Influence of thermal annealing on microstructure, energetic landscape and device performance of P3HT:PCBM-based organic solar cells

Thermal annealing alters the morphology of organic donor-acceptor bulk-heterojunction thin films used in organic solar cells. Here, we studied the influence of thermal annealing on blends of amorphous regio-random (RRa) and semi-crystalline regio-regular (RR) poly (3-hexylthiophene) (P3HT) and the fullerene derivative [6,6]-phenyl-C60-butyric acid methyl ester. Since the P3HT:PCBM blend is one of the most studied in the OPV community, the existing research provides a solid foundation for us to compare and benchmark our innovative characterization techniques that have been previously under-utilized to investigate bulk heterojunction organic thin films. Here, we combine advanced novel microscopies and spectroscopies, including polarized light microscopy, photo-deflection spectroscopy, hyperspectral photoluminescence imaging, and energy resolved-electrochemical impedance spectroscopy, with structural characterization techniques, including grazing-incidence wide-angle x-ray scattering, grazing-incidence x-ray diffraction, and Raman spectroscopy, in order to reveal the impact of thermal annealing on the microstructural crystallinity and morphology of the photoactive layer in organic solar cells. Coupled transfer matrix and drift-diffusion simulations were used to study the impact of the density of states on the solar cells’ device performance parameters, namely the short-circuit current (J SC), open circuit voltage (V OC), fill factor (FF), and power conversion efficiency (PCE).

In a bulk heterojunction (BHJ) organic solar cell, controlling the nano-morphology of the heterojunction plays a critical role in exciton dissociation and charge transport [23][24][25][26][27][28][29].In most cases, chain alignment and morphology are interrelated, and it is difficult to control both factors simultaneously and favorably [30][31][32].Moreover, the morphology of the BHJ layer depends on the device architecture [33][34][35].In BHJ solar cells, the hole mobility is on the order of 10 −4 cm 2 (V•s) −1 , which has been attributed to interchain transport and disorder that controls the charge carrier mobility [36].Several techniques have been used to promote nanoscale phase separation, polymer crystallization, and chain alignment in BHJ solar cells and FETs [37][38][39][40].Previous work suggests that polymer chain alignment and crystallization can readily be controlled by spin coating conditions, solution concentration and composition, film drying conditions, surface modification, and selectivity of the transport layer [24,[41][42][43][44][45].Post-processing thermal annealing is one of the most commonly used methods to alter the crystallization of semi-crystalline organic thin films [46][47][48][49].
In this work, thin film blends of amorphous regio-random (RRa) and semi-crystalline regio-regular (RR) poly (3-hexylthiophene) (P3HT) with [6,6]-phenyl-C 60 -butyric acid methyl ester (PC 60 BM) were prepared.The loading of PCBM in RRa and RR P3HT was reported previously, showing complex structure pathways [50].In previous work, we presented a study on P3HT:PCBM-based organic solar cells, using several optoelectronic and spectroscopic measurements to investigate their photovoltaic properties.Photoluminescence (PL) studies demonstrated the highest film crystallinity for annealed films, which was confirmed by atomic force microscopy (AFM) [51].An enhancement in crystallinity was primarily responsible for the improved photovoltaic characteristics upon annealing.Building on the previous work, we leverage several advanced characterization techniques to probe the structural and energetic properties of the P3HT:PCBM blends and investigate their impact on solar cell device performance both experimentally in fabricated thin films and devices and theoretically via drift-diffusion calculations.Using energy-resolved electrochemical impedance spectroscopy (ER-EIS), the density of states (DOS) of the different systems was mapped and compared to energy levels obtained from cyclic voltammetry (CV).Grazing incidence wide-angle x-ray scattering (GIWAXS), grazing-incidence x-ray diffraction, polarized light microscopy (PLM), and Raman spectroscopy were used to investigate the degree of crystallinity in the films, while photothermal deflection spectroscopy (PDS) was used to quantify the degree of energetic disorder in the films.Additionally, spectroscopic ellipsometry measurements were performed to obtain the thin films' optical constants, namely, the refractive index and extinction coefficients, which were used as inputs for the coupled transfer matrix-carrier drift-diffusion simulations.

Materials
Two commercially available P3HT polymers were purchased from 1-materials (P3HT-I) and BASF P200 (P3HT-II).Herein, we refer to regio-random P3HT as P3HT-I and to regio-regular P3HT as P3HT-II.These two commercial P3HTs were chosen because they are widely available from suppliers and are commonly used by the OPV community.As an acceptor material, PC 60 BM was purchased from Solene.In order to facilitate the charge extraction properties of the solar cell devices, PEDOT:PSS (P VP AI 4083) was used as a hole transport layer (HTL) as bought from Clevios, and pre-structured indium tin oxide (ITO) substrates were used as received from Xinyan Technology.

Solar cell fabrication
The solar cells were fabricated in a conventional architecture using the following layer stack: glass/ITO/PEDOT:PSS as (HTL)/photoactive layer/Al.The details of the solar cell preparation can be found in the supporting information (SI) section 1.To study the effect of thermal annealing, the complete devices were thermally annealed at 150 • C for 10 min, and all the samples were compared with as-cast or non-annealed samples.

Current-voltage characterization
The current-voltage (I-V) curves were acquired in the dark and under one sun illumination (AM 1.5) using a solar simulator (WAVELABS) and were recorded with a Keithley 2400 Source-Meter-Unit (SMU).The intensity of the light source was calibrated by a standard silicon (Si) photodiode to confirm the AM1.5 condition and an intensity of 100 mW•cm −2 .

Device simulation by SETFOS
1D numerical drift-diffusion simulations were performed using Setfos 5.2 (FLUXiM AG) to predict current density voltage (J-V) curves for single-junction BHJ devices.The optical constants (refractive index and extinction coefficient) were obtained from ellipsometry measurements, as described below.The simulations were conducted using the absorption, drift-diffusion, and optimization packages of Setfos.The remaining simulation parameters are included in the SI.

Ellipsometry
The optical constants (refractive index, n, and extinction coefficient, k) for the active layers were collected by variable-angle spectroscopic ellipsometry (VASE) with an M-2000 ellipsometer (J.A. Woolam Co., Inc).The active layers were cast on clean silicon substrates coated with SiO 2 (∼500 nm).The VASE measurements were performed with incident angles ranging from 50 • to 70 • in steps of 5 • relative to the samples.The software CompleteEASE (J.A. Woolam Co., Inc) was used to process all collected data by first fitting the transparent regime (wavelengths longer than 900 nm) to a Cauchy model to determine the film thickness and then interpolating the visible range optical constants from a B-spline model.An effective medium approximation was assumed for the active layer, which was thus modelled as one material.

Energy-resolved electrochemical impedance spectroscopy
The DOS was characterized by the recently introduced ER-EIS method [52,53].The DOS function, g(E) in organic semiconductors is obtained by measuring the charge transfer resistance, R ct , of a semiconductor/electrolyte interface via the following relation: where, E F = eU is the electrochemical potential adjusted by the external voltage U, e is the elementary charge, k et is the charge-transfer rate constant, [A] is the concentration of the electrolyte redox (donor/acceptor) species in the interface region of the solid/liquid contact, and S is the sample area.The reciprocal value of R ct provides direct information about the electronic DOS at the energy adjusted with an external voltage.The ER-EIS measurements were done in the glove box with nitrogen (N 2 ) atmosphere employing the electrochemical cell with a standard three-electrode configuration.The electrochemical cells with a volume of 200 µl were created by gluing plastic micro cylinders on the P3HT:PCBM layer deposited on the ITO substrate.A solution of 0.1 M tetrabutylammonium hexafluorophosphate (purchased from Sigma Aldrich) in acetonitrile was used as the supporting electrolyte.The working electrode's potential with respect to the reference Ag/AgCl electrode was controlled via a potentiostat, and a platinum (Pt) wire served as the counter electrode.The potential recorded for the reference Ag/AgCl electrode can be recalculated to the local vacuum level, assuming an Ag/AgCl energy vs. vacuum value of −4.66 eV.Impedance was measured with an impedance/gain-phase analyzer (Solartron analytical model 1260).The frequency was set to 0.5 Hz, and the rms value of the AC voltage was 100 mV.

Cyclic voltammetry
Cyclic voltammetry measurements were performed in a BioLogic VMP3 multichannel potentiostat in a three-electrode configuration.A platinum wire and Ag/AgNO 3 electrode were used as counter and reference electrodes, respectively.A 0.1 M solution of tetrabutylammonium tetrafluoroborate in dry acetonitrile was used as a supporting electrolyte.The working electrode was indium tin oxide (ITO), on which P3HT:PCBM blend layers were spin-coated.The scan rate was set to 50 mV s −1 with a maximum potential of 1.5 V. Solutions, and films were prepared inside the glovebox.Ferrocene was used as an internal reference (Fc/Fc + ), with the reported 4.8 V as the ionization energy of Fc/Fc + .The E oxd values were calculated using the measured Ferrocene in the same setup and conditions as the samples.

Grazing-incidence wide angle x-ray scattering
The GIWAXS measurements were carried out with a SAXSLAB laboratory instrument using a CuKα x-ray source (8.05 keV, 1.54 Å) and a Pilatus 300 K detector.The incident angle was 0.2 • for the GIWAXS measurements.The sample-to-detector distance was set to 95 mm.The transformation to q-space, radial cuts for the in-plane and out-of-plane analysis, and azimuthal cuts for orientation analysis were processed by the MATLAB-based package GIXSGUI.

X-ray diffraction
The XRD measurements were carried out by a Bruker D8 Ultra equipped with a Cu-Kα tube of a radiation source (λ = 1.5418Å).The data collection was done from 3 • to 25 • (2θ) at a rate of 0.01 • per second.

Polarized light microscopy
A Nikon Eclipse optical microscope with a Xenon lamp source was used to image the degree of crystallinity of the active layer of thin films.The excitation path was polarized in-plane, and a second polarizer (i.e.analyzer) oriented 87 • relative to the excitation polarization was used to collect the light reflected from the samples.Films were imaged with a 50× objective.The brightness and contrast of the PLM images were adjusted in microsoft power point.

Raman spectroscopy
The Raman spectra were acquired using a WITec alpha 300 apryon confocal Raman microscope at 473 nm excitation, using 10x and 100x objectives in a backscattering configuration.The WITec Project FIVE software was used to analyze the measured Raman spectra, and the fits and background correction was performed with Fityk.The power of the laser ranged from 0.1 to 0.3 mW for excitation with the 473 laser.For the single spectra, the acquisition time was 25 s and averaged 100 times.For the Raman mapping, the acquisition time was set to 0.3 s.

Photothermal deflection spectroscopy
The PDS was performed using a home-built, transverse PDS set-up.Light from a 250 W tungsten-halogen lamp (Newport 66996-250Q-R1) was sent through a monochromator (LOT MSH-300) and acted as a pump source, allowing for selective excitation across the ultraviolet to near-infrared spectral region.An optical chopper operating at a frequency of 4 Hz modulated the monochromatic pump light, which was then focused on the sample.The sample was placed in a cuvette and immersed in a chemically inert liquid, namely Perfluorohexane (SigmaAldrich 281 042 99% C 6 F 14 ).A 633 nm cw-laser (Thorlabs HR S015) was used as the probe beam, which was sent to grace the sample's surface.Using a silicon quadrant detector (Thorlabs PDQ80A) and a lock-in amplifier (Stanford Research Systems SR830), the deflection of the probe beam was measured.

Hyperspectral photoluminescence
Hyperspectral PL imaging was done using a hyperspectral microscope (Photon etc).The active layer films were deposited via spin-coated on a quartz substrate.The samples were excited from the top surface with a 532 nm laser and focused through a microscope objective (20 times) with an exposure time of 5 s and power of 1500 W. Hyperspectral images were analyzed using PHySpecV2 software provided by Photon etc [54,55].

Results and discussion
Generally, thermal annealing can improve the photovoltaic performance of P3HT:PCBM blends obtained by spin-coating from solution.Thermal annealing alters the morphology, resulting in demixing of the polymer and acceptor components in the thin film blend.Additionally, thermal annealing enhances the crystallization of both components in the blend, which is beneficial for charge carrier transport and more efficient charge carrier extraction in organic solar cells.

Photovoltaic properties
The chemical structures of RRa and RR P3HT and PC 60 BM are shown in figure 1(a).The J-V curves of the non-annealed solar cells under illumination, shown in figures 1(b) and (c), indicate that the devices have no apparent current saturation.The photovoltaic parameters for the devices are summarized in table S1 in the SI.A weak s-shape of the J-V curve is observed for the non-annealed P3HT-I:PCBM solar cell, indicating the presence of a counter diode caused by either charge carrier accumulation at the interface or low charge carrier mobility (table S2 in SI).The non-annealed devices of both types of P3HT show insufficient component demixing, and the low device performance can be attributed to poor donor (D) and acceptor (A) percolation pathways in the photoactive layer, leading to low charge carrier mobility [16,56].In the case of post-annealing, the morphology is altered, and the mobility is improved, as indicated by the increased J SC .
Using the n and k values from spectroscopic ellipsometry measurements as inputs, J-V characteristics determined experimentally under solar-like illumination were reconstructed using coupled transfer matrix-drift-diffusion simulations.The electrical characteristics of the photoactive layer (e.g. carrier mobility and recombination constants) were used as fitting parameters for the J-V simulations and are summarized in table S2.The simulated J-V curves (solid lines) agree well with the experimentally measured J-V curves.

Density of states distribution
Bulk-heterojunction (BHJ) thin film organic solar cells need separated donor (D) and acceptor (A) domains in order to efficiently separate and extract charge carriers.Since the transport of electrons and holes occurs in different domains, charge carrier recombination occurs only at the D-A interface.A description of the charge transport and recombination in the BHJ cell must consider the electronic structure of the materials given by the DOS distribution.The relevant DOS is derived from the properties of D-A blends in the cell.The transport levels are mainly attributed to the polymer donor's highest occupied molecular orbital (HOMO) and the acceptor's lowest unoccupied molecular orbital (LUMO).Furthermore, the DOS is an essential parameter that links the structural properties of the BHJ photoactive layer to the electronic properties of the solar cells [57].
Several factors, such as device fabrication, material composition, and the degree of phase separation in the photoactive layer, influence the physical structure and, thus, the electronic properties.Therefore, a comprehensive understanding of the device requires studying the impact of atomic and nanoscale structure on the DOS and, ultimately, the optoelectrical properties of solar cells.
Drift-diffusion simulations were used to investigate the impact of the DOS on the photovoltaic properties using the experimentally obtained optical parameters.Using the parameters optimized for the experimentally measured J-V curves (from figures 1(b), (c) and table S2), we repeated the simulations by varying the effective DOS at the HOMO and LUMO levels from 10 18 cm −3 to 10 22 cm −3 for each active layer system (figure 2).These simulations serve as a model to illustrate the impact that DOS can have on the device performance parameters.The DOS had a large impact on the J-V curves of each active layer system, especially on the V OC .The photovoltaic parameters (J SC , V OC , fill factor (FF), and power conversion efficiency) as a function of the effective DOS varied from 10 18 cm −3 to 10 22 cm −3 are shown in SI, figure S2.
The J SC for non-annealed devices were slightly affected due to the increased effective DOS, while almost no changes were observed for annealed devices.Among the essential photovoltaic parameters, V OC strongly depends on the DOS in organic solar cells [58].Previous research has demonstrated that the energy difference between the donor's HOMO and the acceptor's LUMO influences the maximum V OC value [59,60].The relation between energy levels and the V OC can be determined by the following relation; qV OC = HOMO donor − LUMO acceptor − ∆ where ∆ is the loss term, which can be linked to radiative and non-radiative recombination polaron pairs [61].
There is agreement that the distribution of localized states is decisive in defining V OC , despite continuing discussions over the physical mechanisms contributing to the loss term, ∆ [62].Tail states (higher DOS) cause a lower effective energy gap and a reduced V OC .In the donor photo-generated holes reside in tail states, bringing electrons and holes energetically closer.This characteristic in organic solar cells has been considered using numerical device modelling and experimental data [63].The effect of energetic disorder on the V OC applying numerical simulation was well analysed by Blakesley and Neher [64].They noted that a broader DOS reduces V OC as energetic disorder increases, resulting in lower FF.
It was reported that the ER-EIS method is instrumental not only to detemine the DOS distribution of neat organic films but also their blends [57,[65][66][67].Therefore, this method was employed for P3HT-I:PCBM and P3HT-II:PCBM blends on both non-annealed and annealed films to delineate DOS changes due to variations in the fabrication process of the active layer.The DOS spectra g(E) on linear-linear scale are shown in SI, figure S3, from which positions of HOMO P3HT and LUMO PCBM are extracted from the onset peaks (indicated with dashed lines).The annealed devices had larger DOS values compared with the non-annealed devices, which resulted in lower device V OC values (table S1), in agreement with the simulation results described above.The values of the effective bandgap calculated as a difference of HOMO P3HT and LUMO PCBM are summarized with the HOMO P3HT and LUMO PCBM positions in table 1.For comparison, the extracted HOMO levels were found to be in good agreement with those measured with cyclic voltammetry (CV), as seen in figure S4 and table S3 in the SI.
There is an increase of the HOMO P3HT energy from −5.75 eV for the P3HT-I:PCBM non-annealed layer to −5.21 eV for the annealed layer and a decrease in LUMO PCBM energy from −3.56 eV to −3.76 eV.Consequently, the effective bandgap was reduced by 740 meV, from 2.19 eV to 1.45 eV upon annealing.In measured devices, the higher effective bandgap resulted in a higher V OC , with the non-annealed device exhibiting a V OC of 718 mV, while the V OC for the annealed device was 642 mV.The P3HT-II:PCBM layer showed a relatively higher HOMO P3HT and lower LUMO PCBM compared to the P3HT-I:PCBM layer, with a increase of the HOMO energy from −5.41 eV for the non-annealed layer to −5.23 eV for the annealed layer, and a decrease in LUMO energy from −3.79 eV to −4.35 eV.
To anticipate the origin of the observed HOMO and LUMO bands, it is helpful to depict the DOS distributions along with the parent neat films on the log-linear scale (figure 3).These comparisons imply that in annealed P3HT-II:PCBM film the effective band gap is determined by the HOMO P3HT and LUMO PCBM positions of the neat films.The HOMO position of the non-annealed P3HT-II:PCBM blend film moves to lower (deeper) energy with increased disorder.Comparing the two non-annealed films, more disordered film, namely P3HT-I:PCBM, exhibited a lower HOMO level, which indicated that increasing disorder leads to lower HOMO levels.However, it can be associated with the HOMO P3HT .On the other hand, the increased disorder shifts the LUMO position to lower energies correlating with the peaks at −4.15 eV and shoulders at −3.7 eV and −3.2 eV in the LUMO of the neat PCBM.It was suggested that the peak and shoulders are connected with the crystal-like domains and more disordered PCBM [65].The rationale behind this is that upon increased ordering, the ionization energy (IE) and electron affinity (EA) of a solid decrease and increase, respectively.In addition, the observed changes in the electron structure of the investigated system are consistent with the dependence of the V OC and the total voltage loss on fullerene aggregation, donor crystallinity, phase separation, and the effective band gap [58].

Structural properties
The molecular structure of non-annealed and annealed P3HT:PCBM blend layers was investigated by GIWAXS measurements.Figure 4(a) shows the 2D GIWAXS data of the blends, and the corresponding out-of-plane (OOP) and in-plane (IP) line profiles are given in figure 4(b).The non-annealed samples have the (100) peak located at 0.37 A −1 in the IP direction and a broad PCBM scattering ring around 1.3 A −1 , agreeing well with previously reported results [68,69].Additionally, an OOP (010) peak is located at around 1.71 A −1 , revealing the π-π stacking crystallinity of P3HT.The P3HT-II:PCBM films show more pronounced Bragg peaks due to the higher degree of crystallinity.Moreover, for the P3HT-II:PCBM film, the IP (200) peak located at 0.74 A −1 is also visible.
After thermal annealing, the films display a noticeable difference in the GIWAXS data and their scattering profiles due to the improved crystal parts of the films.The (010) peaks differ for different samples in the OOP direction, suggesting an altered π-π stacking distance.As expected, the crystal numbers and size are larger for the P3HT-II blend due to its semi-crystalline properties.Additionally, thermal annealing increased the size of the crystals in both blends.No noticeable shifts can be observed for (100) and (200) peaks; however, the intensity of these peaks increased, suggesting that the crystal number also increased after thermal annealing.
Grazing incidence XRD measurements were performed to study the conformation of the P3HT's crystal structure and corroborate GIWAXS results.Figure 5 shows the XRD patterns for the non-annealed and annealed P3HT blends with a pronounced peak at 2θ • = 5.51 ± 0.01 for all samples.Upon annealing, a shift in peak position, a substantial increase in peak intensity, and a reduction in peak width were observed.
For P3HT, the stacking structure is defined by the polymer backbone, π-π stacking direction, and lamellar direction.In the P3HT-PCBM blends, the P3HT stacking structure changes only in the lamellar crystal structure or a-direction upon annealing [70,71].
Furthermore, as shown in table S4, an increase in d-spacing and crystallite size upon annealing is observed, which leads to an enhancement in the intermolecular transfer of charge carriers and higher mobility [72,73].In conclusion, the annealed P3HT-II:PCBM blend showed the highest degree of crystallinity, and thermal annealing enhanced the crystallinity of both P3HT blends, agreeing well with previous GIWAXS and Raman results.

Morphological properties
While GIWAXS gives a very direct and quantitative understanding of the crystallinity and molecular orientation of organic semiconductor thin films, it normally requires the use of a synchrotron facility and  intense (often damaging) high-energy x-ray probes.However, the crystallinity and molecular orientation of molecular thin films can also be qualitatively evaluated using the relatively unexplored technique of PLM (figure 6) [74].By using a set of orthogonal polarizers in the excitation and collection paths, most of the incident light is filtered out from the image, and only the local optical birefringence is imaged, providing enhanced contrast relative to standard bright-field optical microscopy (SI figure S5).Since organic semiconductors can have high degrees of anisotropy based on their molecular orientations and degrees of crystallinity, PLM can qualitatively image the morphology of organic thin films in a rapid and non-destructive manner.The most amorphous films, i.e.P3HT-I:PCBM (non-annealed) (figure 6(a)), appeared uniformly in PLM images, not exhibiting any striking features.Once annealed (figure 6(b)), visible PCBM crystallites were observed [75], with varying degrees of brightness and contrast, which arose from differences in the alignment of the crystallites relative to the polarizers.Additionally, some texturing appeared in the background of the PLM image, which was attributed to changes in the local crystallinity of the mixed P3HT:PCBM region.
For the semi-crystalline P3HT-II:PCBM film, the PCBM crystallites were not present prior to annealing (figure 6(c)), but there was significantly more texture to the overall film compared to P3HT-I:PCBM, indicating an overall increase in the crystallinity.Upon annealing, there were two drastic changes observed: (1) an increase in the number and elongation of the PCBM crystallites, with crystallites aligned with the polarizers appearing dark and those at acute angles appearing much brighter, revealing information about the packing of PCBM molecules; (2) the overall background had drastically more texture, suggesting an overall increase in the crystallinity of the mixed P3HT:PCBM domains (figure 6(d)).
Raman spectroscopy is a widely employed technique to investigate chemical structure, polymorphism, crystallinity, and molecular interactions [76][77][78][79][80][81][82][83][84].In particular, resonant Raman spectroscopy has been previously used to investigate molecular order in the non-annealed and annealed films of P3HT:PCBM [82,85].Under resonant excitation at 473 nm, well-defined Raman peaks for P3HT can be seen between 724 and 1460 cm −1 (see SI figure S6 and table S5).Comparing the Raman spectra between the non-annealed and annealed P3HT:PCBM blends, as shown in figure 7, the peak position of the symmetric C=C stretching mode of the P3HT polymer chain at ∼1450 cm −1 shifted to lower wavenumber upon annealing, for the P3HT-I:PCBM film, the peak shifted from 1454 cm −1 to 1446 cm −1 (figure 7(a)); whereas for the P3HT-II:PCBM film, the peak shifted from 1450 cm −1 to 1447 cm −1 (figure 7(b)).Additionally, annealing induced a reduction in the full-width-half-maximum (FWHM) of this peak, reducing from 46 cm −1 to 28 cm −1 for the P3HT-I:PCBM film and from 44 cm −1 -31 cm −1 for the P3HT-II:PCBM film.The observed changes in peak position and FWHM of the peak corresponding to the C=C stretching mode indicate an enhancement in molecular order upon annealing.
The degree of molecular order in the P3HT:PCBM films can be estimated by fitting the C=C peak, since the peak can be fitted as a superposition of the two phases, namely the disordered RRa and the ordered RR P3HT phases [82,84].By comparing the peak height of the contribution from the RRa phase to the contribution from the RR phase, an increase in molecular order was observed upon annealing.Figure S7 in SI shows an example of the fitting for the C=C peak, and a summary of the fits can be found in table S6.
In addition to using Raman spectroscopy to correlate device performance to chemical structure, Raman spectroscopy can be combined with optical imaging to obtain high-resolution compositional maps on the sub-micrometer scale, which can provide insight into local variations in morphology and thickness, for example [83,[85][86][87][88].To investigate the local variations in composition and morphology in the annealed P3HT-I:PCBM film, confocal spectroscopic mapping was performed by collecting 100 × 100 spectra in a ) was performed to investigate the surface composition of the films.From the cluster analysis, two distinct Raman spectra were detected, as can be seen in figure 8(c).The red region corresponds to the mixed P3HT:PCBM region, as the Raman peaks for P3HT can clearly be seen, while the blue region corresponds to PCBM aggregates.
To investigate local variations in morphology, the peaks around ∼1450 cm −1 were fit, and the Raman images were processed to show local distributions in peak position and peak width, shown in figure S8.The distribution of the Raman peak position was found to be ±5 cm −1 , and only slight contrast can be observed in the film, indicating a homogeneous distribution of P3HT and similar crystallinity.The distribution of the Raman peak width was found to be slightly broader; however, the optical contrast of the Raman image is consistent throughout the film.The optical images for the other P3HT:PCBM films, as well as the Raman images and cluster analysis for the annealed P3HT-II:PCBM film can be found in figures S9-S13 in the SI.

Optical properties
PDS is a highly sensitive optical technique used to study the degree of energetic disorder in semiconductor thin films by probing the presence (absorption) of sub-gap states [89,90].Figure 9 shows the PDS spectra for the four P3HT:PCBM blends.From the spectra, the Urbach energies were calculated using the following Urbach equation [91]; where α 0 is the absorption coefficient at the gap energy E g , E denotes energy, and E U is the Urbach energy.For all samples, the gap energy was found to be around 1.96 and 1.94 eV, for P3HT-I and P3HT-II, respectively (figure S15, in SI).
Upon annealing, the calculated Urbach energies for the P3HT-I:PCBM and P3HT-II:PCBM films decreased from 500 meV to 59 meV and from 152 meV to 52 meV, respectively.The Urbach energy  Finally, hyperspectral PL imaging was used as a non-destructive technique combining spectroscopy and luminescence imaging since it is particularly useful for investigating spatial variations in solar cells [92,93].In this study, the hyperspectral PL imaging method was utilized to map the surface emission properties on each point of the various P3HT PCBM blends.For the non-annealed films, as shown in figure 10, the spatial distribution of the PL emission peaks is homogeneous, i.e. featureless, indicating well-mixed layers of P3HT and PCBM.PCBM crystals can be noticeably seen in the annealed films, as evident by their redshifted emission peaks compared to the rest of the film [67][68][69].In the darker regions, especially in the annealed P3HT-II:PCBM film, some contrast can be seen, which indicates disordered and ordered P3HT phases.The average PL spectra for the probed areas of the films can be found in figure S16.

Conclusions
Comprehending and controlling the photoactive layer morphology of thin film organic solar cells is crucial to fully harness their photovoltaic potential.Device performance is intricately linked to thin film morphology, and being able to unravel morphology at different scales and understand its impact on device performance remains a complex task.In this work, we applied several spectroscopic techniques in combination with drift-diffusion simulations to investigate the influence of crystallinity on solar cell device performance and energetic disorder in blends of P3HT:PCBM.Thermal annealing significantly improved the crystallinity of the P3HT:PCBM blends as evidenced by GIWAXS, XRD, polarized light microscopy, Raman, and hyperspectral measurements, which led to a lower effective band gap as revealed by ER-EIS and drift-diffusion simulations, respectively, and ultimately higher device performance.

Figure 6 .
Figure 6.Polarized light microscopy images for the non-annealed and annealed P3HT-I:PCBM (a, b) and P3HT-II:PCBM (c, d) blends at 50x magnification, respectively.The scale bar in all images is the same.

Figure 8 .
Figure 8.(a) The optical image with the red square indicating the scan region, (b) cluster analysis on the Raman data where the red area represents the mixed P3HT:PCBM region and the blue region represents the PCBM crystals and (c) distinct Raman spectra of the P3HT-I:PCBM annealed film.

9 .
(a) Photothermal deflection spectroscopy (PDS) spectra for the four P3HT:PCBM thin films, (b) Urbach energy determined from Urbach fits to the tail of the absorption below the bandgap energy.

Figure 10 .
Figure 10.Hyperspectral images for the four blends.In the annealed films, clear aggregates of PCBM are visible.