Photovoltaic (PV) energy harvesting is not always cost-effective, as the efficiency of conventional panels decreases significantly when they are not directly illuminated and optimally oriented toward the south. To address these limitations, a variety of strategies have been developed to enhance PV energy harvesting or to reduce associated costs. In cities, where areas with direct illumination are limited but diffuse light is abundant, luminescent solar concentrators (LSCs) offer a promising alternative.

LSCs are devices made of high-clarity polymer or glass plates doped with fluorescent systems that can harvest part of the solar radiation and channel it toward their edges, thus achieving a concentration effect (Castelletto and Boretti 2023). This is because the plate acts as a waveguide by means of total internal reflection for the emitted radiation, effectively trapping a large fraction of it. Thin PV cells placed at the edges can convert the concentrated radiation to electric energy. In addition, as fluorescence emission occurs at longer wavelengths those absorbed by the system, LSC technology can enable a more effective use of the UV and blue portions of the solar spectrum by shifting the photons toward lower energies (downshifting), where the external quantum efficiency (EQE) of silicon cells is higher (> 450 nm) (Picchi et al. 2023). Notably, compared to standard PV modules, LSCs maintain nearly constant performance when transitioning from direct to diffuse light, ensuring more stable and predictable energy production. Owing to their good transparency in the visible range, LSCs can be integrated into façades, rooftops, or even windows without compromising aesthetics or daylight transmission.

The major loss mechanism of an LSC arises from transmitted solar radiation. Typically, bulk LSCs based on organic dyes absorb no more than 25% of the incident photons, concentrated within the absorption range of the dye. In particular, nearly half of the solar power lies in the infrared region (λ > 700 nm), where most organic dyes are transparent, further contributing to transmission losses. As a result, a significant fraction of sunlight passes through the slab unused, either because its wavelength lies outside the absorption spectrum of the fluorophores or it cannot be converted by the attached PV cells. For instance, Lumogen F Red 305 (LR305), the benchmark fluorophore for LSC applications, displays an absorption spectrum in polymethylmethacrylate (PMMA) extending up to 640 nm. An integration of the AM 1.5G spectrum reveals that, after accounting for reflection losses (≈ 4%) and neglecting host absorption, about 60% of incident photons are transmitted by the sheet. Even under the idealised scenario of absorbing all photons below 900 nm, transmission losses would still amount to approximately 25%.

To increase photon harvesting, mixtures of dyes have often been used (Hermann 1982; Burgers et al. 2006; Bailey et al. 2007; Lyu et al. 2019). Combining donor-acceptor pairs allows for extended absorption, with energy transfer processes, such as Förster-resonance energy transfer (FRET), enabling donor-to-acceptor excitation. While this approach broadens the absorption bandwidth, the overall efficiency strongly depends on the quantum yield (QY) of the emitters: introducing low-QY dyes can reduce device performance despite the extended spectral coverage (Gryczynski et al. 2005; Richards and McIntosh 2006; Zhang et al. 2022).

An alternative strategy is stacking multiple LSC layers with complementary fluorophores separated by air gaps (Goetzberger and Greube 1977). In this design, sunlight first encounters dyes with shorter emission wavelengths, while subsequent layers absorb both the non-absorbed photons and the down-shifted emission lost through the escape cone. This system prevents low- dyes from degrading the photon transport within high- layers (Wilson 2010). Early studies already demonstrated efficiency gains using this concept, such as an increase from 10 to 15.8% (Stahl et al. 1986) and more recently multilayer thin-film system achieving 10–14% higher efficiencies than single-dye LSC (Carlotti et al. 2016). Beyond organic dyes, stacked LSCs incorporating down-converting and quantum-cutting materials such as quantum dots (QDs) or ytterbium-doped perovskites have also shown promising results, including near-200% through quantum cutting (Cohen et al. 2019; Richards and Howard 2023). Recent work further confirmed the potential of QD-based multilayer devices, reporting = 9.1% for a three-layer system, optimisable up to 13% by adjusting QD concentrations (Lin et al. 2024).

This paper explores a multilayer LSC system based on two organic fluorophores, DQ1 and BDT-H2, whose synthesis was previously described in the literature (Papucci et al. 2021; Bartolini et al. 2023). These materials were selected for three reasons: a) their reasonably large in thin-film LSCs, reaching 6.2% and 8.4%, respectively; b) their high in PMMA at low concentrations, making them suitable for bulk systems; c) the spectral alignment between the emission peak of DQ1 (546 nm in PMMA), and the absorption maximum of BDT-H2 in the same polymer (530 nm in PMMA); d) the low self-absorption of these dyes, which makes them appealing for the application in larger area devices. In a stacked design, placing the DQ1 layer on top enables its emission within the escape cone to be reabsorbed by the BDT-H2 below, potentially attaining higher efficiencies.

Finally, we introduce an additional element of innovation and sustainability: the use of chemically regenerated methyl methacrylate (r-MMA) as the monomer to produce acrylic sheets. This approach not only ensures high optical quality of the LSC sheets but also contributes to circularity in material use, enabling more sustainable large-scale production of LSC devices. Indeed, the use of r-MMA reduces the global warming potential of the LSC production by approximately 75% compared to using synthetic methyl methacrylate (Picchi et al. 2024a).

2,9-Bis(2,6-diisopropylphenyl)-4,7,11,14-tetraphenoxyanthra[2,1,9-def:6,5,10-d'e'f']diisoquinoline-1,3,8,10(2H,9H)-tetraone (LR305) was purchased from abcr GmbH, Germany. DIAKON poly(methyl methacrylate) (Mw ≈ 95 kg/mol, Lucite International, The Netherlands), chemically regenerated methyl methacrylate (r-MMA, Madreperla, Italy) and synthetic methyl methacrylate (MMA, 3ci, Italy) were provided by I&S srl, Florence, Italy. 4,4'-(2,3-Diphenylquinoxaline-5,8-diyl)bis(N,N-diphenylaniline) (DQ1) and 2,6-bis(4-(diphenylamino)phenyl)-4,8-bis((2-ethylhexyl)oxy)benzo[1,2-b:4,5-b']dithiophene 1,1,5,5-tetraoxide (BDT-H2) were synthesized elsewhere (Papucci et al. 2021; Bartolini et al. 2023).

The fabrication of r-PMMA slabs was conducted via cell casting in accordance with the following general procedure: r-PMMA slabs were obtained by pouring a solution comprising 80 wt% r-MMA, DIAKON poly(methyl methacrylate) (20 wt%), the fluorophores, and 0.1 wt% AIBN into 15 × 15 × 0.3 cm³ moulds. LR305, DQ1 and BDT-H2 were added as the fluorophores at the concentrations of 200, 300, 400 ppm (12, 18, and 24 mg, and 9, 13.5, and 18 mg for 3- and 2-mm thick slabs, respectively). A total resin weight of 60 g was prepared for pouring, slightly exceeding the required amount. The moulds were assembled using two glass sheets separated by an elastic poly(vinyl chloride) gasket and secured with metallic clamps. Different gaskets were employed to obtain different final thicknesses of the slabs. The sealed mould was placed in a water-filled tank overnight at a temperature of 50°C, and it was then transferred to an oven for a curing of 2 h at 90°C followed by 2 h at 120°C. At the end of the curing process, the two glass sheets were separated, the PMMA slab was cut to size by laser cutting, and polished.

UV-vis absorption and transmission spectra were acquired at room temperature with a Cary 5000 Series UV-vis NIR spectrophotometer (Agilent Technologies, USA).

Emission spectra were acquired using a Jobin-Yvon Fluorolog-3 spectrofluorometer (Horiba, Japan) with an integration time for the analysis of 0.1 s. The selected excitation wavelengths were close to the absorption peak at longer wavelengths. Solid samples were rotated 30° with respect to the excitation beam, and the detector was set in right-angle (RA) mode.

Quantum yield measurements were carried out using a 152 mm Quanta-ϕ F-3029 integrating sphere (Horiba, Japan) equipped with a solid sample holder and connected to the Fluorolog-3 spectrofluorometer by optical fibres and a FL-3000 fibre-optics adaptor (Horiba, Japan). The excitation was set at low wavelengths to allow detection of emission peaks with full extension. The detector was set in right-angle (RA) mode. For slab samples, a fragment was ground using an MF10 grinder (IKA, Germany). Sufficient powder was then transferred to cover the bottom layer of the solid sample holder. The reported 𝑄𝑌 values were calculated as the average of three values calculated through the Fluoressence software and derived from distinct measurements on powder samples. No reabsorption correction was applied to the recorded spectra.

Recently published laboratory protocols and consensus statements were followed to determine the performance of the LSC (Debije et al. 2021; Yang et al. 2022). More details are included in the Supporting Information section.

Referring to ASTM G154 standard (ASTM Standard F1980-02 2006), a home-made setup was used for the photo-stability experiments, which was composed of a LED tower (Cicci research s.r.l., Grosseto, Italy) as light source and an integrating sphere connected to a CMOS-based spectrometer (CCARK.A.4 Spectroradiometer, Fiber Optic VIS/NIR spectrometer, 2048 pixels, grating VA 360-1100nm, slit-50, OSC, DCL- UV/VIS) as the detector. The sample was placed on a controllable hot stage to adjust its temperature during the experiment. The hot stage was set at 80°C, thereby ensuring a constant temperature of 70°C on the top surface of the sample. The experimental temperature (70 ± 1°C) was checked using a K-Type thermocouple thermometer in contact with the top surface of the sample. The end of the optical fibre for photon detection was placed at a distance of < 1 cm from the polymer sample with an angle of ca. 35° to the support surface. A cylindrical cardboard sleeve was placed around the sample to ensure temperature homogeneity and maintain the light-sample distance at 10 cm. During the tests, the entire set-up was covered with a dark blanket to exclude external radiation. Two of the LEDs, designated “Far UV” (peak wavelength = 369 nm) and “UV” (peak wavelength = 391 nm), were selected as incident light sources (95% irradiation in the range 361–406 nm). Where not specified in the text, the emission intensity was adjusted to achieve UV irradiance equal to that of sunlight (33.5 W/m²). The calibration was conducted using an integrating sphere, which was placed at the same distance from the source as the sample.

In order to rationally investigate the properties and performances of the stacked devices, we initially studied the optical characteristics and the performances of 5 × 5 × 0.3 cm LSC devices comprising either the DQ1 or BDT-H2 emitter, obtained via cell-casting process, at three concentrations (200, 300, 400 ppm) with an approximate thickness of 3 mm. Examples of these devices are reported in Fig. 1. The molecular structures of the two dyes are shown in Figure S1.

As briefly mentioned in the introduction, one of the main reasons for the choice of these dyes is related to the particular promising match between the emission of the yellow emitting dye (DQ1) and the absorption profile of the red dye (BDT-H2), as one can observe from the spectroscopic characterization of the LSC devices reported in Fig. 2. In particular, DQ1 showed an absorption band centred at 428 nm, in line with what was observed in thin films PMMA LSCs (Papucci et al. 2021), with absorbance increasing with concentration; the emission peak was found 526 nm for the sample with a concentration of 200 ppm, shifting to 529 nm as the concentration was increased to 400 ppm. This small red-shift of the maximum can be ascribed to inner filter effects due to self-absorption, which, despite being small, it is non-negligible in these systems. Notably, the absorption band of BDT-H2 has a maximum corresponding to the emission peak of DQ1, i.e. at 531 nm, similar to what was already reported for thin film devices (Bartolini et al. 2023). The emission of BDT-H2 shifts from 630 to 635 nm as the concentration increases, remaining in the optimal range for mono-Si solar cells. We also characterized the fluorescence s of the LSCs, as reported in Table S1. In this regard, DQ1 showed ≈ 90% in PMMA, decreasing to about 83% for the most concentrated sample. In comparison, BDT-H2 displayed a slightly lower in the range 60–65%, with only small variations with the dye concentration.

The optical characteristics of the two fluorophoric systems investigated suggest that they can be suitable for application in stacked systems: indeed, this configuration could compensate for the shortcomings of the two fluorophores, namely the comparatively low of BDT-H2 and the narrow absorption profile of DQ1. In the event of mixing dyes in a single sheet to achieve FRET, the two emitters under analysis would not be optimal, as the resulting system would suffer from the lower of the longer wavelength absorber, i.e., BDT-H2.

To evaluate the efficacy of these dyes as fluorophores in LSC application, we performed the characterization of the photonic performances of single-layer devices in terms of internal () and external () efficiencies. These parameters represent the fraction of photons emitted from edges compared to the number of absorbed photons and of the total incident photons, respectively (see Supporting Information for a more detailed explanation). The results of the photonic characterisation are shown in Fig. 3 in comparison with LR305 which is considered the state-of-the-art organic fluorophore for LSC application (Picchi et al. 2024a). The of both fluorophores reflects the behaviour of the s, with DQ1 demonstrating a capability to exploit the absorbed photons comparable to that of LR305, and BDT-H2 showing lower performances of about 10%. Nonetheless, with regard to the conversion of incident photons, BDT-H2 exhibits a markedly higher than DQ1 (8.6% vs 4.2% at 400 ppm). This is primarily attributable to a photon absorption rate that is nearly threefold higher (7.7 × 10²⁰ vs 2.7 × 10²⁰ photons per m²), which overcomes the deficiency.

Subsequently, 400 ppm was identified as the optimal concentration for BDT-H2 to be employed as the red layer in the stacked design, and a 5 × 5 cm LSC with a thickness of 2 mm was manufactured at that concentration. Conversely, DQ1 was employed to manufacture 5 × 5 cm LSCs at three concentrations (200, 300, 400 ppm) with the same 2 mm thickness. The reduction in thickness is essential to facilitate the stacking of two slabs in a multilayer configuration, thereby ensuring a geometric gain () that is appropriate for a lab-scale LSC device. Furthermore, the final thickness of the multilayer is not significantly different from that of the devices previously discussed, allowing more significant comparisons. It is recommended in the literature that an air gap be left between layers to direct photons in a more favourable direction by exploiting refraction at PMMA-air-PMMA interfaces (Richards and McIntosh 2006). In this case, the layers were placed on top of one another (Figure S2) without any material that allowed optical contact, resulting in the presence of a small air gap.

We then investigated the photonic efficiencies of the individual 2 mm LSCs and of the following stacked configurations, indicating them as top layer/bottom layer with the concentration as subscript: (i) H2₄₀₀/DQ1₂₀₀; (ii) H2₄₀₀/DQ1₃₀₀; (iii) H2₄₀₀/DQ1₄₀₀; (iv) DQ1₂₀₀/H2₄₀₀; (v) DQ1₃₀₀/H2₄₀₀; (vi) DQ1₄₀₀/H2₄₀₀.

In the case of the monolayer LSCs, the values of are essentially identical to those observed in the devices with a thickness of 3 mm (Table 1). When considering multilayer configurations, we observed that the efficiency in exploiting the absorbed photons () was intermediate between those of each fluorophore when employed as single-layer LSCs. remains closer to that of BDT-H2, which is likely due to its role in the majority of the absorption. The effect of a smaller thickness on of monolayer LSCs is more evident, since a smaller number of fluorophore molecules implies fewer absorbed photons to be re-emitted. fell by ≈ 1% for BDT-H2 and for by 0.3–0.5% DQ1.

The integration of the BDT-H2 layer atop the DQ1 layer, forming a stacked system, resulted in a notable improvement (8.3–8.8%) in compared to the single BDT-H2 layer with a thickness of 2 mm (7.7%) and reached a level comparable to the 3 mm layer at 400 ppm (8.6%). Nevertheless, there is no significant advantage to be gained from placing the BDT-H2 layer on top of the DQ1 LSC. Indeed, the BDT-H2 layer absorbs the majority of incident radiation below 600 nm and subsequently re-emits with a lower QY than the DQ1.

Conversely, the placement of the DQ1 layer atop the BDT-H2 layer yielded up to 9.2%, which is definitely higher than the value of 8.6% recorded for the single-layer 3 mm-thick LSC containing 400 ppm of BDT-H2. It is evident that the manufactured multilayer device, with a thickness of 5 mm, cannot be directly compared with devices of 3 mm thickness, due to its different . In fact, when multiplying to to calculate the concentration factor, a value of 0.23 ± 0.01 is found, whereas it was 0.37 ± 0.05 in the case of the monolayer. Although the fabrication of the DQ1₃₀₀/H2₄₀₀ multilayer LSC requires greater resources and time than the production of a single-layer device, the obtained is among the highest reported for lab-scale devices in the literature since the new laboratory protocols were published (Debije et al. 2021; Yang et al. 2022), together with the 9.1% reported by Lin et al.(Lin et al. 2024)

A deeper understanding of the reasons underlying the high performance of multilayer LSCs can be achieved by analysing the edge-emitted and transmitted photon rate spectra (in nm^− 1 s^− 1) of the device. For clarity, only the results for the DQ1₃₀₀/H2₄₀₀ stacked device, together with the single-layer slabs comprising BDT-H2 400 ppm and DQ1 300 ppm, are shown in Fig. 4; reversing the order of the stacked device does not affect the transmission properties. As one can observe in Fig. 4a, DQ1 absorbs all the incident photons below 450 nm, while BDT-H2 absorbs the entire radiation between 450 and 600 nm. Although only a small fraction of photons is transmitted through the BDT-H2 slab between 320 and 450 nm, when the two layers are stacked, the device absorbs all the sunlight below 600 nm under 1 Sun irradiation. Upon integration, of the total rate of incident photons on a 5 x 5 cm² LSC (7.13 × 10¹⁸ photons s^− 1), 30% (2.15 × 10¹⁸ photons s^− 1) are in the 330–640 nm range and 27% are absorbed (1.92 × 10¹⁸ photons s^− 1). This value equates to absorption efficiency, .

A comparison of the edge-emitted radiation from the various systems reveals notable differences (Fig. 4b). The placement of a layer comprising BDT-H2 on a DQ1 slab (H2₄₀₀/DQ1₃₀₀) results in an edge emission that is similar in profile to that of BDT-H2 itself, but is characterized by a higher intensity. Furthermore, a low intensity peak was observed at 550 nm, which was attributed to the emission of the DQ1 bottom layer. Conversely, when the configuration is reversed (DQ1₃₀₀/H2₄₀₀), the top DQ1 layer exhibits an emission profile and intensity comparable to that of the individual layer. However, a prominent emission peak at 690 nm, attributed to BDT-H2 fluorescence, is observed, also in this case characterized by a higher intensity than that of BDT-H2 by itself. These observations suggest that the DQ1₃₀₀/H2₄₀₀ configuration allows the of DQ1 to be utilised efficiently, with the maximum emission intensity achievable, and then the transmitted radiation is converted into fluorescence by BDT-H2 contained in the bottom layer. A rough estimate of this contribution can come from the integration of the edge-emitted spectra, which gives information on the number of photons involved in the various processes. The photon rate absorbed by DQ1, but that would have been absorbed by BDT-H2 anyway without stacking the layers, is 5.70 × 10¹⁷ photons s^− 1 (6.15 × 10¹⁷ photons s^− 1 absorbed by DQ1 itself minus 4.51 x 10¹⁶ photons s^− 1 in that range not absorbed by BDT-H2 itself). A conversion of them with an of 45% compared to 30% (taken from the data of the individual DQ1 and BDT-H2 layers) leads to an excess of 8.55 × 10¹⁶ photons s^− 1. This number would correspond to an increase of 1.2%.

In addition to this increase, the stacked configuration allows the BDT-H2 layer to use the part of the emission of DQ1 which would have been lost from the slab via escape cone losses (i.e., rays approaching the slab-air interface with angles too small to undergo total internal reflection). A simulation study revealed that the distribution of escape losses is not uniform between the front and rear directions (Wilson 2010). Indeed, in an LSC with 350 ppm LR305 and 3 mm thickness the front escape cone loss is approximately 16% larger than the rear escape cone loss. Similarly, Debije et al. observed comparable outcomes (Debije et al. 2008). The distribution of first-generation fluorescence is more concentrated towards the front surface of the sheet, as this is where the majority of incident sunlight is absorbed. Consequently, photons directed towards the rear surface have a greater probability of being reabsorbed by the dye in comparison to photons directed towards the front. Therefore, if 25% of the emission is directed to the escape cone, and 46% of the total escape cone constitute the rear-escaped photons, this implies that 11.5% of the DQ1 emission (6.15 × 10¹⁷ absorbed photons of the single layer multiplied by a of 0.88) could be reused by BDT-H2, which equates to 6.22 × 10¹⁶ photons s^− 1. It is evident that, if we assume total absorption of the photons in question, only a proportion equal to could effectively be edge-emitted. This equates to 1.87 × 10¹⁶ photons s^− 1, which corresponds to an increase in of 0.3%.

By totalling these three contributions, the increase in should have been 1.7%, whereas the experimental increase was 0.6%. It should be noted that this discussion does not consider a number of potential photon loss modes, including reabsorption and host absorption. Additionally, the standard deviation associated with has not been considered, which could lead to a maximum empirical increase of 1.9%. The objective of these calculations and discussion is to provide a relative scale for the weight of the different contributions to the enhancement in the stacked configuration.

To better address real-world applicability of the systems proposed, we also characterized the photostability of the different polymer/dye systems under intense radiation (see Materials and Methods section). Our measurements showed that BDT-H2 at a concentration of 400 ppm exhibited remarkable photostability, maintaining 97% of the initial emission intensity after more than 1750 hours of simulated ageing (Fig. 5). Although the photodegradation of DQ1 400 ppm occurs at a higher rate, rendering that fluorophore less appealing for upscaling, the photostability of BDT-H2 is comparable to that of the most durable emitter found in the literature and greater than that of LR305, which lost ≈ 10% of emission intensity over the same period of simulated irradiation (Picchi et al. 2024b). It is worth mentioning that the LSC devices prepared in this study were fabricated employing chemically recycled MMA (r-MMA), which was reported to lower the photostability of dyes due to the presence of impurities originating from the depolymerization process (Picchi et al. 2024a).

Finally, envisioning practical applications of LSC technology, which relies on panels much larger than those investigated at a lab scale (from 25 × 25 cm up to 100 × 100 cm, compared to the small 5 × 5 cm), one has to make considerations about losses related to the self-absorption of the fluorophores employed. Indeed, a fraction of the internally reflected photons may be reabsorbed by other luminophore molecules if the absorption and emission spectra overlap. Reabsorption is more likely to occur in large-area devices, where the optical paths are longer, and is detrimental to LSC efficiencies. The probability that a first-generation fluorescence photon will reach the edge of the slab without undergoing reabsorption is given in Figure S3 for DQ1 and BDT-H2 concentrations of 200, 300, and 400 ppm and compared with LR305 and other emitters already characterised in previous publications from our group (Picchi et al. 2024a; Picchi et al. 2024b). The procedure of the reabsorption probability determination is described in the Supporting Information (Wilson et al. 2010). The least reabsorbing of the characterised dyes was DQ1, followed by BDT-H2. Photons emitted by these fluorophores have a probability of over 65% to pass through a 2.5 cm-path in r-PMMA without incurring in reabsorption (at a concentration of 200 ppm). On the contrary, LR305 was the most reabsorbing sample, displaying values from 0.4 to 0.5. For this reason, while considered as the state-of-the-art for its high QY and efficient light absorption, its performance worsens with upscaling the panel size, and thus other organic emitters become competitive.

In this study, we investigated the dyes DQ1 and BDT-H2 for the fabrication of bulk LSC made of r-PMMA in the single- and multi-layer configurations. For these devices, employing dye concentrations between 200 and 400 ppm, notable ≈ 90% s and close to 45% were found for DQ1, while a remarkable equal to 8.6% was achieved by BDT-H2 in single layer LSCs. These promising performances, along with the good overlap of the emission peak of DQ1 with the absorption band of BDT-H2, both centred at around 530 nm, were considered favourable for a multilayer configuration. Once stacked, the LSCs performed more efficiently when the DQ1 layer was placed on top of the BDT-H2, achieving an of 9.2%, comparable with the best performing examples from the literature. The higher observed in the multilayer system can be attributed to three primary factors, in order of importance: (i) the more efficient conversion of incident photons within the DQ1 absorption range; (ii) the exploitation by BDT-H2 of photons emitted by DQ1 in the bottom escape cone; and (iii) the greater overall absorption provided by the overlap of the two absorption spectra of the dyes. BDT-H2 also exhibited remarkable photostability, a property that would be highly desirable for upscaling. The low reabsorption, in conjunction with the complementary absorption features, notable photostability, and high quantum yield of the two fluorophores, renders the proposed stacked system appealing and competitive to LR305, especially for large-scale installations.