Atomic layer deposition of Cu₂O using copper acetylacetonate

Copper oxide occurs in two oxidation states, Cu₂O and CuO. Both are semiconductors with an intrinsic p-type character resulting from Cu vacancies.¹ While cupric oxide (CuO) has an indirect bandgap of 1.24 eV, cuprous oxide (Cu₂O) has a reported forbidden direct bandgap of 2–2.5 eV.^2–5 In addition to its wider bandgap, Cu₂O possesses higher conductivity due to fully occupied Cu 3d¹⁰ states, which leads to a high hole mobility of up to 256 cm²(V s)⁻¹.^1,6,7 As a result of these attractive properties as well as its nontoxic, earth-abundant nature, Cu₂O has found applications in the fields of water splitting, thin-film transistors, photovoltaics, and catalysis.^8–12 Various methods are available to deposit Cu₂O, including sputtering,^13,14 solgel synthesis,¹⁵ thermal¹⁶ and e-beam evaporation,⁹ chemical vapor deposition,^17–19 and atomic layer deposition (ALD).^20–22 The latter stands out as a result of its ability to produce thin, uniform films with excellent conformality and thickness control, a crucial factor for electronic devices requiring thin, pinhole-free layers.

Early work on ALD of copper-containing materials focused on deposition of Cu metal films for interconnect applications, in which copper oxide was sometimes formed as an intermediate step.^23,24 More recently, greater attention has been given to ALD of copper oxides themselves. A summary of previously reported precursors and processes for ALD of copper oxides is presented in Table I. Deposition of Cu₂O has been reported using a range of precursors, including Cu(I)(hfac)₂+H₂O,²⁵ Cu(OAc)₂+H₂O,²⁰ n(Bu₃P)₂Cu(acac)+H₂O/O₂,²⁶ Cu(dmap)₂+H₂O,²⁷ and Cu(sBu-Me-amd)]₂+O₂ plasma.²⁸ Although these processes resulted in the formation of Cu₂O, problems related to fluorination,²⁹ limited thermal stability,²⁰ and narrow temperature windows³⁰ have been reported, which are associated with the copper precursors. Other processes resulted in the formation of CuO rather than Cu₂O.^31,32 To overcome such issues, copper(II) acetylacetonate [Cu(acac)₂] seems a promising precursor option due to its low cost, good stability, and relatively broad temperature window overlapping with other common processes, which facilitates possible ternary compound formation.^23,33,34

The use of Cu(acac)₂ as an ALD precursor was first reported for deposition of Cu metal, with H₂ as the reactant.³⁵ More recently, Alnes et al. reported ALD of cupric oxide (CuO) from Cu(acac)₂ and O₃ at temperatures of 150–240 °C. No deposition was obtained with H₂O alone; however, when air accidently reached the reactor, Cu₂O was reportedly formed.³³ Tripathi et al. also reported the formation of Cu metal using Cu(acac)₂ as the precursor. The authors investigated a combination of H₂O and a reducing agent to produce the metallic film.²³ When a lesser amount of the reducing agent was introduced, the authors found some residual Cu₂O. Density functional theory studies suggest that the exchange reaction between H₂O and Cu(acac)₂ should be energetically feasible, but that the level of reactivity is less than that for O₃.³⁴ This might be expected to lead to a less oxidized state, i.e., Cu₂O. This would be consistent with results for other Cu precursors, where the reaction with O₃ generally produces CuO,³⁶ while the reaction with H₂O produces Cu₂O.^22,37 Notably, the analogous process with Cu(acac)₂ and H₂S has been employed to obtain copper-sulfide-based materials.³⁸ 

On the basis of this evidence, we have investigated a novel process to deposit Cu₂O films on silicon substrates employing Cu(acac)₂ and H₂O/O₂ as a precursor and a reactant, respectively. The combination of H₂O and O₂ was chosen based on the aforementioned report in which an accidental combination of Cu(acac)₂ and H₂O/air was found to result in Cu₂O.³³ Indeed, the combination of H₂O and O₂ is known to have a synergistic effect in some cases. For example, another work reported significantly higher growth per cycle (GPC) of In₂O₃ using H₂O in combination with O₂ when compared to using either alone.³⁹ We show that this process indeed results in deposition of Cu₂O. Process development and materials characterization are presented.

Copper oxide thin films were deposited using a Beneq TFS 200 ALD reactor. Cu(acac)₂ (>98% purity, Strem) and a combination of H₂O and O₂, dosed simultaneously, were selected as precursors. Due to the relatively low vapor pressure of the precursor (8.83 kPa at 140 °C),⁴⁰ Cu(acac)₂ solid powder was heated to 140 °C, which is sufficient to generate some vapor through sublimation.^33,38 Nitrogen (99.99%) was used as a carrier and purge gas at a flow rate of 300 SCCM. Purge times were kept constant at 1 s for Cu(acac)₂ and 3 s for H₂O/O₂. In addition to standard pulse/purge steps, we added an additional pressurization step to improve transport of Cu(acac)₂ into the reactor. Before each pulse of Cu(acac)₂, the precursor canister was pressurized with N₂ gas for 0.5 s at a flow rate of 300 SCCM. A similar procedure has been reported for this same precursor and reactor model in a previous work.³⁸ 

Films were deposited on single-side-polished ∼1–2 cm² (100) silicon substrates at 200 °C. This temperature was chosen because it is in the middle of the ALD temperature window previously obtained for deposition of CuO using the same precursor with ozone,³³ which is limited at the lower end by the high temperature required for sublimation and at the higher end by decomposition of the precursor at 240 °C. It is also a common temperature for other ALD processes and, therefore, suitable as a baseline for the development of ternary compounds. The wafers were cleaned using the standard Radio Corporation of America (RCA) procedure followed by a dip in 1% HF solution. Film thickness and optical properties were measured ex situ using a J. A. Woollam M2000D spectroscopic ellipsometer. Optical properties were first determined from measurements of the thickest films, as described in the Results and Discussion section, and then used to determine film thickness for thinner films. Film crystallinity was characterized using grazing incidence x-ray diffraction (GIXRD) on a PANalytical X’Pert PRO MRD system with Cu Kα radiation and an incident angle of 0.6°. Fourier-transform infrared spectroscopy (FTIR) was performed using a Bruker Vertex 80 tool in the 400–4000 cm⁻¹ range. Finally, film morphology and roughness were analyzed using atomic force microscopy (AFM). A Bruker dimension icon scanning probe microscope system was used with a SCANASYST-AIR tip (spring constant 0.2–0.8 N m^–1) operating in a tapping mode at a resonant frequency of 45–95 kHz.

We investigated the self-limiting behavior of the deposition process by independently varying the pulse times of the precursor and reactants. The pulse times for both Cu(acac)₂ and H₂O/O₂ were fixed at 3 s while varying the other. We present a schematic illustration of the expected surface reactions occurring during a single deposition cycle [Fig. 1(a)], together with a process timeline representing two deposition cycles, which gives the ranges of parameters studied, in Fig. 1(b).

Figure 2 shows the saturation curve for both the precursor and reactant pulses, determined from ex situ ellipsometry of films deposited using 400 and 500 cycles, respectively. The GPC reaches saturation in both cases, at a pulse time of 3 s for Cu(acac)₂ and 2 s for H₂O/O₂, indicating ALD-like behavior. Some degree of variability was observed, as indicated by the standard deviation shown as error bars in the figure (derived from multiple samples at different positions in the reactor).

The self-limiting behavior results in a GPC of ∼0.06–0.07 Å/cycle in saturation. This is lower than that reported for the reaction with ozone,³³ which is in accordance with previous literature that indicates a lower reactivity of Cu(acac)₂ with H₂O compared to O₃. This is also consistent with the Cu(dmap)₂ and [Cu(^sBuAMD)]₂ cases.^27,32,41,42

To verify the linearity of growth behavior, we investigated film thickness as a function of the number of cycles. Figure 3 demonstrates a linear growth of thickness vs cycle count up to a thickness of ∼25 nm. There is some indication of a possible reduction in growth rate above 800 cycles; however, given the limited number of data points and relatively large standard deviation among different samples at each point, we cannot confirm this hypothesis. The GPC derived from the linear fit is 0.07 Å/cycle. Notably, there is no apparent nucleation delay on the H-terminated Si(100) surface. In addition, we checked whether the GPC was dependent on the surface finish of the substrate as it is well-known that different functional groups on the starting surface can influence the film growth. Thus, apart from RCA-cleaned (HF last) substrates, we also tested the deposition process on silicon substrates (i) where the oxide layer resulting from the RCA clean was retained and (ii) after etching by tetramethyl ammonium hydroxide. We found that the growth behavior was independent of the surface finish (data not shown).

In addition to verifying the ALD features of the process, it is important to determine the nature of the deposited films and, in particular, which copper oxide phase is obtained. Figure 4 shows the GIXRD spectra as a function of the number of cycles. No peaks related to the film appear up to 600 cycles, or ∼4 nm, indicating that the film remains amorphous or poorly crystalline. GIXRD peaks related to the Cu₂O (111) planes appear (weakly at first) at 800 cycles (∼7 nm), while the (200) and (220) peaks become visible at 1400 cycles and above, indicating the formation of a polycrystalline film (COD#9007497).⁴³ The lack of peaks related to Cu or CuO indicates that only the cuprous phase was obtained. As discussed above, this is likely related to the lower oxidation strength of the H₂O/O₂ reactant. Note that the full width at half maximum (FWHM), for the most intense peak, decreases with increasing cycle count, indicating an increase in crystallite size for thicker films.

We propose the following growth mechanism of Cu₂O based on a previous first-principles study.³⁴ Initially, when reacting at the surface, one acetylacetonate (acac) ligand dissociates from the precursor and leaves the system during the first purge, leaving behind a Cu(acac)-terminated surface. When water interacts with the Cu(acac), a proton is transferred from H₂O, initially forming Cu[H(acac)]. The Cu–O bond is then broken, eliminating the other acac ligand and forming Cu–OH.³⁴ Although the role of O₂ is still unclear, we argue that it might play a catalytic role in ligand dissociation, facilitating the deposition process.

FTIR was used to further examine the structural and compositional properties of the deposited films. Figure 5 shows the FTIR absorbance spectra of films deposited with a range of thicknesses by varying the number of cycles. These correspond to the XRD data shown in Fig. 4. The sharp peak at 621 cm^–1, which appears at 1400 cycles or more, is associated with a Cu–O stretch of Cu₂O.^4,44 As the number of cycles increases, the peak becomes sharper and more intense, which indicates an increase in crystallinity and thickness, respectively. This result correlates well to the appearance and growth in intensity of the Cu₂O (111) peak in the GIXRD measurements. Additionally, we identified a CuO peak at 489 cm^–1 for the thickest film.^45,46 This could possibly indicate a mixture of Cu₂O and CuO phases in the film. However, a more plausible explanation is the formation of CuO at the surface due to postdeposition oxidation in atmosphere, which is known to take place for Cu₂O films.²⁰ The FTIR also shows peaks in the range of 1500–1600 and 3000–3600 cm^–1, attributed to unreacted acetylacetonate and OH groups, respectively.^47,48 Since the intensity of these peaks increases with film thickness, we infer that these groups are located in the bulk of the film rather than at the surface. These residual carbonyl and hydroxyl groups are likely the result of an incomplete ligand exchange between Cu(acac)₂ and H₂O/O₂. It may be possible to address this incomplete exchange by lengthening the precursor or reactant pulse times.

The optical properties of the deposited films were determined using spectroscopic ellipsometry. To fit the measured data, we employed a two-step approach. In the first step, the thickness was first determined by fitting the data in the wavelength range where the films were transparent (<2 eV) using a simple Sellmeier model. Having determined the thickness in this way, the optical constants (refractive index n, and extinction coefficient k) were then determined by wavelength-by-wavelength direct inversion over the full range. This approach allows small features in the dielectric function to be resolved, without the need to make assumptions about the appropriate optical model. Figures 6(a) and 6(b) show the resulting optical constants and corresponding absorption coefficient, respectively. The refractive index (n = 2.4 at 632.8 nm) and extinction coefficient values are in accordance with previous findings in the literature for a Cu₂O film of similar thickness, prepared by evaporation of Cu and subsequent thermal oxidation, which were also obtained by ellipsometry.² Note that all optical data are for the film deposited using 3600 cycles (∼25 nm thickness).

Although the nature of the band structure and optical transitions in Cu₂O have been well established in some of the earliest semiconductor literature, there seems to be some confusion in the more recent literature as to whether the first energy transition is direct allowed or direct forbidden, with indirect transitions also reported. We extracted the bandgaps’ energy based on the Tauc equation with n = 3/2 (direct forbidden bandgap) for the first transition and n = 1/2 (direct allowed) for the second transition, in accordance with the expected nature of the first and second energy transitions for Cu₂O.^49,50 For the second transition, the background absorption contribution resulting from the first transition was first subtracted from the total absorption coefficient in order to obtain a reliable estimate for the direct allowed energy gap. The resulting Tauc plots are shown in Figs. 7(a) and 7(b), while the contribution of each transition to the total absorption coefficient is illustrated in Fig. 7(c). The good linearity obtained for both Tauc plots supports our assumption regarding the nature of the transitions.

We find a direct forbidden bandgap of 2.07 eV for the first transition, followed by a direct allowed bandgap of 2.60 eV for the second transition. In reality, each of these transitions is expected to represent a combination of two closely spaced (Δ_so= 0.1338 eV) transitions resulting from spin–orbit splitting of the highest valence band. These are resolvable at low temperature but can be expected to be effectively merged by broadening effects at room temperature. At 4.2 K, the energy of the lower of the two transitions in each case is 2.17 eV (direct forbidden) and 2.624 eV (direct allowed), respectively.^49,51 Comparing to these values, the values obtained for our films at room temperature are slightly lower (0.1 and 0.02 eV, respectively), which is in excellent agreement with previous reports regarding the temperature dependence of the bandgap of Cu₂O.⁵² 

The importance of clearly establishing both bandgaps relates to the potential applications of Cu₂O. While the energy of the first (direct forbidden) transition corresponds to the energy difference between the highest valence band and lowest conduction band edges (electronic band gap) and, therefore, determines the band offsets when the material is employed in a heterojunction structure, the energy of the second (direct allowed) transition indicates the photon energy corresponding to the onset of strong optical absorption. Thus, for optoelectronic applications such as solar cells, in which both band alignment and optical absorption/transparency are critical, a better understanding and differentiation of the two transitions is important.

The morphology of the films was investigated using AFM (Fig. 8). The first few hundred cycles (especially the 200 and 600 cycle samples) show an unclear structure with very small features, typical of an amorphous film. Upon increasing the number of cycles, larger grains start forming and a more well-defined structure indicates some degree of crystallization, especially for 3600 cycles (∼25 nm). When comparing the AFM results with the GIXRD and FTIR data, we can see a clear correlation between the appearance of crystalline peaks, narrowing of FWHM, and grain growth. Additionally, the RMS surface roughness of the films increased with thickness from ∼1 nm for the thinnest films before stabilizing at ∼2 nm around 2200–3600 cycles (Fig. 9). The roughness values are comparable with other results in the literature for Cu₂O films produced by ALD, with higher and lower values being reported.^20,22,27 When compared to other techniques, such as sputtering, the values reported here are considerably smaller for a similar thickness,⁵³ which shows the benefits of ALD for fabricating Cu₂O thin films.

We have successfully produced Cu₂O thin films using Cu(acac)₂ as a precursor and water/oxygen as reactants, with a GPC of ∼0.07 Å/cycle at 200 °C. The process exhibits self-limiting growth without a nucleation delay on Si substrates. GIXRD and FTIR show amorphous or weakly crystalline films up to 600 cycles (∼7 nm) and polycrystalline ones for 1400 cycles (∼12 nm) or more. In addition, FTIR reveals the persistence of carbonyl and hydroxyl groups in the bulk of the film, indicating incomplete ligand exchange, a possible tradeoff of using reactants with lesser oxidizing power that needs to be further addressed. Ellipsometry confirms optical properties consistent with thin-film Cu₂O. A direct forbidden bandgap of 2.07 eV and a direct allowed gap of 2.60 are found for a 26 nm film, in excellent agreement with expected values for Cu₂O. AFM results are in accordance with GIXRD and FTIR, showing an undefined amorphous structure initially, with clear grain growth as thickness increases. Surface roughness of only ∼2 nm for a 25 nm thick film indicates a highly conformal process with the potential to produce high-quality, pinhole-free films for demanding optoelectronic applications. Future work on the effect of deposition temperature will be important to better understand the process and how it affects the final properties of the films. The use of H₂O/O₂ as reactants in a purely thermal process and compatibility with the common ALD process temperature of 200 °C suggest good potential for future use in deposition of Cu-based ternary compounds.

The authors thank James Cotsell for his tireless work in maintaining the ALD system. They also acknowledge the Research School of Chemistry, ANU, for allowing access to their facilities. This work has been supported by the Australian Renewable Energy Agency (ARENA) through the Australian Centre for Advanced Photovoltaics (ACAP).

The authors have no conflicts to disclose.

Gabriel Bartholazzi: Conceptualization (lead); Data curation (equal); Formal analysis (equal); Methodology (lead); Visualization (lead); Writing – original draft (lead); Writing – review & editing (equal). M. M. Shehata: Conceptualization (supporting); Investigation (supporting); Visualization (supporting); Writing – review & editing (supporting). Daniel H. Macdonald: Funding acquisition (lead); Resources (lead); Supervision (supporting); Writing – review & editing (supporting). Lachlan E. Black: Conceptualization (supporting); Formal analysis (supporting); Funding acquisition (lead); Investigation (supporting); Methodology (supporting); Resources (lead); Supervision (lead); Writing – review & editing (supporting).

The data that support the findings of this study are available within the article.