Epitaxial growth and superconducting properties of thin-film PdFe/VN and VN/PdFe bilayers on MgO(001) substrates

Introduction

Since its invention, rapid single-flux quantum (RSFQ) logic [1,2] based on superconducting digital electronics has been seriously considered as an alternative to semiconductor electronics for supercomputing applications [3-5]. Merging it with magnetism [6-8] has given a birth to superconducting spintronics [9,10]. The latter concept was implemented in the US Cryogenic Computing Complexity (C3) Program [11-13] with the goal “to demonstrate a small-scale computer based on superconducting logic and cryogenic memory that is energy-efficient, scalable and able to solve interesting problems”, opening prospects of reaching 100 PFLOPS/s with about 200 kW of electric power consumption including the cryogenic cooling. Niobium-based Josephson junction technology is currently implied to be used for the logics fabrication, however, hybrid Josephson junctions incorporating magnetic components are also considered for the mainframe computation components [9,14-19], and cache and main memories [8,20-25]. It is argued that the use of magnetic Josephson junctions in single-flux quantum electronics significantly reduces the number of junctions and interconnects in the circuits [26] and also has other important advantages such as wide operation margins and low bit-error rate [27]. The magnetic material has to be magnetically soft, tunable and weak in the sense of small spin-polarization of the conduction band [10,28]. The latter provides a large superconducting coherence length and hence bypasses a necessity to deposit flat, nanometer-thick continuous layers expected for strong elemental ferromagnets. A combination of niobium as a superconductor with a Pd_1−xFe_x alloy as a soft and weak ferromagnet was considered as material of choice for superconducting magnetic random access memories (MRAM) [8,29,30]. However, no further developments towards a prototype using a Pd_1−xFe_x alloy have been demonstrated. There are indications of non-homogeneous, nanoclustered magnetism in Pd_0.99Fe_0.01 films grown on niobium [31], which may cause a shortening of the spin-memory length [32] and a reduction of the Josephson critical current.

In general, the metallic Nb lattice (body-centered cubic with a_Nb = 329.4 pm) poorly matches that of the palladium-rich Pd_1−xFe_x alloys (face-centered cubic with a₀ = 389 pm). Therefore, a good crystallinity of the layer stack can hardly be expected. In the resulting polycrystalline films, crystallite boundaries and crystal lattice imperfections can lead to the segregation of iron impurities and to nanoclustering of the alloy. Following the development of a way to grow single-crystalline, magnetically homogeneous epitaxial Pd_1−xFe_x films on MgO(001) single-crystalline substrates [33], we propose fully epitaxial Pd_1−xFe_x/VN and VN/Pd_1−xFe_x (x ≤ 0.08) building blocks as an alternative choice for superconducting MRAM materials, in which vanadium nitride (VN) serves as the superconductor. The magnetic anisotropies of a 20 nm thick Pd_0.96Fe_0.04 film of the first-generation epitaxial sample of VN/Pd_0.96Fe_0.04 on MgO(001) were studied by using a ferromagnetic resonance technique in [34].

Results and Discussion

Sample preparation

Single-crystalline MgO(001) (henceforth designated MgO) epi-polished substrates (CRYSTAL GmbH, Germany) with a size of 10 × 5 × 0.5 mm³ were annealed at 800 °C for 5 min in the ultrahigh vacuum (UHV) molecular beam epitaxy (MBE) chamber with a residual pressure below 10⁻¹⁰ mbar (SPECS, Germany). Then, depending on the desired structure, either the Pd_1−xFe_x alloy layer or the VN layer was deposited. The Pd_1−xFe_x layers were grown by means of UHV MBE following a three-step procedure described in detail in [33]. Metallic Pd (99.95% purity, EVOCHEM GmbH, Germany) and Fe (99.97% purity, ChemPur GmbH, Germany) were co-evaporated from the pre-calibrated high-temperature effusion cells to obtain the desired Pd_1−xFe_x composition.

Vanadium nitride layers were synthesized by using reactive DC magnetron sputtering (MS) in the UHV chamber with a base pressure of p ≤ 5 × 10⁻¹⁰ mbar (BESTEC, Germany). During this process, the substrate had a temperature of 500 °C. A mixture of high-purity (99.9999%) argon (Ar) from a gas chromatography purification system and high-purity (99.9999%) nitrogen (N₂) at a composition of Ar/N₂ = 60:40 was used as plasma gas for the reactive synthesis of VN. During the deposition process, the pressure of the Ar/N₂ gas mixture in the chamber was automatically kept at 6 × 10⁻³ mbar. A metallic vanadium disk of 99.95% purity (GIRMET Ltd, Russia) was used as a target. The magnetron power was 50 W, the distance between the target and the substrate was 20 cm, and the deposition rate was 0.2 nm/min.

To grow heterostructures, the samples on the molybdenum holder were moved without breaking vacuum via the UHV transfer line between the MBE and MS deposition chambers as well as the analysis chamber (SPECS, Germany).

To perform a comparative study allowing to see only the proximity effect of the ferromagnetic layer on the properties of the superconducting VN layer, the latter was deposited in one run for all studied samples. To do this, we mounted two 10 × 5 × 0.5 mm³ MgO substrates close and parallel to each other on the sample holder and used a system of two orthogonal shutters in the MBE chamber. After depositing a 20 nm thick Pd_0.96Fe_0.04 layer onto one substrate (with the second being blocked by the shutter), the sample holder was moved to the magnetron chamber, and a 30 nm layer of VN was grown on both substrates. Then, the holder was moved back to the MBE chamber, and a 12 nm thick Pd_0.92Fe_0.08 layer was deposited to a half of both samples using the second shutter. The thicknesses of the Pd_1−xFe_x layers were adjusted to possess identical magnetic moments. In situ tests of crystallinity, VN stoichiometry and resulting composition of Pd_1−xFe_x were taken at each deposition step using low-energy electron diffraction (LEED) and X-ray photoelectron spectroscopy (XPS). Finally, all structures were capped with 10 nm layer of undoped Si by magnetron sputtering to prevent sample deterioration. Thus, a VN film and stacks of Pd_0.96Fe_0.04/VN and VN/Pd_0.92Fe_0.08 (the first component in a stack being directly deposited to MgO) have been obtained with the identical properties of the VN layer in each sample.

Crystallinity and epitaxial growth

The crystallinity and the epitaxial growth of the thin films were examined in situ by using LEED (SPECS, Germany). LEED images were taken of the pristine MgO(001) substrate after annealing (Figure 1a), after the deposition of VN(30 nm) on MgO (Figure 1b), after the deposition of Pd_0.92Fe_0.08 on VN (Figure 1c) and after the deposition of VN on Pd_0.96Fe_0.04 (Figure 1d). Figure 1b indicates that the individual VN thin film has grown cube-on-cube epitaxially (for an individual Pd_1−xFe_x film see the full crystallinity analysis in [33]). Figure 1c,d shows that the Pd_0.96Fe_0.04/VN and VN/Pd_0.92Fe_0.08 heterostructures are pass-through epitaxial. This is, first of all, due to the good lattice match between MgO, VN and Pd: a_MgO = 421.2 pm, a_VN = 413.7 pm [35] and a_Pd = 389.1 pm. Thus, the lattice mismatch between MgO and VN is only about 1.7%, and between Pd and VN it is as small as 5.95%.

The in situ LEED analysis was corroborated with ex situ X-ray diffraction (XRD, BRUKER D8, Germany) measurements using Cu Kα (λ = 1.5418 Å) radiation in the Bragg–Brentano geometry with a scanning rate of 0.002°/s in the 2θ range from 17° to 82° and a step width of 0.0153°. Room-temperature XRD patterns of the pristine MgO(001) substrate, the VN thin film on MgO, Pd_0.96Fe_0.04 on MgO and the Pd_0.96Fe_0.04/VN heterostructure are shown in Figure 2. The θ–2θ scans clearly indicate the single-crystalline structure of the VN and Pd_0.96Fe_0.04 thin films and of the Pd_0.96Fe_0.04/VN heterostructures. The (002) reflex of the MgO substrate, the (002) reflex of the VN film (30 nm), and the (002) reflex of the Pd_0.96Fe_0.04 (20 nm) film can be easily identified.

The significant peak broadening of the diffraction maxima of VN and Pd_0.96Fe_0.04 is primarily due to small coherent scattering range τ along the normal to the film plane (Scherrer broadening); XRD data with accounting for the instrument function [33] yields estimates of τ ≈ 22.0 nm for Pd_0.96Fe_0.04, τ ≈ 12.6 nm for Pd_0.92Fe_0.08 and τ ≈ 30.4 nm for VN, which agree quantitatively with the film thickness values d(Pd_0.96Fe_0.04) ≈ 21.5 nm, d(Pd_0.92Fe_0.08) ≈ 12.5 nm and d(VN) ≈ 29.8 nm, respectively, measured ex situ with a BRUKER DektakXT stylus profiler by using the shadow mask method. Thus, LEED and XRD measurements confirm that the VN thin film and the Pd_0.96Fe_0.04/VN and VN/Pd_0.92Fe_0.08 heterostructures have grown cube-on-cube epitaxially and that all samples are single crystalline.

Stoichiometry and chemical composition

The stoichiometry and chemical composition of the VN and the Pd_1−xFe_x layers were analyzed in situ using XPS. The measurements were carried out in the UHV analysis chamber (base pressure p < 3 × 10⁻¹⁰ mbar) equipped with a Mg Kα X-ray source operated at 12.5 kV and 250 W, and a Phoibos-150 hemispherical energy analyzer (all from SPECS, Germany). Figure 3a,b shows the XPS spectra of the as-deposited VN/Pd_0.92Fe_0.08 thin film heterostructure. The binding energies of the Fe 2p_1/2, Fe 2p_3/2, and Pd 3d_3/2 and Pd 3d_5/2 states are 721.0, 707.7, and 340.2 and 335.0 eV, respectively, which agrees well with literature data [33,36].

Figure 3c,d shows the XPS spectra of the VN thin film on MgO. The binding energies of the V 2p_1/2, V 2p_3/2 and N 1s states are 521.1, 513.6 and 397.4 eV, respectively, which are very close to that given in the literature for crystalline VN [37,38]. The presence of a characteristic satellite at a binding energy of ca. 515 eV is a fingerprint of V in a nitride compound [37]. The chemical composition of the as-grown VN and Pd_1−xFe_x layers was analyzed with the CasaXPS software [39]. According to the XPS data, the stoichiometry of synthesized layers was Pd/Fe = 96:4, V/N = 52.5:47.5 and Pd/Fe = 92:8, respectively, with an accuracy of ±0.5%. Neither impurities nor surface contaminations were detected (compare with [40]). All recorded high-resolution XPS spectra of VN and Pd_1−xFe_x films were calibrated to the binding energies of crystalline VN at 513.6 eV and of metallic Pd at 335.0 eV [33,37], respectively.

Magnetic moment measurements shown in Figure 4 confirm the composition of Pd_0.96Fe_0.04 and Pd_0.92Fe_0.08 through the ferromagnetic transition temperature T_C ≈ 125 K and T_C ≈ 240 K, respectively [41].

Temperature dependence of resistance and superconducting transition

A physical property measurement system (QUANTUM DESIGN PPMS-9, USA) was used for studying the temperature dependence of the electrical resistance of the VN thin films and Pd_0.96Fe_0.04/VN and VN/Pd_0.92Fe_0.08 heterostructures in the temperature range of 4.2–300 K. A four-probe resistance measurement scheme was used. Figure 5 shows the measurement results as a function of the temperature for the epitaxial VN film and the heteroepitaxial Pd_0.96Fe_0.04/VN and VN/Pd_0.92Fe_0.08 samples. Table 1 contains the data on the residual resistance ratio RRR (i.e., the ratio of room temperature resistance, R_300K, to the resistance at 10 K, R_10K), the superconducting transition temperature (mid-transition criterion) and the width of the superconducting transition (10–90% criterion) for the VN thin film and the heterostructures with Pd_1−xFe_x.

The temperature dependence of the resistance of the VN thin film is of metallic type and exhibits two temperature intervals, one above 250 K and another one in the range of 80–180 K, of quasi-linear temperature dependence with different temperature coefficients of resistivity (TCR), i.e., 9.7 × 10⁻³ Ω/K and 2.1 × 10⁻² Ω/K, respectively, marked by red straight lines over the green line in Figure 5a. It is similar to the R(T) behavior of VN/MgO(011) samples in [42], which was explained by a change in the electron/phonon scattering amplitude upon the structural phase transition from cubic to tetragonal at T_s = 250 K. Below 50 K the R(T) dependence saturates approaching the residual resistance originating, in general, from impurities and imperfections. Further cooling results in the phase transition to the superconducting state as it is shown in Figure 5b. The RRR value of 5.2 and the room-temperature resistivity of 42.5 μΩ·cm for the 30 nm thick VN film are among the best values obtained to date [42-45], indicating the high purity and structural quality of our VN film.

The superconducting transition temperature T_c of the VN film is 7.7 K (see Table 1), which is well above the temperature of liquid helium, LHeT = 4.2 K. Figure 5b shows a very sharp resistive transition at T = 7.7 K with a small width of 25 mK, which is quite remarkable compared to an elemental niobium film of the same (30 nm) thickness deposited in the same chamber and under vacuum conditions (ΔT_c [Nb(30 nm)]) = 10–23 mK).

Combining the VN film into a heterostructure with a palladium-rich Pd_1−xFe_x alloy leads to a lowering of T_c because of the proximity effect [28]. This may shift the material operation temperature close to or even below the LHeT. With the iron content x in Pd_1−xFe_x alloy below 0.08 its magnetic properties meet all the requirements for the F-layer in superconducting spintronic S/F/S-type structures, i.e., it is a weak ferromagnet with a low coercive field [41]. It is important that magnetic properties of epitaxial Pd_1−xFe_x films are precisely controlled with the iron content x [41], and a perfect cube-on-cube epitaxy is realized with either the MgO(001) substrate or with the superconducting VN layers in any sequence. Figure 5b shows that 12 nm thick layer of Pd_0.92Fe_0.08 alloy adjacent to the 30 nm VN film lowers T_c from 7.7 K to 6.1 K, which is well above the LHeT. Moreover, Figure 5b demonstrates that the transition stays sharp: the maximum ΔT_c increases only to 50 mK, and there is no tail towards lower temperatures. Also, there is a room to optimize the superconducting parameters of the VN film towards an increase in T_c by about 1 K [43,44]. In our opinion, the results hint at a possible use of heteroepitaxial combinations of nitrides as superconductors and palladium-rich Pd_1−xFe_x alloys as weak tunable ferromagnets to improve the operation characteristics of superconductor–ferromagnet–insulator heterojunctions for superconducting spintronics applications. For example, cubic superconducting MoN_x, which is a Josephson junction technology material [4,5,46], exhibits a good epitaxial match with Pd_1−xFe_x alloys, a₀(MoN) = 416.3 pm.

Conclusion

Fully epitaxial single-crystalline thin films of VN and heteroepitaxial structures of Pd_1−xFe_x/VN and VN/Pd_1−xFe_x (x = 0.04, 0.08, respectively) were grown on single-crystalline MgO(001) substrates using a combination of UHV molecular beam epitaxy and magnetron sputtering. The obtained 30 nm thick VN films exhibit a sharp superconducting transition with T_c = 7.7 K and ΔT_c = 25 mK. The heteroepitaxial Pd_0.96Fe_0.04/VN and VN/Pd_0.92Fe_0.08 structures reveal a superconductor–ferromagnet proximity suppression of the transition temperature to T_c = 6.1 K. This is, however, well above the liquid helium temperature of 4.2 K and, therefore, suitable for superconducting spintronics. The superconducting transition stays sharp with a somewhat larger width of ΔT_c = 50 mK. Moreover, there is no resistive tail towards lower temperatures. These results, in our opinion, indicate that fully epitaxial Pd_1−xFe_x/VN and VN/Pd_1−xFe_x thin film stacks can be considered as building blocks for superconducting spintronics elements.