Atomic Layer Deposition of Photoconductive Cu2O Thin Films.

Herein, we report an atomic layer deposition (ALD) process for Cu2O thin films using copper(II) acetate [Cu(OAc)2] and water vapor as precursors. This precursor combination enables the deposition of phase-pure, polycrystalline, and impurity-free Cu2O thin films at temperatures of 180-220 °C. The deposition of Cu(I) oxide films from a Cu(II) precursor without the use of a reducing agent is explained by the thermally induced reduction of Cu(OAc)2 to the volatile copper(I) acetate, CuOAc. In addition to the optimization of ALD process parameters and characterization of film properties, we studied the Cu2O films in the fabrication of photoconductor devices. Our proof-of-concept devices show that approximately 20 nm thick Cu2O films can be used for photodetection in the visible wavelength range and that the thin film photoconductors exhibit improved device characteristics in comparison to bulk Cu2O crystals.

Introduction

Copper(I) oxide (n>an class="Chemical">Cu2O) is a p-type semiconductor material that crystallizes in a simple cubic structure.[1] The p-type conductivity of Cu2O originates from intrinsic Cu vacancies in the crystal lattice, which facilitates the formation of acceptor levels approximately 0.25–0.45 eV above the valence band maximum.[2,3] Other properties of Cu2O related to its semiconducting character include an optical band gap of 2.1–2.2 eV for the bulk material,[4,5] a high absorption coefficient in the visible wavelength range,[4] hole mobility ranging from 40 to 120 cm2/(V s), and electron diffusion length reaching up to 5 μm.[6−11] Moreover, Cu2O is nontoxic, low-cost, and consists of earth-abundant elements. Due to this favorable combination of chemical, electronic, and optical properties, Cu2O has been studied for a wide range of applications, including heterojunction[12,13] and homojunction solar cells,[14,15] photoelectrochemical water splitting,[16,17] thin film transistors,[10,18] nonvolatile memories,[19,20] photodiodes, and photoconductors.[11,21−23] Furthermore, copper(I)-based ternary oxides, such as CuAlO2[24] and SrCu2O2,[25] are promising candidates for transparent p-type conductive oxides,[26] and delafossite class CuMO2 (M = Ta, Fe, V) compounds show promise in solar energy applications.[27] Consequently, several methods have been developed for synthesizing Cu2O and for modifying its chemical, electronic, and optical properties. These methods include solvothermal methods,[28] electrodeposition,[14,16,29] sol–gel synthesis,[30] thermal oxidation of copper,[4,5,31] sputtering,[2,7] chemical vapor deposition (CVD),[32−34] and atomic layer deposition (ALD).[23,35−39] From these methods, the solvothermal approach produces freestanding, one-dimensional Cu2O nanostructures, whereas the rest of the techniques are better suited for synthesizing Cu2O thin films. Among the thin film synthesis methods, ALD stands out due to its unique ability to produce conformal, high-quality and pinhole-free films uniformly on large area substrates.[40,41]

The ability of ALD to produce high-quality thin films originates from a depon>an class="Chemical">sition mechanism where the film growth proceeds via saturative and self-limiting surface reactions between alternately supplied gaseous precursors.[42] This concept enables several unique features, such as control of film thickness on a subnanometer level, fabrication of nanolaminates, flexible introduction of dopants,[43] and the deposition of conformal films on three-dimensional structures, such as high aspect ratio vias and trenches.[40,44] Previous reports on ALD of Cu2O have been based on the use of CuCl + H2O,[35] (Bu3P)2Cu(acac) + H2O/O2 (Bu = butyl, acac = acetylacetonate),[36] Cu(dmamb)2 + H2O (dmamb = 1-dimethylamino-2-methyl-2-butoxy),[23,37,38] and Cu(dmap)2 + H2O (dmap = 1-dimethylamino-2-propoxide).[39] The deposition process based on CuCl + H2O is limited to temperatures of 350 °C and higher as CuCl needs to be heated to 340 °C to obtain sufficient vapor pressure for gas phase deposition. It was also noted that this deposition process is not suitable for obtaining phase-pure Cu2O films due to the reductive thermal decomposition of CuCl that leads to the formation of metallic copper in the films.[35] In the case of the (Bu3P)2Cu(acac) + H2O/O2 process, film deposition was achieved at significantly lower temperatures, 110–130 °C.[36] This deposition process exhibited good ALD characteristics, especially on SiO2 substrates. Further improvement in ALD of Cu2O was achieved using an aminoalkoxide copper precursor, Cu(dmamb)2, in combination with water vapor.[37] The optimal deposition temperature for this process with respect to saturation, film crystallinity, and surface roughness was shown to be 140–160 °C. Films deposited from Cu(dmamb)2 and H2O were polycrystalline, phase-pure Cu2O and contained negligible amounts of carbon and nitrogen impurities as analyzed using X-ray diffraction (XRD) and Rutherford backscattering spectrometry. In addition, the Cu2O films deposited using the Cu(dmamb)2 + H2O process had a direct allowed optical band gap value of 2.52 eV and hole mobility of 8.05 cm2/V.[37] Finally, another aminoalkoxide copper(II) compound, Cu(dmap)2, was used with water vapor at 110–200°C.[39] The focus of this study was the deposition and characterization of 5 nm thick films using quartz crystal microbalance (QCM), X-ray diffraction, and X-ray photoelectron spectroscopy (XPS) techniques. Cu2O films deposited using the Cu(dmap)2 + H2O process were reported to be X-ray amorphous and to oxidize upon exposure to air, which was attributed to the small thickness of the films. Interestingly, for both the Cu(dmamb)2 + H2O and the Cu(dmap)2 + H2O processes, the obtained films were of the Cu(I) oxide (Cu2O) phase even though the oxidation state of copper in these precursors is +2. This implies that copper is reduced during the deposition process. In CVD experiments based on aminoalkoxide copper precursors, the reducing agent was identified to be the aminoalkoxide ligand.[45,46] Therefore, it is conceivable that hydrogenated aminoalkoxide ligands can act as in situ formed reducing agents in ALD as well. However, this reaction mechanism has not been verified experimentally.[39]

In this study, we have used copper(II) acetate, pan class="Chemical">Cu(OAc)2, in ALD to deposit Cu2O thin films. Cu(OAc)2 is a commercially available bulk chemical that exists both in the hydrated form, Cu(OAc)2·H2O, as well as in the anhydrous form.[47] In the solid state, both the hydrated form and the anhydrous variant exist as dimers.[48] A clear advantage of using copper(II) acetate in Cu2O film deposition is the low cost of the precursor, in particular if the hydrated variant is considered.

Heating hydrated Cu(OAc)2·n>an class="Chemical">H2O to approximately 110–160 °C results in the loss of water of crystallization and the subsequent formation of anhydrous Cu(OAc)2.[47,49] At temperatures where volatilization starts to occur, the thermal properties of Cu(OAc)2 are complicated by both the reduction of copper and the formation of cluster molecules in the gas phase.[50] A thermogravimetric study performed under a dynamic N2 atmosphere showed that heating Cu(OAc)2 to 200 °C in oxygen-free conditions results in the reduction of copper from Cu2+ to Cu+ and the evaporation of copper(I) acetate, CuOAc.[47] The reduction of Cu(OAc)2 to CuOAc upon heating in vacuo was observed also in another thermogravimetric study.[51] There, it was also noted that the residue of heated Cu(OAc)2 contained both Cu2O and metallic Cu. This indicates that the reduction of Cu(OAc)2 to CuOAc is likely to proceed via a two-stage, solid state comproportionation reaction. In the first stage, Cu(OAc)2 is partially decomposed to form metallic copperThe in situ formed metallic copper then reacts with Cu(OAc)2 and volatile CuOAc is formed via comproportionationThe existence of this reaction pathway is supported by the fact that CuOAc can be obtained from Cu(OAc)2 by heating the latter to 180–190 °C together with copper dust in reduced pressure.[52] If Cu(OAc)2 is heated to 290 °C and above in oxygen-free conditions, complete decomposition of the compound occurs[47] and results in the formation of metallic copper and volatile species deriving from the acetate ligand, such as acetic acid, acetone carbon dioxide, and water.[53]

The reduction of Cu(OAc)2 to n>an class="Chemical">CuOAc has also been shown to occur by means of mass spectroscopy.[50] In an extensive study by Didonato and Busch, it was demonstrated that a majority of the volatile species forming from Cu(OAc)2 contain copper as Cu+, which is in agreement with the findings of the thermogravimetric experiments described in the literature.[47,51] The mass spectrometric study also established that volatile copper(I) species that result from heating Cu(OAc)2 form cluster molecules that contain either two, three, or four copper atoms.[50] This implies that CuOAc species in the gas phase are either dimeric, trimeric, or tetrameric. In the context of ALD, the thermally induced reduction of Cu(OAc)2 to CuOAc is significant because it enables the deposition of Cu(I) films from a Cu(II) precursor without the need for a reducing agent.

To examine the use of Cu(OAc)2 as a pren>an class="Chemical">cursor in ALD, we have conducted a complete process development study using water vapor as the co-reactant. This precursor combination produces Cu2O films at deposition temperatures of 180–220 °C. The deposition chemistry described herein fulfills the characteristics of an exemplary ALD process, namely saturation with respect to both precursors, accurate thickness control, and the deposition of highly crystalline, impurity-free films. Moreover, we demonstrate that this ALD chemistry can be used to fabricate photoconductor devices and that these devices exhibit photoconductivity and appreciable spectral responsivity under illumination in the visible wavelength range.

Results and Discussion

Film Deposition Experiments

In the film deposition expn>eriments, both n>an class="Chemical">Cu(OAc)2·H2O and Cu(OAc)2 were tested for their suitability in ALD. No significant differences in the films deposited from Cu(OAc)2·H2O compared to the films deposited from anhydrous Cu(OAc)2 were observed. An apparent reason for this finding is that Cu(OAc)2·H2O is converted to anhydrous Cu(OAc)2 in the process of heating the copper precursor to its ALD source temperature, 175–185 °C. Close to the source temperature, the reduction of Cu(OAc)2 starts to take place, and CuOAc is evaporated. Because several thermal events occur before the source temperature is reached, Cu(OAc)2·H2O does not appear as an ideal ALD precursor. However, no adverse effects related to the conversion of Cu(OAc)2·H2O to Cu(OAc)2 were detected in the film deposition experiments. Moreover, film deposition did not occur without the co-reactant (H2O), i.e., when the pulsing sequence [Cu(OAc)2·H2O—purge] was repeated for 2000 cycles, which indicates that the water of crystallization does not contribute to the film deposition. In the following discussion, we will refer to the copper precursor as Cu(OAc)2 because the water of crystallization in Cu(OAc)2·H2O is removed at temperatures relevant to film deposition. Similarly, as the evaporating copper species is Cu(I) acetate, we refer to the copper precursor in the gas phase as CuOAc.

Film deposition was studied in the temperature range of 180–240 °C. Deposition experiments related to the effect of deposition temperature and saturation were performed using 1000 deposition cycles. For the depositions done at 180 °C, a source temperature of 175 °C for Cu(OAc)2 was used to maintain a thermal gradient between the copper precursor and the substrates. For the film deposition experiments at 200–240 °C, a source temperature of 185 °C was used. As seen from Figure a, the growth per cycle (GPC) values for films deposited by 1000 cycles were 0.13 ± 0.01 at 180 °C and 0.11 ± 0.01 at 200–240 °C.

Figure 1

Growth characteristics of Cu2O films deposited from Cu(OAc)2 and H2O. (a) GPC at deposition temperatures of 180, 200, 220, and 240 °C, (b) the effect of Cu(OAc)2 pulse length, (c) the effect of H2O pulse length, (d) film thickness vs the number of deposition cycles at 200 °C. The dashed lines in panels (a)–(c) are to guide the reader’s eye. The solid line in panel (d) is a linear fit to the data points. The R2 value describing the goodness of fit is 0.99365.

The GPC values obtained for Cu2O in the tempn>erature range of 180–220 °C are low, which renders the depon>an class="Chemical">sition of films thicker than 100 nm impractical. However, we note that the GPC values obtained for the Cu2O deposition process described herein are comparable to those based on (Bu3P)2Cu(acac) + H2O (<0.10 Å on SiO2 substrates),[36] Cu(dmamb)2 + H2O (0.13 Å at 140–160 °C),[37] and Cu(dmap)2 + H2O processes (0.12 Å at 120–160 °C).[39] Films deposited at 240 °C were visibly metallic due to the thermal decomposition of CuOAc. As the aim of this work was to deposit Cu2O films, deposition experiments at temperatures above 240 °C were not conducted.

Saturation experiments were performed with 1000 deposition cycles at 200 °C. As seen from Figure b,c, the saturative growth mode characteristic to n>an class="Disease">ALD was obtained for both Cu(OAc)2 and H2O using 2.0 s pulses. For films deposited by 1000 cycles, 3.0 s long purge times for both precursors were sufficient for obtaining uniform films over 5 × 5 cm2 substrates (Figure S1). Finally, the thicknesses of Cu2O films deposited at 200 °C with an increasing number of deposition cycles were measured (Figure d). Atypically, for an ALD metal oxide process, the GPC of the Cu2O films did not remain constant as the number of deposition cycles was increased. The application of 500 and 1000 deposition cycles at 200 °C resulted in films with thicknesses of 6.5 and 11.3 nm, respectively, which averages to a GPC of 0.12 Å. With 3000, 5000, and 7000, the thicknesses of the obtained films were 24.2, 43.0, and 58.0 nm, respectively, which averages to a GPC of 0.08 Å. This finding entails that the deposition of Cu2O using this process is substrate-enhanced in the early stages of film growth.[54,55] A probable reason for this behavior is that the adsorption density of the copper precursor is higher on the native oxide-terminated Si than on the Cu2O film itself. We hypothesize that this is due to the lability of hydroxyl groups on Cu2O surfaces, as Cu(I) hydroxides are known to decompose to oxo-terminated Cu2O at elevated temperatures.[56]

According to field emission scanning electron microscopy (FESEM) and atomic force microscopy (AFM) analyses (as disn>an class="Chemical">cussed later), films deposited with 500 cycles on Si substrates are not continuous. Consequently, the effective surface area and the number of adsorption sites at the early stages of the film growth is higher compared to the stage where a Cu2O film fully covering the underlying Si substrate has formed. Despite the nonconstant GPC over an increasing number of deposition cycles, the thickness of Cu2O films deposited at 200 °C could be accurately controlled in the range of 5–60 nm, which signifies that the ALD chemistry described herein is a viable candidate to be used for applications in which sub-100 nm films are required. The good thickness control of this ALD process was also proven by depositing an approximately 50 nm thick Cu2O film on a trench structure with an aspect ratio of approximately 2:1. For this trench structure, 100% conformality was achieved using a pulse and purge sequence of 2 s Cu(OAc)2 pulse—7 s purge—3 s H2O pulse—7 s purge (Figure S2).

Film Crystallinity and Structure

The crystallinity of the films was studied with X-ray diffraction in the grazing incidence geometry. Grazing incidence X-ray diffraction (GI-XRD) patterns of films deposited with 5000 cycles at 180–240 °C are shown in Figure a and the effect of the film thickness on the crystallinity of films deposited at 200 °C is shown in Figure b.

Figure 2

GI-XRD diffractograms of (a) Cu2O films deposited with 5000 cycles at 180–240 °C and (b) the effect of film thickness on crystallinity at a deposition temperature of 200 °C.

GI-XRD diffractograms of (a) pan class="Chemical">Cu2O films deposited with 5000 cycles at 180–240 °C and (b) the effect of film thickness on crystallinity at a deposition temperature of 200 °C.

The peaks in the diffractograms shown in Figure are indexed according to the ICCD cards for cubic n>an class="Chemical">Cu2O (ICCD PDF 5-667) and cubic Cu (ICCD PDF 4-836). According to GI-XRD analysis, the films deposited at 180–220 °C are polycrystalline, phase-pure cubic Cu2O. The most intense reflections are those assignable to the (111) and (200) planes of cubic Cu2O. Based on Rietveld refinement,[57] the unit cell dimensions in Cu2O films deposited at 180–220 °C are within ±0.01 Å of the reported literature value of 4.2685 Å.[58] As discussed above, films deposited at 240 °C were visually metallic due to the thermal decomposition of the copper precursor. The GI-XRD pattern of films deposited at 240 °C confirmed the formation of metallic copper (Figure a). However, these films were not pure metallic copper, as reflections assignable to both metallic Cu and Cu2O could be observed in the diffractogram. The adhesion of the films deposited at 240 °C was poor, which signifies that this ALD chemistry is not suitable for depositing metallic copper thin films on oxide surfaces. The saturation studies and GI-XRD experiments indicated that the optimal deposition temperature for this ALD process is 200 °C. Therefore, this temperature was used to deposit films for further characterization, including morphology studies, compositional analysis, and study of optoelectronic properties.

The morphology of the Cu2O films was analyzed using FESEM and AFM. Plane-view micrographs of Cu2O film deposited on Si substrates with 500, 1000, 3000, 5000, and 7000 cycles are presented in Figure . As observed from Figure a, films deposited over 500 cycles on Si are not continuous. Upon increasing the deposition cycle count to 1000, the underlying Si substrate is no longer observed, which indicates the formation of a continuous Cu2O film (Figure b). Films deposited with 1000, 3000, and 5000 cycles contained larger grains that emerge from the underlying polycrystalline layer. For films deposited over 7000 cycles, the larger grains were no longer observed, and a more uniform surface morphology was obtained.

Figure 3

Representative plane-view FESEM micrographs of Cu2O films deposited at 200 °C over a cycle number of (a) 500 (6.5 nm), (b) 1000 (11.3 nm), (c) 3000 (24.2 nm), (d) 5000 (43.0 nm) and (e) 7000 (58.0 nm). The 200 nm scale bar applies to all images.

Representative plane-view FESEM micrographs of pan class="Chemical">Cu2O films depopan class="Chemical">sited at 200 °C over a cycle number of (a) 500 (6.5 nm), (b) 1000 (11.3 nm), (c) 3000 (24.2 nm), (d) 5000 (43.0 nm) and (e) 7000 (58.0 nm). The 200 nm scale bar applies to all images.

The surface roughness of the Cu2O films was quantified using AFM. The average root-mean-square roughness (Rq) values of Cu2O films deposited with 500, 1000, 3000, 5000, and 7000 cycles are listed in Table . Top-down view of AFM images of these films is presented in Figure S3. According to the AFM results, the surface roughness of Cu2O films deposited at 200 °C increases with the increasing film thickness. This effect is common for polycrystalline ALD films and can be attributed to a film growth mechanism where the grain size grows with increasing film thickness.[55]

Table 1

Root-Mean-Square Surface Roughness (Rq) Values of Cu2O Films Deposited at 200 °C

cycle count	500	1000	3000	5000	7000
film thickness (nm)	6.5	11.3	24.2	43.0	58.0
Rq (nm)	4.0	1.9	3.4	7.7	9.3

In comparison to other ALD pan class="Chemical">Cu2O processes, the deposition process presented herein produces films of similar surface roughness. For the Cu(dmamb)2 + H2O process, Rq values of 4.8 and 6.1 nm were reported for 30 nm thick Cu2O films obtained at 180 and 240 °C, respectively.[37] These values are comparable to the root-mean-square roughness of 3.4 nm measured for a 24 nm thick Cu2O film deposited at 200 °C. Notably, the surface roughness of the film deposited by 500 cycles is 4.0 nm, which corresponds to approximately 60% of the nominal thickness of this film. Since 500 deposition cycles at 200 °C produces isolated islands instead of continuous films, high surface roughness is to be expected.

To study the early stage growth in more detail, the surface roughness of films deposited with 200, 400, 600, and 800 cycles was mapped. From the AFM images presented in Figure , it can be observed that the film depon>an class="Chemical">sited with 200 cycles consists primarily of isolated copper oxide islands on the Si substrate surface, whereas 400 and 600 deposition cycles lead to the formation of a surface texture consisting of larger grains, and subsequently, to increased surface roughness (3.6 nm for 400 cycles and 4.4 nm for 600 cycles). After 800 deposition cycles, the film surface starts to smoothen, as evidenced by a decrease of Rq to 3.5 nm. In comparison to a film deposited over 1000 cycles (film thickness 11 nm, Rq = 1.9 nm), the sample deposited with 800 cycles has a higher surface roughness, which implies that at 200 °C continuous Cu2O layers are formed at a film thickness of approximately 11 nm. We base this conclusion in the observation that a minimum in surface roughness is observed at 1000 deposition cycles (Figure S4).

Figure 4

Top-down view of AFM images of Cu2O films deposited over (a) 200, (b) 400, (c) 600, and (d) 800 cycles at 200 °C. The 1 μm scale bar and the 25 nm height scale apply to all images.

Top-down view of AFM images of pan class="Chemical">Cu2O films depopan class="Chemical">sited over (a) 200, (b) 400, (c) 600, and (d) 800 cycles at 200 °C. The 1 μm scale bar and the 25 nm height scale apply to all images.

Compositional Analysis of the Films

The composition of a pan class="Chemical">Cu2O film deposited at 200 °C with 7000 cycles was studied using the combination of time-of-flight elastic recoil detection analysis (ToF-ERDA) and XPS. ToF-ERDA is an ion-beam technique that can be used to quantitatively detect all elements and that can be also used for depth profiling with a resolution in the order of nanometers, which makes this method useful in compositional analysis of thin films.[59] According to ToF-ERDA measurements, the Cu/O ratio in the analyzed film was 1.96, which signifies that the films are nearly stoichiometric Cu2O (Table ).

Table 2

Elemental Composition (atomic %) of a Cu2O Film Deposited over 7000 Cycles at 200 °C as Determined Using ToF-ERDA

Cu	O	H	C	Cu/O ratio
65.8	33.6	0.4	≤0.2	1.96

The composition of the films differs from the stoichiometric value of 2.0, as the films were also found to contain minor amounts of light impurity elements, 0.4 atom % of pan class="Chemical">hydrogen, and ≤0.2 atom % of carbon. According to the ToF-ERDA depth profile presented in Figure a, the hydrogen in the films is concentrated on the surface, while the bulk of the films is free of impurities. This suggests that the surface impurities are hydroxides that form upon exposure to ambient moisture.

Figure 5

ToF-ERDA and XPS analysis of Cu2O films deposited at 200 °C. (a) ToF-ERDA depth profile, (b) high-resolution photoelectron spectrum of the Cu 2p3/2 binding energy region, (c) Auger electron spectrum in the Cu LMM KE region, (d) high-resolution photoelectron spectrum of the O 1s binding energy region.

ToF-ERDA and XPS analypan class="Chemical">sis of Cu2O films deposited at 200 °C. (a) ToF-ERDA depth profile, (b) high-resolution photoelectron spectrum of the Cu 2p3/2 binding energy region, (c) Auger electron spectrum in the Cu LMM KE region, (d) high-resolution photoelectron spectrum of the O 1s binding energy region.

We emphasize that the amount of n>an class="Chemical">hydrogen and carbon impurities observed in the Cu2O films is exceptionally low for ALD metal oxide films. The low level of hydrogen and carbon impurities incorporated in the films indicates that the film-forming surface reactions between CuOAc and water vapor proceed in a clean manner without ligand decomposition. This is to be expected, as the most probable film-forming mechanism in ALD of metal oxides when using water vapor as the co-reactant is ligand exchange.[60] In the case of a metal acetate precursor such as CuOAc, the by-product of a ligand exchange reaction is acetic acid, which has a high vapor pressure and is thus unlikely to remain on the film surface.[61] The existence of the ligand exchange mechanism is supported by an in situ reaction mechanism study performed by employing quadrupole mass spectrometer (QMS) and quartz crystal microbalance (QCM) techniques (Figure S5). The Supporting Information also contains further discussion on the question of the reaction mechanism.

To complement the ToF-ERDA expn>eriments, the n>an class="Chemical">Cu2O films were analyzed with respect to the oxidation states of copper and oxygen using XPS. Peak fitting of the photoelectron spectra was performed using the parameters published by Biesinger et al.[62,63] The XPS survey spectrum and a spectrum covering the entire Cu 2p binding energy region are shown in Figure S6. The most intense peak in the Cu 2p3/2 binding energy region was found at 932.4 eV, which lies 0.2 eV higher than the reported literature value for Cu2O, 932.2 eV (Figure c).[63] Moreover, the main peak for Cu 2p3/2 at 932.4 eV shows asymmetry toward higher binding energy values. We assign the minor peak shift and asymmetry to partial surface oxidation, i.e., the formation of a Cu2+ terminated surface. Accounting for the minor oxidation of the film surface, the best fit to the main Cu 2p3/2 peak was obtained by deconvolution to Cu2O (the major peak at 932.4 eV) and CuO (the broad peak at 933.6 eV). The partial oxidation of surfaces is a well-known phenomenon for several thin film materials, including Cu2O.[64,65] However, it has been shown that the surface oxidation of nanocrystalline Cu2O in ambient conditions proceeds only to a minor extent,[65−67] as can also be observed for the films deposited in this work.

Further evidence for the deposition of pan class="Chemical">Cu2O films with minimal surface oxidation was obtained from the Cu LMM spectrum (Figure c), which showed an intense peak assignable to Cu2O at 916.7 eV.[63] Compared to the Cu 2p binding energy region, the Cu LMM peak serves as a reliable fingerprint for identifying between metallic Cu and Cu2O. Although the Cu 2p3/2 binding energy values of Cu and Cu2O are separated only by 0.4 eV, the literature values for LMM kinetic energy of Cu (918.6 eV) and Cu2O (917.0 eV) are separated by 1.6 eV and are therefore more suitable for identification between the metal and oxide.[63] Moreover, the Auger parameter, i.e., the sum of the Cu 2p3/2 main peak binding energy and the Cu LMM main peak kinetic energy is 1849.1 eV, which is in excellent agreement with the literature value of Cu2O (1849.2 eV).[63]

To complete XPS analysis, the pan class="Chemical">O 1s spectra were investigated (Figure d). Deconvolution of the asymmetric peak found in the O 1s binding energy range showed maxima at 530.3 and 531.6 eV. The peak at 530.3 eV corresponds to the lattice oxide of Cu2O, whereas the peak at 531.6 eV can be ascribed to a defective oxygen component in Cu2O, or to surface hydroxyl species.[62,63] Both of these options are possible sources for the O 1s signal present at 531.6 eV since the films were found to be slightly understoichiometric and contain excess hydrogen on the film surface (Figure a).

Optical and Electrical Characterization

Cu2O thin films hold potential for optoelectronic applications that rely on visible light because Cu2O has a high absorption coefficient and an optical band gap in the visible range.[4] To estimate the optical band gap of our Cu2O films, we employed the Tauc plot method.[68,69] Furthermore, to show that ultra-thin ALD films can be used as photoconductors, we prepared simple test structures (Figure a) and characterized the devices with respect to responsivity and time constants.

Figure 6

(a) Schematic presentation of the photoconductor test structure. Inset: digital optical microscopy image of the photoconductor structure from the glass side. (b) Photoconductor responsivity as a function of illumination wavelength and an absorbance spectrum of a Cu2O film deposited by 3000 cycles at 200 °C. (c) Dependence of photocurrent and responsivity on light power density. (d) Time-dependent photocurrent response of the photoconductor illuminated for 1 min. Inset: time-dependent photoresponse with chopped illumination (10 s on, 10 s off). In (b)–(d) the photoconductor was under a 10 V bias. In (c) and (d) a 405 nm laser was used for illumination.

(a) Schematic presentation of the photoconductor test structure. Inset: digital optical microscopy image of the photoconductor stn>an class="Chemical">ructure from the glass side. (b) Photoconductor responsivity as a function of illumination wavelength and an absorbance spectrum of a Cu2O film deposited by 3000 cycles at 200 °C. (c) Dependence of photocurrent and responsivity on light power density. (d) Time-dependent photocurrent response of the photoconductor illuminated for 1 min. Inset: time-dependent photoresponse with chopped illumination (10 s on, 10 s off). In (b)–(d) the photoconductor was under a 10 V bias. In (c) and (d) a 405 nm laser was used for illumination.

Tauc plot analysis for a n>an class="Chemical">Cu2O film deposited with 3000 cycles at 200 °C revealed a direct forbidden optical band gap of approximately 2.1 eV (Figure S7), which is in agreement with values reported for Cu2O thin film samples.[4] Characterization of the photoconductor structures showed that the spectral responsivity of the devices follows the optical absorbance of the Cu2O films (Figure b) and that the responsivity increases with increasing photon energy when the incident illumination power density is kept constant. Further photocurrent measurements revealed a definite dependence between responsivity and the illumination power density (Figure c). Fitting a power law function ΔI = aP, where ΔI is the photocurrent, P is the incident illumination power density, anda and b are constants yields ΔI ≈ P0.5. As evident from Figure c and the fitted power law exponent, the dependence of responsivity on illumination power density is nonlinear, which is typical for photoconductor structures. The highest responsivity, approximately 90 mA/W, was observed when using 405 nm illumination at an incident power density of 1 mW/cm2. The responsivity of our devices based on 24 nm thick Cu2O films is comparable to values reported for photodetectors based on electrodeposited, approximately 1 μm thick Cu2O films.[70]

In photodetection, time constants for the response (τr) and decay (τd) are of interest. While our proof-of-concept photoconductor devices displayed promising respn>onn>an class="Chemical">sivities, temporal photocurrent measurements revealed slow switching characteristics (Figure d). The photocurrent did not saturate even after 1 min of illumination and the time needed for recovery was correspondingly long, in the order of 5 min. The response and decay processes can be represented by a biexponential model incorporating fast (τr1, τd1) and slow (τr2, τd2) components. From biexponential curve fits shown in Figure d, we obtain time constants of τr1 = 2 s, τd1 = 22 s, τr2 = 24 s, and τd2 = 352 s. We propose that the slow switching behavior originates from the persistent photoconductivity (PPC) effect intrinsic to Cu2O. Two models have been proposed to explain the PPC effect in Cu2O, that is, a model based on dissociation and association of copper vacancy complexes by Kuzêl,[71] and an electron trapping model by Tapiero et al.[72] The PPC effect causes the conductivity of Cu2O to increase slowly under illumination and the increased conductivity to persist even when the illumination is stopped. For photoconductor devices, this implies an increase in the baseline current under illumination (Figure d inset). In bulk crystals of Cu2O, the increase in conductivity after illumination can persist for several days.[73] However, in the case of our ultra-thin photoconductors, the nanometer level thickness of the device contributes to a significantly faster PPC discharge. The PPC effect in Cu2O can also be discharged by heating the material to 120 °C and above,[73] however, such an approach is impractical for photodetection. Alternatively, introduction of dopants to the Cu2O lattice can be used to suppress the PPC effect. Isseroff et al. suggested that lithium doping can suppress PPC in Cu2O,[74] which in turn could enable photodetection at operating temperatures close to RT, and furthermore lead to improved response and decay times in photodetector devices based on Cu2O. As ALD is known for its capability to produce doped transition metal oxide thin films of high-quality,[43] we expect more detailed studies on this topic to emerge in the future.

Conclusions

In this work, we have studied the applicability of copper(II) acetate as a n>an class="Chemical">copper precursor in ALD. Deposition of high-quality Cu2O thin films was achieved in the temperature range of 180–220 °C when water vapor was used as the co-reactant. The ALD process described herein exhibits saturative growth characteristics, enables accurate thickness control, and can be used to deposit conformal films on trench structures (AR 2:1). GI-XRD, ToF-ERDA, and XPS measurements showed that the films deposited at 200 °C are polycrystalline, phase-pure, and nearly stoichiometric Cu2O with less than 0.5 atom % hydrogen and carbon impurities. As the deposition temperature range of this ALD process matches well with processes reported for other transition metal oxides, the Cu(OAc)2 + H2O ALD process can enable new routes for depositing ternary and quaternary Cu+-based materials, such as CuAlO2 and SrCu2O2. Finally, as a proof-of-concept, we have demonstrated that ultra-thin Cu2O films made using ALD exhibit appreciable photoconductivity and photoresponsivity in the visible wavelength region. The use of Cu2O thin films for photodetection applications is, however, obstructed by the persistent photoconductivity phenomenon, which manifests as the slow response and decay times. According to our experiments, the adverse effect of persistent photoconductivity is less prominent in Cu2O thin films than in bulk crystals of Cu2O. Further material engineering efforts, such as cation doping may improve the performance of Cu2O-based devices and should thus be explored.

Materials and Methods

Film Deposition

Cu2O thin films were depon>an class="Chemical">sited using a commercial, hot-wall F-120 ALD reactor operated in the cross-flow configuration.[75] Nitrogen (AGA, 99.999%, O2 ≤ 3 ppm, H2O ≤ 3 ppm) was used as both carrier and purging gas at a flow rate of 400 sccm. The reactor pressure during the depositions was in the order of 10 mbar. Films were deposited on 5 × 5 cm2 squares of soda lime glass (SLG) and native oxide-terminated Si(100) (Okmetic Oy, Vantaa, Finland). The SLG substrates were cleaned using ultrasonication with successive baths of an alkaline cleaning solution, absolute ethanol, and deionized water, whereas the Si substrates were used as received. Copper(II) acetate was evaporated from an open glass boat held inside the ALD reactor at either 175 or 185 °C depending on the deposition temperature. Both the monohydrate, Cu(OAc)2·H2O (Sigma Aldrich, ≥99.0%) and anhydrous Cu(OAc)2 (Sigma Aldrich, 99.99%) were used in the deposition experiments. Water vapor was led to the reactor through a needle valve from an external reservoir held at room temperature.

Film Characterization

All characterization studies were performed using films deposited on Si(100) substrates unless otherwise noted. Film thickness was measured with X-ray reflectivity using a PANalytical X’Pert Pro MPD diffractometer equipped with an X-ray source emitting Cu Kα radiation (λ = 1.54 Å). The same diffractometer was also used to perform X-ray diffraction (XRD) experiments. The XRD measurements were performed in the grazing incidence (GI) geometry at an incidence angle of 1°.

Electron microscope images of the films were obtained using a Hitachi S4800 field emisn>an class="Chemical">sion scanning electron microscope (FESEM). Atomic force microscopy (AFM) experiments were performed using a Veeco Multimode V tool equipped with a Nanoscope V controller. The AFM images were captured in the intermittent contact mode (tapping mode) in air using Si probes with a maximum nominal radius of 10 nm (Bruker). To remove artefacts caused by sample tilt and scanner bow, the images were flattened using an algorithm included in the Nanoscope 1.6 software package (Bruker). Film roughness was calculated from 2 × 2 μm2 images as root-mean-square (Rq) values.

The film composition was studied un>an class="Chemical">sing time-of-flight elastic recoil detection analysis (ToF-ERDA). The ToF-ERDA measurements were conducted using a 50 MeV 127I9+ ion beam in a configuration described in full elsewhere.[76] The chemical state of the films was studied using X-ray photoelectron spectroscopy (XPS). The XPS spectra were collected with an Omicron ARGUS spectrometer operated at a pass energy of 20 eV. The thin film samples were excited with X-rays emitted from a standard Mg source (Kα line, photon energy of 1253.6 eV). No sputtering was performed prior to the measurements. The binding energy scale was calibrated using the C 1s peak of ambient hydrocarbons located at 284.8 eV. Peak fitting was performed using the CasaXPS software package (www.casaxps.com).

Absorbance spectra of Cu2O films depon>an class="Chemical">sited on SLG substrates were measured in the wavelength range of 200–800 nm using a Shimadzu UV-2600 spectrophotometer. Tauc plots constructed from the absorbance data were used to estimate the optical band gaps of the films. The band gap analysis was performed for a direct forbidden band gap, as prompted by Malerba et al.[4]

Photoconductor Fabrication and Electrical Measurements

Photoconductor structures were made by evaporating n>an class="Chemical">copper electrodes through a source–drain shadow mask (Ossila E321, channel width 1 mm, channel length 30 μm) onto Cu2O films deposited on SLG substrates by 3000 cycles at 200 °C (film thickness of approximately 24 nm). Electrical measurements were performed using a Keithley 2450 sourcemeter. Laser diodes emitting at 405, 450, 532, 635, 780, and 980 nm (Thorlabs) were used to illuminate the devices from the film side. The incident power density on the device surface was controlled with neutral density filters. Due to the persistent photoconductivity effect in Cu2O, responsivity values were calculated from the difference between the photocurrent and dark current obtained from temporal photocurrent measurements (sample illuminated for 10 s and kept in the dark for 10 s).