LETTER pubs.acs.org/JPCL
Hybrid Graphene/Titania Nanocomposite: Interface Charge Transfer, Hole Doping, and Sensitization for Visible Light Response Aijun Du,*,† Yun Hau Ng,‡ Nicholas J. Bell,‡ Zhonghua Zhu,§ Rose Amal,*,‡ and Sean C. Smith*,† †
Centre for Computational Molecular Science (CCMS), Australian Institute for Bioengineering and Nanotechnology (AIBN) and School of Chemical Engineering, The University of Queensland, QLD 4072, Brisbane, Australia ‡ ARC Centre of Excellence for Functional Nanomaterials, School of Chemical Sciences and Engineering, University of New South Wales, Sydney, NSW 2052, Australia §
bS Supporting Information ABSTRACT: We demonstrated for the first time by large-scale ab initio calculations that a graphene/titania interface in the ground electronic state forms a charge-transfer complex due to the large difference of work functions between graphene and titania, leading to substantial hole doping in graphene. Interestingly, electrons in the upper valence band can be directly excited from graphene to the conduction band, that is, the 3d orbitals of titania, under visible light irradiation. This should yield well-separated electronhole pairs, with potentially high photocatalytic or photovoltaic performance in hybrid graphene and titania nanocomposites. Experimental wavelength-dependent photocurrent generation of the graphene/titania photoanode demonstrated noticeable visible light response and evidently verified our ab initio prediction. SECTION: Surfaces, Interfaces, Catalysis
graphene/titania nanocomposites,1619 in which high stability and enhanced photovoltaic properties have been demonstrated due to the unique electronic structure of graphene. However, the underlying mechanisms for this behavior remain wholly unclear. Herein, we report for the first time large-scale density functional calculations to characterize the interface between graphene and a rutile TiO2(110) surface. We find that there is significant charge transfer from graphene to titania, leading to efficient hole accumulation (p-type doping) in the graphene layer and shifting of the Fermi level 0.65 eV below the Dirac point in titaniasupported graphene. Significantly, our calculations reveal that an interface charge-transfer mechanism can mediate the direct excitation of electrons from graphene into the titania conduction band (CB) (the 3d orbitals of titanium) under visible light irradiation. Importantly, the resulting electronhole pairs are then well-separated, leading, in principle, to enhanced photovoltaic efficiency through reduced pair recombination. In fact, our wavelength dependence study of the photocurrent generated by the graphene/titania nanocomposite derived from a
G
raphene has been the subject of intensive research in the past 5 years due to its intriguing electronic structure, which is expected to be important for practical applications in nanoelectronics.18 A prerequisite for the development of graphene-based electronics is the reliable control of the charge carrier type and its density. Graphene displays intrinsic n-type character, and it is very important for charge carriers to be tuned from electrons to holes, that is, hole doping. Up to now, some novel strategies on p-type doping of graphene have been developed. For example, Chen et al. reported surface transfer p-type doping in tetrafluoro-tetracyanoquinodimethane-modified graphene with limited thermal and chemical stability in air or solution.9 Bismuth- and gold-doped graphene has demonstrated efficient hole doping with excellent perspectives for large-scale production.10 Most recently, a promising all-organic route to doping graphene has been proposed to realize band gap tuning of epitaxial graphene on a silicon carbide surface.11 Metal oxides such as TiO2 are large-band-gap semiconductors, only showing photocatalytic activity under ultraviolet irradiation.12,13 Carbon nanotubes have been used to interface with titania to achieve extended photocatalytic activities well beyond that of titania materials.14,15 Most recently, many experimental efforts have been also focused on the synthesis of hybrid r 2011 American Chemical Society
Received: February 28, 2011 Accepted: March 24, 2011 Published: April 01, 2011 894
dx.doi.org/10.1021/jz2002698 | J. Phys. Chem. Lett. 2011, 2, 894–899
The Journal of Physical Chemistry Letters
LETTER
Figure 1. (a) Model for simulating the interface between graphene and the rutile TiO2(110) surface. (b) 3D charge density difference for the graphene/ titania nanocomposite with an isovalue of 0.005 e/Å3. Green, blue, and red balls represent C, Ti, and O atoms, respectively. Yellow and orange isosurfaces represent charge accumulation and depletion in the space.
photocatalytic reduction method has evidenced this interface charge-transfer mechanism by exhibitng considerable visible light response under monochromatic light illumination. A 6 6 (following the conventional definition of the zigzag/ armchair nanoribbon) single graphene layer containing 72 carbon atoms is used to match a 5 2 nine atomic layer stoichiometric TiO2(110) surface slab containing 180 Ti and O atoms with three bottom layers fixed at bulk position in a supercell (14.68 12.89 25 Å3). The whole system contains totally 252 atoms with 1248 valence electrons. The local density approximation20 was chosen because long-range van der Waals interactions are expected to be significant in this system.21 All of the calculations are performed by using the plane wave basis VASP code22,23 implementing the projector augment wave method.24,25 The vacuum space perpendicular to the TiO2(110) surface is around 15 Å, which is enough to separate the interaction between periodic images. A MorkhostPack mesh of kpoints, 4 4 1 and 8 8 1 points,26 is used, respectively, to sample the two-dimensional Brillouin zone for geometry optimization and for calculating the density of states. The cutoff energies for plane waves are chosen to be 500 eV, and the convergence tolerance of force on each atom during structure relaxation are set at 0.005 eV/Å. To understand the origin of the interface charge-transfer complex in the graphene/titania nanocomposite, work functions for the graphene layer and TiO2(110) surface are calculated by aligning the Fermi level relative to the vacuum energy level. The frequency-dependent dielectric matrix is calculated for the TiO2(110) surface and for the hybrid graphene/TiO2(110) nanocomposite. The imaginary part is determined by a summation over empty states using the equation27 00
εRβ ðωÞ ¼
rutile nanopowder (Aldrich) and graphene oxide derived from a modified Hummers’s method29 were dispersed in absolute ethanol under N2 purging and were exposed to UV irradiation for 3 h using an Oriel 450 W xenon arc lamp to obtain RGO/ rutile composites. The ratio of graphene oxide to TiO2 was determined to be 5 wt %. Thin film with 2 mg cm2 of RGO/ rutile and rutile only were drop-cast on FTO glass slides and employed as photoanodes in photoelectrochemical cells to determine the benefit of composite material versus rutile alone. Half cell reactions were conducted using 0.1 M KOH as the electrolyte and Pt and Ag/AgCl as the counter and reference electrodes, respectively. Short-circuit photocurrent generation (under 0 V bias) was recorded on an Autolab PGSTAT302N potentiostat using a Newport integrated monochromator. Figure 1a presents a top view of the graphene/titania interface model used in our calculations. Geometry optimizations were first performed for the graphene/titania interface utilizing the conjugate gradient method. The equilibrium distance between the graphene layer and the top of the TiO2(110) surface is calculated to be 2.75 Å. The interface adhesion energy was obtained according to the following equation Ead ¼ Ecomb Egraphene ETiO2 ð110Þ
where Ecomb, Egraphene, and ETiO2(110) represent the total energy of the relaxed graphene/titania nanocomposite, pure graphene sheet, and clean TiO2(110) slab, respectively. The interface binding energy is calculated to be 1.69 eV for this model interface, which indicates high stability. To characterize the change of electronic structure at the graphene/titania interface, three-dimensional charge density difference plots are calculated by subtracting the electronic charge of a hybrid graphene/TiO2 nanocomposite from the separate graphene layer and TiO2(110) surface, as shown in Figure 1b. Clearly, there is significant charge transfer from the graphene layer to the TiO2(110) surface in the ground electronic state, implying efficient hole accumulation in the titania-supported graphene. Further charge analysis based on the Bader method30 indicates that there is an average charge transfer of around 0.02 e per carbon atom from graphene to titania. To understand the origin of such an interface charge-transfer complex, work functions for the graphene layer and TiO2(110) surface were calculated by aligning the Fermi level relative to the vacuum energy level (see Figures S1 and S2 in the Supporting Information). They were found to be 4.75 and 7.45 eV for
4π2 e2 1 lim 2 2ωk δðεck ενk ωÞ q f 0 q c, ν, k Ω
∑
Æμck þ eRq jμνk æÆμck þ eRq jμνk æ
ð2Þ
ð1Þ
where the indices c and ν represent conduction and valence band (VB) states, respectively, and μck refers to the cell periodic part of the orbitals at the k-point. A large number of empty CB states (1300 bands in our calculation) are included for the summation in eq 1. The reduced graphene oxide/rutile (RGO/rutile) composite photoanode was prepared based on a photocatalytic reduction of graphene oxide using rutile TiO2.16,28 Commercially available 895
dx.doi.org/10.1021/jz2002698 |J. Phys. Chem. Lett. 2011, 2, 894–899
The Journal of Physical Chemistry Letters
LETTER
Figure 2. DOS plots for (a) free-standing and (b) TiO2(110) surface-supported graphene. The short dotted lines indicate the Fermi level.
graphene and the TiO2(110) surface, which are in good agreement with previous studies.31,32 The spontaneous interfacial charge transfer from graphene to titania can be simply rationalized in terms of the large difference in these work functions. Additionally, it is worth noting that dipole correction may artificially favor charge transfer in the insulator. However, our further calculations in the presence and absence of dipole correction indicate that interface charge transfer will not be significantly affected. Other significant aspects of the interfacial charge-transfer complex between graphene and the TiO2(110) surface can be evidenced by comparing the density of states (DOS) plots for free-standing and titania-supported graphene, as shown in Figure 2a and b, respectively. The total DOS and projected DOS on Ti, O, and C atoms for the hybrid graphene/titania nanocomposite have been plotted, as presented in Figure S3 in the Supporting Information. Our calculations indicate that the Fermi level is down-shifted by around 0.65 eV relative to the Dirac point. According to the linear dispersion of the DOS close to the Dirac point,33,34 the hole concentration in titania-supported graphene can be estimated by the following equation 1 Nh ¼ ðjEF ED jÞ2 πðpνF Þ2
Figure 3. The calculated imaginary part of the dielectric function of the rutile TiO2(110) surface (red solid line) and the hybrid graphene/ TiO2(110) nanocompoiste (green dot line) for the polarization vector perpendicular to the surface.
The charge-transfer complex formed at the interface between titania and graphene is also expected to mediate photocatalytic activities under visible light, as demonstrated in some other similar systems.38,39 To examine such an effect, the imaginary part of the dielectric function for a TiO2(110) surface and a hybrid graphene/titania nanocomposite were calculated, as shown in Figure 3. Optical absorption for pure titania materials occurs in the UV light region, which is attributed to the intrinsic transition from the O2p orbital to Ti3d orbital. The absorption edge shown in Figure 3 should have a rigid shift (1.52.0 eV), as reported in ref 11, because LDA significantly underestimates the band gap. However, the shape of the optical absorption curve will remain nearly unchanged because dispersion in the CB will not be affected by LDA. As can be seen clearly from the calculated imaginary part of the dielectric function, the red shift of absorption edge is as large as 0.50.7 eV for a hybrid graphene/titania nanocomposite compared to that for pure TiO2(110) surface. Concomitantly, the hybrid graphene/titania nanocomposite may be expected to display the enhanced photocatalytic activities under the visible light irradiation. To understand the origin of visible light absorption, the electron density distributions for a series of VB and CB states have been calculated, as plotted in Figure 4. Here, VB maximum
ð3Þ
where Nh is the hole density per unit area, νF is the Fermi velocity of graphene (106 m/s), and EF and ED are the energy positions of the Fermi level and Dirac point, respectively. From eq 1, the hole density is roughly estimated to be around 3.0 1013 cm2, suggesting efficient hole doping in graphene when interfaced with titania materials possessing a high work function. It is well-known that the local density approximation (LDA) underestimates the absolute size of band gap, and thus, the reported electron transfer may be overestimated.35 To explore this effect, calculations based on the addition of the on-site Coulomb correction to LDA (LDA þ U)36 have been carried out. Parameters U and J are chosen to be 6.0 and 0.5 eV (Ueff = U J = 5.5 eV), which is more reasonable, as reported in a previous study.37 We found that the equilibrium distance are nearly unchanged, although the TiO2(110) ). Remarkably, the interface charge surface is slightly expanded (0.1 Å transfer between graphene and titania still remains significant, as shown in Figures S4 and S5 (see Supporting Information for more details). Apparently, the predicted interfacial charge transfer is a robust prediction supported by both methods used. 896
dx.doi.org/10.1021/jz2002698 |J. Phys. Chem. Lett. 2011, 2, 894–899
The Journal of Physical Chemistry Letters
LETTER
Figure 5. Photocurrent action spectra of the TiO2-only (rutile) film compared with those of RGO/rutile nanocomposite films prepared by the photocatalytic reduction method at a 0 V bias condition.
reports revealing the positive effect of incorporating graphene to enhance the transport of excited electrons from TiO2.40,41 Another significant feature in this figure is the presence of considerable visible response at a wavelength greater than 500 nm as TiO2 is not band-gap-photoexcited. Note that a bare RGO film generates no photocurrent at all wavelengths. The visible light response of RGO/rutile is presumably attributed to the electron transferred from graphene to the CB of TiO2 and subsequently to the FTO electrode, as calculated in the ab initio model. Most studies available to date report that enhancements of performances of graphene/semiconductor nanocomposites, either in photoelectrochemical or photocatalytic reactions, are mainly attributed to the role of graphene as the electron shuttle to carry excited electrons from the semiconductor rather than sensitizing the semiconductor.17,39,40 This is, therefore, the first to reveal the function of graphene to sensitize TiO2, while an appropriate interface interaction is achieved in the form of a nanocomposite. Experimentally, the long-range arrangement of graphene on a TiO2(110) surface flatly may be challenging. However, the visible light responses predicted here are in good agreement with experimental measurement. Therefore, we believe that our planar model is reasonable to describe the graphene/titania interface because pure rutile titania powder contains more than 50% of the (110) facet. In summary, large-scale density functional calculations have been performed to understand interface interactions between a graphene monolayer and a rutile TiO2(110) surface. We found that the graphene/titania interface in the ground electronic state forms a charge-transfer complex. This can be understood by the large difference of work functions between graphene and titania and implies efficient hole doping in titania-supported graphene. Most interestingly, valence electrons may be expected to become directly excited from graphene into the titania CB under visible light irradiation, producing well-separated electronhole pairs. Combined with the experimental proof from the wavelengthdependent photocurrent study that unveils the visible light response of graphene/TiO2, the new role of graphene as the sensitizer to TiO2 is proven to be genuine and should find use in wider photovoltaic and photocatalytic applications
Figure 4. Plots of the electron and hole density distributions for the VB and CB states with an isovalue of 0.005 e/Å3. Green, blue, and red balls represent C, Ti, and O atoms, respectively.
and CB minimum are determined by the highest-occupied and lowest-unoccupied bands, respectively. The electronic charge predominantly localizes around the 2p state of O and C atoms in both VB3 and VB2. The 3d orbitals of Ti become partially occupied in VB1 and in the VBM states due the interface charge transfer from graphene to TiO2. The CBM state and CBþ1 are comprised largely of the Ti 3d orbitals and partly of graphene 2p orbitals. In CBþ2, electronic charge distribution on graphene will decrease to 0. It is well-known that band gap for pure titania material is around 3.0 eV, which is in the UV light region, and the normal photoexcitation is from the 2p state of oxygen atoms (VB4) to the 3d orbital of Ti atoms (CBþ2). When interfacing with graphene, new visible light absorption bands involving transitions between the high-lying VB states and the low-lying CB states will appear in the middle of the band gap of titania. Under visible light irradiation, electrons in graphene as shown in VB3, VB2, VB1, and VBM are likely to be directly excited into surface Ti 3d orbitals, which constitute the major electron density component of the low-lying CB states, as shown in Figure 4. Clearly, the nascent electronhole pairs are then well-separated, potentially leading to significantly reduced charge recombination and thus enhanced photovoltaic properties. It should be pointed out that direct optical excitation within titania in the visible light region also becomes a possibility as a result of the down-shifting of the Ti 3d orbital energies induced by the complexation with graphene. Here, we should note that excitedstate and hybrid functional calculations are important to further explore these phenomena but are as yet unreachable for such a large interface model. However, the implications of the above analysis for enhanced phtocatalytic activity are clear and are expected to be robust. The interface charge interaction between graphene and TiO2 at the ground electronic state as calculated above is validated experimentally by a photocurrent action spectrum shown in Figure 5. Rutile TiO2 with a band gap energy of 3 eV exhibits an onset photocurrent at around 420 nm, while RGO/rutile enhances the photocurrent generated at wavelength