Halide perovskite photovoltaics

Halide perovskites are a new class of materials that can be processed at low cost yet exhibit excellent optoelectronic properties – a combination rarely seen in semiconductors. This review highlights their exciting promise for inexpensive and efficient solar photovoltaic technologies. The work details the rapid progress in efficiency of solar cells – with efficiencies already rivalling commercial leading technologies such as silicon – and describes the physics of the materials and devices that lend themselves to photovoltaic applications. This work also describes the challenges to realise their true potential as disruptive photovoltaic technologies that could play a key role in decarbonisation in the coming decades, including stabilising long-term operation and furthering understanding of this fascinating class of semiconducting materials.


Photovoltaics playing a key role in driving decarbonisation
Solar photovoltaic (PV) technologies -which convert incident solar light to electrical power -could provide large fractions of humanity's power requirements.The total annual solar insolation hitting the earth is over 10,000 times greater than the entire world's energy consumption, showing the enormous promise for solar.Almost all decarbonisation models require PV to play a prominent role; for instance the International Energy Agency's net-zero report mandates that PV should make up the largest proportion of total energy by 2050 [1].To date, there is just over 1 TW of installed PV capacity worldwide providing around 3.6% of global electricity generation [2] -a feat in and of itself.However, to achieve decarbonisation fast enough, solar PV will need to be deployed at a rate exceeding 1 TW per year [2,3] thus highlighting the scale of the challenge.The lower the cost of PV systems, the faster and more widely the technology can be deployed and thus the more solar PV can contribute to decarbonisation.There are additional benefits: for example, it has been estimated that if 27% of the US electricity was provided by PV in 2050 this alone would give US$167B in health benefits and save almost 60,000 lives through better air quality [4].Furthermore, solar PV can provide energy security and equality -averaged over the year, the sunniest country (Azerbaijan) gets only 4x more sunlight than the cloudiest (Norway), compared to oil where the discrepancy is CONTACT Samuel D. Stranks sds65@cam.ac.uk 1,000,000x between countries with the least and most oil; lower GDP countries also get more sunlight on average than richer countries, making solar PV a great leveller [5].
The cost of PV has dropped dramatically -PV module prices have dropped 100x since 1980 and almost 10x since 2010 (Figure 1(a)) [5].This has been primarily driven by cost reductions in raw materials and manufacturing (economies of scale) and gains in the power conversion efficiency (PCE) of sunlight to electricity (research and development).The cost of a PV system is now dominated by the 'balance of system costs' including the installation, wiring, inverters, and other components, rather than the solar modules themselves (Figure 1(b)).Therefore, the most practical way to drive further tangible reductions in the cost of PV is to increase efficiency.Today's marketleading PV technology -honed over decades -is based on crystalline silicon with > 95% of the market share [6].However, the record lab cell power-conversion efficiency of ∼ 27% [7] is now close to its performance limits ( ∼ 29%) [8].This efficiency plateau, along with high-temperature and energy-intensive manufacturing, means scope for further reductions in PV system cost and thus achieving sufficiently rapid deployment with silicon alone is far from certain [3,9].Further, supply chain and more recent geopolitical issues leading to deglobalisation in many sectors further calls into question whether PV can be globally deployed fast enough in the coming decade.We thus need to urgently develop higher efficiency PV technologies with lower systems Copyright IOP Publishing.Reproduced with permission.All rights reserved [5].(b) Breakdown of the cost of a benchmark residential PV system in the USA in Q1 2022.Image adapted from [10].
costs ensuring faster and more modular deployment of this green energy technology.

Halide perovskites: a new world of semiconductors
Halide perovskites show enormous promise as low-cost yet high-performance photovoltaic technologies [11], with potential to push efficiencies beyond silicon including through adaption in new device configurations, and new manufacturing paradigms.More generally, they show exceptional optoelectronic properties that are seeing them being developed in a variety of applications including light-emitting diodes (LEDs) and X-ray and photodetectors.These materials were worked on extensively for transistor applications in the 1990s and 2000s but overlooked for solar photovoltaic applications until the first peer-reviewed report by Miyasaka and coworkers in 2009 [12].Only in 2012, when the first highperformance ( > 10% lab cells) and more stabilised cells were reported [13,14], did a new wave of worldwide interest in developing them for solar cells kick-start.What has ensued since has been an explosion in both technological and fundamental scientific understanding of these fascinating semiconductors, bucking the trend of traditional semiconductors by being crudely and inexpensively processed yet exhibiting exceptional performance.

General material properties: crystal structure, band structure and defects
A perovskite is any material that shares the same type of crystal structure as calcium titanate, i.e. the ABX 3 crystal structure.The structure consists of a 3D network of corner sharing BX 6 octahedra, where the A-site cations are in 12-fold coordination with the anions, while the B-site cations are in 6-fold coordination (Figure 2(a)).Many perovskites occur as natural minerals, and oxide perovskites have been used as superconductors and in batteries over decades.The Goldschmidt tolerance factor t [15], initially defined for oxide perovskites, predicts the likelihood of a given combination of ions fitting into a 3D perovskite structure at a particular temperature and pressure, where R A,B,X is the radius of the A-, B-or X-site ions.
The ideal perovskite has a cubic structure with t = 1 and typically would retain a cubic structure over 0.9 ≤ t ≤ 1, but lower symmetry structures do in general exist with 0.71 ≤ t ≤ 0.91 yielding, for example, orthorhombic, rhombohedral and/or hexagonal structures.The octahedra tilt when the size of the A-site ions is too small to fully occupy the available volume at a critical value of t, with Glazer proposing the nomenclature for analysis of such small angles of tilt [16].The specific perovskites that are generating excitement as new and efficient photovoltaic semiconductors are 3D metal halide perovskites with the A-site being a short organic (e.g.methylammonium, MA -CH 3 NH 3 + or formamidinium, FA -HC(NH 2 ) 2 + ) or inorganic Cs + cation, or mixtures thereof, the B-site being divalent metal ions (e.g.Pb or Sn) and the X-site being halides (typically I but also small fractions of Br and/or Cl).It is noted that there is a much larger family of halide perovskites, including layered 2D perovskites where a longer chain organic cation spaces between the metal halide inorganic sheets, double perovskites, and confined nanostructured systems including nanocrystals and nanoplatelets, each with their own interesting materials physics.Many of these materials are less suited to photovoltaic uses due to larger bandgaps, limited transport or indirect bandgaps leading to poor absorption, and the focus of this work will herein be restricted to 3D halide perovskites.
The first alkali halides were synthesised and reported in 1893 by Wells [19].The first crystallographic study on a halide perovskite CsPbX 3 (X = Cl, Br, I) was carried out by Møller in 1958, who observed that these materials are photoconductive suggesting that they behave as semiconductors [20] -paving the way for their future application as efficient photovoltaic absorber layers.The first organic-inorganic hybrid perovskite, MAPbX 3 , was reported by Weber in 1978 [21], and over the last decade this composition has become the Drosophila of halide perovskite photovoltaics.MAPbI 3 exhibits a tetragonal structure at room temperature (Figure 2(b)) [17], transitioning to a cubic structure above 330 K and to the lower symmetry orthorhombic phase below 150 K [22].Other analogues including FAPbI 3 and CsPbI 3 present much greater challenges to stabilise in their strongly absorbing 'black phases' at room temperature, with both in general exhibiting non-photoactive 'yellow' phases [23].Indeed, the optically active black cubic phase α-FAPbI 3 is attained at 420 K [24] and α-CsPbI 3 at > 570 K [25], whereas both exhibit wider bandgap δ-phases at room temperature being hexagonal and orthorhombic, respectively.Nevertheless, MAPbI 3 is less thermally stable than the FA -or Cs-based analogues at temperatures reached during operation of solar cells ( > 85°C) owing to the volatility of the MA cation [11], and thus there is strong motivation to stabilise the FA and/or Cs phases at lower temperatures -approaches that will be discussed later.
The band structure of halide perovskites is dictated primarily by the B-X metal halide octahedral cages [18].In the case of MAPbI 3 (and more generally for Pb-halide perovskites), the top of the valence band is derived from I 5p states and, to a smaller extent, Pb 6s states, whilst the top of the conduction band is derived primarily from Pb 6p and I 5p states (Figure 2(c)).Electronic states associated with the A-site cation are located far from the bandgap [18].However, the A-site does indirectly impact band structure by influencing the octahedral cage; for example, the alloying of smaller A-site cations such as Cs with FAPbI 3 induces structural octahedral tilts that ultimately enlarges the bandgap by decreasing metal-halide overlap [26].Continuous bandgap tunability can in general be achieved by substituting any of the chemical elements of the ABX 3 structure.Tuning the X-site from I to Br in FAPbI 3 (including mixtures of halides), for example, causes the bandgap to increase from ∼ 1.5 eV to ∼ 2.3 eV [27] (Figure 3(a)), again owing to decreased metal-halide overlap and hence deepening of the valence band energy [28] (Figure 3(b)).Bandgap tuning can also be achieved by tuning the B-site through use of Sn instead of Pb, or mixtures of metals [29,30].Adding Sn to Pb reduces the bandgap (Figure 3(c)), down to a minimum for these materials of ∼ 1.2 eV with mixtures of Pb/Sn metals close to 50/50, but further increases in Sn fraction increases the bandgap [31].The origin of this 'bandgap bowing' effect remains contested, with proposed explanations including relativistic effects, steric effects of the ions or energy mismatches between Pb / Sn orbitals [32].This property of bandgap tunability is one of the remarkable enabling features of halide perovskites, in principle allowing photovoltaic absorbers to be tunable from the ultraviolet through to the near infrared.
Halide perovskite thin films -the form used for thinfilm PV applications -are typically deposited by solution processing from inks containing precursors of inorganic and/or organic halide salts dissolved in suitable solvents (typically polar aprotic solvents such as dimethylformamide) [33].For MAPbI 3 , this typically constitutes MAI and PbI 2 precursors.In typical lab-based settings, these are formed by casting down the inks onto transparent substrates such as glass in inert atmosphere conditions (e.g.nitrogen glove box), and spin-coating at several thousand revolutions per minute to form a uniform precursor film (Figure 4(a)).The spin-coating may be accompanied by addition of an anti-solvent or gas quench whilst spinning to induce faster crystallisation [34].The precursors nucleate and then fully crystallise into the perovskite structure by annealing the substrates -typically at 100°C-150°C -to form high-quality, highly absorbing films with polycrystalline grain structure (Figure 4(a)).Thermal evaporation of precursor salts by sublimation under vacuum is another way of processing high quality, controlled films (Figure 4(b)), with the added advantage of avoiding solvents that may be toxic or that redissolve underlying layers in device stacks that otherwise limit the types of material able to be processed.Other deposition approaches make use of two-steps in which the inorganic and/or organic salts are cast sequentially [35] -by either vapour, solution or combinations thereof -and the interconversion of the components during annealing forms the perovskite [33].In each of these processing methods, the annealing temperatures are very low compared to more standard semiconductor processing -for example other established PV materials such as crystalline silicon ( ∼ 900°C [36]) or CdTe ( ∼ 600°C [37]).This opens up processibility of PV cells on lightweight and flexible substrates such as polymers or metal foils [38], leading to new application areas such as vehicle-integrated PV to boost the range of electric vehicles, aerial vehicle and space applications.It also enables high-throughput roll-to-roll processing, which could enable high-volume module fabrication (Figure 4(c)) [39].High-quality single crystals can also be grown by carefully controlling interdiffusion of the precursors components, for example through the inverse temperature crystallisation method [40], though these have been by and large more useful for fundamental studies than practical PV applications.

General photovoltaic principles and technologies
In a crystalline silicon system, silicon wafers are cut from highly purified ingots and made into solar cells, which are linked together to form a module (Figure 5(a)); a typical panel would have 60 cells connected in series each with length dimension of ∼ 16 cm.These modules are then protected with encapsulation layers, a backsheet, glass front and metal frames (Figure 5(b)).These modules are arranged together in an array, together with wiring and electronics, to comprise the solar PV system.In thin film technologies such as halide perovskites which can be deposited directly onto substrates, modules can in principle be fabricated monolithically without needing to link cells together (Figure 5(a)).
In general, a solar cell operates by employing an absorbing semiconducting layer able to efficiently harvest photons across the solar spectrum, which generate energised electrons and holes (Figure 5(c)).These charge carriers need to be transported to opposite electrodes where they are collected -generating both a photo-current and photo-voltage, and hence generating power to drive an external circuit (Figure 5(d)).The ideal solar absorber and device stack would facilitate efficient processes at each step: harvesting of photons, transport of charges across the semiconductor, and collection of charges at the electrode.
The photocurrent density J SC generated by a solar cell at short-circuit depends on the incident light and the quantum efficiency of the cell η(E), the latter being the probability that an incident photon of energy E will generate one electron to the external circuit: Here, S(E) is the incident spectral photon flux density (i.e. the number of photons of energy in the range E to E + dE incident on unit area in unit time) and q is the electronic charge.η(E) is dependent upon the absorption coefficient of the absorber material, the efficiency of transport of the charges to electrodes, and the efficiency of charge collection in the device [42].
A solar cell is typically characterised by its current density-voltage (J-V) characteristics, where the net current density under illumination as a function of applied bias J(V) is approximated by the sum of the dark current density J dark and the short-circuit photocurrent density given in an ideal diode by: where J S is the reverse saturation current in the diode, k B is Boltzmann's constant and T is temperature [42].Whilst the sign convention for electronic devices is for net current to be negative and thus the J-V curve is in the fourth quadrant representing generation of power, in the PV field, this is often represented for convenience in the first quadrant (Figure 6(a)).
Another important metric for PV operation is the open-circuit voltage V oc , which occurs when the contacts are isolated and there is maximum potential difference across the cell.In this case, the dark current and shortcircuit photocurrent cancel out, giving in the case of an ideal diode: A solar cell operating regime in which power is generated will be in the range 0 < V < V oc , with the overall power density generated given by: P = JV. ( P is maximised at the maximum power point, corresponding to V MP and J MP , and the fill factor (FF), indicates the 'squareness' of the J-V curve, and is given by: The overall power conversion efficiency (PCE) is therefore defined as the power density delivered at maximum power point as a fraction of the incident light intensity P in : PCE can thus be determined by extracting key parameters V OC , J SC and FF from a J-V curve.The standardised '1 sun' illumination intensity used for assessing solar cell efficiencies is defined as the Air Mass 1.5 spectrum with an incident power density of 100 mW/cm 2 (Figure 7(a), blue), representing terrestrial conditions; different spectra are considered for other applications, such as the space spectrum (AM0) when considering PV for space applications (Figure 7(a), red), or spectra relevant for indoor photovoltaics.
To understand efficiency limits, one must consider how an absorber with finite bandgap can maximally harvest photons across the solar spectrum.Photons of energy below the bandgap will not be absorbed and will not contribute to photocurrent; photons with energy above the bandgap will be absorbed and contribute to photocurrent though excess energy above the bandgap will be lost to heat as carriers relax to the band edge (Figure 7(b)).To maximise photo-current, one would in principle select a low bandgap to harvest as many photons as possible (Figure 7(c)).However, the open-circuit voltage is dictated by the quasi-Fermi-level splitting of the absorber and hence the bandgap, and so to maximise voltage one needs to maximise the bandgap (Figure 7(d)).This leads to a compromise to maximise power density (product of current density and voltage), and the maximum efficiency limit for a single junction solar cell is ∼ 33% with a bandgap of ∼ 1.3 eV [44] (Figure 7(e)).Here, this is termed the radiative efficiency limit as the only loss mechanism is radiative recombination -an unavoidable process in a material absorbing light that also needs to emit light to be in thermodynamic equilibrium.An interesting corollary of the radiative efficiency arguments is that to maximise efficiency of a solar cell, one must maximise the luminescence efficiencyi.e. a good solar cell should also be a good LED [45].This is manifested in the fundamental relationship between electroluminescence (EL) quantum efficiency η EL and loss in open-circuit voltage from the radiative limit V rad OC , where any deviation from 1 in η EL leads to a non-radiative loss Thus, for every order of magnitude of loss in electroluminescence efficiency, there is a ∼ 65 mV loss in V oc .
An ideal solar cell can be modelled with an equivalent circuit in which a current generator is in parallel to a diode (Figure 6(b)).This provides a framework for understanding loss pathways and thus deviation in efficiency from the ideal case -as is the case in realistic solar cells.Series resistance losses arise from low material conductivity including through use of thick absorber layers and contact resistance.Low shunt resistance leads to shunting currents that may arise from shorting through the device, for example pinholes in the absorber layer that arise from issues with device preparation.Further losses are seen when charge carrier recombination rates outcompete charge transport lifetimes, leading to low charge collection efficiencies.Non-radiative recombination arises from processes such as carrier trapping through defects in the material, which in particular leads to voltage losses as carriers lose their energy to heat.The study of carrier trapping is a rich area and a key focus for PV device improvement.
There is a plethora of absorber layer PV technologies [7], with the key established technologies shown in Figure 8. Established PV technologies include crystalline silicon, with lab-record efficiencies now at 26.8%; despite being an indirect absorber, the efficiencies have been honed over decades and the cells exhibit extremely good diode characteristics.The most efficient single junction technology is based on III-V GaAs absorbers, with efficiency of 29.1%, an electroluminescence efficiency exceeding 30% [47] and a remarkable open-circuit voltage loss from the radiative limit of < 40 mV [48].Such semiconductors are processed epitaxially at very high purity to inhibit formation of defects that are otherwise highly detrimental to performance.Therefore, processing costs, especially for GaAs, remain prohibitively high for widespread terrestrial deployment and as such GaAs and related technologies have been mostly limited to use in space applications where cost is less critical.Commercial thin film technologies include CdTe (22.1%) and Copper Indium Gallium Selenide (CIGS, 23.6%), with the former now comprising around 5% of the PV market share but the latter now falling increasingly out of production.Finally, emerging PV includes organic semiconductors which promise low-cost production, but efficiencies remain more modest at ∼ 19.2%, albeit rising substantially over recent years.
A very promising route to efficiencies beyond the ∼ 33% radiative limits for single-junction cells is multijunction configurations in which different absorber layers harvest complementary regions of the solar spectrum.In the case of a tandem solar cell with two absorber sub-cells, the top cell harvests the higher energy (longer wavelength) photons, and the lower energy (longer wavelength) photons are transmitted and absorbed in the bottom cell (Figure 9(a)).For a tandem solar cell, the theoretical efficiency limits push to ∼ 45%, and in the hypothetical case in the limit of infinite absorber sub-cells this in principle reaches ∼ 87% under concentrated sunlight [49].These cells may be operating in parallel in a 4-terminal configuration (Figure 9  must match in both sub-cells [50] (Figure 9(c)).Multijunction cells especially utilising III-V semiconductors have led to the most efficient solar cells to date, with three junction cells demonstrating efficiency of 39.5% [51], but remain extremely expensive and unlikely to be utilised for terrestrial PV energy applications.Other concepts to higher efficiency include solar concentrators and clever conversion of particular colours of light (e.g., upconversion) to maximally harvest energy from the incident photon flux.

Perovskite solar cells -historical evolution, operation and state-of-the-art
Halide perovskites were explored for transistor applications through the 1990s and 2000s [52].However, they were overlooked for solar cells until 2009 when Miyasaka and co-workers reported the first perovskite solar cell with an efficiency of 3.8% [12], followed in 2011 by a report by Park and co-workers who boosted the efficiency to 6.5% [53].In those reports, a dye-sensitised solar cell architecture was employed in which the perovskite was coated on a metal oxide scaffold to act as an absorbing dye, harvesting photons and injecting photo-excited electrons into the metal oxide electrode (Figure 10(a)).These cells had limited performance and were very unstable as they employed a liquid electrolyte hole transporting material, which rapidly degraded the perovskite absorber layer.Critical breakthroughs came in 2012 when both Snaith, Miyasaka and co-workers [14] as well as Park, Graetzel and co-workers [13] pushed efficiencies towards and over 10% by replacing the liquid hole transporter with a solid-state hole transporter, leading to much greater stability and performance, and catalysing the boon of the field over the subsequent years.
In the 2012 report by Snaith and co-workers [14], they demonstrated that the perovskite could also operate when deposited on an insulating metal oxide scaffold -in which the perovskite absorber itself performs the role of both light absorption and charge transport to electrodes [55].This remarkable result paved the way for planar heterojunction perovskite solar cells in which the dye-sensitised solar cell architecture is replaced with an architecture resembling other thin film PV technologies, where the absorber thin film layer is simply sandwiched between two collecting electrodes (Figure 10(b)).Such a simplified architecture reduces interfaces compared to the mesoporous metal oxide analogues, which otherwise contribute to power losses.This device architecture remains the layout of choice for perovskite solar cells and its simple layout is one of the key reasons for the performance increases over the years.In addition, the field has evolved from MAPbI 3 as the absorber composition to FA-rich samples in which FAPbI 3 is alloyed with fractions of Cs, MA (A-site) and/or Br (X-site) [27,34,56], along with dramatic improvements in film quality and deposition methods [33] (Figure 10(c)).Finally, the efficiency enhancements have also been driven by significant work on optimising interfaces with the charge-collecting p-and n-type collecting layers as well as the choice of charge collecting materials.
Perovskite solar cells operate in n-i-p or p-i-n structures where the perovskite itself acts as a low-doped, nearintrinsic (i) semiconductor absorber layer sandwiched between n-and p-type contacts [57].This operation differs from typical p-n junction solar cells such as silicon where a depletion region and built-in field assist in separating charges across the junction; indeed, consensus is that the ionic nature of perovskites screens such fields and so resulting fields across the absorber layer are low in magnitude [58], and diffusion rather than drift dominates charge collection.The highest performing cells have n-i-p architectures where the perovskite is deposited on an n-type selective collecting layer for electrons (typically high-quality SnO 2 deposited on fluorinated tin oxide, FTO, on a glass substrate), and then a p-type collecting layer is deposited on top (typically a doped layer of small molecules such as Spiro-OMeTAD or conducting polymer), followed by metal electrodes (Figure 11(a), inset) [59].A p-i-n 'inverted' architecture is also growing in popularity, with the perovskite typically deposited on a carbazole-based self-assembled monolayer (e.g.2-PACz) that is anchored onto an indium tin oxide (ITO) layer on glass, and an electron-collecting layer such as Figure 10.(a) The first perovskite solar cells adopted a dye-sensitised solar cell configuration in which the perovskite is adsorbed on a mesoporous metal oxide scaffold, with the perovskite acting solely as the absorber.Reprinted with permission from [12].Copyright 2009 American Chemical Society.(b) The evolution of sensitised solar cells (left to right), from mesoporous dye-sensitised configurations, to mesoporous perovskite solar cells, and finally to planar heterojunction perovskite solar cells.Reprinted with permission from [54].
Copyright 2013 American Chemical Society.(c) Over time, through improved processing approaches and compositions, absorber films have become higher quality with careful defect control.Reprinted with permission from John Wiley and Sons [33].C 60 is deposited on top, followed by metal electrodes [60] (Figure 11(c)); operating principles are similar to the ni-p systems and efficiencies are almost on par with n-i-p structures [61].The record certified lab-cell power conversion efficiency for single junction perovskite solar cells to date is now 25.8% in an n-i-p architecture employing a high-quality FAPbI 3 -rich absorber layer [7,62] and a highly optimised SnO 2 n-type layer [59,62] (Figure 11(b  and c)), a remarkable rise from 3.8% in just over a decade.Such record cells employ a high-quality composition in which defects and interfaces have been well managed by passivation [62].For lower bandgap Pb/Sn systems, the record to date is 23.7% [63] (not yet independently certified at time of writing), again achieved through suppression of unwanted defects.
Perovskite-silicon tandem solar cells in which a wider bandgap halide perovskite sub-cell is combined with a silicon solar cell show exceptional promise for high performance (Figure 12(a)) [36,64,65], with the sub-cells joined by an appropriate recombination (tunnel) junction.These devices typically employ a p-i-n perovskite sub-cell.Recently, De Wolf and co-workers reported a certified record of 33.7% [7], an astounding efficiency which surpasses every other tandem PV technology to date.These tandem devices are showing high potential for commercialisation [66].A related architecture is an all-perovskite tandem cell in which two perovskite sub-cells are employed without the need for silicon [67].Typically, the low-bandgap cell is a Pb/Sn absorber ( ∼ 1.2 eV bandgap) and the wide bandgap cell is a FArich absorber with some mixed halides to increase the bandgap ( ∼ 1.6-1.8eV) (Figure 12(b and c)).The current record certified all-perovskite cell is 28.0%[68], but likely to continue to increase with further fast-paced research.A recent report of high-performance triplejunction all-perovskite solar cells also shows promise for even higher performance [69], although utilising ever more complicated device stacks.Projected practical efficiency limits for both all-perovskite and perovskitesilicon solar cells are ∼ 35%-38% [70], representing a very significant boost over the practical efficiency limits of silicon cells due to the Auger recombination limit at ∼ 29% [8], ushering in a new paradigm for highperformance PV.
Finally, operational stability of devices is an important consideration.Silicon PV technologies are typically underwritten for 25 years and so set a high bar for new PV technologies.It is impractical to consider such extended testing to validate a new technology before commercialisation, and thus accelerated testing including continuous operation, elevated temperature, and temperature/humidity cycling are considered to provide some assessment [72].The operational stability will ultimately depend on a combination of intrinsic processes in the devices and the quality of the packaging to prevent ingress of water and oxygen.The first perovskite solar cells were extremely unstable, with combinations of ion migration and degradation due to any exposure to moisture or oxygen leading to device instabilities when operating [73], including hysteresis between forward and backward J-V scans [74].The replacement of MAPbI 3 with more robust FA-or Cs-rich compositions has dramatically improved intrinsic material stability especially at elevated temperature and under operating conditions [27,56].Together with defect control, improved interfaces, and contact layers, the operational stability demonstrations of perovskite solar cells over recent years are extremely promising for longer-term operation, with single junction cells passing rigorous tests including longterm maximum power point operation (Figure 13(a))  and damp heat tests (Figure 13(b)), including in largerarea mini-module format (Figure 13(c)).Nevertheless, passing such tests does not necessarily validate longterm operation nor prove commercial viability, and a lot more work still needs to be done to stabilise longterm operational performance in a variety of devices and applications (see Section 4).

Enabling properties of halide perovskites for photovoltaic applications
Three particularly striking features stand out about halide perovskites materials when considered for optoelectronic -and specifically photovoltaic -applications.First, they show strong, tuneable absorption of light as direct bandgap materials.The absorption coefficient of MAPbI 3 is larger than established PV materials CdTe and GaAs across a large portion of the solar spectrum, with values of 10 5 cm −1 across a wide spectral range leading to strong absorption of many of the incident solar photons even with absorber layers only several hundreds of nanometres thick [78] (Figure 14(a)).Such strong absorption means device stacks including all contact layer can be thinner than 1 μm, compared to silicon which has an indirect bandgap requiring ∼ 200 μm of material to sufficiently absorb sunlight; this opens up applications in lightweight, high-performance perovskite PV that is not as feasible as with silicon.The tuneable bandgap allows the absorption edge to be changed and thus the absorbing colour, and thus allows them to be used in multi-junction cells discussed above, when coupled with multiple perovskites or other layers including silicon or CIGS [65], or in coloured or semi-transparent PV for example in building-integrated PV applications.
The absorption spectra of halide perovskites also show very low energetic disorder in the sub-bandgap region.The Urbach energy is a parameter that quantifies the energetic disorder in the band edges of a semiconductor and is extracted by fitting the absorption edge to an exponential function representing a tail of states.The steepness of the absorption edge therefore depicts the quality of the semiconductor -the sharper the absorption edge, the higher the quality.Halide perovskites exhibit a low Urbach energy of ∼ 13 meV, obtained from sensitive absorption measurements including photothermal deflection spectroscopy (Figure 14(b)) [79,82].This value is competitive with the highest quality PV materials in GaAs (∼ 8 meV [83]) and crystalline silicon ( ∼ 10 meV) [84].This reinforces the fact that halide perovskites are relatively 'clean' semiconductors even in spite of a moderately large defect density.
Second, absorption of incident light leads to free electrons and holes rather than strongly bound excitonic states that are difficult to dissociate.The exciton binding energy in halide perovskites has been reported through magneto-optic measurements to be ∼ 15 meV at low temperature (Figure 14(c)), with this value decreasing to negligible values at room temperature [80].Thus, excitons readily dissociate at room temperature (k B T ∼ 25 meV) into free electrons and holes that can be independently and effectively transported and collected at electrodes.This avoids the use of additional charge separation layers within the absorber layers that otherwise lead to further interfaces and energetic losses; for instance, additional layers are required for some other emerging PV materials such as organic semiconductors where exciton binding energies are several hundreds of milli-electronvolts and free charges are not readily generated without specially designed charge separating interfaces.
Finally, these photo-generated electrons and holes can be efficiently collected by device electrodes in planar thin film device structures, with minimal unwanted nonradiative recombination of carriers.The diffusion length of photo-excited electrons and holes in halide perovskites -the average distance charge carriers travel before recombination events -was shown in 2013 to be over a micrometre in MAPbI 3 -like perovskites (Figure 14(d)) [81].In high quality films and single crystals these values can be multiple microns [85,57].This result means that most carriers, when photo-generated on one side of the device film, will be able to traverse the ∼ 500-nm thick absorber layer to be collected at an electrode before recombining, thus contributing to photocurrent (for example maximising short-circuit current).This allows the use of simple planar heterojunction device stacks for efficient cells.This situation is in contrast to more complicated bulk heterojunctions of intermixed materials or networks of disordered metal oxide particles required for efficient charge collection in other emerging PV materials including organic or dye-sensitised solar cells, respectively, in which charge transport distances are much less than the absorber thicknesses (cf. Figure 10(b)); indeed, this result renders halide perovskites as true thin film PV technologies.
The low level of non-radiative recombination in halide perovskites leads to high photo -and (PL) electroluminescence (EL) quantum efficiencies [86], which is a prerequisite for high photo-voltages in solar cells [46] (cf.Equation 8); non-radiative recombination is an avoidable loss in which photo-excited carriers lose their energy to heat.Peak EL efficiencies in operating solar cells have been reported to be > 10% [87], rivalling those of the most efficient solar cells in GaAs ( ∼ 30%) and approaching the limits when considering outcoupling of photons from thin film device stacks [88], with internal luminescence quantum efficiencies approaching 100%.
The long diffusion lengths and high radiative efficiencies in halide perovskite absorber materials are attributed to the low levels of deep trap states.Many point defects that form even in very high densities, such as halide vacancies, do not introduce deep states in the material and rather introduce states either within the bands or shallow states just near the bands [89,90], and thus carriers either do not trap within them or, if they do, the carriers can thermally de-trap back to the bands (Figure 15(a)).Meanwhile, defects that lead to deep trap states typically have very high formation energies and thus appear in low densities (Figure 15(b)) [91].Thus, although defect densities can be large ( ∼ 10 16 cm −3 ), a large fraction of them are relatively benign [90].This is contrasted with traditional semiconductors such as crystalline silicon or GaAs in which such high defect densities would be catastrophic for device performance; these materials are thus grown with extreme precision and temperature to keep defect densities low ( < 10 10 cm −3 ) [92].The performance of halide perovskite PV cells approaching these traditional semiconductors is particularly remarkable when considering they are processed at low temperature by low-cost solution or vapour meansrelatively crude methods compared to highly specialised and expensive molecular beam epitaxial growth of GaAs, for example.It is even more remarkable when considering the vast spatial heterogeneity in chemical, structural, morphological and optoelectronic properties in halide perovskites [93,94] -with properties varying even in the best devices on length scales from millimetres scales down to nanometre scales (Figure 15(c)), but not leading to high densities of unwanted recombination sites.This tolerance to spatial disorder and defects is a key property for their performance [95], and fascinating from both a fundamental scientific and technological perspective.This overall tolerance -in that defects formed are generally not problematic -renders them much more radiation-hard against high energy protons and other particles seen in space than more traditional PV materials such as GaAs, showing very good promise for use in lightweight, high-performance space photovoltaic applications [96].

Photophysical understanding of halide perovskites
In general, the recombination of charge carriers in a 3D halide perovskite absorber layer (and indeed any semiconductor in which recombination of free electrons and holes dominates) is given by: where n is the electron charge carrier density (for simplicity it is here assumed that electron density equals hole density), k 1 is the first-order Shockley-Read-Hall (SRH) trapping (non-radiative) rate constant, k 2 is the secondorder band-to-band (radiative) recombination rate constant, and k 3 is the third-order Auger (non-radiative) recombination rate constant [98] (Figure 16(a)).The photoexcited carrier density dictates which of the terms dominate recombination: in MAPbI 3 films, the k 1 term typically dominates at photoexcitation densities < 10 15 cm −3 , k 2 dominates in the range ∼ 10 15 -10 17 cm −3 , and k 3 dominates in the regime > 10 17 cm −3 [86,99].This model is able to describe time-resolved PL measurements in thin films, where at low carrier densities recombination is quasi-first-order dominated by the k 1 term but at higher carrier densities it deviates from this as second-order recombination dominates [100,101] (Figure 16(b)).Global fits to such data allow extraction of, among other quantities, the trap density -in this case for MAPbI 3 this is ∼ 10 16 cm −3 , though for well passivated samples these values can be < 10 14 cm −3 [98].The increasing dominance of the bimolecular recombination regime as carrier density increases is also reflected in increasing photoluminescence quantum efficiency (PLQE) -the ratio of the radiative processes of the rate equation to the total (radiative and non-radiative) processes -up until the point Auger recombination dominates and PLQE drops again (Figure 16(c)).Other more extended models that include carriers remaining in trap states for long periods, including possibility to de-trap back to bands, which leads to electron and hole populations in general being non-equal, have also been proposed to describe the carrier populations in halide perovskites, providing good agreement with experimental data [102,103].Whilst many perovskite PV absorbers are either intrinsic or only weakly doped, the lower bandgap Pb/Sn systems can be p-type and thus recombination processes are typically quasi-first-order, limited by the minority carriers (electrons) [30,32].In more complicated alloyed Pb-based alloyed systems with mixtures of A-site FA/Cs and/or MA and/or mixtures of X-site halides Br/I, photo-excitation initially yields bimolecular recombination of electrons and holes, but holes rapidly accumulate in high concentrations on local regions, leading to a photo-doping effect in which the radiative recombination is limited by electrons, becoming quasi-firstorder after tens of nanoseconds [104].When a radiative recombination event occurs, the emitted photons can be reabsorbed in the material  9. Adapted with permission from the American Physical Society [100].(c) PL quantum efficiency as a function of excitation density, highlighting the different recombination regimes.Reprinted with permission [86].Copyright 2017 American Chemical Society.(d) Calculated number of photon recycling events as a function of thickness of the absorber at maximum power (N mpp ) and open-circuit (N OC ).Adapted with permission from the American Physical Society [106].
before being emitted from the film, thus leading to further recombination events.Many of such recombination/reabsorption events can occur before the photon is out-coupled from the film, a phenomenon termed photon recycling, and is particularly prolific in films in which the radiative recombination efficiency is high; the effect has been widely reported in III-V PV materials [45].This effect is significant in halide perovskite PV absorbers in which there is strong overlap between the emission and absorption spectra [105], with ∼ 10 photon recycling events calculated to be occurring per absorbed incident photon at open circuit [106] (Figure 16(d)).Such a phenomenon maintains high carrier densities in the absorber films and can lead to higher open-circuit voltages in solar cells than in the absence of photon recyling [45].Photon recycling and related effects that delay or impede the escape of photons from the film such as waveguiding also complicate the determination of the bimolecular recombination constant k 2 in the models described earlier; one therefore typically distinguishes the bimolecular recombination constant measured through such means as an 'external' k 2 value, from the intrinsic 'internal' bimolecular recombination constant that can be computed by accounting for photon recycling and out-coupling of photons from the film (including effects such as waveguiding) [88].
Charge-carrier diffusion processes compete with recombination processes.Indeed, the charge-carrier diffusion length -a critical parameter for determining charge carrier collection efficiency in perovskite solar cells -depends on the ratio of charge-carrier mobility and recombination rate [108].The intrinsic mobility of halide perovskites such as MAPbI 3 is influenced by interactions between electrons and lattice vibrations (phonons), as indicated by the charge-carrier mobility in MAPbI 3 increasing strongly with decreasing temperature due to suppression of interactions between electrons and phonons as the phonon mode occupancy reduces [109] (Figure 17(a)).Specifically, Fröhlich interactions, involving the macroscopic electric field generated by a longitudinal optical (LO) phonon, have been found to be the dominant mechanisms for charge transport at room temperature, whereas acousticphonon deformation-potential scattering is relatively weak for MAPbI 3 [108].The dominance of Fröhlich coupling was ascertained through assessments of the temperature-dependent emission broadening in FAPbI 3 [110], consistent with ab-initio calculations including Figure 17.(a) Mobility as a function of temperature determined from optical-pump -THz probe measurements (black dots), with the dashed line representing the theoretical T −3/2 dependence predicted for band-like transport.Reprinted from [109].(b) 3D images as a function of time after excitation visualising carrier diffusion vertically through the film, highlighting carriers in regions of poor vertical diffusion first diffusing laterally and then vertically through higher quality regions, thus leading to high charge collection efficiencies at the opposite electrodes.Adapted with permission from Springer Nature [113].
electron-phonon coupling being reported to be significantly reduced below the characteristic temperature corresponding to the energy of the relevant LO phonon (11 meV) [111].Mobilities reported for halide perovskites are ∼ 1-100 cm 2 /V/s measured through contactless methods including terahertz and microwave conductivity spectroscopy and contact methods including Hall effect, space-charge limited current and use of fieldeffect transistors [108].Such values are near the intrinsic limits for halide perovskites and thus exhibit intermediate mobilities, much lower than GaAs ( 1000 cm 2 /V/s) but higher than other emerging materials including organic semiconductors ( 0.1 cm 2 /V/s).This limited intrinsic mobility is rationalised due to the enhanced Fröhlich coupling for halide perovskites compared to GaAs [112].
Extrinsic effects including material imperfections, such as grain boundaries, energetic disorder, or impurities, also further influence carrier transport parameters and depend on material processing and composition [108].Local diffusion measurements through microscopy luminescence measurements reveal that, even in highly efficient solar cells, the local 'vertical' diffusion length through the depth of the film varies wildly across the microscale from 0.01 to 0.3 cm 2 /V/s [113] due to morphological variations.However, charge collection efficiencies remain high because charges in regions of poor diffusion can diffuse laterally inter-grain to regions of better diffusion before diffusing down to the opposite electrode for collection (Figure 17(b)).Thus, despite only moderate mobilities, the interplay between transport and recombination rates of halide perovskites does not limit thin film perovskite PV device performance provided overall deep trap densities are suitably low.

Ongoing challenges and opportunities
Halide perovskites show enormous promise for photovoltaic applications, but a number of grand challenges must be solved to see them making a sizable contribution to decarbonisation as quickly as possible, and to realise their full range of applications.
A key challenge is to realise the highest efficiencies particularly when considering the different bandgaps needed, for example, for tandem solar cells; the highest efficiencies have been achieved for single junction cells of bandgap ∼ 1.5-1.6 eV, but efficiencies for bandgaps of ∼ 1.6-1.8eV and ∼ 1.2-1.3eV still lag behind their intermediate bandgap counterparts.Non-radiative recombination losses still remain across all bandgaps, with headroom to further increase luminescence quantum efficiencies towards the levels of GaAs.When viewed on the microscale, the PL is spatially heterogeneous even in high-quality perovskite devices, with dark regions in PL corresponding to non-radiative power losses [114] (Figure 18(a)).These dark regions in FA-rich samples have been attributed to nanoscale trap clusters with deep lying states [115] that relate to local phase impurities such as hexagonal polytypes [116]; the formation of these unwanted phases can be inhibited somewhat through inducing slight octahedral tilt in the perovskite structure [117].Judicious use of passivation agents has been essential to mitigating non-radiative recombination pathways by managing defects.Two key promising routes have emerged in the field: use of ammonium salts to form 2D perovskite surface layers on the 3D absorber surface [76,118], and use of Lewis bases to passivate interfaces and grain boundaries throughout the bulk and surfaces of the absorber [75,119].Passivation will increase the PL quantum yield and, if considering low carrier densities below the trap density, will increase the PL lifetime (Figure 18(b)).
Even once defect densities in the absorber layer itself are minimised, further losses may be expected from interfaces with selective contact layers.Many contact layers lead to drops in open-circuit voltage (or other metrics) compared to what would be expected from the quasi-Fermi-level splitting of the absorber [120] (Figure 18(c)), and thus further management (e.g. through passivation) or design of suitable charge-collecting layers will be essential.Any passivation approaches must manage defects without blocking charge collection [118].In traditional semiconductor systems, interfaces and carrier extraction are also optimised through carrier doping, but this is a much greater challenge in halide perovskites where the effect of any dopant is compensated by ionic rearrangement in the materials; few reports have managed to provide controlled doping at such interfaces [121].Finally, the use of interlayers that not only passivate defects but also lead to strong adhesion of layers, preventing mechanical fracture points and delamination during long-term operation [122], will be very valuable to the community.Further work in the field on passivation and doping treatments will be essential for further driving down power losses particularly in the wide range of bandgaps under consideration.
Arguably the biggest grand challenge for halide perovskites photovoltaics is operational stability.The ionic nature of halide perovskites leads to facile ion migration under light and/or bias [123,124], presenting a challenge for stable operation.Halide ions in particular are the most mobile, proceeding through vacancy-assisted migration [125], whereas metal and A-site cations have higher activation energies for transport.This movement of ions within the semiconductor leads to ions accumulating at charge-collection interfaces, impacting charge collection; the fact these ions move under bias leads to variable hysteresis in the J-V curve as ions may have different distributions through the film and at interfaces depending on the scan conditions (i.e.scan rate, scan directions etc.) [74,126].Such hysteresis effects are less prevalent in the most stable devices employing robust compositions and high-quality contact layers, suggesting that appropriate interface and bulk management can mitigate any negative impacts of ion migration, even if the phenomenon still occurs.Furthermore, the halide ions are readily oxidised in the presence of photo-excited holes [127], leading to redox reactions that induce both halide movement but also degradation reactions; for example, oxidation of iodide to atomic iodine leads to formation of iodine gas I 2 , which leaves the material leading to material loss and device performance reduction [115,127].Such processes have been shown to seed at nanoscale phase impurity sites in FArich absorbers in which iodide defects are particularly high in density (Figure 19(a)), further highlighting the need to inhibit formation of these unwanted phases.Again, mitigation of these reactions can be achieved through appropriate passivation to control defect densities [128].The use of mixed halides (e.g.Br and I) to achieve wider bandgaps for tandems represents a further challenge as the halides segregate under operation, leading to distinct iodide-rich regions in the film, thus dynamically changing the bandgap (Figure 19(b)).Films with bromide contents relative to iodide at < 20% can in general be stabilised, but halide segregation generally inhibits the use of higher Br fractions to achieve wider bandgaps, thus representing a challenge to reach the ideal bandgaps for tandem solar cells that cannot be attained with pure FA/iodide compositions ( ∼ 1.7-1.8eV).This photo-induced halide segregation has been proposed to proceed through the oxidation of halides [129], though other mechanisms including polaron formation [130] and composition/electronic/structure-linked band offset differences [131] are also under active discussion.Any passivation and/or interface optimisation approach to mitigate these effects must be suitably robust for longterm operation, including accounting for the thermal generation of further halide defects, and any defects appearing at interfaces.Machine learning approaches will be important parts of the tool kit to ultimately understand and even predict stability pathways in materials and full devices [132,133].Finding appropriate universal passivation agents that both improve performance and stability, as well as elucidating fundamental degradation pathways, remain active areas of study for the field.
Stabilising desired phases and bandgaps remains a challenge in the field.The most promising absorbers for efficiency and stability are FA-rich, with fractions of Cs and/or MA.Alloying of the A-site cation is essential to stabilise the black absorbing FAPbI 3 phase at room temperature as it induces octahedral tilt to prevent conversion to non-absorbing phases [117].Even in the most efficient FA-rich samples, there is still some alloying (e.g. with small amounts of MA through addition of MAClmuch but not all of which is removed as a volatile byproduct) or octahedral tilt induced from other additives; these are thus not pure cubic α-FAPbI 3 , but rather tetragonal phases [87,117].Likewise, CsPbI 3 has a very desirable bandgap for tandems ( ∼ 1.7 eV [27]) but is notoriously difficult to stabilise in its black absorbing phase at room temperature.Some success has been found through interfacial management and inducing strain in the material [137].Efforts to stabilise the phase and operation of pure FA -and Cs-perovskites remains a key focus of the field.Successful approaches might include use of other additives and templating agents able to induce and retain formation of stabilise phases, as well as judicious use of lower-dimensional phases such as 2D perovskites [137].
A key challenge for the field is to define appropriate standards for operational stability testing especially for accelerated testing.Other PV-industry standard tests defined by the International Electrotechnical Convention (IEC), such as 1000 h at maximum power point or damp heat testing, are also useful in assessing initial stability, but these are typically used in the PV industry for individual module quality control and not for assessing a new technology as a whole, so they on their own are not sufficient to validate long-term operation.Indeed, many bespoke tests need to be defined for the perovskite technology itself.The community has defined several standards inspired by earlier standards defined for the organic PV community [72], based on Summit on Organic Photovoltaic Stability (ISOS) protocols, making an excellent base, but these will need to continue to be refined, strengthened, and broadened as the field evolves.
For instance, several degradation pathways found for silicon solar cells were only discovered after many years of field operation, allowing feedback to cell and module design only many years after initial fabrication.One example is the reverse bias test, a phenomenon which can occur when a cell is partially shaded in the field, which thus far for halide perovskites presents an unresolved stability challenge [138].This motivates the need for both standardised indoor but also outdoor testing of perovskite solar cells to capture the variation in performance seen in outdoor conditions.At present, only a few reports describe outdoor tests (e.g. Figure 19(c)) and more will be needed as the field further matures.
Another key challenge is to translate the record cell efficiencies, achieved at lab scale (typically < 1 cm 2 ), to larger areas relevant to mini-modules and ultimately full modules.As cell area increases, the efficiency tends to decline, a phenomenon that is universal to all solar cell technologies; for instance, whilst lab record silicon cells are ∼ 27%, the best modules in the field operate at ∼ 20%-23%.The decrease in efficiency with scaling in the case of perovskites is primarily due to the increase in series resistance caused by sheet resistance of the transparent electrode as the area expands, as well as geometric fill factor losses as some fraction of the module is necessarily covered by electrodes.To mitigate this efficiency loss, the structure of perovskite solar modules needs to be redesigned to minimise the impact of series resistance.Scaling up perovskite films and implementing scalable deposition methods for the electron-and holeselective transport layers, as well as back electrodes, will be critical for the successful scaling of the perovskite PV technology [139].Several scalable solution-and vapourbased techniques have been developed to deposit highquality perovskite layers [140], where a key target is to control the morphology and quality of perovskite layers over multiple length scales (nanoscale up to metre scale).Another key target is to ensure reproducible depositions within and between batch processing -a challenge to date at lab-level due to sensitivity of the perovskites to many local processing conditions [141], especially in typical academic labs where multiple processes, materials and solvents are being utilised in the same lab or glove box simultaneously.Nevertheless, notable accomplishments in the scale up of perovskite PV across different sizes include 16.1% for small modules with an area of 804 cm 2 by Panasonic Corporation (Osaka) in 2020 [142], 20.2% for modules with an area of 20.0 cm 2 by Microquanta Semiconductor in 2021 [143], and Oxford Photovoltaics in 2023 certifying an efficiency of 28.6% for a perovskitesilicon device on a full commercial-sized silicon solar cell [144].These results show promise for the ongoing commercialisation of this nascent technology [145].
Continued efforts are also needed to develop effective and robust encapsulation techniques that are costeffective to ensure the long-term durability of perovskite solar modules under operational conditions -both for rigid substrates but also barrier films compatible with flexible and lightweight applications.This will be important not only for protecting the device stacks from the elements, but also for containing any potential lead leakage throughout the operational lifetime of a module.Whilst lead is present, it is in trace amounts in the very thin absorber layers.There is precedent for heavy metals in PV -with more Pb in the solder of silicon cells than in perovskite panels, and cadmium is a key component of extensively commercialised CdTe solar cells -meaning that appropriate management can allow these technologies to be widely deployed.Recent promising avenues include incorporation of lead sequestration agents in the device stacks and/or encapsulating glass [146] that immobilise lead in insoluble forms at any point of module failure.Nevertheless, strategies to recover and recycle the modules at end of life, including full life cycle assessments [147], need further exploration and may form a key part of sustainable deployment models of perovskite PV.
Finally, there is a need to continue to understand the properties of halide perovskites more fundamentally, that will in turn inspire new science and engineering in this space.One aspect that is not suitably understood is strain [148]; while it is known that some aspects of strain can be detrimental in halide perovskites, controlling and exploiting strain may enable far more stabilised materials but also more exotic light-harvesting applications potentially exploiting, for example, the flexophotovoltaic effect in which strain induces a bulk photovoltaic effect [149].Furthermore, the tolerance of halide perovskites to heterogeneity is remarkable and can act as a blueprint for other semiconducting materials with similar properties -be they perovskites, perovskite-inspired or unrelated materials.Machine learning will provide further resource to screen wider material families in this regard [150].These efforts may realise an even broader family of materials suited for low-cost but efficient photovoltaics, as well as other optoelectronic applications including detectors and LEDs.

Conclusion
Halide perovskites are remarkable semiconductors that are pushing the efficiency records for both singlejunction and tandem solar cells whilst being compatible with scalable and inexpensive deposition routes.A sustained push to understand these materials, as well as their performance and stability limitations, will lead to exciting innovations in a technology poised to play a crucial role in decarbonising electricity.
a not-for-profit developing education for school-age children around climate change solutions.

Figure 1 .
Figure 1.(a) Cost of PV module over time in 2017 US dollars.Copyright IOP Publishing.Reproduced with permission.All rights reserved [5].(b) Breakdown of the cost of a benchmark residential PV system in the USA in Q1 2022.Image adapted from [10].

Figure 2 .
Figure 2. (a) ABX 3 perovskite crystal structure.Adapted from [11].(b) Powder X-ray diffraction pattern and fitting of MAPbI 3 revealing the I4/mcm space group.Adapted with permission from the Royal Society of Chemistry [17].(c) Electronic band structure and partial density of states of MAPbI 3 in the low-temperature orthorhombic phase calculated by density functional theory (DFT).Adapted with permission from Springer Nature [18].

Figure 3 .
Figure 3. (a) Bandgap tunability through tuning the I/Br ratio in FAPbI 3 .Adapted with permission from the Royal Society of Chemistry [27].(b) Deepening of the valence band when moving from iodide to bromide perovskites.Adapted with permission from the Royal Society of Chemistry [28].(c) Bandgap tunability through Sn-based perovskites.Reprinted with permission from Springer Nature [29].

Figure 4 .
Figure 4. (a) Spin-coating halide perovskite thin films from precursor solutions.Inset: the initial as-cast film (left) is annealed at 100°C to form a highly crystalline, strongly absorbing film.(b) Vapour deposition of halide perovskite by thermal sublimation of the precursor salts in high vacuum.Reprinted with permission from [41].Copyright 2020 American Chemical Society.(c) A photograph of an operating roll-to-roll machine coating perovskite layers.Reprinted with permission from [39].Copyright 2018 American Chemical Society.

Figure 5 .
Figure 5. (a) Silicon solar cells linked together into a module, which can in turn be combined into an array.Thin film technologies can be fabricated monolithically directly into modules.(b) A typical solar PV module.Copyright IOP Publishing. a and b reproduced with permission.All rights reserved.Also adapted with permission from the Royal Society of Chemistry [5].The general operation of a solar cell involves (c) absorption of incident light to generate photo-excited electrons and holes and (d) transport of these charges to electrodes to generate photo-voltage and photo-current to do work on an external circuit.

Figure 6 .
Figure 6.(a) Current density-voltage (J-V) curves of a solar cell allowing extraction of important metrics, with the maximum power shown in shaded region.(b) Equivalent circuit model of an ideal solar cell, with additional elements to represent a non-ideal solar cell shown in red.

Figure 7 .
Figure 7. (a) Standard solar spectrum representing solar irradiance for terrestrial (blue, AM1.5, 1000 W/m 2 ) and space (AM0, 1366 W/m 2 ).Data from [43].(b) Schematic showing that for a finite bandgap how photons with energy less than the bandgap energy are not absorbed by the material, but photons above the bandgap are absorbed but excess energy lost to heat as carriers relax to the band edge before being transported.(c) Photo-generated current density increases with decreasing optical bandgap.(d) The maximum open circuit voltage is defined by the quasi-Fermi-level splitting of the material, and scales with bandgap.(e) The maximum theoretical efficiency of a solar cell as a function of bandgap.Image from Wikipedia (CC BY-SA), data calculated from adaptions to [44].

Figure 8 .
Figure 8. Lab record certified single junction and perovskitesilicon tandem solar cell efficiencies across different established PV technologies.Uncertified cells are denoted with hollow symbols.Data from [7].
(b)) or connected in series in a monolithic stack (2-terminal) where currents

Figure 9 .
Figure 9. (a) Conceptual schematic of a tandem solar cell with wide and low bandgap sub-cells, harvesting complementary regions of the solar spectrum.Comparison of a (b) 4-terminal (4 T), with independent electrical connection to both cells, and (c) 2terminal (2 T) series-connected tandem concepts.b and c adapted with permission from John Wiley and Sons [50].

Figure 11 .
Figure 11.(a) J-V curve and (b) external quantum efficiency (EQE) of a champion state-of-the-art n-i-p perovskite solar cell.Adapted with permission from Springer Nature[62].Inset of a: representative device architecture.Adapted with permission from John Wiley and Son[59].(c) Architecture of p-i-n perovskite solar cells, with a carbazole-based SAM layer anchored onto the ITO.Reprinted with permission from the Royal Society of Chemistry[60].

Figure 12 .
Figure 12.(a) Schematic of a perovskite/silicon solar cell.Adapted with permission from Scientific American, a division of Springer Nature America[36].(b) All-perovskite solar cell with narrow bandgap (NBG) and wide bandgap (WBG) perovskite absorbers, yielding (c) high performance J-V characteristics.Adapted with permission from Springer Nature[71].

Figure 13 .
Figure 13.(a) Stability test of a perovskite solar cell under continuous maximum power point tracking and full sunlight with 1,3bis(diphenylphosphino)propane (DPPP) passivation treatment.Adapted with permission from the American Association for the Advancement of Science (AAAS) [75].(b) Damp heat test (continuous operation at 85°C and 85% relative humidity) of perovskite solar cells with 2D passivation.Adapted with permission from the AAAS[76].(c) Perovskite mini-module (photograph in inset) under continuous operation near maximum power point.Adapted with permission from Springer Nature [77].

Figure 14 .
Figure 14.(a) Absorption coefficient of various PV materials.Reprinted with permission from Springer Nature [78].(b) Absorption edges and assessment of Urbach energy for perovskite, GaAs and c-Si absorbers.Reprinted with permission from [79].Copyright 2014 American Chemical Society.(c) Magneto-optic measurements of halide perovskites at low temperature, allowing extraction of exciton binding energy R * and effective mass m * .Reprinted with permission from Springer Nature [80].(d) Time-resolved PL quenching measurements, which are fitted with a diffusion model to extract diffusion coefficients (D) and diffusion lengths L D .Adapted with permission from the AAAS [81].

Figure 15 .
Figure 15.(a) Schematic showing that many of the defects states in halide perovskites are within the bands, or are energetically shallow.Adapted with permission from the AAAS[97].(b) Calculated defect formation energies under iodide-rich conditions for various defects in halide perovskites, where many of the deep defects have high formation energies.Reprinted with permission from[91].Copyright 2015 American Chemical Society.(c) Electron microscope images at different length scales demonstrating the remarkable heterogeneity in morphological and crystal properties even in high-performing halide perovskite films.Reprinted with permission from Springer Nature[93].

Figure 16 .
Figure 16.(a) Schematic summarising recombination processes in Equation 9. Reprinted with permission from [107].Copyright 2020 American Chemical Society.(b) Time-resolved photoluminescence decays showing the different recombination regimes, with solid lines corresponding to fits to the data to the recombination model in Equation9.Adapted with permission from the American Physical Society[100].(c) PL quantum efficiency as a function of excitation density, highlighting the different recombination regimes.Reprinted with permission[86].Copyright 2017 American Chemical Society.(d) Calculated number of photon recycling events as a function of thickness of the absorber at maximum power (N mpp ) and open-circuit (N OC ).Adapted with permission from the American Physical Society[106].

Figure 18 .
Figure 18.(a) Confocal PL map of a MAPbI 3 film, demonstrating the spatial heterogeneity in luminescence efficiency.Adapted from [86].(b) Time-resolved PL lifetimes at low fluence, showing the increase in lifetime (decrease in first-order trapping rate) with DPPP passivation.Adapted with permission from the AAAS [75].(c) Quasi-Fermi-level splitting (QFLS) of perovskite films with various contact layers, demonstrating the non-radiative losses associated with the interfaces.Adapted from [120].

Figure 19 .
Figure 19.(a) Local scanning electron diffraction images for an FA-rich absorber before (left) and after (right) extended illumination in high vacuum.The local diffraction patterns allow assignment of polytype phase impurities (shaded yellow), which are the sites that degrade.Adapted with permission from Springer Nature [134].(b) Illumination over time of a mixed halide perovskite sample leads to halide segregation, with distinct PL peaks corresponding to low-bandgap iodide-rich impurities.Adapted with permission from the Royal Society of Chemistry [135].(c) Outdoor testing of perovskite-silicon tandem solar cells over a 1-year period in KAUST, Saudi Arabia.Power Generation Density (BGD).Reproduced with permission from Elsevier [136].