₄ layers are shown in the second column in Fig. 2. For the H/L sample the pattern is consistent with a pure c_⊥-orientation, and a- or b- axis parallel to the image, aligned with in-plane main [100]/[010] substrate directions. For the L/L sample, the first layers at the interface with LaNiO₃ start growing with c_⊥-orientation, but after 5–8 nm additional domains with a different orientation start to nucleate, resulting in a FFT which consists of a mixture of c_⊥- and a_⊥-orientations. The existence of a c_⊥-component was already observed in the XRD pattern of this film. However, the existence of a mixed a_⊥-orientation was not detected because of the overlap with the substrate peaks. For the L/H bilayer the majority of the La₂NiO₄ film seems to be formed by narrow columns of a_⊥-oriented domains, some of them with the c-axis parallel to the image (c_//). This orientation is identified at atomic scale by the arrangement of bright vertical atomic planes corresponding to the a/b plane of the Ruddlesden–Popper structure, while other domains do not show such contrast, presumably because in this case the c-axis lays perpendicular to the image. The corresponding FFT is consistent with pure a_⊥-orientation, as already shown by the ϕ-scans in Fig. 1b. The third column of Fig. 2 corresponds to colour-filtered images obtained by selecting the spots in the FFT for either LaNiO₃ phase (red), c_⊥-La₂NiO₄ (green) and a_⊥ (c_//)-La₂NiO₄ (blue). The H/L sample shows full c_⊥-La₂NiO₄ (full green colour), while the L/L sample shows mixed c_⊥- and a_⊥- La₂NiO₄ (green with blue patches). The L/H sample shows a columnar arrangement with an intimate mixture of blue and red stripes, with nanometer size domains, corresponding to a_⊥-La₂NiO₄. The red stripes correspond in this case to a_⊥-La₂NiO₄ domains with c-axis pointing perpendicular to the image, which are not distinguished from a standard perovskite structure (red colour), but obviously do not correspond to LaNiO₃ although they show the same FFT pattern. The right column in Fig. 2 summarizes these observations in the form of a schematic of the samples' microstructures. The regions where La₂NiO₄ domains with different orientation meet together form grain boundaries. In the pure c_⊥-axis oriented film (H/L) there are not grain boundaries in the La₂NiO₄ layer, whereas in the mixed a_⊥/c_⊥-oriented La₂NiO₄ layer (in the L/L sample) there are vertical and horizontal (001)/(100) boundaries between a_⊥ and c_⊥ domains. In the pure a_⊥ La₂NiO₄ layer (in the L/H sample) the in-plane rotation by 90° of their corresponding c-axis defines also vertical (001)/(100) domain boundaries. The vertical and horizontal domains have been marked with green and red colours, respectively, in Fig. 2 (right column).

To conclude this section, different growth conditions (pO₂) induce remarkable differences not only in the oxygen concentration profiles across the bilayer, with possible gradients at the LaNiO_3−δ/La₂NiO_4+δ interface, but also in the bilayer nano-structure, with the appearance of a particular type of (001)/(100) grain boundaries within La₂NiO₄. These differences will surely affect the electrical properties of these bilayers and could ultimately dictate the memristive behaviour of Pt/LaNiO₃/La₂NiO₄/Pt devices, as will be discussed in the following section.

Role of the pO₂ on the electrical properties of Pt/LaNiO₃/La₂NiO₄/Pt memristive devices

Fig. 3 presents an overview of the electrical response of the three bilayers using circular Pt contacts with a diameter of 200 μm. The voltage was applied to the Pt electrodes directly on top of LaNiO₃ bottom layer (bright part of the image in Fig. 3a), while the Pt contacts on top of La₂NiO₄ (dark side of Fig. 3a) were grounded, as shown in the schematic of Fig. 3b. The shape of each voltage sweep cycle versus time is depicted in Fig. 3c. The I(V) measurements were reproducible when cycling within a fixed voltage range.

Fig. 3 (a) Optical image of the top surface of the Pt/LaNO₃/La₂NiO₄/Pt sample showing the circular Pt electrodes. (b) Schematic of the measuring configuration from grounded top Pt electrode on the La₂NiO₄ layer to biased Pt electrode on the LaNiO₃ bottom layer. (c) Applied triangular voltage sweeps. (d)–(f) Evolution of the I(V) characteristics of the H/L, L/L and L/H bilayers, respectively, when increasing the maximum voltage from left to right (±1 V, ±5 V, ±10 V and ±15 V). The L/L sample shows clear analog-type switching characteristics. (Arrows indicate the direction of the voltage sweep).

The I(V) curves measured for the three samples are shown in Fig. 3d–f, corresponding to every fifth cycle at the ±1 V, ±5 V, ±10 V and ±15 V sweeps. Interface-type memristive devices typically show rectifying I–V characteristics in their initial resistance state (IRS). This nonlinear behaviour is usually explained by the presence of a Schottky-like barrier at the metal/oxide interface necessary for the emergence of the memristive behavior.⁵ Both the H/L and the L/L samples show nonlinear I(V) characteristics in their IRS (I–V sweeps between ±1 V). Since Pt forms an ohmic contact with both LaNiO₃ and La₂NiO₄ (as shown in Fig. S4†) the nonlinearity can only be explained by the presence of a Schottky-like barrier at the LaNiO₃/La₂NiO₄ complex-oxide interface. The increased nonlinearity measured for the L/L sample could signify the presence of a larger Schottky barrier. On the other hand the L/H sample shows almost linear I(V) characteristics between ±1 V and should therefore present the smallest LaNiO₃/La₂NiO₄ contact resistance. As will be later explained, this difference is likely due to the high electronic conductivity expected for the La₂NiO₄ top layer of the L/H sample due to its a_⊥ orientation, high oxygen overstoichiometry (δ) and concomitant high concentration of electronic hole carriers.

Other interesting differences can be noticed between the samples when increasing V_max gradually from ±1 V to ±15 V. Both the HRS and the LRS for all three samples present some form of hysteresis which is most pronounced at V_max = ±10 V. It should be noticed that the small hysteretic behavior disappears when V_max is increased to ±15 V for the H/L and L/H samples, as clearly shown in the resistance versus voltage plots in Fig. S5.† The disappearing of the hysteresis seems to be irreversible as the subsequent application of a lower voltage does not restore the original opening in two resistance states (not shown here). On the other hand the L/L sample progressively shows an increasingly rectifying behaviour with two very distinct resistance states. The Low Resistance State (LRS – higher current branch), shows little rectification within the ±2 V sweeping range, i.e. before the devices switches to the High Resistance State (HRS). Conversely, the HRS shows highly rectifying I(V) characteristics with an increased threshold voltage when compared to the LRS. This hysteretic behavior and the distinct LRS and HRS states are well maintained even for the largest voltage sweep range (±15 V). The evolution of the HRS and LRS values (measured at 500 mV immediately after each sweep) as well as the corresponding HRS/LRS ratios are shown in Fig. S6† for the three bilayer samples. For the H/L and L/H samples both the HRS and LRS show very low values (<10³ Ω) below V_max = ±7 V. For larger voltage sweep ranges, in the H/L sample both states increase to values up to 10⁴ Ω, while for the L/H sample they remain relatively low. However, in both cases the HRS/LRS ratios stay between 1 and 3 for the whole measured range. Interestingly the L/L sample shows a completely different behaviour where both the HRS and LRS values continuously increase all the way from the lowest values of V_max = ±1 V until V_max = ±15 V. At V_max = ±7 V, the LRS reaches a stable value of around 10⁴ Ω, while the corresponding HRS continues increasing above 10⁵ Ω. Therefore, the corresponding HRS/LRS ratio shows a pronounced increase from 1–2 below V_max = ±7 V up to 20–30 at V_max = ±15 V.

Focusing on the LaNiO₃/La₂NiO₄ interface we interpret the resistive switching behavior based on the mixed ionic-electronic conducting character of La₂NiO_4+δ, and its anisotropy, and we present two possible scenarios or mechanisms which can explain the observed memristance. The layered Ruddlesden–Popper structure of La₂NiO_4+δ is expected to induce a high anisotropy both in the electronic transport and oxygen diffusion properties.^10–13 At room temperature the resistivity along the c-axis was measured to up to three orders of magnitude larger than that within the basal plane,^12,13 mainly attributed to differences in mobility due to the variation in interplanar separation within the crystallographic structure. In addition, as a p-type conductor, the resistivity highly depends on the oxygen content, and more particularly to the interstitial oxygen content, which is directly related to the concentration of electronic hole carriers. We recently showed that for epitaxial La₂NiO_4+δ thin films prepared by Pulsed Injection Metal Organic Chemical Vapor Deposition, we could easily change the oxygen interstitial concentration by annealing in different gas atmospheres (from H₂/Ar to O₂), leading to changes in the c-axis out of plane cell parameter from 12.62 to 12.69 Å.⁹ The resistivity values of the different films showed three orders of magnitude increase in resistivity (from 5.3 × 10⁻³ to 5.7 Ω cm) when decreasing the oxygen content (smaller cell parameter). Thus, in the case of the L/H sample studied here a higher electronic conductivity is expected across the top La₂NiO_4+δ film due to: (i) an increased oxygen content (compared to the other two samples) leading to a larger hole carrier concentration, and (ii) its a_⊥-orientation (opposite to the other samples), and thus along the fastest electronic transport path in this anisotropic material. This could significantly reduce the carrier depletion width on the La₂NiO_4+δ side of the interface Schottky contact, therefore reducing the overall contact resistance. Consequently, this prevents the building up of a memristive behavior in this bilayer (small HRS/LRS ratio at V_max = ±10 V that then disappears at higher voltages).

In the following we focus only on the H/L and L/L samples which, do present a Schottky-like characteristics and present the first potential mechanism. Besides the drift of highly mobile electronic charges, typically forming a depletion zone at the interface, in a material such as La₂NiO_4+δ the application of an external electrical field above a certain voltage facilitates the drift of charged oxygen defects (negatively-charged interstitial oxygen, , in this case).¹⁰ Then, the accumulation of oxygen interstitials at the interface with LaNiO₃ may produce a change in the Schottky barrier as well as in the material properties in this location, which result in a change in the contact resistance and ultimately in the overall resistance of the device. Indeed, in the L/L sample, that shows the largest memristive behavior, a strong nonlinearity was observed at low voltage, probably caused by this potential barrier. The resistance-change mechanism in Pt/LaNiO₃/La₂NiO₄/Pt could therefore be explained as follows: initially, the device is in a HRS. When applying a high enough positive voltage (on the LaNiO₃ side) the negatively-charged oxygen ions drift towards the LaNiO₃/La₂NiO₄ interface, locally changing the carrier density (the mobile interstitial oxygen ions, , act as hole dopants in La₂NiO₄) and reducing the charge depletion width in this location. The device now finds itself in its LRS as the flow of current is now facilitated through a thinner potential barrier (additional conduction mechanisms can be present, i.e. tunneling). The opposite occurs when applying a high enough negative voltage. Oxygen interstitials in La₂NiO₄ drift away from the LaNiO₃/La₂NiO₄ interface reestablishing the barrier width, and therefore the HRS.

Although, in the previous explanation, the oxygen interstitial accumulation region has been limited to the La₂NiO₄ part at the interface with LaNiO₃, a possible exchange of oxygen ions between the LaNiO_3−δ and the La₂NiO_4+δ layers, may result in an extension of the depletion region through an internal redox mechanism, also modifying the contact resistance. This corresponds to the second mechanism proposed. This recombination at the interface of LaNiO_3−δ and the La₂NiO_4+δ layers has been observed at 600 °C by Moreno et al.,¹⁵ and it is likely that the exchange of oxygen ions between the two materials can occur at room temperature, assisted by the electric field and by Joule heating. Finally, it is possible that the two mechanisms evoked previously are cohabitating, both playing a role in the measured change in resistance of the device.

While the hysteretic behavior is much more pronounced for the L/L bilayer, the comparison of the film's microstructure between H/L and the L/L samples, as observed by HRTEM and depicted in the schematic in Fig. 2, does not reveal any relevant difference at the LaNiO₃/La₂NiO₄ interface itself. Both bilayers show sharp interfaces between c_⊥-LaNiO₃ and c_⊥-La₂NiO₄ layers across the entire interfacial region. However, the differences in their respective LaNiO₃ and La₂NiO₄ cell parameters seem to indicate a variation in their corresponding oxygen defect concentrations, higher concentration in the LaNiO₃ side for the L/L sample (larger cell parameter, see Table 1) and higher concentration in the La₂NiO₄ layer for the H/L sample (larger cell parameter, see Table 1). The reduced concentration of within LaNiO₃ in the H/L sample, is expected to limit the amount of oxygen ions that can be incorporated into LaNiO₃ (from the La₂NiO₄ film) and their drift within the film towards the Pt electrode. This could be one of the reasons for the smaller HRS/LRS ratios measured for this sample compared to the L/L sample. The second main difference between these two bilayers is found in the nano-structure of the top part of the La₂NiO₄ film. A mixed orientation of the La₂NiO₄ domains in the L/L sample generates a high density of (001)/(100) boundaries within the La₂NiO₄ layer. It is likely that these internal (001)/(100) nano-domain boundaries form additional barriers, effectively blocking both electronic and ionic drift particularly those boundaries laying perpendicular to the electric field (red colour in the schematic presented in the right column of Fig. 2), and therefore, accumulating traps in the volume of the La₂NiO₄ layer. If the presence of these domain boundaries helps increasing the rectifying and hysteretic I(V) characteristics, this could bring an additional strategy for engineering film microstructures with enhanced and better controlled memristive properties.

Conclusions

High crystal quality preferentially oriented LaNiO₃/La₂NiO₄ bilayers were grown using PLD under varying pO₂ conditions. The results confirmed that the oxygen partial pressure used during the growth of the films clearly impacts their microstructure, severely affecting the domain orientation of the La₂NiO₄ layer. At the same time electrical measurements carried out on Pt/LaNiO₃/La₂NiO₄/Pt devices fabricated out of these bilayers confirmed important pO₂-dependent differences in the resistive switching behavior of these complex-oxide stacks. Two switching scenarios have been proposed to explain the memristive behavior of the bilayers. The proposed mechanisms could be occurring independently or simultaneously, depending on the voltage range. While the first one has been attributed to the electric field-induced drift of oxygen interstitial ions in the La₂NiO₄ layer, accumulating at the LaNiO₃/La₂NiO₄ interface, and modulating the depletion width of the Schottky barrier present; the second one would be related to the exchange of oxygen ions between the top La₂NiO₄ and the bottom LaNiO₃ films. This study also reveals the presence of a high density of boundaries between domains with c_⊥ and a_⊥-orientation, under particular film growth conditions, which could also play a significant role in the accumulation of oxygen interstitials in the volume of the La₂NiO₄ layer, possibly improving the memristive properties of these novel bilayer nickelate-based devices.

Author contributions

Conceptualization: KM, MB and JS; thin film deposition: CW, JMCR, VL; TEM analysis: BB, formal electrical characterisation: CW, KM; writing –review & editing: KM, MB, JS.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

This work has been financed by the ANR funded project “MICROSWITCH” (ANR-14-ACHN-0012) and by the “FUN-TO-BE” CNRS-CSIC supported International Program for Scientific Cooperation (PICS, no. PICS07294). Part of this work was carried out through the access to PLD and HRTEM facilities granted by the NFFA-Europe Infrastructure (proposal ID 886) under Horizon 2020 EU Funding Program. JS acknowledges the grant PID2019-108573GB-C21 funded by MCIN/AEI/10.13039/501100011033. This work was also supported in part by the Spanish Ministry of Economy, Industry and Competitiveness (projects MAT2016-77100-C2-1-P), and by the Catalan AGAUR agency (project: 017-SGR-579). ICN2 is funded by the CERCA programme/Generalitat de Catalunya and by the Severo Ochoa programme (SEV-2017-0706).

References

1. I. Riess, Solid State Ionics, 2003, 157, 1–17.
2. J. A. Kilner and M. Burriel, Materials for intermediate-temperature solid-oxide fuel cells, 2014, vol. 44.
3. C. Lenser, R. Dittmann and J. P. Strachan, in Resistive Switching, Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim, Germany, 2016, pp. 437–456.
4. C. Baeumer, T. Heisig, B. Arndt, K. Skaja, F. Borgatti, F. Offi, F. Motti, G. Panaccione, R. Waser, S. Menzel and R. Dittmann, Faraday Discuss., 2019, 213, 215–230.
5. K. Maas, E. Villepreux, D. Cooper, C. Jiménez, H. Roussel, L. Rapenne, X. Mescot, Q. Rafhay, M. Boudard and M. Burriel, J. Mater. Chem. C, 2020, 8, 464–472.
6. K. Maas, E. Villepreux, D. Cooper, E. Salas-Colera, J. Rubio-Zuazo, G. R. Castro, O. Renault, C. Jimenez, H. Roussel, X. Mescot, Q. Rafhay, M. Boudard and M. Burriel, Adv. Funct. Mater., 2020, 30, 1909942.
7. A. Aguadero, J. A. Alonso, M. J. Martínez-Lope, M. T. Fernández-Díaz, M. J. Escudero and L. Daza, J. Mater. Chem., 2006, 16, 3402–3408.
8. A. Chroneos, D. Parfitt, J. A. Kilner and R. W. Grimes, J. Mater. Chem., 2010, 20, 266–270.
9. J. M. Bassat, P. Odier, A. Villesuzanne, C. Marin and M. Pouchard, Solid State Ionics, 2004, 167, 341–347.
10. M. Burriel, G. Garcia, J. Santiso, J. A. Kilner, R. J. Chater and S. J. Skinner, J. Mater. Chem., 2008, 18, 416–422.
11. M. Burriel, H. Téllez, R. J. Chater, R. Castaing, P. Veber, M. Zaghrioui, T. Ishihara, J. A. Kilner and J.-M. Bassat, J. Phys. Chem. C, 2016, 120, 17927–17938.
12. J. M. Bassat, F. Gervais, P. Odier and J. P. Loup, Mater. Sci. Eng., B, 1989, 3, 507–514.
13. J. M. Honig and D. J. Buttrey, in Localization and Metal-Insulator Transitions, Springer US, Boston, MA, 1985, pp. 409–418.
14. T. Yu, Y. F. Chen, Z. G. Liu, X. Y. Chen, L. Sun, N. Ben Ming and L. J. Shi, Mater. Lett., 1996, 26, 73–76.
15. R. Moreno, P. García, J. Zapata, J. Roqueta, J. Chaigneau and J. Santiso, Chem. Mater., 2013, 25, 3640–3647.
16. J. Zhu, J.-W. Lee, H. Lee, L. Xie, X. Pan, R. A. De Souza, C.-B. Eom and S. S. Nonnenmann, Sci. Adv., 2019, 5, eaau8467.
17. E. Mikheev, B. Hoskins, D. Strukov and S. Stemmer, Nat. Commun., 2014, 5, 3990.
18. E. Detemple, Q. M. Ramasse, W. Sigle, G. Cristiani, H.-U. Habermeier, B. Keimer and P. A. van Aken, J. Appl. Phys., 2012, 112, 013509.
19. S. Bagdzevicius, M. Boudard, J. M. Caicedo, X. Mescot, R. Rodríguez-Lamas, J. Santiso and M. Burriel, Solid State Ionics, 2019, 334, 29–35.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/d1ta10296f

---