Area-selective atomic layer deposition of cobalt oxide to generate patterned cobalt films

As the critical dimension of electronics approach sub-10 nm scales, current patterning techniques enabled by lithographic and etch processes are experiencing more and more challenges.^1–4 The development of advanced lithographic processes, such as extreme ultraviolet lithography, and advanced etch processes, such as multiple patterning, are enabling further scaling of device dimensions.^5–10 A disadvantage with these advanced technologies is their high cost.¹¹ At the same time, the development of alternate patterning techniques, like bottom-up patterning, is drawing increased attention. Bottom-up patterning utilizes interactions between molecules to assemble nanostructures thereby reducing manufacturing costs.^12–14 One example of bottom-up patterning is area-selective deposition (ASD).^15–19 ASD is based on the principle that film nucleation and growth only happen at reactive sites rather than inert sites and employs specific surface chemistries to deposit material on desired regions of substrates selectively.²⁰ This selectivity gives ASD the potential to realize self-alignment while reducing lithography and post-deposition etch steps.¹⁶ 

Atomic layer deposition (ALD) is a self-limiting chemical reaction process, in which individual reactants saturate and are purged out of the reactor in sequence.²¹ Due to the self-limiting behavior, ALD can produce highly uniform and conformal films over three-dimensional structures with precise thickness control, which meets the demands of sub-10 nm node fabrication.²¹ Therefore, ALD is a promising approach to achieve ASD and there have been increasing reports on area-selective atomic layer deposition (AS-ALD).^{15–17,19,22}

In AS-ALD, there are regions of different reactivities toward the ALD precursors on prepatterned substrates. ALD precursors adsorb on the reactive regions and films grow there. Area-activation and area-deactivation are the two common ways to create regions of different reactivities on substrates toward the ALD precursors.²³ In area-activation, desired regions of non-reactive substrates are functionalized by reactive groups, while in area-deactivation, desired regions of reactive substrates are passivated by non-reactive groups.^24,25

The selection of, and patterning of, passivation materials is the key to AS-ALD enabled by area-deactivation.²³ Self-assembled monolayers (SAMs), such as octadecyltrichlorosilane, are the most-widely used passivation materials in current ASD studies due to their proven passivation effectiveness and self-alignment behavior to certain surface sites.^26–31 But SAMs have their limitations. First, to produce a defect-free SAM layer that can block nucleation and growth, extended SAM deposition times as long as 24 h may be required.^29,32 Besides this, SAMs are just monolayers and if a thick ALD film is needed, it can grow up and beyond the feature boundary, preventing desired feature formation. SAMs have the potential to be degraded or even damaged and so lose their passivation capabilities during repeated ALD growth cycles, which leads to SAM failure in ASD of thick films.^33,34 Additionally, SAM patterning approaches have not been fully developed. While micropatterns can be generated by methods like microcontact printing, nanopatterning of SAMs is still difficult.^24,27

To overcome these limitations of SAMs and target the applications that SAMs cannot meet, considerable research attention has focused on another group of passivation materials—polymer films.^19,35–39 Defect-free polymer films can be coated onto substrates within a few minutes, and polymer film thickness can be tuned from nanometers to micrometers enabling ASD of thick films. Another important advantage is the multitude of patterning approaches for polymer films producing both micropatterns and nanopatterns, including photolithography and directed self-assembly.⁴⁰ A number of polymers have been identified as effective growth inhibitors to certain materials, such as polyimide and polymethacrylamide to platinum deposition,^25,39 poly(vinyl pyrrolidone) to noble metal deposition,³⁸ and poly(methylmethacrylate) to TiO₂ deposition.¹⁹ 

However, there are very limited reports exploring the feasibility of using polystyrene as a passivation material for ALD. Materials whose nucleation and growth are facile on surfaces featuring hydroxyl groups should not grow on a PS film as it consists of only carbon and hydrogen. PS films also possess no glass transition temperature after UV-induced crosslinking,⁴¹ enabling PS films to work over a wide temperature window (<240 °C). Besides, as a photosensitive polymer and a common component of copolymers, PS can be patterned by photolithography and directed self-assembly. These strengths make PS films a promising passivation material candidate.

Due to their various properties, cobalt oxides and cobalt metal find applications in many fields. With a bandgap of 2.4 eV, cobalt (II) oxide has been used for photoelectrochemical water oxidation.⁴² Cobalt spinel has a bandgap of 1.6 eV and has been applied in sensors, spintronics, and catalysis.^43–47 Due to its superior wetting behavior on copper, cobalt metal has been studied as a Cu electromigration barrier and a next-generation liner material in the back end of line (BEOL) process.^48,49 As ferromagnetic materials, Co and Co alloys like CoFe and CoFeB have been studied as the fixed and free layer for magnetic tunneling junction (MTJ) devices.^50–54 

In previous work, we demonstrated a low temperature ALD approach to growing carbon-free CoO film and the methods for transforming CoO to Co films.⁵⁵ Other groups have also reported ALD growth of cobalt and nickel oxides and their subsequent reduction to metal layers.^56–58 The ALD growth temperature for CoO is 170–180 °C, and polystyrene films are stable at this temperature. In this work, we report the AS-ALD of cobalt (II) oxide and the further reduction to cobalt metal by using polystyrene as a passivation material. The polystyrene is patterned using UV-crosslinking and using self-assembly of polystyrene-block-polymethylmethacrylate (PS-b-PMMA).

Four-in. wafers of SiO₂/Si(001) were prepared by a thermal oxidation method, which produced an amorphous 300 nm-thick SiO₂ layer on the Si substrate. The 0.5 mm-thick wafers were then cut into 20 × 20 mm² pieces. MgO(001) substrates of 10 mm × 10 mm × 0.5 mm are 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.

Polystyrene (purchased from Aldrich with average Mw = 200 000) was dissolved in propylene glycol methyl ether acetate (Sigma-Aldrich, 99.5%) to prepare a 2 wt. % polystyrene solution. The 2 wt. % polystyrene solution was spincoated onto the cleaned SiO₂/Si or MgO(001) substrates to form a 40 nm-thick polystyrene film on the substrates. The PS/SiO₂/Si or PS/MgO(001) samples were then exposed to 185 and 254 nm ultraviolet (UV) light in an ambient or a N₂ environment with a shadow mask on the top of the samples. The shadow mask was a slim-bar 1000 mesh copper TEM grid with a pitch of 25 μm, bar width of 6 μm and hole width of 19 μm. The PS regions that were not shadowed by the mask were exposed to UV light and crosslinked while the PS regions that were shadowed by the mask were not crosslinked. After UV-light exposure, the samples were rinsed by toluene (Fisher Chemical, 99.9%) to remove the PS regions that were not crosslinked and cleaned by deionized (DI) water to form PS micropatterns on the substrates.

PS nanopatterns were formed by using self-assembled PMMA-b-PS diblock copolymers. Random copolymer poly(styrene-co-methyl methacrylate) terminated with an α- hydroxyl-ω-tempo moiety was purchased from Polymer Source. The average molecular weight (Mn) of the random copolymer was 6400 with 59.5 mol. % styrene. The random copolymer was dissolved in toluene to prepare a 1 wt. % random copolymer solution. Poly(styrene-b-methyl methacrylate) diblock copolymer was also purchased from Polymer Source. The Mn of styrene was 55 000, and the Mn of methyl methacrylate was 22 000. The diblock copolymer was dissolved in toluene to prepare a 1 wt. % diblock copolymer solution. The 1 wt. % random copolymer solution was first spincoated onto the cleaned SiO₂/Si or MgO(001) substrates to form a 50 nm-thick polymer layer. Then the samples were heated on a hot plate at 230 °C for 10 min and this resulted in an approximately 3 nm-thick brush layer being end-grafted onto the substrates. After rinsing the samples with toluene to remove the non-grafted random copolymer, only the 3 nm-thick end-grafted brush layer remained on the substrates. The 1 wt. % diblock copolymer solution was then spincoated onto the brush-layer-coated substrates to form a 50 nm-thick polymer layer. The samples were heated on a hot plate at 230 °C for 5 min during which PMMA formed cylinders aligned perpendicular to the substrate and PS formed a continuous matrix. The PMMA cylinders and underlying brush layer were selectively removed by CO₂/Ar-based plasma etch to generate nanopatterns of PS with 20–30 nm holes and 40–50 nm pitch on the substrates. All plasma etch tests were performed in a commercial, 300 mm, capacitively coupled plasma reactor (Lam Research Flex^® Series dielectric etch system). Prior to all plasma etch tests, sample coupons (∼600 mm²) were thermally pasted (Type 120 silicone, Wakefield Solutions) onto the center of 300 mm ArF resist carrier wafers.

After pretreatment, the substrates were loaded into the ultra-high vacuum system and then transferred in situ to the ALD chamber for CoO deposition. The CoO deposition details have been described elsewhere.⁵⁵ Bis(N-tert butyl, N′-ethylpropionamidinato) cobalt (II) was used as the cobalt source and water as oxidant source. Each cycle consisted of a 2 s dose of the cobalt precursor, a 20 s purge of Ar, a 1 s dose of H₂O, and a 20 s purge of Ar. Experiments were not performed to establish the dose time that was sufficient for saturation coverage at 180 °C. After CoO growth, the samples were removed from the growth system and O₂ plasma etching was employed to remove the remaining PS from the samples.

Aluminum metal deposition was performed on some samples in a DCA 600 MBE system with a base pressure of 5 × 10⁻⁹ Torr. Al reacted with air after unloading samples to the atmosphere and formed Al₂O₃ as a capping layer on metallic cobalt films for ex situ characterization. The deuterium atom reduction of CoO was conducted in a custom-built vacuum chamber. The deuterium reduction was performed by dissociating D₂ over a tungsten filament at a D₂ pressure of 10⁻⁵ Torr such that a mixture of D₂ and D were incident on the CoO surface.⁵⁵ 

An in situ 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 of the XPS system is calibrated using a two-point measurement such that the Ag 3d_5/2 core level is at 368.26 eV and the Fermi edge of Ag is at 0 eV. XPS peak positions for the samples were shifted by taking the CoO O 1s elemental peak to be at 530 eV.⁵⁹ Ex situ scanning electron microscopy (SEM) and atomic force microscopy (AFM) were used to characterize the metal oxide and metal patterns. SEM was conducted by a ZEISS Neon 40 SEM equipped with Bruker EDS. AFM was conducted by a Veeco Icon AFM system at tapping mode with Bruker TESPA AFM tips.

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

To determine the passivation capability of polystyrene for inhibiting CoO nucleation and film growth of CoO on oxide substrates, 40 nm-thick PS film was spincoated onto SiO₂/Si(001) and MgO(001) substrates. The PS/SiO₂/Si and PS/MgO substrates were subjected to 300 CoO ALD cycles. Figure 1 shows the Co 2p x-ray photoelectron (XP) spectrum of the PS/SiO₂/Si sample after 300 CoO ALD cycles and there is no Co signal. 300 CoO ALD cycles should lead to 9 nm of CoO on SiO₂.⁵⁵ 

The oxide substrate reactivity toward the CoO ALD process was unaffected by having PS applied to it and subsequently removed from it with toluene. Figure 1 also presents the Co 2p XP spectrum after 300 ALD cycles on a SiO₂/Si sample that had the 40 nm-thick PS removed with toluene. 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 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 The resulting film thickness is about 9 nm, which is the thickness found for 300 ALD cycles on clean, untreated SiO₂/Si substrates.⁵⁵ 

The passivation capability of UV-induced, crosslinked PS was explored by exposing PS/SiO₂/Si to UV light for 1 h in a N₂ environment. After the UV-light exposure, the sample was rinsed with toluene and there was no thickness decrease of the PS film after the toluene rinse, indicating the 1 h of UV-light exposure successfully crosslinked the PS film and made the film resistant to the toluene solvent. 300 CoO ALD cycles were performed on the crosslinked PS/SiO₂/Si sample. Figure 1 shows the Co 2p XP spectrum after the CoO ALD cycles. Compared to the as-coated PS sample, there is a weak, very noisy signal as shown in the inset of Fig. 1. This noisy signal may come from a small amount of the ALD precursor that adsorbed on the crosslinked PS surface as discussed later. These results demonstrate that both as-coated PS and crosslinked PS are very effective in blocking CoO ALD on oxide substrates.

Micropatterns and nanopatterns of polystyrene were prepared to demonstrate the area-selective deposition of CoO on oxide substrates. PS micropatterns were created by photolithography. PS is sensitive to deep UV light and can work like a photoresist. Figure 2(a) shows the slim-bar 1000 mesh copper TEM grid shadow mask used in this work. During UV-light exposure, the PS regions that were not shadowed by the mask were exposed to UV light and crosslinked and the PS regions shadowed by the mask were not crosslinked. After UV-light exposure, the samples were rinsed with toluene to remove the non-crosslinked PS. A key parameter determining the quality of PS patterns is the crosslinking environment. Figures 2(b) and 2(c) represent the AFM images of PS patterns on a SiO₂/Si substrate created by exposing PS/SiO₂/Si to UV light for 1 h in an ambient environment and in a N₂ environment, respectively, followed by the toluene rinse. The squares are PS and the bars among the squares are the underlying SiO₂/Si substrate. The PS patterns exposed in an ambient environment exhibited severe damage and the PS squares featured sharp edges and hollow centers. On the other hand, the PS patterns exposed in a N₂ environment were uniform squares. The difference comes from the presence of O₂ during the UV-light exposure.

Polystyrene can undergo side scission reactions in the presence of O₂ under UV exposure that, as illustrated in Fig. 3, damages the film.⁶¹ These scission reactions are expected to incorporate oxygen into the hydrocarbons that remain in or on the PS film. At the same time, O₂ could be converted to O₃ by UV-light exposure and O₃ could also oxidize the PS surface. Therefore, in order to produce high-quality crosslinked PS without side reactions and pattern degradation, the UV-light exposure should be conducted in an oxygen-free environment. Even under N₂ purge, we cannot guarantee that there was a completely oxygen-free environment in our UV illumination experimental apparatus. The possibility of UV-based PS degradation reactions could explain the weak Co 2p signal in Fig. 1 (inset).

Figure 4 shows the SEM and AFM images of the PS micropatterns used in this work to guide the AS-ALD of CoO. 12 nm-thick CoO was grown on the PS template/SiO₂/Si by ALD at 180 °C. The relatively dark regions in Fig. 4(a) are the PS squares while the relatively bright regions are the SiO₂/Si substrate because SiO₂/Si is more conductive than PS. The relatively bright regions in Fig. 4(b) are the PS squares while the relatively dark regions belong to the SiO₂/Si substrate.

The PS template was then removed by O₂ plasma etching. Figures 5(a) and 5(b) show the SEM and AFM images of the CoO patterns on the SiO₂/Si substrate, respectively, after the PS template was removed. In Fig. 5(a), the relatively dark regions are the CoO bars while the relatively bright regions belong to the SiO₂/Si substrate. In Fig. 5(b), the relatively bright regions are the CoO bars, which are higher than the SiO₂/Si substrate, while the relatively dark regions belong to the SiO₂/Si substrate. CoO only grew on the regions that were not covered by the PS squares shown in Fig. 4. The CoO bar patterns are uniform across the patterned area and their feature size (∼6.5 μm wide) is close to the original feature size (6 μm) of the shadow mask. The CoO patterns were confirmed by elemental analysis as shown in Fig. 6. Figures 6(a) and 6(b) are the elemental map and line analysis of the CoO patterns, respectively, in which the Co element was scanned; both indicate the Co element is only present in the bar regions.

The CoO was reduced to Co by 2 h atomic deuterium reduction at 220 °C followed by a 2 nm Al capping layer as described elsewhere.⁵⁵ Upon exposure to the ambient environment, the Al layer oxidized to form an Al₂O₃ capping layer protecting metallic Co from reoxidizing. Figures 5(c) and 5(d) show the SEM and AFM images of the Al₂O₃-capped Co patterns on a SiO₂/Si substrate, respectively. In Fig. 5(c), the relatively dark regions are the Al₂O₃-capped Co bars while the relatively bright regions belong to the Al₂O₃-capped SiO₂/Si substrate. The relatively bright regions are the Al₂O₃-capped Co bars while the relatively dark regions belong to the Al₂O₃-capped SiO₂/Si substrate in Fig. 5(d). The Co patterns are almost the same as the CoO patterns in the x- and y-directions and there is no deformation caused by reduction of CoO to Co.

A uniformly thick Co film formed by reducing 4.5 nm CoO is estimated to be ∼3.3 nm thick. We have shown that films reduced for times shorter than 1 h are continuous, fully cover the substrate and are nonuniform in thickness.⁵⁵ Films reduced for times on the order of 1 h dewet from the surface and likely form a continuous network that does not fully cover the substrate. It is likely the magnetization hysteresis loop reported in Fig. 7 for Co on a patterned SiO₂/Si substrate is of a continuous Co network that does not fully cover the substrate. The hysteresis loop was measured by VSM with the magnetic field applied parallel to the plane of the thin films at room temperature (25 °C). The reduced Co pattern is ferromagnetic and with a coercivity of 440 Oe.

The coercivity of Co films is affected by many factors, such as thickness, film orientation, and grain size, and ranges from several Oe to approximately one thousand Oe.^62–65 Cobalt coercivity increases as the particle/grain size increases.⁶³ Co films formed by reducing unpatterned CoO displayed increased coercivity with increased dewetting caused during reduction of the CoO.⁵⁵ The coercivity of 440 Oe reported herein for the patterned Co lines is consistent with blanket films and indicates the AS-ALD process did not affect the film properties.

Similar results for AS-ALD of CoO and the resulting Co patterns were also realized on the MgO(001) substrate. 18 nm-thick CoO was grown on the PS template/MgO(001) by ALD at 180 °C and then the PS template was removed by O₂ plasma etching. Figures 8(a) and 8(b) show the SEM and AFM images of the CoO patterns on the MgO(001) substrate after the PS template was removed, respectively. In Fig. 8(a), the relatively dark regions are the CoO bars while the relatively bright regions belong to the MgO(001) substrate. In Fig. 8(b), the relatively bright regions are the CoO bars while the relatively dark regions belong to the MgO(001) substrate. The CoO only grew on the bar regions that were not covered by the PS squares. The CoO bar patterns are uniform across the patterned area and their feature size (∼6 μm wide) is very close to the original feature size (6 μm wide) of the shadow mask. The CoO patterns were confirmed by elemental analysis results as shown in Fig. 9. Figures 9(a) and 9(b) show elemental mapping and line analysis of the CoO patterns, respectively, in which the Co element was scanned. Both indicate that the Co element is only present on the bar regions.

The CoO on MgO(001) was reduced to Co by 2 h atomic deuterium reduction at 220 °C and the Co layer was not covered by Al capping layer in this experiment. Due to the lack of an Al capping layer, the Co lines were reoxidized to CoO during ex situ SEM and AFM characterization. There should not be much difference between the surface morphology of Co films and reoxidized CoO films.⁵⁵ Figures 8(c) and 8(d) show the SEM and AFM images of the cobalt patterns after they reoxidized upon exposure to ambient on the MgO(001) substrate. In Fig. 8(c), the relatively bright regions are the CoO bars while the relatively dark regions belong to the MgO(001) substrate. In Fig. 8(d), the relatively bright regions are the CoO bars while the relatively dark regions belong to the MgO(001) substrate. The CoO patterns are almost the same as the as-deposited CoO patterns in the x- and y-directions and there is no deformation caused by reduction and subsequently reoxidation.

PS nanopatterns were created by self-assembly of a 50 nm-thick PMMA-b-PS diblock copolymer films. PMMA formed cylinders aligned perpendicular to the SiO₂/Si substrate within the PS matrix. There was a 3 nm-thick brush layer (PMMA-PS random copolymer) between the diblock copolymer film and the SiO₂/Si substrate to neutralize the substrate. The PMMA cylinders and underlying brush layer were selectively removed by CO₂/Ar-based plasma etching to generate nanopatterns of PS with 20–30 nm holes and 40–50 nm pitch on the substrate, as shown in Fig. 10. Figures 10(a) and 10(b) are top-down and cross-section SEM images of the as-prepared PS hole-patterns, respectively. The relatively bright regions belong to the PS matrix and the relatively dark regions are the empty holes. The critical requirement for CoO ALD is a starting surface that is not covered with any residual brush layer polymer. Therefore, the complete removal of PMMA cylinders and underlying brush layer by plasma etching is the key to the successful AS-ALD of CoO within the holes. When the holes were not fully cleared, the residue of PMMA or the brush layer blocked the nucleation and film growth of CoO and no CoO was observed after 400 CoO ALD cycles (sufficient to deposit 12 nm of CoO) as shown in Fig. 10(c) after PS removal. On the other hand, when the holes were excessively over-etched, the PS matrix was also etched too much and the remaining PS matrix was very thin. In this case, there was lateral overgrowth of CoO over the top of the PS matrix after 400 CoO ALD cycles as shown in Fig. 10(d) after PS removal. When the etch was controlled within a proper over-etch window, polymer-free holes were generated while minimizing PS matrix damage. Figure 10(e) shows the top-down SEM image of the CoO dots grown in the holes after 400 CoO ALD cycles. The holes were filled with CoO. Then, the CoO dots remaining on the SiO₂/Si substrate after removing the PS matrix with an oxygen-based plasma etch are shown in Fig. 10(f).

We report the AS-ALD of cobalt (II) oxide on polystyrene-patterned SiO₂/Si and MgO(001) substrates at 180 °C. The resulting CoO patterns are carbon-free and smooth, and the CoO patterns can be further reduced to produce metallic Co patterns without deformation by using atomic deuterium reduction at 220 °C. Polystyrene is very effective in inhibiting the nucleation of film growth of CoO. The polystyrene was patterned using UV-crosslinked 40-nm PS films or using self-assembled 50-nm polystyrene-polymethylmethacrylate films. The unexposed PS in UV-crosslinked PS films was dissolved away with toluene. The PMMA component in self-assembled PS-PMMA films was removed by plasma etching to expose the underlying oxide surface. In both patterning approaches, an oxide surface was exposed upon which CoO grew.

We show PS to be a promising passivation material for further area-selective deposition studies. It can be easily patterned to realize microstructures and nanostructures by established methods. Further, uncrosslinked PS can be removed with toluene under conditions that do not damage the substrate for subsequent ALD.

This work was supported by the National Science Foundation (NSF) under Cooperative Agreement No. EEC-1160494. Tanmoy Pramanik of the S. K. Banerjee group at UT Austin is appreciated for performing the vibrating-sample magnetometer measurement. The magnetic characterization was supported by the National Science Foundation (No. NNCI-1542159), Texas Nanofabrication Facility.