Operando characterization of batteries using x-ray absorption spectroscopy: advances at the beamline XAFS at synchrotron Elettra

The beamline XAFS at Elettra is dedicated to XAS. It operates from tender to hard x-rays, namely from 2.4 to 27 keV. Within this energy range, the K-edges of elements from sulfur to silver are accessible and the L-edges of the elements from yttrium to bismuth can be studied, together with those of the lanthanides. The photon source of the beamline is a bending magnet with a critical energy of 3.2 keV and 5.5 keV for Elettra running at 2 GeV and 2.4 GeV, respectively.

Being general purpose, the beamline XAFS fulfills the needs of a wide community of users in many different fields. Examples range from catalysis [27–30] to environmental science [31–35], and from fundamental physics [36] to archaeometry [37].

The beamline XAFS can perform XAS in transmission, fluorescence and converse electron yield modes. Its assets are to perform XAS with an excellent signal-to-noise ratio, stability, versatility and reliability, and a high automation level. The optical scheme (figure 1), including the mask, collimating mirror and double crystal monochromator, is described in [38]. In 2012 the optics of the beamline were completed with the addition of a pair of SiO₂ mirrors. Their aim is to suppress the contribution of higher harmonics when working at energies below 4 keV and therefore to enhance the performance of the beamline in the lower range of operation energies. The chamber described in [39] was installed in the experimental hutch downstream from the exit slits parallel to the monochromatic beam at 6 mrad with respect to the ground. The mirrors were recommissioned and tested. Figure 2 shows the intensity of the I₀ detector while rocking the first crystal of the monochromator around the angle corresponding to the energy of 2400 eV without and with the SiO₂ mirrors placed at ${{\theta}_{c}}=8$ mrad with respect to the direct beam. The peak at around  −570 V in the blue curve indicates that photons of energy corresponding to 7200 eV (2400 eV  ×  3) and diffracted by the Si(1 1 1) monochromator crystals are mixed with those of 2400 eV and counted by the I₀ detector. The red curve shows that the higher harmonic contribution is suppressed thanks to the lack of reflectivity at 8 mrad of SiO₂ in this energy range.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

In order to evaluate the efficiency of the SiO₂ mirrors we measured the K-edge (${{E}_{\text{K}}}=7112$ eV) x-ray absorption spectrum of an iron foil by scanning the monochromator around an angle corresponding to an energy of $E={{E}_{\text{K}}}/3=2370.6$ eV. The left panel of figure 3 shows that without SiO₂, even a detuning as large as 90% of the maximum of the rocking curve does not suppress the third harmonic, as evidenced by the presence of the Fe edge, whereas the data recorded using the the SiO₂ at both ${{\theta}_{c}}=8$ mrad do not evidence the presence of the Fe K-edge (figure 3: right panel). This shows that the SiO₂ effectively suppresses the third harmonic, although a quantitative evaluation of the possible third harmonic photons cannot be assessed.

Zoom In Zoom Out Reset image size

The experimental station of the beamline XAFS at Elettra allows us to host different multipurpose apparatus easily. Three ionization chambers (I₀, I₁ and I₂) are permanently installed for data collection in the transmission mode. In addition, two chambers with motorized sample holders allow us to accommodate multiple samples for ambient temperature and reference measurements. A silicon drift detector is installed perpendicularly to the x-ray beam direction for measurements in fluorescence mode.

Different models of in situ electrochemical cells were used for the operando XAS studies reported here. They must reproduce the same behavior as those used to test the electrochemical processes and, at the same time, they must fulfill the requirements dictated by the x-ray optics. In particular, the cell must be provided with windows that are transparent to x-rays, and at the same time it should be impermeable and corrosion-resistant.

Typical pouch cells used for operando studies of batteries are shown in figure 4. The one on the left is characterized by a large dimension of the cathode. In this particular case, the cathode material had previously been coated in an Al foil collector and assembled together with a Li anode and Celgard separator. This cell can be used at energies above 8 keV without the need for special x-ray windows along the x-ray path. If needed, a small tube can be used as a sink for the gas evolution which can take place during the electrochemical processes. On the right we show a different model of pouch cell that is used for measurements at energies of less than 4 keV, in which part of the external wrapper has been substituted with a mylar of 2 μm thickness to ensure good transparency to tender x-rays.

Zoom In Zoom Out Reset image size

A different type of electrochemical cell that we used for operando studies of batteries is a Swagelok cell that was designed for operando experiments at synchrotron facilities for both x-ray diffraction and x-ray absorption, but which was recently also used for Raman spectroscopy (figure 5). This cell is described in detail in [40]. For XAS, this cell can be used in the transmission and fluorescence modes. In this cell the x-ray window is given by a Be window that was originally of 100 μm thickness. In order to use the cell in the low energy region (below 4 keV) it can be modified by decreasing the thickness of the Be down to 13 μm while at the same time decreasing its diameter to prevent mechanical failure.

Zoom In Zoom Out Reset image size

In order to perform measurements at energies below 4 keV the electrochemical cells can be placed inside a chamber that is connected directly to the beamline and equipped with electrical connections. The fluorescence detector can be placed inside, as shown in figure 6. The whole system can operate in vacuum or in a gas overpressure. For in situ studies of batteries the chamber is usually filled with He at 10% above atmospheric pressure to avoid absorption of the air while at the same time avoiding the lack of electric contact within the cell induced by vacuum conditions.

Zoom In Zoom Out Reset image size

Among the cathode materials used for rechargeable batteries, the use of an open framework structure with sufficiently large interstitial spaces for the fast intercalation/release process of Li⁺ and other cations is one of the most interesting chemical challenges. Metal hexacyanoferrates (MeHCFs) were recently proposed as a host for lithium and/or sodium ion batteries [41, 42]. These structures are related to Prussian blue (PB) and are characterized by a rigid three-dimensional cubic network of repeating -NC–Fe–CN–Me–NC- units, where Me is a transition metal. Iron and Me sites are typically octahedral in the so-called soluble structure, and the sites at the cube center are occupied by water (zeolitic water) as well as the countercations necessary to achieve charge neutrality. The cube size, which is twice the Fe–Me distance, is about 10 Å and guarantees enough space in the zeolitic channels for the insertion/release of ions (like Li⁺, Na⁺ and K⁺) inside the structure.

We used Fe hexacyanocobaltate (K_0.44Fe_1.56Co(CN)₆) as the active material for the positive electrode of the Li-ion battery. We conducted the study on pristine electrodes and in the operando mode during galvanostatic cycling versus Li metal. The cathode was obtained by tape casting a slurry containing the pure active material (80%), 10% polyvinylidene fluoride (PVDF) and 10% carbon black in N-methylpyrrolidone (NMP) on a flat teflon surface. The anode comprised a Li foil and 1 M LiPF₆ in EC:PC:3DMC (ethylene carbonate, propylene carbonate, dimetlylcarbonate) was used as the electrolyte. The battery delivered 77 and 59 mAh g⁻¹ during charge and discharge, respectively.

The battery was assembled in the cell described in figure 5 without modification, i.e. with a 100 μm thick Be window. It was charged and successively discharged at a C/30 rate, i.e. with the insertion of a mole of Li per mole of active material in 30 h. The data were recorded alternately at the Fe K-edge (7112 eV) and the Co K-edge (7709 eV) in the transmission mode. Each spectrum lasted approximately 45 min. We recorded 73 spectra, 42 during the charge and 28 in the subsequent discharge. In this way we were able to accurately monitor the local charge and structure modification upon the potassium release and the lithium insertion process. Figure 7 shows the voltage profile of the charge and discharge curves of the iron hexacyanocobaltate cathode, as obtained during the operando XAS experiment. In the first half of the charge of the battery the voltage curve showed a curved-shape behavior. This can be attributed to an activation process of the electrode material, most probably related to the irreversible dehydration of the pristine iron hexacyanocobaltate.

Zoom In Zoom Out Reset image size

Figure 8 shows selected Fe K-edge XANES spectra obtained during the first full electrochemical cycle. These are characterized by pre-edge features at $E\sim 7115$ eV due to the transition from 1s to 3d bound states [43]. The main edge resonance is associated with the transition from 1s to the continuum and involves multiple scattering resonances of the photoelectrons [44]. The overall XANES region depends on the interatomic distances and the coordination geometry.

Zoom In Zoom Out Reset image size

During the charge the K⁺ cation is extracted from the hexacyanocobaltate structure according to the reaction K_0.44Fe_1.56Co(CN)$_{6}\to 0.44$ K⁺  +  Fe_1.56Co(CN)₆  +  0.44 e⁻. In this way the Li⁺ insertion can take place in the successive discharge according to the reaction  x  Li⁺  +  Fe_1.56Co(CN)₆  +  x e$^{-}\to$ Li_xFe_1.56Co(CN)₆. The changes in the XANES spectra during the charge and discharge indicate that the Fe is electrochemically active. In particular, the ferrous iron of the pristine electrode is oxidized to trivalent iron, as evidenced by the shift to higher energy of the onset of absorption (figure 9). During the discharge, a backward shift of the onset of absorption is observed, indicating the reduction of Fe. It must be noted that the process is not fully reversible, as indicated by the different features of the XANES spectra of the pristine material and final product. On the other side the XANES spectra at the Co K-edge do not show any modification during the electrochemical cycle and reveal that Co is not taking part in the process in the explored electrochemical potential.

Zoom In Zoom Out Reset image size

In order to understand the structural changes of the electrode material during the operation of the battery we performed a full EXAFS analysis using the GNXAS package [45, 46]. In this method the EXAFS signal is decomposed into a sum of several contributions, namely the n-body terms. The two-body terms are associated with pairs of atoms, and probe their distances and variances. The three-body terms are associated with triplets of atoms and probe angles and bond–bond and bond–angle correlations. The four-body terms are associated with chains of four atoms, and probe distances, angles in-between, bond–bond and bond–angle correlations. Given the presence of both Fe and Co in the electrode, a multiple edge approach was used. The multiple edge fitting allows us to obtain a more constrained structural description by doubling the number of experimental observations while using the same structural parameters [47]. The fitting procedures at the Fe and Co K-edge were conducted while including the relevant set of multiple scattering paths that originates from the typical structure of hexacyanoferrates and are discussed in detail for a similar system in [48].

Figure 10 shows the moduli of the Fourier transform of the EXAFS spectra at the Co and Fe K-edges (top and bottom panels, respectively) during the charge and discharge (left and right panels, respectively) of the battery. For Fe data the fits suggest that the Fe–N bond distance decreases slightly during the first charge, from 2.07 to 2.05 Å and sets back to the value of 2.07 Å at the end of the discharge. The increase of the intensity of the first peak of the modulus of the FT is due to an increase of the structural order upon the release of K⁺, evidenced by a decrease of the Debye–Waller factor. Regarding the Fe–C bond length, this remains constant during the charge but decreases during the discharge. At the Co site, the Co–C and Co–N bond lengths are the same in all the investigated spectra, and the same holds true for the corresponding Debye–Waller factor. The contribution in the modulus of the FT at  ∼4.3 Å is attributed to an intense multiple scattering signal due to the Fe–C–N–Co linear chains of the Fe-haxacyanocobaltate structure. From both the Co and the Fe point of view it is evident that the number of linear chains increases during the first charge. This, in combination with the decrease of the Debye–Waller factor for the Fe–N pair, reflects a rearrangement of the structure to a more interconnected one during charge, i.e. upon K⁺ release.

Zoom In Zoom Out Reset image size

Li-S batteries, with their high energy density and cost-effectiveness, are potential candidates for commercialization among all the post-Li-ion batteries. In fact, the transition from an intercalation compound to an elemental cathode, e.g. sulfur or oxygen, theoretically raises the storage capacity by more than one order of magnitude. However, the electroactive elements such as sulfur and oxygen are insulators and need to be housed inside an electronic and ionic conductor. In the case of Li-S batteries, additional issues arise through the unavoidable formation of various intermediate polysulfides during the reversible conversion of elemental sulfur into Li₂S. Improper confinement of the sulfur/polysulfides leads to large losses in the active mass of the chalcogen, resulting in continuous capacity fading. In this context, it is essential to comprehend aspects such as the interactions among sulfur, the host matrix, polysulfides and electrolytes in order to develop new materials for usable and marketable batteries.

Sulfur is a highly reactive, heterovalent element that occurs in a wide range of formal oxidation states, ranging from  −2 in some sulfides to  +6 in sulfates. K-edge XANES spectra are sensitive to the oxidation state, chemical environment, type of bonding and group symmetry, and can be used as a fingerprint to identify individual sulfur compounds. Generally, the pre-edge and white-line peaks can be attributed to transitions of sulfur 1s core electrons to the lowest symmetry-available unoccupied antibonding states. These are atomic or molecular orbitals with a substantial p-orbital character [49–51]. Post-edge features are often attributed to multiple scattering resonances [49, 52], but also to further molecular orbital transitions [53–55]. Therefore, XAS is an appropriate tool for characterizing the redox chemistry and the modifications of the sulfur species of Li-S batteries during cycling, as well as monitoring polysulfide formation and evaluating possible interactions of sulfur and polysulfides with the host matrix and the electrolyte.

We used different in situ and ex situ analytical tools to analyze Li-S batteries at different stages of charge and discharge as well as in operando conditions with the aim of understanding the mechanism of the Li-S battery and the impact of different chemical environments on its electrochemical properties. Different spectroscopic techniques, including UV-Vis, Raman spectroscopy and XAS, were coupled with galvanostatic charging/discharging of the battery.

Restricting the discussion to XAS characterization, two milestones were achieved prior to performing operando measurements of actual batteries, made using a different combination of cathodes and electrolytes. First, there was the optimization of a reliable setup for XAS measurements at the S K-edge. This included the refurbishment of the experimental hutch of the beamline XAFS (with the installation of SiO₂ mirrors for harmonic suppression) and the He chamber for low energy measurements, as well as the modification of the electrochemical cells, as discussed in section 2. The second milestone consisted of building a set of measured spectra for the different battery components (sulfur and electrolyte) and products of the electrochemical reaction (polysulfides and crystalline Li₂S).

All the measurements were performed in the fluorescence detection mode using either the cell shown in figure 4 (right panel) or the one shown in figure 5. Fluorescence S K-edge data are affected by flaws, which, if not taken into account, lead to mistaken interpretations of the results. In fact, the shape of the spectra can be altered significantly by self-absorption effects and by an increase of penetration depth. Self-absorption effects spoil the XANES part of the spectrum in particular. The intensities of its features decrease nonlinearly with increasing concentration of the photoabsorber (figure 11: left panel). To a first approximation, the intensity of the spectral features is dependent on the sulfur oxidation state. This is due to the fact that the size of the white line, and therefore the respective transition probability, is proportional to the number of available final states and, thus, should increase with increasing valence. The alteration of the intensity of the spectral features also yields an erroneous estimate of the energy of the absorption onset evaluated by the maximum of the derivative of the absorption (figure 11: right panel).

Zoom In Zoom Out Reset image size

The increase of the penetration depth of the incident beam in the sample with increasing photon energy mostly affects the EXAFS part of the spectrum. The number of atoms illuminated by the incident beam increases at higher photon energies and consequently the fluorescence signal increases proportionally. An example is shown in figure 12 (top). A standard normalization procedure leads to mistaken EXAFS oscillation intensities (figure 12: bottom panels) and therefore erroneous evaluation of intensity-dependent parameters such as the coordination numbers and Debye–Waller factors. Correct EXAFS oscillation intensities are obtained by subtracting a linear function determined in the pre-edge region, and dividing by the post-edge background, approximated by a smooth best-fit spline function.

Zoom In Zoom Out Reset image size

Examples of the operando XAS characterization of Li-S batteries at the sulfur K-edge are reported in some studies [17–26]. Here we report the XANES study of a Li-S cell in which the cathode composite contained zeolite as the additive to adsorb polysulfides. The addition of zeolite is supposed to block the diffusion of polysulfides, making it possible for XAS to give a view of the whole system, although the penetration depth of x-rays in the S K-edge energy range is limited to approximately 5 μm. More specifically, the cathode composite had the following composition: 20% S  +  3.5% zeolite  +  61.5% CB  +  15% teflon. The electrolyte comprised 1 M of LiTFSI (lithium bistrifluoromethanesulfonimidate) in sulfolane. The open-circuit voltage was 2.5 V, the discharge and charge were performed at the C/20 rate, and each spectrum lasted 65 min.

Figure 13 shows the XANES data during the operation of the battery. From the sulfur K-edge XANES spectra measured in the operando mode it is possible to determine the change of composition of the cathode. The different sulfur compounds formed at the cathode can be identified by the characteristic energy positions of the sulfur edge and pre-edge resonances. Elemental sulfur has a main resonance at 2472 eV, while the onset of absorption in sulfate species is shifted to higher energy at 2479 eV. The lithium polysulfide molecular systems produced in the electrochemical reaction, in which sulfur has an oxidation state of S²⁻, are characterized by a pre-peak at 2470.2 eV together with a S⁰ peak at 2472 eV. The ratio between the pre-peak and the peak is inversely proportional to the chain length [19, 23, 56]. The XANES spectrum of the crystalline form of Li₂S is significantly different from the spectra of long- or short-chain polysulfides [17, 21, 56, 57]. In the present case the main contribution to the spectra comes from the S⁶⁺ compounds present in the electrolyte. The resonance relative to elemental sulfur is present in the initial state and the increasing shoulder during the discharge indicates the formation of Li polysulfides. At the end of the discharge the features in the energy range 2470 eV–2475 eV indicate the formation of crystalline Li₂S. During the charge the peak relative to elemental sulfur reappears, although the system is shown to not be completely reversible. In order to retrieve the quantity of the different components during the discharge and the charge of the battery, LCF was performed using the Athena program [58]. For the LCF we used reference spectra of the electrolyte, elemental sulfur, mid-long chain lithium polysulfides and Li₂S in a similar environment and at a similar sulfur concentration. Figure 14 shows the relative quantities of the S species during the discharge/charge, together with the electrochemical curve. A drop of S is observed in the high voltage plateau concomitantly with the appearance of polysulfides. Nonetheless, the quantity of S never goes to zero and in the low voltage plateau the quantity of S remains constant. In this part of the discharge it is the polysulfide quantity that varies with inverse proportionality to the Li₂S quantity, which is maximum at the end of the discharge as expected. The trend is symmetrically inverted for polysulphides and Li₂S during the charge, although the pristine state is not recovered.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

LCF shows that it is the electrolyte that makes the dominant contribution to the XAS spectra. This does not hinder the evaluation of the redox chemistry of sulfur during the operation of the battery from the XANES part of the spectrum. This holds true because the sulfur and its reduced compounds have an onset of absorption at energies well below and easily distinguishable from the onset of absorption of the sulfate species of the electrolyte. However, the EXAFS signal is actually dominated by the presence of the sulfates in the TFSI anion of the Li salt dissolved in the electrolyte. This dominant signal impedes study of the transformation of the local environment of the sulfur species in the electrode. This point is particularly important if one wants to suppress, as far as possible, the shuttle mechanism of the polysulphides. The measurements presented here confirmed that zeolites have an affinity with polysulphide adsorption, since during the discharge process a much larger quantity of sulfur species remained on the back side of the electrode. However, even though these results are promising, they cannot be considered to be complete without a more precise study of the possible short-range interaction of these species with the zeolite additive. This can only be achieved with a full EXAFS study. S K-edge EXAFS analysis was only made possible by switching from the commonly used electrolyte salt to a sulfur-free one [21]. This provided complementary information on the neighbors around the sulfur center, which could not be gained solely through XANES. In particular, EXAFS analysis allowed us to detect sulfur and Li₂S components, whereas the polysulfides were determined from the variation in the average coordination number of the S–S bonds during the discharge process. In addition, the presence of unknown sulfur species was ruled out, as was specific interaction of the polysulfide species with the matrix or with other species in the electrolyte [21].