We have developed a dry transfer method that allows graphene to be transferred from polymer- thyl-methacrylate (PMMA)/Si (silicon) substrates on commercially available hexagonal boron ni- tride (hBN) crystals. With this method we are able to fabricate graphene devices with little wrin- kles and bubbles in graphene sheets, but that do not degrade the electronic quality more than the SiO2 substrate does. For hBN to perform the function described above substrate cleanliness is critical to get high quality graphene devices. Using hBN as a substrate, graphene exhibits enhanced mobility, reduced carrier inhomogeneity, and reduced intrinsic doping compared to graphene on SiO2 substrate.
The search for new substrates that improve graphene properties has been an intense focus of research since gra- phene was first isolate in 2004 [
The quality of substrate-supported graphene devices has not improved since the first observation of the ano- malous quantum Hall effect in graphene and its bilayer [
In this work, we address a mechanical transfer method to deposit graphene on boron nitride substrates in order to fabricate graphene based devices. The hBN crystals were commercially available (Momentive, Polarthem grade PT110) that are smaller size of order of ~45 µm than non-commercial ultrapure hBN. Different manners of exfoliation of both hBN and graphene were developed, aiming to optimize the best way of exfoliation. Finally we will characterize the graphene transferred using Raman microscopy, atomic force microscopy and electronic transport measurements in fabricated graphene devices.
We first deposit thin hBN crystals (Momentive, Polartherm grade PT110) on a Si wafer with 285 nm of SiO2 by mechanical exfoliation of hBN powder using a Blue Medium Tape P/N 18074 (Semiconductor Equipment Corp.) keeping the substrate and tape for 4 minute. By optical microscopy few layer hBN flakes were initially identi- fied. The surface of every hBN flake is characterized by atomic force microscopy (AFM) (BRUCKER-Icon) to ensure it is free of contaminants or step edges, and also to measure its thickness. Number of layers in a hBN flake can be confirmed by measuring their thickness. Since 0.35 nm is the lattice constant for hBN in C direction, therefore, AFM may be useful for this purpose. It has been deduced from AFM images that a thickness of 20 - 30 nm corresponds to light blue flakes of hBN, as shown in
(Color online) Optical image of two suitable hBN flakes exfoliated on Si/SiO2 substrate with thickness of (a) ≈22 nm and (b) ≈13 nm
deposited on bi-layer PVA/PMMA. The Raman data are recorded by using a micro-Raman spectrometer (Reni- shaw InVia) setup with a laser excitation of 632.8 nm (EL = 1.96 eV), using a 100´ objective, with a laser spot size ≈ 1 µm. We used a laser power about 10 mW such that heating effects can be neglected [
A crucial breakthrough in this paper is the development of a special method to transfer graphene from one substrate to an exact location on another substrate (in this case an hBN flake). In this method extreme care is taken to reduce water residues and this is named “dry transfer method”. We adapted the dry transfer method to our needs. A dry transfer method consists mainly of four steps as illustrated in
(Color online) Schematic illustration of the dry transfer method to fabricate graphene-on-hBN devices. (a) PVA/PMMA film with graphene is glued in the plastic window; (b) Next the bi-layer is put in water, in order to dissolve the PVA film, releasing the PMMA with graphene from the Si sub- strate; (c) The plastic window is attached to aluminum slide with a hole and the hBN flake is exfoliated on Si/SiO2; (d) The aluminum slide is mounted to a micromanipulator from a mask aligner that allows us to align graphene with the hBN flake
slowly to 400˚C in a tube oven (1" Lindberg Blue M) in forming gas environment and anneal them there for ∼3 hours before slowly cooling down to room temperature. This treatment is very effective in removing polymer residues from the transfer.
After transfer process, we fabricate the electronic device using standard beam lithography (Raithe_Plus) and deposit chromium/gold (5 nm/60 nm) contact material using e-gun evaporation system. The polymer mask used for processing was PMMA 950K dissolved in anisole (7%). Lift-off was done in acetone, followed by isopro- panol rising.
We made approximately 40 exfoliations on SiO2/Si substrates and with helping of the optical microscopy, with good resolution; they were observed different sizes and colors. They were chosen flakes of light and dark blue colors and 20 - 40 µm size. These flakes are also thick enough to compensate for the SiO2 roughness, and thin enough to allow good backgate functionally, when are used as electronic devices. After finding suitable flakes AFM, images of the hBN were made to check surface roughness and possible contaminants. Approximately 1 out of 10 flakes have the correct thickness; appear clean and flat under the AFM. The relatively low chance of having a flat and clean flake could possibly explained by the fact that the hBN crystal will likely break at the weak spots in the crystal during mechanical exfoliation. These weak spots could be made out of misaligned crystal planes or places where dirt is located inside the crystal. See
Graphene layer number deposited on bilayer PMMA/PVA film is confirmed by Raman spectroscopy. Raman spectra show clear difference between samples of different layer thicknesses. In particular, the intensity ratio of 2D band peak over G band peak, I2D/IG, is used to identify single layer graphene from bilayer and few layer graphene. As shown in
(Color online) AFM images show a comparison between a dirty (a) and a clean (b) hBN flake. (c) Image AFM of a hBN flake after thermal annealing in Ar/O2 atmosphere. Scale bar is 5 µm
indicating that graphene flake in (a) has no more 4 layers and graphene sheet in (b) must have more than 3 layers. Moreover, FWHM (Full width half maximum) of 2D band increases as the layer thickness increases, which is associated with the electronic band structure near the K and K’ points. It is important to observe that the D mode (defect related mode) do not appear in these spectra, indicating the good quality of our exfoliated graphene sheets. It is also observed that the background signal due to PVA/PMMA film is absent in these spectra.
(Color online) Raman spectrum of two graphene flakes showing the G band peak and the 2D band peak. In (a) no more than 4 layers and (b) more than 4 layers
Graphene that is transferred onto hBN before annealing show typical wrinkling as show in
The Raman spectrum of this graphene flake taken after transfer is very similar to that taken before transfer (pristine) which provided evidence that the Ar/H2 process successfully removed all organic contaminants. These Raman spectra are composed of the known spectra of hBN and graphene (
(Color online) (a) Optical image of dry transferred graphene before annealing; (b) An AFM image that shows wrinkles already present on the graphene surface; (c) After annealing treatment both wrinkles and bubbles have still ap- peared
In order to show the substrate dependence of the Raman lines of graphene, we compare in
(Color online) (a) Measured Raman spectra of graphene trilayer transferred on hBN/SiO2 (blue color) and the spectrum of hBN/SiO2 alone (red color). This sample was heat treated in Ar/H2 at 400˚C after transferred. For comparison in (b) shows the 2D band Raman spectra of graphene trilayer on hBN and on SiO2 substrate
detectable defect-related D band, suggesting that this thermal annealing not damage graphene, and raising the possibility that our cleaning procedure can be applied on hBN and graphene, and facility the preparation of de- vices involving multiple layers [
We found that anneal step in Ar/H2 flow degrades our contacts so we omitted this step here. Also no other an- nealing steps have been used; keeping the fabrication process the same as for SiO2-based devices. Typically the resistance between two contacts is in order of ~1.0 - 2.5 KΩ. Next we performed electrical transport measure- ments in order to estimate the charge mobility. We contacted Cr/Au electrodes to five wrinkled MLG flakes. A typical device is shown in the inset of
(Color online) Resistance as a function of the back- gate voltage for graphene trilayer on h-BN after thermal an- nealing. Inset shows the graphene flake fabricated by e-beam lithography in a Hall bar geometry. Scale bar is 10 µm
mobility is determined as µ = 1/neρ, where ρ is the resistivity of the flake. The field-effect mobility of this de- vices reaches 7200 cm2/V∙s at 4 K at high density of carries of 18.5 ´ 1011 cm−2. This mobility is about one order of magnitude higher than previously reported values for supported values [
A methodology for fabricating graphene electronic devices on commercially obtained hBN is presented. We show that commercially obtained hBN, which is available in large amounts, offers a good alternative to non- commercial ultrapure hBN. We described a mechanical transfer method to deposit graphene on hBN substrates in order to fabricate graphene devices employing a mechanical co-lamination and transfer process. Additionally, since the fabrication recipe we developed requires only one cleaning step, it allows for fast device preparation of graphene on hBN with a little bubbles and wrinkles. Electrical transport measurements of the graphene devices on hBN have mobilities and carrier inhomogeneity that are almost an order of magnitude better than devices on SiO2. We think that this work can contribute in performing of atomic engineering of graphene on hBN, and sheds light on fundamental research as well as electronic applications based graphene/hBN heter-structures in the route of continual investigations of single-layer, bilayer and few-layer graphene.
The authors would like to acknowledge to Professor R. Ando and Professor Y. Pusep for their valuable supports to this research. CNPq and FAPESP (Brazilian agencies) for financial support.