PHYSICAL REVIEW E 88, 062119 (2013)
Chloride ions induce order-disorder transition at water-oxide interfaces Sanket Deshmukh,1 Ganesh Kamath,2 Shriram Ramanathan,3 and Subramanian K. R. S. Sankaranarayanan1,* 1
Center for Nanoscale Materials, Argonne National Laboratory, Argonne, Illinois 60439, USA 2 Department of Chemistry, University of Missouri-Columbia, Missouri 65211, USA 3 School of Engineering and Applied Sciences, Harvard University, Cambridge, Massachusetts 02138, USA (Received 3 May 2013; published 11 December 2013) Water can form quasi-two-dimensional ordered layers near a solid interface. The solvation dynamics and ionic transport phenomena through this ordered water structure is of direct relevance to a variety of problems in interface science. Molecular dynamics simulations are used to study the impact of local fluctuation of the chloride ion density in the vicinity of an oxide surface on the structure and dynamics of water layers. We demonstrate that local increase in chloride ions beyond a threshold concentration near the water-MgO (100) interface introduces an order-disorder transition of this two-dimensional layered network into bulklike water, leading to increased diffusional characteristics and reduced hydrogen bonding lifetimes. We find that the extent of this order-disorder transition can be tuned by modifying the defect chemistry and nature of the underlying substrate. The kinetic fluidity resulting from order-disorder transition at high chloride ion concentration has significance for a broad range of phenomena, ranging from freezing point depression of brine to onset of aqueous corrosion. DOI: 10.1103/PhysRevE.88.062119
PACS number(s): 61.20.Ja, 07.05.Tp, 68.03.−g, 82.45.Bb
I. INTRODUCTION
The behavior of water at electrochemically active oxide interfaces is of broad relevance to problems ranging from aqueous corrosion and catalysis to environmental chemistry, to name a few [1]. The structure and dynamics of interfacial water are very different from bulk [2]. Near any solid interface, the underlying processes such as diffusive and vibrational motions of water molecules and translational motions of the hydrogen-bonded (HB) network are strongly hindered [3]. Water, therefore, forms well-defined and ordered layers that are quasi-two-dimensional in nature [4]. The highly directional nature of the HB network near interfaces dictates water’s unique chemical and physical properties and ultimately controls many important ionic transport processes. As an example, the atomistic mechanism of halide ion initiated aqueous corrosion of oxidized metal surfaces remains an active area of research. Macroscopic theories and several prior experiments suggest the oxide breakdown to be correlated to a critical chloride ion concentration; its atomistic origin, however, has been a matter of intense debate [5,6]. While existing corrosion models focus on either ion adsorption or transport through passive oxides [7,8], ionic solvation of water near passive oxides could also play an important role [9]. In aqueous corrosion problems, aggressive chloride anions can profoundly alter the dynamics of rearrangement of HB networks and local interfacial ordering [10]. In general, questions pertaining to solvation dynamics near electrochemical interfaces, ion transport across electrochemical double layers, evolution of the near-surface point (charge) defects, as well as compositional modulation in a typical electrochemical environment remain unanswered [11–13]. The solvated ions have important functional roles in screening or moving charge, and influencing the kinetics of breaking and structural relaxation of ion-water and water-water HB in these solutions. These in turn greatly influence the mobility of both the ions and
*
Corresponding author: [email protected]
1539-3755/2013/88(6)/062119(5)
water molecules. Although the properties of bulk water have been well explored [14,15], for interfacial water near charged interfaces such as oxides, the physical picture of HB network connectivity, dynamics, and its response to aggressive anions is still emerging [16]. In this work, we demonstrate how chloride ions profoundly affect the solvation dynamics of water proximal to a model passive oxide from atomistic simulations. Here, we argue that a chloride ion concentration-dependent order-disorder transition in the interfacial layers facilitates the halide ion adsorption onto the passive oxide and can initiate corrosion at the atomistic scale. We demonstrate an atomistic level correlation between the concentration of chloride ions, nature of the solid surface and the molecular order, diffusional characteristics, and HB characteristics of interfacial water. Dynamical correlation functions indicate that chloride ions introduce structural and morphological differences in the quasi-two-dimensional water layer near the MgO(100) interface. Our simulations provide a possible molecular level explanation for the experimentally observed critical chloride ion concentration limit in aqueous corrosion problems. II. COMPUTATIONAL MODEL AND SIMULATION DETAILS
We carry out extensive molecular dynamics (MD) simulations using a polarizable shell model of water molecules near an MgO(100) slab (see Fig. S1 in Ref. [17]). In our MD simulations, we adopt a constant molecule number, pressure, and temperature (NPT) ensemble with periodic boundary conditions in the parallel (x and y) directions and analyze a range of chloride ions (number of ions NCl− ∼ 1–10) in water at room temperature. The simulations were carried out on a model passive oxide system, i.e., a MgO (100) slab containing 2048 Mg2+ ions and 2048 O2− ions. The thickness ˚ The region above the MgO(100) of the MgO slab was ∼28 A. slabs was filled with water molecules. A simulation cell of ˚ with water thickness of ∼44 A ˚ was used. ∼34 × 17 × 72 A The number of water molecules filled (836 molecules) was
062119-1
©2013 American Physical Society
DESHMUKH, KAMATH, RAMANATHAN, AND SANKARANARAYANAN
PHYSICAL REVIEW E 88, 062119 (2013)
chosen to initially equal the bulk density of water and were taken from NPT equilibrated bulk water. The effect of chloride ion concentration was studied by randomly inserting Na+ and Cl− pairs (1, 3, 5, 7, and 10) into the aqueous layer. Na is the counterion used to balance the charges on the chloride ion. The simulations were performed at room temperature (300 K). Reflecting boundary conditions are imposed on the atoms and molecules that might reach the simulation box limit. Equations of motion are integrated with time steps t = 1 fs. The standard Ewald summation method was used to treat Coulombic interactions [18]. We carried out MD simulations over a time scale of 50 ns for each chloride ion concentration. III. RESULTS AND DISCUSSION
To examine the effect of chloride ion concentration on interfacial water structure, we plot the atomic density profile of oxygen in water along the surface normal direction as a function of number of chloride ions (Fig. 1). For water in proximity to the interface, noteworthy in the limit of low NCl− are the layering effects, which arise from interface boundary conditions. In the absence of chloride ions, the density profile shows distinct and well-defined peaks and represents a quasitwo-dimensional ordered system [19]. Layering, observed at low chlorination levels (NCl− < 5), generally hinders water molecules from leaving the layer and facilitates them to form a relatively ordered structure compared with typical bulk liquid water. As NCl− in proximity to the oxide increases, we observe a decrease in the extent of layering as indicated by the merging/broadening of the peaks and a corresponding significant drop in the peak intensities. To further quantify the effect of chloride ions on the structural order, we compute the order map based on translational (t) and orientational (q) order parameters to evaluate the evolution of structural order near the oxide-water interface (Fig. 2) [15]. t measures the tendency of pairs of water molecules to adopt preferential separation. It vanishes for an ideal gas and is large for an ordered structure. q measures the extent to which a molecule and its four nearest neighbors adopt a tetrahedral
FIG. 1. (Color online) Effect of chloride ion concentration on the atomic density profile of water molecules perpendicular to the oxide-water interface. Water forms well-defined ordered layers near the passive oxide. An increase in the chloride ion concentration leads to significant reduction of this layering near the passive oxide.
FIG. 2. (Color online) Order map based on translational and orientational order parameters calculated as a function of chloride ion concentration for water layers near three different interfaces. The stoichiometric defect-free MgO shows an order-disorder transition with increasing chloride ion concentration. Earlier onset of this transition is seen for defective MgO when compared to defect-free MgO. An inert graphene surface (for reference) shows very little change in the order parameters with an increase in chloride ion concentration.
arrangement (Ref. [17]). At low NCl (3 Cl–), the interfacial water layers have a high degree of translational order and low degree of tetrahedral order. At 298 K, starting from pure water, we observe that an increase in NCl− leads to a significant increase in q, whereas t is much less affected. In this range, water has the least orientational order owing to the two-dimensional HB network. Further increase in NCl− leads to an increase in q. The order parameter t is strongly coupled to q in this intermediate concentration range (3 < NCl− 7), causing t also to decrease with increasing NCl− . In this range, HB determines both the mutual orientation and the separation between water molecules. Increasing the NCl− weakens the coupling between t and q. The subsequent increase in NCl− leads to an increase in q with no significant change in t. A maximum in q and minimum in t is attained over the simulated concentration range. At highest NCl− , the t and q order parameters begin to approach that of bulk water [20]. The chloride ion induced disorder is rather striking considering that salts such as NaCl are almost insoluble in ordered structures like ice [21], whereas they are highly soluble in water. In an ice-water system, this leads to the interesting phenomena of chloride ion rejection from the ice into water (brine rejection). MD simulations by Vrbka and Jungwirth have demonstrated that the propensity of water to solvate chloride ions leads to their expulsion from an ordered ice structure into liquid water [22]. The quasi-twodimensional structure near the passive oxide is expected to have a similar tendency to expel chloride ions (especially at high NaCl concentration) and maintain its ordered structure, as in an ice-water system. We, however, observe a unique concentration-dependent order-disorder transition. Clearly, the presence of charged interfaces plays a key role. The surface preference of the chloride ions due to long-range Coulombic interactions seems to overcome the tendency to expel these ions into the bulk of the liquid. As a result, in the ordered layers
062119-2
CHLORIDE IONS INDUCE ORDER-DISORDER . . .
PHYSICAL REVIEW E 88, 062119 (2013)
FIG. 3. (Color online) (a) Diffusivity of chloride ions and (b) water molecules as a function of Cl concentration. (c) The effect of chloride ion concentration on HB characteristics of interfacial water near passive oxide is shown. High NCl results in a progressively faster decay indicating slower HB dynamics in the interfacial layers, whereas at low NCl , a slower decay dynamics corresponding to well-ordered water is evident.
near charged interfaces, one can have high enough chloride ion concentration to induce the order-disorder transition. We also show that altering the nature of the interface allows us to tune this order-disorder transition. We configured two additional surfaces, (a) graphene (which is typically inert) and (b) a defective MgO surface, and compare these with that for a defect-free MgO surface (see supporting information for details). The order map of graphene shows no significant change of the ordered interfacial layer with NCl− . Upon introducing Schottky-type defects, there is an increased surface preference of chloride ions. The order map in Fig. 2 therefore shows an early onset and higher extent of disorder. Prior experimental and theoretical work corroborates the increased adsorption tendency of chloride ions onto defective oxides [23–25]. To investigate the relationship between NCl− , molecular order, and solvation dynamics, we show in Fig. 3(a) the calculated dependence of the diffusion coefficient D of chloride ions and water molecules near the defect-free passive oxide. The MD simulations suggest atomic diffusivity of interfacial water depends strongly on the NCl− (Fig. 3(b) and Fig. S2 in Ref. [17]). We find that the diffusion coefficients of water near MgO(100) at low NCl− are fairly small, comparable to those in MgO(100) without any Cl–ions. At high chlorination levels (>5), however, diffusion of water becomes significant, suggesting facile rearrangement of atoms in the layered structure and significant disruption of the quasitwo-dimensional interfacial water structure. The significant loss of molecular order as a result of rapid kinetic diffusion in turn suggests that rapid adsorption of chloride ions to the passive oxide should be possible at high chlorination levels. The diffusivity variation complies with the HB lifetimes (τ HB ) (Ref. [17]). In bulk water, τ HB are ∼5–6 ps, which is reminiscent of the intermittent collective motion of water molecules as a result of the extensive HB network rearrangement dynamics of liquid water [26]. Near the oxide interface and without any chloride ions, the HB correlation function decays slower with lifetimes of ∼410 ps, which implies characteristics of 2D layered water [Fig. 3(c)].
This slow dynamical behavior indicates energetically stable configurations of interfacial water. This is consistent with the findings of Balasubramanian et al., who observed the residence times and τ HB among interfacial water molecules to be much higher than bulk [27]. The variation in τ HB at low chlorination levels can be explained by the fact that interfacial water molecules are strongly hydrogen bonded to the passive oxide surface and hence have much higher τ HB than pure bulk water. With an increase in NCl− (1–3), the HB correlation function decays slightly faster, leading to reduced τ HB . A strong decrease is seen when NCl− exceeds 5, which coincides with the disruption of the ordered interfacial water. τ HB values at these high chlorination levels are ∼10–30 ps. The local increase in chloride ion concentration beyond a threshold near the water-MgO(100) interface thus introduces an order-disorder transition of this two-dimensional layered network into bulklike water, leading to increased diffusional characteristics and reduced hydrogen bonding lifetimes. These findings are further substantiated by our potential of mean force calculations [28] ([17]), which predict the free energies of interaction of MgO with Cl–. Figure 4 suggests that the free energy of interaction increases with a higher number of chloride ions in the system. Clearly, the chloride ion prefers to be in the vicinity of the MgO surface rather than away from it, especially at high concentrations. The rank order for the different chloride ion concentrations based on the free energies of interaction (G) of chloride ion with the MgO surface decreases in the following order of affinity: 10Cl–> 7Cl–> 5Cl–> 3Cl–> Cl–associated with free energies of MgO-Cl solvation of − 33.0, − 26.05, − 19.7, − 14.05, and − 10.5 kcal mol−1 . Increase in chloride ion concentration results in more favorable interactions of chloride ion with the MgO surface, as suggested by the potential of mean force calculations. The increase in the free energy of interaction is significant at high chloride ion concentrations compared to low concentrations, suggesting that the observations based on the dynamics are supported by energetic considerations too. The influence of chloride ion on the nature of aqueous interfacial layers is readily established by visualizing
062119-3
DESHMUKH, KAMATH, RAMANATHAN, AND SANKARANARAYANAN
structures. Figures 5(a) and 5(b) show typical water molecule arrangement near the passive oxide for a representative low (NCl− = 1) and high Cl concentration (NCl− = 10), respectively. We divide the proximal water into three slabs next to the interface (z ∼ 0:0.5 nm, 0.5 < z < 1.0, and 1.0 < z < 1.5 nm), respectively. These definitions are based on the location of the minima in ρ(z) observed for the zero chloride concentration. Figure 5(a) shows water molecules in the various defined slabs for NCl− = 1. Molecules next to the walls arrange in a crystal-like structure and do not diffuse. In this layer, the oxygen of water molecules arranges in the same lattice as that corresponding to MgO(100). The crystal-like structure of this slab is, therefore, templated by oxide. The extent of ordering decreases with distance from the interface. Water in the third slab is liquid-like; it shows no long-range order and molecules can readily diffuse. Figure 5(b) shows the typical arrangements of water molecules in various slabs for NCl− = 10. We observe several isolated defects, primarily vacancies. These are transient and appear and disappear randomly across the slab. As NCl− increases, the water structure is more disordered. Consistent with the order map in Fig. 2, our visual inspection of MD trajectories suggests that the extent of disorder is a discontinuous function of NCl− and abruptly decreases when NCl− > 5. Comparing the evolution of interfacial water, from Fig. 5(a) to Fig. 5(b), it seems that an increase in NCl− is associated with an order-to-disorder transformation analogous to “melting” of the interfacial water layers. Although no cavitation is visible, there is a significant loss of ordering near the interface. Our simulations of the MgO-water-Cl system at 278 K also show similar loss of ordering with an increase in NCl− (Fig. S3 in Ref. [17]). In electrochemical aqueous corrosion, at the macroscopic level, the critical chloride content represents a threshold concentration of Cl− ions sufficient to cause active corrosion or induce pitting [29]. Our simulations suggest that depending on the nature of the substrate, an analogous threshold exists at the atomistic scale, beyond which solvation dynamics near the passive oxide is significantly affected. Local increase in chloride ion content significantly increases the diffusional characteristics of water, reduces the hydrogen bonding life-
PHYSICAL REVIEW E 88, 062119 (2013)
FIG. 5. (Color online) Simulation snapshot at t = 10 ns illustrating the structural change in ordering of interfacial water layers as a function of chloride ion concentration. Molecules are located in (a) a slab of width 0.5 nm next to the interface (0 < z < 0.5 nm), (c) 2nd layer at (0.5 nm < z < 1 nm), and (e) 3rd layer at (1.0 nm < z < 1.5 nm). (a, c, e) Water layers for 1 NaCl. (b, d, f) The same as (a, c, e) but for water molecules with 10 NaCl, respectively. At low NCl− , water molecules next to the interface arrange in an ordered (crystal-like) structure. The extent of order decreases with increasing distance from the interface. As NCl− increases, we observe an order-disorder transition in the interfacial layers.
times, and disrupts the layering of interfacial water. This kinetic fluidity resulting from order-disorder transition at high NCl− in turn facilitates rapid adsorption of chloride ions to the oxide, which might subsequently initiate corrosion at the atomistic scale. It is important to note that the ion adsorption and their near-surface distribution are affected by factors such as the Lennard-Jones parameters of the ions. A change in the potential well depth ε influences strength of the repulsive interactions, whereas a change in the Lennard-Jones diameter (σ ) influences the packing density. The results presented in this work focus on a particular type of halide ion (chloride ion). If we choose a different ion (from the Hofmeister series, such as F−, Br− , or I− ), then the concentration dependence on order-disorder transition is likely to vary accordingly. IV. CONCLUSIONS
FIG. 4. (Color online) Potential of mean force calculations to determine the free energies of interaction of chlorine with the MgO surface. The potential of mean force is plotted as a function of the number of chloride ions.
To conclude, we present atomistic simulation results showing a strong correlation between chloride ion concentration and disorder at a water-oxide interface. An increase in local concentration of chloride ions near a charged interface such as MgO(100) leads to significant disruption of the quasi-two-dimensional water layers, which approaches bulklike behavior with increasing chloride concentration. The associated decrease in HB lifetime and increased diffusional characteristics of water is postulated to originate from the complex interplays between ion-water, ion-oxide, oxide-water and water-water interactions. Although the simulation findings are discussed in the context of aqueous corrosion phenomena, concentration-dependent order-disorder transitions observed near the aqueous-oxide interface might be a general phenomenon of relevance to other electrochemical corrosion and energy problems.
062119-4
CHLORIDE IONS INDUCE ORDER-DISORDER . . .
PHYSICAL REVIEW E 88, 062119 (2013)
ACKNOWLEDGMENTS
Use of the Center for Nanoscale Materials was supported by the U.S. Department of Energy, Office of Science, Office
[1] H. J. Bakker, Nat. Chem. 1, 24 (2009). [2] Y. He, A. Tilocca, O. Dulub, A. Selloni, and U. Diebold, Nat. Mater. 8, 585 (2009). [3] A. Debnath, K. G. Ayappa, and P. K. Maiti, Phys. Rev. Lett. 110, 018303 (2013). [4] S. A. Deshmukh and S. K. R. S. Sankaranarayanan, Phys. Chem. Chem. Phys. 14, 15593 (2012). [5] P. Marcus and J.-M. Herbelin, Corros. Sci. 34, 1123 (1993). [6] J. A. Bardwell, B. MacDougall, and G. I. Sproule, J. Electrochem. Soc. 136, 1331 (1989). [7] D. A. Little, M. A. Jakab, and J. R. Scully, Corrosion 62 300 (2006). [8] L. J. Criscenti, R. T. Cygan, A. S. Kooser, and H. K. Moffat, Chem. Mater. 20, 4682 (2008). [9] S. Park, D. E. Moilanen, and M. D. Fayer, J. Phys. Chem. B 112, 5279 (2008). [10] F. Bruni, S. Imberti, R. Mancinelli, and M. A. Ricci, J. Chem. Phys. 136, 064520 (2012). [11] G. Gilmer, Science 208, 355 (1980). [12] B. Jeon, S. K. R. S. Sankaranarayanan, A. C. T. van Duin, and S. Ramanathan, Philos. Mag. 91, 4073 (2011). [13] S. K. R. S. Sankaranarayanan and S. Ramanathan, J. Phys. Chem. C 114, 6631 (2010). [14] O. Mishima and H. E. Stanley, Nature (London) 396, 329 (1998). [15] J. R. Errington and P. G. Debenedetti, Nature (London) 409, 318 (2001).
of Basic Energy Sciences, under Contract No. DE-AC0206CH11357. S.R. acknowledges the ONR for support under Grant No. N00014-10-1-0346.
[16] D. E. Moilanen, D. Wong, D. E. Rosenfeld, E. E. Fenn, and M. D. Fayer, Proc. Nat. Acad. Sci. USA 106, 375 (2009). [17] See Supplemental Material at http://link.aps.org/supplemental/ 10.1103/PhysRevE.88.062119 for computational details and additional discussion. [18] P. P. Ewald, Ann. Phys. (Leipzeg) 64, 253 (1921). [19] J. Gao, W. D. Luedtke, and U. Landman, Phys. Rev. Lett. 79, 705 (1997). [20] S. Chatterjee, P. G. Debenedetti, F. H. Stillinger, and R. M. Lynden-Bell, J. Chem. Phys. 128, 124511 (2008). [21] V. F W. Petrenko and R. W. Whitworth, Physics of Ice (Oxford University Press, Oxford, UK, 1999). [22] L. Vrbka and P. Jungwirth, Phys. Rev. Lett. 95, 148501 (2005). [23] B. S. Jeon, S. Sankaranarayanan, A. C. T. van Duin, and S. Ramanathan, ACS Appl. Mater. Interfaces 4, 1225 (2012). [24] C. L. Chang, S. K. R. S. Sankaranarayanan, M. Engelhard, V. Shutthanandan, and S. Ramanathan, J. Phys. Chem. C 113, 3502 (2009). [25] G. S. Frankel, J. Electrochem. Soc. 145, 2186 (1998). [26] M. Matsumoto, S. Saito, and I. Ohmine, Nature (London) 416, 409 (2002). [27] S. Balasubramanian, S. Pal, and B. Bagchi, Phys. Rev. Lett. 89, 115505 (2002). [28] J. Henin, G. Fiorin, C. Chipot, and M. L. Klein, J. Chem. Theory Comput. 6, 35 (2010). [29] C. Punckt, M. B¨olscher, H. Rotermund, A. Mikhailov, L. Organ, N. Budiansky, J. Scully, and J. Hudson, Science 305, 1133 (2004).
062119-5
Chloride ions induce order-disorder transition at water-oxide interfaces Sanket Deshmukh,1 Ganesh Kamath,2 Shriram Ramanathan,3 and Subramanian K. R. S. Sankaranarayanan1,* 1
Center for Nanoscale Materials, Argonne National Laboratory, Argonne, Illinois 60439, USA 2 Department of Chemistry, University of Missouri-Columbia, Missouri 65211, USA 3 School of Engineering and Applied Sciences, Harvard University, Cambridge, Massachusetts 02138, USA (Received 3 May 2013; published 11 December 2013) Water can form quasi-two-dimensional ordered layers near a solid interface. The solvation dynamics and ionic transport phenomena through this ordered water structure is of direct relevance to a variety of problems in interface science. Molecular dynamics simulations are used to study the impact of local fluctuation of the chloride ion density in the vicinity of an oxide surface on the structure and dynamics of water layers. We demonstrate that local increase in chloride ions beyond a threshold concentration near the water-MgO (100) interface introduces an order-disorder transition of this two-dimensional layered network into bulklike water, leading to increased diffusional characteristics and reduced hydrogen bonding lifetimes. We find that the extent of this order-disorder transition can be tuned by modifying the defect chemistry and nature of the underlying substrate. The kinetic fluidity resulting from order-disorder transition at high chloride ion concentration has significance for a broad range of phenomena, ranging from freezing point depression of brine to onset of aqueous corrosion. DOI: 10.1103/PhysRevE.88.062119
PACS number(s): 61.20.Ja, 07.05.Tp, 68.03.−g, 82.45.Bb
I. INTRODUCTION
The behavior of water at electrochemically active oxide interfaces is of broad relevance to problems ranging from aqueous corrosion and catalysis to environmental chemistry, to name a few [1]. The structure and dynamics of interfacial water are very different from bulk [2]. Near any solid interface, the underlying processes such as diffusive and vibrational motions of water molecules and translational motions of the hydrogen-bonded (HB) network are strongly hindered [3]. Water, therefore, forms well-defined and ordered layers that are quasi-two-dimensional in nature [4]. The highly directional nature of the HB network near interfaces dictates water’s unique chemical and physical properties and ultimately controls many important ionic transport processes. As an example, the atomistic mechanism of halide ion initiated aqueous corrosion of oxidized metal surfaces remains an active area of research. Macroscopic theories and several prior experiments suggest the oxide breakdown to be correlated to a critical chloride ion concentration; its atomistic origin, however, has been a matter of intense debate [5,6]. While existing corrosion models focus on either ion adsorption or transport through passive oxides [7,8], ionic solvation of water near passive oxides could also play an important role [9]. In aqueous corrosion problems, aggressive chloride anions can profoundly alter the dynamics of rearrangement of HB networks and local interfacial ordering [10]. In general, questions pertaining to solvation dynamics near electrochemical interfaces, ion transport across electrochemical double layers, evolution of the near-surface point (charge) defects, as well as compositional modulation in a typical electrochemical environment remain unanswered [11–13]. The solvated ions have important functional roles in screening or moving charge, and influencing the kinetics of breaking and structural relaxation of ion-water and water-water HB in these solutions. These in turn greatly influence the mobility of both the ions and
*
Corresponding author: [email protected]
1539-3755/2013/88(6)/062119(5)
water molecules. Although the properties of bulk water have been well explored [14,15], for interfacial water near charged interfaces such as oxides, the physical picture of HB network connectivity, dynamics, and its response to aggressive anions is still emerging [16]. In this work, we demonstrate how chloride ions profoundly affect the solvation dynamics of water proximal to a model passive oxide from atomistic simulations. Here, we argue that a chloride ion concentration-dependent order-disorder transition in the interfacial layers facilitates the halide ion adsorption onto the passive oxide and can initiate corrosion at the atomistic scale. We demonstrate an atomistic level correlation between the concentration of chloride ions, nature of the solid surface and the molecular order, diffusional characteristics, and HB characteristics of interfacial water. Dynamical correlation functions indicate that chloride ions introduce structural and morphological differences in the quasi-two-dimensional water layer near the MgO(100) interface. Our simulations provide a possible molecular level explanation for the experimentally observed critical chloride ion concentration limit in aqueous corrosion problems. II. COMPUTATIONAL MODEL AND SIMULATION DETAILS
We carry out extensive molecular dynamics (MD) simulations using a polarizable shell model of water molecules near an MgO(100) slab (see Fig. S1 in Ref. [17]). In our MD simulations, we adopt a constant molecule number, pressure, and temperature (NPT) ensemble with periodic boundary conditions in the parallel (x and y) directions and analyze a range of chloride ions (number of ions NCl− ∼ 1–10) in water at room temperature. The simulations were carried out on a model passive oxide system, i.e., a MgO (100) slab containing 2048 Mg2+ ions and 2048 O2− ions. The thickness ˚ The region above the MgO(100) of the MgO slab was ∼28 A. slabs was filled with water molecules. A simulation cell of ˚ with water thickness of ∼44 A ˚ was used. ∼34 × 17 × 72 A The number of water molecules filled (836 molecules) was
062119-1
©2013 American Physical Society
DESHMUKH, KAMATH, RAMANATHAN, AND SANKARANARAYANAN
PHYSICAL REVIEW E 88, 062119 (2013)
chosen to initially equal the bulk density of water and were taken from NPT equilibrated bulk water. The effect of chloride ion concentration was studied by randomly inserting Na+ and Cl− pairs (1, 3, 5, 7, and 10) into the aqueous layer. Na is the counterion used to balance the charges on the chloride ion. The simulations were performed at room temperature (300 K). Reflecting boundary conditions are imposed on the atoms and molecules that might reach the simulation box limit. Equations of motion are integrated with time steps t = 1 fs. The standard Ewald summation method was used to treat Coulombic interactions [18]. We carried out MD simulations over a time scale of 50 ns for each chloride ion concentration. III. RESULTS AND DISCUSSION
To examine the effect of chloride ion concentration on interfacial water structure, we plot the atomic density profile of oxygen in water along the surface normal direction as a function of number of chloride ions (Fig. 1). For water in proximity to the interface, noteworthy in the limit of low NCl− are the layering effects, which arise from interface boundary conditions. In the absence of chloride ions, the density profile shows distinct and well-defined peaks and represents a quasitwo-dimensional ordered system [19]. Layering, observed at low chlorination levels (NCl− < 5), generally hinders water molecules from leaving the layer and facilitates them to form a relatively ordered structure compared with typical bulk liquid water. As NCl− in proximity to the oxide increases, we observe a decrease in the extent of layering as indicated by the merging/broadening of the peaks and a corresponding significant drop in the peak intensities. To further quantify the effect of chloride ions on the structural order, we compute the order map based on translational (t) and orientational (q) order parameters to evaluate the evolution of structural order near the oxide-water interface (Fig. 2) [15]. t measures the tendency of pairs of water molecules to adopt preferential separation. It vanishes for an ideal gas and is large for an ordered structure. q measures the extent to which a molecule and its four nearest neighbors adopt a tetrahedral
FIG. 1. (Color online) Effect of chloride ion concentration on the atomic density profile of water molecules perpendicular to the oxide-water interface. Water forms well-defined ordered layers near the passive oxide. An increase in the chloride ion concentration leads to significant reduction of this layering near the passive oxide.
FIG. 2. (Color online) Order map based on translational and orientational order parameters calculated as a function of chloride ion concentration for water layers near three different interfaces. The stoichiometric defect-free MgO shows an order-disorder transition with increasing chloride ion concentration. Earlier onset of this transition is seen for defective MgO when compared to defect-free MgO. An inert graphene surface (for reference) shows very little change in the order parameters with an increase in chloride ion concentration.
arrangement (Ref. [17]). At low NCl (3 Cl–), the interfacial water layers have a high degree of translational order and low degree of tetrahedral order. At 298 K, starting from pure water, we observe that an increase in NCl− leads to a significant increase in q, whereas t is much less affected. In this range, water has the least orientational order owing to the two-dimensional HB network. Further increase in NCl− leads to an increase in q. The order parameter t is strongly coupled to q in this intermediate concentration range (3 < NCl− 7), causing t also to decrease with increasing NCl− . In this range, HB determines both the mutual orientation and the separation between water molecules. Increasing the NCl− weakens the coupling between t and q. The subsequent increase in NCl− leads to an increase in q with no significant change in t. A maximum in q and minimum in t is attained over the simulated concentration range. At highest NCl− , the t and q order parameters begin to approach that of bulk water [20]. The chloride ion induced disorder is rather striking considering that salts such as NaCl are almost insoluble in ordered structures like ice [21], whereas they are highly soluble in water. In an ice-water system, this leads to the interesting phenomena of chloride ion rejection from the ice into water (brine rejection). MD simulations by Vrbka and Jungwirth have demonstrated that the propensity of water to solvate chloride ions leads to their expulsion from an ordered ice structure into liquid water [22]. The quasi-twodimensional structure near the passive oxide is expected to have a similar tendency to expel chloride ions (especially at high NaCl concentration) and maintain its ordered structure, as in an ice-water system. We, however, observe a unique concentration-dependent order-disorder transition. Clearly, the presence of charged interfaces plays a key role. The surface preference of the chloride ions due to long-range Coulombic interactions seems to overcome the tendency to expel these ions into the bulk of the liquid. As a result, in the ordered layers
062119-2
CHLORIDE IONS INDUCE ORDER-DISORDER . . .
PHYSICAL REVIEW E 88, 062119 (2013)
FIG. 3. (Color online) (a) Diffusivity of chloride ions and (b) water molecules as a function of Cl concentration. (c) The effect of chloride ion concentration on HB characteristics of interfacial water near passive oxide is shown. High NCl results in a progressively faster decay indicating slower HB dynamics in the interfacial layers, whereas at low NCl , a slower decay dynamics corresponding to well-ordered water is evident.
near charged interfaces, one can have high enough chloride ion concentration to induce the order-disorder transition. We also show that altering the nature of the interface allows us to tune this order-disorder transition. We configured two additional surfaces, (a) graphene (which is typically inert) and (b) a defective MgO surface, and compare these with that for a defect-free MgO surface (see supporting information for details). The order map of graphene shows no significant change of the ordered interfacial layer with NCl− . Upon introducing Schottky-type defects, there is an increased surface preference of chloride ions. The order map in Fig. 2 therefore shows an early onset and higher extent of disorder. Prior experimental and theoretical work corroborates the increased adsorption tendency of chloride ions onto defective oxides [23–25]. To investigate the relationship between NCl− , molecular order, and solvation dynamics, we show in Fig. 3(a) the calculated dependence of the diffusion coefficient D of chloride ions and water molecules near the defect-free passive oxide. The MD simulations suggest atomic diffusivity of interfacial water depends strongly on the NCl− (Fig. 3(b) and Fig. S2 in Ref. [17]). We find that the diffusion coefficients of water near MgO(100) at low NCl− are fairly small, comparable to those in MgO(100) without any Cl–ions. At high chlorination levels (>5), however, diffusion of water becomes significant, suggesting facile rearrangement of atoms in the layered structure and significant disruption of the quasitwo-dimensional interfacial water structure. The significant loss of molecular order as a result of rapid kinetic diffusion in turn suggests that rapid adsorption of chloride ions to the passive oxide should be possible at high chlorination levels. The diffusivity variation complies with the HB lifetimes (τ HB ) (Ref. [17]). In bulk water, τ HB are ∼5–6 ps, which is reminiscent of the intermittent collective motion of water molecules as a result of the extensive HB network rearrangement dynamics of liquid water [26]. Near the oxide interface and without any chloride ions, the HB correlation function decays slower with lifetimes of ∼410 ps, which implies characteristics of 2D layered water [Fig. 3(c)].
This slow dynamical behavior indicates energetically stable configurations of interfacial water. This is consistent with the findings of Balasubramanian et al., who observed the residence times and τ HB among interfacial water molecules to be much higher than bulk [27]. The variation in τ HB at low chlorination levels can be explained by the fact that interfacial water molecules are strongly hydrogen bonded to the passive oxide surface and hence have much higher τ HB than pure bulk water. With an increase in NCl− (1–3), the HB correlation function decays slightly faster, leading to reduced τ HB . A strong decrease is seen when NCl− exceeds 5, which coincides with the disruption of the ordered interfacial water. τ HB values at these high chlorination levels are ∼10–30 ps. The local increase in chloride ion concentration beyond a threshold near the water-MgO(100) interface thus introduces an order-disorder transition of this two-dimensional layered network into bulklike water, leading to increased diffusional characteristics and reduced hydrogen bonding lifetimes. These findings are further substantiated by our potential of mean force calculations [28] ([17]), which predict the free energies of interaction of MgO with Cl–. Figure 4 suggests that the free energy of interaction increases with a higher number of chloride ions in the system. Clearly, the chloride ion prefers to be in the vicinity of the MgO surface rather than away from it, especially at high concentrations. The rank order for the different chloride ion concentrations based on the free energies of interaction (G) of chloride ion with the MgO surface decreases in the following order of affinity: 10Cl–> 7Cl–> 5Cl–> 3Cl–> Cl–associated with free energies of MgO-Cl solvation of − 33.0, − 26.05, − 19.7, − 14.05, and − 10.5 kcal mol−1 . Increase in chloride ion concentration results in more favorable interactions of chloride ion with the MgO surface, as suggested by the potential of mean force calculations. The increase in the free energy of interaction is significant at high chloride ion concentrations compared to low concentrations, suggesting that the observations based on the dynamics are supported by energetic considerations too. The influence of chloride ion on the nature of aqueous interfacial layers is readily established by visualizing
062119-3
DESHMUKH, KAMATH, RAMANATHAN, AND SANKARANARAYANAN
structures. Figures 5(a) and 5(b) show typical water molecule arrangement near the passive oxide for a representative low (NCl− = 1) and high Cl concentration (NCl− = 10), respectively. We divide the proximal water into three slabs next to the interface (z ∼ 0:0.5 nm, 0.5 < z < 1.0, and 1.0 < z < 1.5 nm), respectively. These definitions are based on the location of the minima in ρ(z) observed for the zero chloride concentration. Figure 5(a) shows water molecules in the various defined slabs for NCl− = 1. Molecules next to the walls arrange in a crystal-like structure and do not diffuse. In this layer, the oxygen of water molecules arranges in the same lattice as that corresponding to MgO(100). The crystal-like structure of this slab is, therefore, templated by oxide. The extent of ordering decreases with distance from the interface. Water in the third slab is liquid-like; it shows no long-range order and molecules can readily diffuse. Figure 5(b) shows the typical arrangements of water molecules in various slabs for NCl− = 10. We observe several isolated defects, primarily vacancies. These are transient and appear and disappear randomly across the slab. As NCl− increases, the water structure is more disordered. Consistent with the order map in Fig. 2, our visual inspection of MD trajectories suggests that the extent of disorder is a discontinuous function of NCl− and abruptly decreases when NCl− > 5. Comparing the evolution of interfacial water, from Fig. 5(a) to Fig. 5(b), it seems that an increase in NCl− is associated with an order-to-disorder transformation analogous to “melting” of the interfacial water layers. Although no cavitation is visible, there is a significant loss of ordering near the interface. Our simulations of the MgO-water-Cl system at 278 K also show similar loss of ordering with an increase in NCl− (Fig. S3 in Ref. [17]). In electrochemical aqueous corrosion, at the macroscopic level, the critical chloride content represents a threshold concentration of Cl− ions sufficient to cause active corrosion or induce pitting [29]. Our simulations suggest that depending on the nature of the substrate, an analogous threshold exists at the atomistic scale, beyond which solvation dynamics near the passive oxide is significantly affected. Local increase in chloride ion content significantly increases the diffusional characteristics of water, reduces the hydrogen bonding life-
PHYSICAL REVIEW E 88, 062119 (2013)
FIG. 5. (Color online) Simulation snapshot at t = 10 ns illustrating the structural change in ordering of interfacial water layers as a function of chloride ion concentration. Molecules are located in (a) a slab of width 0.5 nm next to the interface (0 < z < 0.5 nm), (c) 2nd layer at (0.5 nm < z < 1 nm), and (e) 3rd layer at (1.0 nm < z < 1.5 nm). (a, c, e) Water layers for 1 NaCl. (b, d, f) The same as (a, c, e) but for water molecules with 10 NaCl, respectively. At low NCl− , water molecules next to the interface arrange in an ordered (crystal-like) structure. The extent of order decreases with increasing distance from the interface. As NCl− increases, we observe an order-disorder transition in the interfacial layers.
times, and disrupts the layering of interfacial water. This kinetic fluidity resulting from order-disorder transition at high NCl− in turn facilitates rapid adsorption of chloride ions to the oxide, which might subsequently initiate corrosion at the atomistic scale. It is important to note that the ion adsorption and their near-surface distribution are affected by factors such as the Lennard-Jones parameters of the ions. A change in the potential well depth ε influences strength of the repulsive interactions, whereas a change in the Lennard-Jones diameter (σ ) influences the packing density. The results presented in this work focus on a particular type of halide ion (chloride ion). If we choose a different ion (from the Hofmeister series, such as F−, Br− , or I− ), then the concentration dependence on order-disorder transition is likely to vary accordingly. IV. CONCLUSIONS
FIG. 4. (Color online) Potential of mean force calculations to determine the free energies of interaction of chlorine with the MgO surface. The potential of mean force is plotted as a function of the number of chloride ions.
To conclude, we present atomistic simulation results showing a strong correlation between chloride ion concentration and disorder at a water-oxide interface. An increase in local concentration of chloride ions near a charged interface such as MgO(100) leads to significant disruption of the quasi-two-dimensional water layers, which approaches bulklike behavior with increasing chloride concentration. The associated decrease in HB lifetime and increased diffusional characteristics of water is postulated to originate from the complex interplays between ion-water, ion-oxide, oxide-water and water-water interactions. Although the simulation findings are discussed in the context of aqueous corrosion phenomena, concentration-dependent order-disorder transitions observed near the aqueous-oxide interface might be a general phenomenon of relevance to other electrochemical corrosion and energy problems.
062119-4
CHLORIDE IONS INDUCE ORDER-DISORDER . . .
PHYSICAL REVIEW E 88, 062119 (2013)
ACKNOWLEDGMENTS
Use of the Center for Nanoscale Materials was supported by the U.S. Department of Energy, Office of Science, Office
[1] H. J. Bakker, Nat. Chem. 1, 24 (2009). [2] Y. He, A. Tilocca, O. Dulub, A. Selloni, and U. Diebold, Nat. Mater. 8, 585 (2009). [3] A. Debnath, K. G. Ayappa, and P. K. Maiti, Phys. Rev. Lett. 110, 018303 (2013). [4] S. A. Deshmukh and S. K. R. S. Sankaranarayanan, Phys. Chem. Chem. Phys. 14, 15593 (2012). [5] P. Marcus and J.-M. Herbelin, Corros. Sci. 34, 1123 (1993). [6] J. A. Bardwell, B. MacDougall, and G. I. Sproule, J. Electrochem. Soc. 136, 1331 (1989). [7] D. A. Little, M. A. Jakab, and J. R. Scully, Corrosion 62 300 (2006). [8] L. J. Criscenti, R. T. Cygan, A. S. Kooser, and H. K. Moffat, Chem. Mater. 20, 4682 (2008). [9] S. Park, D. E. Moilanen, and M. D. Fayer, J. Phys. Chem. B 112, 5279 (2008). [10] F. Bruni, S. Imberti, R. Mancinelli, and M. A. Ricci, J. Chem. Phys. 136, 064520 (2012). [11] G. Gilmer, Science 208, 355 (1980). [12] B. Jeon, S. K. R. S. Sankaranarayanan, A. C. T. van Duin, and S. Ramanathan, Philos. Mag. 91, 4073 (2011). [13] S. K. R. S. Sankaranarayanan and S. Ramanathan, J. Phys. Chem. C 114, 6631 (2010). [14] O. Mishima and H. E. Stanley, Nature (London) 396, 329 (1998). [15] J. R. Errington and P. G. Debenedetti, Nature (London) 409, 318 (2001).
of Basic Energy Sciences, under Contract No. DE-AC0206CH11357. S.R. acknowledges the ONR for support under Grant No. N00014-10-1-0346.
[16] D. E. Moilanen, D. Wong, D. E. Rosenfeld, E. E. Fenn, and M. D. Fayer, Proc. Nat. Acad. Sci. USA 106, 375 (2009). [17] See Supplemental Material at http://link.aps.org/supplemental/ 10.1103/PhysRevE.88.062119 for computational details and additional discussion. [18] P. P. Ewald, Ann. Phys. (Leipzeg) 64, 253 (1921). [19] J. Gao, W. D. Luedtke, and U. Landman, Phys. Rev. Lett. 79, 705 (1997). [20] S. Chatterjee, P. G. Debenedetti, F. H. Stillinger, and R. M. Lynden-Bell, J. Chem. Phys. 128, 124511 (2008). [21] V. F W. Petrenko and R. W. Whitworth, Physics of Ice (Oxford University Press, Oxford, UK, 1999). [22] L. Vrbka and P. Jungwirth, Phys. Rev. Lett. 95, 148501 (2005). [23] B. S. Jeon, S. Sankaranarayanan, A. C. T. van Duin, and S. Ramanathan, ACS Appl. Mater. Interfaces 4, 1225 (2012). [24] C. L. Chang, S. K. R. S. Sankaranarayanan, M. Engelhard, V. Shutthanandan, and S. Ramanathan, J. Phys. Chem. C 113, 3502 (2009). [25] G. S. Frankel, J. Electrochem. Soc. 145, 2186 (1998). [26] M. Matsumoto, S. Saito, and I. Ohmine, Nature (London) 416, 409 (2002). [27] S. Balasubramanian, S. Pal, and B. Bagchi, Phys. Rev. Lett. 89, 115505 (2002). [28] J. Henin, G. Fiorin, C. Chipot, and M. L. Klein, J. Chem. Theory Comput. 6, 35 (2010). [29] C. Punckt, M. B¨olscher, H. Rotermund, A. Mikhailov, L. Organ, N. Budiansky, J. Scully, and J. Hudson, Science 305, 1133 (2004).
062119-5
