In the journey to develop safe, high energy density batteries, many systems are currently being studied, such as Li–sulfur, Li–O₂, or all-solid-state Li devices (Whittingham, 2004; Li et al., 2017; Bruce et al., 2012). A common feature of those batteries is the use of a lithium (Li) metal negative electrode. Indeed, Li metal is one of the most promising electrode materials, thanks to its high capacity (3,860 mAh g⁻¹), low density (0.534 g cm⁻³), and low electrochemical potential (−3.04 V vs. standard hydrogen electrode) (Cheng et al., 2017). The industrialization of Li metal electrodes with a conventional liquid electrolyte has been hindered by the heterogeneities of the Li electrodeposit onto the Li surface, which limits the Coulombic efficiency (ratio between the discharge and charge capacity) and battery cycle life, and may grow sufficiently in the form of dendrites to induce a short circuit (Aurbach et al., 2000; Li et al., 2014).

To move toward an optimized Li–electrolyte interface, understanding the origin and growth of dendrites is of the utmost importance (Li et al., 2014). Indeed, depending on the nature of the electrolyte and the passive layers (Cheng et al., 2016) formed on the Li metal, different families of dendrite morphologies are observed, such as mosses, needles, and globules (Harry et al., 2014; Wood et al., 2016; Cheng et al., 2017). Furthermore, the dendrite microstructure depends on the operating and cycling conditions such as temperature or current density (Bai et al., 2016).

The most usual techniques to study dendrites are based on optical microscopy (Brissot et al., 1999a), electron microscopy (Gireaud et al., 2006), atomic force microscopy (Morigaki and Ohta, 1998), or magnetic resonance imaging (Ilott, 2016), to name a few. However, most of them are two-dimensional and invasive by nature and may alter the probed materials and interfaces. Noninvasive techniques based on X-ray imaging which take advantage of the prevalent interaction of X-ray with the outer electron shell of atoms are now more accessible, thanks to the development of dedicated synchrotron X-ray beamlines and the widespread diffusion of laboratory tomographs (Maire et al., 2001; Pietsch and Wood, 2017). X-ray tomography is a well suited to study, qualitatively and quantitatively, soft material interface. As an example, Balsara et al. studied model Li symmetric cells and Li-based batteries using synchrotron hard X-ray to report on dendrite growth and volume (Devaux et al., 2015; Harry et al., 2014). Therefore, X-ray tomography is widely reported for studying battery materials (Wood, 2018; Vanpeene et al., 2020; Magnier et al., 2020). In addition, the temporal and spatial resolution obtained from synchrotron sources are high enough to be compatible with electrochemical processes, leading to the development of in operando imaging via dedicated electrochemical cells (Schröder et al., 2016; Grey and Tarascon, 2017; Sun et al., 2016; Foroozan et al., 2020).

Despite the high spatial and temporal resolution of X-ray imaging techniques, some limitations arise due to the low sensitivity toward light Li compared to materials with a high atomic number, which are typically encountered in current collectors (Al and Cu) or active materials (Co, Fe, Mn, and Ni). As a complementary noninvasive technique, neutron imaging, which is based on the specific interactions of atomic nuclei with a neutron beam, gives useful additional information, thanks to a high penetration depth capability with regard to materials comprising high atomic number elements as well as high Li visibility and sensitivity to its isotopes (Strobl et al., 2009; Banhart et al., 2010). The theoretical linear neutron attenuation coefficient of ⁶Li is about 15 and 2,000 times larger than that of natural Li and ⁷Li, respectively (Sears 1992). As an example of in operando neutron radiography, Wang et al. (2012) imaged a planar battery made of a highly oriented pyrolytic graphite electrode with ⁶Li as the counter electrode at a resolution of 50 µm. By taking advantage of the beam attenuation induced by ⁶Li atoms, the graphite electrode areas enriched with ⁶Li were followed over the course of a galvanostatic experiment.

Neutron tomography was first applied to commercial batteries either to quantify gas in an alkaline cell or to probe the internal volume modification of the battery cell upon cycling (Manke et al., 2007; Butler et al., 2011; Senyshyn et al., 2012). So far, neutron tomography remains an ex situ and in situ technique as multiple hours are needed to perform a sample acquisition. The Li spatial distribution is typically quantified within the cathode (Nanda et al., 2012; Ziesche et al., 2020a), anode (Zhang et al., 2017), and battery cell (Ziesche et al., 2020) by neutron tomography. Riley et al. (2010) performed in situ neutron tomography at a resolution of 150 µm to report on the jellyroll volume change in the Li primary cell. To get better quantitative data, several optimization strategies were proposed by the authors, such as designing a specific electrochemical cell comprising deuterated electrolytes and isotope-enriched Li electrodes. Indeed, ⁶Li and ⁷Li are strongly attenuating and almost transparent to the neutron beam, respectively.

To get specific information within a battery by neutron imaging, dedicated cells and setups (Owejan et al., 2012) are needed to increase the spatial and temporal resolution, that is, get information in times compatible with typical beamtime allocations to move toward in operando analysis. Several strategies have already been explored, such as an electrochemical cell casing made of a neutron-transparent polytetrafluoroethylene (PTFE) casing (Owejan et al., 2012), an adapted pouch bag cell (Knoche et al., 2016), or a coin cell (Sun et al., 2017).

To combine X-ray and neutron information, Sun et al. (2019) designed a cylindrical polyamide-imide cell casing having an inner and outer diameter of 3 and 6 mm, respectively. The cell is optimized to image Li–air batteries for in operando synchrotron X-ray (∼1.2 µm resolution) and in situ neutron (∼13 µm resolution) tomography experiments. The complementarity of the two techniques revealed the degradation mechanisms at play at the Li/separator interface. Moreover, Song et al. (2019) performed in operando neutron radiography and in situ tomography on a Li/LiMn₂O₄ battery using a TiZr-based cell to study the Li dynamic distribution upon the dendrite short circuit. The electrochemically active assembly was made of a neutron-transparent ⁷Li metal and deuterated electrolyte in order to enhance the contrast between the Li electrodeposits and the ⁷Li electrode and to reduce the scattering of hydrogen.

In this work, by taking advantage of the capability of a neutron imaging beamline and of the Li isotope sensitivity, we report an in situ neutron tomography analysis of Li electrodeposits. A Swagelok-type electrochemical cell with a PTFE casing specifically designed for neutron imaging comprises a Li symmetric cell made of a natural Li (92.4 wt % ⁷Li and 7.6 wt % ⁶Li) electrode, a ⁶Li (95 wt % ⁶Li and 5 wt % ⁷Li) electrode, and a deuterated liquid electrolyte. ⁶Li was selected to increase the contrast of the deposits from the deuterated electrolyte. The ⁶Li electrode (anode) was oxidized, and ⁶Li was plated onto the opposite natural Li electrode (cathode) to induce dendrite growth. The neutron images are then compared to X-ray tomography analysis of the same cell acquired both at an X-ray synchrotron beamline and in a laboratory X-ray tomograph with a voxel size of 0.64 and 2.0 µm, respectively. In addition, the deposit volume is quantified here and positively compares to the measurement obtained using X-ray techniques. Despite the needed optimization to get further image quality and accelerated acquisition time, this in situ neutron tomography experiment is a proof of concept showing the possibility of studying battery failure modes via neutron imaging.

To validate the design of the electrochemical cell, a series of uncycled electrochemical cells was imaged by X-ray tomography at the laboratory source to get insight into the quality of the active cell assembly (⁶Li/electrolyte/Li). Figure 1 shows a tomography of a cell cross section acquired at room temperature. In this figure, the grayscale value for each pixel is the attenuation coefficient of the corresponding portion of space. So, for example, heavy elements such as Al that strongly absorbs X-rays appear brighter. The cell casing made of PTFE is transparent enough to the X-ray so that the active cell and the Al parts (piston or shim) can be distinguished. Due to the slightly higher absorption of PFA than Li and the polychromatic nature of the laboratory X-ray source, reconstruction artifacts likely due to the cone beam or beam hardening (Maire and Withers, 2014) are observed. Indeed, the upper Li electrode seems to exhibit two different attenuation values, that is, the pixels corresponding to the Li electrodes close to the Al materials appear more attenuating than they should be. In addition, the natural Li electrode located on the top of Figure 1 has similar gray values to the bottom one made of ⁶Li. As expected, X-ray imaging is not sensitive to the Li isotopes, making it impossible to distinguish between these two electrodes (⁶Li vs. natural Li). The presence of two domains located within the PFA tube can also be noticed, one with gray values close to those of the PFA and ascribed to the liquid electrolyte, while the other one presents gray values similar to those of the Li electrodes, suggesting the presence of an Ar bubble. This is explained by the delicate cell assembly that may entrap Ar bubbles in the liquid electrolyte. Despite the presence of a bubble in Figure 1, the Li electrodes are not perfectly flat as the surface of both Li electrodes facing the liquid electrolyte tends to adopt a spherical cap shape. This feature observed on all cell replicates is probably due to the softness of Li compressed within the PFA tube due to the spring pressure and to a mediocre adhesion of Li with the Al parts. The study of the uncycled cells by X-ray tomography highlights the importance of the 3D nature of these analyses, as the inner cell geometry must be accounted for in electrochemical and impedance studies. Indeed, the dome surfaces lead to an effective surface and an effective cell constant that will differ from the geometrical ones. We can consider that the two electrodes formed equivalent spherical cap spheres within the PFA tube, having a radius (r) equivalent to the PFA inner tube radius, a height (h), and separated by a distance (l) equivalent to the PFA tube height. In such a configuration, the effective cell constant (k) is estimated using the equipotential current line (Hong et al., 2019) between two domes given by the following equation: k = ∫ h l − h 1 π . r . d x x = 1 π . r . ln ( l h − 1 ) . (1) By measuring the distances r, h, and l by laboratory X-ray tomography imaging on a series of four uncycled cell replicates, the average effective cell constant k is estimated to be 6.9 ± 0.1. Impedance spectroscopy was then performed on the uncycled cell replicates at 25°C. Considering the effective surface, Figures 2A,B show the impedance spectra in Nyquist coordinates along with the characteristic frequencies of two typical cells, denoted cell_A and cell_B, respectively. Each spectrum presents two semicircle loops, one in the high-frequency range between 7 MHz and 100 kHz and the other one at lower frequencies. The high- and low-frequency loops are ascribed to the electrolyte (R _el) and Li/electrolyte interface resistances, respectively (Bouchet et al., 2003). The spectra were fitted with an electrical equivalent circuit composed of elements in series corresponding to the cable resistance and inductance, and the electrolyte response as R _el in parallel with a constant phase element, followed by another resistance in parallel with a constant phase element corresponding to the Li/electrolyte interfaces. The equivalent circuit is displayed in the inset of Figure 2A. The electrolyte and interface resistances of cell_A are higher than those of cell_B by a factor of 7 and 2, respectively. These higher resistances are ascribed to the presence of an Ar bubble entrapped within the PFA tube due to a contact surface reduction, which also induces a frequency dispersion as the bubble possesses different dielectrical properties. Regarding the electrolyte resistance of cell_B, the ionic conductivity (σ) at given temperature (T) is calculated using the following equation: σ ( T ) = k R e l ( T ) . (2) By combining Eqs 1, 2, the ionic conductivity of cell_B at 25°C is found to be 6.6 × 10⁻³ S cm⁻¹. For comparison, the conductivity of the sole deuterated liquid electrolyte (1M LiTFSI/dDMC) was determined using the conductivity cell for microsamples (see electrolyte conductivity in the experimental section). Supplementary Figure S2 shows σ of the deuterated electrolyte as a function of the inverse of the temperature between −25 and 60°C. In the temperature range explored, the conductivity presents two different domains with a transition at about −5°C, corresponding to the electrolyte crystallization. At 25°C, the conductivity is 6.6 ± 0.2 10⁻³ S cm⁻¹. This value is consistent with the one reported by Dahbi et al. (2011) of about 7.0 × 10⁻³ S cm⁻¹ for a non-deuterated 1M LiTFSI/DMC electrolyte at 24.8°C. Therefore, the conductivity of cell_B at 25°C lies within the error bar of the value measured using the conductivity cell. Consequently, the analysis of the electrolyte conductivity of the uncycled ⁶Li–Li cell by impedance spectroscopy in combination with X-ray tomography validates the design of the electrochemical cell and allows the identification of assembly defects (deformed electrodes and entrapped gas bubbles) that should be taken into account for a future cell assembly optimization.

The cell replicates were then cycled at a constant low current of 8.4 µA in order to move ⁶Li (anode) onto the natural Li electrode (cathode). Considering the effective surface, cell_B was cycled at 0.53 mA cm⁻². For cell_A, the bubble within the liquid electrolyte induces complex current lines between the two electrodes, and thus an uncertain current density value. The result of the galvanostatic step performed on cell_A and cell_B is provided in Figure 3, showing the cell voltage (E in V vs. Li⁺/Li) as a function of time (t). The voltage profile of cell_B is almost constant at 0.1 V vs. Li⁺/Li for 16 h, and then, it increases until a plateau at about 0.45 V vs. Li⁺/Li. For cell_A, the voltage profile first presents a sharp increase after 5 h with a peak at about 1.7 V vs. Li⁺/Li at t = 15 h, before decreasing to 0.2 V vs. Li⁺/Li at t = 27.5 h. For t > 27.5 h, the cell voltage becomes similar to that of cell_B, with E being constant at 0.2 V vs. Li⁺/Li, followed by a slight increase up to a final value of about 0.53 V vs. Li⁺/Li, in the same range as the one reached by cell_B. The voltage fluctuation and high polarization of cell_A during the first 27.5 h is attributed to the presence and the movement of the Ar bubble within the PFA tube during the electrochemical step, leading to a geometrical factor modification coupled with an increase of the Li/electrolyte interface due to heterogeneous electrodeposits (Wood et al., 2016).

A cell voltage behavior similar to the one observed for cell_B was reported in the works of Brissot et al. (1999b) and Rosso et al. (2001), focusing on the study of Li dendrite growth mechanisms in Li symmetric cells comprising a polymer electrolyte. The authors show that below a critical current density (J ^*), the ionic concentration profile, and thus the concentration gradient, throughout the cell evolves to a steady state. The cell voltage attains a constant value which eventually drops to zero when a dendrite makes contact between the two electrodes. For J < J ^* , the morphology of the electrodeposits observed by Brissot et al. (1999a) is tortuous, forming mostly mossy Li with the presence of needle-like dendrites. This behavior was partly confirmed by Bai et al. (2016) on a Li symmetric cell comprising a liquid electrolyte with an electrode interdistance in the range of several millimeters where the electrodeposited Li morphology remains mostly mossy if the current density is below J ^* . Based on the work of Brissot et al. (1999b), J ^* is defined as follows: J ∗ = 2. F . C . D ( 1 − t + ) . l , (3) with F being the Faraday constant, C being the concentration of salt in the electrolyte, D being the ambipolar diffusion coefficient, and t ⁺ being the cationic transference number.

By considering that the ionic transport properties of the deuterated liquid electrolyte are similar to those of its non-deuterated counterpart, we can take C, t ⁺, and D values equal to 1,000 mol m⁻³, 0.46, and 3.46 10⁻⁶ cm² s⁻¹, respectively (Borodin and Smith, 2006). Using Equation 3 and considering a distance l equal to the typical PFA tube length of 2 mm, J ^* is 5.8 mA cm⁻². Consequently, the imposed current density of cell_B corresponds to 9% of J ^* , and the voltage behavior falls within the J < J ^* case. For cell_A, despite the effect of the Ar bubble, its cell voltage profile also corresponds to the J < J ^* case. Indeed, even by considering an active surface as being a minor fraction of the effective surface, the current density remains below the J ^* value. Moreover, for cell_B, the total amount of charges passed during the 54.7 h of the galvanostatic step corresponds to a theoretical displacement of 139 µm of ⁶Li based on the Faraday law and considering the effective surface. This value is close to the ⁶Li electrode thickness used in cell_B (145 µm), which means that a quasi-total ⁶Li oxidation was performed during current imposition. Therefore, for both cells, their current densities used for the Li transfer from the anode to the cathode lie in the J < J ^* case, and the length of the PFA tube permits the oxidization of most of the anode (⁶Li) onto the cathode (natural Li) without dendrite short circuit.

At the end of the galvanostatic step, a final impedance measurement was performed on cell_A and cell_B, and the corresponding spectra are added in Figures 2A,B, respectively. For cell_A, the total cell resistance is diminished by a factor of 3 compared to its initial state, with a strong reduction of the interface resistance by a factor of 18 in the lower frequencies range. This situation is certainly due to an increase in the active surface on the cathode via the growth of mossy or dendritic electrodeposits (as shown later by neutron imaging in Figure 4) and a modification of the solid electrolyte interphase (Cheng et al., 2016) on the anode (Rosso et al., 2006). In addition, the electrolyte loop is depressed with a resistance value similar to the initial one, which is ascribed to the presence of both Li deposits and bubbles within the liquid electrolyte, inducing a strong dispersion in the characteristic frequencies. For cell_B, the total cell resistance increased by a factor of 1.4 with the main change observed for the electrolyte resistance (high-frequency loop), while the characteristic frequencies of the two loops are modified. This behavior may be due to a cell constant modification during the galvanostatic step. Based on Figure 1, the ⁶Li takes the form of a dome on the top of the Al current collector, so the cell constant will gradually change toward a higher constant value while the anode (⁶Li) is oxidized, and conversely, electrodeposits are formed on the cathode, until most of the ⁶Li is consumed, inducing a hole in the anode.

After the galvanostatic step, the electrochemical cells were taken to the neutron tomography beamline for in situ imaging at room temperature. Therefore, a series of radiographies was first acquired at a large field of view and at different rotation angles in order to quickly inspect each cell. Except cell_A, the radiographies of cell_B and other replicates revealed a damaged assembly and thus were discarded for further tomography imaging. For cell_A, a typical radiography is reported in Figure 4A, where at each pixel, the gray value is proportional to the neutron attenuation of the corresponding region of space. In this image, neutron-transparent materials such as Al (piston and shim) and PTFE (cell casing) appear brighter (given that most of the neutrons are transmitted to the detector) than materials with stronger neutron attenuation such as the stainless steel spring or Li electrodes (Sears 1992). As expected, the Al and PFA materials are suitable for neutron imaging as their attenuation is essentially negligible. The top and bottom Li electrodes appear black in between the Al parts and the PFA tube. In addition, the PFA inner surface is visible from the overall background because of the presence of absorbent materials within the deuterated electrolyte from top to bottom, as detailed in Figure 4A: the natural Li electrode (the cathode in the galvanostatic step) with a spherical cap shape as observed in the uncycled cell by X-ray tomography (Figure 1), the ⁶Li deposits in the electrolyte, and a less attenuating portion at the bottom, suggesting the presence of a bubble. It can be noted that at the bottom of the PFA tube, the ⁶Li electrode is almost absent from the radiography, in agreement with its quasi-total oxidation during the electrochemical step (Figure 3). Moreover, at the interface between the PFA tube and the Al parts on both sides of the cell, a highly attenuating material is covering the Al surface and its edges. This is attributed to a chemical and electrochemical degradation of Al areas in contact with Li metal. Tahmasebi et al. (2019) recently reported on the degradation of the Al surface associated with the formation of the LiAl (β) phase, either during electrochemical lithiation or by simple contact between Al and Li metal. Cell_A was subsequently imaged in tomography for 18 h. Figure 4B shows a typical vertical slice of the tomography, showing a cross section of cell_A. The values of the voxels in a tomography are related to the attenuation coefficients of the corresponding volume of space, with higher values for highly attenuating materials, which is the opposite of the radiography images. As observed in the radiography, both Al parts in contact with Li are highly attenuating, indicating the presence of the LiAl phase. At the bottom part, the few ⁶Li anode remainders appear in bright white. The gray values of the ⁶Li throughout the reconstructed volume are the highest of all observed materials in the active cell, with some gray saturation at some voxels as shown in the bottom right corner. As expected, the gray values associated to the natural Li electrode (the bended electrode at the top of Figure 4B) are lower than those of the ⁶Li, as its theoretical neutron attenuation coefficient is 15 times lower (Sears 1992). This result shows that the Li isotope sensibility to neutrons is an effective tool to differentiate between Li electrodes (⁶Li vs. natural Li) after an electrochemical step. Moreover, inside the electrolyte volume in Figure 4B, a highly neutron-attenuating, irregularly shaped domain is observed and attributed to ⁶Li deposits, appearing in light gray and white, originating from the natural Li electrode and spreading throughout the electrolyte. In this experiment, the attenuation values from the ⁶Li deposits are in the same range as those of the natural Li electrode and of the Li reacted with Al, leading to difficulties in distinguishing between each of these domains. Indeed, even if the voxel size is larger than the characteristic diameter of mossy Li or Li dendrites (Bai et al., 2016), the ⁶Li deposits are visible because of the partial volume effect (Santago and Gage, 1995; Kaestner et al., 2013). Typically, the attenuation coefficient of voxels partially occupied by both deposits and electrolyte will be intermediate between the attenuation of the individual components. The ⁶Li deposits are also surrounding a dark gray and black volume with gray values close to those of the background air and having a bulb shape confirming the presence of a bubble.

From the tomography imaging, the active cell volume can be segmented into different components, based on their gray level. The Al domain can be easily segmented as it has significantly lower attenuation values. The segmentation of the other domains containing Li (deposits, cathode, and the LiAl alloy) is more delicate, given the comparable attenuation values. The thresholding limits between these are therefore more arbitrary. Figure 4C corresponds to the segmentation of the slice shown in Figure 4B, while a 3D volume rendering is shown in Figure 4D. The ⁶Li deposits within the electrolyte take the form of a heterogeneous ramified structure. Besides this graphical representation, the segmentation allows a quantitative analysis. The overall deposit volume can be calculated by summing up all the voxels attributed to this material. The result of 0.48 ± 0.10 mm³ is given in Table 1. This value is overestimated as the image resolution is larger than typical electrodeposit dimensions (Bai et al., 2016). It follows that, depending on the threshold value chosen for the segmentation, an overestimation of the deposit volume is probable, which tends to increase with the amount of porosity. So, even if the microstructure has typical dimensions smaller than those of the tomography resolution, this proof-of-concept experiment highlights the potential of neutron imaging in determining the overall shape and nature of heterogeneous electrodeposits as well as cell assembly defects within an electrochemical cell comprising X-ray attenuating materials.

To get further insight into the fine structure of the deposits, cell_A was then imaged at a synchrotron X-ray tomography beamline. A typical tomography image of the entire cross section of the active cell is shown in Figure 5A. Thanks to the voxel size of 0.64 µm, the ⁶Li deposits are clearly seen as they attenuate less X-rays than the liquid electrolyte. The ⁶Li deposits develop a needle-like dendritic morphology in accordance with the finding of Brissot et al. (1999a). In addition, the LiAl domain is easily discernible at the top and bottom parts of the cell, showing interface degradation. A segmentation process was then performed on the X-ray images, and the dendrites in the electrolyte were segmented based on their gray level (very dark dendrites). However, synchrotron tomography has a very coherent X-ray source, leading to dark/light fringes due to phase contrast (Vanpeene et al., 2020). As seen in Figure 5A, the interfaces of the curved Li foil have different gray levels. An automatic grayscale segmentation of the curved foil was therefore unsatisfactory and was thus manually corrected. An illustration of this process is provided in Figure 5B, and the resulting 3D rendering volume is provided in Figure 5C. From this segmentation process, the dendrite volume was calculated to be 0.14 ± 0.01 mm³, a value also included in Table 1. A factor of 3.4 separates the dendrite volumes measured by neutron and synchrotron X-ray analysis, which is mainly due to the difference in imaging resolution.

For completeness, cell_A was imaged at the laboratory X-ray tomograph with a voxel size of 2.0 µm. In Figure 6A, a cross-sectional image corresponding to the same area as the one reported in Figure 5A is shown. Similar observations of the cell can be made compared to the synchrotron X-ray and neutron-based images. It should also be noted that the cell was imaged at the neutron and X-ray beamlines, and laboratory tomograph within few weeks. The morphology of the electrodeposit is then preserved over time. Therefore, post-cycling analyses can be performed with confidence several days or weeks after the electrochemical experiment. After the scan, the segmentation process was performed and the dendrite volume was reconstructed in 3D as depicted in Figures 6B,C, respectively. An Li dendrite volume of 0.15 ± 0.02 mm³ was calculated, a value also reported in Table 1, which is in the same range as the one obtained from synchrotron X-ray imaging. This result is in agreement with the fact that the signal-to-noise and contrast-to-noise ratios of a laboratory tomograph are lower than those for synchrotron beamlines, leading to lower resolution images with lower contrasts (Vanpeene et al., 2020). The dendrite volumes measured with the X-ray sources are similar, while at a larger voxel size, by neutron tomography, the measured dendrite volume is higher by at least a factor of 3 with higher uncertainty. This difference is attributed to lower resolution of the neutron tomography images, leading to the selection of voxels comprising a mixture of Li dendrite and electrolyte (partial volume effect) during the segmentation process. While the neutron imaging resolution is unavoidably lower because of the technical constraints of the technique, the capacity of neutrons to discern different isotopes can be exploited to pinpoint the origin and the composition of the Li comprising the dendrites. It would then be of interest to distinguish the chemical composition within dendrites between Li originating from the anode (⁶Li in this study) and that from the Li salt of the electrolyte (mainly in ⁷Li here).

We report an in situ neutron tomography imaging of Li electrodeposits in a cycled ⁶Li–Li cell by taking advantage of the capability of the neutron imaging beamline and the intrinsic contrast of Li isotopes to neutrons. This proof-of-concept experiment shows that the overall morphology of the deposits can be captured by neutron tomography. The deposit volume could be measured and compared to those obtained using X-ray techniques. To go further, several experimental parameters need to be optimized to move toward a higher resolution and contrast in shorter acquisition times. For the proposed electrochemical cell, the use of a neutron-transparent buffer metal layer in between the Al parts and Li metal while maintaining chemical and electrochemical stability is a necessity, as is the reduction of the cell diameter. As for the beamline, the foreseen increase in capabilities will permit the closing of the resolution and contrast gap between neutron and laboratory X-ray tomograph. In addition, the study of uncycled cells also highlights the importance of 3D analyses in order to validate the design of electrochemical cells as they allow the identification of assembly defects such as deformed electrodes or entrapped gas bubbles that must be taken into account for electrochemical and impedance measurements. In situ neutron tomography permits the envisioning of the thorough study of battery failure modes in electrochemical cells comprising X-ray attenuating materials. The capacity of neutrons to discern different isotopes can then be a tool used to get insight into the origin and the composition of the Li deposits by distinguishing Li from the anode (⁶Li is this study) and that from the Li salt electrolyte (mainly in ⁷Li here).