Bulk-like dielectric and magnetic properties of sub 100 nm thick single crystal Cr2O3 films on an epitaxial oxide electrode

The manipulation of antiferromagnetic order in magnetoelectric Cr2O3 using electric field has been of great interest due to its potential in low-power electronics. The substantial leakage and low dielectric breakdown observed in twinned Cr2O3 thin films, however, hinders its development in energy efficient spintronics. To compensate, large film thicknesses (250 nm or greater) have been employed at the expense of device scalability. Recently, epitaxial V2O3 thin film electrodes have been used to eliminate twin boundaries and significantly reduce the leakage of 300 nm thick single crystal films. Here we report the electrical endurance and magnetic properties of thin (less than 100 nm) single crystal Cr2O3 films on epitaxial V2O3 buffered Al2O3 (0001) single crystal substrates. The growth of Cr2O3 on isostructural V2O3 thin film electrodes helps eliminate the existence of twin domains in Cr2O3 films, therefore significantly reducing leakage current and increasing dielectric breakdown. 60 nm thick Cr2O3 films show bulk-like resistivity (~ 1012 Ω cm) with a breakdown voltage in the range of 150–300 MV/m. Exchange bias measurements of 30 nm thick Cr2O3 display a blocking temperature of ~ 285 K while room temperature optical second harmonic generation measurements possess the symmetry consistent with bulk magnetic order.

One issue hindering electric field manipulation of magnetic order in thin film Cr 2 O 3 is the existence of twin domain boundaries that result from the growth on elemental metal electrodes, particularly in films below 250 nm 12 which are necessary for technological adoption. The relatively conductive twin boundaries lead to high leakage current and reduce dielectric breakdown voltage down below the critical magnetoelectric switching voltage. In previous reports, this issue has been circumvented by utilizing Cr 2 O 3 films of large thickness 12,16 but at proposed device scales this is not a viable solution. Using a V 2 O 3 electrode layer, a metallic oxide isostructural with Cr 2 O 3 , has been shown to reduce or even eliminate twin domains and thereby reduce the leakage current of a 300 nm thick Cr 2 O 3 film in comparison with metal electrodes 8,17 . In this work, we investigate the DC dielectric and magnetic properties of very thin (30-60 nm) single crystalline Cr 2 O 3 films on V 2 O 3 thin film electrodes at room temperature. Leakage data shows robust bulk like behavior for 60 nm thick samples with electrodes below 60 µm in diameter and 6-8 orders of magnitude lower leakage current than twinned Cr 2 O 3 films grown on (111)-oriented Pt electrodes. Finally, investigation of the magnetic properties of single crystal Cr 2 O 3 thin films using an exchange coupled ferromagnetic layer and optical second harmonic generation indicates bulk like behavior around room temperature in films at 30 nm thickness.  12 , which is ~ 20% of the bulk breakdown field (1,000 MV/m) 16 . These values, however, are 3 × higher than the breakdown fields observed in the Cr 2 O 3 /Pt heterostructures containing twin domains. Dielectric breakdown of 30 nm thick single crystal Cr 2 O 3 happens at slightly lower values when compared to thicker films (~ 90 to 160 MV/m) yet remains higher than that observed in the thicker and twinned Cr 2 O 3 /Pt heterostructures.
Magnetic properties. We next consider the magnetic properties of the single crystal Cr 2 O 3 films. To probe the intrinsic magnetic and magnetoelectric order of the 30 nm single crystal Cr 2 O 3 thin film on V 2 O 3 at room temperature, we employed transmission optical second harmonic generation (SHG) measurements (Fig. 3a). www.nature.com/scientificreports/ Below the Néel temperature, bulk Cr 2 O 3 possesses the magnetic point group 3 ′ m ′ that allows the existence of both magnetic ( χ m ) and electric ( χ e ) dipole susceptibility tensors [19][20][21] . The magnetic and electric contributions to SHG signals can be expressed as:  www.nature.com/scientificreports/ where E i is the electric field of the incident beam. When fixing the analyzer along the y-direction ([1100 ]) (or x-([1120])), sin φ s (or cos φ s ) goes to 0, and therefore the SHG intensity signal coming from the magnetic ( I MD ) and electric ( I ED ) dipole contributions will follow sin 2 (2 φ ) (or cos 2 (2 φ )) and cos 2 (2 φ ) (or sin 2 (2 φ)), respectively. The existence of the magnetic dipole signal is attributed to centrosymmetric point group 3 m and is expected in single crystal bulk Cr 2 O 3 . The electric dipole signal, however, is a proof of the existence of non-centrosymmetric magnetic order and only exists below the Néel temperature. The rotational-anisotropy data presented in Fig. 3b, following the cos 2 (2 φ ) (sin 2 (2 φ )) dependence on the incident polarization φ when the SHG analyzer is fixed along the y-axis (x-axis), proves the presence of the electric dipole contribution to the SHG signal and confirms that 30 nm thick single crystal Cr 2 O 3 films on V 2 O 3 electrodes possess the magnetic symmetry consistent with the magnetic order in bulk Cr 2 O 3 at room temperature. Using interface exchange coupling with a thin (4 nm) Permalloy (Py) layer, we probe the blocking temperature of the exchange bias heterostructure to approximate the Néel temperature. Figure 3b shows exchange bias and coercive fields as a function of temperature after cooling from 350 to 10 K in a ± 5,000 Oe in-plane training field. As the SHG of our thin films is consistent with the c-axis antiferromagnetic anisotropy of bulk Cr 2 O 3 thin films, the applied in-plane magnetic field while cooling the sample through the Néel temperature is expected to induce a slight canting of the Cr moments in-plane, consistent with the observed in-plane exchange coupling with Py (Supplementary Figure S2). The exchange bias and coercivity enhancement extracted from M(H) curves disappear at ~ 285 K and ~ 295 K, respectively. These data reveal a blocking temperature that is lower than the bulk Néel temperature (307 K), however, this measured blocking temperature is significantly higher than the blocking temperature of films with comparable and greater thickness reported in the literature 22,23 . The blocking temperature can be qualitatively explained using Meiklejohn-Bean model with the competition between interface exchange coupling ( fJ eb -where f is a factor between 0 and 1 and represents the degree of interface spin disorder and often assumed to be 1), and the product of magnetic anisotropy energy ( K AF ) and thickness ( t AF ) of the antiferromagnetic layer 22,24 . Thus the increase in blocking temperature might be a result of a change in the magnetic anisotropy energy due to the reduced epitaxial strain from the Al 2 O 3 substrate 22 , however, the measured blocking temperature maybe lower than the Néel temperature of the thin film. Regarding the low temperature behavior, the exchange bias is ~ 0 and begins to increase up to ~ 35 Oe at 150 K where afterward it begins to decrease with increasing temperature up to the blocking temperature. Meanwhile, the coercivity monotonically decreases with increasing temperature. It is reported that there is a change in the Cr 2 O 3 crystal structure at low temperature that is thought to lead to an in-plane tilting of the magnetic order 9 or a structural rearrangement at the (0001) surface of Cr 2 O 3 25,26 which then affects both its antiferromagnetic structure and surface magnetism. In the Meiklejohn-Bean model, the ratio ( R ≡ K AF t AF fJ eb ) determines the exchange bias and coercive field behavior 24 . The competition between the change in spin structure (affects K AF , J eb ) and the surface reconstruction (affects J eb , f ) will directly impact the low temperature behavior of the exchange bias field. In order to clarify the situation, future work focusing on isolating these factors is needed.

conclusion
In conclusion, by using V 2 O 3 as an epitaxial buffer layer, crystallographic twinning of Cr 2 O 3 thin films can be eliminated leading to near bulk dielectric and magnetic behavior in Cr 2 O 3 films with thickness well below 100 nm. Leakage measurements performed on very thin single crystal Cr 2 O 3 films, along with electric breakdown tests, as a function of capacitor area suggest the need to further improve film quality or develop additional dielectric layers to mitigate the dielectric parasitics observed at larger capacitor size and enable magnetoelectric characterization with transport methods. Our investigation of magnetic properties in 30 nm thick films indicate bulk magnetic and magnetoelectric order at room temperature and indicate that the magnetoelectric switching of very thin single crystal Cr 2 O 3 films on V 2 O 3 electrodes may be possible at room temperature. As a Metal-Insulator-Transition material, the growth of V 2 O 3 has been intensively discussed elsewhere [28][29][30] . The challenge, however, lies in finding compatible conditions for the growth of Cr 2 O 3 . We found that the presence of oxygen (10 mTorr) as the background gas causes the formation of V 2 O 5 . The V 2 O 5 was found to melt even at a substrate temperature (650 °C) lower than its melting point (690 °C). Reduction of the substrate temperature in oxygen, however, favors the formation of the VO 2 phase. By switching to argon gas at the same temperature (400 °C), we achieved V 2 O 3 thin films with desired topography, x-ray diffraction pattern, and resistivity. The substrate temperature for the growth of the Cr 2 O 3 layer was chosen to achieve high crystallinity and low surface roughness while mitigating the oxidation of the underlying V 2 O 3 layer from the oxygen in the ablation plume of the Cr 2 O 3 target. We optimized our conditions for Cr 2  www.nature.com/scientificreports/ (700 °C) to achieve good crystallinity. However, at that temperature, we speculate that there is a reaction happening at the interface between V 2 O 3 and Cr 2 O 3 , since V 2 O 3 is easy to be oxidized in the presence of oxygen, especially at elevated temperature. The growth temperature was therefore systematically reduced. These above conditions were selected after tuning deposition conditions using surface roughness from AFM and XRD measurements (Peak position, intensity, and oscillation fringes) were used to assess conditions and feedback the growth. From the presence of oscillation fringes around V 2 O 3 peak, the quality is considered as comparable to previous report of a good crystallinity in thin film 30 . Circular Ti (6 nm)/Pt (120 nm) top electrode capacitors of diameter from 10 to 150 mμ were defined using a standard liftoff process and PLD deposition.
X-ray diffraction. 2θ − ω and φ scans were performed using a Rigaku Smart Lab diffractometer (Cu K α radiation and equipped with a Ge (220) × 2 monochromator on the incident) side to assess the orientation and crystallinity of the films.
Atomic force microscopy. The surface topography of the films is observed using an NT-MDT NTEGRA atomic force microscope (AFM).
Magnetometry. Vibrating sample magnetometry was performed using a QuantumDesign Dynacool Physical Property Measurement System (PPMS). electrical measurements. Leakage currents through thin film Cr 2 O 3 are investigated using a Radiant Technologies Precision Multiferroic II with minimum current detection of 1 pA for a 2 s integration period. An electric field of approximately 1.4 to 1.7 MV/m is applied to detect leakage current. Breakdown tests are performed using a Keithley 2420 with a detection limit of 500 pA. These measurements are performed at ambient conditions in a probe station using W tips with a tip diameter of 5 µm.
Second harmonic generation measurement. A rotational-anisotropy SHG measurement was performed with the beam at normal incidence. The transmitted SHG intensity is collected with a single photon count detector as a function of the azimuthal angle ϕ i between the incident electric polarization and the in-plane crystalline axis [ 1120 ] and the analyzer angle φ s between the selected SHG electric polarization and the in-plane crystalline axis [ 1100 ]. The incident ultrafast light source was of 800 nm wavelength, 40 fs pulse duration and 200 kHz repetition rate, and focused onto a 20 μm diameter spot on the sample with a fluence of ~ 0.25 mJ/cm 2 .

Data availability
The data that support the findings of this study are available from the corresponding author upon reasonable request.