All fabrication was performed on Si wafers with 285 nm of thermally grown oxide to

Transcription

1 Supporting Information: Substrate preparation and SLG growth: All fabrication was performed on Si wafers with 285 nm of thermally grown oxide to aid in visual inspection of the graphene samples. Prior to the evaporation of the Cu/Ni layer, all wafers were rigorously cleaned in heated base and acid baths following procedures equivalent to those for CMOS processing. The wafers were then immediately loaded into an e-beam evaporator for metal deposition. 5 nm of Ni were evaporated as an adhesion layer, followed by 495 nm of Cu without breaking vacuum. Growths were carried out in a low pressure hot-walled chemical vapor deposition system. Substrates were cleaved into narrow strips, followed by immersion in acetic acid at 35 C for 10 minutes in order to remove most of the copper oxide. The samples were then quickly loaded into the quartz reaction tube and pumped to a base pressure of 10 mtorr. A constant flow of H 2 ( sccm) was then introduced into the chamber at a pressure of 2 Torr while the reaction tube was heated to the growth temperature of 1000 C at a rate of ~ 40 C/min. After reaching this temperature, 875 sccm of CH 4 was flowed for the growth step and the total pressure was maintained to 11 Torr. Standard growths lasted between 10 and 20 minutes, after which the system was slowly cooled at a rate of ~ 20 C/min without altering the gas flow. After reaching 200 C the system was then purged and pressurized with 1000 sccm of Ar. Micro Raman spectra of the grown samples were taken to ensure the quality of the resulting SLG. Device fabrication:

2 Our substrates with grown SLG were spin coated with photoresist. Standard photolithography defined a pattern in photoresist in the shape of the devices. A 30 second oxygen plasma (60 mtorr, 50 sccm O 2, 150W RF power) was used to pattern SLG by etching away areas unprotected by photoresist. We then etched the unwanted Cu/Ni with a continuously refreshed dilute solution of FeCl 3 :HCl:H 2 O in order to slowly etch the excess Cu/Ni, while removing the etch products. The Cu/Ni etch was timed to allow for the etchant to undercut the photoresist and define the SLG channel, while leaving behind two appropriately sized SLG/Cu/Ni pads to be used for making electric contact with the device. Immediately following the etch, substrates were gently flushed with deionized water for several seconds. Samples were then carefully blown dry with N 2 before being placed into a vacuum chamber and heated to ~70 C for 15 minutes. The photoresist layer was then vigorously stripped with acetone and a subsequent isopropyl alcohol rinsing. Top gate fabrication: A 100 nm thick layer of SiO 2 was evaporated directly onto the devices via e-beam evaporation. An additional step of standard photolithography was used, with a lift-off resist/photoresist bilayer. An adhesion layer of Cr (5 nm), followed by 45 nm of Au were then evaporated to define the top gate electrodes. Lift-off was carried out in an NMP based solvent. In order to facilitate uniform electrical contacts with the SLG/Cu/Ni electrode pads, windows in the evaporated SiO 2 layer were opened using standard photolithography followed by a wet oxide etch (30:1 buffered oxide etch). Comparison of growth materials:

3 Figures S1 and S2 below provide a comparison of growths on Cu foils (left images) and evaporated Cu thin films (right images). Both growths result in visible features over the entire substrate (Fig. S1). Although the average feature size is smaller for evaporated thin films, the Raman spectra presented in Fig. S2 are very similar. The overall quality of our grown films is also very similar to reports on Cu foils. Figure S1. Side by side comparison of growths on a Cu foil (left) and an evaporated Cu thin film (right). Feature sizes appear to be much larger for Cu foils than for Cu thin films.

4 Figure S2. Side by side comparison of Raman spectra for Cu foil (left) and an evaporated Cu thin film (right). Cu background has been subtracted. Fig. S3 shows 2d Raman maps of the D, G, and 2D bands at a constant scale. The D bands, although present, are of very low intensity especially when compared to the intensities of the 2D bands. Fig. 1(c) is reproduced below as Fig. S3(d) for comparison. The majority of the data fits well to an applied Gaussian (μ = 0.40; σ = 0.06). This single peak area accounts for at least 93% of our dataset, from which we estimate as much coverage of SLG. Figure S3. a)-c) 2d Raman maps of D, G, and 2D bands. d) 2d map of the G/2D ratio, reproduced from the main manuscript.

5 120 μ = 0.40 σ = Figure S4. Histogram and Gaussian fit of the data presented in Fig. S3(d). At least 93% of the area can be attributed to SLG. Effect of film thickness and Ni adhesion layer In our experiments 500 nm of Cu was used to grow SLG. Thicker Cu films produced similar quality graphene, whereas thinner films were destroyed during the growth process. At a film thickness of 100 nm the Cu balled up as well as evaporated off, as shown in Fig. S5 below.

6 Figure S5. Optical image of sample substrate after growth on a thin film. The integrity of the film is lost and only Cu islands remain. Fig. S6 shows Raman spectra comparisons of Cu (left) and Cu/Ni (right) growths. There is no significant difference between the two indicating there is no Ni present on the surface. Figure S6. Comparison of growths on Cu (left) and Cu/Ni film (right). Spectra have no significant differences indicating presence of Ni.

7 Oxide removal and Cu etching/pr stripping: As mentioned in the manuscript, removal of the copper oxide is essential to the production of a uniform graphene sheet. Fig. S7(a) provides an example of a device fabricated on a substrate that was reduced for an insufficient amount of time prior to growth. A Raman spectrum taken on a crack in the Cu film before fabrication is shown in the inset. The presence of a much larger D peak, as well as a significant reduction in the 2D/G ratio, indicate a severe decrease in SLG quality. As noted, the order of the final two steps in the fabrication process is important. Damage can occur to the SLG sheet if photoresist is removed before etching Cu/Ni. In addition, the control of the Cu/Ni pad etch rate becomes much more difficult. Fig. S7(b) presents an image in which the order of these steps was switched. Although it is clear that the channel survived (connecting strip), the larger graphene sheets were torn considerably. Figure S7. a) Contrast enhanced image of a device fabricated on an insufficiently reduced substrate. The copper oxide (blue lines) in the area of graphene

8 identically follow the grain boundaries of the residual copper pad. Inset: Raman spectra taken on grain boundaries suggest the presence of a lower quality film. b) Contrast enhanced 20x brightfield optical image of a sample fabricated by stripping photoresist before etching the copper. The graphene sheet has been significantly damaged and a large amount of photoresist residue (blue sheet) has been left behind. Cu/Ni etch solution: A custom stock solution of FeCl 3 :HCl:H 2 O was used in the etch step for all devices presented in this paper. In order to create this solution FeCl 3 H 2 O was combined with concentrated HCl and de-ionized H 2 O in a ratio of 3.5g:10mL:100mL. This solution provided accurate control of the Cu/Ni undercut while maintaining a reasonable etch rate.