Atomic layer deposition of cobalt oxide on oxide substrates and low temperature reduction to form ultrathin cobalt metal films

Cobalt oxides have a wide range of applications, such as photocatalysts, electrodes, optical sensors, magnetic detectors, and catalysts.^1–7 Thin films of cobalt oxides have been grown on different substrates exhibiting a variety of properties.^2–6,8,9 Cobalt (II) oxide has a bandgap of 2.4 eV, which has been used for visible light driven photoelectrochemical water oxidation.² The growth of CoO has been performed using different growth techniques, such as molecular beam epitaxy (MBE),^10–12 pulsed laser deposition,^13,14 chemical vapor deposition (CVD),^15–17 and atomic layer deposition (ALD).^2,9,18–21 Among them, ALD offers film uniformity and conformality over large area substrates, which are desired for most applications.

Compared to cobalt oxides, cobalt metal is a widely studied material for microelectronics and memory applications.^22–30 Co is well known to strongly adhere to Cu, and Co and Co alloy thin films have been studied as a Cu electromigration barrier and next-generation liner material in the back end of line interconnects.^22–25 Additionally, Co and Co alloys, such as CoFe and CoFeB, are used as magnetic materials for the magnetic tunnel junction (MTJ).^26–30 A basic MTJ is composed of two ferromagnetic layers (known as the fixed and free layers) separated by an ultrathin insulating tunneling barrier.²⁸ While magnetic moment orientation of the fixed layer remains unchanged, the magnetic moment orientation of the free layer can be switched during device functioning. If the magnetic moments of the two layers are parallel, it leads to a low resistance state of the MTJ, while the antiparallel configuration leads to a high resistance state. Crystallographic structure, particle size, and texture deeply affect magnetic properties, such as anisotropy, coercivity, and the magnetization reversal process. Therefore, the study of magnetic behavior of Co and Co alloys has attracted considerable attention.^31–34 Cobalt metal can be used as either the fixed or free layer, depending on its structure and particle size.^32–37 

The deposition of Co and Co alloy thin films has been performed by both physical vapor deposition and CVD techniques.^29–37 With downscaling of electronic device feature sizes, thin film uniformity and conformality become increasingly important, and this makes ALD a very attractive technique for cobalt metal film deposition, especially ultrathin cobalt metal films in interconnects or the MTJ stack. One advantage of ALD is precise thickness control since the precursors adsorb and/or react until a monolayer forms.³⁸ The self-limiting behavior of ALD produces smooth and conformal films to the underlying substrates. When compared to sputtered Co-based magnetic films, a promising advantage of ALD films is that patterned films can be directly grown by area-selective ALD, thus eliminating alignment and postdeposition etching steps.^39–41 This has the potential to minimize the film quality degradation coming from structural or compositional damages caused by etching.

There are limited reports of cobalt metal ALD.^41–49 While most Co ALD precursors are organometallic compounds, relatively high energy is needed to convert the Co²⁺ or Co³⁺ to Co⁰, and the relatively high energy normally comes from either a plasma or high temperature. Plasma-enhanced ALD of cobalt metal has been reported by using precursors like Co₂(CO)₈, CoCp₂, and Co(MeCp)₂ and plasmas like H₂ and NH₃, but a plasma may damage the substrate surface and the anisotropy of plasma bombardment compromises the conformality of ALD.^42–44 Thermal ALD of cobalt metal has been achieved at a temperature above 260 °C by using bis(N,N′-di-i-propylacetamidinato)cobalt(II) and bis(N-t-butyl-N′-ethylpropanimidamidato) cobalt(II) precursors and H₂, but there is incorporation of C and N in the Co film due to precursor decomposition at high temperature.^41,45,46 Recently, thermal ALD of cobalt metal has been demonstrated at a relatively low temperature below 200 °C, but these processes cannot grow Co on oxide substrates due to the selective nucleation behavior of the precursors toward metal substrates.^47–49 

Another strategy to produce metal films is by reducing metal oxides using reducing agents.^8,9,50,51 In this study, we investigate the feasibility of producing Co metal films by reducing cobalt oxide. The reduction of CoO on oxide substrates by using hydrogen or deuterium gas at 400–500 °C reported herein likely results in discontinuous cobalt islands instead of continuous cobalt films. This dewetting phenomenon is common during high temperature processing of metals on oxide substrates, especially for ultrathin metal films.^8,9,52–55 Driven by minimization of the total energy of the film-substrate interface, film surface, and substrate surface, the edges of metastable films have the tendency to shrink and form islands when enough atom mobility is offered by thermal energy.⁵² Dewetting normally starts at the edges of voids, which can be as small as several nanometers. During reduction, the loss of oxygen atoms creates voids within the films, which facilitates the onset of dewetting. In this work, we report a method combining a low temperature (∼180 °C) ALD process and low temperature (∼200 °C) reduction processes to grow impurity-free and ultrathin cobalt metal films that are continuous on oxide substrates.

Four-inch wafers of SiO₂/Si(001) were prepared by a thermal oxidation method, which produces an amorphous 300-nm-thick SiO₂ layer on Si(001) substrates. The 0.5-mm-thick wafers were then cut into 20 × 20 mm² pieces. MgO(001) 10 mm × 10 mm × 0.5 mm substrates were purchased from MTI Corporation. The substrates were ultrasonically cleaned with acetone, isopropyl alcohol, and deionized water for 5 min each, followed by UV/ozone treatment for 15 min to remove residual carbon contamination. The substrates were loaded into a UHV system that includes an X-ray photoelectron (XP) spectrometer, a molecular beam epitaxy chamber, an ALD chamber, and the reduction chamber; this system allows in situ sample transfer between chambers.⁵⁶ The ALD system used for CoO deposition is a custom-built, hot-wall, stainless steel, rectangular 20-cm-long chamber, with a reactor volume of 460 cm³.⁵⁶ Ultrahigh purity argon was used as a purge/carrier gas. The substrate temperature was monitored with a reference thermocouple in the ALD chamber that was previously calibrated against an instrumented wafer.

Bis(N-tert-butyl-N′-ethylpropionamidinato) cobalt (II) (Strem, 98%) and deionized water were used as coreactants for the ALD growth. During CoO growth, the Co precursor and water were held at 80 °C and room temperature (25 °C), respectively, while the substrate was maintained at 180 °C. Water dosing was regulated using an in-line needle valve. Each ALD cycle of CoO growth consisted of a 2 s dose of Co, a 20 s purge of Ar, a 1 s dose of H₂O, and a 20 s purge of Ar. The deposition of Al metal was performed in a DCA 600 MBE system with a base pressure of 5 × 10⁻⁹ Torr.⁵⁷ The deuterium gas (D₂) and deuterium atom reduction were conducted in a custom-built vacuum chamber. Atomic deuterium was generated by a 400-W Osram Xenophot bulb, which had part of the glass enclosure removed to expose the tungsten filament. A current of 5.2 A was supplied by a DC power supply (KEPCO, MSK10-10M) to the tungsten filament and the filament temperature was ∼1800 K, as measured by a pyrometer. At this temperature, the tungsten filament can crack molecular deuterium to generate atomic deuterium.⁵⁸ The deuterium gas (Matheson, 99.999%) pressure was controlled by a leak valve. The sample was positioned approximately 4 cm above and faced toward the tungsten filament. A pyrolytic boron nitride heater from Momentive was 2 cm above the sample. The sample temperature was monitored with a reference thermocouple in the vacuum chamber that was previously calibrated against an instrumented wafer.

X-ray reflectivity (XRR) was used to determine the CoO film thicknesses and growth rate. XRR was conducted by a Panalytical X’PERT Pro diffractometer using a sealed tube Cu Kα radiation source (λ ∼ 1.5406 Å) operating at 40 kV and 30 mA. The CoO film crystallinity and orientation were determined by in situ reflection high energy electron diffraction (RHEED). RHEED was performed using a Staib Instruments RHEED gun operated at 18 keV energy and 3° grazing incidence. The VG Scienta R3000 X-ray photoelectron spectroscopy (XPS) system with a monochromated Al Kα source at 1486.6 eV was used to determine film stoichiometry and composition, and the oxidation states of cobalt. The absolute energy scale of the analyzer on the XPS system is calibrated using a two-point measurement with the Ag 3d_5/2 core level at 368.26 eV and the Fermi edge of Ag at 0 eV. Because of sample charging, all XPS peak positions were shifted by taking the CoO O 1s elemental peak to be at 530 eV.⁵⁹ An ex situ Veeco Icon atomic force microscopy (AFM) system is used to characterize the film morphology in tapping mode with Bruker TESPA AFM tips.

To reveal the chemical composition as function of depth, we used time of flight secondary ion mass spectrometry (TOF-SIMS), a technique capable of ultrahigh elemental and surface sensitivity. The instrument employed was a TOF.SIMS 5 (ION-TOF GmbH, 2010) equipped with a pulsed (20 ns) Bi⁺ analysis ion gun (30 keV ion energy, 3.5 pA measured sample current) and a Cs⁺ (500 eV ion energy, 40 nA measured sample current) sputtering ion beam. During depth profiling, the sputtering beam was raster scanned over an area of 300 × 300 μm² while the analysis beam was raster scanning an area of 100 × 100 μm² centered within the regressing sputtered area. All detected ions had negative polarity and the vacuum was maintained at 7.5 × 10⁻¹⁰ Torr. The instrument provided a mass resolution >5000 (m/dm) for all detected masses. Due to the insulating nature of the Al₂O₃ and SiO₂ films, a constant energy (21 eV) electron beam was shot on the sample for charge compensation. In addition, Al foil strips were used across the sample surface to increase the charge collection efficiency.

Magnetic property measurements were carried out with a Quantum Design physical property measurement system combined with a vibrating sample magnetometry (VSM) option. Magnetization hysteresis loops were measured with the magnetic field applied parallel to the plane of the thin films. The samples were at room temperature (25 °C) for all measurements.

Figure 1(a) shows the Co 2p XP spectra of 4.5-nm-thick CoO films grown on SiO₂/Si and on MgO(001) at 180 °C. The binding energies of the main peaks are at 780.5 and 796.5 eV for Co 2p_3/2 and Co 2p_1/2 levels, respectively, and at 786.4 and 803.0 eV for the Co 2p_3/2 and Co 2p_1/2 for satellite peaks, respectively. The 2p binding energy position in conjunction with the very strong satellite at ∼6 eV higher binding energy is consistent with Co being in the +2 valence state with high spin.⁶⁰ Figures 1(b) and 1(c) show RHEED images of CoO thin films grown on SiO₂/Si and MgO(001) substrates, respectively. The centered-ring patterns [as shown in Fig. 1(b)] indicate that the CoO thin films grown on SiO₂/Si have polycrystalline microstructure. The RHEED images in Fig. 1(c) show clear dotted streaks, which means that the CoO films grown on single crystal MgO(001) are epitaxial with the underlying substrate and have some surface roughness. The growth rate of CoO was found to be ∼0.3 Å/cycle on both SiO₂/Si and MgO(001) substrates.

The formation of CoO films on SiO₂/Si and MgO(001) is observed over the 170–270 °C temperature range. At temperatures higher than 305 °C, cobalt metal formation was observed instead of CoO along with carbon incorporation into the cobalt metal. Figure 2(a) displays Co 2p XP spectra of 200-cycle CoO films grown at 270, 305, and 345 °C. While the film grown at 270 °C shows the presence of CoO, the films grown at 305 °C and 345 °C clearly show the presence of Co metal. Figures 2(b) and 2(c) represent RHEED images of Co metal grown on the MgO(001) substrate at 345 °C and 305 °C, respectively. The ring-centered patterns shown in Figs. 2(b) and 2(c) indicate that the thin Co films grown at 305 °C and 345 °C are polycrystalline on MgO(001). However, the spotted patterns are well ordered, meaning that the Co films have a certain orientation preference.

Figure 3 presents the C 1s spectra for films deposited at 270–345 °C. There is ∼31% carbon incorporated into the 305 °C and 345 °C cobalt films as estimated using XPS (Fig. 3). The C 1s XPS intensity remains the same after a 15 min vacuum anneal at 500 °C [as shown in Fig. 3(b)]. This is different from the residual C (<2%) detected on the surface of as-deposited CoO films, which disappears after a slight vacuum anneal, as shown in Fig. 3(a). Carbon incorporation into Co films has also been reported using the same precursor [bis(N-tert-butyl-N′-ethylpropionamidinato) cobalt (II)] and hydrogen as a reducing agent.^45,46

Films deposited at 180 °C, a temperature that leads to CoO, do not have hydroxyl groups incorporated in or on the films. Figure 4 presents a representative O 1s XP spectrum of the ALD CoO film grown on MgO(001) at 180 °C. The O 1s peaks at 530 and 529.5 eV for CoO and MgO, respectively, are consistent with the literature.^59,61 There is no indication of a peak around 532 eV associated with OH that other groups have reported following CoO deposition.^9,62

CoO films were reduced using either deuterium gas, atomic deuterium, or a thin layer of Al that acted as an oxygen scavenging layer. The reduction temperature has a significant impact on the resulting film morphology and dewetting is mitigated at the lower temperatures needed with atomic deuterium or Al. Figure 5 shows the Co 2p XP spectrum of a Co film after deuterium gas reduction of 4.5-nm-thick CoO grown on SiO₂/Si. The D₂ reduction was conducted at 420 °C with $PD2=1×10−3Torr$ for 30 min. 420 °C is approximately the threshold temperature at which the reduction occurs in our experimental apparatus and at our D₂ pressure. The absolute D₂ (or H₂) pressure may be a factor as Väyrynen et al. have reported CoO reduction to Co at 250 °C using 10% forming gas at approximately atmospheric pressure.⁹ After reduction, the binding energies of the main peaks are 778.1 and 793.2 eV for Co 2p_3/2 and Co 2p_1/2 levels, respectively, which agrees with the peaks of Co metal.⁶³ There is a shoulder peak at 780.5 eV, which indicates there is still some CoO in the film. Based on peak fitting of the Co 2p_3/2 peak against that of CoO and metallic Co, the film contains 98% Co and 2% CoO.

The D₂-reduced Co film reoxidized after exposure to the ambient environment as shown in Fig. 6(a). The film that was 98% metallic cobalt before removing the film from the UHV system reoxidized after 4 h of air exposure, as confirmed by the shift of Co 2p_3/2 main peak from 778.1 to 780.5 eV. In an effort to protect the cobalt from reoxidizing, we deposited a 2-nm-thick Al capping layer on a D₂ gas-reduced 4.5-nm-thick CoO sample using the MBE system. The metallic Al layer (∼73 eV) oxidized to an Al₂O₃ (∼75 eV) capping layer upon exposure to the ambient environment, as shown in Fig. 6(b). This Al₂O₃ capping layer is effective in protecting the reduced Co from reoxidizing. As shown in Fig. 6(c), the capped cobalt was still predominantly composed of metallic cobalt after 48 h of air exposure.

The morphology of cobalt layers was studied with AFM and these measurements required the samples to be removed from the UHV system. Figure 7(a) presents AFM scans of a 4.5-nm-thick CoO film on SiO₂/Si and Figs. 7(b) and 7(c) present the changes to the sample after reduction with D₂. Figure 7(b) corresponds to a reduced CoO film that was exposed to ambient after which it reoxidized to CoO. Figure 7(c) corresponds to a reduced CoO film that had 2 nm of Al deposited upon it; the Al oxidized to Al₂O₃ and the cobalt remained reduced beneath the Al₂O₃. The reoxidized CoO has a root mean squared roughness (R_rms) of 3.66 nm [Fig. 7(b)] and the Al₂O₃-capped, reduced cobalt has a R_rms of 3.50 nm [Fig. 7(c)]. Figures 7(b) and 7(c) are presented to illustrate that morphology changes upon reduction are reasonably reflected in a reoxidized layer and the additional step of depositing 2 nm of Al may be unnecessary to study the layer morphology changes.

The CoO films have a typical roughness of 0.4–0.5 nm on SiO₂/Si and MgO(001). The R_rms of the SiO₂/Si substrate is 0.35 nm and the CoO film [Fig. 7(a)] has a roughness of 0.42 nm. Upon reduction at 420 °C, the roughness increased substantially to 3.50 nm and a discontinuous Co layer formed. This dewetting phenomenon is common in high temperature processing of metals,^52–55 and it is undesired when continuous Co films are required.

To attenuate dewetting and produce smoother and continuous Co metal films, we studied two low temperature reduction processes, D atom reduction, and use of an O-scavenging layer. Atomic deuterium is more reactive than D₂ gas, therefore in principle, the threshold temperature for CoO reduction by D atoms should be lower than that by D₂ gas.⁶⁴ We performed a series of CoO reduction studies with D atom exposure at different temperatures with $PD2=1×10−5Torr$ for 30 min and found that the threshold temperature is about 220 °C in our experimental apparatus. Figure 8(a) shows the Co 2p XP spectrum of the Co film after D atom reduction at 220 °C of a 4.5-nm-thick CoO film grown on SiO₂/Si. The binding energies of the main peaks are 778.1 and 793.2 eV for the Co 2p_3/2 and Co 2p_1/2 levels, respectively, and there is no shoulder peak at 780.5 eV, which means 100% CoO has been reduced to Co metal. Figure 8(b) shows the C 1s XP spectrum for the Co metal film produced by D atom reduction, from which we can conclude that there is no carbon incorporation in the film.

The density difference between CoO and Co leads to a thickness contraction of 26.3% upon reducing CoO to Co if the film contracts only in the out-of-plane direction. A continuous, uniformly thick, and pinhole-free Co layer would be ∼3.3-nm-thick after complete reduction of a 4.5-nm-thick CoO film if it only contracted perpendicular to the surface. We consider Co films to be continuous when the R_rms does not exceed 1 nm, which is about one-third of the expected thickness, and the R_rms is not greater than approximately two times the CoO film R_rms. Figure 8(c) is the AFM image of the uncapped Co metal film produced by D atom reduction; the R_rms of this film is 0.70 nm after it reoxidized. We also deposited a 2 nm Al capping layer on a D atom-reduced 4.5-nm-thick CoO sample. Figure 7(d) is the AFM image of an Al₂O₃ (2 nm Al)/Co (reduced by D atom)/SiO₂/Si sample; the R_rms of this film is 0.86 nm. Since the R_rms changes from 0.42 nm for CoO to 0.86 nm for Co, this Co metal film is likely continuous.

Similar morphology results are found upon reduction of CoO on MgO(001) substrates with D₂ or atomic D. The MgO(001) substrate has an R_rms of 0.18 nm. Figure 9(a) represents the morphology of as-deposited 4.5-nm-thick CoO films on MgO(001) and the film roughness is 0.52 nm. After a 30 min D₂ gas reduction at 500 °C with $PD2=1×10−3Torr$⁠, the AFM results in Fig. 9(b) (R_rms of 2.03 nm) reveal the dewetting problem with the reoxidized Co film featuring void regions, while after a 30 min D atom reduction at 250 °C with $PD2=1×10−5Torr$ to produce a Co film, the resulting reoxidized Co film is continuous with a roughness at 0.90 nm, as shown in Fig. 9(c).

We also explored a second, low-temperature method to reduce ALD-grown CoO films to Co. We capped the CoO with metals that have a high affinity for oxidation, such as Al, which reacted with oxygen in CoO and formed a capping oxide on the Co metal film. A 4.5-nm-thick CoO film was grown on the SiO₂/Si substrate by ALD at 180 °C and the sample was then transferred in situ to the MBE chamber where a 3-nm-thick Al metal film was deposited on it at the substrate temperature of 200 °C. The choice of 3 nm for the Al layer thickness is based on complete reduction of CoO and to ensure the Al₂O₃ layer would effectively prevent any underlying Co from reoxidizing.

The Co 2p XP spectrum after capping CoO with Al at 200 °C is shown in Fig. 10(a). The binding energies of the Co 2p_3/2 peak at 778.1 eV and the Co 2p_1/2 peak at 793.2 eV confirm a metallic cobalt formed from the cobalt oxide.⁶⁰ This suggests that the Al reacts with CoO to form Al₂O₃ and Co. The Gibbs free energy ΔG_r⁰ (298 K) of the reaction between 3 mol CoO and 2 mol Al is −939.8 kJ/mol,⁶⁵ which means that the reaction is favored even at room temperature. The reaction between Al and CoO is also demonstrated by the Al 2p XP spectrum shown in Fig. 10(b). The small peak at a binding energy of ∼73 eV is for Al metal, while the larger peak at a binding energy of ∼75 eV is for Al₂O₃.⁶⁶ The existence of Al metal after reaction indicates the complete reduction of CoO. There is a minor peak at a binding energy of ∼73.6 eV, which agrees with the Al 2p peak position for CoAl₂O₄,⁶⁷ indicating the formation of a CoAl₂O₄ interlayer. The formation of a CoAl₂O₄ interlayer is reported when there is diffusion between Co and Al₂O₃.⁶⁸ Figure 10(c) shows that the morphology of this Al₂O₃ (3 nm Al)/Co/SiO₂/Si sample has an R_rms of 0.83 nm, from which we can infer the underlying Co film is continuous.

Figure 11 presents XP spectra for the Co 3s feature at 100.8 eV for 4.5-nm-thick CoO films on SiO₂/Si(001) substrates. Spectra are also presented after reducing films with atomic D at 220 °C [Fig. 11(a)] and D₂ at 420 °C [Fig. 11(b)]. After reduction, an additional feature appears at 103.5 eV that is associated with Si 2p from the SiO₂ substrate. The spectra in Fig. 11(a) are for reduction conditions that lead to R_rms < 1 nm. The Co 3s signal dominates the substrate Si 2p signal before reduction in both cases, which is consistent with electron effective attenuation length estimates of 4.5 nm CoO attenuating ∼94% of the underlying substrate signal.⁶⁹ Assuming contraction of the CoO to Co metal perpendicular to the substrate, a 3.3 nm Co metal film would attenuate ∼92% of the underlying substrate signal. After reduction, the substrate Si 2p signal appears at 103.5 eV in both cases, while the Co 3s metal peak appears at 100.8 eV. In both reduction spectra, the Si 2p signal is greater than what is expected for a uniformly thick 3.3 nm Co metal film. The relative intensity of the Si 2p signal compared to the Co 3s signal is larger for D₂ gas reduction, while the D atom reduction has a lower intensity. Combining these results with TOF-SIMS results discussed below indicates the D atom reduction generated a thin, nonuniform Co metal layer on the SiO₂/Si(001) substrate. The increase in Si 2p signal intensity in Fig. 11(b) strongly suggests voids formed in the Co metal film reduced in D₂ gas at 420 °C. It was not possible within the body of work to determine if a continuous Co network formed that did not fully cover the SiO₂/Si(001) substrate in a manner similar to 16 nm Co films reported on TiN,⁹ or if discontinuous Co islands formed following 420 °C reduction.

Previous work has shown that CoO ALD films grown at 170–180 °C from cobalt bis(diisopropylacetamidinate) were epitaxial and crystalline on SrTiO₃-templated Si(001) and polycrystalline on anatase TiO₂.² AFM and cross-sectional scanning transmission electron microscopy revealed that surface roughness increased with film thickness and that the films were dense and pinhole-free with sharp interfaces to the substrate.² The 4.5-nm-thick CoO films reported herein have R_rms values of ∼0.5 nm, considerably less than the 0.92 R_rms value for 8-nm-thick CoO on SrTiO₃-templated Si(001). Given the absence of an identifiable Si 2p feature in Figs. 11(a) and 11(b) prior to reduction, and by extrapolation of the results reported in Ref. 2, it is reasonable to conclude the 4.5-nm-thick CoO films are smooth, continuous, and void-free.

The issue of continuous versus discontinuous Co layers forming during CoO reduction was examined with TOF-SIMS. A 4.5 nm CoO film was reduced using atomic D at 220 °C for 60 min at $PD2=1×10−6Torr$ and capped with 2 nm of Al to prevent reoxidation of the Co after removing the sample from the growth system. (Note this reduction condition is for a longer time and at a lower pressure than the D-atom-reduced results presented herein.) This film has a R_rms of 0.98 nm. Continuity of the Co metal film can be probed by monitoring when the Al₂O₃ capping layer disappears relative to where the Co layer disappears. The depth profiles (as reflected with the sputtering time) of the species of interest are shown in Fig. 12. Presenting a strong signal at the surface, Al₂O₃ was selected as a proxy for the Al film, thus demonstrating the full Al layer oxidation. Co and Si were selected as markers for the Co and SiO₂ layers, respectively. Al₂O₃⁻ and Co⁻ secondary ions were chosen as the TOF-SIMS species markers for Al and Co, respectively, due to the lack of nearby peaks that might have convoluted the ion signal. The log representation of the profiles shows intermixing between the Al₂O₃ and Co layers, which proves that there is a discontinuity in the Co layer. Depth profiling revealed the Al₂O₃⁻ signal attenuated faster than the Co⁻ signal. Further, the Co⁻ signal remains past the sputtering time the Al₂O₃⁻ signal reaches very low values. Thus, the Co metal film is continuous, but nonuniform in thickness. Additional deconvolution of the layer corrugation and sputtering effects is further required to comment on the Co film nonuniform thickness (ongoing work). Based on the TOF-SIMS and XPS results, we suggest 4.5 nm CoO films that are reduced under mild conditions and generate films with R_rms ≤ 1 nm are continuous and fully cover the substrate surface.

Motivated by potential use of Co layers in devices, such as the MTJ, we explored the magnetic characteristics of the Co layers that were formed under conditions that led to excessive dewetting, i.e., high temperature reduction with D₂, and that were designed to minimize dewetting, i.e., reduction with atomic D or an Al capping layer. Figure 13 displays the magnetization hysteresis loops of three samples as a function of magnetic field applied parallel to the substrate surface. All the measurements were conducted within one day after the samples were removed from the UHV system and the Co films stayed reduced when they were measured.

Sample A is a 4.5 nm CoO film grown on SiO₂/Si and then reduced by D₂ gas at 420 °C and finally covered by a 2 nm Al capping layer. Sample B is a 4.5 nm CoO film grown on SiO₂/Si and then reduced by atomic D at 220 °C and finally covered by a 2 nm Al capping layer. Sample C is a 4.5 nm CoO film grown on SiO₂/Si and then reduced by a 3 nm Al capping layer at 200 °C. The measured coercivities of samples A, B, and C are 480, 90, and 20 Oe, respectively. The two samples, B and C, produced by low temperature reduction processes that lead to R_rms ≤ 1 nm have much smaller coercivity than sample A, which is produced by high temperature D₂ gas reduction. This may be explained by microstructure differences between the continuous nature of Co films produced by the low temperature reduction processes and the nature of the Co films produced by the high temperature reduction process.⁷⁰ The difference between the coercivity of the Co film produced by deuterium atom reduction and the coercivity of the cobalt film produced by an oxygen scavenger layer reduction may come from the presence of a potentially diffusion-induced Co_xAl_yO_z interlayer. The deterioration of the coercivity due to the diffusion-induced interface layer is demonstrated in a hard/soft multilayer in literature.⁷¹ 

We report the growth of carbon-free CoO on SiO₂/Si and MgO(001) substrates by ALD in a temperature range of 180–270 °C. While the CoO films grown on SiO₂/Si are polycrystalline, the CoO films grown on single crystal MgO(001) are crystalline and epitaxial. High temperature thermal reduction of CoO causes severe dewetting. We demonstrate the reduction of CoO to form smooth and continuous carbon-free Co metal films by two low temperature processes, using a deuterium atom source and using an Al metal layer as oxygen scavenger. The Co films produced by the low temperature reduction processes have smaller coercivity than the cobalt films produced by the high temperature thermal reduction. These results provide a process to form Co metal films from ALD-grown CoO and indicate routes to tune the microstructure and magnetic properties through the reduction method and temperature.

This work was supported by the National Science Foundation (NSF) (Grant No. EEC-1160494). The magnetic characterization was supported by the NSF (No. NNCI-1542159) at the Texas Nanofabrication Facility. The authors wish to thank A. Dolocan for assistance with the TOF-SIMS analysis.