Formation of Fe and Ni substituted LiMn_2–XM_XO₄ nanopowders and their crystal and electronic structure and magnetic properties

It is well established that LiMn₂O₄ undergoes a charge ordering transition accompanied by an orthorhombic distortion due to the J-T cooperative effect, close to room temperature [61]. The observed 3 × 3 × 1 superstructure is one of the factors limiting cycling performance of Li-ion batteries.

Substitution of manganese with nickel in spinel lithium manganates has been reported earlier in nanosized form for x = 0.5 [30, 42] as well as polycrystalline phases for x = 0.05, 0.07 and 0.5 [21, 51, 52] and it was sufficient to block the long range J-T transition in favor of the cubic spinel structure. At the same time, a study of the local environment using Raman spectroscopy [30] revealed line splitting characteristic of a lower symmetry space group (P4₃32), which could indicate structural distortion due to cation ordering, similar to the one observed in Mg doped compounds [52]. However, in the latter paper the authors did not find any signature of long range cation ordering in the neutron diffraction pattern of the LiMn_1.5Ni_0.5O₄ [52].

The lack of long range cation ordering in nanosized LiMn_1.5Ni_0.5O₄ was also corroborated in our earlier paper [42]. In the current study, the observation was extended to the whole substitution range x = 0.1 to 0.5.

In all cases, lattice parameters of the Ni doped samples were shorter than for the undoped LiMn₂O₄, although they differed on absolute scale (Table 1). It must be noted that the lattice parameters of LiMn₂O₄ depend on conditions of synthesis, which was reported earlier [62].

Table 1

Profile parameters obtained from the Rietveld refinement of the Ni doped series (CDS is Coherent Domain Size; estimated shape anisotropy is reported in curly brackets).

The decrease of the lattice parameter with increased level of nickel doping in the octahedral cation site might be at first surprising as the ionic radius of Ni²⁺ (0.69 Å) is larger than the radii of Mn³⁺ 0.58 Å in low spin (LS) and 0.65 Å in high spin (HS) forms [63].

The increase of the lattice parameter by a larger size of Ni²⁺ is compensated by the simultaneous conversion of the Mn³⁺ into Mn⁴⁺ which is expected to maintain the charge neutrality. The ionic radius of Mn⁴⁺ is the smallest of all of the radii and equal to 0.53 Å. Roughly estimating, on average and assuming that Mn³⁺ is only in HS form, substitution of x = 0.5 of Ni²⁺ should eliminate Mn³⁺ and produce a net decrease of lattice parameter of (0.69 – 0.65) – (0.65 – 0.53) = – 0.08 Å. The observed shortening of the lattice from 8.22 Å to 8.17 Å, i.e. only 0.05 Å might be explained by possible lithium overstoichiometry which was reported earlier by Strobel et al. [64]. The location of Li⁺ in the octahedral site would shift the charge balance for the undoped sample into the Mn⁴⁺ side, therefore the effect of doping would be lower. The large spread of lattice parameters caused by the nonstoichiometric lithium has already been shown [42] for LiMn₂O₄ synthesized by a sol-gel method, similar to the route used in this paper. However, as it has been already pointed out by Berg et al. [21], the decrease in lattice parameters might be also caused by stronger Ni–O bonding.

The lattice parameters in the Ni doped series decreases monotonically up to x = 0.3 where a second phase starts to appear, which indicates that the solid solution is close to the solubility limit. The second phase with reflections at d = 2.40 Å, 2.07 Å and 1.47 Å is visible up to x = 0.5 and cannot be indexed by lowering of the symmetry of the spinel phase. The limit is also visible in the lattice parameter plot (Fig. 2a), where the dependence flattens out for x = 0.3 and 0.4 proving that no more material is incorporated into the lattice. The last sample with Ni content x = 0.5 regains the trend but with a smaller slope which can be again explained by the appearance of the second phase. The restoration of the structural quality for x = 0.5 can be explained on the basis of the argument given already by Branford et al. [52]. The authors in question noted that the cation structural unit (Fig. 8 therein) was built of 4 cations which stabilized the configurations of 1:3 or 2:2 (1:1). In our case, the x = 0.5 sample is an example of 1:3 ratio (0.5 Ni:1.5 Mn).

The onset of phase separation is also reflected in the increasing lattice strains and growing size anisotropy (Table 6: rapid increase of the S₂₂₀ coefficient and K₄₁ parameter). In all cases, the sizes of the coherent scattering domains are under 100 nm. The refinement of the oxygen position was also performed, but the results reported in Table 6 and Table 7, because of its smaller reliability as oxygen atoms, are poorly seen by X-ray diffraction (Table 6) and depend on other elements in the unit cell. For the same reason, isotropic atomic displacement parameters (ADPs) are also reported in Table 6 and Table 7.

Table 2

Profile parameters obtained from the Rietveld refinement of the Fe doped series (CDS is Coherent Domain Size; estimated shape anisotropy is reported in curly brackets).

Table 3

Chemical compositions and magnetic properties of the Ni series.

Table 4

⁵⁷Fe Mössbauer hyperfine parameters of LiMn_1.5Fe_0.5O₄ where A is abundance, IS – isomer shift, QS – quadrupole splitting.

Table 5

Chemical compositions and magnetic properties of the Fe series.

Table 6

Structural parameters for the LiMn_2–xNi_xO₄ series.

Table 7

Structural parameters for the LiMn_2–xFe_xO₄ series.

All samples in the iron doped series were obtained as single phase cubic materials with the spinel space group Fd3m. Lattice parameters were found to increase with an increasing Fe concentration (Table 2) and for the x = 0.5 sample were similar to those reported earlier in the literature, like sample L16A [42].

The lattice parameter increase with higher iron content (Fig. 2b) is only possible if Fe³⁺ substitutes Mn⁴⁺ or Mn³⁺ but in a low spin state (LS). It is evident if one notices the ionic radii series (all values in Å) Mn⁴⁺ (0.53) < Fe³⁺ (LS 0.55) < Mn³⁺ (LS 0.58) < Fe³⁺ HS (0.645) < Mn³⁺ (HS 0.65) [63].

Despite the monotonic changes in lattice and anion parameters, the samples with the highest iron content stand out from the rest. It is reflected in several of their properties:

(a) the lattice parameters deviate from the trend set out by x = 0.1 to 0.3 (Fig. 2b),

(b) the coherent domain size as well as the respective strain are twice as much (Table 2 and Table 7).

The above facts suggest that there is a qualitative difference between samples x = 0.1 to 0.3 and x = 0.4, 0.5, although the XRD analysis does not indicate any presence of a phase transition. It might be just a mere consequence of a preference to form better crystallized domains or an effect of iron clustering.

The oxygen position for the iron doped series (Table 7) was refined using nominal iron content and octahedral position, which was corroborated by Mössbauer spectroscopy and is also reported in Table 6 and Table 7.

SEM images show the morphologies of grains in the examined compounds at magnification of 50000× (Fig. 3). The obtained powders can be considered as nanocrystalline materials because a large number of the grains are close to 100 nm. The exceptions are the samples doped with Ni and Fe with x = 0.4.

Ratios obtained using the data from each standard were averaged. Table 3 gives the final values of moles Mn/mol Li and moles Mn/mol Ni for each sample. Expanded uncertainties were determined by adding uncertainties from counting statistics, reproducibility of Mn (from the standard deviation of two values from each gamma-ray line/sqrt(2)), and the standard uncertainty: standard deviation of values obtained from the two standard pellets/sqrt(2), and multiplying by a coverage factor of 2. Mn:Ni ratios for all samples are in agreement with the values calculated from the nominal stoichiometry. Mn:Li ratios are lower than stoichiometric values for all samples, however this may be due in part to poor counting statistics for Li due to the low neutron capture cross section and small sample size.

The fitting parameters of the Mössbauer spectrum of LiMn_1.5Fe_0.5O₄ (Fig. 4) are listed in Table 4. The data displayed give clear evidence that Fe³⁺ ions occupy two different octahedral sites (D1, D2) that differ in values of QS and abundances. As the origin of the difference is not recognizable from Mössbauer measurement, we assume that the Fe³⁺ ions occupy two sites with slightly different symmetry because they have different QS values.

Fig. 5 shows the examples of the XPS spectra for the samples LiMn_2–xNi_xO₄ and LiMn_2–xFe_xO₄ in a wide energy range up to 1400 eV. Besides the lines of elements Li, Mn, Fe, Ni and O forming the compounds, there are extra lines visible from other impurities, coming from the technological process. The C1s lines are especially intense due to the presence of organic technological remains of precursors, as was observed in the literature [62]. The chemical formulas of the measured samples were estimated from the XPS analysis (Table 3 and Table 5). They were used for the calculation of the magnetic moments. Fig. 4 presents deconvolutions of the XPS Mn3p lines into two doublets corresponding to the Mn³⁺ and Mn⁴⁺ contributions and the Li line. In Table 3 and Table 5, the ratios of Li:Mn and Mn³⁺:Mn⁴⁺ are collected.

From XRD measurements the decrease of the lattice parameter with increasing Ni amount was observed. The radius of Ni²⁺ (0.69 Å) is larger than that of Mn³⁺LS (0.58 Å) and similar to Mn³⁺HS (0.65 Å). Therefore, a transformation of Mn³⁺ into Mn⁴⁺ (0.53 Å) would explain the decrease of the lattice parameter and the charge balance in the compounds. Such decreased ratios Mn³⁺/Mn⁴⁺ are shown in Fig. 6a and Table 3.

For the Fe series, the opposite behavior is observed: the lattice parameters weakly increase with increasing x and Mn³⁺ dominates (Fig. 6b and Table 5). The ionic radii form two groups. The first group contains Fe³⁺ (LS 0.55 Å), Mn⁴⁺ (0.53 Å), and Mn³⁺ (LS 0.58 Å), whereas Fe³⁺ (HS 0.645 Å) and Mn³⁺ (HS 0.65 Å) make the second one. It is less probable that Fe³⁺ replaces Mn⁴⁺ due to the charge and radius mismatch. The magnetic measurements indicate that for the samples with x = 0.1 and 0.2, Mn³⁺ LS is substituted by Fe³⁺ LS. For the samples with x = 0.3, Mn³⁺ HS is replaced by Fe³⁺ HS, while for x = 0.4 a mixture of the above states is present. Such separation can explain the nearly constant lattice parameter with increasing x.

The XPS O1s spectra, besides the line (1) related to the compounds (Fig. 7), exhibit some extra lines which can be ascribed to oxygen contamination, defects, etc. Extra lines at lower binding energy are also observed.

The XPS lines of the Fe2p and Ni2p, for the samples with x = 0.5, visible in Fig. 8a and Fig. 8b are related to the Fe³⁺ (in accordance to Mössbauer results) and Ni²⁺ states.

The XPS valence bands exhibit characteristic features of the Mn3d and O2p (Fig. 9). Both valence bands include a mixture of Ni3d or Fe3d states. These results are similar to the results concerning LMO [42, 62] and presented by Wu et al. [65]. The features A and B are related to Mn3d states. The peak A is due to the presence of e_g states of Mn³⁺ HS, while the peak B is attributed to the t_2g states of Mn³⁺ and Mn⁴⁺. The features C and D are ascribed to O2p. The features E and F may be due to some oxide contamination.

Measurements of the magnetic susceptibility χ showed magnetic irreversibility, zero-field-cooled and field-cooled for the samples containing Ni and Fe. Fig. 10 shows several examples of magnetic susceptibility runs. No signature of Jahn-Teller transition was observed in the substituted compounds measured in this work.

The applied magnetic field was 1 T. The Ni substituted samples start ordering below 150 K. The magnetic susceptibility measurements of the majority of the samples can be better fit according to the relation using the Neél theory: 1χ=1χ0+TC−σT−θ$$\begin{array}{l} {\displaystyle \frac{1}{\chi}=\frac{1}{\chi_{0}}+\frac{T}{C}-\frac{\sigma}{T-\theta}} \end{array}$$(1) where C is the Curie constant, θ is the paramagnetic Curie temperature, χ₀ and σ are parameters evaluated from the fit, and T is the temperature in Kelvin. This indicates uncompensated antiferromagnetic character. For the samples with x = 0.4 and 0.5 the inverse of the susceptibility was fitted according to the Curie-Weiss law: 1χ=1χ0+CT−θ$$\begin{array}{l} {\displaystyle \frac{1}{\chi}=\frac{1}{\chi_{0}+\frac{\mathit{C}}{T-\theta}}} \end{array}$$(2)

The effective magnetic moment depends on the amount of manganese and the ratio Mn³⁺:Mn⁴⁺ and Ni²⁺ amount (Table 3). Mn(3d⁴)³⁺ with HS (high spin) has a magnetic moment of 4.90 µ_B, while Mn(3d⁴)³⁺ with LS (low spin) and Mn(3d³)⁴⁺ have magnetic moments of 2.83 µ_B and 3.87 µ_B, respectively. According to the XPS results, the Ni magnetic contribution is Ni(3d⁸)²⁺ equal to 2.83 µ_B.

For the samples doped with Fe, measurements at magnetic field of 1 T allow fitting the curves for x = 0.1 – 0.4 using the Neél theory (Fig. 10). The calculations of the effective magnetic moments for the mixture of Mn³⁺ LS and HS and Fe³⁺ LS (1.75 µ_B) and HS (5.92 µ_B), gave good agreement with experimental values (Table 5).

The XPS valence band characters for the Ni and Fe substituted samples reflect magnetic states of the samples (Fig. 9). For the samples containing Fe for x from 0.1 to 0.3 the features A, B, C and D (described in Fig. 9b) are well separated. These samples measured at room temperature are in the paramagnetic state. The samples with x = 0.4 and 0.5 are in ordered state. They have the highest admixture of Fe3d states in the valence band. However, the features A to D are difficult to distinguish. Both above mentioned reasons can be responsible for the change of the shape of the valence band states.

For Ni containing samples, the shapes of the valence bands are similar, blurred (not visible Mn3de_g states (feature A). This may reflect not only the paramagnetic state but the admixture of the second phase, as well.

There are some theoretical papers claiming that existence of Mn³⁺ LS is possible under high pressure for cubic spinel structure [66]. However, in case of the examined samples the primitive cell volumes for LiMn₂O₄ (138.2 ų) and LiMn_1.5Ni_0.5O₄ (136.2 ų) are located in the region where the energies of the Mn³⁺ LS and HS are comparable.