Color-neutral, semitransparent organic photovoltaics for power window applications

Significance

We demonstrate a semitransparent organic photovoltaic cell that achieves a power conversion efficiency of 10.8% and visible transparency of ∼50% using a nonfullerene acceptor (NFA) featuring strong near-infrared (NIR) absorption and simple synthesis. Contrary to expectations, stronger NIR absorption and closer molecular packing are obtained by employing an additive in these partially, instead of fully fused, rigid NFAs. By combining NIR-absorbing material sets with an optical outcoupling structure as well as transparent electrode, we overcome the trade-offs between efficiency, transparency, and device appearance. These results surpass other semitransparent solar cell technologies based on organic and other thin-film materials systems, showing a promising future for ST-OPVs as power-generating windows and other solar energy harvesting applications.

Abstract

Semitransparent organic photovoltaic cells (ST-OPVs) are emerging as a solution for solar energy harvesting on building facades, rooftops, and windows. However, the trade-off between power-conversion efficiency (PCE) and the average photopic transmission (APT) in color-neutral devices limits their utility as attractive, power-generating windows. A color-neutral ST-OPV is demonstrated by using a transparent indium tin oxide (ITO) anode along with a narrow energy gap nonfullerene acceptor near-infrared (NIR) absorbing cell and outcoupling (OC) coatings on the exit surface. The device exhibits PCE = 8.1 ± 0.3% and APT = 43.3 ± 1.2% that combine to achieve a light-utilization efficiency of LUE = 3.5 ± 0.1%. Commission Internationale d’eclairage chromaticity coordinates of (0.38, 0.39), a color-rendering index of 86, and a correlated color temperature of 4,143 K are obtained for simulated AM1.5 illumination transmitted through the cell. Using an ultrathin metal anode in place of ITO, we demonstrate a slightly green-tinted ST-OPV with PCE = 10.8 ± 0.5% and APT = 45.7 ± 2.1% yielding LUE = 5.0 ± 0.3% These results indicate that ST-OPVs can combine both efficiency and color neutrality in a single device.

Transparent solar cells are attractive energy-conversion devices for integration onto window panes, skylights, and building facades, providing an opportunity for increasing solar energy harvesting on building surfaces (1–12). Compared to inorganic semiconductors, the narrow excitonic absorption bands of organic semiconductors offer opportunities for organic photovoltaics (OPVs) as power-generating windows since many organics selectively absorb outside of the visible wavelengths. Semitransparent OPV (ST-OPV) efficiencies have hovered around 7% for the last few years, which is deemed too low for substantial market penetration. Additionally, only a few ST-OPVs have achieved visible transparency ∼50%, which is critical for many power-window applications. The performance of ST-OPVs is ultimately limited by the trade-off between power-conversion efficiency (PCE) and average photopic transmission (APT; see Methods), which is the perceived transparency of the appliance. As a consequence, to our knowledge their light-utilization efficiencies, LUE = PCE × APT, have recently reached 4.3% (SI Appendix) (13–17). Another challenge faced by an ST-OPV is that the most efficient devices exhibit an unwanted tint (14, 18–22). Except for relatively few instances, it is important to develop power-generating windows with aesthetically acceptable neutral colors that are easily applied in the broadest possible applications.

To date, demonstrations of efficient, color-neutral ST-OPVs have primarily focused on designing materials with strong near-infrared (NIR) absorption (6, 14, 15, 18, 20, 23–28), incorporating multijunction device structures to minimize thermalization losses (14, 28–31), and employing antireflection coatings (ARC) or aperiodic dielectric reflectors (ADR) to enhance absorption (3, 19, 32, 33). For example, Zhang et al. developed a color-neutral ST-OPV with an LUE = 1.95% by using ternary blends of two polymer donors with an NIR nonfullerene acceptor (NFA) to balance absorption in the visible region (34). Recently, our group demonstrated a nearly color-neutral ST-OPV with a PCE = 5.8%, APT = 45%, which combine to LUE = 2.56% under 1 sun, AM 1.5G spectral illumination by employing a bilayer outcoupling (OC) coating on the exit surface (13). This device combined a two-component NFA blend absorbing between wavelengths of 600 and 900 nm with an ultrathin metal anode. The absorption of the active layer covers wavelengths λ < 900 nm, leaving substantial solar energy at longer wavelengths unused. Furthermore, the thin metal films are rough, and their transmittance is spectrally dependent across the visible, often lending a tinted appearance.

Here, we demonstrate an ST-OPV that achieves PCE = 10.8 ± 0.6% and APT = 45.7 ± 2.1%, leading to LUE = 5.0 ± 0.3. The device employs an NFA molecule with strong NIR absorption, that requires only a few steps in its synthesis. Contrary to expectations, strong intermolecular π–π interactions and close molecular packing are observed in these simple NFAs that feature partially covalently fused ring backbones rather than rigid, fully fused rings (Fig. 1). By combining the elements of NIR-absorbing material sets with an OC structure on the exit surface and transparent electrodes, we overcome the trade-offs between efficiency, transparency, and device appearance. A color-neutral ST-OPV using a transparent indium tin oxide (ITO) anode exhibits PCE = 8.1 ± 0.3%, APT = 43.3 ± 1.5%, and LUE = 3.5 ± 0.1%. Commission Internationale d’eclairage (CIE) chromaticity coordinates of (0.38, 0.39), a color-rendering index of CRI = 86, and a correlated color temperature of CCT = 4,143 K are obtained for simulated AM1.5 illumination transmitted through the cell. These results suggest that there is a promising future for ST-OPVs employed building integrated applications.

Fig. 1.

(A) Molecular structural formulae of the SBT-FIC, A078, and A134. (B) UV-vis absorption spectra of SBT-FIC, A078, and A134 in toluene solution, and (C) in thin films.

Results

The molecular structural formulae of the three NFAs studied are shown in Fig. 1A. One NFA, SBT-FIC, features a fully fused molecular backbone. The other two NFAs [A078 (22) and A134] with partially fused cores are isomers of SBT-FIC comprising four thiophenes, two cyclopentadienes, and one benzene ring. The details of molecular design and the synthetic routes to these molecules are provided in SI Appendix, Schemes S1–S3. In contrast to SBT-FIC that requires 10 synthetic steps starting from 4,8-dihydrobenzo[1,2b:4,5-b′]dithiophene-4,8-dione, the costs of A078 and A134 are potentially lower since they entail only 4–6 synthetic steps starting from 2,5-dibromoterephthalic acid diethyl ester. In addition, the low synthesis complexity, high yield, less toxic precursors, and inexpensive starting materials enable A078 and A134 to be economically produced at large scale.

The ultraviolet-visible (UV-vis) absorption spectra of the NFAs are shown in Fig. 1 B and C. Surprisingly, thin films of A078 and A134 exhibit significant bathochromic shifts of ∼135 nm compared to SBT-FIC with an absorption peak at λ_max = 900 nm. Contrary to expectations, closer molecular packing leading to the spectral shifts is obtained when a rotation is imposed by the single bond between indaceno[1,2-b:5,6-b′]dithieno[3,2-b]thiophene (IDT) and the flanking thiophene in A078 and A134. Density-functional theory (DFT) calculations at the generalized gradient approximation/triple zeta polarized set level show that the S···S distance between 2-ethylhexanethiol group on the thiophene and the IDT (3.18 Å) unit is far shorter than the sum of the van der Waals radii of the two sulfur atoms (3.68 Å) (35). This reveals that noncovalent S···S interactions rigidify the conjugated structure (36), resulting in a reduction of the torsion angle between the thiophene and IDT from 20° to 2.5°(SI Appendix, Fig. S1).

Cyclic voltammetry in SI Appendix, Fig. S2 gives the highest occupied (HOMO) and lowest unoccupied (LUMO) molecular orbital energies of E_H = −5.81 (±0.02) and E_L = −4.15 (±0.03) eV for SBT-FIC, −5.58 (±0.02) and −4.06 (±0.03) eV for A078, and −5.54 (±0.02) and −4.05 (±0.03) eV for A134, respectively. Both A078 and A134 show a lower HOMO–LUMO energy gap (∼1.40 eV) than SBT-FIC (∼1.65 eV), which is consistent with optical measurements. Moreover, A078 and A134 exhibit shallower LUMO energies compared with SBT-FIC, which leads to an improvement of V_OC in OPVs.

The NFAs blended with PCE-10 were employed in OPVs with the structure ITO/ZnO (30 nm)/active layer (∼100 nm)/MoO₃ (20 nm)/Ag (100 nm) (see fabrication details in Methods). Their current density–voltage (J−V) characteristics are plotted in Fig. 2A, with a summary of performance measured under 1 sun intensity, simulated AM1.5G solar spectral illumination in Table 1. Here, PCE = 13.0 ± 0.4% is achieved in the A078-based device, with V_OC = 0.75 ± 0.01 V, J_SC = 24.8 ± 0.7 mA cm⁻², and FF = 0.70 ± 0.04. In contrast, the A134-based OPV exhibited PCE =7.6 ± 0.2% with V_OC = 0.75 ± 0.01 V, J_SC = 16.7 ± 0.5 mA cm⁻², and FF = 0.61 ± 0.03. For the PCE-10:SBT-FIC device, PCE = 7.8 ± 0.3% with V_OC = 0.70 ± 0.01 V, J_SC = 17.2 ± 0.7 mA cm⁻², and FF = 0.65 ± 0.02. Interestingly, the 1-phenylnathalene (PN) additive results in dramatic improvements in efficiency of the A078 and A134 devices compared with SBT-FIC, which is due to the improved molecular packing of the A078 and A134 as well as more favorable molecular orientations in the blends (see below, this section). Note that the PCE-10:A134 device shows a lower PCE compared to the PCE-10:A078 OPV due to the crystallinity of A134, leading to its lower solubility.

Fig. 2.

(A) Current density (J)–voltage (V) characteristics, and (B) EQE spectra of organic photovoltaic cells based on PCE-10: SBT-FIC (1:2, wt/wt), PCE-10:A078 (1:2, wt/wt), and PCE-10:A134 (1:2, wt/wt).

Table 1.

Operating characteristics of OPVs under simulated of AM 1.5G, 100 mW cm⁻², illumination

Device*	JSC†, mA/cm2	VOC, V	FF	PCE, %	Eloss, eV
PCE-10:SBT-FIC (wo/PN)	18.1 ± 0.6 (17.2)	0.70 ± 0.01	0.62 ± 0.03	7.9 ± 0.3	0.79
PCE-10:SBT-FIC (w/PN)	17.2 ± 0.7 (16.5)	0.70 ± 0.01	0.65 ± 0.02	7.8 ± 0.3	0.79
PCE-10:A078 (wo/PN)	22.2 ± 1.0 (21.3)	0.76 ± 0.01	0.56 ± 0.02	9.5 ± 0.3	0.55
PCE-10:A078 (w/PN)	24.8 ± 0.7 (24.3)	0.75 ± 0.01	0.70 ± 0.04	13.0 ± 0.4	0.55
PCE-10:A134 (wo/PN)	18.5 ± 0.5 (17.4)	0.75 ± 0.01	0.45 ± 0.03	6.2 ± 0.3	0.54
PCE-10:A134 (w/PN)	16.7 ± 0.5 (16.1)	0.75 ± 0.01	0.61 ± 0.03	7.6 ± 0.2	0.54

^*The values in parentheses are calculated from the integral of the EQE spectra.

^†The average value is based on measurement of 16 devices with donor: acceptor blending ratio of 1:2.

Fig. 2B shows the external quantum efficiency (EQE) spectra of the several devices. The significant improvement in J_SC for the A078 compared to the SBT-FIC OPV is attributed to its ∼200-nm absorption redshift that provides solar spectral coverage further into the NIR. The long-wavelength cutoff of A078 and A134 at λ = 1,000 nm is compared to ∼800 nm for SBT-FIC. The EQE of the A078 OPV reaches 80%, between λ = 700 and 900 nm, while leaving a transparency window between the visible wavelengths of 400 and 650 nm. The integrated J_SC are shown in Table 1, which are consistent with the solar simulation measurements.

To better understand the performance of these devices, the absorption profiles of the NFA neat films and PCE-10:NFAs blends with and without the PN additive are shown in SI Appendix, Fig. S3 and Fig. 3 A–C. The PCE-10:SBT-FIC film absorption shows little change by employing PN. In contrast, a new, pronounced aggregation peak around 900 nm is found in both PCE-10:A078 and PCE-10:A134 blends, indicating that the additive enhances intermolecular π–π interactions on partially fused acceptors rather than on the polymer donor.

Fig. 3.

(A) UV-vis absorption spectra of SBT-FIC and PCE-10:SBT-FIC (1:2, wt/wt) blend films. (B) A078 and PCE-10:A078 (1:2, wt/wt) blend films. (C) A134 and PCE-10:A134 (1:2, wt/wt) blend films with and without the PN additive. (D) Corrected polar (100) diffraction peaks and (E) in-plane (black line) and out-of-plane (red line) X-ray scattering patterns extracted from GIWAXS data of PCE-10:SBT-FIC, PCE-10:A078, and PCE-10:A134 blends with and without the PN additive.

The morphological properties are investigated using grazing-incidence wide-angle X-ray scattering (GIWAXS), with a summary of parameters obtained in Table 2 and SI Appendix, Fig. S2. A078 shows a broad (100) diffraction peak at 0.31 Å⁻¹ with a lamellar coherence length of L_c = 7.5 nm, while a narrower and sharper diffraction peak (100) at 0.36 Å⁻¹ is observed in A134 with increased L_c = 15 nm (SI Appendix, Fig. S4). This suggests A134 has increased ordering compared to A078, afforded by replacing the bulk p-hexylphenyl sidechain with the compact linear alkyl chains (37). On the other hand, SBT-FIC shows a diffraction peak at 0.34 Å⁻¹ with the smallest lamellar coherence length of L_c = 3.7 nm due to its amorphous nature. The (010) diffraction peaks of PCE-10:A078 and PCE-10:A134 in Fig. 3E at 1.79 and 1.82 Å⁻¹ (due to NFAs) are shifted and show an increased coherence length (24 vs. 52 Å, for A078) and (30 vs. 63 Å, for A134) when employing the PN additive. In contrast, L_c of PCE-10 is unchanged, which further confirms that morphological differences arise primarily from the NFAs rather than the donor. There is no obvious diffraction peak and coherence length variation for both the donor and acceptor in PCE-10:SBT-FIC blends by employing PN. These results explain the significant improvement in PCE for the A078 and A134 OPV by using the additive, leading to improved stacking of large and ordered aggregates compared to SBT-FIC. Additionally, a dependence on molecular orientation from edge-on to face-on is found in using PN. For PCE-10:A078, the face-on/edge-on ratio extracted from the pole figures of (100) peaks increases from 2.37 to 3.64 (Fig. 3D). Since the face-on orientation is favorable for intermolecular charge transport, it helps to describe why the highest efficiency is achieved in the A078 device.

Table 2.

Morphological parameters extracted from GIWAXS measurements

Device	Peak*, Å−1		π–π stacking distance, Å		FWHM†, Å−1		Coherence length, Å		Face-on/edge-on ratio
	Donor	NFA	Donor	NFA	Donor	NFA	Donor	NFA
PCE-10:SBT-FIC (wo/PN)	1.70	1.61	3.69	3.90	0.26	0.56	24	11	0.65
PCE-10:SBT-FIC (w/PN)	1.70	1.59	3.69	3.95	0.27	0.56	23	11	0.71
PCE-10:A078 (wo/PN)	1.67	1.79	3.76	3.51	0.53	0.26	12	24	2.37
PCE-10:A078 (w/PN)	1.70	1.83	3.69	3.43	0.50	0.12	13	52	3.64
PCE-10:A134 (wo/PN)	1.69	1.82	3.72	3.45	0.50	0.21	13	30	2.34
PCE-10:A134 (w/PN)	1.70	1.87	3.69	3.36	0.47	0.10	13	63	2.57

^*The value extracted from the (010) peak.

^†Full width at half-maximum.

The A078 OPV is further exploited by ST-OPVs with the structure ITO/ZnO (30 nm)/PCE-10: A078 (95 nm)/MoO₃ (20 nm)/Ag (16 nm). The J-V, optical transmission, and EQE spectral characteristics are shown in Fig. 4, with results summarized in Table 3. The ST-OPV showed the LUE = 2.8 ± 0.1%, with PCE = 11.0 ± 0.7% and APT = 25.0 ±1.3%. Although PCE > 10% is obtained, the low APT limits its applications for architectural glass that requires APT ∼50% (1). To solve this problem, a structure was designed to control the device optical properties to achieve maximum transmission in the visible while reflection is maximized in the NIR. An optical OC coating consisting of four layers: CBP (35 nm, index of refraction, n_CBP = 1.90)/MgF₂ (100 nm, n_MgF2 = 1.38)/CBP (70 nm)/MgF₂ (45 nm) was deposited onto the Ag anode surface, and an ARC consisting of a bilayer of 120-nm-thick MgF₂ and 130 nm of low refractive index SiO₂ (n_SiO2 = 1.12) (38) was deposited onto the distal surface of the glass substrate. The ST-OPV with an OC and ARC shows APT increases from 25.0 ± 1.3% to 45.7 ± 2.1%, representing a nearly 80% improvement than that of a cell without these layers. As shown in Fig. 4C, the PCE of the ST-OPV is almost as same as its initial value, with only a slight decrease in J_SC (20.4 ± 0.8 vs. 20.9 ± 1.2 mA cm⁻²). Hence, the LUE = 5.0 ± 0.3% is achieved with the OC and ARC coatings, which to our knowledge, is the highest reported efficiency for ST-OPVs.

Fig. 4.

(A) Schematic of the semitransparent device showing optimized layer thicknesses and compositions. (Right) Detailed layer structures of the OC and AR layers. (B) Current density vs. voltage characteristics, (C) EQE spectra, and (D) measured optical transmission and reflection of the optimized semitransparent cells with and without the OC and ARC layers.

Table 3.

Operating characteristics of semitransparent, neutral-colored OPVs

Device*	JSC, mA/cm2	VOC, V†	FF	PCE, %	APT, %	LUE, %	CIE	CCT
Ag wo/OC and ARC	20.9 ± 1.2	0.75	0.70 ± 0.03	11.0 ± 0.7	25.0 ± 1.3	2.8 ± 0.1	(0.27, 0.34)	9,021
Ag w/OC and ARC	20.4 ± 0.8	0.75	0.70 ± 0.03	10.8 ± 0.5	45.7 ± 2.1	5.0 ± 0.1	(0.33, 0.39)	5,585
ITO wo/OC and ARC	14.3 ± 0.5	0.73	0.68 ± 0.04	7.1 ± 0.4	46.7 ± 1.0	3.3 ± 0.1	(0.34, 0.40)	5,266
ITO w/OC and ARC	16.3 ± 0.4	0.73	0.68 ± 0.04	8.1 ± 0.3	43.3 ± 1.5	3.5 ± 0.1	(0.38, 0.39)	4,143

^*The average value is based on measurement of eight devices.

^†Error is ±0.01 V.

The device appearance was examined using AM1.5G simulated solar illumination. The transmitted light of the OC and ARC-coated device has 1931 CIE chromaticity coordinates of (0.33, 0.39) with CCT = 5,585 K. Note that the high reflectivity of the ultrathin Ag cathode at λ > 600 nm gives the device a green tint shown in SI Appendix, Fig. S5, which is apparent from its color coordinates. In contrast to Ag, ITO has a higher transparency with a flat in transmission spectrum across the visible. Using an ITO cathode and anode results in a more neutral hue. The ITO-based ST-OPVs with the following structure: MgF₂ (120 nm)/ITO glass/ZnO (30 nm)/PCE-10: A078 (105 nm)/MoO₃ (20 nm)/sputtered ITO (140 nm)/MgF₂ (145 nm)/MoO₃ (60 nm)/MgF₂ (190 nm)/MoO₃ (105 nm) exhibit J-V, transmission, and EQE spectral characteristics in Fig. 5 and SI Appendix, Fig. S6, with results summarized in Table 3. Compared to the Ag-based ST-OPV, the ITO-based device shows differences in FF and V_OC due to its higher work function and sheet resistance (∼50 Ω/sq). The largest differences are found in J_SC and PCE. As the device becomes increasingly transparent, the reflection from the ITO anode into the thin active region is decreased, eliminating the double pass of photons. To minimize the loss of low-energy photons, the OC coating is designed with a transmission maximum in the visible while being more reflective at longer wavelengths. Therefore, the device with the OC has a 15% higher J_SC and PCE compared with the uncoated ITO device, although the visible transparency is nearly unchanged. The OC-coated ITO device exhibits an LUE = 3.5 ± 0.1% with PCE = 8.1 ± 0.3% and APT = 43.3 ± 1.5%, and it has a near-neutral appearance. Fig. 5C shows the MacAdam ellipses drawn along the Planckian locus in the 1931 CIE color space. The chromaticity coordinates of (0.38, 0.39) are achieved in the OC-coated ITO device, shown by the blue boxes used for binning white light-emitting diode illumination sources. Moreover, this ST-OPV achieves a color-rendering index of CRI = 86 and a CCT = 4143 K. This high CRI indicates that illumination through the OPV window accurately renders the color of an object (Fig. 5D). The transmission spectra are dependent on light incidence angles >60° from normal, which can be adjusted for each latitude of window application by appropriate design of the optical coatings in SI Appendix, Fig. S7.

Fig. 5.

(A) Current density vs. voltage characteristics, (B) optical transmission and reflection of ITO-based semitransparent cells with and without the OC and ARC layers, (C) MacAdam “ellipses” along the Planckian locus. The black boxes are the American National Standards Institute, ANSI C78.377 standard for variations of an acceptable lighting source at a particular CCT) and the blue boxes show bins used to group illuminants whose CCT and CRI fit within approximately a three-step ellipse. Chromaticity coordinates of the transmission spectra of the ITO cathode device with an OC and ARC (blue cross), ITO device without OC and ARC (red circle), 16-nm Ag device with OC and ARC (orange triangle) using an AM1.5G solar reference input spectrum. (D) Photograph of the outdoor image through the (Left) ultrathin Ag and (Right) ITO semitransparent device.

Discussion

The foregoing results point to a trade-off between transparency and efficiency that can be minimized by the appropriate choice of absorbing materials, transparent contacts, and optical coatings. From these results, we can estimate the performance achievable for ST-OPVs based on thermodynamic limitations for ideal OPVs calculated by Giebink et al. (39). To arrive at a practical estimate, we assume that EQE = 90% is constant across the visible spectrum up to the optical energy gap of the absorbing layer, and FF = 0.75––values that are close to the state-of-the-art OPV performance. The E_loss is defined as the difference between the lowest absorbed photon energy and V_OC (40). Fig. 6A plots the calculated efficiency as a function of the optical energy gap with different E_loss (41). The visible transparency of T = 50% (T₅₀) provides a neutral density appearance that is typical of many windows used in residences and commercial buildings, and T = 100% (T₁₀₀) corresponds to a completely transparent fixture. Interestingly, compared to the T₅₀ device, the efficiency of T₁₀₀ is not significantly reduced. For example, for E_loss = 0.5 eV and an energy gap of E_g = 1.4 eV, we obtain PCE = 15.5% for T = 50%, and 10.5% for T = 100%. This is indicative of the disproportionate amount of solar flux available in the NIR, and the importance of employing cells that are designed to preferentially harvest long-wavelength radiation. Indeed, the ease of tailoring cell transparency is a particular advantage of OPVs compared to inorganic thin-film solar cells. Therefore, OPVs with appropriate optical designs can achieve a combination of both high transparency and efficiency.

Fig. 6.

Calculated PCE vs. optical energy gap with energy loss as parameters for a (A) cell transparency of T = 50% and (B) T = 100% at wavelengths <650 nm.

Another finding is the remarkable bathochromic shift of the absorption spectra of partially fused compared to fully fused molecules achieved by employing a solvent additive. As previously, molecules with rigid and planar conformations allow parallel p-orbital interactions to extend their effective conjugation and facilitate π-electron delocalization between molecules (42, 43). This, in turn, leads to a decrease in the bond-length alternation and reduction of the energy gap, E_g (44). However, a lower E_g is obtained in molecules (A078 and A134) with apparent rotational disorder rather than rigid molecules with fully fused rings (SBT-FIC). DFT calculations show that intermolecular S···S interactions in A078 and A134, result in a larger torsion angle than otherwise would exist between the central fused-ring core and the flanking thiophenes. This allows for substantial overlap between neighboring molecules in the solid state, giving rise to more ordered molecular packing and a reduced E_g. To test the importance of these noncovalent interactions, we studied three additional materials systems in SI Appendix, Fig. S8 and Table S3 (23, 44–48). Similar to the S···S interaction, the O···S interaction in IEICO-4F (23) also shows a bathochromic shift of absorption compared to the fully fused molecule BT-FIC (46). In contrast, IEIC (48) with a 2-ethylhexanethiol group leads to a higher rotational disorder and larger E_g compared to the covalent rigid molecule CBT-IC (47). These results are consistent with the conclusion that noncovalent conformational locks provide a driving force to planarize and rigidify π-conjugated backbones that lead to a reduction in band gap (49–51).

With rapid development of new NFAs, ST-OPVs exhibit considerable potential for achieving high performance and a pleasing appearance. However, their relatively low stability remains a barrier for their commercialization. For example, the most stable NFA devices reported degrade within only a few years (extrapolated to 80% of their initial PCE) (52). This is compared to the extrapolated intrinsic lifetimes of fullerene-based cells of thousands of years, as recently reported by our group (53). Nevertheless, stretchable, foldable, lightweight and mechanically resilient form factors enabled by OPVs will encourage continued developments in reliability as these devices open new applications for attachment to irregular surfaces on textiles and robotic machinery (54, 55).

Conclusions

We demonstrated an ST-OPV with PCE = 10.8 ± 0.6%, APT = 45.7 ± 2.1%, and LUE = 5.0 ± 0.3% by utilizing a partly fused NFA-based, NIR absorbing donor–acceptor in a bulk heterojunction. The S···S interaction between 2-ethylhexanethiol group and central IDT unit planarizes and rigidifies the π-conjugated molecules. Consequently, the partly fused NFAs aggregate, leading to a reduced E_g with absorption peak at λ > 900 nm. Interestingly, the intermolecular π–π interactions of the partly fused NFAs are enhanced by employing a PN additive that contributes to an increased J_SC and FF. Furthermore, with the application of a transparent ITO anode, we demonstrated an ST-OPV with both neutral color and high efficiency. The optimized device exhibits PCE = 8.1 ± 0.3%, APT = 43.3 ± 1.5%, and LUE = 3.5 ± 0.1%, with CIE coordinates of (0.38, 0.39), CCT = 4,143 K, and CRI = 86.

Methods

Materials.

All devices were grown on patterned ITO substrates with sheet resistances of 15 Ω/sq. The acceptors (4,4,10,10-tetrakis(4-hexylphenyl)-5,11-(2-ethylhexylsulfanyl)-4,10-dihydro-dithienyl[1,2-b:4,5b′]benzodi-thiophene-2,8-diyl)bis(2-(3-oxo-2,3-dihydroinden-5,6-difluro-1-ylidene)malononitrile) SBT-FIC; 2-((E)2-((5-(7-(5-(((Z)-1-(dicyanomethylene)-5,6-difluoro-3-oxo-1,3-dihydro-2H-inden-2ylidene)methyl)-3-((2-ethylhexyl)thio)-thiophen-2-yl)-4,4,9,9tetrakis(4-hexylphenyl)4,9-dihydrosindaceno[1,2b:5,6b′]dithiophen-2-yl)-4-((2ethylhexyl)thio)thiophen-2yl)methylene)-5,6-difluoro-3-oxo-2,3-dihydro-1H-inden-1ylidene)malononitrile A078; 2-((E)2-((5-(7-(5-(((Z)-1-(dicyanomethylene)-5,6-difluoro-3-oxo-1,3-dihydro-2Hinden-2ylidene)methyl)-3-((2-ethylhexyl)thio)thio-phen-2-yl)-4,4,9,9-tetrakis-hexyl4,9-dihydrosindaceno[1,2b:5,6b′]dithiophen-2-yl)-4-((2-ethylhexyl)thio)thiophen-2yl)methylene)-5,6-difluoro-3-oxo-2,3-dihydro-1H-inden-1ylidene)malononitrile A134 were synthesized in our laboratories. Other materials were purchased from commercial suppliers: MoO₃ (Acros Organics); 4,4′-Bis(N-carbazolyl)-1,1′- biphenyl (CBP, Luminescence Technology Corp.); MgF₂ (Kurt J. Lesker Corp.); Ag (Alfa Aesar); and Poly[4,8-bis(5-(2-ethylhexyl)thiophen-2-yl)benzo[1,2-b;4,5-b′] dithiophene-2,6-diyl -alt-(4-(2-ethylhexyl)-3-fluorothieno[3,4-b]thiophene-)-2-carboxylate-2–6-diyl)] (PCE-10, 1-Material).

Device Fabrication and Characterization.

Prepatterned ITO-coated glass substrates with sheet resistances of 15 Ω/sq were purchased from Lumtec. The substrates were cleaned using a detergent (tergitol solution) and solvents followed by CO₂ snow cleaning and exposed to UV ozone for 15 min before film growth. The ZnO layer precursor solution was spin cast and then thermally annealed at 150 °C for 45 min in air. The active layer, PCE-10:NFA (1:2 wt/wt), was spin-coated from toluene solution (total concentration 14 mg mL⁻¹), followed by the 1-phenylnathalene additive (0.5%). The vacuum-deposited MoO₃ and Ag were grown at ∼0.6 Å/s in a vacuum chamber with a base pressure of 2 × 10⁻⁷ Torr. The areas of the OPVs are defined by the patterned ITO cathode and the Ag anode deposited through a 50-µm-thick shadow mask, resulting in a device area of 0.04 cm². ST-OPVs used the same fabrication procedures as the opaque cells. The ITO anode was deposited at a rate of 1 Å/s from a In₂O₃:Sn target by direct current magnetron sputtering in a chamber with a base pressure of 1 × 10⁻⁶ Torr. The doping density was varied by adjusting the flow of oxygen into the chamber while the substrate holder was rotated at 10 rpm. The OC coating was grown by vacuum thermal evaporation (VTE) in a chamber with a base pressure of 1 × 10⁻⁷ Torr at 1 Å/s for MgF₂, 0.6 Å/s for CBP and 0.6 Å/s for MoO₃. The ARC was grown onto the glass substrate after the devices were completed. MgF₂ was deposited by VTE at a rate of 1 Å/s, and the SiO₂ was grown by electron-beam deposition on the substrate held at an angle of 85° to the source. Glancing incidence deposition results in a porous film with a refractive index of 1.1.

The current density–voltage (J-V) characteristics and EQEs of the cells were measured in a ultrapure N₂ atmosphere. The EQE measurements were performed using a 200-Hz chopped, monochromated, and focused beam from a Xe lamp. The beam is focused to underfill the device area. The current from the devices and from a National Institute of Science and Technology-traceable Si reference detector were recorded using a lock-in amplifier. Light from a Xe lamp filtered to achieve a simulated AM 1.5G spectrum (American Society for Testing and Materials, ASTM G173-03) was used as the source for J-V measurements. The spectral mismatch factors are calculated to be from 1.006 to 1.009. The lamp intensity is varied using neutral density filters and calibrated by a National Renewable Energy Laboratory certified Si reference cell. The cells were measured using a 3.24-mm² metal mask at intensities from 0.001 to 1 sun (100 mW/cm²). ST-OPVs were measured from the ITO side with no object behind the cells. Errors account for measurement variations from three or more cells. There is also a systematic error of 5% for J_SC and PCE.

Optical and Electrochemical Characterization.

The reflection spectra of the devices were determined using an F20, Filmetrics thin-film measurement instrument integrated with a spectrometer and light source. The layer thicknesses and refractive indexes were measured using spectroscopic ellipsometry (WVASE32, J. A. Woollam). The absorption and transmission spectra were measured using UV-vis spectrometer (Perkin-Elmer 1050). Optical simulations of the single junction based on the transfer matrix method used MATLAB along with the measured J-V characteristics of each cell. The four-point probe method (FPP-5000, Miller Design & Equipment) was used for sheet-resistance measurements. Cyclic voltammetry employed acetonitrile with 0.1 M of tetrabutylammonium hexafluorophosphate at a scan rate of 100 mV s⁻¹. ITO, Ag/AgCl, and Pt mesh were used as the working, reference, and counter electrode, respectively. All measurements were performed at a scan rate of 100 mVs⁻¹.

GIWAXS.

GIWAXS patterns of the thin films were performed at the Stanford Synchrotron Radiation Light Source beamline 11–3 (56) in a He-filled chamber with an X-ray energy of 12.7 KeV (at the critical angle of 0.13° of the films) and LaB₆ was used for geometry calibration. Samples for GIWAXS were prepared on top of Si (100) substrates. The raw two-dimensional X-ray data were processed with a modified version of NIKA into one-dimensional scattering profiles I(q).

Average Photopic Transparency.

The APT is calculated using

$$\text{APT}=\frac{{{\displaystyle \int }}^{\text{}}T\left(\lambda \right)P\left(\lambda \right)S\left(\lambda \right)d\left(\lambda \right)}{{\displaystyle \int }P\left(\lambda \right)S\left(\lambda \right)d\left(\lambda \right)},$$

where $\lambda$ is the wavelength, $T$ is the transmission, $P$ is the normalized photopic spectral response of the eye, and $S$ is the solar irradiance.

Data Availability.

All study data are included in the article and SI Appendix, including cyclic voltammetry measurements, genetic algorithm calculations, and synthesis details.