Patterned OLEDs: effect of substrate corrugation pitch and height

An ongoing OLED challenge is cost-effective enhancement of light extraction, i.e., increasing the external quantum efficiency (EQE ∼20% in conventional devices). OLEDs on corrugated substrates often show enhanced EQEs providing insight into light emission processes. In particular, patterned plastic substrates directly imprinted easily at room temperature and amenable to low-cost R2R production are ideal for studying/optimizing various structures, further elucidating the extraction process. We show new semi-quantitative data of the effect of the pitch (a) and height/depth (h) of plastic substrate patterns on the OLEDs’ stack and EQE, focusing on new designs, interestingly, some showing surprisingly enhanced EQEs that were neither reported nor discussed before. These include: (i) shallow (h < 200 nm) convex polycarbonate with a ∼ 750 versus∼400 nm, where the h gradually decreases as the OLED stack is built and (ii) concave PET/CAB with large a (∼2.8 and ∼7.8 μm), where the EQE enhancement of conformal OLEDs may be due largely to scattering. EQEs of green, blue, and white phosphorescent OLEDs were measured. OLEDs on substrates with narrow a ∼ 400 nm and low h < 200 nm showed no enhancement, resembling flat devices. In contrast, OLEDs on substrates with comparable or smaller h, but larger a ∼ 750 nm show significant EQE enhancement despite h reduction across the stack. Green OLEDs with a ∼ 750 nm and h ∼ 160 to ∼180 nm, showed EQEs ∼30%, reaching ∼58% with substrate mode extraction. Surprisingly, fully conformal OLEDs on a PET/CAB substrate with a ∼ 7.8 μm showed blue and white EQEs reaching ∼33%, without substrate mode extraction. The enhancing patterns increase the OLEDs’ EQE by reducing surface plasmon excitation and internal waveguiding. The experimental results for OLEDs on substrates with a < 2 μm are supported by scattering matrix simulations that assume conformal stacks, incorporating diffraction for internal losses reduction. EQE enhancement not predicted by simulations may be due additionally to scattering mostly for substrates with a significantly larger than the emitting wavelength.


Introduction
OLEDs have matured in display applications, and are emerging in solid state lighting (SSL) and bioelectronic sensing/medical applications.They provide a diffuse light that is suitable for large area illumination.
Extensive efforts to improve EQE have been reported.To extract the externally waveguided light, microlens arrays (μLAs) of various shapes and sizes were deployed on the back of the substrate to typically achieve 1.7x enhancement in the EQE [15,16], A 2x enhancement in the electroluminescence (EL) was obtained with a high quality μLA of a much larger area than the OLED pixel [15].Alternatively, substrates embedded with high-index TiO 2 nanoparticles [17], or high index substrates were used to replace glass [18].
Recovering the light internally waveguided or lost to surface plasmons in a scalable cost-effective approach remains challenging.Earlier efforts include the use of colorless polyimide, including with air voids as scattering centers [19,20], and subanode designs with a scattering grid layer [21], including with a μLA [22].In a complex design, a silver nanowire mesh in nano-imprinted PET substrate with a poly (3,4-ethylenedioxythiophene): polystyrene sulfonate (PEDOT:PSS) anode and an aperiodic nanostructure in the white OLED (WOLED) stack + μLA resulted in EQE max = 46.3%(2.6x that of flat OLEDs on glass/ITO) at 1,000 Cdm −2 [23].Additional approaches to enhance the EQE are using OLEDs based on emitters with preferential horizontal transition dipole moments [24][25][26][27][28] and manipulation of the refractive index of the organic layers [14,29].
A buckling structure mitigated surface plasmon losses in a green OLED [30].A thicker n-doped ETL (doped to reduce the higher resistance associated with thicker layers) that increases the distance between the emitting layer and the metal cathode also weakens this loss channel [12,13].
More recent studies reported significantly enhanced EQEs.EQE of 78% for a WOLED [31] was achieved using a vacuum nanohole array and an external half-spherical lens, which is relatively costly and unscalable.An EQE max = 70% was reported for a green OLED (∼50% for WOLED) using a subanode microlens array with hemispherical structures etched in glass and planarized by a high index (n ∼ 1.8) spacer on which the OLED was fabricated.An index matching fluid (IMF) interfacing the substrate/photodetector, which extracts virtually all of the substrate mode photons but is unscalable, was employed for demonstrating this high EQE [32].
Patterned OLEDs also provide enhanced EQEs [14,29,[31][32][33][34].While patterned OLEDs are not ideal, their use provided insight into the role of optics and photonics in light emission.Plastic corrugated substrates can be fabricated inexpensively in various designs for further optimization, including when such patterns are planarized by high index layers, as reported recently [33,34].These recent results on planarized OLEDs showed enhancements also with unexpected designs, such as a planarized random, shallow structure, which indicates the need for further related studies of EQE enhancement.A planarized plastic structure with a ∼ 4.25 μm and h = 1.8 μm yielded an EQE that with an IMF exceeded 60%.
In addition to being light and amenable to R2R fabrication, and consequently very cost effective, some plastic substrates have a relatively high refractive index, e.g., polycarbonate (PC), where n PC ≈ 1.6, which can be increased by suitable doping with high-index nanoparticles.As a result, they may advance flexible SSL technology and importantly, wearable sensors and medical devices.This paper's objective is to explore the role of the conformality degree of OLEDs fabricated on a range of corrugated substrates with a = 410, 750, 2800, and 7800 nm and h = 160, 180, 320, and 500 nm.It focuses on light outcoupling from new conformal patterned to partially planarized OLEDs (not reported so far) fabricated on patterned flexible plastic substrates, whose a and h are tracked semi-quantitatively using focused ion beam (FIB).These inexpensive, flexible, periodically patterned enhancing plastic substrates that are amenable to R2R processing, had convex or concave structures.They are shown to significantly increase EQE, surprisingly in both pitch-dependent fully conformal devices with large a, and even in OLEDs where h decreases gradually as OLED layers are added on relatively short pitch and/or height substrates.Such results have not been reported or discussed so far and they may be significant in determining optimal substrate designs for applications such as SSL and wearable sensors.We present examples for such changes in h.For example, a green OLED on a substrate with a ∼750 nm and h ∼160-180 nm showed EQEs exceeding 30%, and ∼58% when extracting the substrate waveguided photons.In contrast, the performance of OLEDs on a substrate with a ∼400 nm and comparable h values resembled that of flat devices.An EQE max ≈33% wasobtained for a fully conformal blue OLED on concave PET/CAB (CAB is cellulose acetate butyrate) with a ∼ 7.8 μm and h ∼320 nm.An EQE max ≈32% (a 2x enhancement relative to a device on flat PET/CAB) was obtained for a fully conformal white OLED on that same substrate (not including extraction of the substrate mode).
As mentioned, patterned substrates typically reduce loss to surface plasmons at the metal cathode and disrupt the internal waveguiding in the organics+anode by diffraction and scattering, which is often ignored [21,[31][32][33][34]. Evidence for scattering is provided also by the near Lambertian angular emission profile [35] and in a recent study of OLEDs on patterned plastic planarized by a high index layer [34].
The OLEDs were broadly characterized via optoelectronic and structural analyses.The results are discussed in terms of device performance, conformality/uniformity and the effect of a and h.
Simulation of light-outcoupling from flat OLEDs using an electromagnetic dipole emission model [12,13], incorporating the Purcell (i.e., optical cavity effect) enhancements of fields at the dipole location, demonstrated that η out was strongly dependent on the ETL thickness, with two outcoupling peaks (η out ∼20%-27%) for ETL thicknesses of ∼λ/4 and 3λ/4.The simulated spectral power of the dipole emission within the OLED provided insights into loss mechanisms, demonstrating high plasmonic losses for thin ETLs, and high internal waveguiding losses for thicker ETLs.However, a periodic corrugation can mitigate internal losses by reducing this optical cavity effect: the cavity length l (along z, i.e., the distance between two points on opposite ends of the device, at given (x,y) coordinates) is constant if the device geometry is perfectly conformal.This is not always the experimental case, since in some of our corrugated OLEDs there is a reduction of the corrugation amplitude with distance from the substrate.Moreover, the positions of the E and H field nodes and antinodes vary with x and y (as well as z), and may not be coincident with the position of the emissive zone that also varies with x, y, and z.
We turn now to the description and discussion of the results.Unless specified otherwise, the results were obtained with no substrate mode extraction structures, i.e., no microlens arrays or IMF.

Results and discussion
OLEDs and organic photodetectors on plastic substrates are already used successfully in sensing and biomedical applications [36][37][38][39], however, OLEDs on plastic, especially patterned, are currently not ideal for SSL, though as shown, their study can elucidate the extraction processes and lead to optimized buried (planarized) designs [34].The patterned devices result in scattering, which is sometimes undesired (e.g., in automotive applications), and the plastic is permeable to water vapor and oxygen, an issue which will be resolved once appropriate thin barrier layers become more available.Plastic substrates can contain defects or contaminations, which result in filamentary shorts that burn out as the voltage applied to the OLED increases above ∼4 V.
The devices used in this study were fabricated on substrates with various designs (PC or PET, convex, concave, with large or narrow a, and of different h values); PEDOT:PSS spin-coated at 6000 rpm, 30 s (proven to be optimal for patterned substrates [40]) served as the anode.Such an anode prepared on flat glass or plastic substrates resulted in a ∼35 nm thick layer [40] and minimal change in the corrugation heigh following its application on a substrate with a ∼750 nm and h ∼320 nm [35].
The substrates and OLED stacks were imaged via atomic force microscopy (AFM) and FIB analysis for determining a and the evolution of h as the OLED stack is built.

Convex PC substrates with h <200 nm
The PC substrates were convex and dome-shaped nano-patterned (figure 1), with h < 200 nm and a ∼400 nm or ∼750 nm.We tested such h values as they were found to be close to optimal in simulations discussed below.Efforts were directed to generate smooth structures devoid of any sharp boundaries that can lead to shorts.The substrates' nano-patterns were expected to increase forward light extraction; the results, however, were mixed.For example, green OLEDs showed maximal EQEs of ∼30% for Substrate A (h ∼160 nm, a ∼750 nm), an enhancement of ∼1.5x relative to a reference device on a flat PC.A similar Substrate B (h ∼ 180 nm, a ∼750 nm), used for checking reproducibility and evaluate uncertainty range, showed an EQE of ∼32%, a ∼1.8x enhancement; the EQE approached ∼60% upon extraction of the substrate mode.Unlike Substrates A and B, Substrate C (h ∼ 180 nm, a ∼410 nm) of comparable h, but shorter a showed no enhancement in EQE.The substrates' structures, EQEs, and enhancements are summarized in table 1.As shown below, the simulations agree with the experimental results for a ∼750 nm to ∼3 μm.We observed reproducible EQE enhancements for OLEDS on the similar substrates A and B despite the reduction in h as the stack is fabricated.This is also in agreement with the simulations that indicate insensitivity to h > ∼100 nm, though the FIB provided only semiquantitative values.We were previously successful in generating efficient patterned OLEDs [33,35] when using a thin ∼35 nm PEDOT:PSS layer as an anode in devices fabricated on PC substrates with a ∼750 nm and a relatively large h of 300 < h 400 nm [33,35].A thicker ∼70 nm spin-coated PEDOT:PSS layer, while serving as an excellent alternative anode to planar ITO on glass [40], was less conformal on the patterned substrates and resulted in inferior outcoupling enhancements.The increased light extraction was due largely to reduced surface plasmon losses and mitigated internal waveguiding due to strong diffraction (see results of simulations below) and possibly also due to random changes in the incident angle at the organic+anode/substrate interface (scattering), not included in the simulations.
Figure 1 shows a 3D AFM image and section analysis of the nano-patterned PC Substrate A. As seen, the substrate features are smooth with h ∼160 nm. Figure 1 also shows the FIB data of the OLED fabricated on this substrate.FIB analysis was used for gauging the conformality of the OLED stack and changes through it.As shown below, the EQE was ∼30%, a 1.5x enhancement relative to a device on planar PC.
As seen in figure 1, FIB clearly shows the general, relative trend in the corrugation height throughout the OLED stack.It provides a semi-quantitative estimate of the actual pattern dimensions due to some damage during the ion milling and as the milling can occur at any plane on the top or slope of the features.Previously we showed that depending on the corrugation features, a thin PEDOT:PSS layer (∼35 nm) does not reduce h significantly (by <10% for a substrate with a ∼750 nm and h ∼320 nm) [33,35].However, as we show next, for devices on substrates with various designs (PC or PET, convex, concave, with large or narrow a, and of various h), the conformality is not always maintained and it is related to the patterns' attributes, a and h, as well as likely the diameter of the features.
As mentioned, FIB analysis of the complete OLED on Substrate A indicated progressive reduction in the corrugation height as OLED layers are added.The value of h in the bare substrate A appears as ∼105 nm, while AFM analysis indicated that h ∼160 nm.This means that the FIB milling occurred on the slope.
The bottom bright thin layer in the FIB image (figure 1) is the PEDOT:PSS anode.As the anode covers the troughs more than the peaks it reduced the corrugation height from ∼105 nm in the bare substrate to ∼85 nm.The following darker layer is the organics, which is followed by the bright Al cathode.The corrugation height at the organics/Al interface is only ∼65 nm.This indicates an additional reduction of ∼23% by the organics and slightly more at the top of the Al, where the estimated height is ∼60 nm.
Figure 2 shows the AFM and FIB data for Substrate C (h ∼180 nm, a ∼ 410 nm).The FIB data is not sharp, but the strong reduction in corrugation height at the top of the Al is clearly seen.On a similar substrate of a ∼450 nm and h ∼150 nm, also presented in figure 2, a significant reduction in h was observed after PEDOT:PSS application and the surface was rough.These data are consistent with the optoelectronic results shown below.b Previous data on green OLEDs on various substrates typically showed higher EQE enhancement than blue and white devices.This may be related to better short-tern stability of green OLEDs.
As seen in figure 2, at low a of ∼410-450 nm and h ∼150-180 nm, the corrugation is significantly reduced resulting in devices whose performance is comparable to that of a planar OLED.Table 1 summarizes the data for the OLEDs on Substrates A to E (see below).For Substrates A and B, the results interestingly show reproducibly enhanced performance even with decreasing corrugation height across the stack.In contrast, no enhancement was observed for Substrate C with a ∼410 nm with the similar h <200 nm.Clearly, the reduced corrugation throughout the stack in the former is still sufficient to reduce the internal waveguiding in the organics + anode, as well as the surface plasmon losses at the organic/Al interface.For a ∼400 nm, however, the reduction in the relatively shallow h substrate is significantly stronger, and hence the device performance is not enhanced, remaining comparable to that of a planar device.We note that tens of pixels are fabricated on each substrate.The EQEs for ∼10 pixels measured in sequence are typically ±5%-6% from the average; there is some device deterioration during the measurement of the non-encapsulated devices.
The thickness of the layers between the electrodes of the blue device as measured by the thickness monitor in the deposition chamber was ∼165 nm; with the PEDOT:PSS layer the thickness is estimated to be ∼195-200 nm, the same value that was obtained in the FIB measurements.For the WOLED, the FIB organic thickness was ∼200 nm and that of Al ∼100 nm.That is, both OLEDs on these substrates were conformal with largely uniform layers.This situation is unlike that observed for OLEDs on substrates with low h and much lower a, where h clearly decreased as the OLED stack developed.
As figure 6 shows, EQE max of the blue OLED on Substrates E was ∼33%, reflecting enhancements of 1.6x.The conformality is due to the very mild slopes of the patterned features.Figure 6 also shows the attributes of a WOLED fabricated on Substrate D that showed a 2x enhancement in EQE max (∼32% on the patterned substrate versus ∼16% on flat PET/CAB) with a fully conformal device structure (figure 5).The luminous efficiency was 95 cd /A versus 47 cd/A for the flat device.The respective power efficiencies were 66 and 32 lm W −1 .As mentioned, an issue with current plastic substrates is the presence at times of defects or contaminations, which result in filamentary shorts that burn out as the voltage applied to the OLED increases above ∼4 V.This results in the slope seen in the efficiencies versus brightness (voltage).
The enhancement factors for the blue and white OLEDs are similar to those of the green devices.Also, for devices that show ∼30% EQE, the addition of IMF typically doubles the EQE.
Table 1 summarizes AFM and FIB dimension measurements and the EQE enhancements of the various devices.With the data obtained so far it is clear that for obtaining a conformal OLEDs, substrates with large a (e.g., 2, 2.8 and 7.8 μm) are needed; obviously, for sufficiently large a there will be no enhancement.Note that simulations did not account for non-conformal devices even for low a & h; optimal a & h were calculated to be around a 1-2 μm and insensitive to heights larger than ∼100 nm (h > 100 nm for a ∼1 μm) for Ag cathodes [41,43].The results summarized in table 1 clearly demonstrate the significant outcoupling enhancements for OLEDs fabricated on different plastic substrate designs and the relation to a and h.As mentioned, conformality and uniformity are achieved for devices on substrates with large a, up to 7.8 μm.Devices on substrates with a ∼2 to ∼7.8 μm showed conformal structures for 250 h 500 nm.Narrower a (∼400 and 750 nm) and h < 200 nm showed varying h across the OLED stack, with interestingly still significant increased outcoupling, except for substrates with both low a and h (∼400 nm; <200 nm), where the OLEDs were practically planarized.
As shown next, simulations indicate that 0.8-2 μm pitch values are close to optimal.The simulations, however, assume a totally conformal OLED stack, which, as shown above, is not always the situation in devices built on shallower corrugations <200 nm, and in particular for smaller a.The simulations address also designs with μm-size pitch, indicating, as expected, reduced EQE enhancements for such designs with large μm-size pitch that experimentally lead to conformal OLED stacks.Experimentally, no enhancement in EQE was observed for devices on a substrate of a ∼15 μm.We note that as we progress toward the flat OLED limit with a >> λ, the first order diffraction has small wave-vectors G that are too small to diffract waveguided/ plasmonic modes to the air cone, and the observed outcoupling enhancement may be partially due to high order diffraction.Since high-order diffraction is not a large effect, it is possible that outcoupling is also due to scattering (i.e., incidence and reflection at random angles) by the internal structure of the patterned OLED.Although the corrugations have large a >> λ , there are also small imperfections occurring at the length scales of λ that can cause scattering.

Simulations: optimal substrate designs and outcoupling predictions for conformal devices
We have developed a rigorous theory of the light emission from corrugated OLEDs by generalizing the theory of emission from flat OLEDs [13,44] to a periodically corrugated OLED, where the detailed numerical implementation was discussed [41,43].We utilized the advanced scattering matrix method [45,46] to calculate the outcoupling from conformally corrugated OLED stacks, where the dipole source (in the emissive layer) is also conformally corrugated and hence can reside at different distances from the substrate.Experimental values of the wavelength dependent complex refractive indices n(λ)+ik(λ) were adopted for the materials in the OLED stack.Maxwell's equations were solved in Fourier space to compute the E and H fields within the different OLED layers and in the free space region, incorporating the Purcell enhancements at the location of the dipole.The E and H fields were expanded in reciprocal lattice vectors G that represent the 2-dimensional periodicity of the lattice.For each of the three polarizations of the dipole, we simulated the power emitted inside the OLED P in (k || ) as a function of the parallel wavevector of the light (k || ), and the power emitted to air P out (k || ), from which the total power emitted inside P tot and outcoupled to air P air are obtained by an integration over k || and a sum over the polarizations.The ratios of these powers is η out = P air /P tot .
We utilized the scattering matrix simulation method for predicting η out from patterned convex conformal OLEDs on a polycarbonate substrate (n = 1.58) with a triangular lattice of nanocones as a function of pitch 250 a 8000 nm and various heights h at different emission wavelengths.The schematic configuration used for the simulations is polycarbonate substrate/polycarbonate nanocones/organic layers (HTL, ETL)/Al cathode, with more details provided in [41].Figure 7 shows the dependence of η out on a, for green OLEDs.
The outcoupling simulations utilized the organic HTL layer thicknesses of the green OLEDs of 140 nm for the HTL, which is close to ∼λ/(2n org ), which at a green wavelength of 530 nm is ∼150 nm (as n org = 1.76).These HTL values are known to yield the highest power output to the thick substrate [12,13].In addition, the ETL, terminated by an Al or Ag cathode, was in the range of λ/(4n) (i.e., ∼75 nm for n org = 1.76 at 530 nm), corresponding to a maximum of the field at the location of the emitting dipole and a node at the Al or Ag cathode.After allowing for a small penetration of the field into the Al cathode (> 5 nm) [47], we took the average value of the ETL thickness to be 70 nm and varied the ETL thickness from 60 to 80 nm for each pitch value.We initially fixed the corrugation height h at 200 nm similar to the experiment, and examined η out as a function of a for OLEDs on a convex corrugated PC substrate, similar to substrates A and B (table 1) at green wavelengths (λ = 530 nm).Since the corrugated substrate of nanocones with the triangular lattice symmetry was anisotropic we simulated the parallel component of the emitting photon wave-vector k || along the x, y, and 45°to the x or y  axes (figure 7(a)).The variances in η out at each pitch a arise from the range of ETL thicknesses in the calculation.The behavior of these outcouplings is similar for larger a but differ at a < 500 nm, where η out for k || along x is considerably lower.The average of these outcouplings is shown in figure 7(b).With these parameters, the computed η out reaches a maximum near 65% for a ∼ 0.8-2 μm (figure 7(a)).For larger a, η out decreased gradually towards the flat value at large a > 5 μm.Qualitatively similar conclusions are found for blue (470 nm) and red (610 nm) wavelengths.
The simulated η out ∼ 64% for a ∼750 nm (see the difference between light extraction with and without IMF in simulations in ref [34]) is higher than that of the measured EQE max ∼ 30%-32% (Substrates A, B, no IMF, see table 1).Simulations with the convex corrugated OLED stack with the Ag cathode yield similar outcoupling results [41] for green OLEDs.At the green wavelength of 520 nm, the electromagnetic skin depth of Al is d (Al)∼7 nm whereas the skin depth of Ag is d(Ag)∼ 13 nm.These differences in the skin depth between Ag and Al and the consequent small differences in reflectivity are not significant, and lead to very similar results for the Al cathode in this work as compared to the previous results [41] for Ag cathodes.
The corrugations cause strong diffraction of the trapped plasmonic and waveguided modes to the air cone, leading to substantial enhancement from the η out ∼ 0.2 for the flat green OLED for optimal ETL values (near λ/ (4n)).As the pitch a increases, the G vectors decrease in magnitude and higher order diffraction (larger G) is needed to diffract waveguided and plasmon modes to the air cone.Higher order diffraction is weaker causing a decrease of the outcoupling.Consequently the waveguiding and plasmonic losses increase considerably for larger pitch a.The discrete G-vector lattice also creates the anisotropy of the simulated emission, especially at small pitch (a < 500 nm), where just the first shell of G-vectors causes diffraction.As we showed, the nonconformality of the OLEDs presented experimentally affects η out depending on the corrugation dimensions.
Since the pitch values near a ∼ 0.8-2 μm provided enhanced η out , we proceeded to simulate the variation of the corrugation height -and found that η out was insensitive to the corrugation height (figure 8), with waveguiding losses also remaining at ∼35% and plasmonic losses substantially lower (∼5%).Although simulations are shown (figure 8) for one of the experimental values of a (750 nm), these trends are similar at other pitch values a.
The experiment and theory qualitatively agree for devices on patterned substrates with up to a ∼3 μm.Scattering, not included in the simulations, is speculated to explain EQE enhancement for a ∼7.8 μm.
As mentioned, scattering is supported by the observed Lambertian angular emission profile.Macroscopically thick (0.7 mm) polycarbonate substrates have interfaces with roughness and non-uniformity over several nanometers.The FIB images of the green OLED on substrate A (figure 1) and the OLED on substrate C (figure 2) display significant nonuniformity of the different interfaces.For larger pitch values (e.g.substrate D a ∼7.8 μm) the scale of non-uniformity can approach the wavelength of light, thereby making the effect of scattering more pronounced at larger pitch values.The effect of scattering is best viewed in a real-space ray-tracing framework.As each light ray traverses through a non-uniform interface, it does not follow the path prescribed by Snell's law but introduces a random change of angle δ away from the prescribed path.Furthermore, as a light ray traverses multiple interfaces the randomization angles accumulate.When the random change in angles are small we may be expected to be in a weak scattering regime and the addition of a scattering component η scatt can be added to the prediction of the diffractive model.For the larger pitch structures with a ∼7.8 μm (substrate D), when the predicted η out ∼ 0.18, is added to an estimated scattering contribution η scat ∼ 0.17 we estimate an outcoupling η out values of ∼0.35.The measured EQE of 0.33 lies slightly below this estimated η out as it is expected.This analysis illustrates that the effect of scattering can be very significant for larger pitch structures.
The simulations do assume a totally conformal structure with the same corrugation heights in each layer.This is not the case experimentally especially for shallower corrugations e.g.OLEDs on the PC substrate (figure 2), where the corrugations are reduced at the Al cathode.Developing a model with a non-conformal stack is very challenging for the present scattering matrix method where Maxwell's equations are solved for an ideal structure in Fourier space.Investigation of non-conformality is better in a real-space OLED model, which is an aspect for future work.

Conclusions
OLEDS and organic photodetectors on plastic were already demonstrated successfully in bio/medical applications.They are also useful in studying optical processes in OLEDs on various substrates.We demonstrated typical ∼1.5x to 2x outcoupling enhancements by using new nano-and micron-size corrugated PC and PET/CAB substrates for partially planarized and conformal green, blue, and white OLEDs.
A PC substrate with a convex structure, corrugation height h ∼180 nm, and pitch a ∼750 nm showed an EQE of ∼32% (x1.8 relative to a flat device) that increased to ∼59% upon extraction of the substrate mode.This was achieved interestingly despite an increasing reduction in h as the OLED stack is fabricated.In contrast, similar OLEDs on a PC with the same h ∼180 nm, but narrower a ∼410 nm showed no EQE enhancement.Based on FIB and AFM analyses, in this latter case, a significantly stronger reduction in h likely occurred already by PEDOT: PSS application.Scattering matrix simulations indicate that diffraction results in calculated η out > 60% for OLEDs on patterned substrates with a ∼800 nm to ∼2 μm, by mitigating internal waveguiding and plasmon losses.
PET/CAB substrates with large μm-size periods (a ∼2.8 μm, h ∼500 nm, and a ∼7.8 μm, h ∼320 nm) showed conformal OLED structures verified via FIB analysis.Enhanced EQEs of ∼32%-33% were obtained for blue and white OLEDs; up to ∼2x enhancement.These enhancements cannot be explained by diffraction with simulations showing a decrease in η out for a >2.5 μm.For patterned devices with a significantly larger than λ the enhancements are likely due to reflection/scattering at interfaces and from imperfections/defects that are on the scale of the optical wavelength.We note that scattering by internal structures is seen also in the near-Lambertian emission profiles we reported previously.
The demonstrated enhanced EQEs of ∼30%-33% were achieved without additional extraction means such as a microlens array on the air-side of the substrate, or IMF at the substrate/photodetector interface.Addition of IMF further enhanced the outcoupling to EQE approaching ∼60%.Simulations, which support the experimental results for OLEDs on relatively small pitch substrates, showed that outcouplings are highest for ∼800 a 2 μm with maximal expected η out and EQEs exceeding 60%.

Materials
Materials used in this study were PEDOT:PSS (from H. C. Starck), LiF, BPhen, and FIrpic (from Sigma-Aldrich), MoO 3 (from Sterm Chemicals), and HAT-CN, TAPC, CBP, 3TPYMB, TmPyPB, mCP, Ir(ppy) 3 , and PO-01 (from Luminescence Technology Corporation.)The substrates, PC and PET/CAB, with various pattern designs, were fabricated by MicroContinuum, Inc PEDOT:PSS was used as an anode; it has become increasingly attractive as an anode with the commercial availability of high conductivity PEDOT:PSS and with conductivity-and OLED-enhancing approaches obtained via mixing or treating with additives such as ethylene glycol (EG) [40,[48][49][50][51][52].ITO was also tested in some structures, however, to preserve the pattern in the plastic it was deposited at room temperature, which sometimes increased its sheet resistance and generated defects.
4.2.Corrugated PC and PET/CAB fabrication: Details of fabrication of various patterned substrates were provided previously [35] In brief, MicroContinuum Inc proprietary processes enabled fabrication of corrugated PC at room temperature.The process was very efficient and inexpensive requiring only a single template for each pitch to produce substrates with various corrugation heights.
The double-layer corrugated PET/CAB enabled separate optimization of each layer for a broader range of substrate attributes.[53][54][55].

PEDOT:PSS film fabrication and characterization
The PEDOT:PSS anode was fabricated by spin coating at 6000 rpm for 30 s (optimized) a mixture of PEDOT: PSS solution with filtered 6 v% EG and 1 v% Capstone FS35 fluorosurfactant.The film was annealed at 80 °C for 2 h; its morphology was imaged by tapping mode AFM (TESPA).
To optimize the PEDOT:PSS anode, the material's solution is often mixed with an additive such as alcohol and/or fluorosurfactants.This improves adhesion of the PEDOT:PSS to the plastic substrate.[35,40,56].While alcohol addition can reduced the anode's conductivity, addition of a fluorosurfactant, such as Zonyl FS30 or Capstone FS35, even at low concentration improves the PEDOT:PSS adhesion without reducing the conductivity [57,58].As reported optimized PEDOT:PSS was mixed with ethylene glycol and a one of the fluorosurfactants [35].

FIB Imaging
Device structure and OLED layer conformality were obtained using a FEI Helios DualBeam FIB/SEM for milling, imaging, and analytical information as described by Hippola et al [35].The FIB system utilizes gallium ions for precise milling and high-resolution 3-D microscopy.

Figure 1 .
Figure 1.Top: AFM image (left) and section analysis (right) of convex PC Substrate A (a ∼ 750 nm, h ∼160 nm): Left -5×5 μm 2 image; Right: full vertical scale -400 nm.Bottom: FIB image of a green OLED fabricated on Substrate A. Note that the analysis was performed at the slope of the feature (see text).
Figures 3 and 4 demonstrate the attributes of green OLEDs fabricated on a flat PC and on Substrate A, B and C. As seen in figure 3, despite the significant reduction in h, The EQE of a green patterned device (Substrate B h∼ 180 nm, a ∼750 nm) increased by ∼1.8x to ∼32% relative to a flat OLED on PC.The EQE increased to ∼60% with extraction of the substrate mode adding IMF at the substrate/photodetector interface.The enhancement was ∼1.25x relative to a reference device + IMF.

Figure 4
compares two devices on substrates of comparable h ∼180 nm but different pitch values of a ∼750 nm (Substrate B) and a ∼410 nm (Substrate C).Both devices were fabricated in the same batch, i.e., in the same evaporation run, which makes the comparison more reliable.Devices fabricated in different batches may show small variations.The results for Substates A & B are still within experimental error.While the device on Substrate C showed the same EQE as flat OLEDs, an enhanced EQE approaching ∼60% was obtained for the device on Substrate B that was of larger a and comparable h.As noted in the figure, the EQE increases x1.8 and x1.3 relative to the EQE of the flat device without and with added IMF, respectively.

Figure 2 .
Figure 2. Top: Left -AFM section analysis of convex PC Substrate C of average h ∼180 nm and a ∼410 nm (full vertical scale 300 nm); Right -FIB image.The corrugation at the Al top is clearly reduced; the EQE of a green OLED was unchanged relative to a flat device.Bottom: Left -AFM section analysis after application of PEDOT:PSS on a PC substrate with a ∼450 nm and h ∼150 nm.Center and Right -AFM before and after PEDOT:PSS application respectively, where h reduced significantly from the original ∼150.

Figure 3 .
Figure3. Green OLED on Substrate A. Top: left -EQE of patterned device A with no substrate mode extraction.Right -J-L-V versus voltage for the green on reference PC and patterned PC with PEDOT:PSS anode.The data for a device on glass/ITO reference is also shown.Bottom: EQE of the patterned OLED with added IMF at the substrate/PD interface.The corrugation increased EQE to ∼29% and the IMF addition to ∼58%.

Figure 4 .
Figure 4. Green OLEDs on Substrates B (a ∼750 nm), and C (a ∼410 nm).Top left: J-L-V versus voltage for a green device on reference PC and Substrate B. Top right: EQEs of the patterned OLEDs without addition of IMF.Bottom: Substrate B; EQE increased by the corrugation to 32% and to ∼58% with IMF .The reference device and that on Substrate C show comparable results.

Figure 5 .
Figure 5. Top two sets: AFM images and section analyses of substrates D (h ∼320 nm, a ∼7.8 μm) and E (h ∼500 nm, a ∼2.8 μm).Bottom sets: FIB results for a WOLED on Substrate D and a blue OLED on Substrate E. Both devices were fully conformal.

Figure 6 .
Figure 6.(A) Normalized EL spectra of WOLEDs on Substrate D and on a flat PET.(B) J-L-V of these OLEDs.(C) EQE of a WOLED fabricated on Substrates D (h ∼320 nm, a ∼7.8 μ) and blue OLED on Substrate E (h ∼500 nm, a ∼2.8 μm) in comparison to a device on flat plastic.

Figure 7 .
Figure 7. Simulated η out from green OLEDs at λ = 530 nm as a function of pitch a for an Al cathode; h = 200 nm.At each pitch a the ETL thickness is varied in a 20 nm range (60-80 nm) around the λ/4 to obtain η out at that a.(a) The simulated η out for corrugated OLED with k || along x, z and at 45°to x or y versus the flat OLED.(b) The averaged η out simulated over the 3 directions, along with the estimated power losses in waveguided modes and plasmonic excitations.The optimal η out for flat OLEDs is shown for comparison.The variances in η out at each pitch a arise from the range of ETL thicknesses in the calculation for (a) and (b).

Figure 8 .
Figure 8. Simulated outcoupling for a corrugated plastic substrate of pitch a = 750 nm as a function of the corrugation height, with a Al cathode The ETL is varied in a small range (60-80 nm) for each corrugation height.

Table 1 .
Summary of AFM and FIB measurements, as well as EQE enhancements.
a ∼60% EQE is typical with added IMF when the EQE w/o it is ∼30%-33%.