Journal of New Materials for Electrochemical Systems 9, 291-295 (2006) © J. New. Mat. Electrochem. Systems Molecular Dynamics Simulation of Hydrogen Fluoride Mixtures with 1-Ethyl-3-Methylimidazolium Fluoride: Dynamic Features M. Salanne, ∗ C. Simon and P. Turq Université Pierre et Marie Curie-Parisn 6, CNRS, ESPCI, UMR 7612, laboratoire LI2C, case courier 51, 4 place Jussieu, 75252 Paris Cedex 05, France Received: March 01, 2006, Accepted: April 25, 2006 Abstract: Imidazolium fluoride and hydrogen fluoride mixtures are interesting intermediate systems having features of room temperature ionic liquids (RTIL), hydrogen-bonded molecular liquids, and molten salts. Furthermore, they could be used in a wide variety of electrochemical applications, as media for fluorine generation, as solvent for organic chemistry or as electrolytes in batteries or fuel cells. It is possible to model these mixtures with classical interaction potentials known to describe well HF, and combination of LennardJones potentials for the molecular cations. Molecular Dynamics simulations (MD) have been conducted to investigate both structural and transport properties of these mixtures. In this paper, the values of the self-diffusion coefficients are especially commented, and compared with experimental results. Impact of structure of the first solvation shell on the dynamics properties of the ions is also depicted. Keywords: Room temperature ionic liquids, Polyfluoride anions, Molecular dynamics, Diffusion coefficients These features have been attributed to the formation of so-called polyfluoride anions, which are formed of a central fluoride anion surrounded by one or several HF molecules. Evidence for the existence of these species was given by IR spectroscopy [6]. Hagiwara et al. observed unambiguously the formation of the bifluoride F(HF)− , μ-fluoro-bis(fluorohydrogenate) F(HF)2 − and μ3 -fluorotris(fluorohydrogenate) F(HF) 3 − ions. The general topology of those three anions is shown on figure 1. Our previous simulation results [7] agreed with such a speciation and we found that in EMIF-2.3HF the percentage of those 3 species were respectively of 10%, 40% and 50%. In their ab initio molecular dynamics study of KF-2HF, Klein et al. also obtained more complicated ions [8]. Unfortunately, IR spectroscopy is not sufficient to confirm or dismiss the existence of these later species. NMR experiments carried out on those liquids only showed one signal for either fluorine or hydrogen atoms because the exchange of HF molecules is fast compared to the NMR time resolution. The importance of the polyfluoride species in those liquids can be illustrated in writing the reactions involved in the use of EMIF-2.3HF as an electrolyte in a fuel cell: 1. INTRODUCTION Mixtures of fluorinated ionic liquids with hydrogen fluoride (HF) have always raised huge interest in the field of electrochemistry. Among them, the melt KF-2HF is the key system for the production of fluorine through electrolysis [1,2]. Recently the interest in room temperature ionic liquids (RTIL) has severely increased, and mixtures of RTIL with HF were synthetized [3]. Although they were not tested in an industrial sized process, many applications of these liquids, and particularly of the 1-ethyl-3methylimidazolium fluoride mixture with 2.3 equivalents of HF (EMIF-2.3HF), can be previewed. For example, Hagiwara et al. showed that such liquids could be advantageously used as electrolytes in fuel cells. Their physicochemical properties [4] are indeed very interesting: EMIF2.3HF has a conductivity of 100 mS.cm−1 and a viscosity of 4.85 cP at 298K,whereas in “normal” RTILs (i.e. not mixed with HF) σ and η are respectively of the order of 13.5 mS.cm−1 and 31.8 cP (these values correspond to the 1-ethyl-3-methylimidazolium tetrafluoroborate) at such a temperature [5]. ∗ To whom corresponding to: E-mail: [email protected]; Phone: +33144273265; Fax: +33144273228 291 292 M. Salanne et al. / J. New Mat. Electrochem. Systems Table 1. Simulation parameters a (angstroms) b (angstroms) c (angstroms) Temperatures System 1 System 2 System 3 System 4 37.0331 37.0331 37.0331 37.0331 32.1957 32.4866 32.9702 33.2856 41.7108 41.7108 41.7108 41.7108 298 K 320 K 340 K 353 K cal=4.184 J) and distances in angstroms): Figure 1. Typical conformation of the polyfluorides (Black: Fluorine atoms, white: Hydrogen atoms). These represent only instantaneous geometry, and not averaged or most probable geometry. Ions undergo constantly streching, bending, torsions and vibrations. It would be spurious to deduce symmetry of the ions in the liquid on the single basis of these snapshots. Anode: H2 +8F(HF)2 − − → 6F(HF)3 +2e− − − Cathode: 1/2O2 +6F(HF)3 +2e− → H2 O+8F(HF)2 The liquid at composition EMIF-1.3HF can also be used in such an application. In that case the reactions would involve the F(HF)− and F(HF)2 − anions. On the basis of molecular dynamics, deep insight of the speciation could help understanding and modeling kinetics of those processes, and could contribute to the optimization efforts of those electrolytes regarding each specific application. Furthermore molecular dynamics can provide a picture of the diffusion of the species in those liquids, which is also necessary. 2. MODEL AND METHODS 2.1. Interaction potentials We chose the so-called HF3 model developed by Klein et al. for the HF molecules [9]. This model is a three site description and the bond length is 0.9118 Angstroms. Both the hydrogen and fluorine atom carry a charge of +0.68 e and a compensating charge of —1.36 e is placed between them on the molecule axis, at a distance of 0.165 Angstroms from the fluorine ion. The molecules are kept rigid, and the non-Coulombic terms of the intermolecular potential between two of them are of the following form (where all the potentials are expressed in kcal.mol−1 (1 VF F (r) = 2 × 105 exp(−4.25r) − V Disp (r) (1) V Disp (r) = f (r)(220/r6 + 400/r8 + 4500/r10 ) (2) f (r) = exp(−(4.7/r − 1)2 ) if r ≤ 4.7 = 1 else (3) VF H (r) = 2[exp(−10.6(r − 1.6)) − 2 exp(−5.3(r − 1.6))] (4) VHH (r) = 600 exp(−3.34r) (5) To model 1-ethyl-3-methylimidazolium molecules, we used the AMBER force field [10] developed by Cornell et al. In this representation, the potential energy takes the following form: X X Vtotal = Kr (r − req )2 + Kθ (θ − θeq )2 bonds angles X + Vn /2[1 + cos(nϕ − γ)] Xdihedrals 12 6 + [Aij /rij − Bij /rij + qi qj /rij ] (6) i<j All the values used for the various parameters are given in reference [7]; they were mostly taken from a previous study on 1-ethyl-3-methylimidazolium tetrachloroaluminate by de Andrade et al [11]. This combination of potentials appeared to provide excellent description of the structural properties of the liquid [7]. 2.2. Computational details The MD simulations were performed using DLPOLY [12], version 2.14. The Verlet algorithm was used with a timestep of 1 fs to integrate the equations of motion. The system was studied at four different temperatures: 298K, 320K, 340K and 353K. The simulation cell shapes are specified in table 1. The boxes contained 192 EMIF pairs and 440 HF molecules, yielding a HF/EMIF ratio of 2.3 and a total of 5160 atoms. In all cases the density is set to the experimental one [4]. In all runs the pressure fluctuated between 0 and 1 kbar. Five successive 200 ps simulations at increasing temperatures, starting at 2 K, were used to thermalize the system before to perform a 1 ns simulation in the microcanonical ensemble. A cut-off radius of 16 Angstroms was used for the calculation of the short-range interactions, while the long-range electrostatic interactions were treated using the Ewald summation method [13]. Molecular Dynamics Simulation of Hydrogen Fluoride Mixtures with 1-Ethyl -3-Methylimidazolium Fluoride: Dynamic Features / J. New Mat. Electrochem. Systems Figure 2. Mean square displacements of the hydrogen atoms belonging to the cations, as a function of time,at 4 temperatures: 353 K (top curve), 340 K, 320 K and 298 K (bottom curve).The slope of the linear part (t > 75 ps) is proportional to 6 times the self-diffusion coefficient. 293 Figure 3. Arrhenius plots of the self-diffusion coefficients obtained from simulations for hydrogen and fluoride atoms. The activation energies are deduced from the slopes. Table 2. Activation energies (kJ.mol−1 ) 3. RESULTS AND DISCUSSION 3.1. Self-diffusion coefficients of cations and anions Hagiwara et al. have used the pulsed gradient spin-echo NMR technique to measure self-diffusion coefficients [14]. In order to compare our results to the experimental ones, we calculated the diffusion coefficients of hydrogen atoms in the cations and of both the hydrogen and fluorine atoms in the anions. For each temperature, we computed the mean-square displacement (MSD) of each of these species over the full trajectory. The diffusion coefficients are then obtained by plotting the MSD versus time, as shown for the hydrogen atoms of the cations at the various studied temperature on figure 2: the slope of the linear part of the curve is then equal to 6 D. The values obtained for the diffusion coefficients are shown on figure 3: They are roughly one order of magnitude smaller than the experimental ones. However, this difference may be attributed to the model used for the 1-ethyl-3-methylimidazolium cations: in other simulation studies using this model the same kind of deviations from experiments can be observed: calculated diffusion coefficients are generally smaller than the experimental ones. Voth et al. showed that using an electronically polarizable model improved the agreement between experiment and simulation[15]. But, to our knowledge, our study yields the higher values obtained by simulation for an ionic liquid at room temperature. This is consistent with experiments since EMIF-2.3HF has one of the lowest viscosities and of the highest conductivities and diffusion coefficients of this class of products. To go beyond this qualitative agreement between our simulations and experiments, the arrhenian behavior of the Experiments[14] Simulation H cation 16.4 18.4 H anion 16.2 17.2 F anion 14.7 15 system was examined. Plotting log D versus 1/T , we obtained straight lines, allowing determination of activation energies Ea with the relationship: D = D0 exp(−Ea /RT ) (7) Our results for the various atom studied are given in table 2, to be compared with experimental ones. The agreement is excellent, suggesting that the problem with our diffusion coefficients comes from D0 in the previous equation. 3.2. Separating the contribution from the various polyfluorides An advantage of the molecular dynamics simulations is their atomic scale that allows distinguishing the various contributions to observable quantities. For the present system, we can for example calculate the diffusion coefficients of the different polyfluoride anions. An HF molecule is considered bonded to a given fluoride ion if the distance between the hydrogen atom and the fluoride ion is smaller than a cutoff value of 2.1 angstroms. The choice of this criterion is fully discussed in reference [16]. Experimentally, the exchange of HF molecules between two polyfluorides is very fast compared to the NMR timescale and thus their individual contributions cannot be determined. The Arrhenius plots of the diffusion coefficients of the 294 M. Salanne et al. / J. New Mat. Electrochem. Systems Figure 4. Arrhenius plots of the self-diffusion coefficients of the various polyfluorides. At all temperatures, selfdiffusion of higher weight ions is faster than the one of lighter ions. various anions are given on figure 4. It appears that they increase with the number of HF molecules in the cluster: at all the studied temperatures, D(F(HF)3 − ) > D(F(HF)2 − ) > D(F(HF)− ). While surprising, this result is consistent with conductivity results obtained by Hagiwara et al [6]. These authors indeed studied the influence of the HF fraction x (with 1.0 < x < 2.6) in EMIF-xHF and showed that the conductivity increases with x. The diffusion coefficients can be related to the electrical conductivity by the Nernst-Einstein equation: X σ N E = βe2 ρi qi Di (8) where ρi , qi and Di are respectively the number density, charge and diffusion coefficient of species i. In the EMIFxHF melt, the increase of x is associated with an increase of the proportion of F(HF)3 − ions. The higher diffusion coefficient of this species is then sufficient to explain the rise of the electrical conductivity of the liquid. 3.3. Coupling the diffusion to the liquid structure The later result can be surprising as the size of the anion of course increases with the number of HF molecules bonded to the central fluoride anion. One would be tempted to think that a smaller molecule would diffuse easier. Such results are obtained in high temperature molten salts. For example, Madden et al. showed that in the LiCl/KCl eutectic [17], La3+ or Y3+ cations diffuse much slower than the smaller Li+ and K+ . But in such liquids, the smaller mobility of big cations is due to their high coordination number: they diffuse with an intact coordination shell whereas smaller ions diffuse as isolated spherical particles. Thus we calculated the coordination numbers of the 3 varieties of anions present in our system; the results are summarized in figure 5. Figure 5. Percentage of anions with a given coordination number (i.e. number of cation surrounding a given anion). F(HF)- is surrounded by a higher number of cations (its most probable coordination numbers are 3, and then 4) than F(HF)3 - (whose most probable coordination number is only 2). The coordination number decreases when passing from F(HF)− to F(HF)3 − , which is coherent with the screening of the central charge by the HF molecules in the later. The most probable coordination numbers are then 3 to 4 for F(HF)− and 1 to 2 for F(HF)3 − . It is then clear that the later can diffuse more easily, whereas the F(HF)− ions rattle in a cage of imidazolium cations. 4. CONCLUSION In a previous study, we have shown that our model provided excellent structural results, matching very well Xray diffraction experimental results. In the present paper, we have furthermore demonstrated that the model is in qualitative agreement with experiments regarding transport properties, namely self-diffusion coefficients. Quantitative agreement is also obtained for activation energies. The molecular details provided by the simulation allow deeper insight to the contributions of the various species present in the liquid: it is even possible to relate the selfdiffusion coefficients of each polyfluoride ion with their environment. F(HF)3 − , which is the anion containing the more HF molecules, is counter-intuitively the more mobile species: This is due to its smaller coordination number. This latest feature also explains the experimentally observed increase of conductivity with the HF proportion in EMIF-xHF. 5. ACKNOWLEDGEMENT The authors would like to thank professor Rika Hagiwara for fruitful discussions. Molecular Dynamics Simulation of Hydrogen Fluoride Mixtures with 1-Ethyl -3-Methylimidazolium Fluoride: Dynamic Features / J. New Mat. Electrochem. Systems REFERENCES [1] H. Moissan, “Le Fluor et ses composés”, G. STEINHELL, Paris, France, 1900. [2] M. Jaccaud, R. Faron, D. Devilliers, R. Romano, in “Ullmann’s Encyclopedia for Industrial Chemistry”, Ed., Weinheim, VCH, 1988, p293. [3] R. Hagiwara, T. Hirashige, T. Tsuda, Y. Ito, J. Fluorine Chem., 99, 1 (1999). [4] R. Hagiwara, T. Hirashige, T. Tsuda, Y. Ito, J. Electrochem. Soc.,149,D1 (2002). [5] A. Noda, K. Hayamizu, M. Watanabe, J. Phys. Chem. B, 105, 4603 (2001). [6] R. Hagiwara, Y. Nakamori, K. Matsumoto, Y. Ito, J. Phys. Chem. B, 109, 5445 (2005). [7] M. Salanne, C. Simon, P. Turq, J. Phys. Chem. B, 110, 3504 (2006). [8] T. von Rosenvinge, M. Parrinello, M.L. Klein, J. Chem. Phys., 107, 8012 (1997). [9] M.L. Klein, I.R. Mc Donald, J. Chem. Phys., 71, 298 (1979). [10] W.D. Cornell, P. Cieplak, C.I. Bayly, I.R. Gould, K.M. Merz, D.M. Ferguson, D.C. Spellmeyer, T. Fox, J.W. Caldwell, P.A. Kollman, J. Am. Chem. Soc., 117, 5179 (1995). [11] J. de Andrade, E. Boes, H.J. Stassen, J. Phys. Chem. B, 106, 3546 (2002). [12] T.R. Forester, W. Smith, DLPOLY program (1995). [13] P. P. Ewald, Ann. Phys., 36, 253 (1921). [14] Y. Saito, K. Hirai, K. Matsuura, R. Hagiwara, Y. Minamizaki, J. Phys. Chem. B, 109, 2942 (2005). [15] T. Yan, C.J. Burnham, M.G. del Popolo, G.A. Voth, J. Phys. Chem. B, 108, 11877 (2004). [16] C. Simon, T. Cartailler, P. Turq, J. Chem. Phys., 117, 3772 (2002). [17] R. Brookes, A. Davies, G. Ketwaroo, P.A. Madden, J. Phys. Chem. B, 109, 6485 (2005). 295
© Copyright 2026 Paperzz