Interface Property-Functionality Interplay: Suppresses Bimolecular Recombination Facilitating Above 18% Efficiency Organic Solar Cells Embracing Simplistic Fabrication

The donor/acceptor interface properties play vital roles not only for singlet exciton dissociation but also to suppress the free charge recombination enabling state-of-the-art device fill factors (FFs).


Discussion:
The absorption onsets are consistent with an earlier report suggesting that the bulkier outer side groups present in PC6 and P2EH can reduce aggregations causing absorption hypsochromic shift, 1 justifying their relative values for open circuit voltage (V OC ). It must also be noted that such aggregation size does not directly translate to crystallinity. 2 The optical gap is estimated as the absorption and photoluminescence (PL) spectra intersection in the energy (E) scale. The PL spectra are obtained using either 633 nm or 514 nm excitation wavelengths and divided by E 2 before normalization to account for the Jacobian correction. Discussion: The transport energy levels, known as highest occupied molecular orbital (HOMO) and lowest unoccupied molecular orbital (LUMO), are typically identified through both solutionstate and solid-state (e.g., thin films). Cyclic voltammetry (CV) is commonly employed for solutions wherein there is absence of order while photoelectron spectroscopy is for thin films in which molecules are capable of displaying more compact packing.

Photovoltaic bandgap
Herein, ultraviolet photoelectron spectroscopy (UPS) is used to determine the HOMO levels ( Fig.   S3). Upon closer inspection, it is speculative that all the NFAs have two apparent onset regions while only one is recognizable from the donor PM1. When considering the lower energy onset, the estimated HOMO levels are in close agreements with those previously reported for solution state using CV. 3 Conversely, much deeper HOMO levels are obtained from the higher energy onset. The energy level of the polymer PM1 is not significantly influenced by disorder. This behavior is also observed in other high-efficiency polymers such as PM6. 3 For the case of the NFAs, the two apparent onset regions can be associated with the disorder of different phases present. In thin films, neat NFAs can have both aggregate phase (i.e., more compact) and aggregate-aggregate interface (i.e., more randomly oriented). The higher energy onset associates with a more disordered phase owing to good match with CV measurements. Hence, the lower energy onset represents the less disordered phase. To substantiate this assignment, the actual device metrics were taken into consideration. After charge dissociation, polarons from D/A interface will transport through NFA-rich domains and polymer network until extraction/recombination. Consequently, the quasi-Fermi level splitting (QFLS) of the free electron and hole polarons will depend on NFA-rich domains LUMO and PM1 HOMO, which can be approximated as the device V OC . As the PM1 HOMO is already extracted from UPS, the NFArich domains LUMO can then be estimated. Further, the NFA transport gap (E transport ) can be defined from the offset of the identified LUMO and UPS HOMO while the optical gap (E opt ) is determined from absorption and photoluminescence spectra intersection (Fig. S1)   Discussion: Unlike Y6, the asymmetry in the Y and Z directions for PC6 and P2EH is exacerbated by the distortion caused by the benzene present on their outer side chains, resulting in a much larger dipole moment in the respective directions of the molecule. P2EH has a broader outer side chain compared to PC6, resulting in a relatively larger dipole moment in the Y direction. NFAs neat films upon burn-in degradation. Hence, the crystallinity of NFA-rich domains in blends must also be stable under such conditions. By turning into the PL spectra wherein the contribution from D/A interfaces are more pronounced, it can also be observed that PM1:Y6 exhibits a significant redshift. From the electroluminescence measurements (EL) in Fig. S11, the CTS energy increases for the burn-in Y6 blend which approach the singlet energy. Hence, the singlet emission will be more visible which peaks at lower wavelength and leads to the apparent redshift in PL. Both PC6 and P2EH-based blends are also sought to have a decent interface and nanomorphology changes but are less sensitive to PL and absorbance. Hence, EL was further taken advantage of to understand the D/A interfaces by probing emissions directly from such interfaces.            Table S1. Summary of photovoltaic performance under 1 Sun illumination

Supplementary Tables and Discussion
BTP-4F (Y6) 16 S4). The slight change in V OC can be a consequence of a slightly higher P2EH bandgap (Table   S3). Thus, the only parameter that substantially reflects the trend in PCEs is FF. Take note that the EQE and absorption spectra for P2EH display the largest blueshift followed by PC6, suggestive of less aggregation behavior than Y6. However, it cannot justify the FF variations. Smaller aggregates increase the D/A interface area, which may induce negative impacts on the free charge recombination upon a certain threshold, thus lowering the FFs. On the other hand, it can promote more efficient charge dissociation. For the case of this study, FFs are enhanced while charge dissociation is barely influenced, probably because Y6 charge transfer process is already very efficient. Additionally, when the donor polymer is changed to PTQ10, the same relative device metrics as those obtained from the PM1 donor can be observed. Discussion: At SC, photogenerated free charges from exciton dissociation will be swept toward the electrodes by the built-in electric field. The free charge transport is very efficient under this condition such that the influence of energetic defects/traps and other obstacles for charge transport and collection are minimal. Thus, the dissociation of excitons (i.e., neutrally charged) will dominantly impact P diss. On the contrary, at MPP, the built-in electric field is already almost completely vanished by the externally applied bias, leading to less efficient free charge transport and collection.
The P2EH-based device exhibits marginally lower G max , this can be attributed to the slight blueshift in its absorption and optical gap (Fig. S1). On the other hand, the charge dissociation is comparably efficient for all systems while both PC6 and P2EH-based devices display an apparent improvement for charge transport/collection, as suggested from the value of P coll .  (Fig. S2) and the limits of integration can be arbitrarily chosen, herein they correspond to the 50% probability points. This bandgap is a property of a fully operational (i.e., complete) device, a more detailed discussion concerning these properties is available in earlier works. 7 On the other hand, the more commonly reported optical bandgap (E opt ) is based on thin films which is generally estimated from the intersection of absorption and Jacobian-corrected photoluminescence (PL) spectra.
The Urbach energies (E Urbach ) are obtained by exponential fitting of sEQE spectra in the onset region (Fig. S8). It must be noted that the absorption coefficient (α) is proportional to EQE. From the inset of Fig. 2a, it is illustrated that the absorption exponential tail consists of contributions from the D/A interface (or CTS) and acceptor-rich domains, as observed from several studies in fullerene-based systems. 8 Although it is usually not the same case for NFA-based systems owing to low CTS absorption cross-section that largely overlaps with the acceptor absorption edge. To the best of our knowledge, sEQE is by far one of the most sensitive and commonly used technique to fit the absorption onset features as compared to standard UV-Vis-NIR absorption setups. Values of E Urbach comparable to ambient thermal energy (i.e., 25 meV) suggest that energetically trapped charge carriers can be thermally de-trapped, potentially enhancing charge transport and mobilities.
Surprisingly, there are no apparent differences in the E Urbach for all the systems considered even after the burn-in degradation (Fig. S8) despite photovoltaic measurements indicate that charge recombination increased after the burn-in. It is indisputable that E Urbach is an essential parameter to mitigate charge recombination, as cited by several reports, but it is not sufficient to explicitly clarify how remarkable FFs (i.e., > 80 %) can be achieved nor the burn-in efficiency loss mechanisms.  This is consistent with previous reports demonstrating that the higher energy peak will be more evident upon increasing the disorder. 19 Consequently, the Gaussian centered at C1 are assigned for the disordered phase (i.e, aggregate-aggregate interfaces) while those centered at C2 is for the NFAs self-assembly. The Gaussian at C1 may not be relevant in understanding their photovoltaic performance since the D/A interface will dominate in the actual blend. On the other hand, C2 will depend on the aggregation behavior and the nature of interacting chromophores in the NFA selfassemblies. 19 Nevertheless, it was found from Urbach tails and GIWAXS that all the NFAs considered herein self-assemble into a highly ordered fashion with crystalline characteristics. Furthermore, it is also possible that other vibrionic peaks will convolute with the C2 Gaussian. In OSCs, generating free electrons and holes upon exciton dissociation will lead to induced dipoles and electric field, causing the Stark effect. Thus, EA features are also commonly used to understand charge transfer processes. In this study, the EA (ΔR/R) spectra of devices consisting of donor and acceptor blends are obtained via reflectance geometry measurements at a bias of -2 V and amplitude of peak-peak voltage of 4 V of a sine electric field, R is the reflection intensity. It is also shown in Fig. S6-3  NFAs. 21,22 This is known to be beneficial for charge transfer and dissociation. 21 Additionally, the solubility in polar organic solvents (e.g., chloroform and chlorobenzene) can be enhanced which influences the solid-state assembly. 23 Here, rEL is EL/E where E is the photon energy (eV), f is a pre-exponential factor, λ RO is CTS reorganization energy (eV), Φ T is thermal energy (eV), and E CTS is the CTS energy (eV). It must be noted that further division of rEL by E 2 is typically necessary for Jacobian correction, 30 depending on the type of spectrometer used. It is also worthy to deconvolute the blend EL spectra and disentangle contributions from CTS and acceptor singlets. 31 Hence, spectra are fitted with two or three Gaussians, depending on the relative fitting coefficients, while the Gaussian with the lowest energy is assigned to CTS.
As seen in Fig. S5, simulations showed that larger λ RO would cause broader and redshifted CTS emissions. However, there is an insignificant effect of increasing temperature beyond ambient conditions (300 K) up to certain values that are practically reachable upon continuous 1 Sun illumination. Former studies in fullerene-based systems which can clearly distinguish the absorption exponential tail features and specific energetic disorder corresponding to D/A interface and NFA-domains found that this λ RO is a direct measure of D/A interface disorder. 8 Consequently, λ RO is a useful metric to understand the relative energetic disorder of such molecular interfaces.
Page 41 of 51 The D/A interface energy referred to in this work may not be directly equivalent to E CTS . Such interface energy is established from HOMO/LUMO of disordered and ordered phases of the molecules which translates to transport energy levels associated with mobile free charges. On the other hand, the E CTS is based on energy levels prior to recombination wherein internal relaxations and other energy exchanges have occurred thereby not directly representing mobile charges.
Nevertheless, the E CTS values remain a convenient initial indication of relative D/A interface energy.

Note S4. Understanding the transient absorption spectral features
Upon considering the visible detection range (450 -800 nm), the selective excitation of PM1 displays positive features at 500 -650 nm while negative features at 660 -800 nm, as shown in Fig. S14-1(a,b). Since the positive features are around the PM1 absorption range and there is no apparent spectral shape variation with delay time, it can be assigned to PM1 ground state bleach (GSB). After photon absorption, the ground state will be de-occupied until replenished by charge recombination and internal conversions. Thus, there will be a negative change in absorption (ΔA) or equivalently positive change in transmission (ΔT/T), defined as GSB.
To understand the negative features, the PM1:PCBM blend is also measured. For the actual blend, the electron transfer kinetics (i.e., from PM1 to NFA) can be understood from the PM1 singlet GSB decay in early time scales which will be followed by the rising PM1 + polaron GSB in the same wavelength range. The polarons located at the donor/acceptor will give rise to similar GSB as with the corresponding singlets but lack SE as they are non-emissive. As shown in Fig. S14-3(b), such electron transfer (in combination with energy transfer) is suggestive to be ultrafast and cannot be probed by our setup owing to the instrument response limit. This is already expected based on several reports in other high-performing polymer donors and justifying the almost completely quenched donor exciton PL in blends. Hence, the 400 nm excitation of PM1:Y6 will give rise to much longer presence of the 550 -650 nm GSB features due to PM1 + , as shown in Fig. S14-3(a). To support this, the Y6 can be selectively excited in the blend using an 800 nm pump. After Y6 singlet dissociation, there will be hole transfer to the polymer donor PM1, giving rise to PM1 + population. Indeed, even at 0.5 ps after Y6 excitation, the PM1 + GSB is already evident (Fig. S14-3 Additionally, in Fig. S14-3(b), the PM1 + GSB features at 500 -600 nm become more in-shape at time scales wherein most of the NFA singlets have decayed. This is due to the characteristic absorption of the considered NFA singlets causing PIA-like features that drags portions of the PM1 + GSB in early time scales (<1 ps) and decays analogous to the NFA singlet GSB.
Consequently, the PM1 + GSB peak position is monitored only after 5 ps wherein the signals are already largely dominated by polarons thereby can provide more meaningful information about their energy transition upon migration to purer domains at longer time scales.
The NIR detection range (820 nm -1200 nm) is used here to probe the PIA of NFA electron polarons (NFA -). The NFA singlet excitons are found to live only up to less than 0.5 ns but the polaron PIA remains even at 7 ns ( Fig. S15-1). Similar with previous reports for Y6 blends, 3 it is clearly observable that only the blends exhibit recognizable NFA -PIA when compared to pure acceptors as a consequence of exciton dissociation (Fig. S15-2). The PIA features at 1050 nm -1150 nm are then almost exclusive to electron polarons. It must also be noted that the tail of hole polarons PIA (Fig. S15-3) is also visible on the same range. But, since the peak of hole polarons is closer to 900 nm (overlapping with shorter-lived NFA singlet PIA) while the observed polaron peak is closer to 1000 nm, then it is suggestive that the detection from 1050 -1150 nm is dominated External quantum efficiency. The EQE spectra were measured using a Solar Cell Spectral Response Measurement System QE-R3011 (Enlitech Co., Ltd.). The light intensity at each wavelength was calibrated using a standard monocrystalline Si photovoltaic cell. Transient absorption spectroscopy. Near-infrared (850-1600 nm) femtosecond transient absorption spectroscopy was performed using an apparatus previously described. 33 Briefly, 50% of the output of a 1 kHz, 1W, 100 fs Ti:sapphire laser system with a 827 nm fundamental (Tsunami oscillator/Spitfire amplifier, Spectra-Physics LLC) was used to pump a commercial collinear optical parametric amplifier (TOPAS-Prime, Light-Conversion LLC) tuned to 800. The pump was depolarized to suppress effects due to polarization-dependent dynamics and attenuated to the specific energy density. The pump was focused to a 1 mm diameter spot at the sample position.
The probe was generated using 10% of the remaining output to drive continuum generation in a proprietary crystal and detected on a commercial spectrometer (customized Helios, Ultrafast Systems LLC). The films were measured at room temperature under a vacuum of 10-3 torr to minimize potential sample degradation due to air exposure. A similar setup is used for the visible probe (450 -800 nm), as previously described. 34 Time-resolved photoluminescence. The sample was excited with a Ti:sapphire oscillator at 805 nm with a repetition rate of 76 MHz. The time resolved photoluminescence at 850-860 nm was collected and detected by a streak camera (Hamamatsu).
Device capacitance. E4980A Precision LCR Meter was used to measure the capacitance-voltage characteristic of each cells at under dark. The high terminal was connected to the 1kHz common anode. The low terminal was connected to the cathode of the target cell. The low terminal was chosen to be connected with the smallest area of interest because the ammeter is fixed at low terminal side of the machine such that the configuration can minimize the interference from cells.
Density functional theory simulations. The calculation of geometry optimization was performed by the Gaussian 16 program 35 using the B3LYP-D3/6-311(d, p) with dispersion corrections. 36,37 The dipole moment was calculated using the B3LYP-D3/def2-TZVPD containing the diffuse basis to obtain accurate values. 38 The Multiwfn program was used for result visualization. The calculations are carried out on molecules with full-side chains.