Solvent Responsive Self‐Folding of 3D Photosensitive Graphene Architectures

Stimuli responsive self‐folding structures with 2D layered materials (2DLMs) are important for flexible electronics, wearables, biosensors, bioelectronics, and photonics. Previously, strategies have been developed to self‐fold 2D materials to form various robots, sensors, and actuators. Still, there are limitations with scalability and a lack of design tools to obtain complex structures for reversible actuation, high integration, and reliable function. Herein, a mass‐producible strategy for creating monolayer graphene‐based reversible self‐folding structures using either gradient or differentially cross‐linked films of a negative epoxy photoresist widely used in microfluidics and micromechanical systems, namely, SU8 is demonstrated. Wafer‐scale patterning and integration of complex and functional devices in the form of rings, polyhedra, flowers, and bidirectionally folded origami birds are achieved. Also, gold (Au) electrodes to realize functional graphene–Au Schottky interfaces with enhanced photoresponse and 3D angle sensitive detection are integrated. The experiments are guided and rationalized by theoretical methods including coarse‐grained models, specifically developed for the tunable mechanics of this photoresist that simulate the folding dynamics, and finite element method (FEM) electromagnetic simulations. This work suggests a comprehensive framework for the rational design and scalable fabrication of complex 3D self‐actuating optical and electronic devices through the folding of 2D monolayer graphene.


Introduction
Self-folding of 3D integrated microstructures from wafer-scale integrated twodimensional (2D) precursors has enabled the creation of a range of 3D reconfigurable devices of broad relevance for optics, electronics, sensing, microfluidics, robotics, and biomedical engineering. [1][2][3][4][5][6] Different approaches, such as folding and buckling have been used to assemble 3D mesostructures. [7] Recently, the incorporation of 2D ultrathin high-performance materials such as monolayer graphene has been a significant focus, especially to realize light-weight, transparent, and flexible devices. [8][9][10] Researchers have demonstrated the folding, rolling, or twisting of 2D ultrathin materials like graphene with precise control. [11,12] For example, Joung et al. self-assembled 100 micron-scale 3D polyhedra with graphene and graphene oxide using capillary forces associated with polymer reflow. [13] Xu et al. have reported thermo-responsive, reversible self-folding graphene by coating it with an ultrathin layer of poly(N-isopropyl acrylamide) (PNIPAM), [14] while Miskin et al. have created cell-size origami pH responsive machines using graphene-silicon dioxide bimorphs. [15] Despite these impressive demonstrations, the realization of wafer-scale integration of on-chip or free-standing and reversibly reconfigurable integrated polymer-functional 2DLM hybrid devices has proven challenging. This is the focus of our study.
SU8 is an epoxy-based ultraviolet (UV) cross-linkable negative photoresist based on a commercial resin and the number 8 refers to the eight epoxy groups in its structure. [16] The extent of crosslinking of SU8 can be controlled by tuning the UV dose, and Jamal et al. first demonstrated that this feature could be used to create solvent responsive self-folding devices using gradient cross-linked films. [17] Since then, various self-folding devices have been realized using SU8 hybrids [11,18] including using 2D materials such as MoS 2 . [19,34] In this study, our scientific objectives are to thoroughly understand the folding mechanics of differentially cross-linked SU8 and the interaction of light with flexible graphene-Au-SU8 3D microstructures. Using experiments and simulations, we introduce several new ideas and demonstrations for self-folding SU8 graphene microstructures. First, we show that both bilayer and gradient cross-linked SU8 can be used for wafer-scale fabrication of self-folding graphene hybrids. For the first time, we systematically vary the extent of SU8 cross-linking by tuning UV dose, carefully measure moduli of such cross-linked SU8 flims and develop a physics-based coarse-grained material model that captures the effect of UV light on the material mechanics and volume change. This model is used in folding simulations to rationalize experimental curvature and fold angle. Second, we provide the first evidence that complex 3D shapes with bidirectional curvature, such as origami birds, can be created using this approach. It is important to note that our method utilizes highly scalable and cost-effective multilayer very-large-scale integration (VLSI) approaches with a well-established epoxy photoresist. Hence, our approach allows facile integration with electrodes or potentially other electronic, optical, or microfluidic modules. Third, using a variety of sacrificial layers, we integrate on-chip interconnects with these self-folded structures such that part of the structure can be released while other parts of the electrodes remain on chip, which is important to create solder or wire bonded connections for electronic applications. Fourth, we demonstrate multilayer electrode integration and the first reversibly configurable photosensitive graphene-gold (Au)-SU8 devices that show enhanced and angle sensitive photoresponse. Together, our studies demonstrate a new framework for mass fabrication and integration of 3D transparent soft high-quality chemical vapor deposition (CVD) grown monolayer or few layer graphene (FLG) hybrid functional devices of importance for photonics, wearables, and robotics.

Rational Design of 3D Self-Folding SU8 Structures
We investigated two methods to enable reversible folding of differentially cross-linked SU8 films, namely bilayer ( Figure 1a) Mechanism and versatility of self-folding SU8 films. a) Self-folding SU8 bilayer where the SU8 bottom layer is fully cross-linked and the top SU8 layer is partially cross-linked. The SU8 bilayer film folds upward on the solvent exchange between acetone and water. b) Self-folding SU8 with low UV dose gradient cross-link density across the single SU8 film. The SU8 film folds downward on the solvent exchange between acetone and water. The bilayer SU8 beams' total thicknesses were, c) 20 μm, and d) 10 μm, and UV r ¼ 0.5. Controlled folding of bilayer SU8 stars, e) UV r ¼ 0.8, the thickness was 10 μm. f ) UV r ¼ 0.5, the thickness was 10 μm. g) UV r ¼ 0.5, the thickness was 5 μm. h) SU8 ribbons folded into i) helices, and j) an SU8 star folded into k) a square pyramid using the gradient method. Scale bars are, c-g) 500 μm, h,i) 800 μm, and j,k) 300 μm.
www.advancedsciencenews.com www.advintellsyst.com and gradient (Figure 1b) modalities. For both approaches, we deposited a 50 nm thick thermally evaporated copper (Cu) sacrificial layer on top of a wafer or a glass slide. For structures formed by the bilayer method, we patterned SU8 bilayer films with a fully cross-linked bottom layer and a partially cross-linked top layer via photolithography to enable bending away from the wafer (Figure 1c-g). We spin-coated the first layer of SU8 2000 series photoresist on top of the Cu-coated substrate. We fully crosslinked it by exposure to a high dose (240 mJ cm À2 ) of UV irradiation through a photomask defining the pattern of SU8. We then spin-coated the second layer of SU8 of the same thickness on top of the first pattern and partially cross-linked it with a low dose (100-192 mJ cm À2 ) of UV irradiation. Then we developed the patterns in SU8 developer, which removed the uncross-linked SU8 portions. After differential bilayer cross-linking and patterning, we dissolved the sacrificial Cu layer. Then, we conditioned the SU8 bilayer patterns by immersion in acetone, typically for 3-5 min, to create the self-folding precursors. These conditioned structures reversibly fold and unfold on solvent transfer from water to acetone. By varying the thickness and type of pattern, we can assemble curved beams with different radii (Figure 1c, d) and various 3D shapes (Figure 1e-g). For example, bilayer beams where the thickness of the top layer and the bottom layer is 10 μm curve to a lesser extent ( Figure 1c) than bilayer beams where the layer's thickness is 5 μm (Figure 1d). We attribute this to the increasing bending stiffness with increasing thickness. We observed that we could also increase folding extent by decreasing the ratio of the bilayer's UV exposure dose ratio. We define this ratio as UV r ¼ UV 1 /UV 2 , where UV 1 is the UV exposure dose for the top layer and UV 2 is the UV exposure dose for the bottom layer. For example, when UV r is decreased from 0.8 to 0.5, the extent of folding of the SU8 star patterns increases (Figure 1e,f ). We also created SU8 stars with both decreased thickness (thickness was halved) and a lower ratio (UV r ¼ 0.5) to assemble the tightest folded structures (Figure 1g).
For self-folding with gradient cross-linked SU8, we typically spin-coated 10.7 μm-thick SU8 on a 50 nm-thick Cu sacrificial layer-coated wafer. In contrast to the bilayer cross-linked film, only one layer of SU8 was used, and the entire film was exposed to low dose 120 mJ cm À2 UV irradiation through a photomask and developed. This photopatterning defined the shape of the SU8 sample, and, at the same time, generated a cross-link gradient across the thickness of the SU8 layer, where the top portion of the SU8 film gets more cross-linked than the bottom part. [17] After differential cross-linking, we released the SU8-patterned films from the substrate by dissolving the sacrificial Cu layer. After release, we conditioned the films in acetone which removes uncross-linked material. When the conditioned hydrophobic SU8 sample is subsequently exposed to water, it de-solvates, and the lower cross-linked side undergoes greater contraction, causing the film to curve or fold towards the less cross-linked side ( Figure 1b). These differentially cross-linked regions exposed to low dose 120 mJ cm À2 UV light can be interspersed by regions of SU8 that are fully cross-linked by exposure to highdose UV light of 240 mJ cm À2 to create complex shapes such as helices (Figure 1h,i) and square pyramids (Figure 1j,k).
An important contribution in this work is the systematic rationalization of the self-folding of SU8 using both mechanical measurements and simulations. We measured the modulus of SU8 beams cross-linked with different UV doses using nanoindentation. We observed that the Young's modulus of SU8 increased from 4.99 to 6.12 GPa when the UV dose was increased from 120 to 240 mJ cm À2 , and we can fit a curve for the modulus versus exposure dose based on our data ( Figure 2a). We developed a coarse-grained dynamics simulation model (Note S1, Figure S1-S3, Supporting Information) to rationalize the relationship between the radius of curvature (ROC) as a function of UV exposure intensity (I 0 ), thickness (t) of a single layer in bilayers ( Figure 2b) and intensity gradient (dI/dt) for gradient cross-linked SU8 beams (Figure 2c). The graphs illustrate the Individual points are measured values and the straight line indicates a theoretical fit to these points and the modulus value at the threshold exposure from the SU8 datasheet. b) A plot of the averaged ROC for a bilayer rectangular SU8 beam with dimensions 250 Â 500 μm, as a function of the thickness (t) and exposure intensity (I 0 ) of the top layer. The red indicates the bottom SU8 layer (fully exposed to UV with 240 mJ cm À2 ), and the blue layer is the top SU8 layer (exposed to UV with I 0 ). c) A plot of the ROC for a gradient cross-linked rectangular SU8 beam with dimensions 250 Â 500 μm, as a function of exposure intensity (I 0 ) of the top layer (color in red with energy of I 0 ) and gradient intensity decreasing along the thickness given by (dI=dt). The red layer indicates the portion of the SU8 beam closer to the UV light source, and the blue layer is the portion of the SU8 beam away from the UV light source. In both panels, b,c), the inset images show beam geometries at different plot parameters, obtained from individual simulations. Each set of the red dots/lines is obtained from an individual folding simulation with the lines indicating the standard deviation from the fixed end to the ribbon's free end. A linear interpolation gives the surface on the scattered data points obtained from simulations and the yellow colors indicate high ROC (averaged values) or low curvature while the blue ones show low ROC (averaged values) or high curvature.
www.advancedsciencenews.com www.advintellsyst.com criteria for achieving different fold angles based on relevant design parameters including thickness and extent of crosslinking. Essentially, for bilayer films, we observe that as I 0 increases, the ROC increases or the curvature decreases. We also observed a rich set of geometric folding shapes with different thicknesses. For gradient cross-linked films, we observed that beams bent with an opposite sign of ROC. The extent of curvature (lower ROC) increases with decreasing exposure intensity (I 0 ) and increasing intensity gradient (dI/dt). These graphs provide design criteria to achieve controlled bending and resulting geometries for SU8 microstructures; also, the simulations can reproduce the folding shapes observed in experiments ( Figure S3, Supporting Information).

Graphene into 3D Shapes with the Self-Foldable SU8 Structures
Importantly, we also show that the self-foldable SU8 structures can serve as a support layer for transforming flat monolayer graphene into 3D shapes. The integration process for free-standing structures involved the following steps ( Figure 3a). First, we transferred high-quality CVD grown monolayer graphene from a Cu catalyst-coated wafer to our Cu sacrificial layer-coated Si substrate using the polymethyl methacrylate PMMA) method. [20] We observed the classic peaks for monolayer graphene and SU8 in the Raman spectra of the transferred graphene on SU8 ( Figure S4, Supporting Information). [21][22][23] We patterned the graphene using photolithography and plasma etching as needed. We then spin-coated and differentially cross-linked the SU8 and conditioned the samples as described previously.
We observed the self-rolling of graphene-SU8 structures with reversible rolling and unrolling in water and acetone (Figure 3b-e). We observed no significant difference in the curvature of SU8 with and without graphene. Our integration process for self-rolling graphene-SU8 is a wafer-scale process, and we could integrate other elements such as Au lines or patterns to create functional electronic or optical devices.

Self-Rolling 3D Graphene-SU8 Photodetectors
Graphene has been extensively studied for electronic and optical applications due to its unique physical properties, including excellent mechanical strength and stability, [24] ultra-broadband light absorption due to the zero bandgap structure, and high carrier mobility due to its atomically thin low-dimensional structure. [25,26] Of relevance to optoelectronics, the high charge carrier mobility of graphene up to 10 5 cm 2 (Vs) À1 at ambient temperature indicates the potential to apply graphene in highfrequency and high-speed devices. The high optical absorption coefficient of 7 Â 10 5 cm À1 from ultraviolet to ultra-infrared (300-2500 nm) shows an enormous latent capacity for optoelectronics compared to conventional semiconductors with the "long-wavelength limit". [27][28][29][30] However, even with its high optical absorption coefficient, the light absorption and light-matter interaction of graphene are low for planar graphene-based devices since it is atomically thin. For example, the light absorptivity of single-layer graphene is %2.3%, limiting the photodetection applications of graphene. In this study, we leverage the high optical transparency of SU8 and focus on fabricating 3D selffolding graphene-based optical devices of relevance to flexible optical devices and wearables. To overcome the limitations of poor absorptivity of single-layer graphene, we created multirolled 3D graphene structures. [31] Our photodetection concept relies on the graphene-Au interface (Figure 4a and Figure S5, Supporting Information). When we irradiated this photodetector interface with a 488 nm laser, and connected the two electrodes to a voltmeter, we recorded a photovoltage. To investigate the origin of this photovoltage, we first scanned the laser in the direction perpendicular to the two electrodes on a flat graphene-Au-SU8 photodetector and measured the photovoltage when each Au electrode was illuminated. We observed no photovoltage when the laser spot was in the middle of the two electrodes; the photovoltage increased in magnitude when the laser spot was on either of the two electrodes, but the photovoltage had opposite signs on the two electrodes. Moreover, we also observed that the photovoltage was larger when the laser was incident directly from the graphene side ( Figure S6, Supporting Information) compared with laser illumination from the SU8 side (backside). We attribute this reduced photovoltage on illumination from the SU8 side to the loss due to absorption of light in the SU8 film. We confirmed that the photovoltage generated in this study is mainly caused by the light absorption in the overlap area of Au and graphene, which is when the symmetry of the fabricated structure gets broken.
Based on this observation, we conclude that unlike prior work on graphene photodetectors, [27,32] the mechanism for www.advancedsciencenews.com www.advintellsyst.com photovoltage generation in this study is the combined outcome of photovoltaic (PV) and photo-thermoelectric (PTE) effects (Details in the Figure S7, Supporting Information) related and associated with the injection of hot carriers across the potential barrier on the Au-graphene interface. [33] We reported on a similar effect, which we observed previously in MoS 2 /Au devices. [34] An advantage of using graphene over MoS 2 in this study is the ultra-broadband light absorption in graphene, which is attractive for designing versatile optoelectronic devices. We compared the graphene-Au-SU8 beams's photoresponse and observed a larger photovoltage in the rolled up versus flat state (Figure 4b-d). For instance, the photovoltage is around 28 μV for illumination at 5 mW laser power (488 nm) for flat graphene-Au-SU8, and the value is 104 μV for the folded one, which is 3.7 times higher. We attribute this higher voltage to high absorption of rolled graphene which agrees with simulations that show a linear increase in absorption with increasing number of rolls (Note S2, Figure S8, Supporting Information).
We also observed that the photovoltage generated is broad in the visible spectral range, and depends on the laser wavelength. As shown in Figure S7, Supporting Information, the generated photovoltage, with the same laser power using three different lasers (488, 532, and 600 nm) is highest at 488 nm, and lowest at 600 nm. As shown by dashed curves in Figure S7, Supporting Information, we can interpret the spectral dependence as the sum of two effects-a relatively weak PTE voltage and a stronger PV response with the threshold around 540 nm corresponding to a barrier height of %2.3 eV. [35]

Chip-Integrated 3D Graphene-SU8 Structures and Photodetectors
A highlight of our assembly process is that we can create complex origami-inspired designs including birds (Figure 5a-c) and chipintegrated structures where portions of the structure are attached to the substrate while other parts are released (Figure 5d-g). These complex structures were assembled by first patterning the Cu sacrificial layer and graphene and controlling the UV exposure in specific regions so that the SU8 microstructure could selectively fold in some regions whereas other parts remained flat and pinned down. For the bird structure, the core center is pinned on the chip, due to regions of partial and full cross-linkage to enable selective release and curving ( Figure S9, Supporting Information). We envision that such complex shapes could be important for soft robotics with the possibility for optical remote energy harvesting using the graphene-Au interface.
We also note that on-chip assembled designs are important in optoelectronics. To illustrate this feature, we demonstrate angle-resolved photodetection using a self-folded SU8 graphene photodetector array with multiple panels that fold along different angles on-chip ( Figure S10, Supporting Information, Figure 5f-k). When illuminated by light, they show different photoresponses based on the laser's angle and structure. For example, the photovoltage is 2.6 times higher with laser illumination at 60 and 30 than the 0 measurement (Figure 5j). To estimate the variation of light absorption with incident angles, we simulated a master curve of photoresponse variation with angle that illustrates the functional form of the variation and can be used to determine the angle-resolved response (Figure 5k, Note S3, Figure S11, Supporting Information).

Conclusion
In summary, we have described a highly parallel process for the assembly of 3D flexible graphene microstructures. As compared to prior approaches, our methodology is significant because it offers, a) both free-standing and on-chip integration, b) highly parallel integration of flexible and transparent 3D graphene devices, and c) reversible reconfiguration. It is important also to note that as compared to alternative strategies to fold up 3D graphene and MoSe 2 photodetectors such as based on prestretched elastomers, [36] SiN x , [37] and Si/SiO 2 , [38] this self-folding mechanism with photopatternable polymer is reversible. SU8 is a popular photoresist that can be spin coated, and is optically transparent, and relatively flexible. The SU8 self-folding process is highly parallel, reliable, and reproducible.
Moreover, our structures are stable in air, verified for over 6 months. We have demonstrated complexity and functionality for photodetection by the assembly of complex shapes. These include, a) curved beams of importance in mechanics and curvature estimation, b) polyhedra and spiral ribbons of relevance to bioelectronics, chiral metamaterials, and optics, and c) bird-shaped assemblies that highlight the capacity to www.advancedsciencenews.com www.advintellsyst.com create flexible and transparent bioelectronics of importance to robotics. We note that since 2D layered materials are ultrathin, integrated devices composed of these materials are being actively pursued as light-weight alternatives to silicon-based modules in flying or swimming robots. Experimentally observed measurements are rationalized using simulations that establish a framework for the design of specific shapes and curvatures. We note that compared to other stimuli such as light or electricity, chemical solvent driven self-folding is attractive for marine, biological, and environmentally triggered applications. The principle of self-folding is based on chemical solvent driven differential swelling. This principle can be tailored to other polymers with different affinity for aqueous or organic solvents, which is essential to realize for autonomous actuation in specific chemical or biological environments. We anticipate that this approach could be utilized to create a range of 3D microstructures composed of 2DLMs and is of broad relevance to 3D integrated designs for wearables, autonomous chemically triggered actuators, flying, or swimming robots, energy harvesting devices, and biosensors.

Experimental Section
Materials: CVD graphene was purchased from ACS Material, LLC (Pasadena, CA, USA). SU8 2000 series photoresist and developer were purchased from MicroChem (Newton, MA, USA). Acetone, ferric chloride (FeCl 3 ), and PMMA were purchased from Sigma-Aldrich. Chromium and gold evaporation sources were purchased from Alfa Aesar.
Nanoindentation: The hardness and elastic modulus were measured using the MTS Nanoindenter XP. A Berkovich tip with 2000 nm displacement into the sample was used and nine spots were averaged with a drift rate of %0.3 nm s À1 during measurements. Optical and SEM images of the self-folding of a graphene-Au-SU8 origami bird from, b) flat into c) its 3D shape. Optical and SEM images of the self-folding of a graphene-Au-SU8 flower from, d) flat into e) its 3D shape. f ) Schematic of self-folded 3D graphene-Au-SU8 arrays on-chip. g) SEM image of the self-folded graphene-Au-SU8 array. h) Optical image of the measurement set-up. i) Schematic of the angular-dependent photoresponse measurement for the chip-integrated graphene-Au-SU8 3D self-folded photodetector. j) Angular-dependent photovoltage response of a single graphene-Au-SU8 3D self-folded photodetector. (0 , 30 , and 60 ). k) COMSOL simulation of the variation of light absorption as a function of incident angles. Scale bars are b-e) 500 μm and g) 200 μm.
www.advancedsciencenews.com www.advintellsyst.com Raman Spectroscopy: The Raman spectrum shown in Figure S4, Supporting Information, was measured using a Horiba Jobin Yvon LabRAM HR800 Raman microscope (Edison, NJ) with a 532 nm excitation line. Spectra for each sample were plotted by measuring and averaging the Raman scattering of at least five different spots on each sample.
Fabrication of Graphene-Au-SU8 3D Structures: First, a 50 nm-thick Cu sacrificial layer was deposited and then patterned on either a thermal oxide-coated (SiO 2 ) silicon (Si) wafer (Cu/SiO 2 /Si) or glass slide. Then monolayer graphene was transferred on top. Cu was utilized as a sacrificial layer because it was readily dissolved in FeCl 3 without damaging the Au, graphene, or SU8 polymer in our integrated structures. Also, the 50 nm thickness was chosen based on optimization of the Cu layer etch time and microstructure release time.
The monolayer graphene was transferred from the Cu foil on which it was grown by CVD using the following method. First, PMMA was spin coated on top of the graphene on one side of the Cu foil at 3000 rpm. Afterward, the PMMA-coated Cu foil was baked, first at 85 C for 15 min and then at 135 C for 15 min. The graphene was etched on the other side of the Cu foil by O 2 plasma treatment and then the Cu foil was etched using 1 mol L À1 FeCl 3 solution. After several washes of the remaining PMMA-graphene film with deionized (DI) water, the PMMA-graphene film was transferred onto the Cu/SiO 2 /Si-coated wafer and then the PMMA film was dissolved in acetone. Then electrodes (5 nm Chromium, 40 nm Au) were fabricated on top of the graphene layer. Finally, the SU8 structures were fabricated on top of the patterned electrodes/graphene/Cu/SiO 2 /Si stack. The details of the different exposure energies used for each layer are in the results and discussion section. To actuate the patterned graphene-Au-SU8 into 3D structures, first the Cu sacrificial layer was etched with FeCl 3 solution, then the patterned graphene-Au-SU8 was conditioned in acetone. The graphene-Au-SU8 was taken out from acetone and immersed into water, to trigger self-folding into 3D structures.
For the fabrication of chip-integrated 3D graphene-Au-SU8 microstructures, a patterned Cu sacrificial layer was created as a pinning area between the 3D structures and the wafer. Details can be found in Figure S9, Supporting Information.
Photoresponse Measurement: The free-standing graphene-Au-SU8 microstructures were irradiated with a laser of a specific wavelength (488, 532, or 600 nm) (CW Solid-State Lasers, Coherent Inc.) The chip-integrated graphene-Au-SU8 microstructures were irradiated with a particular wavelength (633 nm) laser diode (Thorlabs Inc.), and the laser power and on/off intervals were digitally controlled using analog modulations. The voltage between the two electrodes was measured using a voltmeter (Keithley 2000 Multimeter) and the current between the two electrodes was set to zero so that the open-circuit photovoltage was measured. A LabVIEW software module was used to record the photovoltage in real-time.
Mechanics Modeling: A coarse-grained model was used with MATLAB to simulate the folding mechanism of both the gradient SU8 and the bilayer SU8. Details of the model can be found in Note S1, Supporting Information.
Electromagnetic Modeling: A COMSOL model was used to perform electromagnetic simulations of light absorption of the graphene rolls and graphene-Au-SU8 structures based on the finite element method. For more details, please see Note S2 and S3, Supporting Information.

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