You are on page 1of 5

commentary

The light and shade of


perovskite solar cells
Michael Grätzel
The rise of metal halide perovskites as light harvesters has stunned the photovoltaic community. As the
efficiency race continues, questions on the control of the performance of perovskite solar cells and on its
characterization are being addressed.

A
lkali-metal lead and tin halides analogue CH3NH3PbBr3 as sensitizers
had been synthesized already in for liquid-electrolyte-based DSSCs13,
18931, yet the first crystallographic the power-conversion efficiency (PCE)
studies that determined that caesium lead measured under standard characterization
halides had a perovskite structure with the conditions reached a mere 3.8%, and the
chemical formula CsPbX3 (X = Cl, Br or I) device stability was poor due to the rapid
were only carried out 64 years later by the dissolution of the perovskite in the organic
Danish scientist Christian Møller 2. He also solvent. However, by changing both the
observed that these coloured materials electrolyte formulation and the method
were photoconductive, thus suggesting of depositing the perovskite, the group
that they behave as semiconductors. In of Nam-Gyu Park was able to increase
1978, Dieter Weber replaced caesium with the device performance and stability
methylammonium cations (CH3NH3+) attaining a PCE of 6.5%14. The authors
to generate the first three-dimensional noted that CH3NH3PbI3 was a superior
organic–inorganic hybrid perovskites3,4. light harvester compared with the more
The general crystal structure of these commonly used molecular N719 ruthenium
Figure 1 | Crystal structure of cubic metal halide
materials is shown in Fig. 1. Caesium sensitizer, meaning that a thinner absorbing
perovskites with the generic chemical formula
ions or small organic cations such as layer was sufficient to obtain good
ABX3. Organic or inorganic cations occupy
methylammonium and formamidinium light-conversion performance.
position A (green) whereas metal cations and
occupy the cuboctahedral voids formed A key advance was subsequently made by
halides occupy the B (grey) and X (purple)
by the 12 nearest-neighbour halide ions. replacing the liquid electrolyte with a solid-
positions, respectively.
The structural characterization of hybrid state hole conductor (or hole-transporting
perovskites of the formula CH3NH3BX3 material, HTM), that is, CsSnI3–xFx (ref. 15)
(with B = Sn(ii) or Pb(ii), and X = Cl, a timescale of hundreds of nanoseconds, or spiro-MeOTAD16,17. Not only did the
Br or I) has also been reviewed in more resulting in long carrier-diffusion conversion efficiency double, but the cell
recent studies5. lengths — that is, the average distance stability also improved greatly as a result
Methylammonium lead iodide, that can be covered by carriers before they of avoiding the use of a liquid solvent. At
CH3NH3PbI3, has both interesting optical recombine — ranging between 100 nm this stage, the embodiment of the PSC was
and electronic properties that have been and 1,000 nm (refs 11,12). Despite some an exact mimic of the solid-state DSSCs
actively investigated during the past of these appealing properties that were (Fig. 2). The N719 dye15 or the CH3NH3PbI3
two decades5–7. It is a semiconducting already known for more than 20 years, perovskite nanoparticles16,17 assume the
pigment with a direct bandgap of 1.55 eV the extraordinary potential of hybrid role of the sensitizer injecting electrons
corresponding to an absorption onset of perovskites in photovoltaic applications in a mesoscopic TiO2 scaffold and holes
800 nm (ref. 6), which makes this material was only revealed less than 5 years in a solid-state HTM. Both the TiO2 and
a good light absorber over the whole visible ago by researchers working on liquid- the HTM act as selective contacts through
solar emission spectrum. The excitons electrolyte-based dye-sensitized solar cells which the charge carriers produced
produced by light absorption have a (DSSCs)13,14. These two papers and the by photoexcitation of the perovskite
weak binding energy of about 0.030 eV, three publications15–17 that followed in 2012, nanoparticles are extracted.
which means that most of them dissociate reporting on the use of tin or lead iodide A change in paradigm occurred with the
very rapidly into free carriers at room perovskites in a solid-state version of the discovery by Lee et al.17 that a mesoporous
temperature8. The electrons and holes DSSC, set off the current meteoric rise of scaffold made of Al2O3 instead of TiO2
produced in this material exhibit a small perovskite solar cells (PSCs). produced similar — if not better —
effective mass9 resulting in high carrier conversion efficiencies, even though Al2O3
mobilities that range from 7.5 cm2 V–1 s–1 for Ascension of perovskite solar cells is unable to assist in electron extraction due
electrons9 to 12.5 cm2 V–1 s–1 – 66 cm2 V–1 s–1 In 2009, when Kojima et al. introduced to its large bandgap. This suggested that the
for holes10. Their recombination occurs on CH3NH3PbI3 and the larger-bandgap perovskite itself transported the electrons

838 NATURE MATERIALS | VOL 13 | SEPTEMBER 2014 | www.nature.com/naturematerials

© 2014 Macmillan Publishers Limited. All rights reserved


commentary

to the front collector electrode composed


of a compact blocking layer of TiO2
h+
deposited on a fluorine-doped tin oxide Sensitizer
conducting glass (Fig. 3a). These researchers TiO2
proposed that the electrons percolate along HTM
h+ Blocking layer
the surface of the Al2O3 nanoparticles
TCO
through an ultrathin conformal coating e– Back contact
of CH3NH3PbI3. However, until now, the
role — if any — of the mesoscopic Al2O3
scaffold in these devices, for which the term
‘meso-superstructured’ has been coined, e– Glass substrate (front contact)
remains obscure.
At the same time, it became clear from
the studies of Etgar et al.18 that CH3NH3PbI3 Figure 2 | Schematic cross-section of a solid-state mesoscopic solar cell using N719 or CH3NH3PbI3 as a
could also assume the role of a hole sensitizer. The perovskite nanoparticles are precipitated from solution onto a mesoscopic TiO2 film acting
conductor, obviating the need for employing as an electron-extraction layer. Spiro-MeOTAD was mostly used as an HTM16,17. The external electric
an additional HTM. Although the initial circuit contacting the solar cell is shown in black. In the close-up view on the left, the light radiation
conversion efficiency for these ‘HTM-free’ (yellow wavy arrow) hitting the sensitizer generates electrons (e–) and holes (h+) that are transported in
architectures was only about 5%, it has the TiO2 and HTM layers, respectively. TCO, transparent conducting tin oxide layer.
been growing steadily over the past two
years. Today, PSC embodiments based on
HTM-free mesoscopic CH3NH3PbI3/TiO2 common solvent such as γ-butyrolactone, γ-butyrolactone, dimethylsulphoxide
heterojunctions that use a double layer of dimethylformamide or dimethylsulphoxide. and toluene to control the nucleation
TiO2 and ZrO2 as scaffold, and carbon as a However, this approach typically led to of CH3NH3PbI3 crystals within the
back contact have reached a certified PCE of uncontrolled morphological variations, mesoporous TiO2 scaffold and the
12.8% and — contrary to many HTM-based which resulted in poor reproducibility subsequent growth of a dense CH3NH3PbI3
PSCs — exhibit excellent stability under of photovoltaic performance. To gain a capping layer. The latter plays a crucial role
long-term light soaking 19. better control of the crystal formation in enhancing the absorption of photons
The finding that CH3NH3PbI3 can act as and growth a sequential deposition by the perovskite in the wavelength
a hole conductor in a PSC prompted a new method was developed23. Lead iodide region above 550 nm, thus substantially
development culminating in the realization (PbI2) was first loaded by spin coating increasing the photocurrent. In this issue
of nanocomposite solar cells20, where the from dimethylformamide onto the of Nature Materials is an Article by this
TiO2 scaffold is fully infiltrated with the mesoporous titanium dioxide film and group26 in which they describe in detail
perovskite light harvester (Fig. 3b). This then exposed to a solution of CH3NH3I in their method of solvent-controlled crystal-
allowed the mesoscopic CH3NH3PbI3/TiO2 isopropanol; this led to the formation of growth engineering used to attain these
film thickness to be reduced to 200–300 nm the final perovskite pigment penetrating high PCE values.
without sacrificing photocurrent. Note the porous titania film. Strikingly, on the
that this PSC configuration featured a mesoscopic length scale the conversion More exciting discoveries
poly(triarylamine) HTM instead of the occurred within seconds and permitted The recent intense research efforts have
previously employed spiro-MeOTAD20. much better control over the perovskite unravelled some extraordinary properties of
Application of energy-dispersive morphology than with the single-step these metal halide perovskites. For instance,
spectroscopy and X-ray diffraction route. An image of the cross-section of the a giant low-frequency dielectric constant
depth profiling confirmed that the pores device is shown in Fig. 4. This technique exceeding 1 million was recently observed
of the TiO2 film were fully infiltrated greatly increased the reproducibility of the in CH3NH3PbI3 films under illumination,
with the CH3NH3PbI3, whereas the solar cell’s performance and boosted the whereas the value was 1,000 times smaller
poly(triarylamine) penetrated the scaffold PCE to 15%. Subsequently, the PCE of a in the dark27. As a consequence of the high
to a much smaller degree. A high PCE perovskite photovoltaic cell prepared by a dielectric constant, these materials show
of 12% was achieved, suggesting that the modified two-step procedure was certified a very large capacitance that is likely to be
carrier collection by hole transport through to be 15.45%. the cause of the slow impedance response
the perovskite was very effective. Transistor Much progress has also been made in in the mHz frequency range observed with
measurements confirmed that CH3NH3PbI3 the preparation of planar PSCs, where these systems28. The observation of a giant
behaved as an ambipolar semiconductor, various methods have been applied dielectric constant suggests that a very
being able to transport both electrons ranging from dual-source high-vacuum strong polarization of the perovskite lattice
and holes in keeping with the results of evaporation21,22 to sequential liquid–vapour occurs in response to the electric field, that
Lee et al.17 and Etgar and colleagues18. This phase deposition24 and low-temperature is, the photovoltage, which is generated
paved the way towards PSC configurations solution casting 25. under illumination. This polarization is
with a planar geometry 21,22, as shown Today, the best-performing cells have likely to originate from the displacement of
in Fig. 3c. reached a certified PCE of 17.9%, which is CH3NH3+ cations, which is also at the origin
more than four times higher than the value of the ionic conductivity of metal halide
Engineering the deposition process attained in 2009; an unprecedented and perovskites10,29. As a consequence of the
The results described previously relied stunning rise in efficiency for a photovoltaic displacement, ion-induced screening of the
on the single-step deposition of the technology. The group of Sang Il Seok has Coulombic attraction between the charge
perovskite pigment using a mixture of PbX2 achieved these latest advances by judicious carriers generated by light is enhanced
and CH3NH3X (X = Cl, Br or I) from a selection of solvent mixtures based on and electrons and holes can separate more

NATURE MATERIALS | VOL 13 | SEPTEMBER 2014 | www.nature.com/naturematerials 839

© 2014 Macmillan Publishers Limited. All rights reserved


commentary

will entail a 60-mV loss in VOC according to


the formula:
a

Perovskite VOC = VOC–ideal + (kT/q)ln(ηext)(1)
Al2O3
HTM where k is the Boltzmann constant, T is
Compact TiO2 layer temperature, q is the elementary charge
TCO and ηext is the external quantum efficiency
Back contact of the photovoltaic cell at VOC obtained
Glass substrate (front contact) under standard measurement conditions. In
line with these theoretical predictions the
electroluminescence quantum yield of PSCs
b has been found to correlate with its VOC
(W. Tress, N. Marinova, O. Inganäs, M. K.
Perovskite
Nazeeruddin, S. M. Zakeeruddin and M. G.,
TiO2
manuscript submitted for publication).
HTM
Compact TiO2 layer
TCO
Open issues and pitfalls
Back contact
During the impending phase of
Glass substrate (front contact)
consolidation, a number of questions
concerning the true potential of PSCs
for widespread deployment will have
to be addressed.
c

Perovskite Stability. Very few stability studies have


HTM been performed so far 19,23 and at this stage
Compact TiO2 layer it is unclear whether PSCs can meet the
TCO stringent international norms for outdoor
Back contact photovoltaic applications. Therefore,
extensive stability tests are urgently
Glass substrate (front contact) warranted. These include long-term light-
soaking tests (1,000 h in full sunlight) as
well as damp heat tests (1,000 h at 85%
humidity and 85 °C). It is well known that
Figure 3 | Evolution from a mesoscopic to a planar embodiment of the perovskite solar cell.
CH3NH3PbI3 degrades in humid conditions
a, ‘Meso-superstructured’ PSC employing a film of Al2O3 nanocrystals covered with a conformal
and forms PbI2 at higher temperatures due
overlayer of CH3NH3PbI3 perovskite. The latter acts as a light harvester as well as an electron conductor.
to the loss of CH3NH3I. These instabilities
b, Nanocomposite embodiment of a PSC where the mesoscopic TiO2 scaffold is infiltrated by the
could hamper outdoor applications.
perovskite. The perovskite assumes the dual role of light absorber and hole conductor. The structure
is akin to p–i–n solar cells. c, Cross-section of a planar heterojunction solar cell lacking the TiO2
Toxicity. Lead compounds are very toxic
mesoporous scaffold.
and harmful to the environment. The
health risks and dangers of CH3NH3PbI3
easily; this leads to a net improvement metal halide perovskite lasers is their wide are currently being investigated by
of the performance of PSCs. A similar tunability over the visible range, which is a consortium of experts within the
screening effect is also operative in DSSCs readily achieved by replacing the iodide framework of the integrated European
where it arises from ions present at the with bromide ions. Joule 7 project Future Nano-Needs
interface of the nanocrystalline TiO2 film The characterization of the (FNN). CH3NH3PbI3 in contact with
with the electrolyte30. electroluminescence properties of fully polar solvents such as water can convert
Films of CH3NH3PbI3 also exhibit operational PSCs allowed further insight to PbI2, a carcinogen, that is moderately
lasing at surprisingly low optical pumping to be gained on the rather high open- water-soluble and whose use is banned in
thresholds, indicating a high level of circuit photovoltage, VOC, observed in these many countries. Therefore it seems that
crystallinity and small concentration of devices, typically ranging between 0.9 V identifying lead-free perovskites or other
bulk impurities or defects in the perovskite and 1.15 V. Owing to entropic losses, the pigments that can replace CH3NH3PbI3 is
material31,32. As a consequence, the maximum obtainable VOC of solar cells necessary for the widespread deployment
radiationless decay of excitons that shortens measured under standard characterization of PSCs. The tin analogues CH3NH3SnI3
their lifetime seems to be mainly controlled conditions is about 300 mV below the (refs 34,35) and CsSnI3 (ref. 36) show
by surface states31. This is astonishing in bandgap. Hence, for CH3NH3PbI3 with a promise, but their extreme sensitivity
view of the fact that the films are prepared bandgap of 1.55 V the maximum VOC is to oxygen presents a severe caveat that
by solution processing, which usually 1.25 V. However, as explained theoretically will have to be addressed to enable
engenders a large concentration of bulk by Robert Ross in 196733, this ideal value, practical applications.
defects. High-quality crystalline perovskite VOC-ideal, is reached only if the external
layers that are suitable for laser applications quantum efficiency, ηext, of the light Pitfalls in measuring conversion
can therefore be obtained by low-cost emission by the solar cell measured at VOC is efficiencies. The appearance in the
deposition methods. A further advantage of 100%. Each decrease in ηext by a factor of 10 photocurrent–voltage curves of strong

840 NATURE MATERIALS | VOL 13 | SEPTEMBER 2014 | www.nature.com/naturematerials

© 2014 Macmillan Publishers Limited. All rights reserved


commentary

hysteresis — which consists of the Au on Al2O3 the CH3NH3SnI3 gave a negligible


anomalous dependence of the PCE on HTM photovoltaic output, with the TiO2 scaffold
the voltage scan direction and speed — a maximum PCE of 6.4% was obtained.
TiO2/CH3NH3Pbl3
calls into question published PCE values Importantly, the VOC = 0.88 V measured
that have not been certified. The results with the CH3NH3SnI3/TiO2 nanocomposite
by Seok et al.26 show that the hysteresis FTO was only 350 mV below the bandgap
effect is particularly severe for planar 200 nm energy, approaching the VOC_ideal value
Glass
device configurations. Hence, PCE values of 0.93 V for CH3NH3SnI3. According to
previously reported in literature for planar equation (1), this implies that the external
architectures should be carefully checked. Figure 4 | Cross-sectional scanning electron quantum efficiency for luminescence at
The authors also show the appearance of microscopy picture of a mesoscopic perovskite VOC should be about 10%, which is high
hysteresis for thicker mesoporous TiO2 solar cell prepared by the sequential deposition for a device exhibiting carrier lifetimes
layers. In this case, however, it has been method. CH3NH3PbI3 densely infiltrates the TiO2 in the picosecond time domain. The fact
reported that by reducing the voltage scaffold and forms an interpenetrating network that VOC values near the theoretical limit
scan speed it is possible to fully suppress junction on the nanometre length scale. FTO, can be reached with CH3NH3SnI3/TiO2
this anomalous behaviour even in solar fluorine-doped tin oxide. Figure reprinted from nanocomposites clearly shows that,
cells having 3-μm-thick mesoscopic ref. 23, Nature Publishing Group. contrary to a contention expressed
oxide scaffolds19. Independently of the earlier 17, employing a mesoporous TiO2
cell embodiment used, it is good practice scaffold does not necessarily entail a
to carefully verify the presence of any present, mesoscopic systems with a certified loss in VOC.
hysteresis by collecting the photocurrent– PCE of 17.9% have a large lead over planar Finally, it should be noted that the
voltage curves in both forward and embodiments, whose reported efficiencies line between planar and mesoscopic
reverse voltage scanning directions and are inferior to this record and remain cell configurations is often difficult to
as a function of scanning speed. Failure uncertified so far. Moreover, as pointed draw. Frequently ‘planar’ embodiments
to do so leads to large overestimations of out above, the hysteresis issue calls into are supported on a nanocrystalline
the device PCE26,28. A further source of question the reported PCE values for planar film of TiO2, ZnO or the C60 derivative
error lies in measuring cell efficiencies on devices. The use of a mesoporous TiO2 [6,6]-phenyl-C61-butyric acid methyl ester
devices smaller than 0.1 cm2. It has been scaffold to host the perovskite material (PCBM) as an electron-capture layer. Also,
shown that for such small cells optical strongly attenuates the hysteresis effects planar perovskite films are often porous,
artefacts can inflate the values of the and increases dramatically the carrier- especially if they are made by solution-
short-circuit photocurrent 37. collection efficiency, which is given by processing at low temperatures and hence
equation (2): can form mesoscopic junctions with the
Doping of CH3NH3PbI3 with chloride hole conductor. On the other hand, the
ions. It has been suggested that replacing ηcoll = (1 + (d/LD)2)–1  (2) efficiency of PSCs based on mesoscopic
PbI2 with PbCl2 as a starting material to TiO2 scaffolds is boosted by placing a
form CH3NH3PbI3 can lead to doping of where d 2 is the mean square displacement compact perovskite overlayer on top of the
the perovskite structure with chloride necessary for an electron (or exciton) to CH3NH3PbI3/TiO2 nanocomposite26. Thus,
ions17. However, no chloride ions could be reach the perovskite/TiO2 interface from it may well be that a marriage of the planar
detected in the CH3NH3PbI3 crystals formed the point where it is photogenerated, and and mesoscopic architecture will win
during the reaction despite extensive LD is the diffusion length. The mesoporous in the end.
analytical efforts. Based on careful X-ray architecture typically reduces d to less
diffraction and elemental analysis, the than 10 nm, which means that a diffusion Outlook
group of Peter Erk has recently confirmed length of 100 nm (ref. 11) would be enough At present, the PSC field is in a
the absence of any chloride doping in to reach a carrier-collection efficiency dynamic state and reports on further
CH3NH3PbI3 crystals formed from PbCl2 higher than 99%. By contrast, for a improvements in performance are
in this manner (keynote lecture, 6th planar film with a thickness of 250 nm — expected in the near future, where
International Conference on Hybrid and which corresponds to the minimum reaching a PCE of 20% seems now to be
Organic Photovoltaics, HOPV 14, 11–14 CH3NH3PbI3 thickness to achieve good a realistic goal. Furthermore, PSCs are
May 2014, Lausanne Switzerland). Instead optical absorption of photons in the attractive candidates for use as top cells
of substituting iodide in the CH3NH3PbI3 600–700 nm wavelength range, where in two-level tandem configurations using
lattice, the chloride ions become part of the solar photon flux is most intense — crystalline silicon or copper indium gallium
an amorphous lead-containing phase that the collection efficiency would be less selenide-based photovoltaic devices as
melts at around 103 °C. In view of these than 50%. Thus the prime advantage of bottom cells. For silicon-based tandems,
unambiguous results, the use of formulas the mesoscopic PSC embodiment over PCEs of 28% to 30% seem to be feasible.
such as CH3NH3PbI3–xClx to indicate a the planar configuration is that it can However, large-scale deployment of
chloride-doped state of the lead iodide reach carrier-collection efficiencies (and perovskite cells will depend on whether
perovskite is no longer justifiable and therefore external quantum efficiencies for stability and toxicity issues can be solved.
should be avoided. photocurrent generation) close to 100%, The amazing advances made recently
even with systems where the exciton or with perovskite photovoltaics will
Planar versus mesoscopic architectures charge-carrier diffusion length is much undoubtedly spur researchers to expand
The simple structure of the planar shorter than the photon absorption length. their investigations to other inorganic or
embodiment for a PSC shown in Fig. 3c This has been confirmed in a recent organic pigments, for which the power of
suggests that ultimately all devices may study by Noel et al.35 on PSCs using tin mesoscopic solar-cell architectures offers
adopt this configuration. However, at perovskites as a light harvester. Whereas new and promising opportunities. ❐

NATURE MATERIALS | VOL 13 | SEPTEMBER 2014 | www.nature.com/naturematerials 841

© 2014 Macmillan Publishers Limited. All rights reserved


commentary

Michael Grätzel is at the Laboratory of Photonics 13. Kojima, A., Teshima, K., Shirai, Y. & Miyasaka, T. 32. Deschler, F. et.al. J. Phys. Chem. Lett. 5, 1421−1426 (2014).
and Interfaces, Swiss Federal Institute of Technology, J. Am. Chem. Soc. 131, 6050–6051 (2009). 33. Ross, R. T. J. Chem. Phys. 46, 4590 (1967).
14. Im, J.‑H., Lee, C.‑R., Lee, J.‑W., Park, S.‑W. & Park, N.‑G. 34. Hao, F., Stoumpos, C. C., Cao, D. H., Chang, R. P. H. &
CH-1015 Lausanne, Switzerland. Nanoscale 3, 4088–4093 (2011). Kanatzidis, M. G. Nature Photon. 8, 489–494 (2014).
e-mail: michael.graetzel@epfl.ch 15. Chung, I., Lee, B., He, J., Chang, R. P. H. & Kanatzidis, M. G. 35. Noel, N. K. et al. Energy Environ. Sci. http://dx.doi.org/
Nature 485, 486–489 (2012). 10.1039/c4ee01076K (2014).
References 16. Kim, H.‑S. et al. Sci. Rep. 2, 591 (2012). 36. Kumar, M. H. et al. Adv. Mater. (in the press).
1. Wells, H. L. Z. Anorg. Chem. 3, 195–210 (1893). 17. Lee, M. M., Teuscher, J., Miyasaka, T., Murakami, T. N. & 37. Ito, S. et al. Thin Solid Films 516, 4613–4619 (2008).
2. Møller, C. K. Nature 182, 1436 (1958). Snaith, H. J. Science 338, 643–647 (2012).
3. Weber, D. Z. Naturforsch. 33b, 1443–1445 (1978). 18. Etgar, L. et al. J. Am.Chem. Soc. 134, 17396–17399 (2012). Acknowledgements
4. Weber, D. Z. Naturforsch. 33b, 862–865 (1978). 19. Mei, A. et al. Science 345, 295–298 (2014). I am grateful to my co-workers at the Laboratory of
5. Mitzi, D. B. Synthesis, Structure and Properties of 20. Heo, J. H. et al. Nature Photon. 7, 487–492 (2013). Photonics and Interfaces at EPFL, and to all academic
Organic–Inorganic Perovskites and Related Materials: 21. Liu, M., Johnston, M. B. & Snaith, H. J. Nature 501, 395–398 (2013). partners with whom I have cooperated in the field
Progress in Inorganic Chemistry Vol. 48 (ed. Karlin, K. D.) 22. Malinkiewicz, O. et al. Nature Photon. 8, 128–132 (2014). of perovskite solar cells, in particular N.-G. Park,
1–121 (J. Wiley & Sons, 1999). 23. Burschka, J. et al. Nature 499, 316–319 (2013). (SKKU Korea), S. I. Seok (KRICT Korea), S. Mhaisalkar,
6. Baikie, T. et al. J. Mater. Chem. A 1, 5628–5641 (2013). 24. Chen, Q. et al. J. Am. Chem. Soc. 136, 622–625 (2014).
N. Mathews and T. C. Sum (NTU Singapore), H. W. Han
7. Umari, P., Mosconi, E. & De Angelis, F. Preprint available at 25. Yella, A., Heiniger, L.‑P., Gao, P., Nazeeruddin, M. K. &
(HUST China) and H. Bolink (Univ. of Valencia
http://arxiv-web.arxiv.org/abs/1309.4895 (2013). Grätzel, M. Nano Lett. 14, 2591–2596 (2014).
8. Ponseca, C. S. et al. J. Am. Chem. Soc. 136, 5189–5192 (2014). 26. Jeon, N. J. et al. Nature Mater. 13, 897–903 (2014).
Spain). Financial support by the Swiss National Science
9. Mitzi, D. B. J. Chem. Soc. Dalton Trans. 1, 1–12 (2001). 27. Juarez-Perez, E. J. et al. J. Phys. Chem. Lett. 5, 2390−2394 (2014). Foundation and the European Research Council through
10. Stoumpos, C. C., Malliakas, C. D. & Kanatzidis, M. G. 28. Dualeh, A. et al. ACS Nano 8, 362–373 (2014). the Advanced Research Grant (ARC no. 247404)
Inorg. Chem. 52, 9019–9038 (2013). 29. Pellet, N. et al. Angew. Chem. Int. Ed. 53, 3151–3157 (2014). ‘Mesolight’ and the Global Research Laboratory (GRL)
11. Xing, G. C. et al. Science 342, 344–347 (2013). 30. Grätzel, M. Nature 414, 332–344 (2001). programme, funded by the National Research Foundation
12. Stranks, S. D. et al. Science 342, 341–344 (2013). 31. Xing, G. et al. Nature Mater. 13, 476–480 (2014). in Korea are specifically acknowledged.

842 NATURE MATERIALS | VOL 13 | SEPTEMBER 2014 | www.nature.com/naturematerials

© 2014 Macmillan Publishers Limited. All rights reserved