Prabhat
Prakash
ac,
Shylendran
Ardhra
a,
Birane
Fall
b,
Michael J.
Zdilla
*b,
Stephanie L.
Wunder
*b and
Arun
Venkatnathan
*a
aDepartment of Chemistry and Centre for Energy Science, Indian Institute of Science Education and Research Pune, Dr Homi Bhabha Road, Pashan, Pune 411008, India. E-mail: arun@iiserpune.ac.in
bDepartment of Chemistry, Temple University, 1901-N 13th St., Philadelphia, PA 19086, USA. E-mail: mzdilla@temple.edu; slwunder@temple.edu
cMaterials Science and Engineering, Indian Institute of Technology, Gandhinagar, Gujarat 382355, India
First published on 1st March 2021
A new type of crystalline solid, termed “solvate sponge crystal”, is presented, and the chemical basis of its properties are explained for a melt- and press-castable solid sodium ion conductor. X-ray crystallography and atomistic simulations reveal details of atomic interactions and clustering in (DMF)3NaClO4 and (DMF)2NaClO4 (DMF = N-N′-dimethylformamide). External pressure or heating results in reversible expulsion of liquid DMF from (DMF)3NaClO4 to generate (DMF)2NaClO4. The process reverses upon the release of pressure or cooling. Simulations reveal the mechanism of crystal “juicing,” as well as melting. In particular, cation–solvent clusters form a chain of octahedrally coordinated Na+–DMF networks, which have perchlorate ions present in a separate sublattice space in 3:
1 stoichiometry. Upon heating and/or pressing, the Na+⋯DMF chains break and the replacement of a DMF molecule with a ClO4− anion per Na+ ion leads to the conversion of the 3
:
1 stoichiometry to a 2
:
1 stoichiometry. The simulations reveal the anisotropic nature of pressure induced stoichiometric conversion. The results provide molecular level understanding of a solvate sponge crystal with novel and desirable physical castability properties for device fabrication.
In this work, we consider an example from the class of salt-organic cocrystals (or salt-solvate cocrystals) of Zdilla, Wunder and co-workers: hybrid inorganic–organic systems where an inorganic salt forms soft-solid cocrystals between an organic molecule and a completely dissociative ion pair. This new class of materials exhibits Li+ ion or Na+ conduction in the solid-state.18,19 The immediate goal behind the synthesis of such crystals was to enable a weak interaction between alkali metal ions and their solvent matrix and to isolate and restrict the movement of anions to enhance positive-ion conduction for practical battery electrolyte applications. However, in one of the cases—where N,N-dimethylformamide cocrystalizes with sodium perchlorate as (DMF)3NaClO4,18—its crystal structure and morphologies have been observed to exhibit novel molecular properties, such as liquid-like grain boundaries and stimuli-responsive, reversible stoichiometric conversion to a different cocrystalline solvate species, where the application of pressure and/or temperature expels the organic solvent reversibly (the latter being the subject of this report). These cocrystals present a rare combination of coupled structural and thermomechanical behaviour and combine useful properties of other individual materials in a single example. For instance, solvent expulsion in these crystals could be comparable to a previously water respiring polymer composite.20 Switchable coordination (later described in results and discussion) could be similar to elastomer actuators with switchable covalent bonds,21 and changes in the unit cell accompanying solvent loss is related to single-crystal-to-single-crystal transformations.22,23
In addition to the unique thermomechanical behavior, these cocrystals of DMF and NaClO4 exhibit a low Ea barrier for Na+ ion hopping (25 kJ mol−1, from impedance spectroscopy) and have ionic conductivity at room temperature in the range of 10−4 S cm−1 to 10−3 S cm−1.18 Scanning electron microscopy (SEM, Fig. 1a) suggests the presence of an inherent liquid-like region at the surface/interface of the cocrystals of these electrolytes, which also results in self-binding grains with low boundary resistance and which do not require sintering. This nanolayer results from the decreased lattice energy at the surface of the crystal, which aids in grain binding.18,24,25 Unlike other solid electrolytes, this grain boundary facilitates—rather than impedes—the conduction of ions across or around grains, in addition to bulk-phase ion conduction. Unlike other adiponitrile or isoquinoline based cocrystalline electrolytes,25,26 the DMF based electrolytes possess one-dimensional channels of closely spaced Na+ or Li+ ions, e.g., (DMF)3NaClO4 (ref. 18) and DMF·LiCl.19 As viewed from single-crystal XRD, the cocrystals of (DMF)3NaClO4 have a one-dimensional channel of Na+ ions where the distance of successive Na+ ions is 3.23 Å, with 3:
1 DMF
:
Na+. Such sufficiently small interionic distances can facilitate ion conduction via vacancy-site-induced jumps, when a cation vacancy is created during ion transport from the electrode/electrolyte or intergranular interface.
The pressure and thermal effects in the mixed inorganic-salt-in-organic-solvent matrix presented in this case are highly quantifiable as these directly manifest in a separate stoichiometry of the cocrystals. Hence, explanation of the thermomechanical behaviour in the DMF–NaClO4 cocrystals at macroscopic and molecular scales is a key objective of this study. Advancements in the development of electrolytes and cocrystals have been accelerated with computer simulation methods like molecular dynamics (MD) simulations, which elucidate the mechanisms of thermal decomposition and ion conduction. In previous work,24 thermal behavior and ion conduction in a cocrystalline electrolyte DMF·LiCl for lithium ion batteries were measured experimentally and modelled using classical MD simulations and gas phase DFT calculations. While MD simulations provided a molecular-level understanding of melting/decomposition and of a surface nanoliquid layer facilitating grain binding in pressed pellets, DFT calculations provided atomic scale explanation of ionic clusters on the surface and in the bulk phase that contribute to ionic conductivity. We report here a combined experimental and theoretical study of a new physical property of this press-castable stimuli-responsive cocrystal: (DMF)3NaClO4 (Fig. 1b), termed the “solvate sponge crystal”. Under an applied pressure, this crystalline structure “juices” to release one equivalent of liquid DMF and forms crystalline (DMF)2NaClO4 (Fig. 1c). Upon the release of pressure, the liquid DMF is reabsorbed and (DMF)3NaClO4 is re-formed in the desired pressed shape, giving a solid pellet. This behaviour is observed macroscopically (visually) as well as at the molecular level using analytical tools (XRD and thermal analysis) and computation, which all support that the introduction of pressure or temperature alters the stoichiometry of the electrolytes.
(DMF)3NaClO4 was prepared using a previously published protocol.18
(DMF)2NaClO4 was prepared by supercooling a solution of NaClO4 salt in anhydrous DMF in liquid nitrogen. 2.9 grams of NaClO4 (0.02 mol) and 3 ml of DMF (0.039 mol) are placed in a heavy-walled pressure flask and, using the SOP for potentially explosive mixtures, (ESI†) heated to 70 °C with stirring to increase the solubility. After all the NaClO4 dissolved in the solvent, the solution was rapidly frozen by immersion in liquid nitrogen, resulting in the formation of a crystalline pellet. After thawing, a few single crystals were removed from the precipitate for single-crystal structure determination. The remaining residue was removed by decanting the mother liquor, and the residue is rinsed repeatedly with diethyl ether (Et2O), giving (DMF)2NaClO4 (2:
1) (Fig. S1†). The material was always contaminated by (DMF)3NaClO4 (3
:
1). The protocol gave quantitative yield of the mixture.
![]() | ||
Fig. 2 TGA and DTGA of (DMF)nNaClO4. TGA for 3![]() ![]() ![]() ![]() ![]() ![]() |
![]() | ||
Fig. 3 Thermal ellipsoid plots of five adjacent asymmetric units of (DMF)3NaClO4 (top)18 and (DMF)2NaClO4 (bottom) with Na⋯Na interatomic distances illustrated. Ellipsoids set at 50% probability level. Hydrogen atoms omitted for clarity. |
Based on the 100 K crystal structure of the 3:
1 cocrystals, every pair of Na+ ions is bridged by a facial array of three DMF-κ2O oxygen atoms making each Na+ ion six-coordinate, and ClO4− ions occupy an interstitial channel in the crystal (Fig. 3, top). In the 2
:
1 stoichiometry, one equivalent of DMF has been lost from each Na+ ion, and this ligand is replaced by perchlorate, which moves to a bridging position in a κ2O,O′ geometry, and forms a three-atom bridge across neighboring sodium ions (Fig. 3, bottom). The loss of DMF and assimilation of the perchlorate anion into the coordination sphere result in a 27% reduction of the molar volume of the crystal. The TGA result of (DMF)2NaClO4 is shown in Fig. 2. An early two-stage loss of DMF corresponds to the complete loss of DMF (54% by mass), after which the TGA shows similar results to pure NaClO4. At 550 °C, decomposition of NaClO4 ensues, leaving NaCl (an 83% total mass loss for (DMF)3NaClO4, and a 78% loss for (DMF)2NaClO4). A stoichiometric reaction scheme for these gradual mass losses is provided in Scheme 1. A list of chemical and physical properties of 3
:
1 vs. 2
:
1 stoichiometric cocrystals of DMF–NaClO4 is presented in Table S1.†
![]() | ||
Scheme 1 Stoichiometric changes in the cocrystals of DMF![]() ![]() |
![]() | ||
Fig. 4 Hand pressing of solvate sponge crystals (DMF)3NaClO4 between glass slides to expel liquid DMF and generate (DMF)2NaClO4. (A) Temple T background image. (B) Fresh, dry crystals of (DMF)3NaClO4 between glass slides. (C) Initial pressing to crush the crystals, mostly obscuring the Temple T. (D) After a few minutes of pressure, the crystals become a translucent slush, with the Temple T visible beneath. See the entire process in ESI Movie 1.† |
Juicing of (DMF)3NaClO4 crystals also occurs thermally, and studies of the temperature dependence of these two phases of DMF–solvated NaClO4 suggest a dynamic equilibrium between the 3:
1 and 2
:
1 solvates and offer clues to system thermodynamics. The 3
:
1 stoichiometry is always formed when crystals are grown from excess DMF solvent at low temperature. However, when heated past its melting temperature to 80 °C and re-cooled, the 2
:
1 solvate, (DMF)2NaClO4/DMF(l), slush forms initially, as evidenced by the PXRD pattern of this post-heated mixture after cooling to room temperature (Fig. 5, top).
Upon the release of pressure or heat from the DMF(l)/(DMF)2NaClO4 slush, the liquid DMF is reincorporated into the crystal to regenerate solid (DMF)3NaClO4 based on XRD analysis. At room temperature, over the course of 20 h, the 3:
1 cocrystal is partially re-formed from the slush, which shows a mixture of the 2
:
1 and 3
:
1 solvates (Fig. S2†). When cooled to 0 °C, reversion of the slush to 3
:
1 is much faster and a mixture of 3
:
1 and 2
:
1 solvates is apparent immediately (Fig. 5, middle). After 20 h at 0 °C, the material completely reabsorbs all the liquid DMF, reverting to the 3
:
1 mixture (Fig. 5, bottom). At an even lower temperature of −40 °C, the reabsorption of liquid DMF from the melt is observed by XRD immediately and the 2
:
1 mixture is not detected (Fig. S2†). These results demonstrate that the 3
:
1 phase formed at low temperature is more thermally stable than the 2
:
1 mixture and that the juicing of the crystal to form the 2
:
1 mixture is endothermic. The remarkable reversibility of this solvate sponge crystal suggests that the two phases are in thermal equilibrium. Reversibility implies that the juicing process to form the DMF(l)/2
:
1 mixture is more entropically favorable, such that the phases are in equilibrium at room temperature (ΔH > 0, ΔS > 0, ΔG = 0). This pressure- and temperature-dependent behavior is hypothesized to be responsible for facilitating the formation of highly ionically conductive pressed or melt-cast pellets with good intergrain conductivity in the 3
:
1 (DMF)3NaClO4.18
![]() | ||
Fig. 6 Na⋯O(ClO4−) and Na+···OClO3−···Na frameworks showing an abundance of Na–ClO4 clusters at T = 325 K. |
The abundance of these Na–ClO4 networks as a function of temperature is quantified from cluster analysis using the simulation trajectory (Fig. 7a and b). The cluster analysis shows that {Na⋯3O(DMF)}n networks of 96 atoms (consisting of Na and O(DMF) atoms) form parallel to the z-axis (c-crystallographic direction) of the simulation box at T < 300 K. This network also represents the prevalence and stability of the Na+ ion channel in the z-direction. At low temperatures (T < 300 K), the size of the largest Na–DMF and Na–Na clusters is large (Fig. 7a) and the total number of these clusters is small (Fig. 7b). The size of the largest of Na–DMF and Na–Na clusters decreases with increasing temperature and the number of these clusters increases, which affirms the visual observation of breaking of Na–DMF and Na–Na networks at T > 300 K (Fig. S4†). Conversely, at T < 300 K, many (∼2000) small (<10, atoms at most) Na–ClO4 clusters exist in the system (most of which constitute a single contact), which increase in size and decrease in number as the temperature increases, suggesting the formation of large Na–ClO4 clusters (size >1000 atoms) at higher temperatures.
![]() | ||
Fig. 7 (a) Size of the largest cluster and (b) number of total clusters (counting clusters of size of one atom also) of Na⋯O(ClO4−) (≤2.2 Å), Na⋯Na (≤3.5 Å) and Na⋯O(DMF) (≤3.0 Å) during simulated heating of model P; the distribution of the number of clusters with respect to their size is provided in Fig. S6,† (c) calculated coordination number of Na+ ions by oxygen ligand atoms of DMF and ClO4− anions in the cocrystalline (DMF)3NaClO4 model P at constant temperatures, simulated under the NpT ensemble; snapshots of the supercell of (DMF)3NaClO4 simulated as model V at (d) 100 K, (e) 300 K, and (f) 400 K. Color scheme: spheres (atoms): yellow—Na, red—O(ClO4), green—O(DMF), cyan—Cl(ClO4−); tubes (dynamic bonds): blue—Na⋯Na, red—Na⋯O(ClO4−), green—Na⋯O(DMF); lines—DMF. From (d) to (f), the increase in red tubes and corresponding decrease in blue and green tubes accompanies the displacement of DMF by ClO4−. |
The simulated heating approach is used to calculate the interplay of non-bonded interactions during melting, which is also consistent with the loss of DMF ligands and replacement of these contacts by perchlorate bridges. To understand the distribution of clusters with respect to their size, model P was simulated under the isothermal–isobaric ensemble at constant temperatures: T = 100 K (20 ns), 233 K (40 ns), 273 K (40 ns), 298 K (40 ns), 325 K (40 ns), and 350 K (20 ns). The histograms of the distribution of different sized clusters at various constant temperatures (Fig. S5†) show that the Na–Na and Na–DMF clusters are predominantly abundant at low temperatures. In contrast, most of the Na–ClO4 clusters are monoatomic (i.e. only 0 or 1 interionic contact) at low temperatures, which increases to a size of 10 atoms (i.e. three to four ion pairs per cluster) at T = 325 K. The cluster analysis also suggests that Na–ClO4 clusters, which form after the Na⋯O(DMF) networks collapse, are small in size, indicating the solvation of small clusters of ion pairs in DMF, rather than phase separation.
To understand the nature of pair interactions in the structure of (DMF)3NaClO4, the radial distribution function (RDF) is calculated at various temperatures for model P †. The RDFs suggest that in the first solvation shell of Na+ cations, DMF molecules occupy the shell at a distance of 3 Å to 3.5 Å, at T = 100 K, whereas ClO4− anions do not interact with the Na+ primary solvation shell and only interact at a distance > 5 Å. However, the Na⋯O(DMF) coordination number decreases as the temperature increases from 233 K to 298 K (Fig. 7c). In the coordination sphere of the Na+ cation, ∼2 O(DMF) is replaced by ∼2 O(ClO4−) contacts at a distance of 2 Å, at 298 K. This implies that (DMF)2NaClO4 could also form from (DMF)3NaClO4, provided that the residual DMF is removed from the system. The simulations do not model the exclusive formation of crystalline (DMF)2NaClO4 within these short timescales but show that under dynamical conditions, the cluster analysis and calculated coordination numbers indicate that the 3:
1 complex is most stable at low temperature and that DMF replacement by perchlorate is increasingly favorable at higher temperatures (endothermic). As the crystals melt, at T = 325 K, 350 K, coordination of ClO4− supersedes the coordination of DMF around Na+ cations. Overall, the Tm,sim = 325 K (= 52 °C) predicted from annealing simulations, cluster analysis and RDFs matches closely with the previously reported18 experimental Tm,exp = 55 °C (Fig. S7†).
While model P mimics the interior behavior of the cocrystals, the nature of the surface of (DMF)3NaClO4 was modeled using model V. Model V was constructed by placing the 6 × 6 × 12 supercell into a larger box of 15 × 15 × 18 nm3 with sufficient vacuum present at the either side of the supercell to avoid any significant interactions with its periodic image. From the simulated annealing of model V from T = 100 K to 500 K, with a heating rate of 20 K ns−1, a visual inspection was sufficient to extract these valuable insights: (1) at T = 100 K, the surface of cocrystals is fluid, with the presence of Na–ClO4 clusters and free DMF molecules on the surface, which are rare in the bulk (Fig. 7d); (2) at room temperature, the surface becomes even more abundant with Na–ClO4 clusters and free DMF, and new Na–ClO4 networks also become significant in the bulk (Fig. 7e); (3) at T = 400 K, the DMF molecules aggregate on the surface and some of them evaporate (analogous to thermal juicing and the observed mass loss during TGA) and more and larger Na+···ClO4− clusters form, resulting from DMF loss (Fig. 7f). This behavior mimics well the thermal juicing and ultimate decomposition behavior of (DMF)3NaClO4 observed experimentally.
![]() | ||
Fig. 8 Model V simulated under semiisotropic coupling, at Px = Py = 100 bar, Pz = 1 bar, at (a) t = 0, (b) t = 20 ns; at Px = Py = 1 bar, Pz = 100 bar, at (c) t = 0, (d) t = 100 ps, (e) t = 1 ns. The compressibility of cocrystals is higher in the z-direction (bottom) compared to that in the xy-directions (top) as it only takes a few picoseconds to compress the crystals along z (note that due to hexagonal crystallographic symmetry, the x and y directions are equivalent). The simulation trajectory of compression events is provided as ESI Movie 3.† |
A quantitative demonstration of pressure effects (similar to thermal effects) on cocrystals is shown from cluster size and number analysis during compression in Fig. S8 and S9.† A cluster histogram is shown in Fig. 9 which exhibits the effect of pressure anisotropicity in a time-averaged manner. The average size of Na–DMF clusters is ∼40 atoms (counting only Na and O atoms) at 1 bar pressure. On applying a high pressure in the x and y directions (Px = Py = 100 bar, Pz = 1 bar), the average cluster size decreases to 10 atoms, suggesting low-to-moderate chain-breaking in these crystallographic directions (Fig. S8a†). On applying a high pressure in the z direction (Px = Py = 1 bar, Pz = 100 bar), the average cluster size is ∼6 atoms, suggesting small fragments forming with significant Na–O(DMF) chain breaking. The sizes of Na–ClO4 clusters, which average ∼2.5 formula units initially, increase significantly on applying pressure from the z direction compared to applying pressure from the x and y directions (Fig. S8b†). Combining the size and numbers of clusters on a histogram (Fig. 9) suggests that high pressure in the z direction is responsible for more severe chain breaking and also more expulsion of DMF molecules from the Na+ coordination sphere in the cocrystals. The removal of these DMF molecules results in the formation of new Na–ClO4 clusters, which is consistent with the observed experimental conversion of the 3:
1 stoichiometry to 2
:
1 under applied pressure. Similar to the thermal conversion case, the pressure induced complete conversion is difficult to observe computationally on such short timescales based on the limitations of simulation models, but providing visulaization for inferences on the molecular-level details on the process of thermal or pressure-induced juicing.
Footnote |
† Electronic supplementary information (ESI) available: SOP for potentially explosive mixtures, additional TGA, DSC, XRD data, additional MD simulation data, and X-ray crystallographic tables. CCDC 2018918. For ESI and crystallographic data in CIF or other electronic format see DOI: 10.1039/d0sc06455f |
This journal is © The Royal Society of Chemistry 2021 |