Interfacial Degradation and Pattern Evolution of Exfoliated Graphene by Cyclic Mechanical Loading

The interfacial interactions between 2D materials and polymer substrates receive increasing interest due to the surge of flexible electronics, multifunctional coatings, and nanocomposites. Although the strain effect on electrical, optical, and mechanical properties of 2D materials is extensively investigated, understanding the interfacial mechanics of 2D material‐polymer systems by dynamic loading is still a challenge. Here, the interfacial degradation and pattern evolution of mechanically exfoliated single‐ and few‐layer graphene on polydimethylsiloxane (PDMS) substrates by cyclic mechanical loading, are reported. It is found that the tensile strain leads to interfacial slippage between graphene and PDMS, whereas the compressive strain can be transferred to graphene with a transfer efficiency above 80%. Through cyclic loading, the graphene surface is seriously deformed by formation of multiple instability patterns including wrinkles and cracks. The morphological characteristic and evolution mechanism of the wrinkles and cracks are analyzed and discussed in detail. The interfacial adhesion energy is evaluated by wrinkle profiles and it decreases from ≈23 to ≈2 mJ m−2 as the cycle number increases. This work can promote better understanding of the interfacial effect of 2D materials on polymer substrates and controllable fabrication of various wrinkled or crumpled surfaces of 2D functional materials by cyclic mechanical loading.


Introduction
In the past decade, 2D materials have attracted considerable attention owing to the unique electrical, optical, thermal, and www.advmatinterfaces.de tension, the Poisson's effect-induced lateral compression can also cause straight wrinkles along the loading direction, [21] leading to the formation of a mosaic wrinkle pattern after strain releasing. [22] Recently, the cyclic loading of 2D materials has also attracted intensive attention due to the potential applications in stretchable electronics and artificial skins. [23][24][25] Wang et al. investigated strain transfer behavior of graphene-polymethyl methacrylate (PMMA) interface by cyclic loading of 0.7% strain. [23] Cui et al. reported fatigue property and damage mechanism of freestanding graphene and graphene oxide by atomic force microscopy-based fatigue testing. [24] The same group also reported fold formation and fatigue fracture of graphene by exerting a larger strain. [25] However, many fundamental issues including the efficiency of strain transfer, mechanism of interfacial degradation, evolution of surface morphology, and influence of material size and thickness by dynamic loading are still unclear up to now. Here, we report the interfacial effect and pattern evolution of single-and few-layer graphene on polydimethylsiloxane (PDMS) substrates by cyclic loading under large strain condition. The novelty of this work is concluded as follows. First, although the interfacial slippage between graphene and PDMS occurs during tensile loading, a high transfer efficiency of compressive strain (above 80%) has been achieved. Second, the interfacial adhesion energy is deduced by wrinkle profiles, and it decreases greatly during the cyclic process. Third, the cyclic loading of single-layer graphene is performed, and the influence of graphene layer number on pattern evolution is also investigated. It is anticipated that the technique of cyclic mechanical loading is universal and can be applied to other 2D materials.

Strain Transfer Efficiency of Graphene on PDMS
The graphene was first attached to a commercial Si wafer with 300 nm thickness SiO 2 surface layer by mechanical exfoliation, as shown in Figure 1a. Then, it was transferred to a PDMS substrate, as shown in Figure 1b. The AFM images of the graphene on SiO 2 and PDMS substrates are shown in Figure 1c,d, respectively. The values of average surface roughness are both equal to ≈0.6 nm, indicating that the graphene on both substrates has a relatively flat surface and a low defect level. Figure 1e shows an enlarged AFM image near the graphene-SiO 2 boundary. The average step height is detected to be ≈3.0 nm, as shown in the inset of Figure 1e. The Raman spectra of PDMS and graphene are shown in Figure 1f. The thickness of graphene at position 2 is twice larger than that at position 1 due to the simple fold of graphene at edge. As the thickness of graphene increases, the peak intensity of PDMS decreases gradually due to the attenuation of laser intensity when penetrating from the graphene. Figure 1g,h show the optical images and Raman spectra of other graphene flakes with different layer numbers. The G band and 2D band of graphene can be seen clearly and no D band is observed both on the SiO 2 and PDMS substrates, indicative of a high quality of the mechanically exfoliated graphene. [26][27][28] Through the combination of Raman spectroscopy and AFM imaging, the layer number of graphene can be well determined (for details, see Section S1, Table S1 and Figures S1 and S2, Supporting Information). In this study, we will focus on the interfacial mechanics of single-and few-layer graphene on the PDMS substrate.

Figure 2a
shows the schematic of uniaxial mechanical loading of graphene on the PDMS substrate. After graphene transferring, the PDMS is first uniaxially stretched to a 50% strain and then the strain is released slowly to complete a cycle of mechanical loading (namely 1st cycle or C1). This process goes on and a desired cycle number (namely Nth cycle or CN) can be achieved. Figure 2b shows the optical images of a single-layer graphene flake (the inset of Figure 1g) during the first stretching process. Our experiment shows that the multiand few-layer graphene can be perfectly transferred onto the PDMS substrate while the single-layer graphene usually generates obvious wrinkles or folds during transferring mainly due to the extremely small bending stiffness (for details, see below). During the stretching process, the graphene surface keeps stable and its length is almost unchanged, as shown in Figure 2d. Many previous studies showed that only a very small strain (≈0.8%) can be transferred to the graphene from a polymer substrate by uniaxial stretching or bending. [15][16][17] Li et al. adopted a simple method to improve interfacial interaction and eliminate slippage or decoupling by encapsulating single-layer 2D materials in flexible substrates through spincoating. [29] Wang et al. reported that the transferred tensile strain can be enhanced to be ≈2% when a formvar resin layer was adopted to improve the interfacial adhesion between the graphene and polymer substrate. [30] As the uniaxial mechanical strain is beyond the critical value, the interfacial sliding occurs and the graphene may undergo a resilience, but the graphene still adheres to the polymer substrate.
When the strain of stretched PDMS is released gradually; however, the graphene surface is deformed obviously, as shown in Figure 2c. The graphene length decreases gradually with a approximately linear fashion when releasing, as shown in Figure 2d. The compressive strain can be expressed as where L is the final length and L 0 is the initial length. It is clear that the compressive strain is negative because L < L 0 . For simplicity, we use the absolute value of compressive strain in this paper. During the releasing process, the PDMS length decreases from 36 to 24 mm while the graphene length decreases from 30.4 to 21.9 µm. The mechanical strain and graphene strain can be calculated to be ≈33.3% and 28.0%, respectively. That is, the compressive strain can be effectively transferred to graphene from PDMS substrate with a transfer efficiency of ≈0.84. Note that the graphene width is almost unchanged during the first cyclic process (see Figure 2d) mainly due to the very small width of this graphene. The transfer efficiency of compressive strain is strongly dependent on the graphene size. [31]

Evolution of Wrinkles During Uniaxial Compression
As the shape and size of graphene have an important role in the strain transfer efficiency; here, we adopt a long and wide graphene flake, as shown in Figure 3a. The pristine length and width of this graphene are ≈34.4 and 26.4 µm, respectively. After the PDMS is stretched to a 50% strain, the graphene length is almost unchanged due to the interfacial sliding. [15][16][17] However, the graphene width is obviously decreased due to the Poisson's effect of PDMS, which is quite different from the long but narrow graphene shown in Figure 2. This means that the compressive strain also can be effectively transferred to the graphene by uniaxial stretching. Driven by the transverse compression, buckle-delaminations (called as wrinkles in this paper according to most of the literature) along the loading direction form, consistent with the previous studies. [21,22] Then, the strain of stretched PDMS is released gradually. Figure 3b-e shows the morphological evolution of the graphene during the releasing process taken by the AFM. The optical images of the graphene during the first cycle are shown in Figure S3, Supporting Information. Here, we adopt the mechanical strain to characterize the compressive state of graphene during releasing. It is clear that straight wrinkles perpendicular to the loading direction form during the releasing process. [18][19][20] Figure 3f shows that the surface of pristine graphene is flat. As the compressive strain increases, the number and average height of the wrinkles both increase. On the other hand, the www.advmatinterfaces.de wrinkles along the loading direction formed during stretching remain; although, their height may be somewhat decreased, as shown in Figure 3g. This results in the spontaneous formation of orthometric wrinkle network on the graphene surface. Figure 3h shows that the graphene length decreases from pristine value (34.4 µm) to 25.0 µm during the releasing process. The graphene strain can be calculated to be ≈27.3% and the transfer efficiency of compressive strain is ≈0.82. They are both very close to the values of the long but narrow singlelayer graphene shown in Figure 2. On the contrary, the graphene width increases slightly from 23.2 to 25.3 µm during this process. Note that the width at unloaded state (25.3 µm) is still smaller than the pristine one (26.4 µm) mainly due to the remaining wrinkles along the loading direction, showing an irreversibility of graphene surface under mechanical loading.

Graphene Fracture by Cyclic Loading
After the graphene has completed a stretching/releasing process ( Figure 3), we further investigate its morphological evolution by cyclic mechanical loading, as shown in Figure 4a. More AFM images are shown in Figure S4, Supporting Information, and the optical images are shown in Figure S5, Supporting Information. It is clear that the graphene is seriously deformed by cyclic loading. Figure 3 has shown that the graphene length decreases drastically while the width only changes slightly after the first cycle (C1). During the subsequent cyclic process, the graphene length and width both decrease quickly first and then the decay rates slow down gradually. Last, the graphene length and width approach to respective saturated values, as shown in Figure 4b. The graphene width has larger decreased amplitude compared to the length during the subsequent cyclic process. At C10, a horizontal crack and a vertical crack are simultaneously nucleated at the graphene interior. Their lengths increase drastically and reach to respective saturated values quickly, as shown in Figure 4c. The saturated length of vertical crack is larger than that of horizontal crack. On the other hand, the crack widths increase comparatively slowly (especially for the horizontal crack), as shown in Figure 4d. The saturated width of vertical crack is much smaller than that of horizontal crack.
It is well known that the material fracture can be classified into three types: Modes I, II, and III. [32,33] Mode I fracture (opening mode) requires the applied stress being normal to the crack surface. Mode II fracture (shear mode) requires the applied stress being parallel with the crack direction. Mode III fracture (out-of-plane shear or tear mode) requires the shear stress at the crack front being normal to the material surface. Figure 3 shows that the bottom-right corner of the graphene is separated from the main flake during the first stretching process, which is a typical signature of Mode I fracture. During the subsequent cyclic process, both the vertical and horizontal cracks can be attributed to the combination of the Mode I and Mode II fracture. For vertical (horizontal) cracks, the Mode I (Mode II) fracture is dominant because they are perpendicular (parallel) to the loading direction. However, the Mode II (Mode I) fracture also has an obvious influence on the crack formation due to the Poisson's effect-induced uniaxial tension (perpendicular to the loading direction) during the releasing process. Our experiment shows that the graphene contraction (or interfacial sliding) is mainly determined by two factors: strain level and pre-existing surface wrinkle. During the first cyclic process, the www.advmatinterfaces.de graphene contraction along the horizontal direction is great due to the larger strain in this direction ( Figure 3). Subsequently, the graphene contractions along the horizontal and vertical directions are almost uniform (Figure 4b). After crack formation; however, the graphene contraction along the vertical direction is obviously larger than that along the horizontal direction (Figure 4b,d). This phenomenon can be attributed to the preexisting wrinkles which are predominantly along the vertical direction.
As the cycle number increases, more horizontal and vertical cracks form at the graphene interior and edge (Figure 4a). Some cracks connect with each other to form longer and wider cracks. Last, the graphene is split into small pieces or narrow stripes and its shape is almost unchanged thereafter. Figure 4e shows the evolution of area ratio of deformed graphene to pristine graphene during the cyclic process. The area ratio of deformed graphene is defined as the proportion of the pixels in the graphene region by binarization method. During the first cycle (C1), the area ratio decreases drastically from 100% to 68.8% due to the remarkable decrease of graphene length (Figure 3). At the early stage of cyclic loading, the area ratio decreases quickly due to the simultaneous diminution of graphene length and width (Figure 4b). Then, the decay rate slows down gradually but the area ratio does not reach to a saturated value until the largest cycle number used in our experiment (C450). This can be attributed to the continuous evolution of cracks inside the flake beause the graphene length and width are both unchanged finally.

Wrinkle Evolution and Interfacial Degradation by Cyclic Loading
To discover the evolutional details of wrinkles, we have taken enlarged AFM images at the early cycle stage, as shown in Figure 5a. We find that the wrinkle morphology, position, and number can change obviously during cyclic loading. For example, a new wrinkle forms in the middle of two already existing wrinkles (see blue arrows in Figure 5a). One wrinkle is split into two parts or two wrinkles are merged into one (see black arrows in Figure 5a). Furthermore, the width and height of a single wrinkle are usually not uniform and they are position-dependent. This nonuniformity becomes more prominent as the cycle number increases. They may originate from the horizontal wrinkles and subsequently formed cracks. The line profiles in Figure 5b show that the wrinkles have the typical profile feature of buckle-delaminations observed in traditional films [34,35] or 2D materials. [10][11][12][19][20][21][22] Note that such profile is completely different from that of periodic wrinkles in a filmsubstrate bilayer system where the film is well adhered to the substrate during strain loading due to the strong interfacial adhesion. [36][37][38][39] Although the wrinkle profile may change obviously during the cyclic process, it still can be well fitted by a sine curve, as shown in Figure 5c. The wrinkle width and maximum height are defined as w and δ, respectively. As the wrinkle profiles are not uniform, here we adopt average width and height of wrinkles by measurement of more than 20 wrinkle profiles for each We suggest that once two or more cracks merged, the length of an initial crack kept constant. That is, the measured length does not include the length of the merged crack.

www.advmatinterfaces.de
sample. The evolutions of average wrinkle width w and height δ during the cyclic process are shown in Figure 5d. As the cycle number increases, the wrinkle average width increases monotonously while the average height increases first and then decreases steadily. The schematic for evolution of wrinkle profile is shown in the inset of Figure 5e. During the cyclic process, the aspect ratio of wrinkles increases from ≈0.09 to ≈0.12 first (C1-C3) and then decreases steadily to ≈0.05 at C40, as shown in Figure S6a, Supporting Information It is well known that the interfacial adhesion is strongly dependent on the wrinkle profile. According to the continuum mechanics theory, the adhesion energy Γ between the film (or 2D material) and substrate can be calculated as [34,40,41] In this study, the Young's modulus and Poisson's ratio of graphene are E = 1000GPa and ν = 0.165 from the literature. [42,43] The graphene thickness is ≈3 nm from the Raman spectroscopy and AFM imaging. Then, the adhesion energy can be calculated by the wrinkle profile during cyclic loading, as shown in Figure 5e. The initial adhesion energy is ≈23 mJ m −2 , which is close to the value of graphene-polymer interface measured by various techniques. [44,45] It is also very close to the interfacial adhesion between other 2D materials and elastomers measured by buckling modes. [40,41] As the cycle number increases, the adhesion energy decreases steadily to ≈2 mJ m −2 , showing a strong degradation of graphene-PDMS interface by cyclic mechanical loading.
Furthermore, the maximum tensile strain on the wrinkle peak can be calculated based on the wrinkle profile [19,34,46] h w (1 ) In our experiment, the initial maximum tensile strain is ≈0.7%, and then, it decreases steadily to ≈0.2% at C40, as shown in Figure S6b, Supporting Information. Such strain is very small compared with the ultimate fracture strain of graphene (≈25%). [42,43] On the other hand, the tensile strain exerted on the graphene by mechanical loading (≈0.8%) is also much smaller than the ultimate fracture strain. Obviously, the cyclic loading is susceptible to generating fatigue fracture at small strain condition (far below the fracture strength). [24,25] The fatigue cracks are not caused by the wrinkle profile but are originated from the impurities and dislocations in the graphene. [47,48] Figure S7, Supporting Information shows the enlarged AFM images of graphene during the late cycle stage when fatigue cracks are dominant. After the graphene is split into small pieces, the initial vertical wrinkles change into orthometric wrinkle network gradually due to the imposed biaxial compression by cyclic loading.
Based on Figures 4 and 5, we conclude that the resulting stable graphene flake with fragmented feature can be attributed to the following reasons. First, the shrinkage of graphene flake

www.advmatinterfaces.de
greatly decreases the contact area between the graphene and PDMS (Figure 4e). Second, the enlargement of wrinkle width (with interfacial delamination) further decreases the contact area ( Figure 5d); thus, decreasing the strain transfer efficiency by mechanical loading. Third, the interfacial strength becomes weak significantly by cyclic loading (Figure 5e). Fourth, the strain transfer efficiency has a strong size effect and it decreases greatly as the graphene size decreases. [31] Fifth, the shrinkage of graphene flake may increase the equivalent thickness; thus, increasing the resistance for further deformation.

Uniaxial Loading of Single-Layer Graphene
To further understand the influence of graphene thickness (or layer number) on the instability patterns by mechanical loading, we have prepared single-layer graphene flakes on the SiO 2 substrate, as shown in Figure 6a. The Raman spectra of these graphene flakes are shown in Figure S8, Supporting Information. During transferring onto the PDMS substrate, the single-layer graphene flakes undergo an obvious surface instability due to the extremely small bending stiffness, as shown in Figure 6b,c. The instability patterns include folding at edge, wrinkling with interfacial delamination, and cracking at the edge or interior. It is interesting that a distinct shaded region can be seen clearly in the AFM image, and it strictly coincides with the initial shape of graphene flakes on the SiO 2 substrate. The line profile shows that the shaded region has a depth of ≈20 nm, as shown in Figure 6d. That is, the graphene flakes are in fact located into shallow pits with ≈20 nm depth after transferring on the PDMS substrate. As the Young's modulus of graphene (≈1 TPa) is extremely larger than that of PDMS (≈1 MPa), the graphene has a higher brightness or z-value due to the constant force mode of AFM. Therefore, the AFM profile cannot mirror the actual step height at the graphene/PDMS boundary as shown in Figure S9, Supporting Information. Similar shaded region can be also observed in few-layer graphene on PDMS after mechanical loading (Figures 3 and 4). We suggest that when the graphene contacts with the PDMS substrate, the PDMS surface under graphene shows a slight depression due to the extreme modulus mismatch between the graphene and PDMS. After etching of SiO 2 and transferring of graphene, the depression permanently remains on the PDMS surface. Figure 2 shows the evolution of single-layer graphene (the upper flake in Figure 6a) during the first cycle (C1). The graphene length (along the loading direction) is almost unchanged during stretching while it decreases obviously during releasing. Driven by the horizontal compressive strain during releasing, more wrinkles along the vertical direction form after C1. Figure 6e shows the morphological evolution of the singlelayer graphene during the cyclic loading taken by the AFM. The optical images of dynamic evolution are shown in Figure S10, Supporting Information. We find that the graphene is seriously deformed by formation of wrinkles and cracks, similar to the experimental result of few-layer graphene shown in Figure 4a. Figure 6f shows that as the cycle number increases, the graphene length and width both decrease quickly first, and then they reach to respective stable values, which is similar to the evolutional behavior of few-layer graphene. However, the prominent evolution of graphene sizes only occurs within the initial ten cycles for single-layer graphene while it occurs within the initial 100 cycles for few-layer graphene (Figure 4b). Note that the original surface deformations of single-layer graphene have no obvious influence on the cycle number because the www.advmatinterfaces.de few-layer graphene can reach similar surface structures only after several cycles (Figure 4a). This indicates that the resistance of deformation decreases greatly with decreasing layer number of graphene because the transfer efficiency of compressive strain is similar for single-and few-layer graphene. The extremely small bending stiffness of single-layer graphene should be responsible for the behavior of quicker dynamical evolution under cyclic loading.

Conclusion
In summary, the interfacial degradation and pattern evolution of single-and few-layer graphene on the PDMS substrate by cyclic mechanical loading are described and discussed in detail. It is found that the graphene length along the loading direction is almost unchanged during the stretching process, whereas it decreases with a linear fashion during the releasing process. The transfer efficiency of compressive strain is above 80% and it is insensitive to the layer number of graphene. Driven by the vertical and horizontal compressive strains in sequence, orthometric wrinkles form after the first cycle. During the subsequent cyclic process, the graphene surface is seriously deformed due to the formation of vertical and horizontal cracks, which are resultant from the combination of Mode I and Mode II fracture. The graphene length, width, and area ratio all decrease quickly first and then reach to stable values gradually. The single-layer graphene has a quicker dynamical evolution compared to few-layer graphene. The interfacial adhesion energy decreases about one order of magnitude (from ≈23 to ≈2 mJ m −2 ) as the cycle number increases. The attenuation of interfacial adhesion and diminution of contact area between the graphene and PDMS result in the decrease of strain transfer efficiency greatly and formation of fragmented graphene flake finally. This work provides a deep insight into the strain transfer, interfacial degradation, and pattern evolution of 2D materials on polymer substrates by cyclic mechanical loading, which can help to optimally design 2D material-based flexible electronics devices. Furthermore, the strain-modulated technique can be developed to fabricate various wrinkled or crumpled surfaces of 2D functional materials, which are beneficial for photoelectric detection, catalysis, sensor, wetting, friction, and so on.

Experimental Section
Preparation of Substrates: Commercial silicon wafer with a 300 nm thickness SiO 2 layer atop (purchased from Zhejiang Lijing Photoelectric Technology Co., Ltd.) was cut into 10 × 10 mm 2 size by a glass knife, and then was ultrasonically cleaned using alcohol for 5 min to remove the contaminants of the SiO 2 surface. The pre-polymer and cross-linker of polydimethylsiloxane (PDMS, Sylgard-184, Dow Corning) were evenly mixed based on the standard mass ratio of 10:1. After degassing, the liquid PDMS was poured into a petri dish and was cured at 70 °C for 5 h. The thickness of solidified PDMS was ≈1.2 mm. The PDMS was cut into 30 × 12 mm 2 rectangular sheets by a sharp knife for graphene transfer and mechanical loading.
Exfoliation and Transfer of Graphene: Graphene flakes were prepared by mechanical exfoliation of highly oriented pyrolytic graphite (HOPG) using the 3M tape method and were deposited directly on the 300 nm SiO 2 /Si substrates. Then, the PDMS sheet was gently placed on the SiO 2 surface with deposited graphene flakes at room temperature. The entire structure was submerged in 1 mol L −1 KOH solution for 20 min, resulting in etching of SiO 2 and adhering of graphene to the PDMS surface. The PDMS was rinsed with deionized water for ten times to remove any KOH residue. Last, the sample was put into a vacuum drying oven at 30 °C for 10 min to remove any water residue. The transfer technique used in this study was similar to some previous works. [20,49] Here, the step of spin coating a layer of PMMA was lacking to eliminate PMMA residue and simplify experimental procedure. Although single-layer graphene may generate some wrinkles and cracks during transferring, the same transfer technique was adopted to keep the same interface condition with few-layer graphene.
Mechanical Loading: The PDMS with transferred graphene flakes was mounted on a custom-designed uniaxial stretching device, as shown in Figure S11a, Supporting Information. The length of PDMS between two clamps was fixed at 24 mm. Then, the PDMS was uniaxially stretched to the length of 36 mm, corresponding to a tensile strain of 50%, as shown in Figure S11b, Supporting Information. Then, the strain of PDMS was released gradually to complete a cycle of mechanical loading, as shown in Figure S11c, Supporting Information. This process went on and a desired cycle number was achieved, as shown in Figure S11d, Supporting Information. To eliminate the influence of strain rate on the interfacial degradation and pattern evolution of graphene, the loading/unloading procedure was uniform for all the samples in this work. In detail, the PDMS was loaded or unloaded step by step with a fixed variation of 1 or 2 mm PDMS length during the first cycle. For subsequent process, the mechanical loading/unloading was successive and the time for completing each cycle was ≈10 s. Due to the excellent elasticity, the PDMS could be fully recovered after hundreds of cycling (with 50% strain) in this experiment.
Definition of Strain: The graphene length along the loading direction was almost unchanged during the stretching process due to the interfacial sliding, while it decreased obviously during the releasing process. To describe the strain transfer more effectively, the loaded condition (50% PDMS stretching) was defined as the initial state and the unloaded condition (PDMS releasing) as the final state. Thus, the compressive strain of PDMS was fixed at 33.3% and the compressive strain of graphene was calculated by with initial length L 0 and final length L. For simplicity, the absolute value of compressive strain was used in this paper. The ratio of graphene strain to PDMS strain was defined as the transfer efficiency of compressive strain.
Characterization: The surface morphologies and instability patterns of graphene were detected by an optical microscope (Olympus BX41) and an atomic force microscope (AFM, JPKSPM) using quantitative imaging mode with PPP-FMAuD tips. The Raman spectra were carried out by a Raman spectrometer (Horiba Jobin Yvon, LabRAM HR EVO) equipped with a charge coupled detector (CCD), a 100× (NA = 0.9, Olympus) objective, and a grating (1800/mm@500nm). The laser wavelength was 532 nm and the laser power was 26 µW. Note that the loading experiment could be roughly divided into two parts: the stretching/ releasing process (Figures 2 and 3) and the cyclic process (Figures 4-6). During the stretching/releasing process, both the optical and AFM images were taken at different strain states. During the cyclic process, the optical and AFM images were taken after the cycle was complete (i.e., the PDMS was recovered to its original length).

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