3D height-alternant island arrays for stretchable OLEDs with high active area ratio and maximum strain

Abstract

Stretchable optoelectronic devices are typically realized through a 2D integration of rigid components and elastic interconnectors to maintain device performance under stretching deformation. However, such configurations inevitably sacrifice the area ratio of active components to enhance the maximum interconnector strain. We herein propose a 3D buckled height-alternant architecture for stretchable OLEDs that enables the high active-area ratio and the enhanced maximum strain simultaneously. Along with the optimal dual serpentine structure leading to a low critical buckling strain, a pop-up assisting adhesion blocking layer is proposed based on an array of micro concave structures for spatially selective adhesion control, enabling a reliable transition to a 3D buckled state with OLED-compatible processes. Consequently, we demonstrate stretchable OLEDs with both the high initial active-area ratio of 85% and the system strain of up to 40%, which would require a lateral interconnector strain of up to 512% if it were attained with conventional 2D rigid-island approaches. These OLEDs are shown to exhibit reliable performance under 2,000 biaxial cycles of 40% system strain. 7 × 7 passive-matrix OLED displays with the similar level of the initial active-area ratio and maximum system strain are also demonstrated.

Introduction

Stretchable optoelectronic devices have attracted considerable attention as a versatile platform that has potential applications in wearable displays^1,2,3,4, skin electronics^5,6,7, robotics^8,9, and healthcare technologies^10,11,12. In stretchable devices, one of the crucial requirements is that the device performance should be maintained without functional failure under stretching deformations, as well as under bending and flexural deformations^{4,13,14,15,16}. To achieve this important goal, the stretchable systems often rely on a complementary integration incorporating both performance-preserving rigid islands and interconnectors capable of dissipating stress under non-linear deformation^{8,17,18,19,20,21,22,23,24}. However, such approaches in two-dimensional (2D) space inevitably sacrifice a planar density of active components to increase the maximum applicable strain of the interconnector (Supplementary Note 1 and Supplementary Fig. 1 for details). In order to overcome this critical limitation, several groups have recently introduced advanced techniques for realizing three-dimensional (3D) mesostructures across a diverse array of materials, employing a mechanically guided assembly process based on the principle of out-of-plane buckling in elastomeric substrates^{25,26,27,28,29,30,31}. In particular, approaches based on a biaxial release of thin active films integrated on pre-strained elastomeric structures were shown effective in realizing deterministic 3D assembly via consistent morphological modifications^25,32,33. In the conventional mechanically guided assembly method, however, the relatively low bending stiffness of the serpentine structure, which is preferred for enhancing the maximum applicable strain, could present an inherent obstacle to the realization of robust 2D-3D transformation, as it makes it difficult to overcome the interfacial adhesion energy between the elastomeric substrate and the thin film^34,35. It has thus been challenging to realize 3D-structure-based stretchable devices that can attain a high initial density and a high level of the maximum applicable strain at the same time. Furthermore, the 3D methodologies reported to date involved high-temperature operations (> 180 °C), plasma or UV ozone process, or wet processes following the transfer process, which would degrade the performance of devices involving materials with low thermal budget or solvent resistance such as organic semiconductors^{25,26,27,28,29,32,33}.

Here, we propose a strategy for a mechanically guided assembly with a spatially selective adhesion control that can lead to a predictable and reliable 2D-to-3D transition while avoiding wet processes and high-temperature environment in the post-transfer process for compatibility with organic light-emitting diodes (OLEDs). A joint theoretical and experimental study is performed to identify a dual serpentine structure to be optimal for an efficient out-of-plane interconnector buckling that induces little deformation in the rigid islands while ensuring sufficiently low stress to be applied on the interconnector itself. To ensure the 2D thin-film membranes overcome the interfacial adhesion energy necessary for the delamination of rigid islands interconnected with serpentine structure from the elastomeric substrate, we propose a pop-up assisting adhesive blocking layer (PA-ABL) that utilizes the effective interfacial contact-area modulation enabled by the presence of an array of micro-scale concave structures. With this proposed approach, we demonstrate stretchable OLEDs with an 85% active-area ratio in the initial, non-stretch (buckled) state that present only a small discrepancy of OLED performances under 2000 biaxial cycles of 40% system strain.

Results

Overview and mechanical design of the proposed stretchable OLEDs with 3D pop-up architecture

Figure 1a depicts a schematic diagram of the proposed high-density stretchable OLEDs with a high initial active-area ratio (${\gamma }_{a0}$) based on 3D height-alternant island arrays. The system consists of (i) a bottom elastomer layer, serving as the physical support for the overall system; (ii) an adhesive layer that affixes a part of the membrane onto the bottom elastomer; (iii) a thin membrane that hosts an array of OLED pixels, contact pads, and stress-relieving interconnectors; and (iv) pop-up assisting adhesive blocking layer (PA-ABL) that enables selective adhesion reduction. With the help of the adhesive layer coated on the elastomer, the pre-patterned PA-ABL and the OLED devices are sequentially transferred onto the elastomer that is biaxially stretched by equal displacement ($\varDelta x=\varDelta y$). Note that islands with OLEDs that do not face the adhesive directly due to the presence of the PA-ABL are not bonded to the elastomer unlike the other parts of the membrane, forming a selectively bonded structure. While the main function of the PA-ABL is to prevent the adhesive layer from facing the structure on its top directly, it plays an additional yet critical role in assisting the structure on its top to easily pop up. This pop-up assisting nature is attributed to the dimpled surface morphology of the proposed PA-ABL that significantly reduces the attractive interaction between the ABL and the layer on its top, as will be discussed in detail in the upcoming section. In this way, the PA-ABL over the adhesive layer functions to disengage specific regions of the OLED devices, enabling 3D height-alternant island arrays, as shown in Fig. 1b–d when the biaxial stretching is released ($\varDelta x=\varDelta y=0$). The proposed architecture utilizes a bottom-emission OLED device with polyimide as a substrate, whose key multilayer structure is shown in the cross-sectional scanning electron microscope (SEM) image (Fig. 1e).

Fig. 1: Stretchable OLEDs utilizing a pop-up assisting adhesive blocking layer (PA-ABL).

a Conceptual illustration of the assembly strategy with PA-ABL and an OLED device on an elastomeric substrate. The adhesive layer is formed on the elastomeric substrate prior to symmetric biaxial pre-stretching ($\varDelta x=\varDelta y$). b The configuration image of stretchable OLEDs after biaxial strain release. The inset shows the scanning electron microscope (SEM) image of a fabricated stretchable OLED. c The schematic image of a unit OLED device in a pop-up state. d The role of PA-ABL in both 2D planar and 3D pop-up states. The proposed device is triggered by the biaxial strain release, resulting in a transition to a 3D pop-up state. The PA-ABL prevents an OLED from being lifted from facing the adhesive layer directly while significantly reducing attractive interactions between the OLED and the PA-ABL via decreased contact area resulting from its dimpled surface morphology. e The cross-sectional SEM image of the OLED device.

The initiation of out-of-plane buckling occurs when the compressive strain (${\varepsilon }_{{{\rm{comp}}}}$) surpasses the critical buckling strain (${\varepsilon }_{{{\rm{cri}}}}$), inducing a transition from the non-buckled state to the buckled state³⁶ (Fig. 2a and Supplementary Fig. 2) and thereby leading to a 3D pop-up structure with deterministic, freestanding topologies. It is preferable to keep ${\varepsilon }_{{{\rm{cri}}}}$ not too high to avoid the need for excessive mechanical energy to initiate the transition to the buckled state. However, the relatively low bending stiffness inherent to serpentine structure tends to result in a high ${\varepsilon }_{{{\rm{cri}}}}$, underscoring the importance of identifying the ideal mechanical structure and reducing the total adhesion energy (${\varGamma }_{{{\rm{adh}}}}$) between the 2D rigid islands to be lifted and the PA-ABL (Supplementary Fig. 3 and Supplementary Note 2 for detailed calculations) for an effective evolution to an intended 3D buckled state^34,35.

Fig. 2: Analytical model and finite element analysis of out-of-plane buckled serpentines in the proposed stretchable OLEDs.

a The shape and key geometric parameters of the proposed dual serpentine. The radius (R), width (W), and two length scales (L_r and L_s) are shown. $\overline{{AA}}^{\prime} \,$ indicates a hinge section within which the top portion contains the most vulnerable point (P_weak) of the serpentine interconnector. The compressive strain $({{\varepsilon }}_{{{\rm{comp}}}})$ of the serpentine interconnector is defined as ${\varepsilon }_{{\rm{comp}}}{=}({L}_{{{\rm{sp}}}}{-}{{L}}_{{{\rm{s}}}}^{\left({{\rm{proj}}}\right)})$/${{L}}_{{{\rm{sp}}}}$. The operational stretching sequence consists of a sequence from the initial state in the popped-up condition to the state where system strain is applied. b The Von-Mises stress at P_weak ($={{\sigma }}_{{{\rm{von}}}}^{{{\rm{max}}}}$) is estimated as a function of (R, W) at compressive strain (${{\varepsilon }}_{{{\rm{comp}}}}$) of 80%. The three cases I_A, I_B, and II are positioned at (R, W) = (60 μm, 40 μm), (35 μm, 35 μm), and (65 μm, 65 μm), respectively. c The spatial distribution of Von-Mises stress (${{\sigma }}_{{{\rm{von}}}}$) at ${{\varepsilon }}_{{{\rm{comp}}}}$ = 80% obtained by FEA in Cases I_A and II. d The spatial distribution of z-axis displacement (${{u}}_{{{\rm{z}}}}$) between Cases I_A and I_B obtained at ${{\varepsilon }}_{{{\rm{comp}}}}$ = 15%. e The thicknesses of aluminum (Al) and pV3D3 in the multilayer configuration can result in the bending rigidity equivalent to that of a single layer (PI) with half of ${{t}}_{{{\rm{PI}}}}$. The thickness of Al₂O₃ is fixed at 60 nm. f The maximum stress applied to PI and Al vs. system tensile strain (${{\varepsilon }}_{{{\rm{sys}}}}$). Inset: schematic diagram showing three distinct states denoted as states A, B, and C according to ${{\varepsilon }}_{{{\rm{comp}}}}$ and ${{\sigma }}_{{{\rm{max }}}}$.

For the mechanical design of the proposed stretchable OLEDs, the stress analysis was first done based on the buckling mechanics of single-layer interconnectors, to gain an insight without the complexity introduced via multilayer topology. From this analysis, the serpentine structure with a period (N) of 0.5 was determined to be the most advantageous for the efficient formation of a 3D pop-up structure, considering that ${\varepsilon }_{{{\rm{cri}}}}$ gets smaller as N decreases (Supplementary Fig. 3a and Supplementary Note 2)^34,35. Among potential serpentine configurations with N = 0.5, the dual serpentine structure is shown in the inset of Figs. 1b, 2a was chosen in which anchor points are connected in symmetry to the central regions of the rigid islands with curved sides pointing outward. This structure was shown advantageous in keeping the distortion low on rigid islands, compared to other configurations such as single serpentine structures or dual serpentine structures with the anchor points near the edges of the rigid islands and the curved sides pointing inward. (Supplementary Fig. 4). Furthermore, compared to a long hinge structure, this design enhances the maximum applicable strain, allowing the proposed system to extend further beyond the 2D planar state (${\varepsilon }_{{{\rm{comp}}}}=0$) depicted in Fig. 2a.

The radius of the arc (R) and the width (W) of the dual serpentine interconnector (Fig. 2b) were then varied to identify geometries that can ensure the reliable transition between the non-buckled and the buckled states as well as tolerance against high compressive strain to achieve high initial rigid-island area ratio $({{{\rm{\gamma }}}}_{{{\rm{I}}}0})$, which is defined by the area of the rigid island area (${A}_{{{\rm{I}}}}$) to the unit cell area of the array and is related to ${{{\rm{\gamma }}}}_{a0}$ by ${{{\rm{\gamma }}}}_{a0}={({A}_{{{\rm{I}}}}^{({{\rm{active}}})}/{A}_{{{\rm{I}}}}){{\rm{\gamma }}}}_{{{\rm{I}}}0}$ where ${A}_{{{\rm{I}}}}^{({{\rm{active}}})}$ is the active luminous area within a rigid island. Note that achieving high ${{{\rm{\gamma }}}}_{{{\rm{I}}}0}$ is thus a prerequisite to achieving high ${{{\rm{\gamma }}}}_{a0}$ in the schemes relying on rigid island platforms. We searched for conditions where the Von-Mises stress (${\sigma }_{{{\rm{von}}}}$) stays below the fatigue strength (${\sigma }_{{{\rm{f}}}}$) of polyimide (PI) at the most vulnerable point (${{{\rm{P}}}}_{{{\rm{weak}}}}$) of the serpentine interconnector (i.e., ${\sigma }_{{{\rm{von}}}}^{\max }$ < ${\sigma }_{{{\rm{f}}}}^{{{\rm{PI}}}}$), which is in the top portion within the hinge section shown as a cross-sectional area along A-A’ in Fig. 2a. At this point, both bending-induced and torsional stress were shown to be concentrated. Calculation of ${\sigma }_{{{\rm{von}}}}^{\max }$ was performed analytically using an algorithm based on elasticity theory³⁶ and Castigliano’s theorem with ${\varepsilon }_{{{\rm{comp}}}}$ set at 80% (see Supplementary Note 3 and Supplementary Figs. 5, 6 for details on calculation). The result indicates that ${\sigma }_{{{\rm{von}}}}^{\max }$ primarily reflects normal stress induced by the bending moment of the entire serpentine structure, which turns out to be dominant over the shear stress resulting from the torsional moment of the arms composing the serpentine (Supplementary Fig. 7). It is noted in Fig. 2c that ${\sigma }_{{{\rm{von}}}}^{\max }$ tends to decrease as R and W decreases in the range of R and W studied in this work. To meet the criteria of ${\sigma }_{{{\rm{von}}}}^{\max }$ < ${\sigma }_{{{\rm{f}}}}^{{{\rm{PI}}}}\approx$ 120 MPa, in particular, (R, W) should be in the region ‘I’ in Fig. 2c, the boundary of which is defined by the line connecting (R, W) $\approx$ (30$\,{{\rm{\mu}}}{{\rm{m}}}$, 74$\,{{\rm{\mu }}}{{\rm{m}}}$) and (R, W) $\approx$ (74$\,{{\rm{\mu }}}{{\rm{m}}}$, 30$\,{{\rm{\mu }}}{{\rm{m}}}$) (Supplementary Fig. 8 for the measured ${\sigma }_{{{\rm{f}}}}^{{{\rm{PI}}}}$ of 120 MPa).

To illustrate its importance, the mechanical properties of the serpentine structures were further evaluated with finite element analysis (FEA) for 3D configuration in three different cases represented as i) star (= Case I_A), (ii) triangle (= Case II), and (iii) square (= Case I_B) in Fig. 2c. Categorization of Cases I_A and I_B vs. Case II is based on whether ${\sigma }_{{{\rm{von}}}}^{\max }$ stays lower or higher than ${\sigma }_{{{\rm{f}}}}^{{{\rm{PI}}}}$ to emphasize that ${\sigma }_{{{\rm{von}}}}^{\max }$ < ${\sigma }_{{{\rm{f}}}}^{{{\rm{PI}}}}$ condition should be fulfilled. Within Region ‘I’, Cases I_A and I_B are further distinguished to illustrate the importance of securing a sufficiently low critical buckling strain by the difference in the critical buckling strain, which increases as the width decreases, as detailed in Supplementary Note 2. Case I_A, which is the most suitable among the three, was chosen, in addition to the aforementioned criteria, upon consideration of a critical dimension for PI that can be reliably achieved via the patterning method used in this work. First of all, it is confirmed in Fig. 2d, comparing the spatial distribution of ${\sigma }_{{{\rm{von}}}}$ between Case I_A and Case II, that ${\sigma }_{{{\rm{von}}}}^{\max }$ values obtained by the analytical model match relatively well with those obtained by FEA. It is also noted that ${\sigma }_{{{\rm{von}}}}^{\max }$ of Case I_A (analytical model: 91 MPa, FEA result: 98 MPa) is indeed below ${\sigma }_{{{\rm{f}}}}^{{{\rm{PI}}}}$ (Fig. 2d, top); in contrast, ${\sigma }_{{{\rm{von}}}}^{\max }$ (analytical model: 134 MPa, FEA result: 127 MPa) of Case II exceeds the fatigue strength of PI (Fig. 2d, bottom), making Case II subject to the risk of fatigue-induced failure.

Finite element analysis (FEA) further shows that it is not just ${\sigma }_{{{\rm{von}}}}^{\max }$ that matters in realizing a 3D pop-up structure. Figure 2e compares the spatial distribution of z-axis displacement (${u}_{{{\rm{z}}}}$) between Cases I_A and I_B obtained at ${\varepsilon }_{{{\rm{comp}}}}$ = 15% (see Supplementary Fig. 9 for z-axis displacement analysis), both of the cases fulfill the ${\sigma }_{{{\rm{von}}}}^{\max }$ < ${\sigma }_{{{\rm{f}}}}^{{{\rm{PI}}}}$ criteria, but Case I_B fails to make a transition into the buckled pop-up state due to an increase in ${\varepsilon }_{{{\rm{cri}}}}$ attributed to its low W, while Case I_A can transition into the buckling state (${u}_{{{\rm{z}}}}$ = 164.4 μm), allowing for OLED membranes to pop up even at relatively low compressive strain (see Supplementary Note 2 for a detailed calculation of ${\varepsilon }_{{{\rm{cri}}}}$). For ${\varepsilon }_{{{\rm{comp}}}}$ > 15%, the overall small dimension of Case I_B does not allow for further compression without causing physical contacts between the edges of the serpentine structure. This case study well illustrates the importance of choosing optimal geometric parameters, such as that of Case I_A, so that a high ${\varepsilon }_{{{\rm{cri}}}}$ can be avoided to effectively realize 3D pop-up structure.

While the analysis introduced so far is based on a single layer of PI, the actual interconnecting region consists of a multilayer assembly of PI/encapsulation/Al/encapsulation. In spite of this discrepancy, the analysis and its results obtained from a single layer could be kept valid by identifying a multilayer configuration of which the effective bending rigidity (=$\overline{{EI}}$) is equivalent to that of the single PI layer (=${E}_{{{\rm{PI}}}}{I}_{{{\rm{PI}}}}$). (Fig. 2f and Supplementary Note 4 for details). The thickness of Al₂O₃ was fixed at 60 nm to ensure a sufficient level of gas barrier properties. The thickness of poly(1,3,5-trimethyl-1,3,5-trivinyl cyclotrisiloxane) (pV3D3)^37,38 (${t}_{{{\rm{pV}}}3{{\rm{D}}}3}$) set at 162 nm and that of Al (${t}_{{{\rm{Al}}}}$) set at 150 nm were found to be among the combinations that meet ${E}_{{{\rm{PI}}}}{I}_{{{\rm{PI}}}}=\,\overline{{EI}}$. This process was shown to preserve ${\sigma }_{{{\rm{von}}}}^{\max }$$({\varepsilon }_{{{\rm{comp}}}})$ well as shown in Supplementary Fig. 10, confirming the validity of our design approach.

As a last step, to determine the overall operation range and understand the effect of the system operation on local mechanical movement, the stress analysis was performed as a function of system tensile strain (${\varepsilon }_{{{\rm{sys}}}}$), defined as

$${\varepsilon }_{{{\rm{sys}}}}=\frac{{{L}_{{{\rm{s}}}}^{\left({{\rm{proj}}}\right)}-L}_{{{\rm{s}}}0}^{({{\rm{proj}}})}}{{L}_{r}+{L}_{{{\rm{s}}}0}^{({{\rm{proj}}})}}$$

(1)

where ${L}_{{{\rm{s}}}}$ and ${L}_{{{\rm{r}}}}$ are the lengths of the serpentine electrode and the rigid island, respectively, as denoted in Fig. 2a. The superscript ‘(proj)’ and ‘0’ in the subscript indicate the effective length of the parameters projected onto the plane of rigid islands and the value obtained in the initial non-stretched condition (${\varepsilon }_{{{\rm{comp}}}}$ = 80%), respectively. As shown in Fig. 2g, there are three distinct states (see the FEA-predicted images in the Inset: State A, State B, and State C). State A corresponds to a 3D pop-up configuration with the highest spatial density of rigid islands, State B corresponds to a planar configuration with the smallest stress, and State C corresponds to a semi-2D configuration in which serpentine electrodes are further deformed to accommodate higher system tensile strain. Note that, with the bottom elastomer stretched, the proposed dual serpentine interconnector remains in a non-buckled state to form a ‘2D planar’ state (State B) or further stretched state (State C). When a compressive strain (${\varepsilon }_{{{\rm{comp}}}}$) of 80% is applied by releasing the strain applied to the elastomer, it transitions to a buckled state (State A). (Fig. 2a and the inset of Fig. 2g.) Transitioning among three distinct states, the designed device has an operational range of ${\varepsilon }_{{{\rm{sys}}}}$ from 0% to 40%. In this operational range, ${\sigma }_{{{\rm{von}}}}^{\max }$ remains below ${\sigma }_{{{\rm{f}}}}^{{{\rm{PI}}}}$ (≈ 120 MPa) and that of aluminum (${\sigma }_{{{\rm{f}}}}^{{{\rm{Al}}}}\approx$ 1.75 GPa), which were verified through repeated tensile stress tests (Supplementary Fig. 8).

Reduction of the interfacial adhesion energy for reliable 3D pop-up operation

Figure 3a outlines the proposed OLED-compatible integration method essential for constructing a 3D pop-up structure (see Supplementary Figs. 11, 12, and Methods for full details). The elastomer (Dragon Skin 10, Smooth-on Inc.) coated with the adhesive layer in its central region is pre-stretched using a customized biaxial stage. Then, an aligning block is placed to serve as a mechanical guide that ensures the OLED-containing membrane and the pop-up assisting adhesion blocking layer (PA-ABL) are aligned onto the desired region. The PA-ABL, composed of PDMS, is transferred before the OLED layer so that only the selective portion of the adhesive layer may be exposed to the OLED membrane. Finally, the OLED layer is transferred, and the biaxial pre-strain is released to induce out-of-plane buckling to complete a 3D pop-up structure in which every other rigid island is lifted up while the others remain in the original plane.

Fig. 3: Overview of the assembly method and the importance of a pop-up assisting adhesion-blocking layer (PA-ABL).

a Illustration of the fabrication process comprised of a sequence involving the transfer of the PDMS layer, followed by OLED transfer and the release of pre-strain. Shown on the left is the biaxial stage with the aligning block used for precise control during the assembly process. b Maximum stress in PI (red) and Al (blue) versus ${{{\varepsilon }}}_{{{\rm{comp}}}}$ obtained for the non-buckled state (circles) and the buckled state (squares). c The measured effective work of adhesion (G_c) vs. the open-area ratio (${{{\eta }}}_{{{\rm{open}}}}$) of the PA-ABL, presented as the mean and standard deviation of three measurements. d The measured number of popped-up islands (N_up) vs. ${{{\varepsilon }}}_{{{\rm{comp}}}}$ applied to serpentine connectors. e The false-color SEM images of the fabrication results after the sequential transfer of the OLED device layer and the PA-ABL obtained at ${{{\eta }}}_{{{\rm{open}}}}$ = 0, 41%, and 88%. The images were taken under ${{{\varepsilon }}}_{{{\rm{comp}}}}$ = 80% enabled by equal biaxial strain release. The left schematic images show the selective bonding area (shown with the white outlines in the bottom image) and the blocking (non-bonding) area between the OLED device layer and the PA-ABL. In the blocking area, the PA-ABL covers up the adhesive layer such that the island with the selected OLED is not glued and can eventually be lifted when a sufficient ${{{\varepsilon }}}_{{{\rm{comp}}}}$ is applied.

${\varepsilon }_{{{\rm{cri}}}}$ increasing with ${\varGamma }_{{{\rm{adh}}}}$ as discussed in Supplementary Note 2, the PA-ABL can play a pivotal role in enabling a 3D pop-up structure at low ${\varepsilon }_{{{\rm{cri}}}}$ by reducing the adhesion energy (${\varGamma }_{{{\rm{adh}}}}$)³⁴. The lack of PA-ABL would tend to yield relatively high ${\varepsilon }_{{{\rm{cri}}}},$ leading to the non-buckled state over a wide range of ${\varepsilon }_{{{\rm{comp}}}}$, which can be detrimental to the system reliability because ${\sigma }_{\max }$ of PI and Al in the non-buckled case could be hiked even over a small increase in ${\varepsilon }_{{{\rm{comp}}}}$ (Fig. 3b).

Our study further reveals that a simple planar PA-ABL would be far from ideal in reducing ${\varGamma }_{{{\rm{adh}}}}$ sufficiently. A double cantilever beam (DCB) test confirms the reduced ${\varGamma }_{{{\rm{adh}}}}$ with the proposed micro-concave structures³⁹ (Supplementary Fig. 13 for DCB load-displacement results). As shown in Fig. 3c, the critical strain energy release rate (G_c), measured as the effective work of adhesion, is as large as 259 mJ/m² for the flat PA-ABL. On the other hand, it gets reduced by seven times to 33 mJ/m² once the PA-ABL has the open (i.e., non-contact)-area ratio (= ${\eta }_{{{\rm{open}}}}$) of 88%. Such a high degree of ${\eta }_{{{\rm{open}}}}$ is realized in this work by an array of micro-concave structures patterned on the top surface of a PDMS layer (Supplementary Fig. 12 for the detailed fabrication procedure). The number of rigid islands transitioning from a planar state to a pop-up state under ${\varepsilon }_{{{\rm{comp}}}}$ induced by biaxial pre-strain release supports the important role of reduced G_c (Fig. 3d). For a flat PA-ABL surface, the out-of-plane buckling in the interconnector was not triggered even with ${\varepsilon }_{{{\rm{comp}}}}$ = 80%. For the surface with ${\eta }_{{{\rm{open}}}}$ = 41%, the number of rigid islands transitioning to the pop-up state (= N_up) gradually increased from ${\varepsilon }_{{{\rm{comp}}}}$ ≈ 10%. For ${\eta }_{{{\rm{open}}}}\,$= 88%, on the other hand, N_up was observed to increase even at ${\varepsilon }_{{{\rm{comp}}}}$ as low as 5% and, moreover, its increase over ${\varepsilon }_{{{\rm{comp}}}}$ was sharp enough to ensure a clear, effective transition even at low ${\varepsilon }_{{{\rm{comp}}}}$. The false-color SEM images of the stretchable OLED devices in State A with ${\varepsilon }_{{{\rm{comp}}}}$ = 80% clearly illustrate the significance of the proposed PA-ABL with an array of micro-concave structures for proper, fail-free operation of stretchable OLEDs based on 3D pop-up architectures (Fig. 3e). The PA-ABL with a flat top surface, on the other hand, suppresses out-of-plane buckling of the serpentine structure and results in an intertwined appearance.

Mechanical and optoelectrical behavior under biaxial stretching

The mechanical performance was first evaluated using the proposed 3D pop-up structure composed of PI, electrodes (Al), and encapsulation layer (Fig. 4a), which corresponds to the multilayer structure of the interconnector regions in the stretchable OLEDs under study. With repetitive symmetric biaxial tensile stretching with the system strain (${\varepsilon }_{{{\rm{sys}}}}$ = ${\varepsilon }_{{{\rm{x}}}}\,$= ${\varepsilon }_{{{\rm{y}}}}$) of 40%, the structure having the PA-ABL with ${\eta }_{{{\rm{open}}}}\,$= 88% exhibited electrical resistance (R) that maintained its initial value (R₀) well with only a small increase below 5% even after 2000 cycles. On the other hand, the structure having the PA-ABL with ${\eta }_{{{\rm{open}}}}\,$= 41% showed R that linearly increased until ca. 500 cycles, after which its increase got steep and eventually became too large due to a mechanical fracture. These results will illustrate the importance of securing sufficiently high ${\eta }_{{{\rm{open}}}}$, being consistent with the discussion made for Fig. 3b–d. The current (I)-voltage (V) characteristics were measured for the structure with ${\eta }_{{{\rm{open}}}}$ = 88% at ${\varepsilon }_{{{\rm{sys}}}}$ of 0%, 10%, 20%, 30%, and 40%, which correspond to ${\varepsilon }_{{{\rm{comp}}}}$ of 80%, 50%, 20%, − 9%, and − 39%, respectively. (Fig. 4b) The invariance is observed in the measured I–V characteristics over a wide range of ${\varepsilon }_{{{\rm{sys}}}}$, further demonstrating the mechanical reliability enabled by the proposed pop-up structure with the PA-ABL having large ${\eta }_{{{\rm{open}}}}$.

Fig. 4: The mechanical stability and optoelectrical characteristic of stretchable OLEDs under stretching operation.

a Resistance changes of a serpentine structure composed of PI layer, encapsulation layer, and Al layer, measured at the initial state during cyclic tests of equal biaxial stretching with the PA-ABL having open-area ratios (${{{\eta }}}_{{{\rm{open}}}}$) of 41% and 88%, respectively. b Current-voltage characteristics of a 3D pop-up structure fabricated using the PA-ABL with ${{{\eta }}}_{{{\rm{open}}}}$ = 88%. Measurement was done for equal biaxial strains of 0, 10%, 20%, 30%, and 40%. c The photographs of the proposed stretchable OLED devices with PA-ABL having ${{{\eta }}}_{{{\rm{open}}}}$ = 88% under the applied equal biaxial strain of 0, 25%, and 40%. d Current density (J)-luminance (L)-voltage (V) characteristics and (e) Current efficiency vs L of the stretchable OLEDs under various equal biaxial strains. f Normalized J, L, and current efficiency at a driving voltage of 6 V under cyclic stretch-release test for a biaxial strain of 40% (a.u., arbitrary unit). g (Left) Schematic diagram that illustrates convex deformation. (Right) Maximum strain of the elastomer surface versus radius of curvature ratio (r/R) obtained from FEA results. (Bottom) The photographs of working stretchable OLEDs under convex deformation with various r/R ratios. h System maximum strain (${{{\varepsilon }}}_{{{\rm{sys}}}}^{{{\rm{max }}}}$) and initial rigid-island area ratio (${{{\gamma }}}_{{{\rm{I}}}{{\rm{0}}}}$) of previously reported 2D rigid island platforms^{22,23,51,52,53,54,55,56,57,58,59} and the proposed stretchable OLEDs. The color map indicates the interconnector strain (${{{\varepsilon }}}_{{{\rm{i}}}}$) that is required to achieve (${{{\gamma }}}_{{{\rm{I}}}{{\rm{0}}}}$, ${{{\varepsilon }}}_{{{\rm{sys}}}}^{{{\rm{max }}}}$) in 2D rigid island platforms. The ‘star’ mark corresponds to (${{{\gamma }}}_{{{\rm{I}}}{{\bf{0}}}}$ = 85% ${{{\varepsilon }}}_{{{\rm{sys}}}}^{{{\rm{max }}}}$ = 40%), achieved with the proposed 3D platform. If this level of performance had been achieved with the conventional 2D rigid island schemes, the required interconnector strain would have been as high as 512%.

The photographs in Fig. 4c present the proposed stretchable OLED composed of a 5 × 5 array using the PA-ABL with ${\eta }_{{{\rm{open}}}}$ = 88%. It can be seen that it operates quite in a uniform manner without dark spots or noticeable visual imperfection under a constant current of 0.2 mA at ${\varepsilon }_{{{\rm{sys}}}}$ = 0%, 25%, and 40% (see Supplementary Movie 1 for pop-up operation under biaxial stretching from ${\varepsilon }_{{{\rm{sys}}}}$ = 0% to 40%). Most of all, they show the 3D configuration anticipated from the FEA results in the three distinct states, highlighted by ${{{\rm{\gamma }}}}_{{{\rm{I}}}0}(={{{\rm{\gamma }}}}_{a0}\,,\because \,{A}_{{{\rm{I}}}}^{\left({{\rm{active}}}\right)}=\,{A}_{{{\rm{I}}}})$ of 85% at ${\varepsilon }_{{{\rm{sys}}}}$ = 0% and a maximum system strain of 40% (Supplementary Fig. 14 for the measured ${{{\rm{\gamma }}}}_{{{\rm{I}}}}$‘s). Although these photographs, taken with a camera equipped with a macro lens, clearly reveal the height differences within the 3D height-alternant structure, these variations are not readily perceptible to the human eye. This is due to the limited human binocular vision, which allows for the typical minimal perceivable depth of about 1-2 mm (refer to Supplementary Note 5 for further discussion.). Most importantly, as the resolution increases and thus both the islands and the serpentines become scaled down, it becomes even less likely that these height differences will be noticed. Figure 4d, e presents the electroluminescent (EL) performance of the stretchable OLEDs under ${\varepsilon }_{{{\rm{sys}}}}$ ranging from 0% to 40% in 10%. The low leakage current observed in the measured current density (J)―voltage (V) characteristic can be attributed to the smooth surface with root-mean-square (RMS) roughness of about 1 nm, as indicated in the Atomic force microscopy (AFM) data in Supplementary Fig. 15. The EL performance of the OLEDs showed only a slight variation under different system strain, with a current efficiency of 55 cd/A at 100 nits in the initial state and a slightly reduced current efficiency of 54 cd/A under a biaxial strain of 40%, partly due to the rigid island-based approach where the strain on the active region is kept minimal and, most of all, due to the advantage of the proposed 3D pop-up structure. The cyclic biaxial stretch-release, test performed at a rate of 10 mm/s with a biaxial strain of 40%, indicates that the EL performance, measured at the initial state with a constant voltage of 6 V (Fig. 4f and Supplementary Fig. 16 for measurement setup), also maintains its characteristics well even after 2000 cycles; the current density and luminance get reduced only by 1.9% and 2.7%, respectively, from their initial values, and the current efficiency decreases merely by 0.8% (see Supplementary Fig. 17 for SEM images of the device after biaxial stretching cycles and Supplementary Movie 2 for cyclic operation). Supplementary Fig. 18 presents the simulated spatial distribution of shear stress at the key interfaces for the adhesives and the measured lap shear strength of the adhesive. The simulation results indicate that the adhesive used in this work can withstand operational stresses without failure, even at a relatively high compressive strain of up to 80%.

As the demands for integrating displays over various objects increase, so does the importance of displays that can conform over 3D objects beyond mere flatness^40,41,42. That is, when operating stretchable OLEDs, deformation over a curved surface can be as important as stretching in a two-dimensional plane. The graph in Fig. 4g shows the FEA results for the maximum strain induced by convex deformation at various radius of curvature ratios (r/R) depicted in the schematic diagram. The maximum of analyzed strain (${\varepsilon }_{{{\rm{x}}}}^{\max }$, ${\varepsilon }_{{{\rm{y}}}}^{\max }$) on the surface of the elastomer reaches from ca. 4% at r/R = 0.2 (mild convex deformation) to ca. 25% at r/R = 1 (steep convex deformation) (see Supplementary Fig. 19 for detailed FEA results). The fabricated device showed stable operation and little variations in EL performance under convex deformation across r/R values ranging from 0.2 to 1 (Refer to the photographs in Fig. 4g and Supplementary Fig. 20), affirming the reliability and structural integrity of the proposed stretchable OLEDs even under convex deformation. Mechanical analysis and the results of experiments that compare the performance evolution over time among glass-reference OLEDs and stretchable OLEDs — with and without the 2000 stretch-and-release cycles — indicate that (i) the 1.5 dyad encapsulation layers provide a sufficient level of protection against ambient air for the purpose of this study; and that (ii) the encapsulation properties are maintained well even after the repeated stretching cycles (Supplementary Figs. 21, 22). While the application of an encapsulation layer with 2.5 dyads will further enhance the barrier properties for long-term protection of organic devices, it should be noted that the strain across the encapsulation stacks can increase, particularly at the edges. Nevertheless, our simulation study suggests that high strain is localized to a very small region along the edge, as shown in Supplementary Fig. 22b, and that the majority of the island area experiences far less strain than the crack onset strain of Al₂O₃. As long as the active region remains within the ‘safe zone’ (e.g., within the central 0.95$\times$0.95 mm²), adequate protection against ambient air can still be considered feasible, from the practical perspectives, with a minimal decrease ${{{\rm{\gamma }}}}_{a0}.$ (Refer to Supplementary Fig. 22e for further details.).

A stretchable passive matrix OLED with a high active area ratio

To illustrate the benefits of the proposed pop-up approach for stretchable displays, we have configured our proposed stretchable OLEDs in the form of a 7 × 7 passive matrix (PM) array, which simultaneously exhibits high, ${\gamma }_{{{\rm{I}}}0},\,{\gamma }_{a0}$, and ${\varepsilon }_{{{\rm{sys}}}}^{\max }$. The OLED device with a 3D height-alternant island array was fabricated using PA-ABL with η_open = 88%. Tested by fabricating several samples and examining the distribution of critical buckling strain, the pop-up mechanism enabled with PA-ABL was shown to be reproducible in a consistent manner (Supplementary Fig. 23). Figure 5a shows the overall system setup that consists of (i) a computer-controlled, Arduino (Arduino®, Arduino mega 2560) based driver; (ii) a customized flexible printed circuit board (FPCB) connecting a stretchable PMOLED and the driver with Ag paste; and (iii) a biaxial translational stage holding the stretchable PMOLED at a given system strain. The PMOLED structure includes vertical scan lines linked to the cathode and horizontal data lines connected to the anode. The 7 × 7 pixel array includes both bonded and pop-up pixels, as shown in detail in Supplementary Fig 24.

Fig. 5: Demonstration of Stretchable Passive-Matrix OLED with 3D Pop-Up Structure.

a The overall system setup for the implementation of the stretchable passive-matrix OLED display. On the right side, a simplified schematic diagram is presented that shows the bonded and pop-up pixels composing the 7 × 7 arrays, whose detailed layer configurations are provided in Supplementary Fig. 24. b Schematic diagram of the system setup for demonstrating the passive matrix. c Photographs displaying the characters in “3D OLED” one by one when biaxial system strains of 0%, 25%, and 45% are applied.

We demonstrated dynamic operation of the proposed PMOLEDs by sequentially activating the characters ‘3’, ‘D’, ‘O’, ‘L’, ‘E’, and ‘D’ (Fig. 5b and Supplementary Movie 3) at biaxial system strains of 0%, 25%, and 45%. Each character was activated through specific sets of voltage streams applied to the data and scan lines (Supplementary Fig. 25). It is noteworthy that, even for PM operation, this process not only achieved ${{{\rm{\gamma }}}}_{{{\rm{I}}}0}$ and ${{{\rm{\gamma }}}}_{a0}$ as high as 85% and 82%, respectively, but also demonstrated operation at a maximum strain of 45%, highlighting the significant benefits of the present technology for stretchable displays. In addition, OLEDs on the proposed platform were shown to cause little heat accumulation and exhibit spectral consistency during the continuous operation over time (Supplementary Fig. 26). Unlike previous works adopting 3D architectures for stretchable devices^27,43,44, our methodology neither uses post-transfer wet processes and thermally-induced bonding processes^27,43 nor has a much smaller active luminous area with respect to the rigid island area even for matrix-type display applications^27,44. (Supplementary Table 1 for detailed comparison).

To extend this platform into more sophisticated display applications, it is crucial to assess whether our processes are compatible with high pixel-per-inch (ppi) resolutions and active-matrix (AM) displays. Currently, state-of-the-art stretchable display prototypes, demonstrated by major display manufacturers, feature AM displays with the minimum dimension of polyimide islands corresponding to 100 ppi (254 µm) and 200 ppi (127 µm)^45,46. Based on this polyimide-based island technology, the methodology proposed here should also be compatible with a similar level of pixel density and AM configuration. Furthermore, it is important to note that a single rigid island may host multiple pixels. Considering the number of electrode lines required for either PM or AM operation per pixel, configuring four pixels per island is reasonable when arranging the individual arms of the dual serpentine interconnectors in both horizontal and vertical directions to have one (PM) or multiple (e.g., two each in AM) electrode line(s) per arm.

Regarding AM operation, it is crucial to recognize that the adequacy of available electrode lines is not the only concern. The technology should also be compatible with the preparation of the in-pixel components such as thin-film transistors (TFTs) and capacitors. Polyimides, in fact, have been a standard substrate material used in industry for displays adopting flexible AMOLEDs, not just for prototype stretchable displays. This is due to the excellent chemical resistance and relatively high serviceable temperature of polyimides that enable the preparation of low-temperature poly-Si (LTPS) or oxide semiconductors for TFTs and the photolithographic patterning of in-pixel components (TFTs, capacitors, electric lines, etc.). As our devices are initially built on polyimide substrates and the transfer process occurs after the fabrication of all essential components, including encapsulation, any standard process used in the fabrication of flexible or stretchable PM or AMOLED in industry can be applied with little modification, enabling more straightforward industrial adoption.

Discussion

In summary, we proposed a strategy involving 3D pop-up structures to realize stretchable OLEDs that can simultaneously achieve a high initial density of rigid islands and a high maximum interconnector strain. Utilizing a mechanically guided assembly method and OLED-compatible fabrication strategy, the proposed OLEDs were designed to get stretched up to a system biaxial strain (${\varepsilon }_{{{\rm{sys}}}}$) of 40% while making effective transitions among the initial non-stretched state (${\varepsilon }_{{{\rm{sys}}}}$ = 0%) with the highest spatial density of rigid islands, a state corresponding to a planar configuration with the smallest stress (${\varepsilon }_{{{\rm{sys}}}}$ $\approx$ 25%), and a state having a semi-2D configuration in which the serpentine electrodes are further deformed to accommodate higher system strain (${\varepsilon }_{{{\rm{sys}}}}$ $\approx$ 40%). Two essential factors for the successful realization of the proposed 3D pop-up structures were as follows: (i) the mechanical design of out-of-plane buckled interconnectors based on an optimal dual serpentine structure that ensures rigid islands to be lifted effectively at a relatively low compressive strain while maintaining the maximum stress applied onto the most vulnerable portion (${\sigma }_{\max }$) to remain below the fatigue strength values for the substrate (PI) and the aluminum electrodes across the serpentine interconnectors; (ii) substantial reduction of interfacial adhesion energy (${\varGamma }_{{{\rm{adh}}}}$) between the 2D membranes to be popped up and the regions underneath them, which was enabled by the proposed pop-up assisting adhesion blocking layer (PA-ABL) that selectively covers up adhesive layers. The top surface of the proposed PA-ABL was designed to have a dimpled surface with an array of concave micropatterns, realizing the open-area ratio (${\eta }_{{{\rm{open}}}}$) of up to 88% so that the adhesion energy between the PA-ABL and the rigid island on its top can be reduced to a significant degree. With the proposed PA-ABL, a reduction in ${\varGamma }_{{{\rm{adh}}}}$ by more than five times was demonstrated with respect to that of a planar adhesion blocking layer. Comparison made between the present and previous works in terms of the maximum system strain and the initial rigid-island area ratio (${\gamma }_{{{\rm{I}}}0}$) suggests that it is difficult, in conventional 2D rigid island platforms, to achieve a maximum interconnector strain (${\varepsilon }_{{{\rm{i}}}}$) far exceeding the empirical threshold of ${\varepsilon }_{{{\rm{i}}}}=140\%$ line in Fig. 4h while maintaining a high ${\gamma }_{{{\rm{I}}}0}\,$. Note that ${\varepsilon }_{{{\rm{i}}}}$ in 2D, rigid island platforms are related to ${\varepsilon }_{{{\rm{sys}}}}$ by ${\varepsilon }_{{{\rm{sys}}}}=\left(1-\sqrt{{\gamma }_{{{\rm{I}}}0}}\right){\varepsilon }_{{{\rm{i}}}}$. The level of ${\gamma }_{{{\rm{I}}}0}$ (= 85%) and the maximum system strain (= 40%) achieved with the proposed stretchable OLEDs can be considered significant in that it could be achieved only with interconnectors that can sustain strain of 512% if it were done with a 2D rigid island platform (Fig. 4h). To further enhance the maximum strain of the dual serpentine structure proposed in this study, consideration can be given to reducing the thickness of polyimides. This is partly because both tensile properties and pop-up states in this structure are predominantly governed by buckling mechanisms that can be facilitated by a thinner substrate¹⁸. However, any reduction in thickness should be approached with caution, considering the risks associated with the transfer and laser lift-off processes.

While this work presents a solution to overcome the fundamental limitations of 2D rigid island platforms in terms of ${\gamma }_{{{\rm{I}}}0}$ and ${\gamma }_{a0}$, it does not fully address the challenges posed by the stretching-induced increase in the lateral gap between islands and the resulting image distortion in stretchable displays (e.g., resolution loss, aspect ratio change, etc.). Nevertheless, the high initial pixel density that becomes available with the proposed technology can be advantageous in mitigating stretching-induced image distortion, for instance, by introducing surplus pixels that are inactive in the initial state but selectively activated upon stretching to compensate for distortion.

The proposed stretchable OLEDs with the 3D height-alternant island array structure, composed of a 5 × 5 array using the PA-ABL with ${\eta }_{{{\rm{open}}}}$ = 88%, were shown to operate in a uniform manner at various system strains, as well as under tight convex deformation, which illustrates the versatile nature of the proposed stretchable OLEDs and affirms their reliability and structural integrity. A 7 × 7 PMOLED operation was also demonstrated with$\,{\gamma }_{{{\rm{I}}}0}$ 85%, ${\gamma }_{a0}$ 82%, and ${\varepsilon }_{{{\rm{sys}}}}^{\max }$ of up to 45%. With the high initial active area ratio and a semi-2D operation where square lattice geometry is maintained while being stretched, the proposed stretchable displays will be preferred in applications wherein displays are used mostly in the initial state but have to deal with occasional stretching, as in the wearable applications. With (i) the comprehensive yet step-by-step design strategies that are compatible even with materials having a low thermal budget and low chemical resistance and (ii) the use of materials like polyimide and its laser lift-off / patterning processes that are already employed in the AMOLED display industry, we believe our methodology can open up an effective and practical pathway towards the realization of stretchable OLEDs or many other optoelectronic systems that can fulfill the demanding requirements of various applications.

Methods

Fabrication of the proposed stretchable OLEDs

The fabrication process of stretchable OLEDs can be majorly categorized into three stages: (i) fabrication of the OLED device layer, (ii) fabrication of the pop-up assisting adhesive blocking layer (PA-ABL), and (iii) transfer and 3D morphing processes.

Fabrication of OLED device layers

On a 4-inch glass wafer, polyamic acid (KPI-1200, Komec) was formed via spin coating and subsequently converted into a 2.5 μm-thick polyimide (PI) film through a thermal imidization process at 280 °C for an hour. A 100 nm-thick aluminum was deposited onto the PI thin-film via a sputtering process, and patterned by a conventional photolithography process. With the aluminum patterns serving as an embedded mask, the PI film was patterned through a reactive ion etching (RIE) process under an oxygen plasma condition (RF power: 300 W, O₂ gas inflow: 25 sccm), after which the aluminum patterns were removed using its etchant (Supplementary Fig. 11). A 160 nm-thick pV3D3 film was deposited through a free-radical polymerization reaction, utilizing an initiator of tert-butyl peroxide (TBPO) and a monomer of 1,3,5-trimethyl-1,3,5-trivinylcyclotrisiloxane (V3D3) in a custom-built iCVD chamber. On top of the pV3D3 layer, a 60 nm thick-Al₂O₃ layer was deposited through an atomic layer deposition (ALD, Lucida D100 NCD Co.) at a low temperature of 70 °C, using a trimethylaluminum (TMA) precursor and H₂O reactant. Another 160 nm pV3D3 layer was deposited on the Al₂O₃ layer, using the same iCVD process as before. The pV3D3/Al₂O₃/pV3D3 structure was intended for a bottom encapsulation, and the layers for the OLED device were sequentially deposited using a shadow mask technique (refer to Supplementary Fig. 27 for detailed design of the shadow masks). Initially, a 100 nm-thick indium zinc oxide (IZO) layer as an anode, was deposited selectively onto the rigid island using a RF sputtering process in a vacuum chamber at a pressure of 5.8 × 10⁻³ torr. A 150 nm-thick aluminum layer was then, deposited utilizing a vacuum thermal evaporator (HS-1100, Digital Optics & Vacuum) selectively onto the interconnector area to provide electrical connections between the IZO patterns. On top of the IZO layer, organic semiconductor layers were deposited using the vacuum thermal evaporator, in the sequence of a 140 nm-thick 4,4′-cyclohexylidenebis [N,N-bis(4- methylphenyl)benzenamine] (TAPC) layer as a hole injection layer, a 10 nm-thick tris(4-carbazoyl-9-ylphenyl) amine (TCTA) layer as a hole transport layer, a 30 nm-thick TCTA: Ir(dmppy-ph)₂tmd: 4, 6-bis(3,5-di(pyridin-3-yl)phenyl)−2-methylpyrimidine (B3PYMPM) (48: 48: 4 wt. %) layer as an emission layer, and a 70 nm-thick B3PYMPM layer as an electron transport layer. Lastly, LiF (1 nm)/Al (150 nm) layers were deposited as a top cathode. In the 5 × 5 rigid island configuration, the horizontal lines of the top cathode and vertical lines of the bottom anode overlap in the rigid island area (Supplementary Fig. 27). A top encapsulation structure, which was the same structure as the bottom encapsulation, was formed on top of the OLED device. In preparation for the transfer process, the OLED device layer was delaminated from the glass wafer using a laser-lift off (LLO) technique via an excimer laser equipment (KORONA^TM, AP systems) (Supplementary Fig. 11). Before transferring onto an elastomeric substrate, the cathode electrode pad and anode electrode pad of the OLED device were connected to a customized FPCB using Ag paste. Supplementary Fig. 28 shows the difference in EL performance between the OLED devices before and after the delamination process. The difference in performance mainly originated from a thermal degradation induced by the LLO process.

Fabrication of adhesive blocking layer

For the fabrication of the PA-ABLs with open-area ratio (${\eta }_{{{\rm{open}}}}$) values of 0, 41%, and 88%, three templates with different surface shapes were prepared corresponding to each ${\eta }_{{{\rm{open}}}}$ values: a flat glass substrate covered by a 1 µm-thick fluorinated polymer layer for ${\eta }_{{{\rm{open}}}}$ of 0, a sapphire substrate with a 1.4 µm minor axis, 1.8 µm major axis microlens array (MLA) surface (HNPS01, LUMTEC, Taiwan) covered by the 1 µm-thick fluorinated polymer layer for ${\eta }_{{{\rm{open}}}}$ of 41%, and the sapphire substrate with the MLA surface covered by a 100 nm-thick Al₂O₃ layer and fluorinated self-assembled monolayer (SAM) for ${\eta }_{{{\rm{open}}}}$ of 88% (Supplementary Fig. 12). All template substrates were sized 25 mm by 25 mm square. The fluorinated polymer layer (NovecTM 1700 Electronic Grade Coating, 3M^TM) was prepared by spin-coating followed by a curing process at room temperature for 10 min. The Al₂O₃ layer was deposited by the ALD process. The fluorinated SAM was introduced using vaporized Trichloro (1H, 1H, 2H, 2H perfluorooctyl) silane (Sigma-Aldrich) in a vacuum desiccator. On top of the templates, A PDMS layer was spin-coated and cured at 100 °C for 30 min to be the PA-ABLs. The selection of PDMS was driven by its low modulus of elasticity and its compatibility with the reactive ion etching (RIE) patterning process using a metal mask. The thickness of the PA-ABL (= ca. 2.6 µm) was chosen by considering that PA-ABL should be thin enough so as not to make the PA-ABL/adhesive/elastomer assembly too stiff over tensile strain and that it should be thicker than the height of the concave lens array to be formed (Supplementary Fig. 29). The fluorinated polymer and the fluorinated SAM were introduced to reduce the surface energy of the templates, thus the PDMS based PA-ABL can be easily delaminated and transferred to the bottom elastomeric substrate coated with an adhesive, as described in the “Transfer and 3D morphing processes” section. For the PDMS layers to perform as PA-ABLs, which enables selective bonding between the OLED device layer and the bottom elastomeric substrate with an adhesive, they were patterned using the RIE process (RF power: 300 W, O₂ gas inflow: 25 sccm, CF₄ gas inflow: 10 sccm), through a metal shadow mask (Supplementary Fig. 12).

Transfer process and 3D morphing process

As a bottom elastomeric substrate, a 1 mm-thick elastomer (Dragon Skin 10, Smooth-on Inc.) film was formed through a spin-coating and curing process at room temperature for 6 h. An adhesive material (DOWSIL^TM SE 9186) was spin-coated on the elastomeric substrate. A film tape (PI Film Tape 5413, 3 M) was used to cover the edges of the elastomeric substrate before the adhesive coating and removed together with the adhesive on it. This is to define the adhesive area on which the PA-ABL and the OLED device layer would be transferred. The elastomeric substrate was then, stretched by 30 mm in biaxial directions using a customized biaxial stage equipped with two motorized linear translators (T-LSR075D, Zaber Technologies Inc.), and an aligning block was positioned on the top. The 25 mm by 25 mm square substrate with the PA-ABL was positioned within the aligner, the PA-ABL was transferred onto the elastomeric substrate with the adhesive surface, and thus the adhesive turned out to be exposed only in the area without the PA-ABL. The OLED device layer was then, transferred onto the bottom elastomeric substrate aligned with the PA-ABL using the aligning block. During the transfer process, the OLED device layer was selectively bonded where PA-ABL was not applied thus, the adhesive was exposed. Considering the bottom emitting direction of the OLED device, the OLED device layer was attached in the upside-down direction. After the curing process of the adhesive at room temperature, the biaxial pre-strain applied to the elastomeric substrate was released, and consequently, the stretchable OLED device, initially in a 2D state, morphed into a 3D pop-up structure. Furthermore, to perform the electroluminescence characterization of the stretchable OLEDs, silver paste (P-100, CANS) was used to connect the OLED contact pads to the customized flexible printed circuit board (FPCB) for seamless external connection. While it was applied manually in this work, automatic dispensing equipment may be used to reduce the potential inconsistency of manual application and ensure uniform connectivity, crucial for reliable operation. Alternatively, anisotropic conductive film (ACF) may also be used⁴⁷.

Finite element analysis (FEA) for out-of-plane buckled serpentine and convex deformation

The mechanical behaviors considering a serpentine interconnector under out-of-plane buckling and convex deformation of the stretchable OLED were analyzed by FEA using computational module software (COMSOL^TM Multiphysics). The module analysis, based on geometric nonlinearity, was carried out under conditions stipulating a minimum element size of 5 μm and 15 μm for the serpentine interconnector and convex deformation, respectively, utilizing a stationary solver. The elastic moduli used in the analyses were 4 GPa for PI (E_PI), 70 GPa for aluminum (E_Al), 134.4 GPa for aluminum oxide (E_Al2O3), and 0.86 GPa for pV3D3 (E_pV3D3). For predicting the mechanical behavior of the elastomer, the analysis was conducted based on the Ogden model, which typifies a near incompressibility of hyperelastic materials.

Adhesion measurement and preparation of specimen

The double cantilever beam (DCB) fracture mechanics test was conducted to quantify the effective work of adhesion (G_c) between the pV3D3 layer, that is, the top surface of the OLED device layer, and the PA-ABL make of PDMS. Two distinct types of specimens were prepared, each having either the surface of the pV3D3 layer or the PA-ABL. For the specimen with the pV3D3 surface, a 160 nm-thick pV3D3 layer was deposited using the iCVD process on a 25 μm-thick PI film, which served as a carrier substrate (see Supplementary Fig. 30 for illustration of specimen preparation). The other specimen with the PA-ABL was prepared to have the ${\eta }_{{{\rm{open}}}}$ values of 0, 41% and 88% on a flat glass or on sapphire glasses with an MLA surface, as described in the “Fabrication of PA-ABL” part in Methods. An adhesive glue (DOWSIL^TM SE 9186 clear) was spin-coated on the back surface of the three PA-ABLs, and then 25 μm-thick PI films were attached to the adhesive glue as a carrier substrate. After the curing process of the adhesive glue at room temperature for 10 h, the PA-ABLs were delaminated from the flat glass and the sapphire glasses, with the support of the PI films. G_c between the two specimens of the pV3D3 surface and the PA-ABL surfaces was measured using high-precision DCB test equipment (DTS Company, Delaminator Adhesion Test System). The prepared specimens were bonded with a 275 μm-thick Si wafer (size: 10 mm × 40 mm) as a back-up substrate to apply sufficient bending strain energy for the separation of the target pV3D3/PA-ABL interface. After adhering to the pV3D3 and PA-ABL interfaces, the specimen was mounted onto the loading grips, and a load (F) ― displacement (d) curve was obtained through the operation of a load cell and a linear actuator (Supplementary Fig. 13). G_c, was then extracted from the F―d curve using beam bending mechanics^39,48,49_. Note that G_c is not determined solely by the load value but is significantly influenced also by the crack length. G_c can be calculated using the following equation.

$${G}_{{{\rm{c}}}}=\frac{12{P}_{{{\rm{c}}}}^{2}{a}^{2}}{E^{\prime} {b}^{2}{h}^{3}}{\left(1+0.64\frac{h}{a}\right)}^{2}$$

(2)

where a is the crack length, C is the elastic compliance, E’ is the plane-strain modulus of the beam, B is the specimen width, h is the half height for the substrate, and P_c is the critical load where the load versus displacement curve begins to decrease. Since the measurement area of the DCB test was sufficiently large relative to the feature size of the micro-scale concave structure, the G_c values derived from the DCB test can be considered as representative values and used for the analysis of the OLED device layer and the PA-ABL interfaces in the proposed stretchable OLEDs.

Measurement of pop-up rigid island counts

To observe the 3D morphing process of the stretchable OLED, the initial pre-strain applied to the elastomeric substrate was gradually released with the interval corresponding to the compressive strain of serpentine (${\varepsilon }_{{{\rm{comp}}}}$) of 5%. Orthogonal view images were captured using camera equipment (ILCE-7M3, Sony) equipped with a macro lens (FE 90 mm F2.8 Macro G OSS, Sony). The number of pop-up rigid islands in the fabricated stretchable OLEDs was counted from the orthogonal view images, as a function of ${\varepsilon }_{{{\rm{comp}}}}$.

The electroluminescent measurements of the stretchable OLEDs

The optoelectrical performance of the stretchable OLEDs, including current density (J) and luminance (L), was measured using CS 2000 equipment (KONICA MINOLTA, Inc.) and a source meter (Keithley 2400, Keithley Inc.) under ambient conditions. The current efficiency (η_CE) was calculated based on the J and L values. A system strain (${\varepsilon }_{{{\rm{sys}}}}$ = ${\varepsilon }_{{{\rm{x}}}}\,$= ${\varepsilon }_{{{\rm{y}}}}$) of up to 40% was applied to the stretchable OLEDs during the electroluminescent measurement, using the customized biaxial stage.

FEM-based mechanical simulation for array-based 3D pop-up simulation

A simulation study of the array-based 3D pop-up structure on the elastomer was performed using static analysis in ANSYS^TM simulations. To achieve ideal strain uniformity on the elastomer, it was compressed in both vertical and horizontal directions equally. The stress-strain curves of an elastomer (Dragon skin 10) were measured using motorized force test equipment (ESM303 tensile tester, Mark-10), and results from these measurements were utilized as input data for simulations. The mechanical properties of elastomers were accurately modeled using Yeoh’s 2nd-order hyperelastic model derived from these curves⁵⁰. The Young’s modulus and Poisson’s ratio for PDMS were set at 1.2 MPa and 0.49, respectively, while for the adhesive material (DOWSIL^TM SE 9186), these were set at 1.1 MPa and 0.49. The three-dimensional structure was constructed using a 5 µm-thick layer of polyimide, which has an effective bending rigidity comparable to that of OLED materials.