The interfacial charge transfer between mechanically exfoliated few-layer graphene and Cr2O3 (0001) surfaces has been investigated. Electrostatic force microscopy and Kelvin probe force microscopy studies point to hole doping of few-layer graphene, with up to a 150 meV shift in the Fermi level, an aspect that is confirmed by Raman spectroscopy. Density functional theory calculations furthermore confirm the p-type nature of the graphene/chromia interface and suggest that the chromia is able to induce a significant carrier spin polarization in the graphene layer. A large magnetoelectrically controlled magneto-resistance can therefore be anticipated in transistor structures based on this system, a finding important for developing graphene-based spintronic applications.
Induced spin polarization in graphene has potential for enhanced narrow-channel, spin polarized conduction in a spin field effect transistor (spin-FET).1–4 While effective spin polarized carrier injection into graphene is a challenging problem,5 an alternative is to exploit a magnetic insulator substrate or overlayer that can induce spin polarization in the graphene by interfacial spin interactions.1,6–10 Cr2O3 (0001) is a magnetoelectric, antiferromagnetic insulator with voltage switchable boundary magnetization;11–14 while the bulk of the material exhibits no net magnetization, the 0001 surface possesses a well-defined moment whose direction may be reversed controllably by the application of an electric field. The presence of a significant net (nonzero) spin polarization at the Cr2O3 (0001) interface is consistent with theoretical expectations15 and has been demonstrated by spin polarized photoemission,11 spin polarized inverse photoemission,12,13 x-ray circular dichroism,13 and spin polarized low energy electron microscopy.14 The huge surface spin polarization of chromia can be preserved at a buried interface, that is to say, the polarization may be retained even with an overlayer present.14,16 Chromia is thus a promising magnetoelectric gate dielectric that has the potential to induce spin polarization in a graphene overlayer due to the proximity effect.6,7 Given the aforementioned advantages, the graphene-on-Cr2O3 (0001) system could offer a route to a nonvolatile magnetoelectric spin valve or spin FET.2,3,7,8
In this work, we report a systematic study of interfacial charge transfer in graphene/Cr2O3 (0001) heterostructures. Scanning probe microscopy and Raman studies reveal an interfacial charge transfer between these materials and point to p-type doping of the graphene with up to a 150 meV shift in the Fermi level. The charge transfer effect and the induced spin polarization of the graphene are investigated using density-functional theory (DFT). These calculations show that the charge transfer is relatively small, while the induced spin polarization is extremely high in the vicinity of the Fermi level of graphene, implying that a large magnetoelectrically driven magneto-resistance can be expected for this system. This magnetoelectric interface may therefore be important for developing spintronic applications.
The graphene flakes studied here were mechanically exfoliated directly onto single-crystal Cr2O3 (0001). Atomic force microscopy (AFM), electrostatic force microscopy (EFM), and Kelvin probe force microscopy (KPFM) studies were then carried out on these samples, using a Bruker Multimode 8 AFM system. The topography of the graphene on chromia was examined by tapping- and contact-mode AFM (AFM tip: Bruker SCM-PIT), from which the thickness of graphene can be evaluated. The EFM and KPFM measurements were performed using conductive AFM tips (SCM-PIT, Bruker, USA) close to the tip resonant frequency of 75 kHz. For the KPFM measurement, the lift height was 20–40 nm, and the AC bias was 0.5 V. All experiments were performed at room temperature (∼297 K), which is about 20 K lower than the Neel temperature of the single-crystal Cr2O3 (0001). The graphene on Cr2O3 heterostructure samples were heated to 120 °C prior to the EFM and KPFM measurements to minimize the possible water adsorbates from the ambient, which are known to hole dope graphene.17 We kept the annealing temperature low in our studies, in order to avoid aggressive thermal treatment induced structural distortion.18
To implement the DFT calculations, we constructed a Cr2O3 (0001) slab model with the outmost layers consisting of Cr atoms, in accordance with previous studies,8,19 as well as with an oxygen terminated surface for comparison. The total thickness of this slab was about 10 Å. For studies of the heterostructure formed with graphene on top of the Cr2O3 (0001) slab, we adopted a 2 × 2 graphene supercell to match the 1 × 1 Cr2O3 (0001) slab. The lattice mismatch at the graphene/Cr2O3 (0001) interface is taken to be about 0.7%. Since the electronic properties of graphene are much more sensitive to the choice of lattice constants than those of the Cr2O3 (0001) slab, we applied a strain of −0.7% to the lattice constants of the latter to match to the graphene supercell. The vacuum lengths between neighboring supercells were taken to be longer than 15 Å to avoid wave function overlap.
All calculations were performed within the framework of the Vienna ab initio simulation package (VASP),20,21 a first-principles plane-wave code based on spin-polarized DFT. The exchange correlation was treated with the Perdew-Burke-Ernzerhof (PBE) functional,22 and the projector augmented wave (PAW) method was used to describe the electron-ion interaction. The plane wave basis was set with an energy cutoff of 500 eV in the calculations. The van der Waals (vdW) corrections and dipole corrections were performed for the heterostructure.23–26 To correct the strong on-site electronic correlation, the DFT + U method was used for the Cr atoms with U − J = 4 eV. The Brillouin zone integration was performed using Monkhorst-Pack 7 × 7 × 1 grid for geometry optimizations and 9 × 9 × 1 for static electronic structure calculations.27 Geometric structure relaxation ended once the force on each atom was less than 0.01 eV/Å, and the energy criterion was 10−5 eV. Bader's atom-in-molecule method, which is based on charge density topological analysis, was performed to analyze charge population.28 Since the actual graphene on chromia samples were realized through exfoliation, the precise registry of the graphene with respect to the chromia is not known. Several models were run for both oxygen and chromium terminated surfaces, with different various graphene registrations of the graphene with the chromia, confirming that the obtained results are largely independent of registry (supplementary material).
Figure 1(a) shows a tapping-mode AFM topography image of a few layer graphene (FLG) flake on Cr2O3 (0001). The majority of the graphene surface is flat, with isolated “puddle” areas with lower height signals. These features cannot be directly correlated with the morphology of the chromia surface and are not observed in the contact mode image [Fig. 2(a)], suggesting different types of graphene/chromia interface interactions, which may originate from a variation in the interfacial adsorbate layer. Similar puddles can also be resolved in the EFM measurements. Figures 1(b)–1(d) show the EFM phase mapping of the same region for DC tip voltages of 0 V, +10 V, and −10 V, respectively. We compared the signal profiles along the same line at different bias voltages [Fig. 1(e)]. For consistency, we chose a region where no “puddle” features are present. At zero tip bias, there is no apparent difference in the EFM phase signal between the graphene and chromia surfaces, except at the graphene boundaries. In contrast, a clear negative phase shift is observed on graphene at a non-zero tip bias, suggesting an enhanced tip-sample interaction. This is understandable as the mobile carriers in graphene can respond to the AC modulation of the tip bias, while the chromia surface is insulating. A notable asymmetry in the EFM response of graphene has been observed between the tip biases of +10 V and −10 V, which indicates the presence of a non-zero doping for the graphene sample on Cr2O3 (0001). The larger shift (by ∼3°) observed at −10 V suggests that the electrostatic interaction between the biased tip and graphene is enhanced at negative bias. This observation is consistent with a p-type doping of the graphene on the Cr2O3 (0001) surface, where a negative (positive) tip bias would increase (suppress) local hole density. Similar asymmetry in the EFM phase shift is also observed for the puddled areas.
In Fig. 2(a), we show the contact-mode AFM of a close-up view of the area in Fig. 1(a). The RMS roughness of the few layer graphene is 2.1 nm, much smoother than the chromia surface (RMS roughness of 3.5 nm). By averaging a large area, we extracted an average height of 4 nm for the few layer graphene flake. We expect the actual sample thickness to be much lower, as the graphene sheet does not fully conform to the rough Cr2O3 surface, which is evident from the signal line profile [Fig. 2(a)] and consistent with the different surface roughness values extracted from the graphene and Cr2O3 surfaces. We then performed the Kelvin probe force microscopy (KPFM) measurement in the same area. Figure 2(b) shows that the surface potential of the few layer graphene flake on Cr2O3 (0001) is about 50 mV higher than that of the pristine Cr2O3. As KPFM probes the work function difference,29 we expect the surface potential of undoped graphene [work function of 4.6 eV (Ref. 30)] to be 200 meV higher than Cr2O3 [work function of 4.8 eV (Ref. 31)]. The much lower surface potential difference confirms that the doping of the graphene is indeed p type, which decreases the Fermi level placement by 150 meV. The positive charge character in graphene is consistent, also, with what is predicted in a band-based model.32 Using an effective hole mass of 0.1me,33 where me is the free electron mass, we estimated the doping level to be 1.2 × 1013/cm2. Such a doping level is significantly higher than what is expected from ambient water or oxygen adsorbates18 and can only originate from the interface interaction.
The induced doping in the few layer graphene is also reflected in the Raman spectra of this sample, as shown in Fig. 3. The inset of Fig. 3 shows the fits to both G and 2D bands. The G band peak position is about 1583.8 cm−1 and the full-width-half-maximum is about 15.7 cm−1. The 2D band shape can be well described by two peaks, 2D1 and 2D2, which is the characteristic spectrum of few layer graphene.34,35 The corresponding peak positions of 2D1 and 2D2 are 2692.6 cm−1 and 2727.3 cm−1, respectively. Both of the 2D peaks of the sample exhibit a significant blue shift (by 5–10 cm−1) compared with few layer graphene on weakly interacting substrates.35,36 Such a blue shift can only originate from a doping effect18,37 and further confirms the interfacial charge transfer.
The p-doped charge transfer inferred above is moreover consistent with the results of our DFT calculations of monolayer graphene/chromia heterostructures (Fig. 4), which reveal a transfer of 0.0008 electrons per carbon atom, or a 2D density of 3 × 1012/cm2, from graphene to chromia. For monolayer graphene, this doping level corresponds to a Fermi level of ∼200 meV, which is comparable with that extracted from KPFM on few layer graphene. Several different trials, with different registrations of the graphene with respect to the chromia, have been considered but do not significantly affect the result of the calculations. Although the charge transfer varies from carbon atom to carbon atom (Fig. 4), the band structure of the chromia is n-type, while the graphene adlayer is p-type, in the calculated band structure (supplementary material).
In spite of the relatively small charge transfer inferred above, the DFT calculations reveal that the boundary polarization of Cr2O3 (0001) induces a very high level of spin polarization in graphene. Figure 5 plots the density of states (DOS) of the pristine chromia surface and chromia with a graphene overlayer. It is clear that the spin density is predominately determined by the Cr atoms in the chromia surface layer and that the outmost Cr-layer retains spin-up ordering that leads to the graphene developing a net spin-up density. This is further confirmed by plotting the density of states difference and the band structure around the Fermi level for the system (supplementary material). This large carrier spin polarization makes graphene-on-chromia a promising material candidate for constructing magnetoelectric transistors.
In conclusion, the presence of induced p-type doping in graphene/few-layer graphene on Cr2O3 (0001) has been confirmed by EFM and KPFM and shown to be consistent with a band model of the doping polarity based on DFT. These results suggest that a large magnetoelectrically controlled magneto-resistance can be anticipated in transistor structures based on the graphene/Cr2O3(0001) system, a discovery that could be important for developing graphene-based spintronic applications.
See supplementary material for the more complete density function theory overview of the electronic structure of graphene on chromia.
The authors would like to thank Dmitri Nikonov and Jun Yan for the valuable discussions and Yongfeng Lu for access to the Raman facility. Shi Cao would like to acknowledge Dr. Yuan Huang for his suggestions on the preparation of high quality graphene flakes on Cr2O3(0001). This research was supported by the National Science Foundation, through Grant Nos. NSF-ECCS-1508541 and ECCS-1509221, and the Nebraska Materials Research Science and Engineering Center (MRSEC) Grant No. DMR-1420645, as well as by the Nanoelectronics Research Corporation (NERC), a wholly owned subsidiary of the Semiconductor Research Corporation (SRC), through the Center for Nanoferroic Devices, an SRC-NRI Nanoelectronics Research Initiative Center. Work by Z.X. and X.H. was supported by the U.S. Department of Energy (DOE), Office of Science, Basic Energy Sciences (BES) under Award No. DE-SC0016153 (scanning probe studies).