Author Correction: Highly efficient, heat dissipating, stretchable organic light-emitting diodes based on a MoO3/Au/MoO3 electrode with encapsulation

Dae Keun Choi1,2,3,a, Dong Hyun Kim1,2,3,a, Chang Min Lee1,2,3,a, Hassan Hafeez1,2,*, Subrata Sarker1,2,3, Jun Su Yang1, Hyung Ju Chae1,2,3, Geon-Woo Jeong1,2,3, Dong Hyun Choi1,2,3, Tae Wook Kim1,2,3, Seunghyup Yoo4, Jinouk Song4, Boo Soo Ma5, Taek-Soo Kim5, Chul Hoon Kim6, Hyun Jae Lee6, Jae Woo Lee7, Donghyun Kim7, Tae-Sung Bae8, Seung Min Yu8, Yong-Cheol Kang9, Juyun Park9, Kyoung-Ho Kim10, Muhammad Sujak10, Myungkwan Song11, Chang-Su Kim11,*, and Seung Yoon Ryu1,2,3,*

. Optical and electrical properties analysis of the respective layers. a-b, The ultraviolet photoelectron spectroscopy (UPS) analysis in both cases between glass and Norland optical adhesive (NOA63) substrate presented a similar trend with a small difference in values. We speculate that the surface roughness of the substrate or exposure to the environment might have affected the UPS measurement because the work function (φ) values were slightly deviated. Interestingly, when an outer molybdenum trioxide (MoO3) layer (15 nm) was added before the deposition of Au (i.e. MA), the φ was increased to ~5.23 eV due to Fermi-Level pinning and the higher φ of the MoO3 layer (~6.6 eV 1, 2 ). The φ of the electrode was further enhanced to ~5.25 eV by the deposition of an inner MoO3 (5 nm) in addition to the MA structure (i.e. MAM) and is more beneficial for device performance, hence, improving the electrical and optical properties. 3,4,5 c, The glass and NOA63 demonstrated an almost equally high transmittance of about 90% due to the intrinsic properties. When the Ag was deposited on both substrates, the transmittance considerably dropped to about 60% at ~400 nm and to ~23%, 6 at ~700 nm, which induces the microcavity effect in the organic light emitting-diodes (OLEDs) devices due to the high reflection as reported by Jalil et al. 7 However, when a MAM electrode was deposited on either NOA63 or the glass substrate, a higher transmittance was obtained that was ~45% and ~70% at about 400 and 650 nm, respectively. Here, the transmittance of the NOA63 with nanoparticles (NPs) was also found to be ~20% less than NOA63 without NPs, which was offset by the ~20-30% haze effect (discussed later in Supplementary Figure 14). d, In the MAM aspect, Hong et. al. 8 reported that by optimizing the thickness of the outer oxide layer, the surface plasmon generated by the sandwiched metal could be coupled with the air using Bragg's scattering, thus increasing the overall transmittance. e-f, The thickness of the sandwiched metal layer is also crucial as it causes a trade-off between transmittance and conductivity as reported by Wrzesniewski et al. 9 It was observed that the 14 nm Au layer presented a very low sheet resistance of ~19.2 Ω/sq, which made it the perfect candidate to be used in the MAM electrode structure. g, MoO3 ( where T is the transmittance and is the sheet resistance. The figure of merit value suggest a fair balance or an optimized performance, between a high transmittance (70~80%) and a low sheet resistance (~ 20 Ω/sq).
where ρ is the density, D is the thickness of the CaO layer which changes with the absorption of moisture, m is the molar concentration, and t is the time passed. The values of ρ and m of water and CaO are known, thus Equation (2) becomes: Here, D can be analyzed by measuring the transmittance of the layer, which increases with the passage of time due to moisture absorption and is given by: The deposited thickness was ~500 × 10 . We analyzed the transmittance values of the devices with different substrates and encapsulations and put them to Equation (4). The values of WVTR obtained for the devices are summarized in the table of Supplementary Figure 3. The device A showed the highest WVTR of 4×10 0 g m -2 day -1 and the WVTR value of device B was improved when 3M tape was used together with NOA63 at the side passivation (2.0 × 10 -1 g m -2 day -1 ). We speculate that the 3M tape alone as a sealant was not the proper material to use for the purpose, as it could have allowed moisture to penetrate into the substrate. Comparing the WVTR of device C (3.6 × 10 -2 g m -2 day -1 ) and device D (1.8×10 -2 g m -2 day -1 ), epoxy was found to be better in blocking the moisture at side than NOA63. Thus, it may be concluded that the side encapsulation is much more important than the penetration from top and bottom films. From the above comparisons, we identified the encapsulation materials in three different paths -bottom (substrate), top (encapsulation), and side (sealant) -in the order of decreasing ease of moisture penetration as red solid lines, and red dot lines. Even though the encapsulations (device B, C and D) suggested improved WVTR than that of device A, those devices were not fully stretchable. There is a trade-off relation between encapsulation quality and stretchability. Therefore, it needs further study to implement edge encapsulation using stretchable sealant for GSOLEDs. c,d, As suggested in the Figure 2, the device lifetime (LT) under various encapsulation schemes and water immersion condition have improved and were comparable, respectively (shown in Supplementary Figures 3c,   d). All of data trend was found to be coherent with the conditions. It may be concluded that the "side passivation" from a combination of 3M tape and NOA63 is quite critical as well as protective against "film penetration" through NOA63 films in a vertical direction. Device A encapsulation withstood for about 7 min during the water immersion, while Device B and C with NOA63 side passivation survived for over 1~3 hours. The shorter LT0 of the device A encapsulation as compared to its LT50 (Supplementary Figures 3c, d) is due to side penetration of water, which was confirmed by WVTR measurement (Supplementary Figure 3b). Even though the side encapsulation with epoxy (Device D) was quite effective, there is a trade-off between stretchability and encapsulation. It needs further study to improve the stretchable encapsulation for GSOLEDs. were deposited on the NOA63 using thermal evaporation followed by NOA63 encapsulation (~9.2 µm). The thin substrate and encapsulation ensured the formation of a sandwich structure that could compensate for the small bending strain and the adjustment of mechanical neutral plane (MNP) at the high Young's modulus (YM) layers of the device. 18 Thus, the crucial layers responsible for the performance of the OLED were preserved from any compression or tensile stresses during stretch-release or bending cycles. The encapsulation also ensured the device was moisture resistant and aided the performance of the GSOLEDs during water-immersion ( Fig. 2c and Supplementary Figure 3). b, The schematic illustration of the side view for the GSOLED device demonstrates the thickness of the layers and the overall layout with and without NPs. The release of pre-strain resulted in the formation of buckles in the devices shown in the image. The NOA63 encapsulation was performed with a small N2/Air gap thus forming a protective cover from moisture contents enabling the device to be water-proof.  magnification along with confocal microscopy (CM) images. a, The OM images of the MAM electrodes at 5× and 10× magnification present the surface morphology after removal of pre-strain. It was observed that the random buckles were formed due to the contraction stresses in the layers. On average the size of the buckles could be estimated to be ~100 µm at maximum and they were closely packed (i.e. high density/number). The buckles formed in the Ag electrode layer were of much larger dimension. This could have occurred because the Ag might have poorly adhered to the NOA63 substrate, causing random parts of the Ag layer to delaminate. This factor affected the performance of the devices because the efficiency had deteriorated. OM images of NOA63 demonstrate that the buckles were of much smaller dimensions, which indicates that the polymer layer had the lowest resistance to the induced stresses. Short wavelength (NOA63 without NPs) and long wavelength (NOA63 with NPs) buckles acted as the main driving force for various complex mechanics of the devices including stretching, folding, twisting and waterproofing. b, The NOA63 buckled substrate was also analyzed using CM images for the dimensions of the peaks and valleys. The analysis was found to be in agreement with the OM analysis. The maximum height of the buckles from the valley was found to be ~220 µm while from the flat area it was ~178 µm, therefore, the average size of the fabricated buckles could be analyzed to be 200 ± 21 µm. With these dimensions, the substrates were conveniently stretched up to 100% area (2D) and were found to be reasonable for stretchable applications. c, It could be observed that the top surface of NOA63 without NPs demonstrate a somewhat smooth surface, however, some bumps were observed in ultra-high resolution field emission scanning electron microscopy (UHR FE-SEM) (Supplementary Figure   8c) and CM images.     Moreover, addition of NPs to the NOA63 film improved the device efficiency due to increased out-coupling.
Interestingly, the glass-based device (700 µm) presented the efficiency roll-off as the luminance was increased, while thin NOA63 based devices displayed slightly increased efficiency without the efficiency roll-off over 100 cd/A and EQE 20% at 20,000 nits. To explain this phenomenon, the temperature of the thin NOA63 and thick glass substrate surface was directly measured by an infrared (IR)-camera as shown in Fig. 5b and Supplementary Figure 13. ( where , , and and i are the initial stiffness matrix, differential load stiffness matrix, eigenvalues and buckling mode shape, respectively. M and N are the degree of freedom and i is i th buckling mode. This means that they are not implemented in FE simulation using a static solver because there is no load or moment component that will cause bending in the lateral direction. Therefore, we used linear buckling analysis which can simulate wrinkled structures in the device. 26 First, the eigenvalue problem was solved in FE simulation with consideration of structure and mechanical properties of the device. The buckled shape of the devices was analyzed with a nontrivial solution obtained from the govern equation. Randomly wrinkled OLED was modelled by controlling displacement boundary conditions in the pre-stretch state. 27 Then, the biaxial stretching of 3M elastomer was simulated by importing the results of the buckling analysis. Simulation model for each strain (0%, 30%, 65%, and 100%) to reflect the real situation when the device is mounted and stretched on the stretching jig. We have evaluated the mechanical simulation for the stress on various strains (0%, 30%, 65%, and 100%) of the GSOLED with 3M elastomer and reflected boundary conditions of gripping edges as in actual experiment for stretching model to overcome the limitation of low convergence and the restriction of stretching model. Figure 11. Bending strain calculation on thin NOA63 (9.2 µm) substrate. a, Shows the schematic of the device with encapsulation. b, The relationship between the bending strain and substrate thickness was plotted using the obtained data which indicates that the bending strain on the buckled devices would increase with an increase in substrate thickness. The bending strain on the devices was calculated using the physical properties of the substrate i.e, YM, thickness of the polymer substrate, and the overall OLED device thickness and YM parameters.
where tOLED and tsub are thickness of OLEDs and substrate and R is the bending radius, respectively. And, and are where Q, k, A, x and T are the transferred energy, heat conductivity, area, thickness and temperature, respectively.   stretching at different percentages (0-100%) was applied to the fabricated GSOLED devices with Ag electrodes and the EL intensity was analysed. At 0% applied strain i.e. the buckled form of the device, the emission wavelength was found to be at ~560 nm which was shifted towards a slightly longer wavelength (yellow shift, ~570 nm) at 30% strain.
When the devices were stretched to 65%, we speculate that the microcavity effect became dominant. This caused a further shift in the emission spectrum towards the longer wavelength (~590 nm) and this change in the emission colour could also be spotted from the camera images (shown as inset). d-f, Further, the CIE 1931 colour analysis was also found to be consistent with the obtained EL profile where with an increase in the applied strain an increase in the emission wavelength was observed. The viewing angle could also generate the microcavity and EL intensities are changeable as shown in Supplementary Figure 19c. g-i, The CIE colour system presented a relatively greener emission at 0% which became yellowish with the increase in strain percentage. With further increase in the applied strain to 100%, the emission wavelength was further yellow-shifted but with a minor difference, which indicates that the maximum effect of the microcavity was attained within the strain range of 65-100%. The EL intensity profile demonstrates the emission spectra for MAM based GSOLED devices. It could be observed that whether a 1D pre-strain was applied or 2D strain, the OLED devices presented a similar emission at ~510 nm.
The EL spectra is in accordance with the camera images (shown as inset) of the pixels, where an identical luminance was maintained. The analysis suggests the absence of the microcavity effect in MAM devices and indicates that the MAM electrode was suitable for the fabrication of color-stable stretchable devices irrespective of the nature of the buckling mechanism i.e, 1D or 2D buckles. Furthermore, the color was stable, even at different strains, and is one of the most desirable properties in a geometrically stretchable OLEDs. This is because the previously presented topemission buckled devices demonstrated an undesirable color-shift in the emissions with respect to the applied strain.
c, The stable emission by the MAM GSOLED device was also analysed at variable angles (0-70°), where a negligible shift in the emission spectra was observed at any given angle. d, e, The CIE 1931 color coordinate analysis at various strain percentages (0-100%) showed a green emission for the MAM GSOLED devices which remained unchanged irrespective of the buckling mechanisms utilized i.e, 1D or 2D microcavity effect and the analysis is found in accordance with the EL spectra. f, Further, the CIE 1931 at variable angles (-70 ̶ 70°) also remained unchanged and was in accordance with the EL spectra. g-i, Respectively, the CIE 1931 for x and y-coordinates were without any changes in any specific axis and stable at any given variable angle. The analysis strongly indicated that the MAMbased OLED could provide an efficient solution for stretchable and wearable devices with high color stability. where the straight position of the finger presented a yellow color that shifted to green upon bending, which was contrary to the green-yellow shift in the EL spectra of the devices due to the viewing angle of the observer. The x-CIE (x) and the y-CIE (y) are defined as, where mx and my are the slopes of x and y obtained by applying the fitting the values, t is the applied strain, and bx is the intercept point of the slope on the y-axis. Similarly, the respective values for the y-CIE are presented in Equation (11).
Here, the values of mx and my were found to be 0.000418 and -0.000327, and bx and by were 0.399 and 0.588, respectively. Hence, at any given strain value, the projection of the changes in x-CIE or y-CIE could be predicted, and vice versa. As from the experimental values, the x-CIE increases and y-CIE decreases with t, and the value of t at any given point in this relation remains constant, e.g, 25% for both x and y-axis, hence we could correlate the equations with tc (constant strain). At any given value of strain (tc), the x-axis is directly proportional and the y-axis is inversely proportional. Hence, the equation for each coordinate can be expressed as Although the derived relationship was found to have some deviation in the results as compared to the experimental results, it could provide a rough estimation of the applied strain with respect to the measured x-CIE and y-CIE coordinates and could be helpful for understanding the dynamics of the Ag strain sensor. Optical simulation modelling is shown with the chromic shifts observed for the Ag electrode. As, the standard emission wavelength of the device is ~520 nm at normal angle (0°) using the 105 nm thick device, the observed emission for the Ag electrode was blue-shifted (towards the shorter wavelength region) with respect to an increase in measurement angle. This blue-shift in the emission follows the Fabry-Perot microcavity equation 33 , given as: where λm is the resonant wavelength, n is the refractive index, d is the physical thickness of the layer, θ is the angle in the space layer (measurement angle), are the changes in the light phase when reflected from the interfaces a and b, respectively, and m is an integer (mode) number in the resonant condition. As discussed in the main manuscript, when θ increases, the value for cosθ decreases, which would consequently decrease the resulting λm. This implies that the emission wavelength would present a blue-shift, which was observed for the Ag electrode OLEDs.  the pitch of 200 nm shows the enhanced extracted emission as well as the out-coupling efficiency. However, the extracted emission spectrum of the bump structure with a pitch of 800 nm in (h) is similar to the one without bump structures in (i). This indicates that a surface with a high surface roughness is effective at extracting the light from the OLED active layer by light scattering. 34 In addition, the total device performance with the MAM electrode was better than that of the Ag electrode due to better charge injection and balance with the MoO3 thin dipole layer. Highly efficient light extraction from the OLED can be achieved with light scattering by the rough surface of NOA63 and better charge injection.