luminescent solar concentrator

7/21/2019 luminescent solar concentrator

1/24

2012 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim12

REVIEW

wileyonlinelibrary.com Adv. Energy Mater. 2012, 2, 1235

www.MaterialsViews.comwww.advenergymat.de

1. IntroductionWith the realities of dwindling oil reserves now affecting muchof everyday life, world governments have increasingly turnedtowards renewable energy technologies in an attempt to ll theever-increasing gap. In particular, there is an increasing interestin bringing solar energy systems to the built environment.Buildings account for about 40% of total energy use, including70% of total electricity use and 40% of emissions in moredeveloped countries, [1] and about 25% of all energy use glo-bally,[2] much of this related to our inability to control sunlight:

tremendous energy resources are used toboth heat and cool buildings and providearticial lighting systems. [1] The sceneof an ofce with shuttered windows andburning lights in the daytime is familiar toeveryone.

The European Committee has decreedthat all new buildings be near-zero energyby 2020.[2] This demands that architectsand the building industry integrate energysaving and energy generation into thedesign of new buildings. This puts consid-erable pressure on the architect wishingto meet these requirements: how to createbuildings both pleasing to the eye, yet ener-getically neutral? To give the maximumfreedom in their designs, the devicesincorporated to save and/or generateenergy have to be easily implemented intothe design and adaptable to the situation.Adaptability could also make retrottingof existing structures during renovationseasier and cheaper. One energy source that

is readily available in a built environmentis the sun, which is a clean, safe, inex-haustible, and reliable energy source.

Adapting traditional silicon-basedphotovoltaics (PVs) for urban integrationhas proven difcult. Some examples are

shown in Figure 1a. PV cells remain relatively expensive.[3] Often the modules are heavy and limited in coloration (usuallyblack or dark blue) and shape (rectangular): variability in bothsize and shape are generally lacking in the marketplace.

Several techniques have been adopted to add color to photo-voltaic cells, including adding uorescent dyes to the encapsu-lation layer,[4] adjusted color-reecting lms, [5] use of organicphotovoltaics[6] and other techniques, all with their own advan-tages, but also carrying disadvantages. For example, color-reecting lms can reduce performance [5] and organic cells stillsuffer from reduced efciencies and lifetimes, although progressis being made. [79] If transparency is desired for a window tting,for example, this capability is limited in traditional solar panels,and requires use of thin-lm cells, standard solar cells arrangedwith physical spaces between them (see Figure 1b), or cells withholes drilled through them. Furthermore, PV cells respond opti-mally to direct sunlight, while in the built environment muchof the sunlight is diffuse due to scattering and reections byother objects such as trees, other buildings, and even clouds,and shading is common, which reduces the performance of

Dr. M. G. Debije, P. P. C. VerbuntFunctional Organic Materials & DevicesDepartment of Chemical Engineering and ChemistryEindhoven University of Technology5600 MB Eindhoven, The NetherlandsE-mail: [email protected]

DOI: 10.1002/aenm.201100554

Michael G. Debije* and Paul P. C. Verbunt

Thirty Years of Luminescent Solar Concentrator Research:Solar Energy for the Built Environment

Research on the luminescent solar concentrator (LSC) over the past thirty-odd years is reviewed. The LSC is a simple device at its heart, employinga polymeric or glass waveguide and luminescent molecules to generateelectricity from sunlight when attached to a photovoltaic cell. The LSC hasthe potential to nd extended use in an area traditionally difcult for effectiveuse of regular photovoltaic panels: the built environment. The LSC is a devicevery exible in its design, with a variety of possible shapes and colors. The

primary challenge faced by the devices is increasing their photon-to-electronconversion efciencies. A number of laboratories are working to improve theefciency and lifetime of the LSC device, with the ultimate goal of commer-cializing the devices within a few years. The topics covered here relate to theefforts for reducing losses in these devices. These include studies of novelluminophores, including organic uorescent dyes, inorganic phosphors, andquantum dots. Ways to limit the surface and internal losses are also dis-cussed, including using organic and inorganic-based selective mirrors whichallow sunlight in but reect luminophore-emitted light, plasmonic structuresto enhance emissions, novel photovoltaics, alignment of the luminophoresto manipulate the path of the emitted light, and patterning of the dye layer toimprove emission efciency. Finally, some possible glimpses of the futureare offered, with additional research paths that could result in a device thatmakes solar energy a ubiquitous part of the urban setting, nding use assound barriers, bus-stop roofs, awnings, windows, paving, or siding tiles.
http://doi.wiley.com/10.1002/aenm201100554


2/24

13wileyonlinelibrary.com


2012 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Adv. Energy Mater . 2012, 2, 1235

suitable for building integration tend to be restricted to lowerconcentration ratios. [14]

1.1. Introduction to the Luminescent Solar Concentrator

The desire, then, is to take advantage of a reective or focusingsolar concentrator (collecting light from a larger area andfocusing it onto a smaller area) while avoiding the disadvan-tages (the need for direct sunlight, tracking, and often awkwarddish-like shapes and large sizes) for integration in urban envi-ronments. An alternative to traditional semiconductor devicesthat takes the advantages of concentrator systems but whichcould still be used in the built environment could be the lumi-nescent solar concentrator (LSC) ( Figure 2), originally intro-duced more than three decades ago. The earliest reference toan LSC type device architecture appeared as a grant proposalin 1973 by Lerner, [15] soon followed by papers in the generalliterature. [1618] There was a great urry of activity in the eld,

an extended PV system signicantly. [10,11] Even many of thenew thin-lm technologies will be a challenge to adapt to thedemands of the architect and society.

In order to combat the high costs associated with solar-energy generation, a multitude of designs have been introducedto concentrate sunlight from large areas onto smaller, moreefcient PV cells, thereby reducing the surface coverage of theexpensive PV cells.[12] There are predominantly two variationson the concentrator system, those relying on light focusing, andthose with associated reectors for concentrating the sunlight.Both concentrating systems demonstrate high degrees of lightconcentration and a concurrent reduction in the price of thegenerated solar electricity. [13]

There are many utilization limitations imposed by these con-centrator systems. For one, they often need to be installed witheither single- or double-directional tracking systems to trace thepath of the sun across the sky, which adds to their expense andcomplexity. They tend to perform less well in indirect sunlight,and thus are restricted then to areas with minimal cloud coverand shading. Often the installations are quite extensive, requirea large area of land, suffer from the not in my back yard effect,and are not easily adaptable to a built-up area. Alternativeconcentrators using nonimaging, nontracking, atter devices

Paul P. C. Verbunt receiveda Bachelor of AppliedSciences degree inChemistry from the FontysHogeschool Technische

NatuurwetenschappenEindhoven, in 2002. After acouple of years in industryhe received a M.Sc. degreein Chemical Engineeringand Chemistry from theEindhoven University of

Technology in 2008. He is currently active as a PhDstudent in the Functional Organic Materials and Devicesgroup of the Chemical Engineering and Chemistry depart-ment at the Eindhoven University of Technology, lead byProf. Dirk Broer. His current research interests are focusedon luminescent solar concentrators.

Michael G. Debije received aM.Sc. in High Energy Physicsfrom Iowa State University,Ames, Iowa in 1994 and aPh.D. in Biophysics fromthe University of Rochesterin Rochester, NY, USA in2000. After completing apostdoc at the InterfacultyReactor Institute at the DelftUniversity of Technology inthe group of John Warman

in 2003, he joined the staff of the Functional OrganicMaterials and Devices group under Prof. Dirk Broer at theEindhoven University of Technology, and is responsible forthe Energy cluster within the program.

Figure 1. a) Photograph of silicon-based panels mounted on a rooftop(photo by Gray Watson). b) Photograph of an atrium element using semi-transparent silicon-based solar panels (photo courtesy Sapa BuildingSystems).


3/24

14 wileyonlinelibrary.com


REVIEW

2012 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Adv. Energy Mater. 2012, 2, 1235

of the waveguide to collect the light and convert it to electricity.Figure 3 summarizes the functionality of the device.

Other at-plate concentrating systems that do not rely onluminescence have been proposed. These systems have avariety of forms, for example, relying on diffractive elements, [27] light trapping, [28] special backside mirrors, [29] or refractive-index

variations.[30]

However, each of these designs presents their ownchallenges to produce electricity efciently enough to be eco-nomically feasible. This paper will conne itself to describingresearch on systems incorporating luminescent materials intothe waveguide.

The portion of the emitted light that is trapped inside thewaveguide depicted in Figure 3 is determined by the refractiveindex of this waveguide. According to Snells law, all photonsapproaching an interface between a material and air at an anglehigher than the critical angle will be totally reected. This crit-ical angle is dened as

c = sin 1

1

n (1)

where n is the refractive index of the waveguide. This meansthat for an isotropic emitter and a waveguide with a refractiveindex of 1.51.6, approximately 75% of all emitted photons willbe internally reected. These photons will waveguide towardsthe side of the LSC where a signicant fraction is coupled outof the waveguide by a photovoltaic cell which transforms thephotons in to electrons. However, the efciency of an LSC isnot only dependent on the trapping efciency of the waveguide.One description of the optical efciency, opt, of an LSC is

opt = (1 R )P TIR abs PLQY Stokes host TIR self (2)

where R is the reection of solar light from the waveguide sur-face, P TIR is the total internal reection efciency, abs is thefraction of solar light that is absorbed by the dye, PLQY is thephotoluminescent quantum yield of the used luminophore(s), Stokes is the energy lost due to the heat generation duringthe absorption and emission event, host is the transport ef-ciency of the waveguided photons through the waveguide, TIR

leading to a plethora of patents. The drop in oil prices in the1980s led to a near-abandonment of the research. Recent surgesin oil prices and increased awareness of the effects of powergeneration from traditional fuels on the global environmenthave revitalized research on the LSC, and it has again become

the interest of many groups around the world.The past ve years have demonstrated a considerable reawak-ening of interest in the LSC. This has been due to a numberof factors, including the increasing awareness by the generalpublic of the dwindling fossil-fuel availability, limitations on theperformance and thus capability of deployment in urban areasfor traditional PV systems and the desire from the architecturaland building sectors for more freedom in their design choices.

The LSC promises potential low costs of production. [19,20] They could provide adaptability to the needs of the architect inthat they can be made in a variety of colors, shapes, and trans-parencies, could be made exible, and should weigh less thansilicon PV panels, for example, which could make LSCs moreviable for mounting to the side of a building. [21] In addition, theLSC could function well in both direct and diffuse light, [22,23] of particular interest for countries with frequent cloud coverageor areas of persistent shady conditions, such as are typical incities, for example. If the photon in/photon out efciency ishigh enough, the cost of electricity generated by the LSC couldbe competitive.[19] While the efciency of an LSC will be lowerthan an equivalent area of a silicon PV due to decreased spec-tral usage, the reduced cost and tremendous exibility in designcould make them a viable alternative for the urban area whereit would be too expensive, inappropriate, or impractical to usestandard solutions.

1.2. Working Principle of the Luminescent Solar Concentrator

The basic LSC design allows sunlight to penetrate the top sur-face of an inexpensive plastic or glass waveguide. This light isabsorbed by luminescent molecules (which could be organicdyes, inorganic phosphors, or quantum dots, as we describe laterin this review), which are embedded in the waveguide, appliedin a separate layer on the top or bottom of the waveguide, orcontained in a liquid solution between two glass plates. [17,2426] The absorbed light is re-emitted at a longer wavelength anda fraction of the re-emitted light is trapped in the waveguideby total internal reection, becoming concentrated along theedges of the plate. Small PV cells can be attached to the edges

Figure 2. Photograph of four luminescent solar concentrators illumi-nated by ultraviolet light.

Figure 3. The luminescent solar concentrator. Incoming sunlight (greenarrow) enters the waveguide and is absorbed by a luminophore. Thelight is re-emitted at a longer wavelength (red arrow) and a fractionbecomes trapped by total internal reection. Attached to the edge(s)of the waveguide are photovoltaic cells for conversion of the light intoelectricity.


4/24




is the reection efciency of the waveguide determined by thesmoothness of the waveguide surface, and self is the transportefciency of the waveguided photons related to re-absorptionof the emitted photons by another luminophore. [17] The losseswhich inuence all these efciencies will be described in thesecond part of this review as well as the solutions proposed by

several research groups over the past thirty years.Thermodynamics limit conventional geometric concentra-tors, which consequently are inefcient at concentrating diffuselight.[31,32] These devices demonstrate maximum concentrationat close to the square of the refractive index of the material inwhich the concentrator is embedded. Therefore, these concen-trators are not very useful in a built environment, where thesunlight is very diffuse due to scattering from buildings, cars,or trees.

The absorption and re-emission of photons in LSCs changesthe entropy of the system; the maximum concentration of anLSC is therefore dependent on the heat generation during theabsorption and emission event, and thus from the Stokes shiftof the luminescent molecules. The maximum concentrationcan be approximated by:

C e 3

2e 3

1exp

e 1 e 2kT 0

(3)

where e2 and e1 are the photon energies of the emitted andabsorbed photons, respectively, k is Boltzmanns constant, andT 0 is the ambient temperature. [31,32] For a luminophore with aStokes shift of 0.2 eV, the maximum concentration is approxi-mately 2000 times.

Considerable effort has been put forth developing a widerange of models using many different approaches to describethe results of the existing devices, as well as to predict optimal,

higher-efciency LSC designs. Random-walk theory has beenused to illustrate re-absorption, [33] and Monte Carlo simula-tions have been used to investigate the performance of single-and double-lm stacked LSCs. [23,34] Different thermodynamicmodels have examined waveguide losses [32,35] which reportresults in very good agreement with experiments on test mod-ules. Recently, Meyer et al. have compared the re-absorptionprobability model reported by Weber et al. to ray-tracing resultsof luminophore-impregnated thin-lm and liquid-concentratorsystems. [36] Other work has used ray tracing to predict theperformance of LSCs, including lled [3740] and thin-lm [41] waveguides.

Each of the theoretical approaches has their advantages anddisadvantages. The thermodynamic modeling requires min-imal input and respond with swift answers but is often lim-ited to simpler geometries and limited luminescent species. [40] Ray tracing allows much more freedom in device design andnumber of luminophores and other details, but is quite com-putationally demanding. Potentially high performance levelsfor the LSC have been predicted by, for example, detailed bal-ance theory of a single stage uorescent collector using a high-efciency luminophore and wavelength-selective surface reec-tors. Such a device was predicted to demonstrate efciencies ofup to 90% that of a directly illuminated cell of equivalent size. [42]

There have also been efforts to use the concentrated lightfrom an LSC to generate high temperature thermal heat. In

these experiments, LSC plates were attached to either side ofa surface oxidized copper pipe inside an evacuated glass tubewhich acted as the collector. When exposed to external condi-tions of 980 W m 2 total illumination, a stagnation temperatureof 535 C was reached, and on a cloudy day (ambient tempera-ture of 26 C) the temperature of the absorber reached in excess

of 250 C.[43]

2. Losses of Luminescent Solar Concentratorsand Proposed SolutionsThe LSC has not yet been extensively commercialized, prima-rily owing to their modest efciencies. [44,45] A diagram depictinga number of identied loss mechanisms for LSCs taking intoaccount the factors described in Equation (2) above may begraphically depicted in Figure 4.

The rst loss of the LSC is light emitted by the luminophoreunder an angle which is refracted out of the waveguide through anescape cone rather than being reected internally (Figure 4, 1).The second loss is due to the re-absorption of emitted photonsin the waveguiding mode by subsequent luminophores due toan overlap of the emission and absorption bands of the lumi-nophores (limited Stokes shifts, Figure 4, 2). The third loss ofan LSC is a combination of losses related to the luminophoreused: one is that the luminescent molecules have limited spec-tral absorption bands which lead to incomplete incident lightabsorption (Figure 4, 3a). This light simply passes through thewaveguide and is lost through the bottom surface. Another lossrelated to the luminophore is the absorption of high energyUV photons that leads to either direct photodegradation ofthe molecules (Figure 4, 3b) such that in time a fraction of themolecules will be broken down and total emission of the LSC

will decrease, or degradation of other molecular species withinthe vicinity of the luminophore that subsequently react withit and degrade the luminophore performance. Some of theabsorbed photons are not re-emitted by the luminophores dueto limited emission quantum yield, but instead lost as heat andvibrations (Figure 4, 3c).

The PV cell at the waveguide edge has a nonuniform spec-tral response, with a fraction of incident photons being lost dueto the nite conversion efciency of the PV cell leading to a

Figure 4. Loss mechanisms in luminescent solar concentrators: 1) Lightemitted outside capture cone; 2) re-absorption of emitted light by anotherluminophore; 3a) input light is not absorbed by the luminophore; 3b) lim-ited luminophore stability; 3c) internal quantum efciency of the lumino-phore is not unity; 4) solar cell losses; 5a) reection from the waveguidesurface; 5b) absorption of emitted light by the waveguide; 5c) internalwaveguide scattering; 5d) surface scattering.


5/24



REVIEW


fourth loss (Figure 4, 4). Even optically perfect waveguides cansuffer from losses; a small part of the input light is lost throughreection from the surface of the waveguide, due to Fresnelreections (Figure 4, 5a). Waveguides used for LSCs can causelosses through parasitic absorption, especially in the nearinfrared (Figure 4, 5b). Imperfections of the waveguide surface

can cause photons in the waveguide mode to leave the surfaceand be lost (Figure 4, 5d), while imperfections in the waveguidebulk also scatter waveguided photons (Figure 4, 5c), leading toadditional surface losses.

In the next section of this paper we give an overview ofresearch done on LSCs towards reducing these losses sincetheir discovery in the 1970s.

2.1. Surface Loss

Photons emitted by luminescent molecules inside the escapecone will be lost through the surfaces (loss 1 in Figure 4).Measurements suggest that 4055% of all absorbed energy islost in this way (this translates into a 5070% loss of photons)in a 5 cm 5 cm device;[46] these high loss results have beensupported by simulations. [40,47] Surface loss is key in LSCs, andin the last few years multiple groups have done research onminimizing them by two main processes: aligning the lumino-phores and the application of selective mirrors.

2.1.1. Aligned Luminophores

Organic luminophores are often dichroic in absorptionand transmission, [4852] which opens new possibilities incontrolling the spatial distribution of emitted light, pro-vided that the physical ordering of the dyes is macroscopi-

cally controlled. The alignment of dichroic dyes in liquidcrystalline (nematic; LC) materials has been previouslyinvestigated[5356] and it has been shown that macroscopicalignment of the dichroic dyes in the LC host resulted inanisotropy and dichroism in both absorption and emission.There are a number of possible ways of aligning the lumino-phores, such as using self-aligning fluorescent nanorods, [57] incorporating the materials in a polymeric sheet which canbe physically stretched to impose alignment, [58] or mixturein a host that itself may be spontaneously aligned, such as aliquid crystal.[59] See Figure 5 for a depiction of some of thedye orientations possible using liquid crystals.

Aligning organic dyes perpendicular to the plane of thewaveguide surface (homeotropically, Figure 5b) using a host(photopolymerizable) liquid crystal leads to an emission prima-rily in the plane of the waveguide, resulting in a sharp decrease

of surface loss to less than 10%, [60,61] conrmed by simulationsusing collimated light. [47] However, in this conguration theluminophores have quite a low absorption, as their transitiondipole for absorption is parallel to the direction of the incominglight. Lower absorption naturally leads to lower edge emission,although the trapping efciency of emitted photons increases

from 65% to over 80% when vertically aligned luminophoresin LSCs are excited by an isotropic light source. [62,63] The ef-ciency of transmission of the emitted photons to the edge of thewaveguide can increase with close-to homeotropic alignment,as the path length through the luminophore-containing regionof the thin absorbing/emitting layer is minimized. [64] Com-pletely homeotropic alignment, however, could be detrimental,as most of the emitted light would be directed within the thinluminophore layer, and would result in excessive re-absorptionevents.

Liquid crystals have also been used to align the lumino-phores parallel to the waveguide surface, also known as planaralignment (Figure 5a). In this conguration, one can relativelyeasily determine the degree of alignment of the luminophoresby determining the dichroism parameter in absorption, Ra, forthe embedded luminophores through use of the equation:

R a =A par A perp

A par + 2 A perp (4)

where Apar and Aperp are the peak absorbance of the lumino-phore measured using incident light polarized parallel andperpendicular to the alignment direction of the liquid crystals,respectively. A system in perfect alignment would demonstratean Ra of 1, while a completely isotropic system would have anRa of 0.

When uorescent organic dyes exhibit a high degree of

alignment in absorption (that is, Ra >

0.5), emission is directedsuch that > 30% more energy is issued from the two edgesparallel to the alignment direction of the dye molecules whencompared to the edge output from any side of an LSC usingno dye alignment (isotropic). [59] The total emission from allfour waveguide edges is the same for both planar and isotropicdevices, suggesting the surface losses have not been reduced.An LSC congured with this planar alignment could be usedas an energy harvesting polarizer [63] for use in display applica-tions, for example. Additionally, if light is emitted primarilytowards just two edges, the number of PV cells on the LSC canbe reduced to two or even one, reducing the cost of the LSC stillfurther. Surface losses, as one might expect, are not reduced inthis conguration. [60]

An obvious next step is to consider combining the reducedsurface losses of the homeotropic orientation and the enhancedemission directions of the planar samples and apply the dyesin a tilted alignment (Figure 5c). [60,65] There are a number ofpotential ways to achieve tilted alignments in liquid-crystal sys-tems. One technique is to use modied alignment layers thatcan be specically tuned by rubbing pressure or processingtemperature to achieve the desired angular tilt [66] or by usingliquid-crystal phases that themselves have a natural tilt, such asthe Smectic C phase. However, to date these approaches haveproven difcult to produce sufciently defect-free alignmentover an extended area: work on this continues.

Figure 5. Schematic depiction of red luminophores aligned in nematicliquid crystals on top of a waveguide. The black arrow depicts the averagealignment direction: a) planar alignment, b) homeotropic alignment, andc) tilted alignment at 45 .


6/24




2.1.2. Selective Mirrors

A second way to reduce surface loss is by the applicationof wavelength-selective mirrors, the purpose of which is toallow incoming light to enter the waveguide and be absorbed,but to be reective for wavelengths emitted by the lumino-

phore[40,42,45,6776]

(see Figure 6 for the illustrated concept).Simulations have suggested that, in some cases, an ideal top-surface reector could increase edge output more than 50%. [40] Wavelength selective mirrors so far employed have been madefrom chiral nematic (cholesteric) liquid crystals, [75] inorganicdichroic Bragg reectors [40,67] and Rugate lters. [73]

A cholesteric liquid crystal is a planar nematic liquid crystalthat has been doped with a chiral molecule. The chiral mol-ecule induces a twist between adjacent layers of the liquidcrystal, the twist direction (either right- or left-handed) beingdependent on the nature of the chiral dopant. This results inthe formation of a type of helical structure through the depth ofthe layer which is capable of reecting a narrow bandwidth oflight of the same polarization handedness as the helix (that is,right-handed cholesterics will reect right-circularly polarizedlight). The wavelength of light reected is dened by the pitchof the helix, given by

= n av p sin (5)

where is the central reection wavelength, nav is the averagerefractive index of the liquid crystal host, p is the pitch, or thedistance it takes for the liquid crystal director of a plane of mole-cules in the helix to undergo a 360 twist, and is the incidenceangle of the light with respect to the normal. [77] The reectionbandwidth of the cholesteric material will be given by

= (n e n o) p (6)

where ne and no are the extraordinary and ordinary refractiveindex of the host LC. [78] For most standard LCs, the bandwidthis on the order of 5075 nm in the visible part of the spectrum.By addition of photo-responsive end groups, it is possible to

crosslink the liquid crystals to form solid lms. See Figure 7 fora depiction of the cholesteric liquid crystal.Naturally, a single cholesteric layer would only be able to

reect half of the incident sunlight (which is equally dividedbetween right- and left-circular polarizations), so to get com-plete reection of incoming light a second cholesteric layeremploying a left-handed helical structure would need to beadded. An advantage of these cholesteric systems is their easeof application: they may be cast or sprayed from solution andself-assemble into the reective state.

By applying full, narrow band reecting cholesterics to thetop surface of Lumogen F Red 305-containing waveguides (acommercial dye from BASF), up to 30% of the light that hadpreviously escaped the surface was turned into edge emission,translating into a 12% LSC output improvement using theorganic reectors. [75] In these experiments, the reectors wereseparated from the waveguide by an air gap. The reectionband which produced the highest enhancement of the edgeoutput was considerably red-shifted with respect to the emis-sion wavelength of the dye. It was shown the fractional increasein emission was higher for waveguides containing a lower dyeconcentration.

Similar enhancements to LSC edge outputs were determinedusing inorganic reectors. [45,73] In this work, waveguides con-taining the dye BA241 were topped by a Rugate lter, which isa photonic band stop lter with a sinusoidal varying refractiveindex, separated by an air gap from the top of the waveguide.

The reection bandwidth of the lter used was on the order of150 nm. By applying the lter to the waveguide, the relative ef-ciency increased by 20% (see Figure 8).

An advantage to using the organic reectors is that they arecast from solution and spontaneously form the reective layer:this is generally a much simpler and less expensive processthan the application of multilayer inorganic Bragg reectors.However, as mentioned earlier, the reection bandwidth ofstandard cholesteric liquid crystals is limited. In general, theemission spectra of the luminophores is often much broaderthan the 75 nm reection band generally present in cholestericLCs. One obvious way to increase performance of these organicreectors is by broadening the reection band. [76] The reectionbandwidth of a cholesteric liquid crystal is determined in partby the difference between the extraordinary and ordinary refrac-tive indices of the liquid crystal used, which is generally lim-ited. Thus, one way to increase bandwidth is to use host liquidcrystals with a higher n (birefringence), but the bandwidth isstill limited to perhaps 150 nm by the birefringence of the hostLCs, generally not extending beyond n = 0.4.

Another option is to layer like-handed cholesterics, eachoffset from its neighbor by 75 nm, creating a stack. Of course,given the single-handedness of the cholesteric, this wouldnecessitate at least four layers (two stacked left-handed andtwo stacked right-handed) to achieve a 150 nm reection band.A third option is to employ gradient-pitch cholesterics. [78] In

Figure 6. a) Standard LSC. Input light (black arrow) is absorbed andemitted by the luminophore outside the capture cone of the waveguide

(gray arrow) and is lost through the surface. b) The same LSC with appli-cation of the selective reector. Incident light (black arrow) is absorbedby the luminophore. The longer-wavelength emitted light (gray arrow)is directed outside the capture cone, but is reected by the selectivereection layer.


7/24


8/24




polymeric plate, thereby increasing the Stokes shift and reducingthe overlap between the absorption and emission spectra. [92]

Another option to reduce the re-absorption losses in devicesusing more than one luminophore is to physically stack anumber of waveguides, each containing a luminophore thatabsorbs in a different part of the spectrum, on top of one

another.[17,20,93]

This is particularly useful when one wishes toinclude one of the luminophores with absorption approachingthe infrared, which to date have generally demonstrated loweruorescent quantum yields. In this way, by stacking the moreefcient luminophores on top, the light is collected, convertedand transported to the attached cell without ever encounteringthe reduced-efciency luminophore. The added benet is thatluminescence emitted through the escape cone could be cap-tured by a luminophore in an adjacent waveguide rather thanbeing lost. [93]

To reduce encounters of emitted light with the lumino-phores, one may attach the luminophores in a thin layer (fromsubmicrometer thickness to hundreds of micrometers) to thesurface of the waveguide rather than lling the luminophorewithin the bulk of the waveguide, [20,26,40,41,48,49,9498] or as mul-tiple thin layers stacked on one another. [99] In this way, emis-sion light may be transported predominantly in the clear hostmaterial, and only encounters the luminescent layer with everysecond internal reection. Theoretically, the thin layers shouldperform as well as the lled waveguides. [100] In practice, caremust be exercised as the limited solubility of the luminophoresoften results in an underperformance of the thin layers due toaggregation of the luminophores, creating nonemitting absorp-tion centers. [41]

An advantage in utilizing the thin luminophore-lled layer ontop of a blank waveguide is in production: one could envisagethe large-area, inexpensive application of thin layers on either

glass or polymer hosts using a variety of techniques (includingspin coating, [41] bar coating, casting, [40] solgel techniques, [101] doctor blading, spraying, LangmuirBlodgett techniques, [26] orprinting). The thin layers could consist of luminophores in acr-ylates,[102] cellulosetriacetate,[40] polymerized liquid crystals, [59] or a host of other materials.

It is also possible to reduce the losses due to the Stokes shiftby using more than one luminophore, and employing Frsterresonance energy transfer (FRET). FRET is the direct exchangeof energy from an excited molecule to another nearby moleculewithout the emission of a photon. FRET is very sensitive todistance and other factors: the probability of a transfer beingrelated to the orientations of the interacting molecules as wellas the distance (with a sensitivity related to 1/ R6 where R is thedistance between the molecules).

By mixing a number of luminophores at high concentra-tions it was possible, through a chain of virtual emissionsand absorptions, to transfer short-wavelength input light intolong-wavelength output light at relatively high efciency. [20,103] However, the high luminophore concentrations necessary forthe FRET effect are often not achievable unless the moleculesare brought closer together by other means. [104,105] Anothermaterial design taking advantage of FRET use ordered dyenanochannel antennae (based on zeolites), which could allowenhanced directionality of light emission as well as reducedre-absorption losses. [106,107]

Another option to try to reduce re-absorption was to physi-cally space thin luminophore layers with stretches of emptywaveguide using spatially separated patterns of luminophores,so that the number of re-encounters that emitted light couldhave with other molecules was reduced: a depiction of the func-tionality of such a device is shown in Figure 9.[108] The trans-

port efciency of the photons through the LSC increased as theoverall uorescent-dye coverage decreased from 100% to 20%,and was relatively independent of the shape of the dye pattern.However, reduction in total light absorption due to the physicalgaps between absorption regions resulted in a total systemdecrease in output, as one might expect.

In order to increase the amount of light absorbed in apatterned system, a lens array on top of the LSC is being devel-oped that would increase the system output. [109] The lenseswould focus incoming light from a wide area onto small, pat-terned regions of dye on the waveguide surface, leading toimproved dye absorption. Current lens designs have demon-strated relatively efcient focusing of light over a range of 30 .By using illumination through the lenses themselves to inducecrosslinking of the dye-containing acrylate layer, one would beable to tailor the dye pattern to the expected solar light posi-tions and so effectively minimize the area coverage require-ment, as well as eliminate the difculty of correctly aligninga separately produced lens array with a dye pattern. While thisdevice demonstrates increased efciency under direct lighting,measures still need to be taken to improve performance indiffuse light.

A very recent approach has used resonance shifting, wherespacially conned emission from a bilayer cavities interacts off-resonance with the luminophore layer upon subsequent inter-actions. This is accomplished by careful variation of the lumino-phore thin-lm thickness across the area of the waveguide. The

thickness variation is on the order of a few tens of nanometersand is most appropriate for samples with luminophore layerson the order of a micrometer in thickness. In this way, initialwork appears to allow light propagation with practically no

Figure 9. Schematic demonstrating the functionality of the patterneduorescent dye waveguide. Top: Function of the standard thin-layer LSCsystem. Incident light (green) is absorbed by a uorophore and emitted ata longer wavelength (red, gray solid lines) and has a signicant probabilityof re-encountering a subsequent dye molecule and being reabsorbed, andpossibly again re-emitted (dotted red, gray lines). Bottom: With the pat-terned dye layer, some of the emission light (red) will travel along thewaveguide and never re-encounter the dye layer, leading to enhancedprobability of avoiding re-absorption.


9/24



REVIEW


re-absorption losses, and as such promises potentially signi-cant performance improvements if the design may be simplyand reproducibly applied. [110]

2.3. Luminophores: Limited Absorption, Stability,

and Luminescence Efciency

Luminophores are the driving force for concentration in lumi-nescent solar concentrators. Due to their absorption and re-emission, luminophores decouple the incoming sunlight fromthe emission light, overcoming the limitations of geometricaloptics in trapping light within waveguiding modes. The absorp-tion and re-emission of the luminophores alter the entropyof the LSC, so higher light concentrations can be reached, inaccordance with thermodynamic laws (tendue). [31,32] Accordingto the laws governing tendue (Equation (3)) the maximumconcentration achievable by the waveguide is determined by theStokes shift of the luminophore used. Besides the maximumconcentration of an LSC, the luminophore is the most deter-mining factor of light transport efciency. In Goetzbergersequation (Equation (2)) for LSC efciency, the fraction of solarlight absorbed, the photoluminescent quantum yield, the energylost due to the heat generation during the absorption and emis-sion events, the transport efciency of the waveguided photonsrelated to re-absorption of the emitted photons, and the spec-trum of the photons reaching the PV cell are all determined bythe characteristics of the luminophore. Combining the needsfor maximum concentration, maximum efciency, and max-imum lifetime, the luminophore is the single most importantcomponent in the device. An effective luminophore must meetall the following requirements:

Broad spectral absorption High absorption ef ciency over the whole absorption

spectrum Large Stokes shift (no or low overlap in absorption and emis -

sion spectra) High luminescent ef ciency (quantum yield) Matching the emitted photons to the spectral response of the

PV-cell ( 1.14 eV for silicon) Solubility in the host matrix material

A wide variety of luminophores have been studied in aneffort to meet these requirements. What follows is an overviewof the major efforts in these areas, broken down into the fol-lowing categories: 1) organic dyes, 2) quantum dots, and 3) rare

earth ions.

2.3.1. Organic Dyes

Organic dyes are conjugated organic molecules, where thecore of the molecule is planar with all atoms of the conjugatedchain lying in a common plane and linked by -bonds. The -electrons form a cloud above and below this plane alongthe conjugated chain. Absorption bands of these organic dyes arethe promotion of these -electrons from a ground energy stateto an excited higher energy state. [111] The transition momentof these absorption events is mostly parallel to the conjugated

plane, mostly the molecular axis of the dye, but some transitionsare perpendicular to the molecular axis. These perpendic-ular transition moments are observed at short wavelengthabsorptions. The main absorption wavelength ( abs) can becalculated by

abs =8 mC 0

h

L 2

N + 1 (7)

where m is the electron mass, C 0 is the velocity of light in avacuum, L is the chain length of the conjugated plane, h isPlancks constant, and N is the number of electrons. Theabsorption band of organic dyes is mainly determined by thechain length and the number of electrons in the conjugatedplane of the molecule. [111]

Since the rst papers on LSCs, organic dyes have been inves-tigated as luminophores of choice due to their solubility, highuorescence yields, and large absorption coefciencts. The dyesinvestigated as possible components of LSCs belong to the fol-lowing classes of molecule: bipyridines, [25] coumarins, [40,48,49,112116] dicarbocyanine iodides, [48] dicyano methylenes, [20,48,49,113,117] lac-tones,[114] naphtalimides, [114,116,118] oxazines,[48] perylenes andperylenebisimides, [40,46,93,96,114116,118123] perylenebisimidazoles, [93] phtalocyanines,[124] phycobilisomes, [125] porphyrins, [20,124] pyr-romethenes, [114] rhodamines, [16,18,48,49,111114]sulforhodamines, [48,121] tertiary amine derivates of tetra-cyano-p-quinodimethane, [126] thi-oxanthenes, [114] (iso)violanthrones, [119] and some unspecied dyesincluding BASF K1,[111,121,127] BASF K27,[127] and BASF Lpero.[127] Characteristic examples of these dyes are depicted in Figure 10.

As can be seen, there have been a wide variety of dyesexplored for use in LSCs over the past few decades, and wewill go into more detail on many of these molecules in thefollowing paragraphs.

The most commonly used dye types for LSCs have been therhodamines, coumarins, and perylene(bisimides) derivatives.Rhodamines, like Rhodamine 6G, and coumarins, like Cou-marin 6 (also known as Coumarin 540), belong to the groupof dyes mentioned in the earliest stages of LSC research, [16,112] while the perylenes and perylenebisimides are mostly men-tioned in the more recent papers on LSCs, and were rstdescribed for use in LSCs in the late eighties. [96,119] Rhodam-ines are known for their high quantum yields and high molar-extinction coefcients, but also for their small Stokes shifts.These small Stokes shifts lead to much re-absorption and con-sequently to increased LSC losses, both surface and quantum(nonradiative) losses. For Rhodamine 6G, the luminescencewas reduced when incorporated in poly(methyl methacrylate)(PMMA) in comparison to the solution luminescence, but nosuggestion has been made to explain this behavior as this iscontrary to common performance of organic uorophores in apolymer matrix. [113] In other types of dyes, like Coumarin 540and dicyanomethylene (DCM), the luminescence increaseswhen they are incorporated in a solid PMMA matrix, probablycaused by an increased molecular rigidity, limiting nonradiativerelaxation of the molecule in the excited state [113] or isolationfrom radicals or other impurities. [128] Rhodamines 590, 575,6G and B show very limited photostability [111,114] in compar-ison to other types of dye molecules like perylenes and somecoumarins.


10/24




Figure 10. Examples of organic dyes used in LSCs: a) a bipyridine derivate, b) coumarin 6, c) a dicarbocyanine derivate (DODCI), d) a lactone, e) DCM(4-dicyanomethylene-2-methyl-6-p-dimethylaminostyryl-4H-pyran), f) oxazine 720, g) a naphtalimide derivate, h) phtalocyanine, i) hematoporphyrin,j) a tertiary amine derivate of tetra-cyano-p-quinodimethane (DEMI), k) pyrromethene 580, l) thioxanthene, m) sulfoRhodamine B, n) Rhodamine 6G,o) 3,9-diisopropionicacid-4,10-dicyanoperylene, p) perylene-3,4,9,10 tetracarboxylic acid-bis-(2-6diisopropylanilide) (Perylenebisimide), q) perylene-1,7,8,12-tetrachloro-3,4,9,10 tetracarboxylic acid-bis-(2-6diisopropylanilide) (Perylenebisimide), r) perylene-1,7,8,12-tetraphenoxy-3,4,9,10 tetracar-boxylic acid-bis-(2-6diisopropylanilide) (Perylenebisimide), s) a derivate of an (iso)violanthrone, and t) a derivate of perylenebisimidazole. [119]


11/24



REVIEW


Coumarins can have a larger Stokes shift compared toRhodamines. [48] The overlap factor, f a, of the absorption andemission spectra of Coumarin 540A is 0.12, while for Rhod-amine 6G this overlap factor of the absorption and emissionspectra is 0.48.[113] There have been coumarins tested withmoderate to very good quantum yields: Coumarin Red G has

a quantum yield of over 80%[116]

(87% was also reported[40]

),while the quantum yield of coumarin 540A and CRS040 havebeen reported as nearly unity (98%). [40,114] Even though thephotostability of coumarins have proven to be better thanof rhodamines, they still have been reported to have reducedstability in comparison to perylene-based dyes. [40]

Perylene dyes and their derivates, like perylene bisimides,perylenebisimidazoles, and (iso)violanthrones, are known fortheir intense uorescence and good photostability. [119] Peryleneitself has low photostability, due to easy electrophylic substitu-tion. To improve the properties of this dye, new side groupswere added: 3,4,9,10-tetracyanoperylene has been synthesized,the cyano groups are electron acceptors and when added at thesepositions around the perylene core they improve the photosta-bility. The solubility of this tetracyanoperylene is moderate, there-fore 3,9-diispropionicacid-4,10-dicyanoperylene (Figure 10o)with a quantum efciency of 91% and good photostability wassynthesized. [119]

Perylene bisimides are known for their intense uores-cence and good photostability, but generally exhibit low solu-bility. To improve the solubility of these perylene bisimides,ortho-alkylated aromatic bulky groups are added to thebisimides (Figure 10p). These groups are twisted 90 withrespect to the perylene core due to steric hindrance, whichincreases the solubility of the dyes. Addition of large sidegroups on the perylene core (1,7,8,12 positions; Figure 10qshows a chlorinated perylene core) cause the two sides of

the perylene core to twist by 42 with respect to each other.This twist bathochromically shifts the absorption and emis-sion bands of the perylene dye. Furthermore, the tertra-phenoxy-substituted perylene bisimides (Figure 10r) havegood solubility in organic matrices due to the phenoxy sidegroups, especially in comparison to the chlorinated perylenebisimides (Figure 10q). Component 10r has an absorptionband peaking at 578 nm and an emission band peaking at613 nm with a quantum yield of 96% with both good solu-bility[129] and photostability.

The violanthrones [119] and perylene-bisimidazoles(Figure 10t)[93] have more red-shifted absorption and emissionbands compared to component 10r, with reasonable quantumyields of, respectively, 55% and 80%. If the conjugated core ofthe perylene is shortened to a naphthalene core, the absorptionand emission bands are blue-shifted. An example of this typeof dye is the BASF Lumogen F Violet 084, which is a naphtal-imide. [118] This work demonstrates that by changing the sidegroups or changing the length of the core of an organic dye,a perylene in this case, one can alter all important propertiesof a dye, allowing for engineering of more suitable and robustmolecules.

As described earlier, perylene based dyes are known for theirhigh quantum efciency [118] and for their large spread in spec-tral absorption and emission. Combining multiple perylenedyes in one single device could thus lead to both broad spectral

absorption (the whole visible spectrum) and reasonably highquantum yields. [118]

Two different types of dicyano methylenes have beenused in LSCs, DCM (4-(dicyanomethylene)-2-methyl-6-(p-dimethylaminostyryl)-4H-pyran) [48,49,113,130] and DCJTB(4-dicyanomethylene)2-t-butyl-6-(1,1,7,7-tetramethyljulolidyl-9-

enyl)-4H-pyran).[20]

DCM has a broad absorption spectrum anda large Stokes shift with a reasonable quantum yield ( 80%in PMMA),[48,113] but the photostability is limited. DCJTB hasbeen used in combination with a platinumporphyrin derivate,whereby DCJTB transfers its absorbed energy to the platinumcomplex. This combination is calculated to form an LSC with a6.8% power conversion if paired with GaAs PV cells. [20]

Other dyes have also been investigated in LSCs, as men-tioned before, but the published research on these dyes ismore limited. Bipyridines [25] have a large Stokes shift, dueto excited-state intramolecular proton transfer and twistedintramolecular charge transfer, but the quantum yield is low(up to 3.6% in butanol). Tertiary amine derivates of tetracyano- p-quinodimethane like DEMI (4-[1-cyano-3-(diethylamino)-2-propenylidene]-2,5-cyclohexadiene-1-ylidenepropane dinitril)are red-light absorbing dyes, but their photostability is low.Nature-based luminophores, like phtalocyanines, [124] hemat-oporphyrin, [124] and phycobilisomes [125] have also been pro-posed for LSCs. The phtalocyanines and hematoporphyrinhave intense absorptions and good thermal stabilities, but thequantum yields are 20% and 8% respectively. The quantumyield of the phycobilisomes is below 50%.

Sulforhodamines, like sulforhodamine-B and sulforhod-amine-101 have, similar to the rhodamines, small Stokesshifts, [48] resulting in many re-absorption events. Oxazine 720,oxazine 750, DODCI (a dicarbocyanine iodide: 3,3-diethy-loxadicarbocyanine iodide) and DOTCI (a dicarbocyanine

iodide: 3,3-diethyloxatricarbocyanine iodide) suffer fromsimilar small Stokes shifts. [48] Thioxanthene dye D 315orange, Lactone dye D 838 yellow, and Pyrromethene dye 580have been shown to be relatively unstable towards illumina-tion with visible light. [40,114] IR-144 (a dicarbocyanine) is onlymentioned in the paper of Batchelder; [48] it appears to havea reasonable Stokes shift and absorption coefcient, but nofurther information on the performance of this dye in LSCsis available.

One of the disadvantages of organic dyes in general is thelimited breadth of their spectral absorptions. To increase thisbreadth a combination of several dyes can be used. These dyescan be used in a stack of LSCs with one dye in or on eachwaveguide[20,49,93,113] or all dyes in one LSC.[40,48,49,103,112,118,120] In the rst case each plate will act as an LSC itself. Lightemitted in the escape cone by one LSC can penetrate anotherLSC where it can be re-absorbed (if the light is in the absorp-tion band of the dye of the second LSC) and partly re-emittedin the waveguide mode. This is described in more detail in thepart of this review focusing on the re-absorption problem of theLSC. In the case of all dyes in one LSC, all dyes will absorbenergy and all this absorbed energy is transferred to the dyewith the smallest bandgap, via nonradiative or radiative transfer.For FRET the dye molecules have to be very close to each other,typically less than 10 nm, so this only happens in LSCs withvery high dye concentrations.


12/24




Numerous studies have been performed on the photosta-bility of these organic dyes:[40,49,96,111,114,116,119121,126128,131] much depends on the processing conditions and polymericenvironment of the uorophore. Photodegradation of dyes inpolymers can occur in two ways: 1) by direct interaction of thedye molecule with the sunlight, leading to decomposition, or

2) by the attack on a dye molecule by an active species formeddue to photodecomposition of a residual molecule in the polymermatrix or by singlet oxygen. [114] These residual moleculesare mostly materials used during polymerization or processingof the matrix material. Two examples of reduced luminescencein LSCs caused by other molecules present in the polymermatrix under solar illumination are photoreduction, where thedye in its excited state takes an electron from an electron donor,forming a nonradiative stable anion, and photo-oxidation wherethe dye in its excited state donates an electron to an acceptor,forming a nonradiative cation. [119] Direct interaction of the dyemolecule with the solar light can lead to molecular changes,caused by high-energy photons. Photochemical decomposi-tion of the dye molecule can lead to a reduction in absorptionand thus also a reduction in uorescence or to a blue-shift inabsorption and emission. [111] An extensive comparison of thephotostability between several types of dyes, including rhodam-ines, coumarins, perylenes, a naphtalimide, a thioxanthene, alactone, and a pyrromethene, showed that the perylenes fromthe Lumogen Series of BASF had the best photostability, bothunder illumination of UV light as well as visible light. [114]

There are several options beyond the engineering of morerobust dyes towards extending the photostability of the LSCdevice. Introducing UV absorbers and hindered amine lightstabilizer (HALS) molecules could reduce the photodegrada-tion of the dyes under UV illumination. Using copolymers ofPMMA has also been shown to increase the photostability of the

dyes.[131]

Decomposition of the dye molecules is much faster ina singlet oxygen environment than in a nitrogen environment,with a small recovery of luminescence occurs after some timein the dark. [119,43] Thus, it may be that during dark periodsduring the night could recover some of the luminescence lostdue to photodegradation in the daytime.

2.3.2. Quantum Dots

Quantum dots (QDs) are nanostructures from semiconductingmaterials with dimensions in the order of 10100 nanometers.The size of the dots is in the order of the de Broglie wavelengthof the electron. As a consequence of their restrictive size, excitedelectrons are conned in the semiconductor, which exhibitsoptical and electrical properties similar to those of atoms.

QDs promise several advantages over organic dyes if used inluminescent solar concentrators. The absorption threshold ofthe QDs may be tuned by judicious choice of the particle dia-meter (see Figure 11). Colloidal InP QDs, for example, have beenshown to be capable of absorbing the whole visible spectrum. [133] QDs may also have large Stokes shifts, which are determinedby the spread in dot size. [134,135] Their crystalline semiconductorcomposition should make them more stable than organic-baseddyes.[88] Quantum dots as luminophores in luminescent solarconcentrators were rst suggested and modeled by the group ofBarnham, [35,87,88] which predicted a concentrator efciency up

to 20% when type IIIV solar cells were used in combination tohigh quantum efciency (QE = 1) QDs.

As mentioned above, QDs promise good photostability, butoutside a solid matrix they are quite sensitive to oxygen andlight,[38,136138] which becomes a challenge for module engi-neering. Several research groups claimed that the incorpora-tion of QDs into a solid matrix can lead to a blue-shift in boththeir absorption and emission, caused by surface oxidation ofthe QD during the manufacturing process. [89,136,139] In additionto a blue-shift in emission wavelength, the emission intensityin QD solar concentrators also is dependent on the nature ofthe host matrix. For example, CdSe/ZnS QDs in a solid matrixlose 22.596% of the emission intensity they exhibited in solu-tion. [89] Several factors can cause this loss, including increasedscattering due to particle clustering owing to decreased solu-bility, and matrix absorption of the emission light.

The spectral emission of QDs may be tuned by adjusting theQD size. With increasing QD size, the bandgap will decrease.The cluster size distribution of the dots can be tuned in situduring the production of the QD concentrator. By increasingthe annealing temperature of CdS quantum dots in silicondioxide lms, the photoluminescence spectra red shifted, attrib-uted to an increasing CdS dot cluster size. [140] These resultswere supported by calculations and direct X-ray diffraction

Figure 11. Top: Fluorescence of CdSe with different emissions ( ex = 365 nm). Bottom: Emission spectra of the CdSe quantum dots at differentgrowth temperatures and at the same reaction time of 10 h (A, 45 C; B,60 C; C, 80C; D, 100 C; E, 120C; F, 150 C). Reproduced with permis-sion. [132] Copyright 2008, American Chemical Society.


13/24



REVIEW


(XRD) measurements of the QDs, determining that the clustersize grew at increased annealing temperature. [139] Schuler alsoshowed that CdS-rich siliconoxide layers show an absorptionband close to the bandgap of bulk CdS, while low concentrationlayers had bandgaps of higher energy, corresponding to nite-well and tight-bonding calculations. [141,142] Naturally, by altering

QD materials one may also dene alternative bandgaps. PdS usedas a QD material, for example, has an absorption band up towavelengths in the IR part of the spectrum. [91] In addition to alarger spectral absorption, these QDs also have a larger Stokesshift, leading to less overlap between the absorption and emis-sion band. Where the CdSe/Zns ( abs 600 nm) QDs used byShcherbatyuk [91] have a Stokes shift of 23 nm and an absorp-tion of 0.68 (normalized to the peak absorption) at the peakemission, the PbS ( abs 750 nm) QDs had a Stokes shift of122 nm and a normalized absorption of 0.24 at the emissionpeak. A disadvantage of these PbS QDs is that the molar extinc-tion coefcient is one order of magnitude lower than that ofthe used CdSe/ZnS QD (2 104 L mol 1 cm 1 for PbS and3 105 L mol 1 cm 1 for CdSe/ZnS). Red-absorbing cadmium-based QDs also show a decrease in absorption coefcient, [143] but because the spectral absorption is higher for these materialsthe total absorption efciency is still higher than of cadmium-based QDs absorbing only up to the blue or yellow part of thespectrum. Kennedy et al. showed that the optical absorptionefciency for NIR, orange, and green-emitting Cd-based QDswas, respectively, 23.1%, 21.7%, and 11.6%.

The increase in the Stokes shift for the lower bandgap QDsleads to less re-absorption of waveguiding photons, which inturn may also reduce the surface losses of the QD solar con-centrator (QDSC). NIR-emitting QD solar concentrators weremodeled to have 43 1% surface loss in comparison to thegreen-emitting QD solar concentrators, which had 58 5%

surface loss. For 60 mm 60 mm 3 mm QD concentrators,the reduced re-absorption translated into a modeled opticalefciency of 13.2% for the NIR-emitting device, in comparisonto 5.0% for one that was green-emitting, assuming similarquantum yields of unity in each device. [143] PbS QDs can alsoshow similar results. The absorption efciency of the PbS QDsis 40% in comparison to 22% for the CdSe/ZnS QD used, butthe luminescent quantum yield for the PbS QDs is lower (30%)than that of the CdSe/ZnS QDs (50%). As a result, the resultantoptical efciency for the PbS QD solar concentrator was set at12.6%.[91] The emission spectra of the QDs used in these exper-iments demonstrated that increased pathlength of the emittedlight through the waveguide resulted in a shape change, withthe short wavelength peak decreasing with respect to the longerwavelength peak, demonstrating that re-absorption still playsan important role in these QD concentrators. [91,140]

QD-based concentrators have been compared directly withorganic dye containing LSC via both modeling and experi-mental studies. One such work compared CdSe/ZnS QD andLumogen Red F 300 containing concentrators. [89] The best QDconcentrator described in this paper reached to 58% of the per-formance of the device containing Red 300 containing, mainlya result of the lower quantum yields (QY 0.10.6) of the dotscompared to that of the organic dye (uorescence quantumyield FQY 1). Another work has compared the optical ef-ciency of CdSe/ZnS QD containing concentrators with several

other organic dye containing concentrators, including Rhod-amine B, a red uorine (Red F), and several other laser dyes(LDS698 and LDS821).[90] Simulations and experimentsdescribed in this work showed the optical efciency of the QDconcentrator was within 10% of the optical efciency of the RedF- and Rhodamine B-containing LSCs. The optical efciencies

of the LSCs containing the LDS series dyes were comparable tothat of the QD concentrator. These relatively low optical deviceefciencies reported were again attributed to the relativelylow quantum yield of the QDs. Increasing the quantum yieldof the dots closer to that of Rhodamine B (stated to be 95%)would lead to increased optical efciencies of the device: as itwas, the QD-based LSC performed at 60% of the optical ef-ciency of the Rhodamine B-containing LSC. Calculations doneon PbS containing QD concentrators predict such devices couldachieve power-conversion efciencies (PCEs) more than doublethe PCE of a Rhodamine-containing LSC, at 3.2% and 1.3%,respectively.[91]

It should be noted there is some concern about the use ofthe QD LSC due to the potentially toxic nature of many of thematerials used in the QD. [144146] There appear to remain con-siderable uncertainties as to the true toxicity of these materials,and the level of toxicity often is related to not only chemicalcomposition, but also processing conditions, environmentalfactors, and a number of other details. Nevertheless, there iscontinued research aimed at reduction of the potential harmfuleffects of QDs, such as the use of jelly dots [147] or on QDsbased on more benign materials, like silicon. [148,149]

2.3.3. Rare Earth Ions

Rare earth ions (sometimes complexed with ligands) are inves-tigated as luminophores for usage in LSCs primarily because of

their promise of high photostability and their large Stokes shift,although the presence of organic ligands may compromise theeffective lifetime of the molecules. Levitt and Weber [18] alreadydescribed in 1977 the use of neodymium (Nd 3+)-doped glasses asmaterials for LSCs. Neodymium mainly absorbs around 580 nm,but also absorbs photons at longer wavelengths. Emittedphotons have wavelengths around 880 nm ( 4F3/2 4I9/2) and1060 nm (4F3/2 4I11/2). The photons emitted at 880 nm can bereabsorbed ( 4I9/2 4F3/2) and the energy of the photons emittedat 1060 nm is slightly lower than the bandgap of silicon. Thesecharacteristics of neodymium lead to low efciencies of Nd 3+-doped glasses as LSCs. [150]

To increase the efciency of neodymium-doped LSCs,codoping with ytterbium (Yb 3+ ) has been discussed. [150] Energyabsorbed by the neodymium ions is transferred to the ytter-bium ions, this accomplishes two goals: it will circumvent thelimitations due to self-absorption, and the neodymium emis-sion at 1060 nm will decrease due to depopulation of the 4F3/2 state of neodymium by the 2F5/2 state of ytterbium. Emissionfrom ytterbium ions is around 970 nm ( 2F5/2 2F7/2), whichis slightly higher in energy than the bandgap of silicon, but theresponse of silicon to photons with this wavelength is still high.The energy transfer from neodymium to ytterbium depends onthe type of glass: in borate tellurite and germanite glasses theenergy transfer efciency can reach up to 90%. [150,151] Ytterbiumitself can also be used as a luminophore in LSCs, but the solar


14/24




absorption of these ions only occurs in the NIR ( 2F7/2 2F5/2 in 4f 13 conguration) or in the UV (4f 12 conguration), so largesunlight absorption is only possible if ytterbium-doped glassesare codoped.[152] In the case of neodymium, the neodymiumions are codopants for the ytterbium. A disadvantage of neo-dymium and ytterbium is their low absorption efciency. [153]

Uranyl ions (UO 22+

) have been reported[153]

in LSCs becauseof their high absorption efciency. Uranyl ions have ve ordersof magnitude higher absorption efciency than neodymiumions, but they absorb only in the blue part of the spectrum(maximum absorption around 430 nm). The uorescence ofuranyl ions is maximum at 500530 nm, [153] and the quantumyield is 0.67.[154] Uranyl ions have also been used as codo-pants in combination with neodymium, [155] and ytterbium andneodymium. [156]

Chromium(III) ions have a large spectral absorption withpeaks at 450 nm ( 4A2 4T1) and 650 nm ( 4A2 4T2) and socould be useful for LSCs, but a major drawback is the limitedquantum efciency (up to 25%). [156] Higher quantum yieldshave been found in quartz-like (FQY = 50%), pentalite-like(FQY = 75%), and gahnite (FQY = 100%) glass ceramics. [157,158] Chromium(III) ions can be used for codoping neodymium- andytterbium-doped glasses, as they increase the absorption rangeof the glass and then transfer their energy to the high-quantum-yield neodymium and/or ytterbium ions. The energy-transferefciencies from chromium(III) to neodymium and ytterbiumhave been determined at 92% and 88% respectively in lithiumlanthanum phosphate (LPP) glasses. [156] Other rare earth ionscan also be used in LSCs. [154,159]

To increase the absorption of rare earth ions, organic ligandscoordinating to the ions have been proposed. [85,156,160] In such acomplex, the ligands absorb energy and the electrons go fromtheir S 0 to their S 1 state. From the single S 1 state, energy is

transferred to the triplet state of the ligand (T 1) via intersystemcrossing (ISC). From this triplet state, energy is transferred tothe rare earth ion. Due to all the energy transfers, these com-plexes have a very high Stokes shift ( > 200 nm). Moudam etal.[160] synthesized a Eu 3+ complex (Eu(hexauoroacetylacetonate)3(bis(2-(diphenylphosphino)phenyl)ether oxide). This complexabsorbs light in the UV region ( < 350 nm) and emits it at 613 nmwith an FQY of 86% in PMMA. In 2011, Katsagounos [161] showed four europium-based complexes with increased lumi-nescence compared to the single europium ion and using thesedyes in LSC-type downconverters increased the efciency ofmulticrystalline-silicon solar cells up by 17%, for the complexwith a pyridine derivate as a ligand.

The spectral absorption of Eu-complexes is not very broadbecause low-energy photons create a triplet state lacking suf-cient energy for transfer to the europium ion, but may have alarge Stokes shift (see Figure 12).[85] Other rare earth ions likeneodymium and ytterbium can be excited by lower-energy tripletstates, creating NIR-emitting complexes with broad absorptionin the visible region of the solar spectrum. A problem occurringwith these ions is that they can be deactivated by surroundingvibrational states of OH, NH, and CH bonds. Thereforethe ligand has to been uorinated or deuterated, because CFand CD absorptions occur at lower energies. The emission ofthis uorinated ligand and ytterbium complex peaks at 970 nmwhen excited at 320 nm. In peruoromethylcyclohexane, the

luminescent QY of the ytterbium complex with the uorinatedligand was determined to be 2.6%; not high enough for LSCapplications.

2.3.4. Enhanced Absorption by Scattering Back Layers

To aid in luminophore absorption, it is standard practice toapply a rear layer to an LSC to act as a reector. The reectingback layer effectively doubles the path length of incident lightthrough the dye layer for enhanced absorption. Some of theinitial experiments used a silver mirror, [17] but such a mirrorabsorbs in the visible range and can result in signicantabsorptive losses for longer transmission distances. To avoidabsorptive losses, most recent work has employed a whitescatterer. [44,45,103,130,162165] The effectiveness of the scattererdepends on both the size of the waveguide (the scatterers havebeen shown to be benecial for waveguides tens of centimeterslong) and the concentration of dye within the waveguide. [165] The scatterer also may direct that fraction of incident light thatcannot be absorbed by the dye directly at the PV cell, allowing itto generate electricity. The separation of the scatterer from thewaveguide by a low refractive index layer is quite important, asone desire to maintain waveguided light in the trapping modesof the waveguide: every encounter with the attached scatteringlayer redistributes the light, and a signicant fraction of thisredirected light will be outside the waveguide modes of thesystem.

It was demonstrated that the effectiveness of the scatterer isgreatest near the edges of the device, where direct scattering ofnonabsorbed light into the solar cell directly is prevalent. How-ever, over long distances the direct scattering component is lessimportant, and the primary advantage given by the scatteringlayer is the return of unabsorbed light back into the waveguide

for possible absorption on the second pass, but even this effectcan be minimized in highly dye-doped waveguides. [165] Thissuggests a possible transparent LSC design with only the closeedges using a scattering layer.

Figure 12. Absorption and emission spectra of Eu 3+ complex (EuhD) withthe structure Eu(hfac) 3(DPEPO) where hfac = hexauoroacetylacetonateand DPEPO = bis (2-(diphenylphosphino)phenyl)ether oxide. [85]


15/24



REVIEW


The functionality of the rear layer may be enhanced by addinga luminophore to the scattering layer. This can potentiallyimprove performance by converting light otherwise lost dueto lack of absorption into light of longer wavelengths that canbe absorbed by the luminophores within the waveguide [163,166] while still maintaining the benets of scattering described

earlier.The use of a scatterer rather than a reector becomes par-ticularly important in the case one uses a top wavelength-selective mirror, as described in Section 2.1.2 above. Aslight that is reected by the selective reector is not in thewaveguiding mode, this means it will encounter the backlayer of the waveguide every bounce. If this rear reector isa silver mirror, the absorption losses from the silver woulddramatically reduce the amount of energy that can be emittedfrom the waveguide edge, and so use of a metallic back mirrorshould be avoided in this case. Because of the relative con-tribution of the scattered light is high when compared todye-emitted light for smaller samples, [40,165] the relative con-tribution of scattered light somewhat obscures the contribu-tion from the selective reector. Thus, the true enhancementof performance expected from the selective reectors willbecome more obvious for larger (that is, greater than 25 cm 2)waveguides.[40]

2.3.5. Enhanced Fluorescence by Metal Nanoparticles (Plasmonics)

A newer research area for LSCs is in the eld of using plas-monics to enhance the system response. [98] It has beenshown in many research publications that when a uorescentmolecule is brought within close proximity of a small metallicnanoparticle, such as silver, there may be a considerableenhancement of the uorescence. [167171] There has been pre-

vious application of surface plasmonics in PVs to enhancedevice performance. [172178]Very recently, experiments have coupled such plasmonic

photovoltaic cells to LSCs, with predicted large enhance-ments of the efciency. [179] The introduction to plasmonicstructures within the LSC waveguide could open up the useof a considerable range of dyes that had heretofore beenrejected on the basis of low quantum efciencies or photo-stability, as both of these important dye characteristics couldbe improved by the application of some type of plasmonic-based system.

2.3.6. Enhanced Absorption and Emission by Connement Effects

Other designs may be contemplated that will reduce the limita-tions of the dye materials through connement effects. Recently,a slot waveguide using a nanometer-sized low-refractive-indexslot sandwiched by two high-refractive-index regions was cal-culated to enhance emission by the luminophore due to thePurcell effect, to increase the effective absorption length ofluminescent centers and improve their uorescence quantumyield.[180] The physical requirements of this device, however, arequite extreme, with a 10 nm slot demanding use of materialswith refractive indices > 2. While immediate use of such anarchitecture is not likely, the design does provide an interestingframework to build upon.

2.3.7. So, Which Luminophore to Choose?

At this point in our examination of the history of luminophoreresearch in LSCs, we have emphasized mainly the performancecharacteristics of the dyes in terms of their ability to generatethe maximum light output of an LSC. However, the exibility ofthe device allows for a broader outlook than simply maximumperformance. When one considers the importance of aestheticsof the LSC, one realizes there should not be a search for thesingle, all-purpose luminophore, but there should be a rene-ment of an array of high-performance molecules displayinga broad range of characteristics. For instance, for artistic rea-sons it is of interest to have materials of good performance ina variety of colors/shades. As shown in Figure 13, being able towork with a variety of colors adds exponentially to the applica-bility of the LSC.

To this point, there is no clear winner among the variousluminophores, nor even has a class of luminophores beenidentied as the obvious choice. Research continues apace onorganics, quantum dots and phosphors. There may be instances

as well where choices will be made based on the specic cri-teria of the application: does one prefer high uorescence yield,pleasant aesthetics, or enhanced photostability? Obviously,there is still plenty of room for innovation in this area.

2.4. Photovoltaic Losses

The standard silicon-based PV has a bandgap correspondingto a photon of around 1100 nm ( 1.1 eV). Photons with ener-gies above this threshold may still be processed by the solarcell, of course, but the excess energy of the photon is wasted,and converted most often into heat, and there is a reductionin the response of the cell for these shorter wavelengths. Ener-gies lower than this bandgap are not sufcient to generate anycurrent.

However, an LSC does not emit a spectrum remotely sim-ilar to the solar spectrum. Rather, it emits a narrow range of

Figure 13. Video still depicting the painting of uorescent dyes on a clearPMMA plate forming an LSC. The full video may be found as SupportingInformation (courtesy Eindhoven University of Technology and Peer + ).


16/24




wavelengths, most often centered at red and NIR wavelengths(630720 nm at the maximum). This gives the advantage thatthe LSC should remain relatively cool under standard operatingconditions, as they do not absorb much of the infrared light.In consequence, the light incident on the edge mounted photo-voltaic will contain minimal infrared components, preserving

the high performance[22]

as generally photovoltaics perform lesswell when heated. [181]In addition to reduced heat loads on the cell, by matching the

spectral response of the solar cell to the output spectra of thewaveguides it may be possible to obtain enhanced performance,although the overall effect may be limited due to the generallybroad response of many cells to light over the common wave-lengths emitted by luminophores. [130] However, this tuningcould nd application in amorphous silicon [182,183] and varioustype IIIV cells. Using larger-bandgap solar cells would deliversimilar currents, but at larger V oc.[40]

Downconversion, which is the process of converting a highenergy photon into a lower energy photon with a wavelengthmore readily utilized by the photovoltaic, has been studiedusing luminophores attached directly to the surface of the PVcells[184187] or within the encapsulation layers of the PVs. [4] Thegoal of these layers is generally to shift the incoming blue por-tion of the spectrum to longer wavelengths better suited to thePV. In other cases, colors are applied to the photovoltaic sur-face for more aesthetic reasons, to add alternative colors to thePV panels. [188] A similar application of this type, but this timeusing small, sliver silicon-based photovoltaics located withinthe waveguide itself, promise improved performance. [117]

To better exploit the emission spectrum in the LSC,researchers have used type IIIV PV cells based on GaAs andInGaP [44] and obtained record-setting efciencies. If these cellscould be produced economically, [189] it could hold promise for

widespread adoption of the LSC in future. Another option couldbe the use of organic-based PV cells, which often have a sweetspot in the spectral range where the LSC emits. [190]

Another option to enhance the response of photovoltaiccells is to use individual dye-lled LSC waveguides, where eachwaveguide is separately attached to a solar cell specic for thatemission wavelength. In the so-called luminescent spectralsplitter (LSS) concept[191] illustrated in Figure 14, the incominglight is collected over a large area and funneled into a smaller

region where it is absorbed by a stack of what are essentiallyLSCs, each attached to a separate photovoltaic cell tuned tothe specic wavelengths emitted by the LSC. In this concept,the LSCs themselves do not need to transport the light overextended distances. Yet another option is to employ an LSCwith edge-mounted photovoltaic directly on top of a secondphotovoltaic: in this conguration, the light that passes throughthe LSC due to its limited absorption range will be collected bythe underlying cell. [192]

2.5. Waveguide Losses

Around 4% of incoming sunlight is reected from each of thewaveguide surface (the refractive indices of poly(methyl meth-acrylate) (PMMA) and poly(carbonate) (PC) being betweenabout 1.49 and 1.58) and never enter the waveguide, and couldthus be considered a loss (Figure 4, loss 5a). While antire-ection coatings are very common in PV cells, they have notyet been applied to LSCs. As the LSC relies on total internalreection from two smooth surfaces, textured systems as usedin many antireective coatings may not be a viable option. [193] Rather, coatings utilizing different refractive indices can reducethese reective losses and can be applied to polymeric mate-rials.[194] It remains to be seen how these coatings affect LSCperformance.

Naturally, it would be a great advantage to be able to pro-

duce the LSC panels on the size of square meters in manyapplications, as this could allow for a more seamless look,and reduces the amount of wiring that needs to be done thus

Figure 14. Luminescent spectral splitter. Reproduced with permission. [191] Copyright 2011, Elsevier.


17/24



REVIEW


lowering installation costs, among other things. However,there are all the losses described in this section that must beaccounted for which places severe limitations on the maximalsize for the device. The drop-off of edge output with distancehas been reported. [35,164,165] These transport losses are gener-ally the result of two main loss mechanisms: re-absorption of

emission light by subsequent dye molecules (described earlier,Figure 4 loss 2) and parasitic absorption of the emission light bythe host waveguide material (Figure 4, loss 5b). Other featuresare certainly at play as well. For example, imperfections in thesurface of the LSC waveguide will also lead to losses (Figure 4,loss 5d), and the presence of dust or minor imperfections canresult in the scattering of light outside of the waveguidingmode and lost to the environment. [195] Generally, the devicesshould be kept as scratch free as possible, suggesting the useof hardcoats or the like for particularly soft waveguide mate-rials. There is also potential for a signicant degree of scatterfrom unwanted scattering centers located within the waveguide(Figure 4, loss 5c).[195] These scattering and absorption eventsdominate for large waveguides: re-absorption by the dyes isprimarily a short-range effect, [196] although the absorption tailof the luminophore may have quite a large impact over longdistances. [83,164]

One desires to extend the dye absorption as far to the infraredas possible. However, this is complicated in that the polymericwaveguides, which are predominantly made of PMMA orpolycarbonate, become parasitic, and absorb strongly at wave-lengths beyond 880 nm. [164,197199] Often the additives designedto improve various characteristics of the host matrix in thesolid state (such as altering coloration, stability, or hardness)or the presence of unreacted monomers in the host materialcan have a large impact on the devices performance. [43,102,116] For example, an additive that shows only a small absorption

when measured through the width of the waveguide can havea severe impact on the edge output of the same object, for thepathlength is magnied many tenfold. [102,130] In addition, as thewaveguides age due to exposure to the elements and in partic-ular ultraviolet light, generates brittleness, opacity, and a host ofreaction species that generate a yellow tint to the plate and actas absorptive light traps.[200] To combat these loss mechanisms,research into copolymer systems has demonstrated enhancedphotostability over single component systems. [131] During thetransition from the excited state to the ground state of the dyemolecule, electron transfer takes place from the dye to the maincopolymer (here polystyrene-co-methylmethacrylate) chain,which may increase the photostability of the dye.

In general, the waveguide materials generally used for LSCwork tend to be chosen with economy, rather than absolute bestperformance, in mind. However, much optically clearer mate-rials are available in the marketplace, and could be consideredfor use in the LSC. For example, PMMA and peruorinated-polymer-based optical bers may demonstrate considerablyenhanced transmission characteristics. [201]

An obvious way of decreasing the waveguide losses isapplication of waveguides using higher refractive index mate-rials. The most prevalent materials used for the LSC havebeen PMMA and glass with refractive indexes on the orderof 1.49. Higher refractive index materials used or consideredhave included polycarbonate ( n 1.59),[102] and special glasses

(n = 1.51.8).[20,38,102] Glass samples allow for potentially lessabsorption in the emission regions of the dye than polymericwaveguides,[38] but many of the professed advantages of theLSC (for example, reduced weight and ease of handling) wouldbe nullied.

Another waveguide material that has been considered for its

exibility has been polysiloxane.[94,202]

While the durability ofsuch materials might be a question, a range of novel applicationscould be considered with the addition of exibility to the LSCrange of characteristics. The emission of such polysiloxane-basedsystems has been comparable to that of lled polymerwaveguides of a certain dye concentration (see Figure 15).This suggests that if the dyes were more soluble, such rubberywaveguides could be a viable option.

An additional aspect of waveguide design that needs to beconsidered is that edge emission from rectangular waveguideedges is nonuniform over the exit surface. Intensity of theemission can vary 20% between the center and corner of thewaveguide edge.[73,203] This variation may also cause additionallosses, as this means illumination of the attached photovoltaicis not uniform, and nonuniform illumination of a photovoltaiccell will result in decreased performance of the cell. [204206] Inorder to improve the light concentration at the edges and reducethe size of the attached solar cells, tapering of the waveguideedges has been suggested. [207]

As one of the main goals of the LSC is to reduce the sur-face area of photovoltaic cells, it would be preferable to reducethe number of edges with solar cells attached from four tofewer. Congurations using two-edge coverage (both oppositeand orthogonal cell placement) and single-edge arrangementsare common. When reducing the edge coverage of solar cells,reective mirrors are often added to the PV-less edges, and afraction of this reected light reaches the photovoltaic cell. [38,40]

For both performance and aesthetic reasons, alternativeshapes of the waveguide have been studied. The shapes ofthese waveguides also inuence the edge-emitted-light distri-bution. [22,40,81,203,208] While the circular LSC could provide the

Figure 15. Edge outputs of LSCs made from exible polysiloxanewaveguides compared to lled polycarbonate LSCs (purple crosses) withthe dye distributed throughout the bulk of the polysiloxane waveguidewith (green triangles) and without (red squares) added gold atoms, andin with the dye cast as thin layers atop a clear polysiloxane waveguide(blue diamonds). [202]


18/24


19/24



REVIEW


4. Current Efciency Achievementsand ConclusionOne challenge the LSC has to confront in order to positionitself in the global solar-energy generation space is the mis-

luminescent solar concentrator

Documents

Transcript of luminescent solar concentrator