Manganese pyrosilicates as novel positive electrode materials for Na-ion batteries

A carbon-coated pyrosilicate, Na2Mn2Si2O7/C, was synthesized and characterized for use as a new positive-electrode material for sodium ion batteries. The material consists of 20–80 nm primary particles embedded in a z10 nm-thick conductive carbon matrix. Reversible insertion of Na ions is clearly demonstrated with z25% of its theoretical capacity (165 mA h g ) being accessible at room temperature at a low cycling rate. The material yields an average potential of 3.3 V vs. Na/Na on charge and 2.2 V on discharge. DFT calculations predict an equilibrium potential for Na2Mn2Si2O7 in the range of 2.8–3.0 V vs. Na/Na, with a possibility of a complete flip in the connectivity of neighboring Mn-polyhedra – from edge-sharing to disconnected and vice versa. This significant rearrangement in Mn coordination (z2 Å) and large volume contraction (>10%) could explain our inability to fully desodiate the material, and illustrates well the need for a new electrode design strategy beyond the conventional “down-sizing/coating” procedure.

A carbon-coated pyrosilicate, Na 2 Mn 2 Si 2 O 7 /C, was synthesized and characterized for use as a new positive-electrode material for sodium ion batteries. The material consists of 20-80 nm primary particles embedded in a z10 nm-thick conductive carbon matrix. Reversible insertion of Na + ions is clearly demonstrated with z25% of its theoretical capacity (165 mA h g À1 ) being accessible at room temperature at a low cycling rate. The material yields an average potential of 3.3 V vs. Na + /Na on charge and 2.2 V on discharge. DFT calculations predict an equilibrium potential for Na 2 Mn 2 Si 2 O 7 in the range of 2.8-3.0 V vs. Na + /Na, with a possibility of a complete flip in the connectivity of neighboring Mn-polyhedrafrom edge-sharing to disconnected and vice versa. This significant rearrangement in Mn coordination (z2Å) and large volume contraction (>10%) could explain our inability to fully desodiate the material, and illustrates well the need for a new electrode design strategy beyond the conventional "down-sizing/coating" procedure.
Once overlooked in favour of high energy density Li-ion batteries (LIBs), research in the eld of non-aqueous Na-ion batteries (NIBs) is now undergoing a revival by virtue of its economic advantages and the relative abundance of Na compared to Li. 1-3 Layered oxides (NaMO 2 , with M ¼ Co, Fe, Mn, etc.) are by far the most extensively investigated candidates as positive electrode materials. 4  , etc. polyanions have recently been reported and characterized as positive electrodes in NIBs with the ambition of overcoming the performance, cost and safety limitations of the layered oxides. 5 Generally, polyanion-based Mn compounds exhibit a higher cell potential (>0.5 V) than their Fe-counterparts, yet pose greater challenges in terms of electrochemical performance, typically resulting from the Jahn-Teller (J-T) distortion of Mn 3+ (3d 4 valence electron conguration), leading to poor electronic conductivity and sluggish structural conversion between their (de)lithiated/(de)sodiated phases. 6 Though nano-sizing of the active particles and the use of electronically conducting coatings 7 can oen resolve the rst challenge, the structural issues still remain. If the crystal structure framework is not able to withstand the strain caused by the J-T distortion of Mn 3+ , this might even result in signicant atomic rearrangement, making it energetically favourable for Mn to migrate into alkali ion sites and/or dissolve in the electrolyte. 8 Although these possibilities are certainly discouraging, the potential rewards for taming Mn in these electrochemical applications remain hugethereby justifying continued effort within the battery materials research community.
The structural stress on Na-ion (de)insertion could be greatly alleviated through the use of some polyhedral linkage in the structures. Recent 11 We report herein the synthesis of a novel carbon-coated, nanosized pyrosilicate Na 2 Mn 2 Si 2 O 7 and a preliminary study of its electrochemical properties in Na half-cells. Na 2 Mn 2 Si 2 O 7 has a theoretical capacity of 165 mA h g À1 , equal to that of other state-of-the-art Li-based compounds in this family of materials; typically, LiFePO 4 . 12 To the best of our knowledge, pyrosilicates have remained unexplored in terms of electrochemical properties until very recently, 13 unlike their orthosilicate counterparts (A 2 MSiO 4 , where M ¼ Mn, Fe, Co; A ¼ Li, [14][15][16][17]. A few relevant pyrosilicates, such as Na 2 Mn 2 Si 2 O 7 and Na 2 Zn 2 Si 2 O 7 were reported more than ve decades ago; they were synthesized hydrothermally under alkaline conditions at both high pressures and high temperatures. 21 We speculate here that this family of compounds could provide new opportunities in terms of cost-effective and environmentally sustainable materials for both Li-and Na-ion secondary batteries through careful control of composition, particle size and conductive coating. The target phase, carbon-coated Na 2 Mn 2 Si 2 O 7 ("Na 2 Mn 2 Si 2 -O 7 /C") was synthesized through a solvothermally assisted sol-gel technique based on the well-known mechanism of acidcatalyzed hydrolysis and condensation of tetraethyl orthosilicate (TEOS); see the ESI. † The structure was rened successfully (R p ¼ 3.87%, R wp ¼ 5.00%, c 2 ¼ 3.73) from powder XRD data ( Fig. 1), using a monoclinic structural model derived earlier from single-crystal XRD data. 22 A model involving mixed site occupation of Mn-and Na-sites 23 did not result in a signicant improvement in the t. The atomic parameters of the rened structural model are reported separately in Table S2 in the ESI. † Impurity-free manganese silicates are known to be difficult to synthesize, especially in the presence of carbon. 24 Indeed, the phase purity of Na 2 Mn 2 Si 2 O 7 /C is here z94 wt%, with z6 wt% of Na 2 MnSiO 4 impurity. The uncoated sample was free from impurity phases (Fig. S1 in the ESI †). An Electron Paramagnetic Resonance (EPR) spectrum of Na 2 Mn 2 Si 2 O 7 /C ( Fig. S6 in the ESI †) displays a single Lorentzian absorption line (width of 21 mT and a g-factor of 2.01) indicating the presence of Mn 2+ speciesin good agreement with an earlier report. 25 The structure of Na 2 Mn 2 Si 2 O 7 is made up of layers of Mn ions separated by [SiO 3 -O b -SiO 3 ] 6À silicate dimers (illustrated as conjoined blue tetrahedra in Fig. 1 inset). These dimers serve as spacers, effectively opening up the structure and accommodating the Na ions. The conjoined silicate tetrahedra have a close to staggered conguration relative to one another, and have a non-straight Si-O b -Si bond angle of 120.9 . The Si-O bond lengths are distributed over the range 1.46-1.74Å with the longest bond to the bridging O b as found in earlier studies of other pyrosilicates. 26 The Mn ions assume two distinct crystallographic positions within each Mn layer: tetrahedral in a corner-sharing conguration, and square-pyramidal (Mn-O < 2.5Å) with both corner and edge-sharing connectivity to their neighbouring coordination polyhedra (Fig. 1 inset).
The carbon content of the Na 2 Mn 2 Si 2 O 7 /C compound was estimated by thermogravimetric analysis (TGA) to be $13 wt% (Fig. S3 in the ESI †). Raman spectroscopy conrmed the presence of typical pyrolytic carbon on the surface of Na 2 Mn 2 Si 2 O 7 /C with maxima centred at 1594 cm À1 and 1345 cm À1 , attributed to the so-called G ("Graphitic") and D ("Disorder") bands 27,28 (Fig. 2a).
The spectrum of pristine Na 2 Mn 2 Si 2 O 7 shows only weak G and D bands, apparently due to minimal amounts of pyrolysis by-products from the acetate precursor. The additional peaks and bands below 1000 cm À1 are assigned to the pyrosilicate phase. 29 The fact that these characteristic peaks and bands are not visible in the Na 2 Mn 2 Si 2 O 7 /C sample demonstrates complete coverage of Na 2 Mn 2 Si 2 O 7 surface with highly Raman active carbon. The main peak at 691 cm À1 is ascribed to a symmetric stretching vibration n s (Si-O b -Si) of the pyrosilicate phase and serves as a ngerprint for the presence of the Si 2 O 7 groups. 30,31 The presence of the pyrosilicate bands is also conrmed by IR spectroscopy, where the bands related to the various symmetric stretching modes within the Si 2 O 7 groups appear at 689, 895, 927, 964 and 997 cm À1 for both C-coated and pristine specimens (Fig. 2b). An additional IR peak centred at 830 cm À1 can be assigned to an antisymmetric stretching vibration n as (Si-O b -Si) of the Si 2 O 7 groups, which is typical of non-straight Si-O b -Si bridges (Si-O b -Si angle < 180 ) and appears to be inactive in Raman. 32 The characteristic separation (D) between the symmetric (n s ) and antisymmetric (n as ) Si-O b -Si bridge stretching modes in Fig. 2b is then 141 cm À1 . As suggested by Lazarev, 33 this separation typically increases as the Si-O b -Si angle increases. Finally, the purity (within the detection limit) of both products is conrmed by the absence of IR absorption bands at higher wavenumbers, which might have arisen  from residues of their xerogel precursors and/or water in the nal powders.
Scanning electron microscopy (SEM) shows that Na 2 Mn 2 -Si 2 O 7 /C is made up of micron-sized agglomerates (>1 mm) of primary particles with typical dimensions $100 nm (Fig. S4 †). More accurate determination of the size of the primary particles is hindered by the presence of the surrounding carbon.
Using volume-weighted integral breadth (LVol-IB) in the XRD data, a crude evaluation of the average crystallite size gives $46 nm.
High resolution transmission electron microscopy (HRTEM) of the Na 2 Mn 2 Si 2 O 7 /C sample conrms that the secondary particles comprise agglomerated crystallites with typical size 20-80 nm (Fig. 3b and c). Characteristic d-spacings for Na 2 -Mn 2 Si 2 O 7 are d 110 z 7.3Å, d 111 z 4.5Å and d 230 z 3.2Å (Fig. 3b  and c). It is also evident that the crystallites are effectively embedded in a matrix of conductive carbon, which apparently inhibits the growth of the crystallites beyond nano-size under these synthesis conditions. The carbon is mainly amorphous (dominant sp 3 hybridization), although turbostratic ordering (sp 2 hybridized) is also observed in parts of the Na 2 Mn 2 Si 2 O 7 /C composite, in line with our Raman observations. A Selected Area Electron Diffraction (SAED) pattern originating from a larger area of the composite material shows that both the main phase and the minor impurity product are monoclinic. As such, they contain a large number of observable (and partly overlapping) diffraction lines, as evidenced by the numerous reciprocal lattice points in Fig. 3d. Totally correct indexing of the diffraction pattern is inhibited by instrument resolution, though the largest d-spacings (those closest to the primary beam) are uniquely attributed to the primary phase, Na 2 Mn 2 -Si 2 O 7 , by virtue of its signicantly larger unit cell (see dashed semicircles in Fig. 3d).
The typical size of the primary particlesspanning tens of nanometersalong with the limited thickness (z10 nm) of the conductive carbon coating should ensure short diffusion pathways for the mobile Na + ions and thereby enhance the overall electronic conductivity of the entire composite electrode. Both factors are deemed necessary in order to mitigate the adverse effects of low conductivity in such polyanionic materials. Note that Na-ion conductivity in Na 2 Mn 2 Si 2 O 7 is not expected to be a performance limiting factor considering the similarity of the Bond Valence Sum (BVS) based Na-ion migration pathways to those in Na 2 Fe 2 Si 2 O 7 , 13 see Fig. S9. † Na half-cells with Na 2 Mn 2 Si 2 O 7 /C and Na 2 Mn 2 Si 2 O 7 cathodes both exhibit an open circuit voltage (OCV) of about 2.5-2.6 V vs. Na + /Na. However, Na 2 Mn 2 Si 2 O 7 shows poor charging performance, as expected in the absence of a proper conductive coating. It was therefore excluded from our further studies.
The cells with Na 2 Mn 2 Si 2 O 7 /C clearly cycled better providing approximately 25% of the theoretical capacity at a low cycling rate (C/100); see the inset in Fig. 4.
The differential capacity plot in Fig. 4 reveals two oxidation maxima at z3.2 and z3.4 V in the rst charge and a broadened peak around 3.4 V in subsequent cycles. Only one maximum near 2.8 V is observed during discharge, together with a sloping voltage prole below 2.8 V (see the inset). Increasing the upper voltage cut-off beyond 4.0 V makes it possible to access z50% of the theoretical capacity (Fig. S2 inset in the ESI †). However, this was detrimental in terms of higher polarization and could well cause the onset of electrolyte oxidation. Such sluggish cycling behaviour cannot be ascribed to poor electronic contact  in the active material or to long diffusion times for the Na + ions, since our HRTEM data show the opposite. Alternatively, structural rearrangement, occurring in the desodiation-induced amorphization of Na 2 MnSiO 4 , 16 could lead to such electrochemical behavior.
Density Functional Theory (DFT) calculations were used to probe in more depth the possible structural origins of such a slow reaction kinetics and the local structural rearrangement occurring under sodium removal from Na 2 Mn 2 Si 2 O 7 . In a fully sodiated structure (Fig. 5a), the tetrahedrally coordinated Mn1 sites are z4Å apart, while the ve-fold coordinated Mn2 sites form an edge-sharing dimer. However, a completely desodiated structure (Fig. 5d) shows an inversion of the Mn1-Mn1 and Mn2-Mn2 connectivity: the Mn1 sites switch to ve-coordination and edge-sharing, whereas the Mn2 sites break up their edge-sharing conguration through a rearrangement of the coordinating oxygen positions. This is highlighted by the change in the Mn2-O3 distance from z4.3Å in the fully sodiated (Fig. 5a) to z2Å in the fully desodiated structure (Fig. 5d). Note that large changes in the metal-oxygen distance (up to 1.2Å) were reported on the extraction of guest ions for structurally related Li 2 MnP 2 O 7 . This was argued to be the main cause of the sluggish cycling performance in this system. 11 Here, we observe even larger changes in the bonding environment of the Mn ions; displacements over a range of z2Å occur on full extraction of Na (Fig. S8 in the ESI †). Accordingly, it is not surprising that large overpotentials occur to overcome the excessive stress built-up in the structure, especially considering the substantial volume contraction (Dvol ¼ À12%) and monoclinic distortion (Db ¼ +15 ) accompanying this conversion (Fig. S7 in the ESI †).
Structural features of this kind could also account for similarly modest electrochemical behaviour in the isostructural Na 2 Fe 2 Si 2 O 7 system. 13 Simulated XRD patterns from the DFT models for x ¼ 2, 1(Na2), 1(Na1) and x ¼ 0 suggest a gradual increase of the (020) reection intensity relative to the other peaks (Fig. S2c †). This is consistent with ex situ XRD data collected for a cycled electrode recovered aer charging to 4.2 V using constant current followed by holding the voltage at this constant cut-off value (Fig. S2b †). Such an increase in the intensity of (020) reection originates from the increase in the ordering of the layer structure of Na 2 Mn 2 Si 2 O 7 under desodiation (Fig. 6).
The calculated OCV values shown for the extraction of Na1 (with the Na2 site occupied, Fig. 5b) and Na2 (with the Na1 site occupied, Fig. 5c) are 2.8 V and 2.9 V vs. Na + /Na, respectively. We see here that the conguration in which the Na2 site is still occupied (Fig. 5b) corresponds to a smaller structural rearrangement (judging from the changes in the Mn-Mn and Mn-O distances) compared to the initial structure and is therefore the more probable de-sodiation path in the rst charge half-cycle (see S2.7 in the ESI † for further details).
Nevertheless, it is likely that Na1 site occupation (Fig. 5c) can occur at higher overpotentials, and this can also lead to structural perturbation, as reected in experimental differential capacity data (Fig. 4) which shows a smearing out of the two distinct oxidation processes during subsequent cycling. We speculate, however, that this fully desodiated structure (Fig. 5d) cannot be formed under the test conditions because the available capacity is <50% of the theoretical value, despite the calculated OCV value of z3 V vs. Na + /Na (Fig. 5a-d). In practice, only one sodium per formula unit is extracted from Na 2 Mn 2 Si 2 O 7 .
To conclude, manganese pyrosilicate was successfully prepared for the use as a new positive electrode material for sodium ion batteries. Despite its very attractive working voltage of $2.8 V and high theoretical capacity of 165 mA h g À1 , the electrochemical activity of this compound, as exhibited in Na half-cells, appears rather moderate. DFT calculations reveal the possibility for complete ip in the connectivity of neighbouring Mn-polyhedrafrom edge-sharing to disconnected and vice versa. Such a large rearrangement of the Mn coordination (z2Å) accompanied by a substantial volume contraction (>10%) could well explain our difficulties in completely desodiating the material. Clearly, some alternative to a conventional "down-sizing/ coating" strategy is needed in these and also other polyanion structures exhibiting similar structural freedom. Further studies of the pyrosilicate family using, for example, structure stabilizing dopants, could open new opportunities for cost-effective and environmentally friendly materials for Li-and Na-ion batteries.

Conflicts of interest
There are no conicts to declare.