Introduction

Graphene, a two-dimensional (2D) monolayer of sp²-hybridized carbon atoms arranged in a hexagonal lattice, has garnered significant attention due to its extraordinary electrical, mechanical, thermal, and optical properties1,2,3,4,5,6,7,8,9,10. Its high carrier mobility7,8, mechanical flexibility11,12,13, optical transparency9,10, and large surface area have enabled the emergence of novel device concepts such as transparent electrodes14,15, high-speed transistors16,17, photodetectors, and flexible optoelectronic devices18,19,20,21. Among these applications, transparent conducting electrodes (TCEs) based on monolayer graphene grown by chemical vapor deposition (CVD)22 have emerged as a compelling alternative to conventional indium tin oxide (ITO)23 due to their superior flexibility, abundance of constituent elements, and potential for large-area integration.

Compared to other TCE candidates such as metal lattices24, nanowires25, carbon nanotubes (CNTs)26,27, multilayer graphenes28, or reduced graphene oxide (rGO)29, CVD-grown monolayer graphene, which can be synthesized with high quality and can be fabricated in a large area, exhibits higher transmittance and mechanical robustness, particularly under bending stress or repeated deformation. However, integrating CVD graphene into functional microelectronic or optoelectronic devices still faces significant challenges. A major bottleneck is the patterning process of microelectrodes, which often degrades the structural and electrical integrity of the graphene.

Conventional photolithographic techniques, although widely adopted in microfabrication, are ill-suited for graphene. The multi-step process including photoresist coating, UV exposure, development, etching, and resist stripping, often induces delamination, introduces surface contaminants, or damages the carbon lattice structure, thereby deteriorating the electrical properties of the patterned graphene electrodes23,30,31. To mitigate these issues, various modified approaches have been explored in the fields of photolithography32,33,34, shearing35,36,37, print transfer38,39,40, and patterned growth41, such as the use of buffer layers30, modified photoresists42, or dry etching techniques, but these strategies either fail to fully prevent contamination or introduce additional process complexity. Furthermore, some patterning techniques, such as direct laser writing or transfer printing38,39,40, offer partial solutions but often compromise spatial resolution, scalability, or uniformity.

Recently, patterning methods based on selective growth of graphene on pre-patterned metal substrates have shown promise for fabricating graphene microstructures without lithography41. However, these methods typically result in multilayer graphene structures with limited spatial resolution and poor transferability, particularly when the feature size approaches the sub-10 µm scale. During the transfer of selectively grown graphene, structural distortions such as folding or tearing can easily occur, which hinders their practical application in high-density devices41,43. Therefore, a reliable, high-resolution, and scalable patterning method that preserves the intrinsic electrical and structural properties of monolayer graphene remains a critical unmet need.

In this study, we present a one-step free-patterning graphene (OFP-G) technique, which enables the direct fabrication of sub-5 µm resolution micro-sized patterns on CVD-grown monolayer graphene without employing photoresists, etchants, or sacrificial buffer layers. This method utilizes a controlled combination of pressure, temperature, and electric field in a high-vacuum environment to selectively convert sp² C = C and C–C bonds to oxygen-containing bonds such as C–O and C = O. This selective bond conversion at designated regions suppresses delamination and contamination while preserving the high electrical conductivity of the graphene elsewhere. We demonstrate that OFP-G achieves robust micro-sized patterning with excellent reproducibility, mechanical adhesion, and electrical performance. Raman spectroscopy, X-ray photoelectron spectroscopy (XPS), and field-emission scanning electron microscopy (FE-SEM) confirm the selective chemical modification and structural integrity of the patterned regions. Our approach offers a scalable, reproducible, and contamination-free pathway for patterning high-resolution graphene, and opens new possibilities for the integration of graphene in flexible and transparent electronics.

Experimental section

Monolayer graphene

Monolayer graphene was synthesized on a 25-µm-thick copper foil (annealed, uncoated, 99.8%, Alfa Aesar, USA) using a thermal-CVD system equipped with a quartz glass tube with an internal diameter of 10 cm and a length of 100 cm. First, the copper foil was positioned at the center of the quartz tube, which was enclosed by a heating furnace, and the temperature was increased to 1000 °C under an Ar (100 sccm, 99.99%) atmosphere. Subsequently, the copper foil was annealed with a 50 sccm flow of H2 at 1000 °C for 20 min, after which CH4:H2 (10:50 sccm, 99.99%) was introduced for 10 min to grow graphene on the copper foil at 1000 °C. Following growth process, the furnace was allowed to cool naturally to room temperature under continuous Ar flow. A detailed time–temperature diagram of the CVD process is provided in Fig. S1.

Wet transfer

The graphene layer was grown on both sides of the copper foil and then poly(methyl methacrylate) (PMMA) was spin-coated on one side as a mechanical support during transfer. The graphene layer on the other side was removed by oxygen plasma treatment. The sample was then floated on the surface of a 0.1 M ammonium persulfate (APS) solution, and the copper foil was etched to obtain the graphene/PMMA layer. Next, the floating graphene/PMMA layer was cleaned in deionized (DI) water and transferred to a silicon target substrate (525 ± 25 µm, p-type, STC.) containing a 300-nm-thick silicon oxide layer. Finally, the substrate was immersed in acetone to remove the PMMA layer. Graphene was transferred to the target substrate (Fig. 1a) and then patterned in the OFP-G method.

Fig. 1: One-step free-patterning method for graphene fabrication and characterization.
figure 1

a Schematic representation of one-step free-patterning approach for the fabrication of graphene. b Macroscale graphene pattern fabricated using the proposed method. Cross-sectional images showing (c) the sample composition and (d) the interfacial interactions at the bonding interface. e Process for fabricating microscale graphene patterns, accompanied by a schematic of the channel structure on the glass substrate. f OFP-G channel with corresponding top and cross-sectional views

Full size image

One-step free patterning of graphene method

The OFP-G method developed in this study employs an advanced anodic bonding process to achieve selective graphene patterning through controlled C-C-, C = C to C-O bond conversion. As illustrated in Fig. 1a and S2, the process utilizes a Suss Micro Technology SE8e bonding system where a Na2O/K2O-rich Pyrex glass substrate (Borofloat 33, 500 ± 20 μm) serves as the anode and a graphene-transferred silicon substrate acts as the cathode. After transferring monolayer graphene grown on Cu foil onto a SiO2 substrate, a pre-etched glass substrate, corresponding to the area intended for bonding, is carefully positioned on top. Mechanical pressing is applied over the contact region, while an electrostatic force acts across the substrate. These forces are applied simultaneously under moderate thermal heating in a low-vacuum environment. The conditions are maintained for 15 min. After the temperature decreases to 90°C under ambient pressure, the bonded sample is retrieved from the apparatus. Under controlled conditions of 50 mTorr vacuum and 380 °C temperature like Table S1, the glass substrate enters a conductive solid electrolyte state where alkali oxides decompose into mobile Na⁺/K⁺ ions and reactive O²⁻ species. When 1,000 V potential is applied across the 12.5 N/cm² pressed interface, the migration of alkali ions toward the cathode creates oxygen-rich regions that facilitate localized conversion of sp²-hybridized carbon bonds to C-O configurations.

The macroscale pattern produced by the OFP-G method divides the graphene area into bonded and unbonded regions, as illustrated in Fig. 1b, which also presents a representative sample and its fabrication process. For macroscale patterns, computer-numerical-controlled laser machining (HR-1260-4sp system) fabricates glass patterns complementary to desired geometries as shown in Fig. 1b. The bonded interface morphology is shown in Fig. 1c, while atomic-scale visualization of the interfacial bonding mechanism in Fig. 1d demonstrates the selective oxidation process that preserves conductivity in non-bonded regions.

Microscale channels down to 5 μm width are achieved through the hybrid approach combining 300 μm sandblasted through-holes with reactive ion etched (CHF3/Ar) interconnects of 1 μm depth as shown in Fig. 1e. The completed structures are presented in Fig. 1f, demonstrating the method’s capability to produce both the overall device architecture and fine channel features through this anodic bonding approach.

A key advantage of the OFP-G method over conventional photolithographic etching techniques is its ability to avoid graphene contamination and structural damage. As illustrated in Fig. S3, the conventional approach involves coating graphene with a photoresist (PR), patterning via development, and etching away unwanted graphene using O2 plasma, followed by PR removal process. However, residual PR often remains on the graphene surface, leading to contamination that degrades electrical performance. Additionally, during PR removal, small graphene area such as especially channels as narrow as 5 μm is delaminated or damaged, compromising the integrity of the micro-sized pattern structures. The OFP-G method circumvents these issues by employing direct bonding without the need for resist-based processing, ensuring cleaner interfaces and more reliable pattern fabrication, as evidenced in Fig. 1c, f.

Computational methods

Our first-principles calculations were performed by using a density functional based tight binding (DFTB) quantum simulation method based on the density functional theory (DFT)44 two-center approach in the Materials Studio 2020 package of BIOVIA Ltd. The first step is to create a sawtooth graphene nanoribbon containing 1 × 5 graphene cells. In addition, the lattice constant is increased to avoid interaction between periodic mirrors. A DFTB+ geometry optimization task was used to optimize the geometric configuration of the nanoribbons, where the Slater-Koster library was mio and smearing was 0.01 Ha. The optimized nanoribbon structures were used to model ideal graphene nanoribbons and graphene nanoribbons containing C-O bonds with 1 × 19 graphene cells. Electrodes were added at both ends of the structure as source and drain. The behavior of the electron transport was calculated by DFT in combination with the NEGF form implemented in the SIESTA package. As part of the calculations, each electrode will be created as a half-cycle structure in order to calculate the charge, potential and Fermi energy levels. Natural defects such as vacancies and interstitials were not considered in this study.

Results and discussion

First principles computational modeling

We performed comprehensive molecular dynamics simulations to investigate the interfacial bonding characteristics and electron transport properties of graphene nanoribbons (GNs) using Materials Studio software. Molecular simulations have shown great potential in dealing with interfacial interactions and revealing microscopic mechanisms41. We constructed simulation models of ideal GNs and GNs containing C-O bonds with a 50 Å vacuum layer to eliminate periodic boundary effects. The geometrically optimized models (top and side views) are shown in Figs. 2(a–d), S4. We included several key configurations. The two-dimensional supercell size of the GN is 7.178 Å × 55.38 Å. The ideal GN contains 200 carbon atoms (Fig. 2a); OFP-GN contains 160 carbon atoms and 40 C-O bonds (Fig. 2b); GN containing 1 column of C-O bonds consists of 190 carbon atoms and 10 C-O bonds (Fig. 2c), and in addition, GN with 3 and 5 columns of C-O bonds have 170, 150 carbon atoms and 30, 50 C-O bonds.

Fig. 2: Molecular dynamics simulation model and electron transport properties.
figure 2

a Ideal GN, b OFP-GN, c single-column C-O bond on GN, d three-column C-O bonds on GN, e five-column C-O bonds on GN at the break down region, respectively, f Electronic transport characteristics

Full size image

The analysis of the electrostatic potential provides further insights into the electron transport properties of these structural models. A cross-sectional slice along the X and Y planes of the structures was used to generate planar electrostatic potential diagrams, as illustrated in Fig. 2. In these diagrams, blue represents negative potential regions, while red corresponds to positive potentials. Examination of the electronic band structure within the energy range of (−2 eV, 2 eV) revealed that bands along the width of the ribbon remain predominantly flat. In contrast, beyond this energy range, the bands become increasingly dispersed. This indicates that electron density within this low-energy window remains localized along the ribbon width, whereas at higher energies, it becomes more delocalized. Furthermore, our findings confirm that, within the low-energy regime, electrons are effectively confined within the GN, as previously observed in the similar study42.

The pristine GN consists exclusively of carbon atoms, exhibiting a characteristic electrostatic potential distribution with positive values at the center and negative potentials at the edges, reaching approximately −1.108 eV. The transport coefficient curve of the pristine GN follows a step-like profile, where the transport values directly correspond to the number of available energy bands in the nanoribbon’s electronic structure as depicted by the black line in Fig. S4. Specifically, the maximum and minimum transport values are approximately 7.0 and 1.0, respectively, with an average value of around 3.5. In contrast, the OFP-GN proposed in this study, illustrated in Fig. 2b, incorporates C-O bonds selectively positioned along the top and bottom edges to modulate electron transport pathways. Electrostatic potential analysis reveals a marked deviation from the pristine GN, with oxygen atoms at the edges exhibiting significantly enhanced negative potentials, reaching up to −3.442 eV. Meanwhile, the central regions maintain electrostatic potentials comparable to those of the pristine GN, around 0.853 eV, suggesting that the overall electronic characteristics remain largely preserved. Furthermore, the electron transport coefficient curves (Fig. S4) demonstrate that OFP-GN retains the step-like structure observed in pristine GNs, albeit with a slightly expanded transport range. The maximum and minimum transport values increase to approximately 8.9 and 1.0, respectively. Notably, the electron transport properties of the graphene channel formed by C-O bond incorporation remain effectively intact, exhibiting a modest enhancement relative to the pristine GN.

More interestingly, the simulations demonstrated that C-O bond incorporation at mid-ribbon positions created localized potential perturbations that progressively degraded electron transport as bond density increased like Fig. 2 (c–e). The electron potential results show that the presence of C-O bonds directly causes a change in the potential, with positive potentials of maximum values of about 4.940 eV, 11.400 eV, and 11.920 eV at the positions where C-O bonds are present, respectively, while the potentials at other sites are close to those of the ideal GN and OFP-GN. The results show that the effect of C-O bonding on electron transport on GN is only in the region where it exists and does not change the electron transport properties of carbon atoms in other regions. We measured average transport coefficients of 0.7, 0.7, and 0.5 for structures containing 1, 3, and 5 columns of C-O bonds, respectively (Fig. 2f). Importantly, these effects remained highly localized, with minimal impact on electron transport properties in regions distal to the bonding sites. In other words, this indicates that the presence of C-O bonds at intermediate positions leads to instability in electron transport and, as the number of C-O bonds increases, further suppresses electron transmission. In essence, introducing C-O bonds in the mid-ribbon region disrupts electron flow, effectively mimicking the patterning of graphene into disconnected segments. This suggests that C-O bonding can be strategically utilized to engineer novel pattern geometries with precisely controlled electronic properties. These findings provide a new perspective on the underlying mechanism of the OFP method and serve as a guideline for subsequent experimental designs.

Surface characterization of OFP-G

Raman spectroscopy has been widely used to characterize the key properties of graphene such as doping levels, structural defects, and the layer number. In this study, Raman spectra were acquired using a confocal Raman microscope (NRS-3100, JASCO, Japan) with a 532 nm excitation laser. Figure 3a presents the Raman spectra of graphene samples obtained before and after the bonding process. Raman mapping was performed over a 45 µm × 45 µm area on both the wet transferred graphene before bonding and the un-bonded area as a pattern region after bonding process. The spatial distributions of the intensity ratios I(D)/I(G) and I(2D)/I(G) were analyzed to evaluate defect density and layer number, respectively43.

Fig. 3: Raman analysis of graphene patterns before and after anodic bonding.
figure 3

a Raman mapping of I(D)/I(G) and I(2D)/I(G) ratios over a 45 µm × 45 µm area of the graphene pattern before and after the anodic bonding process at non-bonded region. Histograms of positions and FWHM of (b) G peak and (c) 2D peak, comparing the graphene surface before and after bonding, respectively at non-bonded area. The red and green dashed lines in the histograms indicate the average peak positions. d Raman spectra and (e) correlation intensity ratios between I(D)/I(G) and I(D’)/I(G) at different graphene regions of before bonding, non-bonded area, and bonded area. All Raman measurements were conducted under consistent environmental conditions

Full size image

As shown in Fig. 3a, prior to bonding process, I(D)/I(G) ranged from 0.004 to 0.19, and I(2D)/I(G) ranged from 1.51 to 3.51. After bonding process at the un-bonded area, these values were 0.06–0.16 and 0.75–2.72, respectively, indicating that the graphene structure remained largely intact post-bonding. The I(D)/I(G) values decreased and converged due to thermal annealing effects during bonding, which may have facilitated defect healing.

Further analysis of the G and 2D peak positions and their full width at half maximum (FWHM) was conducted to assess doping and strain states in the graphene films as shown in Fig. 3b, c45,46,47,48. The average G and 2D peak positions shifted from 1587 cm⁻¹ and 2679 cm⁻¹ before bonding to 1595 cm⁻¹ and 2685 cm⁻¹ after bonding, respectively. This blue shift indicates p-type doping, likely caused by adsorbate-induced charge transfer, which dominates over compressive strain effects induced during transfer and bonding. Moreover, the FWHM of the G peak increased by 16 cm⁻¹ after bonding, indicating a reduction in charge doping, consistent with desorption of adsorbates from the graphene surface.

Raman spectra were also collected from the non-pattern area (bonded area) before and after bonding process like Fig. 3d, e. The comparison includes spectra from as-transferred graphene (black), pattern area after bonding (red), bonded graphene area on the Si substrate (blue) and glass substrate (green), respectively. While high-quality graphene properties were preserved in the pattern region as un-bonded area even after bonding process, prominent D peaks at 1390 and 1376 cm⁻¹ and well-defined D’ peaks at 1622 and 1625 cm⁻¹ emerged in the Si and glass-contact regions. The significant intensity of the D peak reflects a pronounced increase in defect density, while the blue-shifted G and D’ peaks suggest strong p-type doping, likely originating from trapped charges at the graphene–oxide interface49,50.

These observations align with previously reported Raman features of graphene oxide (GO), where blue-shifted G and D’ peaks are attributed to the presence of isolated C = C double bonds and the formation of C–O bonds within the oxidized lattice51,52,53. The emergence of the D’ peak and G band shift in the bonded substrate-contact regions supports the partial oxidation of graphene, likely due to chemical interactions with underlying SiO2 or glass substrates. Briefly, graphene bonds changed from C = C bonds to C–C and C–O bonds.

Quantitative analysis of I(D)/I(G) and I(D′)/I(G) intensity ratios (Fig. 3e) reveals that the I(D)/I(D′) ratio increased from 1.18 before bonding to 1.9 and 2.1 at bonded area in Si and glass contact regions, after bonding process, respectively, which is indicative of increased disorder and bond energy alterations consistent with GO formation50. Furthermore, the blue shift of the 2D peak corroborates the presence of p-type doping in the bonded regions.

To investigate changes in surface properties induced by the bonding process, we conducted water contact angle measurements (Phoenix 10, Surface Electro Optics (SEO), Korea) and X-ray photoelectron spectroscopy (XPS, MultiLab 2000, Thermo, USA) at various regions before and after bonding, as shown in Fig. 4. The water contact angle on the graphene surface within the pattern region as un-bonded area exhibited only a slight decrease from 89.55° to 86.72° post-bonding, suggesting that the surface characteristics of graphene were largely preserved during the OFP-G processing. This observation is consistent with the Raman spectroscopy results.

Fig. 4: Surface characterization of graphene patterns before and after anodic bonding.
figure 4

a Water contact angle measurements, b XPS analysis: (i) before bonding, (ii) after bonding (non-bonded area), (iii) silicon side of the bonded area, and (iv) glass side of the bonded area. c Histogram showing variations in the water contact angle and the C-O/HOPG peak area ratios

Full size image

In contrast, the contact angles on the Si and glass sides of the non-pattern regions as bonded area decreased significantly to 67.04° and 68.46°, respectively as shown in Fig. 4a, c, indicating an increase in surface energy following the bonding process. Complementary XPS analysis was performed at the same locations like Fig. 4b, c. The XPS spectra revealed two prominent peaks at 284.5 eV and 286.1 eV, corresponding to C–C and C–O bonding states in graphene, respectively. Notably, the C–O peak intensity at 286.1 eV was substantially higher on the Si/glass side in the non-pattern region as bonded area compared to the pattern region as un-bonded area, indicating the formation of oxygen-containing functional groups post-bonding. Assuming a surface atomic density of graphene of 3.82 × 10¹⁵ cm⁻²54, the C–O/C–C peak ratios for the Si and glass substrates in the bonded non-pattern areas were determined to be 0.73 and 0.74, respectively. These ratios correspond to C–O bond densities of 2.79 × 10¹⁵ cm⁻² and 2.83 × 10¹⁵ cm⁻², respectively. Collectively, these findings confirm that the graphene in the pattern region remained chemically intact post-bonding, while in the bonded non-pattern regions, chemical modification occurred via the formation of C–O bonds at the graphene–substrate interface. These results are in good agreement with the Raman and contact angle analyses.

Demonstration and analysis of OFP-G

Macroscale patterns using the OFP-G method

Using the OFP-G method, we fabricated three distinct types of graphene patterns with varied pattern geometries, as illustrated in Fig. 5. The electrical properties of these patterns were characterized via two-point probe measurements using a Keithley 2000 source meter (Tektronix, Inc., Korea). As shown in Fig. 5a and b, the first configuration was designed to evaluate electrical isolation across the bonded interface. Prior to bonding, the average resistance across the graphene pattern was measured to be 2.712 kΩ. Following bonding, the resistance reached an unmeasurable level (overflow), indicating complete electrical insulation across the bonded region. This result confirms that the bonding process effectively prevents electron conduction through the designated isolation zone.

Fig. 5: Schematic illustration and optical images of graphene patterns fabricated by the OFP-G method.
figure 5

a, b Part 1: Electrical breakdown test before and after bonding. c, d Part 2: Graphene patterns with varying widths (2.5, 5.0, and 7.5 mm) at a fixed length of 25 mm. e, f Part 3: Graphene patterns with varying lengths (25, 45, and 65 mm) at a constant width of 5 mm. Insets depict LED lamp brightness variations corresponding to different resistance of the graphene patterns, respectively

Full size image

To investigate the influence of patterns geometry on resistance, graphene patterns with identical lengths of 25 mm but varying widths such as 2.5, 5.0, and 7.5 mm were fabricated like Fig. 5c, d. The measured resistances were 12.6, 8.5, and 4.4 kΩ, respectively, demonstrating an inverse relationship between pattern width and resistance. Similarly, a second set of patterns with constant width of 5.0 mm and varying lengths with 25, 45, and 65 mm exhibited increasing resistances of 8.5, 15.1, and 23.2 kΩ, respectively as shown in Fig. 5e, f, in agreement with the theoretical direct proportionality between resistance and pattern length.

To provide a visual representation of the electrical behavior, a constant voltage was applied across each graphene pattern, and resistance variations were qualitatively assessed by monitoring the brightness of a connected LED lamp like Fig. 5b, d, f and videos S1, S2, and S3. As shown in the inset of Fig. 5b, a distinct decrease in brightness—and complete extinction of the LED—was observed following electrical breakdown, which corresponds to the abrupt increase in resistance due to interruption of electron flow across the bonded interface. In contrast, in Fig. 5d and f, the LED brightness increased progressively as the pattern width increased and length decreased, respectively, consistent with the expected decrease in resistance. These visual demonstrations corroborate the quantitative resistance measurements and effectively illustrate the relationship between pattern geometry and electrical performance. Collectively, these results validate the effectiveness of the OFP-G method in achieving spatially controlled patterning of graphene. The ability to fabricate electrically isolated or connected regions with high fidelity demonstrates the method’s potential for the scalable fabrication of functional graphene-based circuits and microsystems.

Microscale patterns using the OFP-G method

The versatility and applicability of the proposed OFP-G method were demonstrated by fabricating microscale graphene patterns with widths of 5 and 10 µm, as illustrated in the top-view images (Fig. 6a). As described earlier in Section 2.1, these microscale graphene channels were fabricated using the newly proposed OFP method. The fabricated structures were characterized by field-emission scanning electron microscopy (FE-SEM, JEM-2100F(HR), JEOL Ltd., Japan) to investigate the side view of the OFP-G channel, with the cross-sectional images presented in Fig. 6a.

Fig. 6: Comparison of graphene patterns fabricated by OFP and general etching methods.
figure 6

Top view and cross-sectional SEM images of the graphene patterns with 5 μm and 20 μm width fabricated by a suggested OFP method and b general etching method. The red dashed inset shows a magnified view of the graphene channel structure. Optical microscopy images (b-2, b-3) show the top-view of the channel and pattern configuration with polymer residues. c Raman mapping results of general etching method for Si-, G-, and 2D-peaks over the graphene channel area (225 μm × 225 μm, scale bar: 50 μm). d I–V characteristics and e histogram of resistance distribution of OFP-G channels, respectively. f I-V curves for graphene channels fabricated using the general etching method, showing negligible conduction and insets display magnified current scales to highlight minimal variation

Full size image

Two through-holes were fabricated on the glass substrate side using the sandblasting process, followed by reactive ion etching (RIE) to selectively etch the microchannel region, as described in Section “Monolayer graphene”. This approach enabled precise definition of the channel structure, which was crucial for the subsequent formation of the OFP-G patterns. After fabricating the graphene patterns via the OFP process, nickel was deposited through the holes on the glass substrate to enable electrical signal measurements. The inset in Fig. 6a (highlighted with a red box) shows the OFP-G channel formed between the two through-holes, where the channel structure was maintained without collapse under electrostatic forces even after the OFP-G process.

For comparison, microscale patterns based on monolayer graphene with same dimensions were fabricated using conventional etching techniques, as shown in Figs. S2, S4. In this method, a 170 nm-thick SiO2 layer was etched by deep RIE, and the photoresist (PR) layer was subsequently removed to define the microchannels like Figs. 6b-2, b-3 and S2. Optical microscopy (BX53M, Olympus, Japan) and SEM were used to capture top and cross-section images and assess pattern fidelity. Irregularly shaped patterns were consistently observed across more than eight fabricated samples for each microchannel size, as shown in Figs. 6b-2, 6b-3, and S4, indicating non-uniform channel formation regardless of the intended dimensions. Additionally, we observed the presence of PR residues on the top of the non-etched regions and irregular microchannels by cross-sectional FE-SEM images.

To confirm the presence of graphene, Raman mapping was performed on the graphene microchannel (Fig. 6b-3) fabricated by conventional methods. The resulting Raman spectra (Fig. 6c) exhibited a strong silicon peak at 520.1 cm⁻¹, but neither the G peak at 1580 cm⁻¹ nor the 2D peak at 2679 cm⁻¹, indicating that the graphene layer was removed because tearing phenomenon while the etching and PR removal processes. These observations are consistent with the findings in Fig. S5, further confirming the limitations of conventional etching in fabricating patterning of graphene microchannels. In the case of graphene patterns fabricated using the OFP method, Raman mapping could not be directly observed because the graphene was bonded between the glass substrate and the SiO2 substrate, and the 500-µm-thick glass substrate prevented direct optical access to the graphene layer.

Therefore, I–V curve measurements were conducted to analyze the electrical characteristics of microscale graphene channel fabricated by both the OFP and the conventional method. Following the fabrication of through-holes by the sand blast, the entire electrode metal was deposited onto the surface as shown in Fig. 6a. A 150 nm-thick nickel layer was deposited using an e-beam evaporator (KVE T-C500200, Korea Vacuum, Korea)55. Electrical properties of the microscale graphene patterns were characterized using a semi-automatic probe station (SUMMIT 11742B/4156 C, Cascade/Keysight, USA). The I–V characteristics of the fabricated patterns are shown in Figs. 6d and e. The patterns fabricated using the OFP-G method exhibited highly linear and symmetric I–V curves, indicating excellent ohmic contact between the graphene channel and the Ni electrodes. The near-linear behavior across the entire voltage range confirms minimal Schottky barrier formation, while the symmetry between positive and negative bias suggests negligible contact asymmetry or defect-induced rectification. Moreover, the patterns displayed notable conductivity, with significant current levels even at low bias, further demonstrating the high quality of the electrical connection. Size-dependent variations were also observed: the 5 µm-wide patterns exhibited a resistance of 11.5 ± 2.8 Ω, while the 20 µm-wide patterns showed a lower resistance of 9.4 ± 0.4 Ω. The resistance distributions for different pattern sizes are summarized in Fig. 6e. The purple bars represent the distribution of the 5 µm patterns, with the estimated peak indicated by a blue line, while the green bars correspond to the 20 µm patterns, with the estimated peak marked by a red line. The resistance distribution of the 20 µm patterns was relatively narrow, whereas the distribution for the 5 µm patterns was broader, indicating lower reliability for the smaller patterns.

In contrast, patterns fabricated using the conventional method exhibited minimal current variation over the applied voltage range, as shown in Fig. 6f. Specifically, as depicted in the inset graphs, the measured current values remained extremely low, near the noise level with approximately ± 5E-13 A, regardless of the applied voltage. This result indicates that the graphene channels fabricated by the conventional method failed to form continuous electrical pathways, leading to an absence of measurable conductivity. Thus, unlike the OFP-G method, the conventional fabrication approach could not produce electrically functional microscale graphene patterns. In addition, we conducted a comparative analysis of the graphene patterning method presented in this work against recent cutting-edge research results from multiple perspectives as shown in Table S2. This comparison objectively highlights the advantages of the OFP-G method, including its one-step, residue-free process, excellent conductivity retention, capability for large-area electrode fabrication, high-resolution patterning, and relatively low cost, while also transparently acknowledging potential limitations, such as specific equipment requirements. Overall, while the conventional fabrication method struggled to produce electrically functional microscale graphene channel, the OFP-G method offers outstanding potential and reliability for the fabrication of microscale monolayer graphene channel.

Conclusions

In this study, we proposed a novel and scalable approach, termed OFP-G, for patterning both macroscale and microscale monolayer graphene pattern. By selectively converting C = C bonds to C–O bonds only in the bonded regions, the method enables precise patterning without introducing additional substances such as PR and PMMA to the non-bonded pattern areas, thereby preserving the intrinsic electrical properties of graphene. The high-temperature conditions during the OFP process also assist in removing transfer-related residues, resulting in clean and high-quality graphene patterns. The fabricated patterns demonstrated reliable, size-dependent electrical performance consistent with theoretical expectations. Furthermore, first-principles calculations confirmed that the formation of localized C–O bonds modulates the electrical characteristics of graphene, providing a theoretical basis for the observed behavior. Altogether, these results validate the OFP-G method as a reliable strategy for fabricating high-performance monolayer graphene patterns, offering strong potential for future applications in graphene-based microelectronics and MEMS devices.

Supporting information

CVD-grown graphene (Fig. S1); OFP-G method (Fig. S2 and Table S1); General etching method fabrication process (Fig. S3); First principles simulation model and electron transport properties (Fig. S4); Image result of graphene channel by general etching method (Figure S5); Recent cutting-edge studies for horizontal comparison (Table S2) (PDF); Video S1: Break down of electron migration pathways by OFP-G method (AVI); Video S2: Different width of electron migration pathways by OFP-G method (AVI); Video S3: Different length of electron migration pathways by OFP-G method (AVI).