Tuning the electrical properties of the p-type transparent conducting oxide Cu1−xCr1+xO2 by controlled annealing

Off-stoichiometric copper chromium oxide delafossite received lately a great interest due to its high p-type electrical conductivity and adequate optical transmittance in the visible range. However, for a suitable integration in active devices such as p-n junctions, transistors or optoelectronic devices, the electronic properties must be efficiently tailored. Here, post-deposition thermal treatment is proven as an adequate approach for finely controlling the electrical properties of this former degenerate semiconducting material. The energetics of the annealing process are investigated using two different approaches, as a function of the annealing temperature and as a function of the annealing time, allowing the accurate determination of the activation energy of the annealing of defects. By using this method, the electrical carrier concentration was varied in the 1021 – 1017 cm−3 range while the recorded changes in the drift mobility covered three orders of magnitude. Moreover, we demonstrate the ability to accurately manipulate the Fermi level of such materials, which is of great importance in controlling the carrier injection and extraction in optoelectronic active layers.

In the field of transparent conductive oxides (TCOs) copper-based delafossite materials are promising candidates as p-type transparent semiconductors closing the gap of optical and electrical properties towards current standard n-type semiconductors (a transmittance greater than 80% in the visible range and an electric conductivity up to 1000 S cm −1 ). The interest in these peculiar delafossite compounds was ignited after the report of CuAlO 2 as a first p-type semiconductor with decent optical transparency 1 in the visible range and was reinforced after reporting a breakthrough of electrical conductivity up to 220 S cm −1 obtained for Mg-doped CuCrO 2 2 . Various copper-based delafossites CuMO 2 (M = Cu, Cr, Ga, In, Fe, B) were thoroughly studied 3 in the effort to understand the rationales of the p-type conductivity and the electrical transport mechanism within for optimizing subsequently theirs electrical and optical properties. Copper vacancies 4 or oxygen interstitials 1,5 were mainly suggested as the p-type doping source whilst small polaron 6,7 or band conduction models 8,9 were proposed to elucidate the conduction mechanism in such materials. Moreover, recent reports had shown large conductivity (larger than 10 S cm −1 ) and adequate transparency for highly off-stoichiometric copper chromium delafossites. [10][11][12][13] In these particular compounds the crystalline phase of delafossite is preserved although a copper deficiency up to 33% is observed.
In our previous work 14 we reported the synthesis and the characterization of such off-stoichiometric Cu-Cr-O delafossite thin films with conductivities greater than 100 S cm −1 and optical transmittances up to 50%. Finite lines of copper vacancies chains randomly distributed within crystalline grains were observed (in Transmission Electron Microscopy) and furthermore suggested as the possible source of high doping in as-deposited films. A peculiar stoichiometry with a 33% copper deficiency compensated by a surplus of 33% chromium (i.e. Cu 0 . 66 Cr 1.33 O 2 ) was found in both as-deposited and annealed (900 °C) films whilst the delafossite crystalline structure remains unaltered. However, after annealing a drop of the carrier's concentration from 10 21 to 10 17 cm −3 or even lower was measured and associated with the healing of the chained defects. We then suggested that the main driving force leading the changes in the defect chemistry is the "healing" of defects, a process driven by short-range structural changes. In the present paper we further investigate how a controlled post-deposition thermal treatment can be used as a very effective approach for tailoring electrical properties of Cu 0.66 Cr 1.33 O 2 . Such fine tuning is required for the electrical engineering of transparent devices such as p-n junctions or p-type transistors 15 . In order to achieve this goal, the metastable nature of the Cu-vacancies chains described above was investigated. Two different types of thermal treatments were used in the present study in order to investigate the thermodynamics of the annealing process: different annealing times at a fixed temperature (900 °C) or a fixed amount of time (900 s in this case) at various temperatures (from 650 °C to 850 °C). The first treatment demonstrates conditions for a fast process which is more adequate to technological applications where long lasting processes might be considered costly. The temperature is situated safely lower than 1100 °C, the reported limit of the materials stability for the copper delafossite phase 16 . The second treatment, involving lower temperatures, allows a better control due to the smooth variation of electrical properties. The experimental results show that the controlled thermal treatment can be used as a versatile process for controlling the holes concentration, the electrical mobility or even the electronic work function, all being key features for the engineering of electronic properties of transparent solid state devices.

Results and Discussions
We start by investigating the chemical composition of as-deposited and annealed films for various annealing time intervals in order to confirm the stability of the material's chemistry upon thermal treatment. Figure 1(a) depicts XPS results for as-deposited films and for films annealed for 30 s and respectively 4000 s. The XPS spectra look similar, suggesting no major changes of chemical concentrations. Besides XPS characteristic peaks for Cu (2p,2s) , Cr (2p,2s) and O 1s , Auger O KLL Cu LMM and Cr LMM peaks are present in the spectra 17 . The positions of Cu 2p peaks (1/2 at 932.6 eV and 3/2 at 952.5 eV) do not vary upon annealing. The distance between them is 19.9 eV, a clear indication of the delafossite phase. No satellite peak of Cu is observed, confirming that only Cu in +1 oxidation state is detected by XPS 13 . The Cr 2p peaks are observed at binding energies of 576.6 eV (3/2) and 585.6 eV (1/2) respectively. The distance between Cr 2p and O 1s remains at a constant value of 45.3 eV for all samples. Moreover, the Auger Cu LMM peak observed at a binding energy of 568.6 eV confirms the purity of our delafossite phase 18 .
The chemical compositions for as-deposited and annealed films are shown in Fig. 1(b). Concentrations around 16, 33 and 50% are measured (within the XPS measurement accuracy) for Cu, Cr and O respectively. This peculiar stoichiometry may lead to deviations in the partial charge of each atom as compared to the stoichiometric CuCrO 2. The bounds are known to be partially ionic and partially covalent and consequently the oxidation state of each atom in the crystal is not directly equal to the most stable oxidation state of each atom 19 . The X-Ray Diffraction, XRD, ( Figure S1) analysis confirms furthermore the preservation of the delafossite phase after thermal treatments along with an increase of the crystalline grains' size and a relaxation of c lattice parameter ( Figure S2). The change of the grains' size is also visible in AFM scans presented in Figure S3 from the supplementary information section.
Six samples with similar (averaged value: 15.6 ± 6.9 S cm −1 ) conductivities (Table 1) were used for each thermal treatment study (one was kept as reference in each case). Electrical conductivity values of tens of S cm −1 are obtained using the set of deposition's parameters described within the experimental methods section. This level of conductivities is usually reported for off-stoichiometric delafossites thin-films fabricated using non-equilibrium chemical methods as dynamic liquid injection chemical vapour deposition 10,11,14 or chemical spray deposition 12,13 . These values are significantly higher than those reported for intrinsic 20,21 or some doped 22,23 CuCrO 2 delafossites thin-films.
Five of the samples were heated for 900 s at 650, 700, 750, 800 and 850 °C, respectively. For the first sample heated at 650 °C no change was observed after 900 s and consequently the time was furthermore increased up to 3600 s when a three times diminution of the electrical conductivity (σ 0 / σ f ) was finally observed. This is in agreement with the work of Gotzendorfer et al. 23 where changes in electrical properties of CuCrO 2 were observed starting from temperatures around 620 °C. Second set of samples were heated at 900 °C for 30, 60, 200, 1000 and 4000 s respectively. For the last sample the measured conductivity was beyond the sensitivity of our apparatus (10 −4 S cm −1 ). For each sample the Seebeck coefficient and the conductivity were measured before and after the thermal treatment and the results are presented in Fig. 2 and Table 1. The first observation is the negative slope of ∆V th /∆T further proving the p-type nature of our films. For as-deposited samples values of S around 100-130 µV K −1 were measured. For the sample heated at 650 °C a similar value (118 µV K −1 ) of the Seebeck coefficient ( Fig. 2(a)) is measured evidencing a slow healing of Cu-vacancies chains defects at this temperature. Beyond 700 °C important changes appear after 900 s of annealing. The electrical conductivity decreases monotonously with the annealing temperature until a diminution of 50 000 times, as measured in the case of the sample heated at 850 °C. The Seebeck coefficient follows an exponential decay (Fig. 2(c)) with the annealing temperature. This coefficient is directly related to the carriers' concentration which is further dependent on the concentration of defects. In the material investigated here, we have previously demonstrated that the defects are the chains of Cu vacancies observed by TEM experiments 14 . Therefore we may correlate the decrease of the carriers' concentration with the annihilation of these chains of Cu vacancies. A further indication will be presented later when analysing the thermodynamics of the annihilation reaction. The exponential decay confirms the metastable nature of defects and that a thermal activated law governs their healing. Similar healing of defects is observed when using the fast annealing process. The conductivity decreases by two orders of magnitude during short (30-60 s) thermal treatments, followed by a continuous decrease with the annealing time down to the 10 −4 S cm −1 range for the sample heated during 4000 s. The values of the Seebeck coefficient can be regrouped in three main regimes as depicted in Fig. 2 . Again the dependence of the Seebeck coefficient on the annealing time is exponential (Fig. 2(d)). It was previously reported 26,27 that performing Hall measurements on copper p-type delafossites is challenging due to very low values of mobility. Furthermore collected Hall voltage values are comparable with the background signal. Hall measurements at high magnetic fields (up to 9 T) were performed on our as-deposited sample and resulted in highly scattered data points 28 . An averaged value of 0.01 cm 2 V −1 s −1 can be extracted after making a large set of assumptions. This averaged value matches the mobility value as we extract it from our Seebeck measurements. This makes us confident that the combined Seebeck and conductivity measurements provide reliable values of electronic properties of our materials. From the electrical conductivity and the Seebeck coefficient measurements we extract the carriers' concentration of thin films. It is important to distinguish on one side the origin of holes in our system, which are related to peculiar defects (here chains of Cu vacancies), and on the other side the localisation of the free (mobile) holes in our system, which is related to the orbitals constituting the valence band maximum. In Cu-based delafossite materials, it has been widely assumed that the valence band is mainly made from d orbitals of Cu atoms, leading to a transition of the oxidation state of Cu I to Cu II . In conclusion whatever the origin of the holes (chains of Cu vacancies, isolated Cu vacancies, O interstitials, extrinsic dopants etc.), the free holes concentration corresponds to Cu atoms with + II oxidation state acknowledging that the oxidation state is an approximation and does not corresponds to the partial charge of atoms. Using this approach within the Heikes formalism, often used in the case of degenerate semiconductors 26 , the following formula of the Seebeck coefficient is obtained: where g 1 = 1 and g 2 =4 are electron degeneracies and of Cu + and Cu 2+ while [Cu + ]/[Cu 2+ ] is the ratio of Cu having +I and +II oxidation states, acknowledging that the oxidation state doesn't correspond to the spatial charge of atoms. This formula allows the calculation of Cu 2+ fraction sites and furthermore p, the carriers concentration. Finally, the mobility μ can be estimated using σ = peµ. The results are presented in Table 1. It should be mentioned here that for conductive samples (σ > 1 S cm −1 ) the band degenerate 29 and the small polaron models lead to similar values 14 . Moreover for lower conductive samples a good agreement is observed using both small polaron and non-degenerate models 30 . The logarithmic dependence of the carriers' concentration and drift mobilities is depicted in Figs. 2(e and f) for both annealing processes. The decrease of the carriers' concentration is efficiently tailored by controlling the annealing temperature, as in this case a smoother variation is observed. However, if low values are aimed at, the fast annealing may be privileged. An interesting observation is related to the drift  mobility. During both thermal treatments a minimum is observed: after 900 s of heating at 800 °C or after 30 s of heating at 900 °C. This behaviour might be related to the atomic rearrangement during the annealing process. In these conditions, the number of grain boundaries may increase to a maximum and furthermore scattering the carrier flow. The results obtained after the two thermal processes described above are also used to investigate the energetics of doping defects. Assuming a first order kinetics law, the holes concentration p can be expressed as a function of the annealing time t in a general way as: where N a is the acceptor dopant concentration, N res is the residual acceptor concentration after an infinite annealing, N i = N 0 + N res is the initial carrier concentration (as-deposited) and k is the rate constant. The assumption of a first order kinetic is validated by the observed linear dependence when plotting ln(p) vs. t for the data taken at 900 °C ( Fig. 3(a)). Since the data point at 1000 s is already very close to the residual holes concentration p res , only data up to 200 s of annealing are fitted. Consequently a rate constant k = 0.040 ± 0.005 s −1 is obtained. Assuming that the reaction is thermally activated, the reaction constant k can be expressed as a function of the annealing temperature T using an activation energy E a : where A is a constant and k B is the Boltzmann constant. Relating equations (2) and (3), we can express the carrier concentration as a function of the annealing temperature T and annealing time t: The residual dopant concentration can be neglected and the relationship can be thus expressed: Taking the as-deposited carrier concentration as the initial dopant concentration N i = N 0 = 1.68×10 21 cm −3 , we can plot − ln(ln N lnp(T, t)) 0 at fixed annealing time (t = 900) s as a function of 1/T (Fig. 3(b)). The data fit well with the activation energy model, and an activation energy of E a = 1.35 ± 0.07 eV is obtained. In the case of ZnO 31 the activation energy for the annihilation of Frenkel pairs (annihilation of vacancies with nearby interstitials) is about 1 eV while the energy barriers for the migration of zinc and oxygen vacancies are 1.4 eV and 2.4 eV, respectively. Based on this quantitative comparison and on our previous microscopic analysis of defects showing the presence of Cu vacancy chains in as-deposited material, we suggest that the changes in electrical conductivities in our system are triggered by Ostwald ripening mechanisms based on the dissolution of shorter chains of vacancies into single Cu vacancies migrating to grain boundaries or longer chained vacancy defects. In our case, the activation energy would preferably correspond to the energy of migration of single Cu vacancies through the crystal. A detailed analysis of the structural evolution of defects goes beyond the present study but it is noteworthy that our measured activation is in the same range as the migration energy of cations vacancy in oxide semiconductors 32,33 .
where χ is the electronic affinity (2.1 eV for copper delafossites 34 ), Eg is the band gap 3,2 eV 2,12,14 . This value is insignificantly affected by annealing 14 and thus holds for both cases of annealed and as-deposited (transmittance spectra and Tauc plots for as deposited and annealed samples are presented in Figure S5). The results are depicted in Fig. 4(a), where the work-function referenced to the vacuum (knowing that Φ HOPG = 4.6 eV) is shown as a function of the carriers' concentration. For as deposited samples, the Fermi level is situated at 90 meV above the valence band maximum, an ordinary value for degenerate semiconductors. Upon thermal treatment the Fermi level can be tailored upwards to 1.08 eV above the valence band maximum; this corresponds to the samples annealed for 4000 s at 900 °C ( Fig. 4(b)). The Fermi level is thus gradually tailored within this range of energy (90 meV to 1.08 eV) when the thermal treatment is tuned.

Conclusions
We proved that the controlled thermal treatment is an efficient approach for controlling electronic properties of Cu 0.66 Cr 1.33 O 2 , a highly conductive p-type transparent oxide. The holes concentration, the electrical mobility and the electronic work function are key features for transparent solid-state devices and for a suitable integration in any active devices (such as p-n junction, transistors or optoelectronic devices) these properties must be tailored in order to tune the required properties of such devices. Therefore we showed that the carrier concentration was smoothly varied from 10 21 to 10 17 cm −3 , the electrical conductivity swaps down 5 orders of magnitude while the recorded changes in drift mobility covered three orders of magnitude. Moreover, we accurately control the Fermi level which is of critical importance in controlling the carrier injection and extraction in electronic devices. Depending on the aimed final values we proposed a time driven or a temperature driven thermal treatment. From the value of 90 meV for as-deposited samples, the Fermi level can be tailored upwards up to 1.08 eV above the valence band maximum; this is achieved via the annealing of as-deposited sample for 4000 s at 900 °C. The energetics of singular Cu-vacancies chain defects responsible of the high conductivity in this peculiar material is also studied. An activation energy model is used to extract the averaged activation energy of the annihilation of Cu-vacancies chains. We found E a = 1.35 eV which is very similar to the energy of the migration of Zn-vacancies in n-type ZnO-based TCOs.

Methods
Thin films with a thickness around 200 nm were deposited on Al 2 O 3 c-cut substrates using a Dynamic Liquid Injection -Metal Organic Chemical Vapour Deposition system (DLI-MOCVD, MC200 from Annealsys) whilst bis 2,2,6,6-tetramethyl-3,5-heptanedionate compounds were used as precursors for copper and chromium. The deposition parameters are: temperature substrate = 450 °C; oxygen flow = 2000 standard cubic centimetres (sccm); nitrogen flow = 850 sccm; total process pressure = 12 mbar. A detailed description of the deposition approach fabrication can be found elsewhere 35 . The annealing processes were performed in a Rapid Thermal Annealing reactor (Annealsys) under Argon, under vacuum or at high oxygen partial pressure (in same conditions as during the deposition) and the changes in conductivity were not significantly depending on the annealing atmosphere. For the present work the annealing was performed under N 2 + O 2 mixture for the fast annealing and under vacuum for the annealing at various temperatures. Electrical properties were measured using four probes in linear configuration while for the Seebeck effect measurement a homemade system was used with a copper wire as reference. A detailed description of this system is presented on supplementary information section ( Figure S5). For X-Ray Photoemission Spectroscopy (XPS) analysis a Kratos Axis Ultra DLD system using a monochromated (Al Kα: hν = 1486.7 eV) X-ray source was used. Kelvin Probe Force Microscopy (KPFM) measurements have been performed on a Bruker Innova using the amplitude modulation to determine the contact potential difference between the Pt-Ir coated tip and the sample surface. The analysis is done in one pass mode, the first resonance amplitude is used to track the topography while the KPFM signal is acquired at a second frequency (23 kHz). Freshly cleaved Highly-Oriented Pyrolytic Graphite (HOPG) is used as reference. Measurements are performed under dry N 2 atmosphere in order to avoid water condensation on the surface.