Stener
Lie
a,
Qingde
Sun
a,
Pritish
Mishra
ab,
Patrick Wen Feng
Li
b,
Anupam
Sadhu
a,
Teddy
Salim
b,
Shuzhou
Li
b,
Geoffroy
Hautier
c and
Lydia Helena
Wong
*b
aEnergy Research Institute at NTU, Nanyang Technological University, 637141, Singapore
bSchool of Materials Science & Engineering, Nanyang Technological University, 50 Nanyang Ave, 639798, Singapore. E-mail: lydiawong@ntu.edu.sg
cThayer School of Engineering, Dartmouth College, Dartmouth College, Hanover, NH 03755, USA
First published on 17th February 2025
Achieving fully transparent electronic devices requires improving p-type transparent conducting materials (TCMs) to match their n-type counterparts. This study explores novel p-type TCMs using high-throughput screening via an automatic spray pyrolysis system. The performance of conducting wide bandgap chalcogenide based on CuS can be improved by incorporating various cations, with Mg emerging as the most promising candidate. The optimized CuS–Mg films exhibited superior transparency and conductivity, comparable to state-of-the-art p-type TCMs. Density functional theory (DFT) calculations linked the inverse correlation between transparency and conductivity to changes in Cu 3d and S 3p orbital coupling with varying Mg content. The best CuS–Mg composition demonstrated high hole concentration (5 × 1021 cm−3), low sheet resistance (266 Ω □−1), and high transparency (∼75%). The transmittance increased by ∼30% compared with pristine CuS. The successful application of a p-CuS–Mg/n-CdS heterojunction as a semi-transparent photodiode highlights its potential for smart displays and window-integrated electronics. This study demonstrates the value of combining experimental and theoretical methods for accelerated material discovery.
New conceptsThis study presents a high-throughput screening approach using an automated spray pyrolysis system to discover novel p-type transparent conducting materials (TCMs) by combining wide bandgap chalcogenides with CuS. This method not only eliminates the downtime typically associated with traditional material screening techniques but also significantly reduces fabrication time while optimizing processing parameters. By integrating experimental and computational techniques, the approach provides a deep understanding of the mechanisms governing conductivity and transparency in the materials. In contrast to conventional methods that achieve high-performing p-type TCMs through doping of wide bandgap oxides, this approach starts with a highly conductive p-type material, CuS, and enhances transparency through the incorporation of Mg, which weakens p–d coupling and induces amorphization. This process results in a CuS–Mg candidate with a figure of merit among the highest reported for p-type TCMs. This strategy not only introduces a novel mechanism to explain the observed properties but also establishes a versatile framework for accelerating material discovery and optimization across various fields. |
CuS has emerged as a promising candidate for conducting material and has been employed in various optoelectronic devices, including solar cells and photodetectors.18–20 Also known as the covellite phase, it possesses a hexagonal structure with CuS4 tetrahedral and CuS3 triangular coordination. CuS is an intrinsically p-type material with a bandgap ranging from 1.5–2.3 eV and exhibit high conductivity (∼103 S cm−1).21 In addition, CuS is non-toxic, abundant, and cost-effective, making it an ideal material for large-scale production. However, even in its nanostructured morphology, CuS exhibits a relatively narrow band gap (∼2.4 eV),22,23 resulting in poor transparency, particularly in the visible range. Therefore, it is crucial to develop a strategy to enhance its transparency and enable it to be an efficient transparent conductor. One approach to improve the transparency of CuS is to reduce its thickness to be ultrathin. For instance, CuS nanowires have demonstrated very high transparency of up to 80%.24 However, this method presents challenges, such as the requirement for encapsulation or protection layers such as polymer layers to prevent the mechanical failure of the CuS nanowires. Moreover, the fabrication methods for ultrathin CuS are not easily scalable and reliable for industrial processes. Another strategy involves the incorporation of CuS with wide band gap binary sulfide compounds to create composite films that combine the conductivity of CuS and the higher transparency of the binary sulfide. ZnS is the most well-known compound used in combination with CuS to form (CuS)x:(ZnS)1−x composite films.22,25,26 However, as the proportion of ZnS increases, the conductivity decreases significantly, despite the improvement in transparency. Although other wide band gap chalcogenides may also serve as suitable combination options, there are currently no reported literature findings showing comparable performance.
Therefore, in this study, we employed a high-throughput screening approach utilizing an automated spray pyrolysis system (schematic of the system is shown in Fig. S1, ESI†) to explore alternative combinations of cations with CuS to fabricate conducting wide band gap chalcogenides. This method not only eliminates the downtime associated with traditional spraying techniques but also significantly reduces the fabrication time by two-thirds. The downtime reduction encompasses processes such as the hotplate's ramping up and down, sample loading, and precursor loading, streamlining the experimentation process. In selecting suitable cations, we initially considered the wide band gap binary chalcogenide compounds, such as BaS (3.8 eV),27 MgS (4.5 eV),28 SrS (4.2 eV),29 or CaS (4.4 eV).30 Additionally, we conducted screening based on cation oxidation states, prioritizing cations with an oxidation state of +2, akin to ZnS. Further refinement involved exploring cations recognized as dopants known to enhance the conductivity of certain p-type Transparent Conductive Materials (TCMs). For example, Mg is a well-known dopants in Cu2O,31 LaCuOSe,32 Cr2O3,33 and CuCrO2;34 Sr acts as dopant in La-based TCMs such as LaCuOS,35 LaCrO3;36 while Ca is also a dopant for SrCu2O2.37 Incorporating such cations into the CuS lattice was anticipated to modify the electronic structure, thereby enhancing transparency without compromising conductivity significantly. Ultimately, we selected seven cations—Ba, Cd, Mg, Mn, Sr, Ca, and Zn for our investigation. Among these, Mg emerged as a promising candidate, prompting further investigation into optimal process fabrication parameters using the high-throughput technique. Subsequently, we evaluated the potential of CuS–Mg films as a p-type TCM, employing a normal spray pyrolysis setup with precise control. To elucidate the optoelectronic changes, a combination of computational studies through DFT and characterizations such as X-ray Photoelectron Spectroscopy (XPS) and AC Hall measurements were performed. The CuS–Mg film demonstrated performance comparable to highly effective p-type TCMs reported in the literature. This comprehensive approach not only presents a novel methodology but also establishes Mg-doped CuS as a promising transparent conductor.
Band gap (eV) | Average Rsh (Ω □−1) | Average visible transmittance (AVT) (%) | Transmittance at 550 nm (%) | FOMH (T10/Rsh) (× 10−6 Ω−1) | FOMG 1/(Rsh·ln![]() |
|
---|---|---|---|---|---|---|
CuS | 2.54 | 90.18 | 32.08 | 42.13 | 0.13 | 9751.97 |
CuS–Ba | 2.37 | 2295.20 | 34.42 | 37.65 | 0.01 | 408.51 |
CuS–Cd | 2.54 | 9552.5 | 48.26 | 54.08 | 0.07 | 143.69 |
CuS–Mg | 2.87 | 205.45 | 56.38 | 63.03 | 15.79 | 8492.83 |
CuS–Zn | 2.91 | 476.99 | 56.97 | 61.55 | 7.55 | 3726.44 |
CuS–Mn | 2.75 | 187.08 | 53.17 | 63.22 | 9.65 | 8461.22 |
CuS–Sr | 2.52 | 5507.50 | 45.74 | 52.37 | 0.07 | 232.12 |
CuS–Ca | 2.57 | 5.00 × 10−8 | 57.54 | 67.70 | 7.95 × 10−6 | 3.62 × 10−3 |
The reduction of crystalline CuS can also be explained through DFT calculations, which show an increase in the formation energy of Cu1−xMgxS as the Mg content increases (Fig. 2(c)) in the covellite phase. The calculations reveal that at high Mg content, the covellite phase is far from the ground state, indicating its instability. The incorporated Mg did not form crystalline phase which might be due to the instability of MgS upon exposure to moisture as the fabrication method in this study was performed in air.50
To further verify the effect of Mg incorporation onto the phase formation, X-ray photoelectron spectroscopy (XPS) measurements were conducted on CuS and CuS–Mg at Mg/(Mg + Cu) = 0.4 where transformation to amorphous phase starts to occur. The XPS results provide further insights into the chemical states of each element (tabulated at Table S2, ESI†). The wide spectrum (see Fig. S9a, ESI†) of CuS is dominated by Cu and S signals, along with a small amount of C, N and O. These are possibly due to organic contamination. In contrast, CuS–Mg shows significant Mg, O, and Cl signals, suggesting the possible deposition of unreacted metal chloride precursors or formation of oxides. To further analyze this, the XPS spectra of the main elements are fitted and evaluated. Fig. 3 shows that both samples exhibit similar Cu 2p spectra, which can be deconvoluted into two sets of Cu+ peaks corresponding to the covellite (CuS) and chalcocite (Cu2S) phases.51,52 In addition, the absence of satellite features in both Cu 2p spectra suggests the lack of Cu2+ compounds, which has been previously reported for covellite CuS.52 Furthermore, the S 2p spectra in Fig. 3 also verify the presence of S22− (CuS) and S2− (Cu2S) species alongside organic thiol compound from residual thiourea. In fact, the peak fitting of Cu 2p and S 2p spectra reveals that more Cu2S phase could be detected in the Mg-modified CuS, indicating the influence of Mg in destabilizing the formation of CuS. This result suggests that Mg affects the formation of CuS by inducing amorphization and indirectly promoting more Cu2S formation in the system, as also observed in the XRD analysis. The additional peaks at 168.8 eV in the S 2p spectrum of CuS–Mg can be ascribed to the sulfonated (–SO3H) group. Fig. S9 (ESI†) presents the Mg 2p and Cl 2p spectra of CuS–Mg. The main peak Mg 2p around 50.7 eV is likely to be attributed to the MgCl2 or MgO, while the minor peak at 49.8 eV can be assigned to metallic Mg.53,54 As for Cl 2p, the peaks around 199 eV match with metal chloride.55
In addition, the observed XPS also indicate the elemental composition of the film as calculated in Table S3 (ESI†). The chemical composition was quantitatively determined from the background-subtracted peak area for the core levels of interest and the corresponding relative sensitivity factor (RSF). The XPS results reveal a significantly higher amount of Mg, as well as a Cu-poor composition, in Mg-modified CuS. However, the S/Cu ratio of 0.98 was obtained when the contribution of the sulfonate species was disregarded in the quantitative analysis. Beside XPS, SEM-EDS was also conducted (tabulated in Table S3, ESI†) to understand the bulk composition on the film since XPS is more surface-sensitive. The films were deposited on Si wafer for this analysis to eliminate elemental contributions from various cations, including Mg present in soda-lime glass.56 The ratio between S and Cu is close to 1 for both samples. However, in the CuS–Mg samples, there is a notable increase in the Mg/(Mg + Cu) ratio, which changes from 0.4 in the precursor to 0.6 after deposition. This increase in Mg content is in line with XPS result where more Mg are being deposited. The discrepancy in the composition of the film based on SEM-EDS and XPS compared with the precursor composition suggests that there are different deposition rates of Mg and Cu during the spray process, even though the Mg and Cu precursors are mixed and sprayed together. Furthermore, it is probable that the deposited Mg is not in the form of a binary sulfide, as the sulfur content maintains its ratio with Cu. The higher amount of Mg and sulfur from XPS compared with SEM-EDS suggests that there is a segregation of Mg-rich regions and an excess sulfur on the surface of Mg-modified CuS. To investigate this, STEM-EDS was performed on the CuS–Mg sample as shown in Fig. S10 (ESI†). It confirms the formation of Mg-rich regions on the top layer and CuS-rich region on the bottom layer of the film.
Furthermore, we also conducted a systematic computation of band gaps across a spectrum of compositions within the crystalline region (0 < x < 0.4) of Cu1−xMgxS to elucidate the progressive enhancement in transmittance upon introducing Mg (see Fig. 4). Due to the significant underestimation of semiconductor band gaps by DFT using local density or generalized gradient approximation (LDA/GGA), we have considered band gaps based on the Heyd–Scuseria–Ernzerhof functional (HSE-band gaps) for evaluating the band gaps of Cu1−xMgxS. We observed that the distribution of Cu and Mg within the covellite structures plays a pivotal role in shaping the band gap behavior. When small quantities of Mg are introduced, a random distribution can lead to a slight reduction in the band gap due to the bowing effect. The bowing effect describes the non-linear deviation of the band gap in mixed materials (A1−xBx) compared to the linear interpolation between the band gaps of the pure components (A and B). This deviation arises due to the bowing parameter, which accounts for factors such as lattice mismatch, strain, and electronic interactions. However, it is the specific ordered configurations that result in a band gap increase. For instance, in the case of x = 0.17 (Cu0.83Mg0.17S), an ordered configuration involving the replacement of Cu by Mg at the Cu1 site within the covellite structure may prevail, resulting in a noteworthy band gap increase to 1.58 eV compared to pure CuS (1.0 eV). Additionally, for x = 0.33 (Cu0.67Mg0.33S), wherein two Mg atoms substitute for Cu atoms, the configuration featuring one Mg atom at the Cu1 site and another at the Cu2 site may also prevail, leading to a substantial band gap increase to 1.82 eV. These distinct compositions (x = 0.17 and x = 0.33 ratios) as well as pure CuS are visually presented in Fig. 4(a) for comparative analysis. The unoccupied states of the top valence band and the neighbouring valence bands, highlighted in the orange dashed frame region as shown in Fig. 4(b), facilitate inter-band transitions. As a result, our calculated band gap of CuS exhibits some discrepancy with experimental values (∼2.5 eV). This observation aligns with findings from other DFT calculations on CuS.57 In addition, although STEM-EDS and XPS results indicate that excess Mg is present, the DFT calculations successfully explained the increase trend in transparency and band gap with higher Mg content. This suggests that ordered substitution is the predominant structure in CuS–Mg films with excess of Mg exist as Mg-rich layer on top of film. In terms of performance as TCMs, CuS–Mg with a precursor ratio of 0.4 exhibits the best performance among these films and is superior to pure CuS in terms of FOMH, which is comparable to those of high-performing p-type TCMs.58 As for the FOMG, this ratio gives the closest FOMG value to CuS. This shows that the addition of Mg at this ratio can improve the transparency of CuS without significantly degrading the conductivity.
Thickness (nm) | Average Rsh (Ω □−1) | Conductivity (S cm−1) | AVT (%) | Absorption coefficient (cm−1) | FOMH (T10/Rsh) | FOMG 1/(Rsh·ln![]() |
Hall measurement | ||
---|---|---|---|---|---|---|---|---|---|
(w/o glass) | (w/o glass) | (× 10−6 Ω−1) (w/o glass) | (× 10−6 Ω−1) (w/o glass) | Hole density (cm−3) | Hole mobility (cm2 V−1 s−1) | ||||
CuS | 260 | 153.36 | 250.79 | 52.42 (57.59) | 17![]() ![]() |
10.22 (26.17) | 10![]() ![]() |
6.72 × 1020 | 3.89 |
CuS–Mg | 164 | 266.29 | 228.98 | 68.43 (75.65) | 16![]() ![]() |
84.55 (230.534) | 9899.08 (13![]() |
5.19 × 1021 | 0.49 |
Following that, the carrier concentration and mobility of the films were also investigated by AC Hall-effect measurements. Both materials exhibited p-type characteristics with a large hole concentration in the range of 1019–1021 cm−3 which is comparable with n-type TCO like ITO. A slight decrease in the mobility value can be observed in the CuS–Mg film, but it is still comparable to those from the high performing p-type TCMs, such as CuAlO2,59 Li:Cr2MnO4,60 and Mg:Cr2O3,42 NiO,61 LaCrO3,36 CuS–ZnS.22,25,62 The p-type conductivity of CuS–Mg originates from CuS, which has unoccupied valence bands above the Fermi level.63 When Mg is incorporated into CuS, substituting for Cu, it fills some of these unoccupied valence bands. However, it still retains a partially occupied VBM (Fig. S13, ESI†), which exhibits natural p-type conductivity. As more Mg is incorporated into the lattice, it will fill more unoccupied valence bands, resulting in a reduction in hole concentration. Moreover, increased Mg incorporation in the lattice leads to a larger hole effective mass (Fig. S13e, ESI†), thus lowering hole mobility. The introduction of Mg in Cu1−xMgxS is shown to weaken the p–d coupling, leading to flatter valence bands. For example, Cu0.67Mg0.33S has a flatter valence band compared to CuS, as depicted in Fig. 4(b) and (c). However, at smaller amount of Mg (x < 0.67), the increase in hole effective mass is not significant, which aligns with slight mobility reduction and the comparable conductivity of our best-performing CuS and CuS–Mg samples. As the Mg content increases, the increase in the hole effective mass becomes more pronounced, which is detrimental to conductivity. When x reaches 1 (i.e., MgS), as shown in Fig. 4(d), only the nonbonding state S 3p contributes to the valence band maximum (VBM), resulting in a significantly flat valence band. Consequently, as more Mg is incorporated into the lattice, the conductivity of deteriorates. The segregation of Mg might play a role to this as less Mg being incorporated into the lattice in the covellite phase and the reduction of Mg incorporation in the lattice will lead to more unoccupied valence bands and can generate more holes.
The energy band alignment for the champion films is also investigated through ultraviolet photoelectron spectroscopy (UPS) measurements (Fig. 5(a)). The UPS spectra reveal that both the secondary energy cut-off (ESECO) and the position of the highest occupied molecular orbital (EHOMO) are shifted to higher binding energies. Taking the band gaps into account (see Fig. S14, ESI†), the band gaps increase with Mg incorporation which is ascribed to a downward shift for valence and conduction bands. The VBM and CBM shifts are in agreement with the theoretical calculation of band structure evolution from pure CuS to Cu0.67Mg0.33S to MgS in Fig. 4. In pure CuS, the VBM is contributed by the antibonding state of Cu 3d and S 3p, exhibiting a strong p–d coupling at the Γ point, while the CBM is contributed by the antibonding state of Cu 4s and S 3p. This strong p–d coupling is primarily responsible for its narrow band gap and small hole effective mass in CuS. Upon incorporating Mg, a significant downward shift of the VBM and a slight downward shift of the CBM are observed, leading to an increased band gap in Cu0.67Mg0.33S. The substantial downward shift of the VBM can be attributed to the weakening of S p and Cu d coupling.64 When it comes to MgS, the absence of p–d coupling results in a VBM with a low energy level, primarily composed of the nonbonding S 3p orbital. Consequently, the band gap and hole effective mass reach their maximum values. We depicted the schematic diagram of the orbital interaction evolution from CuS and MgS to Cu1−xMgxS in Fig. S15 (ESI†). In the Cu1−xMgxS system, VBM exhibits a substantial significant downward shift compared to CuS, primarily due to the presence of non-bonding S 3p states, similar to those found in MgS. This alignment is in line with the experimental measurement of band alignment for CuS and CuS–Mg, as shown in Fig. 5(b). The Fermi level of CuS is closer to the valence band, which is the characteristic of a p-type material. On the other hand, CuS–Mg is observed to have a smaller work function and a shallower Fermi level compared to CuS, which can be related with lower p-type conductivity. This finding suggests that even though the carrier concentration increases with the addition of Mg, the increase comes from the metallic Mg in the film which inadvertently reduces the proportion of holes concentrations which prompts the Fermi level to shift upward closer to the midgap.
Furthermore, we also compared the performance of our champion CuS–Mg with the p-type TCMs in the literature. Fig. 5(c) shows the relationship between inverse of absorption coefficient and conductivity of various TCMs. The grey lines depicted the FOMG. In addition, Fig. S16 (ESI†) shows the comparison of CuS–Mg with other TCMs based on its transmittance at 550 nm and average sheet resistance with the grey lines depicting the FOMH. The full list is shown in Table S5 (ESI†). Our CuS–Mg shows a comparable high FOMG with the best reported TCMs and therefore highlights the great potential of this materials as a high performing p-type TCM.
Furthermore, the semi-transparent junctions were also prepared by using FTO (fluorine-doped tin oxide) instead of Ag electrodes, resulting in the configuration of FTO/CuS–Mg/CdS/FTO. The transmittance spectra and schematic of band alignment of these films are shown in Fig. S18 (ESI†). As expected, the p–n junction for CuS–Mg exhibits higher transparency compared to CuS, with an average visible transmittance of approximately 66.59% compared to 52.42% for CuS/CdS. The introduction of FTO leads to a decrease in the overall transmittance by approximately 13% for both CuS–Mg/CdS and CuS/CdS stack, resulting in transmittance values of 53.07% and 38.49%, respectively. The band alignment shows that both CuS and CuS–Mg are suitable to form a type II (staggered type) p–n junction.68 From the band diagram, the turn-on voltage can be estimated as the difference between the Fermi levels.69,70 Assuming the Fermi level of CdS is close to its conduction band, the turn-on voltage can be approximated to be 0.4–0.5 eV, which aligns with our experimental observations. In this semi-transparent junction, similar rectifying diode behavior is also observed for CuS–Mg/CdS, albeit with a lower rectification ratio of approximately 8 at 5 V bias, and thus a poorer diode performance, as shown in Fig. 6(b). The devices demonstrate higher current compared to the Ag-electrode devices, which might be attributed to the reduction in the contact resistance between the film and FTO electrode as the FTO cover the film whereas Ag electrode is just a point contact. Additionally, the deposition techniques for the films are different, which affects the junction properties. Another notable observation is that the CuS–Mg device exhibits photodiode behavior, while the CuS/CdS device maintains similar ohmic behavior. When a photodiode with p–n junction is illuminated, the I–V characteristic should be shifted according to the photocurrent and reverse current. By measuring the reverse current response under illumination, it can be deduced that it originates from p–n junction rather than the individual layer of CdS or CuS–Mg. With a −5 V applied bias, the dark reverse leakage current of the diode is only around 3.7 × 10−3 A. However, the reverse leakage current rapidly increases to 4.5 × 10−2 A upon illumination. The reasonably high photo-to-dark-current contrast ratio suggests that the CuS–Mg/CdS junction in this study has the potential as photodetector for UV pressure sensor in soft robotics or smart window applications.
Additionally, we have fabricated photovoltaic devices using our p-type TCMs as one of the electrodes. The full configuration of the device is glass/p-type TCMs/PEDOT:PSS/perovskite/PCBM-BCP/Ag, with the perovskite composition being triple cation perovskite Cs0.05(MA0.17FA0.83)0.95Pb(I0.83Br0.17)3. Both CuS and CuS–Mg show photovoltaic response with average power conversion efficiencies of 1.8% and 3.7%, respectively. The detailed performance parameters, including open-circuit voltage (Voc), short-circuit current (Jsc), and fill factor, are shown in Fig. S19 (ESI†). The CuS–Mg device shows improved performance compared to the CuS device. Optimization in morphology, device structure and uniformity could be done to further improve the performance. Nevertheless, this proof-of-concept device demonstrates that our p-type material can function as electrode in optoelectronic devices.
Furthermore, our investigation extends beyond CuS films alone. The photodiode utilizing n-type CdS in combination with CuS–Mg demonstrated superior performance than pure CuS. This observation suggests that, with further optimization, the CuS–Mg technology has the potential to serve as a novel p-type TCM with broad applications in transparent electronics and optoelectronic devices such as photodetectors for smart displays, window-integrated electronic circuits and sensors in soft robotics. In conclusion, our study not only presents a valuable contribution to the field but also opens new avenues for advancements in transparent electronics through the seamless integration of high-throughput screening and theoretical validation methodologies.
Footnote |
† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d4mh01501k |
This journal is © The Royal Society of Chemistry 2025 |