Focussed Review of Utilization of Graphene-Based Materials in Electron Transport Layer in Halide Perovskite Solar Cells: Materials-Based Issues

The present work applies a focal point of materials-related issues to review the major case studies of electron transport layers (ETLs) of metal halide perovskite solar cells (PSCs) that contain graphene-based materials (GBMs), including graphene (GR), graphene oxide (GO), reduced graphene oxide (RGO), and graphene quantum dots (GQDs). The coverage includes the principal components of ETLs, which are compact and mesoporous TiO2, SnO2, ZnO and the fullerene derivative PCBM. Basic considerations of solar cell design are provided and the effects of the different ETL materials on the power conversion efficiency (PCE) have been surveyed. The strategy of adding GBMs is based on a range of phenomenological outcomes, including enhanced electron transport, enhanced current density-voltage (J-V) characteristics and parameters, potential for band gap (Eg) tuning, and enhanced device stability (chemical and environmental). These characteristics are made complicated by the variable effects of GBM size, amount, morphology, and distribution on the nanostructure, the resultant performance, and the associated effects on the potential for charge recombination. A further complication is the uncertain nature of the interfaces between the ETL and perovskite as well as between phases within the ETL.


Introduction
Rapidly increasing global demands for clean energy supplies have created an imperative for increased utilization of renewable energy sources. Solar power is one such resource that can play such a major role. However, as a potential next-generation power supply, photovoltaic technologies must meet the requirements of cost-efficiency as well as environmental stability. Single-crystal and polycrystalline silicon solar cells are the most mature commercial photovoltaic devices, having achieved high efficiencies of >22% and considerable stability [1]. There also are low-cost commercial thin-film solar cells that have been developed as substitute materials, including amorphous silicon (a-Si:H), with a peak efficiency of 14%, and cadmium telluride (CdTe), with a peak efficiency of 22% [2]. Other emerging alternative thin-film solar cells utilize organic materials as light harvesters, and these can contribute to the reduction of noble element consumption and minimize production costs.
Metal halide perovskite solar cells (PSCs) were reported in 2009 and, since then, have gained increasing attention owing to their excellent power conversion efficiencies (PCEs) and low-cost fabrication [2,3]. Perovskite materials have the formula ABX 3 , where A and B are different cations and

GBMs
In order to enhance the optoelectronic properties of PSCs, graphene-based derivatives have been investigated and utilized extensively [44,45]. Graphene (GR) is a 2D material with a single layer of sp 2 -bonded carbon atoms that are tightly connected in a hexagonal honeycomb lattice. It exhibits remarkable electrical, optical, and mechanical properties as well as good stability, making it a good candidate to enhance the performance of PSCs [46]. Despite these excellent properties, pure GR exhibits zero band gap (E g ), which precludes its use as a semiconductor. However, chemical synthesis rarely produces pure GR, so what often is reported as GR is slightly impure. Compared to graphene, graphene derivatives are E g -tunable and relatively low-cost materials that involve less complex fabrication methods. The most commonly used graphene derivatives are, as shown in Figure 1a, graphene oxide (GO), reduced graphene oxide (RGO), and graphene quantum dots (GQDs) that contain hydroxyl and carboxyl functional groups [47][48][49], which enable them to display hydrophilicity instead of hydrophobicity that is seen in graphene. These functional groups are beneficial to dispersibility in suspensions during processing using water and organic media of high dielectric constants [50]. The flexibility of GBMs has allowed them to be used in different parts of PSCs, including electrodes [51,52], perovskite layer [53,54], charge-transfer layer [55][56][57][58][59], and as interfacial layer [60][61][62].
There have been extensive works on GBMs for the ETL [17][18][19][20][23][24][25][26][27]39,43,55,56,60,61,[63][64][65][66][67][68][69][70][71][72][73][74][75][76][77][78][79][80][81] that is sandwiched between the transparent conducting oxide (TCO) anodic substrate (layer 1) and the perovskite (layer 3), as shown in Figure 1b. ETL materials, such as TiO 2 , SnO 2 , ZnO, PCBM, SrTiO 3 + Al 2 O 3 , and Fe 2 O 3 , have been modified using GR, GO, RGO, and GQDs. The improvement in the PCE due to the modification by GBMs in ETLs is summarized in Figure 1c. There are eight reported GBM-modified ETLs in the green area of Figure 1c, which indicates the most effective ETL configurations using GBMs. Of the high-performance solar cells with PCEs ≥ 19%, Tavakoli et al. [43] and Mahmoudi et al. [55] reported a PCE > 30% with the use of a GBM sandwich layer. GQDs and GR have been shown to be the best GBM candidates, as exemplified by PCE improvements from 14.0% to 19.2% [77] and 13.82% to 19.81% [43]. In addition, PSCs with SnO 2 GQDs have yielded a PCE as high as 21.1% [19]. Figure 2 illustrates the energy levels of PSC component layers, including ETLs that have been modified using GBMs relative to the GBM-free analogues. As the ETL is required to extract and transport electrons as well as to block hole transport from the perovskite layer, materials suitable for ETLs should possess CBM or lowest unoccupied molecular orbital (LUMO) levels slightly lower than those of the perovskites; this situation enables electron transfer from the former to the latter. At the same time, ETLs should possess valence band maximum (VBM) or highest occupied molecular (HOMO) levels considerably lower than those of the perovskites; this situation prevents holes from entering the ETLs. The addition of GBMs can be used to tune the band energy levels of ETLs, viz., Energies 2020, 13,6335 4 of 24 engineer the E g , in order to achieve these ends in PSCs. For the entire cell, the direction of electron transport also requires the electrode (cathode) VBM to be above that of the HTM and that of the HTM to be above that of the perovskite as well as the CBM of the TCO (anode) to be below that of the ETL. Of course, this configuration also meets the requirement for the counter-transport of holes. exhibits zero band gap (Eg), which precludes its use as a semiconductor. However, chemical synthesis rarely produces pure GR, so what often is reported as GR is slightly impure. Compared to graphene, graphene derivatives are Eg-tunable and relatively low-cost materials that involve less complex fabrication methods. The most commonly used graphene derivatives are, as shown in Figure 1a, graphene oxide (GO), reduced graphene oxide (RGO), and graphene quantum dots (GQDs) that contain hydroxyl and carboxyl functional groups [47][48][49], which enable them to display hydrophilicity instead of hydrophobicity that is seen in graphene. These functional groups are beneficial to dispersibility in suspensions during processing using water and organic media of high dielectric constants [50]. The flexibility of GBMs has allowed them to be used in different parts of PSCs, including electrodes [51,52], perovskite layer [53,54], charge-transfer layer [55][56][57][58][59], and as interfacial layer [60][61][62].   [17][18][19][20]23,[25][26][27]39,43,55,56,60,[63][64][65][66][67][68][69][70][71][72][73][74] (TCO = transparent conducting oxide, HTM = hole transport material, green area = most efficient configurations for GBM-modified ETLs).

TiO 2
TiO 2 is the most widely used ETL material for n-i-p structure PSCs [5]. TiO 2 exists as two main polymorphs, anatase and rutile, where the latter is more thermodynamically stable [82]. Although the anatase → rutile phase transformation temperature is variable, the commonly cited temperature is 600 • C. Two of the key influences on this temperature are impurities and dopants, so the composition of the adjacent perovskite layer is relevant because its cations tend to promote (A site cations of 2+ valence) or inhibit (B site cations of 4+ valence) this phase transformation. Both polymorphs have been reported to be present in PSCs [83][84][85][86]. Which polymorph is present is important because rutile has a lower E g (~3.0 eV) than that of anatase (~3.2 eV) but the electron-hole recombination rate of the latter is slower [82].
TiO 2 can be presented in ETLs in two nanostructural forms, viz., compact (dense) TiO 2 (c-TiO 2 ) and mesoporous TiO 2 (m-TiO 2 ) [87]. In ETLs, TiO 2 is present as either a bilayer structure of compact plus mesoporous TiO 2 or a monolayer of compact TiO 2 alone. The c-TiO 2 layer acts as a hole blocker [28] while the m-TiO 2 offers the advantage that penetration of the perovskite liquid sol-gel precursor in the mesopores results in increased contact surface area and hence enhanced electron transfer [15]. The use of bilayer TiO 2 is a common and effective way to configure high-performance PSCs of high PCE, as reflected by reduced forward and reverse current-voltage (I-V) hysteresis [88,89]. High PCEs result from these nanostructural characteristics as well as the favorable TiO 2 E g of~3.0 eV or~3.2 eV and CBM level of about -4.4 eV, which allow easy extraction and transfer of electrons from the perovskite layer and higher electron lifetimes. Additional advantages are relatively simple fabrication processes (e.g., spin coating of dispersions or sol-gel dispersions [15,28,83]) and low fabrication costs (e.g., ≤1 h at 100-500 • C sintering). However, TiO 2 has a low bulk electron mobility (1 cm 2 /V s); TiO 2 polymorph not specified [90]) and TiO 2 may reduce the stability of the adjacent perovskite layer owing to the Energies 2020, 13, 6335 6 of 24 former's photocatalytic properties under UV illumination [91]. Consequently, the modification of TiO 2 to overcome these two shortcomings has been the subject of recent research.

Compact TiO 2 Layer
Wang et al. [63] were the first to report the use of a TiO 2 -GR ETL layer in an n-i-p structure PSC. The addition of graphene to the c-TiO 2 layer of bilayer TiO 2 not only enabled the low-temperature fabrication at 150 • C but it also improved the PCE performance to a maximum of 15.6% from the 10.0% obtained from TiO 2 alone. The optimal amount of 0.6 wt % graphene addition contributed to an average open-circuit voltage (V oc ) of 1.05 V, a high short-circuit current density (J sc ) of 21.9 mA/cm 2 , and a high fill factor (FF) of 0.73.
Ryu et al. [18] were the first to coat TiO 2 with GQDs (14, 10, or 7 nm size). The devices consisted of an TCO/ETL/perovskite/HTM/electrode configuration using the materials FTO/c-TiO 2 /perovskite/ Spiro-OMeTAD/Au for an n-i-p structure PSC. An average PCE of 17.80% was achieved using 14 nm GQDs, representing an~8% improvement relative to the reference PSC average of 16.48%. O 2 plasma treatment was found to be beneficial to improved bonding between the TiO 2 and GQDs, as suggested in the corresponding FTIR spectra shown in Figure 3a. After O 2 plasma treatment, the appearance of two absorption bands at 1640 cm −1 and 1070 cm −1 is attributed to carboxylate bonding between the TiO 2 and GQDs. This carboxylate bonding would facilitate the electron transfer from the GQDs to the c-TiO 2 .
Energies 2020, 13, x FOR PEER REVIEW 6 of 24 The introduction of different sizes of the GQDs also contributed to improved recrystallization of the perovskite [18]. Figure 3b shows increased (110) perovskite X-ray diffraction (XRD) peak intensities of the TiO2-GQD ETLs relative to the reference ETL. The effects of the GQDs on the electron transport properties also were analyzed and these revealed that the GQD CBM level could be tuned and lowered to closer proximity to the TiO2 CBM level, which had a significant impact on control of the charge transfer behavior and efficiency. Further, the lower CBM level was effective in electron extraction, as revealed by an increased quenching level in steady-state photoluminescence (PL) spectra. The resultant current density (J) increased proportionally with GQD size, as shown in Figure  3c, and the external quantum efficiency (EQE, which also is known as the incident photon-to-current The introduction of different sizes of the GQDs also contributed to improved recrystallization of the perovskite [18]. Figure 3b shows increased (110) perovskite X-ray diffraction (XRD) peak intensities of the TiO 2 -GQD ETLs relative to the reference ETL. The effects of the GQDs on the electron transport properties also were analyzed and these revealed that the GQD CBM level could be tuned and lowered to closer proximity to the TiO 2 CBM level, which had a significant impact on control of the charge transfer behavior and efficiency. Further, the lower CBM level was effective in electron extraction, as revealed by an increased quenching level in steady-state photoluminescence (PL) spectra. The resultant current density (J) increased proportionally with GQD size, as shown in Figure 3c, and the external quantum efficiency (EQE, which also is known as the incident photon-to-current efficiency, IPCE), as well as the integrated photocurrent measurements, as shown in Figure 3d.

Mesoporous TiO 2 Layer
The utilization of GR and GBMs to modify the semiconducting properties of m-TiO 2 also has been reported. Han et al. [23] reported an m-TiO 2 -RGO nanocomposite as ETL in n-i-p structure PSCs with improved electron transfer properties. The m-TiO 2 -RGO was fabricated by mixing a TiO 2 -ethanol slurry in various TiO 2 /RGO volume ratios and spin coating. The successful incorporation of RGO in m-TiO 2 was confirmed by micro-Raman microspectroscopy using the RGO peaks at 1350 cm −1 (D band) and 1580 cm −1 (G band). The optimal concentration of RGO was found to be 0.4 vol%, which resulted in a PCE and J sc increases from 11.5% to 13.5% and 19.6 mA/cm 2 to 21.0 mA/cm 2 , respectively.
The improved PCE after the addition of RGO was attributed to the resultant lower resistivity and reduced charge recombination [23]. Hall measurements showed that the resistivity of the composite with 0.4 vol% RGO decreased by an order of magnitude. Reduced series resistance and charge recombination were demonstrated by electrochemical impedance spectroscopy (EIS). Further, the work function (Φ) of the RGO of −4.4 eV was lower than and close to the CBM of perovskite (−3.93 eV), thus facilitating electron transfer from the perovskite to the RGO.
Sidhik et al. [17] modified m-TiO 2 by adding graphene nanoplatelets (GNPs) and spin coating the resultant dispersions. A PSC with 2 wt % GNP achieved a maximal PCE of 19.23% and a fill factor of 80% owing to the CBM band alignment and consequent efficient electron transport. The authors speculated that the GNPs passivated the surface defects of TiO 2 and so reduced charge recombination. This was examined through the XPS data shown in Figure 4a for the O 1s peak of TiO 2 , which was deconvoluted into the I Ti-O peak for lattice oxygen at 530.32 eV and the I o peak for an oxygen vacancy at 532.6 eV. Their varying ratios reveal that the lowest I O /I Ti-O was for the 2 wt % GNP ETL. This maximum was supported by photoluminescence (PL) data, shown in Figure 4b, which also suggested that the addition of 3.0 wt % GNP resulted in the generation of an unspecified type of surface defect. It also is possible that the higher amount of GNPs resulted in agglomeration and consequently reduced surface coverage. A further potential benefit of the coverage by GNPs was a reduction in surface roughness owing to the filling of the troughs, which would have increased the extent of the effective m-TiO 2 /perovskite surface apposition and consequent electron transport and PCE.
Cho et al. [65] fabricated mesoporous ETLs from spin coated mixtures of TiO 2 and RGO, sintered them, and then treated the surface with Li-TFSI in order to facilitate electron injection into the ETL. The PSCs exhibited a maximal PCE of 19.54%. The incorporation of RGO led to quenching in time-resolved PL, indicating efficient electron extraction by the ETL owing to a synergistic effect between the Li-TFSI and RGO. The PSC also exhibited enhanced V oc and FF without impacting on the carrier-transfer rate. Biccari et al. [20] compared the characteristics of two configurations of ETLs in MAPbI 3 PSCs, both of which included a perovskite/m-TiO 2 interlayer of spin coated lithium-neutralized graphene oxide (GO-Li). The m-TiO 2 interlayers differed in that one consisted of spin coated m-TiO 2 and the other consisted of spin coated m-TiO 2 -GR. These devices exhibited well crystallized perovskite identified as tetragonal at all temperatures. The PL spectra at the same excitation energy of the configuration with m-TiO 2 -GR revealed an order of magnitude lower trap density. Additionally, the carrier collection efficiency of the configuration with m-TiO 2 -GR approximately doubled. Agresti et al. [61] used an FTO/c-TiO 2 /m-TiO 2 -GR/GO-Li/perovskite/Spiro-OMeTAD/Au configuration for PSCs of 50.6 cm 2 area module which achieved enhanced performance and stability. The PCE increased from 11.6% for a PSC without GR or GO-Li to 12.6% with these materials. The long-term performance was very good on the device with only m-TiO 2 -GR, retaining~91% of the initial PCE after 1630 h (in the dark, dry conditions (humidity rate < 30%) and at open circuit).
which was deconvoluted into the ITi-O peak for lattice oxygen at 530.32 eV and the Io peak for an oxygen vacancy at 532.6 eV. Their varying ratios reveal that the lowest IO/ITi-O was for the 2 wt % GNP ETL. This maximum was supported by photoluminescence (PL) data, shown in Figure 4b, which also suggested that the addition of 3.0 wt % GNP resulted in the generation of an unspecified type of surface defect. It also is possible that the higher amount of GNPs resulted in agglomeration and consequently reduced surface coverage. A further potential benefit of the coverage by GNPs was a reduction in surface roughness owing to the filling of the troughs, which would have increased the extent of the effective m-TiO2/perovskite surface apposition and consequent electron transport and PCE.  Shen et al. [39] infiltrated m-TiO 2 with a GQD dispersion (1 mg/mL) by pre-sintering the m-TiO 2 layer, soaking in the GQD dispersion for 10 s, spin coating once or twice with the same dispersion, and heating at 100 • C for 30 min. Modification of the ETL of the PSC in this way enhanced electron extraction and so facilitated charge transport. Again, the deposition of GQDs smoothed the surface of the m-TiO 2 , decreasing the surface roughness from 24.8 to 22.5 nm. However, the authors observed that the J-V performance and parameters decreased when two GQD coatings were applied. The Nyquist plot suggested that the reason for this was enhanced recombination. However, it also is possible that excessive amounts of GQDs, which act as electron traps, establish high electron concentrations that effectively reverse the direction of electron diffusion.
Lamanna et al. [79] recently reported a comparison of a two-terminal perovskite/silicon heterojunction (JHT) tandem solar cell, as shown in Figure 4c, with and without GR added to both the m-TiO 2 and c-TiO 2 layers. These data showed that the PCE increased from 24.2% to 26.3%, as shown in Figure 4d, with retention at 25.9% after stabilization (inset). The V oc increased from 1.76 V to 1.80 V, the J sc remained essentially unchanged, and the FF increased from 0.73 to 0.75. The voltage scans showed only marginal hysteresis, with 25.0% (forward) and 26.3% (reverse).
As shown in Figure 4e, the authors [79] also prepared a bifacial PSC of FTO/c-TiO 2 /m-TiO 2 / perovskite/HTM/ITO configuration; again, the comparison was between PSCs with and without GR added to both the m-TiO 2 and c-TiO 2 layers. These data showed that a PCE increase from 11.3% to 12.6%, a V oc increase from 1.08 V to 1.12 V, a slight J sc decrease from 19.96 mA/cm 2 to 19.85 mA/cm 2 , and an FF increase from 0.53 to 0.57. The addition of GR also decreased the extent of hysteresis. voltage scans showed only marginal hysteresis, with 25.0% (forward) and 26.3% (reverse).

SnO 2
SnO 2 also has been explored as an ETL since it exhibits excellent optical transmission, a suitable E g , a relatively low CBM level, and high electron mobility [92][93][94][95]. ETLs of compact SnO 2 thin films as planar layers from sol-gel solutions have achieved a maximal PCE of 21.6% [96,97]. SnO 2 has advantages over TiO 2 in the form of lower processing temperatures (55-180 • C) [93,98,99] and the absence of UV-induced degradation [100]. However, surface defects, such as oxygen vacancies, resulting from the synthesis of SnO 2 may hinder electron transfer and reduce the performance of PSCs [101]. GBMs have been added to SnO 2 to passivate surface defects and tune the semiconducting properties, including the electrical conductivity and energy levels [24,26,56,93,98,99].
Hong et al. [26] reported a composite of SnO 2 and nitrogen-doped graphene oxide (NGO) as an ETL in planar n-i-p structure PSCs, where the surface oxygen vacancies were passivated by NGO. Fabrication involved a dispersion of NGO, which acted as an oxidizing agent, and SnCl 2 ·2H 2 O, which was dissolved in ethanol, after which the spin-coated layer was calcined at 180 • C. A peak PCE of 16.54% was obtained by a PSC with 5 vol% NGO and enhanced V oc and FF of 1.17 V and 0.75, respectively, were determined. The oxidation of Sn 2+ to Sn 4+ to form SnO 2 was confirmed by the trends in deconvoluted Sn 3d 5/2 XPS data, as shown in Figure 5a,b, and by the deconvoluted O 1s XPS data, as shown in Figure 5c,d. These results demonstrate that GBMs are not limited to chemically passive agents for increased electrical conductivity. They also may act as oxidizing agents, as in NGO, or are reducing agents, as in GR. The SnO 2 -NGO ETL exhibited higher electrical conductivity, lower recombination rate, and more efficient charge extraction relative to the SnO 2 ETL.
Xie et al. [56] were the first to report GQDs as additives to SnO 2 in order to overcome the trap states in low-temperature solution-processed (i.e., defective) SnO 2 . ETLs modified with 1 wt % GQDs achieved an average PCE of 19.2 ± 1.0% with reduced hysteresis, which was attributed to facilitated electron transport and reduced charge recombination. As shown in Figure 5e, the J-V characteristics for ETL-only devices of Ag/SnO 2 /ITO and Ag/SnO 2 -GQD/ITO configuration showed that the electrical conductivity under the illumination of the latter was greater than that of the former by a factor of~20; the same factor was obtained for comparison of the Ag/SnO 2 -GQD/ITO configuration for illuminated vs. dark conditions. The improved electrical conductivity was attributed to the role of the GQDs in providing segregated trap states in SnO 2 , which accumulate large isolated concentrations of free electrons, thereby acting as localized reservoirs that avoid recombination and facilitate efficient transport of electrons. Kelvin probe force microscopy (KPFM) measurements showed that the E F of SnO 2 -GQD was reduced from 4.35 to 4.01 eV under illumination by simulated sunlight (AM 1.5 G); this moved the E F closer to that of the CBM of SnO 2 , as suggested by Figure 5f. In contrast, equivalent Energies 2020, 13, 6335 10 of 24 KPFM measurements for SnO 2 revealed only a small shift in the E F . These results also indicated that ETLs of SnO 2 -GQD exhibited increased V oc relative to that of SnO 2 .
Fabrication involved a dispersion of NGO, which acted as an oxidizing agent, and SnCl2·2H2O, which was dissolved in ethanol, after which the spin-coated layer was calcined at 180 °C. A peak PCE of 16.54% was obtained by a PSC with 5 vol% NGO and enhanced Voc and FF of 1.17 V and 0.75, respectively, were determined. The oxidation of Sn 2+ to Sn 4+ to form SnO2 was confirmed by the trends in deconvoluted Sn 3d5/2 XPS data, as shown in Figures 5a and 5b, and by the deconvoluted O1s XPS data, as shown in Figure 5c,d. These results demonstrate that GBMs are not limited to chemically passive agents for increased electrical conductivity. They also may act as oxidizing agents, as in NGO, or are reducing agents, as in GR. The SnO2-NGO ETL exhibited higher electrical conductivity, lower recombination rate, and more efficient charge extraction relative to the SnO2 ETL.  Zhou et al. [24] fabricated spin-coated composites of SnO 2 and GQDs of 5 and 10 nm diameters (SnO 2 -5G, SnO 2 -10G) as ETLs, which resulted in PCE increases (0.5 wt % GQD additions) from 16.7% to 19.2% and to 18.0%, respectively, in PSCs. The optical and semiconducting properties as a function of GQD Ø and addition level were examined for ETL-only (Au/SnO 2 -GQD/FTO) devices and PSCs. Despite the complex stoichiometry of the (Cs, organic)Pb(Br,I) 3 , Figure 5g shows that the perovskite (deposited on the ETL) was nearly homogeneous. However, these data showed that the amount of grain-boundary PbI 2 , as shown in Figure 5h which may passivate defects, is reduced by the GQDs, which also was observed by Sidhik et al. [17].
Concerning the effects of GQD size, the electrical conductivity of the SnO 2 -G10 ETL was optimal [24]. However, the recombination rate for the SnO 2 -G5 ETL was optimal. However, all the band energy levels decreased with decreasing GQD size.
Concerning the effects of the GQD amount, a high concentration may be harmful to the performance of PSCs [24]. It was observed that the PCE, V oc , and J sc improved with increasing concentration of GQDs up to 0.5 wt %, after which the values of these parameters decreased, as shown in Figure 5i. Charge recombination on the interface also may be worsened with the additions of high concentrations of GQDs, as demonstrated by the Nyquist plots from EIS in Figure 5j. The interfacial recombination resistance (R rec ), which corresponds to the diameter of the typical semicircle at low frequency (<103 Hz), decreases when the concentration of GQDs exceeds 0.5 wt %. This indicates inefficient suppression of charge recombination or the possibility of the formation of new charge-recombination centers.
More generally, the utilization of GQDs significantly improved the efficiency of electron extraction and PCEs of the PSCs, with negligible change in the absorption of the perovskite layer. More specifically, the best device (0.5 wt % GQD of 5 nm Ø), with a PCE of 19.2%, exhibited a high V oc of 1.07 V, high J sc of 23.9 mA/cm 2 , high FF of 0.75, and reduced hysteresis.
Zhao et al. [60] fabricated ETLs using SnO 2 mixed with graphene-modified by N,N -bis-[2-(ethanoic acid sodium)]-1,4,5,8-naphthalene diimide (NDI) surfactant. The maximal PCE was 20.2% and the FF also was high at 0.82. The use of graphene-NDI (GR-NDI) facilitated the physisorption across the SnO 2 /perovskite interface by van der Waals bonding, which enhanced the semiconducting properties of the ETL. Zhu et al. [25] reported SnO 2 -GR as ETLs in PSCs with enhanced electron mobilities and reduced charge recombination, which resulted in highly stable devices with minimal hysteresis under humidity. These characteristics were attributed to the hydrophobicity of graphene and its passivation of the surface trap sites, the former of which would retard the degradation of the perovskite.

ZnO
ZnO is a promising semiconducting material that has the potential to replace TiO 2 as the ETL in PSCs. It has an E g of~3.3 eV, which is similar to that of TiO 2 , but higher bulk electron mobility (~200 cm 2 /V s) [102]. Another advantage of ZnO is that it can be synthesized at room-temperature [103,104], which reduces the cost and enables its deposition on flexible devices. However, it has been reported that ZnO-based PSCs exhibit thermal instability, which can lead to severe degradation of the perovskite at temperatures > 100 • C [105]. This degradation has been reported to arise from interaction between the perovskite and the surface hydroxyl groups and residual acetate ligands present on the ZnO [37,105]. Hence, it appears that GBMs may be successful at passivation since their addition has overcome this instability issue, resulting in enhanced electron transport [43,81].
Tavakoli et al. [81] reported using a quasicore-shell structure of ZnO-RGO quantum dots (QDs) as ETL to overcome the thermal instability resulting from the effect of room-temperature hydrolysis of the ZnO nanoparticles on the perovskite. The ZnO-RGO composite layer contributed to the PSC's retention of 90% of its original PCE after 30 days in ambient conditions following encapsulation, which is superior to that of ZnO nanoparticles alone, as shown in Figure 6a. This was attributed to the decreased concentration of hydroxyl groups on the ZnO surface owing to the RGO modification, as confirmed by XPS analysis. Further, improvements in electron transfer efficiency and more rapid electron extraction were additional benefits. Tavakoli et al. [43] reported the deposition of monolayer-GR by low-pressure chemical vapor deposition (LPCVD) on copper foil. The monolayer-GR then was transferred to the top of ZnO ETL as an interfacial layer in order to eliminate the instability caused by ZnO. The result was the achievement of thermal stability through successful hindrance of perovskite decomposition by prevention of direct contact between the two materials. This resulted in long-term stability of the PSCs, as shown in Figure 6b. The GR interlayer also contributed to faster electron extraction and transfer, resulting in PCEs with increased J sc and higher PCE with reduced hysteresis. of the ZnO nanoparticles on the perovskite. The ZnO-RGO composite layer contributed to the PSC's retention of 90% of its original PCE after 30 days in ambient conditions following encapsulation, which is superior to that of ZnO nanoparticles alone, as shown in Figure 6a. This was attributed to the decreased concentration of hydroxyl groups on the ZnO surface owing to the RGO modification, as confirmed by XPS analysis. Further, improvements in electron transfer efficiency and more rapid electron extraction were additional benefits. Tavakoli et al. [43] reported the deposition of monolayer-GR by low-pressure chemical vapor deposition (LPCVD) on copper foil. The monolayer-GR then was transferred to the top of ZnO ETL as an interfacial layer in order to eliminate the instability caused by ZnO. The result was the achievement of thermal stability through successful hindrance of perovskite decomposition by prevention of direct contact between the two materials. This resulted in long-term stability of the PSCs, as shown in Figure 6b. The GR interlayer also contributed to faster electron extraction and transfer, resulting in PCEs with increased Jsc and higher PCE with reduced hysteresis. GBMs can serve as means for efficient passivation for ZnO ETLs in the forms of both core-shell structures or monolayers, although this would be impacted by the surface chemistry of ZnO, which is highly dependent on the method and chemicals of synthesis. Alternative to these GBMs as semiconducting passivating interlayers are PCBM and polyethylenimine (PEI), which can be deposited by sol-gel [37]. These represent simple and cost-effective materials and methods to passivate ZnO and facilitate electron transfer.
Other reports of GBM-modified ZnO ETLs focused on the improvement of electron transport and the PSC performance. Ahmed et al. [75] reported that ZnO-GO exhibited lower resistivity than ZnO and hence lowered the series resistance of PSCs. Chandrasekhar et al. [76] reported the fabrication of ZnO-GR ETL by spray coating at 150°C on top of a ZnO compact layer. In comparison to ZnO, ZnO modified with 0.75 wt % GR contributed to improved crystallinity of perovskite thin films while the grain size increased from~100 nm to~200-250 nm, with decreased surface roughness. The EQE increased from~55% to >75%, clearly showing the benefit of GR modification on the semiconducting properties of the PSCs. However, a negative effect from the large concentration of GR (1.0 wt %) was observed in the form of decreased PCE and less quenching in the PL spectra.
Chandrasekhar et al. [74] reported the fabrication of a high-efficiency PSC using an ETL of ZnO modified with nitrogen-doped-GR (NGR). Wurtzite ZnO nanorods were deposited on a ZnO seed layer Energies 2020, 13, 6335 13 of 24 by chemical bath deposition, followed by high-temperature sintering (350°C). Increasing concentration of NGR was reported to increase both the crystallinity and grain size of the perovskite, the latter of which is shown in Figure 6c-e. Increasing the NGR content to 0.8 wt % caused the PCE to increase from 12.87% to 16.82%, the J sc to increase from 17.38 mA/cm 2 to 21.98 mA/cm 2 , the FF to increase from 0.71 to 0.74, the V oc to decrease slightly from 1.04 to 1.02 V, and the PSC hysteresis to disappear. The nanorod arrays on the seed layer were considered to act as a quasimesoporous ETL layer, similar to those observed in TiO 2 that also reduced the hysteresis [89].
Taheri-Ledari et al. [80] reported an HTM-free PSC with a composite ETL of ZnO with RGO and CuInS 2 quantum dots that showed good device stability. The RGO was synthesized in situ with ZnO using the hydrothermal method. The addition of RGO, which has a suitable CBM of −4.4 eV, enhanced the electron transfer and contributed to the improvement in the photovoltaic performance of PSCs. A mixture of 80 wt % ZnO + 16 wt % CuInS 2 + 4 wt % RGO sintered at 300 • C resulted in a uniform flat microstructure of 3-5 µm grains yielded a high PCE of 15.74%, with good stability.
Finally, ZnO of different nanostructures has been reported to be an effective ETL in the forms of nanoparticles, nanorods, nanowires, nanosheets, and 3D hierarchical structures [87]. However, there is a dearth of data for GQD modification of ZnO structures for ETLs and these GBMs would be expected to be able to make a positive impact on both electron transfer and overall PSC performance.

PCBM
Organic materials, such as fullerenes and their derivatives, have been investigated as ETLs in p-i-n type PSCs because they offer simplified fabrication, low-temperature processing, low costs, and feasibility for coatings on flexible substrates [106]. The complex fullerene [6,6]-phenyl-C 61 -butyric acid methyl ester (PCBM) is an organic ETL material that has been used extensively in PSCs. It has been reported that PCBM passivates the perovskite layer and suppresses hysteresis in PSCs [107,108]. Recently, GBMs have been mixed with PCBM to enhance the photovoltaic properties of PSCs. This effectively improved not only the semiconducting properties of the ETL but also the stability of the devices [70,72].
Yang et al. [70] reported data for PCBM and GQD-modified PCBM (PCBM-GQD), which were ETLs in p-i-n structure PSCs. The use of 0.5 wt % GQD increased the PCE from 14.68% to 17.56%. The electrical conductivity increased from 0.151 mS cm −1 to 0.422 mS cm −1 . PCBM-GQD also showed enhanced electron extraction, which resulted in higher V oc and FF values. The unsealed PSC also exhibited stability owing to the maintenance of the PCE at~80% and the FF at~0.9after 300 h of continuous simulated solar illumination in the air at~45% relative humidity. In contrast, the reference PSC with PCBM ETL showed a >50% decrease in PCE and a~30% decrease in FF under the same conditions, as shown in Figure 7a,b. It was suggested that enhanced stability may result from the suppression of the dimerization of PCBM under light exposure through the addition of GQDs.
Kakavelakis et al. [72] reported the performances of RGO-modified PCBM in ETLs in p-i-n structure PSCs, which enhanced the electrical conductivities and stabilized the PCBM/perovskite interface in PSCs. At the optimal concentration of 5 wt % RGO in PCBM, the electrical conductivity increased from 0.109 to 0.495 mS·cm −1 , the J sc increased from 20.65 to 22.92 mA/cm 2 , and the photo-charging effect during light soaking also reduced. The V oc and FF of the unsealed PSCs showed only very minor changes after 100 h illumination under AM 1.5G simulated solar light, as shown in Figure 7c,d, indicating only minor perovskite degradation and stabilized PCBM/perovskite interface.
Bi et al. [71] reported data for PCBM and PCBM modified with 2 wt % N-doped graphene (PCBM-NGR) in p-i-n structure PSCs of FTO/NiMgLiO/MAPbI 3 /PCBM-NGR/CQDs/Ag configuration, which showed significant improvements in stability owing to successful hindrance of the diffusion of iodide ions (the natures of 20 nm thickness NiMgLiO and the 10 nm Ø carbon quantum dots (CQDs) were not described). As the NGR nanoparticles exhibited a dimension of~100 nm Ø, this large size would facilitate passivation and consequent suppression of penetration into the ETL. The long-term stability of sealed PSCs was tested under both illuminated and dark conditions, the results of which are shown in Figure 7e. These data showed that the PSCs with PCBM-NGR ETLs were stable and exhibited PCEs >15% under both conditions. Testing of the ETL-only devices revealed that the PCBM-NGR (relative to PCBM) exhibited superior semiconducting properties and more homogenous nanostructures. The electrical conductivity increased from 4.1 to 23.1 nA (bias 1 V, current through 150 nm ETL film thickness). More rapid electron extraction and slower carrier recombination rates were demonstrated by PL and time-resolved photoluminescence (TRPL). The impact on morphology was characterized by AFM, where fewer PCBM agglomerates and lower roughness were observed, where the roughness decreased from 6 to 1 nm.
Energies 2020, 13, x FOR PEER REVIEW 13 of 24 enhanced electron extraction, which resulted in higher Voc and FF values. The unsealed PSC also exhibited stability owing to the maintenance of the PCE at ~80% and the FF at ~0.9after 300 h of continuous simulated solar illumination in the air at ~45% relative humidity. In contrast, the reference PSC with PCBM ETL showed a >50% decrease in PCE and a ~30% decrease in FF under the same conditions, as shown in Figure 7a,b. It was suggested that enhanced stability may result from the suppression of the dimerization of PCBM under light exposure through the addition of GQDs. Kakavelakis et al. [72] reported the performances of RGO-modified PCBM in ETLs in p-i-n structure PSCs, which enhanced the electrical conductivities and stabilized the PCBM/perovskite interface in PSCs. At the optimal concentration of 5 wt % RGO in PCBM, the electrical conductivity increased from 0.109 to 0.495 mS·cm −1 , the Jsc increased from 20.65 to 22.92 mA/cm 2 , and the photocharging effect during light soaking also reduced. The Voc and FF of the unsealed PSCs showed only Kim et al. [78] developed GNPS that were selectively edge-functionalized (EF) with hydrogen (GNP-H) or fluorine (GNP-F) prepared by ball milling of graphite under H 2 or XeF 2 gas. These nanoparticles were dispersed in isopropanol (5 mg/mL) and spin coated on the PCBM layer of p-i-n structure PSCs of ITO/HTM/MAPBI 3 /PCBM/EF-GNP/Al configuration. The unsealed PSCs with ETLs of GNP-F showed good stability, maintaining 82% of the initial average PCE of 13.4% after 30 days in ambient air with~50% relative humidity under illuminated (simulated sunlight, AM 1.5G) and dark conditions, as shown in Figure 7f. This excellent stability is attributed to the hydrophobicity of the fluorine, which protected the perovskite layer from moisture.

Other ETL Materials
There is only limited information on other ETL materials with GBMs beyond those discussed in the present work. These appear to be limited to reports on ETLs based on Fe 2 O 3 [77] and compact SrTiO 3 (STO) with mesoporous Al 2 O 3 [55].

Summary and Outlook
In summary, the present work reviews the major studies of GBM-containing ETLs. The main bases for these components of PSCs are TiO 2 , SnO 2 , ZnO, and PCBM. Table 1 summarizes PSCs with ETLs modified by GBMs of the highest PCEs. Table 2 provides an extensive summary of the performances of PSCs with GBM-modified ETLs.
GQDs are suitable for addition to the dense microstructures of metal oxide semiconductors, such as TiO 2 , SnO 2 , and ZnO. The PSC performances using such materials are controlled largely by the size of the GQDs. In contrast, nanoparticles of GR, GO, and RGO, which are larger than GQDs, are more effective in mesoporous microstructures. In this case, the PSC performances are dominated largely by the concentrations of the GBMs rather than the sizes.
A single layer of GBM, typically deposited on the top of the ETL and functioning as an interfacial layer, represents a beneficial configuration for the improvement semiconducting properties of the ETL and hence the charge transfer across the ETL/perovskite interface through passivation of the ETL surface and enhancing interfacial bonding. Since a thin and uniform GBM layer is challenging to deposit using a suspension, vacuum deposition represents a promising method for such deposition.
From examination of the principal reports on GBM-modified ETLs and consideration of the relative merits and shortcomings of these devices, a number of conclusions can be made concerning the role of the GBMs. Owing to their outstanding semiconducting properties, their addition to ETLs can be expected to offer several advantages: • Enhanced electron transport from increased electrical conductivity and electron mobility of the ETL; • Enhanced electron transport from improved ETL/perovskite effective interfacial apposition; • Enhanced J-V characteristics and parameters of PSCs largely owing to more effective electron extraction and transport deriving from the ETLs; • Enhanced J-V characteristics of PSCs through reduced hysteresis owing to leveraging of surface trap sites created by GBMs; • Potential tuning of the E g and the energy levels to optimise charge transfer to and from the ETL; • Enhanced device stability (i.e., protection of the perovskite from water vapour-hydrophobicity) from passivation of the ETL at the ETL/perovskite interface; • Enhanced device stability through prevention of alteration of perovskite by inhibition of diffusion and counter-diffusion of mobile ions across the ETL/perovskite interface; • Enhanced device stability through increased crystallinity of perovskite from the action of GBMs as nucleating agents.     These multiple factors are responsible for the variable effects of GBM size, amount, and distribution on the performance through alteration of the nanostructure and associated features, including the impact of surface coverage and agglomeration on the electron diffusion path lengths and directions, interfacial electron transfer, and electron mean free path. Ultimately, these factors are directly relevant to the probability that the kinetics of charge transport are extended such that charge recombination can take place.
Although GBMs generally are assumed to be physisorbed on nanoparticle surfaces, the observation of chemical effects suggests that it may be possible to chemisorb GBMs. This outcome would be likely to generate robust nanostructures of enhanced electron transfer capacities. Modification of the surfaces of GBMs by, for example, redox, functionalization, or intercalation, also represents a strategy to improve the stability of PSCs.
The near-universal method of fabrication of these PSCs is spin coating, which is simple, rapid, and inexpensive while yielding high-quality layered nanostructures. The raw materials include both dispersions and sol-gel solutions. However, there may be size limitations to spin coating, where the film consistency would tend to decrease with increasing size and the cost would increase similarly. Emerging methods, such as 3D printing, screen printing, atomic layer deposition (ALD), and cold-spray deposition, offer promising trajectories.
Author Contributions: X.D. undertook the research, prepared the first draft of the manuscript, and contributed to all subsequent drafts; P.K. contributed to all preliminary drafts; C.C.S. prepared the penultimate and final drafts on the basis of the preliminary drafts; J.L. contributed to all preliminary drafts; J.S.Y. supervised the research and contributed to all preliminary drafts. All authors have read and agreed to the published version of the manuscript.
Funding: This research received no external funding.