Skip to main content
NCBI home page
Royal Society of Chemistry logoLink to Royal Society of Chemistry
. 2026 Sep 22;19(20):7125–7137. doi: 10.1039/d6ee02958b

Holistic design strategy of ferrocene-based interlayers for ultralow-noise perovskite photodetectors and high-efficiency solar cells

Eunyoung Hong a,g,†, Francesco Vanin a,b,†, William D J Tremlett a,†, Patipan Sukpoonprom c, Shuai Li b, Qi Liu d, Somlak Ittisanronnachai e, Longren Li a,g, Zhuoran Qiao a,g, Francesco Furlan f,g, Davide Nodari a,g, Xiao Cheng Zeng d, Pichaya Pattanasattayavong c, Artem A Bakulin a,g, Nicholas J Long a,✉, Zonglong Zhu b,✉, Nicola Gasparini a,g,✉
PMCID: PMC13618407  PMID: 42808021

Abstract

Organometallic ferrocene (Fc)-based interlayers have emerged as effective interfacial modulators for next-generation perovskite optoelectronic devices. The thermally stable Fc core and reversible Fc/Fc+ redox couple, combined with carbonyl linkers, enable strong coordination to halide-related defects and tuneable frontier orbital energies, reducing non-radiative recombination and dark current in perovskite solar cells (PSCs) and perovskite photodetectors (PPDs), respectively. However, previous studies have focused on specific applications separately, resulting in a limited understanding of molecular design principles bridging device operating conditions. Here, we holistically investigate phenyl- and furyl-terminated ester- and ketone-linked Fc interlayers in PSCs and PPDs. By incorporating Fc-derivative interlayers, we demonstrate that the identity of the terminal group critically governs interfacial energetics and thereby directs device performance. Phenyl-based Fc interlayers generate a large potential step at the perovskite surface, as evidenced by a work-function (WF) shift. This creates a localized electrostatic barrier to reverse-biased hole injection, suppressing dark current density (JD) and enhancing the signal-to-noise ratio. In contrast, furyl-based analogs cause a negligible WF shift, leaving the interfacial injection barrier unperturbed. Instead, their illumination-induced band bending promotes balanced charge extraction while passivating surface traps. Electrochemical potential differences associated with ketone- and ester-linkers help explain the respective sub-trends, whereby the former enhance PPD performance, whilst the latter are preferred in photovoltaics. Consequently, PSCs modified with ferrocenyl-bis-furyl-2-carboxylate (F2) achieved state-of-the-art power conversion efficiencies (PCEs), reaching 26.51%, while PPDs modified with ferrocenyl-bis-phenyl-2-ketone (Ph1) attained outstanding specific detectivities, reaching 1.1 × 1013 Jones at −0.5 V (690 nm). These design guidelines highlight the broad applicability of organometallics in perovskite-based optoelectronic devices.

Herein, we study ferrocene derivates as interlayers in perovskite photodiodes. We achieved photovoltaics with 26.51% efficiency and photodetectors with a specific detectivity of 1.1 × 1013 Jones.graphic file with name d6ee02958b-ga.webp

Broader context

Perovskite photodiodes provide a versatile platform in which a single device architecture can operate in photovoltaic (PV) or photodetector (PD) modes via bias modulation. The advancement of next-generation perovskite optoelectronics hinges on the use of functionalized interlayers to enhance device performance. Ferrocene (Fc) is a thermally and chemically robust organometallic scaffold with a rigid spacer and tunable substituents, widely employed as an interfacial modifier. Carbonyl-based Fc derivatives chelate undercoordinated Pb2+ and tune frontier orbitals, promoting quasi-ohmic contacts, suppressing recombination and dark current, and enhancing PV/PD performance. However, previous studies have concentrated on specific device applications, leaving comprehensive design principles that encompass both PVs and PDs underexplored. Herein, we systematically compare phenyl- and furyl-based Fc interlayers featuring ketone and ester linkers to reveal distinct structure–function relationships. Specifically, phenyl-based Fc interlayers induce stronger band bending and higher injection barriers, favoring low-noise perovskite PDs (PPDs), whereas furyl-based analogs provide more moderate band bending and superior defect coordination, advantageous for highly efficient perovskite solar cells (PSCs). By applying these group-specific design guidelines, champion PSCs deliver a power conversion efficiency of 26.51% and champion PPDs achieve a specific detectivity of 1.1 × 1013 Jones, thereby demonstrating broad applicability across perovskite optoelectronic architectures.

Introduction

Ferrocene (Fc) is an organometallic building block that has emerged as a versatile molecular modifier for perovskite optoelectronic devices, particularly in perovskite solar cells (PSCs) and perovskite photodetectors (PPDs).1–7 Fc features a sandwich structure in which a Fe2+ ion is coordinated by two cyclopentadienyl (Cp) rings,8,9 imparting remarkable thermal stability and chemical functionality.10,11 The one-electron reversible Fc/Fc+ redox couple can engage with halide-related defects and ion-migration pathways, where several studies have shown that Fc preferentially interacts with iodine vacancies or iodine species at the perovskite surface and subsequently with Schottky-type PbI2 defects, thereby facilitating defect healing and markedly enhancing operational stability.12,13 Furthermore, both Cp rings can be independently functionalized without disrupting the Fc core, enabling tuning of the Fc/Fc+ redox potential over a wide electrochemical window. A growing body of work has reported molecular design strategies based on the functionalization of either one or both Cp rings with electron-donating or -withdrawing substituents to control both spatial configuration and electronic properties for various optoelectronic applications. For example, bisferrocenyl-bis-thiophene-2-carboxylate (Fc2Tc2) has demonstrated effective redox-active defect passivation of undercoordinated Pb2+ ions, resulting in reduced defect density and achieving a power conversion efficiency (PCE) of 26.08% in p–i–n PSCs.14 More recently, attention has shifted towards tunable ferrocenium dopants for hole-transport materials (HTMs) in n–i–p perovskite devices, with 1,1′-dibromoferrocenium (FcBr2+)-based dopants achieving PCEs up to 26.13% with excellent operational stability by regulating the energetic offset and charge-transfer driving forces.15,16 Building on this foundation, we previously demonstrated that carbonyl groups in the form of ketone or ester linkers act as robust chelators for undercoordinated surface Pb2+, while the subsequent shifts in the Fc redox potential help to adjust frontier orbital energies and band alignment, rendering them an appealing interfacial modulator in both PSCs and PPDs. We showed that Fc derivatives bearing heterocyclic substituents can further modulate surface work function and suppress non-radiative defects, enabling p–i–n PSCs with a PCE of 25.16%.17 In contrast, for PD applications, alongside optimization of performance under illumination, stringent suppression of dark current is critical in high-efficiency PPDs as it directly impacts the achievable signal-to-noise ratio.18–21 Furthermore, ensuring robust reverse-bias stability is vital for commercial viability because ion migration under reverse bias and illumination induces localized defect clusters and promotes chemical reactions or ion accumulation at perovskite/transport-layer interfaces.22 Accordingly, recent work has increasingly focused on bulk23–25 and interfacial engineering26–28 of the perovskite layer, often through functional interlayers inserted between the perovskite and the contacts. In line with these advances, we developed Fc derivatives with phenyl-terminated substituents as multifunctional interlayers to suppress dark current while preserving selective charge extraction through energetic modulation under reverse bias, yielding broadband PPDs with a specific detectivity (D*) of 8.1 × 1012 Jones at −0.5 V.17 Moreover, the reversible Fc/Fc+ redox couple enhanced reverse-bias stability by suppressing the formation of I2 and consequent electrode corrosion, affording excellent PPD performance.

Despite these promising demonstrations, prior studies have focused on specific device applications, leaving unified design principles that span across PVs and PDs unexplored. Here, we elucidate how different aromatic substituents in ketone- and ester-linked Fc interlayers govern the performance of PSCs and PPDs. We explore phenyl- and furyl-containing Fc derivatives bearing ketone or ester linkers as interfacial modulators, namely ferrocenyl-bis-phenyl-2-ketone (Ph1), ferrocenyl-bis-phenyl-2-carboxylate (Ph2), ferrocenyl-bis-furyl-2-ketone (F1) and ferrocenyl-bis-furyl-2-carboxylate (F2). Our findings demonstrate that phenyl-based Fc interlayers, which induce pronounced band bending on the perovskite surface by elevating its Fermi level and thereby increasing the larger effective activation energy, are highly effective for PPDs that require enhanced barrier blocking and low dark current. In contrast, the presence of moderate band bending and Lewis-basic oxygen atoms in furyl-based analogs facilitates more efficient coordination with surface defects and ensures balanced charge transfer. This enhances quasi-ohmic contact formation and reduces interfacial charge barriers, which are critical for achieving efficient power conversion in PSCs. Furthermore, the electrochemical potential of the ferrocene center, controlled by ketone and ester linkers, is found to account for the improved performance sub-trends observed in PPD and PSC devices. As a result, our champion PSCs and PPDs achieved state-of-the-art performance, with PCEs of 26.5% and a measured D* of 1.1 × 1013 Jones (690 nm, at −0.5 V), respectively. These results, together with advanced insights into the structure–function relationship between phenyl/furyl substituents and device performance, establish a design rule for the development of Fc-based derivatives in perovskite optoelectronics and related fields.

Results and discussion

Ferrocene interlayers for perovskite solar cells and photodetectors in a single device

In this work, we employed an inverted p–i–n perovskite photodiode architecture, as shown schematically in Fig. 1a. Devices were built on patterned transparent oxide (TCO) substrates, followed by a hole-transport layer (HTL), the perovskite absorber, the ferrocene-based interlayer and an electron-transport layer (ETL). The stack was completed with a hole-blocking layer (HBL) and a top electrode. In this context, perovskite photodiodes function as a cohesive platform to evaluate the operation of a single device in two distinct modes, PV and PD, simply by adjusting the applied bias (Fig. 1b). In the PV mode, photogenerated carriers are mainly separated and collected by the built-in electric field to generate power (blue rightward arrows in Fig. 1c),29,30 whereas PDs operate under reverse bias, where the stronger electric field enhances charge extraction. Applying excess reverse voltage causes a sharp increase in current and device failure.22,31–33 In general, perovskite devices have lower breakdown voltages than crystalline silicon diodes due to the presence of mobile ions.17,34 Although the external electric field predominantly determines the operating regime, the dark current density (JD), defined as the leakage current that flows in the absence of illumination (red leftward arrows in Fig. 1c), also plays a crucial role and must be carefully mitigated. In PV mode, this JD is commonly described as the recombination or dark saturation current, which induces non-radiative losses and therefore reduces both open circuit voltage (VOC) and fill factor (FF). However, under 1-sun conditions and moderate forward bias, its impact is mitigated by the large photocurrent yield and can be further suppressed by introducing an interlayer that passivates surface defects and reduces interfacial recombination.35,36 In contrast, JD is far more critical in PDs as it governs the noise floor. Since photodiodes are typically operated under low-illumination conditions, a large JD can bury subtle photocurrent changes in noise, preventing reliable signal discrimination even when the responsivity is relatively high.37–41 To suppress JD, a charge-injection barrier is typically employed that presents a favorable energetic offset for blocking undesired carrier injection while remaining sufficiently thin to support efficient transport of the majority carriers.17,22,27,42

Fig. 1. Operating principles for PSCs, PPDs and Fc-derived structures. (a) Schematic of device layout for p–i–n photodiodes with a Fc-based interlayer between the perovskite and ETL. (b) Standard operating regimes of PSCs and PPDs shown in light and dark J–V curves. (c) Schematic energy diagram illustrating common photodiode operating mechanisms under PV and PD modes. (d) Ketone- or ester-linked Fc compounds with phenyl (Ph1, Ph2) or furyl (F1, F2) substituents.

Fig. 1

Open in a new tab

Aiming to tailor our synthetic approach to PPD and PSC applications, we targeted phenyl- and furyl-substituted ketone- and ester-linked ferrocene compounds to investigate how aryl terminations and electrochemical potential (highest occupied molecular orbital energy, EHOMO) affect the use of a substituted ferrocene in a specific device application. These substituents were chosen based on prior investigations and are formally compared herein to draw structure–function relationships for future materials development across optoelectronic device classes. Synthetic procedures and structural characterization of these compounds (Fig. 1d) is provided in the methods section.

Device performance: controlling PSCs and PPDs through ring substituents

To investigate how the phenyl or furyl ring substituents of these Fc-based interlayers influence device performance in different perovskite photodiode applications, we evaluated the Fc-modified devices in both PV and PD modes. The influence of the Fc-based interlayers in PV operation was investigated, with the current density–voltage (J–V) characteristics under AM 1.5G illumination shown in Fig. 2a and Fig. S1, and the corresponding device parameters are summarized in Table S1. Stabilized power output (SPO) and photocurrent were obtained by holding the devices at the maximum power point (MPP) voltage determined from the J–V curves for 500 s under continuous 1-sun illumination (Fig. 2b and Fig. S2 and Table S2). The champion control device delivered a PCE of 25.29% with a short-circuit current density (JSC) of 25.86 mA cm−2, a VOC of 1.17 V and FF of 83.34%. Upon incorporation of the furan-decorated Fc interlayers, the PCE increased to 26.51% (JSC = 25.97 mA cm−2, VOC = 1.19 V, FF = 85.90%) for the F2-based device and to 26.11% for the F1-based device (JSC = 26.03 mA cm−2, VOC = 1.19 V, FF = 84.34%). Conversely, devices featuring phenyl-functionalized Fc analogs exhibited diminished device performance, regardless of the nature of their carbonyl-based linkage (Table S1). The enhanced JSC for devices employing furan-based Fc interlayers is validated by the external quantum efficiency (EQE) spectra shown in Fig. S3 and Table S3, with the integrated photocurrent density (Jcal) obtained from the EQE curves aligning well with the JSC values taken from the J–V characteristics (Fig. 2a). The J–V characteristics of the electron-only type devices reveal that the furyl-based Fc interlayers yield a low-voltage ohmic slope comparable to that of the control, indicating that these furyl Fc variants do not impede electron injection (Fig. S4 and Table S4). Moreover, the reduced trap-filled limit voltage (VTFL) observed for F1 and F2 implies a lower density of electronic trap states, consistent with more effective defect passivation and the enhanced PV performance relative to the control.43 In PPD mode, however, the opposite trend was observed. Fig. 2c presents the dark J–V characteristics, where JD was evaluated at −0.5 V to minimize the influence of ion migration while maintaining efficient charge extraction. To further suppress the effect of mobile ions, JD between −0.5 and +0.5 V was measured after stabilizing the current for 1000 s at each bias. The control device exhibited an average JD of 1.4 × 10−9 A cm−2 at −0.5 V, whereas the use of phenyl-based Fc-derivatives led to a pronounced change. Notably, devices treated with Ph2 showed a clear reduction in JD at −0.5 V, to 3.6 × 10−10 A cm−2, and further to 2.9 × 10−10 A cm−2 for the Ph1-modified device. An analogous trend was observed in the noise spectral density (Sn), which is intimately related to JD when shot noise (ishot) dominates but remains distinct from it (Fig. S5 and Table S5).44–46 By definition, Sn quantifies random fluctuations per unit bandwidth and determines the noise floor that ultimately limits the minimum detectable signal. It should be noted that a low JD does not guarantee a low Sn, because additional contributions from flicker noise (1/f noise) and other noise sources can be present.44–46 In this work, Sn was determined by direct noise spectral density measurements under a reverse bias of −0.5 V, where the control device exhibited an Sn value of 1.3 × 10−13 A Hz−1/2. Devices incorporating the furyl-decorated Fc interlayers showed no significant change, whereas Ph2- and Ph1-treated interlayers suppressed Sn to 2.8 × 10−14 A Hz−1/2 and further to 7.9 × 10−15 A Hz−1/2, respectively. All devices exhibit frequency-dependent behavior in the low frequency region (0.01–10 Hz), confirming the presence of 1/f flicker noise. Although the 1/f slope is similar for all samples, Ph1 and Ph2 consistently show a lower Sn across the entire frequency range. For comparison, a commercially available silicon (Si) diode (DET20X2, Thorlabs) demonstrated a Sn of 5.1 × 10−15 A Hz−1/2 at its operational bias, which is closely comparable to that of the Ph1-modified device (Fig. S6 and S7 and Table S6). This reduction in Sn indicates a lower detection limit and, consequently, enhanced sensitivity to weak optical signals for the phenyl-based Fc devices. To verify this, the specific detectivity (D*), a key figure of merit that quantifies a PD's sensitivity to weak optical signals, was calculated as follows:

graphic file with name d6ee02958b-t1.webp 1

where A is the active area of the PD, and SR is the spectral responsivity, defined as the photocurrent response to incident optical power as a function of wavelength and is directly associated with the EQE (Fig. S8). Since the maximum EQE is unity, except in photomultiplication-type photodiodes, the achievable gain in perovskite photodiodes is inherently limited.6 Consistent with this, no notable variations in SR appeared across the conditions in this study. Thus, further improvements in performance should focus on reducing Sn. As shown in Fig. 2d, the Ph2-treated device exhibited a D* of 2.7 × 1012 Jones at −0.5 V, while the Ph1-treated device reached 1.1 × 1013 Jones at 690 nm, over one order of magnitude higher than that of the control device (7.5 × 1011 Jones). Following the Sn trend, the Fc-based devices bearing furan groups showed only a negligible improvement compared to the control device (Table S5). We further evaluated the shot-noise-limited Inline graphic by using the Ishot instead of the Sn, as Inline graphic, where q is the elementary charge and ID is the dark current. As expected, the Inline graphic values exceed those obtained from the direct Sn measurements, with the Ph1-modified device achieving the highest Inline graphic of 4.4 × 1013 Jones (Fig. S9 and Table S7). We further investigated Fc derivatives bearing various substituents to confirm the universal consistency of the insight developed in the current report regarding the importance of substitution for the selected device function (Fig. S10 and Tables S8).

Fig. 2. Device performance of the best-performing control and Fc-modified devices. In the shared p–i–n architecture, device processing was optimized for the intended device application. (a) J–V curves under 1-sun illumination. (b) Stabilized PCE at MPP voltage. (c) Dark J–V curves where the devices’ current response between −0.5 and +0.5 V was measured after stabilization at each bias. (d) Specific detectivity (D*) calculated based on noise spectral density (Sn).

Fig. 2

Open in a new tab

Origins of ultralow dark current in phenyl-based Fc perovskite photodiodes

In general, JD arises from three main channels: bulk thermal generation in the perovskite absorber, interfacial thermal generation, and thermionic-like injection of minority carriers across interfacial barriers.6,19,47 As an identical perovskite layer is employed in all devices in this study, bulk thermal generation is expected to be comparable for devices with and without the Fc-based interlayers. Here, we compared the best-performing Ph1-treated device with the benchmark and F1-modified devices. F1 was chosen as a reference molecule because it shares the same ketone linker as Ph1, resulting in the Fc-based interlayers whose HOMO levels lie deeper than the valence band maximum (VBM) of the perovskite (Fig. S11 and S12 and Table S9). This alignment implies a similar minority charge injection pathway from the ETL into both Fc-based interlayers and suggests that the differences in device behavior primarily originate at the perovskite/Fc interface, governed by the distinct terminal groups. To elucidate the origin of JD, we examined its temperature dependence and interfacial energetics. The temperature-dependent JD was measured in the range of 295–265 K under a constant reverse bias of −0.5 V (Fig. 3a and Fig. S13 and Table S10) and fitted with an Arrhenius model to probe the thermally activated hole injection at each interlayer interface. Arrhenius plots of ln(JD) as a function of 1/T (Fig. 3b and Fig. S14) yield the effective thermal activation energy (Ea)6,47 according to:

graphic file with name d6ee02958b-t6.webp 2

where kB is the Boltzmann constant, T is the temperature, and Ea reflects the effective injection barrier height, i.e. the barrier height that thermionic emission must overcome. As shown in Fig. 3b, Ea values of 1.67 eV and 0.14 eV were obtained for Ph1- and F1-treated devices, respectively, indicating that a significantly higher interfacial barrier to thermally activate hole injection in Ph1. This barrier substantially exceeds the individual energetic offsets in the system, (i) from the ETL to Ph1 (0.25 eV), and (ii) from Ph1 to the perovskite (0.19 eV) (Tables S9 and S11). The origin of the elevated effective hole injection barrier in Ph1 (1.67 eV) compared to that of 0.14 eV for F1 can be attributed to the upward shift of the perovskite surface work function (WF) induced by the deposition of Ph1, characterized by Kelvin probe measurements (Fig. S15). Ph1 shifts the surface WF by +0.43 eV relative to the pristine perovskite surface, creating a Fermi level mismatch across the perovskite/Ph1 interface. At equilibrium, this mismatch drives electron transfer from the perovskite surface into Ph1, generating a positive space-charge region that bands the perovskite band downward at the surface. Under reverse bias, the applied electric field reinforces this built-in band bending, enhancing the suppression of hole injection.48–51 The Ph1-induced band bending serves a dual role: first, the downward band bending raises the effective potential barrier that holes from the ETL/Ag contact must overcome to enter the perovskite bulk, with the additional energetic cost of +0.43 eV relative to devices without a significant WF shift. Second, the downward band bending moves the valence band maximum (VBM) further below the Fermi level at the perovskite surface. According to the Fermi–Dirac distribution (SI Note S1),52 the equilibrium hole density is exponentially suppressed at the perovskite/Ph1 interface as the downward band bending increases, reducing the number of holes available for parasitic hole injection. These two effects account for the significantly higher Ea measured for Ph1 compared with F1, which is consistent with the order-of-magnitude reduction in JD at −0.5 V from 3.59 × 10−9 A cm−2 for F1 to 2.85 × 10−10 A cm−2 for Ph1. In contrast, the lower Ea (0.14 eV) for the F1-based device, which remains below the energetic barrier expected for direct hole injection from the ETL to the perovskite, indicates a negligible interfacial barrier, consistent with the absence of a WF shift, and a mixed contribution from trap-assisted tunnelling and thermionic injection.53,54

Fig. 3. Origins of charge injection and transport in Fc-treated devices in the dark and under illumination. The perovskite composition Cs0.05(FA5/6MA1/6)0.95Pb(I0.85Br0.15)3 was employed as the control device. (a) JD of the control, Ph1- and F1-based devices under a continuous reverse bias of −0.5 V at various temperatures in the range of 295–265 K. (b) Ea of the control, Ph1- and F1-based devices extracted from the Arrhenius plot. (c) Schematics of minority hole injection under a reverse bias of −0.5 V. The Ph1-based device exhibits downward band bending, resulting in a potential barrier sitting at the perovskite/Ph1 interface (3), in addition to the energetic barriers (1) and (2), whereas the F1-based device lacks band bending, presenting only a two-step hole injection barrier. (d) SPV signals of the pristine perovskite, perovskite/Ph1 and perovskite/F1 samples under illumination. (e) Ground-state bleach (GSB) features of the glass/perovskite/Fc-derivative stacks. (f) Normalized ns-TAS dynamics of GSB signals measured using a 450 nm pump. The decay time (i.e. half-lifetime) of the pristine perovskite, Ph1- and F1-based films is 88.2 ns, 46.7 ns, and 67.5 ns, respectively.

Fig. 3

Open in a new tab

However, the same Fc-based interlayers behave differently under illumination. Surface photovoltage (SPV) measurements confirm the presence of downward band bending in all samples, with a larger SPV signal in Ph1 indicating stronger band bending than F1 (Fig. 3d and Fig. S16).55,56 Transient absorption spectroscopy (TAS) was performed on the nanosecond (ns) time scale to monitor the evolution of photogenerated carriers in the perovskite layer with and without Fc-based interlayers (Fig. 3e and f and Fig. S17). The ground-state bleach (GSB) near 750 nm was largest for pristine perovskite, reduced in F1-treated films, and most strongly suppressed in Ph1-treated films, indicating more efficient carrier quenching (Fig. 3e). Normalized ns-TAS dynamics at the GSB peak (Fig. 3f) was faster for F1 (67.5 ns) than pristine (88.2 ns), with Ph1 fastest (46.7 ns). Biexponential fitting of the GSB recovery (Table S12) assigned the faster component to bulk band-edge carrier recombination and the slower component to perovskite/ETL interfacial depopulation.57 Pristine perovskite showed the slowest interface component (630.6 ns), reflecting inefficient interfacial carrier removal without an Fc/ETL contact,58,59 and a 53.2 ns bulk component. With Fc-based interlayer, interface lifetimes shortened to 520.5 ns (Ph1) and 371.6 ns (F1), confirming an additional carrier depopulation pathway. Notably, F1 shows the shortest interface-associated lifetime and the longest bulk lifetime (76.7 ns), reflecting efficient interfacial extraction with reduced bulk recombination consistent with moderate band bending and effective passivation.56,60 In contrast, Ph1's longer interface lifetime (520.5 ns) and shorter bulk component (64.3 ns) suggests near-interface carrier loss, consistent with its stronger band bending, which drives electrons into an already electron-accumulated ETL interface, promoting non-radiative recombination (Fig. S18).

Elucidating structure–function relationships in Fc-based interlayers

Given the substantial variations in device performance across the different Fc-based interlayers, we examined the correlation between their molecular structures and electrochemical properties. Fc derivatives showed a negligible impact on the optical properties of the perovskite layer as seen by the ultraviolet-visible (UV-vis) absorption spectra (Fig. S19).17,43 We conducted cyclic voltammetry (CV) experiments to determine the redox potentials of each Fc moiety and extract their HOMO energies (Fig. 4a). The corresponding energy-level alignments of the Fc derivatives are shown in Fig. 4b. The LUMO energies (ELUMO) were estimated from their optical bandgaps (Fig. S20 and S21). The choice of ketone or ester linker mainly determined the EHOMO (Table S11). The solution-phase redox and NMR behavior further highlight the distinct electronic properties of these Fc derivatives (Fig. S22–S29). Upon I2 addition, Ph2 and F2, unlike Ph1 and F1, undergo a pronounced color change, consistent with Fe3+/Fe2+ interconversion (Fig. S30). In line with this, the Cp and phenyl 1H NMR spectra of the ester-containing Fc variants broaden and disappear, reflecting paramagnetic Fe3+ formation and accelerated nuclear relaxation (Fig. S31–S34).

Fig. 4. Correlation between molecular structure of Fc derivatives and their electrochemical properties. (a) Stacked cyclic voltammograms and (b) calculated EHOMO and ELUMO levels for Ph1, F1, Ph2 and F2. (c) Computed surface coordination geometries of the Fc-based compounds with undercoordinated Pb2+ centers on the (001) PbI2-terminated perovskite surface. XPS spectra of the perovskite/Fc-derivative samples displaying the (d) Pb 4f and (e) I 3d areas. Statistical distribution of (f) JD and (g) PCE for control, Ph1-, Ph2-, F1- and F2-treated devices. Perovskite composition and device architecture were optimized for PD and PV applications, respectively.

Fig. 4

Open in a new tab

To gain molecular-level insights into the Fc-derived compounds, density functional theory (DFT) calculations were performed on the PbI2-terminated (001) surface of FAPbI3 perovskite (SI Note S2). In the optimized structures, Ph1 shows a Pb–O (carbonyl) bond length of 2.52 Å, which is comparable to those obtained for F1 (2.5 Å), Ph2 (2.67 Å), and F2 (2.64 Å), confirming the formation of interactions between the Fc derivatives and the perovskite surface (Fig. 4c). The corresponding binding energies between the perovskite surface and Ph1, Ph2, F1, and F2 were calculated in the most stable conformations to be −2.78, −2.01, −0.92, and −4.44 eV, respectively, reflecting the combined influence of the bridging carbonyl oxygen atoms and the furyl/phenyl terminal groups. We then conducted X-ray photoelectron spectroscopy (XPS) to experimentally probe the interfacial interactions of these Fc derivatives with the perovskite surface. Fig. 4d and Fig. S35 demonstrate the binding energies corresponding to the Pb 4f, I 3d, N 1s and Cs 3d core levels of the pristine and Fc-treated films. As shown in Fig. 4d, the Pb 4f and I 3d core-level spectra of all samples showed characteristic doublet peaks. Notably, only Ph1 exhibited a slight shift of these core levels toward higher binding energy relative to the pristine perovskite film, whereas Ph2, F1 and F2 showed shifts toward lower binding energy (Table S13). However, the direction of these shifts alone cannot be unambiguously interpreted as indicative of stronger or weaker coordination, since the observed core-level binding energies result from a convolution of band bending and local chemical shifts.5,7,12,43,61 Atomic force microscopy (AFM) on glass/ITO/perovskite/Fc samples revealed only negligible differences with and without Fc interlayer and no clear trend in root-mean-square (RMS) roughness (Fig. S36). Additional AFM and Kelvin probe force microscopy measurements (KPFM) were performed to study the morphology and energetics at the perovskite/Fc/C60 interface (Fig. S37, S38 and S39). The morphology of perovskite/Fc/C60 interfaces do not vary significantly as a function of interlayer molecular structure, which is consistent with the ultrathin nature of the Fc interlayers <1 nm.17 KPFM measurements revealed that ketone-linked Ph1 and F1 show negligible contact potential difference (CPD) change relative to perovskite/C60, whereas ester-linked Ph2 and F2 exhibit positive shifts. This likely arises from stronger interactions between the more electron-rich ester-linked Fc units and C60, leading to interfacial WF shifts consistent with reduced hole injection barriers. Lastly, Fig. 4e and Fig. S40 present the statistical distribution of the PD and PV performance for the shared p–i–n architectures optimized for each application, demonstrating that, on average, phenyl-containing Fc devices deliver superior PD performance, whereas their furyl-based counterparts excel in PV operation.

Our comprehensive experimental investigations indicate that phenyl functionalization establishes a strong interfacial potential barrier via a WF shift that suppresses JD injection but affords weaker chemical passivation of perovskite surface traps than furyl functionalization. Although a small fraction of photogenerated carriers may recombine before reaching the ETL, this loss is negligible when compared to the substantial suppression of reverse-biased JD and the resulting enhancement in D*. In contrast, furyl-based devices provide effective defect passivation but lack the dark band-bending-induced barrier required to suppress JD under reverse bias. Consequently, they show superior PV performance (Fig. S41), which we attribute to a moderate degree of band bending under illumination that promotes efficient charge extraction. We also note an important sub-trend from the direct comparison of ketone- and ester-linked derivatives. The more electron-deficient ketone-linked Ph1 outperforms its ester counterpart (Ph2) in PPD applications, while the more electron-rich ester-linked F2 exhibits superior PV performance than F1 (Fig. 4f and Fig. S40). This highlights a broader trend observed for the compounds tested herein, namely, the best performing ferrocenes for PD operation afford the worst PV performance and the best PV metallocenes afford the worst PD performance. Together, these complementary behaviors reveal a fundamental trade-off between hole-injection barrier formation and chemical passivation in Fc-based interlayer design.

Conclusions

In this study, we employed Fc-based derivatives functionalized with either phenyl or furyl groups, linked to the Cp rings via ketone or ester bridges, as interfacial modulators in PSCs and PPDs. Our holistic study elucidates the mechanisms underlying their opposite performance trends. Furyl-based Fc interlayers serve as effective interfacial passivators in the PV operation, whereas phenyl-containing analogs act as charge injection barriers in PDs, achieving a champion PCE of 26.51% (F2) and a D* of 1.1 × 1013 Jones (Ph1), respectively. By combining CV, DFT, and XPS, we confirm that all Fc-based derivatives coordinate with the perovskite surface through the carbonyl oxygens. Temperature-dependent and Kelvin-probe measurements reveal that downward band bending at the perovskite/Fc-derivative interface manifests differently in the dark and under illumination, which significantly alters the interfacial energetics and, consequently, affects hole transport at the interface. Our results indicate that phenyl-based Fc interlayers create a strong interfacial potential barrier in the dark (due to ΔWF) and a depleted hole population, which increases Ea for hole injection, significantly reducing JD and enhancing D*. In contrast, furyl-based counterparts exhibit moderate illumination-induced band bending, facilitating balanced interfacial charge transport alongside perovskite surface defect passivation, which improves the PCEs. Collectively, these findings provide a general design strategy: phenyl substitution is preferrable for highly sensitive PPDs, while furyl substitution is better suited for high-efficiency PSCs. The phenyl vs. furyl substitution on the Fc scaffold underlines the trade-off between hole-injection barrier formation and chemical passivation.

Methods

Materials

Perovskite solutions were prepared by using PbI2 (99.9%, TCI), PbBr2 (>98.0%, TCI), MAI (>99.9%, Greatcell Solar), FAI (>99.9%, Greatcell Solar), MACl (99.0%, Sigma Aldrich), and CsI (>99.99%, Sigma Aldrich). For transport layer materials, [2-(3,6-dimethoxy-9H-carbazol-9-yl)ethyl]phosphonic acid (MeO-2PACz) (98.0%, TCI), buckminsterfullerene (C60) (>99.5%, Oscilla, and bathocuproine (BCP) (>99.5%, Lumtec), tin oxide (SnO2) (15% in H2O colloidal dispersion, Alfa Aesar) were employed. All anhydrous solvents including ethanol (EtOH), chlorobenzene (CB), dimethylformamide (DMF), and dimethyl sulfoxide (DMSO) were purchased from Sigma Aldrich. Ferrocenyl-bis-phenyl-2-ketone (Ph1),62 ferrocenyl-bis-phenyl-2-carboxylate (Ph2),17 ferrocenyl-bis-furyl-2-ketone (F1)43 and ferrocenyl-bis-furyl-2-carboxylate (F2)43 were prepared using procedures already reported.

Compound synthesis

Materials

Aluminium chloride (99%), benzoyl chloride (99%), ferrocene (98%, purified by sublimation prior to use) and 2-furoyl chloride (95%) were purchased from Sigma-Aldrich. Benzoic acid (99%) was purchased from Alfa Aesar. Copper(i) oxide (97%) was purchased from Acros Organics. Furoic acid (>98%) was purchased from Tokyo Chemical Industry Ltd. Anhydrous acetonitrile (CH3CN) and dichloromethane (CH2Cl2) were dried through a solvent purification system under a nitrogen atmosphere (Inert PureSolv) and transferred into Schlenk flasks that were dried under vacuum and purged with N2 prior to use. CH2Cl2 and n-hexane were purchased from VWR Chemicals, and deuterated chloroform (CDCl3) was purchased from Sigma-Aldrich. Diiodoferrocene (FcI2),9 ferrocenyl-bis-phneyl-2-ketone (Ph1),63 ferrocenyl-bis-phenyl-2-carboxylate (Ph2),17 ferrocenyl-bis-furyl-2-ketone (F1)43 and ferrocenyl-bis-furyl-2-carboxylate (F2)43 were prepared using procedures already reported.

Characterization

Nuclear magnetic resonance (NMR) spectra were recorded at ambient temperature on a Bruker Avance 400 MHz spectrometer using 5 mm 507-pp NMR tubes and processed using MestReNova v14.3.2. NMR spectra were calibrated to the residual solvent peaks of CDCl3 at 7.26 and 77.2 ppm for 1H and 13C{1H}, respectively. Peak assignment was aided through the use of COSY, edited HSQC and HMBC experiments. High-resolution mass spectrometry data were recorded by direct injection on a Waters LCT time-of-flight (ToF) mass spectrometer using electrospray ionization (ESI) conducted by Malgorzata Puchnarewicz of the Mass Spectrometry Service at Imperial College London. Elemental analyses were analyzed using a ThermoFlash 2000 Analyzer by the Elemental Analysis Service at London Metropolitan University. Ultraviolet-visible (UV-vis) absorbance spectra of samples in solution were obtained with an Agilent Technologies Cary 60 UV-vis spectrophotometer. All samples (5–1000 µM) were analyzed at room temperature using a quartz cell with a path length of 1 cm in CH2Cl2.

PD device fabrication

Inverted p–i–n structured perovskite devices were constructed on glass substrates etched with indium tin oxide (ITO). The substrates were sequentially washed with acetone and 2-propanol (IPA) using ultrasonication for 10 min in each solvent. They were then dried with nitrogen and subjected to ultraviolet (UV) ozone treatment for 15 min before being transferred to an N2-filled glovebox. The hole-transporting layer (HTL), perovskite, and Fc-derivative layer were progressively spin-coated. The control device involved spinning MeO-2PACz dissolved in EtOH (1 mM) on ITO at 3000 rpm (with an acceleration of 1500 rpm) for 30 s, followed by drying at 100 °C for 10 min. A 1.2 M Cs0.05(FA5/6MA1/6)0.95Pb(I0.85Br0.15)3 solution, with a 1% excess of PbI2 and 15 mol% MACI was synthesized by mixing stoichiometric quantities of PbI2, PbBr2, and CsI in a mixture of DMF : DMSO (4 : 1 v/v), followed by an overnight stirring at 60 °C. Then MAI and FAI were added to the solution, which then was stirred at 60 °C for 1 h. The final perovskite solution was deposited in a two-step process, first at 1000 rpm (with an acceleration of 1000 rpm) for 10 s, followed by at 5000 rpm for 27 s. 150 µL of CB was added at 6 s before the end. Subsequently, the samples were annealed at 100 °C for 30 min. 0.3 mg mL−1 of the Fc-based compounds dissolved in CB were dynamically spin-coated onto the perovskite layer at 4000 rpm (with an acceleration of 4000 rpm) for 40 s and then was annealed at 100 °C for 4 min. The rest of the stacks were thermally evaporated at a base pressure of around 10−6 mbar, with 40 nm of C60, 6 nm of BCP, and 100 nm of Ag deposited sequentially. The active area of each pixel, defined by a shadow mask, was 0.045 cm2. The electron-only devices were constructed on glass/ITO/SnO2/1.2 M Cs0.05(FA5/6MA1/6)0.95Pb(I0.85Br0.15)3/C60/BCP/Ag. Pre-patterned ITO-coated glass were sequentially washed with acetone and 2-propanol (IPA) using ultrasonication for 10 min in each solvent, dried under nitrogen and treated with ultraviolet (UV) ozone. A thin layer of SnO2 was then deposited by spin-coating a 2.5 wt% colloidal SnO2/H2O suspension onto ITO substrates at 4000 rpm for 30 seconds followed by drying at 140 °C for 15 minutes. The 1.2 M Cs0.05(FA5/6MA1/6)0.95Pb(I0.85Br0.15)3 perovskite solution was spin-coated inside a N2-filled glovebox, first at 1000 rpm (with an acceleration of 1000 rpm) for 10 s, and then at 5000 rpm for 27 s. 150 µL of CB was added at 6 s before the end. Subsequently, the samples were annealed at 100 °C for 30 min. 0.3 mg mL−1 of the Fc-based compounds dissolved in CB were dynamically spin-coated onto the perovskite at 4000 rpm (with an acceleration of 4000 rpm) for 40 s and then was annealed at 100 °C for 4 min. Finally, the remaining layers were thermally evaporated at a base pressure of around 10−6 mbar, with 40 nm of C60, 6 nm of BCP, and 100 nm of Ag deposited sequentially.

PV device fabrication

The nickel oxide (NiOx) nanoparticles were dispersed in deionized water at a concentration of 10 mg mL−1 and ultrasonicated at room temperature for 10 min to form a uniform ink, which was filtered through a 0.45 µm hydrophilic PVDF filter before use. For the self-assembled monolayer (SAM) solution, Meo-2PACz and Me-4PACz were individually dissolved in methanol (0.3 mg mL−1) and subsequently mixed in a 1 : 1 volume ratio. 1.5 M perovskite precursor solution with the composition of FA0.95Cs0.05PbI3 was prepared by dissolving corresponding 19.5 mg CsI, 245.1 mg FAI, and 712.3 mg PbI2 (included 3 mol% more) into 1 mL mixed solvent of DMF and DMSO with a volume ratio of 4 : 1. For the surface treatment, individual solutions of each Fc-based material (0.5 mg mL−1 in CB) were prepared.

Glass/FTO substrates (9 Ω sq−1) were sequentially cleaned by sonication with detergent, deionized water, acetone and isopropyl alcohol for 20 min, respectively. Then, the glass/FTO substrates were dried at 60 °C in an oven, and then were treated with oxygen plasma for 10 min. After that, NiOx ink was spin-coated onto plasma-treated glass/FTO substrates at 4000 rpm for 30 s with the post-annealed at 150 °C for 15 min in air and then transferred into a N2-filled glovebox. Then, as-prepared SAM solution was spin-coated in the N2 glovebox at 4000 rpm for 30 s, followed by annealing at 100 °C for 10 min. Next, 40 µL perovskite precursor was spin-coated onto glass/FTO/NiOx/SAM layer at 1000 rpm for 10 s, subsequently at 5000 rpm for 40 s. During the second step, 200 µL of CB was rapidly dripped onto the center of the spinning substrate 10 s prior to the end of the program. The as-prepared perovskite films were annealed on a hotplate at 100 °C for 30 min. To apply the interlayer, the prepared Fc-based solutions were dynamically spin-coated onto the perovskite layer at 5000 rpm for 30 s and then was annealed at 100 °C for 5 min. The rest of the stacks were thermally evaporated at a base pressure of around 10−6 mbar, with 25 nm of C60, 6 nm of BCP, and 100 nm of Ag were deposited sequentially. The photoactive area was determined by the geometric area of the evaporation shadow mask, measured at 0.0750 cm2. To minimize reflection losses, a 125 nm thick magnesium fluoride (MgF2) anti-reflective layer was thermally evaporated (1.0 Å s−1) onto the back (glass side) of the FTO substrate.

Perovskite characterizations

UV-vis spectroscopy

UV-vis analyses were conducted on a Agilent Technologies Cary 60 UV-vis spectrophotometer. All samples were analysed at room temperature using a quartz cell with a path length of 1 cm in CH2Cl2 (for liquid samples) and a quartz substrate (for solid samples).

Kelvin probe measurement and surface photovoltage

The non-contact Kelvin probe measurement was performed by KP Technology KP020 system to determine the work function (WF) of the films under dark environment. The contact potential differences (CPDs) of samples were measured and referenced to an Au of a known WF value (4.74 eV, measured by APS). Surface photovoltage (SPV) measurement was conducted by SPS040 module (add-on to KP020 system). The samples were probed under dark until CPD was stabilized (30–60 min), then illuminated by white light from Quartz Tungsten Halogen (QTH) lamp for 250 s. All measurements were probed under ambient conditions.

X-ray photoelectron spectroscopy

High-resolution X-ray photoelectron spectroscopy (XPS) was performed on perovskite and perovskite/Fc derivatives deposited on ITO substrates. The elemental bonding states of the films were characterized by a Jeol JPS9010MC photoelectron spectrometer. Their characterization involved a monochromated Al Kα X-ray source operating at a 12 kV and a base pressure of 10−7 Pa, with a low-energy electron flood gun. The data were recorded using an energy of 50 eV for survey scans and 10 eV for core-level scans. The analyzed spectra comprised an average of 80 scans. Before sample loading, they were kept in a vacuum antechamber overnight to reduce the water and oxygen molecules on the film surface. All spectra were analyzed by SpecSurf software and referenced to the same adventitious carbon position (C 1s at 285.0 eV).

ns transient absorption spectroscopy

The single-wavelength ns pump–probe measurement employed a 450 nm pulse (47 µJ cm−2) from a ns laser as the ‘pump’ and a femtosecond pulse from commercial PHAROS Modular-design to be the ‘probe’. The lasers supplying pump and probe beams are synchronized, with the time delay being controlled by an electrical delay generator (DG645, Stanford Research Systems). The repetition rate of primary laser was 5000 Hz, while an optical chopper modulated the pump light on/off at 2500 Hz. To maximize the signal-to-noise ratio for a clearer observation of the decay characteristic, 750 nm was chosen to investigate the ground-state bleach (GSB) peaks from films with and without Fc derivatives. The GSB kinetics was monitored by a silicon photodetector (DET10A/M, THORLABs).

Cyclic voltammetry

Cyclic voltammetry analyses were conducted on a Gamry 600TM potentiostat to record the redox potentials associated the ferrocene (Fc) moiety. The samples were analysed in a cell comprising of glassy carbon working, platinum counter and silver pseudo reference electrode. The samples (2 mM) were dissolved in an electrolyte solution of tetrabutylammonium hexafluorophosphate (0.1 M) in anhydrous and degassed CH2Cl2 and analysed at 100 mV s−1. Each sample was referenced to a ferrocene/ferrocenium (Fc/Fc+) internal reference and all scans corrected for solution resistance with values obtained from potentiostatic electrochemical impedance spectroscopy (EIS) measurements.

Atomic force spectroscopy (AFM)

AFM images (Fig. S36 and S37) were obtained using an Agilent 5500 in tapping mode, and the data were analysed by using the Gwyddion software.

AFM and KPFM (Fig. S38 and S39) were conducted through Bruker Dimension ICON, and Ti/Ir coated silicon tips (ASYELELC-01-R2) with a resonance frequency at ∼58–97 KHz were used in scanning Kelvin probe microscopy (SKPM) with the work function internally calibrated to an Au. All AFM-based experiments were performed in air.

Perovskite photodetector (PPD) characterizations

J–V measurements were conducted using Keithley 4200 source-measurement unit (scan rate of 25 mV s−1) An Oriel instruments solar simulator with a xenon lamp was calibrated with a Si reference cell. The measurements were performed under AM 1.5G irradiance of 100 mW cm−2. Spectral responsivity was measured using an integrated system from a Quantum Design PVE300 at a modulation frequency of 90 Hz. The digital oscilloscope (Siglent, SDS6054A) was employed to determine the cut-off frequency. The devices were illuminated with a 530 nm laser that was controlled by a function generator (ThorLabs DC2200) to produce sinusoidal functions with varying frequencies. For determination of the rise and fall time a 1 kHz square wave pulse was applied to the 530 nm LED using the function generator. The devices were coupled with a high-speed current amplifier (Femto, DHPCA-100) to apply −0.5 V. Transient photocurrent data were obtained by using a digital oscilloscope (RIGOL DS4024) with a 515 nm laser as the excitation source and a pulse duration of 190 fs. The laser repetition frequency was set to 5 kHz, with an intensity of 2 µJ cm−2. All measurements were conducted in ambient air.

Noise spectral density measurement

For the control and Fc-treated devices, J–V measurements were carried out in dark by Keithley 4200 source measurement unit in the range of −0.5 to +1.3 V. The devices were connected to a Zurich MFLI 500 through a preamplifier (STANFORD, SR570), which applied a bias of −0.5 V and a gain based on the devices’ JD values in units of A V−1 (Table S14). For the commercially available Si diode (DET20X2, Thorlabs), ID measurements were performed in the range of +8 and +20 V. Based on its value at +10 V, we applied a gain of A V−1 to the device and measured its Sn with a battery connected. The current responses were measured in dark over the time length of 1 min, using a demodulation low pass filter (LPF) order of 3 and an effective −3 dB cut-off bandwidth of 300 Hz. The recorded time-domain data were transformed into frequency components by Fast Fourier Transform (FFT) to obtain noise amplitude spectrum. The measurement was conducted in ambient air at room temperature.

Temperature-dependent dark current measurement

The devices were placed in a Linkam HF600E-PB4 cryostat with an LNP96 liquid nitrogen cooling module, controlling temperature from 295 K to 265 K. JD measurements were performed using a Keithley 4200 Source Measurement unit under continuous −0.5 V reverse bias.

Perovskite photovoltaic (PV) device measurements

The current–voltage (J–V) characteristics of the photovoltaic devices were measured using a Keithley 2400 source meter in a N2-filled glovebox at room temperature. Simulated solar illumination was provided by a xenon lamp solar simulator (Enlitech, SS-F5, Taiwan), calibrated to 100 mW cm−2 (AM1.5G) using a KG2-filtered silicon reference cell. Devices were measured without pre-conditioning, using both a reverse scan (1.20 V to −0.10 V) and a forward scan (−0.10 V to 1.20 V) at a scan rate of 0.02 V s−1 with a 10 ms delay time. The active area was defined and characterized as 0.0450 cm2 by a metal shadow mask. The stabilized power output (SPO) was determined by tracking the steady-state current density at the maximum power point (MPP) voltage extracted from the reverse-scan J–V curve. External quantum efficiency (EQE) spectra were recorded using a QE-R system (Enlitech, Taiwan).

Conflicts of interest

The authors declare no competing interests.

Supplementary Material

EE-019-D6EE02958B-s001

Acknowledgments

N. J. L. acknowledges Imperial College London for funding via the Sir Edward Frankland BP Chair endowment, an Impact Acceleration Award and the Deep-Tech Prime programme and is also grateful for a Royal Society Wolfson Research Merit award. L. L. thank Imperial College London for a President's PhD scholarship. Z. Q. and N. G. would like to thank the EU Horizon project SWIRFlex-DT (101257652). F. F. acknowledges financial support from the Engineering Physical Science Research Council (EPSRC; UKRI3811).

Data availability

The key data backing this work's primary findings are included in the paper and the supplementary information (SI). Supplementary information includes synthetic steps and materials characterisations, DFT calculations, morphology and optoelectronic measurements. See DOI: https://doi.org/10.1039/d6ee02958b.

For further information, please contact the corresponding authors.

References

  1. He R. Pan J. Chen X. Zheng Q. Zou Y. Lv S. Jin Z. Wang Y. Sun W. Wu J. ACS Photonics. 2025;12:4745–4756. doi: 10.1021/acsphotonics.5c01440. [DOI] [Google Scholar]
  2. Webb T. Liu X. Westbrook R. J. E. Kern S. Sajjad M. T. Jenatsch S. Jayawardena K. D. G. I. Perera W. H. K. Marko I. P. Sathasivam S. Li B. Yavari M. Scurr D. J. Alexander M. R. Macdonald T. J. Haque S. A. Sweeney S. J. Zhang W. Adv. Energy Mater. 2022;12(26):2200666. doi: 10.1002/aenm.202200666. [DOI] [Google Scholar]
  3. Bi H. Liu J. Zhang Z. Wang L. Kapil G. Wei Y. Kumar Baranwal A. Razey Sahamir S. Sanehira Y. Wang D. Yang Y. Kitamura T. Beresneviciute R. Grigalevicius S. Shen Q. Hayase S. Adv. Sci. 2023;10(35):2304790. doi: 10.1002/advs.202304790. [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Hu K. Peng J. Wang Q. Deng L. Deng C. Xu M. Zhang Z. Ren H. Yang C. Chen J. Yu H. Sol. Energy Mater. Sol. Cells. 2024;272:112936. doi: 10.1016/j.solmat.2024.112936. [DOI] [Google Scholar]
  5. Li Z. Li B. Wu X. Sheppard S. A. Zhang S. Gao D. Long N. J. Zhu Z. Science. 2022;376(6591):416–420. doi: 10.1126/science.abm8566. [DOI] [PubMed] [Google Scholar]
  6. Nodari D. Qiao Z. Furlan F. Sandberg O. J. Vandewal K. Gasparini N. Nat. Rev. Mater. 2025;2025:1–15. [Google Scholar]
  7. Wang F. Wang T. Ge C. Sun Y. Sun X. Liang X. Duan D. Li Q. Li Y. Zhang F. Yang G. Zhou X. Zhu Q. Lin H. Chen H. Wu T. Hu H. ACS Energy Lett. 2024;9:4283–4292. doi: 10.1021/acsenergylett.4c01433. [DOI] [Google Scholar]
  8. Rausch M. Vogel M. Rosenberg H. J. Chem. Educ. 1957;34:268–272. doi: 10.1021/ed034p268. [DOI] [Google Scholar]
  9. Roemer M. Nijhuis C. A. Dalton Trans. 2014;43:11815–11818. doi: 10.1039/C4DT01787K. [DOI] [PubMed] [Google Scholar]
  10. Sukpoonprom P. Tremlett W. D. J. Qiao Z. Chattakoonpaisarn C. Hong E. Hu B. Forberich K. Han J. Wang J. Ittisanronnachai S. Li L. Vanin F. Pattanasattayavong P. Zhu Z. Bakulin A. Brabec C. J. Baran D. Long N. J. Gasparini N. Joule. 2025;9:102107. doi: 10.1016/j.joule.2025.102107. [DOI] [Google Scholar]
  11. Ye L. Cai Y. Li C. Zhu L. Xu J. Weng K. Zhang K. Huang M. Zeng M. Li T. Zhou E. Tan S. Hao X. Yi Y. Liu F. Wang Z. Zhan X. Sun Y. Energy Environ. Sci. 2020;13:5117–5125. doi: 10.1039/D0EE02426K. [DOI] [Google Scholar]
  12. Chang Q. Wang F. Xu W. Wang A. Liu Y. Wang J. Yun Y. Gao S. Xiao K. Zhang L. Wang L. Wang J. Huang W. Qin T. Angew. Chem., Int. Ed. 2021;60:25567–25574. doi: 10.1002/anie.202112074. [DOI] [PubMed] [Google Scholar]
  13. Lu M. Sun X. Chen C. Nano Energy. 2025;136:110720. doi: 10.1016/j.nanoen.2025.110720. [DOI] [Google Scholar]
  14. Li B. Gao D. Sheppard S. A. Tremlett W. D. J. Liu Q. Li Z. White A. J. P. Brown R. K. Sun X. Gong J. Li S. Zhang S. Wu X. Zhao D. Zhang C. Wang Y. Zeng X. C. Zhu Z. Long N. J. Cite This: J. Am. Chem. Soc. 2024;146:13391–13398. doi: 10.1021/jacs.4c02220. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Vanin F. Webb T. Gao D. Welch K. Tremlett W. D. J. Qian L. Zhang C. Brown R. K. White A. J. P. Gasparini N. Bo L. Zhu Z. Haque S. A. Long N. J. Joule. 2026;0:102310. doi: 10.1016/j.joule.2025.102310. [DOI] [Google Scholar]
  16. Webb T. Vanin F. Gao D. Zhu L. Tremlett W. D. J. Azaden A. Rodgers A. Jacoutot P. White A. J. P. Islam M. S. Long N. J. Zhu Z. Haque S. A. Energy Environ. Sci. 2026;19:605. doi: 10.1039/D5EE05482F. [DOI] [Google Scholar]
  17. Hong E. Tremlett W. D. J. Hart L. Hu B. Qiao Z. Sukpoonprom P. Fearn S. Angela E. Brunetta M. Koutsogeorgis D. C. Kalfagiannis N. Nodari D. McLachlan M. A. Barnes P. R. F. Bakulin A. A. Long N. J. Gasparini N. Adv. Funct. Mater. 2025:2424556. [Google Scholar]
  18. Simone G. Dyson M. J. Weijtens C. H. L. Meskers S. C. J. Coehoorn R. Janssen R. A. J. Gelinck G. H. Adv. Opt. Mater. 2020;8(1):1901568. doi: 10.1002/adom.201901568. [DOI] [Google Scholar]
  19. Sandberg O. J. Kaiser C. Zeiske S. Zarrabi N. Gielen S. Maes W. Vandewal K. Meredith P. Armin A. Nat. Photonics. 2023;17:368–374. doi: 10.1038/s41566-023-01173-5. [DOI] [Google Scholar]
  20. Kumar P. Shukla V. K. Kim M. Singh R. Sens. Actuators, A. 2024;369:115076. doi: 10.1016/j.sna.2024.115076. [DOI] [Google Scholar]
  21. Rogalski A. Hu W. Wang F. Wang Y. Martyniuk P. Materials. 2024;17:4029. doi: 10.3390/ma17164029. [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Li N. Shi Z. Fei C. Jiao H. Li M. Gu H. Harvey S. P. Dong Y. Beard M. C. Huang J. Nat. Energy. 2024;2024:1–11. [Google Scholar]
  23. Hong E. Nodari D. Furlan F. Angela E. Panidi J. McLachlan M. A. Gasparini N. Adv. Opt. Mater. 2024;12(13):2302712. doi: 10.1002/adom.202302712. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Furlan F. Nodari D. Palladino E. Angela E. Mohan L. Briscoe J. Fuchter M. J. Macdonald T. J. Grancini G. McLachlan M. A. Gasparini N. Adv. Opt. Mater. 2022;10(24):2201816. doi: 10.1002/adom.202201816. [DOI] [Google Scholar]
  25. Liang F. X. Wang J. Z. Zhang Z. X. Wang Y. Y. Gao Y. Luo L. B. Adv. Opt. Mater. 2017;5:1700654. doi: 10.1002/adom.201700654. [DOI] [Google Scholar]
  26. Angela E. Nodari D. Furlan F. Panidi J. McLachlan M. A. Gasparini N. ACS Appl. Mater. Interfaces. 2024;16:33838–33845. doi: 10.1021/acsami.4c06447. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Pan Y. Wang X. Xu Y. Li Y. Elemike E. E. Shuja A. Li Q. Zhang X. Chen J. Zhao Z. Lei W. Front. Chem. 2020;8:547012. doi: 10.3389/fchem.2020.00791. [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. An Q. Sun Q. Weu A. Becker-Koch D. Paulus F. Arndt S. Stuck F. Hashmi A. S. K. Tessler N. Vaynzof Y. Adv. Energy Mater. 2019;9:1901257. doi: 10.1002/aenm.201901257. [DOI] [Google Scholar]
  29. Moia D. Gelmetti I. Calado P. Hu Y. Li X. Docampo P. De Mello J. Maier J. Nelson J. Barnes P. R. F. Phys. Rev. Appl. 2022;18:044056. doi: 10.1103/PhysRevApplied.18.044056. [DOI] [Google Scholar]
  30. Gutierrez-Partida E. Rusu M. Zu F. Raoufi M. Diekmann J. Tokmoldin N. Warby J. Menzel D. Lang F. Shah S. Shoaee S. Korte L. Unold T. Koch N. Kirchartz T. Neher D. Stolterfoht M. ACS Appl. Mater. Interfaces. 2025;17:11176–11186. doi: 10.1021/acsami.4c14194. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Wang C. Huang L. Zhou Y. Guo Y. Liang K. Wang T. Liu X. Zhang J. Hu Z. Zhu Y. Adv. Energy Mater. 2022;13(1):4949. [Google Scholar]
  32. Johnson S. Morales D. Fremouw K. Gould I. E. Borsa T. Johnston S. Palmstrom A. DeCrescent R. A. McGehee M. D. Joule. 2025:102102. doi: 10.1016/j.joule.2025.102102. [DOI] [Google Scholar]
  33. Jiang F. Shi Y. Rana T. R. Morales D. Gould I. E. McCarthy D. P. Smith J. A. Christoforo M. G. Yaman M. Y. Mandani F. Terlier T. Contreras H. Barlow S. Mohite A. D. Snaith H. J. Marder S. R. MacKenzie J. D. McGehee M. D. Ginger D. S. Nat. Energy. 2024;2024:1–10. [Google Scholar]
  34. Ahmadi M. Wu T. Hu B. Adv. Mater. 2017;29(41):1605242. doi: 10.1002/adma.201605242. [DOI] [PubMed] [Google Scholar]
  35. These A. Koster L. J. A. Brabec C. J. Le Corre V. M. Adv. Energy Mater. 2024;14:2400055. doi: 10.1002/aenm.202400055. [DOI] [Google Scholar]
  36. Greulich J. Glatthaar M. Rein S. Prog. Photovoltaics Res. Appl. 2010;18:511–515. doi: 10.1002/pip.979. [DOI] [Google Scholar]
  37. Saggar S. Puidokas G. Murawski C. Adv. Opt. Mater. 2026;14(6):e02399. doi: 10.1002/adom.202502399. [DOI] [Google Scholar]
  38. Lin Q. Armin A. Lyons D. M. Burn P. L. Meredith P. Adv. Mater. 2015;27:2060–2064. doi: 10.1002/adma.201405171. [DOI] [PubMed] [Google Scholar]
  39. Xiong S. Li L. Qin F. Mao L. Luo B. Jiang Y. Li Z. Huang J. Zhou Y. ACS Appl. Mater. Interfaces. 2017;9:9176–9183. doi: 10.1021/acsami.6b16788. [DOI] [PubMed] [Google Scholar]
  40. Kublitski J. Hofacker A. Boroujeni B. K. Benduhn J. Nikolis V. C. Kaiser C. Spoltore D. Kleemann H. Fischer A. Ellinger F. Vandewal K. Leo K. Nat. Commun. 2021;12:1–9. doi: 10.1038/s41467-020-20856-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. Pecunia V. Anthopoulos T. D. Armin A. Bouthinon B. Caironi M. Castellanos-Gomez A. Chen Y. Cho K. Clegg C. Fang X. Fendel P. Fowler B. Gelinck G. Gottlob H. Guyot-Sionnest P. Hannebauer R. Hernandez-Sosa G. Hersam M. C. Hirsch L. Ho J. C. Isikgor F. H. Joimel J. Kim H. J. Konstantatos G. Labram J. Lemme M. C. Leo K. Lhuillier E. Lidorikis E. Loi M. A. Malinowski P. E. Merken P. Mueller T. Nasrollahi B. Natali D. Ng T. N. Nguyen T.-Q. Park S. K. Peng L.-M. Samorì P. Sargent E. H. Shen L. Shishido S. Shorubalko I. Sonar P. Stranks S. D. Tedde S. F. Vandewal K. Verhaegen M. Walia S. Yan F. Yokota T. Zhang F. Nat. Photonics. 2025;19:1178–1188. doi: 10.1038/s41566-025-01759-1. [DOI] [Google Scholar]
  42. Li N. Lim J. Azoulay J. D. Ng T. N. J. Mater. Chem. C Mater. 2020;8:15142–15149. doi: 10.1039/D0TC03013A. [DOI] [Google Scholar]
  43. Vanin F. Tremlett W. D. J. Gao D. Liu Q. Li B. Li S. Gong J. Wu X. Li Z. Brown R. K. Qian L. Zhang C. Sun X. Li X. Zeng X. C. Zhu Z. Long N. J. Angew. Chem., Int. Ed. 2025;64(14):e202424041. doi: 10.1002/anie.202424041. [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. Bielecki Z. Achtenberg K. Kopytko M. Mikołajczyk J. Wojtas J. Rogalski A. Pol. Akad. Nauk. 2022;70(2):140534. [Google Scholar]
  45. Wang F. Zhang T. Xie R. Wang Z. Hu W. Nat. Commun. 2023;14:1–9. doi: 10.1038/s41467-023-37635-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
  46. Fang Y. Armin A. Meredith P. Huang J. Nat. Photonics. 2018;13:1–4. doi: 10.1038/s41566-018-0288-z. [DOI] [Google Scholar]
  47. Ollearo R. Wang J. Dyson M. J. Weijtens C. H. L. Fattori M. van Gorkom B. T. van Breemen A. J. J. M. Meskers S. C. J. Janssen R. A. J. Gelinck G. H. Nat. Commun. 2021;12:1–9. doi: 10.1038/s41467-021-27565-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
  48. Cavalcoli D. Cavallini A. Phys. Status Solidi C. 2010;7:1293–1300. doi: 10.1002/pssc.200983124. [DOI] [Google Scholar]
  49. Daboczi M. Hamilton I. Xu S. Luke J. Limbu S. Lee J. McLachlan M. A. Lee K. Durrant J. R. Baikie I. D. Kim J. S. ACS Appl. Mater. Interfaces. 2019;11:46808–46817. doi: 10.1021/acsami.9b16394. [DOI] [PubMed] [Google Scholar]
  50. Hu H. Birkhold S. Sultan M. Fakharuddin A. Koch S. Schmidt-Mende L. ACS Appl. Energy Mater. 2019;2:4045–4052. doi: 10.1021/acsaem.9b00060. [DOI] [Google Scholar]
  51. Imran T. Aziz H. S. Iftikhar T. Ahmad M. Xie H. Su Z. Yan P. Liu Z. Liang G. Chen W. Chen S. Energy Environ. Sci. 2024;17:7234–7246. doi: 10.1039/D4EE02894E. [DOI] [Google Scholar]
  52. Zhou C. X. Sun J. X. Deng Z. J. Zhou S. Semiconductors. 2013;47:1351–1357. doi: 10.1134/S1063782613100291. [DOI] [Google Scholar]
  53. Jang W. Nguyen T. Q. Wang D. H. Adv. Funct. Mater. 2022;33(2):2209615. doi: 10.1002/adfm.202209615. [DOI] [Google Scholar]
  54. Widenhorn R., Blouke M. M., Weber A., Rest A. and Bodegom E., Sensors and Camera Systems for Scientific, Industrial, and Digital Photography Applications III, 2002, pp. 193–201 [Google Scholar]
  55. Cavalcoli D. Cavallini A. Phys. Status Solidi C. 2010;7:1293–1300. doi: 10.1002/pssc.200983124. [DOI] [Google Scholar]
  56. Daboczi M. Kim J. Lee J. Kang H. Hamilton I. Lin C. T. Dimitrov S. D. McLachlan M. A. Lee K. Durrant J. R. Kim J. S. Adv. Funct. Mater. 2020;30(25):2001482. doi: 10.1002/adfm.202001482. [DOI] [Google Scholar]
  57. Li C. Wang A. Deng X. Wang S. Yuan Y. Ding L. Hao F. ACS Photonics. 2020;7:1893–1907. doi: 10.1021/acsphotonics.0c00677. [DOI] [Google Scholar]
  58. Jašinskas V. Franckevičius M. Gelžinis A. Chmeliov J. Gulbinas V. ACS Appl. Electron. Mater. 2023;5:317–326. doi: 10.1021/acsaelm.2c01346. [DOI] [PMC free article] [PubMed] [Google Scholar]
  59. Pan J. Chen Z. Zhang T. Hu B. Ning H. Meng Z. Su Z. Nodari D. Xu W. Min G. Chen M. Liu X. Gasparini N. Haque S. A. Barnes P. R. F. Gao F. Bakulin A. A. Nat. Commun. 2023;14:1–10. doi: 10.1038/s41467-023-43852-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
  60. Henderson C. Luke J. Bicalho I. S. Correa L. Yang E. J. Rimmele M. Demetriou H. Chin Y. C. Lan T. Heutz S. Gasparini N. Heeney M. Bagnis D. Kim J. S. Energy Environ. Sci. 2023;16:5891–5903. doi: 10.1039/D3EE02571C. [DOI] [Google Scholar]
  61. Zu F. Wolff C. M. Ralaiarisoa M. Amsalem P. Neher D. Koch N. ACS Appl. Mater. Interfaces. 2019;11:21578–21583. doi: 10.1021/acsami.9b05293. [DOI] [PubMed] [Google Scholar]
  62. Sarhan A. A. O. Ibrahim M. S. Kamal M. M. Mitobe K. Izumi T. Monatsh. Chem. 2008;140:315–323. doi: 10.1007/s00706-008-0025-2. [DOI] [Google Scholar]
  63. Schwink L. and Knochel P. 10.1002/(SICI)1521-3765(19980515)4:5 [DOI]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

EE-019-D6EE02958B-s001
EE-019-D6EE02958B-s001.pdf (7.1MB, pdf)

Data Availability Statement

The key data backing this work's primary findings are included in the paper and the supplementary information (SI). Supplementary information includes synthetic steps and materials characterisations, DFT calculations, morphology and optoelectronic measurements. See DOI: https://doi.org/10.1039/d6ee02958b.

For further information, please contact the corresponding authors.

Articles from Energy & Environmental Science are provided here courtesy of Royal Society of Chemistry

RESOURCES