Rapid Preparation of Geometrically Optimal Battery Electrode Samples for Nano Scale X-ray Characterisation

Lithium nickel manganese cobalt oxide (NMC811, NEI Corporation) positive Li-ion electrode samples were secured on a petri dish with polyimide tape and fixed to the translation table of a laser micro-machining instrument (A Series, Oxford Lasers Ltd.) containing a 532 nm laser with a spot size of ca. 40 μm. The electrodes were mounted with the current collector facing up to minimize damage to the delicate electrode surface. Where it is desirable to keep the electrodes in a protective atmosphere, the petri dish can be sealed under argon with a glass lid—its optically transparent nature allows transmission of the laser beam with minimal distortion. The laser was programmed to mill a series of lines amounting to a tab with customizable dimensions and the electrode was cut at a speed of ca. 2 mm s⁻¹. The resulting tabs consisted of 2 rectangular sections with the smaller section measuring ca. 0.1 mm × 0.5 mm and the larger section measuring ca. 1.5 mm × 4 mm. Optimization of laser power and milling speed is sample and instrument dependent, and some trial and error is required with the goal of minimizing laser power (to avoid unnecessary beam-included sample damage) whilst obtaining a clean cut. For comparison, a ca. 80 μm diameter electrode pillar of the same NMC811 electrode sample was prepared with the same technique as previously described⁸ using a low laser power.

The laser-cut electrode tabs as described above were dried for 12 h under vacuum at 120 °C in a glass drying oven (B-585, BUCHI Ltd.) before being assembled into coin cells (CR2032, Hohsen Corp.). A half-cell electrode arrangement was used with the NMC811 tab electrode as the working electrode and lithium metal (MTI Corp) as the counter electrode. A tri-layer polyolefin membrane was used as separator (Celgard 2325, Celgard, LLC.) and the electrolyte used was 1.0 M lithium hexafluorophosphate in ethylene carbonate and ethyl methyl carbonate (1.0 M LiPF₆ in EC:EMC (3:7 v/v), Soulbrain MI).

The theoretical capacity of the CR2032 cells was calculated to be ca. 0.121 mAh, based on the mass loading and area of the electrode, and a specific capacity of 190 mAh g⁻¹. The cells were taken through 3 initial cycles using a constant-current-constant-voltage (CC-CV) charge and constant-current (CC) discharge protocol. The current used during the constant current steps was equivalent to a C/20 rate with a cut-off condition of 4.2 V for charge and 2.5 V for discharge. For the constant-voltage charge step, the cells were held at 4.2 V until the current dropped to C/40. Following the initial cycles, a rate capability test was performed on the cells by taking them through various C-rate regimes from C/10 to 2 C and back to C/10, (C/10, C/5, C/2, 1 C, 2 C) for 5 cycles at each C-rate. The same cut-off conditions were used. For comparison, the same electrodes were punched into 15 mm discs, assembled into CR2032 coin cells with the same arrangement and cycled for the 3 initial cycles described above. All electrochemical tests as shown in the figures in the Results section were performed on a multi-channel battery cycler (BCS-805, Bio-Logic SAS).

The pristine laser-cut electrode tabs were mounted on steel dressmakers' pins using a quick setting epoxy (ITW Devcon 5 minute. Epoxy, ITW Devcon) and were left to set over a 24 h period to minimize sample movement. Ex situ X-ray nano-computed tomography (CT) was performed on the laser-milled tab electrode samples using a lab-based nano-CT instrument (ZEISS Xradia 810 Ultra, Carl Zeiss Inc.) containing a rotating Cr anode source set to an accelerating voltage of 35 kV. The beam was quasi-monochromatized at the Cr-Kα emission line of 5.4 keV by a capillary condenser, illuminating the sample with a hollow cone beam focused on the sample, with a Fresnel zone plate as the objective element creating a magnified image on a 1024 × 1024 pixel CCD detector.¹⁰

The lab-based configuration of the ZEISS Xradia 810 Ultra X-ray microscope offers two imaging modes, absorption and Zernike phase contrast, and two fields of view: either a 65 μm FOV with a pixel size of 64 nm or a 16 μm FOV with a pixel size of 16 nm. Large field of view mode with a pixel binning of 1 was used, resulting in a pixel size of ca. 64 nm and a field of view of ca. 65 μm. The sample was rotated through 180° with radiographs collected at discrete angular intervals amounting to 1601 projections. The sample was translated by ca. 58 μm in the vertical direction after the acquisition of each tomogram to obtain four tomograms with overlapping regions of ca. 6 μm between each tomogram.

For comparison, the laser-milled electrode pillar as described previously was also imaged with the same imaging parameters, although no stitching was carried out as the sample was fully within the vertical field of view.

The projection data from each tomogram were reconstructed with proprietary reconstruction software (XMReconstructor, Carl Zeiss Inc.) using a parallel beam reconstruction algorithm.

The four reconstructed tomograms were manually registered by visually aligning the datasets using the Avizo image processing software (2019.2, Thermo Fisher Scientific). The registered tomograms were then merged using the "Merge" module in Avizo to form a combined volume. To ensure that the virtual slices captured from the dataset were orthogonal to the electrode, the combined volume was transformed to correct for sample tilt. The transformed volume was resampled using a Lanczos filter¹¹ producing a final volume that contained 683 × 903 × 1981 voxels.

Image processing was performed in Avizo on the final transformed volume containing 683 × 903 × 1981 voxels. As the aluminium current collector had similar grayscale values to the NMC particles, a sub-volume was extracted with the current collector cropped out. Segmentation was then carried out via a simple thresholding algorithm derived from Otsu's method¹² and available in the "Auto-Threshold" module in Avizo. Small features less than 5 voxels wide were removed and the "Separate Objects" Avizo module was applied to separate neighbour particles that were in contact with one another. Particles that were only partially captured within the volume were removed using the "Border Kill" module and finally, the remaining particles were individually labelled using the "Label Analysis" Avizo module which assigns different label values to each particle.

For comparison, the same image analysis workflow was carried out on the reconstructed electrode pillar obtained with the laser-lathing technique, where the tomogram contained 1004 × 1024 × 1013 voxels (although the useable volume was smaller as the vertical height of the detector was not fully utilized).

As described in the introduction and illustrated in Figs. 3 and 4a, we have previously adapted a laser-milling technique to lathe battery electrode pillar samples to sizes suitable for nano-CT.⁸ With the original sample geometry shown in Fig. 4a, the current collector of the electrode is orthogonal to the centre of rotation of the sample and does not obscure the electrode layer in any of the angular projections. Whilst this is ideal for electrodes with highly attenuating current collectors (i.e. copper or nickel), and for experiments outlined earlier such as in situ compression tests to mimic calendaring,⁹ potential drawbacks include heat-induced sample damage, long sample preparation time and limited imaging volume. From previous attempts, it has been found that the laser-milling technique may not be suited to samples with a low melting point or susceptible to sublimation, for instance sulfur electrodes used in Li-S cells.^16,17

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Furthermore, attempting to introduce laser-milled electrode pillars into in situ environments remains challenging due to various factors. Samples are often mounted onto stainless steel pins with epoxy before being milled into suitable sizes for nano-CT. Epoxy is generally electrically insulating, preventing the passage of current to the electrode, and whilst certain grades of conductive epoxy may be used, the conducting material may not be electrochemically inert, which would influence the validity of the electrochemical data obtained.

For electrodes with lower-Z current collectors (i.e. aluminium, as commonly employed in Li-ion positive electrodes), it is possible to prepare samples for nano-CT with geometries where the current collector is parallel to the centre of rotation of the sample. We have previously prepared wedge-shaped samples using a sharp scalpel for imaging NMC electrodes using micro-CT,¹⁸ achieving a spatial resolution capable of resolving the carbon binder domain (CBD). However, current collector delamination is a severe issue with manual preparation of battery electrodes for imaging, especially for samples that have undergone extensive ageing. Other drawbacks of manual preparation include poorer control of sample geometry and the possibility that the morphology of the sample is influenced by the force exerted by the cutting blade.

A laser micro-machining instrument with micron-level precision was therefore adapted to cut electrodes to sizes suitable for nano-CT. The difference between the conventional laser electrode pillar preparation and new laser tab preparation technique is illustrated in Fig. 3. The tab-shaped geometry of the cut electrode is presented in Fig. 4b, and the high precision attainable with laser cutting is observable in Figs. 4c and 5, which is a mosaic radiograph of a cycled NMC electrode from which a tab was prepared for nano-CT. As the spot size of the laser beam used is approximately 40 μm, a nominal tab width of 100 μm would result in a tab of ca. 80 μm in width, which is an ideal size that approximately fits within the field of view of the ZEISS Xradia 810 Ultra in large field of view (LFOV) mode.

Zoom In Zoom Out Reset image size

It is thus conceivable for the same electrode to be imaged with lab-based nano-CT at its pristine state before being assembled into standard CR2032 coin cells (as illustrated in Fig. 4d) for cycling due to the unique geometry of the electrode tab. The coin cells may then be disassembled after cycling for post-mortem imaging, facilitating time-resolved studies where the same particles may be tracked as a function of cycle life. The advantages of this proposed workflow will be demonstrated in a later section, where virtually no impact on electrochemical performance is observed when compared to coin cells containing standard 15 mm electrodes.

In addition, whilst ex situ investigations are unable to capture dynamic processes occurring over short time-scales (for instance, charge balancing upon cell relaxation may mask some of these transient effects), operando 2- and 3-D measurements are the ultimate goal, directly linking material heterogeneities to degradation by allowing the same sample to be tracked unperturbed across time. Tsai et al. pioneered a microelectrode technique, enabling single-particle TXM imaging of LiNi_0.33Co_0.33Mn_0.33O₂ (NCM) and direct measurement of electrochemical kinetics for the first time.¹⁹

To investigate the electrochemical performance of the laser-milled tab electrodes, CR2032 coin cells were made with the tab electrodes in a half-cell arrangement with lithium metal as the negative electrode. Control coin cells were also made in the same cell arrangement containing standard 15 mm electrodes cut from the same electrode sheet. As evident in Fig. 6a, the tab electrode exhibited excellent electrochemical performance when cycled to 4.2 V vs Li/Li⁺ at a C/20 rate for an initial 3 cycles, with low hysteresis between the charge and discharge voltage profiles.

Zoom In Zoom Out Reset image size

The initial charge capacity of the tab electrodes was ca. 175 mAh g⁻¹, with a discharge capacity of ca. 150 mAh g⁻¹. The control coin cells demonstrated an initial charge capacity of 190 mAh g⁻¹ with a discharge capacity of ca. 160 mAh g⁻¹. Whilst efforts were made to minimize variations in mass loading by cutting electrodes from the same sheets, it is inevitable for there to be some local variations of mass loading even within the same electrode sheet. This is exacerbated by the significantly smaller electrode produced by the laser milling technique, which may account for the variations seen here. Rate capability tests were also performed on the tab electrodes cycled to cut-off voltages of 4.2 V and 4.3 V vs Li/Li⁺, and shown in Fig. 6b. The C-rate was varied between C/10 and 2 C with 5 cycles at each C-rate, and full recovery of capacity was observed upon returning to lower C-rates.

Electrodes produced with the original laser-milling technique result in a tomogram with the current collector in plane to the XY virtual slice as shown in Figs. 7a and 7b, compared to the XZ virtual slice for the laser-milled electrode samples as shown in Figs. 7c and 7d. As discussed previously, samples produced with the laser-milling technique have a centre of rotation that is orthogonal to the current collector plane and therefore the main limiting factor in sample size is transmission, as well as the need to horizontally stitch radiographs during acquisition if a larger sample volume is to be imaged. This is challenging to implement because the sample has to be translated for each angular projection which may induce significant sample drift. On the other hand, laser-milled electrode tabs circumvent these issues as transmission is no longer a controlling factor (i.e. samples have a cross-section of ca. 80 × 80 μm²). To acquire a larger volume, vertical stitching can be used which involves the acquisition of separate tomograms with a translation in sample height after each tomography and is easily implemented.

Zoom In Zoom Out Reset image size

The advantages of acquiring and stitching multiple volumes at high resolution become visibly apparent from Fig. 7, where four nano-CT datasets were acquired and merged, and appear as a fully continuous dataset. The well-aligned optics of the instrument used allows for a minimal overlap between tomograms of less than 10% without any significant artefacts in the regions where the tomograms overlapped. Whilst some variations in intensity are inevitably observed in the transition regions due to intensity fall-off at the edge of each radiograph caused by uneven illumination, these variations are sufficiently minute so as not to affect image segmentation.

A typical image analysis workflow is presented in Figs. 8 and 9, with detailed steps discussed in the Methodology section. Whilst an extremely high resolution is attainable in nano-CT, enabling the visualisation of sub-particle defects, typical commercial electrodes are composed of particles between 1 to 10 μm in diameter and thus only a limited population of particles may be captured within the less than 100 μm field of view of lab-based nano-CT.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

These particles may not necessarily be representative of the entire electrode because there may be localised variations in mass loading amongst other inhomogeneities. Furthermore, particles that are not fully captured within the tomogram are often removed before quantification of the segmented label field, as illustrated in Figs. 9c and 9d, further reducing the population of particles available for analysis.

By allowing an arbitrarily large number of datasets to be vertically stitched, limited only by equipment time, the analysis of a much larger region of an electrode is possible at the highest possible resolution. The larger number of particles captured significantly improves the statistical relevance of any subsequent image analysis and the overall representativeness of the data as obvious when comparing between Figs. 9e and 8d.

As shown in the radiographs (Figs. 5a and 5b) and volume rendering of the stitched tomograms (Fig. 5c), there are no differences in attenuation contrast between the edges of the electrode exposed to the laser compared to the bulk of the electrode. As attenuation contrast is a function of mass density, it can be concluded that there are no density variations induced by laser cutting (i.e. no particles have melted as a result of laser cutting). From Fig. 7c, some loss of particles can be observed at the edge of the electrode along with minor pulverisation of smaller particles in the beam path within a few microns of the edge of the electrode. However, these processes are also observed in conventional die cutting of larger electrodes, where material at the clearance between the die and the punch may be pushed or crushed. Thus, the level of precision of technique results in highly reproducible cutting with negligible heat-induced damage to the sample, and it is anticipated that there would also be minimal effects on the electrochemical properties of the sample.