Lanthanide ions as spectral converters for solar cells Bryan M. van der Ende,*ab Linda Aartsa and Andries Meijerink*a Received 10th July 2009, Accepted 19th August 2009 First published as an Advance Article on the web 10th September 2009 DOI: 10.1039/b913877c The use of lanthanide ions to convert photons to different, more useful, wavelengths is well-known from a wide range of applications (e.g. fluorescent tubes, lasers, white light LEDs). Recently, a new potential application has emerged: the use of lanthanide ions for spectral conversion in solar cells. The main energy loss in the conversion of solar energy to electricity is related to the so-called spectral mismatch: low energy photons are not absorbed by a solar cell while high energy photons are not used e���ciently. To reduce the spectral mismatch losses both upconversion and downconversion are viable options. In the case of upconversion two low energy infrared photons that cannot be absorbed by the solar cell, are added up to give one high energy photon that can be absorbed. In the case of downconversion one high energy photon is split into two lower energy photons that can both be absorbed by the solar cell. The rich and unique energy level structure arising from the 4fn inner shell configuration of the trivalent lanthanide ions gives a variety of options for e���cient up- and downconversion. In this perspective an overview will be given of recent work on photon management for solar cells. Three topics can be distinguished: (1) modelling of the potential impact of spectral conversion on the e���ciency of solar cells (2) research on up- and downconversion materials based on lanthanides and (3) proof-of-principle experiments. Finally, an outlook will be given, including issues that need to be resolved before wide scale application of up- and downconversion materials can be anticipated. I. Introduction Global energy consumption is on the rise and projected to double in 2050 in comparison with worldwide energy consumption rates in 2001.1 Sustainable energy production based on the direct conversion of energy radiated from the sun into useable forms like heat or electricity is expected to gain importance since it may be the only renewable source capable of generating su���cient energy to meet the long-term worldwide energy demand.1,2 Solar energy utilization requires effective means of capture and conversion of the solar radiation, and storage of the acquired energy.1 The capacity of photovoltaic cells to convert sunlight into electricity makes them prime candidates for effective large-scale capture and conversion of solar energy, but at present the contribution of photovoltaic energy is limited due to its relatively high cost per kilowatt-hour.3 A reduction in price may be achieved by either lowering the production cost or increasing the conversion e���ciency. a Condensed Matter and Interfaces, Debye Institute for NanoMaterials Science, Utrecht University, Princetonplein 1, 3584 CC Utrecht, The Netherlands. E-mail: b.m.vanderende@uu.nl, a.meijerink@uu.nl Fax: +31 (0)302532403 Tel: +31 (0)302532321, +31 (0)302532202 b Department of Physics and Astronomy, Trent University, 1600 West Bank Drive, Peterborough, Ontario, Canada K9J 7B8 Bryan M. van der Ende Bryan van der Ende attended Simon Fraser University (Burnaby, B.C., Canada) for his BSc degree, and the University of Guelph (Guelph, Ontario, Canada) for his MSc and PhD degrees. He is currently finishing a post- doctoral fellowship at Utrecht University, (Utrecht, The Netherlands) under the super- vision of Prof. Andries Meijerink. He will soon begin a new postdoctoral position with Prof. Ralph Shiell at Trent University (Peterborough, Ontario, Canada). Linda Aarts Linda Aarts received her BSc degree in chemistry and subsequently completed the Masters programme in Chemistry and Physics at Utrecht University (Utrecht, The Netherlands). Currently she is a PhD student at Utrecht Univeristy and works under the supervision of Prof. Andries Meijerink on down- conversion for solar cells. This journal is c the Owner Societies 2009 Phys. Chem. Chem. Phys., 2009, 11, 11081���11095 | 11081 PERSPECTIVE www.rsc.org/pccp | Physical Chemistry Chemical Physics

State-of-the-art commercial single-junction crystalline and polycrystalline Si solar cells dominate the photovoltaic market.4,5 Crystalline Si wafer cells typically have energy e���ciencies around 15%.3 In a solar cell a single electron���hole pair is generated upon absorbing a photon above the energy band-gap Eg of the semiconductor material.6 Fig. 1 shows a schematic diagram showing the loss mechanisms that limit the e���ciency of single-junction solar cells. The most significant loss mechanisms are due to relaxation of ���hot��� charge carriers that are created upon absorption of a high energy photon (process 1 in Fig. 1) and transmission of photons with energies below the band-gap of the semiconductor material (process 2 in Fig. 1). The excess energy of high energy photons is rapidly dissipated as heat by thermalization of the electron and hole to the edges of the conduction band and valence band. These losses are known as thermalization losses. Especially in solar cells based on semiconductors with a small band-gap, these losses are substantial. The transmission losses contribute more to the losses for wider band-gap solar cells, simply because a large part of the solar spectrum cannot be absorbed.7 Process 3 shows recombination of electron���hole pairs, and this loss process can be minimized through maintaining high minority carrier lifetimes in the semiconductor material. In a detailed balance model developed by Shockley and Queisser,8 the theoretical e���ciency limit for a single-junction solar cell with Eg equal to 1.1 eV can be determined to be 30%. The largest part of the 70% energy loss is related to processes 1 and 2, and is known as the spectral mismatch. There are several methods to reduce the spectral mismatch losses. Two basic approaches can be discerned: adapt the solar cell to better use the solar spectrum or adapt the solar spectrum to better match the solar cell. The first approach has been successfully applied in so-called multi-junction tandem solar cells for which e���ciencies over 40% under concentrated solar light have been realized.9 This is achieved through combining multiple semiconductor materials, each of which absorb different fractions of the solar spectrum.6,10 Tandem solar cells are costly, however, and are only now becoming cost-competitive for terrestrial concentrated solar applications.10 Other more recently proposed options include multiple exciton generation (MEG), and space-separated quantum cutting (SSQC). In both cases absorption of a high energy photon by the solar cell results in the generation of multiple electron���hole pairs in the cell. MEG has been reported in the recent literature for various types of semiconductor nano- crystals (���quantum dots���, e.g. CdSe, PbSe, and PbS).11,12 The MEG process uses the excess energy of charge carriers produced by photons of energy greater than Eg to produce additional charge carriers. It has been estimated that MEG could enhance the e���ciency of single junction solar cells to as much as 44%.13,14 This high gain in e���ciency is however only reached if the MEG e���ciency is close to creating an extra electron hole (e���h) pair for every increase by Eg in the photon energy. Even though e���h pair generation e���ciencies of 700% for photon energies of 8 Eg have been reported, more recent experiments have revealed that the actual e���ciencies are much lower and are in fact very similar to the well-known e���ciencies for e���h generation in scintillator materials and cathode ray phosphors where every extra e���h pair requires an additional 2.5 Eg in the photon energy. This will lead to marginal increases in the solar cell e���ciency and probably already contribute with similar e���ciency in (bulk) semiconductor solar cells for the highest energy (UV) photons.15 Another concept is space separated quantum cutting (SSQC) which has been recently reported to occur in Si nanocrystals.16 SSQC divides an absorbed higher energy photon into two or more e���h pairs through the interaction of two spatially separated neighbouring Si nanocrystals. SSQC could thereby decrease the loss of energy due to thermalization in solar cells, and multiply the number of charge carriers produced per absorbed photon in a solar cell but further research is required to establish the e���ciency of charge carrier generation. An alternative approach to raise the theoretical e���ciency beyond the Shockley���Queisser limit is to adapt the solar spectrum through upconversion or downconversion. A comparative review of each of these methods is given by Strumpel et al.17 In the case of upconversion, two low-energy photons are ���added up��� to give one higher-energy photon,18 thus converting sub-band-gap photons, which are otherwise lost, into supra-band-gap photons, which can be absorbed.19���21 Downconversion, or ���quantum cutting���, is the opposite process to upconversion, whereby one high energy photon is ���cut��� into two lower-energy photons. This process can reduce the energy loss due to thermalization of hot charge Fig. 1 Loss processes in a single-junction solar cell: (1) lattice thermalization losses (2) transparency loss (3) recombination loss (4) junction loss (5) contact voltage loss. Adapted from Richards,7 copyright 2006, with permission from Elsevier. Andries Meijerink Andries Meijerink received his PhD in 1990 at the Utrecht University under supervision of Prof. George Blasse. After- wards he joined the group of Prof. John Wright at the University of Wisconsin in Madison as a post-doctoral fellow. In 1996 he was appointed at the chair of Solid State Chemistry in the Debye Institute of the Utrecht University. He leads an active research group that focuses on the optical spectroscopy of lanthanide ions and of semi- conductor quantum dots. 11082 | Phys. Chem. Chem. Phys., 2009, 11, 11081���11095 This journal is c the Owner Societies 2009

carriers after the absorption of a high-energy photon. If both lower-energy photons can be absorbed by the solar cell, current doubling is obtained for the high energy region of the solar spectrum, which consists of photons with energies exceeding 2Eg.22,23 The final result is similar to MEG and SSQC however, rather than multiple excitons being created by one photon absorbed into the solar cell, they are created by photon-doubling prior to absorption into the solar cell. Downshifting is similar to downconversion, but in the case of downshifting, the external quantum e���ciency can not exceed unity. Downshifting may also raise the e���ciency of the solar cell by converting a (one) higher energy photon into a (one) lower energy photon that is more e���ciently absorbed by the cell,24 but it could never be used to exceed the Shockley��� Queisser e���ciency limit. Various means can be used to achieve this downshifting, employing quantum dots,25,26 lanthanide ions,27,28 and other types of inorganic29,30 and organic31 materials. This perspective, however, will focus on upconversion and downconversion since these mechanism are capable of raising the e���ciency beyond the Shockley���Queisser limit. Fig. 2 shows the standard terrestrial solar spectrum (air mass coe���cientw AM1.5G) and the fraction of the energy that is currently used by single junction c-Si solar cells. The part that is not used is available for upconversion and downconversion. In this spectrum, 32% (149 W m 2) more of the sunlight intensity is accessible through downconversion, and 35% (164 W m 2) of the sunlight intensity is accessible through upconversion. These numbers vary strongly with the air mass coe���cient: there is a significant spectral shift from the ultraviolet to the infrared, with increasing air mass coe���cient.7 Downconversion is therefore expected to provide the greatest benefits for solar radiation of smaller air mass coe���cients, especially for extraterrestrial solar radiation (air mass coe���cient AM0), and for diffuse terrestrial solar radiation. Upconversion is expected to provide the greatest benefits for direct solar radiation, and for solar radiation of a higher air mass coe���cient.7 Richards considers further advantages and disadvantages of upconversion and downconversion.7 For upconversion, a large body of research has already been published, although primarily aimed at the conversion of the output of NIR diode lasers (800���1000 nm) into visible light, rather than converting longer wavelength NIR photons into NIR radiation that can be absorbed by crystalline Si solar cells. An important advantage of upconversion is that an upconversion layer can be applied to the rear of a solar cell without affecting the performance of the device for incident photons with energy E 4 Eg. Any upconversion of transmitted IR radiation into the useful wavelength range that generates photocurrent serves as real gain. The use of a suitable reflector placed behind the upconversion layer ensures that no usable luminescence can escape out the rear of the solar cell. The major disadvantage of upconversion, however, is that it is a nonlinear process which only becomes e���cient at high incident power. For an upconversion process of two steps, the intensity of upconversion light I is related to the incident light intensity lo by I p lo2. For three step upconversion, the relation is I p lo.3 Down- conversion, however, is a linear process and the e���ciency is independent of the incident power. This suggests that it will be easier to gain e���ciency with downconversion when using unconcentrated sunlight as the illumination source. In the case of downconversion the conversion layer is applied on top of the solar cell and a fraction of the luminescence escapes out of the front top surface of the downconversion layer. For luminescent species contained in a host material with a refractive index of n = 1.5, this fraction is 12.7%. Richards thus argues that with front-mounted downconversion, an external quantum e���ciency (EQE) of 115% for the down- converter is required just to break even. Losses may be reduced by applying an anti-reflective coating which specifically reflects the downconverted emission back into the solar cell. This perspective focuses on the use of lanthanide ions to achieve upconversion and downconversion for the enhancement of solar cell e���ciency. Much of the recent research in upconversion and downconversion was sparked by seminal modelling work published by Trupke et al.21,22 in 2002. We will discuss this and more recent work on modelling of the effect of spectral conversion on the e���ciency of solar cells. Further we consider the status of research in upconversion and downconversion systems that could yield significant enhancements in various kinds of solar cells, and we will summarize what efforts have been made so far���and what further can be tried���to achieve significant solar cell e���ciency enhancement from spectral conversion of solar radiation. II. Modelling of upconversion and downconversion in solar cells We begin by considering the detailed balance models of Trupke et al. for upconversion21 and downconversion.22 The detailed balance models follow that of Shockley and Queisser, except for the addition of spectrally converting layer. In the case of upconversion, the converting layer is mounted beneath a bifacial solar cell, and it is assumed to be electronically isolated from the solar cell a perfect reflector is also located at Fig. 2 AM1.5G solar spectrum showing the fraction of terrestrial sunlight that is currently absorbed and effectively utilised by a thick crystalline silicon device, and the additional regions of the spectrum that can contribute to up- or downconversion. Reprinted from Richards,7 Copyright 2006, with permission from Elsevier. w The air mass coe���cient describes the solar spectrum after the solar radiation has passed through the earth���s atmosphere it is the ratio of the solar radiation path length l to the atmosphere thickness lo, for solar radiation incident at an angle y relative to the normal of the earth���s surface: llo 1 = 1/cos y.7 This journal is c the Owner Societies 2009 Phys. Chem. Chem. Phys., 2009, 11, 11081���11095 | 11083