In Situ Determination of the Orientation of the Emissive Dipoles in Light‐Emitting Electrochemical Cells

The orientation of the emissive dipoles in thin‐film devices is important since it strongly affects the light outcoupling and thereby the device emission efficiency. The light‐emitting electrochemical cell (LEC) is particularly interesting in this context because its emissive dipoles are located in a high electric‐field p‐n junction, which is formed in situ by redistribution of bulky ions. This implies that the dipole orientation could be distinctly different in the driven LEC compared to the pristine device. This study develops the destructive‐interference microcavity method for the accurate in situ determination of the orientation of the emissive dipoles during LEC operation and apply it on a common LEC device comprising an amorphous conjugated polymer termed Super Yellow as the emitter. It is found that ≈95% of the emissive dipoles are oriented in the horizontal direction with respect to the thin‐film plane in both the pristine LEC and during steady‐state light emission. This finding is attractive since it enables for efficient outcoupling of the generated photons, and interesting because it shows that a horizontal orientation of the emissive dipoles can remain despite the existence of a strong perpendicular electric field and the nearby motion of bulky ions during LEC operation.


Introduction
Thin-film devices based on emissive organic semiconductors, notably the organic light-emitting diode (OLED) and the light-emitting electrochemical cell (LEC), can deliver soft areal emission from flexible and lightweight device architectures. Although these advantages render such thin-film devices highly attractive for existing and emerging display and illumination applications, [1] a challenge is that their emission efficiency can be severely limited due to poor light outcoupling. This challenge originates in that a large fraction of the photons generated in their emissive material is coupled into various wave-guided and surface plasmon-polariton modes and, therefore, are unable to exit the device structure. [2] A number of functional methods to alleviate these outcoupling losses have been developed, [3] including the employment of microlens arrays and corrugated substrates, [4] the inclusion of periodic nanostructures between the reflective electrode and the neighboring organic layer, [5] and the tuning of the refractive index of the different organic layers. [6] It is also well established that the orientation of the transition dipole moment of the emissive organic semiconductor has a strong influence on the outcoupling, and that its appropriate orientation can significantly improve the emission efficiency. From here on, we will term the transition dipole moment of the emissive organic semiconductor "emissive dipole" for simplicity. For instance, it has been shown that a shift of the orientation of the emissive dipole from isotropic to parallel or horizontal to the thinfilm plane can result in a 50% improvement of the emission efficiency in OLEDs. [7] A shift to a horizontal dipole orientation can also bring the additional advantage that the electronic transport capacity is facilitated by increased intermolecular -orbital overlap. [8] The LEC is distinguished from the OLED by that mobile ions are blended with the organic semiconductor in its active material, [9] which in turn has paved the way for the fabrication of LECs with cost-and material-efficient ambient-air printing and coating methods. [10] When a voltage is applied between the LEC electrodes, these mobile ions redistribute to first enable for the formation of electrical double layers (EDLs) at the electrode/active-material interfaces, and thereafter for the The black arrows represent the dipole orientation, and the thick red lines indicate the directions of the generated light rays. b) Schematic of the spectrogoniometer setup for the measurement of the angle-dependent emission intensity and spectrum. A half-cylinder lens is attached on the transparent LEC substrate, and this assembly is rotated around the x-axis in the (green) y-z plane, with the emission angle indicated. A linear polarizer is included before the detector, which defines the polarization of the detected light. c) Schematic description of the specific simulated glass/indium-tin-oxide (ITO)/active-material/Al LEC device structure, including the steady-state doping structure in the active material as detailed in the text and in the Experimental Section. d) The simulated steady-state external quantum efficiency as a function of the active-material thickness for three different orientations of the emissive dipoles, as specified in the legend. electrochemical doping of the organic semiconductor; p-type at the positive anode and n-type at the negative cathode. These doping processes continue until a thin p-n junction has formed in the active material, at which subsequently injected electrons and holes recombine into excitons (or excited emissive dipoles) that can decay radiatively. [9a,11] This unique in situ transformation of the LEC active material by electrochemical doping is both interesting and challenging from the perspective of the orientation of the emissive dipoles. Notably, the LEC-characteristic motion of bulky ions in the close proximity of the emissive dipoles (i.e. the excited organic semiconductor) during the turn-on phase, and the existence of a large electric field of ≈10 8 V m −1 in the thin and undoped p-n junction at steady state, suggests that it is reasonable to anticipate that the emissive molecules can reorient during LEC operation. [12] Thus, it is highly motivated to study the orientation of the emissive dipoles in the LEC active material both in the pristine state (before a voltage is applied) and during light emission to establish whether the emissive dipoles reorient or remain constant during LEC operation. This piece of information will in addition enable for a more rational LEC design for improved emission efficiency.
Here, we report on the accurate determination of the orientation of the emissive dipoles in LEC devices in the pristine state and during light emission. Specifically, we utilize a combination of angle-resolved photoluminescence spectroscopy (ARPS) and ellipsometry for the study of the dipole orientation in the pristine active material, and an LEC-relevant adaptation of the destructive-interference microcavity method for the determination of the emissive dipole orientation during LEC operation. For this first study, we select a common LEC device architecture featuring an amorphous high molecular-weight conjugated polymer termed Super Yellow as the emissive organic semiconductor. Somewhat unexpectedly, we find that ≈95% of the emissive dipoles are oriented horizontal to the thin-film plane of the device, in both the pristine state and during LEC operation. Our study thus establishes that the horizontal orientation of the Super Yellow polymer is robust to nearby ion motion and to the exposure of a high electric field, and that this device structure consequently features a high light outcoupling efficiency that remains stable during longterm LEC operation. Figure 1a is a schematic of a general LEC architecture, which comprises a reflective electrode and a transparent electrode on a transparent substrate that sandwich a single-layer active material. The figure also depicts three principle and orthogonal orientations of the emissive dipoles in the active material: one vertical dipole (⊥) that emits p-polarized light, and two horizontal dipoles (∥) that either emit p-polarized or s-polarized light. The vertical dipole is oriented perpendicular to the plane of the thinfilm LEC device (i.e., the x-y plane), while the horizontal dipoles are aligned parallel to the same thin-film x-y plane; the orientation of an arbitrarily oriented emissive dipole can be described by a linear combination of these three orthogonal emissive dipoles. The red arrows in the figure depict the generated light rays in the direction perpendicular to the corresponding dipole. Figure 1b presents a schematic description of the spectrogoniometer measurement setup. A half-cylinder lens is attached to the transparent substrate of the device under study, and this assembly is mounted on a rotary stage that is rotated around the xaxis, i.e., in the (green) y-z plane, with the emission angle of the detector with respect to the device normal being . A linear polarizer is included before the detector, which determines whether the p-polarized light rays (E-field component in the green y-z plane) or the s-polarized light rays (E-field component in the yellow x-z plane) reach the detector.

Results and Discussion
In order to elucidate the importance of the emissive dipole orientation for the LEC performance, we performed an initial optical simulation of the common LEC device architecture depicted in Figure 1c. This particular LEC comprised Al as the reflective negative electrode, indium-tin-oxide (ITO) as the transparent positive electrode, glass as the transparent substrate, and the fluorescent conjugated polymer "Super Yellow" as the emissive organic semiconductor blended with a hydroxyl-capped trimethylolpropane ethoxylate (TMPE-OH):KCF 3 SO 3 electrolyte for the active material (but no half-cylinder outcoupling structure). Figure 1c displays the simulated steady-state doping structure in the active material, which comprises a p-type doping region and an n-type doping region sandwiching an emissive p-n junction. The two doping regions feature constant doping gradients, with the highest doping next to the electrode and zero doping concentration at the boundary to the p-n junction. Each doping region was discretized into 10 constant-doping regions for the simulation, and the corresponding doping-dependent optical properties were gleanedfrom from Ref. [13]. The undoped p-n junction is simulated as a distinct (and transparent) region that is positioned in the center of the active material, with a thickness corresponding to 15% of the total active-material thickness, i.e., 0.15•d AM . The distribution of the excited emissive dipoles (or excitons) within the p-n junction was simulated by a peak-shaped function with its peak value in the center of the p-n junction (see Experimental section for further details). Figure 1d presents the simulated external quantum efficiency (EQE), i.e., the ratio between the photons exiting the device structure and the electrons entering the device, of this LEC as a function of the active-material thickness, d AM . The only difference between the three simulations is the orientation of the emissive dipoles in the p-n junction region, which is either: horizontal, isotropic, or vertical (see legend). The observed periodic variation of the EQE (i.e., the emission efficiency) with d AM is due to that the LEC device forms a weak optical microcavity, implying that the photons generated by the excited emissive dipoles in the p-n junction are affected by constructive or destructive interference depending on their original distance from the reflective electrode. [14] The first position of constructive interference for this particular device structure, with a centered p-n junction (and with a horizontal or isotropic dipole orientation), is achieved with d AM = 155 nm, i.e., with the largest number of excited emissive dipoles positioned at a distance of ≈77 nm from the reflective Al cathode. Importantly, Figure 1d clearly visualizes the impact that the orientation of the emissive dipoles has on the emission efficiency.
A horizontal dipole orientation is strongly preferred (except at the destructive interference point), since it results in that the light-emission rays can be directed perpendicular to the thin-film plane of the device and thereby be efficiently outcoupled into the surroundings, where they can be detected by an external observer. In contrast, the vertically oriented dipoles emit preferentially parallel to the thin-film plane, and these light rays are therefore strongly coupled to wave-guided and surface plasmon-polariton modes, [7b] in which they eventually are "lost" as heat by selfabsorption within the device structure. The isotropic dipole distribution comprises a 2:1 ratio of horizontal and vertical dipoles, and thus exhibits an intermediate value for the emission efficiency.
We begin our experimental investigation by determining the emissive dipole orientation in pristine thin films with the method of ARPS, as developed by Frischeisen et al. [2c] and as described in detail in Refs. [2c,f,15]. The far-field emission intensity from a thin film (or a thin-film device) is a function of the emission angle ( ), the emission wavelength ( ), and the anisotropy coefficient of the emissive dipoles (a). It can be described by the linear combination of the contributions from the three orthogonal emissive dipoles depicted in Figure 1a: [16] (1) where ⊥ represents the vertical emissive dipoles, ∥ the horizontal emissive dipoles, and p and s identify whether the light wave is p-polarized or s-polarized. The anisotropy coefficient a is equal to the fraction of vertical emissive dipoles with respect to the total number of emissive dipoles; a = 1 thus corresponds to solely vertical active-material emissive dipoles, a = 0 to only horizontal emissive dipoles, and a = 1/3 to an isotropic emissive-dipole distribution.
The photoluminescence (PL) emission intensity was first measured as a function of and with the setup presented in Figure 1b, and with the "emission spot" also being the "PL excitation spot." Thereafter the same metric was simulated using an optical model based on the transfer-matrix formalism. [15b] We could thereafter determine the emissive dipole orientation by identifying the simulated value for a that resulted in the best agreement between simulation and measurement. However, since the relative contribution of the vertical emissive dipoles to the measured emission intensity (i.e., the first term in Equation (1)) is very minor for a thin film because of poor outcoupling (see Figure 1d), we opted to enhance it by attaching the half-cylinder lens to the glass substrate (see Figure 1c). We further suppressed the horizontaldipole contribution by solely measuring the p-polarized light rays (thus cancelling the third term in Equation (1)). Further details on this procedure can be found in the Experimental Section. Figure 2a presents the derived average value and the standard deviation for the relative number of vertical emissive dipolesi.e., the anisotropy coefficient a-in three different thin films (thickness = 40 nm) on a glass substrate: Super Yellow: 7 ± 3%; Super Yellow blended with the TMPE-OH ion transporter: 8 ± 3%; and the active material (i.e. Super Yellow blended with TMPE-OH and the KCF 3 SO 3 salt): 8 ± 3%. The fact that the films feature a highly anisotropic dipole distribution, with 90-95% of the emissive dipoles being horizontal, i.e., oriented parallel to the thin-film plane, implies that the constituent emitter, the amorphous long-chain polymer Super Yellow, has a strong propensity for orienting parallel to the substrate. We note that Li et al. reported a similarly high anisotropy for a pristine Super Yellow thin film. [17] We also note with interest that the addition of the ion transporter and the salt to the Super Yellow thin film had no measurable influence on its emissive dipole orientation.
We also investigated the pristine thin films with the technique of variable-angle spectroscopic ellipsometry (VASE). The spectral dependence of the ellipsometric parameters was first measured, and thereafter simulated using either an isotropic or a uniaxial dispersion model. We find that the anisotropic model consistently yielded a much better fit than the isotropic model, which further supports our finding of a highly anisotropic distribution of the emissive dipoles in Super Yellow. The best-fit model parameters are presented in Table S1 (Supporting Information).
The VASE experiment also delivered information on the axial dependence of key optical properties of the thin films. Figure 2b presents the wavelength-dependent refractive index, n, (upper panel) and the extinction coefficient, k, (lower panel) in the in-plane (II) and out-of-plane (⊥) direction of the activematerial film. Figure S1 (Supporting Information) presents the corresponding data for the Super Yellow film, and the observed similarity between these data demonstrates that the inclusion of the ion-transporter and the salt had a relatively negligible effect on n and k.
Interestingly, Figure 2b reveals that both n and k are highly anisotropic, with the in-plane component being notably larger than the out-of-plane component. The refractive index n is directly related to the molecular polarizability (as described by the Lorentz-Lorenz equation), with a larger n corresponding to a higher molecular polarizability. [18] Accordingly, the birefringence, defined as Δn = n ⊥ − n ∥ , is a measure of the anisotropy of the molecular polarizability. The thin films exhibit a negative birefringence for a vast majority of the investigated wavelengths, e.g., Δn = −0.17 at = 600 nm for the active material, which shows that the molecular polarizability is higher in the thin-film plane than in the direction perpendicular to the same plane. For polymers, the molecular polarizability, and therefore the refractive index, is commonly larger in the direction of the polymer backbone, since the electrons often migrate easier along the backbone. [18,19] Thus, the observed anisotropy in the refractive index is in good agreement with the preferential in-plane orientation of the Super Yellow polymer deduced from the above ARPS measurement. We note that similar birefringence values have been reported for polymeric thin films based on the same poly(para-phenylene vinylene) backbone as Super Yellow. [20] It should further in principle possible to obtain an estimate of the orientation of the emissive dipoles from the VASE measurement by setting two different equations for the parameter S equal: where k max ∥ and k max ⊥ are the maximum values of the in-plane and out-of-plane extinction coefficient of the absorption band corresponding to the respective dipole transition. [18,19] We find that the derived values for k max ∥ (at = 456 nm) and k max ⊥ (at = 432 nm) yields a value for the anisotropy coefficient of a = 0.22 for the pristine active material and of a = 0.14 for Super Yellow (see Figure S1, Supporting Information). This exercise accordingly provides support for a preferential horizontal orientation of the emissive dipoles in Super Yellow, although it also suggests that the magnitude of this preferential orientation is smaller than predicted by the ARPS measurement. We tentatively assign this discrepancy to that the VASE analysis, as performed herein, is of lower accuracy, since it, for instance, neglects the energetic disorder that is common in organic polymers by only considering the main extinction peak in the calculation.
We now turn our attention to the determination and analysis of the dipole orientation in electrically driven LECs during light emission. Figure 3a and Figure S2 (Supporting Information) present the measured transients for the drive voltage (solid blue circles) and the forward EL intensity (open orange squares) for a glass/ITO/active-material/Al LEC, with an active-material thickness of d AM = 300 nm. The observed initial decrease of the voltage with time is a characteristic indicator of a functional LEC, since the initial formation of injection-facilitating EDLs improves the electron and hole injection, while the subsequent electrochemical p-type and n-type doping (and the concomitant formation of a p-n junction) increases the electron and hole transport.
The determination of the emissive dipole orientation in thin-film devices can be performed with the destructive interference microcavity method. For OLEDs, the attainment of www.advancedsciencenews.com www.advmattechnol.de destructive interference is effectuated during the device fabricationby, e.g., appropriately tuning the thickness of the electron transport layer in between the reflective electrode and the thin emission layer. [2a,b,d,e,7b] . This designed fabrication approach is however not a straightforward option for LECs because of its dynamic-doping operation. We therefore had to determine the p-n junction doping structure required for destructive interference through dedicated in situ measurements during device operation.
Specifically, we systematically determined the p-n junction center position and width during constant-current LEC operation by first measuring the s-polarized EL spectra as a function of emission angle, using the setup depicted in Figure 1b, and by thereafter simulating the same data with the p-n junction position and width as the free parameters. The utilization of the s-polarized emission for this procedure improved the accuracy, since it only includes a contribution from the horizontal dipoles; more specifically, the third term in Equation (1). Therefore, knowledge of the dipole orientation is not required for this fit. Figure 3b presents the measured (left panel) and the simulated (right panel) s-polarized emission intensity as a function the emission wavelength and the emission angle, and the observed excellent agreement between simulation and experiment provides support for the validity of the procedure.
We find that the center position of the p-n junction (as indicated by the green crosses in Figure 3a) migrates significantly during LEC operation, and that it first forms at d pn = 0.61 (corresponding to 61% of the active-material thickness away from the positive anode) and then slowly moves to reach d pn = 0.42 after 8 h of operation. This migration of the p-n junction in LEC devices has been demonstrated to be a manifestation of that the cation/anion mobility ratio dictates the initial p-n junction position, whereas the electron/hole mobility ratio determines its steady-state position; [21] and that, accordingly, these ratios are different in this particular device.
An optical analysis of this device structure, with d AM = 307 nm, yields that its destructive interference (i.e., its optical minimum) is obtained at d pn = 0.51. Therefore, close to this point in time (as indicated by the vertical dashed black line in Figure 3a), we rotated the linear polarizer so that it switched from recording s-polarized light to p-polarized light. This procedure is motivated by Equation (1), which reveals that the switch to detection of p-polarized light enables for the detection of both the vertical dipoles (term 1) and the horizontal dipoles (term 2). The contribution from the vertical dipoles is commonly much weaker than that of the horizontal dipoles, but at the specific point of destructive interference they are of similar magnitude, as visualized in Figure 1d.
By fitting the simulated EL intensity to the measured EL intensity, it is possible to derive the value for the anisotropy coefficient a in Equation (1), and thereby the relative contributions of the vertical and the horizontal dipoles to the EL emission during LEC operation. We have executed eight measurements on three independent LEC devices and find that the value for a is ranging between 0.02 and 0.05. Accordingly, 95-98% of the emissive dipoles in the p-n junction are oriented horizontally after 5 h of LEC operation.
The validity of this in situ derivation of the dipole orientation in driven LECs is visualized in Figure 3c, which presents the mea-sured p-polarized EL intensity as a function of emission wavelength and emission angle in its left panel, and the corresponding simulated data in its right panel, with the derived value for the anisotropy coefficient a in the inset. The consistent observed excellent agreement between the simulation data and the experimental data shows that the derivation of a primarily horizontal emissive dipole orientation in the LEC devices during light emission is robust. We note a minor disagreement between measurement and simulation at large angles (>50°), and tentatively assign this to the employment of the isotropic optical properties of Super Yellow in the simulation (since its birefringent doping dependent values are currently unknown) and to a potential small misalignment of the emission spot with respect to the rotational axis x (see Figure 1b). [15b] An interesting and important finding of this study is that the initial strong horizontal orientation of the emissive dipoles of Super Yellow in the pristine active material (see Figure 2a and the accompanying analysis) is retained during LEC operation, despite the fact that the emissive dipoles are exposed to a strong (≈10 8 V m −1 ) electric field in the p-n junction region and that large ions have been migrating in their close proximity during the initial electrochemical doping process. It is possible that this stability is due to that the Super Yellow emitter is a long chain and entangled amorphous polymer with a high glass transition temperature (T g = 150°C), [22] with effectively zero dipole moment in its ground state. Accordingly, the electric field will only significantly affect Super Yellow in its short-lived excited state, and this time might simply be insufficient for the effectuation of any significant reorientation of the stiff Super Yellow polymer chains. We also remind that the pristine strong horizontal orientation of the emissive dipoles was measured on a 40-nm thin film, whereas the similar strong horizontal orientation during LEC operation was measured at an effective film thickness of 150 nm. This implies that the strong horizontal orientation of the Super Yellow chains is relatively thickness independent.
From a more general perspective, it is interesting that Jimenez-Solano and co-workers recently reported that the orientation of the emissive dipoles in their investigated LECs, in contrast to our findings, was gradually shifting to a vertical orientation during LEC operation. [12] However, it is important to point out that they employed a much smaller ionic transition metal complex for their emitting species, and it seems reasonable that such a small molecule should be more prone to reorientation than a much larger and entangled high-T g polymer. The relevance of T g in this context was further verified in a recent OLED study, which showed that the pristine emissive dipole orientation could be retained for several months, but only if the device was stored below the lowest T g of its active layer. [23] We further note that recent high-efficiency LEC devices have employed so-called thermally activated delayed fluorescence (TADF) compounds for the emissive species, [24] which are more polar in the ground state than, e.g., Super Yellow because of their characteristic donor-acceptor molecular structure. It would be very interesting and timely to study the evolution of the orientation of these TADF emitters during LEC operation, and to establish whether further improvements in device performance can be attained by an inducement and/or stabilization of a horizontal emissive dipole orientation.
We finally wish to reiterate that our finding of a robust horizontal orientation of the emissive dipoles for our investigated polymer LEC device during device operation is good news, since this enables for efficient light extraction, as visualized in Figure 1d. For instance, a transition from the herein determined horizontal orientation of the emissive dipoles to an isotropic orientation would result in a drop of the EQE from 3.3% to 2.1% at the optical maximum, i.e., a relative drop of the emission efficiency by 36%, whereas the transition to a vertical dipole orientation would effectively turn the device dark (EQE = 0.016%).

Conclusions
We report on a noninvasive method for the accurate in situ determination of the orientation of the emissive dipoles in a thinfilm LEC device during light emission. We specifically find that a vast majority (95%) of the emissive dipoles in a common LEC, comprising a high-molecular weight conjugated polymer as the emissive species, remains aligned with the horizontal thin-film plane of the device during long-term operation. We further report that this emissive-dipole orientation is manifested in a significant anisotropy in the refractive index and the absorption. Our findings of a strongly horizontal dipole orientation that remains robust during long-term LEC operation were surprising in the context of that the emissive dipoles are positioned in a strong vertical electric field and also exposed to the nearby motion of bulky ions during the initial LEC operation, but also important since it enables for efficient outcoupling and device operation.

Experimental Section
Film and Device Fabrication: The active material consists of a blend of a fluorescent phenyl-substituted poly(paraphenylene vinylene) conjugated copolymer termed "Super Yellow" (Merck, Darmstadt, DE), the salt KCF 3 SO 3 (Sigma-Aldrich, US), and the ion-transporter hydroxyl-capped trimethylolpropane ethoxylate (TMPE-OH, Sigma-Aldrich, US; M w = 450 g mol −1 ). Super Yellow was used as received, while the salt and the ion transporter were dried in a vacuum oven at 190°C and 50°C, respectively, for 12 h before use.
The active-material constituents were separately dissolved in cyclohexanone (Sigma-Aldrich, USA) in the following concentrations: 15 g L −1 (Super Yellow) and 10 g L −1 (KCF 3 SO 3 and TMPE-OH). The active-material ink was prepared by first mixing these master inks in a solute mass ratio of Super Yellow:TMPE-OH:KCF 3 SO 3 = 1:0.10:0.03, and thereafter diluting this blend with cyclohexanone so that the overall Super Yellow concentration was 12 g L −1 . For the neat film samples, the Super Yellow, Super Yellow:TMPE-OH (mass ratio = 1:0.1) and active-material inks were prepared with a similar procedure, but with a lower overall Super Yellow concentration of 8.5 g L −1 . The entire ink fabrication was carried out under inert N 2 atmosphere ([O 2 ] < 1 ppm, [H 2 O] < 1 ppm).
The neat thin films were prepared by spin coating the corresponding ink at 4000 rpm for 60 s on glass substrates (Eagle XG glass, Thin Film Devices, US) for the ARPS and at 2000 rpm for 60 s on SiO 2 coated (thickness = 500 nm) silicon substrates for the ellipsometry. The spin-coated neat films were dried at 70°C for 2 h, and the dry film thickness was ≈45 nm for ARPS and ≈110 nm for ellipsometry.
The LEC devices were fabricated by first cleaning ITO coated glass substrates (ITO thickness = 145 nm, R s = 20 Ω □ −1 , Thin Film Devices, US) by sequential ultrasonication in detergent (Extran MA 01, Merck), deionized water, acetone, and isopropanol. The active-material ink was spincoated on top of the ITO at 2500 rpm for 60 s and subsequently dried at 70°C for 2 h. The measured dry thickness of the active material (d AM ) ranged between 287 and 315 nm, as measured with a stylus profilometer (DektakXT, Bruker). The reflective Al top electrodes (thickness = 100 nm) were deposited by thermal evaporation under high vacuum (p < 8 × 10 −6 mbar) using a shadow mask. The overlap between the bottom ITO electrodes and the top Al electrodes defined four independent 2 × 2 mm 2 LEC devices on each substrate. A glass lid was attached on top of the Al electrodes using a UV-curable epoxy (Ossila, UK) in order to protect the devices from ambient air. [25] LEC Optical Modeling and Fitting: The LEC devices were simulated with a commercial optical software (Setfos, version 5.2, Fluxim AG, Switzerland). The simulated LEC device structure comprised, from top to bottom, an infinitely thick transparent and incoherent glass layer to mimic the combined effects of the glass substrate and the half-cylinder lens, the ITO anode (thickness = 145 nm), the active material (thickness ranging between 287 and 315 nm), and the Al cathode (thickness = 100 nm). The simulated active material was further divided into a n-type doped region and a p-type doped region sandwiching an emissive p-n junction region. The two doping regions were simulated with constant doping concentration gradients, with the highest doping concentration at the corresponding injecting electrode and a zero doping concentration at the interface to the emissive p-n junction region. The two doped regions were discretized into 10 equally thick sublayers, with the optical properties of the correspondingly doped Super Yellow gleaned from the literature. [13,14b] The emissive p-n junction region is required to be transparent in the simulation, and its emissive excitons are modeled as emissive electrical dipoles. The emissive dipole distribution in the p-n junction region was simulated by a modified Gaussian function: [21a] f where x is the position within the p-n junction (normalized by the total p-n junction width), x 0 the position of the peak of the dipole distribution, and a measure of the width of the dipole distribution. The experimentally determined position for the p-n junction was set by x 0 , while the p-n junction width was set equal to 3 . This choice of parameters resulted in that only a very minor fraction (≈1%) of the emissive dipoles was localized outside of the p-n junction region, and these were eliminated from the simulation.
The position x 0 and the width of the dipole distribution in the interelectrode gap of our investigated LEC devices were derived by minimizing the sum of the root mean squared error (RMSE) between the measured and the simulated s-polarized EL data for the available set of angles and wavelengths: where the s-polarized EL intensity I s , is normalized to the maximum of the s-polarized EL intensity, I s The dipole distribution and the doping profile for the p-polarized spectra were interpolated from the s-polarized fit taken immediately before and after the p-polarized measurement. Therefore, the only free parameter left to fit with the p-polarized spectra was the anisotropy coefficient a, which was obtained following the same protocol as described for the s-polarized spectra, x 0 , and . A dual annealing optimization procedure (Scipy Python package) was used to ensure that the derived simulation data resulted from the global minimum. [26] Variable-Angle Spectroscopic Ellipsometry: The ellipsometric parameters of the thin films were measured by the method of VASE (EP4 ellipsometer, Accurion GmbH, Germany), at multiple incidence angles ranging from 60°to 70°. The same ellipsometric parameters were simulated with a commercial software (EP4Model, Accurion GmbH), and the fitting of the simulated data to the experimental data enabled for the derivation of the refractive index and the extinction coefficient of the thin films. The simulation was performed with either a simple isotropic model, comprising a single Tauc-Lorentz oscillator, or with an anisotropic model, comprising different Tauc-Lorentz oscillators for the in-plane and out-of-plane contributions. The RMSE between the measured data and the best simulated data was lowered from 6.6 to 3.4 following the transition from the isotropic to the anisotropic model. The RMSE was further reduced to 2.6 when the surface roughness was included as a parameter in the simulation, but the values for the refractive index and the extinction coefficient were essentially left invariant by this procedure.
Angular-Resolved Photoluminescence Spectroscopy: ARPS was performed with a custom-built spectrogoniometer as schematically displayed in Figure 1b, but with two lenses included in between the sample and the detector to improve the signal-to-noise ratio. The sample was excited by a dot-laser, with a wavelength of 405 nm and a spot diameter of <1 mm. Further details on this setup can be found elsewhere. [15b] In order to reduce the laser-induced material degradation, the emission was recorded at a few angles separated by 10°, and by using a short spectrometer integration time of 0.5 s. Eight independent ARPS measurements were recorded at different spot positions on each film sample. In order to obtain an overall higher angle resolution while maintaining the low number of measurement steps, four measurements covered the range −75°to + 75°and four measurements reached from −80°to +80°yielding an effective overall step width of 5°. For each measurement, three reference spectra at 10°w ere recorded at the start, middle, and end of the angle sweep. These data revealed the absolute intensity drop, which was fitted by an exponential function for the correction of the angular dependence.
Angular-Resolved Electroluminescence Spectroscopy: The angularresolved EL intensity and EL spectrum of the LEC devices were measured with a custom-built, automated spectrogoniometer setup, as schematically displayed in Figure 1b. The spectrogoniometer setup was controlled by a single-board computer (Raspberry Pi 400) and Python-based virtual instrument. The devices were driven by a constant current density of either 25 or 50 mA cm −2 and with the compliance voltage set to 21 V (Keithley 2400 sourcemeter).
The device under study was placed at the center of the rotary stage, which rotation defined the emission angle. An emission angle of 0°corresponds to the forward emission, and data were recorded for angles ranging from −85°to 85°in steps of 5°. The detector was placed at a distance of 100 mm from the device surface, and the fraction of the light that was collected by the collimating lens (∅ = 6.8 mm, NA = 0.38, F110SMA, Thorlabs, Germany) resulted in a collection angle of 0.004 sr. An optical fiber delivered the collected light to a CCD-array spectrometer (Flame-S, Ocean Optics, USA). A half-cylinder glass lens (radius = 20 mm, focal length = 40 mm) was attached to the glass substrate of the device using indexmatching oil. The symmetry axis of the cylinder was aligned with the rotation axis of the stage. A linear polarizer (LPVISE100-A, Thorlabs) was placed in front of the detector and rotated to collect either the s-polarized or the p-polarized light beams. The s-polarized EL data were recorded until the time of minimum forward EL intensity (corresponding to the optical destructive interference condition), after which the p-polarized EL data were recorded. Eight measurements on three independent LEC devices were performed.
Statistical Analysis: The value for the anisotropy coefficient a from the ARPS measurement is the mean value from eight independent measurements, and the error is one standard deviation (see Figure 2a). The value for a from the angle resolved EL measurement is the range derived from eight measurements on three independent LEC devices. No additional statistical method and specific software has been used.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.