II. EXPERIMENTAL

LiFe_1−x Mn _x PO₄/C was prepared by a solid-state reaction (Zhang et al., Reference Zhang, Wang, Liu and Huang2010, Reference Zhang, Wang, Liu and Huang2011). Stoichiometric amounts of Li₂CO₃ (Shanghai China Lithium, 99.9%), FeC₂O₄·2H₂O (Hefei Yalong, 99%), MnC₂O₄·2H₂O (Hefei Yalong, 99%), NH₄H₂PO₄ (Beijing Chemical, 99.5%), the appropriate quantity of citric acid (Beijing Chemical, 99.5%) and sugar were mixed by using high-energy ball milling using a zirconia container for 5 h. The mixture was then sintered at 650°C for 10 h under argon atmosphere (99.9999%). To confirm the chemical composition of these samples, the Fe/Mn ratios were determined using ICP (IRIS Intrepid II) after the complete dissolution of the powder into a hydrochloric acid solution. All samples are close to the target composition (Zhang et al., Reference Zhang, Wang, Liu and Huang2010, Reference Zhang, Wang, Liu and Huang2011). The structure was analyzed by an X-ray diffraction (XRD) (Rigaku Rint-2400) with CuKα radiation at a scan rate of 0.02° (2θ) s⁻¹.

Neutron diffraction was carried out on powder samples of LiFe_1−x Mn _x PO₄ (x = 0, 0.2, 0.5, 0.8, and 1.0) at room temperature on the diffractometer HRPD (Fischer et al., Reference Fischer, Frey, Koch, Könnecke, Pomjakushin, Schefer, Thut, Schlumpf, Bürge, Greuter, Bondt and Berruyer2000), installed at the Swiss Spallation Neutron Source SINQ of Paul Scherrer Institute (PSI), Switzerland. The incident neutrons of a wavelength λ = 1.154 Å were extracted from a Ge (822) monochromator with an effective intensity of 79%. Data were collected over the 2θ range of 4°–165° with a step increment of 0.05°, and analyzed by the Rietveld technique (Rietveld, Reference Rietveld1969) via the program FullProf [suite version July-2011] (Rodríguez-Carvajal, Reference Rodríguez-Carvajal1993, Reference Rodríguez-Carvajal1997). As the mean neutron scattering length of Mn (−0.373 fm) differs greatly from that of Fe (0.945 fm), the two elements mixing will cancel each other to weaken the intensity of neutron diffraction as shown in Figure 2, we collected the LiFe_1−x Mn _x PO₄ (x = 0.5 and 0.8) NPD data two times under the same condition.

The refinement included the five NPD data in the 2θ range of 10°–155°. The shape of the diffraction peaks were modeled by a pseudo-Voigt function and the background level was described by a 6-Coefficients polynomial function for LiFe_1−x Mn _x PO₄ (x = 0, 0.2, and 1.0) compounds. As the refinement of the LiFe_1−x Mn _x PO₄ (x = 0.5 and 0.8) data was not able to reach a convergent result with the same background level, we selected about 80 background points via user-selected by the program WinPLOTR to refine it subsequently, which proved to be very effective. The angles below 100° in 2θ (°) of LiFe_1−x Mn _x PO₄ (x = 0, 0.2, and 1.0) compounds and 80° of LiFe_1−x Mn _x PO₄ (x = 0.5 and 0.8) were, respectively, made for peak asymmetry corrections. The atomic positional parameters were refined for all atoms except Li. The isotropic displacement parameters were refined for all atoms, and finally, the occupancy parameters were refined for Li and Fe/Mn, respectively.

In order to test the possibility a mixture of site occupation of Li/Fe atoms proposed by LiFePO₄ NPD refined results, an XRD pattern of LiFePO₄ was collected on a Rint-2400 diffractometer (Rigaku Corporation) with CuKα radiation. The diffraction intensity was measured from 10 to 86° using a step interval of 0.02°.

In the XRD data refinement, we first deal with the scale factor, the diffractometer zero point, the background function, and the unit-cell parameters, and then with the profile function, the atomic coordinates, and the isotropic displacement parameters, and finally, with the Li and Fe/Mn atomic occupancies deviated from nominal chemical stoichiometry. There are no excluded regions in the XRD data refinement and the limit angle for asymmetry correction is 38° for the XRD and 73° for the NPD in the Table I refinement process. The results are listed in Table I.

Table I. The occupancy results obtained from the Rietveld refinement of the LiFePO₄ sample.

Refined parameters are in italics.

III. RESULTS AND DISCUSSION

Our previous investigation suggested that the kinetic property of the Mn²⁺/Mn³⁺ redox couple is improved with forming a solid solution and a mesoporous structure and that the reversible capacity and rate performance decrease with the increase of Mn content (Zhang et al., Reference Zhang, Wang, Liu and Huang2010, Reference Zhang, Wang, Liu and Huang2011).

LiFePO₄ NPD data were used to solve the structure by the program FOX (Favre-Nicolin and Černý, Reference Favre-Nicolin and Černý2002) initially. Ab initio crystal structure solution via the program FOX indicates demonstrably that the space group of LiFePO₄ is Pnma. Thus, the olive structure of LiFePO₄ from Inorganic Crystal Structure Database (ICSD) was employed as the initial structure model in the following Rietveld refinement, with Li¹⁺ on the 4a site and Fe²⁺ on the 4c site.

We focused on the occupancy of Li and Fe sites in LiFePO₄, respectively, to test the conclusion of the mixed Li/Fe occupation in a small amount in previous works (Chung et al., Reference Chung, Choi, Yamamoto and Ikuhara2008, Reference Chung, Choi, Lee and Ikuhara2012; Gardiner and Islam, Reference Gardiner and Islam2010; Hoang and Johannes, Reference Hoang and Johannes2011). First, all the occupancies were fixed to 100%, perfect chemical ordering, and the agreement between the computed and observed patterns is already good with R _B = 2.66% and R _wp = 8.65% (χ ² = 1.65). Then, we released to refine all the occupancies of Li and Fe sites, and found that the Li (4a) and Fe (4c) occupancy became less than full occupation, and the R-factors improved quite slightly to R _B = 2.66% and R _wp = 8.54% (χ ² = 1.61). Finally, we refined the mixture rate of Li/Fe atoms with a constraint of full occupancy in total, and the R-factors scarcely changed to R _B = 2.63% and R _wp = 8.55% (χ ² = 1.61). If the mixture of Li/Fe is true, it can be speculated that the similar result should be concluded from XRD since the X-ray atomic form factors are significantly distinguished between Li and Fe. We used the XRD data calculate as the same strategy for above. We also mixed X-rays and neutrons data for a combined refinement. The results are provided in Table I. The mean neutron-scattering length of Li (−0.190 fm) differs greatly from that of Fe (0.945 fm), so that some Fe²⁺ distributed over Li¹⁺ sites will reduce the average neutron-scattering length in the Li¹⁺ sites, which will decrease the result of Li¹⁺ occupancy. On the contrary, the X-ray atomic form factors increase with the addition of atomic number, so that some Fe²⁺ distributed over Li¹⁺ sites will increase the average atomic form factor value of XRD, which will raise the Li¹⁺ occupancy and will make the occupancy value larger than 1.0 as in Table I. It seems to imply that about 2% mixed Li/Fe occupation (antisite defects) occurs, whereas the statistical R-factors just have a tiny improvement. The antisite defects could decrease the ionic conductivity of the cathode material and increase the polarization, because Li¹⁺ ion diffusion occurs preferentially via one-dimensional channels oriented along the [010] direction (b-axis) (Morgan et al., Reference Morgan, Van der Ven and Ceder2004; Islam et al., Reference Islam, Driscoll, Fisher and Slater2005; Nishimura et al., Reference Nishimura, Kobayama, Ohoyama, Kanno, Yashima and Yamada2008); these antisite defects would add additional electrostatic repulsion and impede the diffusion easily. Chung et al. (Reference Chung, Choi, Yamamoto and Ikuhara2008, Reference Chung, Choi, Lee and Ikuhara2012) have directly demonstrated the disordered occupations by Fe atoms on Li sites in LiFePO₄ with high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM). An intrinsic defect type with the lowest energy is the cation antisite defect, in which Li and Fe/Mn ions exchange positions (Gardiner and Islam, Reference Gardiner and Islam2010). The antisite exchange defects have been demonstrated theoretically for LiMnPO₄ and LiFePO₄ by M. Saiful Islam and Michelle Johannes et al. (Gardiner and Islam, Reference Gardiner and Islam2010; Hoang and Johannes, Reference Hoang and Johannes2011), respectively. The observed and calculated NPD and XRD patterns of LiFePO₄ are shown in Figure 3, from which it can be found that the profile calculated is statistically in agreement with the observed data.

Figure 3. (Color online) Rietveld refinement on the XRD pattern (a) and NPD pattern (b) of the LiFePO₄ sample at room temperature. Observed intensity Y _obs and calculated intensity Y _calc are represented by red dot signs and the black solid line. The green bar represents the Bragg peaks position and the blue curve at the bottom represents the residual difference Y _obs − Y _calc.

It was reasonably assumed that the Mn²⁺ replacement takes place at the Fe-sites (Yao et al., Reference Yao, Bewlay, Konstantionv, Drozd, Liu, Wang, Liu and Wang2006; Molenda et al., Reference Molenda, Ojczyk and Marzec2007; Hong et al., Reference Hong, Wang, Wang and Graetz2011) in Rietveld refinement of LiFe_1−x Mn _x PO₄ (x = 0, 0.2, 0.5, 0.8, and 1.0) compounds. Table II gives the structural parameters obtained from the Rietveld refinement of the structures; some structural data obtained from the Rietveld refinement are also provided in Table III. No impurity phase was identified, and the crystal structure was successfully refined with the space group Pnma. The cell parameters a, b, and c increase linearly as the Mn content increases, inducing a linear increase in the unit-cell volume (Table III, Figure 4). This is supported by the XRD results of Padhi et al. (Reference Padhi, Nanjundaswamy and Goodenough1997a). This result is reasonable since the Mn²⁺ (r = 0.80 Å) has a larger radius than that of the Fe²⁺ (r = 0.74 Å).

Figure 4. Structural data of LiFe_1−x Mn _x PO₄ (x = 0, 0.2, 0.5, 0.8, and 1.0) compounds. The lattice parameters of olivine phase for LiFe_1−x Mn _x PO₄ (x = 0, 0.2, 0.5, 0.8, and 1.0) compounds increase linearly as a function of Mn content (x) (a) and the interatomic distances (in Å) of Li–O(2) and Li–O(1) increase, while the interatomic distances (in Å) of Li–O(3) decrease on the addition of Mn (b).

Table II. Structural parameters refined using the NPD data for the LiFe_{1 − x} Mn _x PO₄ (x = 0, 0.2, 0.5, 0.8, and 1.0) compounds.

For each atom the rows are, in order (from the top): x = 0; x = 0.2; x = 0.5; x = 0.8; and x = 1.0, respectively.

Table III. Structural data of LiFe_{1 − x} Mn _x PO₄ (x = 0, 0.2, 0.5, 0.8, and 1.0) compounds from the Rietveld analysis.

The selected interatomic distances and angles for all LiFe_1−x Mn _x PO₄ compounds are, respectively, described in Table III. The interatomic distances (in Å) of Li–O(2) and Li–O(1) increase while the interatomic distances (in Å) of Li–O(3) decrease on the addition of Mn (Figure 4), respectively, which partially explains a higher potential plateau of ~4.1 eV in LiMnPO₄ than ~3.5 eV in LiFePO₄. An XRD study of LiFe_1−x Mn _x PO₄ interatomic distance was not able to demonstrate a linear relationship described by Yao et al. (Reference Yao, Bewlay, Konstantionv, Drozd, Liu, Wang, Liu and Wang2006), because the X-ray is scattered by the extranuclear electron density, but the neutrons are by the nuclei. The Li¹⁺ ion diffusion occurs preferentially via one-dimensional channels oriented along the [010] direction (b-axis) (Morgan et al., Reference Morgan, Van der Ven and Ceder2004; Islam et al., Reference Islam, Driscoll, Fisher and Slater2005; Nishimura et al., Reference Nishimura, Kobayama, Ohoyama, Kanno, Yashima and Yamada2008); so the interatomic distances (in Å) of Li–O(3) decrease while the cell parameters increase, which will make the Li¹⁺ ion diffusion become difficult. As a result, LiMnPO₄ has a higher potential plateau than LiFePO₄, consistent with our previous investigated electrochemical behaviors (Zhang et al., Reference Zhang, Wang, Liu and Huang2010, Reference Zhang, Wang, Liu and Huang2011).

IV. SUMMARY

Five different phase-pure compositions of LiFe_1−x Mn _x PO₄ (x = 0, 0.2, 0.5, 0.8, and 1.0) have been synthesized and studied by NPD. Rietveld refinements (FullProf program suite version July-2011) for the NPD and XRD profiles measured for LiFePO₄, respectively, show that about 2% Fe²⁺ may distribute over Li¹⁺ sites, which could decrease the ionic conductivity of the cathode material and increase the polarization. The a, b, and c cell parameters increase linearly, and the interatomic distances (in Å) of Li–O(2) and Li–O(1) increase, while the interatomic distances (in Å) of Li–O(3) decrease with the addition of Mn, respectively. The interatomic distances (in Å) of Li–O(3) decrease will also make the Li¹⁺ ion diffusion more difficult.