3.1.

Optimization of the Organic Light-Emitting Diodes Planar Stack

As was exemplified in Fig. 3, the outcoupling efficiency strongly depends on the OLED layers’ thicknesses in a planar architecture. The optical optimization of such a stack can be carried out by simulations, yet the proper modeling of the device requires a good knowledge of the optical indices employed (including their anisotropy⁷), as well as of the emitters’ orientation and distribution profile. Alternatively, an empirical optimization can be undertaken by producing several samples with diverse thicknesses. However, this method is demanding in terms of time and materials consumption. Moreover, the batch-to-batch variability may complicate the analysis of the results, which calls for a simplified single-step approach. This can be easily realized using the horizontal dipping (“H-dipping”) technique, regardless of the emitter type considered.³⁵ For the sake of demonstration, H-dipping was exploited to steadily increase the emissive layer thickness from 10 to 120 nm over a multiactive areas device.³⁶ This thickness gradient was directly obtained by continuously increasing the velocity of the coating bar during deposition. Using this one-pass process, a wedge-shaped device was fabricated, and the opto-electrical characteristics of the OLEDs were analyzed as a function of the active layer thickness. This methodology can be generalized to any layer, e.g., to the hole injection one (see Fig. 6). By varying its corresponding thickness, the cavity effect can be optimized in a similar way. However, the device efficiency gain may be limited by an increased amount of parasitic absorption, eventually combined with a degradation of the charge transport. To go further, the H-dipping approach can be modified for the incorporation of advanced materials that feature low absorption and improved electrical properties and that can still be easily wet-processed. Such a material was recently fabricated by making a composite layer consisting of poly(3,4-ethylenedioxythiophene)-poly(styrenesulfonate) (PEDOT:PSS) and silica nanoparticles, as shown in Fig. 6(c).³⁷ Owing to the small nanoparticles’ diameter (only around 7 nm), the modified hole injection layer possessed a haze factor below 1%, and the optical properties of the weak microcavity could be adjusted without introducing any scattering effects or parasitic absorption. The composite layer thickness was then varied between 50 and 400 nm. A luminous efficacy increase of up to 85% was measured for a layer thickness of 100 nm with respect to a 20-nm thick PEDOT:PSS-based reference device. This enhancement was related to a more favorable cavity configuration (improved coupling to the radiation modes), and on the electrical side, to a better hole vertical transport and injection following the inclusion of the nanoparticles network.

Fig. 6

(a) Schematic and (b) photograph of the device including the wedge-shaped composite layer, (c) SEM cross-section view of a $\mathrm{PEDOT}:\mathrm{PSS}/{\mathrm{SiO}}_2$ composite layer and (d), influence of the hole injection layer thickness on the OLED luminous efficacy for both a pure PEDOT:PSS layer (reference) and a composite layer with 72 wt. % ${\mathrm{SiO}}_2$.

The latter aspect was also highlighted in a similar OLED stack but featuring ${\mathrm{SiO}}_2$ (20 to 30 nm in diameter) and ${\mathrm{TiO}}_2$ (20 to 130 nm in diameter) nanoparticles clustered between the emitting layer and the cathode.³⁸ In this configuration, an electroluminescence efficiency increase of up to 300% was measured by using a nanoparticle-free device as a reference. This was not attributed to an optical (scattering) effect. Instead, the enhancement was found to originate from an improved ${\eta }_{\mathrm{int}}$ due to the nanoparticles covering the cathode surface. Indeed, this partial coverage was believed to enhance the fields locally, leading to a better electron injection and, therefore, to an ameliorated charge balance. Thus, nanoparticles-based composite layers can be used as multifunctional elements that simultaneously act as an optical spacer for optimizing cavity effects and improve the interfacial properties of planar OLEDs.

3.2.

Micrometer Scale Corrugations for Waveguide Modes Outcoupling

As discussed in Sec. 1, the anode and organic layers sandwiched between the substrate and the metallic cathode form a slab waveguide. Therefore, a significant part of the dissipated power is injected into the waveguide modes.

In order to limit this effect, a micron-scale optical perturbation of the anode can be exploited. Thus, micropatterned indium-tin oxide (ITO) anodes were realized by combining a photolithography and a wet etching step,³⁹ or alternatively, by sputtering an ITO layer through the gaps created by a self-assembly of microspheres,⁴⁰ prior to the deposition of PEDOT:PSS and of the remaining stack. To increase the refractive index contrast at the anode level, a low refractive index material can be inserted within the micromesh, as proposed in Ref. 41. In the WOLED investigated, a two-dimensional (2-D) square array of micro-columns was introduced within the ITO anode. The material chosen to form those micro-columns was ${\mathrm{MgF}}_2$. In addition to its chemical stability, this material benefits from a low refractive index (around 1.38) compared with the ITO matrix (refractive index ranging from 1.8 to 2.1). As a result, it induces a modification of the modal distribution within the device. The fabrication route of this perturbed anode is shown in Fig. 7. The parameters (periodicity and columns diameter) were first defined in a resist layer by optical lithography, the 2-D array of micro-holes was then transferred into the ITO anode by wet etching, and this patterned matrix was filled by ${\mathrm{MgF}}_2$ via a lift-off step. The duration of the ${\mathrm{MgF}}_2$ vacuum deposition enabled to control the columns' height.

Fig. 7

Fabrication of the ${\mathrm{MgF}}_2$ microcolumns for the extraction of waveguide modes (figure adapted from Ref. 41).

The micro-columns’ height and interdistance were optimized empirically by measuring the efficiency enhancement in comparison to a similar but unperturbed WOLED as a reference. The materials and architecture of the devices were similar to the ones described in Sec. 2. Nine configurations with different spacing values (5, 10, and $15\mu \mathrm{m}$) and column heights (33, 66, and 100 nm) but with a fixed diameter of $2.5\mu \mathrm{m}$ were studied. The results revealed a maximum efficiency enhancement of 38% obtained for a column distance and height of $5\mu \mathrm{m}$ and 66 nm, respectively. Increasing the spacing between the microcolumns reduced the perturbation introduced, therefore reducing its impact on the internal outcoupling. On the other hand, very small spacing values (well below $5\mu \mathrm{m}$) should result in a strong decrease of the active area since ${\mathrm{MgF}}_2$ is an insulator. The existence of an absolute optimum value can then be expected. Similarly, the optimum height should result from a trade-off between a sufficient optical perturbation and good electrical characteristics of the WOLED. It was shown by simulation that in the absence of the microcolumns, a waveguide mode was mainly propagating within the ITO layer. On the contrary, in the perturbed configuration, no waveguide mode existed at the position of the low refractive index columns, and light was preferentially coupled to substrate or radiation modes. Interestingly, adding an MLA to the ${\mathrm{MgF}}_2$-WOLED led to an efficiency increase of 50%, which is a similar enhancement factor as for the unperturbed WOLED with the MLA (see Fig. 5). Thus, it could be concluded that the microcolumns distributed equally the dissipated power of the waveguide modes into the substrate and radiation modes. By using this microphotonics approach, a luminous efficacy enhancement of 38% was obtained without modifying the emission and electrical characteristics of the device.

Unlike the previous example, microspherically textured OLEDs can enhance waveguide modes extraction while increasing the active area of the devices.⁴² In this approach, an ITO-free OLED stack with PEDOT:PSS electrode was deposited on top of silica microspheres with a diameter of $1.55\mu \mathrm{m}$ partially embedded into a 1.55 refractive index photoresist layer. This led to the curved waveguide configuration shown in Fig. 8(a).

Fig. 8

(a) SEM cross-section view of the micro-spherically textured WOLEDs and (b) luminous efficacy of those devices with respect to equivalent but planar devices (adapted from Ref. 42).

In addition to the simplicity of the fabrication process, this method enabled to achieve a luminous efficacy enhancement of up to 3.7 [at 5 V, see Fig. 8(b)] compared with a planar equivalent WOLED, with otherwise unchanged emissive or electrical characteristics. In essence, three factors were evoked for explaining that improvement: $\mathrm{a}\approx 4/3$ increase of the active area by comparison with a planar layout, a better extraction of the waveguide modes allowed by the waviness of the thin film stack, and eventually a more efficient outcoupling of the substrate modes. Indeed, the curved cathode can change the propagation angle of the reflected light, and thereby reduces TIR effects at the glass/air interface.

In Sec. 3.3, we will see that even higher enhancements can be achieved using adapted nanostructures in the vicinity of the active layer, although this may come along with a modification of the emission pattern which needs to be corrected.

3.3.

One-Dimensional High-Refractive Index Nanogratings for Efficient Waveguide Mode Extraction

The advantage of nanostructures, such as one-dimensional (1-D) Bragg nanogratings, with respect to their micrometer-based counterpart is that stronger optical perturbations can be applied to waveguide modes. Periodic wavelength scale patterns were successfully tested in OLEDs in the early 2000s.^43–46 Still today, research efforts are ongoing to maximize the refractive index contrast using innovative solutions, e.g., by inserting high-$n$ nanoparticles in a resist matrix that can be easily nanoimprinted,⁴⁷ and by fabricating periodically arranged air gaps giving rise to a refractive index difference as high as 0.9.⁴⁸ Regardless of the materials involved, two intrinsic limitations of this approach are the angular dependence of the resulting emission spectrum and the wavelength selectivity. Consequently, Bragg nanogratings appear to be more adapted to monochromatic devices than to WOLEDs. In the following examples, we will show that those nanostructures can interact not only with the waveguide modes but also with the substrate modes, and more generally with all the optical channels of the OLED, and demonstrate that they can be efficient and compatible with WOLEDs if combined with a proper external outcoupling method.

The first case discussed in this section is based on an ITO-free OLED incorporating a 1-D Bragg nanograting based on ${\mathrm{Ta}}_2{\mathrm{O}}_5$.⁴⁹ A cross-section view and the fabrication process of this device are shown in Fig. 9.

Fig. 9

Fabrication process of the ${\mathrm{Ta}}_2{\mathrm{O}}_5$ nanopatterned OLED based on a holographic lithography and a dry etching step. A SEM cross-section view of the final device is also reported (adapted from Ref. 49).

${\mathrm{Ta}}_2{\mathrm{O}}_5$ was selected because it is a highly transparent material, and possesses a high-refractive index ($n=2.1$ at 550 nm), which contrasts with that of the PEDOT:PSS anode ($n=1.5$). The period was adjusted for the peak emission wavelength (550 nm) by using Bragg scattering such that the waveguide modes scatter with an angle lower than that of the escape cone.² To test the influence of the grating depth, the latter was varied between 48 and 77 nm. The angle resolved electroluminescence measurements of the yellow OLEDs underlined an emission increasing with the grating depth (see Fig. 10). In particular, it could be shown that for the 77-nm deep grating, the emission was enhanced by up to 300% compared with an unpatterned OLED.⁴⁹ A remarkable feature of Fig. 10 is the presence of emission peaks at $\pm 8\mathrm{deg}$ for the three deepest gratings which result from the efficient outcoupling of the first-order transverse electric (${\mathrm{TE}}_1$) mode. Because the waveguide modes of different orders have less spatial overlap with the grating or a negligible filling factor in the active layer, they did not give rise to intense emission peaks. However, the extraction of the ${\mathrm{TE}}_1$ mode observed at discrete angles was not sufficient to explain the overall emission increase observed. This was attributed to the outcoupling of substrate modes which account for a large part of the confined light in this particular layout. Indeed, the nanograting was also able to redirect the light trapped in the substrate modes.

Fig. 10

Angle resolved emission obtained at 550 nm for the different grating depths investigated and compared to that of an unpatterned reference (figure from Ref. 49).

Thus, it was possible to evidence experimentally the outcoupling of specific waveguide modes and highlight the capability of the grating to tackle simultaneously the internal and external outcoupling. Furthermore, the emitted light intensity was steadily enhanced by increasing the grating depth.

In the second example, we will see that this last property cannot be generalized, and that one needs to consider the complex interplay between the different optical channels in the design of efficient nanogratings. The structure considered for that purpose is based on a 1-D Bragg grating made of ${\mathrm{TiO}}_2$ ($n$ around 2.5 to 3.2 over the visible range) formed on top of the ITO anode of a WOLED using the materials and architecture as described in Sec. 2.² Its fabrication process, using a holographic lithography step, is shown in Fig. 11. As shown in this figure, a MLA was also added on the glass substrate for the two configurations investigated (grating height of 15 or 35 nm).

Fig. 11

(a)–(e) Fabrication of the ${\mathrm{TiO}}_2$ Bragg grating via holographic lithography and vacuum deposition, (f)–(g) deposition of the WOLED stack and (h)–(i) addition of the MLA on the backside of the glass substrate. (g) The cross-section of the device shows a corrugation throughout the whole WOLED (figure adapted from Ref. 2).

The efficiency enhancement factors measured after the introduction of the nano- and microstructures are shown in Fig. 12. The highest efficiency improvement was obtained for the 15-nm grating, due to an efficient outcoupling of the waveguide modes (mainly TE polarized), in combination with the MLA for the substrate modes extraction. Two additional conclusions arise from this figure. First, it can be observed that increasing the grating depth to 35 nm degrades the enhancement factor. Indeed, as was demonstrated experimentally by Hauss et al.,⁴ the introduction of a Bragg grating impacts both the outcoupling from waveguide modes into radiation modes and the incoupling from radiation modes into waveguide modes. In Fig. 12, both configurations have a stronger outcoupling and thus lead to improved efficiencies. However, the 35-nm grating WOLED suffers from a higher incoupling into waveguide modes than for the 15-nm grating case. This effect limits the resulting enhancement factor. In other words, when incoupling mechanisms dominate over outcoupling ones, the introduction of a Bragg grating degrades the efficiency with respect to the unpatterned reference device.⁴ The second comment is related to the different gains obtained after the introduction of the MLA. Those values tend to indicate that the in- and outcoupling of substrate modes into radiation and waveguide modes is also influenced by the geometry of the grating, and that the different optical channels are interdependent. Hence, the grating design is not trivial and should be carried out with optical simulations for optimized utilization conditions. Lastly, it should be pointed out that in all cases, the addition of the MLA enabled to suppress the grating features on the angular emission spectra, giving rise to a uniform white emission.

Fig. 12

Efficiency enhancement factors calculated by using an unpatterned WOLED without the MLA as a reference (adapted from Ref. 2).

To conclude this section, the nanogratings introduced within the OLED stack can be exploited to simultaneously outcouple the waveguide and the substrate modes due to the interplay between the different optical channels. Even though Bragg gratings introduce angular dependencies on the emission spectra, this drawback can be overcome by using light diffusers such as a MLA.

3.4.

Internal Scattering Layers Based on Nanoparticles

With a view to implementing outcoupling structures into an OLED manufacturing process, producing wet-processable and large area layers that do not require any lithography or vacuum-operated step is a major advantage [see the previous example in Fig. 8(a), or the scattering pixels in Fig. 13(a)]. In this context, scattering layers based on nanoparticles are industrially relevant candidates that can improve the outcoupling of substrate modes when deposited onto the substrate,⁵⁰ as well as internal modes (including waveguide and SPP modes provided that their spatial mode overlap with the scattering film is high enough⁵¹) if they are inserted within the OLED thin film stack.^52–55 In general, the nanoparticles’ concentration should be tuned to avoid backward scattering. However, in the case of translucent/bidirectional OLEDs, this effect can be exploited on purpose to simultaneously improve light extraction in both the forward and backward directions, as demonstrated by Chang et al.⁵⁶ In Sec. 3.1, the concept of a hole injection layer (PEDOT:PSS) loaded with silica nanoparticles was introduced to modify the weak microcavity. Indeed, the densely packed nanoparticles embedded in the polymer layer formed an optically homogeneous layer (in the visible regime), whose thickness could be easily adjusted. The morphology of this low-absorbing composite layer can also be modified to introduce volume scattering by increasing the nanoparticles’ diameters (e.g., to 20 or 30 nm, ). By doing so, and after tuning the silica volume fraction, the device shown in Fig. 13(b) could be obtained.⁵⁷ With respect to an equivalent OLED without nanoparticle, its luminous efficiency increased by 66%, resulting from a better ${\eta }_{\mathrm{int}}$ (see end of Sec. 3.1 for explanations), and from an improved outcoupling [Fig. 13(d)]. Due to the low refractive index contrast between the nanoparticles and the polymer matrix, the scattering effect involved most presumably arose from bulk defects, namely air voids distributed throughout the composite layer. Air voids act as scattering centers, and as numerically shown in Ref. 58, their concentration should be optimized to ensure a good trade-off between the scattering properties and the resulting effective refractive index. A similar morphology was reported for silica nanoparticles based layers inserted between the hole injection and emissive layers. In that case, the corresponding devices exhibited a stronger scattering responsible for a noticeable change in the angular emission profile. To reinforce this scattering effect, ${\mathrm{TiO}}_2$ nanoparticles (with a solid volume ratio below 2%, diameters ranging between 20 and 130 nm and a refractive index around 2.5 to 2.7) can be added inside the composite layer without significantly increasing its absorption coefficient [Fig. 13(c)].^59,60 For the best configuration investigated, the sole addition of ${\mathrm{TiO}}_2$ nanoparticles led to a relative increase of 20% in both the electroluminescence and photoluminescence [Fig. 13(e)] measurements. This observation indicates that the luminous efficacy enhancement obtained in such a device mostly originates from optical effects, that is from a more efficient outcoupling of the modes by volume scattering.

Fig. 13

(a) $5\times 5$ array of scattering layers deposited in ambient conditions by screen printing over a $20\times 20{\mathrm{cm}}^2$ glass substrate. Schematics of an OLED based either on (b) a scattering $\mathrm{PEDOT}:\mathrm{PSS}/{\mathrm{SiO}}_2$ layer, or (c) a scattering layer made of $\mathrm{PEDOT}:\mathrm{PSS}/{\mathrm{SiO}}_2/{\mathrm{TiO}}_2$ (SEM cross-sections of the OLEDs shown in insets). The corresponding angularly resolved photoluminescence of those devices is reported in (d) and (e), respectively.

3.5.

Optical Simulations of Internal Scattering Layers Based on Nanoparticles

Numerical simulations of the light propagation in the device are an important tool in quest of an optimal outcoupling structure for a given OLED geometry. For that purpose, the emitting molecule can be modeled by means of a classical point dipole source.³ However, the numerical schemes to simulate OLEDs with internal outcoupling are as manifold as the structures that have been suggested. The finite element method and the finite difference time domain (FDTD) method are frequently used due to their great flexibility.^61–65 However, the numerical effort grows fast with the considered volume and can rapidly exceed the capabilities of state of the art computers. Further modeling approaches include (but are not restricted to): mode-matching techniques,⁶⁶ coupled-mode approaches,^67,68 layered-medium Green’s tensor methods^69,70 or ray optics.⁷¹ Each approach has inherent advantages and disadvantages in terms of required computational time and hardware resources as well as of the accuracy of the resulting efficiency estimations.

As pointed out in Sec. 3.4, scattering layers based on nanoparticles are a particularly relevant class of internal outcoupling structures. Recent efforts in our group have therefore been dedicated to the development of a simulation tool that allows estimating the outcoupling efficiency of devices integrating such layers, out of wave optics calculations.⁷² In contrast to previous numerical studies, coherent multiple scattering and near field effects (such as evanescent coupling to waveguide modes) are taken into account rigorously.

The underlying formalism makes use of the T-matrix formulation^73,74 of the scattering at the individual particles. The particle ensemble is modeled as a coupled system of $N$ individual scatterers. The incident field at each particle contains the scattered fields from all other particles (in addition to the primary dipole excitation). Both the direct interaction as well as the reflections from the layer interfaces need to be accounted for (see Fig. 14). Therefore, the electromagnetic fields are transformed from the spherical wave representation to a superposition of plane waves.^75,76 Their propagation can then be treated analytically by means of the transfer matrix or the scattering matrix algorithm.⁷² With an efficient scheme for the numerical evaluation of the so-called Sommerfeld integrals that occur in the analysis, the interaction of a large number of particles can be accurately simulated.⁷⁷ However, the number of particles still needs to be truncated to limit the necessary simulation time. As a consequence, the outcoupling efficiency is systematically underestimated in the simulations. An elegant way to estimate and correct for this error was proposed in Ref. 78. By repeating the simulation for an increasing width of the scattering structure, the outcoupling efficiency as a function of the lateral dimension of the scattering layer can be fitted by a phenomenological expression that then allows for an extrapolation to infinite scattering samples.

Fig. 14

(a) Schematic view of the coupling between two nanoparticles. The arrows represent the initial dipole excitation (in black), the direct interaction, i.e., the scattered field of one particle that is then incident to the other (in red), as well as the layer system’s reflections of the scattered field that is then incident to the other particle (in blue). (b) Detail of the simulated three-dimensional electric field distribution for a typical OLED structure with a $1.6\text{-}\mu \mathrm{m}$ thick scattering layer including spherical nanoparticles.

As an application example, we show in Fig. 15 simulation results for a realistic layer stack including a ${\mathrm{TiO}}_2$ nanoparticles-based scattering layer. The layer system parameters were chosen to reproduce the OLED stack named “device H” in the study of Chang et al.⁷⁹ In addition, an ideal layer stack of lossless materials was considered to demonstrate the validity of the approach by checking the conservation of energy. All calculations were run at a vacuum wavelength of 520 nm. Figure 15 shows the simulated substrate coupling efficiency, i.e., the power radiated into the substrate layer divided by the total dissipated power, as a function of the number of scattering particles at a fixed particle volume concentration of 3.25 %. Without absorption, all waveguided power is expected to eventually be outcoupled if the lateral extent of the scattering layer tends to infinity. In fact, an asymptotic efficiency around 100% was extrapolated for the ideal device, indicating the validity of the simulations and of the extrapolation scheme. For the realistic device parameters, the outcoupling from the thin film waveguide modes competes with the mode decay due to absorption, resulting in an efficiency well below 100%. In the given example, the substrate coupling efficiency converges to $\approx 75\%$ ($\approx 25\%$) in the case of a parallel (perpendicular) dipole moment orientation relative to the layer interfaces. This is a significant enhancement compared with the theoretical substrate coupling efficiency of 68% (3%) for the case without scattering particles.

Fig. 15

Substrate coupling efficiency plotted as a function of the particles’ number at fixed volume concentration for a typical OLED geometry with scattering particles. The blue and red lines correspond to parallel and perpendicular dipole moment orientation, and the solid and dashed lines correspond to the realistic structure and to a structure without absorption losses, respectively.

We conclude that a prediction of the outcoupling efficiency for OLEDs with internal scattering particle layers based on rigorous wave optics calculations is feasible, paving the way for computer assisted device optimization of OLEDs with an internal scattering particle layer. In fact, the scattering layer and the OLED stack cannot be regarded independently. An OLED layer system that has been optimized with regard to outcoupling efficiency without scattering layer is not necessarily optimal in combination with internal outcoupling structures. Although involving a non-negligible numerical effort, the method presented herein is much more time and cost effective than empirical and experimental surveys. In the future, it will allow for a simultaneous optimization of the scattering layer parameters and of the OLED stack.