Home
Search
Collections
Journals
About
Contact us
My IOPscience
Fluorinated graphene and hexagonal boron nitride as ALD seed layers for graphene-based van der Waals heterostructures
This content has been downloaded from IOPscience. Please scroll down to see the full text. 2014 Nanotechnology 25 355202 (http://iopscience.iop.org/0957-4484/25/35/355202) View the table of contents for this issue, or go to the journal homepage for more
Download details: IP Address: 128.220.8.15 This content was downloaded on 13/01/2015 at 15:46
Please note that terms and conditions apply.
Nanotechnology Nanotechnology 25 (2014) 355202 (8pp)
doi:10.1088/0957-4484/25/35/355202
Fluorinated graphene and hexagonal boron nitride as ALD seed layers for graphenebased van der Waals heterostructures Hongwei Guo1, Yunlong Liu1, Yang Xu1, Nan Meng1, Hongtao Wang2, Tawfique Hasan3, Xinran Wang4, Jikui Luo1 and Bin Yu5 1
Department of Information Science and Electronic Engineering, Zhejiang University, Hangzhou 310027, People’s Republic of China 2 Institute of Applied Mechanics, Zhejiang University, Hangzhou 310027, People’s Republic of China 3 Department of Engineering, University of Cambridge, Cambridge CB3 0FA, UK 4 National Laboratory of Microstructures and School of Electronic Science and Engineering, Nanjing University 210093, People’s Republic of China 5 College of Nanoscale Science and Engineering, State University of New York, Albany, New York 12203, USA E-mail: [email protected] Received 6 May 2014, revised 28 June 2014 Accepted for publication 14 July 2014 Published 12 August 2014 Abstract
Ultrathin dielectric materials prepared by atomic-layer-deposition (ALD) technology are commonly used in graphene electronics. Using the first-principles density functional theory calculations with van der Waals (vdW) interactions included, we demonstrate that single-side fluorinated graphene (SFG) and hexagonal boron nitride (h-BN) exhibit large physical adsorption energy and strong electrostatic interactions with H2O-based ALD precursors, indicating their potential as the ALD seed layer for dielectric growth on graphene. In grapheneSFG vdW heterostructures, graphene is n-doped after ALD precursor adsorption on the SFG surface caused by vertical intrinsic polarization of SFG. However, graphene-h-BN vdW heterostructures help preserving the intrinsic characteristics of the underlying graphene due to inplane intrinsic polarization of h-BN. By choosing SFG or BN as the ALD seed layer on the basis of actual device design needs, the graphene vdW heterostructures may find applications in lowdimensional electronics. Keywords: seed layer, ALD, dielectric growth, fluorinated graphene, hexagonal boron nitride (Some figures may appear in colour only in the online journal) A large number of graphene-based devices, such as field effect transistors [1–6], waveguides [7–9], and plasmonic structures [10–12] require integration of ultrathin high-κ dielectric on graphene to implement the desired functionalities. This is because high-κ dielectrics can enhance dielectric screening effect and increase the carrier mobility in graphene14. Although atomic-layer-deposition (ALD) serves reliably for thin dielectric growth on semiconductors, direct deposition of dielectric materials on graphene by thermal ALD is still challenging [13–15]. This is because the conjugated π-bond surface of pristine graphene is hydrophobic and chemically inert to most of ALD precursor molecules 0957-4484/14/355202+08$33.00
[16–18]. Functionalizing the graphene basal plane is one of the viable choices to achieve uniform ALD dielectric materials [17–22]. Ozone (O3) [17–19] or NO2 [20, 21] vapor phase functionalization, and remote plasma-assisted ALD process [22] allow deposition of ultrathin dielectric materials, however, they cause significant degradation in graphene lattice and decrease of carrier mobility. Wet chemistry functionalization modifies the graphene surface with minimum disruption of the intrinsic properties. This results in low dielectric constant in the ALD layer and gives rise to scalability issue (∼31 nm minimum dielectric thickness) [23, 24]. Dry chemical approaches, such as fluorine functionalization 1
© 2014 IOP Publishing Ltd Printed in the UK
H Guo et al
Nanotechnology 25 (2014) 355202
[25], can promote ALD process but are limited to epitaxial graphene [26]. An alternative approach is to use a non-covalent functionalized seed layer to catalytically react with ALD precursors and subsequently nucleate the growth of the dielectric materials. In this strategy, three different types of seeding layers have recently been explored: polymer buffer[5, 26–28], metal- [29, 30], and metal oxide- [31, 32] seed layers. Despite successfully enabling the deposition of dielectric materials [14, 15], these seeding layers have their own limitations. Polymer buffer seed layers can be used to obtain uniform and conformal dielectric materials [5, 26–28], but with low dielectric constants (∼2). The low decomposition temperature (100 – 200 °C) of polymer layer [5, 26–28] also adds limitations to the subsequent ALD process. Ultrathin metal seed layers (1 – 2 nm), such as Al, can be evaporated on graphene to enable ALD deposition of dielectric materials after full oxidation of Al in the air [29, 30]. However, the obtained Al2O3 are large size clusters with rough surfaces and excessive defects [29, 30]. Metal oxide seed layers can enhance ALD nucleation, but suffer from inherent high density of charged impurities [31, 32]. Therefore, to maximally preserve the intrinsic properties of graphene, searching an ultrathin, uniform, non-covalent-functionalized seed layer for conformal, high-quality dielectric materials growth by ALD is indispensable. Owing to their atomically thin nature, ultra-smooth surface and negligible charged impurities, two dimensional (2D) layered dielectric materials come into the sight. In this work, by DFT calculations, we demonstrate that single-side fluorinated graphene (SFG) [33, 34], a graphene derivative, and hexagonal boron nitride (h-BN) can serve as the ALD seed layer for H2O precursor based dielectric material growth. This is because SFG and h-BN display considerably stronger physical adsorption of H2O molecule than pristine graphene. With ultrathin 2D crystal structure and sizable bandgap [33–39], SFG or h-BN could be employed as the initial dielectric layer atop graphene to offer trap-free interfaces without dangling bonds, contributing to the preservation of intrinsic graphene properties. Large-area SFG has been obtained by directly fluorinating graphene using XeF2 or SF6, etc [33, 34, 40]. By fluorinating the top layer of a bilayer or multilayer graphene, a van der Waals (vdW) graphene-SFG heterostructure [41] is naturally formed with minimum interfacial defects and residues. Due to relatively simple fabrication process [33, 34, 40] and improved interface [41] between graphene and SFG, the graphene-SFG heterostructure is expected to have appealing advantages similar to graphene-h-BN heterostructures. However, h-BN demands additional preparation process with thickness control and complicated, time consuming transfer method onto graphene film. In addition, SFG exhibits enhanced optical transparency with respect to graphene (having transparency of up to ∼99.5% in the visible spectrum) [34]. Fluorinated graphene also inherits the mechanical strength of graphene (sustaining strains of 15%) [34]. Thus, graphene-SFG heterostructure may also be a viable choice for transparent and flexible devices.
In this work, we take one of the most reliable ALD processes, depositing Al2O3 with trimethylaluminum (TMA) and H2O as precursors, as the demonstrating example. Our vdW inclusive DFT calculations predict that monomer H2O molecules have remarkably large physical adsorption energy on SFG and h-BN in contrast to pristine graphene, increasing by 31% and 13%, respectively. Further study shows that SFG and h-BN have stronger electrostatic interactions with H2O monomer than graphene, giving a theoretical support of utilizing SFG and h-BN as the ALD seed layers for H2O precursor based dielectric material growth. The investigation of the electronic property of H2O/TMA adsorbates on these vdW heterostructures demonstrates that h-BN helps preserving the intrinsic graphene properties, while the energy bands of the adsorbates (H2O/TMA) are shifted by SFG, leading to slight n-doping in graphene. Therefore, by choosing SFG or h-BN as a seed layer, one can also meet various electronic doping needs without introducing additional process steps. The Vienna ab initio simulation package (VASP) [42–44] with projector augmented wave potentials [45] is used for our first-principles calculations. A plane-wave cutoff of 400 eV is employed. Only the gamma point is used to sample the Brillouin zone [46, 47]. The vacuum spacing between the slabs is set at 15 Å [46, 47] and a dipole correction along the direction perpendicular to the surface is applied to avoid spurious interactions between periodic images of the slabs [48]. Since the generalized gradient approximation (GGA) is known for being unable to capture the longrange correlation adequately, we use Grimme’s DFT-D2 approach [49] to include the weak vdW interaction in the calculations. A Gaussian smearing parameter of 0.05 eV is used to calculate the partial wave occupancies. For comparison with the DFT-D2 method, the pure Perdew–Burke–Ernzerhof (PBE) exchange-correlation functional within GGA calculation [50] is also performed. We choose four monolayer substrates for demonstration: monolayer pristine graphene, double-side fluorinated graphene (DFG), SFG, and h-BN. There are theoretical reports34 proposing the possibility of SFGs with different stable configurations (CnF). We notice that experiments [51] find fluorination ratio of graphene (F/C) depends on XeF2 exposure time and the choice of substrate. Fluorine content declines by 50–80% over several days after fluorination [51]. Here, we only consider the typical C4F structure as an example for DFT calculations. The ALD precursors, namely, H2O and TMA for Al2O3 deposition, are chosen as the adsorption sources. The electronic self-consistency criterion is set to be 10−6 eV. All structures are optimized until the atomic forces are less than 0.03 eV Å−1. Since C4F has a larger unit cell (ten atoms) than graphene, DFG, and h-BN34, surface supercells of (5 × 5), (5 × 5), (5 × 5), and (3 × 3) are used for H2O adsorption on graphene, DFG, h-BN, and C4F, respectively. These supercells are large enough to model adsorption of an isolated water monomer [46]. For TMA adsorption, larger supercells of (10 × 10), (10 × 10), (10 × 10), and (6 × 6) are simulated. Previous works [52, 53] found that local density approximation (LDA) [54] gives good description of the geometries and electronic structures for the graphene-h-BN 2
H Guo et al
Nanotechnology 25 (2014) 355202
Figure 1. The interaction energy (Eint = ETot − ESub − EH2O/TMA ) of (a) H2O/graphene and TMA/ graphene systems as a function of grapheneH2O (TMA) distance and (b) H2O/C4F’s fluorine side and TMA/C4F’s fluorine side systems as a function of C4F’s fluorine side-H2O (TMA) distance within PBE and DFT-D2 methods. The graphene/C4F’s fluorine side-H2O (TMA) distance is defined as the vertical distance between the carbon hexagon central plane/fluorine central plane and the oxygen (aluminum) atom of H2O (TMA) molecule. The insets are side view schematics of H2O/TMA adsorption on graphene/C4F’s fluorine side. The corresponding detail adsorption configurations are shown in figures 2(c), (g), (e), and (i).
heterostructures. We also implement LDA, as parameterized by Ceperley and Alder [55] to calculate the electronic properties of graphene-SFG and graphene-h-BN vdW heterostructures, as well as with H2O or TMA precursor adsorption on surfaces. (4 × 4) graphene supercells and 8 × 8 Monkhorst–Pack k-point meshes [56] are used for LDA band calculations. For the calculation of the density of states (DOS), we use a 20 × 20 × 1 k-point grid and a Gaussian smearing of 0.14 eV. Electron density images are drawn using the Visualization for Electronic and Structural Analysis tool [57]. The interaction energy (Eint) between H2O/TMA molecules and substrates is computed by Eint = ETot − ESub − EH2 O/TMA , where ETot, ESub, and EH2 O/TMA are the energies of the whole adsorption system, isolated substrate, and isolated H2O or TMA molecule, respectively. Figures 1(a) and (b) show the interaction energy (Eint) between H2O/TMA molecule and graphene/C4F as a function of adsorption distance. The adsorption energy (Eads) is defined as the minimum Eint of all possible stable adsorption configurations. Considering various relative locations of oxygen or aluminum atom on graphene, three different highly symmetric adsorption sites, namely, top, bridge, and center (figure 2(a)), are calculated. Besides, four different adsorption configurations for water monomer are also simulated: one-leg, two-leg, parallel, and reverse (figure 2(b)). The DFT-D2 calculated most energetically stable structures are shown in figures 2(c)–(l). It is noted that C4F has fluorine on one side and no fluorine on the other side. By flipping C4F plane upside down, the adsorption configurations are doubled, as shown in figures 2(e), (f), (i), and (j). The adsorption energies for H2O/TMA on different substrates are compared in table 1. The DFT-D2 results predict that the two-leg on center site model (figure 2(c)) is the most stable configuration for H2O/graphene system with adsorption energy of −151 meV. The calculated adsorption energy is close to other reported DFT-D2 (−139 meV) [58] and optB86b-vdW (−140 meV) [59] while larger than the
Figure 2. (a) Three highly symmetric adsorption sites with respect to
2D substrates: top, bridge, and center. (b) Four different adsorption configurations for H2O molecule: One-leg, two-leg, parallel, and reverse. The most energetically stable structures estimated by the DFT-D2 method for the adsorption systems of (c) H2O/graphene, (d) H2O/DFG, (e) H2O/C4F’s non-fluorine side, and (f) H2O/C4F’s fluorine side. (g), (h), (i), and (j) are the most stable structures for the systems of TMA/graphene, TMA/DFG, TMA/C4F’s non-fluorine side, and TMA/C4F’s fluorine side, respectively. (k) and (l) are the most stable configurations for H2O/BN and TMA/BN systems. 3
H Guo et al
Nanotechnology 25 (2014) 355202
among the four types of seed layers (see table 1). Owing to relatively larger physical adsorption energy than graphene for both H2O and TMA precursors, h-BN is also expected to be a good seed layer candidate for ALD dielectric growth. This also explains successful deposition of oxides on h-BN by thermal ALD in [61]. To gain further insights into the adsorption mechanism of H2O/TMA on different 2D substrates, the charge–density difference, defined by Δρ(r) = ρTot(r) − ρSub(r) − ρH2 O (r), is calculated within the DFT-D2, where ρTot(r), ρSub(r), and ρH2 O (r) are the charge densities of the adsorption system, isolated substrates, and isolated H2O molecule, respectively. As shown in figure 3, all the seed layers are polarized when H2O monomers are adsorbed. Thus the electrostatic interaction plays a vital role in the adsorption process. Since the H2O/graphene system has only negligible positive charge (blue region) accumulated around hydrogen atoms of H2O, the polarization of H2O adsorption on graphene (figure 3(a)) is weaker than C4F’s non-fluorine side (figure 3(c)) and its fluorine side (figure 3(d)). This results in a small adsorption energy −33(PBE)/−151(DFT-D2) meV in H2O/graphene compared to the large adsorption energy −93(PBE)/−179 (DFT-D2) meV in H2O/C4F’s fluorine side and −93(PBE)/ −198 (DFT-D2) meV in H2O/C4F’s non-fluorine side (see table 1). However, we do not observe polarization enhancement on H2O/DFG adsorption system (figure 3(b)). This in turn leads to smaller adsorption energy (−25(PBE)/−115 (DFT-D2) meV) on H2O/DFG than on graphene. The charge density differences of TMA adsorption on various substrates are also plotted in figures 3(e)–(h). TMA/graphene and TMA/ C4F’s non-fluorine side systems have the similar charge distributions (figures 3(e) and (g)), agreeing well with the DFTD2 predicted close adsorption energy of −526 and −514 meV, respectively. Figures 3(f)–(h) illustrate both the TMA on DFG and C4F’s fluorine side have relatively weaker polarization, causing lower adsorption energy −391 and −347 meV, respectively. The charge density differences of H2O and TMA adsorption on h-BN are also exhibited in figures 3(i) and (j), showing stronger polarization than on graphene (figures 3(a) and (e)). According to the table 1 results, both PBE and DFT-D2 approaches demonstrate that H2O has much larger adsorption energy on each side of C4F than on graphene and even on hBN. To confirm our prediction, the plane-integrated electron density difference Δn(z) is calculated in figure 4. Here, we show the charge distribution between H2O and monolayer substrates. Besides, the vertical axis z (Å) is normalized for comparison. The blue and yellow regions denote negative and positive charges (−Qn and +Qp), respectively. Since the electrostatic force is proportional to the charge quantity (F ∝ QnQp), a larger total charge region (including negative and positive region), can represent a stronger electrostatic interaction, which dominates the physical adsorption strength. According to this criterion, the electrostatic interaction in the systems of H2O/C4F’s non-fluorine side (figure 4(b)), H2O/ C4F’s fluorine side (figure 4(c)), and H2O/BN (figure 4(d)) are considerably larger than that in the H2O/graphene system.
Table 1. The adsorption energy (Eads) of H2O/TMA on different 2D substrates including graphene, DFG, C4F’s non-fluorine or fluorine side, and BN, calculated by both PBE and DFT-D2 methods.
H2O/TMA adsorption energy (method)
Eads − H2 O (PBE) meV Eads − H2 O (DFTD2) meV Eads-TMA (PBE) meV Eads-TMA (DFTD2) meV
Graphene
DFG
C4F (nonfluorine side)
−33
−25
−93
−93
−42
−151
−115
−198
−179
−170
−42
−41
−60
−57
−43
−526
−391
−514
−347
−600
C4F (fluorine side)
BN
PBE-D (−90 meV) [46] and vdW-DF2C09x results (−78 meV) [47]. Comparing with DFT-D2 results, our PBE calculations predict the One-leg on top site model to be the most energetically stable, with adsorption energy of −33 meV. This is consistent with the previous reported PBE results (−30 to −32 meV) [46, 47, 60]. The verification of adsorption energy of H2O monomer on graphene by our PBE and DFT-D2 calculations enables us to further study the adsorption of H2O/TMA on the surface of fluorinated graphene and h-BN. Table 1 shows H2O monomer has lower adsorption energy on DFG than pristine graphene, indicating that not all of the graphene fluorinated derivatives can enhance the adsorption of H2O. Since the unit cells of graphene and DFG have spatially inverse and central symmetry, respectively, the positive and negative charge centers of unit cells are not effectively separated. Thus, no intrinsic polarization is induced, which is likely the dominant factor of H2O adsorption inertness. By introducing the intrinsic polarization of 2D seed layer, the possibility of H2O adsorption is enhanced. The calculated results for H2O adsorption on SFG and h-BN are compared in table 1. For the two asymmetric sides with and without fluorine, the adsorption energy of H2O on C4F increases by 180% for PBE and 19% – 31% for DFT-D2 compared with graphene. Besides, hBN also shows adsorption energy enhancement (27% for PBE and 13% for DFT-D2) for H2O in contrast to graphene. Notably, from the DFT-D2 results, the non-fluorine side of C4F gives the largest increase (∼50 meV) in adsorption energy of H2O. However, such improvements are not observed for the cases of TMA adsorption on C4F (see the DFT-D2 results in table 1). This is because TMA is a nonpolar molecule in contrast to polar H2O. Therefore, the intrinsic polarization of C4F substrates plays a less important role in adsorption of TMA than H2O. Nevertheless, h-BN exhibits preference to TMA due to polarization between B and N atoms, with the largest adsorption energy of 600 meV 4
H Guo et al
Nanotechnology 25 (2014) 355202
Figure 3. 2D slices of the charge density difference (obtained with DFT-D2 method) for the adsorption systems of (a) H2O/graphene, (b) H2O/DFG, (c) H2O/C4F’s non-fluorine side, (d) H2O/C4F’s fluorine side, (e) TMA/graphene, (f) TMA/DFG, (g) TMA/C4F’s non-fluorine side, (h) TMA/C4F’s fluorine side, (i) H2O/BN, and (j) TMA/BN.
This explains the corresponding larger adsorption energy of H2O adsorbate on SFG and h-BN than on pristine graphene, as shown in table 1. While meeting the requirements for dielectric material deposition, one important challenge for current ALD seed layers on graphene is to minimize the unexpected lattice damage and doping effects to the underlying graphene. To further understand the potential of SFG and h-BN seed layers
in preserving the intrinsic graphene properties, we choose the graphene-SFG and graphene-h-BN vdW heterostructures as the testing models to investigate the differences before and after H2O or TMA precursor adsorption. Since the LDA is proven to give better description of the interlayer distance and energy gap opening in graphene-h-BN heterostructure than the GGA method ([49] and [50]), we choose the LDA method to recalculate the geometries, band structures, and DOS (see 5
H Guo et al
Nanotechnology 25 (2014) 355202
Figure 4. Plane-integrated electron density difference Δn(z) for four different adsorption systems of (a) H2O/graphene, (b) H2O/C4F’s nonfluorine side, (c) H2O/C4F’s fluorine side, and (d) H2O/BN. For comparison, only the charge distributions between H2O and substrates are shown and the vertical position z (Å) is normalized. The yellow (blue) color represents the electron accumulation (dissipation). The electrostatic interaction in (b), (c), and (d) is stronger than in (a).
figure 5). Due to weak physical adsorption, the contribution bands and DOS of H2O or TMA adsorbates are far away from the Dirac point (the contribution bands of H2O and TMA are denoted as red and blue dots, respectively), having negligible influence on the intrinsic electronic properties of graphene (figures 5(a) and (f)). Similarly, because of the weak vdW interaction, when forming heterostructures with graphene, hBN and C4F cause slight perturbation to graphene electronic properties, opening a bandgap of 56 meV and 30 meV, respectively (figures 5(b) and (c)). Thanks to an in-plane intrinsic polarization, h-BN displays solid ability to preserve the electronic properties of the underlying graphene. Comparing with the exposed graphene-h-BN heterostructure (figure 5(b)), the H2O and TMA adsorbates (figures 5(d) and (g)) on h-BN make negligible change in the vicinity of graphene Dirac point. However, unlike h-BN, C4F has a vertical intrinsic polarization (please see our previous work in [62]) perpendicular to its basal plane and can interact with the adsorbates. Figures 5(e) and (h) show the H2O or TMA’s contribution bands and DOS will shift towards the Fermi energy when adsorption on graphene-SFG vdW heterostructures, leading to n-doping on graphene. In figure 5(i), H2O/DFG/graphene adsorption system is also investigated. In contrast to H2O/C4F/graphene, contribution bands of H2O adsorbate and its DOS shift little among the H2O/DFG/graphene. Since central-symmetry property of DFG induces no polarization, the comparison between H2O/C4F/graphene and H2O/DFG/graphene systems confirms the doping effects are caused by the vertical intrinsic polarization of C4F. Although we only study H2O or TMA monomer adsorption here, the doping effects are expected to be observed in any practical dielectric material deposition. This is because the vertical electric field induced by intrinsic polarization of C4F can interact with impurity or defective states of dielectric materials and change their hybridization with graphene states, resulting in graphene doping. This is somewhat similar to the
explanation of water-adsorption induced doping effects in substrate-supported graphene in [63]. Doping effect considered above is under the LDA optimized stacking order, i.e. AB, while for AA-stacking, similar n-type doping effects for SFG/G substrate are found, whereas graphene is well protected under h-BN. Besides fluorinated graphene, other preprocessing methods like hydrogenated graphene (HG) and defective graphene are also considered. HG is found hydrophobic [64, 65] and unstable during ALD process [25, 66], while graphene with defects provides a large number of nucleation centers for ALD precursors [15, 18]. For other ALD high-κ dielectric growth, for instance HfO2 or ZrO2, typical precursors like HfCl4 or ZrCl4 [67, 68] are simply considered. Due to the polarization nature of Hf–Cl or Zr–Cl bond, if we put the molecule on the boron site of h-BN or carbon side of SFG, strong adhesion will also be observed. For doping effect, results are the same with H2O or TMA on h-BN/G substrate, where h-BN still plays a key role in preserving graphene’s intrinsic properties. However, unlike H2O and TMA on SFG/G substrate, no obvious doping is found and molecular orbitals are deep inside the valence and conduction band of substrate. By using the vdW interaction inclusive DFT-D2 calculation, we demonstrate that H2O monomer exhibits a substantial increase in the adsorption energy on SFG and h-BN, in contrast to the pristine graphene. Due to the relatively strong electrostatic interaction with H2O adsorbates observed in our simulations, we predict that SFG and h-BN will be good candidates as 2D seed layers for ALD dielectric materials growth on graphene. When adsorbed on the SFG surface of graphene-SFG heterostructure, energy bands of H2O or TMA precursors are found hybridized with graphene near the Dirac point, leading to n-type doping in graphene. However, when H2O or TMA adsorbed on the h-BN surface of graphene-h-BN heterostructures, no doping effect is observed. Thus, one can choose SFG or h-BN as a seed layer to meet requirements for various electronic doping. The proposed 6
H Guo et al
Nanotechnology 25 (2014) 355202
Figure 5. The schematic structures (left), energy band (middle), and density of states (right) for the systems of (a) H2O/graphene, (b) BN/ graphene, (c) C4F/graphene, (d) H2O/BN/graphene, (e) H2O/ C4F/graphene, (f) TMA/graphene, (g) TMA/BN/graphene, (h) TMA/C4F/ graphene, and (i) H2O/ DFG/graphene. The contribution bands of H2O and TMA monomers are marked as red and blue dots in each plot of band structures, respectively. The red or blue arrows in (e) or (h) denote that the H2O or TMA contribution band is shifted towards the vicinity of Fermi energy. The green arrows indicate the corresponding DOS peaks.
Acknowledgment
SFG and h-BN seed layers for graphene vdW heterostructures are expected not to be limited to the growth of Al2O3 and may also find broad applications in the integration of high-κ dielectric materials on graphene.
The authors would like to thank Profs E Pop, C Zhou, I Hamada, H Wong, and Dr J Ma for helpful discussion and 7
H Guo et al
Nanotechnology 25 (2014) 355202
comments. This work is supported by National Science Foundation of China (Grant Nos. 61006077 and 61274123), ZJ-NSF (LR12F04001), and US National Science Foundation (Grants ECCS-1002228, ECCS-1028267, CMMI-1162312). Y Xu is supported by ZJU Cyber Scholarship, and Cyrus Tang Center for Sensor Materials and Applications, and Visiting-by-Fellowship of Churchill College, University of Cambridge. T Hasan is supported by the Churchill College, Cambridge and The Royal Academy of Engineering (Graphlex). The authors thank the official license of VASP package from the University of Vienna.
[28] Farmer D B, Chiu H-Y, Lin Y-M, Jenkins K A, Xia F and Avouris P 2009 Nano Lett. 9 4474 [29] Kim S, Nah J, Jo I, Shahrjerdi D, Colombo L, Yao Z, Tutuc E and Banerjee S K 2009 Appl. Phys. Lett. 94 062107 [30] Pirkle A, Wallace R M and Colombo L 2009 Appl. Phys. Lett. 95 133106 [31] Hsu A, Wang H, Kim K K, Kong J and Palacios T 2011 Japan. J. Appl. Phys. 50 070114 [32] Hollander M J et al 2011 Nano Lett. 11 3601 [33] Robinson J T et al 2010 Nano Lett. 10 3001 [34] Nair R R et al 2010 Small 6 2877 [35] Şahin H, Topsakal M and Ciraci S 2011 Phys. Rev. B 83 115432 [36] Wei W and Jacob T 2013 Phys. Rev. B 87 115431 [37] Karlický F, Kumara Ramanatha Datta K, Otyepka M and Zbořil R 2013 ACS Nano 7 6434 [38] Moon J S et al 2013 IEEE Electron Device Lett. 34 1190 [39] Pacile D, Meyer J C, Girit C O and Zettl A 2008 Appl. Phys. Lett. 92 133107 [40] Baraket M, Walton S G, Lock E H, Robinson J T and Perkins F K 2010 Appl. Phys. Lett. 96 231501 [41] Geim A K and Grigorieva I V 2013 Nature 499 419 [42] Kresse G and Hafner J 1993 Phys. Rev. B 47 558 [43] Kresse G and Furthmüller J 1996 Phys. Rev. B 54 11169 [44] Kresse G and Joubert D 1999 Phys. Rev. B 59 1758 [45] Blöchl P E 1994 Phys. Rev. B 50 17953 [46] Ma J, Michaelides A, Alfè D, Schimka L, Kresse G and Wang E 2011 Phys. Rev. B 84 033402 [47] Hamada I 2012 Phys. Rev. B 86 195436 [48] Neugebauer J and Scheffler M 1992 Phys. Rev. B 46 16067 [49] Grimme S 2006 J. Comput. Chem. 27 1787 [50] Perdew J P, Burke K and Ernzerhof M 1996 Phys. Rev. Lett. 77 3865 [51] Stine R, Lee W-K, Whitener K E, Robinson J T and Sheehan P E 2013 Nano Lett. 13 4311 [52] Giovannetti G, Khomyakov P, Brocks G, Kelly P and van den Brink J 2007 Phys. Rev. B 76 073103 [53] Ramasubramaniam A, Naveh D and Towe E 2011 Nano Lett. 11 1070 [54] Perdew J P and Zunger A 1981 Phys. Rev. B 23 5048 [55] Ceperley D M and Alder B J 1980 Phys. Rev. Lett. 45 566 [56] Monkhorst H J and Pack J D 1976 Phys. Rev. B 13 5188 [57] Momma K and Izumi F 2008 J. Appl. Crystallogr. 41 653 [58] Voloshina E, Usvyat D, Schütz M, Dedkov Y and Paulus B 2011 Phys. Chem. Chem. Phys. 13 12041 [59] Li X, Feng J, Wang E, Meng S, Klimeš J and Michaelides A 2012 Phys. Rev. B 85 085425 [60] Kurita T, Okada S and Oshiyama A 2007 Phys. Rev. B 75 205424 [61] Liu H, Xu K, Zhang X and Ye P D 2012 Appl. Phys. Lett. 100 152115 [62] Xu Y, Liu Y, Chen H, Lin X, Lin S, Yu B and Luo J 2012 J. Mater. Chem. 22 23821 [63] Wehling T O, Lichtenstein A I and Katsnelson M I 2008 Appl. Phys. Lett. 93 202110 [64] Hare E F, Shafrin E G and Zisman W A 1954 J. Phys. Chem. 58 236 [65] Zheng L, Li Z, Bourdo S, Watanabe F, Ryerson C C and Biris A S 2011 Chem. Commun. 47 1213 [66] Ryu S, Han M Y, Maultzsch J, Heinz T F, Kim P, Steigerwald M L and Brus L E 2008 Nano Lett. 8 4597 [67] Müller J, Böscke T S, Schröder U, Reinicke M, Oberbeck L, Zhou D, Weinreich W, Kücher P, Lemberger M and Frey L 2009 Microelectron. Eng. 86 1818 [68] Chen R, Kim H, McIntyre P C and Bent S F 2004 Appl. Phys. Lett. 84 4017
Reference [1] Novoselov K S 2004 Science 306 666 [2] Lemme M C, Echtermeyer T J, Baus M and Kurz H 2007 IEEE Electron Device Lett. 28 282 [3] Wang X, Ouyang Y, Li X, Wang H, Guo J and Dai H 2008 Phys. Rev. Lett. 100 206803 [4] Meric I, Han M Y, Young A F, Ozyilmaz B, Kim P and Shepard K L 2008 Nat. Nanotechnology 3 654 [5] Lin Y-M, Dimitrakopoulos C, Jenkins K A, Farmer D B, Chiu H-Y, Grill A and Avouris P 2010 Science 327 662 [6] Schwierz F 2010 Nat. Nanotechnology 5 487 [7] Liu M, Yin X, Ulin-Avila E, Geng B, Zentgraf T, Ju L, Wang F and Zhang X 2011 Nature 474 64 [8] Bao Q, Zhang H, Wang B, Ni Z, Lim C H Y X, Wang Y, Tang D Y and Loh K P 2011 Nat. Photonics 5 411 [9] Bao Q and Loh K P 2012 ACS Nano 6 3677 [10] Koppens F H L, Chang D E and García de Abajo F J 2011 Nano Lett. 11 3370 [11] Ju L et al 2011 Nat. Nanotechnology 6 630 [12] Grigorenko A N, Polini M and Novoselov K S 2012 Nat. Photonics 6 749 [13] Liao L and Duan X 2010 Mater. Sci. Eng. R 70 354 [14] Kim J and Jandhyala S 2013 Thin Solid Films 546 85 [15] Garces N Y, Wheeler V D and Gaskill D K 2012 J. Vac. Sci. Technol. B 30 030801 [16] Xuan Y, Wu Y Q, Shen T, Qi M, Capano M A, Cooper J A and Ye P D 2008 Appl. Phys. Lett. 92 013101 [17] Lee B, Park S-Y, Kim H-C, Cho K, Vogel E M, Kim M J, Wallace R M and Kim J 2008 Appl. Phys. Lett. 92 203102 [18] Wang X, Tabakman S M and Dai H 2008 J. Am. Chem. Soc. 130 8152 [19] Jandhyala S et al 2012 ACS Nano 6 2722 [20] Williams J R, DiCarlo L and Marcus C M 2007 Science 317 638 [21] Lin Y-M, Jenkins K A, Valdes-Garcia A, Small J P, Farmer D B and Avouris P 2009 Nano Lett. 9 422 [22] Nayfeh O M, Marr T and Dubey M 2011 IEEE Electron Device Lett. 32 473 [23] Pirkle A, McDonnell S, Lee B, Kim J, Colombo L and Wallace R M 2010 Appl. Phys. Lett. 97 082901 [24] Garces N Y, Wheeler V D, Hite J K, Jernigan G G, Tedesco J L, Nepal N, Eddy C R and Gaskill D K 2011 J. Appl. Phys. 109 124304 [25] Wheeler V, Garces N, Nyakiti L, Myers-Ward R, Jernigan G, Culbertson J, Eddy C and Kurt Gaskill D 2012 Carbon 50 2307 [26] Huard B, Sulpizio J, Stander N, Todd K, Yang B and Goldhaber-Gordon D 2007 Phys. Rev. Lett. 98 236803 [27] Zheng Y, Ni G-X, Toh C-T, Tan C-Y, Yao K and Özyilmaz B 2010 Phys. Rev. Lett. 105 166602
8
Search
Collections
Journals
About
Contact us
My IOPscience
Fluorinated graphene and hexagonal boron nitride as ALD seed layers for graphene-based van der Waals heterostructures
This content has been downloaded from IOPscience. Please scroll down to see the full text. 2014 Nanotechnology 25 355202 (http://iopscience.iop.org/0957-4484/25/35/355202) View the table of contents for this issue, or go to the journal homepage for more
Download details: IP Address: 128.220.8.15 This content was downloaded on 13/01/2015 at 15:46
Please note that terms and conditions apply.
Nanotechnology Nanotechnology 25 (2014) 355202 (8pp)
doi:10.1088/0957-4484/25/35/355202
Fluorinated graphene and hexagonal boron nitride as ALD seed layers for graphenebased van der Waals heterostructures Hongwei Guo1, Yunlong Liu1, Yang Xu1, Nan Meng1, Hongtao Wang2, Tawfique Hasan3, Xinran Wang4, Jikui Luo1 and Bin Yu5 1
Department of Information Science and Electronic Engineering, Zhejiang University, Hangzhou 310027, People’s Republic of China 2 Institute of Applied Mechanics, Zhejiang University, Hangzhou 310027, People’s Republic of China 3 Department of Engineering, University of Cambridge, Cambridge CB3 0FA, UK 4 National Laboratory of Microstructures and School of Electronic Science and Engineering, Nanjing University 210093, People’s Republic of China 5 College of Nanoscale Science and Engineering, State University of New York, Albany, New York 12203, USA E-mail: [email protected] Received 6 May 2014, revised 28 June 2014 Accepted for publication 14 July 2014 Published 12 August 2014 Abstract
Ultrathin dielectric materials prepared by atomic-layer-deposition (ALD) technology are commonly used in graphene electronics. Using the first-principles density functional theory calculations with van der Waals (vdW) interactions included, we demonstrate that single-side fluorinated graphene (SFG) and hexagonal boron nitride (h-BN) exhibit large physical adsorption energy and strong electrostatic interactions with H2O-based ALD precursors, indicating their potential as the ALD seed layer for dielectric growth on graphene. In grapheneSFG vdW heterostructures, graphene is n-doped after ALD precursor adsorption on the SFG surface caused by vertical intrinsic polarization of SFG. However, graphene-h-BN vdW heterostructures help preserving the intrinsic characteristics of the underlying graphene due to inplane intrinsic polarization of h-BN. By choosing SFG or BN as the ALD seed layer on the basis of actual device design needs, the graphene vdW heterostructures may find applications in lowdimensional electronics. Keywords: seed layer, ALD, dielectric growth, fluorinated graphene, hexagonal boron nitride (Some figures may appear in colour only in the online journal) A large number of graphene-based devices, such as field effect transistors [1–6], waveguides [7–9], and plasmonic structures [10–12] require integration of ultrathin high-κ dielectric on graphene to implement the desired functionalities. This is because high-κ dielectrics can enhance dielectric screening effect and increase the carrier mobility in graphene14. Although atomic-layer-deposition (ALD) serves reliably for thin dielectric growth on semiconductors, direct deposition of dielectric materials on graphene by thermal ALD is still challenging [13–15]. This is because the conjugated π-bond surface of pristine graphene is hydrophobic and chemically inert to most of ALD precursor molecules 0957-4484/14/355202+08$33.00
[16–18]. Functionalizing the graphene basal plane is one of the viable choices to achieve uniform ALD dielectric materials [17–22]. Ozone (O3) [17–19] or NO2 [20, 21] vapor phase functionalization, and remote plasma-assisted ALD process [22] allow deposition of ultrathin dielectric materials, however, they cause significant degradation in graphene lattice and decrease of carrier mobility. Wet chemistry functionalization modifies the graphene surface with minimum disruption of the intrinsic properties. This results in low dielectric constant in the ALD layer and gives rise to scalability issue (∼31 nm minimum dielectric thickness) [23, 24]. Dry chemical approaches, such as fluorine functionalization 1
© 2014 IOP Publishing Ltd Printed in the UK
H Guo et al
Nanotechnology 25 (2014) 355202
[25], can promote ALD process but are limited to epitaxial graphene [26]. An alternative approach is to use a non-covalent functionalized seed layer to catalytically react with ALD precursors and subsequently nucleate the growth of the dielectric materials. In this strategy, three different types of seeding layers have recently been explored: polymer buffer[5, 26–28], metal- [29, 30], and metal oxide- [31, 32] seed layers. Despite successfully enabling the deposition of dielectric materials [14, 15], these seeding layers have their own limitations. Polymer buffer seed layers can be used to obtain uniform and conformal dielectric materials [5, 26–28], but with low dielectric constants (∼2). The low decomposition temperature (100 – 200 °C) of polymer layer [5, 26–28] also adds limitations to the subsequent ALD process. Ultrathin metal seed layers (1 – 2 nm), such as Al, can be evaporated on graphene to enable ALD deposition of dielectric materials after full oxidation of Al in the air [29, 30]. However, the obtained Al2O3 are large size clusters with rough surfaces and excessive defects [29, 30]. Metal oxide seed layers can enhance ALD nucleation, but suffer from inherent high density of charged impurities [31, 32]. Therefore, to maximally preserve the intrinsic properties of graphene, searching an ultrathin, uniform, non-covalent-functionalized seed layer for conformal, high-quality dielectric materials growth by ALD is indispensable. Owing to their atomically thin nature, ultra-smooth surface and negligible charged impurities, two dimensional (2D) layered dielectric materials come into the sight. In this work, by DFT calculations, we demonstrate that single-side fluorinated graphene (SFG) [33, 34], a graphene derivative, and hexagonal boron nitride (h-BN) can serve as the ALD seed layer for H2O precursor based dielectric material growth. This is because SFG and h-BN display considerably stronger physical adsorption of H2O molecule than pristine graphene. With ultrathin 2D crystal structure and sizable bandgap [33–39], SFG or h-BN could be employed as the initial dielectric layer atop graphene to offer trap-free interfaces without dangling bonds, contributing to the preservation of intrinsic graphene properties. Large-area SFG has been obtained by directly fluorinating graphene using XeF2 or SF6, etc [33, 34, 40]. By fluorinating the top layer of a bilayer or multilayer graphene, a van der Waals (vdW) graphene-SFG heterostructure [41] is naturally formed with minimum interfacial defects and residues. Due to relatively simple fabrication process [33, 34, 40] and improved interface [41] between graphene and SFG, the graphene-SFG heterostructure is expected to have appealing advantages similar to graphene-h-BN heterostructures. However, h-BN demands additional preparation process with thickness control and complicated, time consuming transfer method onto graphene film. In addition, SFG exhibits enhanced optical transparency with respect to graphene (having transparency of up to ∼99.5% in the visible spectrum) [34]. Fluorinated graphene also inherits the mechanical strength of graphene (sustaining strains of 15%) [34]. Thus, graphene-SFG heterostructure may also be a viable choice for transparent and flexible devices.
In this work, we take one of the most reliable ALD processes, depositing Al2O3 with trimethylaluminum (TMA) and H2O as precursors, as the demonstrating example. Our vdW inclusive DFT calculations predict that monomer H2O molecules have remarkably large physical adsorption energy on SFG and h-BN in contrast to pristine graphene, increasing by 31% and 13%, respectively. Further study shows that SFG and h-BN have stronger electrostatic interactions with H2O monomer than graphene, giving a theoretical support of utilizing SFG and h-BN as the ALD seed layers for H2O precursor based dielectric material growth. The investigation of the electronic property of H2O/TMA adsorbates on these vdW heterostructures demonstrates that h-BN helps preserving the intrinsic graphene properties, while the energy bands of the adsorbates (H2O/TMA) are shifted by SFG, leading to slight n-doping in graphene. Therefore, by choosing SFG or h-BN as a seed layer, one can also meet various electronic doping needs without introducing additional process steps. The Vienna ab initio simulation package (VASP) [42–44] with projector augmented wave potentials [45] is used for our first-principles calculations. A plane-wave cutoff of 400 eV is employed. Only the gamma point is used to sample the Brillouin zone [46, 47]. The vacuum spacing between the slabs is set at 15 Å [46, 47] and a dipole correction along the direction perpendicular to the surface is applied to avoid spurious interactions between periodic images of the slabs [48]. Since the generalized gradient approximation (GGA) is known for being unable to capture the longrange correlation adequately, we use Grimme’s DFT-D2 approach [49] to include the weak vdW interaction in the calculations. A Gaussian smearing parameter of 0.05 eV is used to calculate the partial wave occupancies. For comparison with the DFT-D2 method, the pure Perdew–Burke–Ernzerhof (PBE) exchange-correlation functional within GGA calculation [50] is also performed. We choose four monolayer substrates for demonstration: monolayer pristine graphene, double-side fluorinated graphene (DFG), SFG, and h-BN. There are theoretical reports34 proposing the possibility of SFGs with different stable configurations (CnF). We notice that experiments [51] find fluorination ratio of graphene (F/C) depends on XeF2 exposure time and the choice of substrate. Fluorine content declines by 50–80% over several days after fluorination [51]. Here, we only consider the typical C4F structure as an example for DFT calculations. The ALD precursors, namely, H2O and TMA for Al2O3 deposition, are chosen as the adsorption sources. The electronic self-consistency criterion is set to be 10−6 eV. All structures are optimized until the atomic forces are less than 0.03 eV Å−1. Since C4F has a larger unit cell (ten atoms) than graphene, DFG, and h-BN34, surface supercells of (5 × 5), (5 × 5), (5 × 5), and (3 × 3) are used for H2O adsorption on graphene, DFG, h-BN, and C4F, respectively. These supercells are large enough to model adsorption of an isolated water monomer [46]. For TMA adsorption, larger supercells of (10 × 10), (10 × 10), (10 × 10), and (6 × 6) are simulated. Previous works [52, 53] found that local density approximation (LDA) [54] gives good description of the geometries and electronic structures for the graphene-h-BN 2
H Guo et al
Nanotechnology 25 (2014) 355202
Figure 1. The interaction energy (Eint = ETot − ESub − EH2O/TMA ) of (a) H2O/graphene and TMA/ graphene systems as a function of grapheneH2O (TMA) distance and (b) H2O/C4F’s fluorine side and TMA/C4F’s fluorine side systems as a function of C4F’s fluorine side-H2O (TMA) distance within PBE and DFT-D2 methods. The graphene/C4F’s fluorine side-H2O (TMA) distance is defined as the vertical distance between the carbon hexagon central plane/fluorine central plane and the oxygen (aluminum) atom of H2O (TMA) molecule. The insets are side view schematics of H2O/TMA adsorption on graphene/C4F’s fluorine side. The corresponding detail adsorption configurations are shown in figures 2(c), (g), (e), and (i).
heterostructures. We also implement LDA, as parameterized by Ceperley and Alder [55] to calculate the electronic properties of graphene-SFG and graphene-h-BN vdW heterostructures, as well as with H2O or TMA precursor adsorption on surfaces. (4 × 4) graphene supercells and 8 × 8 Monkhorst–Pack k-point meshes [56] are used for LDA band calculations. For the calculation of the density of states (DOS), we use a 20 × 20 × 1 k-point grid and a Gaussian smearing of 0.14 eV. Electron density images are drawn using the Visualization for Electronic and Structural Analysis tool [57]. The interaction energy (Eint) between H2O/TMA molecules and substrates is computed by Eint = ETot − ESub − EH2 O/TMA , where ETot, ESub, and EH2 O/TMA are the energies of the whole adsorption system, isolated substrate, and isolated H2O or TMA molecule, respectively. Figures 1(a) and (b) show the interaction energy (Eint) between H2O/TMA molecule and graphene/C4F as a function of adsorption distance. The adsorption energy (Eads) is defined as the minimum Eint of all possible stable adsorption configurations. Considering various relative locations of oxygen or aluminum atom on graphene, three different highly symmetric adsorption sites, namely, top, bridge, and center (figure 2(a)), are calculated. Besides, four different adsorption configurations for water monomer are also simulated: one-leg, two-leg, parallel, and reverse (figure 2(b)). The DFT-D2 calculated most energetically stable structures are shown in figures 2(c)–(l). It is noted that C4F has fluorine on one side and no fluorine on the other side. By flipping C4F plane upside down, the adsorption configurations are doubled, as shown in figures 2(e), (f), (i), and (j). The adsorption energies for H2O/TMA on different substrates are compared in table 1. The DFT-D2 results predict that the two-leg on center site model (figure 2(c)) is the most stable configuration for H2O/graphene system with adsorption energy of −151 meV. The calculated adsorption energy is close to other reported DFT-D2 (−139 meV) [58] and optB86b-vdW (−140 meV) [59] while larger than the
Figure 2. (a) Three highly symmetric adsorption sites with respect to
2D substrates: top, bridge, and center. (b) Four different adsorption configurations for H2O molecule: One-leg, two-leg, parallel, and reverse. The most energetically stable structures estimated by the DFT-D2 method for the adsorption systems of (c) H2O/graphene, (d) H2O/DFG, (e) H2O/C4F’s non-fluorine side, and (f) H2O/C4F’s fluorine side. (g), (h), (i), and (j) are the most stable structures for the systems of TMA/graphene, TMA/DFG, TMA/C4F’s non-fluorine side, and TMA/C4F’s fluorine side, respectively. (k) and (l) are the most stable configurations for H2O/BN and TMA/BN systems. 3
H Guo et al
Nanotechnology 25 (2014) 355202
among the four types of seed layers (see table 1). Owing to relatively larger physical adsorption energy than graphene for both H2O and TMA precursors, h-BN is also expected to be a good seed layer candidate for ALD dielectric growth. This also explains successful deposition of oxides on h-BN by thermal ALD in [61]. To gain further insights into the adsorption mechanism of H2O/TMA on different 2D substrates, the charge–density difference, defined by Δρ(r) = ρTot(r) − ρSub(r) − ρH2 O (r), is calculated within the DFT-D2, where ρTot(r), ρSub(r), and ρH2 O (r) are the charge densities of the adsorption system, isolated substrates, and isolated H2O molecule, respectively. As shown in figure 3, all the seed layers are polarized when H2O monomers are adsorbed. Thus the electrostatic interaction plays a vital role in the adsorption process. Since the H2O/graphene system has only negligible positive charge (blue region) accumulated around hydrogen atoms of H2O, the polarization of H2O adsorption on graphene (figure 3(a)) is weaker than C4F’s non-fluorine side (figure 3(c)) and its fluorine side (figure 3(d)). This results in a small adsorption energy −33(PBE)/−151(DFT-D2) meV in H2O/graphene compared to the large adsorption energy −93(PBE)/−179 (DFT-D2) meV in H2O/C4F’s fluorine side and −93(PBE)/ −198 (DFT-D2) meV in H2O/C4F’s non-fluorine side (see table 1). However, we do not observe polarization enhancement on H2O/DFG adsorption system (figure 3(b)). This in turn leads to smaller adsorption energy (−25(PBE)/−115 (DFT-D2) meV) on H2O/DFG than on graphene. The charge density differences of TMA adsorption on various substrates are also plotted in figures 3(e)–(h). TMA/graphene and TMA/ C4F’s non-fluorine side systems have the similar charge distributions (figures 3(e) and (g)), agreeing well with the DFTD2 predicted close adsorption energy of −526 and −514 meV, respectively. Figures 3(f)–(h) illustrate both the TMA on DFG and C4F’s fluorine side have relatively weaker polarization, causing lower adsorption energy −391 and −347 meV, respectively. The charge density differences of H2O and TMA adsorption on h-BN are also exhibited in figures 3(i) and (j), showing stronger polarization than on graphene (figures 3(a) and (e)). According to the table 1 results, both PBE and DFT-D2 approaches demonstrate that H2O has much larger adsorption energy on each side of C4F than on graphene and even on hBN. To confirm our prediction, the plane-integrated electron density difference Δn(z) is calculated in figure 4. Here, we show the charge distribution between H2O and monolayer substrates. Besides, the vertical axis z (Å) is normalized for comparison. The blue and yellow regions denote negative and positive charges (−Qn and +Qp), respectively. Since the electrostatic force is proportional to the charge quantity (F ∝ QnQp), a larger total charge region (including negative and positive region), can represent a stronger electrostatic interaction, which dominates the physical adsorption strength. According to this criterion, the electrostatic interaction in the systems of H2O/C4F’s non-fluorine side (figure 4(b)), H2O/ C4F’s fluorine side (figure 4(c)), and H2O/BN (figure 4(d)) are considerably larger than that in the H2O/graphene system.
Table 1. The adsorption energy (Eads) of H2O/TMA on different 2D substrates including graphene, DFG, C4F’s non-fluorine or fluorine side, and BN, calculated by both PBE and DFT-D2 methods.
H2O/TMA adsorption energy (method)
Eads − H2 O (PBE) meV Eads − H2 O (DFTD2) meV Eads-TMA (PBE) meV Eads-TMA (DFTD2) meV
Graphene
DFG
C4F (nonfluorine side)
−33
−25
−93
−93
−42
−151
−115
−198
−179
−170
−42
−41
−60
−57
−43
−526
−391
−514
−347
−600
C4F (fluorine side)
BN
PBE-D (−90 meV) [46] and vdW-DF2C09x results (−78 meV) [47]. Comparing with DFT-D2 results, our PBE calculations predict the One-leg on top site model to be the most energetically stable, with adsorption energy of −33 meV. This is consistent with the previous reported PBE results (−30 to −32 meV) [46, 47, 60]. The verification of adsorption energy of H2O monomer on graphene by our PBE and DFT-D2 calculations enables us to further study the adsorption of H2O/TMA on the surface of fluorinated graphene and h-BN. Table 1 shows H2O monomer has lower adsorption energy on DFG than pristine graphene, indicating that not all of the graphene fluorinated derivatives can enhance the adsorption of H2O. Since the unit cells of graphene and DFG have spatially inverse and central symmetry, respectively, the positive and negative charge centers of unit cells are not effectively separated. Thus, no intrinsic polarization is induced, which is likely the dominant factor of H2O adsorption inertness. By introducing the intrinsic polarization of 2D seed layer, the possibility of H2O adsorption is enhanced. The calculated results for H2O adsorption on SFG and h-BN are compared in table 1. For the two asymmetric sides with and without fluorine, the adsorption energy of H2O on C4F increases by 180% for PBE and 19% – 31% for DFT-D2 compared with graphene. Besides, hBN also shows adsorption energy enhancement (27% for PBE and 13% for DFT-D2) for H2O in contrast to graphene. Notably, from the DFT-D2 results, the non-fluorine side of C4F gives the largest increase (∼50 meV) in adsorption energy of H2O. However, such improvements are not observed for the cases of TMA adsorption on C4F (see the DFT-D2 results in table 1). This is because TMA is a nonpolar molecule in contrast to polar H2O. Therefore, the intrinsic polarization of C4F substrates plays a less important role in adsorption of TMA than H2O. Nevertheless, h-BN exhibits preference to TMA due to polarization between B and N atoms, with the largest adsorption energy of 600 meV 4
H Guo et al
Nanotechnology 25 (2014) 355202
Figure 3. 2D slices of the charge density difference (obtained with DFT-D2 method) for the adsorption systems of (a) H2O/graphene, (b) H2O/DFG, (c) H2O/C4F’s non-fluorine side, (d) H2O/C4F’s fluorine side, (e) TMA/graphene, (f) TMA/DFG, (g) TMA/C4F’s non-fluorine side, (h) TMA/C4F’s fluorine side, (i) H2O/BN, and (j) TMA/BN.
This explains the corresponding larger adsorption energy of H2O adsorbate on SFG and h-BN than on pristine graphene, as shown in table 1. While meeting the requirements for dielectric material deposition, one important challenge for current ALD seed layers on graphene is to minimize the unexpected lattice damage and doping effects to the underlying graphene. To further understand the potential of SFG and h-BN seed layers
in preserving the intrinsic graphene properties, we choose the graphene-SFG and graphene-h-BN vdW heterostructures as the testing models to investigate the differences before and after H2O or TMA precursor adsorption. Since the LDA is proven to give better description of the interlayer distance and energy gap opening in graphene-h-BN heterostructure than the GGA method ([49] and [50]), we choose the LDA method to recalculate the geometries, band structures, and DOS (see 5
H Guo et al
Nanotechnology 25 (2014) 355202
Figure 4. Plane-integrated electron density difference Δn(z) for four different adsorption systems of (a) H2O/graphene, (b) H2O/C4F’s nonfluorine side, (c) H2O/C4F’s fluorine side, and (d) H2O/BN. For comparison, only the charge distributions between H2O and substrates are shown and the vertical position z (Å) is normalized. The yellow (blue) color represents the electron accumulation (dissipation). The electrostatic interaction in (b), (c), and (d) is stronger than in (a).
figure 5). Due to weak physical adsorption, the contribution bands and DOS of H2O or TMA adsorbates are far away from the Dirac point (the contribution bands of H2O and TMA are denoted as red and blue dots, respectively), having negligible influence on the intrinsic electronic properties of graphene (figures 5(a) and (f)). Similarly, because of the weak vdW interaction, when forming heterostructures with graphene, hBN and C4F cause slight perturbation to graphene electronic properties, opening a bandgap of 56 meV and 30 meV, respectively (figures 5(b) and (c)). Thanks to an in-plane intrinsic polarization, h-BN displays solid ability to preserve the electronic properties of the underlying graphene. Comparing with the exposed graphene-h-BN heterostructure (figure 5(b)), the H2O and TMA adsorbates (figures 5(d) and (g)) on h-BN make negligible change in the vicinity of graphene Dirac point. However, unlike h-BN, C4F has a vertical intrinsic polarization (please see our previous work in [62]) perpendicular to its basal plane and can interact with the adsorbates. Figures 5(e) and (h) show the H2O or TMA’s contribution bands and DOS will shift towards the Fermi energy when adsorption on graphene-SFG vdW heterostructures, leading to n-doping on graphene. In figure 5(i), H2O/DFG/graphene adsorption system is also investigated. In contrast to H2O/C4F/graphene, contribution bands of H2O adsorbate and its DOS shift little among the H2O/DFG/graphene. Since central-symmetry property of DFG induces no polarization, the comparison between H2O/C4F/graphene and H2O/DFG/graphene systems confirms the doping effects are caused by the vertical intrinsic polarization of C4F. Although we only study H2O or TMA monomer adsorption here, the doping effects are expected to be observed in any practical dielectric material deposition. This is because the vertical electric field induced by intrinsic polarization of C4F can interact with impurity or defective states of dielectric materials and change their hybridization with graphene states, resulting in graphene doping. This is somewhat similar to the
explanation of water-adsorption induced doping effects in substrate-supported graphene in [63]. Doping effect considered above is under the LDA optimized stacking order, i.e. AB, while for AA-stacking, similar n-type doping effects for SFG/G substrate are found, whereas graphene is well protected under h-BN. Besides fluorinated graphene, other preprocessing methods like hydrogenated graphene (HG) and defective graphene are also considered. HG is found hydrophobic [64, 65] and unstable during ALD process [25, 66], while graphene with defects provides a large number of nucleation centers for ALD precursors [15, 18]. For other ALD high-κ dielectric growth, for instance HfO2 or ZrO2, typical precursors like HfCl4 or ZrCl4 [67, 68] are simply considered. Due to the polarization nature of Hf–Cl or Zr–Cl bond, if we put the molecule on the boron site of h-BN or carbon side of SFG, strong adhesion will also be observed. For doping effect, results are the same with H2O or TMA on h-BN/G substrate, where h-BN still plays a key role in preserving graphene’s intrinsic properties. However, unlike H2O and TMA on SFG/G substrate, no obvious doping is found and molecular orbitals are deep inside the valence and conduction band of substrate. By using the vdW interaction inclusive DFT-D2 calculation, we demonstrate that H2O monomer exhibits a substantial increase in the adsorption energy on SFG and h-BN, in contrast to the pristine graphene. Due to the relatively strong electrostatic interaction with H2O adsorbates observed in our simulations, we predict that SFG and h-BN will be good candidates as 2D seed layers for ALD dielectric materials growth on graphene. When adsorbed on the SFG surface of graphene-SFG heterostructure, energy bands of H2O or TMA precursors are found hybridized with graphene near the Dirac point, leading to n-type doping in graphene. However, when H2O or TMA adsorbed on the h-BN surface of graphene-h-BN heterostructures, no doping effect is observed. Thus, one can choose SFG or h-BN as a seed layer to meet requirements for various electronic doping. The proposed 6
H Guo et al
Nanotechnology 25 (2014) 355202
Figure 5. The schematic structures (left), energy band (middle), and density of states (right) for the systems of (a) H2O/graphene, (b) BN/ graphene, (c) C4F/graphene, (d) H2O/BN/graphene, (e) H2O/ C4F/graphene, (f) TMA/graphene, (g) TMA/BN/graphene, (h) TMA/C4F/ graphene, and (i) H2O/ DFG/graphene. The contribution bands of H2O and TMA monomers are marked as red and blue dots in each plot of band structures, respectively. The red or blue arrows in (e) or (h) denote that the H2O or TMA contribution band is shifted towards the vicinity of Fermi energy. The green arrows indicate the corresponding DOS peaks.
Acknowledgment
SFG and h-BN seed layers for graphene vdW heterostructures are expected not to be limited to the growth of Al2O3 and may also find broad applications in the integration of high-κ dielectric materials on graphene.
The authors would like to thank Profs E Pop, C Zhou, I Hamada, H Wong, and Dr J Ma for helpful discussion and 7
H Guo et al
Nanotechnology 25 (2014) 355202
comments. This work is supported by National Science Foundation of China (Grant Nos. 61006077 and 61274123), ZJ-NSF (LR12F04001), and US National Science Foundation (Grants ECCS-1002228, ECCS-1028267, CMMI-1162312). Y Xu is supported by ZJU Cyber Scholarship, and Cyrus Tang Center for Sensor Materials and Applications, and Visiting-by-Fellowship of Churchill College, University of Cambridge. T Hasan is supported by the Churchill College, Cambridge and The Royal Academy of Engineering (Graphlex). The authors thank the official license of VASP package from the University of Vienna.
[28] Farmer D B, Chiu H-Y, Lin Y-M, Jenkins K A, Xia F and Avouris P 2009 Nano Lett. 9 4474 [29] Kim S, Nah J, Jo I, Shahrjerdi D, Colombo L, Yao Z, Tutuc E and Banerjee S K 2009 Appl. Phys. Lett. 94 062107 [30] Pirkle A, Wallace R M and Colombo L 2009 Appl. Phys. Lett. 95 133106 [31] Hsu A, Wang H, Kim K K, Kong J and Palacios T 2011 Japan. J. Appl. Phys. 50 070114 [32] Hollander M J et al 2011 Nano Lett. 11 3601 [33] Robinson J T et al 2010 Nano Lett. 10 3001 [34] Nair R R et al 2010 Small 6 2877 [35] Şahin H, Topsakal M and Ciraci S 2011 Phys. Rev. B 83 115432 [36] Wei W and Jacob T 2013 Phys. Rev. B 87 115431 [37] Karlický F, Kumara Ramanatha Datta K, Otyepka M and Zbořil R 2013 ACS Nano 7 6434 [38] Moon J S et al 2013 IEEE Electron Device Lett. 34 1190 [39] Pacile D, Meyer J C, Girit C O and Zettl A 2008 Appl. Phys. Lett. 92 133107 [40] Baraket M, Walton S G, Lock E H, Robinson J T and Perkins F K 2010 Appl. Phys. Lett. 96 231501 [41] Geim A K and Grigorieva I V 2013 Nature 499 419 [42] Kresse G and Hafner J 1993 Phys. Rev. B 47 558 [43] Kresse G and Furthmüller J 1996 Phys. Rev. B 54 11169 [44] Kresse G and Joubert D 1999 Phys. Rev. B 59 1758 [45] Blöchl P E 1994 Phys. Rev. B 50 17953 [46] Ma J, Michaelides A, Alfè D, Schimka L, Kresse G and Wang E 2011 Phys. Rev. B 84 033402 [47] Hamada I 2012 Phys. Rev. B 86 195436 [48] Neugebauer J and Scheffler M 1992 Phys. Rev. B 46 16067 [49] Grimme S 2006 J. Comput. Chem. 27 1787 [50] Perdew J P, Burke K and Ernzerhof M 1996 Phys. Rev. Lett. 77 3865 [51] Stine R, Lee W-K, Whitener K E, Robinson J T and Sheehan P E 2013 Nano Lett. 13 4311 [52] Giovannetti G, Khomyakov P, Brocks G, Kelly P and van den Brink J 2007 Phys. Rev. B 76 073103 [53] Ramasubramaniam A, Naveh D and Towe E 2011 Nano Lett. 11 1070 [54] Perdew J P and Zunger A 1981 Phys. Rev. B 23 5048 [55] Ceperley D M and Alder B J 1980 Phys. Rev. Lett. 45 566 [56] Monkhorst H J and Pack J D 1976 Phys. Rev. B 13 5188 [57] Momma K and Izumi F 2008 J. Appl. Crystallogr. 41 653 [58] Voloshina E, Usvyat D, Schütz M, Dedkov Y and Paulus B 2011 Phys. Chem. Chem. Phys. 13 12041 [59] Li X, Feng J, Wang E, Meng S, Klimeš J and Michaelides A 2012 Phys. Rev. B 85 085425 [60] Kurita T, Okada S and Oshiyama A 2007 Phys. Rev. B 75 205424 [61] Liu H, Xu K, Zhang X and Ye P D 2012 Appl. Phys. Lett. 100 152115 [62] Xu Y, Liu Y, Chen H, Lin X, Lin S, Yu B and Luo J 2012 J. Mater. Chem. 22 23821 [63] Wehling T O, Lichtenstein A I and Katsnelson M I 2008 Appl. Phys. Lett. 93 202110 [64] Hare E F, Shafrin E G and Zisman W A 1954 J. Phys. Chem. 58 236 [65] Zheng L, Li Z, Bourdo S, Watanabe F, Ryerson C C and Biris A S 2011 Chem. Commun. 47 1213 [66] Ryu S, Han M Y, Maultzsch J, Heinz T F, Kim P, Steigerwald M L and Brus L E 2008 Nano Lett. 8 4597 [67] Müller J, Böscke T S, Schröder U, Reinicke M, Oberbeck L, Zhou D, Weinreich W, Kücher P, Lemberger M and Frey L 2009 Microelectron. Eng. 86 1818 [68] Chen R, Kim H, McIntyre P C and Bent S F 2004 Appl. Phys. Lett. 84 4017
Reference [1] Novoselov K S 2004 Science 306 666 [2] Lemme M C, Echtermeyer T J, Baus M and Kurz H 2007 IEEE Electron Device Lett. 28 282 [3] Wang X, Ouyang Y, Li X, Wang H, Guo J and Dai H 2008 Phys. Rev. Lett. 100 206803 [4] Meric I, Han M Y, Young A F, Ozyilmaz B, Kim P and Shepard K L 2008 Nat. Nanotechnology 3 654 [5] Lin Y-M, Dimitrakopoulos C, Jenkins K A, Farmer D B, Chiu H-Y, Grill A and Avouris P 2010 Science 327 662 [6] Schwierz F 2010 Nat. Nanotechnology 5 487 [7] Liu M, Yin X, Ulin-Avila E, Geng B, Zentgraf T, Ju L, Wang F and Zhang X 2011 Nature 474 64 [8] Bao Q, Zhang H, Wang B, Ni Z, Lim C H Y X, Wang Y, Tang D Y and Loh K P 2011 Nat. Photonics 5 411 [9] Bao Q and Loh K P 2012 ACS Nano 6 3677 [10] Koppens F H L, Chang D E and García de Abajo F J 2011 Nano Lett. 11 3370 [11] Ju L et al 2011 Nat. Nanotechnology 6 630 [12] Grigorenko A N, Polini M and Novoselov K S 2012 Nat. Photonics 6 749 [13] Liao L and Duan X 2010 Mater. Sci. Eng. R 70 354 [14] Kim J and Jandhyala S 2013 Thin Solid Films 546 85 [15] Garces N Y, Wheeler V D and Gaskill D K 2012 J. Vac. Sci. Technol. B 30 030801 [16] Xuan Y, Wu Y Q, Shen T, Qi M, Capano M A, Cooper J A and Ye P D 2008 Appl. Phys. Lett. 92 013101 [17] Lee B, Park S-Y, Kim H-C, Cho K, Vogel E M, Kim M J, Wallace R M and Kim J 2008 Appl. Phys. Lett. 92 203102 [18] Wang X, Tabakman S M and Dai H 2008 J. Am. Chem. Soc. 130 8152 [19] Jandhyala S et al 2012 ACS Nano 6 2722 [20] Williams J R, DiCarlo L and Marcus C M 2007 Science 317 638 [21] Lin Y-M, Jenkins K A, Valdes-Garcia A, Small J P, Farmer D B and Avouris P 2009 Nano Lett. 9 422 [22] Nayfeh O M, Marr T and Dubey M 2011 IEEE Electron Device Lett. 32 473 [23] Pirkle A, McDonnell S, Lee B, Kim J, Colombo L and Wallace R M 2010 Appl. Phys. Lett. 97 082901 [24] Garces N Y, Wheeler V D, Hite J K, Jernigan G G, Tedesco J L, Nepal N, Eddy C R and Gaskill D K 2011 J. Appl. Phys. 109 124304 [25] Wheeler V, Garces N, Nyakiti L, Myers-Ward R, Jernigan G, Culbertson J, Eddy C and Kurt Gaskill D 2012 Carbon 50 2307 [26] Huard B, Sulpizio J, Stander N, Todd K, Yang B and Goldhaber-Gordon D 2007 Phys. Rev. Lett. 98 236803 [27] Zheng Y, Ni G-X, Toh C-T, Tan C-Y, Yao K and Özyilmaz B 2010 Phys. Rev. Lett. 105 166602
8
