Interface Design Principles for High‐Performance Organic Semiconductor Devices

Precise manipulation of organic donor‐acceptor interfaces using spacer layers is demonstrated to suppress interface recombination in an organic photovoltaic device. These strategies lead to a dramatic improvement in a model bilayer system and bulk‐heterojunction system. These interface strategies are applicable to a wide variety of donor–acceptor systems, making them both fundamentally interesting and technologically relevant for achieving high efficiency organic electronic devices.


Cyclic voltammetry measurement on Oligothiophene (O3) for the determination of HOMO-LUMO level
The LUMO level of Oligothiophene (O3) was determined using a combination of cyclic voltammetry (CV) and linear absorption spectroscopy (1,2). O3 was dissolved in anhydrous acetonitrile and 0.1 M tetrabutylammonium perchlorate (TBAP) was used as an electrolyte.
Three-electrode configuration was used with Pt as working electrode, Ag/AgNO 3 as reference electrode and Pt Gauze as counter electrode. Cyclic voltagram was acquired at a scan rate of 100 mV/s. Redox values for Ferrocene were obtained and the oxidation potential (E ferrocene ) was used as calibration standard. The E ox potential was measured by the intersection of the extrapolated tangent lines of the redox curve illustrated in Fig. S3. The HOMO level is given by equation (1).

HOMO= -(E oxd -E ferrocene + 4.8) …………………………(1)
Band-gap energy was estimated by extrapolating tangent lines at the absorption edge of the O3 absorption spectra and using equation (2). The LUMO level for O3 was determined using equation (3). Table S1 summarizes the HOMO and LUMO energy levels for O3 and the different parameters used to determine them.  c Energy gap (E g ) measured according to the band-edge of UV-Vis absorption spectra of the deposited thin film(E g =1240/λ onset eV).

Mtdata/Bphen bilayer device dark IV curve and photocurrent
As described in the main text, the interface modification strategy has been applied to several organic donor-acceptor pairs. Fig. S4 gives an example of them. The bilayer device is comprised with m-Mtdata/Bphen donor-acceptor pair, the interface was modified with LiF for 0~2 nm thickness. The dark current-voltage curve is shown in Fig. S4a   photocurrent shows 5 orders of magnitude of improvement by adding 1 nm LiF, which suggests the CT state recombination can be suppressed in this system as well.

5
Optical absorption of bilayer device without and with LiF as spacer.

Figure S5: (A) and (B)
shows the absolute absorption and IQE measured using an integrating sphere for devices with and without a spacer layer respectively.
Furthermore, we measured the absolute absorption and the internal quantum efficiency (IQE) for our devices. We followed a standard procedure for the absolute absorption measurement by Burkhard et al. (3) Briefly, we took an integration sphere coupled to a CCD camera, and used a Tungsten white-light source for illumination. The white light source was focused into the integration sphere and we measured the spectrum without any sample as reference. Then we measure the spectrum with the sample (ITO/PEDOT/P3HT:C 60 without or with LiF/O3 spacer) in the center of the sphere (the sample is kept in a small angle with the incident light). The absorption is the total transmitted and reflected light that was collected by the integration sphere over the total incident light (without sample). Fig. S5 A shows the measured absorption spectrum for the no spacer device and for the one with a1nm LiF spacer layer. The presence of the spacer layer does not change the absorption. Normalizing our EQE data by the absorption we get IQE (see Fig. S5 B). The device with the LiF spacer layer shows an IQE 2.5 times higher than the no spacer layer device. This clearly suggests suppression of interface recombination by adding LiF is the underlying process that gives the large enhancement and is not due to any changes in the optical density of the film.    results. The transition rates with the interface processes; CT state recombination, CT state dissociation and CT state formation is also observed to be field dependent. As the applied forward bias increases, the change in the local electric field at the interface suppresses CT state dissociation and enhances CT state recombination and CT state formation rates. For simplicity we assume linear field dependence for the corresponding kinetic coefficients in the device model. The increase in the CT state formation rate by the interface layers is specific to this donor-acceptor system. For example, in the Tetracene/C 60 system (6) the dark current was found to decrease with incorporation of a LiF spacer layer. For the P3HT/C 60 system, the LiF or O3 interface layer may improve the interface smoothness and facilitate the in-plane carrier transport required for CT state formation. Under large forward bias, the calculated currents increase more rapidly than the experimental ones. This is probably due to series resistance effects associated with contacts, which are not included in the device model.

Theoretical simulations of ionization potential and electron affinity, singlet and triplet excitations
The ionization potential (IP) and electron affinity (EA) simulations have been calculated using density functional theory (DFT) for three thiolated oligothiophene derivatives (terthiophene, tetrathiophene, and pentathiophene), Fullerene, Iridium complexes [Ir(piq) 2 acac and Ir(ppy) 3 ], and poly(3-hexylthiophene-2,5-diyl) (P3HT) that contains 10 thiophenes. Ground state geometries of neutral molecules, anions and cations have been optimized using 6-31G* basis set and two hybrid functionals, widely used B3LYP and range-corrected ωB97X-D, which is known for its accuracy in large conjugated molecules with delocalized π-electrons. All simulations in this work have been carried out in solution of dichloromethane, whose dielectric constant ε r =9.08   Table S3. Ground-state energies of neutral molecules, cations and anions, computed in solution of dichloromethane using functional B3LYP (top) and ωB97X-D (bottom). Table S4. The ionization potential (IP) and the electron affinity (EA) of thiolated Oligothiophene, P3HT, fullerene, and Iridium complexes. Computations are performed in dichloromethane solution using functional B3LYP and ωB97X-D.
In addition, the lowest then singlet and triplet electronic excitations of the two Iridium complexes of interest in dichloromethane solutions have been calculated using timedependent density functional theory (TDDFT) at the same quantum-chemical levels.

Discussion about charge transfer mechanism for Oligothiophene (O3) and heavy atoms at D-A interface
We consider two mechanisms that could lead to the thickness dependence that we observe for the O3 spacer layers. First, we consider that the thickness dependence of the photocurrent is limited by the exciton diffusion length (L d~1 5 nm). This in our case would result from the total thickness of the P3HT (~10 nm) and the O3 (5 nm). The excitons generated in the P3HT migrate through the O3 to reach the O3/C 60 interface where they dissociate to form the CT state. Here we note that the absorption in O3 is much weaker than in P3HT, and is shown from photocurrent measurements on O3 in Fig. S2 of SI). Thus if the total thickness of the P3HT and O3 is greater than the exciton diffusion length, then a drop in the photocurrent signal would be expected consistent with the experimental observations. However, this mechanism requires that excitons can resonantly diffuse from the P3HT to the O3. The effective optical gap for the O3 is considerably larger than P3HT with negligible spectral overlap (See absorption spectra Fig. S1 in SI). Thus the photoexcited excitons would be required to gain additional energy in order to diffuse into the O3, which is not plausible.

Role of spin-orbit coupling and triplet states in using Metal organic as spacer layers
The energy cascade process can be further facilitated by the formation of long-lived triplet state. The primary single CT state with a hole on P3HT and an electron on Irpiq (or hole on Irpiq) and an electron on C 60 can undergo rapid intersystem crossing to a long-lived triplet CT state due to a close proximity on the heavy atom. (7) Indeed, calculated spin-density plots (in Fig. S8 and Fig. S9) show that the hole is strongly localized on the Ir whereas the electron is essentially localized on the -piq ligands. Stabilization into long-lived triplet state further promotes cascade-like separation of the electron from the hole leading to an increase in the overall photocurrent efficiency. As the thickness of the Irpiq is increased to 1 nm, the photocurrent reaches a peak value and then for and then for >1 nm thickness of Irpiq, begins to drop reaching a level close to that without a spacer layer. Such drastic reduction in the photocurrent efficiency in likely due to quenching of the exciton and reduced charge mobility in the Irpiq layer. Furthermore, the absorption spectrum and the electronic structure calculations of Irpiq (see Table S5 and S8 in SI) suggests a manifold of low-lying intra-molecular singlet states, where transfer of excitons from P3HT is energetically feasible and more probably with increase in the spacer layer thickness. However these intra-molecular singlet states are expected to undergo rapid intersystem crossing into low-lying intra-molecular triplet states acting as quenching sites. (7) Moreover, with increasing spacer layer thickness, an electron and hole in the Irpiq layer will undertake an increasing number of hopping steps to achieve complete spatial separation of carriers to C 60 and P3HT regions respectively. (8,9)Due to highly localized electron and hole states (see Table S5 and S8 in SI) such transport is not very efficient and can lead to a further reduction in the photocurrent with increasing thickness of Irpiq.

Fig. S10 External quantum efficiency (EQE) spectrum for bulk-heterojunction devices with P3HT: ICBA without and with spacer layer modification
External quantum efficiency is shown in Fig. S10 for BHJ devices without and with spacer molecules. By adding 6% O3, the EQE increases by 25%, which is consistent with the light current-voltage curve results. Similarly, the EQE has increased by 20% by adding 4% Irpiq that is consistent with the observation from light current-voltage curves. Note the integrated current density calculated from the EQE (J SC =11.2mA/cm 2 ) spectrum matches with the J SC measured out of the light current-voltage curves. The 6% difference measured from EQE with measured J SC can be due to the spectrum mis-match from the monochromator and the AM 1.5 white light.

A. Annealing experiment
We measured the photocurrent efficiency of a P3HT:O3:C60 device before (see Fig. S13 A) and after annealing (see Fig. S13 B). The blended device was prepared by mixing P3HT and O3(6 wt%, as used in the BHJ device in Fig 5 in manuscript) over night that resulted in a homogeneous mixture ensuring the incorporation of O3 in the bulk phase. The resulting solution was spin coated onto a ITO/PEDOT electrode and the device completed by thermally evaporating C60 followed by LiF(1nm)/Al (100nm). We also fabricated a P3HT/C60 bilayer device in the same method without O3 blending, which served as a control device. We measured the device performance for these devices before and after annealing and the results are illustrated  was precisely placed at the interface as shown in Fig. 1A of the manuscript, where a 350% increase in photocurrent was observed. These results clearly indicate that the presence of O3 at the donor/acceptor phase interface is critical and that it promotes extremely efficient charge separation and hence leads to a high photocurrent.
Next, we annealed both of these devices. The photocurrent increases as expected due to interdiffusion of the donor and acceptor materials. However, the highest photocurrent was observed for devices with O3 indicating the crucial role played by the molecule. The increase in the photocurrent was ~60% in comparison to the device without O3, again consistent with the percentage increase we observed in the BHJ device in Fig.5 of the manuscript.
Although we cannot conclusively determine whether O3 undergoes phase separation, these measurements clearly show that, after thermal annealing, O3 plays a crucial role in promoting charge separation between the P3HT and C60, and the enhancement in the performance is due to the presence of O3 at the P3HT/C60 donor/acceptor interface. Furthermore, the trend observed as a function of loading volume % of O3 is consistent with that observed with the thickness dependence of the O3 in the bilayer devices (see Fig. 2 & Fig. 3 in the manuscript) again suggesting that the mechanisms for charge separation in the bilayer device with O3 at the interface or the BHJ device with O3 blended in are similar.

B impedance spectroscopy
Impedance spectroscopy is a powerful technique to probe interface charge transfer and recombination during device operation under continuous illumination and applied bias [10][11][12].
A typical BHJ device can be modeled using R s (R 1 C 1 (R 2 C 2 )) circuit elements, where R s is the contact resistance, R 1 corresponds to transport resistance in the bulk phase, C 1 represents the geometric capacitance for the device respectively; and R 2 , C 2 describe the donor/acceptor interface property as shown in Fig. S14 A. Under illumination, the RC time constant τ=R 2 •C 2 represents effective charge recombination lifetime. Fig. S14 B shows a plot of the imaginary part of the impedance (Im(Z)) as a function of AC frequency for a no spacer device and devices with O3 (6%) and Irpiq (4%) spacers. These devices were particularly chosen as they demonstrate the highest increase in efficiency. The results show that the peak value of Im(Z) shifts to lower frequency upon incorporation of O3 and Irpiq. The change in time constant suggests a decrease of the interfacial recombination rate that leads to an increase in the efficiency of the devices. Such a phenomenon should not be observed, if O3 or Irpiq were present in the bulk phase. Also, the calculated effective interface recombination lifetime (the product of R 2 and C 2 ) increases and later gradually decreases with change in wt% of the O3 and Irpiq. We believe that at 6% loading for O3, the interface is saturated and further addition leads to its preferential incorporation in the bulk. All other parameters were constant, including the bulk resistance (R 1 ), the geometric capacitance (C 1 )) and the contact resistance.
In conclusion, although the precise location of the O3 is not known, these results again clearly show that the photocurrent enhancement in our devices after incorporation of an O3 spacer layer is due to its presence at the interface and not in bulk, and hence are depicted as shown in Fig. 5 of the manuscript.  200) and (300) are identical, which indicates the addition of O3 didn't change the film morphology Figure S16: (A) Illustrates measurements to determine the inter-diffusion of C60 into the P3HT and (B) and (C) shows the absolute absorption and IQE measured using an integrating sphere for devices with and without a spacer layer respectively.