Antiferromagnetism and p‐type conductivity of nonstoichiometric nickel oxide thin films

Abstract

Plasma-enhanced atomic layer deposition was used to grow non-stoichiometric nickel oxide thin films with low impurity content, high crystalline quality, and p-type conductivity. Despite the non-stoichiometry, the films retained the antiferromagnetic property of NiO. Nickel oxide (NiO) is one of the most studied transition metal oxides due to its versatile chemical and electronic properties, enabling it to be used in a wide variety of applications. In its stoichiometric form, NiO is an antiferromagnetic insulator, with resistivity up to 1013 Ω cm.1 When Ni vacancies are introduced, the nonstoichiometric nickel oxide becomes a semiconductor. Both theoretical and experimental results have shown that in oxygen-rich conditions the formation energy of the Ni vacancies is the lowest for all defects, leading to p-type conduction.2, 3 With a wide band gap (3.5-4.0 eV) and high work function (4.8-6.7 eV), depending on the crystalline structure, composition, and the processing of the films,4, 5 nickel oxide has gained interest especially in optoelectronics, where it is used as a hole transport/injection layer for photovoltaics and light-emitting diodes.6, 7 Nickel oxide thin films have also been shown to have great potential across a wide range of electronics applications from gas sensors to thin film transistors and resistive switching random access memory (ReRAM) devices.8-10 These include flexible electronic devices, in which nickel oxide has been shown to be stable after exposure to tensile or compressive strains.11, 12 Many of these applications rely on nickel oxide being nonstoichiometric, which has a significant effect on the electrical properties. The antiferromagnetic nature of the NiO has made it a common material in spintronics research,13 for example, in the fabrication of spin valves14 and next-generation memory technologies.15 In all of these applications, a well-controlled growth process of the materials is required. Physical vapor deposition techniques, such as pulsed laser deposition and sputtering, can provide excellent control over film stoichiometry16, 17 but can be incompatible with some applications and device integration. Chemical routes, including chemical vapor deposition techniques and solution processing, can provide a straightforward way to coat large areas uniformly at low temperatures, but the growth is governed by the chemistry, with little control over the film composition and properties. In this work, NiO thin films were grown using plasma-enhanced atomic layer deposition (PEALD), and the composition together with the optical, electrical, and magnetic properties were characterized. The magnetic properties of ALD-grown nickel oxide films have been reported before, for example, by Bachmann et al18 indicating room-temperature diamagnetism, but the results were inconclusive. Additionally, we compare the PEALD nickel oxide to the films deposited with other chemical deposition techniques, that is, atmospheric pressure chemical vapor deposition and solution processing for better understanding of the significance of the deposition method on the microstucture, stoichiometry, and consequently, the electrical and magnetic properties (see Supporting Information). Nickel oxide films were deposited at 200°C on Si, SiO2/Si, and borosilicate glass substrates using PEALD with nickel(II) acetylacetonate (Ni(acac)2) and direct oxygen plasma. Ni(acac)2 was chosen as the Ni precursor because it is widely available and inexpensive. Ni(acac)2 does not pose any significant risk to the user or the environment, and has a low reactivity at ambient conditions, which makes it easy to handle. Ni(acac)2 has a low reactivity toward water, but it has been successfully demonstrated in thermal ALD growth of nickel oxide with ozone at temperatures of 175 to 300°C,19, 20 with the lower temperature limit defined by the vapor pressure of the Ni(acac)2. However, the use of ozone is not ideal for large-area deposition as nickel oxide is known to be a highly effective catalyst for ozone decomposition.21 The use of direct O2 plasma also increases the process efficiency, as it enables the use of short plasma pulses. Here, 4 seconds pulses were enough for saturated growth, while in reports of PEALD nickel oxide with remote plasma, pulses of 15 to 30 seconds are required to achieve saturation and uniform growth under moderate vacuum conditions.22, 23 Two film thicknesses, 50 and 170 nm, were grown for different measurements. X-ray diffraction measurements indicated that the films were polycrystalline cubic NiO with clear (200) preferred orientation (Figure 1) and crystallite size of ca. 35 nm, calculated using the Scherrer equation. The lattice constant of a = 0.4185 Å is slightly higher than the literature value for the bulk NiO at room temperature (0.4176 Å), affected by the film nonstoichiometry or the residual stress in the films. A film density of 6.2 g cm-3 was determined using X-ray reflectometry. The film RMS roughness, measured with atomic force microscopy, was ca. 2 nm for the 50 nm thick films (Figure S3). Further investigation of the film microstructure with cross-sectional imaging of the thicker films using helium ion microscopy showed that the films initially grow from the substrate with a random orientation for a few tens of nanometers, after which the grains have (200) preferred orientation, forming large column-like structures (Figure 1). The film composition was determined using both X-ray photoelectron spectroscopy (XPS) and time-of-flight elastic recoil detection analysis (ToF-ERDA). ToF-ERDA depth profiling (Figure 2) showed that the films were nonstoichiometric with significant nickel deficiency, Ni = 40 at.%, O = 57 at.%, corresponding to Ni/O = 0.7. Apart from nickel and oxygen, the films contain a small amount of light element impurities, that is, ∼3 at.-% hydrogen, 0.3 at.-% carbon, and 0.5 at.-% nitrogen (see Table S1 of supporting information). Of these H, C originate from the Ni(acac)2 precursor while nitrogen is incorporated into the films during the pulsing of the O2/N2 plasma gas mixture. The impurity contents are also significantly lower than in films deposited with other chemical methods (see Table S1). The low impurity content also confirmed the suitability of the PEALD method with Ni(acac)2 precursor for the growth of nickel oxide films, as PEALD processes for β-diketonate precursors for growth of nickel oxide films have not been reported before,22-24 and the thermal ALD processes can suffer from high concentrations of carbon impurities.25 XPS results, shown in Figure 2 are in accordance with the ToF-ERDA, showing a nonstoichiometric composition (see Table S1). The spectrum of O 1s binding energy indicates defective oxide content in the films, seen as the peak at 531.4 eV, typically associated with the presence of Ni3+ and corresponding Ni vacancies. Similar features in the O 1s spectrum have also been associated with hydroxide (-OH) species, but the low H content measured with ToF-ERDA does not support this assumption. However, the quantitative determination of the amount of nickel vacancies in the films, that is, Ni2+/Ni3+ ratio, from the XPS spectra is challenging due to the complexity of the Ni 2p peak shapes resulting from multiplet splitting, shake-up, and plasmon loss structure.26 The electrical properties of the films were characterized by Hall effect measurement. The resistivity of 50 nm films was 80 Ω cm, measured using a van der Pauw configuration, which is in accordance with the film composition corresponding with an increase in the free-hole concentration by the Ni vacancy ionization. The resistivity of 170 nm films was ca. 200 Ω cm. The films were confirmed to have p-type conductivity, with a hole density of ∼1017 cm−3 and mobility ∼0.1 cm2 V−1 s−1, independent of the film thickness. However, antiferromagnetic behavior of the films (see discussion next) could add significant uncertainty to the Hall effect measurements. The resistivity values are consistent with what has been reported for ALD- and CVD-grown nickel oxide.27-29 The mobility and carrier density values are in accordance with what has been reported from Hall measurements of vapor deposited nickel oxide thin films,28 but vary by a few orders of magnitude from what has been derived from electrochemical impedance spectroscopy data.6, 29, 30 In addition, a thermal activation energy (Ea) of 0.31 eV was determined using a two-point-probe measurement within a temperature range of 25°C to 130°C and the Arrhenius relation (Figure S7). The Ea indicates the position of the Fermi level from the valence band edge. The obtained Ea is consistent with values extracted from nickel oxide thin films with similar resistivity grown by other chemical methods.31, 32 The magnetic properties of the films were investigated using a magnetometer with ±70 kOe field and 1.8 to 400 K temperature range. The hysteresis loops for the 50 nm nickel oxide film, measured at 2 K temperature, show linear behavior (Figure 3) with nearly zero coercivity (Hc) and remanence (MR). The magnetization does not saturate within the field range of ±70 kOe. The temperature-dependent magnetization curve shows no bi-furcation between the zero-field cooled (ZFC) and 100 Oe field cooled (FC) curve. No remanence magnetization (MR) was observed within the measured temperature range (350-2 K), as seen in the inset of Figure 3. This indicates a strong antiferromagnetic property of the films throughout the temperature range. To further confirm the antiferromagnetism of the nonstoichiometric nickel oxide thin films, the exchange bias (EB) effect was investigated in-plane. For this, 6 nm of Ni80Fe20 (Py) was deposited on PEALD nickel oxide by DC magnetron sputtering. Since Py is ferromagnetic, an EB shift (HE) should be observed if the nickel oxide film is antiferromagnetic. Figure 3 shows a clear EB behavior measured in the Py (6 nm)/Ni1-xO (50 nm) bilayer film. The initial hysteresis loop (ZFC loop) which starts from (0,0) confirms a complete demagnetization of the film by the applied protocol (see Supporting Information) and that no stray field is present in the superconducting coil of the magnetometer. For ZFC, no HE was present. For positive 10 kOe (negative −10 kOe) applied bias field, the FC hysteresis loop shifts toward negative (positive) field direction. The magnitude of EB shift for positive bias field is +HE ≈ −400 Oe and for negative bias field -HE ≈ 500 Oe. Such asymmetry in EB is often observed in antiferromagnetic oxides.33 The ferromagnetic Py saturates at a very high field of ≈ 5 kOe and the coercivity is also high ∼700 to 800 Oe compared with bare Py, as seen in the inset of Figure 3. This is due to the strong EB coupling between the Py and nickel oxide.34 We have shown that a β-diketonate type precursor can be used in an efficient PEALD process to grow nickel oxide thin films. The nonstoichiometric nickel oxide films are highly crystalline and have a dense microstructure, low impurity content, and high nickel vacancy concentration leading to p-type conductivity. The low concentration of hydrogen impurities is important for films used in electrical applications, as it has been shown that in wide band gap oxide hydrogen can act as compensating center that always counteracts the prevailing conductivity.35 In addition to structural and electrical characterization, the magnetic behavior of the films was thoroughly screened. Here, we confirm that the nickel oxide thin films remain antiferromagnetic despite the significant deviation from the NiO stoichiometry. The comparison to other chemical vapor and solution deposition techniques showed that nickel deficient films are grown with these techniques and while the choice of deposition method has an effect on the microstucture and characteristics of the films, the basic properties of the nonstoichiometric nickel oxide, including antiferromagnetism, persist (See Supporting Information). This underlines the potential of ALD and other chemical routes in fabrication of thin films for different device applications requiring p-type conductivity and/or antiferromagnetism. Mr. Chris Amey is acknowledged for the XPS measurements, and Mr. Jerome Innocent for measuring UVVis of the PEALD and AP-CVD samples. M.N. and J.L. M.-D. acknowledge funding from the EPSRC grant EP/P027032/1. T.N.H. acknowledges funding from the EPSRC Centre for Doctoral Training in Graphene Technology (No. EP/L0160871) and Aziz Foundation. T.M. and J.L.M.-D. acknowledge funding from EU grant H2020-MSCA-IF-2016 745886 MuStMAM and Isaac Newton Trust (RG96474). R.L.Z.H. acknowledges funding from the Royal Academy of Engineering via the Research Fellowship scheme (No.: RF/201718/17101), as well as support from Magdalene College Cambridge. The authors declare that there is no conflict of interest regarding the publication of this article. Appendix S1. Supporting Information. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.