Electrochemical in situ reaction cell for X-ray scattering, diffraction and spectroscopy

A versatile electrochemical in situ reaction cell for long-term hard X-ray experiments on battery electrodes is described. Applications include the small-angle scattering, diffraction and absorption spectroscopy of lithium manganese oxide electrodes.

1. Introduction

On-line monitoring of changes in materials undergoing chemical reactions is usually performed by using in situ reaction chambers. In the last two decades, countless X-ray in situ chambers have been designed for battery research applications. Since some of the battery electrode materials are very sensitive to corrosion in the ambient atmosphere, including in particular the alkaline metal-based systems such as lithium batteries, the sealing of the cells plays a very important role.

While many different cell designs have been successfully developed and tested for short-term experiments, only few have been reported that allow for experiments that extend over several days or even weeks. One research group has overcome many of the obstacles related to electrochemical in situ X-ray experiments by switching to energy-dispersive X-ray diffraction, that utilize rather high X-ray energies (Ronci et al., 2001); other groups depend on in situ cells that were manufactured in an industrial plant (Richard et al., 1997) and on a particular diffractometer type (Gérand et al., 1999). Another in situ cell (Koetschau et al., 1995) allowed for long-term experiments, but did not contain lithium metal counter electrodes. The McBreen group has performed in situ X-ray diffraction and absorption, but they measured the cathode of a high-rate industrial battery which could be studied in a short time (Balasubramanian et al., 2001).

Charging and discharging reactions in batteries are pronouncedly limited by diffusion and may take hours. Since the structure of electrodes may be dependent upon the number of charge and discharge cycles, experiments often take as long as weeks or months.

When exposed to the ambient atmosphere, the lithium reacts with the nitrogen in the air to form lithium nitride or, with the humidity to form lithium hydroxide. Some of the electrolytes used in batteries may also form aggressive reaction products; for instance, LiPF₆ forms hydrofluoric acid when in contact with the ambient humidity.

These corrosive reactions are common in batteries and must be minimized by using proper sealing. Regular battery test cells which are not used for in situ studies are usually very carefully designed to satisfy the strict sealing requirements and allow cycling experiments for several months.

In contrast, in situ cells need X-ray windows and thus are more prone to leakage. In some of them, leakage problems are overcome by extensive use of sealants, and those cells are often for one-time use only.

In this paper we present a convenient re-usable cell with beryllium X-ray windows and sealing rings. It allows charging and discharging of the battery for several days without noticeable traces of corrosion of the lithium metal.

The cell's design allows for it to be employed in the transmission mode at any hard X-ray source. In particular, the cell was tested at beamline 12-ID (Braun et al., 2001) at BESSRC-CAT, Advanced Photon Source, Argonne, IL, USA, and at beamlines 2-1 and 2-3 at Stanford Synchrotron Radiation Laboratory, Menlo Park, CA, USA.

To the best of our knowledge, this is the first time that a single in situ electrochemical cell has been used for X-ray techniques so different in nature such as X-ray diffraction, spectroscopy and anomalous small-angle scattering.

2. Design of the in situ cell

The cell design and a photograph are shown in a panel in Fig. 1. A schematic drawing of the flat cylindric cell is given in Fig. 1(a). The main support of the cell is a stainless steel disk (15 cm diameter) and a polypropylene (PP) disk (9 cm diameter), both of 5 mm thickness. With steel bolts (size M 5), both disks are held together and contain the electrode sample, an ion-conducting separator, the electrolyte and a lithium metal counter electrode. Sealing of the cell is necessary to prevent the electrolyte from drying out and to prevent the sample from chemical reactions with the ambient atmosphere. The zone between the steel disk and the PP disk is the sample compartment, which is sealed by three concentric BUNA O-rings which have different diameters. Corresponding O-ring groves are provided in both disks. Fig. 1(b) shows a magnified view of the portion of the cell that contains the electrode assembly. Both disks have a port of 1 mm × 2.5 mm in their center to allow the X-ray beam to pass through the cell. The ports on either side of the cell are capped with a beryllium disk of thickness 380 µm. The interfaces between the beryllium disks and the steel disk and the PP disk, respectively, are also sealed with three small concentric BUNA O-rings. On either side, a 30 mm-diameter aluminium disk holds the beryllium disks in place. In their center, the aluminium disks have a 7 mm-diameter opening to allow the X-ray beam to pass through.

Figure 1 (a) Cross-sectional overview of the in situ cell. Not all of the parts refer to the same scale. (b) Magnified cross section of the central cell parts. (c) Photograph of the in situ cell, showing the PP disk on top of the stainless steel disk. The ruler shows the length scale in inches.

We have carried out a systematic study on how many O-rings are necessary to provide sufficient protection. One single O-ring was clearly insufficient for proper sealing as required here. We found that a set of three O-rings provides very good sealing and is still efficient and feasible in terms of cell machining.

The thicknesses of the disks and the width of the holes determine the aspect ratio which is necessary to cover a desired Q-range for wide-angle X-ray diffraction. In the present case, a 2-angle of about 45 at maximum was permitted which allowed resolution of the [111] Bragg reflection of LiMn₂O₄ at X-ray energies of around 6500 eV. By choosing higher energies, the entire diffractogram becomes compressed and will be resolved in the 2-angle of about 45.

The Be foils are covered with 25 µm-thick Kapton tape to prevent contact with the corrosive electrolyte.

The working electrode is in direct contact with the steel disk to guarantee electrical contact. The PP disk has two feed-throughs for stainless steel bolts which serve as contacts to the lithium counter electrode and an optional reference electrode. Each of the two steel bolts is sealed with a tiny O-ring of diameter 2 mm; these seals are the least reliable parts in terms of contamination from the ambient atmosphere. After several days of operation, the side of the lithium foil which is next to the steel bolts exhibits typical traces of black lithium nitride, which is an indication that nitrogen has diffused around the O-ring. However, this effect was confined to an uncritical area of the cell.

Fig. 1(c) shows a photograph of the assembled cell with the PP disk on top of the steel disk. Two metal strips are contacted by the bolts in the feed-through for connection to the reference and counter electrodes. Numerous drilled holes in the metal disk allow for versatile mounting of the cell at the beamlines.

At SSRL and APS we assembled and sealed the cells in a portable helium-filled glove bag on site in their chemistry laboratories. On one occasion we assembled the cell in our home laboratory at LBNL and brought it to SSRL for experiments.

To enhance the sealing safety of the cell during experiments, we wrapped it in a conventional ziplock bag filled with helium. It turned out, however, that the cell remained fine without significant contamination or corrosion for more than four days without keeping it in helium.

3. Experimental

The first experimental results were obtained with a complete lithium manganese oxide cell. The LiMn₂O₄ spinel powder was synthesized by mixing lithium carbonate Li₂CO₃ and manganese oxide MnO₂ (CMD IC#5, Japan Metals and Chemical) in the stoichiometric amounts and heating it in a muffle furnace with air at 1123 K and subsequent cooling.

The spinel powder was mixed with carbon (graphite from Timcal, and Shawinigan carbon black) and polymer binder (all together 5 wt%) and cast on an aluminium foil of thickness 25 µm, and subsequently sintered at 423 K, resulting in the positive electrode of thickness 40 µm, and containing spinel to about 95 wt%.

Electrodes with a diameter of 25 mm were punched from the foil sheet and vacuum dried for another 24 h, and then stored under argon before being used.

The positive electrode, with a polypropylene separator, electrolyte (LiPF₆ dissolved in dimethyl carbonate) and lithium foil (125 µm) as counter electrode was assembled to form the cell (Braun et al., 2001).

The cell assembly was as follows:

All necessary cell parts, chemicals and tools for assembly were contained in either a helium-filled glove bag (for assembly at the synchrotron) or in an argon-filled glove box in our laboratory at LBNL. The beryllium windows were already assembled, and the O-rings already placed in the grooves. A piece of lithium foil, as large as the spinel electrode, was punched out from fresh supplies and squeezed onto the steel contact bolt of the PP disk; this made the counter electrode. Then, one single spinel/alu-foil assembly electrode was placed in the center of the steel disk, with the spinel side facing up, and the separator put on top of it. A small strip of lithium foil was placed on the separator such that it would later match the position of the reference electrode steel bolt on the PP disk. A second piece of separator prevents contact between the reference electrode and counter electrode. Next, some drops of electrolyte were dropped by pipette on the separator so that the separator and the electrode underneath were fully soaked. Finally, the PP disk was put on top of the steel disk, with all necessary battery cell parts in between, and tightened with bolts. With some practice, the cell assembly takes only a few minutes.

At the synchrotrons, all data were recorded in the transmission mode using gas ionization chambers for the incident X-ray intensity I_o and the transmitted intensity I_t. A manganese foil was used for energy calibration. For the SAXS experiment only, a CCD camera was used to record the scattered intensity.

4. Results

4.1. XANES

X-ray absorption near-edge structure (XANES) spectra were obtained at SSRL, BL 2-1, with the manganese being reduced from an average oxidation state of 3.5 towards 2.0 by inserting lithium electrochemically into the spinel host lattice (deep discharge of the cell). XANES spectra for three different states of charge are shown in Fig. 2.

Figure 2 XANES spectra of Li1+xMn2O4 for three different oxidation stages of Mn. Included as a reference is the XANES spectrum of rhodochrosite (—⋯—⋯, 6530–6580 eV) (after Shiraishi et al., 1997).

Fig. 3 displays the evolution of electric charge in the cell during discharge (lithiation). This curve was obtained by integration of the current through the cell over discharge time, which was monitored with a portable potentiostat from Pine Instruments (Braun et al., 2001). The cell was discharged for approximately 27 h at currents not exceeding 0.2 mA.

Figure 3 Evolution of charge, as obtained from integration of current through the discharge process.

The stoichiometry of LiMn₂O₄ implies that manganese in this compound is present in equal amounts as Mn⁺³ and Mn⁺⁴. An XANES spectrum of this state was recorded prior to discharging of the cell and is shown as the dotted spectrum in Fig. 2. It can be considered as a superposition of spectra with the chemical shift for Mn⁺³ and Mn⁺⁴, averaging to an oxidation state of 3.5.

By electrochemical lithiation, the stoichiometry changes from LiMn₂O₄ to Li_1+xMn₂O₄ with , and, consequently, manganese is reduced from an average of 3.5 towards 3.0, or even lower. For x = 2, the average oxidation state of manganese in Li_1+xMn₂O₄ matches 3.0. The corresponding XANES is the semi-dotted spectrum in Fig. 2, which was recorded approximately 900 min after the discharging was started. This stage is indicated by A in Fig. 3. Within this time, a charge of 144 mA min was accumulated in the cell. Since the amount of spinel in the electrode was about 12 mg, we calculated a total capacity of 200 mA h g⁻¹. The theoretical capacity for the reaction is 148 mA h g⁻¹. The excess amount is probably parasitic charge due to electrochemical side reactions. We continued to lithiate the cell for another 730 min. At this stage (indicated by B in Fig. 3) the cell potential was 1.23 V and the total accumulated charge was 3.1 mA h, which amounts to a fictitious discharge capacity of 258.3 mA h g⁻¹. The corresponding XANES shows the smallest chemical shift so far among the three spectra. We assign it the oxidation state of Mn⁺² because it matches both literature data of a mineral with Mn⁺² [rhodochrosite, MnCO₃ (Shiraishi et al., 1997)] as well as the extrapolated calibration curve of the chemical shift for manganese compounds (Shiraishi et al., 1997). The XANES spectrum of rhodochrosite as obtained by Shiraishi et al. (1997) is included in Fig. 2 as a reference.¹

4.2. XRD

Fig. 4 displays the [111] Bragg reflection of the Li_1+xMn₂O₄ spinel lattice for two different X-ray energies at Q-values of 1.392 and 1.395 Å^-1, respectively. The two upper curves were recorded at 6500 and 6538 eV, the former of them at a potential of 2.97 eV (`open' cell, disconnected from the potentiostat, and the electrode being LiMn₂O₄), the latter at a slightly lower potential. Both curves are shifted by 500 units upwards on the intensity scale for better distinction from the two lower diffractograms, recorded after deep discharge at the same X-ray energies but at a lower cell potential (∼1.54 V).

Figure 4 [111] Bragg reflection in the uncharged (single peak) and deeply discharged (split peak) state for energies at 6538 eV (solid line) and 6500 eV (dotted line).

While discharging, the [111] reflection splits into two peaks at Q-values of about 1.37 and 1.39 Å^-1. This cubic to tetragonal phase transition is known as the Verwey transition and is supposedly caused by the Jahn–Teller effect (Dunitz & Orgel, 1957; Yamada et al., 1999). We notice that the diffractogram of the open cell already shows a slight densification of diffracted intensity at Q = 1.38 Å^-1, indicating that the spinel as prepared is not entirely a pure cubic phase. The presence of two phases may be due to the fact that LiMn₂O₄ undergoes a phase transition from cubic to tetragonal at about 300 K, and our experiments were carried out at room temperature. The critical balance of parameters that affect phase changes and phase purity in lithium manganese oxides is a topic that currently receives much attention from many research groups (Mishra et al., 1999; Kanno et al., 1999, and references therein).

When we plot the two diffractograms, obtained at 6500 and 6538 eV, in the classical way versus the 2-angle, the latter diffractogram will be shifted towards smaller angles. We can compensate for this effect by using the scattering vector Q instead,

The X-ray wavelength is related to the X-ray energy E via E = .

Yet, we see in Fig. 4 that corresponding Bragg reflections are shifted, even if we plot them versus the scattering vector Q. Thus, the shift is a result of structural changes in the electrode material during discharge.

4.3. EXAFS

EXAFS spectra were recorded during discharge of the cell at beamline 2.3 at SSRL. Fig. 5 shows the k-weighted EXAFS oscillations (k) of the absorption curve when the average Mn oxidation state was 3.39. For this single figure we have used the computer program EXAFS pour le Mac (Michalowicz, 1991, 1997) for the pre-edge fit and baseline fit, determination of absorption edge and removal of monochromator glitches. All other EXAFS data were analyzed using the computer program EXAFSPAK (George et al., 2000).

Figure 5 EXAFS k(k) of the sample while Mn was in the oxidation state of 3.39.

The evolution of the Fourier transform (FT) amplitudes of six scans during the entire discharge process is displayed in Fig. 6. In both scans we assigned the manganese oxidation states, derived from the chemical shift of the XANES spectra (as exercised in §4.1). The deepest state of discharge obtained in this particular experiment corresponds to an average manganese oxidation state of +3.09, and to a nominal stoichiometry Li_1.82Mn₂O₄.

Figure 6 Evolution of EXAFS Fourier magnitude for Mn during discharging from average Mn valence 3.5 to 3.09. Magnified inset with FT maxima normalized to unity.

Since these data have not been corrected for the phase shift, distances are given in terms of R + . The peak at 1.92 Å corresponds to the Mn—O distance, and the peak at 2.82 Å corresponds to the Mn—Mn distance.

We first notice that, upon discharging, the intensity of the peaks decreases systematically. The reverse effect, i.e. increasing peak heights upon charging, has already been reported (Nakai et al., 1999; Shiraishi et al., 1997) and attributed to structural ordering of MnO₆ octahedra when manganese approaches a uniform valence state of Mn⁺⁴. The extension of this interpretation for the discharge process is obviously not applicable. This is not surprising because the manganese is being reduced during discharge from an average oxidation state of Mn^+3.5 towards Mn^+3.0, and Mn⁺³ is known to be the high-spin Jahn–Teller ion of manganese with a pronounced tetragonal distortion of either MnO₆ octahedra.

Then, we notice that the peak for the Mn—Mn distance experiences a gradual shift from 2.82 to 2.74 Å during discharge. In the inset in Fig. 6 the FT amplitudes were divided by a constant factor such that the peak for the Mn—O distance is normalized to unity. The range from 1 to 3.5 Å is magnified to show how the peak for the Mn—Mn distance shifts. The normalization also makes clear that the small peak at 3.35 Å is systematically shifted towards 3.16 Å, increasing in intensity, at the cost of intensity of the peak at 2.82 Å (2.74 Å, respectively).

It is remarkable, however, that the Mn—O distance at 1.92 Å turns out to be very robust versus discharging, since this peak experiences no shift at all.

4.4. ASAXS

Small-angle X-ray scattering (SAXS) is used to study the structure of disordered media on a nanoscale, resolving structures with sizes ranging from 1 nm to fractions of a µm.

SAXS can be considered as diffuse scattering of monochromatic X-rays with respect to the (000) Bragg reflex. By convention, small-angle scattering occurs between the angle of specular beam and the half angle of the first Bragg reflex position. SAXS, and small-angle neutron scattering, have already been used for ex situ microstructure analyses of glassy carbon electrodes, such as in electrochemical capacitors (Braun et al., 1999, 2003).

Anomalous small-angle X-ray scattering (ASAXS) is small-angle scattering with at least two different X-ray energies. Scattering curves (SCs) are obtained with an X-ray energy before the absorption edge of a particular element, and right after the absorption edge. This technique is called contrast variation and allows the separation of the scattering contribution from objects containing a particular chemical component from other contributions (Fratzl et al., 1992).

Battery electrodes are relatively complex in terms of microstructure and chemical composition. They contain carbon, active material such as the spinel, and also pores and electrolyte. The X-ray SC obtained by an in situ experiment contains structural information of the electrode, but also of the cell parts.

With SAXS at one single X-ray energy only, no unambiguous distinction can be made between either phases. Tuning the X-ray energy to the manganese K-edge, however, and subtracting the SC taken before the absorption edge of manganese from the SC taken at the absorption edge, yields an SC which contains information about the manganese-rich structures only.

ASAXS was carried out at the Advanced Photon Source (APS, BESSRC-CAT) for a period of 40 h, during which the cell was slowly charged to 4.3 V, and thereafter discharged back to 3.05 V.

Sets of SCs were sequentially recorded at 20 different X-ray energies between 6350 and 6550 eV near the manganese K-shell absorption edge to allow for contrast variation due to the anomalous scattering of the Mn. Owing to the thin design of the in situ cell with beryllium X-ray windows, a single SC could be acquired in only 0.1 s, and the overall acquisition time for a set of 20 SCs was only 1 min. Therefore the ASAXS studies can be considered as a time-resolved experiment. About 60 sets of such 20 SC were recorded during the entire experiment. The SCs at energies of 6400 and 6545 eV were chosen for further data analysis (Braun et al., 2001).

Fig. 7 displays SCs of the entire cell, obtained at four different X-ray energies prior to any charging of the electrode in the cell. The corresponding energies are noted accordingly in the plot. Intensity changes among either four SCs as functions of the X-ray energy are due to X-ray fluorescence and anomalous scattering of the manganese. By appropriate subtraction algorithms (Fratzl et al., 1992; Braun et al., 2001), the scattering of the Mn-containing objects can be obtained.

Figure 7 Scattering curves of the entire in situ cell, obtained at four different X-ray energies.

The inset in Fig. 7 shows one SC of the discharged cell. This curve was analyzed for anomalous scattering as described by Fratzl et al. (1992) and fitted with two Guinier functions (Braun et al., 2001). This SC represents structural data of the Mn in the electrode only. All other contributions, such as from the cell parts, carbon and pores, are subtracted.

Fig. 8 shows the SC of the manganese in the electrode in its uncharged, charged and discharged state. The humps in the data curves account for the presence of structural inhomogeneities, such as nanocrystallites or aggregates or agglomerates. They can be approximated by Guinier functions and analyzed quantitatively. The data points are the actual data, and the drawn lines are sums of two Guinier functions for the charged and discharged electrode. For the uncharged electrode we found that three Guinier ranges were present. For further details of the ASAXS experiment and interpretation of the data, we refer the reader to Braun et al. (2001).

Figure 8 Scattering curves of manganese oxide in electrode matrix for three different states of charge.

5. Summary

A versatile and easy to manufacture, easy to assemble, re-usable electrochemical in situ battery cell for X-ray experiments in transmission was tested with LiMn₂O₄-based electrode material at various synchrotron beamlines. The cell contains a lithium metal counter electrode and can operate safely for more than 40 h without significant traces of corrosion. The thin cell design with beryllium X-ray windows permits a high transmission for quality data on the chemical shift, crystallographic phase transitions and changes of the microstructure during electrochemical reaction.

Footnotes

¹QuickTrace is a freeware Macintosh program that digitalizes original data in Cartesian plots for further processing. In this paper it was used to read in the XANES of rhodochrosite for reference purposes, as obtained by Shiraishi et al. (1997). QuickTrace is available from Dr Rainer Petschick, [e-mail].