Low temperature, area-selective atomic layer deposition of NiO and Ni

The deposition of morphologically continuous, high purity nickel metal has important applications in the semiconductor industry for Ni metal Schottky gate contacts^1,2 and as a precursor for NiSi Ohmic contacts.^2–4 Additionally, nickel oxide thin films are employed within resistive switching heterostructures for nonvolatile memory devices,^5,6 wide bandgap device technology,^7,8 and as hole transport layers in solar cells.^9–13 Outside the semiconductor industry, nickel oxide finds applications in chemical sensing,^14–16 improving electrode performance in batteries,^17–19 enabling supercapacitors,^20,21 metal-organic frameworks,²² catalyzing the photooxidation of water,^23,24 modulating surface hydrophilicity,²⁵ and within transparent heat mirrors to enable energy-efficient glass window coatings.²⁶ 

In recent years, the rising popularity of flexible electronics for enabling next-generation displays and sensors has created a demand for the development of fabrication processes that integrate inorganic devices with polymeric substrates. Here, also nickel oxide finds applications within device heterostructures as well as fabrication schemes such as metal-assisted chemical etching.²⁷ A variety of solution-based methods such as spray pyrolysis, sol-gel, etc., and physical vapor deposition methods such as pulsed laser deposition, sputtering, etc., have been used to deposit nickel and nickel oxide.²⁸ However, atomic layer deposition (ALD) is an attractive choice due to thermal constraints of polymeric substrates, requirements for high-throughput, large-area scalability, and nanoscale spatial control of the deposited material. Furthermore, as a chemical route of deposition, ALD is sensitive to the surface chemical composition of the substrate. Prepatterning of surface moieties then provides a means of tuning the selectivity of the deposition process, enabling bottom-up fabrication.

Thus far, a variety of precursor and coreactant combinations have been investigated for the ALD of oxides, nitrides, and sulfides of nickel. The following is a summary of the reagents used, the deposition temperature, and the material grown. NiO ALD was performed using Ni(acetylacetonate)₂ (N,N,N′,N′-tetramethyl-ethylenediamine) and ozone at 200–275 °C,²⁹ Ni(1-dimethylamino-2 methyl-2-propanolate)₂ and water at 80–240 °C,³⁰ bis(2,2,6,6-tetramethylheptane-3,5-dionato) nickel(II) and water at 230–260 °C,³¹ nickelocene and ozone/dioxygen mixture at 230 °C,³² and Ni(dmamb)₂(dmamb ¼ 1-dimethylamino-2-methyl-2-butanolate) and ozone at 110–240 °C.³³ Recently, bis(N,N′-di-tert-butylacetamidinato) nickel(II) and H₂S were used to deposit nickel sulfide between 125 and 225 °C.^34,35 Additionally, the same nickel reagent was used to deposit nickel oxide at 200 °C using water as the coreactant³⁶ and nickel nitride at 160 °C using NH₃/H₂ as the coreactant.³⁷ Finally, this same nickel reagent has been used to grow NiO at 90 °C using water as a coreactant.³⁸ Due to the wide processing window, the ability to produce a variety of nickel compounds thermally with simple coreactants and the low temperature at which deposition is enabled, we selected bis(N,N′-di-tert-butylacetamidinato) nickel(II), henceforth referred to as Ni(amd)₂ herein, as the precursor of choice for this work.

Thermal ALD of Ni⁰ has also been investigated previously. Yuan et al. have demonstrated Ni deposition within a hot-wire-assisted ALD scheme using nickelocene and NH₃ at 250 °C.^39,40 Do et al. deposited Ni at 200–300 °C using bis(1-dimethylamino-2-methyl-2-butoxy) nickel(II) and H₂.⁴¹ Zhang et al. have shown thermal ALD of Ni using (N,N,N′,N′-tetramethylethylenediamine) [bis(2,4-pentanedionato)] nickel(II) and anhydrous hydrazine between 220 and 280 °C.⁴² Recently, Kerrigan et al. have demonstrated Ni AS-ALD using bis(1,4-di-tert-butyl-1,3-diazadienyl)nickel and tert-butylamine at 180–195 °C.⁴³ In general, compared to the deposition of nickel compounds, the thermal ALD of Ni⁰ requires higher processing temperatures and/or more reactive precursors. Because of the relatively high electropositivity of the nickel metal, nickel reagents require a strong reducing coreactant to enable metal deposition at mild temperatures.⁴³ 

Plasma-enhanced ALD (PEALD) can be utilized to increase reactivity of the reagents and lower the process temperature. Wang et al. describe a PEALD process for Ni at 160–280 °C using nickelocene and NH₃ plasma.⁴⁴ Ahn et al. demonstrated plasma-assisted deposition of the nickel metal between 190 and 250 °C using bis(1-dimethylamino-2-methyl-2-butoxy)nickel(II); both H₂ and NH₃/H₂ plasma were utilized as coreactants with the latter producing films with lower carbon content.⁴⁵ Lee et al. showed ALD of Ni between 150 and 300 °C with bis(dimethylamino-2-methyl-2-butoxo) nickel(II) and either H₂ or NH₃ plasma, with the latter yielding films of lower carbon content.⁴⁶ Park et al. demonstrated Ni film ALD between 125 and 150 °C using bis(1,4-diisopropyl-1,4-diazabutadiene)nickel(II) and NH₃ plasma, reporting up to 13% nitrogen content in the as-deposited films.⁴⁷ Xiong et al. demonstrated the growth of nickel carbide films employing H₂ plasma as the coreactant and using bis(1,4-di-tert-butyl-1,3-diazabutadienyl)nickel(II) at 95 °C.⁴⁸ Guo et al. grew nickel carbide using Ni(amd)₂ over 75–250 °C using H₂ plasma as well.⁴⁹ Due to the types of reactive species generated by the plasma, PEALD of Ni is still vulnerable to incorporating significant carbon, oxygen, and nitrogen from the precursors.

Alternatively, impurity-free nitrides and oxides of nickel may be reduced to yield pure nickel metal, though this route too requires strong reducing agents for low process temperatures. Espejo et al. synthesized NiFe thin films by first depositing Ni_xFe_1−xO_y by alternating doses of nickelocene/ozone and ferrocene/ozone and subsequently reducing the films at 400 °C under 5% H₂/95% Ar and observed significant film dewetting and void formation.⁵⁰ Lindahl et al. demonstrated the fabrication of Ni/NiO structures from the reduction of preformed Ni₃N/NiO; interestingly, they observed that at 1 Torr H₂ and at temperatures as low as 180–200 °C, the NiN₃/NiO film could form nickel metal whereas NiO uncapped by NiN₃ is not reduced under same conditions.⁵¹ Utriainen et al. reported the deposition of nickel metal using bis(acetylacetonate)nickel(II) and H₂ at 250 °C; in addition, they outline a second process under the same conditions where O₃ is instead used as a coreactant to first deposit NiO and then a subsequent 5%/95% H₂/Ar anneal at 230 °C to yield metallic films with pinhole formation.⁵² Li et al. have reported pure nickel films from H₂ annealing of preformed NiN_x films at 160 °C and H₂ plasma treatment at room temperature.³⁷ Chae et al. report the deposition of continuous Ni films at 165 °C by using an ABC-type reaction scheme involving nickelocene dose, argon purge, water dose to form NiO, argon purge, and hydrogen plasma dose to yield metal.⁵³ Due to its relatively high surface energy and diffusivity, the nickel metal may undergo dewetting⁵⁴ and interdiffusion⁵⁵ with nearby materials at higher processing temperatures.

To our knowledge, the deposition of NiO and Ni has not been investigated as part of a plasma-free process at temperatures as low as 100 °C, a low enough temperature to be compatible with roll-to-roll and other flexible electronic technology. Here, we explore a route to impurity-free, continuous Ni films by low-temperature ALD of NiO and subsequent low temperature reduction. We demonstrate that the NiO growth may be promoted on dielectric and organic substrates that normally pose significant nucleation delay by utilizing a buffer layer, suggesting a route to coating Ni on myriad substrates. Additionally, NiO can be selectively deposited against a sp² hybridized carbon-rich resist, in this work polystyrene (PS), enabling area-selective ALD (AS-ALD). Finally, we show that the preformed NiO films may be reduced using H-atoms at 100 °C, yielding ultrathin (<10 nm), continuous nickel metal films.

100 mm silicon thermal oxide wafers (University Wafer, B-doped, 10–20 Ω cm) and 200 mm SiN_x (2000 Å)/Si wafers (provided by collaborators at Raytheon Technologies) were cut into 10 × 10 mm² sample tokens and then cleaned by ultrasonically degreasing with isopropyl alcohol (99%, Fisher Scientific), acetone (99%, Fisher Scientific), and 18.2 M Ω D.I. water (Millipore) before a 15 min UV/ozone treatment (Novascan PSD series).

PS resist patterns were fabricated by a modified version of protocols described elsewhere.^56–58 First, a 2 wt. % solution of PS (M_w ca. 200 000 and 4000 Da, Sigma Aldrich) in propylene glycol methyl ether acetate (≥99.5%, Sigma Aldrich) was prepared. The solution was then spincoated (Smart Coater 100, Weinview) on clean substrates at 2000 rpm for 1 min with 200 rpm/s ramp rate, yielding uncured PS films that are ∼20 nm thick. To crosslink the PS, samples were then mounted on a 2.75 in. CF quartz optical window (2.5 mm thickness) which is then installed on a small (<250 cm³) gasket-sealed purge chamber with a gas inlet/outlet. 10 SCCM N₂ (99.999%, Airgas) flow is maintained in the purge chamber and it is placed in the UV/ozone generator. The UV light (185 and 254 nm) is turned on 15 min after the N₂ flow starts in order to purge oxygen, and the samples are then irradiated for 2 h. To generate patterned PS, a copper transmission electron microscopy (TEM) grid (1000-mesh, Ted Pella) is placed on top of the spincoated PS film to serve as a shadowmask prior to curing. Toluene (99.9%, Fisher Scientific) is then used to rinse away the uncured PS after curing, followed by an isopropyl alcohol rinse. To examine the effect of water exposure to crosslinked PS, PS was also cured in situ within an ultra-high vacuum (UHV) chamber (∼10⁻⁸ Torr) through a 2.5 mm quartz viewport using a commercial mercury lamp (25 W, 185/253.7 nm, G6 Wellness) for 2 h. Crosslinked PS was completely removed by oxygen plasma etching (Samco RIE-1C) using a 3 SCCM O₂ flow at 65 mTorr chamber pressure and 60 W RF power for 90 s.

The substrates were then placed in a UHV system that includes a hotwall ALD reactor whose body consists of a standard 2.75 in. conflat cube fitted with an eight-channel precursor manifold, a PHI model 5600 x-ray photoelectron spectrometer with a Mg Kα x-ray source and PHI 04-303A differential Ar ion gun, a custom-built annealing tube furnace, and a custom-built H-atom doser, enabling in situ sample transfer between chambers. The experiments involving SiN_x substrates and aluminum metal capping layers were performed in another similar UHV system detailed elsewhere^59,60 that also includes a hotwall ALD reactor, an in situ VG Scienta R3000 x-ray photoelectron spectrometer with a monochromated Al Kα source at 1486.6 eV, and a DCA 600 molecular beam epitaxy (MBE) system with a base pressure of 5 × 10⁻⁹ Torr.

NiO was deposited using Ni(amd)₂ (99.999%-Ni Puratrem, Strem) and de-ionized water at a baseline pressure of 1.2 Torr from 100 to 200 °C with argon (99.9999%, Praxair) carrier gas. Ni(amd)₂ was heated to 80–90 °C to facilitate sublimation and the dosing scheme used was 3 s/120 s/0.5 s/120 s Ni(amd)₂/Ar/H₂O/Ar. We found a growth per cycle (GPC) of ∼1.1 Å/cycle at 100 and 190 °C, consistent with NiO ALD GPC of 0.8 Å/cycle at 90 °C³⁸ and 1 Å/cycle at 200 °C³⁶ using this precursor and water. Beyond verifying the ALD temperature window, we found the same GPC at 175 °C when the Ni(amd)₂ exposure time was doubled to 6 s, the Ar purge times increased from 120 to 180 s, and the water exposure time was reduced from 0.5 to 0.25 s when establishing the ALD dosing scheme.

Trimethyl aluminum (TMA) (97%, Sigma Aldrich) and de-ionized water were used to deposit Al₂O₃ with a dosing scheme of 0.5 s/50 s/0.5 s/50 s TMA/Ar/H₂O/Ar. Palladium hexafluoroacetylacetonate [Pd(hfac)₂] and hydrogen(99.999% Airgas) dosed as 2 s/40 s/0.01 s/40 s Pd(hfac₎₂/Ar/H₂/Ar were used to grow Pd capping layers on reduced Ni surfaces at 100 °C.

The reduction of NiO to Ni was performed in two ways: H₂ annealing in a UHV quartz tube furnace or by H-atom exposure in a separating dosing chamber. The annealing furnace is a UHV chamber fitted with a H₂ gas inlet, a quartz tube with a ceramic heating jacket under PID control, and a magnetic arm with a sample stage to facilitate sample transfer. Samples are preheated for 15 min under UHV in the tube furnace before the chamber is backfilled with H₂ to the desired pressure for reduction. H-atom treatment is performed in a UHV chamber consisting of a sample stage with PID-controlled bulb heater for sample heating and the H-atom doser, constructed according to a design described elsewhere.⁶¹ Briefly, the doser involves a leak valve passing H₂ through a W capillary biased at +580 V whose tip is heated by e-beam bombardment from a hot W filament to a temperature sufficient to crack H₂ (1800–2000 K). Beam current was increased until W was detectable on the sample by x-ray photoelectron spectroscopy (XPS), indicating alignment of the capillary with the sample and sublimation of W from the capillary and then the beam current was slightly lowered to 15 mA at which point W was not detectable. During H-atom exposure, the sample surface is positioned 20 cm below the W capillary exit and a water-cooled copper radiation shield sits in between. Despite the shield, without stage heating, the sample temperature equilibrates to ∼70 °C under standard reduction conditions. Samples were preheated to the desired temperature for 15 min prior to H-atom exposure. The H₂ pressure between the leak valve and W capillary entrance was held at 3 × 10⁻⁵ Torr.

Films were characterized ex situ using x-ray reflectivity (XRR) (Rigaku Ultima IV Diffractometer) with GenX (v 3.6.9) to calculate film thickness and atomic force micrography (AFM) (Asylum Research) in tapping mode with Si cantilevers (HQ/NSC15/Al BS/μmasch).

NiO growth was investigated across a range of temperatures from 100 to 220 °C on SiO₂. Beyond 200 °C, films grown appeared to have a significant metal fraction and increased carbon content as detected by XPS, suggesting the decomposition of Ni(amd)₂, consistent with another report³⁸ that found an ALD window of 90–200 °C. Compared with other thermal processes, this oxide growth proceeds at a remarkably low temperature, chiefly due to the relatively high volatility and reactivity of this precursor. This makes Ni(amd)₂ particularly well suited for processing flexible substrates; for example, the glass transition temperature of polyethylene terephthalate, a typical organic roll-to-roll substrate, is ∼120 °C suggesting a maximum process temperature of ∼100 °C, which we use as our operating point herein. Figure 1(a) shows the survey spectra after 150 ALD cycles of NiO on thermal oxide at 100 °C. The C 1s peak, located at 284.7 eV, is small but apparent as shown in the high-resolution spectrum in Fig. 1(b). After 10 min in situ Ar sputtering, the carbon signal decreases significantly indicating the removal of adventitious carbon. After another 10 min sputtering, the Si 2p peak located at 103.5 eV also increases very slightly, suggesting that some nickel oxide was removed as well. Within the bulk of the film, there is essentially no detectable carbon. Thus, even at 100 °C, nickel oxide grows carbon-free as evidenced by XPS, which is critical to the purity and performance of the resulting metal after reduction.

The Ni 2p high-resolution XP spectrum shown in Fig. 2(a) indicates that the deposited nickel is in the 2+ oxidation state, as evidenced by binding energy position of the main peak Ni 2p_3/2 peak at 853.7 eV and the characteristic multiplet split structure of this peak.^62,63 The O 1s spectrum, as shown in Fig. 2(b), is comprised of two peaks; the peak at the lower binding energy of 530.1 eV corresponds to lattice oxygen in NiO. The peak at higher binding energy of 532 eV has been attributed to various species, including defective sites/vacancies or surface hydroxide.^64–66 Although the as-deposited film shows a small, higher binding energy peak, this peak significantly grows after the film is vented, suggesting a change in oxide composition upon venting. As the vacancy concentration can have a significant effect on electronic properties, modulating this parameter may be critically important when NiO is employed in semiconducting^10,67–69 or electrocatalytic⁷⁰ applications. The effects of deposition parameters on the film oxygen content warrants further investigation but are outside the scope of this study.

The growth rate and surface morphology of nickel oxide were investigated by ex situ XRR and AFM, respectively. The as-deposited NiO films tended to give poor reflectivity oscillations; as shown in Fig. 3(a), 300 ALD cycles of NiO deposited at 190 °C produced just enough of a reflectivity pattern to enable a thickness calculation, whereas 150 ALD cycles of NiO (not shown) did not. In contrast, 150 ALD cycles of Al₂O₃ deposited at 100 °C produced a clear reflectivity pattern and providing a calculated GPC of 1.2 Å/cycle was calculated, which is in agreement with literature.⁷¹ The change in the oxygen content upon exposure of nickel oxide films to ambient and the difficulty in obtaining an XRR pattern raised concerns about obtaining an accurate GPC. As Al₂O₃ ALD produces a dense oxide that can function as an effective barrier layer, a thin capping layer of 15 ALD cycles (∼1.8 nm) of Al₂O₃ was grown in situ on top of 200 ALD cycles of NiO deposited at 100 °C to protect NiO from ambient and yields a much clearer XRR pattern. Both the capped and uncapped nickel oxide films showed similar GPC, ∼1.1 Å/cycle. Additionally, the growth rate is observed to be the same at a deposition temperature of 100 and 190 °C. The observed GPC is in approximate agreement with the GPC reported in the literature for NiO ALD using Ni(amd)₂ and water, which ranges from 0.8 to 1 Å/cycle.^36,38 Furthermore, AF micrographs of uncapped NiO [Fig. 3(c)] and capped NiO [Fig. 3(d)] reveal smooth surfaces whose roughness values are comparable to that of a bare thermal oxide surface (not shown) or ALD Al₂O₃ on thermal oxide [Fig. 3(b)]. The calculated agreement of NiO GPC with literature and the atomically smooth film surface together suggest that NiO is growing two dimensionally on the well-hydroxylated thermal oxide substrate.

Some nitride and oxide dielectric surfaces may be without a high density of surface hydroxyl groups by which this ALD process propagates. Thus, there may be a significant nucleation delay if Ni(amd)₂ is unable to react with the surface species and undergo chemisorption. Modification of surface moieties by employing a buffer layer or surface treatment can overcome this limitation. Figure 4 shows the high-resolution XP spectra of a SiN_x surface, a ubiquitous dielectric material in semiconductor processing. The Si 2p peak shows two distinct features at 101.5 and 103.2 eV. The lower binding energy peak may be assigned to bulk Si₃N₄ while the higher binding energy feature likely arises from surface oxynitride SiN_xO_y,⁷² rather than a SiO₂ overlayer, considering that lattice silicon in SiO₂ has a binding energy of 103.6 eV.⁷³ The primary N 1s feature position of 397.8 eV is the characteristic of Si₃N₄⁷⁴ while the main O 1s feature is positioned at 533.1 eV, a slightly higher binding energy than that of lattice oxygen in SiO_2,⁷³ suggestive of an oxynitride environment.⁷² Silazane (Si–N–Si) bridge reconstructions on the silicon nitride surface may oxidize to Si–O–Si bridges, producing a molecular blend of silicon, oxygen, and nitrogen.⁷⁵ Si–H, N–H, and O–H bonds may be present as well depending on the method of deposition⁷⁶ and humidity during oxidation.⁷⁷ 

Figure 5(a) shows the Ni 2p XP spectrum after 300 ALD cycles of NiO on SiN_x at 190 °C. The low signal-to-noise and lack of development of peak shape suggest that the equivalent of ∼10 ALD cycles that would form on SiO₂ (∼1 nm) or less accumulated on the surface, indicating a large nucleation delay. In comparison, a strong Ni 2p peak is observed after 300 ALD cycles of NiO on SiO₂. To improve the surface hydroxyl density, the silicon nitride surface was pretreated with 3 and 6 cycles of Al₂O₃. TMA undergoes a large free energy change (ΔG = −370 kcal/mol, T = 0 °C)⁷⁸ in forming Al₂O₃, making it a very reactive precursor able to nucleate on a variety of metal⁷⁹ and polymeric⁸⁰ substrates that typically pose nucleation difficulty for oxide ALD. As shown in Fig. 5(b), aluminum is detectable by XPS after 3 and 6 ALD cycles with greater number of cycles giving a stronger Al 2p signal—it appears that Al₂O₃ readily nucleates on the nitride surface. Its peak position at ∼74.7 eV indicates that aluminum is in the 3+ oxidation state. 300 ALD cycles of NiO performed on the Al₂O₃ pretreated surfaces produce strong Ni 2p signals but not as strong as NiO grown on SiO₂. Notably, both three- and six-cycle Al₂O₃ pretreatment produced the same intensity and peak shape of Ni 2p after NiO deposition. The difference in Ni 2p intensities between the Al₂O₃-pretreated substrates and SiO₂ was not due to signal attenuation by adventitious carbon as only a trace amount of carbon was detectable by XPS for these films. Perhaps the surface hydroxyl concentration of ALD grown Al₂O₃ is less than well-hydroxylated SiO₂.

Carbonaceous substrates such a graphene and polyethylene are also important dielectric substrates for flexible electronics and lack hydroxyl groups, generally resisting oxide ALD. NiO deposition was also investigated on PS. Figure 6 shows survey XP spectra of 50 ALD cycles of NiO deposited on cured blanket PS as well as on SiO₂. Compared to growth on SiO₂, there is only a residual amount of Ni present on cured PS as evidenced by the presence of small Ni 2p and Ni LMM peaks. After six cycles of Al₂O₃ deposition on PS, Al 2p and 2s peaks are clearly visible by XPS as shown in the spectrum in Fig. 6; again, due to the particularly high reactivity of TMA, alumina can nucleate on PS, whereas nickel oxide cannot. The survey spectrum of 50 ALD cycles of NiO grown on Al₂O₃-buffered PS reveals the presence of a significant Ni 2p peak, indicating a dramatic improvement in nucleation of NiO. Comparing the Ni 2p peak intensities of SiO₂ and Al₂O₃-buffered PS, it appears that less NiO deposited on Al₂O₃-buffered PS than on well-hydroxylated SiO₂. The ability of TMA and water to deposit alumina on a variety of organic substrates has been investigated elsewhere.^81–85 TMA is able to undergo surface reactions with a variety of organic moeties such as esters, vinyls, carbonyls ethers, etc.⁸⁶ However, in the case of materials that lack reactive sites, such as polypropylene and PS, alumina growth can still take place if the substrate is sufficiently porous and able to absorb TMA and water into the near-surface region; it is then possible for alumina clusters to form in these voided regions and eventually coalesce.⁸⁰ This may serve to explain both the ability of TMA to enable NiO growth on PS as well as the discrepancy between the initial growth rate on Al₂O₃-treated PS versus well-hydroxylated SiO₂ and warrants further exploration of the effects of exposure time and crosslinking density on nucleation.

Promotion of reactive surface moieties may be achieved without deposition of an intermediate layer as well. Because of the ability of plasma to generate reactive ions, plasma-enhanced deposition of silicon nitride is capable of producing N–H and O–H species;⁷⁶ furthermore, low-temperature plasma oxidation of Si₃N₄ is capable of creating Si–OH moeties on the surface and, in the presence of water, Si-NH as well.⁸⁷ Figure 7 shows Ni 2p spectra of similar intensities for 150 ALD cycles of NiO grown on SiO₂ as well as on SiN_x treated with a short exposure to oxygen plasma at room temperature. Clearly, exposure to oxygen plasma enabled NiO ALD to proceed on SiN_x as well as it does on SiO₂. As the plasma chemistry is complicated and requires in situ characterization to elucidate mechanistic information, comprehensive investigation of plasma surface treatments is outside the scope of this work. However, the use of –OH promoting buffer layers and surface treatments enable a general pathway to deposit NiO on a wide variety of dielectric surfaces.

As PS poses a significant nucleation challenge to NiO and because it can be readily patterned by using a shadowmask and a simple UV curing process, we investigated AS-ALD against patterned PS as a resist, but commercially available resists are also applicable. Figure 8 shows Ni 2p and survey XP spectra of up to 200 ALD cycles of NiO on blanket PS that was cured ex situ. As shown in Fig. 8(a), a small amount of nickel oxide is detectable by XPS after the first 40 ALD cycles. However, with increasing number of ALD cycles, the Ni 2p peak intensity does not increase appreciably. Furthermore, nickel is detectable in the high-resolution scan but not the lower resolution survey scans. Therefore, the amount of nickel oxide deposited is residual, likely corresponding to fewer than five ALD cycles of NiO on SiO₂. On the other hand, the survey spectra reveal that the starting surface shows a significant O 1s peak, which diminishes after deposition.

PS does not contain oxygen and uncured blanket PS films show a barely detectable O 1s signal by XPS [Figs. 9(a) and 9(b)]. After a 3 h in situ UV cure (P = 3 × 10⁻⁸ Torr), there is no longer any oxygen detectable by XPS [Figs. 9(a) and 9(b)]. However, exposure to ambient conditions results in the accumulation of some oxygen species. A 200 °C cure under UHV [Fig. 9(a)] causes the oxygen signal to diminish greatly. Additionally, 25 min of Ar sputtering (10 × 10 mm raster) completely removes this accumulated oxygen species, suggesting that it is confined to the surface [Fig. 9(c)]. Furthermore, in vacuo cured, oxygen-free PS demonstrates an oxygen signal upon in situ exposure to water [Fig. 9(b)]. These findings together suggest that after crosslinking, PS is likely susceptible to water absorption. Since water is a coreactant in NiO ALD, some amount of water will inevitably accumulate in the near surface region of the PS film, resulting in a superficial, residual amount of NiO deposition.

A PS pad pattern (25 μm pitch, ∼35 nm height) was fabricated on SiO₂, as shown in Fig. 10(a), using a TEM grid as a shadowmask and Hg lamp irradiation. Cure time was varied from 60 to 300 min. The resulting morphology of the cured pattern depends on cure time, as detailed elsewhere.⁵⁶ At lower cure times, PS pads are convex and have significant scalloping at the borders and, therefore, poor line edge fidelity. As cure time increases, the shape of the pad surface tends to become concave and the height of the rim increases. At sufficiently long cure times, crosslinking begins to take place in the shadowed regions leading to pad coalescence due to the lateral diffusion of photogenerated radial species, while at the same time, PS pads degrade. Pattern inversion can also occur.⁵⁶ A cure time of 2h resulted in patterns with the best edge fidelity. Figure 10(b) shows the resulting grid line pattern after 150 ALD cycles of NiO at 100 °C and subsequent oxygen plasma etching to remove the resist pads. The height of the grid lines is approximately 8 nm; from the GPC measured on clean SiO₂, ∼12 nm deposition height is expected, suggesting a nucleation delay of ∼50 cycles, which did not occur in the case of AS-ALD of CoO using a similar process.⁵⁸ Additionally, the NiO line width is also lower than the trench width defined by the pad pattern. Descumming prior to deposition and the use of higher performance resists may ameliorate this delay and improve pattern fidelity.

Nickel oxide was reduced to nickel metal by using H₂ annealing, capping with the Al oxygen scavenging layer as well as low temperature H-atom treatment. The morphology of the resultant metal films depends critically on the process temperature and the interfacial surface energies. The thermodynamic driving force for dewetting and void formation of a material of isotropic surface energy on a rigid substrate is described by Young's equation,⁸⁸ 

where γ_s is the surface energy of the substrate, γ_i the material-substrate interfacial energy, γ_f is the material's surface energy, and θ is the equilibrium contact angle. Of course, real materials demonstrate a wide array of complex morphologies, depending on the kinetic pathways available to enable film restructuring, surface energy gradients, formation of interfacial phases, the size and evolution of grain boundaries in the metal, etc. The basic consideration, however, that a large difference between the free surface energy of the substrate and the metal film drives dewetting is a useful perspective. In the absence of an interfacial phase with the substrate, a bare Ni metal (surface free energy of ∼2.4 J m⁻²)⁸⁹ will undergo dewetting on dielectric substrates such as SiO₂ (hydrated amorphous silica has a surface free energy of ∼0.129 J m⁻²)⁹⁰ due to the relatively high surface energy of Ni. Driven by solid state diffusion, film dewetting can occur at temperatures as low as 0.2T_m, where T_m is the melting point of the pure metal (∼1453 °C for Ni).⁹¹ Furthermore, as the film thickness decreases, the onset temperature for dewetting also decreases.⁸⁸ Thus, lowering the process temperature is critical for yielding void-free Ni thin films.

The Ni 2p and Si 2p XP spectra of 300 ALD cycles (∼33 nm) of NiO (deposited at 190 °C) on SiO₂ before and after sequential H₂ reductions are shown in Figs. 11(a) and 11(b), respectively. The substrate Si 2p peak at 103.8 eV is completely attenuated by the as-deposited film. After an initial reduction at 420 °C, a signal peak emerges in the Ni 2p spectrum at 852.2 eV, indicating the formation of metallic Ni. Considering the relative unit cell sizes of NiO (the hexagonal unit cell with a primitive unit cell volume of 1.1 × 10⁻²⁸ m³)^92,93 and Ni (the fcc structure with a lattice constant of 3.5 Å, suggesting a unit cell volume of 4.3 × 10⁻²⁹ m³),⁹⁴ if film volume contraction from complete conversion of NiO to Ni occurred only in the out-of-plane direction, it would be expected that a 33 nm film may contract to ∼13 nm, a thickness that is still expected to attenuate the underlying substrate signal completely. However, a significant Si 2p peak appears after treatment, suggesting that void formation has occurred in the film. Complete conversion of the film to nickel metal is achieved after a subsequent 30 min H₂ anneal at 435 °C, which also drives the further dewetting of the film as evidenced by the increased Si 2p signal.

Lower annealing temperatures were also investigated; Figs. 11(c) and 11(d) show the Ni 2p and Si 2p spectra of thinner, 150 ALD cycles (∼16.5 nm) of NiO (deposited at 190 °C) before, after a 10 min H₂ reduction at 250 °C, and subsequently at 300 °C. After 10 min reduction at 250 °C, nickel still appears to be in the 2+ oxidation state as evidenced by the Ni 2p main peak position at 853.7 eV. A subsequent anneal at 300 °C for 10 min yields complete conversion of the nickel oxide to metal and an obvious increase in the Si 2p peak.

Complete conversion of thicker films requires higher temperatures and longer time, since there is more material to reduce. The reduction proceeds from the surface and is facilitated by the diffusive mass transport of H₂, reaction byproducts, and intermediates into the bulk. Though thinner films exhibit greater propensity to dewet, the 16.5 nm NiO film reduced at milder temperatures shows a lower Si 2p peak intensity after annealing than the 33 nm NiO film annealed at higher temperatures. Lowering the process temperature is the most critical consideration in suppressing void formation, and under these conditions, H₂ cannot reduce NiO at 250 °C and below.

Film roughness was also investigated at the mildest H₂ annealing conditions by ex situ AFM after capping the annealed films with 3 nm Pd by in situ ALD at 100 °C. A barrier layer is required to prevent the restructuring by reoxidation of the nickel metal films. ALD of palladium may be performed at a temperature as low as 80 °C using Pd(hfac)₂ and H₂ if the substrate can enable surface dissociation of H₂,⁹⁵ which Ni metal can.⁹⁶ Since ALD is capable of delivering atomically smooth and conformal films, it was used to deposit the Pd capping layers, whose morphology serves to represent the morphology of the underlying nickel metal. In separate experiments (not shown), we find that a 3 nm Pd capping layer prevented reoxidation of a fully reduced NiO film (i.e., Ni⁰) when the sample was exposed to ambient for four months demonstrating the pinhole free nature of 3 nm ALD Pd films.

Figure 12 shows the Ni 2p (a) and Si 2p (b) XP spectra of 100 ALD cycles of NiO (deposited at 190 °C) annealed under 50 m Torr H₂ at 300 °C for 30 and 5 min. By comparing the Ni 2p spectra, it is evident that the shorter annealing time resulted in partial reduction of the NiO film. Peak fitting of the Ni 2p spectrum of the partially reduced film was performed using a Shirley background and Gaussian (70%)—Lorentzian (30%) profiles, similar to what has been described elsewhere.⁶³ Two components were used to fit the Ni metal peak contribution and three components for NiO, revealing the film to be ∼26% metallic. Additionally, the Si 2p signal remains completely attenuated in the partially reduced film but appears in the case of complete reduction. Figures 12(c) and 12(d) show the corresponding AF micrographs after capping the metal films with Pd, revealing that the partially reduced film has a significantly lower surface roughness than the fully reduced film. As it is already expected that the Ni metal on SiO₂ will undergo dewetting when surface diffusion can occur, the improved adhesion of the partially reduced metal film to the SiO₂ support is likely due to the presence of an interfacial layer such as partially oxidized Ni that can bridge the bonding characteristics of the metal and the substrate. Therefore, allowing the presence of some nickel oxide at the interface may serve well to improve the wettability of the metal film.

In addition to improve the metal-substrate interfacial bonding to suppress void formation and dewetting, an oxygen scavenging capping layer may be utilized to both reduce NiO to Ni as well as constrain the top interface of the film. In pursuit of this strategy, 150 ALD cycles of NiO were performed on SiO₂ at 190 °C and capped with 3 nm Al in situ by MBE at room temperature followed by a 500 °C anneal to drive the reaction to completion. The amount of aluminum deposited is insufficient to reduce the entire volume of NiO to Ni. Figure 13(a) shows Ni 2p XP spectra before Al deposition and after annealing, revealing complete conversion of NiO to Ni within the depth of the film scanned (∼5 nm). The Al 2p spectrum shown in Fig. 10(b) consists of a main peak around ∼74.7 eV corresponding to Al₂O₃⁹⁷ and a minor peak around 72.6 eV as well as a broad peak centered around 71.4 eV. The latter two features can correspond to the formation of a nickel aluminate interfacial layer, as the Al 2p peak position in mixed nickel aluminum oxides ranges from 72 to 73.9 eV.⁹⁸ Furthermore, Fig. 13(b) also shows the Ni 3p_3/2 feature at 65.8 eV and Ni 3p_1/2 at 66.9 eV, characteristic of nickel metal.⁹⁹ Despite the relatively high annealing temperature, no underlying NiO or SiO₂ is visible by XPS, unlike the case of 435 °C H₂ reduction shown in Fig. 11(b) previously discussed. As such, we may infer that the formation of a nickel aluminate at the top interface aids in suppressing void formation.

If the use of interfacial layers is not possible, then the main strategy to suppress dewetting and void formation is substantially lowering the reduction temperature below ∼200 °C, which, in turn, requires much stronger reducing agents. Hydrogen atoms are a very reactive species and a mixture of H₂ and H can be readily thermally generated from molecular hydrogen. Here, we investigate the treatment of NiO by H-atoms. Through various reduction experiments with H₂ and H-atoms, it was established that in this dosing chamber, the onset of reduction with H₂ begins around ∼175 °C. Figure 14(a) shows the Ni 2p XP spectra of 150 ALD cycles of NiO subjected to sequential reduction at the lowest sample temperature available in this setup: 70 °C, which is the sample temperature due to proximity to the hot H-atom source when stage heating is not applied. After 5 min, a small Ni⁰ shoulder peak appears—reduction by hydrogen atoms proceeds immediately at this temperature. As the annealing time is increased to 10 and 30 min, the Ni⁰ shoulder grows significantly. Between 30 min and 5 h of H-atom treatment, reduction proceeds but appears to approach a saturation thickness. Though the H-atoms are sufficiently energetic for the reaction to proceed, for the reduction to proceed into the bulk of the film, diffusive transport of H-atoms, oxygen, and water must be enabled. Therefore, while stronger reducing agents can lower the annealing temperature significantly, the lower temperature also constrains the depth of reduction. Complete reduction may, however, be achieved for sufficiently thin films. Figure 14(b) shows the Ni 2p XP spectra of 20 ALD cycles of NiO (∼2.1 nm) before and after a 15 min H-atom treatment at 100 °C. The main Ni 2p peak position at 852.6 eV after reduction indicates complete conversion of NiO to Ni. Figure 14(c) shows the AF micrograph after the reduced film was capped with 3 nm Pd, revealing a low surface roughness, comparable to that of ALD-grown NiO, suggesting that the underlying Ni is likely continuous and void free.

Here, we investigate a carbon-free route to ALD of NiO on SiO₂, SiN_x, and PS in a temperature range of 100–200 °C. Although NiO readily nucleates on well-hydroxylated SiO₂, SiN_x and PS pose a significant nucleation challenge, which may be overcome by surface modification such as the use of Al₂O₃ seed layers and oxygen plasma treatment. Furthermore, area selective deposition of NiO against patterned PS resist is demonstrated. NiO can be reduced to Ni by H₂ annealing, but high process temperatures drive void formation in the film. Changing the interfacial surface energy by partially reducing NiO or using an Al capping layer can help suppress dewetting. Alternatively, the process temperature can be reduced below the temperature at which the onset of dewetting occurs if a strong reducing agent such as H-atoms is used. For sufficiently thin films, H-atom annealing can deliver complete conversion of NiO to Ni at 100 °C while preserving the atomic level smoothness delivered by ALD of NiO.

The authors would like to thank Shallaco McDonald and David Gray for their help in customizing and constructing the facilities used. This work was funded by the National Science Foundation (NSF) under Grant Nos. DGE-1610403 and EEC-1160494. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation. J.G.E. acknowledges the support of the Robert A. Welch Foundation.

The authors have no conflicts to disclose.

Himamshu C. Nallan: Conceptualization (equal); Data curation (lead); Formal analysis (lead); Investigation (equal); Methodology (equal); Validation (lead); Writing – original draft (lead); Writing – review & editing (equal). Xin Yang: Conceptualization (supporting); Formal analysis (supporting); Investigation (supporting); Methodology (supporting); Writing – review & editing (supporting). Brennan M. Coffey: Conceptualization (supporting); Formal analysis (supporting); Methodology (supporting); Writing – review & editing (supporting). John G. Ekerdt: Conceptualization (equal); Data curation (supporting); Formal analysis (equal); Funding acquisition (lead); Investigation (equal); Methodology (equal); Project administration (lead); Resources (lead); Supervision (lead); Validation (supporting); Writing – original draft (equal); Writing – review & editing (equal).

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