This journal is© the Owner Societies 2015 Phys. Chem. Chem. Phys., 2015, 17, 1619--1629 | 1619

Cite this:Phys.Chem.Chem.Phys.,

2015, 17, 1619

Organic–inorganic halide perovskite/crystallinesilicon four-terminal tandem solar cells†

Philipp Loper,*a Soo-Jin Moon,b Sılvia Martın de Nicolas,a Bjoern Niesen,a

Martin Ledinsky,ac Sylvain Nicolay,b Julien Bailat,b Jun-Ho Yum,b Stefaan De Wolfa

and Christophe Ballifab

Tandem solar cells constructed from a crystalline silicon (c-Si) bottom cell and a low-cost top cell offer a

promising way to ensure long-term price reductions of photovoltaic modules. We present a four-terminal

tandem solar cell consisting of a methyl ammonium lead triiodide (CH3NH3PbI3) top cell and a c-Si

heterojunction bottom cell. The CH3NH3PbI3 top cell exhibits broad-band transparency owing to its

design free of metallic components and yields a transmittance of 455% in the near-infrared spectral

region. This allows the generation of a short-circuit current density of 13.7 mA cm�2 in the bottom cell.

The four-terminal tandem solar cell yields an efficiency of 13.4% (top cell: 6.2%, bottom cell: 7.2%), which

is a gain of 1.8%abs with respect to the reference single-junction CH3NH3PbI3 solar cell with metal back

contact. We employ the four-terminal tandem solar cell for a detailed investigation of the optical losses

and to derive guidelines for further efficiency improvements. Based on a power loss analysis, we estimate

that tandem efficiencies of B28% are attainable using an optically optimized system based on current

technology, whereas a fully optimized, ultimate device with matched current could yield up to 31.6%.

I. Introduction

The photovoltaic market has been dominated for decades bycrystalline silicon (c-Si) solar cells, which account for a marketshare of about 90%.1 Recently, efficiencies as high as 25.6%were reported for wafer-sized devices,2 approaching the single-junction limit of E29% for c-Si solar cells.3 Additionally, thecosts of photovoltaic installations scale mainly with the systemarea, and the key driver to reduce the price of electricity is thusthe module efficiency (power per area).4 Consequently, cell con-cepts enabling ultra-high efficiencies beyond the c-Si single-junction limit with low-cost fabrication are needed to ensurelong-term competitiveness of photovoltaics with conventionalenergy sources. The most promising way towards this aim is toconstruct a dual-junction tandem solar cell based on c-Si solarcell technology combined with a high-efficiency solar cell with

higher band gap. Until recently, the lack of a suitable high-efficiency top cell compatible with low-cost processing wasinhibiting the fabrication of such c-Si-based tandem devices.However, this situation changed drastically recently with the riseof organic–inorganic halide perovskite solar cells.5,6

Since their first application as photovoltaic materials,7,8

perovskites have demonstrated their photovoltaic potentialwith confirmed solar cell efficiencies of up to 20.1% to date.9

They can be either solution-10,11 or vacuum-processed,12 allow-ing for simple and cost-effective device fabrication, and can bedeposited on glass and also on flexible materials13,14 andprepared at temperatures o150 1C.15,16 Especially interestingfor the application in a Si-based tandem solar cell is thereported band gap of methyl ammonium lead triiodide(CH3NH3PbI3) between B1.50 eV (ref. 17) and 1.57 eV.18 Sofar, CH3NH3PbI3 has been reported to be the perovskite materialwith the best efficiency, exhibiting a steep absorption edge19 andexceptionally low sub-gap absorption.

This renders CH3NH3PbI3 a promising candidate for thehigh-band gap top cell in a dual-junction tandem solar cellwith the bottom cell made of a lower band-gap material such ascopper indium gallium selenide (CIGS) or c-Si.20

The tandem device can either be designed as a mechanicalstack of independently connected cells (four-terminal) or as amonolithic device (two-terminal). In a four-terminal tandemdevice, the perovskite module, deposited on glass, could also

a Ecole Polytechnique Federale de Lausanne (EPFL), Institute of Microengineering (IMT),

Photovoltaics and Thin-Film Electronics Laboratory, Rue de la Maladiere 71b,

2002 Neuchatel, Switzerland. E-mail: philipp.loeper@epfl.ch; Tel: +41-21-6954562b Centre Suisse d’Electronique et de Microtechnique (CSEM), Jaquet-Droz 1,

2002 Neuchatel, Switzerlandc Laboratory of Nanostructures and Nanomaterials, Institute of Physics,

Academy of Sciences of the Czech Republic, v. v. i., Cukrovarnicka 10, 162 00 Prague,

Czech Republic

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c4cp03788j

Received 23rd August 2014,Accepted 14th November 2014

DOI: 10.1039/c4cp03788j

www.rsc.org/pccp

PCCP

PAPER

1620 | Phys. Chem. Chem. Phys., 2015, 17, 1619--1629 This journal is© the Owner Societies 2015

serve as module glass encapsulating the bottom cell. Sucha four-terminal tandem device would require only marginalchanges to the c-Si bottom cell, and could thus facilitatesignificantly the market entry for perovskite photovoltaics inthe short term.21 The monolithic tandem device requires theadaptation of both cells and their interconnection with arecombination junction, representing thus a more advanceddevice to be realized in a longer term.

While the virtues of perovskite/c-Si tandem solar cells havebeen pointed out by several authors,5,20–25 the solutions thathave to be delivered to realize efficient systems have not yetbeen addressed in detail.

In either the four- or two-terminal case, the first critical steptowards a perovskite/c-Si tandem solar cell is the realization ofa top cell with broad-band transparent electrodes and excellentinfrared (IR) transparency.

Semi-transparent perovskite solar cells have been reportedusing a thin (10 nm) Au layer as a back electrode,26 reachingconversion efficiencies of 3% to 7.5% and showing a transmittanceof 20% at 800 nm (ref. 26, ESI†). In another approach, thin layersof aluminium-doped zinc oxide (AZO) and Ag were combined to aAZO/(9 nm Ag)/AZO stack electrode, leading to perovskite solar cellefficiencies of up to 7%, but the transmittance was not reported inthis study.27 For organic solar cells, semi-transparent electrodesbased on silver (Ag) nanowires (NWs) have been employed,28 butrequire yet an additional layer for lateral transport to the Ag NWgrid and ohmic contact with low resistivity.29

Moreover, metal-based electrodes inherently suffer from theirintrinsic absorption, especially in the infrared region.30 There-fore, for high IR transparency, non-metallic electrodes have to bedeveloped. For the application in a monolithic tandem device,the electrode should also be applicable to wafer-sized devices(4100 cm2), and be compatible with industrial metallizationschemes such as screen printing or copper plating.

In this article, we present an IR-transparent CH3NH3PbI3

solar cell featuring a transparent back contact free of metalliccomponents. We use this IR-transparent CH3NH3PbI3 solar cellto realize a four-terminal perovskite/c-Si tandem device. Finallywe perform a detailed opto-electrical analysis of the tandemsystem and assess the efficiency potential of optimized perovs-kite/c-Si tandem solar cells.

II. Broad-band transparentCH3NH3PbI3 solar cells

High-efficiency perovskite solar cells are commonly prepared as alayer stack comprising a transparent conducting oxide electrodeon glass, coated with either a planar12 or a scaffold-structured10

charge (electron or hole) transport layer, the perovskite absorberlayer, the second charge transport (hole or electron) layer, and ametal rear electrode. Light enters the solar cell through the glasssubstrate, such that the metal rear electrode acts as a backreflector to boost the photocurrent. For the four-terminal tandemdevice presented in this article we employ the preparation routepresented by Burschka et al. because of its high efficiency of up to

15%,10 but replace the Au back contact by a dedicated transparenthole-collecting electrode. This cell structure, shown in Fig. 1(a),comprises fluorinated tin oxide (FTO) as TCO on glass, which iscoated by a compact TiO2 (c-TiO2) layer and a mesoporous TiO2

(m-TiO2) layer. After a 2-step preparation of the perovskite layer,2,20,7,70-tetrakis(N,N-di-p-methoxyphenylamine)-9,90-spirobifluorene(spiro-MeOTAD) is spin-coated and serves as a hole transportmaterial (HTM).

Importantly, the transparent rear electrode is not only thefirst necessary development for a four-terminal tandem device,but also the missing building block for the integration in amonolithic device, sketched in Fig. 1(b).

First, we prepared a transparent electrode made of 100 nmindium tin oxide (ITO) by sputter deposition directly on thespiro-MeOTAD layer. Even though the sputter process is opti-mized for soft deposition, the current–voltage (IV) curves,shown in Fig. 2, do not even show rectifying characteristics.Reference samples that received an evaporated Au electrodeinstead of the sputtered ITO on top of the spiro-MeOTAD layer

Fig. 1 (a) Schematic of the perovskite solar cell architecture based onpreparing the perovskite absorber on a compact TiO2 electrode, which isdeposited on FTO-coated glass. Gold (Au) or silver (Ag) is most commonlyemployed as the hole collecting electrode. For the four-terminal tandem,the metal hole collecting electrode has to be replaced by a broad-bandtransparent material. (b) Also, for a monolithic perovskite/c-Si tandem, tobe realized at a later stage, the transparent hole collecting electrode is akey building block as it would enable using the established top cellpreparation route also used in (a).

Fig. 2 Current–voltage curves in dark and under illumination of perovskitesolar cells with ITO (dotted line), MoOx/ITO (line), and MoOx/Ag (dashed line)back contact.

Paper PCCP

This journal is© the Owner Societies 2015 Phys. Chem. Chem. Phys., 2015, 17, 1619--1629 | 1621

exhibited decent cell results. From this, we conclude that thedeteriorated junction characteristics in the case of the ITOelectrodes are likely related to damage caused by the ITO sputterdeposition process, a phenomenon well-known to detrimentallyaffect other solar cell technologies as well, including siliconheterojunction solar cells.31

Recently, efficient hole-collectors for a-Si:H/c-Si heterojunc-tion32 and perovskite solar cells33 were realized based on thinlayers of evaporated molybdenum oxide (MoOx). The MoOx-basedhole collectors were proven to provide a remarkably high trans-parency without inducing sputter damage in the active layersunderneath.32

Inspired by these results, we tested here MoOx layers as hole-collecting buffer layers, capped with 100 nm ITO to providelateral conductivity to the metallization. With this design,rectifying properties were obtained both in the dark and underillumination as shown by the IV curves in Fig. 2, resulting in aconversion efficiency of 6.2% with an active area of 0.2773 cm2.This result was achieved with a 30 nm MoOx buffer layer,yielding an open-circuit voltage (VOC) of 821 mV, a short-circuit current density ( JSC) of 14.5 mA cm�2 and a fill factor(FF) of 51.9%. Fig. 2 also shows the IV curves of a reference cell,which was co-processed with the IR-transparent cell but receivedan evaporated Ag electrode instead of the sputtered ITO. Thiscell exhibits a decent performance ( JSC = 18.51 mA cm�2, VOC =938 mV, and FF = 67%) with an efficiency of 11.6%.

Obviously, replacing the Ag back contact by ITO stronglyaffects all cell parameters. The cell with Ag back contactexhibits higher external quantum efficiency (Fig. S1, ESI†) overthe full spectral range, especially at higher wavelengths (500 nmto 800 nm), demonstrating better light trapping due to the backreflecting properties of the Ag back contact.

The discrepancy in VOC between the two cells can most likelybe explained by the process-induced damage of the ITO sputterdeposition – also known from a-Si:H/c-Si heterojunction solarcells31 – being insufficiently shielded by the MoOx layer.

To analyse the origin of the FF loss, we calculate the cells’series resistances from their dark and light IV curves accordingto Dicker (eqn (3) in ref. 34). The cell with MoOx/Ag backcontact has a series resistance of 7.3 O sq�1, and an associated

FF loss of 11.4%. However, the cell with transparent MoOx/ITOback contact is affected by a series resistance of 19.6 O sq�1,resulting in a FF loss of 24.6%abs. Interestingly, the two cellsdiffer only marginally in their series-resistance-free FF (1.6%abs),indicating that the FF is not affected by the ITO sputter process,but rather the additional series resistance due to lateral trans-port in the ITO.

While the perovskite cell with transparent MoOx/ITO backcontact does not yet attain the performance of high-efficiencydevices, its efficiency is comparable to that of other recentlypresented results obtained with semi-transparent electrodes.26,27

In contrast to the latter, however, the cell presented here featurestransmission not only in the visible spectral region, but alsoin the near-infrared spectral region up to 1200 nm (discussedbelow), which makes this device suitable as a top cell for ac-Si-based tandem device.

Owing to the decent VOC and FF values of the MoOx/Ag referencecell, we expect that the performance of the IR-transparent cellcan be boosted to the level of state-of-the-art perovskite devicesby fine-tuning the ITO deposition.

III. Four-terminal CH3NH3PbI3/c-Sitandem solar cell

As in a tandem device the c-Si bottom cell operates withreduced generated excess carrier density with respect to standardconditions (0.1 W cm�2 irradiance), it has to be adequatelydesigned to convert the light transmitted through the top cellin an optimal way. In brief, the bottom cell should fulfil tworequirements. First, it should exhibit an excellent IR responseto utilize the transmitted light as far as possible. Secondly, itshould feature well-passivated surfaces and contacts to enable ahigh open-circuit voltage even at low excess carrier densities. Theclass of c-Si solar cells that fulfils these two requirements bestare a-Si:H/c-Si heterojunction solar cells because of their lowparasitic absorption in the infrared region35 and also excellentsurface passivation at low excess carrier density.36

Combining the top cell with a transparent MoOx/ITO elec-trode discussed above with an a-Si:H/c-Si heterojunction bottom

Fig. 3 (a) Schematic of the mechanically stacked four-terminal tandem. The system consists of a high-efficiency a-Si:H/c-Si heterojunction solar celland a high-efficiency CH3NH3PbI3 top cell with a metal-free MoOx/ITO transparent electrode. (b) External quantum efficiency and (c) current–voltagecurves of the two individually connected subcells in the four-terminal perovskite/Si tandem solar cell.

PCCP Paper

1622 | Phys. Chem. Chem. Phys., 2015, 17, 1619--1629 This journal is© the Owner Societies 2015

cell, we realized the first perovskite/c-Si tandem solar cell, shownin Fig. 3a. We evaluated the performance of this four-terminaltandem solar cell with EQE measurements of the two subcells,i.e. the top cell and the top-cell-filtered bottom cell, and IVmeasurements of the two cells under the illumination condi-tions in the tandem configuration.

The EQE of the four-terminal tandem solar cell is shown inFig. 3b. The top cell EQE peaks between 400 nm and 460 nmwith 76%, and then decreases monotonically towards theCH3NH3PbI3 band gap at B800 nm. For comparison, we alsoplot in Fig. 3b the absorbance A of a CH3NH3PbI3 layer,calculated with the absorption coefficient a reported in ref. 19in the single pass limit (A = 1 � exp(�a�d)) for a typical absorberthickness of d = 200 nm. The EQE decreases towards the bandgap and its similarity to the single pass limit suggests that thedegree of light trapping in the top cell is low. Photons between500 nm and 800 nm that are not converted in the top cellcontribute at least partially to bottom cell current. The totaltandem EQE is approximately constant up to the CH3NH3PbI3

band gap. For photon energies below the CH3NH3PbI3 bandgap (4800 nm), up to 58% of the photons that are incident onthe tandem system are transmitted through the top cell andutilized by the c-Si bottom cell.

The bottom cell JSC calculated from the EQE measurement is13.7 mA cm�2. Please note that only light with wavelengths>500 nm reaches the bottom cell in the tandem configuration.As the SHJ cell employed exhibits excellent surface passivationits collection efficiency does not depend on the incident spec-trum. To obtain the correct IV curve of the bottom cell at theillumination conditions in the tandem configuration, we there-fore adjusted the illumination intensity using grey filters suchthat the bottom cell JSC matched the value calculated from theEQE measured in tandem configuration. With this current, thebottom cell exhibits an open-circuit voltage of 689 mV and a fillfactor of 76.7%. The IV curves of both subcells of the tandemdevice are shown in Fig. 3c. In a final tandem module, the two(top and bottom) submodules will have different maximumpower points, which makes an individual connection of eachsubmodule to a separate inverter necessary. The total four-terminal tandem device efficiency is thus the sum of theindividual subcell efficiencies. The efficiency sum of the four-terminal tandem device presented here is 13.4%, which is1.8%abs higher than the efficiency of the reference perovskitesolar cell with MoOx/Au back contact mentioned above. Table 1summarizes the parameters of the four-terminal tandem system.

The total EQE of the tandem system is above 70% for l o750 nm, but strongly drops for longer wavelengths as the top

cell does not contribute to photocurrent anymore. We remarkthat without the top cell, the bottom cell features an EQE490% from 400 nm to 1000 nm, virtually utilizing all photonsin this spectral region. The low tandem EQE observed not onlyin the infrared region, but also in the visible region, thereforehints at parasitic light filtering by the perovskite top cell.

We remark that an efficiency gain of 1.8%abs relative to asingle-junction perovskite cell cannot justify the cost of addinga c-Si bottom cell. From the perspective of c-Si photovoltaics,the top cell is far from being sufficiently efficient and trans-parent to boost the efficiency of state-of-the-art c-Si cells.However, it has to be taken into account that the top cellefficiency is severely affected by at least two significant draw-backs: firstly, the JSC (low lightrapping degree), VOC (sputter-induced damage) and FF (additional series resistance) lossesinduced by the MoOx/ITO transparent electrode, and secondlythe parasitic absorption in the top cell. As the IV characteristicsprove that the junction is well formed and rectifying, we expectthat further process and material optimization will likely increasethe top cell VOC and FF towards state-of-the-art performance.

In the meantime, the four-terminal tandem device presentedhere constitutes a valuable test platform for an experimentalpower loss analysis and enables us to assess the tandem deviceefficiency potential.

The preliminary results presented in this article thus pointout not only the complexity of realizing efficient perovskite/c-Sitandem devices, but also which solutions have to be found tocompete with and even surpass high-efficiency c-Si solar cells.

IV. Current loss analysis

In the following, we present a detailed opto-electrical analysisto quantify current and power losses and assess the efficiencypotential of optimized systems.

Fig. 4 depicts the transmittance (TTop) spectra of the per-ovskite top solar cell as well as its 1-reflectance (1 � RTop) andabsorbance (ATop = 1 � RTop � TTop) spectrum. We also replot inFig. 4 the EQEs shown in Fig. 3b for comparison. The EQE and

Table 1 Cell parameters of the transparent and reference perovskite solarcells, the SHJ bottom solar cell, and the resulting tandem efficiency (sum)

CellHolecontact

JSC

(mA cm�2)VOC

(mV)FF(%)

Efficiency(%)

Perovskite reference MoOx/Ag 18.5 938 67.0 11.6Perovskite top MoOx/ITO 14.5 821 51.9 6.2SHJ bottom 13.7 689 76.7 7.2Perovskite/SHJ tandem 13.4

Fig. 4 Transmittance, 1-reflectance and absorbance spectra of the top celland the external quantum efficiency spectra of the top and bottom cells.

Paper PCCP

This journal is© the Owner Societies 2015 Phys. Chem. Chem. Phys., 2015, 17, 1619--1629 | 1623

the absorptance of the top cell exhibit a similar spectral shape inthe visible region, but are offset by about 15%abs. The differencebetween absorptance and EQE can be attributed to parasiticabsorption losses, i.e. absorption processes not contributing tothe top cell photocurrent.

Excellent agreement can be seen between the top cell trans-mittance and the EQE of the bottom cell up to 1000 nm, indicat-ing that the bottom cell efficiently converts the incident photonsto photocurrent. For wavelengths 41000 nm, the two curvesbegin to deviate as the bottom cell EQE decreases near the c-Siband gap. Ideally, the top cell should transmit 100% of theincident photons below the CH3NH3PbI3 band gap. However,the data shown in Fig. 4 demonstrate that the realized top celltransmits B55% of the light between 800 nm and 1200 nm, andabsorbs between 25% and 35% of the incident photons despitethe excellent sub-bad gap transparency of CH3NH3PbI3 itself.19

Fig. 4 thus clearly demonstrates that the tandem performanceis limited by parasitic absorption in the top cell in the visiblespectral region but even more in the infrared spectral region. AsCH3NH3PbI3 itself does not lead to undesired sub-band gap absorp-tion,19 the encountered top cell absorption losses rather have to berelated to the specific top cell architecture. To elucidate the origin ofthe parasitic absorption, we plot the absorptance of incrementallybuilt-up layer stacks in Fig. 5. The layer stacks were preparedidentically to the perovskite cells and resemble the structure,morphology and layer thicknesses of the perovskite cells. The FTOfront electrode alone is highly transparent in the visible spectralrange, but does exhibit pronounced absorption in the infraredspectral region. Adding the compact TiO2 layer leaves the absorp-tance almost unchanged. With the mesoscopic TiO2/perovskitecompound layer, the absorptance is already close to that of thefinished cell in the visible and slightly increases in the infraredspectral region. The spiro-MeOTAD layer and the MoOx/ITO elec-trode (in this device configuration) do not influence the visibleabsorption considerably, but both induce increased infrared absorp-tion. In summary, Fig. 5 shows that absorption losses in the infraredare caused by the top cell components FTO, spiro-MeOTAD andITO. These layers are highly doped, and the absorption can beattributed to free carrier absorption (FCA), which is proportionalto the charge carrier density and the square of the wavelength.37

In spiro-MeOTAD, the high doping is needed for efficient holeextraction.38,39 FTO and ITO have to provide lateral conductivity overseveral millimetres, and a reduced charge carrier density wouldcompromise the fill factor. Consequently, the four-terminal tandemarchitecture faces a trade-off between TCO electrical properties(top cell fill factor), and minimized FCA (bottom cell current).

To quantify the associated power losses, we separate the spec-trally resolved current losses in the top cell from those in the bottomcell according to the procedure outlined in the Appendix. Fig. 6illustrates the current loss spectra of each subcell according tothe loss mechanism (reflection from or parasitic absorption inthe top cell). Reflection by the top cell causes current losses of1.64 mA cm�2 in the top cell and 2.06 mA cm�2 in the bottomcell. The most dominant current loss is caused by parasiticabsorption, which accounts for 3.81 mA cm�2 in the top celland 5.67 mA cm�2 in the bottom cell. In total, the current lossestranslate to efficiency losses of 2.4%abs in the top cell, and4.2%abs in the bottom cell.

V. Efficiency potential

In this section, we consider different model perovskite/c-Sitandem systems, moving from the realized perovskite/c-Sitandem device presented in this article towards more optimizedsystems. We start with the four-terminal tandem device shownin Fig. 3, denominated Case 1 in Table 2. Assuming that theVOC- and FF-related losses caused by the transparent MoOx/ITOelectrode can be overcome by process and architecture optimiza-tions, we recalculate the efficiency with state-of-the-art junctionparameters (VOC and FF) taken from ref. 10. To account for theVOC illumination-dependence we recalculate VOC using the one-diode model40 (see Appendix) using an ideality factor of 1. Thisyields a tandem efficiency of 17.6% (Case 2). As the next steptoward an ultimate, but still realistic limit of the tandemefficiency we use record values for the bottom cell VOC and FF.Si solar cells are already close to their ‘‘technical’’ limit, andpublished world-record devices are therefore good approximatesFig. 5 Absorbance spectra of incremental layer stacks.

Fig. 6 Current loss spectra separated into top cell and bottom cellcurrent losses.

PCCP Paper

1624 | Phys. Chem. Chem. Phys., 2015, 17, 1619--1629 This journal is© the Owner Societies 2015

of a realistic ultimate bottom cell. The world-record c-Si solar cellis a back-contacted solar cell featuring a-Si:H/c-Si heterojunctioncontacts2. For tandem applications, however, the IR responserather than JSC under full spectrum illumination is decisive. Asthe best reported IR performance was realized with a both-sidecontacted SHJ device by carefully designing the rear contact,35

we use the record values for both-side contacted SHJ cells, VOC =750 mV, JSC = 39.5 mA cm�2, and FF = 83.2% for an illuminationof 0.1 W cm�2, as the ultimate bottom cell.41 We account forthe VOC illumination dependence using the one-diode model40

and an ideality factor of 0.7,3 and obtain a tandem efficiency of18.7% (Case 3).

Even though this tandem cell is composed of record VOC andFF values for both subcells, its efficiency still falls short behindsingle-junction c-Si solar cells. This illustrates how severely thecurrent losses due to reflection and parasitic absorption affectthe tandem efficiency. Assuming that these optical losses canbe overcome by an appropriate optical design, including anti-reflective coatings and intermediate light reflectors/couplersbetween the perovskite and the c-Si cell, we add the JSC lossesdetermined in Section IV to the JSC of the respective cell andrecalculate their efficiencies. Such an optimized system, free ofany parasitic absorption attains an efficiency of 27.6% (Case 4).The efficiency improvement relative to Case 3 results from a JSC

gain of 5.45 mA cm�2 in the top cell and 7.73 mA cm�2 in thebottom cell, corresponding to a efficiency gain of 4.04% and4.85%, respectively. Interestingly, the tandem device of Case 4is very close to current matching conditions. As stated above,we consider a perovskite top cell with VOC = 1050 mV and FF =80% to be a realistic performance goal. In a current matchedtandem device, represented by Case 5, an efficiency of 31.62%would be attainable with this top cell. Even with a current lossof 0.5 mA cm�2 for each of the subcells, the tandem efficiencywould still be 430%.

VI. Large-area four-terminal modules

The most straightforward way to realize perovskite/c-Si tandemmodules is to mechanically stack a large-area perovskite modulewith – and couple it optically to – c-Si solar cells. In this devicearchitecture, both subcells can be fabricated independently,yielding thus great engineering flexibility for optimization. Inter-connection would be done by laser patterning the top cell to aperovskite top module as sketched in Fig. 8 (see Appendix), andthen wiring the top and bottom modules. Alternatively to laserpatterning, the top cell could also be contacted with a metal grid,but this solution is likely to be too costly. In any case, however,lateral conductance through the top cell TCOs over one cellstripe width w is needed, and can cause resistive power lossesdepending on the TCO sheet resistance. As free carriers in theTCOs also cause absorption, and thus current losses in thebottom cell, the total tandem efficiency is a delicate balancebetween minimized resistive losses in the top module andminimized current losses in the bottom cell due to parasiticabsorption in the top cell (see the Appendix for details). Thetrade-off between absorption and sheet resistance of a TCO canbe described by the figure of merit hFOMi, defined here as theinverse of the product of the effective top cell sheet resistance RSh

and its weighted sub-band gap absorptance hAi (see Appendix):

FOMh i ¼ 1

RSh � Ah i: (1)

We evaluate the tandem efficiency according to the analyticalmodel developed in the Appendix (eqn (16)) for a photo-inactiveinterconnection width of 100 mm, which is a typical value forstate-of-the-art processing, assuming an average transparency ofthe interconnection area m of 50%. We plot the four-terminaltandem module efficiency using the cell parameters of Case 5(Table 2), optimized with respect to the cell stripe width, as a

Table 2 Photovoltaic conversion efficiencies of perovskite/c-Si tandem solar cells. Case 1 summarizes the demonstrator device discussed in section III.Cases 2 and 3 are theoretical efficiencies calculated with the published record VOC and FF values. For Case 4, the current losses, calculated from thedevice of Case 1, were added to the respective cells. Case 5 corresponds to the ideal current matched tandem discussed in Section I. Values that changewith respect to the preceding case are indicated in red. ‘‘Experiment’’ denotes experimental results presented in this paper

Paper PCCP

This journal is© the Owner Societies 2015 Phys. Chem. Chem. Phys., 2015, 17, 1619--1629 | 1625

function of the top cell TCO sheet resistance and the figure ofmerit in Fig. 7 (color contours). To illustrate the effect of the TCOoptical properties more explicitly, we also plot in Fig. 7 theweighted top cell sub-band gap absorptance hAi (black dashedequiabsorptance lines) according to eqn (1).

For hFOMio 1 sq O�1, the top cell is increasingly transparentfor higher sheet resistances and the tandem efficiency increases.For higher FOM (41.5 sq O�1), the tandem efficiency firstincreases with an increase in the sheet resistance (0 o RSh o15 O sq�1) as the TCO becomes increasingly transparent, andthen decreases due to the resistive losses in the top cell. Anefficiency 430% can be reached with a TCO which yields a hFOMiZ 3.5 sq O�1 for a sheet resistance range of 8–13 O sq�1.

The transparent top cell presented above yields an effectivefigure of merit of 0.07, calculated from the sub-band gapabsorptance shown in Fig. 4 and its average sheet resistanceof 52.5 O sq�1 (average of the employed FTO and ITO layerswith 15 O sq�1 and 90 O sq�1 respectively). We remark that thisvalue is an effective value, which includes absorption of theentire top cell and is moreover affected by absorption enhance-ment due to scattering and multiple reflections.

Finally, we evaluate the properties of single layer TCOsoptimized for solar cell applications. With ITO layers optimizedfor infrared transmittance,42 a hFOMi of 0.97 can be attainedwhen neglecting absorption in any other layers. Even highertransparency can be achieved with hydrogen-doped tin oxide(IO:H)42 or plasma-treated zinc oxide (ZnO).43 These materialsprovide weighted figure of merits of 2.41 and 2.97, respectively,

and according to Fig. 7 would thus enable tandem moduleefficiencies of 30% for a wide range of TCO parameters.

VII. Outlook

The results presented here illustrate that perovskite/c-Si tandemsolar cells have the potential to surpass the c-Si efficiency record(25.6%),2 but they also point out the challenges that are asso-ciated with their realization.

In detail, the following development steps have to be accom-plished to achieve tandem efficiencies of 30%:

(a) Marginal sub-band gap absorption of the perovskiteabsorber material,

(b) Top cell efficiency improved to the level of 418% forwafer-sized devices,

(c) Transparent top cell electrodes that do not compromise VOC

and FF (values in the range of 1000 mV and 80%, respectively),(d) A broad-band (500–1200 nm) transparent top cell result-

ing in a total parasitic absorption loss of o1 mA cm�2, and, forfour-terminal tandem devices, yield a weighted figure of meritof 42.

For the CH3NH3PbI3 material, criterion (a) is fulfilled.19

Criterion (b) is met when considering the recent record valueson small devices,9,44 but module efficiencies have not exceededB5% so far.45

The transition from a top cell with 418% efficiency (b) to atop cell with transparent electrodes showing 18% efficiency (c)can likely be achieved by process and material optimization.Broad band transparency (c), however, is limited by the devicearchitecture, as our results indicate that the standard architec-ture employed here induces parasitic absorption losses of5.45 mA cm�2 and 7.73 mA cm�2, respectively, in the top andbottom cells.

As the four-terminal device requires lateral conductivity, andthus sufficient doping and thickness of the two intermediateTCOs, it is inherently affected by parasitic absorption. More-over, upscaling requires micropatterning of the top cell andinterconnection to minimodules, further increasing opticallosses. These losses can be mitigated by monolithically inter-connecting the two cells, relaxing the requirements to lateralconductance and potentially eliminating completely the twointermediate TCOs (cf. Fig. 1b).

Importantly, the monolithic device also facilitates upscalingbecause it requires lateral conductivity only in the topmost elec-trode. Our results also illustrate the importance of implementinghighly transparent alternatives for the doped layers (FTO, ITO andspiro-MeOTAD), such as tetrathiafulvalene derivative,46 IO:H42 orZnO films.43 The most promising candidates for the bottom cellappear to be Si heterojunction devices. Besides their excellent open-circuit voltages they feature a record-high infrared response.35

Moreover, as a bottom cell they are not affected by optical lossesin the blue region, which is their most important drawback at full-spectrum illumination, and can thus easily be optimized for thespectrum and illumination intensity in a tandem device. Inaddition, Si heterojunction solar cells already feature a built-in

Fig. 7 Four-terminal tandem module efficiency (color contours with whitelabels) and weighted top cell sub-band gap absorptance (black dashed lineswith black labels) calculated with the cell parameters of Case 5 (Table 2) as afunction of the sheet resistance and the weighted figure of merit of the topcell TCOs. The module efficiency was calculated for an interconnectionstripe width and a transparency of 100 mm and 50%, respectively, andoptimized with respect to the cell stripe width. A maximum moduleefficiency 430% is attained at sheet resistances of 8–13 O sq�1 with aFOM 4 3.5 sq O�1.

PCCP Paper

1626 | Phys. Chem. Chem. Phys., 2015, 17, 1619--1629 This journal is© the Owner Societies 2015

recombination junction47 (between doped amorphous Si andTCO) which is required for a monolithic interconnection of thebottom and top cells.

Finally, we remark that for optimum spectral utilization, amaterial with slightly higher top cell band gap (1.7–1.8 eV) andthus also higher VOC would be beneficial. This underlines theimportance of alternative absorber materials such as mixediodide–bromide perovskites.48,49 Using this material system,Noh et al. were able to achieve a band gap of 1.7 eV for acomposition of CH3NH3Pb(I0.75Br0.25)3, a promising result inview of tandem applications.48

IX. Conclusion

In this article, we presented a four-terminal perovskite/c-Sitandem solar cell and derived limiting efficiencies of practicalperovskite/c-Si tandem devices.

The realized four-terminal tandem device is based on a c-Siheterojunction bottom cell and a CH3NH3PbI3 top cell featur-ing a transparent MoOx/ITO hole contact. With the transparentMoOx/ITO hole contact, the top cell features an impressiveinfrared-transparency of up to 455%. The four-terminal tandemdevice yields an efficiency of 13.4%, with similar contributions ofthe top cell (6.2%) and the bottom cell (7.2%). We employed thefour-terminal tandem solar cell as a test device for an experi-mental power loss analysis, splitting the current losses intoreflection- and parasitic absorption-induced losses based on anoptical analysis, and separating them according to the respectivesubcell. Assuming the published record values for the individualsubcells, an efficiency of 27.6% would be attainable with anoptimized system. Anticipating improved values for the top cellvoltage (1050 mV) and the fill factor (80%), we identify a tandemefficiency 430% as a mid-term goal.

AppendixSpectral analysis of current losses

In order to quantify the power losses in the tandem solar cell,we split the photon current losses to those photons that wouldhave contributed to top cell current and to those that wouldhave contributed to bottom cell current if they were not lost byreflection or parasitic absorption.

Photons that are incident on the tandem system can either bereflected by the top cell, or transmitted through it, or generate aphotocurrent, or be absorbed without generating photocurrent.We denote the probabilities of the respective processes as RTop,TTop, EQETop and PATop, with the parasitic absorption PATop

defined as the difference between absorbed photons and photonsthat are absorbed and collected as photocurrent, PATop = ATop �EQETop. These four processes fulfill the identity

1 = RTop + TTop + EQETop + PATop. (2)

For photons that are not reflected, we can calculate the probabilitiesof the other three processes, TTop

0, EQETop

0and PATop

0, by normal-

izing with 1 � RTop (rearranging eqn (1)): TTop0 ¼ TTop

�1� RTop

� �,

EQETop0 ¼ EQETop

�1� RTop

� �, and PATop

0 ¼ PATop

�1� RTop

� �.

In analogy to eqn (1), the identity

1 ¼ TTop0 þ EQETop

0 þ PATop0

(3)

holds for these three expressions. The reflected photon flux,

JR ¼ðRTopfAM1:5Gdl; (4)

can thus be split up according to

JR ¼ðRTop EQETop

0 þ PATop0 þ TTop

0� �

fAM1:5G dl: (5)

fAM1.5G denotes the AM1.5G solar spectrum50 and l is thewavelength. For those photons that are neither reflected norparasitically absorbed, we calculate the probabilities for trans-mission and photocurrent generation by normalization with oneminus the internal fraction of not parasitically absorbed photons,

1� PATop0

(rearranging eqn (2)): Ttop00 ¼ Ttop

0�1� PAtop

0� �, and

EQETop00 ¼ EQETop

0�1� PATop

0� �, for which the identity 1 ¼

TTop00 þ EQETop

00holds. We can then split up the current loss

caused by parasitic absorption in the top cell,

JPA0 ¼ðPATop

0fAM1:5Gdl; (6)

into the parts affecting the top and bottom cells,

JPA0 ¼ðPATop

0EQETop

00 þ TTop00

� �fAM1:5Gdl (7)

Using eqn (4) and (6) we can now separate the lossesaffecting the top cell short-circuit current density,

JLoss;Top ¼ð

RTopEQETop

0 þ PATop0EQETop

00� �

fAM1:5Gdl; (8)

from those that affect the short-circuit current density of thebottom cell,

JLoss;Btm ¼ð

RTopTTop0 þ PATopTTop

00� �

EQEBtmfAM1:5Gdl:

(9)

Eqn (7) and (8) allow us to determine the light-generatedcurrent of the corresponding cell if the current losses weremitigated, J 0

L = JExpSC + JLoss. Here we omit the indices Top and

Btm for better clarity, und use the superscript Exp ( JExpSC , VExp

OC

and ZExp) for the experimentally determined values of the loss-affected device.

We account for the illumination-dependence of the open-circuit voltage using the one-diode model40

VOCðJLÞ ¼mkBT

qln

JL

J0þ 1

� �; (10)

with the light-generated current density JL, the diode idealityfactor m, the absolute temperature T, and the Boltzmannconstant kB. Throughout this paper the temperature is fixedat 25 1C. We model the top cell using an ideality factor of 1 andthe bottom cell using an ideality factor of 0.7. J0 is the diodesaturation current density, which is a lumped expression for all

Paper PCCP

This journal is© the Owner Societies 2015 Phys. Chem. Chem. Phys., 2015, 17, 1619--1629 | 1627

recombination pathways, and is calculated from eqn (10)with the experimental values JExp

SC and VExpOC . For simplicity, we

assume the fill factor to not depend on the illumination.The efficiency of the loss-free cell can then be calculated as

Z0 = FF�J0L�VOC( J0

L)/Pin, (11)

with Pin being the incident light power, and the efficiency losscaused by reflection and parasitic absorption is

Z0 � ZExp = FF�{ J 0L�VOC( J 0

L) � JExpSC �JExp

OC }/Pin. (12)

Four-terminal module model

Thin-film photovoltaic modules are commonly fabricated bylaser patterning of the large-area solar cell layer stack, seeFig. 8. This results in an active area lw for an aperture areal(w + m), with l being the length and w the width of one cellstripe and m the total width lost due to laser scribing. Withstate-of-the-art techniques, the required three laser scribesresult in m E 100 mm. The interconnection between individualcells within the module is ensured by the front and back TCOs,causing a relative power loss of51

DPP¼ 2RSh

J

3Vw2; (13)

due to their respective lateral sheet resistance RSh, and J and Vthe current through and the voltage at the TCO. The moduleefficiency is thus

ZTop;Module

¼ JMPP;TopVMPP;Top 1� 2RShJMPP;Top

3VMPP;Topw2

� �lw

lðwþmÞ

� �Pin:

(14)

JMPP,Top and VMPP,Top are the top cell current density and voltageat the maximum power point (MPP), respectively. The factor of2 comes from the two TCOs. Please note that we neglectshadowing of the top cell by the front TCO, because most TCOsare highly transmittive in the top cell spectral region.

However, the light intensity incident on the bottom cell inthe tandem module is reduced by the top cell due to free carrierabsorption in the top cell TCOs and the partially non-transparent nature of the top cell interconnection. The doping

in the top cell TCOs causes absorption by free carriers and thusreduces the light intensity that is incident on the bottom cell.The efficiency of the bottom cell whose short-circuit current isreduced by the weighted absorptance hAi and the absorptanceof the interconnection areas Am is

ZBtm ¼ JMPP;BtmVMPP;Btm 1� ATop

�� � w

wþm

�

þ 1� Amð Þ m

wþm

�Pin;

(15)

and the tandem module efficiency is

ZTotal = ZBtm + ZTop. (16)

The weighted absorptance hAi is the integrated spectral absor-bance of the top cell below the top cell absorber band gap lGap,weighted with the bottom cell current at each wavelength:

Ah i ¼ð1lGap

ATopEQEBtmfAM1:5Gdl

,ð1lGap

EQEBtmfAM1:5Gdl:

(17)

To calculate hAi we assume the top cell band to be at 800 nmand use the external quantum efficiency EQEBtm of an ideal,110 mm thick c-Si solar cell.3

Sample preparation

Silicon heterojunction solar cells were prepared from phosphorus-doped 1 O cm, 250 mm thick (100) c-Si wafers as described inref. 52.

Perovskite devices were prepared as described in ref. 10except for the back contact. PbI2 was purchased from Sigma-Aldrich and the material CH3NH3I was synthesized followingthe procedure described in ref. 53. A compact TiO2 layer wasdeposited onto the patterned and cleaned FTO (Solaronix, TCO22-15) substrate by spray pyrolysis at 450 1C using titaniumdiisopropoxide bis(acetylacetonate) (75 wt% in 2-propanol,Sigma-Aldrich) diluted in ethanol. A 300 nm thick mesoporousTiO2 layer was then deposited on the c-TiO2/FTO substrate byspin coating followed by annealing at 500 1C. PbI2 was dissolvedin N,N-dimethylformamide at a concentration of 1 M at 70 1C. APbI2 solution was then spin coated on the m-TiO2/c-TiO2/FTOsubstrate at 6500 rpm for 30 s. After drying the PbI2-coated glassat 70 1C for 10 min, the substrate was dipped in a solutionof CH3NH3I in 2-propanol (10 mg ml�1) for 30 s, rinsed with2-propanol, and dried at 100 1C for 10 min. Spiro-MeOTADsolution was prepared by dissolving in 1 ml of chlorobenzene72.3 mg of spiro-MeOTAD (Merck), 28.8 ml of 4-tert-butylpyridine(Sigma-Aldrich), 17.5 ml of a stock solution of 520 mg ml�1

lithium bis(trifluoromethylsulfonyl)imide (Sigma-Aldrich) inacetonitrile and 14 ml of a stock solution of 300 mg ml�1

tris(2-(1H-pyrazol-1-yl)-4-tert-butylpyridine)cobalt(III) bis(trifluoro-methylsulfonyl)imide (FK209) in acetonitrile, and spin coating onthe perovskite/m-TiO2/c-TiO2/FTO substrate at 4000 rpm for 30 s.The cells were then finished with either a MoOx/ITO stack

Fig. 8 Interconnection scheme for thin-film minimodules. Three laserscribes are employed to (1) pattern the front TCO (yellow), (2) pattern theabsorber, and (3) pattern the back TCO to isolate the cells. The photo-active cell width is w, and the width of the photo-inactive interconnectionarea is m.

PCCP Paper

1628 | Phys. Chem. Chem. Phys., 2015, 17, 1619--1629 This journal is© the Owner Societies 2015

electrode or a MoOx/Ag electrode. A metal grid was not applied tothe perovskite solar cell.

Characterization

Reflection and transmission spectra were recorded on a PerkinElmer Lambda 900 photospectrometer using an integratingsphere. The external quantum efficiency of the top cell wasmeasured using an IMT spectral response setup designed for Sithin-film solar cells, and equipped with a grating monochro-mator and a xenon lamp. The bottom cell EQE was measuredusing a commercial PVtools system dedicated for c-Si wafersolar cells. The short-circuit currents were calculated from theEQE measurements. Current–voltage curves were measuredusing a WACOM sun simulator with 0.25 cm2 and 4 cm2

shadow masks to define the active cell areas of the top andbottom cells, respectively. For the top cell I–V curve, theillumination was set to 0.1 W cm�2. To measure the I–V curveof the bottom cell, the illumination was adjusted using greyfilters such that the bottom cell JSC matched the value calcu-lated from the EQE measured in tandem configuration.

Acknowledgements

The project comprising this work was evaluated by the SwissNational Science Foundation and funded by Nano-Tera.ch withSwiss Confederation financing and by the Office Federal del’Energie, Switzerland, under Grant SI/501072-01.

Notes and references

1 SEMI, International Technology Roadmap for Photovoltaics(ITRPV), http://www.itrpv.net/Reports/Downloads/2014/, accessed30.07.2014.

2 K. Masuko, M. Shigematsu, T. Hashiguchi, D. Fujishima,M. Kai, N. Yoshimura, T. Yamaguchi, Y. Ichihashi, T. Yamanishi,T. Takahama, M. Taguchi, E. Maruyama and S. Okamoto,Achievement of more than 25% conversion efficiency withcrystalline silicon heterojunction solar cell, Denver (CO),USA, 2014.

3 A. Richter, M. Hermle and S. W. Glunz, IEEE J. Photovol.,2013, 3, 1184–1191.

4 D. M. Powell, M. T. Winkler, H. J. Choi, C. B. Simmons,D. B. Needleman and T. Buonassisi, Energy Environ. Sci.,2012, 5, 5874.

5 P. Gao, M. Gratzel and M. K. Nazeeruddin, Energy Environ.Sci., 2014, 7, 2448.

6 G. Hodes, Science, 2013, 342, 317–318.7 A. Kojima, K. Teshima, T. Miyasaka and Y. Shirai, Meet.

Abstr., 2006, MA2006–02, 397.8 A. Kojima, K. Teshima, Y. Shirai and T. Miyasaka, J. Am.

Chem. Soc., 2009, 131, 6050–6051.9 NREL Efficiency Chart, http://www.nrel.gov/ncpv/images/

efficiency_chart.jpg, accessed 17.11.2014.

10 J. Burschka, N. Pellet, S.-J. Moon, R. Humphry-Baker,P. Gao, M. K. Nazeeruddin and M. Gratzel, Nature, 2013,499, 316–319.

11 G. E. Eperon, V. M. Burlakov, P. Docampo, A. Goriely andH. J. Snaith, Adv. Funct. Mater., 2014, 24, 151–157.

12 M. Liu, M. B. Johnston and H. J. Snaith, Nature, 2013, 501,395–398.

13 P. Docampo, J. M. Ball, M. Darwich, G. E. Eperon andH. J. Snaith, Nat. Commun., 2013, 4, 2761.

14 J. You, Z. Hong, Y. Yang, Q. Chen, M. Cai, T.-B. Song, C.-C.Chen, S. Lu, Y. Liu, H. Zhou and Y. Yang, ACS Nano, 2014, 8,1674–1680.

15 J. M. Ball, M. M. Lee, A. Hey and H. J. Snaith, Energy Environ.Sci., 2013, 6, 1739–1743.

16 K. Wojciechowski, M. Saliba, T. Leijtens, A. Abate andH. J. Snaith, Energy Environ. Sci., 2014, 7, 1142–1147.

17 H.-S. Kim, C.-R. Lee, J.-H. Im, K.-B. Lee, T. Moehl,A. Marchioro, S.-J. Moon, R. Humphry-Baker, J.-H. Yum,J. E. Moser, M. Gratzel and N.-G. Park, Sci. Rep., 2012, 2, 1–7,DOI: 10.1038/srep00591.

18 G. E. Eperon, S. D. Stranks, C. Menelaou, M. B. Johnston,L. M. Herz and H. J. Snaith, Energy Environ. Sci., 2014, 7,982–988.

19 S. De Wolf, J. Holovsky, S.-J. Moon, P. Loper, B. Niesen,M. Ledinsky, F.-J. Haug, J.-H. Yum and C. Ballif, J. Phys.Chem. Lett., 2014, 5, 1035–1039.

20 H. J. Snaith, J. Phys. Chem. Lett., 2013, 4, 3623–3630.21 M. A. Green, A. Ho-Baillie and H. J. Snaith, Nat. Photonics,

2014, 8, 506–514.22 R. F. Service, Science, 2013, 342, 794–797.23 R. F. Service, Science, 2014, 344, 458.24 M. McGehee, Optimizing Perovskite Semiconductors for Tandem

Solar Cells, MRS Spring Meeting, San Francisco, CA, USA, 2014.25 P. Loper, B. Niesen, S. Moon, S. Martin de Nicolas,

J. Holovsky, Z. Remes, M. Ledinsky, F. Haug, J. Yum, S. DeWolf and C. Ballif, IEEE J. Photovol., 2014, 4, 1545–1551.

26 G. E. Eperon, V. M. Burlakov, A. Goriely and H. J. Snaith,ACS Nano, 2014, 8, 591–598, ESI†.

27 C. Roldan-Carmona, O. Malinkiewicz, A. Soriano, G. MınguezEspallargas, A. Garcia, P. Reinecke, T. Kroyer, M. I. Dar,M. K. Nazeeruddin and H. J. Bolink, Energy Environ. Sci.,2014, 7, 994.

28 Z. M. Beiley, M. G. Christoforo, P. Gratia, A. R. Bowring,P. Eberspacher, G. Y. Margulis, C. Cabanetos, P. M.Beaujuge, A. Salleo and M. D. McGehee, Adv. Mater., 2013,25, 7020–7026.

29 G. Y. Margulis, M. G. Christoforo, D. Lam, Z. M. Beiley,A. R. Bowring, C. D. Bailie, A. Salleo and M. D. McGehee,Adv. Energy Mater., 2013, 3, 1657–1663.

30 J. Gjessing, E. S. Marstein and A. Sudbø, Opt. Express, 2010,18, 5481–5495.

31 B. Demaurex, S. De Wolf, A. Descoeudres, Z. Charles Holmanand C. Ballif, Appl. Phys. Lett., 2012, 101, 171604.

32 C. Battaglia, S. Martın de Nicolas, S. De Wolf, X. Yin,M. Zheng, C. Ballif and A. Javey, Appl. Phys. Lett., 2014,104, 113902, DOI: 10.1063/1.4868880.

Paper PCCP

This journal is© the Owner Societies 2015 Phys. Chem. Chem. Phys., 2015, 17, 1619--1629 | 1629

33 Y. Zhao, A. M. Nardes and K. Zhu, Appl. Phys. Lett., 2014,104, 213906.

34 D. Pysch, A. Mette and S. W. Glunz, Sol. Energy Mater. Sol.Cells, 2007, 91, 1698–1706.

35 Z. C. Holman, A. Descoeudres, S. De Wolf and C. Ballif,IEEE J. Photovol., 2013, 3, 1243–1249.

36 S. De Wolf, A. Descoeudres, Z. C. Holman and C. Ballif,Green, 2012, 2, 7–24.

37 R. A. Smith, Semiconductors, Cambridge University Press,New York, USA, 1978.

38 U. Bach, D. Lupo, P. Comte, J. E. Moser, F. Weissortel,J. Salbeck, H. Spreitzer and M. Gratzel, Nature, 1998, 395,583–585.

39 J. Burschka, A. Dualeh, F. Kessler, E. Baranoff, N.-L. Cevey-Ha,C. Yi, M. K. Nazeeruddin and M. Gratzel, J. Am. Chem. Soc.,2011, 133, 18042–18045.

40 P. Wurfel, Physics of Solar Cells - From Principles to NewConcepts, Wiley-Vch Verlag GmbH & Co KgaA, Weinheim,2005.

41 M. Taguchi, A. Yano, S. Tohoda, K. Matsuyama, Y. Nakamura,T. Nishiwaki, K. Fujita and E. Maruyama, IEEE J. Photovol.,2014, 4, 96–99.

42 L. Barraud, Z. C. Holman, N. Badel, P. Reiss, A. Descoeudres,C. Battaglia, S. De Wolf and C. Ballif, Sol. Energy Mater. Sol.Cells, 2013, 115, 151–156.

43 L. Ding, S. Nicolay, J. Steinhauser, U. Kroll and C. Ballif,Adv. Funct. Mater., 2013, 23, 5177–5182.

44 H. Zhou, Q. Chen, G. Li, S. Luo, T.-b. Song, H.-S. Duan,Z. Hong, J. You, Y. Liu and Y. Yang, Science, 2014, 345, 542–546.

45 F. Matteocci, S. Razza, F. Di Giacomo, S. Casaluci, G. Mincuzzi,T. M. Brown, A. D’Epifanio, S. Licoccia and A. Di Carlo, Phys.Chem. Chem. Phys., 2014, 16, 3918–3923.

46 J. Liu, Y. Wu, C. Qin, X. Yang, T. Yasuda, A. Islam, K. Zhang,W. Peng, W. Chen and L. Han, Energy Environ. Sci., 2014,DOI: 10.1039/c4ee01589d.

47 A. Kanevce and W. K. Metzger, J. Appl. Phys., 2009, 105, DOI:10.1063/1.3106642.

48 J. H. Noh, S. H. Im, J. H. Heo, T. N. Mandal and S. I. Seok,Nano Lett., 2013, 13, 1764–1769.

49 S. A. Kulkarni, T. Baikie, P. P. Boix, N. Yantara, N. Mathewsand S. Mhaisalkar, J. Mater. Chem. A, 2014, 2, 9221.

50 IEC, Photovoltaic devices - part 3: measurement principles forterrestrial photovoltaic (PV) solar devices with reference spectralirradiance data, International Electrotechnical Commission,2008.

51 A. Shah, Thin-Film Silicon Solar Cells, EPFL Press, 2010.52 A. Descoeudres, Z. C. Holman, L. Barraud, S. Morel, S. De

Wolf and C. Ballif, IEEE J. Photovol., 2013, 3, 83–89.53 J.-H. Im, J. Chung, S.-J. Kim and N.-G. Park, Nanoscale Res.

Lett., 2012, 7, 353, DOI: 10.1186/1556-276X-7-353.

PCCP Paper