Near-Infrared Harvesting Transparent Luminescent Solar Concentrators
Article first published online: 7 MAY 2014
© 2014 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Advanced Optical Materials
Volume 2, Issue 7, pages 606–611, July 2014
How to Cite
Zhao, Y., Meek, G. A., Levine, B. G. and Lunt, R. R. (2014), Near-Infrared Harvesting Transparent Luminescent Solar Concentrators. Advanced Optical Materials, 2: 606–611. doi: 10.1002/adom.201400103
- Issue published online: 15 JUL 2014
- Article first published online: 7 MAY 2014
- Manuscript Revised: 10 APR 2014
- Manuscript Received: 9 MAR 2014
- Michigan State University
- National Science Foundation. Grant Number: CBET-1254662
- DuPont Young Professor Award
- transparent luminescent solar concentrators;
- Stokes shift;
- light harvesting
There is a renewed interest in applying luminescent solar concentrators onto power-producing surfaces such as mobile electronics and windows, since it can dramatically improve energy harvesting efficiency, increase electronic device autonomy, and drive down the expensive PV module cost.[1-7] However, the widespread adoption of such devices is typically limited by chromophore absorption and emission in the visible spectrum which leads to a large degree of colored tinting. To overcome these hurdles, we have developed a transparent luminescent solar concentrator (LSC) which can be integrated onto electronic displays, solar windows and other glazing systems and is highly transparent in the visible spectrum. We harness the structured absorption of molecular dyes to produce near-infrared (NIR) LSC architectures that selectively harvest NIR photons by waveguiding deeper-NIR luminophore emission to high efficiency segmented solar cells. These transparent NIR LSCs can eliminate the visual impact and minimize the amount of expensive solar materials required while extending the photon harvesting range into the NIR.
Transparent and semitransparent power-producing surfaces have been developed for improving the utilization of solar energy without disrupting the aesthetic of surfaces. For example, transparent photovoltaics have been fabricated by exploiting the excitonic character of organic semiconductors which selectively absorb near-infrared light and have reached efficiencies in the range of 2–4% for high transparencies.[8-10] While this approach offers the highest potential for combined transparency and efficiency, it has other challenges for scale-up and defect tolerances. Another approach is through the design of transparent luminescent solar concentrators (TLSC). Recently, ultraviolet (UV) harvesting hexanuclear metal halide clusters with downconverted emission into the NIR were demonstrated for UV TLSCs. The large phosphorescent Stokes shift and high quantum yield allowed for power conversion efficiency of >0.5% over large module area (m2) but these LSCs are ultimately limited to efficiencies <5% due to the limited UV fraction in the solar spectrum. To increase the overall potential of these systems, we look to selectively harvest NIR photons, where there is a substantially greater fraction of the solar photon flux (∼74%). NIR fluorescent dyes, especially phthalocyanines, cyanines, and squaraine dyes have been widely used in fluorescence microscopy, bioimaging, organic light emitting diodes and other light emission applications. However, the quantum efficiency has mostly been limited to <40% and most exhibit visible absorption.[12-16] Previous research on NIR-emitting LSCs employing inorganic compounds such as semiconducting quantum dots and nanocrystals as active materials typically have improved quantum yields[17-21] but also present continuous band absorbance (with only minor excitonic features near the band edge). Accordingly, these systems all exhibit visible absorption or coloring despite emitting NIR.
In this work we demonstrate the first transparent NIR-absorbing luminescent solar concentrator with high transparency and minimal tinting. We focus on the development of luminophore blends of cyanine and cyanine salts and have synthesized cyanine salt-host blends with quantum yields of >20%, combined with spectrally-selective NIR harvesting. We show the impact of Stokes shift on the ultimate performance and demonstrate transparent power conversion efficiencies >0.4%. The near-infrared TLSCs based on organic salts provides an alternative strategy for transparent solar harvesting systems that can ultimately enhance the overall system efficiency of combined UV and NIR TLSCs.
We focus on two promising cyanine derivatives: 2-[7-(1,3-dihydro-1,3,3-trimethyl-2H-indol-2-ylidene)-1,3,5-heptatrienyl]-1,3,3-trimethyl-3H-indolium (HITC) iodide (HITCI) and 1-(6-(2,5-dioxopyrrolidin-1-yloxy)-6-oxohexyl)-3,3-dimethyl-2-((E)-2-((E)-3-((E)-2-(1,3,3-trimethylindolin-2-ylidene)ethylidene)cyclohex-1-enyl)vinyl)-3H-indolium chloride (CY) and utilize these systems to explore the impact of the Stokes shift. The CY and HITCI molecular structures are shown in Figure 1(c). The absorption spectra peaks at 742 ± 1 nm for CY and 733 ± 1 nm for HITCI with little visible absorption, and NIR emission peaks at 772 ± 1 nm and 753 ± 1 nm for CY and HITCI, respectively. The Stokes shift, defined as the wavelength difference between the absorption and emission peaks, is 30 ± 2 nm for CY and 20 ± 2 nm for HITCI. The Stokes shift of the two materials helps demonstrate the difference in assembled LSCs performance and is also an important parameter to predict large-area scalability.
We explore the luminescence quantum yield (QY) dependence on the concentration to understand the photophysical behavior of these luminophores. The measured QY and absorption of CY and HITCI in dichloromethane solutions are shown in Figure 2(b) as a function of luminophore concentration. We fit the QY data of Figure 2(b) to a model with a concentration dependent non-radiative rate and calculate the relationships between intrinsic non-radiative rate for isolated molecules knR0 rate for luminescence kR, and concentration quenching scaling constant a. For CY, knR0/kR = 5.8 ± 0.2, a/kR = 1.4 ± 0.1 ml/mg, while for HITCI, knR0/kR = 2.5 ± 0.2, a/kR = 2.2 ± 0.1 ml/mg. For both CY and HITCI, the non-radiative rate (knR0) is much larger than the radiative rate (kR). The larger ratio a/kR for HITCI implies a slightly larger luminophore-luminophore interaction distance for HITCI; compared to the metal halide nanoclusters these concentration dependent terms are nearly an order of magnitude larger and thus require very dilute concentrations to suppress detrimental interactions. The critical concentrations for CY and HITCI, defined here as the concentration where the QY is half of the maximum, are 5 mg/L and 1 mg/L, respectively. In Figure 1(b) we show the transparent LSC waveguide incorporating the CY luminophore. The measured intensity dependence of the assembled CY LSC is shown in Figure 3(a) which is largely dependent on the intensity dependence of the edge-mounted Si cells. The spectrally resolved external quantum efficiency (EQE) of the overall LSC system of different plate sizes are shown in Figure 4(a), which exhibit a peak at around 760 nm, matching the absorption spectrum and the calculated EQE in Figure 4(b). Shown in Figure 3(b) are the current voltage characteristics of the 2 cm × 2 cm × 0.1 cm CY and HITCI LSC systems for two edge mounted Si solar PVs (see Experimental Section for details). The measured short circuit current density (Jsc) of the overall system for CY under 1.0 ± 0.1 sun was 1.2 ± 0.1 mAcm−2, with an open circuit voltage (Voc) of 0.50 ± 0.01 V and a fill factor of 0.66 ± 0.02 leading to an efficiency 0.4 ± 0.03% for the CY luminophore. The calculated short circuit current density from integrating the product of the EQE and the AM1.5G solar spectrum, is 1.0 ± 0.1 mAcm−2, which is within error of the measured photocurrents. The corresponding average visible transmittance and color rendering index for the CY LSC system is 86 ± 1% and 94, respectively, compared to 90 ± 1% and 100 for the glass substrate alone and is slightly better in aesthetic quality compared to the UV-only TLSCs.
In Figure 5 we show area-scaling calculations of optical efficiency of CY LSC with modeled Stokes shift as a function of LSC length. When the Stokes shift (S) is below 30 nm, the critical plate length, defined here as the LSC length where the optical efficiency is half of the maximum, is around 1–2 cm while increasing the S can significantly increase the critical plate length to >1 m for S > 100 nm.
To explore the potential for modifying the two luminophores, the electronic structures of isolated HITC and CY molecular cations were optimized using density functional theory (DFT) with the CAM-B3LYP functional and 6–31G* basis set. The excitation energies, oscillator strengths, and natural transition orbitals were then calculated at the time-dependent (TD-) DFT level of theory[26, 27] in the Tamm-Dancoff approximation using the same functional and basis set.
Luminophore Photophysics: The individual non-radiative rates for luminophore are larger than the radiative rates, leading to moderately-low quantum yields. This is true for many of the demonstrated NIR fluorophores and there continues to be significant effort to improve QY in this spectral range both for medical applications and light emitting diodes.[12, 13, 16] Here, the quantum yield decreases with increasing concentration due to excited state (non-radiative) quenching caused by particle-particle interactions at higher concentrations: these interactions persist even into dilute solutions likely through long-range dipole-dipole interactions such as Förster energy transfer due to the significant absorption-emission overlap. Accordingly, we design our concentrators using thicker layers of dilute concentrations to maintain both high quantum yield and high absorption efficiency, following the Beer Lambert law. Utilizing this criterion we start with a design concentration of 5mg/L for high quantum yield as well as a reasonable thickness, a molar absorption coefficient of 1.45 × 108 Lmol−1m−1 for CY at 760 nm, leading to blend thicknesses around 1mm for near complete NIR absorption. By utilizing thicker layers of dilute concentrations this also aids in reducing reabsorption losses discussed below.
LSC Design: The efficiency of the transparent LSCs is governed by: the solar spectrum absorption efficiency, luminophore photoluminescence efficiency, waveguide (trapping) efficiency, transport (re-absorption) efficiency and solar cell quantum efficiency and photovoltage. The optical efficiency consists of waveguiding efficiency, transport efficiency, and luminescence efficiency. The EQE consists of the optical efficiency and the quantum efficiency of the PV at the emission wavelength. Due to the monochromatic emission nature of the LSC, only single junction PVs can be attached around each individual LSC, which ultimately limits the overall system efficiency without LSC stacking to that below the PV efficiency directly. Reabsorption losses are critically dependent on the quantum yield of the luminophore, the Stokes shift of the dye emission-absorption, and the overall waveguide dimensions. The reabsorption losses limit the performance of LSC fabricated in this work, due to the moderately low Stokes shift. For example, although HITCI exhibits a slightly higher quantum yield than CY, the CY device has enhanced LSC performance due to the larger Stoke shift. Indeed, it has long been recognized that LSC performance is often limited by reabsorption losses, particularly for dyes with modest S and larger plate sizes. The calculated optical efficiency in Figure 5 shows that an increase of Stokes shift from 30 nm to 80 nm can improve the critical length, from 3 cm to 30 cm, where an LSC size of 30 cm would be appropriate for many LSC applications. To fully maximize the large-area scaling of these devices, the design of high quantum yield molecules with a larger Stokes shift is favorable such as with phosphorescent emitters. However, when designing appropriate fluorescent and phosphorescent luminophores for NIR LSCs there is a balance in considering the ideal Stokes shift since we are limited to PVs that have high quantum efficiency at the luminophore emission peak, and also must consider maximizing the bandgap to minimize PV voltage losses. With Si PVs the maximum Stokes shift is limited to <200 nm with the expectation of harvesting a 200–300 nm slice of the NIR spectrum. For GaAs this maximum S is even more restricted.
One approach to obtain better scalability is through the improvement of the quantum yield closer to 100% through optimization of luminophore-host interactions and molecular design. TDDFT shows that the lowest singlet transitions of HITC and CY are very bright, with oscillator strengths of 3.5 and 3.3, respectively. Natural transition orbital (NTO) analysis demonstrates that these bright excitations correspond to ππ* transitions localized on the polymethine chain of both dyes. NTOs corresponding to the dominant excitation amplitude (>0.95) for each dye are presented in Figure 1(c). The predicted vertical excitation energies for these transitions are 2.56 and 2.52 eV for HITC and CY, respectively; the overestimation of the excitation energies is not surprising given that the range-corrected functional used here is known to overestimate the vertical excitation energies of cyanine dyes. To determine a correction to our TDDFT vertical excitation energies, the highly accurate equation-of-motion coupled cluster singles and doubles (EOM-CCSD) level of theory was applied to a similar but truncated cyanine dye. Comparison to the TDDFT-computed excitation energy of the same truncated system suggests that TDDFT overestimates the vertical excitation energies of these dyes by approximately 0.69 eV. Taking this as a correction to the TDDFT vertical excitation energies of HITC and CY yields energies of 1.87 and 1.83 eV, respectively, which are in much better agreement with experiment. Thus we are confident in the assignment of these NIR absorptions to polymethine-localized ππ* transitions. (See Experimental Section for details of these EOM-CCSD calculations.) These calculations help us to understand molecular motif targets to manipulate bandgap, oscillator strength, radiative rates and potentially improve QY.
Our present work and the past work of other groups both point toward rational principles for the design of NIR cyanine emitters with larger Stokes shifts, and thus promising candidates to decrease reabsorption losses. Modified cyanine dyes which exhibit efficient fluorescence with Stokes shifts of ∼100 nm have been reported.[34, 35] These molecules, which emit from intramolecular charge transfer excited states, absorb in the visible range and thus are not useful in the present application. However, their existence demonstrates that the engineering of cyanines with intramolecular charge transfer states may allow us to achieve NIR emitters with large Stokes shifts through substitutions in the central methine group with bridgehead amine groups (that could be applied to the molecules here). Another strategy is suggested by the increased Stokes shift of CY compared to HITC. The side chain or the central methine group of CY must be responsible for this increase, given that these are the only differences between the two luminophores. Considering the flexibility of the side chain, we hypothesize that it may interact with the polymethine chain electrostatically, reorienting to stabilize the redistribution of charge upon excitation in the same way that a solvent molecule would respond to excitation of a solvated dye. Increasing the number and/or polarity of these side chains as well as central methine substitution are two rational strategies to engineer visibly-transparent NIR luminophores with larger Stokes shifts, and will be pursued in future work.
For materials with QY close to 100% each reabsorption event leads to another emission event, reducing transport losses even for small S. However, it should be noted that for each absorption-reemission event this appears as additional scattering since radiative emission is typically isotropic and therefore would eventually lead to greater front/back-side losses for larger plate lengths. Provided high quantum yields can be achieved, multiple cascaded luminophores could also be utilized to increase the effective Stokes shift.[22, 36]
Another important approach to improve the scalability for low Stokes shift materials, is to embed highly-segmented solar cell micro-arrays as meshes throughout the LSC waveguide to essentially create a series of “micro-LSCs“,[37, 38] allowing for minimal reabsorption losses and additional contribution from the segmented PV. Considering this last approach, and combining these demonstrated NIR LSCs with the UV-TLSCs, efficiencies >1% are readily achievable, and efficiencies approaching 10% are possible with enhancements in QY, absorption range, Stokes shifts, and micro-LSCs.
We have designed and fabricated the first visibly-transparent luminescent solar concentrator devices which selectively harvest NIR photons based on fluorescent organic salts. These transparent TLSCs exhibit a non-tinted transparency of 86 ± 1% in the visible spectrum combined with an efficiency of 0.4 ± 0.03% and have the potential for efficiencies up to 10% due to the large fraction of photon flux in the near-infrared. We show through both experiments and modeling that the development of larger Stokes shift luminophores, optimization of luminophore–host interactions and fabrication of embedded segmented-PV configurations could further reduce reabsorption losses and increase system efficiencies. These transparent NIR LSCs provide a new route to transparent light-harvesting systems with tremendous potential for high defect tolerances and processability.
Organic Salt Solution Preparation: 1-(6-(2,5-dioxopyrrolidin-1-yloxy)-6-oxohexyl)-3,3-dimethyl-2-((E)-2-((E)-3-((E)-2-(1,3,3 trimethylindolin-2-ylidene)ethylidene)cyclohex-1-enyl)vinyl)-3H-indolium chloride (CY) (Lumiprobe) and 2-[7-(1,3-dihydro-1,3,3-trimethyl-2H-indol-2-ylidene)-1,3, 5-heptatrienyl]-1,3,3-trimethyl-3H-indolium iodide (HITCI) (Exciton) were characterized as received without further purification. Solutions for optical characterization were prepared by directly dissolving each compound in dichloromethane at various concentrations up to 4 mg/L.
Optical Characterization: Specular transmittance of both solutions and films were measured using a dual-beam Lambda 800 UV/VIS spectrometer in the transmission mode without the use of a reference sample. The emission spectra and quantum yields for various dyes were measured using a PTI QuantaMaster 40 spectrofluorometer with excitation at 730 nm (HITCI) and 700 nm (CY); quantum yield measurements were made using a calibrated integrating sphere attachment.
Module Fabrication: A 5 mg/L CY dichloromethane solution was mixed with (poly)-butyl methacrylate-co-methyl methacrylate (PBMMA) (Sigma-Aldrich) at a volume ratio of 1:1, yielding the target dye concentration (5 mg/L) in the polymer composite film. This mixture was drop cast on 2 cm × 2 cm × 0.1 cm (for efficiency measurements) or 7 cm × 7 cm × 0.1 cm (for EQE measurements) glass substrates and allowed to dry for 2 h for each layer, and repeated 3 times, resulting in a layer thickness of approximately 1mm. On the mounted edge, a laser-cut, 7 cm × 0.1 cm Si cell (Narec Solar) with an efficiency of 15 ± 1% @ 1 sun was attached within a clip that fit around the edge of the waveguide and film. To block edge reflection and simplify the geometric configuration for the EQE measurements, the other three edges were attached with black electrical tape. For the power efficiency measurements, two cells were mounted on orthogonal edges and connected in parallel. The remaining two edges were covered with enhanced specular reflector (Vikuiti, 3M). To avoid any direct illumination of the solar cell, a thin border area around the LSC edges was masked. Due to the illumination area of the solar simulator (67005 Xe arc lamp, Newport) plate lengths for power efficiency measurements were limited to <0.05 m and for EQE measurements <0.2 m.
Module Testing: EQE measurements were obtained by directing a monochromatic excitation beam from a fiber perpendicular to the LSC at various distances (d) from a single edge-mounted Si PV. The measured EQE was corrected by the factor, g = π/tan–1(L/2d), due to the different angle subtended by the solar cell at each spot distance, where g is the geometric collection correction for collecting photons on only one edge of the concentrator (the rest of the edges are blackened), d is the distance from the illumination spot on the LSC and L is the length of the collection edge of the LSC. The EQE was measured utilizing a Newport calibrated Si detector. Current density versus voltage (J–V) measurements were obtained under simulated AM1.5G solar illumination corrected for solar spectrum mismatch and note a good match between the integrated EQE and the measured photocurrent.
Optical Modeling: The PV efficiency (reported for AM1.5G) is normalized by the solar spectrum absorption efficiency and the quantum efficiency at the luminophore wavelength.[11, 40] The reabsorption and forward emission losses are estimated with luminophore emission spectrum, molar absorptivity, concentration, LSC length, and the LSC system optical efficiency, considering reabsorption losses from the overlap in the absorption and emission spectra were numerically evaluated in Matlab as a function of plate size, and luminophore concentration at fixed blend thicknesses.
Electronic Structure Calculations: All DFT calculations were performed with the Q-Chem ab initio electronic structure package[41-43] while EOM-CCSD calculations were performed in GAMESS.[41, 42, 44-46] In order to assess the magnitude of the error in our CAM-B3LYP/6–31G* calculations of the vertical excitation energies of the cyanine dyes, a truncated model dye: (CH3)2N(CH)9N(CH3)2+ was evaluated. This dye contains the same 9-carbon polymethine chain terminated by tertiary amines as HITC+ and CY+, but smaller dimethylamine groups replace the larger heterocycles to make more accurate quantum chemical calculations feasible. The vertical excitation energy of this model was calculated at the same TDDFT level of theory as applied to the larger dyes and with the more accurate equation-of-motion coupled cluster singles and doubles (EOM-CCSD) level of theory with the cc-pVDZ basis set. The computed vertical excitation energies at the TDDFT and EOM-CCSD levels of theory are 3.09 and 2.40 eV, respectively. The difference, 0.69 eV, is subtracted from the TDDFT-computed vertical excitation energies of the larger dyes to estimate the more accurate EOM-CCSD excitation energies of these dyes. Note that the lowest excitation in the truncated cyanine has identical ππ* character to that predicted by TDDFT for the lowest excitations in HITC and CY.
Financial support for this work was provided in part by Michigan State University startup funds, the National Science Foundation (CAREER award, CBET-1254662) and a DuPont Young Professor Award.