Anatase TiO2 Surface Functionalization by Alkylphosphonic Acid: A

Jan 5, 2012 - William E. Ford , Ffion Abraham , Frank Scholz , Gabriele Nelles , Graham Sandford , and Florian von Wrochem. The Journal of Physical ...
2 downloads 0 Views 498KB Size
Article pubs.acs.org/JPCC

Anatase TiO2 Surface Functionalization by Alkylphosphonic Acid: A DFT+D Study Cristiana Di Valentin*,† and Dominique Costa*,‡ †

Dipartimento di Scienza dei Materiali, Università di Milano-Bicocca, via Cozzi 53, 20125 Milano, Italy Laboratoire de Physico-Chimie des Surfaces, UMR CNRS-ENSCP 7045, Ecole Nationale Supérieure de Chimie de Paris, Chimie-Paristech 11, rue Pierre et Marie Curie, 75005 Paris, France



S Supporting Information *

ABSTRACT: We report on a DFT study of the adsorption of nbutylphosphonic acid [CH3(CH2)3PO(OH)2] on the anatase (101) TiO2 model surface. The pure GGA (PBE and rPBE) approach is compared to a hybrid functional (B3LYP). Dispersion forces are taken into account in the Grimme framework correcting both energies and gradients. The surface coverage is varied from 0.25 ML (one phosphonic unit every four surface Ti5c) to 1 ML (one phosphonic unit/Ti5c). The overall picture resulting from these three approaches is qualitatively the same, although quantitative details are different. At low coverage, the alkylphosphonic acid adsorbs in a monodissociated bidentate mode, forming two PO−Ti5c bonds with one additional OH−Os bond, where Ti5c and Os refer to surface titanium and oxygen atoms. At full coverage, the molecules adsorb in a monodentate mode and reorient in order to minimize lateral repulsion and maximize H-bonds. The contribution of dispersion forces to the adsorption energy increases with the coverage. The 0.75 and 1 ML coverages, which are the most favored, correspond nicely to the reported experimental coverages. The results suggest that for entropic reasons the surface could be rather disordered, with coexisting domains of different coverages 0.75 and 1 ML. No additional states in the band gap of the semiconductors are formed for this hybrid organic−inorganic system. The workfunction decreases by 0.7 eV when a full self-assembled monolayer is supported on the surface. SAMs to enhance the cell attachment to Ti surfaces.2 Crucial issues in the biomedical context are the stability of SAMs under physiological and sterilizing conditions. Note that Ti and Ti alloys exposed to ambient conditions and physiological media are covered with a thin native TiO2 oxide film.3−6 This film (whose structure is not known) is largely hydroxylated.3,4,6 There is an increasing number of experimental studies of phosphonic acid SAMs on Ti/TiO2 (see refs 2 and 7−18 for a nonexhaustive list). Helmy and Fadeev have investigated octadecylphosphonic acid (C18H37PO3H2) SAMs formation on anatase TiO2 and found that the SAM order gradually improves with coverage and that highly ordered SAMs are obtained only for high surface coverages, the monolayer density being 4.3−4.8 groups/nm2.15 Gawalt et al. showed by means of diffuse reflectance infrared spectroscopy (DRIFT) that octadecylphosphonic acid (ODPA) forms ordered monolayers on native titanium dioxide surfaces.2 The degree of monolayer ordering was reported to increase with increasing chain length.11 Pawsey et al., Guerrero et al., and Mutin et al. proposed that bond of the organophosphonic acids to the surface is usually bi- or tridental.16−18 The stability of SAMs on TiO2 mainly depends on the type of head groups and on their

1. INTRODUCTION A biomaterial is defined as any natural or man-made material that constitutes part of a living structure or a biomedical device performing a natural function. Such materials are essentially used in medical applications. The biocompatibility of a material is defined independently of where and how the material is to be used. The “ideal” biomaterial should induce little or no immune response in a given organism and should be able to integrate with a particular cell type or tissue. Also, it should be resistant to corrosion, nontoxic, nor induce rejection or allergies. Ti and Ti alloys are typical biomaterials used for hip joints, bone plates, and dental implants, as these materials present most of the characteristics listed above and, additionally, have a high mechanical strength.1 However, Ti, in the same way as gold, is biologically inactive. In order to achieve the desired functionality in vivo, surface treatments of metallic materials have been proposed. In particular, self-assembled monolayers (SAMs) are used to functionalize metal surfaces. SAMs are single-layered organic coatings that are deposited on metal or metal oxide surfaces by the adsorption of organic molecules from a solution. A further step for biomedical applications is the attachment to the supported SAMs of a variety of biomolecules, involving proteins, peptides, DNA, carbohydrates, antibodies, and therapeutics. As a possible application on Ti, arginine-glycineaspartic acid (RGD) peptides were attached to phosphonic acid © 2012 American Chemical Society

Received: April 7, 2011 Revised: December 19, 2011 Published: January 5, 2012 2819

dx.doi.org/10.1021/jp203256s | J. Phys. Chem. C 2012, 116, 2819−2828

The Journal of Physical Chemistry C

Article

phonic acids have a strong interaction with TiO2 surfaces and could be able to form ordered, high density SAMs. Phosphonic acid is an interesting case where the head of the molecule has a significant size, being composed of two O(H) moieties. In this case, on the one side, the formation of strong interactions with the surface, through e.g. multipoint anchoring, is expected, whereas on the other side, the self-organization of the adsorbed layer may be limited by the steric hindrance of the head function. Therefore, it is interesting to study which of the present factors (Ti surface density, head size, or chain stacking properties) or which combination of factors determines the SAMs density (and the resulting properties). In the present paper we propose a DFT study of the adsorption of an alkylphosphonic acid [CH3(CH2)3PO(OH)2] on the anatase (101) TiO2 model surface. In this acid the alkyl chain is sufficiently long to allow for some chain−chain interactions. A pure GGA approach (PBE48,49 or rPBE50,51) is compared to the use of a hybrid functional (B3LYP52,53). We expect that rPBE corrects for PBE overbinding,50 while B3LYP reduces the self-interaction problem.31 Moreover, GGA calculations are definitively computationally less expensive, which allows one to investigate a larger number of configurations. In this study, we also consider the correction for dispersion forces to the total energy and gradients. Recently, few works appeared in the literature that take dispersion forces into account in a DFT approach when studying organic molecules adsorbed on inorganic surfaces.54−57 The reader may be interested in a review of the present advantages and limits of each method, available in refs 56 and 57. Here, we used the dispersion contribution estimation recently proposed by Grimme et al. in the DFT-D scheme.57

affinity toward the metal oxide surface. It has been reported that dodecylphosphonic acid [CH3(CH2)11P(O)(OH)2] SAMs on Ti are stable in ambient air11 and in water.8,19 Besides applications for biomaterials, organic molecules adsorption and self-organization on TiO2 surfaces, especially phosphonic acid, are also of interest in the field of dyesensitized solar cell systems,20 catalysis,21 photoelectrochemistry,22 and optoelectronic.23 DFT approaches have been robustly used by the physicists’ and chemists’ communities during the last 2 decades and are nowadays successfully applied to model the reactivity of complex and large surfaces for applications in, for example, catalysis,24 nanoelectronics,25 geology,26 or biomedicine.27,28 Theoretical works concerning the interaction of metal/alloy surfaces with organic molecules are found in the literature, where the partners building the interface (surfaces, organic molecules) are considered both when separated and when in contact (organic molecules adsorbed at inorganic surfaces). Numerous ab initio studies have been devoted to bulk and surface properties of metals and metal oxides. TiO2 anatase and rutile (the two most common polymorphs) surfaces have been deeply investigated during the past few years.29,30 Particular attention has been focused on the details of the electronic structure of reduced and hydroxylated surface of TiO2.31 Such defects on the substrate surface behave as preferential adsorption sites for molecular adsorption. Due to their well-ordered, oriented, and dense structure, SAMs offer excellent model systems for atomic scale theoretical studies of adsorption of organic compounds. Recent papers have been published on SAMs at gold surfaces (refs 32−35 and references therein). Theoretical studies have confirmed that the intermolecular van der Waals and H bonds interactions between the methylene chains of molecules are responsible for the ordered, close-packed structure of monolayer, whereas the headgroup of the molecule is responsible for the strong metal−molecule interaction that is usually chemisorption, and the tail functional group determines the physical and chemical properties of the modified surface. The study of the interaction of organic molecules with oxide surfaces is also increasingly performed by means of ab initio tools.20,36−47 Some computational investigations on the adsorption of phosphonic,20,42 alkyl phosphonic,43 octylphosphonic,44 methyl phosphonic,45 and dimethyl methylphosphonic46 acids on inorganic surfaces and, in particular, also on TiO2 anatase surface exist in the literature. Nilsing et al. found at the B3LYP level a strong adsorption energy for HPO3H2 phosphonic acid on (101) anatase TiO2 of 180 kJ/mol (1.9 eV), in a monodentate nondissociated adsorption mode.20 In contrast, Luschtinetz et al., exploring a great amount of possible conformations, found, with the DFT-B method, that the most stable configuration of HPO3H2 (with a surface coverage of 0.5 ML) was a bidentate monodissociative one with an energy of adsorption of 255 kJ/mol (2.6 eV).42 Methylphosphonic acid adsorption on (110) rutile was studied by means of STM and DFT.45 As the coverage of adsorbates approaches 0.5 ML, STM images show an ordered 2 × 1 overlayer consistent with LEEDS results. According to DFT, at this coverage the molecule adsorbs dissociatively in the bidentate mode, with a strong adsorption energy of 2.5 eV.45 Finally, the adsorption of dimethyl methylphosphonate on (101) anatase TiO2 was calculated at the B3LYP level and the methyl groups were found to be bidentate dissociative with an energy of adsorption of 25 kcal/mol (1.1 eV).46 Thus, it appears that alkylphos-

2. COMPUTATIONAL METHODS 2.1. VASP Calculations. The periodic calculations were performed using the DFT method based on the GGA approximation employing the PBE48,49 and rPBE50,51 exchange-correlation functionals. These methods were used as implemented in the plane-wave program VASP.58,59 The projector-augmented wave (PAW) potentials60,61 were used for the core electron representation with a PAW core radius of 1.52 Å for oxygen. With plane wave methods, the quality of the basis set is determined by a single parameter, the energy cutoff (Ecut). We used a converged value of Ecut = 400 eV. The integration in reciprocal space was performed with a Monkhorst−Pack grid.62 The k-points grid is 4 × 4 × 2 for the bulk. The bulk cell parameters were calculated as acalc = 3.864 Å, ccalc = 9.707 Å (rPBE) [respectively acalc = 3.871 and ccalc = 9.724 (PBE)], a value in agreement with experimental (aexp = 3.776 Å and cexp = 9.486 Å)63 and other theoretical data with the same method.64 The anatase (101) surface was modeled with a periodically repeated slab. We considered a surface supercell containing 72 atoms in three triatomic layers, of dimension 1 × 2 in the [101̅] and [010] directions, respectively, corresponding to a surface area of 10.45 × 7.73 Å2. The surface slab exhibits four 5-fold coordinated Ti (Ti5c) surface atoms. Such a large cell allows us to study increasing surface coverage, from one to four phosphonic acids/unit cell, as well as different configurations at a given coverage. The Ti and O atoms of the bottom layer were fixed to the bulk positions, the cell size was kept fixed to the PBE (respectively rPBE) optimized bulk lattice parameters, and the other atoms were free to relax. Details of the surface are shown in Figure 1a. A vacuum layer of 30 Å along the z 2820

dx.doi.org/10.1021/jp203256s | J. Phys. Chem. C 2012, 116, 2819−2828

The Journal of Physical Chemistry C

Article

imentally.66,74 Table 1 of the Supporting Information summarizes all those data. 2.2. CRYSTAL09 Calculations. The anatase (101) surface was modeled with the same periodically repeated slab as described above. However, periodic boundary conditions are in two dimensions (2D calculations). The optimized bulk lattice parameters are taken from previous studies,75 in which the same approximations as in this work were used. As above, the atoms in the bottom layer were fixed to their bulk positions during geometry optimizations, in order to simulate the presence of the bulk underneath. The gradients with respect to atomic coordinates are evaluated analytically. The equilibrium structure is determined by using a quasi-Newton algorithm with a BFGS Hessian updating scheme.76 Convergence in the geometry optimization process is tested on the root-mean-square (rms) and the absolute value of the largest component of both the gradients and nuclear displacements. For all atoms, the thresholds for the maximum and the rms forces have been set to 0.000 45 and 0.000 30 au and those for the maximum and the rms atomic displacements to 0.001 80 and 0.001 20 au, respectively. Calculations were performed using the hybrid B3LYP52,53 functional with the Kohn−Sham orbitals expanded in Gaussiantype orbitals (GTO), as implemented in the CRYSTAL09 code77 [Ti 86411(d41), slab O 8411(d1), O8411(d11), C6311(d11), H 511(p1), P 8521(d1) all-electron basis-sets]. The basis set superposition error (BSSE) was evaluated in one sample case by using the counterpoise correction method (see Table 1).78 The percentage of exact Hartree−Fock (HF) exchange in the B3LYP functional is 20%.

Figure 1. (a) Top view of the anatase (101) TiO2 surface. Ti and O atoms are dark and light spheres, respectively. Lattice parameters are B3LYP data. Rows of O2c and Ti5c are represented by red and green, respectively. The repeated supercell is shown with dashed lines. (b) Top view of one n-butylphosphonic acid. (c, d) Schematic view of the M0 and B1 modes of adsorption.

direction perpendicular to the surface (x and y being parallel to the surface) was employed to prevent interactions between the repeated slabs. A k-point mesh of 2 × 2 × 1 was used. A dipolar correction was introduced along the z axis perpendicular to the surface to take into account the spurious polarity induced by the nonequivalence of the top and bottom surfaces. A convergence criterion of forces on relaxed atoms