Carla Gobbo*ab,
Yuancai Gong
*cd,
Alex Jimenez-Arguijo
cd,
Giorgio Tseberlidis
a,
Vanira Trifiletti
a,
Claudia Malerbae,
Matteo Valentinie,
Pau Estarlich
c,
Elaine Armelin
df,
Sonia Lanzalacodf,
Riccardo Po
b,
Simona Binettia and
Edgardo Saucedo
cd
aDepartment of Materials Science and Solar Energy Research Center (MIB-SOLAR), University of Milano-Bicocca, Via Roberto Cozzi 55, 20125 Milan, Italy. E-mail: c.gobbo3@campus.unimib.it
bEni S.p.A - New Energies, Renewable Energies and Materials Science Research Center, 28100 Novara, Italy
cPhotovoltaic Group, Electronic Engineering Department, Polytechnic University of Catalonia (UPC), Av. Eduard Maristany, 16, 08019 Barcelona, Spain. E-mail: yuancai.gong@upc.edu
dUniversitat Politècnica de Catalunya (UPC), Barcelona Centre for Multiscale Science & Engineering, Av Eduard Maristany 10-14, Barcelona, 08019, Spain
eENEA (Agenzia Nazionale per le Nuove Tecnologie, l'Energia e lo Sviluppo Economico Sostenibile) C.R. CASACCIA, Via Anguillarese 301, 00123 Rome, Italy
fIMEM-BRT's Group, Departament d'Enginyeria Química, EEBE, Universitat Politècnica de Catalunya, C/Eduard Maristany, 10-14, Ed. I, 2nd Floor, 08019 Barcelona, Spain
First published on 1st July 2025
Kesterite solar cells are pivotal in advancing flexible photovoltaic devices integrated into buildings and products. High-purity Mo foil is one of the most promising flexible substrates, thanks to its outstanding properties. However, the kesterite/Mo foil interface is extremely reactive and chemically unstable during the high-temperature selenization process required to obtain the kesterite crystalline phase, forming a thick MoSe2 layer. The role of MoSe2 in kesterite solar cells is still under discussion, as it can affect the charge extraction at the back contact and the kesterite grain growth. This work reports on the functionalisation of Mo foil to fabricate flexible kesterite solar cells based on Li-doped and Ag-alloyed Cu2ZnSn(S,Se)4 (Li-ACZTSSe) films grown using the molecular ink method. MoS2, Al2O3, MoO2, and MoO3 were inserted between the precursor layer and the substrate to investigate their impact on the MoSe2 thickness, the morphology and composition of the absorber, interface chemistry, carrier collection at the back contact, and the related photovoltaic parameters. It is demonstrated that MoO3 as an interlayer significant enhances device performance by improving the absorber quality and back contact, achieving an efficiency of 11.2% with a 15 μm thick MoSe2 layer. To the best of our knowledge, this is the first report demonstrating that an over-thick MoSe2 layer is not significantly detrimental to the performance of flexible kesterite-based devices when the Li-ACZTSSe crystallinity and grain growth are improved and its decomposition at the back is prevented.
Several flexible substrates, including metals (e.g., Mo, stainless steel, Ni, Cu, Ti, and Al), plastics (e.g., polyimide), and ceramics (e.g., flexible glass and ZrO2), have been studied and reported for CZTSSe-based devices.3,5 Compared to the others, high-purity Mo foil excels due to its high-temperature resistance, good mechanical strength, high conductivity, and lack of metallic impurities.9 Furthermore, CZTSSe thin films grown on Mo foil exhibit enhanced crystallinity.3,10 Moreover, the Mo foil can act both as a substrate and a back contact.5 Regardless of the substrate type, achieving high efficiencies in CZTSSe solar cells is a significant challenge due to the formation of defects and secondary phases in the absorber,11 and severe recombination at the kesterite/buffer interface,12–14 leading to a large open-circuit voltage (VOC) deficit and efficiency loss. Recently, record efficiencies of 15.8% on conventional rigid soda-lime glass (SLG)15 and 12.8% on flexible Mo foil16 have been achieved using a molecular ink-based absorber synthesis approach. Although the efficiency gap between flexible and rigid kesterite-based solar cells is closing rapidly, as discussed in this work, additional requirements are needed when a flexible substrate is used.1
One of the main challenges with flexible Mo foil is the lack of alkali metal elements in the substrate (present instead in SLG), typically needed to promote kesterite crystallisation, adjust carrier concentration, passivate grain boundaries, and enhance adhesion to the foil.1,17,18 Alkali element doping6,17 (e.g., Na, Li, and K), combined with substitutional alloying (e.g., Ag),19 plays a key role in facilitating grain growth, enhancing absorber crystallinity, and regulating detrimental defects to achieve high-efficiency flexible kesterite-based devices.20 Little research has focused on the combined effect of Li doping and Ag substitution, with most studies centred on SLG-based devices.20
Another crucial factor influencing flexible solar cell performance is the absorber/Mo foil interface, which impacts electronic properties, kesterite morphology, and secondary phase formation. The role of the interplay between formation, thickness, and doping of MoSe2 in flexible Mo foil-based CZTSSe solar cells remains unclear. During the high-temperature thermal process under a large excess of selenium, required to obtain the crystalline CZTSSe phase, forming a thick MoSe2 layer is inevitable.21 As shown in this work, when the Mo foil is used both as a substrate and as a back contact, the formation of the MoSe2 interface layer is even more favoured due to the substrate's crystal-preferred orientation and high reactivity.22 The MoSe2 growth can follow two different reaction pathways, as reported below:
Mo + Se2(g) → MoSe2 | (1) |
2Cu2ZnSnSe4 + Mo → 2Cu2Se + 2ZnSe + 2SnSe + MoSe2 | (2) |
The free energy change determined at 550 °C for reactions (1) and (2) is −186 kJ mol−1 and −100 kJ mol−1, respectively, indicating that the CZTSSe/Mo interface is reactive and unstable under such conditions.21,23 Typically, a thin MoSe2 layer forms a good ohmic contact, while an over-thick one increases the carrier recombination at the rear interface and reduces the hole transport. However, this might be related to its conductivity, as MoSe2 shows both p-type and n-type electrical properties depending on different possible dopants incorporated into the film (e.g., O, Zn, P, Nb, Va, and Ta give p-type doping, while Re and Cs induce n-type doping).24,25 In the first case, the back contact properties of the CZTSSe device are improved, and the electronic structure of MoSe2 is optimised. At the same time, hole transfer is prevented in the latter, and the carrier recombination is enhanced. A variety of back contact diffusion barriers, including metal oxides, sulfides and nitrides, have been investigated to prevent Se diffusion to the back contact in CIGS and CZTS solar cells.26,27 To improve the back contact, many blocking and sacrificial layers, typically with a high work function, such as Al2O3, TiN, TiB2, MoOx, Ag, Bi, MoS2, ZnO, WO3, a-SiC, and CuO, have been deposited between CZTSSe/CIGS and Mo.23,24,28–32 However, although MoSe2 formation and back contact engineering are more critical in Mo foil-based kesterite solar cells, most previous studies have focused on regulating the kesterite/MoSe2/Mo interface on rigid SLG substrates.33–35
In general, previous studies have devoted extensive efforts to improving flexible kesterite solar cells through (1) alkali doping to enhance absorber crystallization and defect passivation on alkali-free Mo foil;36 (2) surface treatments of Mo foil—such as electrochemical polishing, plasma,37 or ozone exposure,38—to reduce roughness and improve substrate wettability; and (3) interface engineering of the kesterite/CdS heterojunction to suppress interfacial recombination.16,39 However, regulation of the back interface in flexible kesterite solar cells remains underexplored. While several reports have demonstrated that inserting barrier or sacrificial layers at the absorber/Mo interface improves performance on rigid SLG substrates, these strategies are not directly transferrable to flexible devices due to the distinct crystallographic texture, reactivity, and thermal expansion behavior of Mo foil compared to Mo coated on glass.
To address these critical issues, this work presents the first systematic study that combines alkali (Li) doping and back contact engineering—through the introduction of four different interlayers (MoS2, MoO2, MoO3, and Al2O3)—to enhance the performance of flexible CZTSSe solar cells. The study systematically investigates these interfacial layers, examining their influence on MoSe2 thickness and growth mechanisms (reaction (1) or (2)) on flexible Mo foil, their effect on carrier transportation at the back contact, their impact on Li-ACZTSSe crystallization and film quality, and their ability to suppress kesterite decomposition at the back interface (reaction (2)). As a result, a flexible kesterite solar cell with 11.2% efficiency was achieved by fabricating a Li-doped kesterite absorber on Mo foil coated with a MoO3 layer. Compared to other interlayers, MoO3 significantly improves film uniformity by enhancing the solution's wettability on the foil, suppresses reaction (2), and promotes uniform and compact grain growth, leading to a higher VOC and fill factor (FF). More importantly, to the best of our knowledge, this is the first report demonstrating that a 15 μm MoSe2 layer is not detrimental to device performance when absorber and back contact quality are sufficiently improved through effective back contact engineering.
In contrast, the sputtered Mo layer, typically deposited as the back contact when SLG is used as substrate, has a (110) preferred orientation at 2θ = 40.6 deg, implying reduced Se reactivity with the Mo layer and limited MoSe2 formation. Therefore, the functionalisation of the Mo foil with MoS2, Al2O3, MoO3, and MoO2 as interfacial layers between the flexible substrate and the absorber was tested to prevent the potentially detrimental reactions (1) and (2), as schematized in Fig. 1d. Phase identification of the introduced interlayers was performed by XRD and Raman spectroscopy, as shown in Fig. S3a–c (ESI†). The Raman spectra of the Li-ACZTSSe synthesised on all the different functionalised flexible substrates are presented in Fig. 2a. Regardless of the interlayer employed, typical kesterite Raman peaks are detected,45 confirming good Li-ACZTSSe quality without evident secondary phases. XRD measurements were then performed as complementary analyses to Raman spectroscopy. The XRD patterns of the Li-ACZTSSe samples on the flexible substrates are presented in Fig. 2b. Good Li-ACZTSSe crystalline quality is confirmed by the presence of sharp peaks related to the kesterite phase, regardless of the interlayer used. Furthermore, it is possible to evaluate through XRD how the different interlayers affect the thickness of MoSe2 on the substrate by assessing the intensity of the (100), (110), and (200) peaks of MoSe2 at 2θ = 31.8 deg, 2θ = 56.0 deg, and 2θ = 66.1 deg, respectively. When MoS2 is intentionally grown on the flexible foil, due to its preferred orientation perpendicular to the plane of the substrate, Se vapour readily diffuses through the S–Mo–S planes to the Mo foil, thus forming a thick MoSe2 layer.46 Reducing the Se diffusion wholly or partially to the flexible substrate can be achieved by introducing Al2O3 with different thicknesses. The thickest Al2O3 interlayer (20 nm) effectively prevents the formation of MoSe2, and the one with intermediate thickness (5 nm) shows weak MoSe2 signals. In comparison the thinnest layer (2 nm) allows MoSe2 growth, but with lower-intensity peaks compared to the other interlayers. In the last case, mechanical defects like cracks in the interlayer might limit diffusion barrier properties due to the deposition of ultrathin Al2O3 on a moderately rough Mo foil surface.24 Interestingly, although it is reported in the literature that MoOx can suppress Se diffusion toward Mo,47,48 both MoO2 and MoO3 do not exhibit blocking layer behaviour and have no effect on reducing the MoSe2 thickness.
The MoSe2 thickness was also determined using GDOES analysis (see Fig. S4a–e, ESI†). Consistent with the XRD pattern, the samples with an Al2O3 interlayer show a thinner MoSe2 layer, followed by the reference sample produced directly on Mo foil. The sample with the MoS2 interlayer exhibits an almost double MoSe2 thickness compared to the reference. On the other hand, the devices with MoO3 and MoO2 interlayers display very similar MoSe2 thicknesses, close to that obtained on bare Mo foil. GDOES measurements also assessed elemental distributions across the device stack (Fig. 3a–e). The Cu, Ag, Zn, Sn, and Se profiles within the absorber appear consistent across all samples, suggesting that variations in the interlayer composition have limited influence on bulk elemental uniformity. In the interphase region between Li-ACZTSSe and MoSe2, signals corresponding to kesterite-related cations are detected in all cases. Notably, samples incorporating bare Mo foil or MoO2 exhibit more pronounced diffusion of the kesterite elements into the back contact (Fig. 3a and e), which is consistent with interfacial reactions that may lead to kesterite decomposition (reaction (2)) and the formation of secondary phases such as ZnSe, SnSe, Mo(S,Se)2, etc. Besides, Cu and Se diffusion toward the back contact can impact the absorber/back contact interface, forming Sn- and Zn-related defects and VOC losses. In contrast, interlayers such as MoS2, MoO3, and Al2O3 have been reported to act as effective diffusion barriers, helping to limit these interfacial reactions and improve absorber stability.32,49,50
![]() | ||
Fig. 3 GDOES profiles of devices with (a) Mo foil, (b) Mo foil/MoS2, (c) Mo foil/2 nm Al2O3, (d) Mo foil/MoO3, and (e) Mo foil/MoO2 as substrates. |
These trends are reflected in the GDOES profiles shown in Fig. 3a and d, suggesting sharper interfacial gradients and reduced diffusion of kesterite metal cations at the back interface in samples containing diffusion-limiting interlayers. The interlayers not only play a key role in the MoSe2 thickness and the Li-ACZTSSe/Mo interfacial region but can also affect the absorber nucleation and growth. As shown in Fig. 4a–e, the grain size of Li-ACZTSSe depends on the interlayer used. Compared to the reference on bare Mo foil, almost no effect is observed when MoS2 and MoO2 are used, suggesting that they induce similar kesterite nucleation and grain growth processes. In contrast, Al2O3 leads to reduced grain size, while MoO3 enhances it. To understand this, wetting contact angle (WCA) measurements were performed, showing that grain growth might be related to the wettability,51 and therefore the surface energy, of the functionalised Mo foil substrates. As shown in the ESI (Fig. S5a–e†), MoS2 and MoO2 exhibit wettability comparable to the bare Mo foil, leading to similar absorber morphology and MoSe2 growth. The lower wettability of the Al2O3 layer might hinder the absorber grain growth during selenization, resulting in non-uniform crystallization. In contrast, the improved Li-ACZTSSe grain size detected when MoO3 is inserted can be related to its enhanced wettability, which might aid grain growth and promote kesterite adhesion to the foil. Moreover, its lower WCA seems to induce a more continuous and uniform bottom precursor layer, with reduced lattice strain, which can facilitate the growth of larger coherent grains in the final bottom absorber layer (Fig. 1a).
![]() | ||
Fig. 4 Top-view SEM images of the absorbers grown on (a) Mo foil, (b) Mo foil/MoS2, (c) Mo foil/2 nm Al2O3, (d) Mo foil/MoO3, and (e) Mo foil/MoO2. |
Another beneficial effect on the Li-ACZTSSe crystalline quality introduced by MoO3 could be the role of oxygen in defect passivation, as during the high-temperature selenization process, it can be supplied from the interlayer and fill chalcogen vacancies.52 Finally, the impact of the different interlayers on majority carrier collection at the back contact and on the final PV performance was studied. Devices with 5 and 20 nm Al2O3 resulted in non-working solar cells due to the introduction of a thicker resistive interlayer. The statistical photovoltaic parameters of devices fabricated without an interlayer (reference) and those incorporating various interlayers are presented in Fig. 5a–d. Fig. 5e and f displays representative J–V curves of the devices, and their PV parameters are summarised in Table 1. When MoS2 is intentionally deposited on the Mo foil, in agreement with the formation of an over-thick MoSe2 layer confirmed by XRD and GDOES results, there is a drop in all the PV parameters (η = 7.8%) due to the severe deterioration of the back contact. The devices with a 2 nm Al2O3 barrier exhibit a significant improvement in short-circuit current density (JSC), attributed mainly to reduced elements diffusion from the absorber to the Mo foil and, consequently, enhanced kesterite quality near the back interface.
Interlayer | JSC (mA cm−2) | Jintegrated (mA cm−2) | VOC (mV) | FF (%) | η (%) | RS (Ω cm2) | RSh (Ω cm2) | J0 (A cm−2) | n |
---|---|---|---|---|---|---|---|---|---|
Reference | 34.6 | 33.9 | 510 | 58.6 | 10.4 | 1.6 | 538 | 1.77 × 10−6 | 2.01 |
MoS2 | 38.3 | 36.0 | 457 | 44.3 | 7.8 | 2.98 | 436 | 1.64 × 10−5 | 2.31 |
Al2O3 | 37.8 | 36.5 | 493 | 47.7 | 8.9 | 3.28 | 245 | 5.70 × 10−6 | 2.13 |
MoO3 | 35.5 | 33.3 | 528 | 59.9 | 11.2 | 1.57 | 632 | 6.20 × 10−7 | 1.86 |
MoO2 | 32.3 | 33.0 | 466 | 54.5 | 8.2 | 2.61 | 265 | 1.04 × 10−6 | 1.77 |
However, due to the insulating properties of Al2O3, resistive losses increase, resulting in a reduced fill factor (FF) and limiting the device efficiency to 8.9%. Indeed, according to previous studies,50,53 this dense and dielectric Al2O3 interlayer effectively blocks Se diffusion and prevents the contact between the absorber and Mo foil. This significantly suppresses MoSe2 formation (reaction (1)) and inhibits kesterite decomposition (reaction (2)) during high-temperature selenization, ensuring the formation of a thin and controlled MoSe2 layer. However, it hinders the direct connection between the absorber and MoSe2, which is essential for the ohmic hole contact. To improve the ohmicity at the back contact, nano-openings/patterns could be created in the Al2O3 passivation layer.50 Moreover, the lower wettability of the substrate, which limits absorber grain growth, contributes to a reduced VOC.
In contrast, MoO3 insertion has a beneficial effect on both the FF and the VOC. However, the improvement in PV performance (η = 11.2%) cannot be attributed to the ability of this interlayer to inhibit MoSe2 formation but rather to its role in better promoting a controlled Li-ACZTSSe film composition and grain growth, which are strongly related to the MoSe2 formation dynamics. This suggests MoSe2 growth mainly through reaction (1), thereby preventing absorber decomposition. Despite its extension, MoSe2 acts as a good hole transport layer (HTL), suggesting that, when MoO3 is used as a sacrificial layer, O-doping from the initial oxide layer might enhance the electrical properties and the p-type behaviour of the resulting MoSe2.54 Indeed, it has been demonstrated in the literature that the passivation of S and Se vacancies by oxygen can enhance the p-type doping of Mo(S,Se)2.52,55 Furthermore, the MoO3 layer can modify the band alignment at the back contact, leading to improved majority carrier extraction.56 Additionally, the VOC enhancement can be ascribed to the higher wettability of the substrate and, therefore, to the improved Li-ACZTSSe grain growth.
In contrast, when MoO2 is employed, severe decomposition of the absorber at the back contact and significant diffusion of kesterite metal cations into the MoSe2 layer result in lower PV parameters (η = 8.2%), suggesting that the Mo–O phase and fine-tuning the oxygen content in the initial oxide interlayer is crucial for achieving good p-type MoSe2 and promoting kesterite crystallisation and grain growth. A single-diode model analysis was also performed (Table 1). Devices with the MoO3 interlayer exhibited a low diode ideality factor (n = 1.86) and the lowest saturation current density (J0 = 6.2 × 10−7), indicating reduced recombination and superior absorber quality. In contrast, MoS2 and MoO2 interlayers resulted in higher n (>2) and J0, suggesting increased recombination respectively due to significant substrate or absorber decomposition at the back contact and non-optimal grain growth during the selenization process. As discussed earlier, the dielectric Al2O3 interlayer yielded intermediate diode parameters and device performance, effectively minimising interface decomposition, elemental diffusion from the absorber to the Mo back contact, and MoSe2 thickness, but introducing significantly higher series resistance (RS). The lowest RS was determined for the MoO3-based devices, thus contributing to improved hole transport and enhanced FF (∼60%). In contrast, MoS2 and MoO2 interlayers considerably raised RS, while the Al2O3 barrier further increased it due to its dielectric nature. Furthermore, the MoO3 sacrificial layer provided the highest shunt resistance (RSh), confirming minimised leakage currents due to improved absorber uniformity and quality. Overall, the MoO3 interlayer optimised back contact properties, reduced recombination, and promoted beneficial grain growth.
These results highlight the critical role of controlled back contact engineering in enhancing flexible kesterite solar cell performance. The changes in the current collection are assessed by EQE and are reported in Fig. 5g. The EQE spectra can be divided into two distinct regions: the buffer and window layer absorption effect zone (300–520 nm) and the exclusive Li-ACZTSSe absorption region (520–1200 nm). All devices exhibit similar spectral responses within the 300–520 nm range, indicating consistent properties of the buffer and window layers across the devices. Similar spectra are obtained when MoO2 and MoO3 are employed, suggesting a comparable band structure. Devices with Al2O3 and MoS2 interlayers show higher carrier collection in the 550–1200 nm range. This enhancement likely arises from a reduction in defects within the absorber and improved interface quality at the Li-ACZTSSe/buffer junction, consistent with the decreased elemental diffusion observed at the back contact. Consequently, a longer electron diffusion length and more efficient minority carrier collection are achieved. Moreover, Al2O3, already well known in the literature as a rear passivation intermediate layer,53 can successfully inhibit the detrimental interfacial reaction between kesterite and flexible Mo foil, thereby preventing the related electron losses. Interestingly, the lower EQE values for both MoO2 and MoO3 layers indicate higher collection losses at the back interface. Therefore, the improved efficiency with MoO3 may result from enhanced crystallinity and grain quality, leading to a higher VOC and FF. This highlights the importance of controlling grain growth and back-contact properties via an interlayer when fabricating kesterite devices on flexible substrates. Further deposition of a thin, nanopatterned Al2O3 layer on Mo foil may offer a promising strategy to suppress excessive MoSe2 formation while maintaining adequate charge transport, thereby improving device performance. Fig. 5h presents the evaluated electronic bandgaps, which, with slight variations, are close to the expected value (1.1 eV) in all cases. In conclusion, the MoO3 interlayer led to a maximum device performance of 11.2%, among the best devices reported in the literature (Fig. 6a and Table 2).
![]() | ||
Fig. 6 (a) Record efficiencies of kesterite solar cells on flexible Mo foil and (b) VOC deficit of the corresponding devices as a function of Eg.16–18,39 The results of this work are highlighted with a red star. |
Substrate | Process | Strategies | Absorber | η (%) | Ref. |
---|---|---|---|---|---|
Mo foil | Molecular ink | MoO3 interlayer | (Ag,Cu)2ZnSn(S,Se)4 | 11.2 | This work |
Mo foil | Molecular ink | Rb-doped ACZTSSe/CdS | (Ag,Cu)2ZnSn(S,Se)4 | 12.84 | 16 |
Stainless steel | Sputtering | SiO2 barrier | Cu2ZnSn(S,Se)4 | 10.3 | 57 |
Ti foil | Molecular ink | Cd alloying | Cu2(Zn,Cd)SnS4 | 6.51 | 58 |
Polyimide | Sputtering | Na–Ge co-doping | Cu2ZnSnSe4 | 4.9 | 59 |
Rigid SLG | Molecular ink | n/a | (Ag,Cu)2ZnSn(S,Se)4 | 15.8 | 15 |
Furthermore, the voltage deficit (VdefOC) of the champion device, evaluated using VdefOC = VSQOC − VmeasuredOC, where VSQOC = 0.932 Eg/q − 0.1667 V,20 is one of the lowest in the literature, as shown in Fig. 6b. Table 2 summarizes representative high-efficiency kesterite solar cells fabricated on various substrates—including rigid SLG, Mo foil, Ti foil, stainless steel, and polyimide—and compares them with the results of this study. As shown, a significant performance gap remains between rigid (15.8%) and the flexible devices (11.2%). This gap is primarily attributed to MoSe2 overgrowth, stress-induced crystallization issues, and stoichiometric imbalances on flexible substrates. To address these challenges, further improvements may involve nano-patterned Al2O3 interlayers, stress-relief buffer layers, alkali co-doping, and passivation of the kesterite/CdS interface. Additionally, as shown in Fig. 7, bending tests were performed on devices with MoO3 as an interlayer. Under bending conditions with a radius of 1.0 cm for 2.5 × 2.5 cm2 devices, the flexible solar cells retained over 90% of their initial efficiency after 500 cycles. Both VOC and FF remained highly stable, with less than 1% loss, confirming the excellent mechanical resilience of the kesterite absorber and the engineered back interface. The primary source of efficiency loss was a decrease in JSC (over 8%), which is attributed to the fixed bending curvature during measurement after cycling. This likely reduced the effective light absorption due to changes in the incident angle and light path, rather than degradation of the device itself. The devices maintained stable and durable performance under the bending condition. The high-efficiency flexible devices were stored under ambient conditions without encapsulation. After 90 days, the devices exhibited less than 2% PCE degradation as shown in Fig. S6,† demonstrating good long-term stability.
The results demonstrate that Li doping significantly improves device efficiency from 8.5% to 10.4%. And each interlayer has a specific impact on the MoSe2 thickness and growth, Li-ACZTSSe crystallisation, grain size and quality, charge extraction at the back contact, and the PV parameters of the device. MoS2 insertion leads to the formation of an over-thick MoSe2 byproduct and a reduction in all the PV parameters. In contrast, Al2O3 is an effective blocking layer, capable of wholly or partially reducing the Se diffusion to the back contact and tuning the MoSe2 thickness. Moreover, Al2O3 inhibits the detrimental interfacial reaction between the absorber and the substrate. However, Al2O3 introduces resistive losses and limits the absorber grain growth due to its lower wettability, leading to non-optimal crystallisation and a reduced FF and VOC. When MoO2 is used, severe kesterite degradation at the rear interface results in lower PV parameters. In contrast, MoO3 is the most promising interlayer and an effective strategy, achieving an efficiency of 11.2%. The improvement in the final device performance can be ascribed to three main factors. (i) MoO3 reduces the rear interface decomposition and mitigates reaction (2) more successfully than other interlayers. (ii) The resulting thick MoSe2 layer is tolerable and not detrimental to PV performance, thanks to MoO3's influence on its growth mechanism, primarily through the direct reaction between Mo and Se (see reaction (1)), and to O-doping. (iii) MoO3 enhances substrate wettability, thus promoting the formation of larger kesterite grains and improving the FF and VOC more effectively than the other strategies explored. These results highlight the crucial role of back contact engineering in enhancing flexible kesterite solar cell performance. To the best of our knowledge, this is the first report demonstrating that an over-thick (15 μm) MoSe2 layer does not negatively affect the performance of flexible kesterite-based solar cells if the absorber grain size, composition, and quality are optimised. Furthermore, compared to the deposition methods of the other interlayers investigated, pre-annealing the flexible Mo foil in air is easier and more straightforward, and does not require sophisticated or expensive equipment, making this process appealing and suitable for industrial scale-up implementation. In conclusion, the findings reported in this work offer promising prospects for further advancements in flexible kesterite-based devices, paving the way for future optimizations and potential large-scale integrated PV applications.
Footnote |
† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d5ta03303a |
This journal is © The Royal Society of Chemistry 2025 |