Computational Modeling of the Structure and Properties of Zr(OH)4

Computational Modeling of the Structure and Properties of Zr(OH)4...

2 downloads 21 Views 6MB Size

Article Cite This: J. Phys. Chem. C XXXX, XXX, XXX−XXX

pubs.acs.org/JPCC

Computational Modeling of the Structure and Properties of Zr(OH)4 Ivan O. Iordanov,† Victor M. Bermudez,*,‡,∥ and Craig K. Knox§ †

U.S. Army Edgewood Chemical and Biological Center, 5183 Blackhawk Road, Aberdeen Proving Ground, Aberdeen, Maryland 21010-5423, United States ‡ Electronics Science and Technology Division, U.S. Naval Research Laboratory, 4555 Overlook Avenue, S.W., Washington, DC, 20375-5347, United States § Science and Technology Corporation, 111 C. Bata Boulevard, Belcamp, Maryland 21017, United States S Supporting Information *

ABSTRACT: Ab initio molecular dynamics (AIMD), based on density functional theory, has been used to develop and test a model for amorphous Zr(OH)4, which is of interest as an agent for the adsorption of toxic gases. Beginning with an idealized and highly ordered structure based on two-dimensional layers of polymeric Zr(OH)4, AIMD at 300 K and above yields an amorphous structure. Disordering is seen to occur concomitantly with a reaction between acidic bridging OH sites and basic terminal OH to produce H2O and Zr−O−Zr bridges. The modeling also shows that Zr(OH)4 can be maintained in an artiﬁcial ordered state either by a regular array of hydrogen bonds within layers or by interaction between layers. Radial distribution functions and distributions of hydrogen-bond parameters are calculated for both the ordered and the amorphous structures, and a comparison between the two sets of results provides insight regarding the nature of the amorphous state. Infrared and proton nuclear magnetic resonance spectra are computed for the amorphous model, and chemical properties are evaluated by examining the acidity of diﬀerent types of OH groups and also the reactions with SO2 and CO2. In particular, various mechanisms leading to sulﬁte formation during exposure to SO2 are examined in some detail. In all cases, good agreement with experiment is found, which indicates that the model is a suitable basis for analyzing the adsorption properties of amorphous Zr(OH)4.

1. INTRODUCTION The eﬃcient removal of hazardous materials from the environment is a subject of continuing and pressing concern. Such species include toxic industrial compounds (TICs) and chemical warfare agents (CWAs) that may have been introduced either accidentally or deliberately into ambient air. Currently the standard material used in air puriﬁcation via adsorption is activated carbon, the applications and limitations of which are well understood.1 It has been shown that amorphous zirconium hydroxide, Zr(OH)4, is remarkably superior to activated carbon in many air-puriﬁcation applications because it not only adsorbs toxic agents but also reacts with them. This results largely from the fact that Zr(OH)4, in the form of a high-surface-area powder, consists of a high density of chemically active OH groups, some of which are acidic in character, whereas others are basic. Coordinatively unsaturated Zr atoms (Zrcus) also act as Lewis acid sites. Attaining a quantitative understanding of the eﬀectiveness of Zr(OH)4 in such applications, in comparison to that of other oxides, is a major factor motivating the on-going studies of this material. The interaction of Zr(OH)4 with TICs such as SO2 (refs 2−4), NO2 (ref 4), NCCl (cyanogen chloride, ref 5), HCl, COCl2 (phosgene), Cl2 (ref 6), and NH3 (ref 7) has been investigated. The use of this material in CWA decontamination has been demonstrated8 for GB (Sarin) and VX, and it has been © XXXX American Chemical Society

found to be particularly eﬀective in the catalytic hydrolysis of VX to form less-toxic compounds. A similar capacity for adsorption and detoxiﬁcation has been shown9 for 2chloroethyl ethylsulﬁde, a simulant for HD (sulfur mustard). It has also been found10 that Zr(OH)4 supported on polymer ﬁber mats is catalytically active in the decontamination of [(CH3)2(H)CO]2(F)PO (diisopropyl ﬂuorophosphonate, a Sarin simulant) via hydrolysis of the P−F bond. It has furthermore been seen11 that mesoporous, high-surface-area Zr(OH) 4 has an exceptionally high capacity for CO 2 adsorption. From a theoretical perspective, the adsorption and reaction of the CWA simulant dimethyl methylphosphonate (DMMP, (CH3O) 2(CH3)PO) on Zr(OH)4 has recently been studied12 using small cluster models. There has also been extensive experimental13−26 and theoretical26,27 work addressing the use of Zr metal-organic frameworks (MOFs) in detoxiﬁcation of CWAs and simulants. The active sites in these crystalline materials are small ZrxOy or ZrxOy(OH)z clusters, with organic ligands, that bear some resemblance to the active sites in amorphous Zr(OH)4. In particular, the theoretical studies show the importance of Zrcus sites (with a coordination Received: November 9, 2017 Revised: January 30, 2018

A

DOI: 10.1021/acs.jpcc.7b11107 J. Phys. Chem. C XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry C

eﬀect on adsorption.39 For example, reaction with a Lewis acid such as SO2 strongly favors terminal OH groups.3 On the other hand, CWAs with a PO group, such as GB and VX, can interact8 with bridging or terminal OH sites depending on the mechanistic details. It is noted that one study9 found a much lower (bridging-OH)/(terminal-OH) concentration ratio of 0.5 using potentiometric titration. This is a wet-chemical method, and it is probable36 that immersion of Zr(OH)4 in an aqueous solution of varying pH changes this quantity from that which pertains to the “dry” conditions that are more relevant to the present study.

number of 7 or less) in the reaction with species containing a phosphoryl (PO) group. Although the potential utility of Zr(OH)4 has been well documented, there is only a limited understanding of the fundamental chemical mechanisms involved and of the relevant structure−property relationships. In the work reported here, a theoretical model is developed for this material that will be used to explore these issues and to serve, in subsequent studies, as an aid in quantitatively interpreting spectroscopic and other data relating to adsorption. After a description of the computational methods, a discussion will be given of the use of ab initio molecular dynamics (AIMD) to formulate a realistic model of Zr(OH)4. Predictions based on this model will then be compared with experimental results relating to structural, spectroscopic, and chemical properties. Although the crystalline forms (cubic, monoclinic, and tetragonal) of ZrO2 have been extensively studied using ﬁrst-principles methods (refs 28−32 and works cited), this is to our knowledge the ﬁrst such treatment of amorphous Zr(OH)4, which is a distinctly diﬀerent material. The point of departure in this study is the idealized structural model (Figure 1) proposed by Clearﬁeld33 and later developed

2. COMPUTATIONAL DETAILS All calculations were performed using density functional theory as implemented, except where noted, in the Quantum Espresso (QE) program suite40 (vers. 5.0 and higher) with the PBE functional. The projector augmented wave (PAW) method was employed with wave function and density cutoﬀs of 40 and 400 Rydbergs (Ry), respectively. The Zr pseudopotential was checked by computing the lattice constants and cohesive energy of bulk Zr metal. The results are a0 = 3.24 and c0 = 5.147 Å for the hexagonal lattice constants and 6.21 eV per atom for the cohesive energy, which are in very good agreement with the experimental lattice constants41 of a0 = 3.2294 and c0 = 5.1414 Å (at T = 4.2 K) and the cohesive energy42 (6.25 eV per atom at T = 0 K). This work entails a study of the adsorption of SO2, and the S and O pseudopotentials were checked by computing the properties of molecular SO2. The results are 1.449 Å and 119.44° for the bond length and angle, which agree well with the experimental values43 of 1.432 Å and 119.54°. As a test of convergence, the adsorption energy (ΔEads) was computed for a range of wave function and density cutoﬀs using DMMP as a test molecule. For nondissociative molecular adsorption, ΔEads was found to be converged to within ∼2 meV at the cutoﬀ values used here. All PAW calculations included a Grimme D2 semiempirical dispersion correction44 to better describe the eﬀects of noncovalent interaction between OH groups. The main purpose of AIMD in the present work is to include, in a realistic way, the eﬀects of temperature on the amorphous structure and reactivity of Zr(OH)4. Born− Oppenheimer (B−O) AIMD calculations within QE employed the Verlet algorithm45 with temperature controlled via velocity scaling. A typical run used a 1.0 femtosecond (fs) time step and extended for >3 picoseconds (ps), with stabilization of the total (ionic + electronic) energy occurring within the ﬁrst ∼0.5 ps. In some runs, H was replaced with D to increase the period of the ν(O−H) stretching mode (τOH ≈ 9.5 fs vs τOD ≈ 13.5 fs), but this had no signiﬁcant eﬀect on the stabilization of the total energy. In the following, however, hydrogen (whether H or D) will be described as “H”. Proton (1H) NMR chemical shifts were obtained using the gauge-including PAW (GIPAW) method.46 Here, structures obtained in AIMD were ﬁrst relaxed in a static (T = 0) calculation with convergence criteria of 1 × 10−5 Ry and 1 × 10−4 Ry/Bohr, respectively, for the total energy and the force on any atom in any direction. For some calculations, B−O AIMD was performed using the CP2K code,47 which employs a hybrid of plane waves and Gaussian basis sets. The PBE functional was used with normconserving pseudopotentials48 and a 400 Ry plane-wave cutoﬀ. The Gaussian basis sets were of double-ζ quality, with polarization functions added, that were optimized49 for molecular calculations. Dispersion was included using the

Figure 1. Schematic model of an ideal, ordered polymeric structure formed from [Zr(OH)4]4 tetramers. H atoms in the OH bridges are not shown. Each Zr atom is joined to four other Zr atoms by two OH bridges. Four of these bridges are shown as heavy lines to emphasize the fundamental structural unit of the polymer. Adapted with permission from ref 37. Copyright 2002 American Chemical Society.

by Mamott et al.,34 Turrillas et al.,35,36 and Southon et al.,37 which is based on a planar polymeric network of [Zr(OH)4]4 tetramer units. Here, Zr atoms form a square-planar array in which each atom is linked to its four Zr neighbors by two OH bridges and is thus 8-fold-coordinated. A typical nanoparticle, grown via sol−gel synthesis, consists37 of from one to three of these polymeric sheets, each with about 50 Zr atoms. Southon et al. characterized their material using X-ray diﬀraction (XRD), extended X-ray absorption ﬁne structure (EXAFS), and Raman spectroscopy. As was recognized by the diﬀerent investigators, it is readily apparent from these results that the ideal structure is only an approximation because XRD and EXAFS indicate an amorphous material. All OH groups in the ideal model, except at the edges of nanoparticles, form bridges between Zr atoms, whereas proton nuclear magnetic resonance (NMR) data8,38,39 reveal the presence of both bridging (Zr−O(H)−Zr) and terminal (Zr− OH) groups with a (bridging-OH)/(terminal-OH) concentration ratio (after mild thermal outgassing in vacuo) in the range of 3.3−3.7 depending on sample preparation. This observation is critical to understanding the chemical functionality of Zr(OH)4 because bridging (terminal) OH groups exhibit an acidic (basic) character, which has a pronounced B

DOI: 10.1021/acs.jpcc.7b11107 J. Phys. Chem. C XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry C Grimme D3 semiempirical approach,50 and the AIMD trajectories were obtained using the Nosé−Hoover thermostat. To test the compatibility of the QE and CP2K results, ΔEads was computed for molecular adsorption of DMMP on a twolayer Zr(OH)4 model (see below) with very similar results (−0.95 and −0.99 eV, respectively, for QE and CP2K). For comparison, a cluster model12 in which DMMP adsorbs via a single hydrogen bond (H-bond) between a bridging OH and the phosphoryl O atom gives ΔEads = −72 kJ/mol = −0.75 eV. (In the present work, a negative ΔEads corresponds to an exothermic process.) The IR spectrum was computed using the Fourier transform of the autocorrelation function of the total system dipole moment, ⟨μ(0)μ(t)⟩, obtained from the CP2K AIMD trajectory. The spectrum is given by51 I(ω) =

ℏω 2 2πkBT

interactions. Because of the large size of the unit cell, all calculations were for the Γ-point only; however, as a test, a calculation for a bilayer using a (3 × 3 × 1) Monkhorst−Pack (MP) grid gave a total energy within 2 meV per unit cell of the Γ-point result. The convergence of the NMR results was also checked at higher cutoﬀ values (60 and 600 Ry, respectively, for the wave function and the density) and a (3 × 3 × 1) MP grid spacing.

3. RESULTS AND DISCUSSION 3.1. OH Content and H2O Production. The AIMD results were obtained to study annealing-induced eﬀects such as the reaction between OH groups to form Zr−O−Zr bridges and H2O. Thermal eﬀects in Zr(OH)4, such as changes in structural, electronic, and optical properties and the evolution of H2O, have been studied experimentally by Soliz et al.57 Most of the present work is focused on the experimentally relevant 300−500 K range, which is typically used in the pretreatment of Zr(OH)4 in adsorption studies, but some results were also obtained at temperatures as high as 1000 K. The ideal, well-ordered model shown in Figure 1 is unstable at room temperature and above and rapidly becomes disordered during AIMD. It is important to note that disordering occurs when the initial state is such that a regular array of H-bonds between neighboring OHs is not present in both in-plane lattice directions. The following section will discuss the eﬀect of such H-bonds on the evolution of structure during AIMD. Disordering is accompanied by relaxation, leading to the conversion of some bridging OH groups to terminal species, i.e., the breaking of one of the Zr−O bonds in a Zr−O(H)−Zr bridge. After several picoseconds at 400 K, few 8-fold-coordinated Zr atoms remain and on average the Zr atoms are 7-fold-coordinated as in monoclinic ZrO2. The Zr atoms also show signiﬁcant out-of-plane displacements, in some cases by more than ∼1.0 Å. Figure 2 shows structures illustrating the increase in the degree of disorder (e.g., out-ofplane displacement) with temperature, which will be discussed quantitatively in the following sections. Table 1 summarizes AIMD results for the evolution of diﬀerent O-containing species at 300−500 K. The relatively short simulation times used here are suﬃcient for the purpose of illustrating the increases in H2O production, (bridging-OH)/ (terminal-OH) concentration ratio, and Zr−O−Zr formation that occur with increasing temperature or with increasing time at a ﬁxed temperature. These are all consistent with experimental data discussed below. Some results are derived from “snapshots” (arbitrarily chosen individual steps in the trajectories), whereas others represent averages over the last 1.0 ps of the indicated total trajectory time. At a given temperature, both sets of results are similar for equivalent elapsed times. The analyses are based on bonding criteria related to interatomic distances. The maximum distance assumed for a bond is somewhat larger than the equilibrium distance at T = 0 because disorder introduces a distribution of distances and some bonds are elongated by vibrations. A Zr atom and an O atom (O and H atom) separated by 2.7 (1.2) Å or less are considered bonded, as are a Zr and a H2O with a Zr−OH2 distance of 3.0 Å or less. On the basis of EXAFS data58 for Zr(OH)4 gels, the maximum distance for Zr−OH2 bonding is taken to be about 10% larger than that presently chosen for covalent Zr−O interaction. The results in Table 1 were not signiﬁcantly aﬀected by small changes in these bonding criteria.

∫ dt e−iωt ⟨μ(0)μ(t )⟩

where kB is the Boltzmann constant and I(ω) is the oscillator strength. Calculation of I(ω) was done using the visual molecular dynamics (VMD) program52,53 with the IR spectral density calculator plug-in. Here, H rather than D was used for hydrogen. Eﬀective atomic charges were estimated using two diﬀerent methods, which gave nearly identical IR intensities. One approach made use of Mulliken population analysis performed using the CPMD code.54 The second involved a Bader “atoms-in-molecules” analysis of the QE wave function using the code55 provided by Henkelman et al. The charges used were −0.53, +0.23, and +1.2 |e| for O, H, and Zr, respectively. An advantage of this method is that anharmonicity is included. Hence, no semiempirical rescaling of the computed frequencies is needed, unlike in calculations based on the harmonic approximation, and the results can be compared directly with those of the experiment. A disadvantage is that the atomic displacements cannot easily be determined, which makes mode assignments diﬃcult. The starting model was a two dimensionally periodic slab (2DPS) constructed using a (4 × 4) unit cell of Zr(OH)4 molecules joined by bridging OH groups, as in the idealized structure (Figure 1). Each layer thus consisted of 16 Zr(OH)4 units per cell with all Zr atoms coplanar. In the starting structure, the unit cell was a perfect square; however, variablecell optimization (at T = 0 K) yielded a structure with slightly diﬀerent lattice constants (14.59 and 14.78 Å) but 90° corner angles. This unit cell was used in all subsequent calculations, and the 1.3% diﬀerence in lattice constants is consistent with that seen56 in the a and b lattice constants of monoclinic ZrO2. A similar variable-cell optimization for an amorphous two-layer 2DPS with mixed terminal and bridging OH groups (which results from the AIMD process described below) gave slightly smaller lattice constants, 14.34 and 14.63 Å, which is an a − b diﬀerence of 1.8%. A 2DPS typically consisted of two layers, although some results were also obtained for one- and three-layer cells. A large vacuum gap (∼30 Å) between successive bilayers in a doublelayer calculation was used to avoid spurious long-range interlayer interactions and to allow space for H2O that desorbs during AIMD (see below) and for the diﬀusion of adsorbates (e.g., SO2). Although a model with so few layers might be considered inadequate in many calculations, the system that is being simulated is in fact thought37 to consist typically of two or three layers without signiﬁcant covalent interlayer C

DOI: 10.1021/acs.jpcc.7b11107 J. Phys. Chem. C XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry C

Zr(OH)4 is given a mild thermal treatment (∼370−450 K in vacuo) to reduce the coverage of H2O and other species adsorbed from the atmosphere without signiﬁcantly aﬀecting the character of the OH groups, the material porosity, or the surface area. However, higher temperatures are employed in experiments where the goal is speciﬁcally to observe the eﬀects of such structural changes on adsorption. As the temperature is raised (with approximately the same time interval at each temperature), or with increasing time at a ﬁxed temperature, AIMD shows the production of an increasing amount of H2O via reaction between terminal and bridging OH groups, i.e., Zr−OH + Zr−O(H)−Zr → Zr−OH2 + Zr−O−Zr. (The Zr−O−Zr bridge is sometimes termed an “oxo” structure.) This reaction can be seen directly in the AIMD trajectory. Some H2O molecules remain adsorbed as Zr−OH2, whereas others desorb (Zr−OH2 → Zrcus + H2O), with the latter being increasingly likely as the temperature is raised. As noted elsewhere8,39 and below, in Section 3.4.3, terminal (bridging) OH groups are basic (acidic). Hence, the H2O-forming process may be viewed as simply an acid−base reaction. A typical trajectory at 400 K indicates a highly dynamical system with OH groups ﬂuctuating between terminal and bridging and reacting to form mobile H2O that adsorbs via a Zr−OH2 bond or else desorbs or moves between Zr sites. Transfer of H between terminal OH groups, mediated by H2O formation and dissociation, can also be observed in a process such as

Figure 2. Structures resulting from AIMD runs for a single-layer 2DPS: (a) 1.9 ps at 300 K, (b) 1.9 ps at 600 K, and (c) 1.4 ps at 1000 K. Desorbed H2O molecules that are far from the slab have been excluded to save space. The gray, red, and yellow spheres are Zr, O, and H, respectively.

Zr−OH 2 + HO−Zr + HO−Zr → Zr−OH + H 2O−Zr + HO−Zr

As noted in the Introduction, the experimentally determined38 (bridging-OH)/(terminal-OH) concentration ratio is in the range of 3.3−3.7 for samples outgassed in vacuo at ≤450 K. The computed ratio for a disordered two-layer model at 400 K is consistent with this result, although the variation with time and temperature is not completely monotonic. This reﬂects the combined eﬀects of Zr−O bond breaking in Zr−O(H)−Zr, which eliminates a bridging OH and forms a terminal OH, and H2O production, which eliminates one OH of each type (see below). It is noted that the calculation is for a model that is laterally inﬁnite without edges or defects. Introduction of edge eﬀects, due to particles of ﬁnite size, or vacancies (by removing single Zr(OH)4 units) would lead to more terminal and fewer bridging OHs and reduce the ratio to well within the observed range or even lower. The AIMD results at 400 K are generally the most useful because, in typical adsorption experiments,

→ Zr−OH + HO−Zr + H 2O−Zr

A similar OH dynamical behavior has also been seen59 in AIMD studies of Zr MOFs; however, further analysis of this aspect is beyond the scope of the present work. Unlike in the real material, some H2O in the AIMD runs remains adsorbed even at a higher temperature because the model is in eﬀect a closed system with no means for H2O to escape into vacuum. Because the interest here is primarily in Zr(OH)4 properties near room temperature, this is not a signiﬁcant issue. Thermogravimetric analysis3,11,57,60 shows that the loss of H2O essentially ceases above ∼700 K, at which point all terminal OH groups have been consumed. This is illustrated3 by the loss of reactivity toward SO2, which requires terminal OH. Following this treatment, XRD shows the onset of diﬀraction features34−36,57,61−64 associated with both tetragonal

Table 1. Evolution of OH Speciesa with Time and Temperature T (K) elapsed time (ps)b bridging OHc terminal OHc bridging/terminal ratio total H2Oc H2O adsorbed as Zr−OH2d Zr−O−Zre

300 2.54 100 24 4.17 2 2 2

(2.62) (98.8) (24.2) (4.08) (2.5) (2.1) (2.5)

400 2.47 93 25 3.72 4 4 6

500 6.92 90 22 4.09 8 6 8

(7.16) (91.4) (21.9) (4.19) (7.3) (5.3) (7.3)

2.43 95 17 5.59 8 6 8

3.22 93 17 5.47 9 7 9

(3.33) (94.8) (18.1) (5.29) (7.4) (5.3) (7.7)

a

Results for a disordered two-layer model, beginning at each temperature with a pristine sample. Numbers without parentheses are based on snapshots after the indicated elapsed times, with two diﬀerent elapsed times being given for 400 and 500 K. Numbers in parentheses are averages over the last 1.0 ps of the stated total trajectory time. bTotal elapsed time of the AIMD simulation at each temperature. cNumbers for each species (bridging OH, etc.) refer to the number per (4 × 4) unit cell including both layers. dWater not adsorbed as Zr−OH2 is considered to be desorbed. e Zr−O−Zr bridges with no O−H. Bridging OH + terminal OH + total H2O + Zr−O−Zr = total number of O atoms in model = 128. D

DOI: 10.1021/acs.jpcc.7b11107 J. Phys. Chem. C XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry C and monoclinic ZrO2. The complete loss of terminal OH and the concomitant onset of crystallization were not observed in the present AIMD simulations. Crystallization proceeds via a nucleation event, which may not occur in MD on a time scale of a few picoseconds. Furthermore, conversion of Zr(OH)4 to ZrO2 in a multilayer model would require the removal of interstitial H2O formed between the layers. For example, after 6.92 ps at 400 K (Table 1), three of the eight H2O molecules per (4 × 4) unit cell lie between the two layers. Removal of interstitial H2O is diﬃcult in a model based on inﬁnite planes because the H2O would need to go through one or more layers. It is also noted that Weakliem and Carter,65 in their empirical force-ﬁeld MD studies of reconstruction on Si and Ge(100) surfaces, observed somewhat anomalous behavior for very thin slabs, suggesting that reconstruction might entail the correlated motion of atoms several layers apart. Similar considerations may also apply to crystallization in the present case. The inability to model the thermal conversion of Zr(OH)4 to ZrO2 is not a cause for concern because the main focus here is on Zr(OH)4 itself at temperatures below 500 K. Thus, no attempt was made in the present study to remove interstitial H2O forcibly to promote crystallization. An exothermic interaction energy between layers of −7.73 eV per (4 × 4) cell was found for the two-layer model, which corresponds to an average of −0.48 eV per interlayer pair of Zr(OH)4 units. This result was obtained using a snapshot of the two-layer model after 6.9 ps at 400 K. The structure was relaxed at 0 K after which the two layers were separated and each relaxed individually. Molecular H2O adjacent to either layer that had formed during MD was included with that layer when separated. Thus, both separated layers had the same net stoichiometry, [Zr(OH)4]16. The layers were not structurally identical and diﬀered slightly in energy when relaxed. The interaction energy was then obtained from the three total energy values. No covalent cross-linking (e.g., reaction of the form Zr−OH + HO−Zr → Zr−O−Zr + H2O) was seen to occur between layers in the AIMD simulations for the temperatures and time scales used here, and no such reaction was observed between terminal OH groups even within a layer. Apparently, two basic terminal OH groups do not react as readily as do a basic terminal and an acidic bridging OH group. The interlayer bonding is instead attributed to a combination of H-bonding and dispersion interactions; however, no eﬀort has been made to assess the relative contributions of these two eﬀects. Experimentally,11 Zr(OH)4 dried at 333 K after synthesis is seen in scanning electron microscopy to consist of aggregates described as “secondary particles”. Adhesion in this case may arise from similar H-bonding and dispersion contributions. An exception to the lack of cross-linking noted above occurred in a case where an AIMD trajectory was run for a long time (>15 ps) at 650 K, during which several Zrcus sites were created as a result of the H2O-forming reaction. A reaction was then seen between a terminal Zr−OH in one layer and a Zrcus in the other to form a Zr−O(H)−Zr bridge, which subsequently transferred the H to a terminal Zr−OH to form H2O. In a further study of possible cross-linking, a calculation was done for the adsorption of a single Zr(OH)4 unit on a single-layer Zr(OH)4 slab. Figure 3 shows the resulting AIMD structure after 3.7 ps at 400 K. A terminal Zr−OH on the slab formed a Zr−O(H)−Zr bridge to the adsorbed Zr(OH)4. A H moved from a bridging OH in the slab to form Zr−OH2 at another slab terminal-OH site, and the resulting bare bridging

Figure 3. Structure formed by a single Zr(OH)4 unit adsorbing on a single-layer Zr(OH)4 slab during AIMD at 400 K. The green arrows show the adsorbed Zr, the H2O that forms as a result of the reaction, and the 3-fold-coordinated O. For clarity, the edges of the slab have been trimmed, and the view is from slightly below the slab.

O formed a bond to the adsorbed Zr(OH)4 to give a 3-foldcoordinated O, which is characteristic of monoclinic ZrO2. This mode of bonding, which may depend on the local structure of the adsorption site in the amorphous model, appears to be stable at 400 K. However, later in the trajectory a terminal OH on the adsorbed Zr was lost through a H2O-forming reaction with a bridging OH in the slab. Although this aspect has not been further investigated here, it is possible that a second slab might react similarly with the adsorbed Zr(OH)4, which would then function as a link between the two, thus forming a quasithree-dimensional structure. Either of these cross-linking mechanisms would be consistent with the observed57 increase in particle size with annealing above 525 K. 3.2. Ordered Structures. To our knowledge, Zr(OH)4 is found to be amorphous in all experimental work reported thus far. Our results agree with this because the original idealized crystalline structure quickly becomes disordered at room temperature. Nevertheless, it is useful to consider whether a diﬀerent starting morphology can lead to a crystalline, or at least a better-ordered, form that remains stable during AIMD. We ﬁnd that such a phase can be obtained if, in the initial structure, the OH groups (all of which are bridging) are appropriately oriented. Speciﬁcally, if the conﬁguration is such that rows of H-bonds to neighboring OHs form in both lattice directions, the resulting structure (Figure 4) is ∼0.18 eV per Zr atom lower in energy than the disordered material. This structure remains ordered during AIMD runs longer than 5 ps at 400 and 600 K but begins to disintegrate above 1000 K, as does the disordered material. The degree of ordering and its relationship to the distribution of H-bond energies will be discussed in the following section. It is noted that “ordered” in the present context refers only to the regular array of H-bonds and to the degree of planarity shown in Figure 4. There remains a ﬁnite variation in interatomic distances, including the Zr−O E

DOI: 10.1021/acs.jpcc.7b11107 J. Phys. Chem. C XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry C

Figure 5. Radial distribution functions calculated for various two-layer structures showing only Zr−O distances. The results represent averages over the respective AIMD trajectories. (a) The 300 K ordered structure as in Figure 4 and (b) 400 K disordered and (c) 1000 K disordered structures resulting from AIMD as in Figure 2.

Figure 4. Top-down (a) and side (b) views of a single-layer 2DPS based on an ordered Zr(OH)4 structure. In (a), a few H-bonds are indicated by heavy green lines. Not all such bonds are illustrated.

bond length, as will be seen in the radial distribution function (RDF) discussed below. The source of this stability appears to be the pairing of bridging OHs to form strong H-bonds that persist with nearly no bond breaking or rearrangement during AIMD. This behavior stands in clear contrast to the dynamic bond breaking and reforming in the disordered phase described previously. Despite the higher stability of the ordered state, there has so far been no direct experimental evidence of ordered Zr(OH)4, which may be due to a kinetic barrier against ordering during synthesis. As noted above, annealing at ≥700 K, which in principle might promote ordering, leads instead to conversion to ZrO2. It is also possible that small regions of ordered Zr(OH)4 are present within the amorphous Zr(OH)4 matrix but have not yet been detected. The most direct experimental probe of such local ordering would be the RDF measured in EXAFS. The calculated RDF of the ordered material, discussed below, shows clear peaks in the second neighbor shell at Zr−O and Zr−Zr distances of ≥3.5 Å, whereas the disordered material shows relatively little regular structure at these distances. It also appears that interaction between layers can act to impede disordering. An initially ordered single layer that, by itself, disordered in AIMD at 300 K was used to construct a two-layer model. On the basis of visual inspection of the Zr outof-plane displacements, the system exhibited less disorder during AIMD at 300 K than did the single layer. However, increasing the temperature to 600 K, or increasing the interlayer separation before AIMD, led to disordering, as shown in Figure 2 for a single layer. To quantify better the degree of order, RDFs and their integrals were calculated by averaging over AIMD trajectories using the VMD code.52,53 A partial RDF showing only Zr−O distances is given in Figure 5. The main issue addressed here is the degree of order rather than the assignment of individual peaks to speciﬁc atom pairs, and this aspect is illustrated more clearly by focusing on Zr−O distances. Figure 5a shows the RDF following AIMD at 300 K for a two-layer model that was

stabilized in an ordered state by strong H-bonding. This is representative of ordered Zr(OH)4, which, as also noted above, has not been found experimentally. Major peaks are seen at ∼2.2, 3.8, 4.7, and 5.7 Å. Weak secondary peaks at about 4.2 and 5.2 Å result from H2O that was included in the model used in the RDF calculations. The ﬁrst peak corresponds to Zr−O nearest-neighbors and the second to Zr−O next-nearestneighbors (∼3.8 Å) for which the distance is close to that of Zr−Zr nearest-neighbors (∼3.6 Å). For amorphous Zr(OH)4 gels,37,58 EXAFS data yield values of 2.1−2.2 and ∼3.5 Å, respectively, for the Zr−O and Zr−Zr nearest-neighbor distances. The main RDF peak at 2.1−2.2 Å, which is the Zr−O bond length in Zr(OH)4, is broadened and shifted slightly lower by disorder. This could be viewed as a relaxation of the nearestneighbor shell accompanying the reduction in the average Zr coordination number from 8- to 7-fold. There are also clear diﬀerences in the peaks at greater distances, which are much broader for the disordered systems, and at 1000 K, all such features are replaced by a structureless band. The integrated RDFs reveal that in the ordered phase Zr is 8-fold-coordinated, as in the ideal lattice (Figure 1), and becomes 7-foldcoordinated on average in the disordered state as some Zr− O(H)−Zr bridges break to form Zr−OH and Zrcus. The lack of terminal OH in the ordered material does not, to our knowledge, agree with any experimental studies of Zr(OH)4 so far, which again suggests that the ordered material is not formed under practical conditions. Turrillas et al.35,36 have reported EXAFS data for Zr(OH)4 that was dried at 403 K after synthesis. The Fourier transform modulus of the Zr K edge EXAFS data shows a Zr coordination number of 7 and Zr−O and Zr−Zr nearest-neighbor distances of about 2.1 and 3.5 Å, respectively, with little or no order beyond this. The results in Figure 5 are in qualitative agreement with these data except that the calculated RDF for the 400 K disordered model shows evidence of structure at greater F

DOI: 10.1021/acs.jpcc.7b11107 J. Phys. Chem. C XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry C

Figure 6. H-bond histogram contour plots of the double-layer disordered (left) vs ordered (right) structures of Zr(OH)4. Theta is the donor− hydrogen−acceptor (D−H···A) angle in degrees, and r is the hydrogen−acceptor (H···A) distance in angstroms. Time averages (over ∼4 ps AIMD trajectories) of the H-bond counts (histogram populations of a given θ−r bin) are plotted as the colored z-contour values. In the top panel, hydrogens in H-bonds were statically deﬁned, whereas in the bottom panel, hydrogens in H-bonds were dynamically deﬁned, as described in Section 3.3.

distances. Turrillas et al.36 note that the structure of Zr(OH)4 depends strongly on the pH of the solution from which the gel was precipitated. A higher pH, with more available OH−, results in more-defective polymeric sheets having more terminal OHs than is the case for a low-pH solution and also yields Zr(OH)4 sheets with more strain. Thus, a detailed comparison of observed and calculated RDFs will require a knowledge of the dependence of the EXAFS data on the pH used in sample growth. 3.3. Hydrogen Bonding. Hydrogen bonding in the twolayer model was analyzed quantitatively using H-bond histogram contour plots displaying the number of H-bonds occurring within a range of angles and distances. The angle (θ) is deﬁned as the O−H···O (D−H···A) bond angle and the distance (r) as the H···O (H···A) bond length, where “D”, “H”, and “A” refer to “donor oxygen”, “hydrogen”, and “acceptor oxygen”, respectively. The goal of this analysis was to understand the relationship between the H-bond strength and disordering in Zr(OH)4. The H-bond counts from the AIMD trajectories were assigned to 2D histogram bins according to their values of θ and r. General criteria66 for a strong H-bond are 150° < θ < 180° and r < 2.0 Å. The H···A distance was chosen for analysis rather than the D···A distance because the latter depends on both the H···A distance and the D−H···A angle. The D−H distance was assumed to be approximately constant at about 0.95 Å. The histogram bins were then averaged over the AIMD trajectory, resulting in average Hbond counts for a given θ−r combination. The eﬀects of H hopping were examined, in which H moves from D to A as the covalent D−H bond becomes a hydrogen bond, whereas the H···A bond becomes covalent. To this end,

H atoms were selected according to two diﬀerent deﬁnitions, namely, static and dynamic. In the static case, only H atoms initially bonded to donor oxygens at t = 0 were considered for a given D−H···A combination, whereas in the dynamic deﬁnition, any H near a donor O at a given time step was considered for a given D−H···A combination. The static deﬁnition essentially freezes the labeling of which H belongs to which D or, alternatively, which O is the donor and which is the acceptor in a D−H···A triplet. If a H hops from D to A during the course of the simulation, the new covalent bonding partner would still be called the acceptor under the static deﬁnition, because of the initial conditions, and the new covalent bond would still be labeled H−A (the former H-bond). In contrast, the dynamic deﬁnition allows for changes in the labels over time so that the new covalent bonding partner would now be labeled the donor, even though initially it was the acceptor. The end result is that under the static deﬁnition a H hopping event appears as a H-bond at very short H−A distances (r ∼ 1 Å) because that H-bond is really a covalent bond. Under the dynamic deﬁnition, the H hopping event is not readily seen in the H-bond histogram plots because H-bonding is more rigorously analyzed to treat such eﬀects. Figure 6 shows plots for the double-layer Zr(OH)4 model. In all plots, hotter colors indicate more H-bonds, and the strong H-bonding region of interest is near the top left, at a short distance (r < ∼2 Å) and a large angle (θ > 150°). The region near the bottom right of each plot is one of smaller angles and greater distances, representing weak or nonexistent H-bonds. Comparing the left and right panels of each ﬁgure, large diﬀerences are seen in the H-bonding of the disordered versus ordered systems. As expected, the ordered system exhibits G

DOI: 10.1021/acs.jpcc.7b11107 J. Phys. Chem. C XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry C

relative intensities at the red end. A very weak structure in the 2900−3000 cm−1 range arises from trace amounts of hydrocarbon impurities. Figure 7c shows the spectrum computed at 300 K for a twolayer 2DPS model using the dipole correlation function (Section 2) for an 18 ps trajectory with a 1 fs time step. The model is nominally dry with no H2O other than what forms during MD via the OH condensation reaction described above. A Gaussian broadening with a 100 cm−1 full width at halfmaximum (FWHM) was applied to approximate the experimental line widths, which result largely from disorder. Plots with a smaller FWHM showed the same general features but with increased “noise”. In the calculation, no modes are seen in the 1300−1700 cm−1 region except near 1600 cm−1, which corresponds to the δ(HOH) bending mode of H2O. Strong features in the 1300−1700 cm−1 range in the IR data are therefore not intrinsic to Zr(OH)4 and are instead due to carbonate and/or bicarbonate species 68,69 arising from adsorption of atmospheric CO2. Because of the high aﬃnity of Zr(OH)4 for H2O and CO2, it is diﬃcult to obtain IR and Raman data that are completely “clean”. In the dipole correlation method, it is not possible to formulate mode assignments by simple inspection of atomic displacements. Therefore, an IR spectrum was calculated for a [Zr(OH)4]4 tetramer (Figure 7 inset) that approximates the basic bonding in amorphous Zr(OH)4. The calculation was performed using the Gaussian program suite70 with the B3LYP hybrid functional, 6-311++G(3df,3pd) basis sets for H and O, and the DGDZVP basis set71 for Zr. A Grimme D3 semiempirical dispersion contribution50 was also included. The normal modes were obtained in the harmonic approximation, and no rescaling of the frequencies was performed. The computed spectrum (Figure 7d) is suﬃciently similar to that of the 2DPS model to permit its use as an aid in mode assignment. Proposed correlations between various peaks in the experimental, 2DPS, and tetramer spectra are shown by the dashed lines and summarized in Table 2.

narrower H-bond distributions and higher counts in the strong H-bonding region, consistent with stabilization by strong Hbonding. In the disordered system, H hopping can be readily seen as a small “island” to the left of the strong H-bonding region, at r ≈ 1 Å, in the top panel (static deﬁnition). This H hopping island disappears, as expected, in the bottom panel (dynamic deﬁnition). These results show that the disordered system exhibits a much broader distribution of H-bonds, some of which are involved in active hopping between O atoms. As noted in Section 3.1, this hopping is observed directly in the AIMD trajectories for disordered Zr(OH)4. 3.4. Comparisons between Theory and Characterization Experiments. In this section, comparisons are presented between various types of chemical, structural, and spectroscopic data and predictions based on the model developed in the preceding sections. 3.4.1. Infrared Spectra. Figure 7a shows a diﬀuse-reﬂectance (DR) spectrum (McEntee, M., Edgewood Chemical and

Table 2. Summary of Mode Assignments in Computed IR Spectraa

Figure 7. (a) Experimental DR spectrum for Zr(OH)4 powder. (b) Experimental ATR spectrum for Zr(OH)4 powder. (c) Computed IR spectrum for a two-layer model at 300 K. A Gaussian broadening of full width at half-maximum (FWHM) = 100 cm−1 has been applied to approximate the experimental broadening, which is due largely to disorder. The quantity plotted is the oscillator strength. (d) Computed spectrum for a [Zr(OH)4]4 tetramer (low-energy end only) with a 20 cm−1 Gaussian broadening. The spectra have been displaced vertically for clarity. The dashed lines show the proposed correlations between tetramer and 2DPS computed features and between peaks in the 2DPS and experimental data. The inset shows the structure of the tetramer with the two H-bonded bridging-OH groups marked by arrows. The relative intensities of diﬀerent spectra are not quantitative.

assignmentb

tetramer

2DPS

ν(Zr−O) in Zr−O(H)−Zr ν(Zr−O) in Zr−OH ω(OH) in Zr−O(H)−Zrd ω(OH) in Zr−O(H)−Zrd ρ(OH) in Zr−O(H)−Zrf ω(OH) in Zr−O(H)−Zrf

495 670 730 787 833 1000

590 830c 950e 1100

a

Spectra are shown in Figure 7. The assignments refer to tetramer modes, and all frequencies are in cm−1. bν = stretching, ω = wagging, ρ = rocking. cThis band involves overlapping contributions from other modes in addition to ν(Zr−O). dThe H is “free”, not hydrogenbonded. eAll three modes (730, 787, and 833 cm−1) contribute to this peak. fThe H is hydrogen-bonded to O in another OH bridge.

Biological Center. Personal communication) for Zr(OH)4 powder after purging in dry N2 at 300 K. Figure 7b shows data (Balow, R. B., Naval Research Laboratory. Personal communication) obtained under similar conditions for a diﬀerent sample using multiple-reﬂection attenuated total reﬂection (ATR) spectroscopy for powder deposited on a ZnSe optical element. The two spectra diﬀer mainly in the H2O content and in the relative intensity of the low-energy end. In ATR spectroscopy, the eﬀective sampling depth increases with the IR wavelength,67 which then leads to an enhancement of

Starting at the low-energy end, the tetramer shows strong modes at ∼495 and 670 cm−1 that correspond to Zr−O stretching in, respectively, Zr−O(H)−Zr and Zr−OH. Raman data for Zr(OH)4 gels37,72 show peaks at about 400 and 540 cm−1. The assignment of the ∼400 cm−1 Raman mode is uncertain, but the peak at ∼540 cm−1 is attributed37,72 to Zr−O stretching in Zr−O(H)−Zr bridges in reasonable agreement with the tetramer result (495 cm−1). The corresponding 2DPS H

DOI: 10.1021/acs.jpcc.7b11107 J. Phys. Chem. C XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry C peak is at 590 cm−1. The 670 cm−1 tetramer peak is proposed to correlate with the strong 2DPS peak at 830 cm−1, which is therefore assigned to ν(Zr−O) in terminal Zr−OH, although other modes may contribute to this peak (see below). One discrepancy is that the 670 cm−1 mode is the strongest feature in this region of the computed tetramer Raman spectrum (Supporting Information), but the experimental Raman data for Zr(OH)4 gels do not reveal a corresponding peak. It is possible that extensive polymerization in the gels reduces the concentration of terminal Zr−OH relative to that in the tetramer and 2DPS models. A group of three weaker tetramer peaks is seen at 730, 787, and 833 cm−1 that together correlate with a shoulder in both the experimental and the 2DPS spectra at ∼950 cm−1. The lower two are associated with wagging modes of bridging OH groups for which the H is free (i.e., not H-bonded), and the 833 cm−1 peak is due to a rocking mode of bridging OH where the H is hydrogen-bonded to the O on the opposing bridge. The two H-bonded bridging OHs are indicated by arrows in the inset. In a wagging (rocking) mode, the H atom moves approximately parallel (perpendicular) to the Zr−O−Zr plane. The tetramer peak at 1000 cm−1, which corresponds to a weak shoulder in the 2DPS spectrum at ∼1100 cm−1, arises from wagging in a bridging OH where the H is hydrogen-bonded. The structure near 1000 cm−1 in the IR spectra of various Zr compounds has sometimes been assigned73 to the stretching mode of the zirconyl (ZrO) group, but no such species exists in the present models. Other work,74 on aqueous solutions of [Zr4(OH)8(H2O)16]8+, assigns a peak in the 1018−1024 cm−1 range to a mode involving H2O, but there is no H2O in the present tetramer model. In practice, more than one species can potentially contribute a mode in the 1000−1100 cm−1 range, depending on the chemical structure of the material. In a large and amorphous structure, the diﬀerent modes will be spread across overlapping energy ranges. However, the general conclusion is that the bending modes of bridging-OH groups contribute structure in the ∼800−1100 cm−1 range, whereas Zr−O stretching modes occur in the ∼600−830 cm−1 range. Bending modes of the terminal OHs in the tetramer model fall mainly below ∼400 cm−1 and may be attributable to the peak at about 300 cm−1 in the 2DPS calculation. Experimental data for Zr(OH)4 that is free of H2O and bicarbonate, and that clearly show ν(O−H) for the diﬀerent OH species, are diﬃcult to obtain. Thus, we turn to data75 for monoclinic ZrO2 surfaces, which show ν(OH) modes at 3774, 3738, and 3675 cm−1. Theoretical results identify these with free (not H-bonded) OH with O bonded to one, two, or three Zr atoms, respectively, although the predicted ranges for the latter two extend over ∼150−200 cm−1. In the present models, there are no 3-fold-coordinated OH groups. H2O adsorbed on OH-terminated ZrO2 yields an additional broad and intense band75 peaking at ∼3200 cm−1 and requires heating in vacuo for complete removal. It was noted that the experimental peak assigned to 3-fold-coordinated OH might receive a contribution from 2-fold-coordinated OH in which the H is hydrogenbonded to O in another OH. It was also found75 that Hbonding between OH groups shifts ν(O−H) 100 to 500 cm−1 to lower energy depending on the strength of the interaction. In the 2DPS calculation (Figure 7c), the highest-energy peak, at 3788 cm−1, is assigned to overlapping free terminal- and bridging-OH stretches. Such free OHs are reduced in number in the experimental data by the presence of adsorbed H2O. The next-highest 2DPS peak, at 3640 cm−1, is then assigned to

weakly H-bonded OH groups. The weak 2DPS peaks at 3160 and 3370 cm−1 are ascribed to adsorbed H2O molecules and to strongly H-bonded OH groups (Figure 6), both of which are few in number in the present disordered model. An exact correlation between the 2DPS and experimental Zr(OH)4 spectra is diﬃcult because of the high concentration of H2O in the latter and also the unavoidable presence of bicarbonate, for which ν(O−H) is expected76 at 3600−3630 cm−1 (if not Hbonded). In the ν(O−H) region, the ATR data show two broad bands at ∼3100 and 3400 cm−1 and also weak and relatively narrow shoulders at 3700 and 3780 cm−1 that indicate a small density of free OHs. The low energy of the lowest band (3100 cm−1 in ATR, 3160 cm−1 in the 2DPS, and 3200 cm−1 for OHterminated ZrO2; ref 75) suggests77 that some H2O is strongly H-bonded. The stronger band, at 3370 cm−1 in the 2DPS (3400 cm−1 in ATR), may correspond to Zr−OH2, which is known to exist in this model. Vibrational spectra were also computed for the ordered twolayer structure described in Sections 3.2 and 3.3. In comparison to that in the disordered model, there was less intensity in the ν(O−H) band assigned to free (not H-bonded) OH, and the ν(O−H) modes associated with H-bonded OH were more redshifted. Both eﬀects are consistent with the stronger and more extensive H-bonding in the ordered structure that is illustrated in Figures 4 and 6. 3.4.2. Proton Nuclear Magnetic Resonance. Experimental 1 H NMR data8,38,39 for Zr(OH)4 exhibit diﬀerent chemical shifts (δ) for H2O and for terminal and bridging OHs. Figure 8 shows a comparison between the data and a spectrum calculated for a snapshot of a three-layer 2DPS after 4.51 ps at 400 K followed by relaxation at 0 K. In NMR, δ is determined by the local electronic environment, which is somewhat diﬀerent for each H atom in an amorphous material such as Zr(OH)4. Thus, the calculated spectrum is quite scattered, with each of the 192 H atoms in the unit cell having a slightly diﬀerent δ. A three-layer model was used to maximize the sample size and thus to improve the statistics, but no signiﬁcant diﬀerence was seen between these results and those for a two-layer model. An internal calibration standard is needed to align the observed and calculated spectra. This was obtained by computing the spectrum for a model with molecular H2O added. The contribution from these additional free H2O molecules, most of which do not interact strongly with Zr(OH)4, falls within a ∼6 parts per million (ppm)-wide range centered at the high-δ end of the spectrum, which is higher than the range computed for adsorbed H2O. The calibration then consists of shifting the calculated δ scale to align this band with that seen for H2O in the experiment. The H2O in the experimental spectrum, which arises from atmospheric adsorption, probably exists in the form of an adsorbed liquidlike multilayer. In performing the alignment, which equates this H2O to the free H2O in the calculation, one neglects the eﬀect78 on δ of H-bonding between H2O molecules in the experiment. The absolute δ values are thus only approximate, but the relative values, which are the main quantities of interest here, are more reliable. With some degree of qualiﬁcation, the calculation agrees well with experiment. The overall width in the calculation (∼14 ppm) is close to that in the data (∼10 ppm), and the bridging and terminal contributions are completely overlapping in both sets of spectra. This suggests that the model is correctly representing the level of disorder in the real material. The I

DOI: 10.1021/acs.jpcc.7b11107 J. Phys. Chem. C XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry C

range and are shifted to smaller δ and into the range of the surface OH groups. The computed average δ for bridging OH is smaller than that for terminal OH by 1.2 ppm, whereas in experiment, it is larger by 1.0 ppm. The experimental result is as expected,8 given the higher Brønsted acidity of the bridging OH, for which the shielding of the H should be smaller than for the terminal OH. However, the source of this discrepancy has not been investigated. More seriously, it might indicate that the 2DPS model incorrectly represents the relative acidity of the two OH types. However, results for reaction with SO2 given below show that this is not the case. These constitute a direct test of acidity because experiment3 shows the SO2 reaction to involve the basic terminal OH virtually to the exclusion of the acidic bridging OH. 3.4.3. Acidity of Bridging vs Terminal OH. Additional calculations were done to verify that a bridging OH is more acidic than a terminal OH, as indicated by experiment. Here, the energy for the dissociation OH → O− + H+ was obtained for the cation and anion at inﬁnite separation, where the OH and the O− are either bridging or terminal. To avoid dealing with a charged unit cell in a 2DPS, a free-standing [Zr(OH)4]12 cluster model (Supporting Information) was analyzed using the Gaussian program suite70 as described in Section 3.4.1 except that 6-311G(d,p) basis sets were used for H and O. Both the neutral and the anionic cluster were fully relaxed, and arbitrarily chosen OH groups were evaluated. The relaxed cluster was completely disordered and resembled a nanoparticle. The result is that removing H+ from a bridging OH requires 0.49 eV less energy than from a terminal OH in a case where neither the OH nor the O− is stabilized by H-bonding to another OH, which is consistent with the higher Brønsted acidity of the bridging OH. The relaxed neutral cluster also exhibited one OH group that was apparently 3-foldcoordinated, although one Zr−O distance (2.46 Å) was larger than the others (2.20 and 2.22 Å). Removing H+ in this case requires 0.63 eV less energy than from a terminal OH, which indicates a higher acidity for a nearly 3-fold- versus a 2-foldcoordinated OH. These results are consistent with those79 for OH acidity versus coordination number in the well-studied case of dry γ-Al2O3. 3.4.4. Adsorption and Reaction of SO2. Calculations addressing the adsorption and reaction of SO2 are useful as a test of the model, especially in view of the intended applications discussed in the Introduction, because there are experimental data2−4 that help identify the reaction product and also provide insight into the possible reaction mechanism. The reaction requires3 terminal OH and produces a sulﬁte (SO32−). Little or no adsorption of SO2 is seen on dry Zr(OH)4 that has been calcined to eliminate the terminal OH. If the Zr(OH)4 is hydrated (i.e., exposed to H2O-saturated air for an extended period) prior to admission of SO2, then XPS data4 indicate the formation of sulfate (SO42−) as well as sulﬁte. However, this aspect will not be treated in the present work, which focuses on nominally dry Zr(OH)4. Temperatureprogrammed desorption (TPD) data2 monitoring the loss of SO2 show a major peak at ∼583 K with additional features at about 653 and 698 K that may be aﬀected by the conversion of Zr(OH)4 to ZrO2 occurring at these higher temperatures (Section 3.1). Essentially all adsorbed and/or reacted SO2 is removed in TPD, leaving no S residue that is detectable in XPS. Loss of SO2 presumably involves (using a bridging sulﬁte as an example) a process such as Zr−O−S(O)−O−Zr → Zr−O−Zr

Figure 8. (a) Experimental (ref 38) and (b) computed proton NMR spectra. The experimental data were obtained for a sample that had been heated to 448 K in vacuo to reduce the coverage of adsorbed H2O. The contributions from molecular H2O, bridging OH, and terminal OH are shown in red, blue, and green, respectively. Each individual resonance in the computed spectrum has been assigned an arbitrarily chosen Gaussian line shape with a FWHM of 0.2 parts per million (ppm). Note the diﬀerent δ scales in the two plots. The horizontal arrows indicate the nominal widths of the two sets of spectra. The (bridging-OH)/(terminal-OH) concentration ratio in the computed result is 4.4.

NMR δ is seen in the calculation, as well as in experiment,78 to be very sensitive to interaction between a given proton and the rest of the system. In the model, all of the H2O molecules are formed as a result of reactions between OH groups that occur during AIMD. After subsequent relaxation at T = 0 K, all H2O molecules are undergoing H-bonding, based on interatomic distances, and several are also bonded to Zrcus via Zr−OH2 dative bonds. None are free in the sense of being desorbed, and as a result, the corresponding resonances are spread over a wide J

DOI: 10.1021/acs.jpcc.7b11107 J. Phys. Chem. C XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry C + SO2, although the exact mechanism is presently undetermined. The data available at present do not permit a conclusive determination of the relative importance of molecular adsorption of SO2 on dry Zr(OH)4 at room temperature. Possible generic sulﬁte species include bidentate (Zr(−O−)2SO), bridging (Zr−O−S(O)−O−Zr), and monodentate (Zr−O−SO2). Theoretical results80 for SO2 on γ-Al2O3 reveal several diﬀerent sulﬁte structures with varying degrees of stability, and IR data81,82 for SO2 on dry but hydroxylated (i.e., OH-terminated) anatase TiO2 nanoparticles indicate that bridging and monodentate are the dominant sulﬁte species. It is noted here that some authors use the terms “bidentate” to describe the structure that here is referred to as “bridging” and “chelated” in relation to that described here as bidentate. In the ﬁrst phase of the work, a single-layer model was used to reduce the computational cost for the climbing-image nudged elastic band (CINEB) calculation83,84 of the activation energy (described below). To enhance reactivity, a vacancy was formed by removing one Zr(OH)4 unit from the 16 unit (4 × 4) cell. This creates an easily accessible region with a high density of terminal OH, and the local structure of this site might also be seen as a reasonable approximation of the edge of a Zr(OH)4 nanoparticle. Before introduction of the defect, the Zr(OH)4 layer was ordered, as described in Section 3.2. The defect then deﬁnes a region within which the terminal OH groups are localized, thus permitting a clear identiﬁcation of the role of such species in the reaction. Creation of the vacancy is exothermic, with an estimated formation energy of ΔEf = −0.50 eV for a single layer with a (4 × 4) cell. However, desorption of an intact Zr(OH)4 molecule was never observed in any of the AIMD simulations. This indicates that intact desorption is less favorable (kinetically and/or energetically) than partial decomposition that releases H2O and/or converts some bridging OH to terminal. The value of ΔEf was obtained using an ordered single layer (Figure 4) that was relaxed at T = 0 K. The exothermic ΔEf presumably results from the relaxation that occurs in the vicinity of the vacancy as well as from the relaxation of the free Zr(OH)4 to a tetrahedral conﬁguration. Because of the attractive interaction between layers discussed in Section 3.1, one expects that creating a defect of this type in a relaxed, disordered bilayer would be signiﬁcantly less exothermic than in an ordered monolayer. The AIMD run began at 300 K, and after 1.2 ps, the SO2 had not approached suﬃciently close to the vacancy to react. However, upon lowering the temperature to 200 K, the SO2 drew closer, and a reaction occurred after about 0.5 ps. Figure 9 shows the structure of the reaction product, a bidentate sulﬁte, and adsorbed H2O. The AIMD trajectory reveals a H-atom transfer mechanism of the form Zr−OH + SO2 + HO−Zr → Zr(−O−)2SO + H2O−Zr. A similar process, in which SO2 reacts with two terminal OH groups to form a sulﬁte and H2O, has been proposed82 in an experimental study of SO2 on hydroxylated anatase TiO2 nanopowders. The adsorption energy is ΔEads = E(V + SO2) − [E(V) + E(SO2)] = −1.29 eV, where E(V) and E(SO2) are the relaxed total energies (at T = 0 K) of the unit cell with a vacancy and of free SO2, respectively, and E(V + SO2) is the relaxed total energy after reaction. The presence of a short (1.70 Å) H-bond between a bridging OH and the free O atom of the sulﬁte, which contributes to the total ΔEads, is also noted. For the temperatures and time scales used to study the SO2 reaction, no H2O was produced via reaction between OH groups. The stability of the reaction product was evaluated using AIMD,

Figure 9. Relaxed product resulting from the reaction of SO2 with terminal OH groups at a vacancy site. Zr, S, O, and H are shown in gray, light yellow, red, and dark yellow, respectively. The circles enclose the H2O and bidentate sulﬁte (Zr(−O−)2SO) reaction products, and the green line indicates a H-bond (O···H distance = 1.70 Å). The view is from an oblique angle so as to show the relevant bonds clearly.

which showed the bidentate structure to persist after ∼9 ps at 300 K. In computing ΔEads, the value of E(V) was obtained for the relaxed structure without SO2 prior to AIMD. If instead SO2 is removed from the reaction product and the resulting structure is relaxed, then a more-stable SO2-free model is obtained. This leads to ΔEads = −1.06 eV, which is somewhat lower than the value of −1.29 eV found above. In the relaxation process, the H2O that formed during the reaction (Figure 9) dissociates and the H that is released returns to the Zr−O site to which it was originally bonded. Thus, the absolute value for ΔEads depends somewhat on how the bare surface is modeled. The computed result concurs with experiment as to the formation of a sulﬁte reaction product with the involvement of terminal OH sites. The three-center nature of the reaction requires that SO2 and the two terminal OH groups all be in a favorable conﬁguration simultaneously. This suggests that, in the absence of the vacancy, much longer AIMD run times would be needed to observe a reaction. This in turn suggests that, in the real material, vacancies and nanoparticle edges, which have a high local density of terminal OH sites, are the kinetically favorable reaction sites for sulﬁte formation. For comparison, a similar AIMD calculation for an ordered bilayer with a vacancy in one layer yielded a monodentate sulﬁte via a somewhat diﬀerent three-center reaction of the form Zr−OH + SO2 + HO−Zr → Zr−O(H)−SO2 + HO−Zr → Zr−O−SO2 + H2O−Zr. Here, the reaction, which is seen at 300 K, proceeds through an intermediate that is clearly observed as a terminal Zr−OH in which the O is also bonded to SO2, i.e., Zr−O(H)−SO2. For the Zr−O−SO2 product, a ΔEads of −1.59 eV was found, as above, where E(V) for the bare surface was computed by relaxing the structure obtained by removing SO2 from the product. This result is signiﬁcantly more exothermic than that for the bidentate sulﬁte described above; however, the eﬀect of diﬀerences in the local structure of the defect in the one- and two-layer models has not been determined. The stability of the monodentate sulﬁte at 300 K K

DOI: 10.1021/acs.jpcc.7b11107 J. Phys. Chem. C XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry C

K shows that the Zr−O−Zr bridge with the O bonded to S is asymmetric with one Zr−O bond longer (2.45 Å) than the other (2.16 Å). In yet another series of calculations, an AIMD trajectory was obtained at 650 K for a disordered bilayer with no defect, during which 20 H2O molecules were produced over a period of ∼21 ps by the bridging + terminal OH reaction described above. These were removed, which left a high density of Zrcus sites, and two SO2 molecules were added between the layers. Here, a rapid reaction with SO2 occurred, within ∼0.5 ps at 300 K, to form sulﬁtes, one of which was a 3-fold-coordinated (Zr− O−)3S species. Structures of this type have been identiﬁed theoretically for SO2 on OH-free γ-Al2O3(110) (DM7 and DM8 in ref 80) and are the most stable species in that case. The other SO2 formed a Zr−O−S(O)−O−Zr bridge. The system studied here is somewhat artiﬁcial in that experiment64 indicates that heating at 650 K leads to the onset of ZrO2 formation rather than to Zr(OH)4 with a high density of Zrcus sites. However, these results show that Zrcus, as well as terminal Zr−OH, can potentially contribute to SO2 chemisorption. The multiplicity of reaction products, mechanisms, and adsorption energies is a consequence of the amorphous nature of the Zr(OH)4 model, in which there is a distribution of reaction-site conﬁgurations with diﬀerent OH density and speciation and Zrcus concentration. This is qualitatively consistent with the observation2 of broad features (FWHM ≈ 100 K) in the TPD of SO2 chemisorbed on Zr(OH)4. A common aspect of all such reactions, however, is that some type of sulﬁte is formed and that one or two terminal Zr−OH sites are involved (in agreement with experiment) unless a high density of Zrcus sites is also available. The present results also indicate that the order of sulﬁte stabilities is bidentate < monodentate < bridging. Conversion of bidentate to monodentate is slow at 300 K, which indicates a signiﬁcant barrier; however, conversion of monodentate to bridging appears to be more rapid if a Zrcus site is available to form the second S−O−Zr bond. It is noted that for SO2 on OH-free γ-Al2O3(100) the monodentate → bridging barrier80 is only 1.19 kcal/mol = 51.6 meV. The ease with which bidentate sulﬁte forms in AIMD via adsorption of SO2 at low temperature suggests a low activation energy (ΔEa). This was studied using the CINEB approach with 15 “images” and the single-layer plus defect model (Figure 9), which gave ΔEa = +0.27 eV relative to the absolute lowestenergy molecularly adsorbed conﬁguration. The calculation was spin-restricted, but spin-unrestricted calculations were also done to check for a paramagnetic transition state (TS). No evidence of spin polarization was found, i.e., each image converged to a closed-shell conﬁguration when initialized as an open-shell system. In the TS, the SO2 shifts slightly such that the H-bond stabilizing the molecularly adsorbed state is weakened (i.e., lengthened). Figure 10 summarizes the results for the reaction and activation energies (not corrected for zeropoint energies). Any ambiguity in deﬁning the bare-surface energy E(V), as discussed above, aﬀects only the absolute energies in Figure 10 and not the energy diﬀerences. Because kBT = 17.2 meV at T = 200 K, it would appear that a ΔEa of +0.27 eV would preclude a reaction on the time scale of the present AIMD calculations. However, in performing the CINEB analysis, it was found that there are other quasi-stable H-bonded states for SO2 that are closer in energy to the top of the barrier (i.e., more like the TS). One, in particular, is 0.21 eV higher in energy than the lowest-energy molecularly adsorbed

was investigated using AIMD. After several picoseconds, rearrangement occurs to form a bridging sulﬁte, Zr−O− S(O)−O−Zr, via reaction with a Zrcus site. After relaxation at T = 0 K, this was found to be more stable by 1.58 eV than the relaxed monodentate structure, i.e., ΔEads = −3.17 eV. This is consistent with theoretical results80 for SO2 on the OH-free γAl2O3 (100) surface, which show the bridging sulﬁte (structure CM8 in ref 80) to be more stable than the monodentate sulﬁte (CM7) by 30.04 kcal/mol = 1.30 eV. The monodentate sulﬁte was used to study the possible formation of a monodentate bisulﬁte, Zr−O−S(O)OH. Here, a H was removed from either of two diﬀerent OH bridges or from an adsorbed H2O (as in Zr−OH2) and bonded to a free O atom of the sulﬁte. During relaxation at T = 0 K, the H returned to the site from which it had been removed or moved to a terminal OH to form Zr−OH2 or the bisulﬁte remained intact. However, the relaxed bisulﬁte (in the case where it remained intact) was less stable than the original monodentate sulﬁte by 0.81 eV. The Zr−OS bond length was 2.19 Å in the monodentate sulﬁte versus 2.45 Å in the bisulﬁte, which suggests a weaker chemisorption bond for the latter. Thus, the formation of monodentate bisulﬁte under the present conditions (i.e., dry Zr(OH)4 at 200−300 K) is exothermic (ΔEads = −0.78 eV) but less so than for either mono- or bidentate sulﬁte. Theoretical results80 for SO2 on the OHterminated γ-Al2O3(110) surface also indicate that formation of a monodentate bisulﬁte (structure HM4 in ref 80) is exothermic (ΔEads = −17.53 kcal/mol = −0.76 eV) but much less so than for sulﬁte. Other types of bisulﬁte (i.e., bidentate and bridging) were not considered, but formation of a bisulﬁte with a free OH group was generally not observed, even as a transient species, in the present AIMD trajectories. On the other hand, an experimental study82 of SO2 adsorption at 300 K on anatase TiO2 nanopowders that were OH-terminated but nominally dry found evidence for bisulﬁte formation, at least during the initial exposures. In the present calculations, an exception occurred when SO2 was introduced between two Zr(OH)4 layers in the vicinity of multiple Zrcus and terminal Zr−OH sites. Here, the SO2 bonded to two Zrcus to form Zr−O−S−O−Zr followed by transfer of a terminal OH to yield Zrcus and a bridging bisulﬁte, Zr−O−S(OH)−O−Zr. This existed as a transient species and later transformed to a sulﬁte. In another calculation, at 200 K for a disordered bilayer with a defect, a bridging sulﬁte of the form (−Zr−)2O−S(O)−O− Zr resulted within ∼1.1 ps. Here, one of the O atoms in the sulﬁte bridge is itself bridging between two Zr atoms, which constitutes a 3-fold-coordinated O atom. A similar structure has been identiﬁed as the most stable product (HM5) in a theoretical study80 of SO2 on OH-terminated γ-Al2O3(110). In the present case, the AIMD trajectory reveals the ﬁrst step to be reaction between SO2 and a highly unsaturated, 5-foldcoordinated Zrcus to form an OS−O−Zr bond, followed by bond formation between S and the O in a Zr−O−Zr bridge. The OS−O−Zr intermediate is similar to the OS−O−Al species (HM3) identiﬁed80 as a stable entity for SO2 reacting with hydroxylated γ-Al2O3(110). No terminal OH is directly involved in the Zr(OH)4 reaction, although it is recalled that the H2O-forming reaction between terminal and bridging OHs is needed for the generation of Zr−O−Zr. The 5-foldcoordinated Zrcus involved in this reaction is a rare species because (Section 3.3) the average Zrcus coordination number in the current model is 7. Relaxation of the reaction product at 0 L

DOI: 10.1021/acs.jpcc.7b11107 J. Phys. Chem. C XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry C

chemical properties, results were obtained for adsorption and reaction of CO2. Experiment shows11 that CO2 at high pressure adsorbs reversibly on Zr(OH)4 (previously degassed by evacuation at room temperature) except for a small fraction that forms a bicarbonate, Zr−O−C(O)OH. In the ﬁrst calculations, CO2 placed close to the external surfaces of a disordered two-layer model desorbed quickly (within ∼100 fs) during AIMD at 300 K. Here, no defect was introduced into the Zr(OH)4 model, unlike in the case of SO2. However, placement between the layers, to preclude rapid desorption, led to bicarbonate formation within ∼1 ps, as shown in Figure 11.

Figure 10. Schematic diagram (approximately to scale) showing the energies for the SO2 reaction that forms a bidentate sulﬁte on a single Zr(OH)4 layer with a vacancy. The energies are in electronvolt, and “TS” means “transition state”. The “bare Zr(OH)4” energy is termed E(V) in the text, i.e., the energy of the bare unit cell with a vacancy. This was obtained for the relaxed structure without SO2 prior to AIMD, and the value used aﬀects only the absolute (not relative) energies.

structure, and it is probable that the reaction proceeds via one of these states rather than through the absolute lowest-energy molecularly adsorbed conﬁguration. Thus, ΔEa = +0.27 eV represents an upper limit on the eﬀective barrier height. Because of the computational cost of a complete CINEB analysis, ΔEa was not obtained for other reactions described above. However, the fact that all occur easily in the 200−300 K range suggests that all involve similarly low barriers. The initial state in the CINEB calculation is molecularly adsorbed SO2 relaxed at T = 0 K to the absolute lowest-energy conﬁguration. Molecular adsorption in the present context refers to adsorption via one or more H-bonds to O atoms in SO2. At the single-layer defect site (Supporting Information), this occurs via formation of a single short OH···OSO bond to a bridging OH, the acidic nature of which makes it a better Hbond donor than a terminal OH. The H···O distance is 1.77 Å, and ΔEads = −0.95 eV, as indicated in Figure 10. The amorphous nature of the material results in a diﬀerent local environment at each adsorption site, which in turn makes it diﬃcult to obtain a single universal ΔEads. At an arbitrarily chosen site farther from the defect, ΔEads = −0.66 eV is found where one O in SO2 H-bonds to a bridging OH and the other to a terminal OH with distances of 1.88 and 1.93 Å, respectively. Adsorption at an arbitrarily chosen site on a bilayer with no defect yields ΔEads = −0.91 eV with one of the O atoms in SO2 forming H-bonds to both a bridging (1.85 Å) and a terminal (1.97 Å) OH. These results might not correspond to the absolute lowest-energy conﬁgurations at the diﬀerent sites and are intended only to show the variation in ΔEads for molecular adsorption with the nature of the adsorption site. These results suggest that SO2 can undergo stable molecular adsorption on nominally dry Zr(OH)4 by bonding either to bridging OH or to both bridging and terminal sites. However, it appears that sulﬁte formation is more exothermic than molecular adsorption and that the eﬀective ΔEa for reaction is low because sulﬁte appears within a few picoseconds at 300 K or below in AIMD. The available XPS data3 show little or no adsorption of any kind for SO2 on Zr(OH)4 that has been calcined to eliminate terminal OH while leaving at least some bridging OH. However, TPD experiments2 indicate a small coverage of “weakly adsorbed” SO2 that is removed by heating at 373 K in ﬂowing N2. Such SO2 might also desorb at room temperature, and therefore not be detected, in the ultrahigh vacuum of an XPS apparatus. 3.4.5. Adsorption and Reaction of CO2. As a further test of the ability of the model to reproduce experimentally observed

Figure 11. Bicarbonate product (not relaxed) formed after ∼1 ps during AIMD at 300 K when CO2 is placed between two Zr(OH)4 layers. For clarity, the inset identiﬁes the atoms in the bicarbonate with the O atoms indicated by green circles.

This suggests that bicarbonate formation in the real material may involve CO2 retained in nanopores and cavities or may require a high CO2 pressure to increase the steady-state coverage (i.e., residence time). It is noted that the experiments described in ref 11 were done at CO2 pressures of up to ∼27 atm (2700 kPa). It is also noted69 that the species formed during exposure of Zr(OH)4 to CO2 under dry conditions may diﬀer from those produced as impurities during synthesis or by exposure to humid room air. As in some other AIMD calculations in the present work, H is replaced with D, although hydrogen will be described as H. The reaction mechanism observed in the AIMD trajectory at 300 K is Zr−OH + CO2 → (Zr)cus + HO−CO2

(1)

HO−CO2 + (Zr′)cus → HO(O=)C−O−Zr′

(2)

Zr−OH + HO(O=)C−O−Zr′ ↔ Zr−OH 2 + O2 C−O−Zr′ (Zr ″)cus

⎯⎯⎯⎯⎯⎯⎯→ Zr−OH 2 + Zr″−O−C(=O)−O−Zr′

(3)

Here, CO2 undergoes an acid−base reaction in which an OH is abstracted from a terminal Zr−OH (eq 1). The bicarbonate species then reacts with a diﬀerent CUS site (Zr′)cus to form the adsorbate (eq 2). The preferential involvement of terminal Zr− OH sites is consistent with experimental and theoretical results85 for CO2 reacting with OH-terminated surfaces of ZrO2; however, the ZrO2 results indicate formation of both M

DOI: 10.1021/acs.jpcc.7b11107 J. Phys. Chem. C XXXX, XXX, XXX−XXX

The Journal of Physical Chemistry C

■

ACKNOWLEDGMENTS This work was supported by the Defense Threat Reduction Agency. Computer facilities were provided by the DOD HighPerformance Computing Modernization Program at the AFRLMSRC, Wright Patterson Air Force Base, and by the Naval Research Laboratory. We are very grateful to R.B. Balow (NRL) and M. McEntee (ECBC) for providing the experimental IR spectra, and many helpful discussions with colleagues at ECBC (W.O. Gordon, C.J. Karwacki, G.W. Peterson, G.W. Wagner) and at NRL (D.E. Barlow, D. Gunlycke, P.E. Pehrsson, I.V. Schweigert) are gratefully acknowledged. Xianwei Sha is thanked for help in installing and implementing the GIPAW capability. Helpful communication with P.D. Southon, during the early stage of this work, is gratefully acknowledged.

bridging bicarbonate and bridging carbonate species. Subsequently, a dynamical process occurs in which H moves back and forth between the bicarbonate and a terminal OH (eq 3). After three such cycles, which occur over an interval of ∼1.5 ps (∼110 ν(O−D) periods) following the initial bicarbonate adsorption, the product shifts to a bridging carbonate via reaction with another CUS site (Zr″)cus. This structure remains essentially stable at 300 K for an additional ∼4.5 ps and then cycles between bridging and monodentate conﬁgurations during an additional 4 ps. The bridge in this case forms between the two Zr(OH)4 layers. The formation of both bicarbonate and bridging carbonate species has recently been observed experimentally69 for CO2 adsorption on nominally dry Zr(OH)4. As in the case of SO2, it is expected that these results will diﬀer in detail depending on the local structure of the reaction site. The calculation was spin-restricted, and no eﬀort was made at this stage to determine the possible eﬀects of paramagnetism in the HO−CO2 intermediate, which involves a single unpaired electron.

■

ASSOCIATED CONTENT

* Supporting Information S

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.jpcc.7b11107.

■

REFERENCES

(1) U.S. Environmental Protection Agency. Adsorption and Desorption of Chemical Warfare Agents on Activated Carbon: Impact of Temperature and Relative Humidity, EPA 600/R-14/261; U.S. Environmental Protection Agency: Research Triangle Park, NC, 2014. (2) Peterson, G. W.; Karwacki, C. J.; Feaver, W. B.; Rossin, J. A. Zirconium Hydroxide as a Reactive Substrate for the Removal of Sulfur Dioxide. Ind. Eng. Chem. Res. 2009, 48, 1694−1698. (3) Peterson, G. W.; Rossin, J. A.; Karwacki, C. J.; Glover, T. G. Surface Chemistry and Morphology of Zirconia Polymorphs and the Influence on Sulfur Dioxide Removal. J. Phys. Chem. C 2011, 115, 9644−9650. (4) Singh, J.; Mukherjee, A.; Sengupta, S. K.; Im, J.; Peterson, G. W.; Whitten, J. E. Sulfur Dioxide and Nitrogen Dioxide Adsorption on Zinc Oxide and Zirconium Hydroxide Nanoparticles and the Effect on Photoluminescence. Appl. Surf. Sci. 2012, 258, 5778−5785. (5) Peterson, G. W.; Wagner, G. W.; Keller, J. H.; Rossin, J. A. Enhanced Cyanogen Chloride Removal by the Reactive Zirconium Hydroxide Substrate. Ind. Eng. Chem. Res. 2010, 49, 11182−11187. (6) Peterson, G. W.; Rossin, J. A. Removal of Chlorine Gases from Streams of Air Using Reactive Zirconium Hydroxide Based Filtration Media. Ind. Eng. Chem. Res. 2012, 51, 2675−2681. (7) Glover, T. G.; Peterson, G. W.; DeCoste, J. B.; Browe, M. A. Adsorption of Ammonia by Sulfuric Acid Treated Zirconium Hydroxide. Langmuir 2012, 28, 10478−10487. (8) Bandosz, T. J.; Laskoski, M.; Mahle, J.; Mogilevsky, G.; Peterson, G. W.; Rossin, J. A.; Wagner, G. W. Reactions of VX, GD, and HD with Zr(OH)4: Near Instantaneous Decontamination of VX. J. Phys. Chem. C 2012, 116, 11606−11614. (9) Giannakoudakis, D. A.; Mitchell, J. K.; Bandosz, T. J. Reactive Adsorption of Mustard Gas Surrogate on Zirconium (Hydr)oxide/ Graphite Oxide Composites: the Role of Surface and Chemical Features. J. Mater. Chem. A 2016, 4, 1008−1019. (10) Kim, S.; Ying, W. B.; Jung, H.; Ryu, S. G.; Lee, B.; Lee, K. J. Zirconium Hydroxide-Coated Nanofiber Mats for Nerve Agent Decontamination. Chem. Asian J. 2017, 12, 698−705. (11) Kamimura, Y.; Endo, A. CO2 Adsorption-Desorption Performance of Mesoporous Zirconium Hydroxide with Robust Water Durability. Phys. Chem. Chem. Phys. 2016, 18, 2699−2709. (12) Schweigert, I. V.; Gunlycke, D. Hydrolysis of Dimethyl Methylphosphonate by the Cyclic Tetramer of Zirconium Hydroxide. J. Phys. Chem. A 2017, 121, 7690−7696. (13) Katz, M. J.; Mondloch, J. E.; Totten, R. K.; Park, J. K.; Nguyen, S. T.; Farha, O. K.; Hupp, J. T. Simple and Compelling Biomimetic Metal−Organic Framework Catalyst for the Degradation of Nerve Agent Simulants. Angew. Chem., Int. Ed. 2014, 53, 497−501. (14) Katz, M. J.; Klet, R. C.; Moon, S.-Y.; Mondloch, J. E.; Hupp, J. T.; Farha, O. K. One Step Backward Is Two Steps Forward: Enhancing the Hydrolysis Rate of UiO-66 by Decreasing [OH−]. ACS Catal. 2015, 5, 4637−4642.

4. CONCLUSIONS A model has been developed for amorphous Zr(OH)4 using AIMD and evaluated by comparison of predicted results with several diﬀerent types of experimental data, including structural, spectroscopic, and chemical measurements. The structural results include RDFs and the temperature dependence of the relative amounts of bridging and terminal OHs. The spectroscopic measurements include IR and 1H NMR, and the chemical results include the relative acidity of the two OH types and the reactions with SO2 and CO2. The good overall agreement between theory and experiment indicates that the model, even in its present form, will be a suitable basis for the analysis of Zr(OH)4 adsorption studies. Future developments can include the introduction of various defects and edge sites, for the purpose of modeling the eﬀects of ﬁnite particle size, and of ambient H2O vapor to simulate hydrolytic reactions. This will provide insight into the importance of humidity in the preparation and processing of this material for use in air puriﬁcation.

■

Article

Raman spectrum of the [Zr(OH)4]4 tetramer; freestanding [Zr(OH) 4 ]12 cluster model; molecularly adsorbed SO2 (PDF)

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. ORCID

Victor M. Bermudez: 0000-0002-3358-5450 Craig K. Knox: 0000-0002-6578-3208 Notes

The authors declare no competing ﬁnancial interest. ∥ Retired, Volunteer Emeritus (V.M.B.). N

DOI: 10.1021/acs.jpcc.7b11107 J. Phys. Chem. C XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry C (15) Katz, M. J.; Moon, S.-Y.; Mondloch, J. E.; Beyzavi, M. H.; Stephenson, C. J.; Hupp, J. T.; Farha, O. K. Exploiting Parameter Space in MOFs: a 20-Fold Enhancement of Phosphate-Ester Hydrolysis with UiO-66-NH2. Chem. Sci. 2015, 6, 2286−2291. (16) Li, P.; Klet, R. C.; Moon, S.-Y.; Wang, T. C.; Deria, P.; Peters, A. W.; Klahr, B. M.; Park, H.-J.; Al-Juaid, S. S.; Hupp, J. T.; et al. Synthesis of Nanocrystals of Zr-Based Metal-Organic Frameworks with Csq-Net: Significant Enhancement in the Degradation of a Nerve Agent Simulant. Chem. Commun. 2015, 51, 10925−10928. (17) Li, P.; Moon, S.-Y.; Guelta, M. A.; Lin, L.; Gómez-Gualdrón, D. A.; Snurr, R. Q.; Harvey, S. P.; Hupp, J. T.; Farha, O. K. Nanosizing a Metal-Organic Framework Enzyme Carrier for Accelerating Nerve Agent Hydrolysis. ACS Nano 2016, 10, 9174−9182. (18) Liu, Y.; Moon, S.-Y.; Hupp, J. T.; Farha, O. K. Dual-Function Metal-Organic Framework as a Versatile Catalyst for Detoxifying Chemical Warfare Agent Simulants. ACS Nano 2015, 9, 12358−12364. (19) López-Maya, E.; Montoro, C.; Rodríguez-Albelo, L. M.; Aznar Cervantes, S. D.; Lozano-Pérez, A. A.; Cenís, J. L.; Barea, E.; Navarro, J. A. R. Textile/Metal−Organic-Framework Composites as SelfDetoxifying Filters for Chemical-Warfare Agents. Angew. Chem., Int. Ed. 2015, 54, 6790−6794. (20) Mondloch, J. E.; Katz, M. J.; Isley, W. C., III.; Ghosh, P.; Liao, P.; Bury, W.; Wagner, G. W.; Hall, M. G.; DeCoste, J. B.; Peterson, G. W.; et al. Destruction of Chemical Warfare Agents Using MetalOrganic Frameworks. Nat. Mater. 2015, 14, 512−516. (21) Moon, S.-Y.; Liu, Y.; Hupp, J. T.; Farha, O. K. Instantaneous Hydrolysis of Nerve-Agent Simulants with a Six-Connected Zirconium-Based Metal−Organic Framework. Angew. Chem., Int. Ed. 2015, 54, 6795−6799. (22) Moon, S.-Y.; Wagner, G. W.; Mondloch, J. E.; Peterson, G. W.; DeCoste, J. B.; Hupp, J. T.; Farha, O. K. Effective, Facile, and Selective Hydrolysis of the Chemical Warfare Agent VX Using Zr6-Based Metal-Organic Frameworks. Inorg. Chem. 2015, 54, 10829−10833. (23) Peterson, G. W.; Moon, S.-Y.; Wagner, G. W.; Hall, M. G.; DeCoste, J. B.; Hupp, J. T.; Farha, O. K. Tailoring the Pore Size and Functionality of UiO-Type Metal-Organic Frameworks for Optimal Nerve Agent Destruction. Inorg. Chem. 2015, 54, 9684−9686. (24) Plonka, A. M.; Wang, Q.; Gordon, W. O.; Balboa, A.; Troya, D.; Guo, W.; Sharp, C. H.; Senanayake, S. D.; Morris, J. R.; Hill, C. L.; et al. In-Situ Probes of Capture and Decomposition of Chemical Warfare Agent Simulants by Zr-Based Metal Organic Frameworks. J. Am. Chem. Soc. 2017, 139, 599−602. (25) Ploskonka, A. M.; Marzen, S. E.; DeCoste, J. B. Facile Synthesis and Direct Activation of Zirconium Based Metal-Organic Frameworks from Acetone. Ind. Eng. Chem. Res. 2017, 56, 1478−1484. (26) Wang, G.; Sharp, C.; Plonka, A. M.; Wang, Q.; Frenkel, A. I.; Guo, W.; Hill, C.; Smith, C.; Kollar, J.; Troya, D.; et al. Mechanism and Kinetics for Reaction of the Chemical Warfare Agent Simulant, DMMP(g), with Zirconium(IV) MOFs: An Ultrahigh-Vacuum and DFT Study. J. Phys. Chem. C 2017, 121, 11261−11272. (27) Troya, D. Reaction Mechanism of Nerve-Agent Decomposition with Zr-Based Metal Organic Frameworks. J. Phys. Chem. C 2016, 120, 29312−29323. (28) Grüning, M.; Shaltaf, R.; Rignanese, G.-M. Quasiparticle Calculations of the Electronic Properties of ZrO 2 and HfO2 Polymorphs and Their Interface With Si. Phys. Rev. B 2010, 81, No. 035330. (29) Jiang, H.; Gomez-Abal, R. I.; Rinke, P.; Scheffler, M. Electronic Band Structure of Zirconia and Hafnia Polymorphs From the GW Perspective. Phys. Rev. B 2010, 81, No. 085119. (30) Gallino, F.; Di Valentin, C.; Pacchioni, G. Band Gap Engineering of Bulk ZrO2 by Ti Doping. Phys. Chem. Chem. Phys. 2011, 13, 17667−17675. (31) Ricca, C.; Ringuedé, A.; Cassir, M.; Adamo, C.; Labat, F. A Comprehensive DFT Investigation of Bulk and Low-Index Surfaces of ZrO2 Polymorphs. J. Comput. Chem. 2015, 36, 9−21. (32) Gerosa, M.; Bottani, C. E.; Caramella, L.; Onida, G.; Di Valentin, C.; Pacchioni, G. Electronic Structure and Phase Stability of Oxide Semiconductors: Performance of Dielectric-Dependent Hybrid

Functional DFT, Benchmarked Against GW Band Structure Calculations and Experiments. Phys. Rev. B 2015, 91, No. 155201. (33) Clearfield, A. Structural Aspects of Zirconium Chemistry. Rev. Pure Appl. Chem. 1964, 14, 91−108. (34) Mamott, G. T.; Barnes, P.; Tarling, S. E.; Jones, S. L.; Norman, C. J. Dynamic Studies of Zirconia Crystallization. J. Mater. Sci. 1991, 26, 4054−4061. (35) Turrillas, X.; Barnes, P.; Dent, A. J.; Jones, S. L.; Norman, C. J. ‘Hydroxide’ Precursor to Zirconia: Extended X-ray Absorption Fine Structure Study. J. Mater. Chem. 1993, 3, 583−586. (36) Turrillas, X.; Barnes, P.; Gascoigne, D.; Turner, J. Z.; Jones, S. L.; Norman, C. J.; Pygall, C. F.; Dent, A. J. Synchrotron-Related Studies on the Dynamic and Structural Aspects of Zirconia Synthesis for Ceramic and Catalytic Applications. Radiat. Phys. Chem. 1995, 45, 491−508. (37) Southon, P. D.; Bartlett, J. R.; Woolfrey, J. L.; Ben-Nissan, B. Formation and Characterization of an Aqueous Zirconium Hydroxide Colloid. Chem. Mater. 2002, 14, 4313−4319. (38) Mogilevsky, G.; Karwacki, C. J.; Peterson, G. W.; Wagner, G. W. Surface Hydroxyl Concentration on Zr(OH)4 Quantified by 1H MAS NMR. Chem. Phys. Lett. 2011, 511, 384−388. (39) DeCoste, J. B.; Glover, T. G.; Mogilevsky, G.; Peterson, G. W.; Wagner, G. W. Trifluoroethanol and 19F Magic Angle Spinning Nuclear Magnetic Resonance as a Basic Surface Hydroxyl Reactivity Probe for Zirconium(IV) Hydroxide Structures. Langmuir 2011, 27, 9458−9464. (40) Giannozzi, P.; Baroni, S.; Bonini, N.; Calandra, M.; Car, R.; Cavazzoni, C.; Ceresoli, D.; Chiarotti, G. L.; Cococcioni, M.; Dabo, I.; et al. QUANTUM ESPRESSO: a Modular and Open-Source Software Project for Quantum Simulations of Materials. J. Phys.: Condens. Matter 2009, 21, No. 395502. (41) Goldak, J.; Lloyd, L. T.; Barrett, C. S. Lattice Parameters, Thermal Expansions, and Grüneisen Coefficients of Zirconium, 4.2 to 1130 K. Phys. Rev. 1966, 144, 478−484. (42) Kittel, C. Introduction to Solid State Physics, 8th ed.; Wiley: New York, 2005. (43) Pauling, L. The Nature of the Chemical Bond, 3rd ed.; Cornell University Press: Ithaca, NY, 1960. (44) Grimme, S. Semiempirical GGA-Type Density Functional Constructed with a Long-Range Dispersion Correction. J. Comput. Chem. 2006, 27, 1787−1799. (45) Hoover, W. G. Molecular Dynamics; Springer-Verlag: Berlin, 1986. (46) Pickard, C. J.; Mauri, F. All-Electron Magnetic Response with Pseudopotentials: NMR Chemical Shifts. Phys. Rev. B 2001, 63, No. 245101. (47) CP2KFor further information, see https://www.cp2k.org/. (48) Krack, M. Pseudopotentials for H to Kr Optimized for GradientCorrected Exchange-Correlation Functionals. Theor. Chem. Acc. 2005, 114, 145−152. (49) VandeVondele, J.; Hutter, J. Gaussian Basis Sets for Accurate Calculations on Molecular Systems in Gas and Condensed Phases. J. Chem. Phys. 2007, 127, No. 114105. (50) Grimme, S.; Antony, J.; Ehrlich, S.; Krieg, H. A Consistent and Accurate Ab-Initio Parametrization of Density Functional Dispersion Correction (DFT-D) for the 94 Elements H-Pu. J. Chem. Phys. 2010, 132, No. 154104. (51) Chandler, D. Introduction to Modern Statistical Mechanics; Oxford University Press: Oxford, 1987. (52) Humphrey, W.; Dalke, A.; Schulten, K. VMD: Visual Molecular Dynamics. J. Mol. Graphics 1996, 14, 33−38. (53) VMDFor further information, see http://www.ks.uiuc.edu/ Research/vmd/. (54) CPMDFor further information, see http://cpmd.org/. (55) BaderFor further information see http://theory.cm.utexas. edu/henkelman/code/bader/. (56) Liu, Q.-J.; Liu, Z.-T.; Feng, L.-P. Elasticity, Electronic Structure, Chemical Bonding and Optical Properties of Monoclinic ZrO2 from First-Principles. Physica B 2011, 406, 345−350. O

DOI: 10.1021/acs.jpcc.7b11107 J. Phys. Chem. C XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry C (57) Soliz, J. R.; Klevitch, A. D.; Harris, C. R.; Rossin, J. A.; Ng, A.; Stroud, R. M.; Hauser, A. J.; Peterson, G. W. Structural Impact on Dielectric Properties of Zirconia. J. Phys. Chem. C 2016, 120, 26834− 26840. (58) Cho, H.-R.; Walther, C.; Rothe, J.; Neck, V.; Denecke, M.; Dardenne, K.; Fanghänel, T. Combined LIBD and XAFS Investigation of the Formation and Structure of Zr(IV) Colloids. Anal. Bioanal. Chem. 2005, 383, 28−40. (59) Ling, S.; Slater, B. Dynamic Acidity in Defective UiO-66. Chem. Sci. 2016, 7, 4706−4712. (60) Huang, C.; Tang, Z.; Zhang, Z. Differences between Zirconium Hydroxide (Zr(OH)4·nH2O) and Hydrous Zirconia (ZrO2·nH2O). J. Am. Ceram. Soc. 2001, 84, 1637−1638. (61) Štefanić, G.; Popović, S.; Musić, S. The Effect of Mechanical Treatment of Zirconium(IV) Hydroxide on its Thermal Behaviour. Thermochim. Acta 1995, 259, 225−234. (62) Turrillas, X.; Barnes, P.; Tarling, S. E.; Jones, S. L.; Norman, C. J.; Ritter, C. Neutron Thermodiffractometry and Synchrotron EnergyDispersive Diffraction Studies of Zirconium Hydroxide Calcination. J. Mater. Sci. Lett. 1993, 12, 223−226. (63) Gómez, R.; López, T.; Bokhimi, X.; Muñoz, E.; Boldú, J. L.; Novaro, O. Dehydroxylation and the Crystalline Phases in Sol-Gel Zirconia. J. Sol−Gel Sci. Technol. 1998, 11, 309−319. (64) Ali, A. A. M.; Zaki, M. I. HT-XRD, IR and Raman Characterization Studies of Metastable Phases Emerging in the Thermal Genesis Course of Monoclinic Zirconia via Amorphous Zirconium Hydroxide: Impacts of Sulfate and Phosphate Additives. Thermochim. Acta 2002, 387, 29−38. (65) Weakliem, P. C.; Carter, E. A. Constant Temperature Molecular Dynamics Simulations of Si(100) and Ge(100): Equilibrium Structure and Short-Time Behavior. J. Chem. Phys. 1992, 96, 3240−3250. (66) Steiner, T. The Hydrogen Bond in the Solid State. Angew. Chem., Int. Ed. 2002, 41, 48−76. (67) Harrick, N. J. Internal Reﬂection Spectroscopy; Harrick Scientiﬁc Corp.: Ossining, NY, 1967. (68) Dobson, K. D.; McQuillan, A. J. An Infrared Spectroscopic Study of Carbonate Adsorption to Zirconium Dioxide Sol-Gel Films from Aqueous Solutions. Langmuir 1997, 13, 3392−3396. (69) Balow, R. B.; Lundin, J. G.; Daniels, G. C.; Gordon, W. O.; McEntee, M.; Peterson, G. W.; Wynne, J. H.; Pehrsson, P. E. Environmental Effects on Zirconium Hydroxide Nanoparticles and Chemical Warfare Agent Decomposition: Implications of Atmospheric Water and Carbon Dioxide. ACS Appl. Mater. Interfaces 2017, 9, 39747−39757. (70) Frisch, M. J.; Trucks, G. W.; Schlegel, H. B.; Scuseria, G. E.; Robb, M. A.; Cheeseman, J. R.; Scalmani, G.; Barone, V.; Mennucci, B.; Petersson, G. A.et al. Gaussian 09, revision D.01; Gaussian, Inc.: Wallingford, CT, 2009. (71) Sosa, C.; Andzelm, J.; Elkin, B. C.; Wimmer, E.; Dobbs, K. D.; Dixon, D. A. A Local Density Functional Study of the Structure and Vibrational Frequencies of Molecular Transition-Metal Compounds. J. Phys. Chem. 1992, 96, 6630−6636. (72) Ambrosi, M.; Fratini, E.; Canton, P.; Dankesreiter, S.; Baglioni, P. Bottom-Up/Top-Down Synthesis of Stable Zirconium Hydroxide Nanophases. J. Mater. Chem. 2012, 22, 23497−23505. (73) Kharitonov, Y. Y.; Rozanov, I. A. Infrared Absorption Spectra of Zirconium Rhodanide Complexes. Bull. Acad. Sci. USSR, Div. Chem. Sci. (Engl. Transl.) 1962, 11, 373−377. Izvest Akad. Nauk S.S.S.R., Otdel. Khim. Nauk 1962, 402−407. (74) Scholz, J.; Scholz, K.; McQuillan, A. J. In-Situ Infrared Spectroscopic Analysis of the Water Modes of [Zr4(OH)8(H2O)16]8+ During the Thermal Dehydration of ZrOCl2·8H2O. J. Phys. Chem. A 2010, 114, 7733−7741. (75) Korhonen, S. T.; Calatayud, M.; Krause, A. O. I. Stability of Hydroxylated (1̅11) and (1̅01) Surfaces of Monoclinic Zirconia: A Combined Study by DFT and Infrared Spectroscopy. J. Phys. Chem. C 2008, 112, 6469−6476.

(76) Baltrusaitis, J.; Schuttlefield, J.; Zeitler, E.; Grassian, V. H. Carbon Dioxide Adsorption on Oxide Nanoparticle Surfaces. Chem. Eng. J. 2011, 170, 471−481. (77) Scherer, J. R. The Vibrational Spectroscopy of Water. In Advances in Infrared and Raman Spectroscopy; Clark, R. J. H.; Hester, R. E., Eds.; Heyden: London, 1978; Vol. 5, pp 149−216. (78) Mallamace, F.; Corsaro, C.; Broccio, M.; Branca, C.; GonzálezSegredo, N.; Spooren, J.; Chen, S.-H.; Stanley, H. E. NMR Evidence of a Sharp Change in a Measure of Local Order in Deeply Supercooled Confined Water. Proc. Natl. Acad. Sci. U.S.A. 2008, 105, 12725−12729. (79) Knözinger, H.; Ratnasamy, P. Catalytic Aluminas: Surface Models and Characterization of Surface Sites. Catal. Rev.: Sci. Eng. 1978, 17, 31−70. (80) Lo, J. M. H.; Ziegler, T.; Clark, P. D. SO2 Adsorption and Transformations on γ-Al2O3 Surfaces: A Density Functional Theory Study. J. Phys. Chem. C 2010, 114, 10444−10454 Note that in this paper exothermic adsorption energies are given as positive quantities.. (81) Nanayakkara, C. E.; Pettibone, J.; Grassian, V. H. Sulfur Dioxide Adsorption and Photooxidation on Isotopically-Labeled Titanium Dioxide Nanoparticle Surfaces: Roles of Surface Hydroxyl Groups and Adsorbed Water in the Formation and Stability of Adsorbed Sulfite and Sulfate. Phys. Chem. Chem. Phys. 2012, 14, 6957−6966. (82) Nanayakkara, C. E.; Larish, W. A.; Grassian, V. H. Titanium Dioxide Nanoparticle Surface Reactivity with Atmospheric Gases, CO2, SO2, and NO2: Roles of Surface Hydroxyl Groups and Adsorbed Water in the Formation and Stability of Adsorbed Products. J. Phys. Chem. C 2014, 118, 23011−23021. (83) Henkelman, G.; Uberuaga, B. P.; Jónsson, H. A Climbing Image Nudged Elastic Band Method for Finding Saddle Points and Minimum-Energy Paths. J. Chem. Phys. 2000, 113, 9901−9904. (84) Henkelman, G.; Jónsson, H. Improved Tangent Estimate in the Nudged Elastic Band Method for Finding Minimum-Energy Paths and Saddle Points. J. Chem. Phys. 2000, 113, 9978−9985. (85) Kouva, S.; Andersin, J.; Honkala, K.; Lehtonen, J.; Lefferts, L.; Kanervo, J. Water and Carbon Oxides on Monoclinic Zirconia: Experimental and Computational Insights. Phys. Chem. Chem. Phys. 2014, 16, 20650−20664. (86) King, G.; Soliz, J. R.; Gordon, W. O. Local Structure of Zr(OH)4 and the Effect of Calcination Temperature from X-ray Pair Distribution Function Analysis. Inorg. Chem. 2018, DOI: 10.1021/ acs.inorgchem.7b03137.

■

NOTE ADDED IN PROOF A recent study86 has analyzed the x-ray pair distribution function for amorphous Zr(OH)4. The results are generally consistent with the present theoretical model. However, Zr is typically found to be linked to 5 other Zr atoms in a plane, in a conﬁguration resembling that found in monoclinic ZrO2. This may aﬀect the spatial distribution of OH groups but is not expected to have a signiﬁcant eﬀect on chemical properties.

P

DOI: 10.1021/acs.jpcc.7b11107 J. Phys. Chem. C XXXX, XXX, XXX−XXX

Computational Modeling of the Structure and Properties of Zr(OH)4

Recommend Documents