Article pubs.acs.org/JPCC
Atomistic Investigation of the Solid−Liquid Interface between the Crystalline Zinc Oxide Surface and the Liquid Tetrahydrofuran Solvent Maria Ilenia Saba,*,†,‡ Vasco Calzia,†,‡ Claudio Melis,†,‡ Luciano Colombo,†,‡ and Alessandro Mattoni*,‡ †
Dipartimento di Fisica dell’Università di Cagliari and ‡Istituto Officina dei Materiali (CNR-IOM) Unità SLACS Cagliari, Cittadella Universitaria, I-09042 Monserrato (Ca), Italy ABSTRACT: Zinc oxide is typically used as an electron acceptor in the preparation of hybrid solar cells. Such hybrids are, in many cases, synthesized from solutions. In this work, we investigate the possible persistence of solvent tetrahydrofuran molecules at the interface with zinc oxide by means of atomistic simulations based on model potentials and density functional theory calculations. We found a strong interaction between the solvent and the zinc oxide that leads to the formation of an ordered layer of tetrahydrofuran molecules bound to the zinc oxide substrate.
■
INTRODUCTION inc oxide (ZnO) is a wide band gap semiconductor of current interest in optoelectronics and photovoltaics.1 It provides very good electron mobility (205 cm2/(V s)), it is nontoxic, and it can be grown in a variety of highly crystalline nanostructures,1,2 which are commonly used as electron acceptors. In combination with organic donors (e.g., conjugated polymers or molecules), ZnO nanostructures are used to synthesize hybrid bulk heterojunctions (BHJs), very promising in photovoltaics for their low cost and for combining the excellent transport properties of ZnO with the processability and the tunable optical properties of the organic components.3 In most cases such hybrids are synthesized from solutions by dissolving ZnO nanostructures and the organic components into suitable solvents without the need of expensive vacuum conditions. For example, by spin-coating4 a drop of solution containing ZnO nanorods and a conjugate polymer (such as P3HT) can be centrifugated over a substrate under air conditions. After solvent evaporation a thin film of organic− inorganic material is deposited. The final microstructure and the photoconversion efficiency of the corresponding hybrid strongly depend on the processing conditions. In particular, the type of solvent adopted can cause a large change (up to two orders of magnitude) in the efficiency.5 Some residual solvent molecules can bind to ZnO and persist even after the syntesis process at the organic/inorganic interface. Solvent contamination of the ZnO/organic interface can possibly affect the binding between the components, the interface morphology, and the stability;6 furthermore they can generate dipoles (in case of polar solvent) that eventually affect the charge separation process. Among the solvents commonly used in combination with ZnO are xylene, dichlorobenzene,
chlorobenzene, tetrahydrofuran (THF), and chloroform. In particular, THF is commonly used in the production of hybrid ZnO-based solar cells because of its low freezing point and the ability to solvate both polar and nonpolar compounds.7 Each THF molecule consists of one oxygen and four carbon atoms (each saturated by two hydrogens as in Figure 1), and it exists in different isoenergetic planar and nonplanar configurations (e.g., twisted or envelope).8 At room temperature THF is liquid, with molecules weakly interacting through Coulombic and dispersive forces. In this work we combine first-principles calculations and model potential molecular dynamics (MPMD) to study the interaction between the solvent THF and ZnO. To this aim we generate atomistic models of the liquid−solid interface between THF and ZnO (101̅0) crystalline surface, we characterize the local order and morphology of the interface, and we calculate the solvent-ZnO adhesion energy. Our analysis provides evidence of the formation of an ordered (i.e., crystalline) layer of THF that spontaneously wets the ZnO surface. Such a wetting crystalline layer is further analyzed by studying the binding of a single THF molecule on ZnO through firstprinciples calculations. We finally compare the energetics of THF/ZnO with other organic/ZnO interfaces used in photovoltaics, and we conclude that the solvent is likely a contaminant of realistic hybrid interfaces.
Z
© 2012 American Chemical Society
Received: March 21, 2012 Revised: May 15, 2012 Published: May 15, 2012 12644
dx.doi.org/10.1021/jp302700b | J. Phys. Chem. C 2012, 116, 12644−12648
The Journal of Physical Chemistry C
Article
Figure 1. Left: Final configuration of a single THF molecule on a ZnO (101̅0) surface, obtained by using DFT techniques and MPMD (inset). Right: Another perspective of the final configuration of the system, obtained by DFT calculations. Charge density isosurfaces on the (100) plane have been superimposed to the atomic configuration.
Figure 2. Some stable configurations of a THF molecule on the ZnO surface.
■
SOLVENT THF INTERACTION WITH ZNO The most energetically stable surface of crystalline ZnO is the nonpolar (101̅0), and, hereafter, we will focus on it because it is the most common in ZnO nanostructures. ZnO nanorods typically used in hybrid BHJs9 exhibit six equivalent (101̅0) surfaces, which, on the atomic scale, exhibit trench grooves alternated with rows of ZnO dimers, both oriented along the [010] crystallographic direction (x axis in Figure 1).10 In most cases the local curvature is unimportant because real nanostructures have planar facets of lateral size larger than 10 nm. Accordingly, in our study we focus on the planar (101̅0) crystalline surface. We generate atomistic models of the perfect ZnO surface by cutting a wurtzite ZnO crystal along the (101̅0) plane and by next relaxing the atomic positions in periodic boundary conditions by density functional theory (DFT), as implemented in the Quantum-ESPRESSO code.11 (See the Methods section). The present surface model reproduces the atomic scale features and the energetics already described in the literature for the ZnO.12 We also generate an atomistic model of ZnO surface at MPMD level13 using the LAMMPS code.14 (See the Methods section.)
We start from the THF molecule in different initial positions and orientations over the surface (with the carbon−oxygen ring parallel and perpendicular to it), and we relax the atomic positions by performing MPMD simulations at low temperature, followed by atomic relaxations based on the conjugate gradient method. In all cases the oxygen atom of THF binds to a zinc atom on the surface. In the lowest energy configuration, the molecule turns out to be quasi-vertical with respect to the surface (see Figure 1, inset left panel), its plane being perpendicular to the [100] crystallographic direction. The Zn−O distance is 1.9 Å, and the calculated adhesion energy is found to be as large as 1.15 eV. This molecule-surface binding is very strong, as proved by 10 ns-long room-temperature MPMD simulations: although several different quasi-isoenergetic configurations are indeed explored (with the molecule quasi vertical, as shown in Figure 2 left and center, or parallel to the surface, as shown in Figure 2 right), desorption is never observed. To validate the MPMD result, the minimum energy molecule-surface configuration (inset Figure 1 left) is further relaxed at DFT level (Figure 1 left). After relaxation, we observe a Zn−O bond of length 2.1 Å due to the electrostatic interaction between the positively charged Zn and the negative oxygen of THF and a partial electronic density overlap (Figure 1 right panel). Both MPMD and DFT calculation show that the molecule prefers the twist geometry with its plane slightly tilted with respect to the vertical. (See Figure 1 left and inset.) The adhesion energy, calculated by including the Grimme correction,15 is as large as 0.97 eV in good agreement with the MPMD result (see above). This large ZnO/THF interaction turns out to be larger than both P3HT/P3HT
■
INTERACTION BETWEEN THE THF MOLECULE AND THE ZNO SURFACE To investigate the THF−ZnO interaction we first study a single THF molecule on a ZnO surface by a combination of MPMD and DFT. The force field and the atomic partial charges are described in the Methods section. In particular, MPMD is used to explore carefully the space of configurations and to find the stable molecule geometry on the surface. DFT is used to validate and refine the MPMD result. 12645
dx.doi.org/10.1021/jp302700b | J. Phys. Chem. C 2012, 116, 12644−12648
The Journal of Physical Chemistry C
Article
cohesive energy (0.1 eV/monomer)16 and ZnO/P3HT interaction (0.56 eV/monomer).13
the wetting layer from the remaining liquid. In conclusion, the interface gives rise to a sharp transition in the THF density corresponding to an order/disorder discontinuity in the molecules distribution. The interface between a van der Waals liquid and a hard wall (i.e., solid a surface) has been previously studied.18 Density fluctuations within the liquid phase are expected depending on its ρ* bulk packing density. ρ* is defined as ρ* = ρTHFκ3, where ρTHF is the liquid density and κ is the van der Waals diameter of the liquid molecules. Liquids that are characterized by high bulk packing density (ρ* > 0.8) show a densified region next to the substrate, followed by an oscillating exponentially decaying density profile. This is the case, for example, of cyclohexane on silicon surface, where a densified close-packed liquid layer is found at z ≈ 0.5 nm.18 Away from this layer, a low-density region of width ∼2 nm follows. The present THF/ZnO case is consistent with the above picture. The region C identified in our investigation corresponds to the densified one (i.e., wetting layer) and L′ to the low density region. At variance with the cyclohexane case we do not observe sizable exponential fluctuations, and we attribute this behavior to the actual bonding between the molecules and the hard substrate. The THF molecules have the same orientation on the ZnO hard surface and give rise to a softer methyl-terminated surface that does not induce fluctuations on the remaining liquid. To investigate further the order of the system, we calculate, in each slice, the average structure factor S(k) of the O atoms along the x direction (Figure 4 top):
■
INTERACTION BETWEEN THE THF LIQUID SOLVENT AND ZNO SURFACE AT ROOM TEMPERATURE The next step is to consider a THF liquid solvent interacting with the ZnO surface at room temperature. To obtain a realistic model of the liquid solvent, we start from a simple cubic crystal formed by 216 THF molecules, and we melt it at high temperature. The liquid is then cooled to room temperature and equilibrated in the constant-pressure, constant-temperature (NPT) ensemble under ambient conditions by using a Nosé− Hoover barostat and thermostat. The equilibrium density of the final liquid is found to be 0.879 g/cm3, in agreement with previous theoretical results8 and close to the experimental value 0.884 g/cm3.17 A portion of this liquid is cut and merged to ZnO, and the resulting solid−liquid system is equilibrated at room temperature for 0.2 ns in a simulation cell as large as 45 × 65 × 90 Å. (See Figure 3.) After a few picoseconds we observe
N
S( k ) =
1 |∑ e−ikxj| N j=1
(1)
where xj are the x coordinate of the jth oxygen, N is the number of oxygens in the slice, and k = 2π/λ is the wave vector. S(k) is further averaged during the molecular dynamics run. A similar calculation is performed for the y direction by considering the yj coordinates of atoms instead of xj in eq 1 (Figure 4 bottom). S(k) ≈ 0 when the distribution of atomic positions is disordered (as occurs in a liquid). Conversely, when atoms are periodically distributed within the slice with period λ, then S(k) ≈ 1 for wavevectors satisfying the Bragg condition (i.e., k = 2π(nλ)−1). Accordingly, to investigate the local crystallinity, we calculate the structure factor as a function of k by repeating the calculations in different regions of the system. As for the region C, containing the wetting THF layer, we find peaks at λ = 3.35 Å and λ = 5.47 Å for S(k) along the x and y directions, respectively. These λ values correspond to the lattice periodicity of our ZnO surface, showing a crystalline order in the wetting layer induced by the ZnO surface. If we consider the slice just above the wetting layer (region L′), then the order is lost and a flat low-value S(k) profile is found (see Figure 4, right). The differences in the S(k) profiles along x and y directions are not sizable, and we conclude that, except for the wetting layer, there is no order in THF even close to the interface. This analysis further confirms that, in terms of structure, the THF/ZnO interface is sharp. We next turn to the energetics of the ZnO/THF system by calculating the adhesion energy within the system. To this aim, we consider a plane (hereafter labeled as A/B) that divides the system into two parts, A and B, and we calculate the work (w) necessary to separate them rigidly at increasing distance z. At infinite distance, this work is by definition the adhesion energy
Figure 3. Density profile of ZnO−THF system with respect to the axys perpendicular to the surface. (For clearness, in the picture we do not represent the hydrogens of THF.).
the formation of an ordered (and hereafter stable) monolayer of THF molecules wetting the ZnO surface. Most of the molecules in the layer are stuck on the substrate as in the single molecule case, with the oxygen of THF bound to the zinc atom on the surface, suggesting that part of the THF molecules efficiently bind to ZnO during synthesis in solution. To characterize the interface and its local structure, we divided the simulated system into slices along the z direction. We set z = 0 Å at the ZnO surface. For each slice, we calculate the density ρ and we obtain the density profile ρ(z) reported in Figure 3. Far from the interface, at z < 0 Å and at z > 5 Å, ρ(z) is constant and similar to the value of the ZnO crystal (ρZnO = 5.6 g/cm3) and that of the THF liquid (ρTHF), respectively. The ZnO/THF interface, defined as the regions where dρ/dz ≠ 0, turns out to be as thin as 1 nm, and it consists of the two regions labeled C and L′ in Figure 3. C region corresponds to the crystalline THF layer wetting the ZnO surface. L′ has width 0.5 nm and it corresponds to region where the liquid density is smaller than ρTHF. A visual inspection of the molecular distribution in L′ shows that there is an empty space separating 12646
dx.doi.org/10.1021/jp302700b | J. Phys. Chem. C 2012, 116, 12644−12648
The Journal of Physical Chemistry C
Article
Figure 4. Structure factor in the x (top) and y (bottom) direction for the wetting layer C (left) and the liquid THF close to the surface L′ (right).
γA/B of the two parts, A and B. In the case where a the molecule is cut by the plane, we attribute the whole molecule to the part (A or B) containing its oxygen. The calculated γA/B is directly related to energies of the generated surfaces (σA and σB); in particular, γA/B = σA + σB. We consider three cuts (see Figure 3): (i) L/L, separating two halves of the bulk liquid; (ii) C/L′, separating the crystalline layer C from the neighboring liquid THF layer (L′); and (iii) Z/C, separating the ZnO crystalline surface (Z) from the wetting layer (C). The L/L work of separation (wL/L) as a function of the distance is reported as a red curve in Figure 5, and it is
calculated without relaxing the atomic positions after the cut, and it corresponds to the unrelaxed adhesion energy. If we relax the surfaces, then we find the relaxed adhesion energy γL/L (∼0.059 J/m2). This value corresponds to a surface tension of THF σL = 0.030 N/m, and it can be compared with the experimental value 0.027 N/m.19 As for the C/L′ (black curve of Figure 5), we find that γC/L′ is about 20% lower than γL/L. This means that the wetting layer locally reduces the adhesion of the liquid. In fact, as a result of the crystallinity of the W layer, all of its molecules expose their hydrophobic methylene groups CH2 to the liquid, and the electrostatic interactions with oxygens are reduced in average. Finally, we consider the Z/C cut. In this case, the Zn−O bonds are broken during the separation process, and γZ/C turns out to be ∼0.64 J/m2, that is, one order of magnitude higher than both γL/L and γC/L′.
■
CONCLUSIONS By means of atomistic simulations based on DFT and model potential methods, we have characterized the THF/ZnO interaction. We found that the THF/ZnO binding energy is ∼1 eV, and the solvent efficiently wets the zinc oxide by forming a crystalline (i.e., ordered) monolayer that likely persists after the THF evaporation at room temperature. The interface between the wetting layer and ZnO is sharp in terms of density and local crystallinity, and it lowers the liquid/liquid interaction close to the wetting layer. Accordingly, to this analysis THF is likely present in hybrids after evaporation during the syntesis processes.
Figure 5. Work of separation for C/L′ (black) and L/L (red) cases. The y axis is normalized with respect to γL/L.
■
normalized to the asymptotic value. This work varies until the two semibulks are interacting. At distances larger than the interaction range (z0 ≈ 7 Å) the work reaches the asymptotic value γL/L. Because of the statistic distribution of molecules in the liquid phase, γL/L is found to depend slightly on the position of the cut. For this reason, we averaged the results over different cuts, and we find γL/L ≈ 0.112 J/m2. This value is
METHODS For the calculation of MDPM, the interactions within ZnO were described as the sum of Coulomb and a Buckingham-type two-body potential.20 As for THF, we adopted the AMBER force field,21 including both bonding (bonds, bending, torsional) and nonbonding (van der Waals plus Coulomb) 12647
dx.doi.org/10.1021/jp302700b | J. Phys. Chem. C 2012, 116, 12644−12648
The Journal of Physical Chemistry C
Article
(15) Grimme, S. J. Comput. Chem. 2006, 27, 1787−1799. (16) Melis, C.; Mattoni, A.; Colombo, L. J. Phys. Chem. C 2010, 114, 3401−3406. (17) Marcus, Y. The Properties of Solvents; Wiley: Chichester, U.K., 1998. (18) Doerr, A. K.; Tolan, M.; Schlomka, J.-P.; Press, W. Europhys. Lett. 2000, 52, 330. (19) Pan, C.; Ke, Q.; Ouyang, G.; Zhen, X.; Yang, Y.; Huang, Z. J. Chem. Eng. Data 2004, 49, 1839−1842. (20) Matsui, M.; Akaogi, M. Mol. Simul 1991, 6, 239−244. (21) Ponder, J. W.; Case, D. A. Adv. Protein Chem. 2003, 66, 27−85. (22) Jakalian, A.; Bush, B. L.; Jack, D. B.; Bayly, C. I. J. Comput. Chem. 2000, 21, 132−146. (23) Hockney, R. W.; Eastwood, J. W. Computer Simulation Using Particles; McGraw-Hill: New York, 1981. (24) Perdew, J. P.; Burke, K.; Ernzerhof, M. Phys. Rev. Lett. 1996, 77, 3865−3868.
contributions. For hybrid ZnO/THF interactions, we used a sum of Coulomb and Lennard-Jones contributions.16 The atomic partial charges were calculated according to the standard AM1-BCC method.22 Simulations were performed by the LAMMPS code.14 The velocity Verlet algorithm with a time step of 0.5 fs was used to solve the equations of motion. A particle−particle particle-mesh Ewald algorithm23 was used for the long-range electrostatic forces, and the van der Waals interactions were cut off at 9 Å. DFT calculations have been performed at DFT level by using the Quantum-ESPRESSO11 code. We used Vanderbilt ultrasoft pseudopotentials with the Perdew−Burke−Ernzerhof (PBE) version of the generalized gradient approximation (GGA) exchange-correlation functional.24 Kohn−Sham eigenfunctions have been expanded on a plane-wave basis set by using cutoffs of 30 Ry on the plane waves and of 300 Ry on the electronic density. The Grimme15 correction was used to include the effects of dispersion interactions. In addition to the 13 atoms of the THF, the surface cell contained a 3 × 2 slab formed by four atomic layers of ZnO (48 atoms) and 30 Å of empty space. The electronic properties of the system have been investigated by analyzing the electronic eigenvalues calculated at the Γ point.
■
AUTHOR INFORMATION
Corresponding Author
*E-mail:
[email protected];
[email protected]. Notes
The authors declare no competing financial interest.
■
ACKNOWLEDGMENTS This work has been funded by the Italian Institute of Technology (IIT) under Project SEED “POLYPHEMO” and Regione Autonoma della Sardegna (L.R.7/2007). We acknowledge computational support by CYBERSAR (Cagliari, Italy), CASPUR (Rome, Italy), and IIT Platform Computation.
■
REFERENCES
(1) Ö zgür, U.; Alivov, Y. I.; Liu, C.; Teke, A.; Reshchikov, M. A.; Doŏan, S.; Avrutin, V.; Cho, S.-J.; Morkoç, H. J. Appl. Phys. 2005, 98, 041301-1−041301-103. (2) Wang, Z. L. J. Phys.: Condens. Matter 2004, 16, R829. (3) Bouclé, J.; Ackermann, J. Polym. Int. 2012, 61, 355−373. (4) Middleman, S.; Hochberg, A. K. Process Engineering Analysis in Semiconductor Device Fabrication; McGraw-Hill: New York, 1993. (5) Kwong, C. Y.; Djurišic′, A. B.; Chui, P. C.; Cheng, K. W.; Chan, W. K. Chem. Phys. Lett. 2004, 384, 372−375. (6) Chang, L.; Lademann, H. W. A.; Bonekamp, J.-B.; Meerholz, K.; Moulé, A. J. Adv. Funct. Mater. 2011, 21, 17791787. (7) Bowron, D. T.; Finney, J. L.; Soper, A. K. J. Am. Chem. Soc. 2006, 128, 5119−5126. (8) Girard, S.; Müller-Plathe, F. Mol. Phys. 2003, 101, 779−787. (9) Said, A. J.; Poize, G.; Martini, C.; Ferry, D.; Marine, W.; Giorgio, S.; Fages, F.; Hocq, J.; Boucle, J.; Nelson, J.; Durrant, J. R.; Ackermann, J. J. Chem. Phys. C 2010, 114, 11273−11278. (10) Melis, C.; Colombo, L.; Mattoni, A. J. Phys. Chem. C 2011, 115, 18208−18212. (11) Giannozzi, P.; Baroni, S.; Bonini, N.; Calandra, M.; Car, R.; Cavazzoni, C.; Ceresoli, D.; Chiarotti, G. L.; Cococcioni, M.; Dabo, I.; Corso, A. D.; de Gironcoli, S.; Fabris, S.; Fratesi, G.; Gebauer, R.; et al. J. Phys.: Condens. Matter 2009, 21, 395502−395521. (12) Kulkarni, A. J.; Zhou, M.; Sarasamak, K.; Limpijumnong, S. Phys. Rev. Lett. 2006, 97, 105502−105505. (13) Saba, M. I.; Melis, C.; Colombo, L.; Malloci, G.; Mattoni, A. J. Phys.Chem. C 2011, 115, 9651−9655. (14) Plimpton, S. J. Comput. Phys. 1995, 117, 1−19. 12648
dx.doi.org/10.1021/jp302700b | J. Phys. Chem. C 2012, 116, 12644−12648
