- Email: [email protected]
Optimisation of a three-colour luminescent solar concentrator daylighting system
ARTICLE IN PRESS
Solar Energy Materials & Solar Cells 84 (2004) 411–426
Optimisation of a three-colour luminescent solar concentrator daylighting system Alan A. Earp*, Geoff B. Smith, Jim Franklin, Paul Swift Department of Applied Physics, University of Technology, P.O. Box, 123 Broadway, 2007 NSW, Sydney, Australia Received 4 November 2003; accepted 16 February 2004 Available online 25 May 2004
Abstract Electrical power consumption in buildings may be reduced considerably by more efficient use of sunlight for indoor lighting. A stack of luminescent solar concentrator sheets has been developed, which utilises three different coloured fluorescent dyes to produce a concentrated near-white light source to be coupled into flexible polymer sheets and transported up to 10 m. Thus in clear sky conditions a remote room with neither walls nor roof in direct contact with the sun, can be illuminated by over 1000 lm of natural light with a luminous efficacy of over 300 lm/W, regardless of the sun’s position. r 2004 Elsevier B.V. All rights reserved. Keywords: Daylight; Luminescent Solar Concentrator; Light guide; Lighting
1. Introduction Daylight is an underused resource that has the potential to improve the quality of indoor lighting, as well as substantially reducing energy costs. Artificial lighting is one of the major sources of electrical energy costs in office buildings, both directly through lighting energy consumption and indirectly by production of significant heat gain, which increases cooling loads [1]. Recent studies have shown that effective use of daylighting in conjunction with artificial lighting can produce significant energy savings in office buildings [2–5]. The excellent colour rendering properties of daylight *Corresponding author. Tel.: +61-2-9514-2225; fax: +61-2-9514-2219. E-mail address: [email protected] (A.A. Earp). 0927-0248/$ - see front matter r 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.solmat.2004.02.046
ARTICLE IN PRESS 412
A.A. Earp et al. / Solar Energy Materials & Solar Cells 84 (2004) 411–426
and its close match to the photopic response of the human eye make it an ergonomic light source that is generally preferred for pleasant working conditions [6]. Thus the use of daylight to supplement artificial lighting has proven beneficial for minimising energy usage and improving lighting quality. Daylighting in perimeter zones of buildings is readily achievable through careful window design [3,7]. For regions further away from the window fa@ade reasonable daylighting is possible with light shelves [8,9] or laser cut light deflecting panels [10]. However, these systems are only capable of transporting daylight over relatively small distances. Various mirror light pipe systems have been proposed for daylighting of deeper interior spaces with light pipes [11–13], but these systems can be fairly complex, and to date are still at a prototype stage. Luminescent solar concentrators (LSCs) also hold considerable potential for application to daylighting, due to their capacity to collect and concentrate sunlight [14]. LSCs have been extensively studied since the mid-1970s, but most of this work has been focused on photovoltaic applications [15–18]. The small amount of work done to date on daylighting with LSCs shows considerable potential, providing the motivation for this study.
2. Three-colour LSC stack operation 2.1. LSC stack design Daylight can be transported to remote areas using a stack of three coloured LSCs connected to clear flexible light guides [19], as shown in Fig. 1. Each coloured LSC consists of a clear polymethylmethacrylate (PMMA) matrix with the dimensions 1.200 m 0.135 m 0.0020 m, and is doped with a coloured fluorescent dye. For the purpose of light transport these collectors are coupled with optically clear glue to clear PMMA light guides up to 5 m in length. Thus daylight can be transported to underground rooms or windowless rooms in the centre of a building. By appropriate selection of dyes, white light output can be produced, giving a good match with the colour of daylight. Optical transport of daylight using an LSC stack operates by the following process. Solar energy enters the stack where it may be absorbed and randomly re-emitted by
Fig. 1. Schematic of a three-colour LSC stack connected to light guides and luminaire.
ARTICLE IN PRESS A.A. Earp et al. / Solar Energy Materials & Solar Cells 84 (2004) 411–426
413
the violet dye in the top sheet. If the emitted photon travels below the critical angle with respect to the top or side surface of the sheet (42o for PMMA), it will leave the collector. Otherwise it will be totally internally reflected to the end of the collector, then through the light guide to the luminaire. Photons that are not absorbed by the violet dye may be transmitted through to the next sheet, where they may be absorbed and re-emitted by a fluorescent green dye. Highly reflective mirrors are fixed to the back edge of each sheet to reflect photons that are originally directed away from the light guides. Some emitted photons may also leave each sheet at its base, and enter the pink sheet at the bottom of the stack, where they may be subsequently re-absorbed and re-emitted. A white base plate is placed underneath the pink sheet to reflect any light reaching the base of the stack, thus increasing the absorption efficiency of the stack. Although total internal reflection effectively transports light to the end of the light guides, a large quantity of this may be trapped (as shown in Fig. 2a). Assuming the dye molecules emit isotropically, approximately 1/8 of the emitted light will be lost at each of the six surfaces of the sheet. In an ideal system with no self-absorption or matrix scattering, the remaining 2/8 of the total emitted light falls within an angular range such that it is totally internally reflected at every surface and is totally trapped within the system. Back edge mirrors reflect light that would ordinarily leave through the back edge, and most of this is transported through the light guides to the luminaire. In summary with no losses, 2/8 of the total emitted light will leave the system at the collection edge, 2/8 is lost through the sides, 2/8 is lost through the top and bottom surfaces and the remaining 2/8 is trapped, and will at some stage reach the collection edge. Provided no air gaps exist between the collector and the light guides, 4/8 of the light can travel to the collection edge. That is to say that approximately half of the light reaching the collection edge will naturally leave through this edge and the other half is trapped inside the light guide. Most of the trapped light can be extracted by installation of an appropriate luminaire [20]. A PMMA luminaire is coupled with optically clear glue to the end of the light guides to extract the trapped light (as shown in Fig. 2b). The luminaire enlarges the aperture at the end of the light guide, creating an additional area in which rays can be reflected. Each of the sides and the back of the luminaire are sandblasted and painted with diffuse white paint so they are rough and highly
Fig. 2. Luminaire for extraction of trapped light: (a) without luminaire 1/4 of emitted rays are trapped and (b) luminaire extracts trapped light.
ARTICLE IN PRESS A.A. Earp et al. / Solar Energy Materials & Solar Cells 84 (2004) 411–426
414
reflective. Trapped light is extracted by the following process. As mentioned above, about half of the rays reaching the smooth front surface of the luminaire will exit on their first pass, and the rest will be totally internally reflected. Most of these totally internally reflected rays then reflect randomly off the rough back surface, highly increasing the probability that they will leave the front surface of the luminaire on their next pass. If a ray is again reflected from the front surface, it may bounce around inside the luminaire a number of times, and each time it strikes the rough back and sides it will reflect randomly, until it either leaves the front surface of the luminaire or re-enters the light guide. If the geometry of the luminaire is well chosen, the fraction of rays re-entering the light guide will be negligible. A simple welldesigned luminaire can extract around 80% of the trapped light actually reaching the luminaire. A single luminaire is used for all three light guides in the stack. Any trapped light extracted at the collection edge adds to the un-trapped light, significantly boosting the output. The luminaire gain factor is defined as the ratio of the output lumens for an LSC stack with a luminaire to the output of the same stack before the luminaire was installed. Maximum luminaire gain is achieved with a perfectly non-scattering light guide and collector system with no self-absorption. Trapped light has a longer path length than other light reaching the luminaire, so losses due to unavoidable attenuation processes will be more significant. Practically a gain factor of around 1.50–1.60 is achievable for a well-designed luminaire. 2.2. Stack modelling theory In order to improve the understanding of the behaviour of LSCs, they can be theoretically modelled using the analytic method [21] or with Monte Carlo simulations [22]. For the lighting application of LSCs, a different kind of model is necessary. The light output of our LSC stack was predicted using the following model, based on previously published theory [23]. Let the fluorescent dye have an emission power spectrum eo(l) (with SI units W/m2/nm, where the subscript ‘‘o’’ denotes the spectrum of emitted radiation prior to encountering other dye molecules). l is the wavelength and a(l) is the dye related attenuation coefficient at some nominal concentration of the dye in a particular matrix. Consider a single dye molecule a distance l from the collection edge. For a LSC of length L, width w and thickness t, the spectral intensity at the collection edge of the collector resulting from the uniform illumination of the sheet by an external source is Z Z p=2 eo ðlÞ Ee 1 L R eðl; LÞ ¼ dl sin y dy eo ðl0 Þdl0 wtðp=2Þ L 0 w Z 0
arcsinðcosw=sin yÞ
ðaðlÞ þ am Þl exp df; ðsiny cos fÞ
ð1Þ
where the angles f and y are standard spherical coordinates (as defined in Fig. 3), w is the critical angle of the matrix material and am represents the matrix attenuation coefficient, assumed here to be independent of wavelength. am is measured via a
ARTICLE IN PRESS A.A. Earp et al. / Solar Energy Materials & Solar Cells 84 (2004) 411–426
415
Fig. 3. Luminescent Solar Concentrator of length L, illuminated by source S(l), produces end emission e(l,L) at collection edge.
logarithmic relationship to the light transport performance parameter L1=2 ; which is described in Section 3.2. The distance travelled by a particular ray emitted from the dye molecule before it reaches the collection edge is l/sin y cos f. For the case in which sunlight with spectral intensity S(C) is distributed uniformly over the top surface of the collector (with area wL and front surface reflectance Rc) at normal incidence, the total power emitted by the dye molecules Ee is given by Z Ee ¼ wLð1 Rc Þ ð1 eaðCÞt ÞSðCÞZe ðCÞ dC; ð2Þ where Rc is the reflectivity of one surface of the collector and Ze the energy-to-energy conversion efficiency of the dye––i.e. the emitted photon energy as a fraction of the incident photon energy multiplied by the quantum efficiency of the dye, Zq. Zq is the ratio of emitted photons to absorbed photons. The base reflector enables a reduction in dye concentration, which reduces transport losses. The back edge mirrors enable an additional integral of the form of Eq. (1), with the limits from L to 2L, corrected by the mirror reflectance. A cover sheet is placed over the LSC stack to prevent weathering and dye degradation due to UV light. It contains a UV blocking dye that absorbs most of the spectrum below 360 nm, and transmits most UV above 380 nm. Approximately 10% of the signal is lost to Fresnel reflection from the cover sheet. The effect of the cover on the output will be discussed in more detail in Section 3. In ordinary operation the top surface of the LSC is uniformly illuminated by sunlight, producing a luminous flux, FL, at the collection edge. Z ð3Þ FL ¼ k eðl; LÞyðlÞ dl; where k is a lumens per watt conversion factor and y(l) is the standard photopic response of the human eye. While the intensity at the collection edge is by far the largest of any LSC surface, smaller luminous fluxes are also produced at each of the sides and the base of each LSC.
ARTICLE IN PRESS 416
A.A. Earp et al. / Solar Energy Materials & Solar Cells 84 (2004) 411–426
In a three-colour LSC stack, the emitted flux leaving the base of each sheet is transferred to the next sheet, where it acts as an additional energy source. If this emitted flux falls within the absorption region of the dye in the next collector, it will also contribute to the total absorbed energy, augmenting the sun spectrum S(C) in Eq. (2). Conversely, each coloured sheet absorbs some of the solar spectrum, transmitting a modified solar spectrum to the sheet below it in the stack. A reflector at the base of the stack redirects light back into the coloured sheets, increasing the possibility of absorption. The effective incident energy available for absorption in one pass of the 2 mm thick pink sheet is shown in Fig. 4 as a solid line. This line is above the AM 1.5 solar spectrum (the dashed line), due to emission from the green sheet in the range 480–550 nm, and the effect of the base reflector at wavelengths where aðlÞ is reduced. aðlÞ for the pink dye is represented by the dotted line (with an arbitrary vertical scale), and it can be seen that the pink absorption largely overlaps with the green emission region. Similarly, the emission of the violet dye falls within the absorption region of the green dye between 420 and 500 nm. Thus the actual absorbed energy by each sheet in a three-colour stack may be more than would be absorbed by the same collectors individually in direct sun. As the dyes show negligible absorption above 600 nm, solar radiation in this region does not contribute to the output lumens in this LSC system. Two physical processes significantly affect the energy output spectrum at the collection edge of the LSC stack: Stokes shift and dye self-absorption. Stokes shift is a purely molecular affect as absorbed photons are emitted with a lower energy, causing a spectral shift to higher wavelengths, as demonstrated for the pink dye in Fig. 5. The dotted line represents the emission spectrum of the pink dye molecule. Self-absorption is observed when emitted photons have a wavelength that falls within the overlapping region of the absorption and emission spectra of the dye (also
Fig. 4. Absorption and incident energy spectra for 50 ppm pink sheet at the bottom of a three-colour LSC stack on a base reflector.
ARTICLE IN PRESS A.A. Earp et al. / Solar Energy Materials & Solar Cells 84 (2004) 411–426
417
Fig. 5. Output spectra showing Stokes shift and self-absorption shifts for two L values with fluorescent pink dye.
shown in Fig. 5). Self-absorption causes the loss of photons in this overlapping region, excluding them from the final output spectrum. Thus as collector length increases, self-absorption causes a red shift in the colour of the output light, and diminishes the area under the emission curve, thereby reducing the intensity of the output energy. A significant spectral shift is observed as the path length is increased to 13.5 cm, represented by small dashes in Fig. 5. At a path length L=1.200 m this shift has reached saturation (represented by long dashes). For longer collectors the losses due to matrix absorption and scattering will be increased. If the dye suffers from photodecomposition, further parasitic dye losses may be experienced as incident light is absorbed by non-fluorescing impurities. These losses have not been included in the model, as the UV-blocking cover should prevent significant photodecomposition from occurring.
3. Results and analysis 3.1. Output colour Many aspects of the light transport in LSCs were modelled using the theory described above, for all three coloured sheets and for the stack as a complete unit. Firstly, the Stokes spectral shift and self-absorption losses were modelled in order to calculate the output spectrum of the three-colour LSC stack, which is displayed in Fig. 6. The solid line represents the measured spectral output of the LSC stack, and the dashed line represents the modelled output. At this stage spectral measurements of the output of the LSC stack have not been made for the entire system, so a reasonable approximation is given here by the superposition of the measured outputs
ARTICLE IN PRESS 418
A.A. Earp et al. / Solar Energy Materials & Solar Cells 84 (2004) 411–426
Fig. 6. Output spectrum of the three-colour LSC stack.
of the three individual sheets. From Fig. 6 it is evident that the stack has three evenly distributed spectral peaks, centering close to the peak photopic response of the eye, represented as a dotted line. Each of the three dyes used in the LSC stack were chosen because of their high quantum efficiencies (Zq>90%), and because the combined colour is very close to white. Pink dye is chosen here over red dyes that have been used in the past, because it has a larger overlap with the spectral response of the eye. Green dye provides the bulk of the luminous output (see discussion in Section 3.4) as its peak output wavelength is very close to the peak spectral response of the eye. Violet, however, has an output spectrum that is very poorly matched with the eye response, and it is only used here to balance the colour, providing a near-white light source. Blue dye with a peak wavelength around 450 nm would be preferable for this purpose, but blue dyes with high quantum efficiencies are not available to our knowledge. In Fig. 7 the CIE colour coordinates are shown for the pink, green and violet sheets, as well as the stack with and without a UV-blocking cover. Points 1 and 2 represent warm white and cool white, respectively, and the stack with no cover almost exactly matches the point on the black body curve representing the colour of daylight. As previously suggested, a higher wavelength blue dye located roughly in line with the end of the blackbody curve would produce a better output colour, more closely matched to the photopic eye response. Alternatively, a green dye with a lower wavelength (towards the top left of the CIE diagram) would be more suitable for producing white output from two colours, and would minimise the need for a violet sheet in the stack. Dye stability is also a major consideration in the selection of dyes for LSCs [24,25]. The use of a violet LSC imposes limitations on the output colour, due to the need for a UV-blocking cover to lengthen the lifetime of the violet dye. If left exposed to the UV radiation from the sun, the violet sheet degrades after only a few months,
ARTICLE IN PRESS A.A. Earp et al. / Solar Energy Materials & Solar Cells 84 (2004) 411–426
419
Fig. 7. CIE colour co-ordinates of the output of pink, green and violet sheets and the LSC stack.
whereas the lifetime is extended to many years by the use of a special UV-blocking cover. With a UV-blocking cover and a violet sheet above them, the pink and green dyes do not significantly degrade. The absorption of the UV-blocking cover overlaps significantly with the absorption spectrum of the violet dye, thereby reducing the violet output and affecting the colour of the final output of the stack. This difference is shown in Fig. 7, where the stack output with the cover is slightly greener than without the cover. In order to modify the output colour to the same white as produced without the stack, the luminous output of the green and pink sheets must be reduced. Hence with the UV-blocking cover over the stack, output intensity must be sacrificed to achieve the desired colour. Actual intensity values for the LSCs and stack are given in Section 3.4. Furthermore, the colour rendering properties of the output light are also slightly reduced by the addition of the cover, as its colour rendering index (CRI) is reduced from 78 to 76. 3.2. Light transport performance Light transport performance in an LSC is governed by a number of parameters including dye self-absorption, total internal reflection losses and matrix losses [26,27]. A simple output performance parameter was devised to describe the overall light transport performance of one sheet. This performance parameter is called the ‘half-length’, denoted by the symbol L1=2 : Half-length is defined as the distance light travels along a sheet of a given length, over which the intensity falls by 50%. In practice, L1=2 varies with the length of the sheet, so the test needs to be done at a fixed length [28]. A length of 1.200 m has been used.
ARTICLE IN PRESS 420
A.A. Earp et al. / Solar Energy Materials & Solar Cells 84 (2004) 411–426
Table 1 Variation of half-length with LSC surface quality Sample
60 ppm violet 120 ppm violet 30 ppm green 60 ppm green 30 ppm pink 50 ppm pink
Half-Length (m) Rough polished edges
Diamond polished edges
0.72 0.87 0.74 0.77 0.56 0.51
1.45 1.74 1.16 1.04 0.97 0.88
The half-length of a LSC is greatly affected by its surface quality. In Table 1 the half-length of 1.20 m-long violet, green and pink LSCs are shown at various concentrations, and with two grades of side-surface finish quality. Half of these collectors have been cut with rough polished edges, and the edges of the other half were diamond polished, giving them a smooth finish. For the violet LSCs the difference in performance is extreme; the observed half-length is doubled when the edges are diamond polished as opposed to rough polished. Less of a difference is noted for the green and pink LSCs, but without a doubt diamond polishing the edges significantly improves the light transport performance. Two further observations can be made about the half-length data in Table 1. Firstly, for these samples half-length increases with decreasing dye wavelength. Secondly, for the violet sheets, the half-length of the high concentration sheet is greater than that of the low concentration sheet. This is contrary to the expected relationship between half-length and dye concentration. Upon closer examination, the low-concentration violet sheet was found to be milky in appearance––a phenomenon often associated with scattering from incompletely dissolved dye. Despite this apparent production fault, the violet sheets have by far the largest halflength of the three colours, indicating the highest light transport efficiency. However, for various reasons outlined in Sections 3.1 and 3.4, they are not ideal for this type of daylighting system. 3.3. Effect of dye residual attenuation Pure matrix losses can be measured from a clear matrix without the presence of the dye. As well as these matrix losses there are weak, almost spectrally flat losses due to the dye, which occur in the wavelength range where the sheet output (as shown in Fig. 5) occurs [28]. This finite loss causes no spectral shift but does impact significantly on final output intensity. It can be observed directly but is difficult to measure accurately in a 2 mm-thick sheet, as it is complicated by fluorescence. A simple wavelength independent parameter, d, has been devised to describe this residual attenuation loss. Let T=To be the transmittance with no residual attenuation across the output spectrum. In practice the total transmission is T(l)=To(l)(1–d). By comparing measured values of L1=2 and output lumens to
ARTICLE IN PRESS A.A. Earp et al. / Solar Energy Materials & Solar Cells 84 (2004) 411–426
421
Table 2 Effect of attenuation factor d on half-length and luminous output of 2 mm thick 120 ppm violet, 50 ppm pink and 60 ppm green sheets (without a luminaire) d (t=2 mm)
Violet 120 ppm
Green 60 ppm
Pink 50 ppm
L1/2 (m)
Lumens
L1/2 (m)
Lumens
L1/2 (m)
Lumens
0 5 104 1 103
2.78 1.51 0.96
109 95 80
1.50 0.77 0.56
1829 1315 1060
3.19 1.62 0.38
915 398 280
Measured values
1.74
80
1.04
1122
0.88
222
simulations using various values of d (see Table 2), d for each dye is estimated to fall within the range 1 104od o1 103. Each of these results were achieved by modelling the half-length with the tails region of the absorption spectra modified so that the shape remained the same, but the average value of d was either zero, 5 104 or 1 103. Each colour dye has a different wavelength range for residual attenuation, so these modifications were made over the appropriate range for each dye. For the violet dye, the tails region is l>420 nm, for green l>520 nm, for pink l>600 nm, and 800 nm is the maximum wavelength in each case. From these results it is evident that the light transport performance of the collectors are extremely sensitive to small changes in d. With d =1 103 the halflength of the pink dye is 12% of the value calculated for the same dye with d=0. The green and violet dyes show smaller reductions, at approximately 35% of the halflength with d=0. Likewise, d=1 103 produces luminous outputs ranging from 30% to 73% of the output with d=0, depending on the dye colour. This extreme sensitivity suggests that these fine details of the dye attenuation spectrum play a major role in determining the light transport efficiency and ultimately the light output of a LSC. From Fig. 8 it can be seen that d also has an effect on the optimum dye concentration. When d is increased, the luminous output and optimum dye concentration both decrease. Hence d is a dye-related constant, and it should be minimised for the best possible light output. 3.4. Luminous output and light-to-light efficiency Modelled and measured luminous outputs and light-to-light efficiencies for each colour LSC and the three-colour stack are shown in Table 3 (the cover sheet is used in all cases), as well as the effect of adding the luminaire to the stack. Here light-tolight efficiency is defined as the number of lumens measured at the luminaire divided by the number of lumens falling on the cover sheet from the sun. In each case the most accurate tails data available have been used for the simulations. In the simulations it was assumed that 50% of the rays reaching the end of the sheet were trapped inside the sheet in the absence of a luminaire. In practice this value is probably closer to 40%, so the modelled output may be slightly underestimated.
ARTICLE IN PRESS 422
A.A. Earp et al. / Solar Energy Materials & Solar Cells 84 (2004) 411–426
Fig. 8. Effect of d on concentration dependence of output of pink dyed sheet.
Table 3 Luminous outputs and light-to-light efficiencies of a three-colour stack and individual LSCs under solar illumination of 100,000 lx (with a UV-blocking cover sheet) LSC type
Measurements
Model
Lumens
Efficiency (%)
Lumens
Violet Green Pink
48 937 239
0.29 5.8 1.5
50 939 237
Stack (no luminaire) Stack (with luminaire)
783 995
4.8 6.1
1274 1656
Efficiency (%) 0.31 5.8 1.5 7.9 10.2
It is clear that most of the output lumens come from the green LSC, as it is best matched with the human eye response (see Fig. 6). In direct sun the green sheet gives approximately 940 lm, while the pink gives around 240 lm, and the violet gives around 50 lm. Thus green is the most important colour for luminous output, and the other colours play lesser roles, contributing more to the colour of the output light than the actual intensity. As discussed in Section 3.1, the special UV-blocking dye in the cover is required to increase the lifetime of the violet dye, but inevitably the violet light output is reduced. For the green and pink LSCs, an 11–13% reduction in luminous output is observed when the cover sheet is used, and this can mostly be attributed to Fresnel reflection losses. For the violet sheet, however, the equivalent reduction in lumens is three to four times more severe, due to the overlapping of the absorption spectra of the special UV-blocker and the violet dye. A blue dye with a higher peak wavelength would be less affected by this UV-blocker, so if the violet were replaced with blue, greater luminous output would be achievable.
ARTICLE IN PRESS A.A. Earp et al. / Solar Energy Materials & Solar Cells 84 (2004) 411–426
423
An alternative approach that is currently being investigated for future designs is to replace the violet LSC with a blue LED [29], which could be powered by a photovoltaic cell. Around 1 W would be required to supply the lumens that are currently contributed by the violet sheet. The pink and green outputs could be brighter without the losses from the violet. Using this system, any colour balancing is done by adjusting the brightness of the LED; so white light output can be achieved without sacrificing any of the lumens from the green and pink. As both the LED and the LSCs are driven by input from the sun, the stack output would be continually colour balanced, irrespective of the brightness of the sun. It can be seen from Table 3 that the luminaire successfully extracts some of the trapped light, although it does not achieve the estimated luminaire gain factor of 1.50. As discussed in Section 2.1, a luminaire should theoretically be able to extract around 80% of the trapped light, although 50% is a more realistic value. For the system presented in this paper, the luminaire was not well designed, so a luminaire gain of around 1.30 was achieved, and this value has been used in the model. Tests on simple sheets have shown that a luminaire gain factor of 1.50–1.60 is achievable with a well-designed luminaire. For the individual LSCs, the modelled results are in fairly good agreement with the measured values, however the modelled LSC stack output is overestimated by about 60%. This suggests that there are other significant sources of loss specific to the three-colour stack that are not fully accounted for in the model. Poor surface flatness and surface dirt may be causing more low angle light loss than expected. Some of the LSCs have up to 6% variation in thickness, and ray tracing simulations have shown that when joined to the PMMA light guide, this variation can cause up to 8% light loss. Further losses were caused by excess adhesive on the surface of the LSCs, residue from the protective sheets used by the manufacturers for shipping. Light can transmit into neighbouring sheets via the glue, causing large losses. This glue is very difficult to remove, and the attempt to clean the surface creates scratches, which cause further light loss. Hence for future LSCs, a different adhesive will be used for these protective sheets––one that does not leave any marks on the LSC surfaces. The above losses can be minimised by careful LSC production, ensuring that the top and bottom sheet surfaces are flat and parallel, clean and free from scratches. Despite all the losses, the LSC stack has quite a good optical performance. Under solar illumination of 100,000 lx an output of around 1000 lm is observed. Theoretically an output of around 1650 lm is achievable with the current luminaire, and around 1900 lm with a luminaire gain of 50%. The output lumens would decrease slightly in cloudy conditions, and if sunlight availability becomes very low, an artificial back-up lighting system will be required. Nonetheless, the LSC stack accepts both specular and diffuse light, making good use of all available sunlight. Although the performance of most daylighting systems is assessed by the lux levels produced, luminous output is given here in lumens, because our LSC system functions like a standard lamp fitting. It is conceived that residential room lighting is the most likely application of this system––e.g. bathrooms/kitchens and the like. For a room of known floor area, the number of LSC daylighting systems required can be
ARTICLE IN PRESS 424
A.A. Earp et al. / Solar Energy Materials & Solar Cells 84 (2004) 411–426
determined by calculating the number of lumens per square meter. For most rooms in an average house, one system per room should be sufficient. As with all daylighting systems, supplementary artificial lighting is required in highly overcast conditions. An LSC stack with high light-to-light efficiency ensures that optimal light output is achieved whatever the sky conditions. Light-to-light efficiency for an LSC is very size dependent, so it can only be compared between LSCs of similar sizes. For this particular stack it was anticipated that an efficiency of 5% would be sufficient for general household room lighting, and a value of 6% was achieved by the prototype. Our model predicts that this can be increased to 10% if residual glue and surface roughness losses are effectively minimised, and with a luminaire gain factor of 1.50 it should be possible to achieve a theoretical efficiency of 12%. Another outstanding feature of the LSC stack is its high luminous efficacy of 311 lm/W. However, this is partly due to the non-neutral slightly greenish colour of the output. If the cover sheet is removed, the output is a neutral white with a luminous efficacy of 294 lm/W. If the violet LSC were replaced with a blue LED, it is estimated that the luminous efficacy would drop to 250 lm/W, but the output would be brighter than the current system. These values of luminous efficacy by far exceed that of natural daylight, which generally falls within the range 100–130 lm/W [30]. When compared to standard artificial light sources such as incandescent light bulbs (16–40 lm/W) and fluorescent lamps (50–80 lm/W), it is clear that the output of the LSC stack is significantly more energy efficient. The above results show that with further improvements, this three-colour LSC daylighting system is capable of producing a high luminous output, with good lightto-light efficiency and luminous efficacy. These improvements are well within reach, and will be included in future prototypes.
4. Conclusions A daylighting system has been produced, which transports sunlight to remote areas of a building using a stack of pink, green and violet LSCs and clear PMMA light guides. In direct sun of intensity 100,000 lx, prototypes with collector area 1.2 m 0.135 m deliver 1000 lm of near-white light with a luminous efficacy of 311 lm/W and a light-to-light efficiency of 6%. Surface effects such as excess adhesive and variations in flatness are thought to be causing unnecessary light loss, which can be avoided by careful LSC production. It is proposed that the colour and intensity of the output light, and the efficiency of the system may be improved if the violet LSC is replaced with either a blue LSC or a blue solar-powered LED. In this case colour matching would be achieved by controlling the LED output, thus maintaining maximum light output from the green and pink LSC stack. A 30% increase in luminous output is observed by the installation of a single luminaire, although it is expected that the efficiency of the luminaire will be significantly improved by modifying the design, leading to a further increase in output. Thus it is demonstrated that the three-colour LSC stack is a
ARTICLE IN PRESS A.A. Earp et al. / Solar Energy Materials & Solar Cells 84 (2004) 411–426
425
suitable system for daylighting in remote areas of buildings, many metres away from the nearest available daylighting aperture. With the suggested improvements it is expected that the performance of the system may be improved significantly.
Acknowledgements The author would like to acknowledge the support of an Australian Postgraduate Award. The work was in part supported by BASF and Skydome industries.
References [1] [2] [3] [4] [5] [6] [8]
[9]
[10] [11] [12] [13]
[14] [15] [16] [17] [18] [19] [20] [21] [22]
[23] [24] [25]
J.C. Lam, D.H.W. Li, Solar Energy 65 (4) (1999) 251–262. C.A. Balaras, K. Droutsa, A.A. Argiriou, K. Wittchen, Energy Build. 34 (2) (2002) 135–153. D.H.W. Li, J.C. Lam, Energy Build. 33 (2001) 793–803. E. Ne’eman, Energy Build. 6 (2) (1984) 97–108. A.J.F. Rutten, Build. Environ. 29 (1) (1994) 105–111. C.L. Robbins, Daylight, Buildings and Design, Van Nostrand Reinhold, New York, 1986, pp. 3–13. R. Compagnon, J.L. Scartezzini, B. Paule, Application of nonimaging optics to the development of new daylighting systems, in: Proceedings of the ISES Solar World Congress 1993, Budapest, Hungary, 1993. J. Page, J. Kaempf, J.L. Scatrezzini, Assessing daylighiting performances of electrochromic glazings coupled to an anidolic device, in: Proceedings of the ISES Solar World Congress, Gothenburg, Sweden, 2003. I.R. Edmonds, Sol. Energy Mater. Sol. Cells 29 (1) (1993) 1–26. W. Pohl, C. Anslem, Natural room illumination using sunlight, in: Proceedings of the World Renewable Energy Congress VII (WREC 2002), Cologne, Germany, 2002. S. Chirarattananon, S. Chedsiri, L. Renshen, Solar Energy 69 (4) (2000) 331–341. A. Zastrow, V. Witter, Materials and optics for solar energy conversion and advanced lighting technology, Daylighting with mirror light pipes and with fluorescent planar concentrators, in: Proceedings of the SPIE, 1986, 227–234. A. Zastrow, V. Wittwer, Optical materials technology for energy efficiency and solar energy conversion SPIE 653 (1986) 93–100. A.M. Hermann, Solar Energy 29 (4) (1982) 323–329. R. Reisfeld, C.K. Jorgensen, Struct. Bond. 49 (1982) 1–36. W.H. Weber, J. Lambe, Appl. Opt. 15 (10) (1976) 2299–2300. R. Soti, E. Farkas, M. Hilbert, Z. Farkas, I. Ketskemety, J. Lumin. 68 (2–4) (1996) 105–114. G.B. Smith, J. Franklin, Sunlight collection and transfer system, US Patent 5,548,490, 1992. J. Franklin, Light systems for transmitting and releasing luminescent radiation, US Patent 6,272,265 B1, 2001. A. Goetzberger, V. Wittwer, Fluorescent Planar Concentrators for Solar Energy Conversion, F.Viewig, Braunschweig, 1979, pp. 427–451. V. Wittwer, A. Goetzberger, K. Heidler, Fluorescent planar concentrator (FPC) Monte Carlo computer model limit efficiency and latest experimental results, in: Proceedings of the Fourth E.C. Photovoltaic Solar Energy Conference, Stresa, Italy, 1982, 682–686. P.D. Swift, G.B. Smith, J. Franklin, Light to light efficiencies in luminescent solar concentrators, in: Proceedings of the SPIE Conference on Solar Optical Materials, Denver, CO, 1999, pp. 21–28. J.S. Batchelder, A.H. Zewail, T. Cole, Appl. Opt. 20 (21) (1981) 3733–3754. A.F. Mansour, Polymer Test. 17 (1998) 333–343.
ARTICLE IN PRESS 426
A.A. Earp et al. / Solar Energy Materials & Solar Cells 84 (2004) 411–426
[26] [27] [28] [29] [30]
W.R.L. Thomas, J.M. Drake, M.L. Lesiecki, Appl. Opt. 22 (21) (1983) 3440–3450. J.S. Batchelder, A.H. Zewail, T. Cole, Appl. Opt. 18 (18) (1979) 3090–3110. A.A. Earp, G.B. Smith, P.D. Swift, J. Franklin, Solar Energy 76 (6) (2004) 655–667. G.B. Smith, J. Franklin, A hybrid lighting system, US Patent pending, 2002. J.C. Lam, D.H.W. Li, Energy Convers. Manage. 37 (12) (1996) 1703–1711.
Solar Energy Materials & Solar Cells 84 (2004) 411–426
Optimisation of a three-colour luminescent solar concentrator daylighting system Alan A. Earp*, Geoff B. Smith, Jim Franklin, Paul Swift Department of Applied Physics, University of Technology, P.O. Box, 123 Broadway, 2007 NSW, Sydney, Australia Received 4 November 2003; accepted 16 February 2004 Available online 25 May 2004
Abstract Electrical power consumption in buildings may be reduced considerably by more efficient use of sunlight for indoor lighting. A stack of luminescent solar concentrator sheets has been developed, which utilises three different coloured fluorescent dyes to produce a concentrated near-white light source to be coupled into flexible polymer sheets and transported up to 10 m. Thus in clear sky conditions a remote room with neither walls nor roof in direct contact with the sun, can be illuminated by over 1000 lm of natural light with a luminous efficacy of over 300 lm/W, regardless of the sun’s position. r 2004 Elsevier B.V. All rights reserved. Keywords: Daylight; Luminescent Solar Concentrator; Light guide; Lighting
1. Introduction Daylight is an underused resource that has the potential to improve the quality of indoor lighting, as well as substantially reducing energy costs. Artificial lighting is one of the major sources of electrical energy costs in office buildings, both directly through lighting energy consumption and indirectly by production of significant heat gain, which increases cooling loads [1]. Recent studies have shown that effective use of daylighting in conjunction with artificial lighting can produce significant energy savings in office buildings [2–5]. The excellent colour rendering properties of daylight *Corresponding author. Tel.: +61-2-9514-2225; fax: +61-2-9514-2219. E-mail address: [email protected] (A.A. Earp). 0927-0248/$ - see front matter r 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.solmat.2004.02.046
ARTICLE IN PRESS 412
A.A. Earp et al. / Solar Energy Materials & Solar Cells 84 (2004) 411–426
and its close match to the photopic response of the human eye make it an ergonomic light source that is generally preferred for pleasant working conditions [6]. Thus the use of daylight to supplement artificial lighting has proven beneficial for minimising energy usage and improving lighting quality. Daylighting in perimeter zones of buildings is readily achievable through careful window design [3,7]. For regions further away from the window fa@ade reasonable daylighting is possible with light shelves [8,9] or laser cut light deflecting panels [10]. However, these systems are only capable of transporting daylight over relatively small distances. Various mirror light pipe systems have been proposed for daylighting of deeper interior spaces with light pipes [11–13], but these systems can be fairly complex, and to date are still at a prototype stage. Luminescent solar concentrators (LSCs) also hold considerable potential for application to daylighting, due to their capacity to collect and concentrate sunlight [14]. LSCs have been extensively studied since the mid-1970s, but most of this work has been focused on photovoltaic applications [15–18]. The small amount of work done to date on daylighting with LSCs shows considerable potential, providing the motivation for this study.
2. Three-colour LSC stack operation 2.1. LSC stack design Daylight can be transported to remote areas using a stack of three coloured LSCs connected to clear flexible light guides [19], as shown in Fig. 1. Each coloured LSC consists of a clear polymethylmethacrylate (PMMA) matrix with the dimensions 1.200 m 0.135 m 0.0020 m, and is doped with a coloured fluorescent dye. For the purpose of light transport these collectors are coupled with optically clear glue to clear PMMA light guides up to 5 m in length. Thus daylight can be transported to underground rooms or windowless rooms in the centre of a building. By appropriate selection of dyes, white light output can be produced, giving a good match with the colour of daylight. Optical transport of daylight using an LSC stack operates by the following process. Solar energy enters the stack where it may be absorbed and randomly re-emitted by
Fig. 1. Schematic of a three-colour LSC stack connected to light guides and luminaire.
ARTICLE IN PRESS A.A. Earp et al. / Solar Energy Materials & Solar Cells 84 (2004) 411–426
413
the violet dye in the top sheet. If the emitted photon travels below the critical angle with respect to the top or side surface of the sheet (42o for PMMA), it will leave the collector. Otherwise it will be totally internally reflected to the end of the collector, then through the light guide to the luminaire. Photons that are not absorbed by the violet dye may be transmitted through to the next sheet, where they may be absorbed and re-emitted by a fluorescent green dye. Highly reflective mirrors are fixed to the back edge of each sheet to reflect photons that are originally directed away from the light guides. Some emitted photons may also leave each sheet at its base, and enter the pink sheet at the bottom of the stack, where they may be subsequently re-absorbed and re-emitted. A white base plate is placed underneath the pink sheet to reflect any light reaching the base of the stack, thus increasing the absorption efficiency of the stack. Although total internal reflection effectively transports light to the end of the light guides, a large quantity of this may be trapped (as shown in Fig. 2a). Assuming the dye molecules emit isotropically, approximately 1/8 of the emitted light will be lost at each of the six surfaces of the sheet. In an ideal system with no self-absorption or matrix scattering, the remaining 2/8 of the total emitted light falls within an angular range such that it is totally internally reflected at every surface and is totally trapped within the system. Back edge mirrors reflect light that would ordinarily leave through the back edge, and most of this is transported through the light guides to the luminaire. In summary with no losses, 2/8 of the total emitted light will leave the system at the collection edge, 2/8 is lost through the sides, 2/8 is lost through the top and bottom surfaces and the remaining 2/8 is trapped, and will at some stage reach the collection edge. Provided no air gaps exist between the collector and the light guides, 4/8 of the light can travel to the collection edge. That is to say that approximately half of the light reaching the collection edge will naturally leave through this edge and the other half is trapped inside the light guide. Most of the trapped light can be extracted by installation of an appropriate luminaire [20]. A PMMA luminaire is coupled with optically clear glue to the end of the light guides to extract the trapped light (as shown in Fig. 2b). The luminaire enlarges the aperture at the end of the light guide, creating an additional area in which rays can be reflected. Each of the sides and the back of the luminaire are sandblasted and painted with diffuse white paint so they are rough and highly
Fig. 2. Luminaire for extraction of trapped light: (a) without luminaire 1/4 of emitted rays are trapped and (b) luminaire extracts trapped light.
ARTICLE IN PRESS A.A. Earp et al. / Solar Energy Materials & Solar Cells 84 (2004) 411–426
414
reflective. Trapped light is extracted by the following process. As mentioned above, about half of the rays reaching the smooth front surface of the luminaire will exit on their first pass, and the rest will be totally internally reflected. Most of these totally internally reflected rays then reflect randomly off the rough back surface, highly increasing the probability that they will leave the front surface of the luminaire on their next pass. If a ray is again reflected from the front surface, it may bounce around inside the luminaire a number of times, and each time it strikes the rough back and sides it will reflect randomly, until it either leaves the front surface of the luminaire or re-enters the light guide. If the geometry of the luminaire is well chosen, the fraction of rays re-entering the light guide will be negligible. A simple welldesigned luminaire can extract around 80% of the trapped light actually reaching the luminaire. A single luminaire is used for all three light guides in the stack. Any trapped light extracted at the collection edge adds to the un-trapped light, significantly boosting the output. The luminaire gain factor is defined as the ratio of the output lumens for an LSC stack with a luminaire to the output of the same stack before the luminaire was installed. Maximum luminaire gain is achieved with a perfectly non-scattering light guide and collector system with no self-absorption. Trapped light has a longer path length than other light reaching the luminaire, so losses due to unavoidable attenuation processes will be more significant. Practically a gain factor of around 1.50–1.60 is achievable for a well-designed luminaire. 2.2. Stack modelling theory In order to improve the understanding of the behaviour of LSCs, they can be theoretically modelled using the analytic method [21] or with Monte Carlo simulations [22]. For the lighting application of LSCs, a different kind of model is necessary. The light output of our LSC stack was predicted using the following model, based on previously published theory [23]. Let the fluorescent dye have an emission power spectrum eo(l) (with SI units W/m2/nm, where the subscript ‘‘o’’ denotes the spectrum of emitted radiation prior to encountering other dye molecules). l is the wavelength and a(l) is the dye related attenuation coefficient at some nominal concentration of the dye in a particular matrix. Consider a single dye molecule a distance l from the collection edge. For a LSC of length L, width w and thickness t, the spectral intensity at the collection edge of the collector resulting from the uniform illumination of the sheet by an external source is Z Z p=2 eo ðlÞ Ee 1 L R eðl; LÞ ¼ dl sin y dy eo ðl0 Þdl0 wtðp=2Þ L 0 w Z 0
arcsinðcosw=sin yÞ
ðaðlÞ þ am Þl exp df; ðsiny cos fÞ
ð1Þ
where the angles f and y are standard spherical coordinates (as defined in Fig. 3), w is the critical angle of the matrix material and am represents the matrix attenuation coefficient, assumed here to be independent of wavelength. am is measured via a
ARTICLE IN PRESS A.A. Earp et al. / Solar Energy Materials & Solar Cells 84 (2004) 411–426
415
Fig. 3. Luminescent Solar Concentrator of length L, illuminated by source S(l), produces end emission e(l,L) at collection edge.
logarithmic relationship to the light transport performance parameter L1=2 ; which is described in Section 3.2. The distance travelled by a particular ray emitted from the dye molecule before it reaches the collection edge is l/sin y cos f. For the case in which sunlight with spectral intensity S(C) is distributed uniformly over the top surface of the collector (with area wL and front surface reflectance Rc) at normal incidence, the total power emitted by the dye molecules Ee is given by Z Ee ¼ wLð1 Rc Þ ð1 eaðCÞt ÞSðCÞZe ðCÞ dC; ð2Þ where Rc is the reflectivity of one surface of the collector and Ze the energy-to-energy conversion efficiency of the dye––i.e. the emitted photon energy as a fraction of the incident photon energy multiplied by the quantum efficiency of the dye, Zq. Zq is the ratio of emitted photons to absorbed photons. The base reflector enables a reduction in dye concentration, which reduces transport losses. The back edge mirrors enable an additional integral of the form of Eq. (1), with the limits from L to 2L, corrected by the mirror reflectance. A cover sheet is placed over the LSC stack to prevent weathering and dye degradation due to UV light. It contains a UV blocking dye that absorbs most of the spectrum below 360 nm, and transmits most UV above 380 nm. Approximately 10% of the signal is lost to Fresnel reflection from the cover sheet. The effect of the cover on the output will be discussed in more detail in Section 3. In ordinary operation the top surface of the LSC is uniformly illuminated by sunlight, producing a luminous flux, FL, at the collection edge. Z ð3Þ FL ¼ k eðl; LÞyðlÞ dl; where k is a lumens per watt conversion factor and y(l) is the standard photopic response of the human eye. While the intensity at the collection edge is by far the largest of any LSC surface, smaller luminous fluxes are also produced at each of the sides and the base of each LSC.
ARTICLE IN PRESS 416
A.A. Earp et al. / Solar Energy Materials & Solar Cells 84 (2004) 411–426
In a three-colour LSC stack, the emitted flux leaving the base of each sheet is transferred to the next sheet, where it acts as an additional energy source. If this emitted flux falls within the absorption region of the dye in the next collector, it will also contribute to the total absorbed energy, augmenting the sun spectrum S(C) in Eq. (2). Conversely, each coloured sheet absorbs some of the solar spectrum, transmitting a modified solar spectrum to the sheet below it in the stack. A reflector at the base of the stack redirects light back into the coloured sheets, increasing the possibility of absorption. The effective incident energy available for absorption in one pass of the 2 mm thick pink sheet is shown in Fig. 4 as a solid line. This line is above the AM 1.5 solar spectrum (the dashed line), due to emission from the green sheet in the range 480–550 nm, and the effect of the base reflector at wavelengths where aðlÞ is reduced. aðlÞ for the pink dye is represented by the dotted line (with an arbitrary vertical scale), and it can be seen that the pink absorption largely overlaps with the green emission region. Similarly, the emission of the violet dye falls within the absorption region of the green dye between 420 and 500 nm. Thus the actual absorbed energy by each sheet in a three-colour stack may be more than would be absorbed by the same collectors individually in direct sun. As the dyes show negligible absorption above 600 nm, solar radiation in this region does not contribute to the output lumens in this LSC system. Two physical processes significantly affect the energy output spectrum at the collection edge of the LSC stack: Stokes shift and dye self-absorption. Stokes shift is a purely molecular affect as absorbed photons are emitted with a lower energy, causing a spectral shift to higher wavelengths, as demonstrated for the pink dye in Fig. 5. The dotted line represents the emission spectrum of the pink dye molecule. Self-absorption is observed when emitted photons have a wavelength that falls within the overlapping region of the absorption and emission spectra of the dye (also
Fig. 4. Absorption and incident energy spectra for 50 ppm pink sheet at the bottom of a three-colour LSC stack on a base reflector.
ARTICLE IN PRESS A.A. Earp et al. / Solar Energy Materials & Solar Cells 84 (2004) 411–426
417
Fig. 5. Output spectra showing Stokes shift and self-absorption shifts for two L values with fluorescent pink dye.
shown in Fig. 5). Self-absorption causes the loss of photons in this overlapping region, excluding them from the final output spectrum. Thus as collector length increases, self-absorption causes a red shift in the colour of the output light, and diminishes the area under the emission curve, thereby reducing the intensity of the output energy. A significant spectral shift is observed as the path length is increased to 13.5 cm, represented by small dashes in Fig. 5. At a path length L=1.200 m this shift has reached saturation (represented by long dashes). For longer collectors the losses due to matrix absorption and scattering will be increased. If the dye suffers from photodecomposition, further parasitic dye losses may be experienced as incident light is absorbed by non-fluorescing impurities. These losses have not been included in the model, as the UV-blocking cover should prevent significant photodecomposition from occurring.
3. Results and analysis 3.1. Output colour Many aspects of the light transport in LSCs were modelled using the theory described above, for all three coloured sheets and for the stack as a complete unit. Firstly, the Stokes spectral shift and self-absorption losses were modelled in order to calculate the output spectrum of the three-colour LSC stack, which is displayed in Fig. 6. The solid line represents the measured spectral output of the LSC stack, and the dashed line represents the modelled output. At this stage spectral measurements of the output of the LSC stack have not been made for the entire system, so a reasonable approximation is given here by the superposition of the measured outputs
ARTICLE IN PRESS 418
A.A. Earp et al. / Solar Energy Materials & Solar Cells 84 (2004) 411–426
Fig. 6. Output spectrum of the three-colour LSC stack.
of the three individual sheets. From Fig. 6 it is evident that the stack has three evenly distributed spectral peaks, centering close to the peak photopic response of the eye, represented as a dotted line. Each of the three dyes used in the LSC stack were chosen because of their high quantum efficiencies (Zq>90%), and because the combined colour is very close to white. Pink dye is chosen here over red dyes that have been used in the past, because it has a larger overlap with the spectral response of the eye. Green dye provides the bulk of the luminous output (see discussion in Section 3.4) as its peak output wavelength is very close to the peak spectral response of the eye. Violet, however, has an output spectrum that is very poorly matched with the eye response, and it is only used here to balance the colour, providing a near-white light source. Blue dye with a peak wavelength around 450 nm would be preferable for this purpose, but blue dyes with high quantum efficiencies are not available to our knowledge. In Fig. 7 the CIE colour coordinates are shown for the pink, green and violet sheets, as well as the stack with and without a UV-blocking cover. Points 1 and 2 represent warm white and cool white, respectively, and the stack with no cover almost exactly matches the point on the black body curve representing the colour of daylight. As previously suggested, a higher wavelength blue dye located roughly in line with the end of the blackbody curve would produce a better output colour, more closely matched to the photopic eye response. Alternatively, a green dye with a lower wavelength (towards the top left of the CIE diagram) would be more suitable for producing white output from two colours, and would minimise the need for a violet sheet in the stack. Dye stability is also a major consideration in the selection of dyes for LSCs [24,25]. The use of a violet LSC imposes limitations on the output colour, due to the need for a UV-blocking cover to lengthen the lifetime of the violet dye. If left exposed to the UV radiation from the sun, the violet sheet degrades after only a few months,
ARTICLE IN PRESS A.A. Earp et al. / Solar Energy Materials & Solar Cells 84 (2004) 411–426
419
Fig. 7. CIE colour co-ordinates of the output of pink, green and violet sheets and the LSC stack.
whereas the lifetime is extended to many years by the use of a special UV-blocking cover. With a UV-blocking cover and a violet sheet above them, the pink and green dyes do not significantly degrade. The absorption of the UV-blocking cover overlaps significantly with the absorption spectrum of the violet dye, thereby reducing the violet output and affecting the colour of the final output of the stack. This difference is shown in Fig. 7, where the stack output with the cover is slightly greener than without the cover. In order to modify the output colour to the same white as produced without the stack, the luminous output of the green and pink sheets must be reduced. Hence with the UV-blocking cover over the stack, output intensity must be sacrificed to achieve the desired colour. Actual intensity values for the LSCs and stack are given in Section 3.4. Furthermore, the colour rendering properties of the output light are also slightly reduced by the addition of the cover, as its colour rendering index (CRI) is reduced from 78 to 76. 3.2. Light transport performance Light transport performance in an LSC is governed by a number of parameters including dye self-absorption, total internal reflection losses and matrix losses [26,27]. A simple output performance parameter was devised to describe the overall light transport performance of one sheet. This performance parameter is called the ‘half-length’, denoted by the symbol L1=2 : Half-length is defined as the distance light travels along a sheet of a given length, over which the intensity falls by 50%. In practice, L1=2 varies with the length of the sheet, so the test needs to be done at a fixed length [28]. A length of 1.200 m has been used.
ARTICLE IN PRESS 420
A.A. Earp et al. / Solar Energy Materials & Solar Cells 84 (2004) 411–426
Table 1 Variation of half-length with LSC surface quality Sample
60 ppm violet 120 ppm violet 30 ppm green 60 ppm green 30 ppm pink 50 ppm pink
Half-Length (m) Rough polished edges
Diamond polished edges
0.72 0.87 0.74 0.77 0.56 0.51
1.45 1.74 1.16 1.04 0.97 0.88
The half-length of a LSC is greatly affected by its surface quality. In Table 1 the half-length of 1.20 m-long violet, green and pink LSCs are shown at various concentrations, and with two grades of side-surface finish quality. Half of these collectors have been cut with rough polished edges, and the edges of the other half were diamond polished, giving them a smooth finish. For the violet LSCs the difference in performance is extreme; the observed half-length is doubled when the edges are diamond polished as opposed to rough polished. Less of a difference is noted for the green and pink LSCs, but without a doubt diamond polishing the edges significantly improves the light transport performance. Two further observations can be made about the half-length data in Table 1. Firstly, for these samples half-length increases with decreasing dye wavelength. Secondly, for the violet sheets, the half-length of the high concentration sheet is greater than that of the low concentration sheet. This is contrary to the expected relationship between half-length and dye concentration. Upon closer examination, the low-concentration violet sheet was found to be milky in appearance––a phenomenon often associated with scattering from incompletely dissolved dye. Despite this apparent production fault, the violet sheets have by far the largest halflength of the three colours, indicating the highest light transport efficiency. However, for various reasons outlined in Sections 3.1 and 3.4, they are not ideal for this type of daylighting system. 3.3. Effect of dye residual attenuation Pure matrix losses can be measured from a clear matrix without the presence of the dye. As well as these matrix losses there are weak, almost spectrally flat losses due to the dye, which occur in the wavelength range where the sheet output (as shown in Fig. 5) occurs [28]. This finite loss causes no spectral shift but does impact significantly on final output intensity. It can be observed directly but is difficult to measure accurately in a 2 mm-thick sheet, as it is complicated by fluorescence. A simple wavelength independent parameter, d, has been devised to describe this residual attenuation loss. Let T=To be the transmittance with no residual attenuation across the output spectrum. In practice the total transmission is T(l)=To(l)(1–d). By comparing measured values of L1=2 and output lumens to
ARTICLE IN PRESS A.A. Earp et al. / Solar Energy Materials & Solar Cells 84 (2004) 411–426
421
Table 2 Effect of attenuation factor d on half-length and luminous output of 2 mm thick 120 ppm violet, 50 ppm pink and 60 ppm green sheets (without a luminaire) d (t=2 mm)
Violet 120 ppm
Green 60 ppm
Pink 50 ppm
L1/2 (m)
Lumens
L1/2 (m)
Lumens
L1/2 (m)
Lumens
0 5 104 1 103
2.78 1.51 0.96
109 95 80
1.50 0.77 0.56
1829 1315 1060
3.19 1.62 0.38
915 398 280
Measured values
1.74
80
1.04
1122
0.88
222
simulations using various values of d (see Table 2), d for each dye is estimated to fall within the range 1 104od o1 103. Each of these results were achieved by modelling the half-length with the tails region of the absorption spectra modified so that the shape remained the same, but the average value of d was either zero, 5 104 or 1 103. Each colour dye has a different wavelength range for residual attenuation, so these modifications were made over the appropriate range for each dye. For the violet dye, the tails region is l>420 nm, for green l>520 nm, for pink l>600 nm, and 800 nm is the maximum wavelength in each case. From these results it is evident that the light transport performance of the collectors are extremely sensitive to small changes in d. With d =1 103 the halflength of the pink dye is 12% of the value calculated for the same dye with d=0. The green and violet dyes show smaller reductions, at approximately 35% of the halflength with d=0. Likewise, d=1 103 produces luminous outputs ranging from 30% to 73% of the output with d=0, depending on the dye colour. This extreme sensitivity suggests that these fine details of the dye attenuation spectrum play a major role in determining the light transport efficiency and ultimately the light output of a LSC. From Fig. 8 it can be seen that d also has an effect on the optimum dye concentration. When d is increased, the luminous output and optimum dye concentration both decrease. Hence d is a dye-related constant, and it should be minimised for the best possible light output. 3.4. Luminous output and light-to-light efficiency Modelled and measured luminous outputs and light-to-light efficiencies for each colour LSC and the three-colour stack are shown in Table 3 (the cover sheet is used in all cases), as well as the effect of adding the luminaire to the stack. Here light-tolight efficiency is defined as the number of lumens measured at the luminaire divided by the number of lumens falling on the cover sheet from the sun. In each case the most accurate tails data available have been used for the simulations. In the simulations it was assumed that 50% of the rays reaching the end of the sheet were trapped inside the sheet in the absence of a luminaire. In practice this value is probably closer to 40%, so the modelled output may be slightly underestimated.
ARTICLE IN PRESS 422
A.A. Earp et al. / Solar Energy Materials & Solar Cells 84 (2004) 411–426
Fig. 8. Effect of d on concentration dependence of output of pink dyed sheet.
Table 3 Luminous outputs and light-to-light efficiencies of a three-colour stack and individual LSCs under solar illumination of 100,000 lx (with a UV-blocking cover sheet) LSC type
Measurements
Model
Lumens
Efficiency (%)
Lumens
Violet Green Pink
48 937 239
0.29 5.8 1.5
50 939 237
Stack (no luminaire) Stack (with luminaire)
783 995
4.8 6.1
1274 1656
Efficiency (%) 0.31 5.8 1.5 7.9 10.2
It is clear that most of the output lumens come from the green LSC, as it is best matched with the human eye response (see Fig. 6). In direct sun the green sheet gives approximately 940 lm, while the pink gives around 240 lm, and the violet gives around 50 lm. Thus green is the most important colour for luminous output, and the other colours play lesser roles, contributing more to the colour of the output light than the actual intensity. As discussed in Section 3.1, the special UV-blocking dye in the cover is required to increase the lifetime of the violet dye, but inevitably the violet light output is reduced. For the green and pink LSCs, an 11–13% reduction in luminous output is observed when the cover sheet is used, and this can mostly be attributed to Fresnel reflection losses. For the violet sheet, however, the equivalent reduction in lumens is three to four times more severe, due to the overlapping of the absorption spectra of the special UV-blocker and the violet dye. A blue dye with a higher peak wavelength would be less affected by this UV-blocker, so if the violet were replaced with blue, greater luminous output would be achievable.
ARTICLE IN PRESS A.A. Earp et al. / Solar Energy Materials & Solar Cells 84 (2004) 411–426
423
An alternative approach that is currently being investigated for future designs is to replace the violet LSC with a blue LED [29], which could be powered by a photovoltaic cell. Around 1 W would be required to supply the lumens that are currently contributed by the violet sheet. The pink and green outputs could be brighter without the losses from the violet. Using this system, any colour balancing is done by adjusting the brightness of the LED; so white light output can be achieved without sacrificing any of the lumens from the green and pink. As both the LED and the LSCs are driven by input from the sun, the stack output would be continually colour balanced, irrespective of the brightness of the sun. It can be seen from Table 3 that the luminaire successfully extracts some of the trapped light, although it does not achieve the estimated luminaire gain factor of 1.50. As discussed in Section 2.1, a luminaire should theoretically be able to extract around 80% of the trapped light, although 50% is a more realistic value. For the system presented in this paper, the luminaire was not well designed, so a luminaire gain of around 1.30 was achieved, and this value has been used in the model. Tests on simple sheets have shown that a luminaire gain factor of 1.50–1.60 is achievable with a well-designed luminaire. For the individual LSCs, the modelled results are in fairly good agreement with the measured values, however the modelled LSC stack output is overestimated by about 60%. This suggests that there are other significant sources of loss specific to the three-colour stack that are not fully accounted for in the model. Poor surface flatness and surface dirt may be causing more low angle light loss than expected. Some of the LSCs have up to 6% variation in thickness, and ray tracing simulations have shown that when joined to the PMMA light guide, this variation can cause up to 8% light loss. Further losses were caused by excess adhesive on the surface of the LSCs, residue from the protective sheets used by the manufacturers for shipping. Light can transmit into neighbouring sheets via the glue, causing large losses. This glue is very difficult to remove, and the attempt to clean the surface creates scratches, which cause further light loss. Hence for future LSCs, a different adhesive will be used for these protective sheets––one that does not leave any marks on the LSC surfaces. The above losses can be minimised by careful LSC production, ensuring that the top and bottom sheet surfaces are flat and parallel, clean and free from scratches. Despite all the losses, the LSC stack has quite a good optical performance. Under solar illumination of 100,000 lx an output of around 1000 lm is observed. Theoretically an output of around 1650 lm is achievable with the current luminaire, and around 1900 lm with a luminaire gain of 50%. The output lumens would decrease slightly in cloudy conditions, and if sunlight availability becomes very low, an artificial back-up lighting system will be required. Nonetheless, the LSC stack accepts both specular and diffuse light, making good use of all available sunlight. Although the performance of most daylighting systems is assessed by the lux levels produced, luminous output is given here in lumens, because our LSC system functions like a standard lamp fitting. It is conceived that residential room lighting is the most likely application of this system––e.g. bathrooms/kitchens and the like. For a room of known floor area, the number of LSC daylighting systems required can be
ARTICLE IN PRESS 424
A.A. Earp et al. / Solar Energy Materials & Solar Cells 84 (2004) 411–426
determined by calculating the number of lumens per square meter. For most rooms in an average house, one system per room should be sufficient. As with all daylighting systems, supplementary artificial lighting is required in highly overcast conditions. An LSC stack with high light-to-light efficiency ensures that optimal light output is achieved whatever the sky conditions. Light-to-light efficiency for an LSC is very size dependent, so it can only be compared between LSCs of similar sizes. For this particular stack it was anticipated that an efficiency of 5% would be sufficient for general household room lighting, and a value of 6% was achieved by the prototype. Our model predicts that this can be increased to 10% if residual glue and surface roughness losses are effectively minimised, and with a luminaire gain factor of 1.50 it should be possible to achieve a theoretical efficiency of 12%. Another outstanding feature of the LSC stack is its high luminous efficacy of 311 lm/W. However, this is partly due to the non-neutral slightly greenish colour of the output. If the cover sheet is removed, the output is a neutral white with a luminous efficacy of 294 lm/W. If the violet LSC were replaced with a blue LED, it is estimated that the luminous efficacy would drop to 250 lm/W, but the output would be brighter than the current system. These values of luminous efficacy by far exceed that of natural daylight, which generally falls within the range 100–130 lm/W [30]. When compared to standard artificial light sources such as incandescent light bulbs (16–40 lm/W) and fluorescent lamps (50–80 lm/W), it is clear that the output of the LSC stack is significantly more energy efficient. The above results show that with further improvements, this three-colour LSC daylighting system is capable of producing a high luminous output, with good lightto-light efficiency and luminous efficacy. These improvements are well within reach, and will be included in future prototypes.
4. Conclusions A daylighting system has been produced, which transports sunlight to remote areas of a building using a stack of pink, green and violet LSCs and clear PMMA light guides. In direct sun of intensity 100,000 lx, prototypes with collector area 1.2 m 0.135 m deliver 1000 lm of near-white light with a luminous efficacy of 311 lm/W and a light-to-light efficiency of 6%. Surface effects such as excess adhesive and variations in flatness are thought to be causing unnecessary light loss, which can be avoided by careful LSC production. It is proposed that the colour and intensity of the output light, and the efficiency of the system may be improved if the violet LSC is replaced with either a blue LSC or a blue solar-powered LED. In this case colour matching would be achieved by controlling the LED output, thus maintaining maximum light output from the green and pink LSC stack. A 30% increase in luminous output is observed by the installation of a single luminaire, although it is expected that the efficiency of the luminaire will be significantly improved by modifying the design, leading to a further increase in output. Thus it is demonstrated that the three-colour LSC stack is a
ARTICLE IN PRESS A.A. Earp et al. / Solar Energy Materials & Solar Cells 84 (2004) 411–426
425
suitable system for daylighting in remote areas of buildings, many metres away from the nearest available daylighting aperture. With the suggested improvements it is expected that the performance of the system may be improved significantly.
Acknowledgements The author would like to acknowledge the support of an Australian Postgraduate Award. The work was in part supported by BASF and Skydome industries.
References [1] [2] [3] [4] [5] [6] [8]
[9]
[10] [11] [12] [13]
[14] [15] [16] [17] [18] [19] [20] [21] [22]
[23] [24] [25]
J.C. Lam, D.H.W. Li, Solar Energy 65 (4) (1999) 251–262. C.A. Balaras, K. Droutsa, A.A. Argiriou, K. Wittchen, Energy Build. 34 (2) (2002) 135–153. D.H.W. Li, J.C. Lam, Energy Build. 33 (2001) 793–803. E. Ne’eman, Energy Build. 6 (2) (1984) 97–108. A.J.F. Rutten, Build. Environ. 29 (1) (1994) 105–111. C.L. Robbins, Daylight, Buildings and Design, Van Nostrand Reinhold, New York, 1986, pp. 3–13. R. Compagnon, J.L. Scartezzini, B. Paule, Application of nonimaging optics to the development of new daylighting systems, in: Proceedings of the ISES Solar World Congress 1993, Budapest, Hungary, 1993. J. Page, J. Kaempf, J.L. Scatrezzini, Assessing daylighiting performances of electrochromic glazings coupled to an anidolic device, in: Proceedings of the ISES Solar World Congress, Gothenburg, Sweden, 2003. I.R. Edmonds, Sol. Energy Mater. Sol. Cells 29 (1) (1993) 1–26. W. Pohl, C. Anslem, Natural room illumination using sunlight, in: Proceedings of the World Renewable Energy Congress VII (WREC 2002), Cologne, Germany, 2002. S. Chirarattananon, S. Chedsiri, L. Renshen, Solar Energy 69 (4) (2000) 331–341. A. Zastrow, V. Witter, Materials and optics for solar energy conversion and advanced lighting technology, Daylighting with mirror light pipes and with fluorescent planar concentrators, in: Proceedings of the SPIE, 1986, 227–234. A. Zastrow, V. Wittwer, Optical materials technology for energy efficiency and solar energy conversion SPIE 653 (1986) 93–100. A.M. Hermann, Solar Energy 29 (4) (1982) 323–329. R. Reisfeld, C.K. Jorgensen, Struct. Bond. 49 (1982) 1–36. W.H. Weber, J. Lambe, Appl. Opt. 15 (10) (1976) 2299–2300. R. Soti, E. Farkas, M. Hilbert, Z. Farkas, I. Ketskemety, J. Lumin. 68 (2–4) (1996) 105–114. G.B. Smith, J. Franklin, Sunlight collection and transfer system, US Patent 5,548,490, 1992. J. Franklin, Light systems for transmitting and releasing luminescent radiation, US Patent 6,272,265 B1, 2001. A. Goetzberger, V. Wittwer, Fluorescent Planar Concentrators for Solar Energy Conversion, F.Viewig, Braunschweig, 1979, pp. 427–451. V. Wittwer, A. Goetzberger, K. Heidler, Fluorescent planar concentrator (FPC) Monte Carlo computer model limit efficiency and latest experimental results, in: Proceedings of the Fourth E.C. Photovoltaic Solar Energy Conference, Stresa, Italy, 1982, 682–686. P.D. Swift, G.B. Smith, J. Franklin, Light to light efficiencies in luminescent solar concentrators, in: Proceedings of the SPIE Conference on Solar Optical Materials, Denver, CO, 1999, pp. 21–28. J.S. Batchelder, A.H. Zewail, T. Cole, Appl. Opt. 20 (21) (1981) 3733–3754. A.F. Mansour, Polymer Test. 17 (1998) 333–343.
ARTICLE IN PRESS 426
A.A. Earp et al. / Solar Energy Materials & Solar Cells 84 (2004) 411–426
[26] [27] [28] [29] [30]
W.R.L. Thomas, J.M. Drake, M.L. Lesiecki, Appl. Opt. 22 (21) (1983) 3440–3450. J.S. Batchelder, A.H. Zewail, T. Cole, Appl. Opt. 18 (18) (1979) 3090–3110. A.A. Earp, G.B. Smith, P.D. Swift, J. Franklin, Solar Energy 76 (6) (2004) 655–667. G.B. Smith, J. Franklin, A hybrid lighting system, US Patent pending, 2002. J.C. Lam, D.H.W. Li, Energy Convers. Manage. 37 (12) (1996) 1703–1711.