Thin film luminescent solar concentrators are promising components of distributed power generation systems for building integrated photovoltaic applications. However, thin film geometries require high luminophore loading fractions to achieve sufficient absorption of sunlight, which, in the case of nanocrystal luminophores, can lead to aggregation and light scattering. In this work, we integrate CdSe/CdS nanocrystals into thin films of poly(cyclohexylethylene) at a range of loading fractions and characterize the composites with a combination of spectroscopic and simulation tools. We find that increased incident sunlight scattering is observed for the increasing luminophore loading fraction, but that the scattering is mostly limited to higher energy sunlight such that visible transmittance and haze of the samples are all greater than 89.7% and less than 8.3%, respectively. We then analyze the refractive index of the composite and show that the increase in loading fraction also affects the propagation of photoluminescence in the film, especially if the refractive index of the film is greater than that of the substrate. These studies show the importance of understanding the optical transport within thin films and provide design criteria to fabricate thin films for future implementation into building integrated photovoltaic applications.

Thin film luminescent solar concentrators (LSCs) have promising applications in building integrated photovoltaic (BIPV) applications as a way to integrate solar harvesting into the facades of urban buildings.1 In a luminescent solar concentrator, a luminophore is embedded into a host material that typically acts as an optical waveguide. Incident sunlight is absorbed by the luminophore and emitted into the modes of the waveguide, where it is concentrated onto a small solar cell mounted to the edge of the concentrator. With careful choice of the luminophore and polymer matrix, these LSCs could act as semitransparent energy harvesting glazings for both opaque and transparent building materials, producing a distributed power generation system without disrupting the view of the occupants.1–10 

Thin film configurations, where the luminophore-polymer composite is coated on a glass substrate, offer advantages for LSC performance. Common loss pathways, including reabsorption, scattering from luminophores, and scattering from imperfections in the polymer matrix, are reduced as a function of the ratio of thicknesses between the nanocomposite and the glass substrate.11 In addition, nanoscale, thin film architectures are compatible with many advanced photonic structures that reduce escape cone losses.12–15 

However, it can be challenging to realize thin film luminescent composites that meet all of the demands of luminescent concentrators, especially when utilizing nanocrystal luminophores. The matrix needs to possess high optical clarity, a high refractive index, be processable into films, be resistant to weathering and ultraviolet (UV) light degradation, and be chemically compatible with the nanocrystal ligands to avoid both aggregation and detrimental optical losses from reduced quantum yields. A key consideration for nanocrystal-based LSCs is the effect of light scattering: while in dye-based LSCs, this may arise from internal defects in the waveguide, in nanocrystal-based concentrators scattering can also arise from the dispersed nanocrystals themselves or from aggregation.16 For thin film configurations that require higher loading concentrations of nanocrystals to achieve sufficient absorption of sunlight, achieving high quality dispersions with minimal light scattering can be a significant challenge.

Many nanocrystal based concentrators are fabricated by embedding nanocrystals in a poly(laurylmethacrylate) (PLMA) matrix that has been cross-linked with ethylene glycol dimethacrylate (EGDMA)17–21 or in a polymethylmethacrylate (PMMA) matrix.4,16 PLMA has been shown to achieve a homogeneous dispersion of quantum dots in the polymer matrix due to the interaction between the long carbon chain of the LMA monomer and the nanocrystal ligands. However, without the cross-linker, PLMA is a viscoelastic liquid with a glass transition temperature well below room temperature, making it susceptible to flow and delamination.22 When the cross-linker is added, the final polymer is no longer soluble, which prevents synthesis of the polymer prior to QD addition and limits large scale solution processing. Another drawback to the PLMA-co-EGDMA and PMMA systems is the presence of absorption in the near infrared (IR), which can interact with infrared-emitting luminophores.

In this work, we study thin film LSCs comprising CdSe/CdS nanocrystals embedded in a matrix of poly(cyclohexylethylene) (PCHE). PCHE is soluble in nonpolar solvents and is an attractive polymer for optical materials. PCHE is derived from the existing polystyrene (PS) material by metal catalyzed hydrogenation.23,24 Dow Chemical has already developed the infrastructure to develop PCHE-based materials for commercial applications, including for optical storage media, and USI corporation has PCHE-containing materials available in the form of elastomers.25 PCHE is prepared from polystyrene (PS), so it may be useful to compare the cost of PS to PMMA. PS is a lower cost material than both PMMA and polycarbonate (PC); after factoring in hydrogenation, the cost may be more comparable. In terms of properties, PCHE exhibits a glass transition temperature (Tg) of about 408 K, well above ambient and significantly higher than PLMA (∼220 K), PMMA (378 K), and the precursor PS (375 K). This affords PCHE a wider use temperature range.26 In addition, PCHE does not absorb strongly in the ultraviolet region and is resistant to oxidation due to the decreased ability of the methine hydrogen to be abstracted from the backbone.27–29 PCHE also displays lower water absorption, density, and coefficient of thermal expansion compared to PMMA, which are all favorable properties for materials in a wide range of applications.26,30–32 PCHE exhibits similar hardness to PMMA, as both have a Rockwell R125 rating.33–35 The Dynstat unnotched impact strength of both PCHE and PS (2.8 kJ/m2) are comparable to that of PMMA (3 kJ/m2).33–35 The impact strength of PS has previously been modified by including elastomeric additives or copolymers to create high impact PS.35 A similar block-polymer approach could be taken to improve the impact strength of PCHE while retaining optical transparency if necessary.26 Finally, the refractive index of PCHE is similar to PMMA and PLMA, ∼1.5, which still allows for ∼75% of the photoluminescence (PL) to be trapped by total internal reflection on each absorption and PL event.

The luminophore in this study consists of CdSe/CdS core/shell quantum dots with a 7.5 monolayer shell. This common nanocrystal platform has been synthesized with quantum yields up to 99% for thin shells and 86% for thicker shells, and exhibits tunable absorption and PL dependent on the thickness of the CdS shell.17,36 These quantum dots have also been synthesized with a narrow PL bandwidth, which makes them advantageous for integration with a variety of nanophotonic structures.37 

We fabricate thin films of CdSe/CdS–PCHE composites with a range of CdSe/CdS loading fractions. Through a combination of spectroscopic and simulation techniques, we show how the refractive index, transmission, reflection, and luminophore PL are affected by the increased loading fraction of luminophores. We find that proper tuning of the loading fraction and thickness of the nanoscale thin films is required to fabricate LSC films that effectively direct light toward the edges of the concentrator and that optical clarity is achieved for the full range of loading fractions studied.

CdSe/CdS nanocrystals were synthesized following the synthesis procedures in Refs. 38 and 39, and as previously described.40 A transmission electron micrograph of the synthesized QDs is shown in Fig. S1 of the supplementary material. These nanocrystals have an average diameter of 8.4 nm with a standard deviation of 0.7 nm as determined by transmission electron microscopy.

Styrene (Sigma-Aldrich, ReagentPlus Grade) was purified by passing through a short column of basic alumina (Sigma). Toluene (Fisher Chemical, HPLC Grade) was stored and delivered through a PPT glass contour solvent delivery system. Lauroyl peroxide (Aldrich, 97%) was recrystallized from pentanes before use. Nuclear magnetic resonance (NMR) spectra were taken in deuterated chloroform on a Bruker Avance 400 MHz spectrometer. Size-exclusion chromatography (SEC) was performed on an Agilent 1260 Infinity liquid chromatograph equipped with a Wyatt DAWN Heleos II 18-angle laser light scattering detector and a Wyatt OPTILAB T-rEX refractive index detector using tetrahydrofuran as the mobile phase.

Preparation of precursor PS

Polystyrene (PS) was prepared by free radical polymerization of the styrene monomer in toluene using lauroyl peroxide (Luperox LP) as an initiator. In a typical polymerization, 6.3 ml of styrene, 2.5 ml of toluene, and 74 mg of lauroyl peroxide were combined in a Schlenk flask with a magnetic stir bar. This flask had previously been flame dried, evacuated, and refilled with argon for 3 cycles. The flask was then heated in an oil bath to 85 °C for 16 h. The reaction mixture was then diluted with tetrahydrofuran before precipitating into methanol. The polymer was recovered as a white solid (3.34 g, 58% yield, Mn,LS-SEC = 78 kg/mol, and LS-SEC = 1.9) and was dried under vacuum at 80 °C for 16 h.

Hydrogenation of PS

The catalytic hydrogenation of the PS was performed in a Parr 1 L pressure reactor. PS (3.2 g) was dissolved in cyclohexane (400 ml) and added to the reactor. 8 g of the catalyst (5% Pd on CaCO3) was added to the reactor, which was then sealed and purged several times with argon. The reactor was then raised to 140 °C and charged with 500 psi (gauge) of hydrogen gas. The reaction was allowed to proceed with mechanical stirring for 20 h. The pressure of hydrogen gas was refilled to 500 psi until the pressure no longer decreased over a 30-min time frame, signaling the end of the reaction. The reactor was allowed to cool to room temperature and then vented. The reaction slurry was collected and filtered to remove the catalyst. The cyclohexane solution was then concentrated by rotary evaporation to afford PCHE (3.2 g, 96% yield, >99.9% saturation by NMR analysis, Mn,LS-SEC = 85 kg/mol, and LS-SEC = 1.55) as a white solid.

An initial PCHE/octane solution was created at a concentration of 200 mg/ml and stirred for 2 h in order to ensure full dissolution of the PCHE in the octane. The concentrated solution was run through an alumina plug to filter any polymer reaction byproducts or impurities. After filtration, a volume of the concentrated PCHE solution was added to octane such that the PCHE was at a concentration of 60 mg/ml. Then, a known mass of the QDs was added to the solution such that the concentration of QDs in octane was 0, 3, 6, 9, or 12 mg/ml.

This solution was then spun on 1 in. × 1 in. glass and silicon substrates using a Laurell WS-650Mz-23NPPB spin coater. The solution was deposited onto the substrate and then spun at 2000 rpm for 20 s with an acceleration of 500 rpm/s and then 500 rpm for 40 s with an acceleration of −500 rpm/s.

Spectroscopic ellipsometry measurements were performed on the PCHE and the CdSe/CdS-PCHE thin film composites on Si substrates using a J.A. Woollam Co., Inc. Variable Angle Spectroscopic Ellipsometer. Measurements were taken at three points on each sample. The PCHE data were modeled using a Sellmeier fit from 350 to 1000 nm. For the composites, a Maxwell-Garnett effective medium approximation was used to model the ellipsometry data and estimate the optical constants. The measured optical constants for the PCHE material were used for the matrix material, and the optical constants for the CdSe/CdS quantum dots in solution were used for the inclusions, following the method described elsewhere.40 More information on the effective medium approximations can be found in the supplementary material.

Transmission measurements of the PCHE and the CdSe/CdS-PCHE thin film composites were performed using a Cary 7000 UV-Vis spectrophotometer. The total transmission through the sample was measured by first attaching the sample to the transmission port of an integrating sphere accessory on the UV-Vis spectrophotometer with a powder PTFE reference on the back port. Diffuse transmission was measured by removing the PTFE reference. These transmission measurements were used to calculate the visible haze and transmittance, which are discussed in more detail in the supplementary material.

Reflection measurements of the PCHE and the CdSe/CdS-PCHE thin film composites were also performed using a Cary 7000 UV-Vis spectrophotometer. Diffuse reflection was measured by attaching the sample to the backside of the integrating sphere accessory of the UV-Vis spectrophotometer. The total reflection was measured by tilting the sample by 3° 10′ so that specular reflection was coupled into the integrating sphere.

Spatially resolved steady-state PL was measured using a Princeton Instruments Isoplane 160 spectrometer coupled to an Axio Observer D1 inverted microscope. Each thin film composite was mounted on a Mad City Labs, Inc. piezo nanopositioner and excited by a 405 nm PicoQuant picosecond pulsed diode laser. The PL was measured at 121 equidistant positions to create a 10 μm × 10 μm grid of measured points. Each measured spectrum was fitted to a Gaussian distribution to identify the peak PL wavelength.

A single photon avalanche diode from Micro Photon Devices with a PicoHarp 300 electronics box from PicoQuant was used to measure the time-resolved PL. The same light source and nanopositioner, which were used for steady-state PL, were used for the time-resolved PL. The PL histograms were normalized and fit to a stretched exponential,

It=et/τβ+Cbkg,
(1)

where I is the PL intensity, t is time, τ is lifetime, β is the stretch parameter, and Cbkg is the background counts.

Figure 1 shows the optical properties of the constituent materials, with the absorption and PL spectra of the CdSe/CdS QDs shown in Fig. 1(a) and the UV-Vis spectrum of a solution of PCHE in Fig. 1(b). Notably, there is a lack of absorption features throughout the visible and near IR in PCHE that could interfere with luminescent light propagation, which would benefit devices made with near-infrared emitting nanocrystals.5,16Figure 1(c) shows photographs of the films with varying concentrations of nanocrystals under UV illumination. The composites were made by mixing the nanocrystals with the solution of PCHE at various loading fractions and spin coating onto glass slides, producing films with an approximate thickness of 350 nm. A table of the average thickness of each film is provided in the supplementary material.

FIG. 1.

(a) Absorption (black) and PL (red) spectra of CdSe/CdS core/shell quantum dots with a 7.5 monolayer shell. (b) Absorption spectrum of reference PCHE in octane. (c) Photographs of CdSe/CdS–PCHE nanocomposites under UV illumination with varying CdSe/CdS loading fractions between 0 mg/ml of octane (left) and 12 mg/ml of octane (right).

FIG. 1.

(a) Absorption (black) and PL (red) spectra of CdSe/CdS core/shell quantum dots with a 7.5 monolayer shell. (b) Absorption spectrum of reference PCHE in octane. (c) Photographs of CdSe/CdS–PCHE nanocomposites under UV illumination with varying CdSe/CdS loading fractions between 0 mg/ml of octane (left) and 12 mg/ml of octane (right).

Close modal

To quantitatively characterize the interaction of the thin film LSCs with incident sunlight, transmission and reflection measurements were performed. Figures 2(a) and 2(b) show the total transmission and reflection, respectively, of the PCHE reference and the four CdSe/CdS–PCHE composites. As the loading fraction of the QDs increases, the transmission of incident light decreases over the spectral range from 370 nm to 600 nm due to the increase in optical density of the luminescent film at higher CdSe/CdS loading fractions. In addition to the decrease in the transmission of incident light, oscillations are observed in the reflection and transmission spectra for the composites. These oscillations arise from thin film interference effects based on the thickness and refractive index of each composite.41 

FIG. 2.

Transmission (left) and reflection (right) measurements made for a PCHE reference (black) and CdSe/CdS–PCHE nanocomposites with a loading fraction of 3 (blue), 6 (red), 9 (green), and 12 mg/ml (purple). Total transmission and reflection are shown in (a) and (b), respectively, while the diffuse component of the transmission and reflection are shown in (c) and (d), respectively. The table in (a) shows the haze and visible transmittance (VT).

FIG. 2.

Transmission (left) and reflection (right) measurements made for a PCHE reference (black) and CdSe/CdS–PCHE nanocomposites with a loading fraction of 3 (blue), 6 (red), 9 (green), and 12 mg/ml (purple). Total transmission and reflection are shown in (a) and (b), respectively, while the diffuse component of the transmission and reflection are shown in (c) and (d), respectively. The table in (a) shows the haze and visible transmittance (VT).

Close modal

For performance as an optical waveguide, it is especially important that the composite has low diffuse scattering, as this disrupts the propagation of luminescent light. Figures 2(c) and 2(d) show the diffuse transmission and reflection, respectively, of the samples. The diffuse transmission and reflection can be thought of as the incident light that is forward scattered for diffuse transmission or back scattered for diffuse reflection. In the samples that we measured, we observe little change in the diffuse component of the reflection between the PCHE reference and the CdSe/CdS–PCHE composite at a loading fraction of 3 mg/ml in octane. However, as the loading fraction of QDs increases, the diffuse transmission and diffuse reflection also increase, which is evidence of aggregation or interaction between the nanocrystals. The increased concentration of QDs will increase the scattering cross section due to the greater refractive index of the CdSe/CdS QDs compared to the polymer matrix.42 

The forward and backward scattering is mostly limited to the higher energy wavelengths, where the eye is less sensitive. Therefore, the scattering does not affect the clarity of the film, as characterized by haze and visible transmittance (VT). Haze and visible transmittance are standard methods of characterizing the properties of a window in the visible regime. Haze is the percentage of light that is diffusely scattered, while visible transmittance is the fraction of visible light transmitted through the LSC, weighted by the solar spectrum and the sensitivity of the eye. These calculations are found in the inset of Fig. 2(a). Even at the highest loading fraction of 12 mg/ml of octane, the haze is only 8.3% and the visible transmittance is 89.7%, well within the thresholds for an optically clear window.43 

In very thin films, the loading fraction of nanocrystals must be high to achieve significant absorption, which in turn modifies the refractive index of the composite and the light propagation in the system. The real and imaginary components of the refractive index of the nanocomposites are shown in Fig. 3, as measured using spectroscopic ellipsometry. The black lines correspond to the volume fractions of nanoparticles in the matrix, estimated using a Maxwell-Garnett effective medium approximation. As the loading fraction of the CdSe/CdS quantum dots increases in the composite, the refractive index of the composite exhibits the spectral characteristics of the CdSe/CdS quantum dots. In bulk composites, the QDs often make up less than 1% by volume of the total composite, which does not significantly affect the refractive index compared to the bulk polymer. However, in this geometry, we can achieve a volume fraction of up to 8.25% with the most concentrated sample. This increased volume fraction leads to an increase in the refractive index of the composite, as well as the spectral oscillation features that are observed in Fig. 2. In addition, the imaginary component mirrors the spectral features of the absorption spectrum. Importantly, the refractive index also changes the light propagation within the film, as discussed later. We note that the volume fractions extracted from spectroscopic ellipsometry are similar to simple mixing ratios for the three lowest concentrations but exhibit some deviation for the highest concentration.

FIG. 3.

Real (a) and imaginary (b) components of the refractive index for the PCHE reference (black) and CdSe/CdS–PCHE nanocomposites with a loading fraction of 3 (blue), 6 (red), 9 (green), and 12 mg/ml (purple). Dotted black lines on both plots show the predicted refractive index given by the Maxwell-Garnett effective medium approximation for QD volume fractions for 0%–10%.

FIG. 3.

Real (a) and imaginary (b) components of the refractive index for the PCHE reference (black) and CdSe/CdS–PCHE nanocomposites with a loading fraction of 3 (blue), 6 (red), 9 (green), and 12 mg/ml (purple). Dotted black lines on both plots show the predicted refractive index given by the Maxwell-Garnett effective medium approximation for QD volume fractions for 0%–10%.

Close modal

It is also important that the PL and radiative lifetime do not change significantly with loading fraction. Figure 4 shows spatially resolved data of both the PL wavelength maximum and the radiative lifetime. These measurements do not appear to show any spatial inhomogeneity in the film over the 100 × 100 μm2 range that was studied. This implies that the increased loading fraction does not induce any large aggregates with size greater than 10 μm that significantly affect the PL properties of the film. In addition, the PL peak wavelength does not appear to significantly shift from sample to sample. The range of peak wavelengths is approximately 0.7 nm, centered around the peak wavelength of 634 nm, which is well within the expected uncertainty of the measurement. In addition to the small range of peak wavelengths measured, there appears to be no obvious trend in peak wavelength as a function of loading fraction. In contrast to the peak PL wavelength, the radiative lifetime of the PL does slightly decrease with increasing loading fraction, decreasing by approximately 1.4 ns over the range of loading fractions studied. One possible explanation is the nonradiative energy transfer from one QD to another due to close proximity at a high loading fraction.44 

FIG. 4.

Spatially resolved peak PL wavelength (left) and PL lifetime (right) measured for CdSe/CdS–PCHE nanocomposites with CdSe/CdS loading fractions of 3 [(a) and (b)], 6 [(c) and (d)], 9 [(e) and (f)], and 12 mg/ml [(g) and (h)], labeled with their corresponding volume fractions, extracted from spectroscopic ellipsometry measurements.

FIG. 4.

Spatially resolved peak PL wavelength (left) and PL lifetime (right) measured for CdSe/CdS–PCHE nanocomposites with CdSe/CdS loading fractions of 3 [(a) and (b)], 6 [(c) and (d)], 9 [(e) and (f)], and 12 mg/ml [(g) and (h)], labeled with their corresponding volume fractions, extracted from spectroscopic ellipsometry measurements.

Close modal

To fabricate a high concentration factor LSC, it is important to balance the absorption of incident light and the reabsorption of PL photons. These factors can compete with one another as the refractive index changes: if the index is high, there may be greater absorption and less outcoupling of luminescent light, but reabsorption losses will also increase. To study these factors, we used electromagnetic simulations to calculate the modes of the waveguides and determine the reabsorption losses. These 2D simulations were performed for a range of CdSe/CdS volume fractions from 0 to 20 vol. % and thicknesses from 100 to 1000 nm. The refractive index of the thin film with a given volume fraction was calculated using a Maxwell-Garnett EMA model as discussed previously. More information regarding the simulations performed can be found in the supplementary material.

Figure 5(a) shows the fraction of incident light that is absorbed by the CdSe/CdS–PCHE thin film weighted by the AM 1.5G solar spectrum over the spectral range from 400 to 700 nm. As expected, the fraction of the incident sunlight that is absorbed by the thin film LSC increases as both the volume fraction of the luminophores and the thickness increase.

FIG. 5.

(a) Predicted absorption of incident sunlight weighted by the solar spectrum over the spectral range from 400 to 700 nm. (b) Predicted absorption of the first 5 modes calculated for a wavelength of 630 nm, which is approximately the peak PL wavelength of the CdSe/CdS QDs.

FIG. 5.

(a) Predicted absorption of incident sunlight weighted by the solar spectrum over the spectral range from 400 to 700 nm. (b) Predicted absorption of the first 5 modes calculated for a wavelength of 630 nm, which is approximately the peak PL wavelength of the CdSe/CdS QDs.

Close modal

To calculate reabsorption, we calculated the first 5 modes (both TE and TM polarization) of the structure for a given thickness and volume fraction at a wavelength of 630 nm, the PL maximum. Then, these modes were excited and tracked over a distance of 10 μm. Selected electric field intensity distributions are shown in Fig. S3 of the supplementary material. The reabsorption of the electric field intensity used to excite the PL modes was calculated from the electric field intensity distribution and the imaginary component of the refractive index of the composite, as shown in Fig. 5(b). We note that reabsorption is not necessarily a loss process as calculated here, as some of these photons could be re-emitted depending on the quantum yield. Some of these re-emitted photons will also be outcoupled from the waveguide. Therefore, we use this metric to indicate, in general, how reabsorption is affected by the volume fraction, although the losses strictly due to reabsorption and nonemission could be found by accounting for the nanocrystal quantum yield. We find that for the largest thicknesses and volume fractions studied, up to 30% of the electric field intensity that excites the PL modes will be reabsorbed within the thin film over this propagation length. As the thickness and volume fraction decrease, the fraction reabsorbed also decreases. In these cases, the modes are less confined in the absorbing CdSe/CdS–PCHE layer and, therefore, are less susceptible to reabsorption losses. Therefore, a tradeoff exists between the absorption of incident sunlight and the reabsorption of PL modes. Of the thicknesses and volume fractions studied, a volume fraction of 6 vol. % and a thickness of 500 nm, as well as a volume fraction of 4 vol. % and a thickness of 800 nm show a good balance of incident sunlight absorption and reduced reabsorption losses that may be useful for future study.

In this work, we prepared thin film LSCs using CdSe/CdS quantum dots embedded in PCHE, an attractive high Tg plastic with favorable optical properties. Using a combination of spectroscopic and simulation tools, we studied how the loading fraction of the QDs in the PCHE/octane solution affects the incident light transmission, film refractive index, and PL. We find that PCHE is a good candidate matrix for these applications, as across the range of loading fractions studied, the composites exhibit high visible transmittance, low haze, and uniform PL peak wavelength.

In addition, the absorption of incident sunlight and reabsorption of PL modes can be balanced by carefully tuning the volume fraction and thickness of the thin film LSC, especially since the refractive index of the composite can be modified to be higher or lower than the index of the glass substrate. We find that while increasing loading fractions lead to increased light absorption, these composites also lead to higher reabsorption of luminescent light, with additional effects arising from the changing refractive index of the composite. This work shows the importance of carefully designing the refractive index and thickness of the thin films to fabricate concentrators with low waveguiding losses for future integration into building integrated photovoltaic applications and nanophotonic architectures.

See the supplementary material for information about CdSe/CdS characterization, Maxwell-Garnett effective medium approximations, and simulation methods.

The authors thank Mayank Puri for CdSe/CdS synthesis and characterization. This work was supported partially by the National Science Foundation under Award No. 1553234. Partial support was received from a Discovery grant from the Institute on the Environment at the University of Minnesota under Award No. DG-0002-17. We also acknowledge partial support from the Minnesota Environment and Natural Resources Trust Fund (M.L. 2018, Chp. 214, Art. 4, Sec. 02, Subd. 07a). Part of this work was carried out in the College of Science and Engineering Characterization Facility, University of Minnesota, which has received capital funding from the NSF through the UMN MRSEC program under Award No. DMR-1420013. The authors acknowledge the Minnesota Supercomputing Institute (MSI) at the University of Minnesota for providing resources that contributed to the research results reported within this paper. URL: http://www.msi.umn.edu.

1.
F.
Meinardi
,
F.
Bruni
, and
S.
Brovelli
,
Nat. Rev. Mater.
2
,
17072
(
2017
).
2.
M. G.
Debije
and
P. P. C.
Verbunt
,
Adv. Energy Mater.
2
,
12
(
2012
).
3.
W. G. J. H. M.
van Sark
,
Renewable Energy
49
,
207
(
2013
).
4.
F.
Meinardi
,
A.
Colombo
,
K. A.
Velizhanin
,
R.
Simonutti
,
M.
Lorenzon
,
L.
Beverina
,
R.
Viswanatha
,
V. I.
Klimov
, and
S.
Brovelli
,
Nat. Photonics
8
,
392
(
2014
).
5.
F.
Meinardi
,
S.
Ehrenberg
,
L.
Dhamo
,
F.
Carulli
,
M.
Mauri
,
F.
Bruni
,
R.
Simonutti
,
U.
Kortshagen
, and
S.
Brovelli
,
Nat. Photonics
11
,
177
(
2017
).
6.
F.
Meinardi
,
H.
McDaniel
,
F.
Carulli
,
A.
Colombo
,
K. A.
Velizhanin
,
N. S.
Makarov
,
R.
Simonutti
,
V. I.
Klimov
, and
S.
Brovelli
,
Nat. Nanotechnol.
10
,
878
(
2015
).
7.
W. H.
Weber
and
J.
Lambe
,
Appl. Opt.
15
,
2299
(
1976
).
8.
K.
Wu
,
H.
Li
, and
V. I.
Klimov
,
Nat. Photonics
12
,
105
(
2018
).
9.
R. R.
Lunt
,
Appl. Phys. Lett.
101
,
043902
(
2012
).
10.
Y.
Zhao
and
R. R.
Lunt
,
Adv. Energy Mater.
3
,
1143
(
2013
).
11.
H.
Li
,
K.
Wu
,
J.
Lim
,
H.-J.
Song
, and
V. I.
Klimov
,
Nat. Energy
1
,
16157
(
2016
).
12.
J.
Gutmann
,
M.
Peters
,
B.
Bläsi
,
M.
Hermle
,
H.
Zappe
, and
J. C.
Goldschmidt
,
Proc. SPIE
8111
,
81110K-1
81110K-9
(
2011
).
13.
J.
Gutmann
,
M.
Peters
,
B.
Bläsi
,
M.
Hermle
,
A.
Gombert
,
H.
Zappe
, and
J. C.
Goldschmidt
,
Opt. Express
20
,
A157
(
2012
).
14.
J. C.
Goldschmidt
,
M.
Peters
,
J.
Gutmann
,
L.
Steidl
,
R.
Zentel
,
B.
Bläsi
, and
M.
Hermle
,
Proc. SPIE
7725
,
77250S-1
77250S-11
(
2010
).
15.
I.
Rousseau
and
V.
Wood
,
Appl. Phys. Lett.
103
,
131113
(
2013
).
16.
S. K. E.
Hill
,
R.
Connell
,
C.
Peterson
,
J.
Hollinger
,
M. A.
Hillmyer
,
U.
Kortshagen
, and
V. E.
Ferry
,
ACS Photonics
6
,
170
(
2019
).
17.
I.
Coropceanu
and
M. G.
Bawendi
,
Nano Lett.
14
,
4097
(
2014
).
18.
D. R.
Needell
,
O.
Ilic
,
C. R.
Bukowsky
,
Z.
Nett
,
L.
Xu
,
J.
He
,
H.
Bauser
,
B. G.
Lee
,
J. F.
Geisz
,
R. G.
Nuzzo
,
A. P.
Alivisatos
, and
H. A.
Atwater
,
IEEE J. Photovoltaics
8
,
1560
(
2018
).
19.
N. D.
Bronstein
,
Y.
Yao
,
L.
Xu
,
E.
O’Brien
,
A. S.
Powers
,
V. E.
Ferry
,
A. P.
Alivisatos
, and
R. G.
Nuzzo
,
ACS Photonics
2
,
1576
(
2015
).
20.
J.
Bomm
,
A.
Büchtemann
,
A.
Fiore
,
L.
Manna
,
J. H.
Nelson
,
D.
Hill
, and
W. G. J. H. M.
van Sark
,
Beilstein J. Nanotechnol.
1
,
94
(
2010
).
21.
G.
Liu
,
H.
Zhao
,
F.
Diao
,
Z.
Ling
, and
Y.
Wang
,
J. Mater. Chem. C
6
,
10059
(
2018
).
22.
S.
Rogers
and
L.
Mandelkern
,
J. Phys. Chem.
61
,
985
(
1957
).
23.
M. D.
Gehlsen
and
F. S.
Bates
,
Macromolecules
26
,
4122
(
1993
).
25.
J. A.
Brydson
,
Plastics Materials
(
Butterworth-Heinemann
,
Woburn, MA
,
1999
).
26.
F. S.
Bates
,
G. H.
Fredrickson
,
D.
Hucul
, and
S. F.
Hahn
,
AIChE J.
47
,
762
(
2001
).
27.
H.
Rachapudy
,
G. G.
Smith
,
V. R.
Raju
, and
W. W.
Graessley
,
J. Polym. Sci., Polym. Phys. Ed.
17
,
1211
(
1979
).
28.
F.
Jing
,
M. R.
Smith
 III
, and
G. L.
Baker
,
Macromolecules
40
,
9304
(
2007
).
29.
K.
Morimoto
and
S.
Suzuki
,
J. Appl. Polym. Sci.
16
,
2947
(
1972
).
30.
M.
N’Diaye
,
F.
Pascaretti-Grizon
,
P.
Massin
,
M. F.
Baslé
, and
D.
Chappard
,
Langmuir
28
,
11609
(
2012
).
31.
W. D.
Drotning
and
E. P.
Roth
,
J. Mater. Sci.
24
,
3137
(
1989
).
32.
J.
González-Benito
,
E.
Castillo
, and
J. F.
Cruz-Caldito
,
Phys. Chem. Chem. Phys.
17
,
18495
(
2015
).
33.
J. F.
Pendleton
,
D. F.
Hoeg
, and
E. P.
Goldberg
, “
Novel heat resistant plastics from hydrogenation of styrene polymers
,” in
Polymerization Reactions and New Polymers
, edited by
N. A. J.
Platzer
(
American Chemical Society
,
Washington, D.C.
,
1973
), pp.
27
38
.
34.
S. F.
Hahn
, in
Modern Styrenic Polymers: Polystyrenes and Styrenic Copolymers
, edited by
J.
Scheirs
and
D.
Priddy
(
John Wiley & Sons
,
West Sussex, England
,
2003
), pp.
533
556
.
35.
Polymers: A Property Database
, edited by
B.
Ellis
and
R.
Smith
(
Chapman & Hall/CRC Press
,
Boca Raton, FL
,
2000
).
36.
D. A.
Hanifi
,
N. D.
Bronstein
,
B. A.
Koscher
,
Z.
Nett
,
J. K.
Swabeck
,
K.
Takano
,
A. M.
Schwartzberg
,
L.
Maserati
,
K.
Vandewal
,
Y.
van de Burgt
,
A.
Salleo
, and
A. P.
Alivisatos
,
Science
363
,
1199
(
2019
).
37.
N. D.
Bronstein
,
L.
Li
,
L.
Xu
,
Y.
Yao
,
V. E.
Ferry
,
A. P.
Alivisatos
, and
R. G.
Nuzzo
,
ACS Nano
8
,
44
(
2014
).
38.
O.
Chen
,
X.
Chen
,
Y.
Yang
,
J.
Lynch
,
H.
Wu
,
J.
Zhuang
, and
Y. C.
Cao
,
Angew. Chem., Int. Ed.
47
,
8638
(
2008
).
39.
C.
Pu
and
X.
Peng
,
J. Am. Chem. Soc.
138
,
8134
(
2016
).
40.
D. B.
Dement
,
M.
Puri
, and
V. E.
Ferry
,
J. Phys. Chem. C
122
,
21557
(
2018
).
41.
J. D.
Rancourt
,
Optical Thin Films: User Handbook
(
SPIE
,
Bellingham, Washington
,
1996
).
42.
C. F.
Bohren
and
D. R.
Huffman
,
Absorption and Scattering of Light by Small Particles
(
Wiley
,
1983
).
43.
ASTM Standard D1003-13,
1949
.
44.
T.
Kang
,
K.
Um
,
J.
Park
,
H.
Chang
,
D. C.
Lee
,
C.-K.
Kim
, and
K.
Lee
,
Sens. Actuators, B
222
,
871
(
2016
).

Supplementary Material