Area-Selective Atomic Layer Deposition of Metal Oxides on Noble Metals through Catalytic Oxygen Activation.

Area-selective atomic layer deposition (ALD) is envisioned to play a key role in next-generation semiconductor processing and can also provide new opportunities in the field of catalysis. In this work, we developed an approach for the area-selective deposition of metal oxides on noble metals. Using O2 gas as co-reactant, area-selective ALD has been achieved by relying on the catalytic dissociation of the oxygen molecules on the noble metal surface, while no deposition takes place on inert surfaces that do not dissociate oxygen (i.e., SiO2, Al2O3, Au). The process is demonstrated for selective deposition of iron oxide and nickel oxide on platinum and iridium substrates. Characterization by in situ spectroscopic ellipsometry, transmission electron microscopy, scanning Auger electron spectroscopy, and X-ray photoelectron spectroscopy confirms a very high degree of selectivity, with a constant ALD growth rate on the catalytic metal substrates and no deposition on inert substrates, even after 300 ALD cycles. We demonstrate the area-selective ALD approach on planar and patterned substrates and use it to prepare Pt/Fe2O3 core/shell nanoparticles. Finally, the approach is proposed to be extendable beyond the materials presented here, specifically to other metal oxide ALD processes for which the precursor requires a strong oxidizing agent for growth.

Introduction

Area-selective deposition plays an increasingly important role in the development of nanostructured materials for semiconductor processing[1,2] and catalysis.[3,4] In semiconductor processing, conventional fabrication of multilayer device structures is facing the challenge of aligning the layers with nanometer accuracy. Use of area-selective deposition allows for control over where deposition takes place without requiring photolithography for every device layer.[1] Consequently, the number of lithography steps can be reduced, which eliminates these alignment errors while lowering the fabrication costs.

In catalysis, there is a desire to create highly controlled bimetallic or core/shell nanoparticles that are monodisperse and of high purity. Conventional synthetic methods have difficulty ensuring a consistent bimetallic composition.[5−8] Large improvements can be expected if the deposition of the second material occurs selectively on the first material only, and not on the surrounding support substrate. This ensures that all particles are covered uniformly, while preventing the formation of monometallic particles of the second material.

Atomic layer deposition (ALD) utilizes self-limiting reactions of precursor and co-reactant gases to achieve highly controlled deposition. ALD has many favorable attributes including high conformality, good spatial uniformity, and Å-level thickness control. Since the chemical reactions occur only on the substrate surface, area-selective ALD can be achieved by either blocking or activating the growth on specific areas or materials. Prior reports have shown that such selectivity can be obtained by chemically modifying surfaces to prevent growth.[1,2,9−15] Alternatively, ALD growth can be activated on an inert substrate by locally catalyzing the surface reactions of an ALD process.[8,16−18] Etching can be added as part of the area-selective ALD process to improve the selectivity.[19,20]

Prior work has shown that area-selective ALD by area activation can be achieved by choosing a suitable co-reactant.[3,8,16−18,21] For example, a mildly oxidizing co-reactant such as O2 gas may allow for area-selective growth on certain surfaces, while more strongly oxidizing co-reactants such as ozone or O2 plasma generally result in growth on any surface and are therefore not suitable for area-selective deposition. Weber et al. used this approach to synthesize Pt/Pd and Pd/Pt core/shell nanoparticles supported on Al2O3 substrates.[16,18] Selective deposition of Pt on Pd particles was achieved by using O2 gas as the co-reactant. The Pd particles are able to catalyze the surface reactions of subsequent Pt ALD. More specifically, noble metal surfaces catalyze the dehydrogenation and combustion of precursor ligands.[8,16,18,22] The latter is driven by dissociative chemisorption of O2 into O* (chemisorbed O). These catalytic surface reactions allow for deposition of Pt only on the Pd particles and not on the surrounding Al2O3 substrate. In a similar way, Pd deposition on Pt particles was enabled by using H2 gas, where this time the Pt particles catalyze the formation of H* reactive sites. Lu et al. expanded this methodology by exploiting the selective chemisorption of O2 and H2 to synthesize bimetallic PdPt, RuPt, and RuPd nanoparticles on Al2O3.[8]

In this work, we extend this approach to area-selective ALD of metal oxides by demonstration of selective iron oxide and nickel oxide deposition on catalytic Pt and Ir substrates. We focus particularly on Pt-Fe based materials since this system has key applications in catalysis and magnetic devices. For example, Pt-Fe can be used for carbon nanotube growth,[23−26] the oxygen reduction reaction,[27] and room temperature CO oxidation.[28] Furthermore, the use of ALD to prepare these materials may allow them to be integrated into next-generation catalyst designs[3] and take advantage of ALD’s favorable attributes for catalyst synthesis.[3,4,21,29−32] As a result of their magnetic properties, Pt-Fe based materials have key applications for creating memory[33,34] and spintronic devices,[35] and are used in medical imaging.[36]

Area-selective ALD is achieved by exploiting the catalytic activation of O2 on Pt and Ir substrates, which enables deposition of iron oxide and nickel oxide from t-butyl ferrocene (TBF) and nickelocene precursors, respectively. These precursors are relatively unreactive and usually require strongly oxidizing co-reactants such as ozone or O2 plasma to combust the ligands and result in deposition.[37,38] However, in this work, we show that deposition can also be achieved with O2 gas when a catalytic substrate such as Pt or Ir is used. On such catalytic metals, O2 gas is catalytically activated through dissociative chemisorption (Figure ). The reactive O* species are in this case formed at the catalytic substrate instead of supplied by the ozone or O2 plasma. On the other hand, deposition does not take place on materials such as SiO2, Al2O3, and Au as these materials do not dissociate the O2 gas. In this way, by using O2 gas, area-selective deposition is achieved on the catalytic Pt and Ir substrates, while no growth is obtained on noncatalytic substrates such as SiO2. In this mechanism of catalytic dissociation by the substrate metal, one might expect growth of the transition metal oxide to terminate when all the surface Pt or Ir sites are covered by the growing film. However, as shown in this work, we observe sustained growth even after depositing a film of 20 nm in thickness. Two potential mechanisms are proposed to explain these results. Using this process, we successfully synthesize patterned bilayer thin films as well as core/shell nanoparticles.

Figure 1

Schematic illustration of the proposed mechanism by which area-selective ALD occurs. As an example, we illustrate Fe2O3 deposition from t-butyl ferrocene (TBF)/O2 on Pt. (a) Pt surfaces allow for the dissociative chemisorption of O2 to O* (chemisorbed oxygen), whereas SiO2 surfaces do not catalyze this reaction. (b) While TBF adsorption may occur on both surfaces, TBF only fully reacts where O* is present and therefore only leads to deposition on Pt. (c) In this way, a film of Fe2O3 can be deposited selectively by ALD on the Pt.

Methods

In this work, we performed several processes in three different ALD reactors, located at Eindhoven University of Technology (TU/e) and Stanford University. The TU/e reactor was used for Fe2O3 film depositions on planar substrates but not on particles. The experiments include in situ spectroscopic ellipsometry (SE) measurements on various substrates (Figure a), and deposition on patterned Pt squares (Figure ). The Stanford reactors were used for Fe2O3 deposition on supported Pt particles (Figure , Figure , and Figure S3), and Fe2O3 and NiO deposition on patterned Ir substrates (Figure S2). We note that the similar results obtained in the TU/e high-vacuum reactor and the Stanford low-vacuum reactors supports the robustness of our approach.

Figure 2

(a) Thickness measured by in situ SE during deposition of Fe2O3 from TBF/O2 on Pt with various TBF exposure times, with growth on Fe2O3 shown for comparison. Almost negligible growth is observed on the Fe2O3 substrate. (a, inset) Saturation curves showing the growth per cycle of Fe2O3 deposited on Pt substrates, extracted at 100 and at 250 cycles from the in situ SE data. The growth per cycle is determined by dividing the thickness at 100 or 250 cycles of deposition by the number of cycles performed. (b) X-ray photoelectron spectroscopy (XPS) scans showing the Fe2p region after 300 ALD cycles of Fe2O3 on Pt, Au, SiO2, and Al2O3 substrates. A clear Fe2p signal is observed on Pt, but no Fe above the detection limit is detected on the catalytically inactive substrates.

Figure 5

AES elemental maps for (a) Pt and (b) Fe of Fe2O3 deposited on patterned Pt substrates. (c) Corresponding SEM image. (d) Line scan showing Pt and Fe counts. The AES maps show excellent correlation between the Pt and Fe signals.

Figure 6

Bright-field TEM images of Pt/Fe2O3 core/shell nanoparticles supported on SiO2 nanospheres prepared by 50 cycles of Pt ALD, followed by (a) 25 cycles and (b) 50 cycles of Fe2O3 ALD from TBF/O2.

Figure 7

(a) ADF micrograph and (b) STEM-EDS line profile of Pt/Fe2O3 nanoparticles supported on SiO2 prepared using 50 cycles of Pt ALD, followed by 50 cycles of Fe2O3 ALD from TBF/O2. The large SiO2 sphere is decorated with smaller Pt nanoparticles that in turn are coated with Fe2O3. The location and direction of the line scan in (b) is marked on the ADF micrograph in (a).

The TU/e reactor is a high-vacuum system that is evacuated by combination of a rotary and a turbomolecular pump to a base pressure of ∼10–6 Torr. The system has been extensively described in previous work.[39] Fe2O3 depositions were performed using t-butyl ferrocene precursor (TBF, 98%, Strem Chemicals) and O2 gas. The TBF was kept at 100 °C to ensure adequate vapor pressure and was dosed using Ar carrier gas through a delivery line heated to 120 °C. Each cycle consisted of 1–20 s TBF exposure (as indicated in the results), 10 s pump time, 10 s O2 exposure, and 10 s pump time. The reactor walls were kept at 100 °C, and the substrate table was heated to 300 °C. Due to poor thermal contact in vacuum, however, the actual temperature of the samples during the process is typically lower.[40] For the area-selective process, a high O2 pressure of 750 mTorr was used to improve the thermal contact, giving a sample temperature of ∼250 °C as measured by spectroscopic ellipsometry (see below).

The Stanford reactors were of two types: a custom-built, low-vacuum system and a commercial Arradiance Gemstar low-vacuum system. Processes were developed for NiO and Fe2O3 depositions on planar samples using the custom system (Figure S2). Typical oxygen pressures on the order of 1–10 Torr were used. Depositions were performed at a 1 Torr N2 operating pressure. For NiO ALD, the substrates were heated to 250 °C and the nickelocene precursor was heated to 70 °C. Each cycle consisted of 7 s nickelocene exposure, 40 s purge time, 3 s O2 exposure, and 30 s purge time. For Fe2O3 ALD, the substrates were heated to 225 °C and the TBF precursor was heated to 95 °C. Cycles consisted of 8 s TBF exposure, followed by 10 s holding time, 10 s purge time, 3 s O2 exposure, and 30 s purge time.

The selectivity of the process was checked in the TU/e reactor by in situ SE on five different substrate materials: (1) Pt, (2) Au, (3) Fe2O3, (4) SiO2. and (5) Al2O3. These substrates will be referenced in the results, but details are described here. (1) The Pt films were 20 nm in thickness and were deposited by plasma-assisted ALD on thermal SiO2.[41] (2) The Au films were 16 nm in thickness and were deposited by electron-beam evaporation. (3) Fe2O3 films were ∼5 nm in thickness (without an underlying Pt substrate) and were deposited by plasma-assisted ALD of Fe2O3 on SiO2, using the process reported by Ramachandran et al.[42] Due to the use of O2 plasma, this is a conventional, nonselective process which allows for deposition on noncatalytic substrates. (4) SiO2 substrates consist of as-received 450 nm thermal SiO2 on Si. (5) The Al2O3 films (20 nm) were deposited on SiO2 by ALD. The in situ SE measurements were performed using a J.A. Woollam, Inc. M2000 ellipsometer (1.2–5.0 eV photon range).[43] Prior to the experiments, the sample temperature was determined by SE. A Si wafer with native oxide was measured using SE for different O2 pressures, and modeled using a J.A. Woollam temperature-sensitive optical model.

For XPS and cross-sectional TEM measurements, planar Fe2O3 films were deposited on electron-beam evaporated Pt films (∼16 nm). Cross-sectional TEM images were obtained at TU/e using a JEOL JEM-ARM200F system. The TEM lamella was cut from the sample by focused ion beam (FIB) milling. Prior to FIB milling, a protective layer of SiO2 was deposited by electron-beam induced deposition. Cross sections were imaged in bright-field and high angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) modes. Energy-dispersive X-ray spectroscopy (EDS) was performed using this microscope (Figure ).

Figure 3

Cross-sectional TEM micrographs of Fe2O3 deposited on a Pt substrate as recorded using (a) bright-field TEM and (b) HAADF-STEM. (c) Elemental concentrations determined using EDS line scan in cross-sectional view, along the vertical direction as indicated by the arrow in (b). The elemental concentrations measured are averaged over a ∼ 40 nm lateral distance. The Si signal for distances > 33 nm results from the protective SiO2 layer deposited prior to FIB milling.

Auger electron spectroscopy was performed at Stanford using a PHI-700 spectrometer. X-ray photoelectron spectroscopy (XPS) measurements were done at TU/e using Thermo Scientific KA1066 and at Stanford PHI Versaprobe III spectrometers with monochromatic Al Kα X-rays in both cases, and using Ar+ sputtering for depth profiling.

Core/shell Pt/Fe2O3 nanoparticles were deposited at Stanford by ALD on Aerosil OX50 silica powder, which was cleaned using UV irradiation and ozone. The powder was contained in a custom stainless steel cup with lid based on a design by Libera et al.[44] A standard Pt ALD process was used to deposit Pt nanoparticles with the Stanford Arradiance system.[45] Fe2O3 was deposited in the Stanford custom system using alternating exposures of TBF and O2 gas, with the silica powder heated to 250 °C. To saturate the high surface area of the powder, for these experiments, cycles consisted of 10–15 s TBF exposure, followed by 20 s holding time, 120 s purge time, 1.5 s O2 exposure, followed by 30 s holding time, and 120 s purge time.

TEM of the prepared nanoparticles was performed at Stanford using an FEI G2 F20 Tecnai TEM (Figure ) and an FEI Titan environmental transmission electron microscope (Figure and Figure S3). Scanning transmission electron microscopy (STEM) and energy-dispersive spectroscopy (EDS) were carried out on the Titan microscope. The SiO2 nanospheres on which the nanoparticles were prepared were dropcast onto Cu TEM grids with lacey or ultrathin C.

Results

ALD Growth and Substrate Selectivity

In situ SE was performed to characterize the selectivity and growth properties of the area-selective Fe2O3 process. Depositions were carried out on a number of different starting substrates to verify the need for a catalytically active film. The substrates include Pt, Au, Fe2O3, SiO2, and Al2O3 (as described in the Methods section). Plots of the thickness as a function of cycle number are presented in Figure a for the Pt and Fe2O3 substrates, using various TBF exposure times. On the Pt substrates, growth of Fe2O3 is clearly observed. The growth rate increases during the initial 100 cycles until a constant growth rate is achieved. This behavior is typical for the growth of polycrystalline films.[46] In the inset of Figure a, the saturation curve for the TBF exposure is shown as determined at 100 cycles or 250 cycles. At low film thicknesses up to ∼5 nm (around 100 cycles), the process appears to be in saturation after 10 s TBF exposure and reaches a growth rate of 0.045 nm/cycle. For thicker films at 250 cycles, however, the growth rate for 20 s TBF exposure (0.070 nm/cycle) is higher than for 10 s TBF exposure (0.057 nm/cycle), suggesting that the process is no longer in saturation. This observation will be explained in terms of a possible reaction mechanism in the Discussion section below. We note that the growth rate of 0.070 nm/cycle is similar to results observed for ALD using TBF and O2 plasma.[42]

It is evident from the growth curves of Figure a that negligible growth occurs on as-deposited Fe2O3 substrates (with no Pt underneath, substrate “3” as described in the Methods section), with a growth rate of only 0.005 nm/cycle for 10 s TBF exposure. This is an important observation as it confirms that the Fe2O3 film itself is not catalyzing the deposition of additional Fe2O3.

Figure b shows XPS scans of the Fe2p region on the Pt, Au, SiO2, and Al2O3 substrates after a deposition of 300 cycles. Fe is detected on the Pt substrate, with the Fe 2p3/2 peak located at 711.0 eV, in good agreement with Fe(III) in Fe2O3 (710.9 eV).[47] In addition, the spin orbit splitting between Fe 2p1/2 and Fe 2p3/2 is 13.6 eV as expected for Fe(III) in Fe2O3.[47] Prior work on ALD of Fe2O3 films using ferrocene precursors found that the films consisted of either α-Fe2O3 or amorphous Fe2O3.[37,42,48,49] Furthermore, no Fe is detected after deposition on the Au, SiO2, and Al2O3 substrates. These results demonstrate that the process has excellent selectivity, with immediate growth on Pt substrates (Figure a) and no indication of any growth on SiO2 or Al2O3 substrates (Figure b). Furthermore, the lack of growth on Au substrates indicates that a metal catalytically active for O2 dissociation, such as Pt or Ir, is required.

We also note another interesting observation: on Pt substrates, the Fe2O3 layer can deposit to at least 20 nm in thickness without noticeable attenuation of the growth rate. This is in contrast to the expectation that the growth will stop once the catalytic surface has been covered. This behavior is discussed in detail in the Discussion section.

Microstructure and Composition

Bright-field and dark-field cross-sectional TEM micrographs of a Fe2O3 film deposited with 300 Fe2O3 ALD cycles on a Pt substrate are provided in Figure . The Fe2O3 layer has completely covered the Pt substrate. However, the layer appears to be polycrystalline, resulting in grain boundaries and a high roughness. The images show a clearly defined interface between the underlying Pt substrate and the Fe2O3 film. Furthermore, Figure c shows the elemental concentrations as measured by EDS in the cross-sectional orientation. EDS likewise supports a clearly defined interface between the Pt and the Fe2O3 layers with no detectable Pt within the Fe2O3 layer. A plan-view scanning electron micrograph is shown in Figure S1, in which the polycrystallinity and resulting roughness is clearly visible.

The stoichiometry and chemical structure of Fe2O3 films on Pt substrates were investigated using XPS depth profiling by Ar+ ion sputtering. Figure shows the concentration of Pt, C, O, and Fe in a 17 nm Fe2O3 film as a function of the sputtering time (representing the depth). The initial surface measurement before any sputtering shows atomic concentrations of 60% O, 20% Fe, 13% C (adventitious), and 6% Pt. Initially, the Fe concentration remains constant while there is a reduction in O concentration most likely resulting from preferential sputtering of O.[50] The small amount of C present on the surface is fully removed after the first sputtering cycle, indicating it is only present on the surface and is the result of adventitious C contamination. The detection of Pt in the initial surface measurement is likely caused by the probing of the underlying Pt film, and not from possible Pt in the Fe2O3 film itself. Due to the high roughness of the Fe2O3 film, there are some regions of the film where the thickness is low enough (∼10 nm) for XPS to detect the underlying substrate. The absence of significant Pt signal in the cross-sectional EDS measurement (Figure c) further supports that the 6% Pt originates from the substrate and not the Fe2O3 film itself.

Figure 4

XPS sputter depth profile of a 17 nm thick Fe2O3 film on a Pt substrate.

Selectivity on Patterned Samples

We employed AES to investigate the selectivity of the deposition process on microstructured patterns. AES maps and line scans of Fe2O3 deposited on SiO2 substrates with patterned Pt squares are presented in Figure . The high spatial correlation between Pt and Fe elemental signals in the AES maps further supports the preference for deposition on Pt over SiO2 surfaces. On the basis of the ratio of observed elemental counts in the Pt and SiO2 regions of the substrate, the selectivity for deposition on Pt versus SiO2 is 2000:1.

The area-selective deposition of NiO and Fe2O3 on SiO2 substrates with patterned Ir was similarly investigated by AES, and the results are presented in Figure S2. There is also a very high correlation between the Fe and Ir, and likewise the Ni and Ir signals. Very low Ni and Fe levels are detected in the SiO2 regions where Ir is not present, giving selectivities of 200:1 for Ni and 900:1 for Fe.

Synthesis of Core–Shell Nanoparticles

The area-selective ALD process reported here is ideal for the preparation of core/shell nanoparticles, as a metal oxide shell can be deposited selectively on catalytically active core particles. As a demonstration, we prepared core/shell Pt/Fe2O3 nanoparticles. The Pt core particles were deposited by performing 50 cycles of Pt ALD on a SiO2 nanosphere support, and subsequently, the Fe2O3 shell was deposited using either 25 or 50 cycles of the area-selective Fe2O3 ALD process. Bright-field TEM micrographs of these nanoparticles are shown in Figure . The micrographs indicate a dark, strongly diffracting core with a lighter, weakly diffracting shell. The core is assigned to Pt and the shell to Fe2O3, as follows. Lattice fringes spaced by 0.23 Å in the dark core correspond to the Pt(111) crystal plane spacing. Strong lattice fringes are not observed in the shell, which may be due to limited crystallinity and the lower atomic number (Z) of Fe. Where fringes are observed, they are spaced by 0.27 Å corresponding closely to the Fe2O3(104) crystal plane spacing which agrees with the XPS results discussed above. As expected, the thickness of the Fe2O3 shell increases in going from 25 to 50 cycles. An annular dark-field (ADF) micrograph and a corresponding STEM-EDS line scan are shown in Figure . The dark-field image shows a strongly scattering core in each particle and a weakly scattering shell, which corresponds to a high Z Pt core and a low Z Fe2O3 shell. During the line scan (direction indicated in Figure a), the Fe signal appears before the Pt signal is observed, also in support of a core/shell structure with an Fe-containing shell and a Pt core. A STEM-EDS map and corresponding ADF micrograph are presented in Figure S3. In the elemental map, there is a high correlation between the Pt and Fe signals supporting area-selectivity. Very few Fe counts are observed in regions where only Si is present.

Discussion

The selective growth of Fe2O3 and NiO on Pt and Ir surfaces was demonstrated. No growth of Fe2O3 and NiO was observed on SiO2, Al2O3, and Au surfaces. The mechanism of this deposition process will now be discussed.

We propose a reaction mechanism closely related to the mechanism for ALD of Pt-group metals, most notably the ALD process for Pt using MeCpPtMe3 and O2 gas which was demonstrated to involve the chemisorption of oxygen.[22,51] In each cycle of this Pt ALD process, oxygen dissociatively chemisorbs on the Pt surface during the oxygen half reaction, and the chemisorbed oxygen oxidizes the ligands from the MeCpPtMe3 precursor. These combustion-like reactions happen during both the MeCpPtMe3 and oxygen half-reactions.[52−54]

Similarly, we propose a mechanism (Figure ) where, in each Fe2O3 or NiO ALD cycle, oxygen dissociatively chemisorbs during the oxygen half-reaction on the catalytic Pt or Ir substrate and not on surfaces inactive for oxygen activation (e.g., Au, SiO2, Fe2O3, or Al2O3). On the catalytic Pt or Ir, dissociatively chemisorbed oxygen species then combust the ligands of the nickelocene and t-butyl ferrocene. The lack of chemisorbed oxygen on the inactive substrates is the reason why no deposition occurs on these surfaces. Importantly, the lack of growth on Au substrates supports the requirement of a catalytically active substrate for the growth, since Au films are not catalytically active toward oxygen dissociation (except in small Au clusters).[55−57]

In this mechanism, one might expect the growth of the transition metal oxide to attenuate when the catalytically active substrate becomes covered by the growing metal oxide film. However, in the case of Fe2O3 on Pt shown in Figure , linear growth is sustained for at least 300 cycles (resulting in >20 nm growth), at which point the Pt substrate is already fully covered by Fe2O3 (as can be concluded from TEM images; see Figure ). Figure a also shows that the growth is almost negligible on Fe2O3 substrates, where the initial Fe2O3 film is grown using nonselective ALD with TBF/O2 plasma. This indicates that the Fe2O3 material itself does not provide a significant catalytic contribution to the growth.

The mechanism of the sustained growth is currently not fully understood. One key hypothesis is that, during the oxygen half-reaction, oxygen diffuses into the Fe2O3 film, reaches the Pt substrate, and is dissociated there. This diffusion is likely enhanced by the presence of grain boundaries resulting from the polycrystalline nature of the film as seen in Figure and Figure S1. If the dissociated O* can diffuse back toward the surface, it could react with the TBF precursor during the subsequent precursor half-cycle. In this way, the Fe2O3 film could act as a reservoir for oxygen. A similar phenomenon was observed during SiO2 ALD on Ag, where Ag2O is formed and releases O species, resulting in increased SiO2 deposition.[58] The saturation behavior of our process as shown in the inset of Figure a provides some evidence for this mechanism. For thin Fe2O3 films up to 100 cycles, saturated growth is observed. In contrast, however, for thicker films such as at 250 cycles, the growth rate increases going from 10 to 20 s of TBF exposure. This supports the proposed mechanism, as thicker Fe2O3 films would have a higher oxygen uptake capacity, allowing for increased amounts of oxygen to be released. Consequently, a higher TBF dose would be required for the growth to saturate. We expect saturation to eventually occur if the TBF exposure time is increased.

An alternative explanation may be the presence of a small amount of Pt in or on the Fe2O3 film. Diffusion of Pt from the substrate toward the top of the Fe2O3 layer may be possible at the elevated growth temperature, and this Pt may be catalyzing the growth. However, no Pt is detected in the Fe2O3 layer in the cross-sectional EDS measurement (Figure c), which makes this explanation less likely. As mentioned previously, the observation of Pt in the XPS depth profiling (Figure ) is likely due to detection of the Pt substrate through thinner regions of the Fe2O3 film. Therefore, we do not expect surface Pt species to play a key role in the continued growth; however, we cannot conclusively rule out the presence of trace amounts of Pt below the EDS detection limit.

The process of area-selective ALD that we demonstrate here should be extendable to a variety of systems. First, we have shown that selectivity can be achieved for two different transition metal oxide ALD systems (Fe2O3 and NiO) on two different catalytic substrates (Pt and Ir). Moreover, the ability to perform these selective depositions in ALD reactors of different designs and on both flat and nanoparticle substrates supports the robustness of our processes. By using Pt-group metal surfaces such as Pt, Ir, Pd, and Ru as the substrate to catalyze the oxidant reaction, we expect that area-selective deposition can be achieved through a similar mechanism for almost any metal oxide ALD system that uses a precursor requiring a strong oxidizing agent for growth (i.e., does not react with molecular O2) such as Co3O4,[59] In2O3,[60] and MoO3.[61] Likewise, there are other materials that can chemisorb and catalyze oxygen dissociation, and this may allow for selective growth even on non-noble metals such as Cu.[62] Furthermore, similar mechanisms may be possible with other co-reactants such as hydrogen,[8] as demonstrated by the selective deposition of Pd on Pt particles via hydrogen dissociation.[16]

Conclusions

We studied the area-selective ALD of Fe2O3 and NiO on Pt and Ir surfaces, using t-butyl ferrocene and nickelocene precursors, which conventionally require strong oxidizing agents. When using molecular O2 gas as the co-reactant, ALD growth can only be obtained on catalytically active surfaces, a phenomenon that was exploited for achieving area-selective ALD. These substrates catalyze dissociative chemisorption of O2, leading to activated O* that participates in the ALD reactions, while no growth occurs on substrates that do not possess catalytic activity for O2 dissociation including Au, SiO2, and Al2O3. Area-selective deposition is thereby achieved through the catalytic activity of the underlying substrate.

Process characterization was performed using in situ spectroscopic ellipsometry. XPS measurements on Au, SiO2, and Al2O3 show no indication of any Fe presence after 300 ALD cycles of Fe2O3 ALD, which confirms the high selectivity of the process. Area-selective ALD was demonstrated by successfully depositing Fe2O3 and NiO on micron-scale Pt and Ir patterns. AES measurements on these patterned samples yielded selectivity values of 2000:1 for Fe2O3 on Pt with respect to SiO2, and 200:1 for NiO on Ir. Furthermore, Pt/Fe2O3 core/shell nanoparticles were synthesized by performing area-selective ALD of Fe2O3 on Pt nanoparticles.

Finally, the demonstrated area-selective deposition of both Fe2O3 and NiO, as well as the similarity of results obtained in different reactors, supports the robustness of the process. This approach holds promise for the area-selective deposition of a wide range of other metal oxides for which the precursor does not react with molecular oxygen, for example, processes relying on O2 plasma or ozone as co-reactant. In particular, the Fe/Pt-based results may be applied for preparation of magnetic or spintronic devices, or catalyst particles for the synthesis of carbon nanotubes.