**Advances in Emerging Solar Cells**

Special Issue Editor **Munkhbayar Batmunkh**

MDPI • Basel • Beijing • Wuhan • Barcelona • Belgrade • Manchester • Tokyo • Cluj • Tianjin

*Special Issue Editor* Munkhbayar Batmunkh Griffith University Australia

*Editorial Office* MDPI St. Alban-Anlage 66 4052 Basel, Switzerland

This is a reprint of articles from the Special Issue published online in the open access journal *Nanomaterials* (ISSN 2079-4991) (available at: https://www.mdpi.com/journal/nanomaterials/ special issues/nano solar cells).

For citation purposes, cite each article independently as indicated on the article page online and as indicated below:

LastName, A.A.; LastName, B.B.; LastName, C.C. Article Title. *Journal Name* **Year**, *Article Number*, Page Range.

**ISBN 978-3-03928-979-0 (Hbk) ISBN 978-3-03928-980-6 (PDF)**

c 2020 by the authors. Articles in this book are Open Access and distributed under the Creative Commons Attribution (CC BY) license, which allows users to download, copy and build upon published articles, as long as the author and publisher are properly credited, which ensures maximum dissemination and a wider impact of our publications.

The book as a whole is distributed by MDPI under the terms and conditions of the Creative Commons license CC BY-NC-ND.

## **Contents**



## **About the Special Issue Editor**

**Munkhbayar Batmunkh** is currently a research fellow at Centre for Clean Environment and Energy (CCEE) at Griffith University, Australia. Dr. Munkhbayar Batmunkh also worked at the University of Queensland (2018–2019) and the Flinders University of South Australia (2017–2018), and was a visiting scholar at Virginia Tech, USA. He completed his Ph.D. study in the School of Chemical Engineering at the University of Adelaide, Australia, in 2017. He obtained his M.Eng. degree from Gyeongsang National University, South Korea, in 2012. He completed his B.S. in Chemistry at the National University of Mongolia, Mongolia, in 2010. Dr. Munkhbayar Batmunkh's research interests focus on the production of functional nanomaterials (e.g., nanocarbons and 2D materials) for energy-related applications such as solar cells and catalysis. He has contributed to the fields by publishing more than 70 refereed journal articles in top-ranking journals.

## *Editorial* **Advances in Emerging Solar Cells**

#### **Munkhbayar Batmunkh**

Centre for Clean Environment and Energy, Griffith University, Gold Coast, Queensland 4222, Australia; m.batmunkh@griffith.edu.au

Received: 4 March 2020; Accepted: 12 March 2020; Published: 17 March 2020

There has been a continuous increase in the world's electricity generation and consumption over the years. Today's energy requirements are principally met by burning fossil fuels. However, in addition to increasing fuel prices, greenhouse gas emissions caused by the fuel-burning process have become a serious issue. As such, the development of renewable and sustainable energy technologies is of great importance. Direct conversion of the sunlight into electricity using photovoltaic (PV) devices is now considered as a mainstream renewable energy source. According to the international energy agency (IEA) [1], the world's total renewable-based power capacity is expected to grow by 50% between 2019 and 2024. Interestingly, solar PV accounts for more than 50% of this rise.

The PV market is currently dominated by technologies based on crystalline (poly + single) silicon. These silicon-based solar cells are a mature technology and can deliver a power conversion efficiency (PCE) of approximately 20% under full-sun illumination. Although significant reductions in the price of silicon PV cells have been observed, these technologies still suffer from high installation costs. Many scientists and researchers in the field of PV have paid particular attention to the development of a viable alternative PV technology. In this regard, emerging solar cells have received intense attention because these classes of solar cells, in comparison to traditional silicon PVs, promise to be less expensive, lighter, more flexible, and portable. Despite these great features, there are several challenges that restrict the possible commercialization of these technologies. This has led to significant efforts being focused on addressing issues associated with emerging solar cells. This Special Issue presents twelve excellent articles, ten research and two review papers, covering perovskite solar cells (PSCs) [2–8], heterojunction solar cells (HJSCs) [9], organic solar cells (OSCs) [10], dye-sensitized solar cells (DSSCs) [11], and PV materials [12,13].

The first report on organic–inorganic hybrid perovskite for solar cells was published in 2009 by Kojima et al. [14], and achieved a PCE of 3.8%. Since then, excellent achievements have been made in the PSC field and the certified efficiency of PSCs has now exceeded 25%, making them the fastest advancing PV technology. In this Special Issue, McDonald et al. [8] provided an excellent overview of PSCs and outlined the recent advances that have been made in nanoscale perovskites such as low-dimensional perovskites, perovskite quantum dots, and perovskite-nanocrystal based solar cells. Chang et al. [7] discussed the hot-carrier characteristics of perovskite light absorbers, which play a critical role in high efficiency PSCs. They also pointed out the practical issues hindering the development of highly efficient perovskite-based hot-carrier solar cells. The authors presented their own perspective on the future development of hot-carrier PSCs.

Although PSCs are very attractive and highly efficient, they suffer from several serious limitations. A typical PSC is fabricated using a transparent conductive electrode such as indium–tin oxide (ITO) and fluorine-doped tin oxide (FTO). However, these transparent electrodes are expensive and have natural brittleness and poor mechanical robustness. Two research articles in this Special Issue reported alternative transparent electrodes to the conventional ITO/FTO. Lu and colleagues [3] demonstrated that the composite electrode of silver nanowires and large area graphene oxide (Ag NWs/LGO) can exhibit comparable device performance to the standard ITO based PSCs. Chen et al. [5] designed a hexagonal Ni (30 nm)/Au (10 nm) mesh that showed a transmittance close to 80% in the visible light

region and a sheet resistance lower than 16.9 Ω/sq. This metal mesh, when used in device fabrication, displayed a PCE of 13.88%, which was comparable to that of the ITO-based PSC.

Phenyl-C61-butyric acid methyl ester (PCBM) is the mostly commonly used electron transporting material in the p-i-n type (inverted) PSCs. However, the energy barrier at the interface between the PCBM layer and metal electrode limits the photogenerated charge extraction and thus results in reduced device efficiencies. In order to tackle this issue, Dong et al. [2] used a room temperature, solution processed Al-doped ZnO (AZO) as an interlayer between the PCBM and Ag electrode. The PSC device fabricated with an AZO interlayer not only exhibited a promising PV efficiency, but also showed excellent device stability. Incorporating additives into the perovskite has been proven to be a promising strategy to enhance the efficiency of PSCs. Wu et al. [4] explored the influence of adding water and potassium halides (KCl, KBr, and KI) into the PbI2 precursor solutions on the PV performance of PSCs. By co-doping with KI and water, they significantly improved the efficiency of CH3NH3PbI3 perovskite based solar cells. In PSCs, hole transporting materials (HTMs) play a critical role in selecting holes and transporting them to the conductive electrodes. High efficiency PSCs rely on expensive HTMs such as 2,2- ,7,7- -Tetrakis[*N*,*N*-di(4-methoxyphenyl)amino]-9,9- -spirobifluorene (Spiro-OMeTAD) and poly[bis(4-phenyl)(2,4,6-trimethylphenyl)amine] (PTAA). In addition to their high costs, the devices fabricated using these HMTs suffer from poor stability in ambient conditions. Therefore, developing a novel HMT is of great interest. Wang et al. [6] designed a new type of HTM, named 4,4- -(9-methyl-9H-carbazole-3,6-diyl)bis(*N*,*N*-bis(4-methoxyphenyl)aniline) (CZTPA), as an alternative to the traditional Spiro-OMeTAD. This new HTM based PSC achieved a PCE of 11.79%, which was comparable to that (11.74%) of the non-doped Spiro-OMeTAD, while showing better stability in ambient conditions.

Solution-processed CdTe based HJSCs have attracted a great deal of attention from the PV community. However, the efficiencies of this class of HJSCs are still very limited. Mei et al. [9] developed an efficient approach to enhance the efficiency of CdTe/TiO2 HJSCs by inserting a thin layer of CdS nanocrystal between the CdTe and TiO2 layers. OSCs have many attractive properties such as high flexibility, solution processability, light weight, and simple manufacturing. In a typical OSC, poly(3,4-ethylenedioxythiophene):poly(styrenesulfonate) (PEDOT:PSS) is used as a HTM. However, the major drawback in using PEDOT:PSS in OSCs is the surface energy mismatch between the PEDOT:PSS and the active layer. To overcome this issue, Ramasamy et al. [10] used oleylamine-functionalized MoS2 in the PEDOT:PSS layer. By using this strategy, they observed a 15.08% enhancement in the device performance. DSSCs are an attractive emerging PV due to their eco–friendliness, ease of fabrication, and cost effectiveness. Designing a new type of dye molecule as a light harvesting material is still a hot area of research. Ambroz et al. [11] developed two bodipy dyes with different carboxylic acids on the meso-position of the bodipy core and used them to sensitize TiO2 photoelectrodes for DSSCs.

Exploring new synthesis methods, properties, and functionalization of PV materials is of great importance. Yang et al. [12] studied the electrical properties of 4H-silicon carbide (SiC) Schottky barrier diodes (SBDs) under high-dose electron irradiation. They used in-situ noise diagnostic analysis to demonstrate the correlation of irradiation-induced defects and microscopic electronic properties. Semiconductor SiC is widely used in electronic devices such as inverters, which deliver energy from PV arrays to the electric grids and other applications. Furthermore, Naffeti et al. [13] used a facile, reliable, and cost-effective metal assisted chemical etching method to fabricate highly crystalline vertically aligned silicon nanowires (SiNWs). SiNWs are widely used not only in solar cells, but also in other applications including lithium-ion batteries, sensors, electronics, and catalysis. SiNWs fabricated in this work [13] showed a strong decrease in the reflectance, demonstrating that these SiNWs are an excellent candidate for PV cells.

Finally, I believe that these articles will be of wide interest for the broad readership of the journal (*Nanomaterials*).

**Funding:** This research received no external funding.

**Acknowledgments:** The Guest Editor would like to thank all authors for submitting their work to the Special Issue. Special thanks also go to all the reviewers for their prompt responses and for making constructive suggestions that enhance the publication quality and impact. I am also grateful to Sandra Ma and the editorial assistants who made the Special Issue creation a smooth and efficient process.

**Conflicts of Interest:** The author declares no conflicts of interest.

#### **References**


© 2020 by the author. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Review* **Nanostructured Perovskite Solar Cells**

**Calum McDonald 1,\*, Chengsheng Ni 2, Paul Maguire 3, Paul Connor 4, John T. S. Irvine 4, Davide Mariotti <sup>3</sup> and Vladimir Svrcek <sup>1</sup>**


Received: 26 September 2019; Accepted: 12 October 2019; Published: 18 October 2019

**Abstract:** Over the past decade, lead halide perovskites have emerged as one of the leading photovoltaic materials due to their long carrier lifetimes, high absorption coefficients, high tolerance to defects, and facile processing methods. With a bandgap of ~1.6 eV, lead halide perovskite solar cells have achieved power conversion efficiencies in excess of 25%. Despite this, poor material stability along with lead contamination remains a significant barrier to commercialization. Recently, low-dimensional perovskites, where at least one of the structural dimensions is measured on the nanoscale, have demonstrated significantly higher stabilities, and although their power conversion efficiencies are slightly lower, these materials also open up the possibility of quantum-confinement effects such as carrier multiplication. Furthermore, both bulk perovskites and low-dimensional perovskites have been demonstrated to form hybrids with silicon nanocrystals, where numerous device architectures can be exploited to improve efficiency. In this review, we provide an overview of perovskite solar cells, and report the current progress in nanoscale perovskites, such as low-dimensional perovskites, perovskite quantum dots, and perovskite-nanocrystal hybrid solar cells.

**Keywords:** solar cells; perovskites; perovskite nanocrystals; perovskite quantum dots; low-dimensional perovskites; nanocrystal solar cells; organic–inorganic hybrid solar cells; lead halide solar cells; hybrid solar cells

#### **1. Introduction**

In the search of high-efficiency, low-cost solar cells, a multitude of new materials and architectures are currently being explored. Over the past decade, organometal halide perovskites (OHPs) have emerged as a highly promising photovoltaic material and have been demonstrated as the active layer in perovskite solar cells (PSCs) with efficiencies over 25% for laboratory-based devices (~0.1 cm2) [1] and around 10–15% in modules [2] and are recently being employed in high-efficiency tandem devices [3]. The performance of PSCs has seen a meteoric rise over the past decade and they are already comparable with or superior to well-established photovoltaic technologies [1]. OHPs are attractive particularly due to their ease of processing [4], large absorption coefficients [5], long carrier diffusion lengths [6], low exciton binding energies [7], and low non-radiative recombination rates [8]. These properties also make OHPs an attractive material for various other optoelectronic devices, such as light emitting diodes [9], lasers [10,11], and photodetectors [12].

OHPs have a perovskite crystal structure with the general stoichiometry ABX3 as shown in Figure 1. The A-site is occupied by a monovalent cation e.g., methylammonium (MA, CH3NH3 +), formamidinium (FA, CH3(NH2)2 <sup>+</sup>), Cs<sup>+</sup> etc. The B-site is usually occupied by a Pb2<sup>+</sup> divalent metal cation and can be substituted by a similarly-sized divalent cation such as Sn2+. The X-site is usually occupied by a halide anion e.g., I−, Cl−, Br−. OHPs with mixed cations and/or anions are now the standard for high efficiency cells, particularly due to improved structural stability [13–15]. Their high compositional tunability, whereby the bandgap can be easily modified through ion substitution [16] and low-cost facile deposition procedures [17] makes OHPs excellent candidates for tandem solar cells, where two materials of different bandgaps are employed in conjunction to absorb different parts of the solar spectrum. OHPs can be employed either as the top cell in a tandem device (with e.g., silicon, cadmium telluride, copper indium gallium diselenide etc. bottom cell) or in a stacked perovskite–perovskite tandem device. The successful fabrication of tandem cells with OHPs has the potential to achieve efficiencies in excess of 40% [3].

**Figure 1.** Cubic perovskite unit cell.

While OHPs have demonstrated remarkable efficiencies in laboratory solar cells, there remains significant challenges regarding long-term suitability and feasibility of commercialization [18]. OHPs are extremely susceptible to moisture-induced degradation, and therefore devices must be fabricated in controlled nitrogen atmospheres to avoid trapped moisture in the active layer. Furthermore, devices must be sufficiently encapsulated to prevent external moisture ingress, and the fragility of OHPs along with weak inter-layer adhesion may demand rigid glass substrates to avoid delamination or fractures in the OHP. Even so, heat and light cycling can still induce degradation in encapsulated devices due to thermal mismatch [19]. The use of encapsulants, which can be expensive, along with rigid glass supports, makes OHPs less attractive due to increased costs [3]. It is therefore highly desirable to develop perovskite materials which are stable and tolerant to moisture and other environmental stresses.

Forming nanostructured OHPs (also referred to as low-dimensional OHPs) can be a potential route towards increasing the stability. So far, various types of low-dimensional OHPs have been demonstrated in solar cells, and typically show far superior stability to bulk OHPs [20–22]. This is achieved particularly due to higher formation energies of the low-dimensional perovskite structure and the possibility of encapsulating low-dimensional OHPs in long-chain polymers, essentially providing a protective barrier to moisture [22]. However, carrier transport tends to be restricted in nanostructured perovskites due to the presence of potential barriers within the nanostructured OHP, while quantum confinement also tends to widen the bandgap towards values typically in excess of 2 eV. This therefore comes at a cost to the performance, with the best nanostructured OHPs performing between 10–18% [20–24].

Considering the recent advances in nanostructured perovskites, here we will provide an insight into the important developments and progress in photovoltaics. First, an introduction to the use of bulk OHPs in solar cells will be provided while discussing the challenges and issues facing these materials in order to provide a context for the recent direction towards nanostructured perovskites. This review will then provide a perspective into nanostructured perovskite solar cells as a possible route towards overcoming the issues pertaining to bulk OHPs. Furthermore, hybrid devices formed with OHPs and nanocrystals (NCs) will be discussed, along with high-stability metal oxide perovskite

nanocrystals. We hope this will provide the reader with a basis for understanding the current status of PSCs and the potential opportunities of stable, low-dimensional perovskites.

#### **2. Overview of Bulk Perovskite Solar Cells**

PSCs were initially inspired by the dye-sensitized solar cell (DSSC), where simply replacing the dye in a DSSC with an OHP immediately yielded efficiencies of ~3% [25]. The OHPs used were either MAPbI3 or MAPbBr3, where MA is the small organic cation methylammonium (CH3NH3 <sup>+</sup>). Since the liquid electrolyte, which is used in DSSCs as a redox mediator, dissolved the OHP, these devices had very short lifetimes on the order of seconds. The rapid dissolution of the OHP was overcome by replacing the liquid electrolyte with a polymer which did not dissolve the OHP. Subsequently, devices were reported using the polymer spiro-MeOTAD for hole transport, quickly achieving efficiencies of ~10% with improved device lifetime [26,27]. It was demonstrated that electron and hole transport occurs in the OHP, indicating that free-carriers are generated in the OHP with long diffusion lengths and lifetimes, contrary to suspicion that photocarriers would be excitonic as for organic solar cells, and therefore the sensitized architecture was in fact not necessary [26].

The main PSC device architectures are shown in Figure 2. The OHP is sandwiched between two selective contacts, an electron transport layer (ETL) such as TiO2, and a hole transport layer (HTL) such as spiro-OMeTAD. Metallic contacts are formed on either side of the transport layers: a window contact is formed using a transparent conducting oxide (TCO) such as indium-doped tin oxide (ITO), and a back contact is formed using either gold, silver, aluminum etc. The first architecture employed in the research timeline was the sensitized architecture using a thick mesoporous layer of TiO2 (Figure 2a). This was quickly replaced with bi-layer devices, where the mesoporous-TiO2 was reduced in thickness and a thicker OHP layer was deposited to allow for greater absorption of light and longer crystalline order with larger grain sizes (Figure 2b). A planar device architecture can also be used, with either n-i-p configuration (Figure 2c) or p-i-n configuration (Figure 2d). The planar device eliminates the necessity for the mesoporous TiO2 layer, further reducing fabrication costs and complexity. Planar devices show greater potential for low-cost roll-to-roll printing of PSCs at low temperatures due to the elimination of mesoporous-TiO2 which must typically be annealed at high temperatures during device fabrication (~500 ◦C) for high-efficiency PCSs, and is therefore unattractive for large-scale production while also eliminating the possibility of fabricating devices on flexible plastic substrates. Furthermore, the high-temperature annealing of TiO2 is not suitable for the fabrication of tandem devices with silicon or perovskite bottom cells since such high-temperature annealing process will damage the silicon bottom cell [3]. Planar devices using an SnO2 electron transport layer can be fabricated via low-temperature methods and demonstrate superior stability to mesoporous-TiO2 devices, however the best efficiency of 21.6% is somewhat lower than mesoporous-TiO2 devices (25.2%) [1,28]. Since PSCs employing mesoporous-TiO2 transport layers have shown greater efficiencies than planar devices thus far [29], ideally low-temperature fabrication techniques should be developed for mesoporous-TiO2 transport layers to enable their incorporation into tandem devices.

**Figure 2.** Various device architectures for organometal trihalide perovskite solar cells. (**a**) Mesoporous sensitized, (**b**) bi-layer, (**c**) n-i-p planar and (**d**) p-i-n planar. ETL, HTL, and TCO stand for electron transport layer, hole transport layer, and transparent conducting oxide, respectively.

#### *2.1. Stability of Perovskite Solar Cells*

While exceptional efficiencies have been demonstrated with Pb-based perovskites [13–15], significant challenges exist such as poor stability, toxicity, and rate-dependent current-voltage hysteresis. Stability is an important consideration when assessing commercialization viability of new materials given that silicon solar cells can easily operate for >25 years, even when exposed to a broad range of temperatures and intense solar irradiance. OHPs tend to degrade rapidly in open air conditions and must be fabricated in controlled atmospheres to avoid moisture contamination. The rapid degradation of MAPbI3 in open-air conditions is shown in Figure 3, where the majority of the MAPbI3 layer degraded to PbI2 within 13 days [30]. Although the exact mechanism of degradation remains unclear; it is generally understood that an intermediate phase is first formed via hydration of the OHP [31,32]. Considering the decomposition of MAPbI3, the hydration of MAPbI3 leads to its conversion to MA4PbI6·2H2O and PbI2, followed by phase separation and the subsequent loss of MA, with the final products being CH3NH3I, PbI2, and H2O [31]. The degradation has been shown first to occur at the grain boundaries and is assisted by the presence of trapped charges which usually exist at defect sites, surfaces, and grain boundaries [33]. Ions can easily migrate within OHPs, causing charge accumulation, phase segregation, lattice distortions, and strain in the perovskite structure [34–38]. The degradation of OHPs is enhanced under illumination, and degradation can be accelerated even under moderate temperatures of ~60 ◦C [39,40]. Furthermore, I2, which is generated within the OHP due to exposure to moisture, can easily migrate and leads to the self-sustaining and irreversible degradation of the OHP [41]. The degradation of OHPs leads to the release of the gaseous products CH3NH2, HX, CH3X, and NH3 (where X is a halide), and the release of these gases can be observed at temperatures below 70 ◦C [42].

**Figure 3.** Degradation of MAPbI3. (**a**) Photographs of MAPbI3 degradation and (**b**) corresponding X-ray diffraction (XRD) spectra of the same samples after 1, 13, and 26 days stored in ambient conditions. The starred peaks in the XRD spectra correspond to PbI2. Reproduced from ref. [30], with permission from John Wiley and Sons, 2016.

Due to the high susceptibility of OHPs to degrade when exposed to moisture, it is therefore necessary to carefully control the atmosphere during fabrication. Entire device encapsulation is necessary to prevent exposure to moisture and mechanical fractures. For encapsulated devices, the formation of bubbles has been observed in the encapsulant layer due to the release of gaseous species. Encapsulation prevents gaseous products from escaping, creating a thermodynamically enclosed system which is expected to reduce the rate of degradation [42]. Encapsulation is therefore essential for several reasons: to prevent the ingress of moisture; to prevent the release of gases; and to prevent the release of toxic materials to the environment. However, due to the thermal expansion coefficient mismatch between the various layers, including the encapsulant, temperature cycling of the PSC (i.e., day and night temperature variations) can lead to significant delamination and device failure. Careful selection of the encapsulant and various device layers is therefore necessary to minimize delamination caused by temperature cycling. This eliminates the possibility of flexible, low-weight

modules, and the low stability and Pb-contamination necessitates careful recycling of PSCs. In spite of these measures, the question of whether the lifetime of OHPs can match silicon PV remains dubious.

#### *2.2. Toxicity of Perovskite Solar Cells*

Pb-containing OHPs' decomposition results in the formation of Pb-halide compounds, metallic Pb, and various carbonated molecules [43]. Although PSCs contain small amounts of Pb (~0.4 g/m2 for a 400 μm-thick OHP layer) [44], the harmful Pb-halides generated via degradation are highly water-soluble and therefore pose a significant risk to the environment [45]. The contamination of Pb can be addressed either by replacing Pb with other non-toxic elements or by stabilizing the structure of the perovskite so as to avoid the formation of PbI2. Unfortunately, computational studies have suggested that there is no viable alternative to Pb in PSCs to achieve the similarly high efficiencies which are in excess of 20% [46]. The high efficiencies of OHPs is attributed to the favorable Pb2<sup>+</sup> orbital hybridization with I- and Br- halide ions which results in high absorption coefficients and long carrier diffusion lengths [47]. Sn is a potential alternative to Pb, and whilst still toxic to animals and humans, it is less harmful than Pb. [43] Sn-OHPs have been produced by the direct replacement of Pb with Sn, but the best efficiency achieved to date is 7.14% [23]. In addition, the stability of Sn-based devices is usually worse than Pb-OHPs due to the tendency of tin to easily oxidize from Sn2<sup>+</sup> to Sn4<sup>+</sup>. This can be mitigated to some extent by the addition of SnF2 and ethylenediammonium during fabrication to inhibit the formation of Sn4<sup>+</sup> [23,48]. While pure Sn-OHPs are unstable, the oxidation of Sn2<sup>+</sup> becomes less energetically favorable when less than 50% of the B-site in the perovskite structure is occupied by Sn2<sup>+</sup> (i.e., MAPb≥0.5Sn≤0.5I3) and the stability is significantly improved [49]. Notably, Zn, which is a 2<sup>+</sup> ion with a slightly smaller ionic radius than Pb, has also been investigated for the partial replacement of Pb and has demonstrated an improvement in the power conversion efficiency (PCE) for small amounts of Zn (~1% to 5%). The introduction of Zn into MAPbI3 leads to the formation of larger grains which are more homogeneous, and layers which are more compact and with fewer pinholes. This is achieved through a lattice contraction induced by the smaller Zn ion, along with stronger coordination with the organic cation, leading to a reduction in the amount of point defects [50–53]. However, this work only serves to reduce Pb contamination without eliminating it entirely, and the contamination of toxic Pb and Sn remains and degradation is still observed [49].

#### *2.3. Hysteresis in PSCs*

A common issue exhibited by nearly all PSCs is a hysteresis present during solar cell characterization. Hysteresis, defined as the dependence of the state of a system on its history, is frequently observed during current density-voltage (J-V) measurements, where a change in the voltage scan direction between forward and backward results in a differing J-V response, as shown in Figure 4a. A device without J-V hysteresis is shown in Figure 4b. The observed hysteresis is largely attributed to ion mobility within the OHP [54–56], whilst other mechanisms have also been proposed, see reference [57]. Hysteresis is problematic as it primarily introduces difficulties in accurately measuring device performance, but can also be indicative of stability issues [41,58]. Recent work [13,15] has shown that high-efficiency mesoscopic devices possess low hysteresis in the forward and backward J-V scans with the same scan rates from 10 mV/s to 50 mV/s; however, hysteresis is still well observed particularly for fast scans [56,59,60]. Selecting appropriate contacts and forming high-quality OHP layers appears to negate most of the hysteresis observed during standard performance measurements with slow scan speeds; however, the J-V character for fast scans is often unreported and ionic motion and charge accumulation are still likely to be present in the perovskite layer. Furthermore, hysteresis is often intensified as devices are scaled to active areas over 1 cm2, particularly due to issues with controlling morphology when depositing OHPs over larger areas [61]. The hysteresis observed in OHPs depends on various measurement conditions during the J-V characterization, in particular: the voltage scan rate and scan range [56,62]; the delay time between applying the bias voltage and measuring the current [63]; and the poling voltage prior to measurement [57]. Hysteresis has also

been shown to vary with the grain size of the perovskite [57,64], the A-site cation [65], and device architecture [62,63].

**Figure 4.** (**a,b**) Current density-voltage curves with forward (R-F) and reverse (F-R) voltage scan direction for a device with hysteresis (**a**) and without (**b**). Reproduced from ref. [66], with permission from The Royal Society of Chemistry, 2017. (**c**,**d**) Time-dependent photocurrent response under reverse and forward stepped scans with (b) 1 s step time and (c) 0.1 s step time. Reproduced from ref. [67], with permission from American Chemical Society, 2015. (**e**,**f**) Current decay after removing device from illumination showing two discharging events occurring over different timescales. Reproduced from ref. [68], with permission from American Chemical Society, 2015.

The hysteresis is well-described by Figure 4c,d whereby the voltage is scanned forward and backward in a stepwise fashion with different delay times between the steps: 1 s in Figure 4c and 0.1 s in Figure 4d [67]. It is clear that at least two processes are involved: one is an ultrafast process which leads to an almost instantaneous (microsecond) change in photocurrent, followed by a slower response on the timescale of milliseconds to seconds. There is a large difference in the forward and reverse J-V scans observed for a 0.1 s voltage step time: this arises because when the step speed is too fast, the photocurrent is not able to stabilize and there is a remnant charge stored in the device. This was further investigated and it was shown that there are at least two ways in which charge is stored in OHPs (Figure 4e,f) [68]. After removing an OHP device from illumination, the photogenerated current decayed from 180 mA/cm<sup>2</sup> to less than 50 μA/cm2 within 50 μs (Figure 4e). This was followed by a second, longer decay event which occurred over the next ~3 s (Figure 4f). Although the peak current in the second decay event (~50 μA/cm2) accounted for less than 1% of the initial photocurrent (~180 mA/cm2), the lifetime of the second current was far longer and therefore the total charge associated with this slower decay was calculated to be ~50 times larger than the charge associated with the initial microsecond-discharge event. Therefore, at least two types of capacitive electronic charges were confirmed in OHPs: the first one is small (~0.2 μC cm−2) and likely due to charge trapping; and the second one is much larger (~40 μC cm<sup>−</sup>2), which could be the result of mobile ions or dipole

realignment [68]. Furthermore, it is also known that large differences in the carrier mobility of the electron and hole transport layers can lead to charge accumulation resulting in hysteresis [68].

Understanding the origin and mechanism of hysteresis could lead to the improvement of the performance and stability of PSCs. The main mechanisms which have been proposed to contribute to the effect are: ion migration [56,67,69], charge trapping and accumulation [70,71], and polarization of dipoles [57,62,72]. These mechanisms are represented in Figure 5 and are described briefly in order of the legend:


**Figure 5.** Schematic of the proposed contributions to hysteresis. ITO, FTO, ETL and HTL stand for indium-doped tin oxide, fluorine-doped tin oxide, electron transport layer, and hole transport layer, respectively. Reproduced from ref. [76], with permission from Elsevier, 2016.

These processes may occur simultaneously, and each process will have a different impact on the hysteresis depending on various parameters such as the device structure, interfacial quality, and the properties of the perovskite layer (grain size, defect density, composition etc.), amongst others.

#### **3. Nanostructured Perovskite Absorbers**

#### *3.1. Introduction*

Non-toxic and/or stable materials with similar properties to bulk Pb-OHPs are a high priority and are currently being explored, such as the replacement of Pb with Sn or Bi [23,77], lead-free halide double perovskites [78], and low-dimensional materials [22]. The efficiencies of these solar cells are often far lower than bulk Pb-OHPs and a large amount of development is still required. Nanostructured perovskites include perovskite quantum dots, nanoparticles, nanosheets, nanorods, and perovskites with nanoscale internal ordering. These materials are often termed low-dimensional

perovskites (LDPs) and can generally be envisioned by reducing the bulk perovskite structure to the nanoscale in at least one structural dimension.

Figure 6 shows schematically how a bulk perovskite with ABX3 structure transforms from a three-dimensional perovskite (3DP) to an LDP. In 3DPs, i.e., the typical bulk perovskites used in record-efficiency devices, each BX6 <sup>4</sup><sup>−</sup> octahedra is connected along all three axes and is anisotropic. It is rather important that this octahedral structure is mostly preserved since the orbital hybridization of B and X sites is responsible for many of the favorable optoelectronic properties of OHPs. For two-dimensional perovskites (2DPs), e.g., nanoplatelets and nanosheets, the BX6 <sup>4</sup><sup>−</sup> octahedra is connected along two axes and consists of 2D slabs of octahedra with the organic cation occupying the A-site in the voids between slabs. Surrounding the nanosheets are organic 'barrier' molecules which prevent the sheets from crystallizing into a larger 3D structure whilst also providing encapsulation and protection against degradation. For one-dimensional perovskites (1DPs), e.g., nanowires and nanorods, the BX6 <sup>4</sup><sup>−</sup> octahedral network extends along only one axis and is encapsulated with organic barrier molecules. For 1DPs and 2DPs, various organic barriers can be selected, and a wide range of choices exist. Hydrophobic organic barriers can be selected which protect the structure against moisture. For zero-dimensional perovskites (0DPs), the BX6 <sup>4</sup><sup>−</sup> octahedra is disconnected in all directions and consists of isolated octahedral clusters stabilized by a cationic sublattice. A distinction is often made between 0DPs and quantum dots (QDs), where for a perovskite QD (PQD), the BX6 <sup>4</sup><sup>−</sup> octahedra remains connected in all three axes and the radius of the particle is below the Bohr exciton radius, whereas for a 0DP each octahedra is completely disconnected from adjacent octahedra, as shown in Figure 6.

**Figure 6.** Overview of the different perovskite dimensionalities. Reproduced from ref. [79], with permission from John Wiley and Sons, 2015.

Low-dimensional materials can also be produced which are not strictly perovskites yet follow a similar set of design rules; being based on a large heavy metal ion bonded ionically with halide ions, and stabilized by a sublattice of 1+ cations: For example, B-site 3+ cations such as Bi3<sup>+</sup> form B2X9 3− bioctahedra instead of a BX6 <sup>4</sup><sup>−</sup> octahedra for 2+ cations, forming the 0DP material (CH3NH3)3Bi2I9. These materials, which can be produced very similarly to standard perovskites (i.e., from solution) whilst also possessing similar properties, are discussed later. The perovskite term is used loosely to describe these materials, as in some cases the perovskite structure is disturbed.

LDPs exhibit quantum confinement effects which are particularly noticeable through a widening of the bandgap [22]. Although 3DPs already have a bandgap close to the optimum value of ~1.4 eV for a single junction solar cell, a wider bandgap is advantageous for forming tandem devices or for indoor photovoltaics [80]. Furthermore, quantum confinement effects introduce the possibility to reduce losses via carrier multiplication which has already been demonstrated in CsPbI3 quantum dots [81] and in the 0DP material (CH3NH3)3Bi2I9 [82]. The effective use of carrier multiplication in a single-junction solar cell can potentially increase efficiency to ~44% [83], far beyond the Shockley-Queisser (SQ) efficiency limit for a single junction cell of ~33% [84]. In addition, both 3DPs and LDPs are capable of incorporating a low concentration of inorganic nanocrystals into their lattice to form internal energy band alignments which can be used to increase carrier collection and absorption. These hybrid devices can potentially harvest a wide range of the solar spectrum through quantum confinement effects without significantly altering the device architecture, and will be discussed later [85,86].

LDPs often exhibit excitonic behavior as carriers become localized. Since LDPs are often stabilized with organic barriers or a cationic sub-lattice which behaves as an insulating spacer layer, this results in a potential barrier surrounding the individual sheets, rods, or clusters. Carriers therefore become localized on the sheets, rods, or clusters, which often inhibits carrier extraction. The strength of the exciton binding energy is strongly dependent on the dimensionality, with 0DPs usually exhibiting the highest exciton binding energies [87,88].

#### *3.2. One- and Two-Dimensional Perovskites*

Along with PQDs, perovskite nanosheets and nanorods are the most successful types of LDPs demonstrating the highest efficiencies in photovoltaic devices. The main advantage of reduced dimensionality is that the OHP can be encapsulated with a more stable long chain organic molecule which reduces the rate of degradation. In reference [22] it was shown via simulations that the stability of MAPbI3 perovskites can be improved by producing a 2D perovskite encapsulated by larger cations. Further to the benefit of the protective ligands, the 2D perovskite structure has a higher formation energy, which therefore yields a more stable perovskite material. A single 2D slab of the perovskite structure, i.e., a monolayer, encapsulated with organic barrier, is termed *n* = 1, as shown in Figure 7. The bandgap is strongly dependent on the number of perovskite slabs (*n*); as *n* increases, the bandgap narrows and the strength of quantum confinement reduces, and the dimensionality tends towards a quasi-2D structure (*n* > ~10), while for very large values of *n* the perovskite tends towards a 3D structure.

**Figure 7.** Reducing the dimensionality of organometal halide perovskites leads to higher stability, but lower device performance. Reproduced from ref. [22], with permission from American Chemical Society, 2016

There are a very large number of organic molecules which can potentially be used as the barrier layer, however thus far only a limited number of molecules have been investigated, e.g.: phenylethyl ammonium (C8H9NH3, PEA) [22], benzyl ammonium (C6H5CH2NH3, BA) [89], 2-iodoethylammonium (IC2H4NH3) [90], polyethylenimine ((C2H5N)n, PEI) [91], 2-thiophenemethylammonium (C5H7NS, ThMA) [92], and 3-bromobenzylammonium iodide (BrC6H4CH2NH2.HI, 3BBA) [24]. The absorption spectra of 2DPs is weakly associated with the selection of the barrier molecule; optical properties are far more dependent on the *n* value [93]. As *n* tends towards lower values, the stability of the 2DP increases [22], yet the device performance tends to decrease dramatically due to the widening of the

band gap and the higher proportion of insulating barrier molecules which have a detrimental effect on carrier transport. Whilst the high in-plane mobility of bulk OHPs is retained along the nanosheets and nanorods, the transport between nanosheets/rods is restricted due to the potential barrier created by the insulating organic barriers which reduces the overall carrier mobility [88]. However, this can be mitigated somewhat by using shorter barrier molecules [94].

In reference [22], the MAPbI3 perovskite was reduced to a 2DP and a quasi-2DP structure using PEA barriers with varying *n* values. A quasi-2DP with *n* = 40 was capable of achieving ~15% efficiency, however the stability of quasi-2DPs is still rather poor. Reducing the *n* value to 6 provided high stability, yet the efficiency fell towards ~5%. It is likely that the low efficiency was due to the disordered nature of the sheets which are not aligned perpendicular to the contacts, inhibiting charge transfer. This is shown schematically in Figure 8a. When nanosheets are oriented horizontally, i.e., parallel to the contacts, the charge carrier transfer is restricted in the vertical direction, and charge carrier extraction in inhibited because the long organic barriers separating the LDP sheets inhibit transfer between the layers.

**Figure 8.** Solar cells based on a perovskite absorber with a two-dimensional network. (**a**) Sheets align parallel with the contacts resulting in low carrier mobility between the contacts and (**b**) sheets align perpendicular to the contacts resulting in favorable out-of-plane mobility between contacts.

Higher efficiencies can be achieved by vertically orientating the inorganic sheets, as shown schematically in Figure 8b, whereby charge transport is less restricted. If the nanosheets/rods are orientated vertically, i.e., perpendicular to the contacts (out-of-plane), charge transport is predominantly along the perovskite structure and carrier extraction is therefore far more efficient since carriers must overcome fewer potential barriers. This was initially demonstrated in BA-capped 2DPs with *n* = 3 and the efficiency was increased to over 12% using a hot casting deposition technique to achieve out-of-plane alignment of the 2D sheets [23]. However, these devices still showed rather poor stability when exposed to 65% relative humidity without encapsulation, while fully encapsulated devices demonstrated impressive stability. This has also been demonstrated in perovskite nanorods, with an increase in efficiency from 1.74% to over 15% following out-of-plane alignment [92]. This was achieved by using a methylammonium chloride (MACl) assisted film formation technique which resulted in vertically aligned perovskite nanorods, demonstrating far improved stability over 3D perovskite. Disordered (unaligned) 2DPs usually show significant hysteresis [95], which is likely due to a bias-voltage induced charging effect caused by the insulating organic molecules and poor charge transport when the 2DP sheets are not vertically aligned. However, the hysteresis is mostly eliminated when the nanosheets are aligned out-of-plane with respect to the contacts since charge transport is less restricted [23].

A problem which must be overcome in 2DPs is a stacking misalignment of the 2DP grains which reduces carrier mobility. It was shown that even when 2DPs are aligned with favorable out-of-plane alignment, stacking misalignments between grains restricts charge transfer between vertically aligned sheets [83]. In order to improve device performance, it is important to minimize stacking misalignment between grains. In addition, it was recently demonstrated that it is essential to use LDPs with at least *n* > 2, as it has been shown that exciton dissociation occurs within the nanosheets of 2DPs due to the presence of lower energy states at the edges of the nanosheets which exit only for nanosheets with *n* > 2 [96]. While these edge states are present for 2DPs with *n* > 2, for n ≤ 2, edge-state exciton dissociation was not observed, and the device performance was significantly lower. These lower energy states exist at the edge of 2DPs and provide a favorable energy pathway for excitons to dissociate into free-carriers with longer lifetimes, which was demonstrated to significantly improve device performance. This work demonstrated that it is imperative to synthesize 2DPs with at least *n* = 3 in order to benefit from the favorable exciton dissociation mechanism, even though thinner nanosheets (*n* ≤ 2) can provide higher stability.

Recent work demonstrated that mixed *n* value 2D perovskites can achieve both favorable carrier transport and band alignment introduced via a unique nanostructuring of the 2D perovskite film, achieving a PCE of 18.2% [24]. The introduction of the barrier molecule 3-bromobenzylammonium iodide (3BBA) leads to the oriented growth of small *n* value 2D perovskites perpendicular to the substrate (*n* ≈ 1–4), followed by the crystallization of large *n* value quasi-2D perovskites in the bulk of the film, shown schematically in Figure 9a and the overall device structure in Figure 9b. This structure also introduces a favorable band alignment as shown in Figure 9c whereby the larger bandgap of the mixed low *n* value 2DPs provides a potential energy gradient driving carriers to the desirable extraction contacts. This demonstrates the remarkable tunability that can be achieved through nanostructuring perovskites to achieve favorable energy band alignment. The devices also showed impressive stability: Unencapsulated devices stored in a dark oven between measurements under ≈40% relative humidity retained 80% of the original PCE after 2400 h. The device could also be submerged underwater for 60 s without any immediate negative effect on the efficiency. It was stipulated that the hydrophobicity due to the presence of iodine in 3BBA results in the enhanced moisture durability of these 2DPs.In general, 2DPs have not been optimized yet via cation engineering to the same extent as 3DPs, which has led to the high performance and improved stability of 3DPs today [13]. Recently, 5% Cs<sup>+</sup> doping in a 2DP demonstrated an efficiency increase from 12.3% to 13.7%, which was attributed to improved crystal quality and low trap defects, increased grain size, and improved carrier transport [97]. Since most 2DPs with low *n* values show wide bandgaps, it is important to engineer 2DPs which absorb in the visible spectrum. Material engineering and optimization as such demonstrates that there is still great potential for work on improving the 2DPs' material properties.

Finally, 2DPs may also find use in improving the stability of 3DPs by acting as a protective capping layer. A 2DP was demonstrated as the capping layer in a 3DP solar cell and displayed over 19% efficiency, along with improved stability over the 3DP alone [98]. Further work in this area showed that the deposition of a hydrophobic 2D perovskite on top of a 3D perovskite not only protects against moisture, but also improves carrier extraction. The formation of the 2D perovskite on the surface of the 3D perovskite consumes detrimental and undesirable non-perovskite phases present at the surface of the 3D perovskite and resulted in faster injection of holes into the HTL [99]. More recently, an ammonium salt post-treatment of a 3D OHP film increased the PCE from 20.5% to 22.3% via the formation of a 1DP passivation layer [100]. Devices retained 95% of the initial PCE after continuous illumination for 550 h. This area of work presents a route towards avoiding the necessity for encapsulants in PSCs, therefore reducing costs and avoiding issues pertaining to thermal expansion mismatch.

#### *3.3. Zero-Dimensional Perovskites*

Pb-based 0DPs have been previously studied but so far seem unsuitable for photovoltaics [101,102]. For example, when the typical perovskite MAPbI3 is transformed into a 0DP with the chemical formula (CH3NH3)4PbI6, the structure is extremely unstable [101]. Alternatively, more stable inorganic Pb-based 0DPs can be produced such as Cs4PbBr6, however, the bandgap is very large: Pb- based 0DPs tend to

have very large bandgaps which are unsuitable for photovoltaics, typically in the UV-range, irrespective of the halide anion selected [102].

**Figure 9.** 2D perovskite solar cells using 3-bromobenzylammonium iodide barrier molecule. (**a**) Schematic of the device nanostructuring, (**b**) schematic of the device architecture, (**c**) energy band alignment relative to the vacuum level in eV, (**d**) current density-voltage measurement, (**e**) incident photon conversion efficiency (ICPE), (**f**) histogram showing reproducibility of the power conversion efficiency (PCE) and (**g**) solar cell stability for devices stored in the dark between measurements under ≈40 relative humidity. Reproduced with modifications for clarity from ref. [24], with permission from John Wiley and Sons, 2018.

Alternatively, Bi-based 0DPs have bandgaps closer to 2 eV and have been demonstrated as the absorber in photovoltaic cells [86,103–105]. Bi, which is adjacent to Pb in the periodic table, has a similar atomic radius to Pb yet with one additional valence electron yielding 3+ instead of 2+, resulting in a B2X9 <sup>3</sup><sup>−</sup> bioctahedral structure rather than the BX6 <sup>4</sup><sup>−</sup> octahedral structure. These perovskite structures have the formula A3B2X9, but can also be expressed as AB2/3X3, i.e., a metal-deficient perovskite. Figure 10 shows the structure of a 0DP with the chemical formula (CH3NH3)3Bi2I9, where Bi2I9 3− clusters are separated by a CH3NH3 <sup>+</sup> cationic lattice. Here, CH3NH3 <sup>+</sup> can be replaced with a range of organic and inorganic cations. Whilst these materials are often referred to as 'perovskites', their crystallographic structure is slightly different to the perovskite structure, whereby the BX6 4- octahedra is instead replaced with a B2X9 <sup>3</sup><sup>−</sup> bioctahedra.

Bi-0DPs have been studied and the best devices have achieved efficiencies of 1.64% [105]. These materials, with bandgaps of ~2 eV, generally exhibit high exciton binding energies (~300 meV) and high effective masses for carriers [88]. Because of the excitonic nature of these materials with quantum confinement effects, 0DPs have been shown to exhibit carrier multiplication [82]. However, due to the high exciton binding energy, the rates of electron-hole recombination is high which limits device performance. 0DPs also exhibit anisotropic carrier mobilities if the cluster is non-symmetrical and/or if the spacing between clusters varies between planes [88]. It is therefore necessary to try to overcome the high exciton binding energy and carrier transport issues by a range of possible methods,

such as modifying the cationic sub-lattice, using semiconducting polymers which enhance carrier mobility between the clusters, or by forming hybrids with inorganic nanocrystals which assist in exciton dissociation.

**Figure 10.** Schematic of the structure of (CH3NH3)3Bi2I9 which forms a zero-dimensional network. Bi2I9 3- clusters are stabilized within a (CH3NH3) <sup>+</sup> ionic lattice. Reproduced from ref. [82], with permission from Springer Nature, 2017.

(CH3NH3)3Bi2I9 solar cells can be processed and stored entirely in ambient conditions and have demonstrated far superior stability to 3DPs, likely due to the formation of a native surface layer of Bi2O3/BiOI which provides self-encapsulation of the perovskite [30]. This layer does not inhibit carrier extraction, and is also likely responsible for the negligible hysteresis observed in these devices [86]. If 0DPs were employed as the wide-bandgap top cell in a tandem solar cell, their high stability can provide encapsulation for the less-stable OHP bottom cell to prevent moisture ingress. Furthermore, the absorption can be modified by incorporating optically active organic molecules or forming hybrids with nanocrystals with suitable band alignment [86], and the large bandgap of 2 eV can be reduced to values as low as 1.45 eV through doping and/or changing the A-site cation [106–108].

Sb-based 0DPs have also been demonstrated with the formula (CH3NH3)3Sb2I9 and have so far achieved higher efficiencies than Bi-0DPs, with the best devices so far achieving 2.77% efficiency [109]. The higher efficiencies of these devices is likely due to the intrinsically lower exciton binding energy of Sb-0DPs [110]. Since the bandgap of Sb-0DPs is still quite large (~1.9 eV), researchers have attempted to lower the bandgap through Sn-doping, and successfully reduced the bandgap to 1.53 eV with 40% replacement of Sb with Sn to form (CH3NH3)3Sb0.6Sn0.4I9. Doping with Sn increased the efficiency of the devices from 0.57% (without Sn, bandgap = 2.0 eV) to 2.7% (40% Sn, bandgap = 1.53 eV). Since the starting efficiency of the undoped Sb-0DP reference device was quite low (0.57%) compared to the highest reported in the literature (~2.77%), it is likely that through device optimization of the Sn-doped Sb-0DP will quickly lead to higher efficiencies in the near future, likely exceeding 5%. These Sn-doped Sb-0DPs demonstrated impressive stability with no change in the XRD spectra after 15 days of exposure to ambient conditions. Although inorganic 0DPs have also been produced with the formula Cs3Sb2I9 and Cs3Bi2I9, these devices tend to show very low efficiencies below 0.1% [111,112], likely due to their large bandgaps and high exciton binding energy, and have therefore not been pursued to the same extent.

#### *3.4. Perovskite Quantum Dot Solar Cells*

High exciton binding energies and inefficient charge transfer are significant issues associated with LDPs which limit carrier extraction, therefore inhibiting device performance. This can potentially be overcome in PQDs through close-packing with electronic coupling between QDs. Colloidal PQDs can be readily synthesized from solution using organic capping molecules, such as oleic acid, oleylamine, octadecene, etc. which prevent the perovskite from forming into a larger crystal [113]. These long chain

molecules must be removed during device fabrication for efficient solar cell performance. However, PQDs with organic A-site cations are often highly unstable, and it is therefore not possible to remove these barrier molecules as they are essential for preventing rapid degradation. As discussed previously, the issue of long chain organic barrier molecules in perovskite nanorods and nanosheets can be overcome by aligning the sheets and rods perpendicular to the contacts, minimizing the number of potential barriers that must be overcome by charge carriers. However, due to the spherical shape of QDs, this type of favorable alignment is not possible, and researchers must therefore look towards inorganic PQDs which do not require encapsulation in protective organic barriers or use a different architecture [85,114].

All-inorganic perovskites can be formed by replacing the A-site with an inorganic cation, such as Cs<sup>+</sup>, e.g., CsPbI3. Inorganic PQDs such as CsPbI3 are the most favorable perovskite material since the bandgap of bulk CsPbI3 is the smallest of the inorganic perovskites (1.73 eV for the cubic phase) [115]. However, accessing the desired cubic phase of CsPbI3 is challenging: For bulk CsPbI3, the orthorhombic phase is thermodynamically preferred at room temperature, but the large bandgap of 2.82 eV renders orthorhombic CsPbI3 unsuitable for photovoltaics [115]. The cubic phase exhibits a more favorable bandgap of 1.73 eV; however, this phase is unstable at room temperature. Forming CsPbI3 quantum dots enabled researchers to achieve the cubic phase at room temperature, as the contribution of the surface energy for CsPbI3 quantum dots was shown to retain the favorable cubic perovskite phase [114].

CsPbI3 PQD solar cells were fabricated with 10.77% efficiency [114]. These devices could be fabricated at ambient conditions and showed impressive stability when stored in a desiccator, with no decrease in performance after 60 days. However, when stored in relative humidity of 40–60% there was a significant decrease in the device performance after just 2 days, although QD devices demonstrated improved stability over bulk CsPbI3. Furthermore, CsPbI3 QD devices showed significant hysteresis, likely due to difficulties associated with charge transfer between quantum dots, ion migration, and charge trapping at QD surfaces.

These devices were later improved by a post treatment of the CsPbI3 QDs, and increased the efficiency to 13.43%, as shown in Figure 11 [21]. This was achieved through efficient QD coupling via a post-treatment of the film, allowing improved change transfer between the QDs in the film. The post-treatment involved soaking the CsPbI3 QD thin film in a formamidinium iodide in ethyl acetate solution for 10 s. The post-treatment creates a coating on the CsPbI3 QDs and does not alter their nanocrystalline character. It was confirmed that the post-treatment improved the carrier mobility from 0.23 to 0.50 cm2 V−<sup>1</sup> s<sup>−</sup>1. However, the poor stability of CsPbI3 at ambient conditions has not yet been addressed, and it is likely that these materials will require encapsulation. Alternatively, a Cs- salt post-treatment was reported achieving PCE of 14.1% [116]. The Cs-salt treatment is performed after the removal of ligands from the CsPbI3 QDs. When the ligands are removed, Cs vacancies are left behind on the CsPbI3 QDs. These vacancies are filled by Cs via a Cs-salt post treatment, resulting in improved free carrier mobility, lifetime, and diffusion length, as well as greater stability over untreated CsPbI3 QDs.

One of the advantages of CsPbI3 QDs is the possibility of carrier multiplication, which has already been demonstrated in CsPbI3 QDs with a high carrier multiplication quantum yield of 98% [81]. While the bandgap for quantum confined materials scales as *E*<sup>g</sup> ~ <sup>1</sup> *<sup>r</sup>* where *r* is the radius, the rate of Auger recombination scales as <sup>1</sup> *<sup>r</sup>*<sup>6</sup> and therefore forming smaller QDs is more favorable for carrier multiplication. The average radius of the QDs in this work was 5.75 nm and the exciton Bohr radius for CsPbI3 QDs is 6 nm. The QDs are therefore in the weak quantum confinement regime, yet still exhibited highly efficient carrier multiplication indicating that strong quantum confinement is not necessary in these materials for carrier multiplication [81].

**Figure 11.** CsPbI3 quantum dot solar cells. (**A**) Schematic of the device structure, (**B**) cross-sectional scanning electron microscopy image, (**C**) current density-voltage scans under solar simulated light, (**D**) stabilized current at a constant voltage of 0.95 V, and (**E**) external quantum efficiency. Reproduced from ref. [21], with permission from AAAS, 2017.

The band energy structure of the active layer can be tuned to achieve improved carrier extraction by using PQDs with varying condition band, valence band, and Fermi level positions. The sequential deposition of PQDs with varying band energy positions has been shown to improve carrier extraction [117] and is reproduced in Figure 12. A schematic of the sequential deposition of PQDs is shown in Figure 12a and the band energy positions of the PQDs studied in this work are shown in Figure 12b. PQDs were synthesized in the series CsxFA1-xPbI3 and PQD heterojunction devices were fabricated with the structure ITO/TiO2/PQDs I/PQDs II/spiro-MeOTAD/MoOx/Al. The best device performance was obtained using either Cs0.5FA0.5PbI3 or Cs0.25FA0.75PbI3 as the bottom layer and CsPbI3 on the top. Devices based on a Cs0.25FA0.75PbI3:CsPbI3 heterojunction were investigated further for optimization. Figure 12c shows the SEM cross section of the device and Figure 12d shows the effect of varying the thickness ratio of Cs0.25FA0.75PbI3:CsPbI3 on the EQE spectra. A ratio of 1:3 (Cs0.25FA0.75PbI3:CsPbI3) retained most of the short wavelength EQE contribution from CsPbI3 whilst also red-shifting the EQE onset slightly. Higher proportions of Cs0.25FA0.75PbI3 lead to a fall in EQE at shorter wavelengths, despite red-shifting the EQE onset more. Figure 12e shows that varying the bottom layer composition, i.e., by fabricating devices with the structure ITO/TiO2/CsxFA1-xPbI3/CsPbI3/spiro-MeOTAD/MoOx/Al for x = 0.25, 0.5 and 0.75 leads to a similar red-shift in the EQE as the bandgap of the CsxFA1-xPbI3 PQDs is decreased. The J-V characteristics are shown in Figure 12f, and ratios of 1:3 and 2:2 achieve the highest PCEs, however due to the large hysteresis present in these devices, the SPO was also presented and revealed that devices with a 1:3 ratio of Cs0.25FA0.75PbI3:CsPbI3 achieved the highest SPO at 15.52%. Finally, bulk heterojunction architecture devices were also fabricated by mixing the PQDs. These devices did not exhibit the same enhanced performance confirming that a bi-layer heterojunction of PQDs is essential for achieving improved carrier collection.

A summary has been provided in Table 1 comparing a selection of the most notable results since 2018 for 0D, 1D, 2D and QD perovskites, as well as also including some of the notable heterojunctions formed between 3D perovskites and LDPs. This table also provides a summary of the stability of the solar cell devices, noting the storage conditions and the solar cell J-V measurement type (i.e. continuous or intermittent, where continuous measurements typically involve the device remaining under constant solar simulated light, whilst for intermittent measurements the device is removed from illumination and stored in specified storage conditions between measurements).

**Figure 12.** Perovskite quantum dot (PQD) solar cells with charge separating heterostructure. (**a**) Schematic of the device fabrication via spin coating, (**b**) energy band structure of the various PQDs used in the study, (**c**) cross-sectional scanning electron microscope of a typical device, (**d**) the external quantum efficiency (EQE) of solar cells made with various ratios of Cs0.25Fa0.75PbI3 to CsPbI3 quantum dots, (**e**) EQE at the absorption edge of various quantum dots in the series CsxFA1-xPbI3 as the bottom layer. (**f**) current density-voltage (JV) curves for the devices shown in (**d**) and (**g**) stabilized power output (SPO) of the varying compositions shown in (**f**). Reproduced from ref. [117], with permission from Springer Nature, 2019.


**Table 1.** A selection of notable reports on low-dimensional perovskite solar cells. QDs, PCE, RT, and RH stand for quantum dots, power conversion efficiency, room temperature, relative humidity, respectively.

<sup>1</sup> *n* values are only applicable for 1D and 2D materials.

#### *3.5. Perovskite-Nanocrystal Hybrid Devices*

The formation of hybrid layers and devices through incorporating nanocrystals into the OHP layer have also been explored, both in bulk 3DPs [85] and in 0DPs [86]. The introduction of quantum-confined NCs to bulk 3D OHPs enables the possibility of carrier multiplication. Thus far, SiNCs have been primarily studied in this context, since most nanocrystals studied for organic-inorganic hybrid photovoltaics are toxic Pb- or Cd-based [20], and it would be counterintuitive to add a toxic material to a lead-free perovskite. SiNCs are an environmentally-friendly material which are non-toxic and can be synthesized through a wide variety of methods [124,125]. The properties of SiNCs can also be easily modified by surface engineering and the absorption and emission properties can be influenced by the surface terminations [125–127]. Surface engineering can also improve carrier transport in SiNCs by passivating surface defects [128]. While SiNCs do present their own challenges, they represent an important model NC material.

It was previously demonstrated that the incorporation of silicon nanocrystals (SiNCs) into the 0DP with the formula (CH3NH3)3Bi2I9 led to an enhancement in the device performance [86]. It was proposed that the SiNCs may act as a dissociation pathway for tightly-bound excitons on the nanoclusters of Bi2I9 3- bioctahedra. An electronic junction formed between the perovskite material and the inorganic nanocrystal can provide an energetically favorable pathway for excitons to overcome the potential barrier created by the cationic sublattice, providing exciton dissociation before the carrier recombines. Once the exciton is dissociated it becomes a free-carrier which can be extracted. This is commonly employed in organic–inorganic hybrid solar cells using SiNCs to enhance exciton dissociation [129]. These types of hybrids may present a route towards significantly improving the efficiency of LDPs.

Hybrid MAPbI3-SiNC devices also exhibit improved device performance and stability [85]. X-ray photoelectron spectroscopy (XPS) indicated that MAPbI3 bonds with SiNCs via intermediate oxide bonds with nitrogen in methylammonium (N–O–Si). The oxidation of SiNCs was also observed in XPS and is likely responsible for the improved stability, whereby SiNCs may act as a 'sponge' absorbing oxidizing species in the MAPbI3 layer resulting in slight oxidation of the SiNCs. Furthermore, hybrid devices with SiNCs exhibited improved device performance after light soaking for 8 min (Figure 13), whilst the performance of MAPbI3-only devices decreased. This is commonly observed in MAPbI3 devices and is attributed to light-activated trap states with inhibited photocarrier extraction [130]. The observation of the inverse behavior in hybrid devices suggests that SiNCs may inhibit defect migration possibly via bonding with the perovskite structure.

**Figure 13.** Perovskite-silicon nanocrystal (SiNC) hybrid solar cells show improved device performance especially after light-soaking. (**a**) Schematic of device structure, and current-density voltage (JV) curves for (**b**) MAPbI3 alone, (**c**) MAPbI3 with p-type SiNCs, and (**d**) n-type SiNCs. Reported from ref. [85], with permission from Elsevier, 2018.

In addition, incorporating nanocrystals into OHPs presents the opportunity to create various types of favorable band alignment between the OHP and the nanocrystal. Coupling the properties of nanocrystals with perovskites can lead to improvements in device performance and opens up an avenue of possibilities to exceed the SQ-limit. Forming an inverted type-I junction can potentially improve carrier collection either through optical coupling or electronic coupling. MAPbI3-SiNC hybrid devices form an inverted type-I band alignment (Figure 14), where wider-bandgap SiNCs were incorporated into the perovskite layer with electronic and/or optical coupling with the OHP, depending on whether or not the SiNCs are oxidized [85]. In an electronically coupled inverted type- I junction, the absorption in the wider-bandgap nanocrystal generates carriers which can be transferred into the adjacent conduction and valence bands of the smaller bandgap perovskite. In an optically coupled system, the nanocrystal behaves as a 'interpenetrated' down converter for high energy photons, where radiative carrier recombination via photoluminescence (PL) results in excitations in the narrow bandgap perovskite. It is therefore important that the peak PL emission is tailored to the bandgap of the perovskite to maximize the conversion efficiency. In MAPbI3:SiNC hybrid devices, it is expected that the structure initially forms an electronically coupled junction whereby carriers generated in the SiNCs can transfer into the OHP. After oxidation, carriers generated in SiNCs are trapped by the oxide potential barrier and recombine via photoluminescence, thus generating an optically coupled junction. It was found using Kelvin probe and XPS that the type-I band alignment is preserved even after the SiNCs became oxidized [85]. These new architectures represent new opportunities for exploring different combinations of materials with perovskite structures.

**Figure 14.** Inverted type-I band alignment: (**a**) electronically coupled and (**b**) optically coupled. Reproduced from ref. [85], with permission from Elsevier, 2018..

#### *3.6. Perovskite Oxide Nanoparticles*

Perovskite oxides (ABO3) are attractive materials for photovoltaics because of the possibility of low-cost, non-toxic photovoltaics with high stability [131]. However, most semiconducting perovskite oxides have large bandgaps (~3–5 eV) due to oxygen-metal transitions with large differences in their electronegativities [132,133], and are therefore generally unsuitable for absorbing light within the solar spectral range. Attempts to reduce the bandgap of perovskite oxides include doping [134], intrinsic defects [135], forming oxynitrides [136], solid solutions [137], and cationic ordering [132].

Perovskite oxides and their derivatives (layered perovskite oxides) represent a large family of materials which exhibit a multitude of properties, and have been investigated for applications including photovoltaics [138]. Perovskite oxides possess a high-degree of flexibility given that 90% of the metallic natural elements in the periodic table can adopt a stable perovskite-type oxide structure [139]. There remains a significant opportunity for exploring the use of metal oxides in photovoltaics to achieve affordable solar cell devices with high efficiency and tunability, whilst easily meeting the often elusive requirement of high stability. The use of metal oxides with highly-tunable absorption properties via the introduction of vacancies [135] and doping [140] would allow for the facile fabrication of multi-junction devices with high stability.

Ferroelectric perovskite oxides have been demonstrated in photovoltaics [133], however they tend to possess large bandgaps (~3–5 eV) and low conductivities, and therefore efficiencies are low (~1%). Plasmonic perovskite oxides have not been explored to the same extent for photovoltaics. Perovskite oxides can be heavily doped to be plasmonic or can be achieved through structural vacancies to strongly modify electronic properties [140]. However, one of the issues associated with plasmonic materials is carrier extraction, and therefore forming extremely thin absorber layers using nano-sized plasmonic oxides is necessary to rapidly extract carriers before recombining.

The perovskite oxide CaMnO3 is an orthorhombic perovskite, and upon reduction in flowing Ar gas the structure can be transformed to an oxygen-deficient perovskite with the structure CaMnO2.5. The structure of CaMnO2.5 is essentially an orthorhombic perovskite with an internal 1D nanostructure ordering as shown in Figure 15a. The introduction of oxygen vacancies removes one oxygen atom from each MnO6 octahedra and results in a square pyramid of MnO5. This structural transformation reveals many interesting properties, such as plasmonic behavior and significantly improved electrocatayltic and photocatalytic activity [141]. CaMnO2.5 can be described as an orthorhombic perovskite because the Ca and Mn perovskite sub-lattice is preserved. Since the resulting powder is phase pure, the oxygen vacancies are expected to be ordered resulting in 1D chains of MnO5 square pyramids [141]. The MnO5 square pyramids are connected along a one- dimensional network extending through the crystal connected by oxygen atoms, which may be favorable for carrier extraction. This oxygen deficiency creates an internal molecular level porosity. The one-dimensional network of MnO5 pyramids may also enhance charge transport and enable efficient carrier separation whereby photoexcited carriers are transported along segregated Mn–O carrier transport channels. CaMnO2.5 displays broad absorption of light from the infra-red through to the visible region of the solar spectrum. Nanoparticles of CaMnO2.5 can be easily produced via a sol-gel process followed by reductive annealing, and then deposited as an ultrathin film either by spray coating or spin coating. The shape and size of the CaMnO2.5 nanoparticles is shown in Figure 15b. CaMnO2.5 nanoparticles have been successfully used to fabricate a photovoltaic cell, and the device performance is shown in Figure 15c. While initial device performance was low, this work serves as a proof of concept and it is likely that the efficiency can be significantly improved, primarily through optimization of the layer thickness and interfacial engineering to improve coupling between CaMnO2.5 nanoparticles and transport layers.

**Figure 15.** (**a**) Structure of CaMnO2.5 reproduced from ref. [141], with permission from American Chemical Society, 2014, (**b**) optical microscope images of CaMnO2.5 after laser fragmentation, the inset shows a high-magnification optical microscope image, and (**c**) current density-voltage characteristic of a CaMnO2.5 solar cell under solar simulated light.

#### **4. Conclusions and Outlook**

This review article has provided a summary of 3D bulk OHPs and an overview of the recent direction and progress towards LDPs. To date, 1DPs and 2DPs have shown the highest efficiencies, yet it is unclear whether these materials will suffer the same long-term stability issues as bulk OHPs. Nanosheets with *n* ≤ 2 tend to show impressive stabilities but suffer from low performance issues,

particularly due to their very large bandgaps. It is still unclear whether 2DPs and 1DPs with *n* > 2 can demonstrate the long-term stability required for commercialization. Furthermore, the issue of stacking faults between grains, which inhibits charge transfer through the layer, must be overcome to increase the efficiency towards 20%. Exploring conductive organic barriers could be a possible route towards overcoming carrier transport issues.

0DPs tend to be highly stable, however their efficiencies are often very low due to issues associated with carrier extraction, where excitons tend to be strongly localized on BX6 <sup>4</sup><sup>−</sup> or B2X9 <sup>3</sup><sup>−</sup> clusters. Methods to enhance exciton dissociation and carrier transport need to be further explored if these materials are to demonstrate noteworthy efficiencies in photovoltaics, particularly through forming hybrids with nanocrystals to promote exciton dissociation, and by exploring various ion substitutions at the A-site to lower the exciton binding energy. Provided that these challenges can be overcome, 0DPs with large bandgaps can be incorporated as a top cell in a tandem solar cell. For use in single junction cells, it is important to explore doping along with varying the A-site ion with the aim of discovering 0DPs with smaller bandgaps, which are currently often >2 eV.

PQDs have shown impressive performance so far, yet the choice of materials is rather limited due to the poor stability of organometal PQDs, which are unstable unless capped with long chain organic barriers which inhibit carrier transport. Inorganic CsPbI3 QDs do not require capping molecules and have demonstrated improved short-term stability along with impressive solar cell efficiencies over 13%. Despite this, CsPbI3 QDs are highly unstable in ambient conditions and encapsulation of the entire solar cell device is essential. As research in this field is still in its infancy, there are limited studies on the stability of CsPbI3 QDs and the extent to which the stability can be improved remains unclear. It is therefore currently difficult to predict the potential for CsPbI3 QDs in photovoltaics.

Hybrid devices can be formed by adding NCs to bulk 3DPs or 0DPs. These hybrid device architectures have been explored using SiNCs, demonstrating an improvement in the device performance due to the possibility of a type-I band alignment which can be optically and/or electronically coupled to improve carrier collection. Furthermore, adding SiNCs indicates a route towards extending the device lifetime, whereby SiNCs are oxidized by the residual moisture in the layer rather than degrading the OHP. This was shown to preserve the favorable type-I band alignment without affecting the device performance.

Due to the significant stability issues suffered by OHPs, occurring both in bulk and lowdimensional forms, we have also briefly introduced the field of perovskite oxide nanomaterials, studying the oxygen-deficient perovskite CaMnO2.5. This material, which absorbs a broad range of light in the solar spectrum from infrared to ultra-violet, has a one-dimensional internal structure which may promote carrier transport. Although the efficiency of the solar cell device is low, there remains significant opportunities for tuning the properties, optimizing devices, and exploring doping to improve device performance.

Finally, while the efficiencies of LDPs are still often far lower than bulk-OHPs, it is encouraging that higher device efficiencies are continually being reported. Provided that these devices can be fabricated with efficiencies of >20%, it is likely that they will be attractive to the market assuming they can be produced at very low cost and with far superior stability to 3D OHPs.

**Funding:** This work was supported by the Japanese Society for the Promotion of Science (JSPS) (17F17815), EPSRC (EP/K022237/1, EP/M024938/1 and EP/R023638/1), the EPSRC Supergen SuperSolar Hub, the Department for Employment and Learning (DEL) of Northern Ireland Studentship, and by the New Energy and Industrial Technology Development Organization (NEDO).

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**

1. National Renewable Energy Laboratory Best Research-Cell Efficiency Chart. Available online: https: //www.nrel.gov/pv/assets/pdfs/best-research-cell-efficiencies.20190923.pdf (accessed on 25 September 2019).


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Review* **The Way to Pursue Truly High-Performance Perovskite Solar Cells**

**Jia-Ren Wu 1,**†**, Diksha Thakur 1,**†**, Shou-En Chiang 1, Anjali Chandel 1, Jyh-Shyang Wang 1,2, Kuan-Cheng Chiu 1,2 and Sheng Hsiung Chang 1,2,\***


Received: 16 August 2019; Accepted: 3 September 2019; Published: 5 September 2019

**Abstract:** The power conversion efficiency (PCE) of single-junction solar cells was theoretically predicted to be limited by the Shockley–Queisser limit due to the intrinsic potential loss of the photo-excited electrons in the light absorbing materials. Up to now, the optimized GaAs solar cell has the highest PCE of 29.1%, which is close to the theoretical limit of ~33%. To pursue the perfect photovoltaic performance, it is necessary to extend the lifetimes of the photo-excited carriers (hot electrons and hot holes) and to collect the hot carriers without potential loss. Thanks to the long-lived hot carriers in perovskite crystal materials, it is possible to completely convert the photon energy to electrical power when the hot electrons and hot holes can freely transport in the quantized energy levels of the electron transport layer and hole transport layer, respectively. In order to achieve the ideal PCE, the interactions between photo-excited carriers and phonons in perovskite solar cells has to be completely understood.

**Keywords:** perovskite solar cells; hot-carrier characteristics; quantized electron transport layer; quantized hole transport layer

#### **1. Introduction**

The bounded electrons of inorganic and organic semiconductors can be efficiently and instantaneously excited from the ground state to the excited state when the photon energy of the incident lightwaves is higher than the absorption bandgap. However, the photo-excited electrons (hot electrons) in the light-absorbing materials (LAMs) have to relax to the meta-stable state (conduction band minimum (CBM) or lowest unoccupied molecular orbital (LUMO)) due to the ultrafast thermallization process [1–4], which results in the intrinsic potential loss and thereby limits the power conversion efficiency (PCE) of single-junction solar cells to be a moderate value of 33.7% [5]. The physical concept of the Shockley–Queisser (S–Q) limit can be understood as the following descriptions. When then LAM has a large bandgap, the broadband sun light cannot be efficiently absorbed by the wide-bandgap material. Therefore, the photocurrent density of solar cells can be increased with a decrease in the absorption bandgap of the active layer. For example, the photocurrent density of single-crystalline Si solar cells (~42 mA/cm2) is always higher than that of single-crystalline GaAs solar cells (~29 mA/cm2) because the absorption bandgap of crystalline Si (1.1 eV) is lower than that of crystalline GaAs (1.43 eV). When the active layer is a wide-bandgap material, the photo-excited electrons have to relax to the meta-stable state, which indicates that the highest potential difference between the cathode electrode and the anode electrode is equal to *Eg*/*e*. *Eg*and *e* are the absorption bandgap of the active layer and the electric charge, respectively. Usually, the potential difference between the cathode electrode and the anode electrode equals to the open-circuit voltage (VOC) which is defined by the current density–voltage (J–V) curve of solar cells. As we know that the VOC of a solar cell is proportional to the absorption bandgap of the active layer. For example, the VOC of single-crystalline

Si solar cells (~0.738 V) is lower than that of single-crystalline GaAs solar cells (~1.127 V). According to the S–Q limit, it is impossible to simultaneously obtain the high VOC and the high photocurrent density (short-circuit current density, JSC), which results in an optimal absorption bandgap of 1.34 eV for the highest PCE of 33.7%.

In the past several decades, physical and chemical scientists were trying to achieve the highest PCE of 33.7% by using the different types of solar cells. When the exciton binding energy of LAMs is lower than the thermal energy, the planar thin-film structure can be used to construct the high-performance solar cells, such as the crystalline Si [6,7], crystalline GaAs [8,9] and crystalline InP [10,11] solar cells. When the exciton binding energy of the LAMs is higher than the thermal energy, the P:N nanocomposite thin-film structures have to be used to increase the photovoltaic performance, such as the organic bulk-heterojunction solar cells [12–14] and dye-sensitized solar cells (DSSCs) [15–17]. Although the PCE of organic photovoltaics (OPVs) and DSSCs is significantly lower than that of the planar thin-film inorganic semiconductor-based solar cells, the cost-effective OPVs and DSSCs still received a lot of attentions in the past two decades. Thanks to the fundamental investigations on the OPVs and DSSCs, the PCE of perovskite solar cells has dramatically increased from 3.8% [18] to 25.2% [19] by using the solution-processed methods.

Itis amazing that the high-efficiency perovskite solar cells can be realized by using thelow-temperature solution-processed methods because the presence of high-density defects in the active layer [20–22] usually can simultaneously reduce the VOC, JSC and fill factor (FF) of solar cells. It is well known that high-efficiency perovskite solar cells can be explained mainly due to the large absorption coefficient [23,24], moderate refractive index [25,26], low exciton binding energy [27,28], long exciton (carrier) lifetime [29,30] and long exciton (carrier) diffusion length [31,32]. In addition, the high PCE of perovskite solar cells also relieson the efficient energy transfer at the perovskite/electron transport layer (ETL) and perovskite/hole transport layer (HTL) interfaces. The highest PCE of perovskite solar cells is theoretically predicted to be an attractive value of 31% [33], which is also limited by the prediction from the S–Q limit.

To pursue truly high-performance solar cells, it is necessary to reduce the intrinsic potential loss via increasing the hot-carrier lifetimes of LAMs. The hot-electron lifetimes of GaAs, Si and InP crystals are 1.5 ps [34], 0.18 ps [35] and 3.4 ns [36], respectively. In general, the lower phonon energy corresponds to the longer hot-electron lifetime [37,38]. The ultrashort hot-electron lifetimes mean that the photo-excited electrons must relax to the meta-stable state to form excitons. In recent reports, the lifetime for the hot electrons in perovskite crystals has been related to the +1 cation [39]. Furthermore, the hot-electron lifetime (diffusion length) of MAPbI3 thin films was determined to be longer than 20 ps (600 nm) by using transient absorbance spectroscopy [40]. The long-lived hot-carrier mediated light emission was also observed in formamidinium tin triiodide perovskites [41]. The existence of long-lived hot electrons means that it is possible to realize truly high-performance solar cells when crystalline perovskite thin films are used as the LAM.

In this review, we discuss the hot-carrier characteristics and the ways for hot-carrier extractions in the energy–space diagrams. A theoretical point of view is proposed in order to understand how the single-junction hot-carrier solar cells can be realized. Finally, the practical issues are discussed in order to assess the possibility for the realization of perovskite-based hot-carrier solar cells.

#### **2. Light-Materials' Interactions: Excited Bounded Electrons**

Photon energy can be efficiently converted to electrical power by using p-type materials due to the high absorption coefficient. Figure 1 shows the carrier dynamics of photo-excited electrons in an energy–space diagram. When the incident photons are absorbed by a p-type material, the electrons in the ground state can transit to the excited state to form hot carriers. The hot carriers can be viewed as the oscillating charged particles, which can coherently and incoherently collide with the lattice vibrations (photons). The coherent collisions between the hot carriers and phonons can result in Raman scattering emissions. The incoherent collisions between the hot carriers and photons can result in the photoluminescence (PL) emissions. In addition, the hot-carrier mediated PL emissions can be observed in the perovskite thin film due to the slow thermalization process [41].

**Figure 1.** Photo-excited carrier dynamics in an energy–space diagram.

In general, the hot-carrier lifetime of organic LAMs is shorter than 100 fs. Therefore, it is not easy to observe the hot-carrier dynamics of organic materials by using the femtosecond time-resolved photoluminescence (FTR-PL) technique due to the limited instrument response function (IRF ~150 fs) [42,43]. The ultrashort hot-carrier lifetime is mainly due to the large optical phonon energy of organic materials [44], which results in an extremely short hot-carrier diffusion length. It means that the hot carriers rapidly decay to the meta-stable state in organic materials via the thermalization (downhill relaxation) process [2,45]. Then, the electrons in the lowest unoccupied molecular orbital (LUMO) and the holes in the highest occupied molecular orbital (HOMO) are mutually attracted to form excitons. In conjugated small organic molecules, the exciton binding energy can be reduced by increasing the conjugation due to the delocalization effect [46]. In general, the exciton-binding energy of organic LAMs can be a wide range from 0.3 eV to 1.0 eV [47,48], which depends on the degree of delocalization of electron-hole pairs. This means that the larger exciton binding energy corresponds to the shorter (smaller) exciton radius (volume). From the concept of allowable excitation density, the smaller exciton volume results in the higher exciton generation rate (absorption coefficient). Due to the high density of excitons, the exciton diffusion length and exciton lifetime of organic materials can be lower than 1 nm and 1 ns, respectively.

It can be predicted that the hot-carrier lifetime of inorganic materials [34–36] is longer than that of organic materials because the optical phonon energy of inorganic materials [49,50] is lower. The optical phonon energy and hot-carrier lifetime of various materials are listed in Table 1 [34–38,49–53]. During the hot-carrier thermalization process in a polar semiconductor as shown in Figure 2, the energy of the hot carriers has to be firstly transferred to the longitudinal optical (LO) phonons. Then, the transition from the LO phonons to acoustic phonons results in the lattice heating. With the propagation of acoustic phonons, the thermal energy can transfer to the surroundings. This means that there are three ways that can be used to slow down the hot-carrier thermalization process. The energy transfer rate from hot carriers to LO phonons is intrinsic fast in non-polar materials [54,55], such as Si and Ge. In a polar CH3NH3PbI3 (MAPbI3) crystal thin film, the energy transfer rate from hot carriers to LO phonons can be delayed due to the formation of hot polarons [55], which results in a long hot-carrier cooling time. As we know that hot polarons are quasiparticles, which describes the interaction between the hot carriers and polar lattices. Therefore, the formation of hot polarons can delay the transfer rate from hot carriers to LO phonons. This was firstly explained due to a phonon bottleneck effect [56]. In addition, the propagation of acoustic phonons in perovskite crystals is theoretically predicted to be slow due to the use the insulating organic cation [57], which also can generate the up-conversion of acoustic phonons to re-heat the hot carriers and thereby increases the hot-carrier lifetime [53]. Up to now, the methods to delay the energy transition from LO phonons to acoustic phonons have not yet been proposed for increasing the hot-carrier lifetime in polar perovskite crystals.

**Table 1.** Optical phonon energies (*E*Phonon) and hot-carrier lifetimes (τhc) of various inorganic materials and organic materials. TPA-TTAR-A: triphenylamine-tetrathienoacene-acceptor.


**Figure 2.** Hot carrier-optical phonon energy transfer and thermalization process.

#### **3. Hot-Carrier Extraction at a Light-Absorbing Material**/**Electron Transport Layer (LAM**/**ETL) Interface**

The hot-electron injection from organic fused thiophene-based dyes to TiO2 nanoparticles (NPs) was investigated for the first time by using a FTR-PL technique [58], which was used to explain the abnormal high VOC of 0.93 V in DSSCs with an iodide/triliodide based electrolyte [50]. The dyes can be adsorbed on the surface of TiO2 NPs due to the electrical attraction between the anchoring group of dyes and the oxygen defect of TiO2 NPs. Therefore, it can be understood that the high VOC is due to the hot-electron injection from the dyes to the higher quantized energy levels of the TiO2 NPs, as shown in Figure 3. In this study, the average diameter of the TiO2 NPs is about 10 nm [59], which is about 3 times of the exciton radius. Therefore, it can be predicted that the quantized energy levels can be created in the TiO2 quantum dots (QDs) [60].

**Figure 3.** Hot-electron injection from dyes to the quantized energy levels of TiO2 quantum dots (QDs).

The efficient hot-electron extraction in OPVs has not yet been reported in literature, which is probably due to the fact that the diffusion process is needed for the hot carriers to reach the region of charge transfer radius [61,62]. For example, the hot electrons in poly(3-hexylthiophene-2,5,-diyl) (P3HT) polymers is rapidly decayed from the excited state to the LUMO energy level due to the ultrafast self-localization process (~100 fs) [2,43], which suppresses the hot-electron diffusion. This means that it is possible to realize organic hot-carrier photovoltaics when the ultrafast self-localization process can be reduced by increasing the delocalization quantum-assisted transport within long polymer chains [46].

The efficient hot-electron extraction in perovskite solar cells has not yet been discussed in literature. However, the abnormal high VOC (=1.61 V) of inverted-type MAPbBr3 based solar cells is probably due to the efficient hot-electron extraction from the MAPbBr3 nano-crystals to the indene-C60-bisadduct (ICBA) ETL [63]. In this study, the poly(3,4-ethylenedioxythiophene):polystyrene sulfonate (PEDOT:PSS (1:6 wt%)) thin film and ICBA thin film are used as the HTL and ETL, respectively. The Fermi level of PEDOT:PSS thin film and the LUMO energy level of ICBA thin film are −5.1 eV [64,65] and −3.9 eV [66,67], respectively, which results in a S.-Q limited VOC of 1.2 eV. The high VOC of MAPbBr3 based solar cells means that the hot electrons have to be extracted by the LUMO+1 and/or LUMO+2 of the ICBA thin film because the PEDOT:PSS thin film is a metal-like conductive polymer [68]. The photovoltaic performances of high-VOC perovskite based solar cells are listed in Table 2 [63,69,70]. The three types of perovskite solar cells both contain bromide elements in the active layer, which suggests that the bromide-based perovskite thin films probably have longer hot-carrier lifetimes. In addition, the long hot-carrier lifetime and diffusion length were observed in a MAPbI3 perovskite thin film by using transient absorbance spectroscopy [40], which means that the realization of high-performance hot-carrier perovskite solar cells is possible.

**Table 2.** Photovoltaic performance of bromide-based perovskite solar cells.


#### **4. Hot-Carrier Extraction at a LAM**/**Hole Transport Layer (HTL) Interface**

As we know that the bounded electrons are not excited from the valence band maximum (EVBM) when the photon energy of the incident lightwaves is higher than the absorption bandgap of materials. Therefore, the hot-hole relaxation process has to be considered in order to realize the hot-carrier solar cells. Figure 4 shows the hot-hole and hot-electron dynamics in the energy diagram. For example, the hot-hole and hot-electron relaxation times of a MAPbI3 thin film are 100–500 fs and 1–5 ps [71], respectively, which means that it is more difficult to collect the hot holes in perovskite thin films by using a hot-hole selective layer due to the sub-picosecond relaxation time. Fortunately, there is experimental evidence of transient absorbance spectra to show that the hot holes in a CsPbI3 thin film can be efficiently extracted by the capping layer of P3HT thin film within a few 100 fs [72]. However, the hot-hole extraction process from the perovskite thin film to the P3HT thin film is not yet completely understood. Further experiments are needed to demonstrate that the hot-hole extraction can increase the VOC of perovskite solar cells.

**Figure 4.** Energy diagram of hot-hole and hot-electron relaxations.

The extraction efficiency of hot holes is also related to the carrier diffusion coefficient of perovskite thin films because the hot holes have to diffuse into the region of the charge transfer radius at the perovskite/HTL interface. According to the Einstein relation (*D* = μ*kBT*/*e*), the carrier diffusion coefficient (*D*) is proportional to the carrier mobility (μ), where *kBT* is the thermal energy and *e* is the electric charge. The carrier drift equation (μ = *e*τ/*m*∗ ) shows that the carrier mobility is proportional (inversely proportional) to the carrier relaxation time (effective mass), where τand *m\** are the carrier relaxation time and carrier effective mass, respectively. Therefore, the smaller hot-hole effective mass corresponds to the longer hot-hole diffusion length and thereby results in the higher extraction efficiency of hot holes. Interestingly, the hole mobility can be higher than the electron mobility when the LAM is CsPbBr3 or CsPbCl3, which is due to the lower hole effective mass [73]. However, the propagation characteristics of hot holes are not yet completely understood.

#### **5. Theoretical Point of View**

The highest PCE of single-junction hot-carrier solar cells was theoretically predicted to be 66% under one sun illumination [74]. Figure 5a shows the energy diagram of a single-junction hot-carrier solar cell. In the single-junction solar cell, the high-energy and low-energy incident photons are absorbed by the electrons in the deeper levels and in the shallower levels of the valence band, respectively. When the electrons in the valence band are excited to the conduction band, the hot

electrons in the higher energy levels and in the lower energy levels both have to be directly extracted in order to avoid ultrafast potential loss. As to the hot holes, they also have to be directly extracted in order to keep the original potential.

**Figure 5.** Energy diagrams. (**a**) Single-junction hot-carrier solar cell. (**b**) Tandem solar cell.

The concept of single-junction hot-carrier solar cells is similar to the tandem (multi-junction) solar cells which has a highly theoretical PCE of 68% under one sun illumination [75]. Up to now, the highest PCE of tandem solar cells is 39.2% under one sun illumination, which is significantly higher than the highest PCE (29.1%) of single-junction GaAs solar cells [19]. When the different absorption ranges in a single LAM are viewed as the individual materials with the different absorption bandgaps, the energy diagram of a tandem solar cell can be plotted in Figure 5b. The PCE of tandem solar cells is strongly related to the performance of the tunnel junctions [76] with an ultrafast carrier dynamic [77]. In single-junction hot-carrier solar cells, the double-barrier resonant tunneling structure was proposed as the energy selective contact of hot carrier solar cells [78] due to the sub-picosecond carrier extraction ability [79]. Therefore, it can be believed that the highly efficient single-junction hot-carrier solar cells can be realized when the ultrafast carrier lifetimes of hot carriers in the LAM can be increased from the sub-picosecond time scale to sub-nanosecond time scale.

In order to efficiently collect the hot carriers at the different energy levels, a multi-band ETL and a multi-band HTL have to be used to extract the dispersive hot carriers under a broadband excitation. T 5ashows the energy diagram of an ideal hot-carrier solar cell. The ECBM of HTL (EVBM of ETL) has to be higher (lower) than the E3 in the conduction band of ETL (E3 in the valence band of HTL) in order to block the hot electrons (hot holes), which can help the collection of hot carriers. If the high-energy hot carriers can be efficiently collected by the E3, the hot electrons (hot holes) have to freely transport within the E3 of ETL (HTL) without the energy transitions from the E3 to the E2 and/or E1. Fortunately, the photo-excited carriers can be collected and transport in the quantized energy levels of Si-doped quantum dots, which increases the VOC from 0.78 V to 0.91 V [80]. Therefore, it can be expected that the hot electrons and hot holes can freely transport in the quantized energy levels of ETL and HTL, respectively. In addition, n-type graphene quantum dots [81–83] and p-type graphene quantum dots [84–86] might have the potential as the multi-band ETL and multi-band HTL of hot-carrier solar cells, respectively.

In addition, a theoretical approach is used to calculate the J–V curves of single-junction solar cells when the selective contacts are used to collect the hot carriers [87]. Their simulation results show that the *e*×VOC of hot-carrier solar cells can larger than the absorption bandgap of LAMs. For example, *e*×VOC of the GaAs solar cell is 1.85 eV which is larger than the absorption bandgap of crystalline GaAs.

#### **6. Experimental Challenges and Opportunities**

Up to now, the LAMs of highly efficient perovskite solar cells were fabricated by using solution-processed methods [26,28,88–91], which means that the defect density of LAMs remains high. Although the shallow defects of perovskite thin films do not significantly influence the photovoltaic performance [92,93], the formation of defects can increase the exciton binding energy and optical phonon energy of perovskite thin films [51,94]. Therefore, the defect-mediated phononic properties of perovskite thin films have to be investigated in order to understand how to realize hot-carrier solar cells.

The grain size of solution-processed perovskite thin films can be increased from several hundred nanometers to several micrometers by adding the small molecules [95] or with the solvent annealing process [96]. The average crystal domain size of perovskite thin films is smaller than averaged grain size, which indicates that the perovskite thin films are composed of multi-crystalline grains [97]. It means that the residual stress in a multi-crystalline perovskite thin film [98] can also influence the phononic properties, which might dominate the hot-carrier characteristics. In other words, the defect-mediated phononic properties [99] and/or crystal distortion-mediated phononic properties [100] have to be considered when the perovskite thin films are fabricated on top of amorphous substrates or poly-crystalline substrates by using solution processes or thermal evaporation methods. However, the preferred oriented perovskite thin films can be fabricated on top of the single-crystalline substrates by using the spin-coating method [101], which was observed by measuring the two-dimensional X-ray diffraction patterns.

Conceptually, the existences of defects and lattice distortions should decrease the hot-carrier lifetimes in the LAMs, which are predicted to impede the development of hot-carrier solar cells. Therefore, the development of single-crystalline perovskite bulks plays an important step for the realization of high-performance optoelectronic devices. Two years ago, the single-crystalline perovskites were grown on top of various substrates, such as indium tin oxide (ITO), quartz and silicon wafer, which were used as the X-ray detector [102]. The strategy is to modify the surface of the substrates with a NH3-Br-teminated self-assembling molecules monolayer, which provides a seeding layer to grow the single-crystalline MAPbBr3. The long carrier lifetime of the single-crystalline MAPbBr3 is 692 ns, which indicates the low defect density. Therefore, the hot-carrier lifetimes of single-crystalline perovskites can be expected to be longer than that of poly-crystalline perovskite thin films.

When the substrate (Au/p+-type wafer) is the anode side, the single-crystalline perovskite has to be grown on top of a quantized HTL. Then, a quantized ETL has to be fabricated on top of the single-crystalline perovskite. A transparent conductive oxide has to be deposited on top of the device as the cathode electrode. Therefore, it can be imagined that the device architecture is Au/p+-type wafer/quantized HTL/single-crystalline perovskite/quantized ETL/transparent conductive cathode. The p+-type wafer has to be a large-bandgap material in order to block the hot electrons from the single-crystalline perovskite. The quantized HTL and the quantized ETL can be a p-type quantum well (QW) structure [103] and a double-barrier resonant-tunneling structure [78], respectively. In addition, the transparent conductive cathode is used to block the hot holes from the single-crystalline perovskite. The potential candidates as the p+-type substrate, HTL, ETL and transparent conductive anode are listed in Table 3. The Au coated p+-type GaN, AlN or SiC wafer can be used as the anode electrode. The epitaxial growth process of GaN/AlN QW [104] or AlGaN QDs [105] on top of the GaN or AlN wafer is a mature technique by using metal organic chemical-vapor deposition (MOCVD) or molecular beam epitaxy (MBE). However, the barrier high and physical size of the QW or QDs has to be investigated and designed in order to be used as the quantized HTL for the collection of hot holes. P-type graphene QDs can be produced by using microwave-assisted heating method [106] or pulsed laser ablation method [107]. And, the p-type graphene QDs thin film can be deposited on top of the p<sup>+</sup>-type wafer by using the spin-coating method. Although, the single-crystalline MAPbBr3 has been demonstrated that can be grown on various substrates with a surface modification method [102]. The contact at the MAPbI3/HTL interface, which should strongly influence the collection efficiency

of hot holes, has not yet investigated. It is predicted that the [6,6]-phenyl-C61-butyric acid methyl ester (PCBM)/bathocuproine (BCP) QW, ZnO QDs [108] or TiO2 QDs [109] can be used to collect the hot electrons from the single-crystalline perovskite. PCBM molecules can be dissolved in low-polarity solvents, such as chlorobenzene and toluene, which can be directly spin-coated on top of MAPbI3 thin films. BCP molecules, ZnO QDs and TiO2 QDs are usually dissolved in isopropanol (IPA) which is a polar solvent. Therefore, the BCP, ZnO QDs or TiO2 QDs thin film cannot be directly spin-coated on top of MAPbI3 thin films. The contact at the hydrophobic ETL/hydrophilic MAPbI3 interface can be improved by slightly roughening the surface of MAPbI3 thin film [91]. Then, the Al-doped ZnO, Ga-doped ZnO or Al-Ga co-doped ZnO thin film can be used as the transparent conductive cathode because of the metal-like electrical conductivity which can directly collect the hot electrons without the additional potential loss. However, the high-quality transparent conductive cathodes are usually deposited by using the radio-frequency magnetron sputtering method, which can damage the MAPbI3 thin film due to the excessive energy bombardment during the deposition process [110]. To resist the excessive energy bombardment, an inorganic thin film has to be used as the buffer layer in between the MAPbI3 thin film and transparent conductive cathode [110].

**Table 3.** Potential candidates as the p+-type substrate, hole transport layer (HTL), electron transport layer (ETL) and transparent conductive anode.


The realization of hot-carrier solar cells has to rely on the long-lived hot carriers in the active layer and the efficient collections of hot carriers by the HTL and ETL. The hot-carrier lifetime and hot-carrier diffusion length of crystalline perovskite thin films can be longer than 100 ps and 600 nm, respectively, which indicates that the hot carriers are possibly collected without potential loss. However, the hot carriers have to be rapidly collected before the ultrafast thermalization process, which means that the collection times of hot carriers have to be faster than the hot-carrier lifetimes. For example, the hot-carrier lifetime of the LAM and the hot-carrier collection time of a double-barrier resonant tunneling structure can be 100 ps and 0.4ps, respectively. Furthermore, the hot-carrier collection efficiency can be calculated by h = 1/t1/(1/t1 + 1/t2) [58], where t1 and t2 are hot-carrier collection time and hot-carrier lifetime, respectively. Therefore, the calculated hot-carrier collection efficiency equals to 99.6%, which means that it is worthwhile to develop a double-barrier resonant tunneling structure as the selective contact of hot-carrier perovskite solar cells. In this review, we suggest that the quantized HTL and quantized ETL can be used to collect the dispersive hot holes and hot electrons, respectively. The collection times of hot carriers will dominate the collection efficiency. Therefore, the barrier height and physical size of the quantized HTL and quantized ETL has to be varied in order to decrease the collection times of hot carriers. Conceptually, the collection times of hot carriers are also influenced by the crystallinity (carrier mobility) of the HTL and ETL. Therefore, the formation of high-quality quantized HTL, quantized ETL and perovskite thin film is necessary in order to realize hot-carrier solar cells.

#### **7. Conclusions**

Recent progress in the understanding of the light-perovskite interactions shows that the realization of highly efficient hot-carrier solar cells is possible because the long-lived hot polarons can be formed and thereby delays the thermalization process. Conceptually, the extraction of hot electrons and hot holes can increase the open-circuit voltage (VOC). The hot-electron injection was firstly observed in the dye-sensitized solar cells, which significantly increased the VOC from 0.75 V to 0.93 V. The long-lived hot electrons in the organic fused thiophene-based dyes were observed by using the femtosecond time-resolved photoluminescence technique.

The highest VOC of CH3NH3PbBr3 solar cells is 1.61 V, which is far larger than the potential difference (1.2 V) between the LUMO energy level of the electron transport layer (ICBA thin film) and the Fermi level of the hole transport layer (PEDOT:PSS thin film). The abnormal high VOC of perovskite solar cells can be explained due to the efficient collection of hot electrons by the LUMO+1 and/or LUMO+2 of the ICBA thin film. In addition, the P3HT polymer thin film was used as the hot-hole selective contact layer, which was observed by using the femtosecond transient absorbance spectra.

To realize truly high-performance hot-carrier perovskite solar cells, it is necessary to simultaneously collect the hot electrons and hot holes without the additional potential loss. We have proposed a device architecture which might be used to achieve the desired power conversion efficiency. The device architecture is Au/p+-type GaN wafer/quantized hole transport layer (QHTL)/crystalline perovskite/quantized electron transport layer (QETL)/transparent conductive cathode. Ideally, the hot holes and hot electrons in the crystalline perovskite have to be collected by the QHTL and QETL, respectively. Therefore, it is necessaryto investigate the ultrafast energy transfer dynamics of hot carriers from the crystalline perovskite to the QHTL and QETL by varying the barrier high and physical size of the quantum wells or quantum dots.

**Author Contributions:** Conceptualization was made by S.H.C., J.-S.W. and K.-C.C. References were collected by J.-R.W., D.T., S.-E.C. and A.C. Draft preparation was made by J.-R.W. and D.T. Writing—Review and Editing was made by S.H.C.

**Funding:** This research was funded by Ministry of Science and Technology, Taiwan. Grant number is MOST 107-2112-M-033-001-MY3.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
