1. Introduction
Cu
3N is a novel semiconductor in which crystal imperfections such as Cu interstitials (Cu
i) and nitrogen vacancies (V
N) give rise to states that are energetically located inside or very close to the conduction and valence bands, respectively [
1], but do not give rise to any mid-gap states. Consequently, it has been suggested to be suitable as a solar cell absorber in view of the fact that it has an indirect energy band gap of ~1.0 eV [
2], but also due to the fact that n- and p-type doping are possible. However, despite the fact that Cu
3N has been described as a defect-tolerant semiconductor, so far no one has fabricated a working p-n junction solar cell using Cu
3N. In the past, Chen et al. [
3] fabricated a Cu
3N p-n homojunction on indium tin oxide, and Yee et al. [
1] fabricated an Al: ZnO/ZnS/Cu
3N p-n heterojunction, both of which exhibited rectifying behavior but no photogenerated current. This has been attributed to the large concentration of Cu
i defects, which capture electrons and result into substantial Shockley–Read–Hall recombination and quenching of the steady-state minority carrier concentration under illumination. In other words, crystal imperfections such as V
N and Cu
i can still reduce the minority carrier lifetime and prevent the extraction of photogenerated electron–hole pairs in Cu
3N. Nevertheless, Cu
3N has been used successfully for energy storage as it has a cubic anti-ReO
3 crystal structure, belonging to the Pm3m space group (number 221) with a lattice constant of 3.8 Å [
4], and a vacant body center that can readily act as a host for Li ions in batteries [
5]. Cu
3N has been obtained by many different methods including reactive sputtering [
6], molecular beam epitaxy [
7], atomic layer deposition [
8,
9] and pulsed laser deposition [
10,
11]. Recently, we converted Cu into Cu
3N under NH
3: O
2 between 400 °C and 600 °C, and observed distinct spectral features and maxima in differential transmission at 500 nm (≡2.48 eV), 550 nm (≡2.25 eV), 630 nm (≡1.97 eV) and 670 nm (≡1.85 eV) on a ps time scale by ultrafast pump–probe spectroscopy (UPPS) [
12]. These correspond to the M and R direct energy band gaps of bulk-relaxed and strained Cu
3N in excellent agreement with density functional theory (DFT) calculations of the electronic band structure [
12]. This observation of the M and R direct energy band gaps in fact confirmed that Cu
3N has a clean energy band gap. More recently, we also showed that iodine-doped Cu
3N, i.e., I: Cu
3N, is a p-type semiconductor and that the extensive incorporation of I in Cu
3N can be used to convert Cu
3N into γ-CuI, which is a p-type transparent semiconductor that was used in conjunction with n-type Cu
3N, for the fabrication of a γ-CuI/TiO
2/Cu
3N p-n heterojunction that exhibited rectifying current–voltage characteristics [
13]. In the past, most have focused on n-type doping of Cu
3N, such as Gao et al. [
14], who showed that the incorporation of Zn resulted into n-type Cu
3N and increased the carrier density from n = 10
17 to 10
21 cm
−3 with a resistivity of 10
−3 Ω cm. In contrast, only a few have considered p-type doping of Cu
3N, such as Matsuzaki et al. [
15], who used NF
3 for the growth of F-doped p-type Cu
3N. The ability to obtain p-type Cu
3N is important, of course, for the realization of p-n junctions, but it should be noted that I-VII γ-cuprous halides such as γ-CuCl, CuBr and CuI are p-type transparent semiconductors with a zinc blende crystal structure and direct energy band-gaps of 3.3, 2.9 and 2.95 eV, respectively. In addition, they have lattice constants close to that of Si [
16] and may be readily combined with Cu
3N for the realization of novel cuprous electronic and optoelectronic devices.
Here Cu
3N has been obtained by aerosol-assisted chemical vapor deposition (AACVD) using CuCl
2 in CH
3CH
2OH and NH
3. AACVD is a low-cost growth method [
17] that has been used for the growth of a broad range of semiconductors [
18] including III-V semiconductors such as InN, GaN and In
xGa
1−xN [
19]. The growth of these III-Vs is carried out using anhydrous N
2 and NH
3, i.e., O
2 and H
2O are generally avoided and eliminated. In the past, McInnes et al. [
19] used 0.1M GaCl
3 and 0.1M InCl
3 in acetonitrile (CH
3CN), N
2 as carrier gas and anhydrous NH
3 to grow In
xGa
1−xN. The total flow rate through the 0.1 M GaCl
3 and 0.1 M InCl
3 solutions was maintained at 529 mL min
−1, and anhydrous NH
3 was used at a high flow rate of 862 mL min
−1, which assisted in promoting the formation of smaller droplets whilst also ensuring an excess of NH
3 for the deposition of In
xGa
1-xN. The deposition was carried out at 600 °C, which exceeds the melting point of both GaCl
3 and InCl
3 and gave layers with a thickness of ~2 μm. Both GaCl
3 and InCl
3 react directly with NH
3, leading to the deposition of GaN and InN, respectively, while CH
3CN [
20] breaks into HCN and CH
4 at elevated temperatures [
21,
22].
AACVD has also been used for the growth of Cu
3N by Yamaguchi et al. [
23], who obtained Cu
3N on α-Al
2O
3 by AACVD at 300 °C using copper (II) acetylacetonate Cu(O
2C
5H
7)
2 that was dissolved in aqueous NH
3. No CuO or Cu
2O was detected in the Cu
3N despite the fact that Cu(O
2C
5H
7)
2 was used in aqueous NH
3. Others such as Park et al. [
8] used metal organic sources of copper such as C
14H
32CuN
2O
2, which contains oxygen in conjunction with NH
3 for the atomic layer deposition of Cu
3N, but metal organic sources are expensive [
9].
No one has previously attempted to grow Cu
3N using CuCl
2 and NH
3 by AACVD or tried to grow Cu
3N on m-Al
2O
3, which is ideally suited for the growth of cubic and tetragonal crystals. It is found that the reaction of CuCl
2 with NH
3 will not give Cu
3N as in the case of halide vapor phase epitaxy (HVPE) of III-V semiconductors such as In
xGa
1−xN. In contrast, the reaction of CuCl
2 with an excess of NH
3 resulted into the deposition of polycrystalline Cu layers consisting of oriented grains on m-Al
2O
3, which have a higher crystal quality compared to Cu obtained by sputtering [
12] or electron beam evaporation [
15] used previously to obtain Cu
3N under NH
3: O
2. Consequently, the Cu layers on m-Al
2O
3 obtained via the reduction of CuCl
2 under NH
3 at elevated temperatures were converted into cubic Cu
3N under NH
3: O
2 at a lower temperature without exposure of the Cu to the ambient. The Cu
3N layers on m-Al
2O
3 have an anti-ReO
3 cubic crystal structure with a lattice constant of 3.8 Å and are n-type with carrier density n = 2 × 10
16 cm
−3 and mobility µ
n = 32 cm
2/Vs at room temperature. The electrical properties are described in conjunction with theoretical calculations of the conduction band potential profile, surface band bending and depletion in the effective mass approximation.
3. Results and Discussion
The reaction of CuCl
2 in CH
3CH
2OH with an excess of NH
3 did not lead to the direct deposition of Cu
3N, as in the case of HVPE of III-V semiconductors such as In
xGa
1−xN, but resulted into the deposition of metallic Cu on m-Al
2O
3 that had a shiny, reflective surface and metallic conductivity. A typical SEM image of the Cu layer obtained on m-Al
2O
3 at 600 °C is shown in
Figure 1c, from which one may observe that the Cu layer is polycrystalline and consists of grains oriented along a single direction. A higher magnification image is also shown in
Figure 1d, from which it is evident that the grains have sizes of ~5 µm, while a side view of the Cu on m-Al
2O
3 is shown in
Figure 1e, showing that columnar growth occurs. The epitaxial growth of Cu on c-Al
2O
3 and a-Al
2O
3 has been investigated extensively for the growth of high-quality graphene [
24,
25], but only a few have considered the growth of Cu on m-Al
2O
3 [
26]. The deposition of Cu on m-Al
2O
3, which contains grooves or steps along specific crystallographic directions, as shown in
Figure 2a, will lead to instabilities and ruptures of the Cu layer at elevated temperatures [
27]. These ruptures occur at high curvature sites, i.e., peaks and ridges, which act as retracting edges leading to a net flux of atoms away from the high positive curvature regions. For sufficiently thin layers, this process will lead to a self-assembly of the Cu grains along a specific direction [
28].
The Cu layers exhibited clear peaks in the XRD, as shown in
Figure 2a, corresponding to the face-centered cubic (fcc) crystal structure of Cu with a lattice constant of 3.6 Å. No peaks belonging to CuO, Cu
4O
3 or Cu
2O are observed in
Figure 2a. Likewise shown are the peaks corresponding to the underlying m-Al
2O
3, which has an oxygen-terminated surface with tetragonal crystal symmetry that is suitable for the epitaxial growth of semiconductors with a cubic crystal structure. It is worthwhile to point out that the deposition of Cu on n-type Si (001) resulted in columnar growth, as shown in
Figure 3a,b. The Cu pillars have a height of ~20 µm, but they are not ordered in any way. No Cu
3N was obtained under an excess of NH
3 by varying the temperature between 300 °C and 800 °C. Instead, the reaction of CuCl
2 in CH
3CH
2OH with NH
3 always led to the deposition of Cu on m-Al
2O
3, which occurs via the reduction of CuCl
2 to CuCl and then into Cu by the H
2 evolving from NH
3.
More specifically, the 0.1 M solution of CuCl
2 is initially converted into a mist of liquid drops and mixed with NH
3, which is soluble in CH
3CH
2OH [
29]. Subsequently, the liquid drops are vaporized at an elevated temperature, and the CH
3CH
2OH gives C
2H
4 and H
2O according to the reaction C
2H
5OH → C
2H
4 + H
2O. No carbon is released from the pyrolysis of C
2H
4 between 500 °C and 800 °C [
30]. Upon vaporization, CuCl
2, which has a melting point of 498 °C, will be reduced to CuCl, which has an even lower melting point of 423 °C [
31], and finally into metallic Cu by the H
2 evolving from the breakdown of NH
3. Before elaborating further, it is useful to note that the thermal breakdown of NH
3 into N
2 and H
2 was investigated as early as 1905 by White et al. [
32], who showed that it depends on the gas flow, i.e., residence time as well as the temperature. In particular, White et al. [
32] showed that a flow of 200 mL/min NH
3 resulted in a dissociation of 5% NH
3 at 600 °C and 10% at 700 °C. However, the breakdown of NH
3 is also promoted catalytically by the deposited Cu at elevated temperatures [
33]. In other words, the Cu deposited on the m-Al
2O
3 will participate actively in the dissociation of NH
3 near the surface, thereby further promoting the reduction of CuCl
2 and deposition of Cu, which has a melting point of 1085 °C. A schematic representation of the proposed reaction mechanism is shown in
Figure 3c. For completeness, it must also be pointed out that the NH
3 will react with CH
3CH
2OH and give ethylamine (CH
3CH
2NH
2) and acetonitrile (CH
3CN), which have boiling points of 20 °C and 82 °C, respectively. CH
3CH
2NH
2 and CH
3CN will dissociate into HCN and CH
4 depending on the temperature and residence time, but they are not expected to influence the overall reaction governing the deposition of Cu. It is also important to mention that the Cu will tend to react with H
2O supplied from the CH
3CH
2OH and give CuO and Cu
2O. However, no oxides are detected in
Figure 2a, so it is very likely that they are reduced to metallic Cu due to the H
2 evolving from the NH
3 over the Cu. This is consistent with the findings of Kim et al. [
34], who showed that CuO is converted into metallic Cu under an excess of H
2 without the formation of intermediate Cu
4O
3 or Cu
2O.
In short, CuCl2 is reduced to CuCl and then into Cu by the H2 evolving from NH3, according to: CuCl2 + H2 → CuCl + 2HCl and 2CuCl + H2 → 2Cu + 2HCl. The HCl reacted in turn with the excess NH3, giving NH4Cl, i.e., NH3 + HCl → NH4Cl, which solidified below its sublimation temperature, i.e., ~340 °C near the cool end of the reactor, very similar to what occurs during conventional HVPE of III-Vs.
The reduction of CuCl
2 into Cu may also be achieved by using H
2 as opposed to NH
3. In order to show this, the 0.1 M solution of CuCl
2 in CH
3CH
2OH was used to deposit a layer of CuCl
2 on 15 mm × 30 mm soda lime glass (SLG) slides by drop-casting, as shown in
Figure 4a. The CuCl
2 layer had a light green color and good uniformity, and a typical SEM image is shown in
Figure 4b. This was converted into Cu under a flow of (i) 10 and (ii) 50 mL.min
−1 pure H
2 at 400 °C for 30 min, as shown schematically in
Figure 4c. The CuCl
2 as-deposited on SLG displayed a crystalline structure and multiple peaks in the XRD, as shown in
Figure 4d, but all the peaks were eliminated after the reduction of the CuCl
2 into Cu.
The Cu deposited on m-Al
2O
3 at 600 °C by AACVD using CuCl
2 and NH
3 has a higher crystal quality compared to the Cu obtained by sputtering, which was nonetheless successfully converted into crystalline Cu
3N under a flow of 300 mL/min NH
3 and 15 mL/min O
2 between 400 °C and 600 °C, as shown previously [
12]. The Cu
3N obtained in this way had an anti-ReO
3 cubic crystal structure, and we observed the M and R direct energy band gaps of Cu
3N by UPPS in excellent agreement with DFT calculations of the electronic structure, confirming that it has a clean energy gap [
12]. Consequently, the polycrystalline Cu layer that was obtained by AACVD on m-Al
2O
3 at 600 °C was converted into Cu
3N under a flow of 300 mL/min NH
3 and 15 mL/min O
2 at 400 °C. The Cu
3N had an olive-green-like color, and a typical SEM image of the Cu
3N layer on m-Al
2O
3 is shown in
Figure 1f. This exhibited peaks in the XRD, as shown in
Figure 2b, corresponding to the anti-ReO
3 cubic crystal structure of Cu
3N with a lattice constant of 3.8 Å.
The reaction of Cu with NH
3 containing O
2 and the formation of Cu
3N can be understood by considering the catalytic oxidation of NH
3 by O
2 in the presence of a catalyst, e.g., Cu, Pt, etc., at elevated temperatures, as described by Carley et al. [
35], who investigated the catalytic reactivity of Cu (110) metal surfaces with coadsorbed NH
3 and O
2. More specifically, Carley et al. [
35] proposed that the oxidation of NH
3 leads to the formation of a stabilized N monolayer on the Cu metal surface, which in turn is responsible for the conversion of the bulk Cu layer into Cu
3N. It should be noted that the reaction of NH
3 with O
2 also gives H
2O according to the reaction NH
3 + O
2 → NO + H
2O, which was observed to condense near the cool end of the reactor upon increasing the gas flow of O
2. The reaction mechanism of the conversion of Cu into Cu
3N is depicted schematically in
Figure 3d. No Cu
3N was obtained from Cu by using only NH
3, in accordance with Matsuzaki et al. [
15]. Moreover, no CuO or Cu
2O peaks are detected in
Figure 2b, but Cu
2O will nevertheless form as native oxide on the surface of the Cu
3N upon exposure to the ambient, as we have shown previously by using Raman spectroscopy [
36]. Before considering the electrical properties of the Cu
3N layers, it is useful to point out that the reaction of CuCl
2 with a smaller flow of 100 mL/min NH
3 at 600 °C mainly led to the deposition of Cu
2O, not Cu
3N.
In order to measure the Hall effect, Ag ohmic contacts were deposited at the four corners of the Cu
3N layers on m-Al
2O
3. It has been shown that Au, Ag and Cu in Cu
3N give rise to a semiconductor-to-metal transition and remarkably constant electrical resistivity over a very broad range of temperatures [
37]. Consequently Ag, Au and Cu may be used for the formation of ohmic contacts on Cu
3N, and in the past, we have shown that Au and Ag deposited on n-type Cu
3N results in the formation of contacts with linear IVs [
13]. The Cu
3N layers on m-Al
2O
3 were found to be n-type and had room temperature carrier densities of 2 × 10
16 cm
−3 with a maximum mobility of 32 cm
2/Vs. The Cu
3N layers are n-type as they are Cu-rich, but also due to the fact that oxygen may be included in the Cu
3N by the preferential formation of interstitial oxygen (O
i) that acts as donors, not as acceptors [
36]. Furthermore, the Cu
3N layers obtained here were found to be persistently n-type, and the carrier density and mobility did not exhibit any changes upon illumination with light of λ = 450 nm under ambient conditions. In other words, the n-type Cu
3N layers did not exhibit any photoconductivity, which may be attributed to recombination via Cu
i and V
N states, in accordance with Yee et al. [
1].
It is worthwhile pointing out here that Matsuzaki et al. [
15] showed that epitaxial Cu
3N layers with a thickness of 25 nm on SrTiO
3 were p-type, due to the upward surface band bending mediated by the chemisorption of O
2−, but switched to n-type upon exposure to ultraviolet (UV) light and reverted back to p-type after terminating the irradiation. In contrast, they observed that the Cu
3N layers remained n-type after exposure to UV light under vacuum, confirming that the adsorbed O
2− is responsible for the surface inversion observed under ambient conditions in air. However, the epitaxial Cu
3N layers of Matsuzaki et al. [
15] were found to be persistently n-type under ambient conditions, with a carrier density of the order of 10
14 cm
−3 and mobility of 100 cm
2/Vs after annealing under NH
3 between 125 and 350 °C, suggesting a change in the composition of the surface and overall band bending. The Cu
3N layers obtained here were found to be persistently n-type and had a room temperature carrier density of 2 × 10
16 cm
−3, perhaps due to the fact that after the conversion of Cu into Cu
3N under NH
3: O
2, the flow of NH
3 was maintained for at least 30 min until the temperature fell well below 100 °C.
However, the properties of Cu
3N layers with a thickness of a few tens of nm will depend strongly on the properties of the surface but also the properties of the underlying substrate that is often overlooked. The Cu
3N layers obtained here are persistently n-type with a carrier density of 2 × 10
16 cm
−3, most likely due to the fact that the thickness of the Cu
3N layers is greater than 1 µm, so it is bulk-like and will not be strongly influenced by properties of the surface or underlying m-Al
2O
3. In thermodynamic equilibrium, the Fermi level (
EF) with respect to the conduction band minimum (
EC) away from the surface and deep in the bulk is determined from:
where
NC is the conduction band effective density of states,
k is Boltzmann’s constant and T the temperature taken to be equal
T = 300 K. The electron density is equal to
n = 2 × 10
16 cm
−3, and the conduction band effective density of states in Cu
3N is given by:
where
mn is the electron effective mass in Cu
3N taken to be m
n= 0.16 m
o [
6], m
o is the free-electron mass and h is Planck’s constant. This gives
NC = 1.6 × 10
24 m
−3 or 1.6 × 10
18 cm
−3, so
EC −
EF = 0.11 eV in the bulk where a flat band condition exists. On the other hand, the energetic position of the Fermi level with respect to the conduction band edge, i.e.,
EC −
EF, at the surface is dependent on the local density and energetic position of any surface states that will be occupied by electrons, which in turn may pin the Fermi level at the surface. According to Navío et al. [
38], the Fermi level at the surface of ultrathin Cu
3N layers is pinned at the middle of the gap, which will give rise to a barrier height of ϕ
b = 0.5 eV. The surface depletion region will extend into the Cu
3N, and the depletion width is:
where
εS =
εRεo,
εR is the static dielectric constant of Cu
3N,
εo the permittivity of free space,
e the electron charge and
ND the donor density taken to be equal to
n = 2 × 10
16 cm
−3. Considering that the static dielectric constant of Cu
3N is
εR ~10 [
39], the depletion width is found to be
w = 0.16 µm, taking into account that the Fermi level at the surface of Cu
3N layers is pinned at the middle of the gap, according to Navío et al. [
38]. However, despite the fact that we did not detect any CuO or Cu
2O in the XRD, a thin layer of Cu
2O will exist on the surface of Cu
3N. According to Hodby et al. [
40], the Fermi level at the surface of Cu
2O is pinned at states residing energetically in the upper half of the band gap ~0.4 eV below the conduction band edge. The native Cu
2O layer of Cu
3N is expected to have a thickness of only a few nm and will be completely depleted, so the depletion width taking ϕ
b = 0.4 eV is found to be
w = 0.15 µm. The conduction band potential profile of the Cu
3N layer including the native Cu
2O layer at its surface is shown in
Figure 5a, where the work function and electron affinity of Cu
3N i.e., ϕ(Cu
3N) = 5.0 eV and χ(Cu
3N) = 3.5 eV [
41] have been considered as well as the work function and electron affinity of Cu
2O, i.e., ϕ(Cu
2O) = 4.8 eV and χ(Cu
2O) = 3.2 eV [
42]. The formation of p-type Cu
2O over the n-type Cu
3N will lead to the confinement of photogenerated electron–hole pairs at the Cu
2O/Cu
3N heterojunction, which will inadvertently result into recombination via states at the interface, thereby suppressing the photoconductivity. This mechanism is different to that put forward by Yee et al. [
1], who fabricated an Al: ZnO/ZnS/Cu
3N p-n heterojunction that exhibited rectifying behavior but no photogenerated current, which was attributed to the large concentration of Cu
i defects that capture electrons and result in substantial Shockley–Read–Hall recombination and quenching of the steady-state minority carrier concentration under illumination. While it is possible that both mechanisms are responsible for the suppression of the photocurrent and photoconductivity in Cu
3N, it is imperative that the surface recombination should be suppressed via the deposition of suitable layers that prevent the formation of Cu
2O that was originally suggested to act as a suitable passivation layer for Cu
3N, similar to that of SiO
2 for Si p-n junction solar cells [
2].