Nanocolloid simulators of luminescent solar concentrator photovoltaic windows

: Transparent luminescent solar concentrator ( LSC ) windows with edge - attached photovoltaic ( PV ) cells have the potential for improving building e ﬃ ciency without com promising aesthetics and comfort. Optimization of such windows requires an inexpensive simulator for experi menting with various designs. We report, for the ﬁ rst time to the best of our knowledge, the simulator of a transparent LSC window in the form of a plastic container ﬁ lled with a colloid of photoluminescent nanoparticles ( NPs ) in an organic solvent ( 1 - propanol ) . The exemplary NPs were pro duced by ball milling of the powder of rare earth ( RE )- doped phosphor NaYF 4 :Yb 3 + ,Er 3 + synthesized by the wet method. The NPs converted the ultraviolet ( UV ) solar spectrum into visible/near infrared ( NIR ) via spectral down - shifting and down - conversion ( quantum cutting ) . With a photolumines cence quantum yield ( PLQY ) of the phosphor < 0.4%, the LSC at a nanocolloid concentration of ∼ 0.1 g solids per 100 mL liquids demonstrated a power conversion e ﬃ ciency of 0.34% and a power concentration ratio of ∼ 0.022 comparable to the LSCs with RE - doped NPs with 200 times greater PLQY. At the same time, the 3cm thick LSC window simulator had ∼ 90% transmittance to the sunlight. The content and concentration of the nanocolloid could be easily modi ﬁ ed to optimize the LSC window performance without a costly window making process.


Introduction
Luminescent solar concentrators (LSCs) operate by the principle of collecting solar radiation over a large area, converting it by luminescence, and directing the generated radiation into a relatively small output target particularly for producing photovoltaic (PV) electricity [1]. The idea of using windows as LSCs with edge-attached PV cells for harvesting solar power ( Figure 1) and thus improving building efficiency without compromising aesthetics [2] has got traction and continues to be the focus of the research community [3][4][5][6][7]. The usefulness can be broadened beyond buildings to, for instance, cell phones and any other display devices that have a clear surface [8]. The extra weight of such LSC displays would not exceed a few grams. The application to the windows in the aerospace field can be also envisioned, but researchers would have to address issues like ruggedness and cost reduction. The LSC window must either have a thin (from several microns to a few millimeters) spectrum converting layer or embedded centers (atoms, molecules, or nanoparticles (NPs)) that capture solar UV photons with high energy hν 1 and re-emit visible or near-infrared (NIR) photons with less energy hν 2 < hν 1 that match the spectral responsivity of commercial silicon-based PV cells (Figure 1). The window will act as a lightguide directing visible/NIR photons to the edges with the PV cells attached. The PV cells will convert the energy of these photons into a photoelectric current. The energy of the otherwise wasted UV component of solar radiation will be harvested and converted into electricity, thus improving the energy efficiency of the building. LSC windows must satisfy contradictory requirements, namely, be clear, colorless, and no-glow to provide maximum comfort to the inhabitants, and simultaneously, convert as much UV light into photoelectricity as possible [4]. The key issue of the LSC window development process is to find the proper spectral converting material that meets the requirements. In search of such materials, it will be helpful to use an inexpensive and simple simulation platform for rapid material testing without making actual windows. The purpose of the simulator is to reduce the parameters for the experimentalists so that fewer experiments are needed subsequently. This article describes such a simulator in the form of a transparent rectangular vessel filled with a liquid that carries the spectral converting centers of the material of interest. As an example, the nanoparticles (NPs) of a rare-earth (RE)-doped compound are considered. Such materials have been intensely studied as optical spectrum converters (both up and down) by the authors of this article in the past [9][10][11][12][13][14][15][16][17][18][19][20][21][22][23][24][25][26][27]. To the best of the authors' knowledge, LSC simulators have been implemented computationally using the ray tracing method [28]. The existing experimental LSC prototypes to be mentioned below could not serve as cost-effective simulators due to their irreversible design and rigidness. In the next section, the materials and methods will be described. The sections on the obtained results, their discussion, and conclusions will follow.
The phosphor was synthesized using the wet method followed by baking the obtained micro-powder in ambient air and reducing it with ball-milling to obtain the nanopowder [9][10][11][12][13][14][15][16][17][18][19][20][21][22][23][24][25][26][27]. In a typical procedure, 2.1 g of NaF (0.05 mol) were dissolved in 60 ml of deionized water. Another solution was prepared by mixing a × 16 mL of 0.2 mol/L aqueous solution of YCl 3 , b × 16 mL of 0.2 mol/L solution of YbCl 3 , c × 16 mL (c = 1 − (a + b)) of 0.2 mol/L solution of ErCl 3 , and 20 mL of 0.2 mol/L EDTA aqueous stock solution to form the metal-EDTA complex. All the chemicals were acquired from Sigma-Aldrich. The complex solution was injected into the  NaF solution quickly, and the mixture was stirred vigorously for 1 h at room temperature. After stirring, the mixture was allowed to stay overnight for the precipitate to settle. The precipitate was collected and washed several times with distilled water and anhydrous ethanol. In the second stage of the process, the precipitate was dried in the open air for 48 h at 60°C to remove traces of water. The resulting powder had the doping rate according to the formula NaY F : (sample D). The freshly made crystalline powder had NaYF 4 host in a cubic α-phase. It was converted into a hexagonal β-phase more suitable for the spectrum conversion function using heat treatment in an open-air furnace in the temperature range of 400-600°C for 1 h. The temperature range and duration of heating were found to be optimal for the crystalline phase change [9][10][11][12][13][14][15][16][17][18][19][20][21][22][23][24][25][26][27]. After that, the phosphor powder was reduced to nanopowder by ball-milling in water using a PQ-NO4 Planetary Ball Mill from Across International [9][10][11][12][13][14][15][16][17][18][19][20][21][22][23][24][25][26][27]. The next step was to extract the NPs from water and dry them before transferring them to an organic solvent. Two approaches were used. In the first method, the water colloid together with zirconia balls was extracted from zirconia cups of the ball mill and sonicated in an ultrasonic bath for 60 min. The water colloid was separated from the balls and allowed to settle for 24 h. The top, unsettled part of the colloid was collected with a syringe and filtered through a 1 µm PTFE filter. The filtered water colloid was poured into an open Petri dish and dried on a hot plate at ∼80°C to eliminate water. The obtained dry crust was scraped from the bottom of the Petri dish. The resulting dry powder consisted of flakes with an average size of the order of 1,000 µm ( Figure 3). In the second method, a drop of oleic acid was added to the water colloid before it was sonicated together with the balls in an ultrasonic bath for 60 min. The goal was to check if oleic acid helped to prevent clustering NPs into bigger particles after drying. The water colloid was separated from the balls and allowed to settle for 24 h. The top, unsettled part of the colloid was again collected with a syringe and filtered through a 1 µm PTFE filter. The filtered colloid was poured into a Petri dish and dried on a hot plate at ∼80°C. After scraping, the dry powder consisted of flakes with 20 times smaller average size than in the case with no oleic acid addedof the order of 50 µm ( Figure 4).
Adding oleic acid to the water colloid of the ball-milled powder significantly reduced the size of the obtained flakes after drying and scraping. To find out if this affected the size of the resulting crystalline NPs aggregated in the flakes, X-ray diffraction (XRD) spectroscopy was conducted with a Bruker D2 Phaser X-ray diffractometer on three powder samples: the dry powder before ball milling; the ball-milled powder with big flakes (no oleic acid added), and the ballmilled powder with small flakes (oleic acid added). The results are presented in Figure 5. As one can see, the XRD spectra of both ball-milled powders, with oleic acid (spectrum 3) and without oleic acid (spectrum 4), had the same width of diffraction peaks, and the peaks were broader than those of the powder before ball milling (spectrum 2). In all three cases, the peaks and their positions corresponded to the hexagonal β-phase of NaYF 4 (plot 1 in Figure 5). Since the peak broadening is inversely proportional to the size of crystals in the powders [30], it could be assumed that the size of the monocrystalline NPs in both ball-milled powders (with small and big flakes) was approximately the same. Adding oleic acid just prevented assembling the NPs into bigger flakes. This assumption was confirmed when both ball-milled powders were dispersed in 1-propanol. The size distribution of NPs was measured using the dynamic light scattering (DLS) method. The 10 mm fluorometric cuvettes filled with both nanocolloids in 1-propanol were put in a Zetasizer DLS instrument (Malvern Instruments). The measurement results in the form of size distributions (assuming the spherical shape of the NPs) are presented in Figure 6. The size of NPs made from the large-flake powder (described by distribution curve 1 in Figure 6) was determined as 172.6 ± 65.06 nm. The size of the NPs from the small-flake powder (described by distribution curve 2) was 228.0 ± 88.7 nm. Size distributions in both cases overlap, and the difference between average sizes of NPs is less than the spread of the distributions.

Optical properties of the spectral converting phosphor
Diffuse reflectance spectra of the synthesized phosphor micropowders with various concentrations of RE ions ( Figure 7) were recorded with a Shimadzu UV-2600-ISR-2600 Plus UV-VIS-NIR spectrophotometer with an integrating sphere (measurable spectral range: 220-1,400 nm). The samples were prepared by compressing the micropowders    into pellets with a 25-T manual hydraulic press. Following Aarts et al. [31], the absorption peaks in the diffuse reflectance spectra were related to the transitions from the ground states to various optically excited states of Er 3+ and Yb 3+ ions as presented in Figure 7. The strongest peak 6 corresponded to the optical absorbance of the Yb 3+ ion due to the transition 2 F 7/2 → 2 F 5/2 . The ion was responsible for the down-conversion NIR emission when relaxing back from excited to the ground state. The ratio of the strength of peak 6 to peak 3 (520 nm) of Er 3+ was approximately 1:1, 10:1, and 30:1 in phosphor A (NaY 0.94 F 4 :Yb 3+ 0.03 Er 3+ 0.03 ), B (NaY 0.78 F 4 :Yb 3+ 0.2 Er 3+ 0.02 ), and C (NaY 0.69 F 4 :Yb 3+ 0.3 Er 3+ 0.01 ), respectively. This was roughly proportional to the ratio between the molar concentrations of Yb 3+ to Er 3+ in the samples. This correlated well with findings in ref. [31].
Photoluminescence spectroscopy of phosphor powders was conducted using excitation with a 372 nm UV diode laser LBX-375-70-CSB-PPA from Oxxius and 488 nm spectral line of an Ar-ion laser. The spectra were taken with a spectrometer AvaSpec-ULS4096CL-EVO with integrating sphere AvaSphere-30 from Avantes. Figure 8 presents the spectra of the powder of phosphor NaY 0.83 F 4 :Yb 3+ 0.14 , Er 3+ 0.03 (sample D), the most efficient among the synthesized ones in terms of photoluminescence. The spectra were adjusted to correspond to the same energy received from both excitation lasers. The peaks corresponded to the same absorption bands as in Figure 7. The new peak 5′ corresponded to the band at 850 nm ( Figure 2). This band did not exhibit prominent absorption in Figure 7. Emission bands around peaks 1-5′ could be attributed to the spectral down-shifting transitions of the excited ion of Er 3+ (color vertical arrows down in Figure 2). Band 6 could be associated with the down-shifting radiation from Er 3+ at ∼980 nm ( Figure 2) and spectral down-conversion (quantum cutting) radiation of Yb 3+ at ∼1000 nm. The photoluminescence quantum yield (PLQY) of the phosphor defined as the ratio of the number N Out of the produced output photons to the number N Pump of pump photons PLQY = (N Out /N Pump ) × 100% obtained using the integrating sphere method [32,33] is presented in Table 1 together with PLQY of other spectrum converting materials. One can see that the phosphor had PLQY (∼4% combined for all bands, see Table 1, column 3) three orders of magnitude less than the recently reported, highly efficient spectrum converters. They belong to the group of Ybdoped perovskites CsPbX 3 :Yb 3+ (X= Br or Cl). The reason for low PLQY is that typical RE 3+ ion, such as Er 3+ , exhibit 4f-4f transitions with well-defined energy levels that are nearly invariable in different hosts due to the shielding of 4f orbitals and poor light absorption coefficient (∼1-10 M −1 cm −1 ). Various light-absorbing sensitizers (such as perovskites) can be used to nonradiatively excite the f electrons of RE 3+ ions and boost PLQY [37]. However, these sensitizers themselves can emit strong visible radiation that can be re-absorbed (dropping the efficiency) and make LSC windows colored and strongly Absorption peak 1 corresponds to the transition in ion of Er 3+ : 4 I 15/2 → 4 G 11/2 (380 nm); 2 -Er 3+ : 4 I 15/2 → 4 F 7/2 (480 nm); 3 -Er 3+ : 4 I 15/2 → 2 H 11/2 (520 nm); 4 -Er 3+ : 4 I 15/2 → 4 S 3/2 (540 nm); 5 -Er 3+ : 4 I 15/2 → 4 F 9/2 (650 nm); 6 -Er 3+ : 4 I 15/2 → 4 I 11/2 and Yb 3+ : 2 F 7/2 → 2 F 5/2 (∼980-1000 nm). All the spectra are normalized to peak 3: its depth is (−1) in arbitrary units. fluorescent in the visible spectrum and thus uncomfortable for building residents. Therefore, it still makes sense to investigate RE-doped compounds with low PLQY in LSC simulators. The final judgment on the suitability of a particular material could be made based on the overall LSC performance.

LSC simulator
To simulate LSCs, a rectangular Plexiglass container was filled with the RE phosphor nanocolloid in 1-propanol as shown in Figure 9. This solvent has a low evaporation rate and low absorption in UV-VIS-NIR. The concentration of the nanocolloid was 0.193 g solids per 100 mL liquids. The container had dimensions 65 mm × 65 mm × 42 mm (2-9/16 × 2-9/16-1 × 15/16 inch). Four AOSHIKE polycrystalline silicon PV cells (5-voltage; current, 30-mA current; dimensions, 53 mm × 30 mm) were attached to the container sides. The PV cells were connected in parallel. The container was filled to the level of the PV cells. Figure 9b shows the near-UV (488 nm) laser beam passing through the nanocolloid in the container without the attached PV cells. In Figure 9c, the beam is visible through a red filter (laser protective goggles) thus proving that the nanocolloid converted the near-UV light into visible re-emitted by the phosphor NPs. Figure 10 shows the block diagram of the experimental setup to investigate the spectrum emitted by the nanocolloid in the LSC toward the edge-attached PV cells. Solar simulator LCS-100 model 94011A-ES from MKS-Newport with Xe-lamp and Air Mass (AM) 1.5G filter was used as a light source. The shutter of the simulator was controlled with a shutter controller 71445 from MKS-Newport. The simulator produced an evenly illuminated square spot on LSC surface with dimensions ∼65 mm × 65 mm. A   household white 40 W incandescent light bulb was also used as an alternative light source with no UV light in the spectrum. A multimode optical fiber terminated with focusing optics was mounted near the edge of the LSC without PV cells attached. The other end of the fiber was connected to the spectrometer AvaSpec-ULS4096CL-EVO.

Experimental setup
The experimental setup for measuring I-V characteristics of the LSC is presented in Figure 11. The setup used the same light sources as in the previous case. There was also a set of PV cells placed under the LSC to measure the light passing through it. The set consisted of six AOSHIKE PV cells connected in parallel and was covered with a mask with 70 mm × 70 mm window matching horizontal dimensions of the LSC. During the measurements of the light passing through the LSC, its PV cells were disconnected from the measurement circuit and replaced by the bottom PV set (the wires are marked by blue dashed lines in Figure 11). The measurement of I-V characteristics was conducted automatically under the computer control with running LabVIEW software developed for MKS-Newport I-V Test Station PVIV-1A [39]. The voltage was automatically applied to the LSC, and the electric current was measured by a computer-controlled Keithley 2400 Source and Measurement Unit. The Unit was connected to the computer through the GPIB-USB interface. The elevation of the solar simulator over the LSC was adjusted to provide the radiation intensity on the surface equal to that of the sun (1.0 Sun) equivalent to a 100 mW/cm 2 power density. The 40 W incandescent lamp was mounted at an elevation of 25 cm over the surface.

Results and discussion
3.1 Solar spectrum conversion in the LSC nanocolloid Figure 12 presents the spectra of the light emanating from the edges of the LSC illuminated by the solar simulator and the light bulb. Spectra 1 are of the radiation directly from the light sources at grazing incidence with respect to the walls of the empty LSC container. Spectra 2 are of the radiation from the LSC filled with pure 1-propanol. Spectra 3 were produced by the LSC filled with the photoluminescent nanocolloid. One can see that spectra 2 are lower than spectra 1. This was due to the total internal reflection of the light on the interface between the side walls of the filled LSC container and the air gap between the wall and the PV cell. The gap was not filled with an index matching substance to minimize the total internal reflection. But spectrum 3 is higher than spectra 1 and 2 approximately 18 and 43 times, respectively, in the case of the solar simulator (Figure 12a). In the case of the light bulb, spectrum 3 is higher than spectra 1 and 2 roughly 3 and 8 times, respectively (Figure 12b). The results indicate that the nanocolloid scattered a significant portion of the incident light toward the edges. This can be also seen in the photographs in Figure 9b and c. The scattering is strong for the light, which has a wavelength λ < πd, where d is the average size of NPs, and beyond that limit rapidly falls to zero [38]. Since d = 228 nm (Figure 6), the light was scattered when λ < 717 nm, on the left edge of the responsivity spectrum of the PV cell (spectrum 4 in Figure 12). To compare the shapes of the spectra in Figure 12 regardless of the intensity, they were normalized as shown in Figure 13. The spectrum of the radiation scattered by the nanocolloid towards the edges when the LSC was illuminated with the solar simulator (spectrum 3, Figure 13a) has less degradation of its long-wavelength part than in the case of the light bulb (spectrum 3, Figure 13b). The latter spectrum has its "red" portion diminished by half. This occurred because the light bulb spectrum was more concentrated in the region beyond 715 nm and thus less subject to the scattering towards the edges. The conversion of the UV solar spectrum into additional visible and NIR light by NPs could also contribute to spectrum 3 of the LSC illuminated with the solar simulator (Figure 13a). This light (of red color) can be seen in the photograph in Figure 9c.

I-V characteristics of LSC PV cells
The I-V characteristics of the PV cells attached to the LSC edges (will be called below as I-V characteristics of the LSC) were measured to evaluate the PV performance. The measurements were conducted using the LabVIEW software for I-V Test Station PVIV-1A from MKS-Newport [39, p. 14]. The error of the measurements was of the order of 0.1%. The results are presented in Figures 14, 15, and Table 2. The data on the generated PV power in Table 2 were computed as areas under the I-V curves [40] using the same LabVIEW program. Figure 14a shows the I-V characteristics of the LSC illuminated with the solar simulator. One can see that filling LSC with pure 1-propanol caused a significant drop of the photocurrent (curve 2) as compared to the empty LSC (curve 1). The PV power generated by the LSC decreased by 89.7% ( Table 2, data row 2, column 3). This was due to the total internal reflection on the interface between the side walls of the filled LSC container and the air gap described in the previous subsection. (b) with the light bulb. Spectra 1 correspond to empty LSC (true spectra of the light sources), 2 -LSC filled with pure 1-propanol, 3 -LSC filled with the NaY 0.83 F 4 :Yb 0.14 Er 0.03 nanocolloid in 1-propanol, 4normalized responsivity of the PV cells. In (a) the spectra were normalized to the intensity of the peak at 398.3 nm. In (b) the spectra were normalized to the intensity of the peak at 565.8 nm. photocurrent (curve 3), and the PV power jumped up by 99.2% ( Table 2, row 2, column 5) relative to the empty LSC. The results correlate with the relationship between the intensity of the spectra in Figure 12a. They can be explained by strong scattering of the "blue" portion (with λ < 715 nm) of the incident solar light toward the edges and by the conversion of solar UV to visible and NIR radiation that matched the spectral response (spectrum 4 in Figures 12 and 13) of the edge-attached PV cells (Figure 1). Figure 14b shows the I-V characteristics of the LSC illuminated with the light bulb. Filling LSC with pure 1-propanol again caused the drop of the photocurrent (curve 2) as compared to empty LSC (curve 1) due to the total internal reflection. The PV power decreased by 85.1% ( Table 2, data row 6, column 3). Filling LSC with the nanocolloid just slightly increased the photocurrent (curve 3) due to the scattering of some visible/NIR light toward the edges. But the power remained less than that of the empty LSC by 63.0% ( Table 2, row 6, column 5). Despite spectrum 3 being generally higher than spectrum 1 in Figure 12b, its "red" portion (near the maximum of spectrum 4 of the PV cell responsivity) was relatively low (as Figure 13b also shows)  due to the scattering effects described in the previous subsection. There was also no UV radiation in the spectrum of the light bulb that the NPs might convert to extra visible/ NIR. The nanocolloid-filled LSC illuminated with the light bulb thus produced less PV power than the empty LSC. Figure 15a presents the I-V characteristics of the PV set (six masked PV cells) placed under the LSC illuminated with the solar simulator. Filling LSC with pure 1-propanol caused the drop of the photocurrent (curve 2) as compared to the empty LSC (curve 1). The PV power dropped by 12.8% ( Table 2, data row 4, column 3), and was probably due to the internal reflection of the light on the interface between the bottom of the LSC filled with propanol and the air gap between the LSC and the PV set under it. This power drop and, correspondingly, the loss of light was seven times less than in the LSC itself (89.7%, Table 2, row 2, column 3). Filling LSC with the nanocolloid slightly increased the photocurrent (curve 3) compared to the case of the LSC filled with pure 1-propanol (curve 2). But the photocurrent was still below the level corresponding to the empty LSC (curve 1). The power was still 10.6% less than in the case of empty LSC ( Table 2, row 4, column 5). Apparently, the nanocolloid reduced the losses of the light passing through the LSC more likely due to scattering and conversion of the UV solar spectrum into additional visible and NIR radiation suitable for the PV. Figure 15b shows the I-V characteristics of the PV set under the LSC illuminated with the light bulb. Filling LSC with pure 1-propanol again caused the drop of the photocurrent (curve 2) as compared to empty LSC (curve 1) due to the total internal reflection. The power decreased by 10.3% ( Table 2, data row 8, column 3). Filling LSC with the nanocolloid reduced the photocurrent even more (curve 3), and the power was dropped by 12.8% as compared to empty LSC ( Table 2, data row 8, column 5). This could be due to additional scattering of the light passing through the nanocolloid (with the spectrum modification described in the previous subsection) with no UV radiation converted to visible and NIR.
The last column of Table 2 summarizes the power measurement results. In the case of the solar simulator, the nanocolloid improved the PV power generated by the LSC by 1825.3% as compared to the LSC filled with pure 1-propanol. As measures of significant improvement 99.2 and 1825% are highlighted with bold in Table 2. At the same time, the nanocolloid brought the increase of the power measured by the bottom PV set by 3.2%. For the incandescent light bulb as an illuminator, the nanocolloid improved the power as compared to the LSC filled with pure propanol by 155.0%. But the bottom PV set showed a drop of power by 1.8%.
Based on the data in Table 2, the power conversion efficiency and related parameters of the LSC were estimated. The power conversion efficiency η PCE was defined as the ratio of the output power generated by the edgeattached PV cells of the LSC P out (14.44 mW, Table 2, data row 2, column 4) to the incident, input power P in : η PCE = (P out /P in ) × 100% [41]. For the solar simulator at 1.0 Sun (100 mW/cm 2 power density) and an illuminated spot size of 65 mm × 65 mm, the incident power was 100 mW/cm 2 × 6.5 cm × 6.5 cm = 4225 mW. The power conversion efficiency was thus 14.44 mW 4,225 mW 100% 0.34%.

PCE
The power concentration ratio C (also called "flux gain" in ref. [42]) was defined as the ratio between the incoming and outgoing power density (radiance) [41] where G was the geometric gain of the LSC defined as the ratio of the LSC area and the area of the attached PV cells; η PV was the power conversion efficiency of the PV cells defined as the ratio between the electric power generated by the cell and the power of the incident light. The geometric gain G of the LSC simulator was G = 42.25 cm 2 / 63.6 cm 2 = 0.66. The power conversion efficiency of the AOSHIKE silicon cells (for solar light) was measured to be ∼10%. Correspondingly, the power concentration ratio was estimated as C = 0.022. The power conversion efficiency of the presented LSC (η PCE = 0.34%) was 4.2 times less than that of the LSC reported in ref. [41] (η PCE = 1.44%). where the authors used microparticles (11 µm average size) of RE-doped photoluminescent compound CaAlSiN 3 :Eu 2+ with two orders of magnitude higher PLQY (83%). The power concentration ratio (C = 0.022) was ∼56 times less than that of the LSC in ref. [41] (C = 1.23). It should be also considered that the concentration of the NPs in the presented LSC simulator (∼0.1 g solids per 100 mL of the host) was six times less than the concentration of phosphor microparticles in ref. [41] (0.6 g per 100 mL of polymethyl(methacrylate) or PMMA host).
Because of the small size of the NPs (228 nm instead of 11 µm) and their low concentration, the 30 mm thick LSC simulator performed well as a transparent window transmitting ∼90% of sunlight (10.6% drop of passing light powersee Table 2, row 4, column 5) while 0.5 cm thick LSC in ref. [41] blocked almost all the incident sunlight (Figures 2 and 3 in ref. [41]). The power conversion efficiency and power concentration ratio were also estimated in the case of the light bulb with no UV light in the spectrum. The total input power coming to the LSC from the light bulb was estimated as = The ratio C (Light bulb) was estimated as = ( )

Light bulb PV
Both performance parameters are 2.56 times greater than those of the nanocolloid LSC with the solar simulator. This was due to a better match between the light bulb emission spectrum and the responsivity spectrum of the PV cells (spectrum 4 in Figures 12 and 13).

Conclusion
The power of UV radiation is ∼8 and ∼6% of the total power of sunlight outside the atmosphere and on the sea level, respectively [43]. The latter percentage is the upper limit for the power to be extracted from the sunlight by a transparent LSC window converting solar UV spectrum down to visible/NIR. The share of the output radiation power at the edges of the LSC simulator to the incident sunlight power can be estimated as [P out /(P in η PV )] × 100% = (C/G) × 100% = (0.022/0.66) × 100% = 3.3%. This value is more than half of the maximum share that can be extracted from UV sunlight by converting it down to visible/NIR. Considering low PLQY of the photoluminescent phosphor, most of the visible/NIR light came to the edges due to selective (for wavelengths < 715 nm) scattering on the NPs, but not due to down-shifting or down-conversion (quantum cutting).
A clear plastic container filled with a nanocolloid converting solar UV spectrum into visible/NIR radiation can work as a flexible experimental simulator of transparent LSC windows. Various types of spectrum converting NPs and PV cells can be investigated with it in the lab environment. Two "clever" solutions were made during the feasibility study of the simulator. The first was the selection of 1-propanol as a solvent. It is polar and easily disperses NPs; it has a low evaporation rate and stays longer in the open container; it etches insignificantly the walls of the Plexiglass container; and it has weak optical absorption in the range from UV to NIR. The second clever solution was the selection of an incandescent light bulb with no significant UV spectrum as a reference light source for sunlight (simulated by a Xe-lamp with AM 1.5G filter). Using the light bulb led to a surprise  result that the proposed LSC simulator could be used to investigate visible sunlight reaching the edges due to scattering on the NPs separately from visible/NIR radiation re-emitted by NPs due to the sunlight spectrum down-shifting and down-conversion. At a nanocolloid concentration of ∼0.1 g solids per 100 mL liquids, the 3 cm thick LSC simulator had a power conversion efficiency of 0.34% and a power concentration ratio of 0.022 while transmitting ∼90% of sunlight. The results point to the feasibility of massive use of transparent LSC windows where the energy of the otherwise wasted (such as UV) light can be harvested while the visible light is comfortably transmitted and utilized by humans.