Precise CO2 Reduction for Bilayer Graphene.

It is of great significance to explore unique and diverse chemical pathways to convert CO2 into high-value-added products. Bilayer graphene (BLG), with a tunable twist angle and band structure, holds tremendous promise in both fundamental physics and next-generation high-performance devices. However, the π-conjugation and precise two-atom thickness are hindering the selective pathway, through an uncontrolled CO2 reduction and perplexing growth mechanism. Here, we developed a chemical vapor deposition method to catalytically convert CO2 into a high-quality BLG single crystal with a room temperature mobility of 2346 cm2 V-1 s-1. In a finely controlled growth window, the CO2 molecule works as both the carbon source and the oxygen etchant, helping to precisely define the BLG nucleus and set a record growth rate of 300 μm h-1.

Introduction

The CO2 reduction reaction (CO2RR) is involved in hydro-deoxygenation (HDO),[1] C–C coupling,[2] and aromatization processes.[3] This reaction has received much attention for its significance in recycling CO2 into value-added chemical feedstock and exploring unique reaction pathways for transforming and utilizing CO2. Armed with state-of-the-art electrochemistry, photochemistry, and thermocatalysis, CO2 can be converted into C1 (methanol, formic acid),[4−7] C2 (ethylene, ethanol),[8−11] and C2+ products (propanol, aromatics).[12−14] Therein, graphene is a kind of highly valuable product, because of its special π-conjugated structure with sp2-hybridized C and intriguing physical properties.[15,16] Bilayer graphene (BLG), with a tunable band gap and bizarre phenomena at a magic stacking angle, holds tremendous promise for next-generation electronic and twistronics devices under 3 nm.[17−20] However, the selective pathway from CO2 to a BLG single crystal has never been activated, due to the difficulty in precisely splitting the inert C=O bond and the perplexing growth mechanism. In addition, in the conventional chemical vapor deposition (CVD) system, once the first graphene layer fully covers the copper surface, nucleation of the second layer becomes restricted, significantly suppressing the BLG’s growth. Reducing CO2 molecules into a high-quality BLG single crystal would broaden the understanding of CO2 reduction and simultaneously harvest a precious electronic material.

With both carbon (C) and oxygen (O) atoms, carbon dioxide (CO2) is a greener and safer C source in graphene growth compared with conventional CH4. After stripping O from CO2 in a controllable manner, the active C fragments can participate in building π-conjugated graphene. On the other hand, the residual O atom has been widely reported to be beneficial to BLG growth.[21−23] In previous studies, using a tandem catalytic CVD method, single-layer graphene (SLG) with a single crystal structure can be synthesized from CO2 feedstock.[24] Our group also demonstrated a tandem electrochemical-CVD system to efficiently convert CO2 into an SLG film,[25] providing a way for tunable CO2 transformation. Nevertheless, there is no BLG reported in these works, due to inappropriate oxygen content caused by uncontrolled CO2 reduction. A moderate amount of O can remove the undesirable carbon and open the carbon diffusion channel for BLG nucleation. By introducing O-rich copper[21,22] and oxygen-containing molecules such as ethanol,[23] both the quality and growth rate of BLG were improved.

Here, we demonstrate a catalytic strategy in CVD systems to precisely reduce CO2 into BLG single crystals by finely tuning the growth window. With CO2 providing both carbon as the BLG growth feedstock and oxygen as an etching reagent, we can realize the growth of large BLG single crystals with unprecedented swift kinetics. We also measure the electrical transport properties based on BLG single crystals; the mobility can reach 2346 cm2 V–1 s–1 at room temperature.

Results and Discussion

The BLG single crystal is grown by a CVD method in a home-built tube furnace equipped with dual heating zones as shown in Figure . Given that the CO2 molecule is too inert by itself to be converted into graphene, 1.5 sccm CO2 is preliminarily activated with 200 sccm H2 to produce more reactive carbon species in the first heating zone. This zone is equipped with commercial catalyst Ni/Al2O3 and operated at an optimal temperature of 300 °C. Components of the gas mixture are quantitatively analyzed in a following tail gas chromatography (GC) setup. The activated feed gas further flows into the second heating zone with a temperature of 1055 °C and is converted into a BLG single crystal on a polished copper foil substrate.

Figure 1

Schematic illustration of the BLG single crystal growth process.

The as-prepared BLG single crystals can be directly observed under optical microscopy (Figure a). Specifically, many hexagonal crystals were stacked on top of a uniform SLG film. Their thickness and stacking mode were determined using Raman spectroscopy according to their full width at half maximum (FWHM) of the 2D peak and the intensity ratio between the 2D peak and G peak (I2D/IG). For a typical hexagonal graphene crystal, the Raman mapping of FWHM and I2D/IG at different regions shows even intensity, indicating that the whole hexagonal crystal has a uniform thickness and stacking mode (Figure b,c). Corresponding values of FWHM of the 2D peak and I2D/IG are ∼53–54 cm–1 and 0.7–0.8, respectively, consistent with characteristics of AB-stacked BLG (Figure d),[26,27] while in areas not covered by the hexagonal crystals, the counterparts are ∼30–35 cm–1 and 1.2–1.5, respectively, manifesting an SLG signature. It is worth noting that although the graphene is grown in an etching CO2 atmosphere, we did not observe any D peaks in both the single-layer and bilayer region. A small amount of CO2 was reported to selectively remove amorphous carbon without compromising the lattice of graphene.[28] BLG is also known to display a higher etching resistance and a slower etching rate than SLG.[29] Thus, we did not observe any obvious defect in our BLG samples. Besides, the ID/IG mapping of a BLG single crystal shows minimum variation, with a ratio close to 0 (Figure S1), suggesting an intact atomic arrangement with negligible defects.[30] The BLG single crystals produced from CO2 display an average lateral size of ∼200 μm, with an optimized value of ∼220 μm (Figure e and Figure S2).

Figure 2

Structure of the BLG single crystal. (a) Optical image of a typical AB-stacked BLG single crystal on a 300 nm SiO2/Si substrate (scale bar: 50 μm). Raman mapping of (b) the FWHM of the 2D peak and (c) the I2D/IG of the BLG single crystal in panel a (scale bar: 50 μm). (d) Raman spectra measured at Area 1 and Area 2 in panel a. (e) Size distribution of the as-prepared BLG single crystals. (f) HRTEM image of the AB-stacked BLG’s edge (scale bar: 5 nm). (g) AB-stacked BLG lattice extracted from the white dash-line box in panel f after the inverse Fourier function transformation (IFFT) treatment (scale bar: 1 nm). (h) SAED pattern of a typical AB-stacked BLG and the corresponding intensity profile of the diffraction spots along (2110), (1100), (0110), and (1120).

Detailed structure information on a BLG single crystal was further studied by transmission electron microscopy (TEM). The edge of the graphene under TEM observation displays a clear bilayer structure (Figure f). The crystal lattice adjacent to the edge exhibits a periodical hexagonal symmetry, whose lattice pattern is in good agreement with that of graphene (Figure g). A high crystal quality in most regions of BLG is evidenced by an orderly lattice arrangement with nearly no defects, correlating with the D peak absence in Raman spectra. Stacking mode and crystal orientation of the BLG can be clearly measured by selected area electron diffraction (SAED). In ∼73% of the BLG regions, the intensities of the diffraction spots at (1100) and (0110) facets are about half of those at (2110) and (1120) facets, strongly supporting the AB-stacked BLG structure (Figure h and Figure S3).[31] We also observed ∼27% BLGs with 30° twisted angles (Figure S4), a ratio in agreement with the Raman spectra. Specifically, we analyzed Raman signals of 100 randomly selected single crystals. AB-stacked BLG is the dominated bilayer product in our system, accounting for ∼72% of the total number of the BLG single crystals based on Raman spectra statistics, while the remaining ∼28% are mostly 30°-twisted BLG single crystals. We can also observe sharp edges of the BLG domains through optical microscopic images.[32] The AB-stacked BLG domains account for ∼75%, and the rest are mostly 30° twisted angles (Figures S5 and S6). It had been reported that the orientation of the second graphene layer can be affected by the environment of graphene growth, such as the variation in gas flow and distinct interface energy between the adlayer graphene and graphene/Cu steps.[32,33] According to theoretical calculations, AB-stacked and 30° are the top two stable structures in twisted BLG,[33] which is consistent with our experimental observations in Figures S3, S5, and S6.

In addition, SAED results at several different spots spreading across a large area (20 × 30 μm2) show that all of the diffraction patterns can be overlapped very well, indicating a single crystal structure nature of the BLG with identical crystalline orientations (Figure S7).

To understand the kinetic process of BLG growth from CO2, we monitored the graphene growth with different time duration under the same condition (Figure a–d, Figure S8). We found that the adlayer graphene formed on the hexagonal SLG islands in the first 10 min (Figure S5). Within 40 min of growth, the SLG islands merged into a fully covered SLG film, while the adlayer graphene gradually grew into a larger single crystal with a size up to 220 μm. By measuring the domain size with the growth time, the rate of the first layer growth was calculated to be 1200 μm h–1 (Figure e). Although slower than the first layer, the growth rate of the adlayer still reaches 300 μm h–1 (Figure f). To the best of our knowledge, it is the highest growth rate of BLG single crystals among reported works (Figure S9 and Table S1).[21,27,34−39] As the reaction continued, the BLG single crystals gradually transformed into multilayer graphene (MLG; Figure S10). At 20 min, the number ratio of BLG’s nucleus to MLG’s nucleus (NBLG/NMLG) reaches 50; the amount of BLG nuclei can still be dominant at 40 min with NBLG/NMLG also being 8 (Figure S11). After 40 min, the density of MLG nuclei kept increasing, while both the density of the BLG’s nuclei and the NBLG/NMLG continuously declined (Figure g). At 40 min, the size and the density of the BLG single crystal reach the optimal values. By measuring the area of BLG single crystals, the coverage of BLG single crystals on the surface of SLG reaches ∼65% after 40 min of growth (Figure S12).

Figure 3

Kinetics of BLG single crystal growth. (a–d) SEM images of graphene on a 300 nm SiO2/Si substrate at 5, 10, 20, and 40 min, respectively (scale bar: 100 μm). (e) Growth rate of SLG at different reaction times. (f) Growth rate of BLG at different reaction times. (g) Density of BLG and MLG nuclei at different growth times.

The catalytic transformation of CO2 into graphene in our system involves many chemical processes such as deoxidation and hydrogenation.[1] In order to further reveal the mechanism of BLG growth, we analyzed the tail gas with GC equipped with a hydrogen flame ionization detector (FID) and thermal conductance detector (TCD). A typical tail gas indicated H2 (97.9%), CH4 (0.6%), CO (0.08%), CO2 (0.04%), and H2O (1.3%) in our growth system (Figure S13). Among these gases, CH4 is the most common carbon source broadly reported in graphene growth.[21,27,34−36] No graphene was grown, and no CH4 was detected in the tail gas when we fed CO2 and H2 into a tube furnace without Ni/Al2O3 activation (Figures S14 and S15). Therefore, CH4 in situ formed from CO2 hydrogenation is most likely the effective carbon source in our experiment as well. In addition, in the case of only CH4 and H2 (1.5:200) fed into our system, only the SLG film was found with no BLG nucleus (Figure S16), indicating that the oxygen atoms originating from the inlet CO2 are also necessary for the BLG growth (the CO2 hydrogenation mechanism can be found in the Materials and Methods section). To ensure that there is no oxygen introduced to the copper foil, Cu foil could also be annealed in a H2 atmosphere instead of O2. Under this condition, BLG single crystals can still be grown as shown in Figure S17, indicating that the possible oxygen in copper foil does not play a significant role in BLG growth.

In our experiment, we observed an enhanced D peak in the Raman spectrum after treating the SLG film with pure CO2 at the growth conditions (Figure S18). This indicates an increase in the number of defects. While SLG was annealed in a pure H2 atmosphere, its Raman spectrum changed little (Figure S19), suggesting a much weaker etching capability of H2 compared to CO2. Meanwhile, it has been reported that H2 tends to etch the edge of graphene,[40] while CO2 can attack both the edge and surface.[41] Given that all of the nuclei of adlayer are located within the first graphene layer, we believe that BLG’s nucleation originates from CO2 etching rather than H2 etching.

During the growth, we found that the growth windows for BLG and MLG are quite close to each other. In order to optimize the system growth pressure (P) for the favored BLG growth, we kept the CO2/H2 ratio at 1.5:200 and studied the BLG growth under different pressures ranging from 800 to 1600 Pa (Figure a–d, Figures S20 and S21). We found that the optimal P is ∼1100–1200 Pa, under which conditions the BLG coverage reached its highest value while the nucleation density of MLG was still relatively low (Figure e). A similar optimal P can be derived under a series of BLG growth experiments performed under different CO2/H2 ratios (1:200 and 2:200) (Figures S22–S25).

Figure 4

Growth mechanism of the BLG single crystal. (a–d) SEM images of graphene on a 300 nm SiO2/Si substrate under 800, 1000, 1200, and 1500 Pa, respectively (scale bar: 100 μm). (e) Density of the BLG and MLG nuclei under different growth pressures. (f) Relative ratio of [O]/[CH4] under different growth pressures. (g, h) Densities of MLG and BLG nuclei under different ratios of [O]/[CH4]. (i) Scheme of the BLG single crystal growth mechanism.

Both etching (O atoms) and growing (CH4) components coexist in our system, which impose opposite effects on graphene formation. To rationalize this competitive relationship, we thereby define a relative concentration ratio of [O]/[CH4], where [O] is determined by inlet CO2 content and system pressure P, and [CH4] is controlled by the conversion yield from CO2 to CH4 with catalyst Ni/Al2O3. A tail gas analysis reveals that the [O]/[CH4] presents a negative correlation with the P (Figure f). A higher P favors a smaller [O]/[CH4] ratio, corresponding to a higher conversion yield of CO2 to CH4. In Figure g, the MLG density reduced monotonically against [O]/[CH4]. By contrast, the nucleation of BLG follows a “volcano” curve versus the relative ratio, with an optimal BLG growth window located at a [O]/[CH4] ratio of 2.3–2.6 (Figure h). Therefore, the growth pressure P plays the most important role in the conversion of CO2 to BLG. Based on our results, we propose a possible role of CO2 in our BLG growth (Figure i). Composed of O and C atoms, CO2 can simultaneously serve as an etching gas and carbon source for graphene growth. First, the oxygen atoms provided by CO2 create defects on SLG, which break the SLG self-limiting growth and host the nucleation sites for BLG. It was reported that such defects with a low energy barrier could serve as the BLG nucleation centers.[21,23,42] Around the defect, Cu–O and C–O bonds can promote decomposition of CH4 and nucleation of the second graphene layer. Therefore, the CH4 produced from CO2 can be deposited on the defect sites to form the second layer. Furthermore, a moderate amount of oxygen atoms inhibit the further growth of MLG, giving rise to a high selectivity toward BLG.

It has been reported that BLG, with weaker temperature-dependent scattering than SLG, exhibits a higher mobility (μ).[43] In order to evaluate the mobility of BLG, we fabricated a back-gate field effect transistor (FET) device based on our BLG single crystals, with the schematic diagram shown in Figure a. A graphene channel was constructed based on the BLG sample. Au/Ti electrodes work as the source and drain and the silicon wafer works as the bottom-gate. Figure b is the optical image of the back-gate FET device. Calculated from the transfer characteristic data in Figure c,[44] the μ is ∼2346 cm2 V–1 s–1 at room temperature. The value is comparable to those of CVD-grown BLG using CH4 as the carbon source.[36,45,46]

Figure 5

Electrical measurement of a back-gate FET device based on BLG single crystals. (a) Schematic of the back-gate BLG FET device. (b) Optical image of the BLG FET device (scale bar: 10 μm). (c) Transfer characteristic of the BLG FET device, with Vd = 1 V.

Conclusions

In conclusion, we have successfully developed a unique catalytic strategy to precisely convert CO2 into a high-quality BLG single crystal with large size, swift growth rate, and high electronic performance. Facilitated by oxygen atoms originating from CO2, the growth rate of BLG is accelerated to ∼300 μm h–1, the fastest kinetics reported to the best of our knowledge. An appropriate concentration of etching oxygen atoms not only creates the precise nucleation centers for BLG but also inhibits the MLG formation. Through this finely tuned precise reduction of CO2, larger BLG can be grown more quickly and reliably, boosting the development for next-generation electronic and twistronic devices.

Materials and Methods

Graphene Preparation

BLG Single Crystal Growth from CO2

A piece of 30 μm thick Cu foil was first electrochemically polished in a polishing solution composed of 25 mL of H2O and 75 mL of H3PO4, under 5 V and 2 A for 30 s on both sides. After being washed with deionized water and dried under N2 flow, the polished Cu foil was placed in the second heating zone of the CVD system; the commercial catalyst Ni/Al2O3 (Sichuan Shutai Co. Ltd.) was loaded in the upstream side of the Cu foil. To prevent impurities from depositing on the copper foil, the space between the catalyst and the copper foil was filled with quartz wool. When the CVD system was pumped to 20 mTorr for 30 min, the Cu foil was put into the heating zone of a furnace. To deplete carbon impurities, Cu foil was annealed in an O2 atmosphere with a partial pressure of ∼5 mTorr for 10 min. Before that, the catalyst was heated at 300 °C in the first heating zone until the pressure remained stable. The reaction was then carried out at 1055 °C with 200 sccm H2 and 1–2 sccm CO2 under 800–1600 Pa for 5–120 min. After growth, the Cu foil was cooled to room temperature under 100 sccm H2 for 30 min.

SLG Growth from CH4

A piece of 30 μm thick polished Cu foil was placed in a quartz tube of our CVD system. When the CVD system was pumped to 20 mTorr for 30 min, the Cu foil was put into the heating zone of the furnace, and then the Cu foil was annealed in an O2 atmosphere with a partial pressure of ∼5 mTorr for 10 min. Then, the reaction was carried out at 1055 °C with 200 sccm H2 and 1.5 sccm CH4 under 1200 Pa for 40 min. After growth, the Cu foil was cooled to room temperature under 100 sccm H2 for 30 min.

Graphene Transfer to Wafer from Cu Foil

The graphene/Cu samples were coated with 2 drops of 6% PMMA solution (poly(methyl methacrylate), Mw 50 000, in anisole) by spin-coating, and then, the samples were etched by 0.5 mol L–1 Marble’s solution (CuSO4·5H2O in deionized water and 37% hydrochloric acid) for 5 min and 0.1 mol L–1 APS solution (ammonium peroxydisulfate in deionized water) for 6 h. After the copper was etched, PMMA-coated graphene samples were cleaned by deionized water three times. The samples were transferred to 300 nm SiO2/Si and further heated at 150 °C for 2 h. PMMA was dissolved in acetone at 70 °C for 1.5 h and acetic acid for 0.5 h, separately, and then, the samples were further rinsed in isopropanol and dried under N2 flow.

Graphene Transfer to TEM Grid from Wafer

The graphene/SiO2/Si samples were coated with 1 drop of 6% PMMA solution (Mw 50 000) by spin-coating and then placed into buffered oxide etch (BOE) solution (8 g of ammonium fluoride in 2 mL of hydrofluoric acid and 12 mL of deionized water). The substrate was etched away after 24 h of treatment, followed by cleaning with deionized water. Then, graphene was transferred to a TEM grid and heated at 85 °C for 10 min. Finally, PMMA was dissolved by hot acetone.

Graphene Transfer to TEM Grid from Cu Foil

The graphene/Cu samples were coated with 2 drops of 6% PMMA solution (Mw 50 000) by spin-coating; after the copper was etched, the PMMA-coated graphene samples were cleaned by deionized water and then transferred to a TEM grid and heated at 85 °C for 10 min. Finally, PMMA was dissolved by hot acetone.

Structural Characterization

The as-prepared BLG single crystal, transferred onto a 300 nm SiO2/Si substrate, was characterized by Raman spectroscopy and optical microscopy (LabRAM HR, 514 nm laser wavelength). The graphene samples were peeled from the SiO2/Si substrate for further TEM characterization (HT7700 EXALENS, 120 kV; and TECNAI G2 F20 S-TWIN, 200 kV) and SEM characterization (NOVA NANOSEM 450, 5 kV). The electrical transport properties were measured by a KEYSIGHT probe station (B1500A) equipped with a PICO probe (ST-20-5).

Growth Mechanism Studies

CO2 Hydrogenation Mechanism

The CO2 hydrogenation includes two processes, as shown below:

The mathematical derivation of the relative oxygen ratio ([O]/[CH4]) is presented as follows:

[CO2]In is the initial concentration of inlet CO2 into the CVD system. The total concentration of oxygen atoms, existing as CO2, CO, and H2O in the system, is determined by the inlet CO2 concentration and pressure as eqs and 2 show.

Collection and Analysis of Intermediate Gases

A pipe was connected at the back of the CVD system before the reaction. During the BLG growth process, we collected 500 μL of the tail gas from the CVD system by a sampler and injected the mixture into a GC equipped with a TCD and FID for further analysis.

Process of Etching Graphene by CO2

A piece of SLG/Cu foil was prepared based on the synthesis method of SLG. When the CVD system was pumped to 20 mTorr, the SLG sample was put into the heating zone of a furnace. Then, the etching process was performed at 900 °C with 1.5 sccm CO2 under 262 mTorr for 1 min. After the process, the sample was cooled to room temperature under 1.5 sccm CO2.

Process of Etching Graphene by H2

When the system was pumped to 20 mTorr, the SLG sample was put into the heating zone of a furnace. The SLG was annealed at 900 °C with 200 sccm H2 under 110 Pa for 1 min. After the etching process, the sample was cooled to room temperature under 200 sccm H2.

FET Device Based on BLG Single Crystals

Fabrication of the FET Device

The BLG single crystal grown on copper was transferred to a SiO2/Si substrate with the PMMA-assisted method. A photoresist was spin-coated on the surface of the BLG sample which was then baked at 170 °C for 3 min. A graphene channel with a width of 5 μm was preserved. Other regions were exposed by electron beam lithography (EBL), and the photoresist was washed away by the developer solution. Unwanted graphene regions were etched away by O2 plasma. The sample was covered again by the photoresist. A pattern of source and drain electrodes were exposed by EBL and the photoresist was washed away, followed by the deposition of 15 nm Ti and 70 nm Au on the pattern by physical vapor deposition (PVD). After being cleaned by acetone, the back-gate FET device based on the BLG single crystal was fabricated.

Calculation of Mobility of the FET Device

The calculation is based on eqs and 5.

Here, dId/dVg is 2.6979 × 10–5, the channel length L is 5 μm, the channel width W is 5 μm, Vd is 1 V, and the capacitance between the channel and the gate per unit area Ci is 1.15 × 10–4 F m–2.

Here, ε0 is 3.9, εr is 8.85 × 10–12, and d is 300 nm.

Safety Statement

Caution! Hydrofluoric acid solution is extremely corrosive. It must be handled very carefully in the hood. Face masks, safety goggles, and double nitrile gloves are needed when using this reagent.

Caution! Hydrogen is classified as a GHS Flammable Gas, Category 1. Safety goggles are mandatory, and the hydrogen alarm next to the CVD system needs to be activated during the graphene growth process.