Enhancing Organic Semiconductor Molecular Packing Using Perovskite Interfaces to Improve Singlet Fission

Singlet fission, a process by which one singlet exciton is converted into two lower energy triplet excitons, is sensitive to the degree of electronic coupling within a molecular packing structure. Variations in molecular packing can be detrimental to triplet formation and triplet–triplet separation, ultimately affecting the harvesting of triplets for electricity in organic photovoltaic devices. Here, six phase‐pure molecular packing structures of 6,13‐bis(triisopropylsilylethynyl)pentacene (TIPS‐pentacene) with varying optoelectronic properties are isolated using 2D lead halide perovskites as tunable, crystalline surfaces for crystallization. Transient absorption spectroscopy reveals that while triplet formation is fast (<100 fs) regardless of template structure, the increased ordering in perovskite‐templated samples speeds up triplet–triplet separation and recombination, providing evidence that the benefits of phase‐purity offset minor variations in molecular packing. Molecular dynamics modeling of the interface reveals that perovskite‐templating allows for closer packing of TIPS‐pentacene molecules for all perovskite templates. With an extensive number of organic molecule‐perovskite pairings, this work provides a methodology to use ordered, periodic surfaces to elucidate structure–property relationships of small organic molecules in order to adjust structural or optoelectronic responses, such as molecular packing and singlet fission.


Introduction
Controlling the molecular packing of small molecule organic semiconductors (OSCs) is vital for their commercialization in optoelectronic devices. [1]2e,3] It is well-established that molecular packing and morphology affect singlet fission (SF), an exciton multiplication process whereby one singlet exciton is converted into two triplet excitons, which has the potential to overcome the Shockley-Queisser efficiency limit for solar cells. [4]F has been observed in many prominent acene-based OSC molecules, such as 6,13-bis(triisopropylsilylethynyl)pentacene (TIPS-Pn), which satisfy the energetic requirement that the triplet state energy is less than or equal to half of the singlet state energy. [5]6b,d] Precise control of crystalline properties is still an active area of research, as optimizing SF properties could have profound ramifications for photovoltaic applications.
Despite the importance of molecular packing on SF, controlling molecular packing has proven difficult.For organic semiconductors, molecular packing results from nonselective van der Waals forces, which has the potential to result in a variety of structures with similar energies.2d,e,8] In the case of TIPS-Pn, thermodynamically stable, metastable, and strain-stabilized structures have been isolated using a range of solution deposition techniques, [1c,9]  each exhibiting different optoelectronic properties.While researchers exploited this nonselective crystal structure formation to tune the degree of intermolecular -orbital overlap to enhance solid-state properties, an added challenge to reduce structural variation and isolate phase-pure crystalline structures was introduced.Methods to precisely control molecular packing and increase phase purity would be beneficial for applications that require reliable carrier transport, including singlet fission.
Substrate functionalization using self-assembled monolayers (SAMs) has been used to control molecular orientation, morphology, crystallization kinetics, and crystal grain size. [10]11c] However, tunability of SAM ligand density is limited due to an inherent lack of long-range ordering arising from nonspecific reaction sites when SAM molecules attach to silicon surfaces, restricting their use to systematically control molecular packing.On the other hand, while crystalline packing of SAMs on gold surfaces is possible, there is no method to reliably tune the SAM packing density and order.
Here, we demonstrate precise control of molecular packing of TIPS-Pn and isolate six phase-pure structures using crystalline perovskite interfaces (Figure 1a).While perovskites are a class of materials of great scientific significance, [12] we exploit their structural tunability as surfaces for OSC crystallization to tune intermolecular overlap (Figure 1b,c).To overcome ligand density uniformity and tunability limitations of traditional SAMs, we use one-layer 2D lead halide perovskites whose composition can be altered to modify the ligand molecule and ligand density by up to 30 percent in a periodic manner.Molecular dynamics (MD) simulations of TIPS-Pn/perovskite interfaces undergo structural changes to TIPS-Pn molecular packing that are qualitatively consistent with experimental data and reveal reduced - stacking distances for perovskite-templated structures, elucidating a potential explanation for the similar optoelectronic behavior observed for different perovskite-templated structures.Finally, we measure singlet fission rates and triplet state lifetimes using transient absorption spectroscopy and show that increased ordering in molecular packing is essential for rapid triplet-triplet separation and transport, which is required to ultimately harness the triplets for electricity.

Tuning Ligand Density at the TIPS-Pn/Perovskite Interface
1D and 2D perovskites preferentially crystallize in horizontal orientation with octahedral sheets parallel to the substrate and periodic ligand termination on the thin film surface (Figure 1d), as confirmed with grazing incidence X-ray diffraction (Figure S1, Supporting Information).Surface chemistry was altered in two ways: changing the halide composition tuned ligand density from 0.025 to 0.033 Å 2 (32 percent relative increase) (Figure 1e; Table S1, Supporting Information), and altering the cation precursors changed the organic ligand terminating the surface.We used phenylmethylammonium (PMA) and phenylethylammonium (PEA) ligands to modify surface chemistry.PEA forms a long-range -stacking network while PMA ligands are more isolated, resulting in different ligand orientations and leading us to expect changes in surface energies. [13]IPS-Pn thin films were deposited on six pre-crystallized 2D perovskite surfaces with varying ligand and ligand density (PEA 2 PbCl 4 , PEA 2 PbBr 4 , PEA 2 PbI 4 , PMA 2 PbCl 4 , PMA 2 PbBr 4 , and PMA 2 PbI 4 ) using solution shearing (Figure 1f). [14]1c] Following TIPS-Pn thin film deposition and solvent vapor annealing, the underlying perovskite layers exhibited no structural changes or visual degradation while still retaining high crystallinity and strong absorption onset characteristic of the perovskites (Figures S2-S4, Supporting Information).

Structural Characterization of Perovskite-Templated Molecular Packings
Polarized optical microscopy showed similar morphologies of TIPS-Pn thin films regardless of perovskite surface (Figure 1g and Figure S5, Supporting Information).Thin films consisted of long crystalline grains aligned in the solution shearing direction (Figure S6, Supporting Information), with crystals hundreds of microns in length and tens of microns in width.TIPS-Pn morphology and crystal alignment were unchanged after solvent vapor annealing, indicating that no recrystallization occurred (Figure S7, Supporting Information).Because SF properties are influenced by both molecular packing and morphology, identical morphologies obtained in this study allow an opportunity to determine the effect of molecular packing and phase purity without morphological variation.
Grazing incidence X-ray diffraction was used to determine the influence of perovskite ligand and ligand density on the molecular packing of TIPS-Pn.Due to the anisotropy of TIPS-Pn thin films, diffraction patterns were collected with the incident X-ray beam along the shearing direction and with 90°rotation with respect to the shearing direction to observe the (101) and (011) crystallographic planes, respectively (Figure 2a,b).TIPS-Pn diffraction patterns are (Tables S2 and S3, Supporting Information) consistent with the published Form I molecular packing structure of TIPS-Pn, as reported by Diao et al. [2e] The highest and lowest TIPS-Pn plane spacing values obtained from interfacing with perovskites were 7.05 and 6.95 Å for the (101) plane, and 6.53 Å and 6.48 Å for the (011) plane.While all structures are within the Form I family, the (101) and (011) plane spacings of TIPS-Pn were uniquely influenced by perovskite surfaces.Perovskite surfaces induced peak shifts in the (101) plane relative to each other (Figure 2c).These perovskite-templated TIPS-Pn structures exhibited discrete (101) plane spacings ranging from 6.95 Å to 7.06 Å (≈1.6% change).Untemplated TIPS-Pn exhibited an average (101) plane spacing of 7.00 ± 0.06 Å, which has a significantly larger standard deviation relative to perovskite-templated structures (Figure 2d).This indicates poor control over TIPS-Pn molecular packing and more structural variation when deposited on silicon.In contrast, TIPS-Pn molecular packing was finely tuned using perovskite surfaces with varying ligand and ligand density, as evidenced by discrete spacings with substantially lower standard deviations.Perovskite-templated structures exhibited high phase purity with reduced structural disorder.The (011) plane of untemplated TIPS-Pn was observed concomitantly with a second peak that is separated in Qz.2d,e] For perovskite-templated TIPS-Pn structures, this degeneracy was removed, which is consistent with increased ordering in these systems.Only minor variations in (011) plane spacings were observed, indicating that TIPS-Pn molecular packing changes occurred mostly along the a-axis of the unit cell (i.e., (101) plane) rather than the b-axis.These results show that TIPS-Pn molecular packing can be finely tuned using tunable, ordered surfaces.
While unit cell parameters for TIPS-Pn molecular packings could not be obtained due to too few independent Bragg peaks (i.e., without perovskite peaks overlapping), analysis of Q xy and Q z components of the Q-vector in reciprocal space for (101) peaks provide qualitative information about relative changes to the TIPS-Pn unit cell.2e,9c] The overall change in Q xy for perovskite-templated samples was 1.7% and high variability was observed again for untemplated TIPS-Pn.Upon inspection of the Q z components of the (101) peak (Figure 2 g), distinct Q z values were observed for perovskite-templated samples, possibly suggesting changes in the tilt of the c-axis with respect to the ab-plane.
No clear trend in TIPS-Pn (101) plane spacing and ligand density was observed.For PEA perovskites, the TIPS-Pn (101) plane spacing increases with halides in the order Cl < I < Br while ligand density increases in the order I < Br < Cl.For PMAterminated perovskites, (101) plane spacing increases in the order Cl < Br < I. Without a clear correlation to ligand density, these results suggest that no direct lattice matching is taking place.Interestingly, perovskites with similar ligand density but different ligand molecules (e.g., PEA 2 PbCl 4 and PMA 2 PbCl 4 ), induced different (101) d-spacings of TIPS-Pn.Taken together, these observations suggest that both ligand and halide selection contribute to defining the electrostatic interaction at the interface that controls the TIPS-Pn molecular packing.
Because perovskite-templated structures were relaxed (i.e., strain-free) using solvent vapor annealing, final structures are expected to be thermodynamically stable with their distinctions stemming from unique interfacial interactions for each TIPS-Pn/perovskite pairing.The nonspecific structure formation as evidenced by increased structural variation in untemplated samples is consistent with there being no particular energetic minimum when interfaced with silicon.While no extensive structural changes were observed in perovskite thin films, slight variations in perovskite absorption peaks were detected (Figure S8, Supporting Information).These shifts likely arose from interactions between TIPS-Pn molecules and the terminating ligands of the perovskite that induced small changes in the octahedral tilt in the perovskite structure.The thickness of TIPS-Pn thin films is 110 ± 30 nm (Figure S9, Supporting Information), demonstrating that TIPS-Pn/perovskite interactions are significant enough to propagate into the bulk of the film to more than 100 nanometers, and still yield phase pure structures.

Effect of Enhanced Phase Purity on Optical Properties
Optical properties of perovskite-templated TIPS-Pn structures were studied using UV-Vis spectroscopy.The lowest-energy transitions between 500 and 750 nm are associated with the S0→S1 transition of TIPS-Pn, [15] as labeled as peaks 1, 2, and 3 in Figure 3a.5a,9c] Because S0→S1 transitions are below the bandgaps of the perovskites, the lowestenergy TIPS-Pn absorbance peaks could be isolated (Figures S10-S11, Supporting Information).All TIPS-Pn thin films exhibit an absorbance peak near 700 nm, denoted as peak 1, which corresponds to a red-shifted absorbance feature arising from - stacking of TIPS-Pn molecules in the solid state that is not present in solution state spectra. [16]Incremental changes to this transition energy were observed when varying the perovskite surface (Figure 3b), which suggests variations in the degree of stacking interactions and is consistent with our structural findings.The transition energies of peak 2 and 3 also showed differences in transition energies, but in contrast to the more discrete transition energies of peak 1, multiple samples exhibited overlapping transition energies (Figure S12a,b, Supporting Information).Sharifzadeh et al. showed in ordered domains of TIPS-Pn that peak 2 arises from transitions from a blending of nearly degenerate states while peak 1 is more strongly associated with the lowest energy singlet excited state arising from -stacking, [5a] which may explain the overlapping transition energies of peak 2. Peaks 4 and 5 show minor variations in transition energy, which is consistent with an intramolecular excitation that would be less affected by - intermolecular interactions (Figure S12c,d, Supporting Information).
Steady-state photoluminescence (PL) measurements of TIPS-Pn thin films are shown in Figure 3c.TIPS-Pn thin films were selectively excited with 560 nm light, which is below the bandgap of perovskites.All TIPS-Pn thin films exhibited weak PL intensity with only one observable peak at ≈650 nm, which is consistent with literature. [17]Perovskite-templated TIPS-Pn samples exhibited reduced PL intensity relative to untemplated samples, suggesting that relaxation dynamics have been altered in more-ordered TIPS-Pn films.Nui et al. showed that TIPS-Pn exhibits more intense PL in disordered or amorphous regions due to disruptions in intermolecular interactions and exhibits reduced PL in regions with increased crystallinity. [18]ur results are consistent with literature in that perovskitetemplated structures exhibited increased ordering and reduced PL.Our results provide evidence that regardless of minor variations in perovskite-templated TIPS-Pn structure, increased ordering plays a significant role in enhancing intermolecular interactions.

Singlet Fission and Triplet Transport
Transient absorption (TA) spectroscopy was used to investigate singlet-fission dynamics in perovskite-templated and untemplated TIPS-Pn films.TA spectra are dominated by a spectral feature peaking near 530 nm associated with the absorption of both correlated and separated triplet pairs in TIPS-Pn (Figure 3d). [19] The complete set of spectra can be found in Figure S13 (Supporting Information).Because TA spectra of TIPS-Pn samples on PMA 2 PbI 4 and PEA 2 PbI 4 perovskite templates were complicated by direct excitation of the perovskite (Figure S14, Supporting Information), we were unable to isolate singlet-fission/triplet-pair dynamics for these samples.
19a] Determination of this timescale is inhibited by a coherent artifact arising from nonlinear optical interactions in the substrate during the period of pulse-pair overlap.The triplet pair absorption intensity continues to rise and redshifts by ≈2 nm in the picoseconds following excitation (Figure 3d and Figure S13, Supporting Information).19a] The triplet feature begins to decay over the next nanosecond, which we attribute to triplet recombination (T 1 +T 1 → S 0 +S 1 ).
TA spectral evolution was analyzed by global spectral analysis subject to the kinetic model given in Equation 1 (i.e., global lifetime fitting.see Methods).Species associated difference spectra (SADS) associated with each kinetic state of Equation 1 are presented in Figure S15 (Supporting Information), with fitting lifetimes reported in Table S4 (Supporting Information).Transients at a probe wavelength of 525 nm are presented in Figure 3e and Figure S16 (Supporting Information).All templated TIPS-Pn exhibited both faster triplet-pair separation and recombination kinetics when compared to the untemplated control.19a] In contrast, triplet-pair separation speeds up by a factor of 2.4 (on average) in perovskitetemplated samples.
Triplet recombination/annihilation is also observed to be faster for perovskite-templated samples.For the Br-based templates, recombination behavior was best fit using a biexponential decay (both components have the same SADS, i.e., they correspond with the same photophysical species).Figure S16 (Supporting Information) compares transient cuts for various data sets at the lowest excitation fluences used in our study.We note that recombination is ≈50% faster than what has been reported [19a,c] previously (2.1 ns) for amorphous and crystalline TIPS-Pn.This is most likely a consequence of the higher fluence needed to conduct our measurements compared to prior studies.The triplet decay lifetime decreases by 50% for our samples at even higher fluences (Figure S17, Supporting Information).
Importantly, triplet annihilation is a second-order process, with a rate that depends on both the transport rate and initial population density of triplets within the film.In the limit that triplet population decay is dominated by annihilation, the inverse of the measured triplet absorption intensity will scale linearly in time with a slope proportionate to the second-order annihilation rate constant.We fit the measured triplet absorption decay for each sample as a second-order process for time delays that follow the initial triplet separation (from ≈ 5-10 ps to >1 ns, see the Experimental Section).As demonstrated in Figure S18 (Supporting Information) we obtain reasonable fits to a linearized expression (Equation 2) and used the slopes to calculate the annihilation rate constant for each sample.Rate constants for the separation of correlated triplets and annihilation of separated triplets are plotted in Figure 3f and listed in Table S5, revealing that both increase with templating.Interestingly, we observe a clear dependence on perovskite template structure for both parameters: While films on lead-chloride templates exhibit a modest increase (30-40%) in both rate constants relative to untemplated TIPS-Pn, films on lead-bromide templates exhibit significant increases (300-400%).This implies that the site-to-site triplet energy transfer that underlies triplet separation and diffusive annihilation is improved with templating, and substantially with lead-bromide templates.In contrast, there is no appreciable sensitivity with change in ligand.
The kinetics of triplet-pair generation and the separation, decoherence, diffusion, and recombination of triplets is well-known to be sensitive to the details of intermolecular interactions. [22]6d,23] Triplet-pair generation in TIPS-Pn films is generally quite fast and efficient, such that little to no impact is expected for modest changes in packing structure.We observed no significant difference in the rate of triplet-pair generation in our films (in both cases, it occurs on ≈100 fs timescales or faster), which is consistent with these previous observations.
In contrast, triplet-pair separation and recombination, which occur on slower timescales, are highly sensitive to the impact of crystal structure on intermolecular triplet energy transfer.19c]  In addition, Grieco et.19a] Similarly, Doucette et al. demonstrated that the triplet-separation rate can be increased when polycrystalline films are compressed in a diamond anvil cell, as compression induces packing modifications that increase pair-wise charge-transfer interactions. [21]We note that the (101) plane spacings for templated films are within the standard deviation of the plane spacing of untemplated TIPS-Pn, yet all templated films exhibit triplet transfer dynamics,that are faster than that of the untemplated TIPS-Pn.This behavior is consistent with the rate of triplet transfer dynamics increasing with increased packing order with contributions also from improved stacking interactions.Further, the large increase in transport (annihilation) rates observed for Br-based perovskite-templated films implies the likelihood of improved intermolecular coupling.

Simulated Annealing of TIPS-Pn/Perovskite Interfaces
We employed molecular dynamics (MD) simulations to better understand the structural origins of the templated TIPS-Pn singlet fission behavior.24a] After developing forcefield parameters that recovered properties of Form I TIPS-Pn and each perovskite, we generated initial periodic interfaces by lattice matching TIPS-Pn with perovskite a and b cell parameters (pgs.S21-S31, Supporting Information).To further optimize a and b cell parameters of the interfaces, anisotropic NPT simulated annealing was performed.Optimized cell parameters were subsequently used in simulated annealing NVT simulations to identify low-energy structural minima for each interface (Tables S7 and S8, Supporting Information and Figure 4a, additional details in the Supporting Information).We could not identify minima for the TIPS-Pn/PMA 2 PbI 4 interface due to irreconcilable lattice mismatching, which would have required a supercell too large for MD simulations.
To validate our computational methods, we compared the simulated and experimentally observed values of (101) spacings.MD simulations qualitatively recovered experimental trends in TIPS-Pn d-spacing (Table S9, Supporting Information) but were systematically lower than experimental values.Using the experimental TIPS-Pn (101) spacing as a reference, MD simulations correctly predict the (101) spacing trend observed using PEA-terminated perovskite interfaces, with higher and lower distortion of the TIPS-Pn lattice for PEA 2 PbCl 4 and PEA 2 PbBr 4 , respectively, and PEA 2 PbI 4 in-between.Calculated (101) spacings for PMA 2 PbCl 4 and PEA 2 PbBr 4 were within 0.004 Å, in agreement with experiment (Figure 2d).While our models do not represent the full complexity of the real interfaces, they recover the experimental trends in (101) spacings, suggesting that the parameters dictating structural changes (i.e., interfacial lattice sizes and intermolecular interactions) are effectively captured.
The large number of atoms in these interfaces (≈1000s, Table S8, Supporting Information) prohibit accurate band structure calculations of these materials.However, previous studies on acene-based molecules demonstrate a strong correlation between - stacking distances and charge transfer character.19a] We proceeded to compute TIPS-Pn - stacking distances using MD interfacial structures as a proxy for singlet fission behavior.Figure 4b shows that all perovskite-templated TIPS-Pn structures relaxed to shorter - stacking distances compared to the Form I phase and that all - stacking distances for the interfaces were similar, despite differences in (101) spacings.These crystal structure changes can occur via a lengthened c vector in the templated TIPS-Pn structure and/or a difference in TIPS-Pn molecular tilt with respect to the a,b plane.Our experimental data corroborates this, as diffraction data indicates changes in both a and c axes of perovskite-templated samples (Figure 2f,g, Supporting Information).With all considered, the faster triplet-triplet separation and recombination observed using TAS is consistent with expected outcomes for enhanced intermolecular interactions and increased triplet energy transfer arising from shorter - stacking distances (observed using MD) and increased ordering (observed using GIXD).We conclude that perovskite-templating allows for improved intermolecular interactions by means of closer packing of TIPS-Pn molecules in conjunction with increased molecular ordering.

Conclusion
Small molecule OSCs, such as TIPS-Pn, adopt different molecular packing structures with varying optoelectronic properties stemming from differences in intermolecular interactions.Singlet fission, an exciton multiplication process, is dictated by the degree of electronic coupling between molecules.While triplet state formation via singlet fission can be robust against minor variations in molecular packing, triplet-triplet separation and diffusion are acutely sensitive to such regions, reducing likelihood that triplets can be harvested for electricity in photovoltaic devices.However, controlling the OSC packing structure and reducing structural disorder in molecular packing structures are both difficult.In this work, we use 2D perovskites, which are highly-ordered materials with modifiable ligand termination and ligand density, as a templating interface for the OSC TIPS-Pn.We show that templating TIPS-Pn on 2D perovskites can both modify the molecular packing of TIPS-Pn and reduce structural disorder in the TIPS-Pn thin films.With enhanced phase-purity, triplettriplet separation and recombination occurs on faster timescales, indicating improved triplet energy transfer between molecules.Interestingly, we observe that templating further increases the rate of triplet-pair separation relative to untemplated crystalline TIPS-Pn under ambient conditions, highlighting that this approach can be used to affect enhanced triplet-energy transfer as is needed to ultimately harvest triplets in materials applications.We used MD simulations of the interfaces to further elucidate how perovskite-templating improves performance, which is via closer packing of TIPS-Pn molecules.
Researchers have established a direct link between singlet fission and structural, morphological, and heterogeneous features.Recent reports have repeatedly emphasized the importance of phase-purity to efficiently separate triplets.However, without methods to achieve phase-pure structures, singlet fission remains promising but limited in application.This work provides an easy-to-adopt method to reduce structural disorder of small molecules using 2D perovskites as tunable, crystalline interfaces, which are produced through simple, rapid fabrication methods.Attempts to use traditional self-assembled monolayers to influence molecular packing have stopped short of ligand density tunability.Perovskite-templating could prove advantageous in other small molecule applications where even modest amounts of disorder can severely hinder device functionality.The number of molecule-perovskite combinations are extensive, making the perovskite-templating method generalizable to application-based and fundamental studies to elucidate structure-property relationships via the fine tuning of crystalline structures.
wavelength that is below the threshold energy required to excite perovskites used in this study.A 560 bandpass filter (FWHM = 10 nm) was used for the excitation light and a 570 nm longpass filter was used for the emission.Background arising from the substrate was subtracted from each spectrum.Absorbance spectra were collected using a PerkinElmer UV/Vis/NIR Lambda 950S spectrometer with an integrating sphere.Two samples were measured for each TIPS-Pn/perovskite layering.
Molecular Dynamics and Simulated Annealing [24a] : AMOEBA forcefield parameters were either generated via previously outlined procedures or referenced from previous works (pgs.S21-S32, Supporting Information).Initial structures for perovskites and TIPS-Pn were developed using previously published crystal structures (Table S1, Supporting Information).Forcefield parameters were validated by NPT simulations of Form I TIPS-Pn and perovskite structures to verify that they recovered experimental densities (Table S5, Supporting Information).All MD simulations were run in OpenMM [26] with CPU acceleration.All NPT MD simulations used a Verlet Integrator with timesteps of 0.5 fs, and NVT simulations used a Langevin integrator with a timestep of 0.5 fs.Constant-temperature simulations used an Andersen thermostat with a collision frequency of 1 ps, and constant pressure simulations used a Monte Carlo anisotropic barostat.Nonbonding cutoffs were set to 1.05 nm to avoid errors with fluctuating box sizes.Periodic boundary conditions were enforced and the Particle Mesh Ewald (PME) method and a mutual polarization scheme was employed with a 10 −5 tolerance.Simulated annealing was used to enable efficient structural sampling on ps timescales.Simulated annealing NVT simulations were preceded by geometry optimization of the interface structures and then initialized at 500 K with a friction coefficient of 1 ps −1 .After 5 ps, the temperature of the simulation was lowered by 33.33 K every 0.25 ps for 10 ps.Simulated annealing NPT simulations used these same parameters.At the end of simulated annealing, the final structure was optimized, and (101) spacing and - stacking distances were computed.Additional simulation details can be found in the Supporting Information (pgs.S32-S36, Supporting Information).
Ultrafast Transient Absorption Spectroscopy to Measure Singlet Fission Rates: The setup for our transient absorption measurements has been described in detail previously [27] and here experimental details were described for the work presented.Ultrafast excitation and probe pulses were generated using the amplified output of a Ti:sapphire laser (Coherent Legend Elite, 3.8 mJ pulse −1 , 1 kHz repetition rate, ≈35 fs pulse duration).Excitation pulses at 700 nm (≈700 μJ cm −2 ) were obtained through second harmonic generation of the NIR signal obtained from an optical parametric amplifier (OPA, Coherent OperaSolo).The time resolution of our experiments was ≈150 fs, and triplet kinetics were monitored out to 1200 ps.Broadband probe pulses were obtained via white light generation in a 2 mm calcium fluoride crystal.The vertically polarized pump was aligned parallel to the -stacking axis (the a-axis of the unit cell), with the probe polarization set to magic angle with respect to that of the pump using a wire-grid polarizer.The pump and probe beams were aligned to overlap in sample films.The relative pulse delay at the sample was controlled with a motorized translation stage.Films were sealed inside an argonpurged chamber to avoid photo-oxidation.Films were also translated sideto-side by < 1 mm in a direction perpendicular to the incident laser beam throughout data acquisition in order to eliminate the potential for photobleaching or photodegradation.A scan at a fixed pump-probe time delay averaged 25 accumulations (averaging a total of 2500 individual spectra).The timescale for acquisition at a given time delay with sample rastering ensured that measurements averaged over any contributions from heterogeneity of thin films.TA spectral evolution was analyzed in more detail by global spectral analysis. [28]Global analysis assumed a kinetic model involving time-resolution-limited appearance of triplet pairs, kinetic interconversion to separated pairs on the timescale  sep , and single or biexponential recombination ( rec ), as captured by Equation ( 1) Kinetic models were convoluted with the instrument time-resolution.Time-dependent fits at 525 nm and species associated difference spec-tra (SADS) obtained from global lifetime analysis are presented in Figurse S16 and S15 (Supporting Information), respectively.Reasonable fits were obtained in all cases, and a summary of the triplet separation and recombination lifetimes and their associated errors obtained are plotted in Figure 3f.Triplet annihilation rates were determined by fitting transient absorption (or optical density, ΔOD) at the triplet absorption peak (525 nm) from time delays t ≈ 10 ps to >1 ns to a second-order expression given by Equation (2) Here, ΔOD 0 is the triplet absorption taken at time delays immediately following correlated pair separation (≈3-10 ps).Weighted linear regression was used to account for variation in in error of 1/ΔOD to obtain best values for B and associated error bars.Conventional annihilation rate constants (k in units of m −1 s −1 ) were calculated via k = 2lB, [21] where  is the extinction of the triplet absorption and l is the film thickness. [29]tatistical Analysis: The (101) d-spacing values are the averages of different samples of TIPS-Pn/perovskite pairings with error bars representing one standard deviation.(101) diffraction peaks were analyzed via a bivariate Gaussian peak fitting using a Matlab toolbox developed by Zhang Jiang called Grazing-incidence X-ray Scattering User interface (GIXSGUI). [25]Sample set is as follows: TIPS-Pn controls (n = 4), TIPS-Pn deposited on PEA 2 PbX 4 (X = Cl, Br, I) (n = 4), TIPS-Pn deposited on PMA 2 PbCl 4 (n = 3), and TIPS-Pn deposited on PMA 2 PbBr 4 /I 4 (n = 2) .TIPS-Pn absorption peaks were fit using a Gaussian peak fitting in Orig-inPro.Absorption peak values represent the averages of three measurements on n = 2 samples per TIPS-Pn/perovskite pairing with error bars representing one standard deviation.TIPS-Pn photoluminescence peaks were measured for two samples for each TIPS-Pn/perovskite pairing with error bars representing one standard deviation.
Transient measurements reported were conducted with one of each sample type.Transient spectroscopic data was chirp-corrected based on the well-defined nonlinear response of a blank silica substrate that appears with pump+probe temporal pulse overlap.No additional preprocessing of TA data was applied.Triplet separation lifetimes and associated uncertainties, as reported in Table S4 (Supporting Information), were determined by global lifetime analysis using Glotaran [28a] subject to the kinetic model presented in Equation 1. Fits are based on (a sample size of) 83 independent wavelengths for each data set.Corresponding triplet separation rate constants and associated uncertainties reported in Figure 3f were obtained from standard error propagation.Annihilation rate constants were determined by fitting the transient absorption decay at 525 nm to Equation 2 using a weighted linear least-squares regression (intercept set to zero) with a standard spreadsheet program (Excel).The standard signal uncertainty iat 525 nm was taken as 0.1 mOD (based on fluctuations from a baseline of OD = 0 in absence of the pump pulse).Error for the signal inverse (as plotted in Figure S18, Supporting Information) were determined by standard uncertainty propagation.B and its associated error were calculated by weighted regression formulae.Relative uncertainty in annihilation rate constants were obtained by propagated uncertainty of B and sample film thicknesses.
The - stacking distances from MD simulations presented here are the averages from five separate simulations of each system, with error bars generated representing one standard deviation calculated from all five runs for each system.In the supporting information, NPT simulations validate the use of the polarizable forcefields made here to simulate TIPS-Pn/perovskite interfaces, and the average volumes reported in those Figure are numerical averages across a single simulation's run.

Figure 1 .
Figure 1.Bilayer thin films of TIPS-Pn and 2D perovskite.a) Schematic of one-layer 2D perovskite and TIPS-pentacene interface.b) Depiction of the TIPS-Pn molecular offset of pentacene backbones as viewed down the a-axis crystallographic direction.c) Schematic - stacking distances between TIPS-Pn molecules (black arrows).Variations in (101) plane spacing will result in changes in both the -stacking and molecular offset, which dictate overall intermolecular interactions.d) The ligand-terminated perovskite surface of PEA 2 PbCl 4 in the a-b plane showing the number of ligands per a-b area.e) Calculated ligand densities of perovskites with varying ligand and halide composition and their respective relative changes in ligand density.f) Processing diagram for the TIPS-Pn/perovskite bilayer thin film fabrication.g) Polarized optical microscopy image of the morphology of TIPS-Pn thin films.

Figure 2 .
Figure 2. Crystal structure analysis of TIPS-Pn (101) and (011) diffraction peaks using grazing incidence X-ray diffraction (GIXD).a) Depiction of GIXD technique with incoming X-rays parallel to crystal alignment in TIPS-Pn thin films.b) Diffraction pattern showing the (101) and (011) plane peaks when X-ray beam is parallel to crystal alignment and with 90°rotation, respectively.c) Incremental peak shift of the TIPS-Pn (101) plane was observed on different 2D perovskite surfaces.d) (101) and e) (011) plane spacings of TIPS-Pn on perovskite surfaces with varying ligand and ligand density.f) The Q xy component of the (101) plane showing variation along the -stacking direction.g) The Q z component of the (101) plane suggests variation in the tilting of the c-axis.

Figure 3 .
Figure 3. Optical characterization and singlet fission of perovskite-templated TIPS-Pn.a) Absorbance spectra of TIPS-Pn showing vibronic progression of lowest energy S 0 →S 1 transition (peaks 1, 2, and 3) and intramolecular transitions (peaks 4 and 5).b) Transition energies of - stacking absorbance peak (peak 1) of TIPS-Pn with varying perovskite surface.c) Photoluminescence of TIPS-Pn thin films showing quenching for perovskite-templated samples.d) Transient absorption spectra of TIPS-Pn on PEA 2 PbCl 4 excited at 700 nm (≈700 μJ cm −2).e) Transient absorption at 525 nm following 700 nm photoexcitation of TIPS-Pn on silica (black diamonds) and PMA 2 PbBr 4 (green circles).Fits from global spectral analysis are plotted as solid lines.Data are plotted on a linear timescale from 0 to 1 ps and logarithmic from 1 to 1000 ps.f) Separation (k sep in s −1 , solid red circles) and annihilation (k an in m −1 s −1 , blue squares) rate constants with associated experimental uncertainties for correlated and separated triplets, respectively, generated by singlet fission in TIPS-Pn on perovskites and silica.Horizontal lines (red dashed and blue dash-dot) reflect these respective rate constants for untemplated TIPS-Pn and are drawn as a guide to the eye.g) Diagram of ordered and h) disordered regions of TIPS-Pn showing that enhanced phase-purity (i.e.reduced disorder) leads to enhanced triplet separation.Blue lines represent pentacene backbone of TIPS-Pn.

Figure 4 .
Figure 4. Molecular dynamics simulations of the TIPS-Pn/perovskite interface.a) Reaction scheme of TIPS-Pn and perovskite structures.b) Calculated - stacking distances Form I TIPS-Pn and each TIPS-Pn/perovskite interfaces.