11790 | J. Mater. Chem. C, 2017, 5, 11790--11797 This journal is©The Royal Society of Chemistry 2017

Cite this: J.Mater. Chem. C, 2017,

5, 11790

Simulation study of environmentally friendlyquantum-dot-based photovoltaic windows

Rostyslav Lesyuk,ab Vladimir Lesnyak, c Axel Herguth,d Dmytro Popovych, b

Yaroslav Bobitski,ef Christian Klinke ag and Nikolai Gaponik *c

We modeled a prototype of a photovoltaic window, a passive source of clean energy, using a Monte

Carlo ray-tracing method. We considered different geometries, material properties, and edge solar cells

to determine the optimal conditions and possible electrical power yield. The modeled photovoltaic

window prototype was based on colloidal luminescent low-toxic I–III–VI quantum dots (core/shell

CuInS2/ZnS nanocrystals) with large Stokes shifts, high quantum yields, and tunable spectral properties.

We also showed the influence of the quantum dot absorption/emission spectra on the resulting

spectrum of transmitted light (i.e., the visual appearance of the photovoltaic window) using a chromati-

city diagram.

Solar energy harvesting is a promising method to guarantee theglobal energy supply in the 21st century. A gradual and steadyincrease in energy production from renewable resources iscurrently occurring. The photovoltaic (PV) energy sector is amongthe most rapidly growing in this area. The installed capacity of PVsolar plants and roof systems has increased 45-fold in the lastdecade, with an average annual increase of more than 46% thatreached 231 GWp of installed capacity producing 253 TW h ofelectrical energy annually by the end of 2015 (corresponding to1.1% of global electricity consumption).1,2 However, the growthrate of installed capacity has slowed recently to 28% in 2015, withthe growth in solar energy decreasing to 32.6% from 91% in2011.1,3 By maintaining these growth rates, and assuming abalance increase in global electricity consumption of B1%annually, solar electricity production could effectively contributeto the electricity balance and even cover our global electricityneeds by 2035 (estimation based on ref. 1). However, to retainthese annual rates, new solar plants with the areas comparable to

countries such as Qatar should be installed by 2030. Further-more, the newly installed area will need to be even larger tocompensate for the degradation of presently installed PV capa-cities. These rates are not achievable with the current level oftechnology and cost. Presumably, a further slowing in solarenergy growth will occur, in which case the overlap of producedsolar energy and consumed electricity will be shifted far intothe future. In contrast, the increasing production of electricalvehicles could gradually move the energy balance of the planetfurther from oil and gas toward the electricity sector, which iscurrently unable to cover these needs.

Breakthrough technologies, including safe nuclear fusion,environmentally friendly fuel cells, wind plants, and low-costsunlight conversion, have received much attention in modernresearch and development. In the 1970s, the concept of lumines-cent collectors (and subsequent luminescent solar concentrators,LSCs) was introduced4,5 as a potential way to produce low-costsolar energy. Since then, extensive analytical and experimentalresearch has been performed, mainly based on the application ofdyes in LSCs.6–9 To improve the performance of LSCs, interestingapproaches such as multidye thin-film LSCs,5,10 photonic bandfilters and reflectors inside the LSCs,11–13 and plasmonicallyenhanced LSCs,14,15 have been proposed. With the utilizationof new materials, including metal oxide and rare earth element-doped glasses and polymers, exotic carbon based light-harvestingantennas, biomaterials (see extensive reviews ref. 10 and 16), andespecially quantum dots (QDs),17 LSC research has been revita-lized and reevaluated. Colloidal QDs have become attractivecandidates as fluorophores in LSCs due to their stability, broadand high absorption, and large Stokes shifts, which reducesparasitic self-absorption.18 Most research in this area has focusedon II–VI QDs and quantum rods (mainly CdSe-based19) due to the

a Institute for Physical Chemistry, University of Hamburg, Grindelallee 117,

20146 Hamburg, Germanyb Pidstryhach Institute for Applied Problems of Mechanics and Mathematics of NAS

of Ukraine, Naukowa str. 3b, 79060 Lviv, Ukrainec Physical Chemistry and Center for Advancing Electronics Dresden (cfAED),

TU Dresden, Bergstr. 66b, 01062 Dresden, Germany.

E-mail: nikolai.gaponik@chemie.tu-dresden.ded Photovoltaic Division, Department of Physics, University of Konstanz,

Universitatsstr. 10, 78457 Konstanz, Germanye Photonics Department, Lviv National Polytechnic University, Bandery str. 12,

79013 Lviv, Ukrainef University of Rzeszow, Department of Mechatronics and Automation,

Rejtana Alley 16C, 35-001 Rzeszow, Polandg Department of Chemistry, Swansea University, Singleton Park, Swansea, SA2 8PP,

Wales, UK

Received 30th June 2017,Accepted 23rd October 2017

DOI: 10.1039/c7tc02945d

rsc.li/materials-c

Journal ofMaterials Chemistry C

PAPER

This journal is©The Royal Society of Chemistry 2017 J. Mater. Chem. C, 2017, 5, 11790--11797 | 11791

well-developed synthetic protocols, extensively studied physicaland chemical properties, and excellent photoluminescence (PL)quantum yields (QYs).20 To overcome limitations resulting fromthe high reabsorption losses of II–VI QDs, Stokes shift engineer-ing approaches have been reported, including the formation of agiant shell around the CdSe core21,22 and changing the lumines-cence mechanism by doping with metal ions.23 As shown byKennedy et al., NIR emitting II–VI QDs can significantly improvethe efficiency of QD-based LSCs.24 In the last decade, I–III–VI-based QDs, such as CuInS2, have become attractive alternativematerials25 for replacing Cd-based QDs in LEDs,26,27 solarlight harvesting applications,28,29 and bioimaging.30 In fact,state-of-the-art I–III–VI QDs possess very large Stokes shifts(B400–500 meV) and high QYs approaching 0.8 in solution,31

and show flexible spectral shifting due to their tunable compo-sition. These materials show excellent promise for achieving aPLQY of 0.5 or higher in the near future using QDs embeddedin transparent matrices.

Recently, this class of low-toxic QDs has been consideredfor application in LSCs. In particular, Klimov et al.29,32 andSumner et al.33 reported extensive studies on the application ofCuIn(SxSe1�x)2-based QDs in LSCs using both analytical andnumerical mathematical modelling, and presented corres-ponding prototype composite materials and devices. In parti-cular, the authors concluded that scattering losses would have agreater role with increasing QD loading for small geometricgain and dominative high waveguide losses as the device isscaled up.33 In a previous report on spectral converters, weshowed that Zn–Cu–In–S (ZCIS) QDs can outcompete CdSe QDsowing to their larger Stokes shifts and corresponding lower self-absorption losses.34 Furthermore, ZCIS is a potentially non-toxic material, which is an important property for large-scaleapplications.

In parallel to LSCs, the harvesting of dissipative sunlightenergy is viable using PV windows, which could be realized inthe form of semitransparent LSCs without a back reflector(Fig. 1a). In contrast to LSCs, little attention has been paid toPV windows. To our knowledge, no special simulations havebeen performed for such systems based on QDs. At locationsexcessively illuminated by solar light, buildings equipped withsemitransparent windows (or walls) could act as a passiveenergy source. Moreover, it is important for the quality ofinterior lighting conditions that these windows retain a naturalspectrum (color) of transmitted light. Several concepts havebeen proposed toward this goal. One such approach is thedirect application of semitransparent PV cells,35 which remainschallenging because of the high costs involved. An alternativeidea is to develop UV and NIR light absorbers in dye sensitizedsolar cells that will transmit visible light without distortion, asproposed by Zhang et al.36 A similar physical effect (UV and NIRlight absorption) using an alternative harvesting concept wasreported by Zhao et al.,37 who created fully transparent andenergy producing windows using a previously reported planarsolar concentration strategy (for example, in ref. 38). In thiscompletely transparent device, molecules absorb small portionsof invisible UV light and reemit this energy as secondary light at

longer wavelengths partially guided to the edges by total internalreflection. At the edges, thin strips of conventional silicon-based solar cells are attached. However, extending the rangeof absorbed light is highly desirable. From this perspective, thedevelopment of colloidal QDs has made this a possibility, whileensuring better long-term stability.

Recently, we have developed a procedure for the fabricationof QD-in-polymer composite materials suitable for large-scalesamples with controllable particle load through QD surfaceengineering and copolymerization.42 In the present study, weused the experimental knowledge and characterization resultsfrom our previous work to analyze the feasibility of CuInS2

(CIS)-based composites with different QYs and geometries forapplication in PV windows using Monte Carlo simulations.Furthermore, we performed a quantitative assessment of thepower generation potential of this technology for two levels ofaverage transparency in the visible optical range (90% and 70%)and analyzed the spectral content of light transmitted throughthe PV window prototype.

Results and discussion

The Monte Carlo or ray-tracing approach9,43–45 consists of thecalculation of a single photon path through the PV window with

Fig. 1 (a) Schematic representation of the modeled PV window prototype.Solar cells are placed along the perimeter of the down-converting plate.(b) External quantum efficiencies of solar cells considered as photovoltaicelements in our calculations. Data for c-Si were measured by ourselves, thosefor a-Si, CZTS, and CdTe cells are adopted from ref. 39–41, respectively.

Paper Journal of Materials Chemistry C

11792 | J. Mater. Chem. C, 2017, 5, 11790--11797 This journal is©The Royal Society of Chemistry 2017

possibilities to be transferred according to ray optics, namely,being absorbed, emitted, reabsorbed, or reflected from theinterface and refracted to the outer medium. Each of theseprocesses has a certain probability that can be calculated usingthe program. The ray interactions with interfaces are describedby the Fresnel equation for unpolarized light, accounting forthe incidence angle and refractive indices of the media. Afterreflection coefficient, R, for the current ray (photon) is calcu-lated (from 0 to 1), the uniformly distributed random valuewithin [0,1] is generated and compared with R to determine thereflection or transmission event. To trace the photon throughthe medium, a one-shot path is generated as a function of arandom number and the extinction coefficient (more detailscan be found in ref. 34 and 43). This path is compared with thedistance to the interface in the direction of the photon, and adecision is made regarding absorption or intersection with theinterface. If the photon is absorbed, the emission probability issampled similarly using the QY value. The wavelength of thenewly emitted photon is assigned according to the normaldistribution of the spectrum with the predefined central wave-length. The model accounts for the experimentally observedspectrum broadening at longer wavelengths. After photontracing is complete, the final light flux profile (spectrum) f(l)at the edges and behind the window can be collected. The shortcircuit current can be calculated as an integral value of photons(converted to free carriers) with respect to the external quantumefficiency (EQE) curve of a solar cell according to the followingequation:46

Jsc ¼ q

ðl2l1

fðlÞEQEðlÞdl (1)

In our model, l1 and l2 correspond to 285 and 1250 nm,respectively.

As photovoltaic elements for the conversion of light trans-mitted to edges, we considered as examples two silicon-basedand two alternative thin-film solar cells, which have industrialhistory and are established technologies or show promisingstability. Importantly, they consisted of abundant materials, asfollows: monocrystalline (c-Si) and amorphous silicon (a-Si) asthe most common outdoor and indoor solar cells, Cu2ZnSnS4

(CZTS) as an environmentally friendly and low-cost absorber,and CdTe, which has the lowest energy payback time reportedto date and little sensitivity to low-light conditions. At present, wecannot predict which of the new emerging solar cell technologies(such as organic, perovskite, dye-sensitized, or QD-based) willachieve stable characteristics and a large-scale market. Accord-ingly, we limited our consideration to the well-established solarcell technologies mentioned above.

For c-Si, we measured the EQE characteristics of a commerciallyavailable cell. The following EQE characteristics were adopted: a-Si(glass/TCO/p–i–n a-Si:H/ZnO:Ag, without nanoparticles) fromref. 39, CZTS (IPCE above 8%) from ref. 40, and CdTe fromref. 41. Corresponding EQE curves are presented in Fig. 1b.

The incident light spectrum was adopted from the ASTMG173-03 (AM1.5G) spectrum of ref. 47 with normalization to

80 mW cm�2 flux. The absorption and emission spectra of ourexperimentally available Zn–Cu–In–S QDs embedded in poly-styrene are shown in Fig. 2b (inset). These spectra are typical forCIS-based systems and exhibit large Stokes shifts and reducedself-absorption compared to most II–VI counterparts, such asCdSe/ZnS QDs.

The input absorption and emission parameters for differentQDs were simulated as a set of spectra based on experimentalspectra of a QD sample with emission maximum at 675 nm andabsorption shoulder at 545 nm by gradually shifting the spectraso that the emission maxima covered a range of interest from475 to 1095 nm. For each spectral profile, the Stokes shift,which was acquired from experimental curves (B440 meV), wasretained. For example, as mentioned above, the emissionspectrum with the maximum at 675 nm corresponds to theabsorption curve with first excitonic shoulder at B545 nm,while the PL maximum at 995 nm corresponds to the excitonicshoulder at 740 nm. The absorption curves were further nor-malized in this way, such that the average optical transmit-tance, T, of the prototype in the range of human eye sensitivity(400–730 nm) remained nearly constant (we chose two caseswith T = 0.7 or T = 0.9). These spectral changes are shown inFig. 2a and correspond to typically observed changes in the sizeand/or composition of CIS-based QDs.31 Through this spectraldesign we were able to define the optimal spectral profile foreach solar cell and study the impact of the spectral propertiesof the PV window prototype on the light flux that passed thePV window and appeared for the observer as a slightly shady,but realistically colored, day light behind the window. Wedefined the optimal spectral profile as that which appears nearlycolorless.

For every emission profile, the short circuit current for solarcells at the four edges of a 100 � 100 mm PV window prototypecan be calculated. Based on the open circuit voltage (Voc) and

Fig. 2 (a) Extinction spectra taken as an input parameter for different modelQDs with emission maxima from 475 to 1075 nm. The average transmittancein the visible region remains constant (90 or 70% for presented simulations).Experimental absorption (b) and PL (c) spectra of several ZCIS/ZnS, CIS, andCISxSe1�x QD solutions showing similar spectral profiles to those simulatedin (a). Inset in (b): typical experimental emission and absorption spectra ofZCIS QDs in polystyrene, as fabricated in our laboratory. Insetin (c): photo of a luminescent composite under UV light, fabricated usingthe surface engineering approach (ZCIS/ZnS QDs in polystyrene42) forpotential application in PV windows.

Journal of Materials Chemistry C Paper

This journal is©The Royal Society of Chemistry 2017 J. Mater. Chem. C, 2017, 5, 11790--11797 | 11793

the fill factor (FF), the maximal power can be evaluated. In oursimulations, we considered different aspect ratios of the lateraldimensions, the thickness of the PV window, and their influ-ence on the overall efficiency. For the power evaluation, we tookinto account that the Voc depended on the short circuit current

Isc according to the relationship Voc /kT

qln

Isc þ IS

IS

� �, and due

to the shunt-resistance effect, the overall efficiency of the cell ingeneral decreased under low-light conditions. For simplicity, inthe present study we assumed that the short circuit scaled strictly

linearly with the light flux, meaning that Isc1

xSun

� �¼ 1

xIsc 1 Sun½ �

for 1 o x o 10, and that only Voc and FF are affected by low-lightconditions. The excitation and emission kinetics of the QDs weresimplified here. We did not introduce the ‘‘off’’ state andassumed that the luminescent intensity had a linear responseunder sun illumination, as reflected in the average QY. QY valueswere iterated from 0.1 to 0.9 (we noted that QY = 1 for colloidalQDs embedded in a polymer is hard to achieve in practice).A state-of-the-art value of 0.5 has recently been reported21 and istaken here as a reasonable average value.

Photons not wave-guided to the edges by total internalreflection (referred to usually as ‘‘cone losses’’) were traced bythe model until they refracted from the PV window. Before this,they can be reabsorbed/reemitted and contribute to the shortcurrent (at the edges) or refract through the top or the frontinterfaces, thus contributing to the color of the PV window. Thisfeature of the model makes it more correct in comparison withthe analytical models, which typically ignore ‘‘cone losses’’. Theinterface between QDs containing composite and the solar cellwas modelled without an antireflective coating. Scattering andabsorption losses in the matrix were considered negligible.Incident photons were traced normally to the front surface.

In the PV window, a spectral conversion (down-shifting)occurs. This process guides light to the edges, which is stronglydependent on the QY of the embedded QDs (Fig. 3). PV windowswith constant transmittance in the visible region can transfer

more energy if luminescent particles emit at longer wavelengths(such as in the NIR), because the light of a broader spectral rangeis absorbed (visible and part of the NIR, as shown in Fig. 2).Reduction of the transmittance from 90 to 70% caused theguided power to increase by a factor of 2.5 for QDs emittingat B700 nm, and by a factor of 3 for wavelengths (l) 4910 nmfor the QY of 0.5. For example, for normal incidence of80 mW cm�2 AM 1.5 solar irradiation, a modeled prototypewith a surface of 100 � 100 mm and a thickness of 1 mm cantransfer 22 mW of power to the edges with T = 90% and QY = 0.1,nearly 115 mW with T = 90% and QY = 0.5, and 260 mW withT = 70% and QY = 0.5.

In Fig. 4, the curves of the short circuit current as a functionof the maximum emission wavelength of the QDs are presentedfor CdTe, c-Si, a-Si, and CZTS edge solar cells. An aspect ratio of0.01 was selected for the modeled PV window (1 mm thickness,100 � 100 mm front surface) in this simulation, which corre-sponds to a geometric gain (G) of 25 (defined as the ratiobetween the front surface area and the area of the four edges,

G ¼ Sfront

Sedges). Dotted graphs in Fig. 4 show statistical dispersion,

which comes from the precision error of our Monte Carlomethod, although the main tendencies can still be clearlyfollowed. Due to the wide spectral range and high conversionefficiency, c-Si was superior to thin-film solar cells. However,due to the distinct dependence of Voc and FF on the light flux,this advantage may disappear under real conditions. For thec-Si solar cell, an optimal spectral range for the QD emissionlay between 875 and 955 nm, while for CdTe and CZTS it was675–755 nm, and 555–595 nm for a-Si. As shown, Isc stronglydepended on the QY. In Fig. 5, these dependencies are sum-marized for 90 and 70% PV window transmittances for dimen-sions of 100 � 100 � 1 mm. The simulations were performedfor each solar cell with corresponding optimal central emissionwavelengths.

We observed the linear dependency of current on QY (Fig. 5).As the simulations were performed for different spectral pro-files for each solar cell (optimal selection for each cell), theincrease in Isc upon changing from 90% to 70% transmittancediffered. This change was most pronounced for the c-Si solarcell (factor of 2.8), while for other solar cells the photocurrentincreased by a factor of 2.17–2.25. This behavior of the c-Si solarcell can be explained by its better performance in the NIRspectral region. According to our simulation method, we usedthe spectral profile for the central emission wavelength 915 nm.To keep the transmittance constant at 70%, absorption wasreduced in the blue range, but increased in the red range andNIR region. In the latter region, the integral amount of photonswas higher, meaning that the overall increase in Isc for the c-Sisolar cell was steeper (Fig. 5b).

Along with the influence of the QD spectral properties on theperformance of the PV window, the geometry of the device alsoalters the output. An increase in the aspect ratio (understood asan increase in thickness of the PV window by constant frontarea) and a corresponding decrease in the geometric gain pumpedges with more photons able to reach them with less reflection

Fig. 3 Power transferred to the edges of the modelled PV windowprototype (100 � 100 � 1 mm) with CIS-based QDs having varied emissionwavelengths and transmittances of 90% and 70% as a function of theQD emission spectrum. Device is illuminated by the 80 mW cm�2 sunflux (AM1.5).

Paper Journal of Materials Chemistry C

11794 | J. Mater. Chem. C, 2017, 5, 11790--11797 This journal is©The Royal Society of Chemistry 2017

events at the front and the back interface. Therefore, morepower can be transferred and harvested. At a lower G, thecurrent in the solar cell reaches saturation. This is exemplifiedby the c-Si based PV window in Fig. 6a. In contrast, an increasein the device thickness influences the concentration effect of

the PV window. Therefore, we observe a decrease in the powerconcentration factor (c-factor, defined as a ratio of flux densi-ties incident onto the edges and front surface, Fig. 6b). In thiscase, a decrease in solar cell efficiency is expected. For example,

Fig. 4 Short circuit current of the solar cells at the edges as a function of QD PL spectrum in the PV window prototype (wavelength represents centralemission wavelength): (a) c-Si, (b) a-Si, (c) CdTe, and (d) CZTS (incident AM1.5 solar spectrum with light flux density of 80 mW cm�2; transmittance of thePV window set to 0.9).

Fig. 5 Photocurrent of edge solar cells as a function of QY for a PVwindow prototype with surface area of 100 � 100 mm, thickness of 1 mm,and geometric gain (G) of 25 (incident spectrum AM1.5, 80 mW cm�2).Data provided for transmittances of (a) 90% and (b) 70%, respectively.

Fig. 6 (a) Photocurrent and (b) power concentration factor (c-factor) of ac-Si solar cell at the edges of a PV window prototype as a function of varyinggeometric gain, G (for 100 mm � 100 mm surface area), at constantemission and absorption profile of CIS QDs (lemission = 755 nm, underAM1.5, 80 mW cm�2 illumination, transmittance of the PV window = 70%).QY varies in the range 0.1–0.9.

Journal of Materials Chemistry C Paper

This journal is©The Royal Society of Chemistry 2017 J. Mater. Chem. C, 2017, 5, 11790--11797 | 11795

an increase in the aspect ratio from 0.01 to 0.02 boosts thephotocurrent by 50%. Meanwhile, the c-factor decreases fromnearly 0.8 to 0.6 for QDs with a central emission wavelength of755 nm and a QY of 0.5. We referred to the study on the effectof low-light conditions on solar cells performance by Randallet al.48 Based on data for several c-Si solar cells and modules,we noted that the efficiency should not decrease more than by2.5% for a c-factor drop from 0.8 to 0.6. Therefore, thickerwindows would be beneficial to the performance. However, thedesign, mass, and price of the matrix (such as polymer or glass)may be an essential restrictive factor. To our knowledge, for large-scale mass production applications, an aspect ratio of more than0.01 is unlikely (for example, 1 cm thickness with 1 m2 surfacearea). Therefore, we adopted this value as a standard for furtherconsiderations. The c-factor dependency on the geometric gain inFig. 6b showed an initial quasilinear increase (defined by geo-metry) with further saturation defined by waveguide losses. Thisbehavior was consistent with analytical models of LSCs32 andprimarily comes from thermodynamic restrictions of the concen-tration for every luminescent moiety. Additionally, we noted that,under the given conditions (Fig. 6b), the c-factor can exceed unityif the QY of QDs containing composite materials reaches 0.7.

For future consumers of PV windows, the transmitted lightspectrum is of high importance. The Monte Carlo tracing is aconvenient instrument for studying changes in the spectrum oftransmitted light behind the window, in combination with thechromaticity diagram. Again, we gradually shifted the spectralprofile of the QDs inside the modeled window and collected theoutput spectra, as shown in Fig. 7a. Furthermore, based onthese spectra (converted to power per area), the CIE† 1931coordinates were calculated.

Modification of the transmitted spectrum and the appearanceof the window color were caused by a high absorption coefficientof the QDs at shorter wavelengths (black arrow, Fig. 7a), and theappearance of QD emission through the ‘‘escape cone’’ (redarrows). Therefore, the incident white light can be essentiallyshifted to green, brown, or red tones. The corresponding evolu-tion of chromaticity coordinates is shown in Fig. 7b. The pointsin the diagram follow an elliptical curve. We noted that the QDsemitting in the spectral region 515–675 nm contributed to theappearance of the color. The closest points to white light in thecolor space corresponded to l 4 795 nm. The spectrum inFig. 7a, which corresponds to a central emission wavelength of475 nm, provides a too low system efficiency, and cannot beconsidered for potential application. By combining these find-ings with the analysis of data shown in Fig. 4, we were able toconclude that, to keep the PV window colorless and ensure highefficiency of the device, we should use QDs emitting at l 4795 nm in combination with silicon edge solar cells or solar cellswith similar spectral properties.

Notably, CIS/ZnS or ZCIS/ZnS QDs do not exhibit an efficientluminescence in the NIR spectral region.49 The bulk energy gapof 1.53 eV can be increased to 2 eV and beyond, causing emissionin the spectral range 600–700 nm. An example is shown in Fig. 2c.

To achieve further emission shifts to the NIR, CIS nanocrystals canbe doped/alloyed with selenium, as reported by Meinardi et al.29

Based on the obtained data for four solar cells, we evaluated themaximum power generated by a CIS-based PV window prototype,according to the expression Pmax = FF�Isc�Voc�b. For the efficiencychange under low light conditions, we introduced correction factorb, adapted from literature. b reflects the consolidated efficiencychange under low illumination conditions, contributed both byFF and Voc. The short circuit current, Isc, was derived from oursimulations for QY = 0.5, c-factor = 0.78 for the aspect ratio = 0.01,G = 25, and FF and Voc taken from the literature for each cell. Wenoted that, in an ideal case, CdTe solar cells attached to the edgesof a 100� 100 mm PV window could produce E41 mW at T = 90%and E95 mW at T = 70% for QY = 0.5 (by 80 mW cm�2 incidentAM1.5 spectrum). The power almost doubled with a QY of 0.9. Thevalues of c-Si corresponded to 34 and 85 mW for QY = 0.5, and58 and 163 mW for QY = 0.9. Although the values of CdTe and c-Siwere similar, CdTe solar cell showed better performance. Onereason for this might be the input data, which used a CdTe EQEcurve adapted from the literature for a cell, but might differ(decrease) slightly in Isc, Voc, and FF in the complete module.For c-Si, we used an EQE curve measured using a commerciallyavailable cell. Another reason could be the different QDs thatshould be used for different systems. For example, for the CdTesystem, the best performance can be achieved with QDs that

Fig. 7 (a) Spectra of transmitted light through the PV window prototype(aspect ratio = 0.01, G = 25, average transmittance = 70%) with varying QDspectral properties. As an example, the spectra for luminescent particlesemitting with QY = 0.5 at lmax = 475 nm, 635 nm, 795 nm, 955 nm, and1075 nm are shown. The thick line is the AM 1.5G Sun spectrum (80 mW cm�2).(b) Positions of the transmitted spectra through a simulated PV window onthe chromaticity diagram (CIE 1931), exemplified for an aspect ratio = 0.01,average transmittance = 70%, and QY = 0.5.

† International Commission on Illumination

Paper Journal of Materials Chemistry C

11796 | J. Mater. Chem. C, 2017, 5, 11790--11797 This journal is©The Royal Society of Chemistry 2017

emit close to 700 nm. However, as discussed, the correspondingspectral profile of the QDs would introduce a color tone to thePV window. For c-Si, the optimal region was 875–955 nm. Alongwith emission, the absorption spectra were changed in ourmethod (see Fig. 2 and method description), which may result inslightly more favorable conditions for CdTe solar cells. Furtherreason is the high Voc and optimized EQE of the CdTe cells,which provide better power conversion. CZTS and a-Si solar cellswould produce lower amounts of energy. Another drawback ofthe a-Si solar cell is poor spectral matching for the condition of acolorless PV window to be fulfilled.

To evaluate the overall system performance for the modelledprototype, the optical efficiency, Z (Z = Pproduced/Pincident), fordifferent system parameters was calculated (Table 1). For maxi-mal produced power, Pmax, at QY = 0.9 the concentration factorexceeded 1 and was close to 1.4, which should cause an increasein the cell efficiency for most semiconductor solar cells, but mayworsen the solar cell long-term performance, especially in sys-tems with an a-Si absorber.50 We found that optimized systemsfor c-Si or CdTe solar cells could have optical efficienciesexceeding 2% in 70% transparent windows, which is promisingfor future applications. As discussed in the literature, the self-absorption effect in solar concentrators and PV windows cansufficiently suppress system performance.19,29,33 In the case ofCIS or ZCIS QDs, this detrimental effect is sufficiently reduced,but not eliminated, due to the large Stokes shift, compared withCdSe/ZnS core/shell QDs.25 However, for the 100 � 100 mmwindow, self-absorption still reduced the energy transfer by afactor of approx. 1.1–1.7 for QY = 0.5, depending on the spectralemission–absorption profile of the given particles and the QY(in our experience, the absorption overlap is higher in the blueregion and decreases toward the NIR). To fabricate the simulateddevice on a 1 m2-scale, substantial experimental effort should bemade toward overcoming parasitic losses, firstly by solving theQY drop problem in QD-based luminescent composites.

Conclusions

We assessed the potential application of CuInS2-based QDsin photovoltaic windows as a passive source of clean energyfor building-integrated technologies. Using Monte Carlo ray

tracing simulations, we defined the optimal spectral regions ofQD emissions for c-Si, a-Si, CdTe, and CZTS solar cells at theedges of an environmentally friendly PV window prototype. Weanalyzed the influence of window transmittance, PLQY of theQDs, and window geometry on the photocurrent and power ofsimulated devices based on spectral profiles of CuInS2 QDsobtained by colloidal synthesis. Our calculations showed that thepower produced by a 100 � 100 mm CIS QD-based PV windowprototype can reach up to 85–95 mW with an optical perfor-mance of 1.1–1.2%, assuming a QY of 0.5 and 70% averagetransmittance in the visible region, which can be increased to175 mW (2.2% optical performance) if the QY can be boosted to0.9 (incident AM1.5G 80 mW cm�2 Sun spectrum). The design ofthe PV window should find a compromise between the tenden-cies toward increased thickness, which improves the photo-response, and decreased concentration factor, which reducesthe solar cell efficiency. By transferring the transmitted spectrumto the color space, we defined the color of the light behind thewindow (as delivered to the consumer) and found that QDs withan emission maximum at l 4 795 nm provided inside illumina-tion conditions closest to white light. This imposed restrictionson the choice of appropriate solar cells attached to the edges. Wefound that the c-Si solar cell was the optimal choice for thispurpose. Finally, the results showed that the modeled PV windowscan be a useful passive energy source. The vital condition isimproving the QY of semitransparent composite materials basedon QDs in specified spectral regions.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

RL and NG acknowledge the Dresden Fellowship Program 2016under Grant Agreement No. F003661-553-61D-1716100. The researchleading to these results received funding from the 7th EuropeanCommunity Framework Programme under Grant AgreementNo. 304980 (ERC Starting Grant ‘‘2D-SYNETRA’’) and from theGerman Research Foundation (DFG) under project LE 3877/1-1.

Table 1 Calculated maximal power and optical efficiency of the modelled prototype PV window for four solar cells at QY = 0.5 and 0.9. For each system(QDs + solar cells), the optimal conditions with respect to generated power were considered

Solar cell Isc [mA] Voc [V] FF b Pmax [mW], QY = 0.5 Pmax [mW], QY = 0.9

CdTe 62 (T = 90%)143 (T = 70%)

0.85 V 0.78 Z1 41.1 (Z = 0.5%)94.6 (Z = 1.2%)

Z75 (Z = 0.94%)Z176 (Z = 2.2%)

CZTS 48 (T = 90%)103 (T = 70%)

0.66 V 0.65 1a 20.6 (Z = 0.26%)44 (Z = 0.55%)

41 (Z = 0.5%)89 (Z = 1.1%)

c-Si 76 (T = 90%)190 (T = 70%)

0.6 V 0.76 0.98 33.9 (Z = 0.42%)85.1 (Z = 1.1%)

58 (Z = 0.73%)163 (Z = 2.1%)

a-Si 28 (T = 90%)64 (T = 70%)

0.81 V 0.47 Z1 10.7 (Z = 0.13%)24.4 (Z = 0.3%)

Z18 (Z = 0.5%)Z40 (Z = 0.5%)

a No data found for CZTS performance under low-light conditions.

Journal of Materials Chemistry C Paper

This journal is©The Royal Society of Chemistry 2017 J. Mater. Chem. C, 2017, 5, 11790--11797 | 11797

References

1 BP Statistical Review of World Energy, June 2016. www.bp.com/statisticalreview.

2 Photovoltaics Report. Fraunhofer Institute for Solar EnergySystems, ISE. November 2016. www.ise.fraunhofer.de.

3 Renewables 2016. Global Status Report. REN21. www.ren21.net, Report, ISBN 978-3-9818107-0-7.

4 W. Weber and J. Lambe, Appl. Opt., 1976, 15, 2299–2300.5 A. Goetzberger and W. Greube, Appl. Phys. A: Mater. Sci.

Process., 1977, 14, 123–139.6 J. Batchelder, A. Zewail and T. Cole, Appl. Opt., 1981, 20,

3733–3754.7 C. L. Mulder, P. D. Reusswig, A. Beyler, H. Kim, C. Rotschild

and M. Baldo, Opt. Express, 2010, 18, A91–A99.8 M. S. de Cardona, M. Carrascosa, F. Meseguer, F. Cusso and

F. Jaque, Appl. Opt., 1985, 24, 2028–2032.9 L. H. Slooff, E. E. Bende, A. R. Burgers, T. Budel, M. Pravettoni,

R. P. Kenny, E. D. Dunlop and A. Buchtemann, Phys. StatusSolidi RRL, 2008, 2, 257–259.

10 B. C. Rowan, L. R. Wilson and B. S. Richards, IEEE J. Sel.Top. Quantum Electron., 2008, 14, 1312–1322.

11 J. C. Goldschmidt, M. Peters, A. Bosch, H. Helmers,F. Dimroth, S. W. Glunz and G. Willeke, Sol. Energy Mater.Sol. Cells, 2009, 93, 176–182.

12 R. Reisfeld, Opt. Mater., 2010, 32, 850–856.13 L. Xu, Y. Yao, N. D. Bronstein, L. Li, A. P. Alivisatos and

R. G. Nuzzo, ACS Photonics, 2016, 3, 278–285.14 S. Chandra, J. Doran, S. McCormack, M. Kennedy and

A. Chatten, Sol. Energy Mater. Sol. Cells, 2012, 98, 385–390.15 M. G. Debije and P. P. Verbunt, Adv. Energy Mater., 2012, 2,

12–35.16 F. Purcell-Milton and Y. K. Gun’ko, J. Mater. Chem., 2012,

22, 16687–16697.17 W. G. van Sark, K. W. Barnham, L. H. Slooff, A. J. Chatten,

A. Buchtemann, A. Meyer, S. J. McCormack, R. Koole, D. J.Farrell and R. Bose, Opt. Express, 2008, 16, 21773–21792.

18 S. Gallagher, B. Rowan, J. Doran and B. Norton, Sol. Energy,2007, 81, 540–547.

19 J. Bomm, A. Buchtemann, A. Fiore, L. Manna, J. H. Nelson,D. Hill and W. G. van Sark, Beilstein J. Nanotechnol., 2010, 1,94–100.

20 P. Reiss, J. Bleuse and A. Pron, Nano Lett., 2002, 2, 781–784.21 F. Meinardi, A. Colombo, K. A. Velizhanin, R. Simonutti,

M. Lorenzon, L. Beverina, R. Viswanatha, V. I. Klimov andS. Brovelli, Nat. Photonics, 2014, 8, 392–399.

22 I. Coropceanu and M. G. Bawendi, Nano Lett., 2014, 14,4097–4101.

23 C. S. Erickson, L. R. Bradshaw, S. McDowall, J. D. Gilbertson,D. R. Gamelin and D. L. Patrick, ACS Nano, 2014, 8, 3461–3467.

24 M. Kennedy, S. McCormack, J. Doran and B. Norton, Sol.Energy, 2009, 83, 978–981.

25 K. E. Knowles, K. H. Hartstein, T. B. Kilburn, A. Marchioro,H. D. Nelson, P. J. Whitham and D. R. Gamelin, Chem. Rev.,2016, 116, 10820–10851.

26 P. Ilaiyaraja, P. S. Mocherla, T. K. Srinivasan and C. Sudakar,ACS Appl. Mater. Interfaces, 2016, 8, 12456–12465.

27 J. Zhang, R. Xie and W. Yang, Chem. Mater., 2011, 23,3357–3361.

28 H. McDaniel, A. Y. Koposov, S. Draguta, N. S. Makarov,J. M. Pietryga and V. I. Klimov, J. Phys. Chem. C, 2014, 118,16987–16994.

29 F. Meinardi, H. McDaniel, F. Carulli, A. Colombo, K. A.Velizhanin, N. S. Makarov, R. Simonutti, V. I. Klimov andS. Brovelli, Nat. Nanotechnol., 2015, 10, 878–885.

30 W. Guo, N. Chen, C. Dong, Y. Tu, J. Chang and B. Zhang,RSC Adv., 2013, 3, 9470–9475.

31 P.-H. Chuang, C. C. Lin and R.-S. Liu, ACS Appl. Mater.Interfaces, 2014, 6, 15379–15387.

32 V. I. Klimov, T. A. Baker, J. Lim, K. A. Velizhanin andH. McDaniel, ACS Photonics, 2016, 3, 1138–1148.

33 R. Sumner, S. Eiselt, T. B. Kilburn, C. Erickson, B. Carlson,D. R. Gamelin, S. McDowall and D. L. Patrick, J. Phys.Chem. C, 2017, 121, 3252–3260.

34 R. Lesyuk, V. Marinov, E. K. Hobbie, A. Elbaradei,I. Tarnavchyk and Y. Bobitski, Sol. Energy Mater. Sol. Cells,2016, 151, 52–59.

35 J. Ajuria, I. Etxebarria, W. Cambarau, U. Munecas, R. Tena-Zaera, J. C. Jimeno and R. Pacios, Energy Environ. Sci., 2011,4, 453–458.

36 K. Zhang, C. Qin, X. Yang, A. Islam, S. Zhang, H. Chen andL. Han, Adv. Energy Mater., 2014, 4, 1301966.

37 Y. Zhao, G. A. Meek, B. G. Levine and R. R. Lunt, Adv. Opt.Mater., 2014, 2, 606–611.

38 M. J. Currie, J. K. Mapel, T. D. Heidel, S. Goffri and M. A. Baldo,Science, 2008, 321, 226–228.

39 X. Chen, B. Jia, J. K. Saha, B. Cai, N. Stokes, Q. Qiao,Y. Wang, Z. Shi and M. Gu, Nano Lett., 2012, 12, 2187–2192.

40 B. Shin, O. Gunawan, Y. Zhu, N. A. Bojarczuk, S. J. Chey andS. Guha, Prog. Photovoltaics, 2013, 21, 72–76.

41 M. Gloeckler, I. Sankin and Z. Zhao, IEEE J. Photovolt., 2013,3, 1389–1393.

42 R. Lesyuk, B. Cai, U. Reuter, N. Gaponik, D. Popovych andV. Lesnyak, Small Methods, 2017, 1, 1700189.

43 S. A. Prahl, M. Keijzer, S. L. Jacques and A. J. Welch,Dosimetry of laser radiation in medicine and biology, 1989,vol. 5, pp. 102–111.

44 B. Richards and K. McIntosh, Prog. Photovoltaics, 2007, 15,27–34.

45 A. Kerrouche, D. Hardy, D. Ross and B. Richards, Sol. EnergyMater. Sol. Cells, 2014, 122, 99–106.

46 U. Rau, Phys. Rev. B: Condens. Matter Mater. Phys., 2007, 76,085303.

47 http://rredc.nrel.gov/.48 J. Randall and J. Jacot, Renewable Energy, 2003, 28, 1851–1864.49 Q. A. Akkerman, A. Genovese, C. George, M. Prato, I. Moreels,

A. Casu, S. Marras, A. Curcio, A. Scarpellini, T. Pellegrino,L. Manna and V. Lesnyak, ACS Nano, 2015, 9, 521–531.

50 R. Ruther and J. Livingstone, Sol. Energy Mater. Sol. Cells,1995, 36, 29–43.

Paper Journal of Materials Chemistry C