Transparent, near-infrared organic photovoltaic solar cells for window and energy-scavenging applications

We fabricate near-infrared absorbing organic photovoltaics that are highly transparent to visible light. By optimizing near-infrared optical-interference, we demonstrate power efﬁciencies of 1.3 (cid:2) 0.1% with simultaneous average visible transmission of (cid:3) 65%. Subsequent incorporation of near-infrared distributed-Bragg-reﬂector mirrors leads to an increase in the efﬁciency to 1.7 (cid:2) 0.1%, approaching the 2.4 (cid:2) 0.2% efﬁciency of the opaque cell, while maintaining high visible-transparency of (cid:3) 55%. Finally, we demonstrate that a series-integrated array of these transparent cells is capable of powering electronic devices under near-ambient lighting. This architecture suggests strategies for high-efﬁciency power-generating windows and highlights an application uniquely beneﬁting from excitonic electronics. 2011

We fabricate near-infrared absorbing organic photovoltaics that are highly transparent to visible light. By optimizing near-infrared optical-interference, we demonstrate power efficiencies of 1.3Ϯ 0.1% with simultaneous average visible transmission of Ͼ65%. Subsequent incorporation of near-infrared distributed-Bragg-reflector mirrors leads to an increase in the efficiency to 1.7Ϯ 0.1%, approaching the 2.4Ϯ 0.2% efficiency of the opaque cell, while maintaining high visible-transparency of Ͼ55%. Finally, we demonstrate that a series-integrated array of these transparent cells is capable of powering electronic devices under near-ambient lighting. This architecture suggests strategies for high-efficiency power-generating windows and highlights an application uniquely benefiting from excitonic electronics. © 2011 American Institute of Physics. ͓doi:10.1063/1.3567516͔ The low energy density of solar illumination necessitates deployment of solar technologies over large surface areas in order to capture enough of the sun's energy to offset a significant portion of nonrenewable energy consumption. The obstacle of large-area deployment could be overcome with development of a low-cost, 1 transparent, photovoltaic ͑PV͒ technology that can be integrated onto window panes in homes, skyscrapers, and automobiles, enhancing the functionality of already utilized transparent surfaces. Presently, window glass used in automobiles and in architectural installations is typically 70%-80% and 55%-90% transmissive, respectively, to visible light, 2 with the reduction in light transmission purposely introduced in the form of absorptive or reflective tinted windows. The additive transparent PV technology can similarly modify the glass transparency, but the nontransmitted light would be utilized for power generation. Although there may be opportunities for inorganic semiconductors in such PV applications, their limited mechanical flexibility, high module cost and, more importantly, the bandlike absorption spectrum limits their potential utility to transparent solar cells. In contrast, the excitonic character of organic and molecular semiconductors results in absorption spectra that are highly structured with absorption minima and maxima that is uniquely distinct from the bandabsorption of their inorganic counterparts. Previous efforts to construct semitransparent devices have focused on the use of thin active layers ͑or spatially segmented films͒ with light absorption focused in the visible spectrum and therefore have been limited to either low efficiencies Ͻ1% ͑Refs. 3 and 4͒ or low average visible-light transmissivity ͑AVT͒ of 10%-35%, 5-8 since both parameters cannot be simultaneously optimized. In this letter we demonstrate heterojunction organic PV ͑OPV͒ cells utilizing a molecular organic donor, chloroaluminum phthalocyanine ͑ClAlPc͒, and a molecular acceptor, C 60 , that show peak-absorption in the ultraviolet and near-infrared ͑NIR͒ ͑wavelength span of = 650-850 nm͒. Combining the OPVs with selective highreflectivity near-infrared mirror coatings we optimize device performance while also permitting high transmission of visible photopic light 9 through the entire device.
The OPVs are fabricated on glass substrates that were precoated with 150 nm thick, patterned indium-tin oxide ͑ITO͒ transparent anode with 15 ⍀ / sq sheet resistance ͑from Kintec Co.͒, ClAlPc ͑from TCI͒, and C 60 ͑from Sigma-Aldrich, sublimed͒ were purified once by vacuum train sublimation prior to loading, while bathocuproine ͑BCP, from Lumtec͒ and MoO 3 were used as purchased. MoO 3 ͑20 nm͒, ClAlPc ͑15 nm͒, C 60 ͑30 nm͒, BCP ͑7.5 nm͒, and a 100 nm thick Ag cathode were sequentially deposited via thermal evaporation at a rate of 0.1 nm/s. For the transparent devices, ITO cathodes were rf-sputtered directly onto the organic layers at a deposition rate of 0.005-0.03 nm/s at a low operating power of 7-25 W, with 10 SCCM ͑SCCM denotes cubic centimeter per minute at STP͒ Ar flow and sputtering chamber pressure of 6 mTorr. The cathode films were deposited through a shadow mask for single devices, defining a 1 mmϫ 1.2 mm active device area; series-integrated OPV cells were grown monolithically via 3-mask procedure to create anode-to-cathode connections of 12 subcells ͑0.7 cm 2 total active area͒. A distributed Bragg reflector ͑DBR͒ utilized as the transparent NIR mirror 10 was grown separately on quartz substrates via sputtering of seven alternating layers of TiO 2 and SiO 2 at the rate of 0.1 nm/s, with layer thickness selected to form a stop band centered around = 800 nm. Broad-band antireflection ͑BBAR͒ coatings 10 ͑from Knight Optical͒ precoated on their own quartz substrates ͑1-side͒ were attached to the DBR and substrate glass via index matching fluid. Transmission data of the assembled devices were obtained at normal incidence with a Cary Eclipse 5000 dual-beam spectrophotometer without reference samples.  Figure 1͑a͒ shows a schematic of the full transparent solar cells. The imaginary part of complex index of refraction, k, of the active layers is shown in Fig. 1͑b͒, along with the J-V curves for the ClAlPc-C 60 control and transparent cells in Fig. 1͑c͒. Because the absorption peak for ClAlPc is positioned in the NIR ͑ ϳ 740 nm͒, incorporation of a NIR reflecting mirror enables simultaneous optimization of the solar cell performance and visible-transmissivity as diagramed in Fig. 1͑a͒. The control device with a thick Ag cathode exhibits a power conversion efficiency ͑ p ͒ of 2.4Ϯ 0.2%, open circuit voltage ͑Voc͒ = 0.80Ϯ 0.02 V, short-circuit current density ͑Jsc͒ = 4.7Ϯ 0.3 mA/ cm 2 , and fill-factor ͑FF͒ = 0.55Ϯ 0.03, which is comparable to previous reports. 14 Replacing the Ag cathode of the control cell with 20 nm thick ITO cathode, the Jsc drops to 1.5Ϯ 0.1 mA/ cm 2 , FF drops to 0.39Ϯ 0.02, and the Voc decreases slightly to 0.69Ϯ 0.02 V leading to p = 0.5Ϯ 0.1%. The FF decreases due to an increase in series resistance from the thin ITO cathode that is observable in the J-V curve under forward bias in Fig. 1͑c͒. For thicker ͑120 nm thick͒ ITO cathode films, the series resistance diminishes and the FF increases to 0.46, close to the value for the control cell, as shown in Fig. 2. The slight drop in Voc, independent of ITO thickness, is likely due to a change in the cathode-anode work function offset. Nonetheless, it is notable that utilizing ITO as both anode and cathode there is enough deposition anisotropy in the work function to support this large Voc and is likely assisted by the large work function of the MoO 3 layer. 15 The Jsc decreases as the cathode is switched from Ag to ITO due to the reduced light reflection at the cathode, which reduces the total absorption across the spectrum in the active layers. As the ITO cathode thickness is increased, and the NIR mirror is introduced, the transparent device photocurrent increases threefold, reaching maximum performance for 120 nm thick ITO ͓see Fig. 2͑b͔͒. Simultaneously, threefold increase in p is measured. Modeling the optical modes in these multilayer dielectric stacks, we find a similar trend in the calculated device performance, determined by the interference of the light inside the device with the light reflected from the ITO cathode. Comparing the transmission of the ITO-only devices in Fig. 3 shows that the absorption for the thinnest ITO cathode ͑20 nm͒ and the optimized thicknesses ITO cathode ͑120 nm͒ appears equivalent. Inspection of the simulations, however, indicates that the NIR field distribution is shifted from within the ITO anode to the ClAlPc active layer as the ITO cathode thickness increases, so that the total transmission appears the same even though the active layer absorption changes substantially. 12 Despite the significant impact on the photocurrent, the AVT shows little variation with thickness of the ITO cathode ͓see Fig. 2͑a͔͒. The optical model predicts AVT values that are in close agreement with experimental data. We note that the high reflectivity of 99% between = 695 and 910 nm, also makes these devices useful for simultaneous NIR rejection in architectural cooling. Additionally, the use of the BBAR coatings next to the DBR ͑outcoupling͒ and below FIG. 1. ͑Color online͒ ͑a͒ Schematic of the transparent photovoltaic architecture. The simple transparent structure without the NIR reflector ͑ITO-only͒ is highlighted with the bracket. ͑b͒ Imaginary part of the complex index of refraction, k, for the active layers with the molecular structures inset: the ClAlPc donor, C 60 acceptor, and sputtered ITO ͑black line͒ cathode. ͑c͒ J-V curves measured under 0.8 sun illumination and ͑d͒ EQE for the control cell and for the transparent devices with the NIR reflector as a function of ITO cathode thickness. the substrates ͑incoupling͒, results in a concomitant increase in the measured AVT by ϳ4% -6% and the quantum efficiency by ϳ2%-3%.
To highlight the transparency of the fully assembled device we show the solar cell array in front of a picture of the "Teton mountains" in Fig. 4. Both picture-detail and colorclarity ͑blue-green͒ are minimally disrupted so that details of the device array pattern are even difficult to discern. Further, we show that integrating these cells in series, we can power a small liquid crystal display ͑LCD͒ clock under nearambient lighting ͓see Fig. 4͑c͔͒, where the red-white MIT logo can simultaneously be seen through the device.
Optimizing the ITO cathode thickness in the ITO-only devices, we obtain power conversion efficiency of 1.3Ϯ 0.1%, with a simultaneous AVT of 65Ϯ 3%. Incorpo-ration of the NIR reflector and BBAR coatings with the optimized ITO thickness ͓see Fig. 2͑a͔͒, improves the power conversion efficiency to 1.7Ϯ 0.1% with an AVT of 56Ϯ 2%. With the NIR mirror, the increase in the power conversion efficiency stems from additional NIR photocurrent in the ClAlPc layer where the EQE shows a near doubling of the peak ClAlPc EQE from 10% to 18% ͓see Fig. 1͑d͔͒. Switching from planar to bulk-heterojunctions in these structures or incorporating tandem stacking of subcells with active layer absorption deeper into the infrared, 16 efficiencies of Ͼ2% -3% may be possible for this material set with nearly identical visible transmission.
In conclusion, we have demonstrated near-infrared absorbing, transparent planar organic solar cells with a maximum power conversion efficiency of 1.7Ϯ 0.1% and average visible transmission of Ͼ55Ϯ 2%, which is sufficiently transparent for incorporation on architectural glass. Here we exploit the excitonic character of organic semiconductors to produce photovoltaic architectures not easily accessible via inorganic semiconductors. By positioning the active layer absorption selectively in the NIR, we are able to optimize the architecture using a NIR reflector composed of a DBR mirror centered at 800 nm that results in a transparent solar cell efficiency approaching that of the nontransparent control cell. Ultimately these devices provide a guide for achieving high-efficiency and high transparency solar cells that can be utilized in windows to generate power, reduce cooling costs, and scavenge energy.