Electrochemical Properties of Li₃Fe_0.2Mn_0.8CO₃PO₄ as a Li-Ion Battery Cathode

Na₃Fe_0.2Mn_0.8CO₃PO₄ was synthesized via a hydrothermal synthesis method.³ Two clear solutions were made with FeSO₄·7H₂O (J. T. Baker) and MnN₂O₆·4H₂O (Sigma Aldrich), and Na₂CO₃ (Alfa Aesar, 99.5%) and (NH₄)₂HPO₄ (Alfa Aesar, 98%), respectively. The combined molar ratio of transition metal cations was 1:7 to carbonate ions and 1:1 to phosphate ions. The solutions were mixed in a glass bottle, heated in an oil bath to 130 °C and magnetically stirred for 44 hours, all in an Ar-flushed glove box. Next, the reaction products were separated from the ion-exchange solution and washed by centrifuging in the native solution, followed by centrifuging in methanol, distilled water, and methanol again. The wet powder was dried overnight in a vacuum oven at 50 °C.

Li₃Fe_0.2Mn_0.8CO₃PO₄ was made from its sodium-containing precursor via a Li-Na ion-exchange method. As-synthesized sodium carbonophosphate was added to 2M LiBr in 1-hexanol.¹⁰ The suspension was stirred magnetically for 5 days at 140 °C in an Ar atmosphere. The product was centrifuged to separate it from the exchange solution, and was then washed in methanol, water, and methanol again. The wet powder was dried in a vacuum oven overnight at 50 °C.

The ion-exchange product was then ball-milled with carbon in a Retsch PM200 planetary ball mill to improve the material's electrochemical performance through decreased particle size and carbon coating.^11,12 Zirconia-lined milling jars were loaded with Li₃Fe_0.2Mn_0.8CO₃PO₄ and carbon (Super P) in a weight percent ratio of 85:15 in an Ar-filled glove box. The sample was milled for 6 hours at 500 rpm, and the resultant powder was also unloaded in the Ar-filled glove box to limit air exposure.

Pristine powder samples of the sodium and lithium carbonophosphates were characterized by X-ray diffraction (XRD) using a Bruker D8 Advance diffractometer (Molybdenum Kα, λ = 0.7017 Å). Elemental composition of the lithium carbonophosphate material was verified by Inductively Coupled Plasma atomic emission spectroscopy (ICP) measurements using a Horiba Jobin Yvon, ICP-AES (ACTIVA-S). The sample powders were dissolved in 2% nitric acid, which itself was prepared from concentrated nitric acid and ASTM grade I deionized water. The diluted nitric acid was also used as a blank. Calibration curves were constructed using five calibration solutions of different concentrations for each measured element in 2% nitric acid. All measurements for each element were repeated three times and had a relative standard deviation of less than one percent.

To electrochemically characterize the Li₃Fe_0.2Mn_0.8CO₃PO₄ powder cathode films were made using the as-prepared sample, which was mixed with polyvinylidene fluoride (PVDF) and carbon (Super P) such that the overall ratio of components is in a 6:1:3 weight percent. The films were made using a doctor-blade method on aluminum foil, with each electrode containing ∼2 mg of active material. CR2016 coin cells were assembled in an Ar glove box with a Li metal foil anode, Celgard 2025 separator (Celgard Inc., U.S.A.), and 1M solution of LiPF₆ in ethylene carbonate/dimethyl carbonate (1:1) as electrolyte. Galvanostatic experiments were conducted at a rate of C/50 on an Arbin Instruments (College Station, TX) battery cycler.

The sodium-containing sample is tan in color, while the lithium-containing sample is light beige, which is consistent with previously reported results.³ The XRD patterns of synthesized phase-pure Na₃Fe_0.2Mn_0.8CO₃PO₄ and Li₃Fe_0.2Mn_0.8CO₃PO₄ samples are shown in Fig. 1. The pre- and post-ion exchange spectra exhibit substantial differences, but are consistent with differences between Li- and Na-containing carbonophosphate phases for other transition metals.³ Substantial peak broadening is seen in the Li₃Fe_0.2Mn_0.8CO₃PO₄ sample, which implies that either a substantial quantity of defects is present in the sample or that there are very small crystal domains in the ion-exchanged powder. Elemental analysis of the Li₃Fe_0.2Mn_0.8CO₃PO₄ sample was conducted using ICP analysis, the results of which are displayed in Table II. It is evident that the Li-Na ion exchange proceeded almost entirely (>99%) to completion. The other element ratios are also present in very close to the desired ratio, with the exception of iron, which is in excess. The most likely explanation is that some amorphous Fe-containing phase formed during the ion exchange reaction in 1-hexanol, as has been previously hypothesized.³

Zoom In Zoom Out Reset image size

Electrochemical testing of the Li₃Fe_0.2Mn_0.8CO₃PO₄ cathodes was conducted at a rate of C/50 at room temperature in a 2.00–4.75 V window. The voltage versus capacity data for one of these tests is shown in Figure 2a. As seen in the figure, the first-cycle charge starts at greater than three volts, and exhibits a charge capacity of 284 mAh/g, which exceeds the material's theoretical capacity of 231 mAh/g. The high starting voltage and the differing behavior in subsequent cycles indicate that the Fe²⁺ in the prepared sample was oxidized during processing. The high charge capacity on the 1^st cycle may stem from the irreversible Li⁺-extraction from an amorphous impurity phase left on the sample surface from synthesis, which is also indicated by the excess of iron seen in the ICP analysis of the ion-exchanged sample. The 1^st discharge shows a capacity of 147 mAh/g, but the capacity of the sample reduces over the course of the first 4 cycles to a reversible value of around 105 mAh/g, or 0.9 Li⁺ per formula unit. This decrease is not believed to be the result of traditional cathodic capacity fade, but instead the result of irreversible cycling of amorphous iron-containing impurity phases between 2.5 V and 2.0 V. These impurities are likely iron-containing hydroxides or oxyhydroxides, which have been previously been shown to intercalate Li⁺ at voltages in this region.¹³ As shown in Figure 2a, the difference in discharge capacity between the 1^st, 2^nd, and 18^th cycles stems primarily from decreased capacity in this low-voltage region.

Zoom In Zoom Out Reset image size

On charges following the first cycle, a distinct plateau is present near 3.0 V, which is the predicted voltage of the Fe²⁺/Fe³⁺ and near that of the Mn²⁺/Mn³⁺ couple.² The plateau approximately corresponds to the theoretical capacity of iron in Li₃Fe_0.2Mn_0.8CO₃PO₄ of 23 mAh/g. Additional plateaus, or other distinct features indicative of further redox activity are less apparent, and are discussed later.

The capacity retention is displayed in Figure 2b. Both the charge capacity and discharge capacity drop significantly over the first 4 cycles; from 284 mAh/g to 120 mAh/g and from 143 mAh/g to 108 mAh/g, respectively. After the fourth cycle the charge and discharge capacity remain relatively constant, and drop only ∼5 mAh/g over the next 20 cycles. In addition, significant coulombic inefficiency is observed, with the charge capacity consistently being about 10 mAh/g higher than the discharge capacity after the fourth cycle. This is likely the result of electrolyte decomposition at high voltage.

The biggest attraction of Li₃MnCO₃PO₄ is the potential existence of two redox couples in the voltage window of commercially used battery electrolytes. The proposed two-electron activity leads to a high theoretical capacity of 231 mAh/g. To improve the performance of the pure Li₃MnCO₃PO₄, Li₃Fe_0.2Mn_0.8CO₃PO₄ was synthesized, and its electrochemical performance is shown above. Of the 208 mAh/g theoretical capacity of the material, only about 105 mAh/g can be reversibly cycled for greater than 20 cycles. Despite performance well under theoretical values, this demonstrates a significant improvement over the previously reported performance of Li₃MnCO₃PO₄, which showed much poorer cyclability. In addition, Li₃Fe_0.2Mn_0.8CO₃PO₄ has a larger energy density than Li₃FeCO₃PO₄. While the reversible discharge capacity for these materials is about the same, the average discharge voltage in Li₃Fe_0.2Mn_0.8CO₃PO₄ is approximately 0.4 V higher, corresponding to an energy density 40 Wh/kg larger than in Li₃Fe_0.2Mn_0.8CO₃PO₄.

To examine the activity of Fe²⁺/Fe³⁺, Mn²⁺/Mn³⁺ and Mn³⁺/Mn⁴⁺ redox couples more in depth, which are all predicted to be accessible in the tested voltage window, the differential charge (dQ/dV) capacity is shown in Figure 3 for several cycles. Two charge peaks and two discharge peaks are clearly visible. The first set of peaks around 3.0 V most likely correspond to overlap of the Fe²⁺/Fe³⁺ and Mn²⁺/Mn³⁺ couples, computed to exist at 3.0 V and 3.2 V, respectively. It has been previously reported that Fe²⁺/Fe³⁺ and Mn²⁺/Mn³⁺ couples move closer together when mixed in single-phase Li(Fe, Mn)PO₄,¹⁴ and similar behavior could be present in this system. The second set of peaks around 4.0 V corresponds to the predicted Mn³⁺/Mn⁴⁺ activity in the active material. The slight shift seen in the second set of peaks over the first 12 cycles likely indicates that structural changes may be occurring in the material. In addition, the area of the peaks is noticeably different upon charge and discharge. Specifically, the area under the 3 V charging peak is smaller than the area of the 3 V discharge peak. Finally, the upturn at the high end of the charging curve on Figure 3 indicates that electrolyte oxidation may be occurring. This idea is reinforced since none of this observed charge capacity is returned on discharge, and explains the coulombic inefficiency seen in Figure 2b.

Zoom In Zoom Out Reset image size

The observed voltage of the Mn²⁺/Mn³⁺ redox couple shows one potential advantage carbonophospate cathodes hold over oxide cathodes. The inductive effect of the carbonate and phosphate polyanions causes the Mn²⁺/Mn³⁺ redox couple's voltage to be higher (3.2 V) than in oxide materials.¹⁵ However, the inductive effect in this mixed-polyanion system is not as large as that in pure phosphates, making the Mn redox couples lie in a useable voltage window, in contrast to LiMnPO₄. In order to improve Li₃Fe_0.2Mn_0.8CO₃PO₄, the capacity fading of the material, and its limited Li⁺ extraction must be addressed. In addition, the material's ionic and electronic conductivity would also need to be improved.

While this work demonstrates preliminary results of carbonophosphates as Li-ion cathode materials, a number of aspects of this materials performance remain to be understood. One notable unexplained cycling feature is the apparent asymmetric capacity seen in the Mn²⁺/Mn³⁺ redox couple on charging and discharging. In addition, the possibility that the increased discharge capacity in early cycles may result from an impurity phase in the material will need to be addressed in future works. Finally, the reason why the inclusion of iron in Li₃Fe_0.2Mn_0.8CO₃PO₄ significantly improves the material is not yet known.

In this work we reported an improvement over the performance of Li₃MnCO₃PO₄ by doping the compound with iron. We demonstrated the synthesis of Li₃Fe_0.2Mn_0.8CO₃PO₄ via hydrothermal synthesis of a sodium-containing precursor followed by a Na⁺-Li⁺ ion exchange reaction. The Fe doping allows to ion exchange to proceed to near (>99%) completion. Li₃Fe_0.2Mn_0.8CO₃PO₄ cathodes show a reversible capacity around 105 mAh/g with excellent cyclability over 25 cycles. The observed initial drop in capacity comes from irreversible cycling of impurities formed during synthesis. Three active redox couples, Mn²⁺/Mn³⁺, Mn³⁺/Mn⁴⁺, and Fe²⁺Fe³⁺ are active in this material. The results open new opportunities to improve the performance of the novel carbonophosphate cathode materials by doping and structural tuning. This is particularly meaningful because as recently reported, besides Fe and Mn, many alkaline earth and transition metals can also form carbonophosphate in pure or doped form.¹⁶

The authors acknowledge the Robert Bosch Company and Umicore for their financial support. This work also used the shared facilities in the Center for Materials Science and Engineering (CMSE) at MIT. We thank Geoffroy Hautier, Rahul Malik, Nancy Twu, and Jinhyuk Lee for their valuable assistance in this work.