1. Introduction

The improvement of high performance lithium-ion batteries requires an optimization of the structure of materials, the knowledge of the mechanism of the intercalation/deintercalation of Li⁺ into/from host materials as well as the kinetics of the electrolytic reaction ^[1]. Since the early report by Novák et al. ^[2], only few articles have addressed an overview of in situ microscopic studies on the structural and chemical behaviors of lithium-ion battery materials ^[3,4,5]. In this respect, in operando or in situ analytical techniques have been widely developed to investigate dynamic systems such as material growth ^[6], insertion/deinsertion reaction ^[7], phase stability ^[8], interface evolution ^[9], compatibility between materials ^[10], electrode degradation ^[11,12], etc.

As a non-destructive and sensitive method, Raman spectroscopy (RS) is a powerful analytical technique that has the potential to study electrode materials under in situ conditions. RS involves an inelastic scattering of the electromagnetic radiation by the matter, characterized by a shift of the frequency of the scattered light with respect to the incident light. This shift is equal to the frequency of the vibrational modes of atoms or molecules that are Raman active, which depends only on their symmetry. Consequently, RS is a primary tool used to probe the structural evolution and the chemical composition of materials composing a lithium-ion battery. It gives information on the local structure (short-range order), in contrast to X-ray diffraction (XRD) that probes the long-range order of the matter. Therefore, RS is a complementary tool to diffraction techniques (X-ray, electron, neutron) to investigate the local structure of transition-metal oxides used as positive electrodes in lithium-ion batteries ^[13]. It proved to be very useful in characterizing the phase transformations in such materials upon cycling ^[14] or heat treatment ^[15]. Such changes can affect only the surface of the particles, in which case they are not detected by diffraction techniques. RS is also a powerful analytical technique for the knowledge of the structure of either disordered or amorphous compounds ^[16,17]. It should be noted that, in general, non-stoichiometry, substitutions or ion exchanges in a non-ideal material are violating the selection rules giving rise to additional bands, wavenumber shifts and/or band broadenings. For instance, Raman spectroscopy was employed to probe the nucleation and growth and structure of pulsed-laser deposited (PLD) V₂O₅ films ^[18]. The spectral features allow to establish a phase diagram mapping the effect of substrate temperature on the microstructure evolution of PLD V₂O₅ thin films. RS has also been widely used to study the surface phenomena of electrodes ^[19,20] and electrolyte including ion-solvent interactions, since the conductivity being proportional to the salt concentration, a correlation between Raman shift and the conductivity of the electrolyte can be established ^[21,22].

In situ Raman spectroscopy is a direct method for real-time study of dynamic reactions during working conditions of electrochemical cells. Four spectral evolutions can be distinguished as (ⅰ) appearance of new band, (ⅱ) shift of vibrational mode, (ⅲ) change in band profile and (ⅳ) change in band intensities (Figure 1). In situ RS has been widely employed to elucidate the structural changes at the level of chemical bonds in electrode materials for lithium-ion batteries during the cycling of electrochemical cell; this analytical method is currently used to describe the variations in the valence state of transition-metal cations, lithium transport mechanism (insertion/deinsertion reaction), cycle-induced degradation, electrode–electrolyte interface evolution and state-of-charge (SOC) distribution as well ^[4,23]. Kinetic characterizations of Li⁺ insertion/extraction into/from composite cathode were also investigated by in situ Raman spectroscopy ^[24]. In situ RS is currently used to investigate the electrode–electrolyte interface in lithium batteries ^[25]. Taking the advantages in the difference in Raman spectra vs. composition or/and local symmetry, it is possible to study the spatial distribution of phases in an electrode by mapping ^[26]. The powerful advantage of in situ micro-Raman spectroscopy for the analysis of materials included in a lithium battery was emphasized by several authors ^[27,28,29].

The purpose of this review is to acknowledge the current state-of-the-art and the progress of in situ Raman spectroscopy that has been made on all the elements of the lithium-ion batteries: cathode and anode materials and the formation of solid electrolyte interphase (SEI) layer. We also discuss the results in the perspective of spatial and real-time investigations by in situ Raman imaging (mapping) during charge/discharge cycling.

---

2. Raman single point and mapping measurements

---

2.1. In situ Raman single point

Typical in situ Raman spectroscopy is carried out using a confocal Raman spectrometer in backscattering geometry. For a description of the micro-Raman set-up, we guide the reader to the review papers Refs. ^[28,29]. Typical home-made electrochemical cells for in situ Raman studies were targeted to meet their experimental measurements ^{[30,31,32,33]}. Burba and Frech ^[30] described a spectro-electrochemical cell that was a modified industrially available coin cell with a 2-mm diameter hole drilled into the coin cell casing (Figure 2a). In situ RS cells are currently equipped with a glass, quartz or sapphire window transparent to the Raman laser light. A 100× magnification long working distance optical objective is used, providing a laser spot of approximately 3 μm² on the sample electrode. Another in-situ cell was fabricated using the Swagelok standard ^[32] with a sapphire optical window (Figure 2b). A home-made in situ cell by Novák et al. ^[34] is presented in Figure 2c. Industry-standard 2032-coin cells (from Pred Materials International) were adapted to carried out in situ Raman mapping of electrode materials (Figures 2d, e). Fang et al. used a MgO window to cover a drilled 1/8 inch diameter hole. State-of-charge maps were collected of a 35 × 35 μm² area considering two spectral features, i.e., Raman peak position and peak intensity ^[35]. Ghanty et al. ^[36] fabricated a in situ Raman pouch cell with an optical sodium borosilicate window covering an aperture in the aluminum-coated polyamide outer shell (Figure 2f).

---

2.2. Optical-skin depth

A Raman spectrum represents the response of nanoparticles contained in the volume defined by the focal spot times the penetration depth δ_p that depends on the probed sample. The axial resolution in opaque materials is determined by the optical-skin depth (δ_p) of the laser beam penetrating into a material. Because of the optical absorption of the light in electrode materials, the observed Raman spectra are induced in a thin surface layer; thus, due to small δ_p value, Raman spectroscopy can be considered as a surface analytical method in most cases. The Raman scattering intensity is then weak because of the small material volume probed (V < 1 μm³). The skin depth is given by the relation:

${\delta _{\rm{p}}} = {\left( {2\lambda /\mu \sigma } \right)^{1/2}}$

(1)

where λ is the laser wavelength, σ the electronic conductivity and μ the magnetic permeability.

Consequently, using laser excitation with higher wavelength increases the penetration depth and an increase in the electronic conductivity, e.g., from semiconductor to metal, results in decrease of δ_p. According to the Beer–Lambert law, the intensity I(ξ) of the laser light inside the materials is an exponential function of the absorption coefficient α of the material as I(ξ) = I₀exp[−αξ]. The optical penetration depth defined as the distance at which the intensity decreases to 1/e (37%), where e is the base of the natural system of logarithms, we have δ_p = α⁻¹. For instance, δ_p ≈ 1 μm in crystalline silicon, δ_p ≈ 100 nm in amorphous silicon and δ_p ≈ 100 nm in Li_xSi ^[38,39,40]. Currently, the experiments are performed at very low laser power W_laser ≈ 0.02 mW using a 0.1% filter that corresponds to a specific power of 600 W·cm⁻². An optical-skin depth is less than 50 nm in highly oriented pyrolytic graphite examined with a 514.5 nm green laser beam ^[41], while δ_p falls to few nanometers in delithiated Li_1−xNi_1/3Mn_1/3Co_1/3O₂. Another experimental case is the low optical-skin depth (δ_p ≈ 30 nm) of carbon coating onto LiFePO₄ particles ^[42,43]. Thus, to probe the electrode materials of LIBs, the choice of laser excitation line at high wavelength λ > 700 nm is preferred, rather than the solid-state 532 nm excitation.

---

2.3. In situ Raman imaging

In situ Raman mapping consists in the spatial characterization of an electrode at a multitude of locations that allows to define the chemical distribution and inhomogeneity of particles during the cell charge–discharge process and electrode degradation as well. A Raman image is formed by hyperspectral data set, in which a pixel is a complete Raman spectrum. Raman micro-spectroscopy as a spatially resolved analytical method is currently used to generate local surface images, to construct state-of-charge maps of electrodes, or to investigate the kinetic aspect of Li⁺ intercalation–deintercalation process in a single particle. The feasibility of Raman mapping was first explored by Panitz et al. ^[44] who investigated the lithium intercalation in graphite electrode under potentiostatic and galvanostatic conditions. Results indicated that, at a potential of 0.2 V vs. Li⁺/Li, the lithium is not homogeneously inserted over the graphite electrode and a new Raman band evolved at ca. 1850 cm⁻¹ at potential 0.18 V vs. Li⁺/Li. The same group developed an in-situ cell to monitor the Li insertion in carbon electrode by changing the optical geometry that improved the signal-to noise ratio by a factor of 20 ^[45]. The mapping of the state-of-charge distribution of LiCoO₂ (LCO) was investigated with a special resolution of few micrometers, which enabled to test the intensity, width and position of the Raman peaks by constructing the Raman images of the A_1g mode (595 cm⁻¹) ^[46]. The remarkable potential of Raman imaging was applied to study the local degradation behavior in composite cathode materials such as LiNi_0.8Co_0.15Al_0.05O₂ and LiNi_0.33Mn_0.33Co_0.33O₂ ^[47]. The authors investigated the nanoscale modifications of the surface composition, structure, SOC, and determined the electronic conductivity at the electrode surface as well. Raman imaging was performed on an area of 52 × 75 μm² at 0.7 μm resolution. An image is formed of color-coded pixels, which represent the relative intensities of specific Raman peaks and the chemical surface composition, i.e., active material and carbonaceous additive (acetylene black, graphite, etc.) characterized by the Raman G- and D-bands. Currently, the Raman spectrum of an electrode particle varies significantly with the position on the surface. For example, the spectrum of LiNi_0.33Mn_0.33Co_0.33O₂ displays bands at 475 and 554 cm⁻¹ associated with the vibrations of the Ni–O bonds having intensities strongly coupled with electronic states that provoke Raman resonance behavior. Consequently, the relative intensities I₄₇₅/I₅₅₄ vary significantly from one particle to another. Fang et al. ^[35] adapted industry-standard coin cells for high spatial resolving in situ Raman mapping. By extraction of the local frequency of the A_1g mode of electrodes with rock-salt-type structure, the authors demonstrated an easy-to-implement electrode design using LiNi_0.5Mn_0.3Co_0.2O₂ (NMC) as active material.

Quantification of the Li⁺ ion transport in liquid electrolyte, i.e., LiClO₄ dissolved in dimethyl carbonate (DMC), was performed by combining high-spatial-resolution confocal Raman microscopy and microfluidic technique. ^[48]. The Raman peak at ca. 523 cm⁻¹ corresponding to the O–C–O deformation mode of DMC develops a sideband with the increase of LiClO₄ concentration. Its analysis using a Voigt profile was used to quantify the ionic diffuse transport.

Using high-speed and high-definition in situ Raman imaging, several groups have investigated the local SOC and state-of-health (SOH) distribution of electrodes. Nishi et al. ^[49] visualized the SOC distribution in LiCoO₂ at various rates and showed the limitation of ionic transport. Nanda et al. ^[50] investigated the SOC inhomogeneity in fresh and aged layered cathodes, while Slautin et al. ^[51] studied the degradation paths in LiMn₂O₄ spinel structures. As an example of experimental procedure, Raman maps of the LiNi_0.8Co_0.15Al_0.05O₂ (NCA) positive electrode were performed by collecting spectra in a range of Raman shift from 250 to 1800 cm⁻¹ across a 10 μm square region of the electrode surface ^[50]. Results obtained from the composite electrode NCA + carbon black + PVdF binder are shown in Figure 3. Typical map is obtained with a minimum of 1000 data points. Each spectrum was treated considering three Raman peaks of NCA centered at ca. 475,550 and 620 cm⁻¹ attributed to the vibrations of Ni–O bonds, while bands of the carbon are observed at ~1350 cm⁻¹ (D band), 1500 cm⁻¹ (amorphous carbon) and 1590 cm⁻¹ (G band).

---

3. Positive electrode (cathode) materials

---

3.1. Rock-salt-type compounds

Structural properties of layered-type compounds such as LiCoO₂, LiNiO₂, LiNi_xCo_1−xO₂, LiNi_0.8Co_0.1.5Al_0.05O₂ as well as LiNi_xMn_yCo_zO₂ with various Co contents were widely investigated by in situ X-ray diffraction ^[52]. These compounds crystallize with a lamellar structure of rock-salt-type of R-3m space group. For instance, the poor electrochemical cycling behavior of Ni-rich NMC (LiNi_0.5Mn_0.2Co_0.3O₂) cathode is due to the structural phase transition between hexagonal phases H2 to another hexagonal phase H3 (at 4.7 V) upon Li⁺ de-intercalation, which can be suppressed in long-term cycleability.

---

3.1.1. LiCoO₂ (LCO)

LiCoO₂ prepared by solid-state reaction or by wet chemistry followed by a post-annealing treatment at 800–900 ℃ (denoted HT-LCO) crystallizes in the rhombohedral system R-3m with the α-NaFeO₂ structure built by alternating ordered layers of CoO₆ and LiO₆ octahedra along the c-axis. At a lower annealing temperature, T_a ≈ 400 ℃, LCO crystallizes in the cubic NaCl-type structure (Fd3m space group) (LT-LCO) with cations distributed among octahedral sites of the cubic oxygen close-packed array. While the HT- and LT-LCO structures cannot be efficiently discriminated by X-ray diffraction, Raman spectroscopy has been successful to determine the phase of LiCoO₂ synthesized in the temperature range 400–900 ℃ ^[16].

---

3.1.1.1. Phase identification of Li_1−xCoO₂

The phase identification of Li_1−xCoO₂ as a function of the Li extraction is currently established via the A_1g and E_g phonon Raman active (gerade) modes (Table 1), which are observed at 596 and 487 cm⁻¹, respectively. The A_1g mode is the symmetrical stretching of Co–O (atomic displacement parallel to c-axis) whereas the E_g mode is the symmetrical deformation (atomic displacement perpendicular to c-axis). Gross and Hess ^[32] demonstrated the presence of the resonance Raman effect for LiCoO₂ cathode under electrochemical conditions upon switching the laser excitation by evaluating the A_1g/E_g integrated intensity ratio. A_1g/E_g increases from 0.9 (for λ = 632.8 nm) to 2.3 (for l = 532 nm) due to resonance process involving the d–d electronic transition from Co-t_2g to Co-e_g bands (optical absorption at 2.1 eV (591 nm)) and under resonance conditions an overtone of the A_1g band occurs ^[12]. The resonance enhancement for LCO materials was also pointed out when using a green laser excitation for in situ studies ^[32].

Several articles have reported in situ RS experiments of LCO electrode materials ^{[2,14,20,31,32,53,54,55,56,57,58,59]} but only few of them mentioned the phase transformation upon lithium extraction/insertion. The early work by Inaba et al. ^[20] showed the change in the Raman spectra of Li_1−xCoO₂ prepared by electrochemical extraction of Li ions. Results indicated a line broadening of a second hexagonal phase and the monoclinic phase. Novák et al. ^[2] addressed the rapid disappearance of the A_1g band with increasing potential (V > 3.95 V). This Raman band emerges again on discharge (at ca. 3.7 V) with a lower intensity, which was attributed to a lithium-deficient LCO electrode.

As shown in Figure 4, the phase evolution during the charge of Li_1−xCoO₂ (lithium extraction) is as follows: (ⅰ) for 0 ≤ x ≤ 0.1 the two phonon (normal) modes at 596 and 487 cm⁻¹ are invariant and correspond to the hexagonal H1 phase, (ⅱ) two new Raman bands located at 473 and 572 cm⁻¹ appear at x > 0.1 due to the formation of the hexagonal H2 phase that coexists with H1 up to x ≈ 0.3, i.e., in a two-phase system, (ⅲ) for x > 0.3 the LCO-type normal modes disappeared while the modes of the H2 phase shift toward lower wavenumbers to 462 and 567 cm⁻¹, respectively, due to the expansion of the Co–O bonds along the c-axis (Figure 5). It is worthy to note that the intensities of the Raman peaks decrease significantly due to the transition from semiconductor to metal for x > 0.15. Novák et al. ^[2] suggested an increase of electrical conductivity by about two orders of magnitude in Li_1−xCoO₂ for x > 0.04 correlated with the change of colour of the particles from dark to light. Similar description was previously reported by Inaba et al. ^[20], while Itoh et al. ^[14] conceived the formation of a Raman-inactive phase. Song et al. ^[15] investigated the phase change of LCO films deposited on to Co substrate using a spot of 1 μm of 514.5 nm laser irradiation by in situ RS. A remarkable phase change of the hexagonal film deposited at 200 ℃ to cubic spinel is due to reaction of LCO films with the Co substrate forming a Co-rich phase or Li deficient phase Li_1−xCoO₂. The laser power (≤50 mW·cm⁻²) provokes a shift accompanied by broadening and depressed intensity of the hexagonal A_1g (589 cm⁻¹) and E_g (478 cm⁻¹) bands, and the occurrence of two new bands at ca. 523 and 681 cm⁻¹ that correspond to the cubic spinel Fd3m phase. In the harmonic approximation, the phonon frequencies are given by:

$\nu \approx {\left( {k/\mu } \right)^{1/2}}$

(2)

where k is the force constant and μ the reduced mass (μ⁻¹ = m_Li⁻¹ + (2m_O)⁻¹). As Li and Co co-exist in the Li sites, μ becomes larger, which results in band softening and broadening. Such a structural transformation has been confirmed by FTIR and RS spectroscopy of sol-gel LCO powders ^[60] showing the growth of the rock-salt structure (R-3m symmetry) upon annealing heat-treatment in the range 400–900 ℃. Upon prolonged cycling, degradation of LCO cathodes is evidenced by disappearance of the peak at 579 cm⁻¹ and the emergence of new bands at 515 and 674 cm⁻¹ assigned to the mode vibration of Li₂O and Co₃O₄, respectively, both being electrochemically inactive components ^[61].

---

3.1.1.2. Surface modification upon cycling

Spatially-resolved in situ Raman diagnostics were performed to analyze the variation of chemical composition of composite electrodes based on the mixture of LCO active particles, carbon additive and polyvinylidene fluoride (PVdF) binder, and the degradation mechanism of materials ^[12,56,58]. Taking into account the resonance enhancement for LCO materials, i.e., maximum for green laser excitation, the wavelength-dependent spatially resolved analysis demonstrated significant variation of chemical composition across the electrode surface ^[32]. Using the same technique, Fukumitsu et al. ^[57] investigated the SOC distribution in cross-section of LiCoO₂ cathode and observed the inhomogeneity of the active particles where Li⁺ ions did not completely return after discharging. Such an inhomogeneous distribution and the mapping of the LCO electrode degradation were also correlated with battery performance by several groups ^{[31,55,56,57]}. Hausbrand et al. ^[12] addressed the degradation aspects upon overcharge of LiCoO₂/PVdF/carbon black composite (85:10:5) electrodes by in situ RS imaging the A_1g phonon band of LCO. After cycling at C/12 rate, the composite electrode showed a redistribution of the chemical composition that consists in the formation of surface layers; thus, the loss of electrical contact between LCO particles were evidenced by mapping, while Gross and Hess claimed no significant structural changes observed from in situ RS experiments ^[32]. Nishi et al. ^[49] investigated the local SOC distribution of LCO cathode. The results showed that individual LCO particles were discharged at various rates with inhomogeneous current distribution and part of them was irreversibly charged upon subsequent cycles due to the increasing electrical resistance between LCO grains. As an example, Figure 6 displays the Raman images recorded during the first and second cycle of LCO (charge at 0.4C and discharge at 1C) in the potential range 2.5–4.4 V vs. Li⁺/Li. This electrode shows an inhomogeneous distribution of the SOC of the particles upon cycling.

Recently, Otoyama et al. ^[59] reported the Raman mapping of the composite electrode, i.e., mixture of active LiCoO₂ and Li₂S–P₂S₅ solid electrolyte before and after the initial charging reaction. While the Raman images show the expected distribution of reactions after full charge test, several LCO particles do not display structural changes. Kostecki et al. ^[47] reported similar studies performed on LiNi_0.8Co_0.15Al_0.05O₂ (NCA) composite cathode material. Observation of the Raman microscopy mapping of the electrode surface showing nonuniform electrode state-of-charge was correlated with the significant impedance rise and capacity fade of the lithium-ion cell graphite//NCA.

---

3.1.2. Mixed layered oxides LiNi_aMn_bCo_1−a−bO₂ (NMC)

Mixed layered oxides, namely lithium nickel-manganese-cobalt oxides LiNi_aMn_bCo_cO₂ with c = 1 − a − b (named NMC) are solid solutions of the binary LiMO₂ (M = Ni, Mn, Co) compounds. They are identified by the cation stoichiometry, for instance NMC532 for LiNi_0.5Mn_0.3Co_0.2O₂. NMC materials crystallize with the α-NaFeO₂ structure belonging to the R-3m (D_3d⁵) space group. The local structure of LiNi_1/3Mn_1/3Co_1/3O₂ (NMC333) synthesized by wet chemistry with various chelate to metal-ion ratio was studied Raman spectroscopy ^[62]. Analysis of the intensity A_1g and E_g modes by integral of the Lorentzian individual band allows to estimate the cationic mixing, i.e., partial occupancy of Ni²⁺ in Li sites. The intensity of the Ni–O and O–Ni–O vibrations (ν₁ and ν₄, respectively) are smaller in the sample with higher cationic mixing. The larger width of these ν₁ and ν₄ vibrations means that the lifetime of phonons is shorter, another evidence of the larger Ni cationic disorder, in agreement with the structural analysis. An important issue for NMC materials is the cation mixing, i.e., filling of Li sites by Ni²⁺ ions. Employing a combination of XRD and spectroscopic methods (Raman and NMR spectroscopy), Ben-Kamel et al. ^[63] reported the effect of varying Co content on the local structure of as-synthesized LiNi_xCo_yMn_zO₂ oxides and showed that the cation mixing effect during cycling decreases with increasing Co content, illustrating the usefulness of Raman spectroscopy in this issue. The effect of high voltage charging on NMC532 cathodes has been investigated by in situ Raman mapping by analyzing the high-frequency phonon mode. Above the potential 4.4 V vs. Li⁺/Li, the Raman component at 595 cm⁻¹ shifts above 600 cm⁻¹, which corresponds to a capacity loss similar to the one observed in spinel LiMn₂O₄ ^[64]. The local charging profiles of NMC532 particles during cycling was studied by construction of high-spatial resolved SOC images of 35 μm square portion of electrode. The delithiation process provokes a downshift in the frequency of A_1g and E_g peaks, accompanied with a decrease in the A_1g peak in intensity and an increase in the E_g peak intensity. The local topography was obtained from data by numerically extracting the frequency that highlights the inhomogeneity in the SOC distribution when the cathode is charged further to 4.21 V ^[35]. Raman mapping was conducted for NMC333 composite cathode in solid-state batteries. Using 75Li₂S–25P₂S₅ solid electrolyte, mapping images displayed uniformly delithiated and lithiated NMC particles ^[65]. The ratio I₅₉₅/I₄₇₄ was used to determine the electrode state-of-charge with I₅₉₅/I₄₇₄ = 4.0 for low SOC and I₅₉₅/I₄₇₄ = 0.85 for high SOC.

In situ Raman spectroscopy of Ni-rich Li_1+x(Ni_yCo_zMn_z)_wO₂ (0.005 < x < 0.03; y:z = 8:1, w is nearly 1) electrodes demonstrated the structural transformation using an original pouch cell with sodium borosilicate glass window ^[36]. Figure 7a presents the deconvoluted Raman spectrum of a NMC811 electrode in the spectral range of 400–650 cm⁻¹, which shows six active Raman modes according to the approach proposed by Julien and co-workers ^[63]. In situ Raman spectra collected during Li⁺ extraction (charge process) in the electrode potential from 3.8 to 4.3 V vs. Li⁺/Li are shown in Figure 7b. Spectral changes are correlated with the structural transitions from initial hexagonal phase H1 to phase H2 and to coexisting phases H1 + H2. The bandwidth (FWHM) of A_1g(Ni) and E_g(Ni) modes remained unchanged upon the potential charge 3.7–3.8 V corresponding to the H1 phase, while increased up to a maximum at 3.9–4.0 V (phases H1 and H1 + H2) and further decreased in the potential range 4.0–4.3 V implying a structural transition from disorder to order that occurs for x > 0.56 in Li_xNi_0.8Mn_0.1Co_0.1O₂ electrode. During the discharge, the Raman response shows quite reversible behavior with low deviations from the original ν and FWHM values.

---

3.1.3. Li-rich layered–layered oxides (LLNMC)

The exact nature of Li-rich layered-layered cathode materials (LLNMC) of formula Li_1+δ(Ni_aMn_bCo_1−a−b)_1−δO₂ or xLi₂MnO₃·(1 − x)LiMO₂ (with M = Ni, Mn, Co) is still the subject of debate; two models are proposed, one considering the structure of a lamellar compound with Li₂MnO₃ domains, the other proposing a solid solution as an integrated structure of Li₂MnO₃ (C2/m) and LiMO₂ (R-3m) layered components. For δ > 0, the overlithiated oxide are also known as high-energy NMC (HE-NMC). Li₂MnO₃ is electrochemically activated during the first charge at a potential plateau of 4.5 V. Depending on the composition and charging conditions, LLNMC is enabling to deliver more than 250 mAh·g⁻¹ at potential of 5 V vs. Li⁺/Li. This class of materials is often noted xLi₂MnO₃·(1 − x)LiMO₂ based on the X-ray diffraction patterns. Among this series, 0.5Li₂MnO₃·0.5LiNi_0.5Mn_0.5O₂ is equivalent to Li_1.2Ni_0.2Mn_0.6O₂ and 0.5Li₂MnO₃·0.5LiNi_0.33Mn_0.33Co_0.33O₂ to Li_1.2Mn_0.54Ni_0.13Co_0.13O₂. Both of them are promising cathode materials. The layered structure of Li₂MnO₃ or Li(Li_1/3Mn_2/3)O₂ is described in the monoclinic C2/m (C_2h³) space group; there is an extended cation ordering in the slabs with the Li/Mn ratio of 1/2 with Li ions fully occupying one 2b Wyckoff site of the slabs along the c-axis, while Mn ions occupy the 4g site. Factor group analysis predicts 15 Raman active modes as 7A_g + 8B_g.

Lanz et al. ^[66] stated that, prior to charging, i.e., in the open-circuit voltage (OCV) state, the LLNMC framework is a mixture of the Li₂MnO₃ and layered LiMO₂ phases observed by in situ XRD and in situ Raman spectroscopy. The local structure of Li-rich NMC with the composition Li_1.20Mn_0.54Co_0.13Ni_0.13O₂ studied by Raman spectroscopy was compared with that of layered compounds LiCoO₂, LiNi_1/3Mn_1/3Co_1/3O₂ and Li(Li_1/3Mn_2/3)O₂ (or Li₂MnO₃). This comparison supports that Li_1.20Mn_0.54Co_0.13Ni_0.13O₂ is a solid solution of Li₂MnO₃ and LiNi_1/3Mn_1/3Co_1/3O₂ instead of separated domains of these components ^[67]. However, an additional Raman peak located at ~660 cm⁻¹ was attributed to the vibrational mode of Mn–O bond also observed in other layered phases such as HT-Li_0.52MnO_2.1(with I4₁/amd symmetry) ^[68] simply shifted by the fact that the Mn–O bond is shorter in LLNMC.

In situ Raman spectroscopy of Li-rich layered oxides xLi₂MnO₃·(1 − x)LiMO₂ (M = Ni, Co, Mn) was carried out for several compositions x = 0.3, 0.5, and 0.7 ^{[37,66,69,70,71,72,73]}. The general trends consisted in the determination of the Li₂MnO₃ activation during the first cycle characterized by a modification of the ionic local coordination and an increase of the ionic disorder of the Li-rich layered structure. Singh et al. ^[69] reported in situ Raman spectra of Li_1.2Ni_0.175Co_0.1Mn_0.52O₂ during galvanostatic charge and discharge process between OCV (~3.0 V) and 4.8 V. During the charge, a new peak emerges at ~544 cm⁻¹ due to the oxidation of Ni²⁺ ions, and a change in the Raman peak at 445 cm⁻¹ (up to 4.1 V) reflects the extraction of Li⁺ ions from the transition-metal layer, contrary to the previous claim of a movement of Li⁺ ions to vacant site at lower potential 2.92 V ^[74]. It was concluded that, in the voltage range 4.1–4.4 V, lithium ions are removed from both the transition-metal layers and lithium layers. At upper state of charge at 4.55–4.6 V, the oxygen removal becomes severe, and associated with the diffusion of the transition metal ions into the vacant sites. In situ Raman spectroscopy was applied to characterize the electrochemical activation of Li₂MnO₃ in LLNMC cathodes. The fingerprint of this activation lies in the main peak at 615 cm⁻¹ (A_g band of Li₂MnO₃ phase), which shifts to higher wavenumbers ~630 cm⁻¹ upon of charging above 4.4 V vs. Li⁺/Li ^[70]. The same group of researchers reported the emergence of the new Raman band at ~545 cm⁻¹ for stoichiometric Li_1.1(Ni_1/3Co_1/3Mn_1/3)O_2.1 during the first charging process at low voltage in the range 4.1–4.3 V vs. Li⁺/Li. The activation of Li₂MnO₃ possibly results in the formation of Li₂O and MnO₂, the former being responsible for the band at ~545 cm^{−1 [66]}. This assumption was previously proposed by Hy et al. ^[72] by studying the oxygen-related surface reactions occurring on Li_1.2Ni₀₂Mn_0.6O₂ using in-situ surface enhanced Raman spectroscopy (SERS). Li₂O was directly detected during the extended voltage plateau at 4.5 V that leads to the hydrolysis of the electrolyte and the formation of Li₂CO₃. Rao et al. ^[71] reported the spectral modifications of the Li_1.2Mn_0.54Ni_0.13Co_0.13O₂ electrode during charging from OCV to 4.7 V (Figure 8). While the Raman spectrum is invariant until the charge at 4.2 V, a new broad peak due to the Ni³⁺–O bonds appears at ca. 542 cm⁻¹ at 4.25 V, which becomes sharp in the potential range 4.35–4.47 V. Other notable changes associated with the removal of Li⁺ ions consist in the vanished 445 cm⁻¹ peak and a shift of the 490 cm⁻¹ peak due to the increase of the interlayer distance. Finally, in the potential region 4.25–4.5 V, a strong increase of the Raman signal background is attributed to the Li₂MnO₃ activation indicating a local environment modification with simultaneous lithium extraction and oxygen release. During discharge, the Raman spectrum returns to the initial situation with suppression of the background intensity of the peak.

Particular attention has been devoted to the effects of cycling on the structure of lithium-rich NMC cathode compounds, and of Li₂MnO₃–LiMO₂ (M = Ni, Mn, Co) composites (LR-NMC), the later being an integrated structure of the C2/m (Li₂MnO₃) and R-3m (LiMO₂) layered components. Several investigators have shown the gradual transformation of Li-rich NMC materials cycled from a discharge state at 2.0 V to the charging voltage as high as 4.9 V ^[75,76]. Using a combination of XRD, TEM and ex situ Raman spectroscopy, Amalraj et al. ^[75] demonstrated a partial layered-spinel structural transition in the initial charge to 4.7 V vs. Li⁺/Li. Similar results on carbon coated Li_1.2Mn_0.56Ni_0.16Co_0.08O₂ were reported by Kumar-Nayak et al. ^[76]. The Li-rich NMC Raman features were interpreted in terms of a gradual layered to spinel transformation provoked by electrochemical cycling.

Long-term cycling of high-energy sol-gel synthesized cathodes, i.e., Li-rich layered-layered oxides xLi₂MnO₃·(1 − x)LiNi_2/3Co_1/6Mn_1/6O₂ (x = 0.3, 0.5 and 0.7) was investigated by in situ Raman spectroscopy for 35 cycles. During the rate capability measurements at different current densities, the A_1g (602 cm⁻¹) and E_g (490 cm⁻¹) phonon modes were unchanged without the occurrence of 630–670 cm⁻¹ peaks corresponding to the phonons mode of the cubic spinel-like phase. The increase of the bandwidth of the phonon modes after electrochemical tests was attributed to the formation of the SEI layer ^[73]. Huang et al. ^[37] reported similar behavior and showed that, using in-situ electrochemical Raman spectroscopy correlated with the differential capacity (dQ/dV) curves, the oxidation reactions Ni²⁺ → Ni³⁺ → Ni⁴⁺ appeared in the potential range 3.70–4.45 V with a good reversibility in Li-rich 0.5LiNi_0.5Mn_0.5O₂·0.5Li₂MnO₃ material. Recently, the study of the fate of Li₂MnO₃ (C2/m) from a commercially produced LLNMC cathode material having the composition 0.49Li₂MnO₃·0.51LiNi_0.37Co_0.24Mn_0.39O₂ (Toda Kogyo Corp. HE5050) adopted the domain model. The best spectral response in terms of signal-to-noise, was obtained using the 785 nm laser excitation rather than lower wavelengths ^[77]. On charge, the C2/m phase was delithiated at the same rate as the R-3m phase, while on discharge the monoclinic component was reformed latter.

---

3.1.4. LiNi_0.8Co_0.15Al_0.05O₂ (NCA)

Aluminum-doped Ni-rich layered structure, i.e., LiNi_0.8Co_0.15Al_0.05O₂ (named NCA) is considered as a promising cathode material for high-power batteries with a high specific discharge capacity of ~180 mAh·g⁻¹ at 1C-rate and a superior thermal stability ^[78]. Increase of the Li⁺ ion diffusivity in the layered R-3m lattice is due to the aluminum substitution which enlarges the c-axis parameter ^[79]. In situ Raman microscopy was investigated to evaluate the kinetic features of Li extraction/insertion of individual NCA particles ^[24], the SOC distribution for deep discharge regime ^[47], and the NCA electrode degradation upon long term cycling ^[11]. Using a sealed spectro-electrochemical cell, in situ monitoring of the SOC of NCA composite electrodes (84 wt% NCA, 4 wt% carbon black, 4 wt% SFG-6 graphite and 8 wt% PVdF) showed that the rate of charge and discharge of a selected particle is dependent of the time of the process and location in the electrode. In situ Raman spectra collected using the 632 nm laser line are dominated by a group of bands in the spectral range 470–580 cm⁻¹ characteristic of the Ni–O bonds in NCA oxides. It is assumed that the integrated peak ratio I₄₇₅/I₅₅₃ is proportional to the amount of lithium in one particle, which allows the measure of SOC in individual particle. However, the plot of I₄₇₅/I₅₅₃ versus electrode potential does not show symmetric behavior during charge-discharge cycle, due to the non-uniform particle distribution ^[24].

Performing Raman microscopy mapping of NCA cathode material in a pouch-type lithium-ion cell, Kostecki et al. ^[47] showed that the SOC distribution of particles was non-homogeneous despite the cycling at deep discharge regime. The NCA electrode degradation upon long term cycling was also studied by the same group ^[11]. Raman image was collected on an area 52 μm × 75 μm using the red excitation source (632.8 nm) with a spot of ~1.2 μm of power adjusted at 0.1 mW exhibiting the Raman features of NCA (peaks at 475 and 554 cm⁻¹) and carbon (D- and G-band at ~1350 and ~1600 cm⁻¹, respectively (Figure 9a). As shown in Figure 9b, the Raman image displays different colored pixels, red, blue, and green corresponding to the integrated band intensities of NCA, and D, G carbon bands of single spectrum, respectively. Figure 9c shows that the NCA spectral response, the intensity band ratio I₄₇₅/I₅₅₄ and the peak FWHM are position dependent. These results evidence the spectral distortion due to the variation of the excited electronic states. Similar results obtained from mapping of Li_1−x(Ni_yCo_zAl_1−y−z)O₂ maintained at different SOCs were explained in terms of a link between the variation of composition and the structural integrity at the single particle level ^[50].

---

3.2. Spinel-type compounds

---

3.2.1. LiMn₂O₄ (LMO)

Among the three-dimensional frameworks, spinel LiMn₂O₄, is an interesting cathode material as it is a low-cost material with low environmental impact. It generally exhibits reduced volume expansion upon lithium extraction/insertion reaction. LiMn₂O₄ grows in the cubic structure (Fd3m space group), in which cations occupy octahedral (Mn: 16d) and tetrahedral (Li: 8a) sites in the cubic close-packed array of oxygens occupying 32e Wyckoff positions. Extraction of Li ions from Li_1−xMn₂O₄ (0 < x < 1) occurs with two voltage plateaus at ca. 3.95 and 4.15 V vs. Li⁺/Li in between the LiMn₂O₄/Li_0.5Mn₂O₄/λ-MnO₂ end phases that correspond to the oxidation of Mn³⁺ to Mn^{4+ [80]}. Li_1−xMn₂O₄ is a stable phase: the removal of Li from Li_1−xMn₂O₄ (oxidation process) occurs with a gradual decrease of the unit cell volume through a two-step redox reaction at the two voltage plateaus. However, the identification of the phases during Li extraction from LiMn₂O₄ to λ-MnO₂ has been debated in several publications involving two- or three-phase structural models ^{[81,82,83,84,85,86]}.

A complete factor group analysis of the vibrational modes of the ideal AB₂O₄ spinel structure was given by White and DeAngelis ^[87] and revisited by Ammundsen et al. ^[88]. In the O_h⁷ symmetry there are five Raman active modes represented by:

${\Gamma _{{\rm{LiM}}{{\rm{n}}_{\rm{2}}}{{\rm{O}}_{\rm{4}}}}} = {A_{1{\rm{g}}}} + {E_{\rm{g}}} + 3{T_{2{\rm{g}}}}$

(3)

which decomposes in A_1g + E_g + 2T_2g as internal modes and 1T_2g as an external (translational) mode. The experimental Raman spectrum displays 6 peaks distributed as follows: (ⅰ) the low-frequency peak at ca. 162 cm⁻¹ is the T_2g(1) translational mode that derives from the Li sublattice vibration, (ⅱ) the E_g and T_2g(2) modes are observed at 434 et 455 cm⁻¹, respectively, (ⅲ) due to the presence of mixed valence state of Mn (Mn^3.5+) in LiMn₂O₄, the A_1g mode related to the Mn–O symmetric stretching mode is split into two components at 577 and 625 cm⁻¹ attributed to the vibrations ν(Mn^+Ⅳ–O) and ν(Mn^+Ⅲ–O), respectively, (ⅳ) the high-frequency band at 654 cm⁻¹ is the T_2g(3) mode predominantly characterized by large oxygen displacements.

Investigations of the phase evolution of LiMn₂O₄ during charge–discharge process have been reported by several groups using in situ Raman spectroscopy ^{[89,90,91,92,93,94]}. The main spectral fingerprints of the spectra of LiMn₂O₄ and λ-MnO₂ consist in intense peaks at 593 and 628 cm^-1. Kanoh et al. ^[89] studied the electro-insertion of Li⁺ in λ-MnO₂ using in situ Raman spectroscopy, and demonstrated that the logarithm of the intensity ratio log(I₅₉₃/I₆₂₈) is an accurate measure of the variation of the lithium content, x, in Li_1−xMn₂O₄ spinel lattice. In their early work, Huang and Frech ^[90] investigated the phase diagram of stoichiometric Li_1−xMn₂O₄ and non-stoichiometry Li_1.1−xMn₂O₄ cathode materials. Raman spectra of the former electrode (0.1 < 1 − x < 1.0) showed a single-phase followed by a two-phase process, and finally by another single-phase reaction, whereas the latter electrode displayed the presence of λ-MnO₂. Modifications of the vibrational properties of a single microparticle of Li_1−xMn₂O₄ were monitored in situ via simultaneous Raman spectroscopy and cyclic voltammetry. In the potential range 4.021–4.118 V, a markedly asymmetric band centered at ca. 600 cm⁻¹ are attributed to the T₂ and A₁ modes of the Li_0.5Mn₂O₄ (half-charged) phase, whereas two prominent bands located at 495 and 588 cm⁻¹ are due to the T_2g(2) and A_1g modes of λ-MnO₂ (full charged), respectively, for electrode voltage higher than 4.131 V ^[92]. Shi et al. ^[93] reported similar Raman experiments on LiMn₂O₄ single crystal microelectrode and showed that the SOC derived from optical data displays two well-defined steps up to fully oxidized material. Anzue et al. ^[93] studied the effect of the excited laser lines of 647.1 nm (1.97 eV), 514.5 nm (2.48 eV) and 457.9 nm (2.71 eV) on in situ Raman spectra of Li_1−xMn₂O₄. They found a resonance enhancement effect for x = 1 (λ-MnO₂). This energy dependence attributed to the variation of the bandgap as a function of the Li content. Using a confocal Raman microscopy method, Slautin et al. ^[51] investigated the degradation paths in LiMn₂O₄ electrodes. Upon cycling, it is shown that the Mn₃O₄ phase is formed and dissolved that induces a local disturbance of the lithiation state.

Figure 10 shows the in-situ Raman spectra recorded using the 532 nm laser excitation line of Li_1−xMn₂O₄ electrode during the charge at C/20 rate as a function of the electrode potential in the range 3.2–4.45 V vs. Li⁺/Li ^[91]. The evolution of the frequency shift of Raman peaks that describes the Li_1−xMn₂O₄ phase diagram in the composition range 0 ≤ x ≤ 0.88 is presented in Figure 11. Three distinct single phases are identified from the spectral features during the charge process: (ⅰ) the phase of the initial compound LiMn₂O₄ (phase 1) is maintained up to x = 0.10, (ⅱ) this phase coexists with an intermediate phase (phase 2) up to x = 0.6, (ⅱ) above which a single intermediate phase (phase 2) is identified in a short range of composition 0.6 < x < 0.66. (ⅲ) a two-phase system between phase 2 and phase 3 is then observed in the range 0.7 ≤ x ≤ 0.8, (ⅳ) the single Li-poor phase (phase 3) is observed for x > 0.80. Note that the intermediate phase 2 has been previously reported from X-ray diffraction ^[86] and by Raman spectroscopy ^[88,91] analysis. Ammundsen et al. ^[88] identified this phase as an ordered structure Li_1−xMn₂O₄ (0.5 ≤ x ≤ 0.7) with lower symmetry (space group F-43m, point group T_d²), due to a distortion of the initial cubic lattice and vacancies in Li tetrahedral sites. According to White and DeAngelis, the ordered cubic spinel (Ⅲ) displays 12 Raman active modes represented by $\Gamma$_Li0.5Mn2O4 = 4A₁ + 3E + 6T₂. It is worthy to make the following remarks: (ⅰ) using neutron and ex situ X-ray diffraction studies, the intermediate phase was only observed by Liu et al. ^[82] due to the difficulty to distinguish bi-phased domains at low C-rate (<C/10). On another hand, Raman spectroscopy is predominantly a weak penetrating probe, which investigates the delithiated phases formed at the surface of particles. (ⅱ) It seems that the phase diagram in the 4-volt region is sensitive to the Li stoichiometry; as shown by in situ X-ray diffraction ^[95] lithium-deficient Li_0.99Mn₂O₄ and lithium-rich Li_1.04Mn_1.93O₄ show three cubic phases with the existence of two-phase domains due to the overlap of the adjacent phase.

Micro-Raman mapping was used to investigate the local structure of Ce-doped LiMn₂O₄ as shown in Figure 12a with the Raman spectra of Li–Mn–O components (Figure 12b). By probing the A_1g mode shift, spectral components observed in Figure 12a are as follows: (ⅰ) the largest area (blue regions) represents the pristine spinel (628 cm⁻¹), (ⅱ) some spots originate from the monoclinic Li₂MnO₃ phase (616 cm⁻¹, white spots), (ⅲ) a small fraction of distorted lattice spinel is shown by green spots (646 cm⁻¹) and (ⅳ) a small fraction of signal comes from CeO₂ particles ^[96].

---

3.2.2. LiNi_0.5Mn_1.5O₄ (LNM)

LiNi_0.5Mn_1.5O₄ spinel (denoted LMN hereafter) is considered to be a promising high-voltage (4.7 V vs. Li⁺/Li) cathode materials for powering hybrid electrical vehicles (HEVs) and plug-in hybrid electric vehicles (PHEVs). However, this material suffers from the presence of Jahn-Teller Mn³⁺ ions, cation ordering, and oxygen vacancies, which affect its high-rate performance. Recent investigations showed that the partial substitution of Ni and Mn by cations, such as Cr, Co, Fe, Nb, Mg, Zn, etc. ^{[97,98,99,100,101,102,103]} stabilizes the disordered LNM phase (Fd-3m space group) and eliminates the impurities of LiNiO_x, leading to electrochemical improvements. Typical Raman spectra of pristine and doped LNM electrodes are presented in Figure 13 with attribution of the spectral features ^[103].

The in-situ XRD studies on the phase evolution of LNM cathodes during cycling reveal that the doped materials have identical phase diagrams showing three cubic phases and the patterns reveal a decrease of the amount of Mn³⁺ Jahn-Teller ions leading to an increase of the rate capability ^[101]. In-situ Raman spectroscopy of doped LiNi_0.5Mn_1.5O₄ spinel cathodes was employed to investigate the valence state variations of nickel and manganese, as well as the local structure change during galvanostatic charge–discharge ^[102,103]. The early report of the in-situ Raman spectra of LiNi_xMn_2−xO₄ (0 < x < 0.5) thin films collected in the potential range 3.4–5.0 V vs. Li⁺/Li showed a Raman band located at 540 cm⁻¹ appearing at ~4.7 V, assigned to the Ni⁴⁺–O bond. It was suggested that, upon cycling, the redox reactions Ni^2+/3+ and Ni^3+/4+ occurred in the potential range 4.4–4.7 V and 4.7–5.0 V, respectively ^[102].

Figure 14 shows the Raman spectra of Cr- and Co-doped samples collected between 3.5 and 4.9 V ^[103]. In the wavenumber range of 100–800 cm⁻¹ the Raman spectra contain the bands of the pristine LNM with eleven features: (ⅰ) the bands at 162 and 220 cm⁻¹ are attributed to the translation mode of molecular vibration, (ⅱ) the 486,593 and 639 cm⁻¹ bands are assigned to the stretching mode of Mn–O bond, (ⅲ) the vibration modes at 408,498,528 and 613 cm⁻¹ originated from the Ni–O bond. In addition, a new band is introduced by the doping at 672 cm⁻¹ and is attributed to A_1g mode of Cr³⁺–O/Co³⁺–O. Near the end of charge process at ca. 4.80 V, several new bands are detected among which (ⅰ) the T_2g(T) band at 170 cm⁻¹ is attributed to the translation mode, and (ⅱ) the T_2g band at 538 cm⁻¹ is due to the presence of Ni⁴⁺–O bond in the delithiated lattice. These Raman bands are clearly observed at V ≥ 4.78 (x ~ 0.32) and V ≥ 4.82 (x ~ 0.28) for Cr- and Co-doped LMN, respectively. The high-wavenumber band at 672 cm⁻¹ assigned to Cr³⁺–O/Co³⁺–O vibration remains fixed during cycling, which proves that the valence state of Cr and Co ions is invariant ^[103].

---

3.3. Vanadium pentoxide (V₂O₅)

V₂O₅ is an attractive electrode material which can accommodate 3 Li per mole of oxide to deliver a specific capacity of 450 mAh·g⁻¹ within the potential range 1.5–4.5 V vs. Li⁺/Li ^[104]. Vanadium pentoxide, V₂O₅, with high-valent V⁵⁺ cations, are layered structures which consist of edge and corner-sharing VO₅ square pyramids attached by weak chemical bonds favourable to Li insertion process. The phase diagram of Li_xV₂O₅ is rather complex with reversible and irreversible phases across the composition 0 < x < 3. Over the composition 0 ≤ x ≤ 3, Li_xV₂O₅ undergoes the phase transitions with the corresponding potential plateaus α → ε, ε → δ and δ → γ at 3.4, 3.2 and 2.3, respectively. The γ-phase is irreversibly transformed into the ω-phase at the potential 2.05 V (x > 2). From XRD patterns, it was shown that, upon lithiation from α to to ε-phase, the lattice parameter, a, decreases continuously, while the lattice parameter b remains almost unchanged. In situ Raman spectra of Li intercalated V₂O₅ have been measured on crystal ^[19,30,105] and thin films ^[106,107] to characterize structural changes in Li_xV₂O₅ according to different C-rates, but also to study the abnormal stress modification upon phase transitions ^[107,108]. The first in situ Raman spectrometry studies of electrochemically lithiated and delithiated V₂O₅ materials were reported by Frech et al. ^[10,30,105]. Findings of Raman patterns of Li_xV₂O₅ electrodes are as follows: (ⅰ) a shift of the low-frequency B_1g/B_3g Raman peak (translational mode) from 144 to 154 cm⁻¹, (ⅱ) a progressive intensity decay of the spectral response in the range 195–715 cm⁻¹ attributed to the reduction of V⁵⁺ to V⁴⁺ generating negative charge carriers, (ⅲ) the subsequent reduction of the optical-skin depth due to the metallic nature of the Li-intercalated V₂O₅ electrode, (ⅳ) the disappearance of the vanadyl stretching mode at 996 cm⁻¹ at electrode potential of ~3.27 V vs. Li⁺/Li, replaced by a new peak at 983 cm⁻¹ observed at ~3.40 V related to the coexistence of the α- and ε-Li_xV₂O₅ phases, (ⅴ) the new peak at 983 cm⁻¹ shifts down to 972 cm⁻¹ for further degree of Li insertion corresponding to the ε-Li_xV₂O₅/δ-Li_xV₂O₅ two-phase system, (ⅵ) finally, the γ-Li_xV₂O₅ phase is characterized by a peak at 975 cm⁻¹ and the disappearance of the peaks at 196,302,484 and 700 cm⁻¹ due to a loss of crystallinity in the γ-phase. Jung et al. ^[108] reported that the Raman patterns are not detected in the γ → ω phase transformation suggesting that the ω-phase has a quasi-amorphous structure. Studies of Li insertion in Li_xV₂O₅ film, suggested that the shift of the low-frequency peak from 144 to 154 cm⁻¹ is due to the increase in the restoring force consequently to the tensile stress in the Li_xV₂O₅ film ^[109]. This assumption was confirmed by a combined approach (micro-Raman and optical stress sensor) correlating the Raman shift of B_1g + B_3g mixed modes with the more tensile stress as a function of the Li content in the Li_xV₂O₅ electrode ^[108]. The initial V₂O₅ orthorhombic symmetry (Pmmn space group) is kept upon Li intercalation in the Li_xV₂O₅ film (0 ≤ x ≤ 0.94) due to a solid-solution behavior leading to the typical Raman signature of the ε-Li_0.94V₂O₅ phase. During electrode charging, the local structure is completely recovered, the reason for the excellent electrochemical reversibility of V₂O₅ thinfilms prepared by atomic layer deposition (ALD).

The changes of the Raman spectra during the discharge process (lithiation) depend on the V₂O₅ synthesis technique ^[108]. In the works by the Pereira–Ramos's group ^{[106,110,111,112]} the Raman spectra differs for ALD samples, for rf-magnetron sputtering films and for sol-gel powders. Baddour-Hadjean et al. assumed that the decrease in the intensity of the low-frequency B_1g/B_3g Raman active mode located at 144 cm⁻¹ reflects the structural disorder occurring upon V₂O₅ lithiation ^[113]. Therefore, the peak height I₁₄₄ variation was correlated with the Li content in the Li_xV₂O₅ cathode and the dependence of the induced disorder during cycling with C-rates. Jung et al. ^[107] analyzed the charge rate dependent stress and structure changes in V₂O₅ electrode synthesized by ALD using an in-situ Raman micro-spectrometer equipped with a 532 green laser combined with a stress monitor. As shown in Figure 15, an exponential dependence of I₁₄₄ on the degree of Li intercalation is observed in the composition range 0 < x < 0.2. When the cathode is discharged at high C-rate (>C/4), the slope of ln(I₁₄₄) increases, which reveals an increase of the structural disorder in Li_xV₂O₅.

---

3.4. Manganese dioxide (MnO₂)

Manganese dioxide, MnO₂ has been extensively studied as an active material for Leclanché dry cells, lithium-ion batteries and pseudocapacitors. The most popular phases used in energy-storage systems are γ-MnO₂ (pyrolusite/Ramsdellite intergrowth), α-MnO₂ (hollandite or cryptomelane phase) and δ-MnO₂ (layered birnessite phase). The crystallization in one of these different phases is determined by the synthesis conditions ^[114].

The charge storage mechanism (Li⁺ ions incorporation) of a pseudocapacitive α-MnO₂ electrode was probed using in operando Raman spectroscopy. Raman spectra were obtained using a 514 nm excitation source and a confocal slit of 5 μm. the MnO₂ electrode was tested with different cations in the electrolytes (Li⁺, Na⁺ and K⁺) in order to validate the best electrolyte of the pseudocapacitor ^[115]. Four major active Raman bands labelled ν₁ to ν₄ located at ca. 640,577,515 and 390 cm⁻¹, respectively, were considered for this purpose, since they describe the intrinsic vibration modes of pristine MnO₂. According to Julien et al. ^[116,117], the ν₁ band (620–650 cm⁻¹) is assigned to the symmetric stretching vibration of Mn–O bond in the MnO₆ octahedral environment; the ν₂ band (570–590 cm⁻¹) is assigned to the Mn–O vibration along the chains of the MnO₂ network, whereas the band at 390 cm⁻¹ originates from the Mn–O bending vibrations. Based on the monoclinic structure (C2/m symmetry) of Li-intercalated MnO₂, the irreducible representations of the Raman active modes are classified as:

${\Gamma _{{\rm{LiMn}}{{\rm{O}}_{\rm{2}}}}} = 10{A_{\rm{g}}} + 8{B_{\rm{g}}}$

(4)

Figure 16 shows the in-situ Raman spectra evolution of MnO₂ film cycled within pseudocapacitive conditions in 2 mol·L⁻¹ LiNO₃ aqueous electrolyte between 0.7, 0 and 0.7 V vs. Ag/AgCl. Changes in the Raman spectrum are as follows. (ⅰ) At the beginning of the cycle (0.7 V), water molecules fill the interlayer spacing with Mn in the highest oxidation state (~Mn⁴⁺). As the cathodic reaction proceeds (0.7 V to 0.0 V), the position of the ν₁ peak shows a red shift while the ν₂ band position underwent a blue shift. As pointed out by Julien et al. ^[118] the band shifts imply a reduction of the interlayer spacing upon the cathodic process due to the replacement of Li⁺ ions for water molecules. (ⅱ) At 0 V, the spectrum is close to that of Li_1.0MnO₂ ^[119]. While the band frequencies ν₁ (616 cm⁻¹), ν₂ (586 cm⁻¹) and ν₃ (480 cm⁻¹) match well with spectral features of Li_1.0MnO₂, the weak ν₄ is observed at 402 cm⁻¹ instead of 420 cm⁻¹, which could be due the lower amount of incorporated Li⁺ ions. (ⅲ) The ν₁ and ν₂ bands become broadened due to the structural distortion induced by the increase of Mn³⁺ Jahn-Teller ions, (ⅳ) the evolution of the peak intensity ratio I(ν₂)/I(ν₁) (polarizability change) is attributed to the exchange of neutral H₂O by charged Li⁺ cations and (ⅴ) during the cathodic process, the weak band ν₃ displays a doublet, while the ν₄ band shows reverse behavior ^[120].

Note that the reversible expansion and shrinkage in lattice spacing during charge transfer at manganese sites upon reduction/oxidation of MnO₂ demonstrated by the Raman spectroscopy is a direct proof that the pseudocapacitance involves intercalation or insertion of cations into the bulk of the oxide structure and is not limited to only the surface in contact with the electrolyte, as it was thought in the early stage of the study of the charge storage mechanism in this material.

The origin of high stability of the birnessite δ-MnO₂ pseudo-capacitive electrode synthesized by electroplating was elucidated using operando Raman spectroscopy at different states of charge/discharge. In situ spectra were collected using a green laser (λ = 514.5 nm) and a long working distance of 50× objective lens. The electrochemical tests of the electrode were carried out galvanostatically between 0 and 1 V vs. Ag/AgCl at a 60 A·g⁻¹ current density in solution containing 1 mol·L⁻¹ Na₂SO₄ as electrolyte ^[121]. During charge/discharge cycling, three major Raman bands ν₁ (677 cm⁻¹), ν₂ (589 cm⁻¹) and ν₃ (507 cm⁻¹) show frequency shifts due to the structural evolution. For increasing applied potential 0.0 → 1.0 V (charge process) the ν₂ band at 589 cm⁻¹ shifts down to 572 cm⁻¹, which corresponds to the expansion of the interplanar spacing due to the modification of the oxidation state of Mn⁴⁺ occurring for extraction of Na⁺ ions and incorporation of water from the solution. During charge, the new peak ν₁ (677 cm⁻¹) is assigned to an intermediate Mn–O oxide state because of the coexistence of Mn⁴⁺ and Mn³⁺ cations. The reverse mechanism occurs during discharge (Na⁺ ions insertion). Moreover, theoretical calculations also demonstrated that the incorporation of Na⁺ into δ-MnO₂ is more energetically favorable than H⁺ at all available sites of the interlayer space.

---

3.5. LiFePO₄ olivine

In the last decade, LiFePO₄ (LFP), member of the family of AB^IIXO₄ olivine orthophosphates, has become an attractive positive electrode material for Li-ion batteries because their low cost and excellent thermal stability ^[122]. According to the group factor analysis of the LFP structure with Pnma space group (D_2h¹⁶ point group), the 36 Raman active modes at the centre of Brillouin zone are classified as:

${\Gamma _{{\rm{LiFeP}}{{\rm{O}}_{\rm{4}}}}} = 11{A_{\rm{g}}} + 7{B_{1{\rm{g}}}} + 11{B_{2{\rm{g}}}} + 7{B_{3{\rm{g}}}}$

(5)

Raman spectrum of LFP includes: (ⅰ) the high-frequency bands in the region 908–1126 cm⁻¹ that are stretching modes of P–O bonds, (ⅱ) vibrations in the medium-frequency range between 487 and 691 cm⁻¹ that are O–P–O bending internal modes of the PO₄³⁻ anion, (ⅲ) external modes in the low-frequency region between 175 and 335 cm⁻¹ and (ⅳ) the four peaks between 402 and 445 cm⁻¹ correspond to the Li cage modes in octahedral coordination with O²⁻ anions ^[123]. Wu et al. ^[124] studied the evolution with x of in situ Raman spectra of Li_xFePO₄ samples (non-carbon coated) with different morphologies, i.e., nanoparticles and bulk particles, during galvanostatic charge/discharge process. The phase change from LiFePO₄ to FePO₄ is characterized by the shifts in the internal modes (stretching region), the appearance of peaks at 175 and 244 cm⁻¹, the intensity growth of external modes and the peak broadening of near 960 cm⁻¹. The symmetric stretch (ν₁ at 1102 cm⁻¹) is displaced to lower frequencies and splits in two bands at 1093 and 1091 cm⁻¹. Among the antisymmetric stretchs (ν₃), the peak at 1023 cm⁻¹ disappears in the spectrum of FePO₄, while a red shift is observed for bands at 1011 and 978 cm⁻¹; the symmetric stretch (ν₁) blue-shifts from 953 to 975 cm⁻¹. Finally, a new peak at 908 cm⁻¹ comes from the red shift of an asymmetry stretch. The largest LFP particles show incomplete delithiation, which could be due to the effect of anti-site defects on the surface preventing Li⁺ ion motion in the 1D channels as evidenced by TEM experiments ^[125]. Siddique et al. ^[126] have combined two in operando techniques, micro-Raman spectroscopy and X‐ray diffraction to study the Li_1−xFePO₄ phase change in samples with distinct length scales, i.e., particle level scale (≈1 μm) and macroscopic scale (≈several cm) using LiPF₆ in EC + DMC solution. In addition to the symmetric stretching mode (ν₁) of PO₄³⁻ anions, the Raman spectrum displays the characteristic bands from the electrolyte: (ⅰ) the peaks at 915 and 895 cm⁻¹ are assigned to C–O stretchs of carbonate groups in non-solvated DMC and EC, respectively, which shift to 932 and 904 cm⁻¹ in solvated solution, (ⅱ) the peak at 714 cm⁻¹ is attributed to the symmetric C = O ring deformation in EC, which moves to 720 cm⁻¹ in solvated EC. In contrast with ex situ measurements ^[127], the in-situ Raman of the delithiated LFP electrode was able to display conclusive interpretation of the peak profile at 950 cm⁻¹ peak nor new bands in the range 910–1050 cm⁻¹. In particular, from the spectroscopic study of ion-solvent interactions, it was shown that the intensity of the C–O stretch is a linear function of the salt concentration that leads a blue shift in solvated products. The relative intensity of solvation, I_r, is expressed as ^[128]:

${I_{\rm{r}}} = {I_{\rm{s}}}/({I_{\rm{s}}} + {I_{\rm{n}}})$

(6)

where I_s is the intensity of solvated molecule and I_n the intensity of bulk non-solvated molecule. Therefore, the variation of the Li⁺ concentration at the LFP particle surface, C_Li, depends on the electrode potential and C-rate. C_Li decreases at the beginning of the charge and reached a minimum for the potential of 3.45 V (potential plateau), then C_Li increases till the charge at 4.0 V.

---

3.6. Li₃V₂(PO₄)₃ (LVP)

Polyphosphate material such as α-Li₃V₂(PO₄)₃ (LVP) has a complex three-dimensional network with a monoclinic structure that provides mobility for three Li⁺ ions, providing a high specific capacity of 197 mAh·g⁻¹. According to the group factor analysis of the LVP structure with P2₁/n space group (C_2h point group), there are 120 Raman active modes at the centre of Brillouin zone, which are classified as:

${\Gamma _{{\rm{ \mathsf{ α} - L}}{{\rm{i}}_{\rm{3}}}{{\rm{V}}_{\rm{2}}}{{\left( {{\rm{P}}{{\rm{O}}_{\rm{4}}}} \right)}_{\rm{3}}}}} = 60{A_{\rm{g}}} + 60{B_{\rm{g}}}$

(7)

Only 19 vibrational modes are observed in the unpolarized Raman spectrum located at ca. 1078, 1059, 1030, 1009,975,652,604,560,505,454,430,375,349,290,255,224,169,134 and 118 cm⁻¹. Burba and Frech ^[129] studied the vibration spectra of electrochemically delithiated Li₃V₂(PO₄)₃ powders (0 ≤ x ≤ 3) and reported the high sensitivity of the (PO₄)³⁻ modes with extraction of Li⁺ ions and change of the vanadium oxidation state. After the first discharge process, Yin et al. ^[130] found the (PO₄)³⁻ band broadening that is assigned to some local disorder in the relithiated lattice. The thermal phase stability of α-Li₃V₂(PO₄)₃ was investigated by in situ Raman spectroscopy by varying the irradiation power intensity from 5 to ~600 kW·cm⁻² (using a 1.3 μm spot diameter in air). The increase of temperature resulted in phase transitions in the sequence ^[8]:

${\rm{ \mathsf{ α} - L}}{{\rm{i}}_{\rm{3}}}{{\rm{V}}_{\rm{2}}}{{\rm{(P}}{{\rm{O}}_{\rm{4}}}{\rm{)}}_{\rm{3}}} \to {\rm{ \mathsf{ β} }} \to {\rm{ \mathsf{ γ} }} \to {\rm{ \mathsf{ α} - LiVOP}}{{\rm{O}}_{\rm{4}}}$

(8)

where the α-LiVOPO₄ material is the oxidation phase under air atmosphere.

---

3.7. β-AgVO₃ (SVO)

Silver vanadium oxide (SVO) has an important technological application in medical power sources as cathode material of lithium batteries for implantable cardio-pacemakers and defibrillators ^[131]. The structural changes and of Li inserted β-AgVO₃ nanoparticle were studied by in situ Raman spectroscopy using the blue excitation line (λ = 488 nm) combined with atomistic simulation to determine the lithium migration pathways ^[132]. Upon Li insertion, the reduction sites in the β-AgVO₃ lattice can be divided into three steps that correspond to the potential plateau in the discharge profile: (ⅰ) the most favorable reduction on sites V1/V4 occurs at higher potential 3.55–3.05 V, (ⅱ) Ag2/Ag3 are reduced and substituted at medium potential 3.05–2.55 V and (ⅲ) finally, Li⁺ ion reduces V2/V3/Ag4 and further reduces V1/V4 in the potential range 2.55–1.75 V vs. Li⁺/Li. In the first step, in which vanadium is reduced from V⁵⁺ to V⁴⁺, the intensities of Raman peaks at 266,530,772,808,847 and 886 cm⁻¹ are enhanced at 3.35 V and are further weakened at 3.05 V; the stretching vibration at 513 cm⁻¹ of V–O bonds in pyramids exhibits a large blue shift to 530 cm⁻¹, while the asymmetric bending mode of (VO₄)³⁻ pyramids at 278 cm⁻¹ has a red-shift to 266 cm⁻¹ related to the distortion of V1/V4 pyramids upon Li introduction in the host lattice. In the second step, the Raman spectrum becomes strongly attenuated due to the loss of resonance and to the increase of electrical conductivity that reduces the optical-skin depth. However, a blue shift of the Raman band at 807 cm⁻¹ (to 811 cm⁻¹) due to the shortened V–O bond is observable. In the third step, no significant modification of the Raman features is detected in the potential range 2.55–2.05 V, while the band at 771 cm⁻¹ shifts to 775 cm⁻¹ between 1.85 and 1.75 V and the peak at 756 cm⁻¹ splits into two components at 750 and 761 cm⁻¹. Moreover, the reduction of vanadium at the V1/V4 site occurring at 1.95 V is characterized by the spit of the band at 730 cm⁻¹ in two well-resolved peaks at 723 and 734 cm⁻¹. These changes of the vibrational modes indicate a significant lattice distortion of β-AgVO₃ with Li insertion.

---

4. Negative electrodes

---

4.1. Carbonaceous materials

Numerous works prior 1998 have been devoted to the studies of local structure of carbonaceous materials used as negative (anodes) materials, for Li-ion batteries. The term carbonaceous include various forms of carbon materials from highly disordered networks (turbostratic graphite, glassy carbons), to defect-free well crystallized crystals, i.e., highly oriented pyrolytic graphite (HOPG) characterized by its L_c and L_a parameters indicating dimensions perpendicular and parallel to the basal plane. A relation that is a measure of the structural order in the graphene sheets of carbonaceous materials has been established between the correlation length L_a (expressed in nm) and the derived spectral parameter, i.e., intensity ratio of Raman carbon-like bands I(D)/I(G) ^[133]:

${L_a} = \frac{{4.382}}{{I\left( D \right)/I\left( G \right)}}$

(9)

In this section, we will not expose in detail the results of in situ Raman but will only give some hints about the latest research.

---

4.1.1. Graphite intercalation compounds (GICs)

The layered structure of graphite consists of regular stacked graphene sheets arranged in hexagonal AB stacking sequence of P6₃/mmc space group (D_2h⁴ point group), of which the factor group analysis gives rise to two Raman active modes for q = 0:

${\Gamma _{{\rm{graphite}}}} = 2{E_{2{\rm{g}}}}$

(10)

The spectrum of graphite exhibits a low-frequency prominent peak E_2g1 mode at 42 cm⁻¹ (rigid-layer shear mode) that originates from the in-plane sliding of graphene sheets, and the predominant peak at 1580 cm⁻¹ attributed to the E_2g2 mode (in-plane C–C stretching mode) that is usually termed as G band. An additional Raman band at ~1350 cm⁻¹ (D band) appears for polycrystalline graphites and non-graphitic carbons, which is the A_1g breathing mode activated by defects and grain boundaries. Overtone bands are also observed at ~2670 cm⁻¹ (2D) and ~3250 cm⁻¹ (2G) ^[134].

Since the early work of the Japanese's group ^[135], the changes in the local structure of graphite intercalation compounds (GIC) have been widely studied by in situ Raman spectroscopy ^{[25,34,44,136,137,138,139,140,141,142,143,144,145,146,147]}. At the open circuit potential (ca. 3.0 V vs. Li⁺/Li), the Raman spectrum exhibits three bands in the region between 1000–3000 cm⁻¹: a weak D-band, a predominant G-band and a weak 2D-band. The staging process of Li intercalation into HOPG and natural graphite was first reported by Inaba et al. ^[135]. Using in situ Raman spectroscopy, they observed that Li intercalation into graphite proceeds via a series of staged GICs phases, noted by a stage index, n, which represents the number of graphene sheets separating the intercalated layers. It is claimed that Raman spectral changes are correlated with the potential plateaus recorded on the charging reaction as follows: a phase transition from dilute stage-1 to stage-4 followed by the stage-2' (LiC₁₈) → stage-2 (LiC₁₂) and stage-2 (LiC₁₂) → stage-1 (golden LiC₆) transitions. This model will be modified in further investigations of Lonza KS‐44 and KS‐6 graphite materials ^[136]. Panitz et al. ^[44] presented in situ Raman patterns and imaging of electrode made of TIMREX SFG 44 synthetic graphite with PVdF binder, which reveal a new band at ca. 1850 cm⁻¹ at potential negative to 0.18 V vs. Li⁺/Li assigned to the complex between Li ions and decomposition of the aprotic solution. This electrolyte/graphite interaction was later reconsidered by Novák et al. ^[137] in terms of SEI. The so-called dilute stage-1 phase of GICs was deeply studied on KS-44 carbon particles (8–50 μm in diameter) embedded in heat-treated Ni foils ^[138]. Kostecki and McLarnon ^[139] investigated the structural degradation of graphite electrodes by portraying the graphite D/G band ratio of the anode via a 50 μm × 75 μm cross-section Raman image.

During lithiation (discharge process from OCV (ca. 3.0 V) to 0.05 V), modifications of the first- and second-order Raman bands may be classified into four steps: (ⅰ) in the potential range 3.0–0.6 V, the D-band intensity vanishes, (ⅱ) a blue-shift of the G-band from 1580 cm⁻¹ to 1590 cm⁻¹ occurs at ~0.19 V along with a weakening of the 2D(1) band intensity due to the formation of dilute stage-1, (ⅲ) in the potential window 0.19–0.10 V, the G-band splits into the E_2g2(ⅰ) (1575 cm⁻¹) and E_2g2(b) (1601 cm⁻¹) bands, accompanied by a major red-shift of the 2D(2) band due to the formation of stage-2 and stage-1 and (ⅳ) finally, a gradual loss of all distinct Raman peaks occurs below 0.10 V vs. Li⁺/Li with the appearance of a weak band located at ca. 1370 cm⁻¹. The formation of highly conductive GIC at low-stage number is at the origin of the loss of all Raman features ^[148]. Note that the formation of the G-band doublet E_2g2(ⅰ) (1578 cm⁻¹) and E_2g2(b) (1600 cm⁻¹) is also observed in the case of the electrochemical intercalation of tetraethylammonium (Et₄N⁺) and tetrafluoroborate (BF₄⁻) into microcrystalline graphite ^[149]. The use of propylene carbonate (PC) provokes graphite exfoliation characterized by a new Raman band (E-band) at high wavenumbers of 1609 cm⁻¹ between 0.9 and 0.005 V ^[144].

In the reverse reaction (lithium de-intercalation) four stages can be also considered: (ⅰ) a broad G-band at 1592 cm⁻¹ and the weak band at 1370 cm⁻¹ recurs at ca. 0.14 V, (ⅱ) the E_2g2(b) band of the stage-2 GIC is formed at ca. 0.15 V from the blue-shift of the G-band from 1592 to 1598 cm⁻¹, (ⅱ) both the G band doublet, E_2g2(ⅰ) (1573 cm⁻¹) and E_2g2(b) (1601 cm⁻¹) and the 2D(2) band reappear in the potential range 0.17–0.22 V and (ⅳ) the sharp singlet G-band at 1586 cm⁻¹ and the 2D(1) band are reformed coming to the delithiated structure between ca. 0.3–1.5 V. Note that the doublet E_2g2(ⅰ), E_2g2(b) is associated respectively with carbon vibrations in interior graphite layers and in-bounding graphite layers adjacent to the intercalate planes. This splitting is attributed to changes in symmetry at the boundary layer and to the change in electronic states upon Li⁺ ion intercalation at stage n > 2. When intercalated graphite appears as stage-1 and -2, the E_2g2(ⅰ) band vanishes because of the absence of graphite interior layer. By taking into consideration the intensity ratio of the E_2g doublet (R), Solin has proposed a quantitative measure of the intercalation stage index, n, by the following equation ^[150]:

$R = \frac{{{I_i}}}{{{I_b}}} = \frac{{{\sigma _i}}}{{{\sigma _b}}}\frac{{n - 2}}{2}\;\;\;\;\left( {{\rm{n > 2}}} \right)$

(11)

where I_i and I_b are the intensities of the interior E_2g2(i) and bounding E_2g2(b) layer modes, respectively, and σ_i/σ_b is a stage-independent ratio of the Raman scattering cross-section from the interior and bounding layers, equal to 1 for lithium intercalation.

Typical in situ Raman spectro-electrochemical measurements of graphite electrode is presented in Figure 17, where spectral features were taken on different points that show the inhomogeneity of the particle distribution in the electrode ^[143]. Detailed real-time Raman spectra from a single KS-44 graphite microflake electrode determined that the transition between dilute stage-1 and stage-4 of lithiated GIC occurs in the potential range 0.174–0.215 V vs Li⁺/Li ^[141]. Graphite surface disorder was detected by the variation in electrode potential of the appearance of the G-band doublet associated to the formation of stage-4 GIC ^[143]. During lithium intercalation in hydrogen-rich amorphous carbon, lithium-carbon interaction leads to an increase of I_G/I_D and formation of Li–C bands at ~480 and ~700 cm^{−1 [151]}. Raman imaging of synthetic graphite LF-18D particles with an average particle diameter of 22 μm (Chuetsu Graphite Works Co.) showed that the Li does not intercalate in a homogeneous manner; it varies with the article morphology and local conditions ^[142]. In situ Raman experiments evidenced the intercalation effect of both large cation (EMI⁺) and anion (TFSI⁻) in KS44 microcrystalline graphite and https://www.aimspress.com/aimspress-data/aimsmates/2018/4/PICACTIF activated carbon. Intercalation is detrimental for graphite with an increased I_D/I_G ratio, while possible intercalation of activated carbon upon cycling is evidenced by the blue-shift of the D-band ^[146]. The surface region of graphitic negative electrode (microcrystalline graphite) was examined by in situ real-time Raman measurements under potential control during Li⁺-ion insertion and extraction ^[33]. A special attention was paid to the double resonance 2D-band that shifts from 2681 to 2611 cm⁻¹. The change of the electronic structure of the intercalated compound and the C–C bonding of stage 3 and 4 was evidenced by the band shape transformation into a single Lorentzian in the potential range 0.24–0.15 V vs. Li⁺/Li. It was concluded that the Daumas–Hérold model can be applied, where the graphene sheets are flexible and deform around Li intercalated domains.

In situ Raman spectro-electrochemical measurements were also carried out to study the degradation of edge-plane graphite negative-electrodes ^[152], the SEI formation ^[153], the influence of the salt in the electrolyte solution on the structural degradation of graphite ^[154], the surface evolution of a single graphite particle with ethylene carbonate/dimethyl carbonate ^[155], the behavior of graphite electrodes in electrolyte solution containing fluorinated phosphoric esters ^[25], the decrease in the surface crystallinity of graphite negative-electrodes at high potentials in LiPF₆-based electrolyte solution ^[156], the correlations of concentration changes of the electrolyte salt with the surface state of a HOPG graphite electrode, the Li⁺ ion intercalation and deintercalation behaviors of graphitized carbon nanospheres ^[157] and the structure of heat-treated graphene nanoflake-based anodes ^[158].

---

4.1.2. Li insertion in other carbonaceous materials

By contrast to graphite, carbonaceous materials with non-graphitic structure exhibit poor galvanostatic responses with slopping potential profiles that originate from disorder. Coke, carbon black and glassy carbon (GC) deliver low-specific capacity, i.e., 223 mAh·g⁻¹ for GC against 372 mAh·g⁻¹ for graphite. The spectral modifications of the Raman E_2g2 band of mesocarbon microbeads (MCMBs) annealed at 2800 ℃ presents a Li insertion mechanism similar to that of graphite with formation of staged phases ^[159]. Totir and Scherson ^[160] investigated the electrochemical lithium intercalation of embedded KS‐44 graphite and MCF28 carbon microfibers and reported spectral patterns in good agreement with similar materials. The graphene oxide (GO) and reduced graphene oxide (rGO) obtained by scanning the cell potential from 0 to −1 V vs. SCE in an aqueous electrolyte were characterized by in situ Raman spectro-electrochemical study, which showed a clear shift of the G-band from 1610 to 1585 cm⁻¹ for GO and rGO, respectively ^[161]. In situ Raman and nuclear magnetic resonance measurements were performed to study the trapped Li ions in the SEI of rGO ^[162]. It is shown that the graphitic G-band weakens and vanishes fast at potential > 0.3 V without staged GIC and suggested that the monolithic defective sites are randomly accommodated by Li⁺ ions, which cover the rGO surface and cause the G-band disappearance.

---

4.2. Lithium titanate Li₄Ti₅O₁₂ (LTO)

Li₄Ti₅O₁₂ (LTO) is an anode material being a zero-strain network with high thermal stability. It crystallizes in the cubic structure with Fd3m space group (O_h⁷ point group) and can be expressed as Li[Li_1/3Ti_5/3]O₄ in the spinel notation; thus, the Raman-allowed phonon peaks are represented by species in Eq 3, while factor group analysis predicts 4 Raman-active modes for the lithiated phase Li₂[Li_1/3Ti_5/3]O₄ at q = 0:

$\Gamma = {A_{1{\rm{g}}}} + {E_{\rm{g}}} + 2{T_{\rm{g}}}$

(12)

The Raman spectrum of the pristine material exhibits three intense Raman bands centered at 671,430 and 232 cm⁻¹ along with two weak Raman peaks at 347 and 271 cm⁻¹. The vibration at 671 cm⁻¹ (A_1g mode) is ascribed to the symmetric Ti–O stretch of TiO₆ octahedra, while the peak at 430 cm⁻¹ (E_g mode) originate from the asymmetric Li–O stretch of LiO₄ tetrahedra and the 232 cm⁻¹ band (T_g mode) from δ(Ti–O). The Raman peak at 232 cm⁻¹ and other weak bands are assigned as T_2g mode ^[163]. During discharge, the lithiated Li_1+x[Li_1/3Ti_5/3]O₄ electrode can accommodate x ≈ 0.94 Li with a CV curve showing a very wide potential plateau around 1.55 V vs. Li⁺/Li. Insertion of Li⁺ ions in octahedral 16c empty sites of the Li_1+x[Li_1/3Ti_5/3]O₄ framework and the reduction of Ti ions in TiO₆ octahedra provoke a gradual blue shift of Raman bands at 430 and 232 cm⁻¹, while the high-wavenumber peak at 671 cm⁻¹ is maintained ^[164]. In situ Raman studies show that, at the end of the discharge, the spinel transforms in rock-salt lattice and, as the electronic conductivity increases considerably due to the presence of Ti⁴⁺ and Ti³⁺ ions, the Raman line intensities gradually weaken with x and finally disappear at x ≈ 0.94 due to the decrease of the optical-skin depth ^[165].

---

4.3. Titanium dioxide TiO₂

TiO₂ has been exhaustively investigated due to its attractive properties; it is low cost, thermally stable, environmentally safe, etc., and it is considered as a promising candidate for use in Li-ion batteries. TiO₂ exists in different polymorphic forms: anatase, rutile, brookite, TiO₂-B (bronze), TiO₂-R (ramsdellite), TiO₂-H (hollandite), etc. ^[166]. It appears that the potential profile of the Li//TiO₂ couple is strongly influenced by the structure; as an example, the working voltage of Li//anatase TiO₂ cell is higher than that of Li//rutile TiO₂ cell. Anatase TiO₂ (tetragonal structure, space group I4₁/amd, D_4h¹⁹ point group) can accommodate reversibly x ≈ 0.5 Li that converts Li_0.5TiO₂ to orthorhombic phase (space group Imma, D_2h²⁸ point group) ^[167]. Only nanocrystalline TiO₂ particles (<7 nm in size) are able to host x ≈ 1 Li ^[168]. The lattice dynamics of pristine anatase TiO₂ has been subject to numerous works (see ^[29] and Refs. herein). Its first-order Raman analysis gives rise to 6 active vibrations at q = 0:

${\Gamma _{{\rm{Ti}}{{\rm{O}}_{\rm{2}}}}} = {A_{1{\rm{g}}}} + 3{E_{\rm{g}}} + 2{B_{1{\rm{g}}}}$

(13)

which have been experimentally observed at 635,515,395,195 and 142 cm^{−1 [169]}. The symmetry change from D_4h¹⁴ to D_2h²⁸ induced by elongated equatorial Ti–O bonds in Li_0.5TiO₂ produces 9 Raman active modes represented by:

${\Gamma _{{\rm{L}}{{\rm{i}}_{{\rm{0}}{\rm{.5}}}}{\rm{Ti}}{{\rm{O}}_{\rm{2}}}}} = 3{A_{\rm{g}}} + 3{B_{2{\rm{g}}}} + 3{B_{3{\rm{g}}}}$

(14)

where B_2g and B_3g modes come from the splitting of the E_g mode. Non-stoichiometry and phonon confinement in nanoparticles (>10 nm in size) produce a blue shift and a broadening of the lowest-frequency E_g vibration. Several studies reported in situ Raman spectro-electrochemistry of Li-inserted anatase TiO₂ ^{[170,171,172,173,174,175,176,177]}. Dinh et al. ^[170] combined in situ Raman and in situ transmittance spectra of electrochromic TiO₂ anatase thin films and demonstrated that the colored film exhibit five characteristic bands at 629,531,316,224 and 176 cm⁻¹. Hardwick et al. ^[173] studied three nanosized TiO₂ anatase powders (80, 15 and 8 nm in size) and, using in situ Raman microscopy, followed the increase in electrical conductivity as lithiation proceeds, with a marked band intensity decrease for x > 0.3 in Li_xTiO₂. In situ Raman spectro-electrochemistry was a powerful tool to detect the contamination of anatase TiO₂ in the TiO₂(B) and TiO₂(rutile) polymorphs during Li⁺ insertion/deinsertion ^[176]. In situ Raman spectro-electrochemistry showed that structural modifications during Li insertion in anatase TiO₂ nanoparticles can be understood by a mechanism of Li-poor tetragonal → orthorhombic → Li-rich tetragonal double phase transformation. For Li_0.09TiO₂ (at 1.75 V electrode potential), a blue-shift of the band at 142 cm⁻¹ is observed (Δν = 5 cm⁻¹) followed by the two-phase process characterized by the occurrence of two bands at 165 and 177 cm⁻¹ that corresponds to the potential plateau at 1.4 V. The additional Raman peaks associated with the end-composition Li_0.5TiO₂ are recorded at 625,555,528,355,339, and 315 cm^{−1 [175]}.

---

4.4. Silicon

Silicon is an attractive anode material because of its low cost, its large specific capacity, i.e., 4200 mAh·g⁻¹ when lithiated to Li_4.4Si (or Li₂₂Si₅), and its relatively low redox potential, i.e., 0.3–0.4 V above Li⁺/Li ^[178]. However, the electrochemical performance of Li–Si alloys degrades due to the large volume expansion, i.e., 420% from Si to Li_4.4Si. During discharge the Li–Si alloy is formed and subsequently a transition from crystalline (c-Si) to amorphous (a-Si) silicon occurs. In this context, few researchers have studied the phase transformation using in situ Raman microscopy ^{[179,180,181,182,183]}. One of the earlier in situ Raman investigation of silicon anode was conducted by Holzapfel et al. ^[184]. Unlithiated Si material exhibits the typical first-order (q = 0) Raman peak at 520 cm⁻¹ (optical phonon) of the crystalline phase that significantly decreases at potential below 0.88 V upon Li alloying. For further lithiation, the Si peak vanishes totally and a broad band grows at ca. 480 cm⁻¹ when the discharge reaches 0.09 V, which could indicate the growth of an amorphous Li_xSi phase ^[179]. Nanda et al. ^[180] characterized the surface of the Si/C composite anode at a microscopic level and monitored the alloying/dealloying behavior of Si/C with Li, using in situ Raman microscopy study. As shown in the Raman spectra of silicon as a function of the first discharge between 2.0 and 0.45 V (Figure 18), the Raman peak of Si peak does not change (discharged at a current density of 90 μA·cm⁻²), while at 0.2 V the signal substantially decreases and disappears below 0.1 V due to the conversion of the crystalline Si into an alloyed Li–Si phase.

Long et al. ^[185] examined the effects of p-type (boron) and n-type (phosphorus) dopants on the lithiation of crystalline Si. The transition to amorphous Si associated with the Li insertion was monitored by in situ Raman spectroscopy. A plot of the height of phonon peak shows that the n-type Si exhibits a phonon decay at potential of 0.09 V, against 0.68 V for the p-type Si. This effect is explained in terms of energy states available to the electron associated with the Li⁺ ion. The lithiation of silicon decorated by plasmonic metals, i.e., Ag and Au, was characterized by in situ Raman microscopy and Raman mapping on a section of 20 × 30 μm² using a laser wavelength of 532 nm ^[186]. The structural evolution, the internal strain and stress in the lithiated and delithiated Si nanoparticles were monitored by both in situ X-ray diffraction and in operando Raman spectroscopy ^[182,187]. Upon the formation of Li–Si alloys, the drop of the phonon intensity results first from the amorphization of the c-Si framework and secondly from the decrease of the optical-skin depth. There are no changes of peak position during the first discharge, while a blue shift occurs during delithiation. In contrast, the upshift that originates from compressive stress is followed by the downshift due to tensile stress during the second cycle and so forth. From the experimental Raman shifts, one can estimate the magnitude of the stress using the relation ^[182]:

${\rm{ \mathsf{ σ} = - 230}}\left( {{{\rm{ \mathsf{ ω} }}_{\rm{s}}}{\rm{ - }}{{\rm{ \mathsf{ ω} }}_{\rm{0}}}} \right)$

(15)

where ω₀ and ω_s are the wavenumbers for the relaxed and of the stressed particles, respectively. Yang et al. ^[188] experienced the lithiation behavior of Cu-nanowire/Si nano-particle composite anodes using in situ micro-Raman mapping, showing that the Li–Si reaction does not occur in a homogeneous manner due to disconnected active particles. Recently, Sakaguchi et al. strenuously determine silicon anode properties ^{[189,190,191]}. The deterioration mechanism of Si anode was analyzed using Raman mapping of Li–Si alloying and dealloying reactions ^[189]. Evolution of the Raman shift of the 520 cm⁻¹ line of phosphorous-doped Si revealed that doping monitored the Li–Si alloying reaction and the formation of the c-Li₁₅Si₄ phase ^[190].

---

4.5. Other anode materials

---

4.5.1. Antimony

The active researches of high-performance anode materials have motivated the consideration of metal-containing compounds in the form of intermetallics such as Sb, InSb, Cu₂Sb, AlSb, etc. that make alloys upon lithiation ^[192]. Antimony is capable of 3Li uptake forming an alloy with a theoretical specific capacity of 660 mAh·g⁻¹ based on the Li₃Sb reaction ^[193]. The lithium alloying of carbon-coated antimony microparticles synthesized by a one-pot sol-gel auto-combustion route was analyzed by in situ Raman spectroscopy using a red laser-line (λ = 632.8 nm) as excitation source ^[194].

Using in situ Raman spectro-electrochemistry in the potential between 1.5 and 0.02 V, a comparison with commercially available Sb nanoparticles of same size (ca. 5–50 μm) shows a superiority of sol-gel prepared Sb (Figure 19). The Raman spectrum exhibits the signal due to D- and G-band of the disordered carbon coating and the phonons of antimony at 150 cm⁻¹ (A_1g mode) and 112 cm⁻¹ (E_g mode), which is sensitive to the confinement and strain of particles. During lithiation, the Raman spectrum of Sb is maintained until ~0.9 V vs. Li⁺/Li and then the intensity of phonon vibrations decreases until smearing out at 0.02 V. Some degree of long range ordering is recovered upon delithiation to 1.5 V with the reemergence of the A_1g peak at 152 cm⁻¹.

---

4.5.2. ZnM₂O₄ (M = Mn, Fe)

Among the ternary Zn-based oxides, the spinel ZnMn₂O₄ material reacts with lithium with both conversion and alloying process, which results in a specific capacity 1024 mAh·g⁻¹. In situ Raman spectra of mesoporous ZnMn₂O₄ microspheres were collected in the potential range 2.0–0.01 V vs. Li⁺/Li (Figure 20) ^[195]. The structural changes were monitored by following the evolution of the three Raman bands located at 681 cm⁻¹ (A_1g symmetry involving motion of oxygen in AO₄ tetrahedra), 386 and 324 cm⁻¹ (involving the vibration of BO₆ octahedra) according to the usual notations for compounds of generic form AB₂O₄. When the Mn³⁺ ions are reduced to Mn²⁺ during the lithiation process at 1.21 V, the A_1g band begins to wane and disappears at 0.16 V, while the strength of Raman bands at 386 and 324 cm⁻¹ decreases. At this potential Zn²⁺ and Mn²⁺ are reduced to Zn⁰ and Mn⁰. During the charge process (delithiation) at 1.25 V, a band grows at 694 cm⁻¹ which is characteristic of the asymmetric Mn–O stretch for R-Mn₂O₃ associated with the oxidation of Mn⁰ to Mn³⁺. Above 1.52 V, a band located at 407 cm⁻¹ assigned to the E₂ mode of ZnO is the fingerprint of oxidized Zn⁰. Therefore, the two weak Raman bands at 324 and 386 cm⁻¹ for ZnMn₂O₄ reappear at the end of the charge process.

Cabo-Fernandez et al. ^[196] investigated the delithiation process of carbon-coated ZnFe₂O₄ (ZFO) nanoparticles. Like ZnMn₂O₄, ZFO is a low-cost anode material with a theoretical specific capacity ~1000 mAh·g⁻¹ that reacts with lithium via a conversion/alloying process. The first-order Rama spectrum displays 5 bands located at 647 cm⁻¹ (A_1g), 467 cm⁻¹ (T_2g), 340 cm⁻¹ (T_2g), 246 cm⁻¹ (E_g) and 221 cm⁻¹ (T_2g). The conversion reaction that occurs at ~0.8 V leads the formation of Zn⁰ particles, which enhance the Raman signal of the SEI components. In the electrode potential between 0.80 and 0.69 V, the Raman spectrum is dominated by the reduction products of electrolyte carbonate solvents which form the SEI.

---

5. Solid electrolyte interface (SEI)

The solid electrolyte interface is a chemical layer formed on the anode side, i.e., lithium metal or graphite, that is indispensable for the electrochemical stability of Li-ion batteries using organic (aprotic) electrolyte. A SEI is the result of solvation of Li ions, i.e., electrolyte reduction because of the sufficiently low potential of graphite, resulting in the formation of a film deposited on the electrode surface in the first cycle of charge/discharge. After the second cycle, the SEI suppresses the electrolyte decomposition acting as an energetic barrier between the Fermi level of the graphite and the high occupied molecular orbital (HOMO) of the electrolyte. Analysis of the SEI using in situ Raman spectroscopy has been reported by few research teams ^{[197,198,199,200,201]} only, because the SEI layer is usually very thin so that the Raman signal difficult to be detected. The composition and formation of the SEI were analyzed using copper electrodes after lithium plating. From Raman spectra Schmitz et al. ^[197] have detected semicarbonates such as Li₂CO₃ and Li₂C₂ as SEI components. In situ Raman micro-spectroscopy and Raman mapping reveal that: (ⅰ) Li₂C₂ is not formed at a potential > 0 V vs. Li⁺/Li, (ⅱ) Li₂C₂ is located on the plated lithium and (ⅲ) Li₂CO₃ is homogeneously distributed over the copper sheet. The presence of Li₂C₂ on the metallic Li surface has been considered as a product of laser degradation rather than an actual SEI species ^[202,203]. The SEI formation onto SiO₂-coated Au nanoparticles was studied using in situ surfaced-enhanced Raman spectroscopy (SERS) in lithium-ion battery ^[201]. The effect of the additive vinylene carbonate (VC) in the ethylene carbonate (EC) electrolyte was investigated by monitoring the band intensity of silicon, which exhibits different amorphization rates between bulk and surface. Park et al. ^[204] investigated the mechanism of the electrolyte–electrode interface reaction using in situ Raman spectroscopy during the second charging process. Due to the low permeability of the SEI, there is a variation of ion concentration during charging and discharging a Li-ion battery ^[205]. This change in concentration nearby the separator has been investigated by in situ Raman spectroscopy using ultrafine microfiber probes ^[206,207]. Recently, Yamanaka et al. ^[22,207] performed such measurements with a graphite//LiFePO₄ cell using an electrolyte solution of 1 mol·L⁻¹ LiClO₄ dissolved in ethylene carbonate (EC) and diethyl carbonate (DEC) solution. Intensities of three Raman peaks were analyzed: at 718 and 730 cm⁻¹ (symmetric ring deformation mode of EC and solvated EC with Li⁺ and 935 cm⁻¹ (symmetric mode of the anion ClO₄⁻). The properties of the SEI were modified by adding film-forming additives to the EC:DEC solution. Based on in situ Raman spectro-electrochemistry, the concerted effect of the SEI and the adjacent separator film generates an effective barrier for Li ions ^[208]. The same research's group ^[209] performed a chronoamperometric pretreatment of HOPG anode at 0.9 V vs. lithium applied for 20 h that decrease the resistance charge transfer and capacitance at the surface of graphite. The main result of this pretreatment consists in the suppression of the concentration changes.

---

6. Conclusions

In this review, we have shown that, among the various in situ methods, Raman spectro-electrochemistry is a highly effective technique in intensive field of researches covering all aspects of battery operation. Regarding the structural response of electrode materials, the contribution of both in-situ Raman spectro-electrochemistry and Raman mapping were determining for the issue of real time charge–discharge cycling because they are powerful tool for studying the short-range order (local environment) in lattices. In situ analyses are carried out to track several effects occurring during charging and discharging process in lithium batteries such as phase stability, structural modifications of the electrode materials during insertion/deinsertion reaction, interface evolution, compatibility between materials, kinetics of Li⁺ ions diffusing in the electrodes et electrolytes, electrode degradation and formation of the SEI layer.

As other in operando analytical methods, in situ Raman spectroscopy is used not only to improve the cycle stability of entire batteries, but to provide experimental evidence of new insights of Li⁺-ion insertion/deinsertion mechanism. However, due to the short optical penetration depth, Raman spectroscopy is essentially a surface analysis technique and the spectral response cannot be extended to the bulk of a sample that makes some difference with results obtained from in situ X-ray diffraction when the rate of analysis is not enough slow.

---

Conflict of interest

The authors declare that there is no conflict of interest regarding the publication of this manuscript.

---