Characterization of Graphene-based FET Fabricated using a Shadow Mask

To pattern electrical metal contacts, electron beam lithography or photolithography are commonly utilized, and these processes require polymer resists with solvents. During the patterning process the graphene surface is exposed to chemicals, and the residue on the graphene surface was unable to be completely removed by any method, causing the graphene layer to be contaminated. A lithography free method can overcome these residue problems. In this study, we use a micro-grid as a shadow mask to fabricate a graphene based field-effect-transistor (FET). Electrical measurements of the graphene based FET samples are carried out in air and vacuum. It is found that the Dirac peaks of the graphene devices on SiO2 or on hexagonal boron nitride (hBN) shift from a positive gate voltage region to a negative region as air pressure decreases. In particular, the Dirac peaks shift very rapidly when the pressure decreases from ~2 × 10−3 Torr to ~5 × 10−5 Torr within 5 minutes. These Dirac peak shifts are known as adsorption and desorption of environmental gases, but the shift amounts are considerably different depending on the fabrication process. The high gas sensitivity of the device fabricated by shadow mask is attributed to adsorption on the clean graphene surface.


Details of fabricating graphene based devices using TEM grids a. Making PDMS
In our study, PDMS (Polydimethylsiloxane) slab was utilized as the supporting substrate for carrying and aligning the TEM grid. The PDMS slab must be flat with uniform thickness as much as possible in order to let the TEM grid close to the graphene and wafer for precise alignment. For this purpose we have made molds by using microscope glass slides (Fisherfinest™ Premium Plain Glass) (Fig. S1a) with 1 millimeter thickness. The mold consists of two or several glass slides which are separated by two pieces of the same glass slides on both edges (Fig. S1b). The distance between two slide glasses is the same as the thickness of a slide.
This distance becomes actually the thickness of the solid PDMS slab. After making the mold, mixture of prepolymers for PDMS (Curing agent and base with ratio 1:10) was poured in to the mold. Heating at 100 o C is required to cure the PDMS. After 2 hours PDMS became solid, and then the mold was disassembled. The advantage of using this mold is that it is simple to make and PDMS slab has smooth surface with uniform thickness, which is very critical for precise alignment with sub-micron accuracy. The PDMS slab was punched to make a hole which was used for window, to allow the metal vapor pass through it.

b. Method to align TEM grid into the desired location
In this study the graphene FETs were supported on Si/SiO2 substrates and bottom-gated using the SiO2 (300 nm thermal oxide) as the gate dielectric and the p + doped Si substrate (ρ ∼ 3 × 10 −3  cm) as the gate electrode. Highly oriented pyrolytic graphite (Natural Kish graphite (Grade 300), Graphene supermarket) was used to produce graphene flakes by mechanical exfoliation of using adhesive tape (Scotch Tape, 3M, Inc). The specifications of the TEM grids are listed in the Table 1. To align TEM grid into the desired location, the TEM grid was attached at a hole (1 mm diameter) punched in a PDMS slab (1 mm thick) as shown in Fig. S3, as the PDMS is sticky to clean metallic surfaces.

c. PMMA residue on the device fabricated by conventional e-beam photolithography
After samples were fabricated with a conventional e-beam lithography technique, the sample was cleaned with dissolving and annealing, but the PMMA residue was not removed completely. Specifically, dissolving was performed with boiled acetone and chlorofom cleaning, and annealing was done with Ar/H2 flow, at 400 o C, overnight. From AFM measurement, residue layer with 1 nm roughness was found, as shown in Fig. S4.

The resistance change depending on gate voltage for different graphene samples in different environments. a. Graphene on hBN devices
We prepared 6 FET samples with graphene on hBN using TEM grid as a shadow mask. All of them showed the same trends in air, low vacuum, and high vacuum, as shown Fig. S5.

b. Graphene on Si devices fabricated using TEM grids
We prepared 2 FET samples with graphene on SiO2 using TEM grid as a shadow mask. Both of them showed the same trends in air, low vacuum, and high vacuum, as shown Fig. S6.

c. Graphene-on-hBN devices fabricated using TEM grids in liquid environment
Graphene-on-hBN device also showed the same results as graphene-on-SiO2, either in water or in ethanol (Fig. S7). Therefore, the behavior in liquid does not depend on the substrates.

Fig. S7
The relationship between total resistance Rtot and back-gate bias Vg of the graphene on hBN device fabricated using TEM grids; (a) in water; (b) in ethanol. All measurements were carried out at room temperature.

d. Extrapolating and estimation of Dirac peaks
Dirac points were estimated from extrapolating the data in Fig. 2(a) into Fig. 2(d). From the data in Fig. 2(d), it is confirmed that the shape and baseline of curves are the same, and only the shifts in x-axis can be found. As the Rtot at Vg= 40V in air in Fig. 2(a) have lower values than the Rtot at Vg= -20V measured at 7hr in Fig. 2(d), the Dirac point in air in Fig. 2(a) should be shifted to about 80 V. By the same manner, the Dirac points at 10 -3 and 10 -5 Torr in Fig. 2(a) are estimated to be about 60, -30 V, respectively. These estimation from extrapolation is added in this revision. Fig. S8 While the shape and baseline of curves in the data in Fig. 2(a) and (d) were similar, but the shifts in x-axis were found. Dirac points were estimated from extrapolating the data in Fig.  2(a) into Fig. 2(d).

e. Time evolution of Dirac peak shift
The time evolution of Dirac point was plotted, as shown in Fig. S9. The tendency of our result is similar to previously published result, 1-4 but the time constant estimated from exponential function fitting was much shorter (time constant = 0.76 hr). This fast adsorption can be attributed to the clean surface of the FET using shadow mask.

f. Sticky residue on graphene
Fig. S10 In order to confirm the adhesive being washed by ethanol, graphene sample was exfoliated with the Scotch tape, and submerged into ethanol. (a) Particles and residues were found on graphene surface from AFM topography. (b) Particularly, strong frictional forces from LFM images were found on graphene surface. This means that the sticky residue on SiO2 was dissolved by ethanol, and some amount of this residue was moved onto the graphene surface.