Unravelling redox processes of Li7MnN4 upon electrochemical Li extraction–insertion using operando XAS

D. Muller-Bouvet a, N. Emery *a, N. Tassali a, E. Panabière a, S. Bach ab, O. Crosnier cd, T. Brousse cd, C. Cénac-Morthe e, A. Michalowicz *f and J. P. Pereira-Ramos a
aInstitut de Chimie et des Matériaux Paris Est, GESMAT, UMR CNRS UPEC 7182, 2 rue Henri Dunant, 94320 Thiais, France. E-mail: emery@icmpe.cnrs.fr
bUniversité d’Evry Val d’Essonne, Bd F. Mitterrand, Département Chimie, 91025 Evry Cedex, France
cIMN, UMR6502 La Chantrerie, rue Christian Pauc, 44306 Nantes, France
dRS2E, FR3459 CNRS, France
eCNES, 118 avenue Edouard Belin, 31401 Toulouse Cedex 9, France
fInstitut de Chimie et des Matériaux Paris Est, UMR CNRS UPEC 7182, 2 rue Henri Dunant, 94320 Thiais, France. E-mail: michalow@wanadoo.fr

Received 1st August 2017 , Accepted 18th September 2017

First published on 18th September 2017


A large data set of XAS (X-ray Absorption Spectroscopy) Manganese K-edge spectra has been collected operando and studied upon the electrochemical oxidation of the promising Li-ion battery anode material Li7MnN4. Using chemometric tools such as PCA (Principal Component Analysis) and MCR-ALS (Multivariate Curve resolution – Alternating Least Squares), three independent environment spectra were insulated. Based on the faradaic yield and well-chosen comparison of absorption spectrum energies within the frame of the coordination charge model, these environments were ascribed to unusual oxidation states allowed by nitride chemistry at a low potential (∼1.2 V vs. Li+/Li), i.e. Mn5+ (3d2), Mn6+ (3d1) and Mn7+ (3d0). Also, their relative amounts are discussed with regard to the long-range structural variation which can be simply described by two successive biphasic domains followed by a solid-solution behaviour. Gathering this long-range and local structure information provides a complete picture of the redox mechanisms occurring in Li7MnN4.


Introduction

Lithium batteries are used in a wide range of applications, from small scale on-board electronics to electric vehicles. However, improvements are necessary to fulfil the requirements for high capacity and/or power applications. Carbon graphite, which is commonly used as a negative electrode, does not support high current densities. Various alternatives are still studied, for example silicon,1 tin-based materials,2,3 phosphides4,5 or transition metal oxides.6 Due to the lower ionicity of the transition metal–nitrogen bond with regard to metal–oxygen bonds, lower operating voltages are usually encountered in lithiated transition metal nitrides compared to their homologous oxides.7 Various phases were evaluated8,9 and among them, Li7MnN4 appears to be the most promising. Li7MnN4, which adopts a derivative anti-fluorite type structure,10–12 displayed an interesting specific capacity, up to 300 mA h g−1 around 1.2 V vs. Li+/Li at a C rate (i.e. 1 F mol−1 in 1 hour), with an excellent cycle life.7,13,14 In addition, using an appropriate ball-milling step to reduce the overall particle size distribution, a capacity retention of 120 mA h g−1 is stabilized at a 5C rate.14,15 These interesting performances can be explained by the excellent reversibility and the limited volume variation induced by redox processes observed through operando X-ray diffraction.16 A reversible three-phase mechanism was evidenced in good agreement with the electrochemical behaviour. Indeed, two successive biphasic domains with two abrupt unit cell variations were observed during the oxidation, according to the two voltage plateaus of the galvanostatic curve. At the end of the charge process, one single phase was observed with a solid-solution like behaviour characterized by a continuous variation of the unique unit cell parameter. XRD gives an indication of possible mechanical strain induced by redox processes involved. However, it does not give information on a possible amorphous contribution to the oxidation state variation of the redox centre, i.e. manganese ions, implied in the reaction. Actually, from the electrochemical point of view, the lithium extraction process leads to a maximum faradaic yield of 1.7 F mol−1, which suggests important variations in the manganese oxidation state. Moreover, manganese ions in Li7MnN4 are in the +5 (3d2) formal oxidation state and higher degrees, up to +7 (3d0), might be involved in such a reversible electrochemical reaction. To shed light on this issue, we performed operando X-ray Absorption Spectroscopy (XAS) experiments at the K-edge of manganese during a full oxidation process. Indeed, operando XAS allows information to be collected on the local environment and the oxidation state of a chosen absorbent without possible pollution induced by ex situ sample preparation.17 In addition, the recording of a continuum of spectra during the oxidation process allows a better description of the occurring redox mechanisms. In this study we use linear chemometric methods, namely Principal Component Analysis (PCA) and Multivariate Curve resolution-Alternating Least Squares (MCR-ALS), to quantitatively describe the evolution of redox species concentrations. These powerful chemometric methods allow the spectra of pure components involved in our time resolved data set to be extracted and their concentration profile evolutions to be revealed.18 In addition to this fine elucidation of the manganese redox behaviour in Li7MnN4 upon electrochemical oxidation, the reversibility of the oxidation process is also demonstrated.

Experimental

Synthesis

A solid-state synthesis method under continuous nitrogen flow has been used to prepare Li7MnN4. A mixture of manganese powder (Alfa-Aesar 99+%, 325 mesh) and lithium nitride Li3N (Alfa-Aesar 99.4%, 60 mesh) has been pressed into pellets in an argon-filled glove box. 15% excess of Li3N is added to compensate the lithium loss during the thermal treatment. Then, the sample was heated up to 750 °C for 10 h under constant nitrogen flow using a stainless steel reactor especially designed to avoid any air exposure during transfers between the glove box and the furnace.

Characterization

The as-prepared powder was analysed by X-ray diffraction using a Bruker D8 advanced diffractometer equipped with a Cu Kα radiation source and a LynxEye detector. An air-tight polymer sample holder was employed to avoid any degradation due to ambient air exposure during the acquisition.

The electrochemical properties were checked using a VSP150 potentiostat (BioLogic) in a CR2032 coin cell. The working electrode was composed of 70 wt% Li7MnN4 powder, 22 wt% acetylene black as a conductive agent, and 8 wt% PTFE as a binder. This paste was pressed on a copper grid (Goodfellow). A lithium disk acts as a counter and reference electrode in such a 2-electrode configuration. The separator is composed of three sheets of Whatman paper (silica fibre) soaked in LP71 electrolyte (Ethylene Carbonate EC[thin space (1/6-em)]:[thin space (1/6-em)]Diethyl Carbonate DEC[thin space (1/6-em)]:[thin space (1/6-em)]Dimethyl Carbonate DMC 1[thin space (1/6-em)]:[thin space (1/6-em)]1[thin space (1/6-em)]:[thin space (1/6-em)]1 in volume, 1 mol L−1 LiPF6, BASF).

Operando X-ray absorption spectroscopy

X-ray absorption spectra at the manganese K-edge were collected at the BM30B/CRG-FAME beamline of the European Synchrotron Radiation Facility (ESRF, Grenoble, France19) operating at 6 GeV and 200 mA. The X-ray absorption spectra were recorded at the manganese K-edge in transmission mode. A Si(220) double-crystal monochromator was used with about 0.35 eV resolution at 6.5 keV. Due to the energy resolution being better than 0.75 eV, i.e., close to the 10−4 intrinsic resolution of the monochromator, it is possible to discern the characteristic peaks in the pre-edge part of the manganese spectra. Metallic Mn foil was used to calibrate the monochromator. The absorption spectra were measured from 6450 to 6588 eV, with a step of 0.3 eV in the XANES region.

The XAS spectra were collected operando using a modified Swagelok-type electrochemical cell with a beryllium window allowing transmission X-ray spectroscopy measurements.20 In this electrochemical cell, the supporting copper grid is arranged on one of the Be disks of the cell which acts as a current collector and as X-ray window. The battery was charged at a slow rate of C/20 (i.e. 1 F mol−1 in 20 h) and discharged at a C/10 rate. The electrochemical curve is displayed in the ESI1. A series of spectra were collected during the electrochemical oxidation with an acquisition time for one spectrum fixed to 11 min. Then, the composition variation occurring during the acquisition of one spectrum approximately corresponds to Δx = 0.011 in Li7−xMnN4.

Data processing of the XANES spectra, including energy calibration and interpolation, pre-edge background subtraction, and post-edge normalization, was performed using the MAX-Cherokee software.21 The oxidation process was analysed using Linear Chemometry methods, including the determination of the number of independent spectra using Principal Component Analysis (PCA22,23), spectral evaluation of pure species present in the time series mixture using the Multivariate Curve Resolution-Alternating Least Squares (MCR-ALS24,25) method and the Linear Combination Least Squares fit26 using Singular Values Decomposition.27 All these linear and statistical analysis methods are specifically implemented for XAFS data analysis in the present version of MAX-StraighnoChaser.21

Results and discussion

Characterization of the pristine powder

The X-ray diffraction pattern of the as prepared powder is displayed in Fig. 1. According to ref. 10,11 and 16 a cubic structure belonging to the space group P[4 with combining macron]3n with a cell parameter of 9.558(1) Å allows us to properly fit the XRD pattern (χ2 = 1.35, Rwp = 2.21%, Rp = 1.68%, the amorphous dome centred around 21° is ascribed to the air-tight sample holder and is included in the background fit). The structure, present in the inset of Fig. 1, is composed of an edge-sharing tetrahedron filled by either manganese or lithium ions. In such a derivative anti-fluorite type structure, 1 cationic site over 8 is filled by a manganese ion with a peculiar order leading to a 4 times larger cell than a typical anti-CaF2 cell. Differences in the electrostatic potential and ionic radii between Li+ and Mn5+ may explain the distortions from the original scheme.
image file: c7cp05207c-f1.tif
Fig. 1 XRD pattern with the associate Rietveld refinement of the as-prepared powder of Li7MnN4. Inset, representation of the cubic structure (SG P[4 with combining macron]3n).

A typical galvanostatic cycle recorded at C/10 is presented in Fig. 2. The oxidation curve consists of two successive voltage plateaus at 1.18 and 1.20 V vs. Li+/Li, involving respectively ∼0.9 and ∼0.6 F mol−1, followed by a continuous and sharp increase of potential up to 1.6 V for an additional 0.3 F mol−1. The final composition Li5.3MnN4 is then obtained. In agreement with previous reports,13–16 this electrochemical fingerprint is consistent with redox reaction (R1).

 
Li7MnN4 ⇆ Li7−xMnN4 + xe + xLi+(R1)


image file: c7cp05207c-f2.tif
Fig. 2 Galvanostatic curve of Li7MnN4 recorded at a C/10 rate.

In addition, the excellent reversibility of the Li-extraction process is highlighted in Fig. 2 since the subsequent discharge curve gives rise to a symmetric potential–composition curve, with the same faradaic yield. This reversibility is well supported by the structural behaviour reported in ref. 16.

Evolution of the XANES spectra recorded during the oxidation

Prior to the chemometric study, the evolution of the XANES spectra recorded at the K-edge of manganese reported in Fig. 3 was qualitatively analysed. The black curve represents the initial Li7MnN4 compound and the plot corresponding to the end of the oxidation (for x = 1.7) is in blue. All the intermediate spectra are displayed in grey except the intermediate composition Li6.1MnN4, i.e. x = 0.9, which is plotted in red. The peculiar composition of Li6.1MnN4 corresponds to the transition between the two biphasic domains evidenced from the electrochemical curve (Fig. 2).

First, between 6538 and 6545 eV, all the recorded spectra display a pre-edge peak, which is characteristic of the tetrahedral environment of the absorbent species (Fig. 3). It is due to a 1s → 3d + 4s electronic transition, allowed in tetrahedral symmetry, where hybridization of 3d and 4s atomic orbitals can occur. Its amplitude is related to the number of empty final d states.28 The augmentation of the observed pre-edge intensity supports the oxidation of manganese ions upon electrochemical lithium extraction.


image file: c7cp05207c-f3.tif
Fig. 3 Evolution of Li7MnN4 Mn K-edge XANES spectra upon electrochemical oxidation. Inset, evolution of the pre-edge energy region of XANES spectra.

Second, the electrochemical oxidation of Li7MnN4, i.e. from black to blue spectra, induces an overall shift to higher energies. The pre-edge peak shifts from 6541.1 eV to 6542.1 eV and from 6542.1 eV to 6542.7 eV, after the extraction of respectively 0.9 and 1.7 Li per unit formula (Table 1 and the inset of Fig. 3). Similarly, the threshold shifts from 6550.8 eV to 6552.6 eV and finally to 6554.6 eV. It is known that the shift of the edge towards higher binding energies indicates an increase in the average valence state of the absorbing ion.29,30

Table 1 Formal oxidation state, coordination charge, pre-edge and threshold X-ray absorption energies of Li7MnN4, Li6.1MnN4, Li5.2MnN4, LiMn2O4, BaMnO4 and KMnO4, and the three calculated environments
Compound Pre-edge peak energy (eV) Threshold energy (eV) Mn average ox. state Coordination charge η
Li7MnN4 6541.1 6550.8 +5 2.67
Li6.1MnN4 6542.1 6552.3 +5.9 3.57
Li5.3MnN4 6542.7 6554.6 +6.7 4.37
LiMn2O4 6548.9 +3.5 1.07
BaMnO4 6542.6 6555.5 +6 4.38
KMnO4 6543.5 6557.3 +7 5.38
Env1 6541.1 6550.9 +5 2.67
Env2 6541.9 6552.0 +6 3.67
Env3 6542.8 6555.2 +7 4.67


Chemometric analysis of the spectra

In the preceding section, we have observed qualitatively the oxidation of the transition metal ion centre affected by the lithium extraction. Due to the high faradaic yield, up to 1.7 F mol−1, at least 3 different oxidation states of manganese must be involved in explaining the redox processes. As described in the previous section, this is reflected in the XANES spectra by an overall shift to higher energies. To determine the exact number of components, i.e. of manganese environments including their oxidation state required to describe our data set and their concentration profile along the oxidation reaction, we employed several chemometric methods: Principal Component Analysis (PCA22,23), Multivariate Curves Resolution-Alternating Least Squares (MCR-ALS24,25), and Linear Least Squares Combination Analysis (LCA,26) (or Linear Least Squares fit) using Singular Values Decomposition of the data matrix.27

The first step, PCA, is of utmost importance. It allows us to determine the number of different environments for the manganese ions. It is worth noticing that, even in a single crystalline phase, 2 different oxidation states and/or 2 different local environments can be encountered. Conversely, two different crystalline phases can involve identical local ion environments. This is why the number of local environments studied using a local probe like XAFS is not necessarily equal to the number of crystalline phases determined using XRD. The number of components that should be retained for analysis is estimated on the basis of the Principal Component Analysis (PCA) results.22,23,31 As shown in the Scree plot (Fig. 4a), singular values fall drastically down to 1.5 for the third component. For the following components, the singular values are almost constant at a minimal value (<1) corresponding to the noise of the data. Thus, the scree plot suggests that 3 components only can determine the variance in the data. This is clearly confirmed by the Scores plots (Fig. 4b), which represent the concentration profile of the data matrix Linear Combination Least Squares Analysis based on the abstract orthogonal PCA components.22 The component 1 profile is constant simply because it is representative of the spectral average. PCA component scores 2 and 3 reflect the smooth chemical evolution of the data. Indeed, the slope change in the concentration profile of the abstract PCA component 3 can be related to the change in the real components’ phase concentrations around x = 0.9. Then, adding a forth component only leads to a randomly disperse plot representative of noise.


image file: c7cp05207c-f4.tif
Fig. 4 (a) Scree plot: number of the singular values versus the component index. (b) Scores plot: the concentration profile of the first four PCA components.

In standard Linear Combination Analysis the spectra of the reaction components are known and the determination of their concentration profile by least squares fit is straightforward. It is not the case in our experiment: the initial state, Li7MnN4, is well known, but neither the intermediate nor the final state can be assumed to be pure. We use Multivariate Curve resolution-Alternating Least Squares (MCR-ALS) to decompose the mixture evolution of a priori unknown “pure” manganese environments and to determine their concentration profile upon electrochemical oxidation of Li7MnN4. Based on the previous PCA analysis, which limits the number of local environments to 3, and by introducing selected constraints and initial guesses in the model (more details are given in the ESI2), a realistic convergence of the ALS process has been obtained.

MCR-ALS leads to the concentration profiles as a function of the electrochemically induced lithium depletion in Li7MnN4 (Fig. 5) of the three independent spectra (Fig. 6, solid line) necessary to describe the full data set (examples of the reconstructed spectrum are given in the ESI3). From the concentration profiles (Fig. 5), two main domains are observed.


image file: c7cp05207c-f5.tif
Fig. 5 Fraction evolution of the three calculated MCR-ALS spectra versus lithium extraction in Li7−xMnN4.

image file: c7cp05207c-f6.tif
Fig. 6 (a and b) Comparison of the three MCR-ALS calculated spectra with selected experimental spectra (Li7MnN4, Li6.1MnN4 and Li5.3MnN4).

Fig. 5 displays a first biphasic region between Env. 1 and Env. 2 for x < 0.85. The first environment (Env. 1), which exhibits a fraction of 1 for x = 0 in Li7−xMnN4, is obviously attributed to pure Li7MnN4. As a result, the calculated spectrum extracted from the MCR-ALS analysis fits well the experimental data of the Li7MnN4 (Fig. 6, black lines). Then, very quickly (at x = 0.05), a second environment (Env. 2) appears and gradually replaces Env. 1 (Fig. 5). At x ∼ 0.85, the fraction of Env. 2 reaches a maximum while the Env. 1 fraction is almost nil. Then, for x > 0.85, a second biphasic domain took place between Env. 2 and Env. 3. At the end of the oxidation process, i.e. for x = 1.7, Env. 3 is the main component with ∼69% and Env. 2 is still present and represents ∼31% of the absorption signal. However, one can notice that Env. 3 appears before the complete disappearance of Env. 1 (from x around 0.4), leading to the simultaneous existence of the 3 environments. This fact is attributed to the dynamic condition of our experiment which can induce small local inhomogeneities in our composite electrode. This point is also supported by the evolution of the Env. 3 concentration profile which displays two different slopes. First, the proportion of Env. 3 increases very smoothly between 0.4 and 0.8 with a maximum of ∼7%. Then, for x > 0.85, the concentration profile rises quickly to reach a maximum of 69% at x = 1.7.

In Fig. 6, the three independent calculated spectra obtained through the MCR-ALS procedure are compared to the three experimental spectra highlighted in Fig. 3 (i.e. Li7MnN4, Li6.1MnN4 and Li5.3MnN4) and peculiar energies of both experimental and calculated spectra are reported in Table 1 (i.e. pre-edge and threshold energies). For Env. 1 and Env. 2, calculated spectra were compared to their maximum of concentration, i.e. to Li7MnN4 and Li6.1MnN4 respectively. The third environment was compared to the last spectrum of the oxidation, which corresponds to the Li5.3MnN4 composition. Qualitatively, we observe an excellent agreement between Li7MnN4 and Env. 1. The spectrum of Env. 2 well describes the Li6.1MnN4 spectrum but the agreement is less good than for Env. 1. Indeed, as shown in Fig. 5, pure Env. 2 is never observed and the maximum fraction in the composite electrode reaches approximately 93% of the experimental absorption spectra. The remaining 7% are due to Env. 3.

The largest discrepancy is between the calculated Env. 3 and the Li5.3MnN4 experimental spectrum. As shown in Fig. 6b, the threshold of Env. 3 is clearly shifted to a higher energy, with a ΔE value of ∼0.6 eV. In addition, the pre-edge peak of the calculated spectrum is more intense and thin than the experimental one. These two simple considerations indicate that Env. 3 is in a more oxidized state than the last recorded spectrum. Then, according to the concentration profiles determined through the MCR-ALS analysis (Fig. 5), the Li5.3MnN4 spectrum is a combination of both Env. 2 and Env. 3. The pre-edge peak of the last experimental spectrum is larger and more asymmetric than the calculated one of Env. 3 simply because Env. 2 is responsible for ∼31% of the absorption spectra.

Oxidation state attribution of the environments

To facilitate the interpretation of the Li7MnN4 Mn K-edge upon delithiation, various phases with well-known oxidation states were used as reference compounds. However, to our knowledge, manganese nitrides are very few and not well characterized. Manganese K-edge spectra of several oxides such as LiMn2O4, BaMnO4 and KMnO4 were recorded as references and compared in Fig. 7 with the three selected experimental spectra. First, as for the nitrides, BaMnO4 and KMnO4 present a pre-edge peak, according to the tetrahedral environment of Mn ions while in the case of LiMn2O4, such a feature is almost absent due to the manganese octahedral environment. The residual intensity observed around 6541 eV in this last model compound is attributed to a small local distortion induced by the Jahn–Teller effect of Mn+3 ions.32,33 Second, for a similar formal oxidation state, oxide spectra are shifted to higher energies with regard to the nitride ones. This is clearly evidenced in Table 1, where the pre-edge peak maximum and threshold energies (the first inflexion point of the edge) are gathered.
image file: c7cp05207c-f7.tif
Fig. 7 LiMn2O4, KMnO4 and BaMnO4 K-edge spectra compared to Li7MnN4, Li6.1MnN4 and Li5.2MnN4 spectra.

According to ref. 34–36, absorption edge energies are sensitive to the effective charge of the absorbent. Indeed, formal oxidation states are not sufficient to explain the different observed energies for nitrides and oxides. For example, comparisons of manganese pre-edge and threshold energies of Li6.1MnN4 and BaMnO4, which displayed similar formal oxidation states (+5.9 and +6), lead to significant shifts: ΔEpre-edge = 0.5 and ΔEthreshold = 2.9 eV. Also, BaMnO4 and Li5.3MnN4, which have different oxidation states, exhibit almost the same pre-edge and similar threshold energies (Table 1). The nature of the metal–ligand bond is therefore responsible for such chemical shifts which can be correlated by the introduction of the coordination charges η (Batsanov method, described in ref. 35). The concept of coordination charge η is a simple mean to take into account the degree of covalence of the metal–ligand bond, which is obviously dependent on the ligand electronegativity, and finally to roughly reflect the net charge of the absorbing atom. It is defined by:

 
image file: c7cp05207c-t1.tif(1)
where m is the formal oxidation state, ck is the degree of covalence and nk the number of k bonds. ck is directly related to the ionicity i and can be written using the Pauling's formula:
 
image file: c7cp05207c-t2.tif(2)
where χa and χb are manganese and ligand electronegativies (χO = 3.50, χN = 3.07 and χMn = 1.6 taken from Allred and Rochow37).

Coordination charges of each composition, either experimental or calculated from MCR-ALS analysis, are evaluated using eqn (1) and (2) and gathered in Table 1. Formal average oxidation states of nitrides are determined from the composition of the pristine powder and the subsequent coulombic titration upon electrochemical oxidation. Since the electronegativity differences between the manganese and the nitrogen or the oxygen are rather important, the coordination charges are all positives, suggesting a charge transfer from the manganese to the ionic species in both cases. Pre-edge and threshold energies are plotted versus η in Fig. 8. Blue squares and red circles are related to the experimental spectra of nitrides and oxides respectively. Open symbols represent the pre-edge energies while full ones are related to the threshold. For both experimental sets of peculiar energies in the absorption spectra, an excellent correlation with the estimated coordination charge is observed. According to ref. 34–36, such a simple method clearly shows the influence of the ligand's electronegativity. Since nitrogen has a lower electronegativity than oxygen, the Mn–N bond is more covalent than the Mn–O one, leading to an overall shift to lower energies.


image file: c7cp05207c-f8.tif
Fig. 8 Threshold and pre-edge energies versus coordination charge η of Li7MnN4; Li6.1MnN4, and Li5.2MnN4 compared to reference oxides LiMn2O4, BaMnO4 and KMnO4.

As described above, the full data set can be decomposed to a linear combination of only three independent spectra, which correspond to three manganese environments with different oxidation states. These data are shown in Fig. 8 (open and full dark blue diamond symbols). The first environment is obviously ascribed to a +5 valence, in agreement with the Li7MnN4 stoichiometry. The last environment (Env. 3) was set to an oxidation state of +7. Indeed, the comparison between the last experimental spectrum and the calculated one (Fig. 6) suggests a higher overall oxidation state for the latter. The pre-edge peak and threshold energies support this assumption, i.e. Mn7+ in a nitrogen tetrahedral environment (Fig. 8). The maximum of Env. 2 is reached for x ∼ 0.85. Since integers were ascribed to Env. 1 and Env. 3, we have settled the oxidation state of the second environment (Env. 2) to +6. The obtained coordination charge is in excellent agreement with pre-edge peak energy but the energy threshold exhibits a larger discrepancy than the two other environments. This may originate from the accuracy of the MCR-ALS spectrum reconstruction.

Reversibility of the process

As seen in Fig. 9, the comparison of the first spectrum of the oxidation process and the last spectrum of the reduction process showed that redox processes involved are also reversible at the local environment of manganese. Such a reversibility, already demonstrated on a larger scale from XRD,16 explains the excellent cycling life of Li7MnN4 in the potential range 1–1.7 V vs. Li+/Li.
image file: c7cp05207c-f9.tif
Fig. 9 Comparison of Li7MnN4 Mn K-edge spectra recorded before and after a full electrochemical oxidation reduction cycle.

Response of Li7MnN4 upon electrochemical oxidation

Based on operando XRD, a structural response has been assigned to the main features of the electrochemical curve.16 Indeed, a three-phase mechanism was evidenced, involving two successive biphasic domains and then, a solid-solution behaviour. On the atomic scale, XRD is a long-range characterisation technique compared to XAS, which provides information on a few coordination shells. In Fig. 10, a schematic representation of the phase fraction evolutions of the 3 crystalline phases involved, built from the interpretation of operando XRD,16 is compared to the environment fraction evolutions determined in this work. From Fig. 10, a complete picture of Li7MnN4 can be built. First, during the first potential plateau, the observed biphasic domain between Li7MnN4 (x = 0) and Li6.1MnN4 (x = 0.87(5)) coincides with the Env. 1 and Env. 2 fraction evolution. Indeed, as mentioned above, the oxidation of manganese in Li7MnN4 directly induced a phase transition. During the second plateau, a different evolution is observed. Indeed, Env. 2 has not disappeared at the end of the second biphasic domain and similar proportions between Env. 2 and Env. 3 are observed. Then, during the last part of the oxidation process, i.e. the potential increase of the electrochemical curve for x > 1.35 F mol−1, a solid solution behaviour was proposed.16 XAS analysis indicates that this third crystalline phase shows a smooth evolution of the proportion of Env. 2 and Env. 3 until the end of the oxidation process, with a proportion of 31%/69% for Env. 2/Env. 3. According to the faradaic yield and the XAS analysis, the oxidation of Li7MnN4 does not lead to the full oxidation of Mn5+ to Mn7+, but to an average valence of +6.7.
image file: c7cp05207c-f10.tif
Fig. 10 Evolution of the phase fractions extracted from the structural mechanism proposed in ref. 16 compared to the environmental fraction evolutions determined from the MCR-ALS analysis.

Conclusions

A complete data set of Mn K-edge XAS spectra has been collected operando upon electrochemical delithiation of Li7MnN4. As indicated in ref. 17, operando techniques allow the avoidance of contamination possibly induced by ex situ preparations, which is a key point for a moisture sensitive material like Li7MnN4.38 In addition, the collection of a large enough data set possible with the operando collection mode permits a fine analysis using various chemometric tools (PCA and MCR-ALS). The reconstruction and the examination of the three pure independent manganese environments involved in the redox processes are provided. In this work, we clearly demonstrate the implication of the three highest oxidation states of manganese (i.e. Mn5+, Mn6+ and Mn7+) in the redox processes, which is remarkable in such a low potential range (0.9–1.7 V vs. Li+/Li). The chemometric analysis presented here might have been facilitated by the presence of intense pre-edge peaks. Indeed, the energy and the intensity of this 1s → 3d + 4p transition, only allowed in a tetrahedral local environment, are very sensitive to the oxidation state. However, this method can also be applied to transition metal compounds with lower oxidation states arranged in octahedral sites, as shown in ref. 18, 39 and 40. As a result, a coherent picture has been built between the manganese local environment and the long-range structural variation proposed in ref. 16. Indeed, the first voltage plateau at 1.18 V vs. Li/Li+, which indicates the oxidation of Li7MnN4 through a biphasic process into Li6.1MnN4, is attributed to the oxidation of Mn5+ (3d2) ions in Mn6+ (3d1). The second voltage plateau at 1.20 V vs. Li/Li+, where Li6.1MnN4 is oxidised in Li5.7MnN4 again through a biphasic process, leads to the oxidation of roughly 50% of Mn6+ in Mn7+. Then, the increase of potential recorded for 1.35 < x < 1.7 in Li7−xMnN4, in line with the observed solid-solution like evolution of the structure,16 accounts for the additional 20% oxidation of Mn6+ in Mn7+. In addition, the excellent reversibility of the Li7MnN4 structural response, which has already been demonstrated on a long-range scale,16 is also fully established on the local scale.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

The authors acknowledge Dr Isabelle Llorens and Dr J.-L. Hazemann (ESRF) for their technical assistance and the ESRF for beamtime allocation. E. P. acknowledge the CNES (Toulouse, France) and the CNRS for providing scholarship funding. Dr R. Baddour-Hadjean and Dr B. Laik are thanked for fruitful discussions on the overall redox mechanism.

Notes and references

  1. M. N. Obrovac and V. L. Chevrier, Chem. Rev., 2014, 114, 11444–11502 CrossRef CAS PubMed .
  2. I. Rom, M. Wachtler, I. Papst, M. Schmied, J. O. Besenhard, F. Hofer and M. Winter, Solid State Ionics, 2001, 143, 329–336 CrossRef CAS .
  3. R. A. Huggins, Lithium alloy negative electrodes, J. Power Sources, 1999, 81–82, 13–19 CrossRef CAS .
  4. D. C. S. Souza, V. Pralong, A. J. Jacobson and L. F. Nazar, Science, 2002, 296, 2012–2015 CrossRef CAS PubMed .
  5. R. Alcantara, J. L. Tirado, J. C. Jumas, L. Montconduit and J. Olivier-Fourcade, J. Power Sources, 2002, 109, 308–312 CrossRef CAS .
  6. M. V. Reddy, G. V. Subba Rao and B. V. R. Chowdari, Chem. Rev., 2013, 113, 5364–5457 CrossRef CAS PubMed .
  7. S. Suzuki and T. Shodai, Solid State Ionics, 1999, 116, 1–9 CrossRef CAS .
  8. N. Tapia-Ruiz, M. Segalés and D. H. Gregory, Coord. Chem. Rev., 2013, 257, 1978–2014 CrossRef CAS .
  9. J. M. Cameron, R. W. Hughes, Y. Zhao and D. H. Gregory, Chem. Soc. Rev., 2011, 40, 4099–4118 RSC .
  10. R. Niewa, F. R. Wagner, W. Schnelle, O. Hochrein and R. Kniep, Inorg. Chem., 2001, 40, 5215 CrossRef CAS PubMed .
  11. M. Nishijima, N. Tadokoro, Y. Takeda, N. Imanishi and O. Yamamoto, J. Electrochem. Soc., 1994, 141, 2966–2971 CrossRef CAS .
  12. J. Cabana, N. Dupré, G. Rousse, C. P. Grey and M. R. Palacín, Solid State Ionics, 2005, 176, 2205–2218 CrossRef CAS .
  13. J. Cabana, C. M. Ionica-Bousquet, C. P. Grey and M. R. Palacín, Electrochem. Commun., 2010, 12, 315–318 CrossRef CAS .
  14. E. Panabière, N. Emery, S. Bach, J.-P. Pereira-Ramos and P. Willmann, Electrochim. Acta, 2013, 97, 393–397 CrossRef .
  15. E. Panabière, N. Emery, S. Bach, J.-P. Pereira-Ramos and P. Willman, J. Alloys Compd., 2016, 663, 624–630 CrossRef .
  16. N. Emery, E. Panabière, O. Crosnier, S. Bach, T. Brousse, P. Willmann and J.-P. Pereira-Ramos, J. Power Sources, 2014, 247, 402–405 CrossRef CAS .
  17. P. P. R. M. L. Harks, F. M. Mulder and P. H. L. Notten, J. Power Sources, 2015, 288, 92–105 CrossRef CAS .
  18. P. Conti, S. Zamponi, M. Giorgetti, M. Berrettoni and W. H. Smyrl, Anal. Chem., 2010, 82, 3629–3635 CrossRef CAS PubMed .
  19. O. Proux, X. Biquard, E. Lahera, J.-J. Menthonnex, A. Prat, O. Ulrich, Y. Soldo, P. Trévisson, G. Kapoujyan, G. Perroux, P. Taunier, D. Grand, P. Jeantet, M. Deleglise, J.-P. Roux and J.-L. Hazemann, Phys. Scr., 2005, T115, 970–973 CrossRef CAS .
  20. J.-B. Leriche, S. Hamelet, J. Shu, M. Morcrette, C. Masquelier, G. Ouvrard, M. Zerrouki, P. Soudan, S. Belin, E. Elkaïn and F. Baudelet, J. Electrochem. Soc., 2010, 157, A606–A610 CrossRef CAS .
  21. (a) A. Michalowicz, J. Moscovici, D. Muller-Bouvet and K. Provost, J. Phys.: Conf. Ser., 2013, 430, 012016 CrossRef ; (b) http://www.icmpe.cnrs.fr/spip.php?article602 .
  22. F. Malinowski and D. Howery, Factor Analysis in Chemistry, Wiley, New York, 1980 Search PubMed .
  23. S. R. Wasserman, P. G. Allen, K. Shuh, J. J. Bucher and N. M. Edelstein, J. Synchrotron Radiat., 1999, 6, 284–286 CrossRef CAS PubMed .
  24. A. de Juan, J. Jaumot and R. Tauler, Anal. Methods, 2014, 6, 4964–4976 RSC .
  25. W. H. Cassinelli, L. Martins, A. R. Passos, S. H. Pulcinelli, C. V. Santilli, A. Rochet and V. Briois, Catal. Today, 2014, 229, 114–122 CrossRef CAS .
  26. P. R. Bevington and D. K. Robinson, Data reduction and error analysis for the physical sciences, McGraw Hill, NY, 2003 Search PubMed .
  27. W. H. Press, S. A. Teukolsky, W. T. Vetterling and B. P. Flannery, Numerical Recipes, Cambridge University Press, 1997 Search PubMed .
  28. C. Hannay, R. Thissen, V. Briois, M.-J. Hubin-Franskin, F. Grandjean, G. J. Long and S. Trofimenko, Inorg. Chem., 1994, 33, 5983–5987 CrossRef CAS .
  29. A. Manceau, A. I. Gorshkov and V. A. Drits, Am. Mineral., 1992, 77, 1133–1143 CAS .
  30. M. Belli, A. Scafati, A. Bianconi, S. Mobilio, L. Palladino, A. Reale and E. Burattini, Solid State Commun., 1980, 35, 355–361 CrossRef CAS .
  31. A. Manceau, M. Marcus and T. Lenoir, J. Synchrotron Radiat., 2014, 21, 1140–1147 CAS .
  32. T. Okumura, Y. Yamaguchi, M. Shikano and H. Kobayashi, J. Mater. Chem. A, 2014, 2, 8017–8025 CAS .
  33. H. Yamaguchi, A. Yamada and H. Uwe, Phys. Rev. B: Condens. Matter Mater. Phys., 1998, 58, 8–11 CrossRef CAS .
  34. C. Mande and V. B. Sapre, Chemical shifts in X-ray absorption spectra, in Advances in X-ray spectroscopy, ed. C. Bonnelle and C. Mande, Pergamon Press, 1982, ch. 17, pp. 287–301 Search PubMed .
  35. S. P. Cramer, T. K. Eccles, F. W. Kutzler, K. O. Hodgson and L. E. Mortenson, J. Am. Chem. Soc., 1976, 98, 1287–1288 CrossRef CAS PubMed .
  36. J. Wong, F. W. Lytle, R. P. Messmer and D. H. Maylotte, Phys. Rev. B: Condens. Matter Mater. Phys., 1984, 30, 5596–5610 CrossRef CAS .
  37. A. L. Allred and E. G. Rochow, J. Inorg. Nucl. Chem., 1958, 5, 264–268 CrossRef CAS .
  38. E. Panabière, N. Emery, S. Bach, J.-P. Pereira-Ramos and P. Willmann, Corros. Sci., 2013, 77, 64–68 CrossRef .
  39. A. Iadecola, A. Perea, L. Aldon, G. Aquilanti and L. Steviano, J. Phys. D: Appl. Phys., 2017, 50, 144004 CrossRef .
  40. A. Rocht, B. Baudet, V. Moizan, C. Pichon and V. Briois, C. R. Chim., 2016, 19, 1337–1351 CrossRef .

Footnote

Electronic supplementary information (ESI) available: (ESI1) – Operando experiment electrochemical curve. (ESI2) – Constraints applied in the chemometric analysis and (ESI3) reconstruction of various spectra using the 3 different environmental components. See DOI: 10.1039/c7cp05207c

This journal is © the Owner Societies 2017