Phenylphosphonate surface functionalisation of MgMn2O4 with 3D open-channel nanostructures for composite slurry-coated cathodes of rechargeable magnesium batteries operated at room temperature

Spinel-type MgMn2O4, prepared by a propylene-oxide-driven sol–gel method, has a high surface area and structured bimodal macro- and mesopores, and exhibits good electrochemical properties as a cathode active material for rechargeable magnesium batteries. However, because of its hydrophilicity and significant water adsorption properties, macroscopic aggregates are formed in composite slurry-coated cathodes when 1-methyl-2-pyrrolidone (NMP) is used as a non-aqueous solvent. Functionalising the surface with phenylphosphonate groups was found to be an easy and effective technique to render the structured MgMn2O4 hydrophobic and suppress aggregate formation in NMP-based slurries. This surface functionalisation also reduced side reactions during charging, while maintaining the discharge capacity, and significantly improved the coulombic efficiency. Uniform slurry-coated cathodes with active material fractions as high as 93 wt% can be produced on Al foils by this technique employing carbon nanotubes as an electrically conductive support. A coin-type full cell consisting of this slurry-coated cathode and a magnesium alloy anode delivered an initial discharge capacity of ∼100 mA h g−1 at 25 °C.

A general approach to enhance the insertion and extraction of Mg 2+ ions is to reduce the particle size of host lattice and minimise the diffusion length of Mg 2+ ions in it. 11,12,14,[18][19][20]23,25,26 Recently, structured MgMn 2 O 4 (theoretical discharge capacity: 270 mA h g À1 ) with continuous three-dimensional (3D) macroand mesopores has been synthesised 20,23 by a propylene-oxidedriven sol-gel method. [27][28][29] Because of the small particle size ($10 nm), high surface area ($100-300 m 2 g À1 ), and controlled bimodal pore size distribution in the micrometre (1-10 mm) and nanometre (10-100 nm) regions, cathodes of the material prepared by the sol-gel method outperform those of conventionally prepared MgMn 2 O 4 . Very recently, this structured MgMn 2 O 4 has been used as the cathode active material in coin-type full cells operated at room temperature. 23 While promising, the hydrophilicity of the structured MgMn 2 O 4 limits its practical application in RMB fabrication: uniform slurry-coated cathodes are essential components, which cannot be prepared easily because of the tendency of the structured MgMn 2 O 4 to aggregate in non-aqueous solvents like 1-methyl-2-pyrrolidone (NMP).
Side reactions during charging are common for transition metal oxide cathodes and present another challenge to their application in RMBs. Side reactions, such as the oxidative decomposition of organic electrolytes on the cathode surface, are usually associated with the high catalytic activities of transition metals 17 and have been suppressed effectively by incorporating inert cations (e.g. Fe) into the spinel host lattice, 17 or by coating cathode materials with less reactive oxides like V 2 O 5 . 21 This study was aimed at developing a facile technique to functionalise the surface of the structured MgMn 2 O 4 and suppress both water adsorption and undesirable electrochemical side reactions, while retaining the discharge capacity. Organic phosphate compounds were selected because of their ability to readily form strong chemical bonds with transition metal ions. [30][31][32][33][34] The high selectivity of the functionalisation is expected to result in the formation of thin uniform monolayer that minimises the hindrance to the insertion and extraction of Mg 2+ ions and passivates the active sites of side reactions. Phenylphosphonic acid was selected as a model compound containing hydrophobic functional groups. Additionally, anchoring phenyl groups to the surface of active materials is attractive, as p interactions between the phenyl groups and carbon-based electrically conductive supports, such as carbon nanotubes (CNTs), help strengthen the contact between them, potentially leading to an increase in the fraction of active material.

Experimental procedure
The structured MgMn 2 O 4 powder was prepared following a reported procedure. 20 Stoichiometric amounts of magnesium and manganese chlorides (18 mmol in total) and citric acid (18 mmol, Fujilm Wako Pure Chemical) were dissolved in 20 mL of ethanol, and propylene oxide (12 mL, Kanto Chemical) was added. The resulting metal-organic complex gel was maintained for 1 day at 25 C, washed with ethanol and acetone to remove byproducts, and subjected to sequential solvent exchange with acetone and cyclohexane three times in 3 days. The resulting wet gel was freeze-dried using liquid nitrogen and heat treated for 5 h at 300 C in air. The specic surface area of the powder evaluated by the Brunauer-Emmett-Teller (BET) method using a nitrogen adsorption isotherm was $100 m 2 g À1 . Ammonium phenylphosphonate (PhPO(ONH 4 ) 2 ) was obtained by adding excess aqueous ammonia (10 wt%) to an aqueous solution of phenylphosphonic acid (Tokyo Chemical Industry) and drying the solution at 80 C. The resulting ammonium phenylphosphonate (1 mmol) was dissolved in 10 g of methanol, along with the structured MgMn 2 O 4 powder (2.5 mmol), and the mixture was stirred for 3 h at room temperature to functionalise the MgMn 2 O 4 with phenylphosphonate groups. The resulting suspension was centrifuged, washed twice with methanol, and dried at 60 C in air to provide the phenylphosphonate-functionalised MgMn 2 O 4 as a powder. The resulting samples were evaluated by powder X-ray diffraction (XRD, SmartLab, Rigaku) and Fourier-transform infrared (FT-IR) spectrometry (FT/IR-4600, JASCO) using an attenuated total reection (ATR) unit with a diamond prism. Thermogravimetry and differential thermal analysis (TG-DTA, DTG-60, Shimadzu) were carried out at a heating rate of 5 K min À1 in air.
Dry composite cathodes were prepared by mixing the pristine or surface-functionalised MgMn 2 O 4 powder, acetylene black (AB, Denka; electrically conductive support), and poly(tetrauoroethylene) (PTFE, Du Pont-Mitsui Fluorochemicals; binder) at a weight ratio of 60 : 30 : 10, and the composite ($2 mg) was pressed on a Pt mesh. Electrochemical measurements of the composite cathode were conducted at 100 C in an Ar-lled glovebox with a three-electrode cell using a Mg ribbon (99.9%, Fujilm Wako Pure Chemical) as the counter electrode, and Ag wire immersed in a triglyme (G3, Kanto Chemical) solution of 0.01 mol dm À3 AgNO 3 (Kanto Chemical) and 0.1 mol dm À3 magnesium bis(triuoromethanesulfonyl)amide ( , and used as the electrolyte solution. Galvanostatic charge-discharge tests were carried out using an electrochemical analyser (HZ-Pro, Hokuto Denko) at 10 mA g À1 in the potential range from À1.6 to 0.6 V vs. Ag/Ag + (from 1.0 to 3.2 V vs. Mg/ Mg 2+ ). It was initiated from the discharge step, and the charge capacity was restricted to 135 mA h g À1 (half of the theoretical capacity of MgMn 2 O 4 ) to minimise the undesirable oxidative decomposition of the electrolyte solution.
Slurries to fabricate coated electrodes were prepared by mixing the MgMn 2 O 4 powder, carbon nanotube (CNT, Cnano; electrically conductive support), and poly(vinylidene diuoride) (PVDF, Kureha; binder) at a weight ratio of 93 : 4 : 3 in NMP (Fujilm Wako Pure Chemical). The slurry was applied on an Al foil and dried at 80 C overnight under vacuum. The resulting slurry-coated cathode (f 9.5 mm) was encapsulated in a 2032type coin cell using a Mg-Al-Zn alloy plate (AZ31, Nippon Kinzoku, 3 wt% Al, 1 wt% Zn, f 9.5 mm, 44 mm thick) as the anode, and glass lter paper (GA-55, Advantec) as the separator. A 0.3 mol dm À3 G3 solution of magnesium tetrakis(hexa-uoroisopropyloxy)borate (Mg[B(hp) 4 ] 2 ) was chosen as the electrolyte solution because of the low overpotentials for the Mg anode dissolution and deposition. [37][38][39] Galvanostatic chargedischarge tests were performed using a battery testing system (HJ1020mSD8, Hokuto Denko) at 25 C and 5 mA g À1 in the 0.1-4.0 V potential range, and the charge capacity was restricted to 135 mA h g À1 . The slurry-coated cathodes were also characterised by scanning electron microscopy (SEM; JSM-6490A, JEOL) and energy-dispersive X-ray spectroscopy (EDS; JED-2300, JEOL). functionalisation with phenylphosphonate groups, several new bands appeared. From the similarity of the spectra between this sample and related phenylphosphonate-functionalised transition metal oxides, absorption bands at $1010 and $1105 cm À1 were attributed to the P-O stretching modes. 30,31,34 Sharp absorption bands at $1146 and $1438 cm À1 were assigned to the phenylphosphonate P-C stretching and n 19b C-C ring modes, respectively. 34,44 The absence of the P]O double bond stretching mode peak at 1200-1250 cm À1 suggests that the phenylphosphonate groups are covalently bonded to the MgMn 2 O 4 surface and that the P]O bonds are converted to P-O-(Mn,Mg) bonds. 30,31,34 A broad peak at 2500-3800 cm À1 was observed in the pristine MgMn 2 O 4 , which was ascribed to an O-H stretching mode with a shape comparable to that of water conned in mesopores. 45 The component at wavenumbers below $3000 cm À1 , which is absent in bulk liquid water, 46 suggests the presence of structured water, i.e., an ice-like hydrogen bonding network between water molecules and pore walls. These observations conrm signicant water adsorption properties in mesopores of the pristine MgMn 2 O 4 . Surface functionalisation resulted in reduced intensity of the broad absorption centred at $3300 cm À1 and the appearance of a narrow absorption band at $3500 cm À1 . The narrow band was assigned to isolated OH groups, which, when coupled with the shi to higher frequencies, indicates a weakening in hydrogen bonding 47 caused by the partial desorption of physisorbed water molecules and isolation of surface OH groups.

Results and discussion
To visualise the effect of surface functionalisation, samples were dispersed in a water-toluene biphasic mixture as shown in Fig. 3. The pristine MgMn 2 O 4 was precipitated at the bottom of the water layer. In contrast, the surface-functionalised MgMn 2 O 4 was mostly suspended at the bottom of the toluene layer. These results conrmed that the surface of the structured MgMn 2 O 4 was rendered hydrophobic aer functionalisation with phenylphosphonate groups.
The coverage of phenylphosphonate groups on the structured MgMn 2 O 4 surface was evaluated using TG-DTA, and the results are shown in Fig. 4. In the pristine powder a two-step weight loss was observed, which can be explained by the desorption of physisorbed water ((250 C) and dehydration of OH groups along the MgMn 2 O 4 grain growth (T350 C). In the functionalised powder, the weight loss by water desorption below $150 C was smaller than in the pristine powder, consistent with the ATR-FT-IR results shown in Fig. 1. However, exothermic peaks associated with the combustion of organic substances were observed at higher temperatures. The weight loss at $450 C ($4%) was attributed to the thermal decomposition of phenyl groups. We assumed that this weight loss was associated with the conversion of C 6 H 5 PO 2 to PO 5/2 , and the residue at 800 C was formally represented as MgMn 2 O 4 $xPO 5/2 . The stoichiometry, x, of the phenylphosphonate groups with    respect to MgMn 2 O 4 was calculated to be x x 0.14, which was equivalent to a surface density of $4 nm À2 for the phenylphosphonate groups when coupled with the MgMn 2 O 4 surface area ($100 m 2 g À1 ). This equates to approximately half of the Mn surface density in MgMn 2 O 4 (e.g. 6.1 nm À2 for the (001) face and 9.6 nm À2 for the (101) face), suggesting that there are approximately two Mn atoms per phenylphosphonate group on the surface. This is consistent with the bridged bidentate coordination mode known for phenylphosphonate-and phenylphosphinate-functionalised surfaces of atomically-at transition metal oxides. [32][33][34] Fig . 5 show galvanostatic charge-discharge curves of dry composite cathodes of the pristine and functionalised MgMn 2 O 4 recorded at 100 C. The current density was normalised with respect to the mass of the MgMn 2 O 4 powder, which included the mass of phenylphosphonate groups in the functionalised one. The initial discharge capacities of the pristine and functionalised samples were $130 mA h g À1 . Aer the second cycle, the functionalised sample exhibited higher discharge capacities, indicating that the functionalisation with phenylphosphonate groups increased the utility of MgMn 2 O 4 . In the pristine sample, charging in the rst four cycles did not reach the cut-off (3.2 V vs. Mg/Mg 2+ ) and ended at the predetermined capacity limit of 135 mA h g À1 . In addition, the discharge capacity was signicantly smaller than the corresponding charge capacity over each cycle, and poor coulombic efficiencies ($0.46-0.67) were achieved. These observations indicate that signicant side reactions occur during charging of the pristine sample. In contrast, in the functionalised sample, charging ended at the charge cut-off voltage, owing to the suppression of the side reactions, and the high coulombic efficiencies were obtained ($0.83-0.98). Fig. 6 shows images of the composite slurry-coated cathodes of the pristine and functionalised MgMn 2 O 4 applied on Al foil. The cathode of the pristine sample was not smooth and exhibited MgMn 2 O 4 aggregates, which were easily detached from the Al foil aer drying. In contrast, the cathode of the functionalised sample was uniform, and its adhesion to the Al foil was good. Fig. 7 shows SEM images of the coated cathodes. The smoothness of the coated cathode of the functionalised MgMn 2 O 4 was much better than that of the pristine MgMn 2 O 4 . Large particles seen in Fig. 7(a) were the aggregates of the pristine MgMn 2 O 4 formed during slurry preparation because such large particles were scarce before mixing. Thus, the surface functionalisation improved the homogeneity of slurries, and it would be responsible for the smoothness and good adhesion of the coated cathodes of the functionalised MgMn 2 O 4 . Fig. 8 shows EDS spectra of the coated cathodes. The P K peak (2.01 keV) was seen in the functionalised MgMn 2 O 4 , conrming the presence of phenylphosphonates. In this cathode, CNTs were used as the electrically conductive support, and the fraction of active material was as high as 93 wt%, which is, to the best of our knowledge, the highest reported thus far in the eld of RMBs. Fig. 9 shows the galvanostatic charge-discharge curves of the coin-type full cells with a slurry-coated cathode of the pristine or functionalised MgMn 2 O 4 and a Mg-Al-Zn alloy anode recorded at 25 C. The full cell with the cathode of the pristine sample   ($3 mm thick, active material loading $0.70 mg cm À2 ) delivered an initial discharge capacity of $50 mA h g À1 . However, the discharge capacity dropped abruptly to $1 mA h g À1 in the second cycle, and the cell broke during subsequent charging. We suggest that the degradation was the result of side reactions during charging, which may have been considerable during the gradual voltage decay at T10 mA h g À1 in the initial charge step. The poor cyclability and large polarisation indicate insufficient current collection arising from inhomogeneous mixing between the MgMn 2 O 4 and CNTs. The weak adhesion between the cathode and the Al foil may also contribute to the degradation of cyclability.
In contrast, the full cell of the surface-functionalised MgMn 2 O 4 ($5 mm thick, active material loading $0.74 mg cm À2 ) delivered an initial discharge capacity of $100 mA h g À1 , even at a low CNT fraction (4 wt%). The initial discharge capacity agreed well with the value ($100 mA h g À1 ) expected for the pristine structured MgMn 2 O 4 with the BET surface area of $100 m 2 g À1 , when evaluated at 25 C using coin-type full cells with dry composite cathodes at the active material : AB : PTFE weight ratio of 60 : 30 : 10. 23 These results demonstrate that the phenylphosphate functionalisation of the structured MgMn 2 O 4 surface signicantly improves the current collection efficiency in slurry-coated cathodes with high active material fractions. The improvement in the electrical contact between the surfacefunctionalised MgMn 2 O 4 and CNTs is attributed to p interactions between phenyl groups and carbon surfaces, increasing their affinity. Surface functionalisation also reduced polarisation during cycling. The suppression of side reactions by surface functionalisation would be another major contributing factor to such improvements seen in the electrochemical properties. The discharge capacity decreased to $50 mA h g À1 aer the second cycle. This capacity fading behaviour is also in accordance with that of the dry composite cathodes of the pristine structured MgMn 2 O 4 in coin-type full cells. 23 The charge curves of the coin-type cell exhibited plateau originating from the oxidative decomposition of electrolytes. In this experiment, a high cut-off voltage (4 V) was necessary to complete charging by compensating the overpotential of Mg deposition on the Mg-Al-Zn alloy anode, which uctuated during cycling. 23 Hence, it was difficult to maintain the cathode potential of the coin-type cell below 3.2 V vs. Mg/Mg 2+ , at which the electrolyte decomposition on the functionalised MgMn 2 O 4 was slow (Fig. 5(b)). This problem may be overcome by the improvement of anodes and electrolytes to decrease the overpotential of Mg deposition.
Thus, the phenylphosphonate functionalisation technique investigated is a highly effective strategy to ensure the fabrication of uniform slurry-coated cathodes of nanostructured hydrophilic transition-metal-based electrode active materials of high surface areas at high active material fractions, by preventing common aggregation issues in non-aqueous slurries for such materials.

Conclusions
A facile method for the surface functionalisation of transition metal oxide cathodes with phenylphosphonates has been developed, making use of the strong affinity between transition metal ions and phosphonate groups. The cathode active mateials were rendered hydrophobic by the presence of phenyl groups at the surface, and their use in RMBs was investigated. This technique suppressed the water adsorption of the hierarchically structured MgMn 2 O 4 , a promising high-voltage cathode active material for RMBs because of its high surface area (T100 m 2 g À1 ) and small particle size ($10 nm) that facilitate the insertion and extraction of Mg 2+ ions. This treatment reduced the aggregation of the structured MgMn 2 O 4 in NMP-based nonaqueous composite cathode slurries employing CNT as an electrically conductive support, and enabled the production of uniform slurry-coated cathodes with improved contact between  MgMn 2 O 4 and CNTs. In addition, the surface functionalisation suppressed side reactions during charging and signicantly increased the coulombic efficiency, while maintaining the discharge capacity. A coin-type full cell consisting of a slurrycoated cathode with active material fractions up to 93 wt%, a Mg-Al-Zn alloy anode, and a Mg[B(hp) 4 ] 2 electrolyte delivered an initial discharge capacity of $100 mA h g À1 at 25 C. This surface functionalisation technique is attractive for the development of practical slurry-coated cathodes of nanosized hydrophilic transition metal oxides for RMBs.

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