Unlocking the energy capabilities of micron-sized LiFePO4

Utilization of LiFePO4 as a cathode material for Li-ion batteries often requires size nanonization coupled with calcination-based carbon coating to improve its electrochemical performance, which, however, is usually at the expense of tap density and may be environmentally problematic. Here we report the utilization of micron-sized LiFePO4, which has a higher tap density than its nano-sized siblings, by forming a conducting polymer coating on its surface with a greener diazonium chemistry. Specifically, micron-sized LiFePO4 particles have been uniformly coated with a thin polyphenylene film via the spontaneous reaction between LiFePO4 and an aromatic diazonium salt of benzenediazonium tetrafluoroborate. The coated micron-sized LiFePO4, compared with its pristine counterpart, has shown improved electrical conductivity, high rate capability and excellent cyclability when used as a ‘carbon additive free' cathode material for rechargeable Li-ion batteries. The bonding mechanism of polyphenylene to LiFePO4/FePO4 has been understood with density functional theory calculations.

L i-ion battery, since its first commercialization in 1991, has drastically transformed and popularized portable electronic devices, and will continue to play a major role in the electrification of road transportation in the future 1 . However, for the realization of the latter, better energy storage materials are needed [1][2][3][4][5][6][7][8][9][10][11][12][13] . LiFePO 4 , an environmentally benign and relatively safe cathode material for rechargeable Li-ion batteries, has attracted a great deal of interest during the last few decades [2][3][4] . Considerable efforts have been devoted to overcoming the intrinsically low electrical conductivity of LiFePO 4 , a drawback that hinders its direct use in Li-ion cells 5,6 . Several strategies, such as doping with foreign metal ions, have been explored [7][8][9] . However, the most common approach remains coating with carbon 8 . Carbon coatings are usually formed during LiFePO 4 synthesis, in which an organic precursor (the carbon source) and the inorganic raw materials are mixed together. The subsequent calcination of the mixture in an inert or reducing atmosphere produces conducting carbon and LiFePO 4 , simultaneously [10][11][12] . Similarly, carbon coatings can also be introduced after LiFePO 4 synthesis, in which an organic precursor and preformed LiFePO 4 are mixed and then calcined 13,14 . The calcination-based strategies are often energy intensive and can be environmentally unfriendly because of the emission of harmful volatile organic compounds from the thermal decomposition of organic precursors 15 . Moreover, carbon coatings on LiFePO 4 produced by heat treatment tend to be irregular, which does not provide a good connectivity for the particles and hence the expected performance for battery applications 16 . To mitigate the negative environmental effects of calcination, conducting polymers have been employed to increase the electronic conductivity and thus improve the performance of LiFePO 4 (refs 17-22). Several methods have been used to produce polymer/LiFePO 4 composites, including electrochemical 19 and chemical 20 polymerization in the presence of LiFePO 4 particles; rapid mixing of conducting polymer colloidal and LiFePO 4 suspensions 21 ; and more recent spontaneous polymerization driven by the oxidation power of partially delithiated LiFePO 4 (ref. 22).
It shall be noted here that the above mentioned carbon and conducting polymer-coating procedures work well mainly on nano-sized LiFePO 4 , ranging typically from 200 to 20 nm (ref. 3). The reason for using nano-sized LiFePO 4 lies in that reducing particle size can shorten the solid-state diffusion distance within LiFePO 4 , which is beneficial to the high-power (or high rate) applications 23 . However, one obvious drawback associated with nano-sized LiFePO 4 is the decreased tap density (and the resultant lower volumetric energy density), which becomes critical when fitting LiFePO 4 -based batteries into the trunks of pure electric vehicles 3,4 . Although the literatures on LiFePO 4 are predominantly based on nano-sized materials, there are indeed some efforts of exploring submicron-and micron-sized LiFePO 4 . For instance, Dominiko et al. 24 showed that a 0.5-mm carbon-free LiFePO 4 has a specific capacity of 72 mAh g À 1 at 1 C; McNeil et al. 25 reported that a carbon-coated LiFePO 4 (0.5-1.0 mm) exhibited a specific capacity of 129 mAh g À 1 at 1 C; Wang et al.
reported that 0.5 mm LiFePO 4 could exhibit an initial capacity of 151 mAh g À 1 at 1 C and 58 mAh g À 1 at 10 C, but with limited cyclability 26,27 ; larger (45 mm) LiFePO 4 particles have been identified to have very poor performance even when coated with conducting carbon 25,26 . So far, no facile procedure has been reported to make high-performance micron-sized LiFePO 4 , possibly due to the limited robustness of the coatings that cannot keep the integrity of the LiFePO 4 particles during discharge and charging, particularly at high rates 27 .
Here we report a room-temperature method that can spontaneously coat micron-sized (B1.01 mm) LiFePO 4 uniformly with a thin conducting polymer of polyphenylene, as depicted in Fig. 1. The coated LiFePO 4 , compared with its pristine counterpart, has demonstrated enhanced electrical conductivity, high rate capability and excellent cyclability when used as a 'carbon additive free' cathode material for rechargeable Li-ion batteries. In addition, the bonding of polyphenylene to LiFePO 4 , Li 1 À x FePO 4 and FePO 4 has been explored by performing density functional theory (DFT) calculations for the adsorption of the phenyl radical on model surfaces of these compounds. It is concluded that phenyl preferentially forms a strong chemical bond to surface O sites under typical experimental and battery operational conditions, which could be disrupted in the unlikely event that the surface of LiFePO 4 becomes completely lithiated.

Results
Reaction of LiFePO 4 and C 6 H 5 N 2 þ BF 4 À . It is known that when an aromatic diazonium salt (ArN 2 þ X À ) is subjected to electrochemical or chemical reduction or thermal decomposition, an aryl radical (Ar Á ) will form, which is reactive and is an effective agent for surface functionalization of many kinds of substrates 28 . The obtained organic layers strongly adhere to the substrates because a covalent bond is thought to form between the substrate surface and the aryl radicals [29][30][31][32] . Although most of the organic layers reported in the literatures are insulating 28 , a few of them are indeed conducting when certain diazonium precursors are used 31,33,34 . One such diazonium salt is benzenediazonium tetrafluoroborate (C 6 H 5 N 2 þ BF 4 À ), which can be easily and economically prepared in a single-step synthesis 31 (Supplementary Figs 1 and 2 for the nuclear magnetic resonance (NMR) and electrospray ionization mass spectrometry (ESI-MS) characterization) and is employed here as the radical-generating agent to functionalize the micron-sized pristine LiFePO 4 . Powder X-ray diffraction (PXRD), scanning electron microscopy (SEM)/ transmission electron microscopy (TEM) and Fourier transform infrared (FTIR) characterizations  showed that the micron-sized LiFePO 4 is of pure phase and contains no adventitious impurities such as Li 2 CO 3 and LiOH.
To demonstrate the feasibility of the spontaneous reaction of LiFePO 4 and C 6 H 5 N 2 þ BF 4 À , an electrochemical measurement was performed. In a three-compartment cell, C 6 H 5 N 2 þ BF 4 À is electrochemically reduced at an Au electrode, where partially delithiated LiFePO 4 (that is, Li 1 À x FePO 4 , x ¼ 0.1) is used as the reference electrode because of its highly stable potential of 3.43 V versus Li þ /Li 35 . Figure 2 shows the cyclic voltammograms for the first 1-5 cycles. One distinct feature of the I-E curves is that the successive peak currents corresponding to the electroreduction of C 6 H 5 N 2 þ to C 6 H 5 Á radical does not decrease drastically, which is different from the insulating film-forming diazonium salts, such as 4-nitrobenzenediazonium tetrafluoroborate 36 , a benchmark compound for electrografting of diazonium salt. This observation is consistent with the conducting nature of the grafted polyphenyl layers on the electrode 31 . Although the reduction peak potential is located at À 0.05 V, the onset potential of the electroreduction of C 6 H 5 N 2 þ BF 4 À is around 0.1 V, which is positive to the Li 1 À x FePO 4 reference. It should be noted that the open circuit potential of the pristine LiFePO 4 electrode against Li þ /Li is in the range of 2.5-3.0 V, thus providing an even greater driving force for the reduction of C 6 H 5 N 2 þ BF 4 À . Recently, Madec et al. 37 have used nitrobenzenediazonium salts to functionalize pristine LiFePO 4 and limited reaction extent has been observed, which could be due to the insulating nature of the polymers formed from the nitrobenzenediazonium precursors.
Following the realization that LiFePO 4 can reduce C 6 H 5 N 2 þ BF 4 À to the C 6 H 5 Á radical, three reactions were conducted with molar ratios of LiFePO 4 :C 6 H 5 N 2 þ BF 4 À of 1:5, 1:1 and 1:0.05, respectively. The obtained products were subjected to structural and compositional analysis with the aim of understanding the extent and kinetics of the reaction. PXRD analysis (Fig. 3a) of the products after 12 h of reaction shows that the reaction, as expected, produces a new phase, that is, FePO 4 , and that the amount of the FePO 4 increases with that of the added C 6 H 5 N 2 þ BF 4 À . However, the excess C 6 H 5 N 2 þ BF 4 À cannot completely transform LiFePO 4 to FePO 4 , which means the reaction could be a self-limiting process. At molar ratios of LiFePO 4 :C 6 H 5 N 2 þ BF 4 À of 1:5 and 1:1, around 48 and 44% of the pristine LiFePO 4 were oxidized to FePO 4 , respectively, as determined by a Rietveld refinement procedure 38 (Supplementary Fig. 6 and Table 1 for details). These two products contain a high ratio of FePO 4 , which is not beneficial to the direct application as cathode material of Li-ion battery because of the severe Li þ ion loss. However, the reaction of LiFePO 4 and C 6 H 5 N 2 þ BF 4 À with molar ratio of 1:0.05 shows a very small amount of FePO 4 is formed (4.3% as measured indirectly by element analysis and online mass spectrometry, instead of Rietveld refinement due to the very weak PXRD signal of FePO 4 phase, see Supplementary Fig. 6 and Table 1 for the fitting results) and thus much less Li is extracted, as seen in Fig. 3a (blue curve). It is this reaction with less C 6 H 5 N 2 þ BF 4 À that we will focus on. First we conducted a quantitative online mass spectrometric investigation (the technical details may be found in refs 39,40) of the reaction of LiFePO 4 and C 6 H 5 N 2 þ BF 4 À . In a typical experiment, 0.162 g of LiFePO 4 was dispersed in acetonitrile in a reaction vial that was incorporated into the purging system of an online mass spectrometer. Before mixing with C 6 H 5 N 2 þ BF 4 À , only the signal of the Ar carrier gas was observed. On addition of C 6 H 5 N 2 þ BF 4 À solution (containing 10 mg of the diazonium salt as indicated by the arrow in Fig. 3b), the signal of N 2 increased abruptly and the reaction finished within 20 min. The evolved N 2 gas was quantified, according to a procedure published previously 39,40 , to be 1.03 ml (0.046 mmol), while the expected N 2 volume was 1.15 ml (0.051 mmol). This discrepancy may be due Electroreduction of 1 mM C 6 H 5 N 2 þ BF 4 À at a 2-mm diameter Au electrode in a three-compartment cell thermostated at 21°C. Supporting electrolyte is 0.1 M TBAClO 4 -acetonitrile and scan rate is 0.  ARTICLE to minor side reactions that do not release N 2 gas 41 . To verify the formation of the conducting polyphenyl polymers on LiFePO 4 , the solid product was analysed with FTIR. Figure 3c shows the FTIR spectra of the LiFePO 4 after reaction, and also the pristine LiFePO 4 (associated bands are marked with *) and C 6 H 5 N 2 þ BF 4 À for comparison purposes. The band in the 2,300-2,130 cm À 1 region corresponding to the stretching of the NN þ bond of the C 6 H 5 N 2 þ BF 4 À is not present in the solid product after reaction, which confirms the loss of N 2 during the reaction, consistent with the online mass spectrometric results. In addition, a new band at 1,240 cm À 1 (marked with þ in Fig. 3c, red curve) found in the spectrum of the solid product after reaction indicates the formation of FePO 4 (refs 42,43). The bands at 1,456 and 1,375 cm À 1 , corresponding to the stretching of C ¼ C bonds in aromatic rings, together with a stronger band at 686 cm À 1 (marked with # in Fig. 3c, red curve) associated with different types of aromatic substitutions 44 , indicate the formation of polyphenyl polymers. The different types of substitutions reflect the non-regiospecific attack of aryl radicals at those sites where the molecules are already attached to the surface, which has been found previously in the multilayers from diazonium reactions 31,36 . Figure 3d shows the TEM micrograph of the polyphenylenecoated LiFePO 4 , in which a coating with a thickness of 2-4 nm has been observed for isolated particles. The formation of the polymer coating has also been proved by a high-angle annular scanning transmission electron microscopy equipped with a energy-dispersive X-ray detector for the elemental mapping of a single LiFePO 4 particle after reaction (Supplementary Fig. 7). Although the PXRD data (Fig. 3a, blue curve) of the LiFePO 4 after reaction with diazonium salt show a very weak signal of FePO 4 , the co-existence of two phases of LiFePO 4 and FePO 4 within a single particle (Supplementary Fig. 8) has been observed by annular bright-field scanning TEM 45 . We also noticed that the extended exposure of the polymer under electron beam causes degradation. Similar phenomena have been observed previously for polyaniline-coated noble metal particles 46 . The weight percentage of polyphenylene in the composite was determined by elemental analysis to be 2.0% (wt%), which is consistent with the value measured by online mass spectrometry 2.1% (wt%). The latter is transformed from a 4.3% (mol%) of LiFePO 4 that is oxidized to FePO 4 , by taking into account that the reaction occurs according to LiFePO 4 þ PhN 2 þ -FePO 4 þ Ph þ Li þ þ N 2 and that the generated Ph forms polymers on LiFePO 4 surfaces. The formation of conducting polymer was further evidenced by the conductivity measurement of pressed powders, in which the polyphenylene-coated samples showed an electronic conductivity of 0.03 S cm À 1 , while pristine LiFePO 4 is less than 10 À 6 S cm À 1 (refs 7,8). Higher conductivity of polyphenylene has previously been reported by Shacklette et al. 47 , who demonstrated that the conductivity of polyphenylene can be increased to 50 S cm À 1 by doping with K þ , and even to 500 S cm À 1 by doping with AsF 5 À .
Electrochemical performance. The obtained polymer/LiFePO 4 composite had a tap density of 2.02 g cm À 3 that was higher than the values reported for nano-sized LiFePO 4 (typically in the range of 1.0-1.5 g cm À 3 ), and were mixed with polyvinylidene difluoride (PVDF) binder (9:1 wt/wt) and casted on an Al foil current collector to make a cathode (mass loading in the range of 2-3 mg cm À 2 ) for Li-ion cells. It is worth noting that the cathode does not contain any conducting carbon such as Super P, an additive that is extensively used in the practical cathode fabrication to ensure electronic conductivity throughout the electrode. When tested at a lower rate of 0.1 C, a capacity of 165 mAh g À 1 (the capacity is normalized to the expected mass of the active material of the cathode after complete lithiation) was achieved, as seen in Fig. 4a,b. This value is close to the theoretical capacity of LiFePO 4 (170 mAh g À 1 ), while very limited capacity is obtained for LiFePO 4 treated in the absence of C 6 H 5 N 2 þ BF 4 À ( Supplementary Fig. 9). When the carbon additive of Super P was used, improved electrochemical performance, relative to that of pristine LiFePO 4 , has been obtained ( Supplementary Fig. 10), which was, however, still inferior to the performance of the polyphenylene coating, particularly at higher rates. Electrochemical impedance spectroscopy has been used to probe the interfacial reaction kinetics, and smaller interfacial reaction resistance has been identified for polyphenylene-coated LiFePO 4 ( Supplementary Fig. 11), which could account for the improved performance at higher rates. The Li þ diffusion within the micron-sized LiFePO 4 has also been measured with the electrochemical impedance spectroscopy, and diffusion coefficients in the range of 10 À 16 -10 À 14 cm 2 s À 1 comparable to that of nano-LiFePO 4 have been obtained 48 ( Supplementary Fig. 11 and Table 2). The performance of polyphenylene-coated LiFePO 4 is significant since the prepared cathode contains no conducting carbon additive, and it shall be emphasized here that conducting carbon additive is not electrochemically active and therefore diminishes the practical energy density of the electrode 13 . The ability to replace carbon with a conducting polyphenylene coating is for these reasons highly advantageous. Cycling at a higher rate of 20 C for 1,000 cycles confirms the stability of the coating in the lithium-ion battery environment (Fig. 4c). Performances of the polyphenylene-coated LiFePO 4 at different temperatures have also been tested, and the results at 0°C demonstrated comparable performance at rates o8 C and somewhat degraded performance at rates 412 C ( Supplementary Fig. 12). The high rate capability of the micron-sized LiFePO 4 can be attributed to the formation of a metastable phase that can effectively decrease the energy battier of the nucleation and growth of a new phase, even the intrinsic Li þ bulk diffusion coefficient of LiFePO 4 is small 49,50 .

The interaction between LiFePO 4 /FePO 4 and phenyl radicals.
To address the issue of how polyphenylene is bonded to LiFePO 4 , we have performed DFT calculations using the phenyl radical as the probe molecule. As noted above, the initial reaction between LiFePO 4 and C 6 H 5 N 2 þ BF 4 À can partially delithiates the former leading to the formation of lithium vacancies predominantly present on the surface region 51  On FePO 4 (010), phenyl is most stable on the O 1 site (Fig. 5a) with a C-O bond length (d C-O ) of 1.420 Å, which is consistent with covalent C-O single bonds. The adsorption energy (DE ads ) is À 2.80 eV, indicating the bond to be chemical in nature. We have not located any previous report of the phenyl adsorption energy on metal oxides for comparison. DFT-calculated DE ads of phenyl on transition metal surfaces range widely from À 1.04 (on Au(111)) to À 2.81 (on Ti(0001)) eV 54 . Van der Waals forces may further stabilize phenyl adsorption 55 , but are not expected to change the site preference of phenyl on these surfaces. Phenyl adsorption on LiFePO 4 (010) with a Li vacancy is weaker than on FePO 4 (010), although phenyl still preferentially binds to an O site (Fig. 5b) Fig. 5c) In contrast, the adsorption of phenyl on Fe sites varies much less across the three model surfaces. Phenyl on the best Fe site (Fe 1 , Fig. 5a) has d C-Fe ¼ 2.239 Å and DE ads ¼ À 0.59 eV on FePO 4 (010); 2.113 Å and À 0.84 eV on LiFePO 4 (010) with a Li vacancy (Fe 1 , Fig. 5b), and 2.136 Å and À 0.94 eV, respectively, on stoichiometric LiFePO 4 (010) (Fe 1 , Fig. 5c).
Our DFT results suggest that the phenyl radical (and thus the polyphenylene coating) is most likely attached to the LiFePO 4 surface via O sites. Although the strength of the phenyl-O bond varies with the extent of surface lithiation, a strong chemical bond is expected under typical experimental and operational conditions as the surface is expected to be always deficient in Li to some extent. However, if the surface were to be completely lithiated, that is, by over-discharging a battery, the bonding of the coating to the surface could be disrupted because the bond strength would be weakened considerably and the preferred bonding site would shift from O to Fe.

Discussion
It has long been desirable to obtain high-performance LiFePO 4 particles with sizes approaching micrometres [24][25][26][27] . In this way, a high volumetric energy density of the Li-ion batteries resulted from a high tap density of micron-sized LiFePO 4 particles can be achieved, which is critically important for electric vehicle applications 1 . Practically, it has been difficult to coat large LiFePO 4 particles with uniform carbon coatings by calcination of carbon sources (usually organic small or macro molecules) and the preformed LiFePO 4 particles, possibly because of the inhomogeneous mixing of carbon sources and the large LiFePO 4 particles, which frequently results in phase-separated mixture of carbon and LiFePO 4 . On the other hand, simultaneous calcination of carbon sources and LiFePO 4 precursors often produces only nano-sized carbon-coated LiFePO 4 particles, because the formation of carbon phase inhibits the further growth of the LiFePO 4 phase, which has been proved in numerous literatures on the synthesis of nano-LiFePO 4 (refs 3,4). Furthermore, as pointed out by Zhou and co-workers 16 , carbon coatings on LiFePO 4 produced by heat treatment tend to be irregular and not well connected to the particles, particularly for large LiFePO 4 particles, and hence do not fully deliver the expected performance for battery applications. Actually, the difficulty encountered when coating large LiFePO 4 particles with a uniform carbon coating by calcination has motivated us to find alternatives to coat large LiFePO 4 particles with other conducting materials in the first place, as exemplified in this work.
To demonstrate the importance of the conductive coatings (either carbon or polymer coatings) on large LiFePO 4 , we have tested the electrochemical performance ( Supplementary Fig. 9) of bare LiFePO 4 by preparing cathode of binder þ LiFePO 4 (1:9 wt/wt). ARTICLE To compare the performances between polyphenylene-and carbon-coated LiFePO 4 , we prepared and tested the cathode of binder þ polyphenylene-coated LiFePO 4 (1:9 wt/wt), and the cathode of binder þ carbon additive þ LiFePO 4 (1:1:8 wt/wt) ( Fig. 4; Supplementary Fig. 10). These results highlight the importance of the high-quality conducting coatings on the electrochemical performance of large pristine LiFePO 4 particles, and that for improving the performance of micron-sized LiFePO 4 particles, polyphenylene clearly outperforms the typical conducting carbon.
On the basis of the electrochemical measurement results of pristine, carbon-coated and polymer-coated LiFePO 4 ( Fig. 4; Supplementary Figs 9 and 10), and particularly on the fact that large LiFePO 4 usually could not be easily coated with high-quality carbon coatings and therefore could only exhibit limited performance even in the presence of conducting carbon 24-27 , we attributed the improved electrochemical performance of polymer-coated large LiFePO 4 particles to the intimate bonding between LiFePO 4 and polyphenylene, a result of the surfaceinitialized electrografting of a diazonium salt, which is also supported by our DFT calculations. Another factor that might attribute to the improved performance of polymer-coated LiFePO 4 is that Li þ diffusion coefficient in polyphenylene (B1.32 Â 10 À 8 cm 2 s À 1 ) 56 is higher than in amorphous carbon coatings (B9.0 Â 10 À 11 m 2 s À 1 ) 57 . At higher rates, the Li þ diffusion within the coatings that separate the LiFePO 4 particles and Li þ containing electrolytes become crucially important, because better Li þ ion conducting coatings can ensure a rapid exchange of Li þ ions between the LiFePO 4 phase and the liquid Li þ electrolyte.
In summary, the energy capabilities of micron-sized LiFePO 4 have been unlocked in this work. Using the intrinsic reducing power of LiFePO 4 towards a diazonium salt of C 6 H 5 N 2 þ BF 4 À , a covalently bonded conducting polymer coating of polyphenylene can spontaneously form on pristine LiFePO 4 . The reaction mechanism has been studied in a detailed way by a range of complementary techniques coupled with theoretical calculations. More importantly, we have shown that the standard carbon coating generated by pyrolysis reaction can be substituted by the polymer coatings without emission of volatile organic compounds. Moreover, the polymer-coated LiFePO 4 can be used directly as 'carbon additive free' electrodes with desired electrochemical performance for rechargeable Li-ion batteries. Combined, the method reported here represents a potential replacement of the industrial standard of carbon coating LiFePO 4 with the spontaneous formation of conducting polymers from diazonium salt reactions.

Methods
Procedure. Carbon-free LiFePO 4 powder was prepared according to a published procedure 18 . In brief, the stoichiometric amount of precursors of FePO 4 (H 2 O) 2 and Li 2 CO 3 were thoroughly mixed together in isopropanol. After drying, the blend was heated at 700°C under reducing atmosphere. The obtained LiFePO 4 particle has an average size of 1.01 mm (Supplementary Table 3). C 6 H 5 N 2 þ BF 4 À was synthesized as follows 31 . A total of 0.01 mol of newly distilled aniline was dissolved in 50 ml solution of 0.1 M of HCl. After cooling the solution at 0°C with ice, a concentrated solution of NaNO 2 (0.015 mol) in water was added for 20-min reaction, then 0.012 mol NaBF 4 was added to precipitate the obtained diazonium cations. After filtration, the product was washed successively with cold water and ether. The powder was dried and kept in a freezer at À 18°C. The reactions of LiFePO 4 and C 6 H 5 N 2 þ BF 4 À with different molar ratios were conducted in acetonitrile. After reaction, the obtained products were subjected to a centrifugation also in acetonitrile. The supernatant after centrifuge was collected and examined by ultraviolet-visible to ensure the complete removal of the possible non-surface confined polymer and the starting material. The electrochemical properties of polyphenylene-LiFePO 4 were determined with CR2032-type coin cells using metallic lithium as the anode. The cathode was made by coating polyphenylene-LiFePO 4 and a solution of PVDF (Kynar 2801; 90:10 wt/wt) in N-methylpyrrolidone onto Al foil.
Characterization. Electroreduction of C 6 H 5 N 2 þ BF 4 À was conducted in an air-tight, three-compartment glass cell with valves to control the gas inlet and outlet. A polycrystalline Au disk electrode (diameter 2.0 mm, CHI Inc.) was used as the working electrode and was polished with 0.05 mm alumina slurry before use. A partially delithiated Li 1 À x FePO 4 (x ¼ 0.1) coated on stainless steel mesh (LiFePO 4 :Super P:PVDF 80:10:10 wt/wt) was used as the reference electrode. All electrochemical measurements were carried out using a Biologic VMP3 electrochemical workstation. The conductivity of the coated material was measured in a D41 À 11C/ZM four-probe resistivity tester. The chemical structure of C 6 H 5 N 2 þ BF 4 À was checked by 1 H NMR (Avance III 400, Bruker) and ESI-MS(Quattro Premier XE system, Waters). PXRD was carried out using a STOE STADI/P diffractometer operating in transmission mode with a primary beam monochromator and position-sensitive detector. Fe Ka1 radiation (l ¼ 1.936 Å) was employed. The details of online mass spectrometry were reported elsewhere 39,40 . The FTIR analysis was carried out with a Nicolet 6700 FTIR in transmission mode. Elemental analyses were performed using a Vario EL analyser. TEM images were recorded with JEOL JEM-2100F TEM operating at 200 kV. The experimental details for annular bright-field scanning TEM and high-angle annular scanning TEM for LiFePO 4 were reported elsewhere 45 .
Theoretical calculations. Periodic DFT calculations were performed using the Perdew-Burke-Ernzerhof functional and the projected augmented wave method as implemented in the Vienna Ab initio Simulation Package (version 5.3) 58 . The Kohn-Sham valence states (Fe(3d4s), Li(2s2p), O(2s2p), P(3s3p), C(2s2p) and H(1s)) were expanded in a plane wave basis up to a kinetic energy of 400 eV. As the electronic structures of both LiFePO 4 and FePO 4 are sensitive to electron correlation effects, the DFT þ U approach was used 59 . The optimized U-values and lattice parameters for the bulk LiFePO 4 and FePO 4 were taken from Zhou et al. 60 . The electronic energies of LiFePO 4 and FePO 4 depend strongly on the magnetic state of the Fe atoms 61 . In accord with previous investigations, we found DFT þ U to favour high-spin ferromagnetic and antiferromagnetic states over other spin orderings. As the binding energies of the phenyl radical on the LiFePO 4 (010) and FePO 4 (010) surfaces were observed to be reasonably insensitive to ferromagnetic/ antiferromagnetic ordering of the Fe ions, we report here results for high-spin ferromagnetic states, where the site projected atomic magnetic moments of Fe remain essentially identical to their computed bulk values (4.3 and 3.7 m B for FePO 4 and LiFePO 4 , respectively). Following relaxation of the bulk atomic positions, symmetric (010) slabs were constructed according to the procedure outlined in ref. 51. The dimensions of each slab correspond to 1a Â 2b Â 2c of the bulk lattice vectors (where a ¼ 10.42, b ¼ 6.07 and c ¼ 4.75 Å for LiFePO 4 and a ¼ 9.99, b ¼ 5.88 and c ¼ 4.87 Å for FePO 4 ), yielding 16 formula units per unit cell. Vacuum space (13 Å) was added along the b direction. The surface Brillouin zone was sampled with a 3 Â 1 Â 3 k-point mesh, which was confirmed to converge the total energies of both surfaces to within 3 meV per formula unit. For FePO 4 (010), the atomic positions of the top layer of Fe and P atoms along with all the O atoms coordinated to them (including subsurface O atoms) were relaxed, while holding the coordinates of the remaining atoms frozen in their bulk positions. For LiFePO 4 (010), the same layer of Fe, P and O atoms as in FePO 4 (010) was relaxed, as well as the top two layers of Li atoms. The vacancy model was created by removing one of the two top-layer Li atoms in the surface unit cell that we used for LiFePO 4 (010), with the surface relaxed again. The phenyl adsorbate was fully relaxed. All geometry relaxations were considered converged once the force in each relaxed degree of freedom fell below 0.03 eV Å À 1 . The adsorption energy of phenyl was calculated as DE ads ¼ E total À E surface À E phenyl . Including the semi-core states of Fe and Li in the valence (Fe(3s3p3d4s), Li(1s2s2p)) and increasing the kinetic cutoff energy to 500 eV were verified to changed adsorption energies by o0.1 eV on all three surfaces.