Visualize Battery Dynamics with Multimodal Imaging

In this perspective, we propose a multimodal approach for visualizing the charge and discharge cycle of batteries. In the first imaging mode, the redox reactions at the positive electrode are imaged by the associated magnetic changes when ions and electrons are removed from (added to) the positive electrode during charge (discharge). In mode two, the flow of active ions and electrons is detected by the Ørsted magnetic field formed by the moving charges. In mode three, the redox reactions at the negative electrode are spatially resolved by tracking the transformation of active ions into paramagnetic metals. We provide finite element simulations of a Li|LLZO|LixCoO2 solid-state battery, which collectively depict the magnetic stray fields arising from these processes as described in the following sections. Following this, we also discuss the prospects of gaining insight into battery operation and degradation using magnetically ordered phases.

Positive electrode reactions

Most active materials in the positive electrodes of batteries contain transition metals like Co, Mn, Ni, or Fe, since their multivalent redox states can accommodate the removal/addition of electrons by changing their 3d electron occupancy. In general, one can classify battery materials as 1D, 2D, and 3D ionic conductors with Table 1 outlining common lithium battery materials as well as their structural, electronic, ionic, and magnetic properties at room temperature. Common positive electrode active materials include 1D polyanion oxides such as LiFePO472, 2D layered oxides such as LiCoO2 and nickel-rich LiNixMnyCo1-x-yO2, and 3D spinel structures such as LiMn2O4. As illustrated in Fig. 3a, the transition metals are generally magnetically active owing to the five correlated and energetically accessible 3d states, which promote rich magnetic phase diagrams for these material classes. The magnetic properties are primarily determined by the number of 3d electrons per transition metal atom and the spin arrangement of these electrons. For positive electrodes, the 3d electron occupancy is dictated by the choice of transition metal ion and its redox state. During battery charge and discharge, the redox state of the transition metal ions changes to balance charges after structurally accommodating or releasing lithium. As a consequence, the magnetic properties can be used as a proxy to monitor redox reactions at the positive electrode and state-of-charge through changes in their spin states, see Fig. 3a73,74,75,76,77. For the layered oxides and LiFePO4, the transition metal ions are octahedrally coordinated with the surrounding ligands, and low- or high-spin configurations form as a balance between the crystal field splitting between eg and t2g states and the energy gained by spin alignment (Fig. 3a).

The spin configurations in the positive electrode materials may be affected by a number of additional factors such as polymorphic transitions, distortions in bond angles, and defect formation. This can be leveraged to gain insights into important aspects such as Li-Ni interdiffusion in positive electrodes and lattice distortions32, while also complicating interpretation.

For the 2D Li-conducting LixCoO2, the lithiation degree is balanced by changing the redox state of the transition metal ion between Co3+ and Co4+. At room temperature, LixCoO2 is paramagnetic at all lithiation stages, but the paramagnetic susceptibility changes significantly with the state-of-charge73 as illustrated in Fig. 3b. The experimental procedure, battery cycling curves, and raw magnetometry data behind Fig. 3b are provided in Supplementary Section S1. The lower value of the paramagnetic susceptibility for electronically insulating LiCoO2 represents Van Vleck paramagnetism73. Upon delithiation, LixCoO2 transitions into a Pauli paramagnet at room temperature with a monotonously increasing paramagnetic susceptibility in most of the reversible state-of-charge window (Fig. 3b)73. Irrespective of the charge state, the magnetic susceptibility of LixCoO2 exceeds that of most elements used in batteries, as outlined in Fig. 3c, which enables the detection and distinction of the LixCoO2 charge state in the presence of other elements. Figure 3c also serves as a guide for selecting current collectors and interfacial layers with the lowest magnetic susceptibilities to ensure minimal influence when performing magnetic imaging. We note that Fig. 3c illustrates the magnetic susceptibility of various elements, which is complemented by Table 1 providing the magnetic susceptibilities of battery materials.

NV magnetometry allows for imaging the local magnetic susceptibility in the presence of an external magnetic field using stray field imaging (process 1 in Fig. 2f), thus spatially resolving the redox process at the positive electrode. In Fig. 4a, we provide finite element calculations78 of the electrochemical processes and their associated magnetic stray field from a Li|LLZO|LixCoO2 solid-state battery. Here, LLZO fills the gaps between the LixCoO2 particles to promote redox homogeneity, but conductive carbon additives are not included in the simulations owing to the high electronic conductivity of LixCoO279,80. The configuration modeled corresponds to cross-sections of conventional solid-state batteries with electrodes on the top/bottom of an electrolyte or in-plane solid-state batteries with ionic transport taking place in the plane of an electrolyte thin film. Supplementary Section S2 describes the model in detail, including its experimental input parameters and geometrical inspiration from X-ray tomography and scanning electron microscopy studies81,82. We find the redox reactivity to be homogeneous when electrochemically charging the battery at low and moderate charge rates of C/10 and 1 C, which corresponds to nominal current densities of 0.09 and 0.94 mA·cm-2, respectively (Supplementary Section S2.5). However, at fast charge rates of 5 C (4.70 mA·cm-2), we observe a large gradient in the lithium concentration where the positive electrode particles closest to the negative electrode are delithiated faster, as displayed after 4 minutes of charging in the second panel of Fig. 4a where LixCoO2 remains in the reversible operating range (x > ½). The ionic conductivity of LixCoO2 varies non-monotonically by more than an order of magnitude depending on the lithium content, but the lithium-ion diffusion remains sufficiently high to promote a relatively homogeneous lithium composition within each LixCoO2 particle. Such redox heterogeneity is commonly observed for various positive electrode active materials6,7,24,48 and is consistent with scanning transmission X-ray microscopy revealing interparticle variations in the lithium content upon a fast 4 C delithiation in a liquid-state battery with Lix(Ni1/3Mn1/3Co1/3)O2 at the positive electrode7.

Applying a sample-wide magnetic field of 100 mT perpendicularly to the LixCoO2 surface induces a stronger magnetization in delithiated regions compared to lithium-rich regions, effectively resulting in a magnetic stray image revealing the underlying state-of-charge (top panel of Fig. 4a). Although this is superimposed with shape effects, the trend is even observed in positive electrode particles buried several hundreds of nanometers below the surface layer such as the lower, right particle in Fig. 4a. The stray field is calculated at a plane located 50 nm above the battery surface, which is consistent with typical NV stand-off distances. Supplementary Fig. S10 displays the results for the external magnetic field and the NV quantization axis both being perpendicular, parallel, or tilted 54.7° with respect to the sample surface, corresponding to diamond tips with common crystallographic terminations. Although the simulated magnetic stray field differs significantly depending on the external field and NV orientation, the stray field is on the order of 10 µT in all cases, which is detectable with CW-ODMR sensitivities of \(\sim\)1 µT·Hz−0.5 and the pulsed detection limits of \(\sim\)50 nT using scanning NV magnetometry with 1 second integration time58,63,83,84. The paramagnetic moments are unidirectional and aligned along the external magnetic field axis, which simplifies the reconstruction of the underlying magnetization from the measured magnetic field images. If the magnetization can be assumed to be 2-dimensional, it can be constructed uniquely from the measured stray magnetic field85, otherwise, the reconstruction becomes more involved41,86,87 and structural knowledge, such as particle sizes and shapes, can be an advantage. Owing to the paramagnetic properties of LixCoO2 in the positive electrode, the magnetic signal strength scales linearly with the applied magnetic field (Fig. 4b), which provides both the possibility of enhancing the signal strengths with larger fields applied using superconducting or permanent magnets as well as performing local ac susceptometry by delivering smaller alternating magnetic fields using the microwave antenna58. The linearity can further serve to establish the paramagnetic origin of the measured magnetic stray fields. We note that, as overviewed in Table 1 and Supplementary Section S1, many positive electrode materials change their magnetic susceptibility as a function of lithium content, which adds to the prospects of visualizing the local state-of-charge in positive electrode materials beyond LixCoO2.

Following the redox reaction at the positive electrode, the electrons flow to the negative electrode through the electrodes and external circuit, whereas ions flow to the negative electrode through the electrolyte. The electronic and ionic currents produce an Ørsted magnetic field (process 2 in Fig. 2f), which is simulated in Fig. 4a. The simulations are performed with a high input current density of 23.50 mA·cm−2 (charge rate of 25 C) as discussed later. The current redistributes near key defects such as lithium dendrites, void, grain boundaries, and chemical inhomogeneities88, which leads to distinct imprints in the stray field. The high electronic conductivity of the lithium metal dendrite focuses most current within the electrolyte in a narrow region, which eventually leads to battery short-circuiting if the dendrite extends to the electronically conducting positive electrode. At the positive electrode, the current distributes in response to the higher ionic conductivity in LLZO compared to LixCoO2, the distribution of grains and grain boundaries, and how the heterogeneous lithium content impacts the highly lithium-dependent ionic conductivities in LixCoO2. The current gradually becomes more focused in the positive electrode and electrolyte as it approaches the dendrite. In the absence of a conductive carbon matrix, electronic current hot spots are formed at the finite-sized contact points between LixCoO2 particles, while ionic current hot spots form in interconnected LLZO particle chains connecting the extended tips of the dendrite. With a nominal current density of 23.50 mA·cm−2, the magnetic signal strengths associated with this battery geometry are on the order of 1–5 nT, with larger signals around the dendrite and smaller signals in inactive battery regions.

Two key advantages of imaging this Ørsted magnetic field are that it directly reveals which defects and local features are important for the battery operation and that the buried current density can be quantitatively reconstructed from the measured stray field during operation85,89. If the current can be approximated to flow in 2D, it can be uniquely reconstructed using an inverse filter method85,89, whereas more complex 3D current distributions require inverse modeling41,90,91. The magnetic signal strength from the inhomogeneous currents at the dendrite, electrolyte, and positive electrode is presented in Fig. 4b as a function of the nominal current density and compared with NV magnetometry sensitivities83,92. The magnetic signals can be strengthened by increasing the current, but the application of high currents is generally limited by the growth of lithium dendrites above a critical current density93. Typical critical current densities in solid-state batteries range from 0.1 to a few mA·cm−2 with a 10-fold increase after negative electrode|electrolyte interface optimization5,94,95,96. In a recent report, a new high-throughput method for characterizing the critical current density has been proposed and tested on ceramic electrolytes withstanding current densities exceeding 300 mA·cm−2 97. At the local scale, extremely large current densities exceeding 10,000 mA·cm−2 have been reported near lithium dendrites5. Longer sample acquisition can be used to resolve the weak Ørsted magnetic fields by preparing the battery in a desired state using direct currents while probing the current distribution using alternating currents. The critical alternating current density is generally inadequately investigated in solid-state batteries, but studies have revealed that alternating currents can improve the performance by mitigating dendrite formation or preheating the battery98,99,100,101,102. Here, the use of alternating currents greatly increased the critical current density compared to direct currents, resulting, for instance, in a more than ten-fold increase when increasing the frequency from 0.1 to 100 Hz101,102. Although more work is needed to elucidate the alternating critical current densities, nominal alternating currents on the order of up to 10-100 mA·cm−2 may be expected in solid-state batteries, with higher values obtainable after employing strategies for enhancing the critical current density93,103. The local current densities near dendrites are expected to be 100–10,000 mA·cm−2. This results in measurement integration times ranging from sub-seconds to tens of minutes per pixel to resolve these currents with scanning NV magnetometry (Fig. 4b). The measurements can be further accelerated using more advanced protocols such as spin memory-based readout sequences, which have been demonstrated to yield a 10-fold improvement in the sensitivity of single and ensemble NVs104,105.

As illustrated in Supplementary Section S2.5, the magnetic signals associated with sourcing currents through only the electronically conducting electrode are 4 orders of magnitude larger than the ionically limited currents simulated in Fig. 4a. The electronic and ionic contributions to the magnetic field may be discriminated by performing temperature-dependent measurements to thermally activate or reduce ionic transport, probe the transport in various frequency regimes, or employing ion- or electron-blocking electrodes106. Further exploration of these possibilities could uncover the individual contributions of buried electronic and ionic currents at the nanoscale, which is generally not possible with conventional analytic tools. This holds great importance not only for understanding the role of electronic leakage in electrolyte grain boundaries for forming dendrites21,107, but also for probing the nanoscale impact of solid-electrolyte interphases and inter-/intraparticle electronic conduction in positive electrodes.

Negative electrode reactions

The use of pure metals as negative electrodes greatly increases the energy density, particularly in anode-free architectures where the negative electrodes are formed by battery cycling rather than during fabrication. When employing metals at the negative electrodes, the reduction of most active ions used for charge transport in lithium and post-lithium batteries concurs with a transition to paramagnetic pure metals. As illustrated in Fig. 3, this includes the majority of monovalent (Li+, Na+, K+), divalent (Mg2+, Ca2+), and trivalent (Al3+) ions.

Applying an external magnetic field of 100 mT produces a magnetic stray field from the paramagnetic lithium metal, which can be used to directly image the local state-of-charge, dendrites, and lithium metal inclusions in the diamagnetic LLZO electrolyte, as shown in Fig. 4a. Due to the stronger magnetic stray field from LixCoO2 compared to the Li metal, the paramagnetic fields from the lithium metal and LixCoO2 are simulated separately and plotted with two color bars. Similar visualization of dendrites and inactive metal inclusions is valid for the paramagnetic metals Li, Na, K, Mg, Ca, and Al present in diamagnetic electrolytes. To date, the magnetic properties of only a few solid electrolytes have been reported (see Table 1). In general, most electrolytes lack 3d transition metal ions, whereby it is reasonable to hypothesize that the majority exhibits diamagnetic characteristics based on their expected spin states108. However, Fe dopants and oxygen vacancies have been observed to induce paramagnetism in LLZO109,110, and hence, we encourage the field to further characterize the magnetic properties of electrolytes to form a solid future basis for visualizing metallic inclusions.

The signal strength as a function of the applied magnetic field is shown in Fig. 4b and compared with the detection limit of NV magnetometers. Although the predicted signals are one order of magnitude weaker for lithium metal compared to LixCoO2, their detection remains feasible in high external magnetic fields. Similar to the induced magnetization of the positive electrode, quantitative reconstruction of the underlying magnetization from the measured magnetic field images is also simplified by the unidirectional magnetization. Furthermore, the magnetic field from the magnetization non-perturbatively penetrates nonmagnetic protective layers and thin current collectors, which makes it possible to visualize redox processes at the negative electrode even when the air-sensitive pure metal negative electrodes are covered by protective layers.

Magnetically ordered phases

Most battery materials are considered either paramagnetic or diamagnetic at room temperature, but magnetically ordered phases may emerge particularly in two cases: First, as outlined in Table 1 (magnetic ground state), many positive electrode active materials order magnetically below transition temperatures on the order of tens of Kelvin. The characteristics of the magnetic order, including the transition temperature111,112, are sensitive to the lithium content and crystalline phase. Scanning magnetometry measurements performed below or in the vicinity of the transition temperature can introduce enhanced signal strength and spatial contrast in the magnetic stray field arising from spatial inhomogeneities affecting the ferromagnetic order or magnetic susceptibility. Cooling further freezes the battery in its present state, allowing for long integration times.

Second, imperfections, defects, and degradation products arise during battery synthesis or cycling, which can produce magnetically ordered states. Key examples include (1) the recent operando detection of Fe3O4 and Fe, which represents an impurity in LiFePO436, (2) Li-Ni interdiffusion in LiNixMnyCo1-x-yO2, whereby Ni located in the lithium layers can ferromagnetically couple two adjacent transition metal layers32,74, and (3) degradation products such as the degradation of LixCoO2 into magnetically ordered Co3O4 and CoO at high voltages and delithiation15,113,114.

NV magnetometry may complement established nano- and microscale imaging techniques for battery characterization, as outlined in Fig. 5a, with more information provided in Supplementary Section S3. In contrast to other imaging techniques, NV magnetometry provides a non-destructive view into the functional properties such as current imaging with sub-surface sensitivity. Scanning NV magnetometry is typically done with a fixed separation of several tens of nanometers between the sample surface and the sensor. However, larger distances to the NV sensor occur if the magnetic field sources are buried underneath current collectors, electrodes, or protective layers. In contrast to, e.g., transmission microscopy and most other scanning probe techniques, the magnetometer can detect features buried several micrometers into the battery, although the signals are weaker and resolved with a poorer spatial resolution on the order of the sensor/source distance (Fig. 4c). However, lacking the extended depth resolution of X-ray techniques, NV magnetometry is expected to support primarily the ex-situ characterization of the ionic, electronic, and magnetic properties of individual battery components as well as in-situ measurements on battery cross-sections, thin-film batteries, or exposed surfaces such as electrode surfaces. Interspersing nanodiamonds in the battery matrix may, however, provide an exception to this as recently demonstrated36, although the transmission of optical light and microwave to the battery interior, the dedicated sample synthesis, and the invasiveness may provide limitations hereof.

The practical realization of NV magnetometry depends on the NV architecture employed, but in all cases, the invasiveness should be investigated and potentially eliminated by optimizing the measurement protocol, particularly when performing operando imaging. For scanning NV magnetometry, an exposed surface with an acceptable roughness on the order of a micrometer and optional electrical connections are needed. Determining only the magnetic state of the individual battery materials is, hence, readily accessible without dedicated sample preparation beyond surface polishing. Advancing to current imaging, an important limitation is that scanning magnetometers are only sensitive to currents parallel to the surface, thus emphasizing the importance of cross-sectional or in-plane geometries as in Fig. 4a. When interspersing nanodiamonds into the battery matrix, care must be taken to minimize the invasiveness of the foreign material and avoid high temperatures in the sample synthesis. Lastly, when employing wide-field imaging, the rigid diamond surface and the sample need to be in close proximity, which can be realized by preparing the battery materials directly on diamond substrates, transferring the battery materials to the diamond, or placing small diamond pieces on the battery surface. Mature transfer techniques of van der Waals materials can further be employed when using spin defects such as boron nitride vacancy centers in few-layer hexagonal boron nitride for magnetometry, which can also benefit from reduced sensor-to-source distances115,116,117.

The subsurface sensitivity to the local distributions of currents and magnetization is inherited from probing the stray magnetic fields, which can be achieved using a range of techniques41. Figure 5b provides a comparison of selected, commercially available techniques, including scanning superconducting quantum interference devices (SQUIDs), wide-field and scanning NV magnetometers, and atomic magnetometers (see Supplementary Section S3 for further information). The improved detection limits of SQUIDs and atomic magnetometers are particularly advantageous when resolving the weak magnetic fields from batteries, but the low operating temperature of SQUIDs and the poor spatial resolution of atomic magnetometers constrain their application. NV magnetometers advantageously combine room temperature operation with nano- or microscale imaging resolution. The lower sensitivity of CW-ODMR provided in Fig. 5b can be improved using pulsed protocols and long integration times, particularly when employing wide-field NV magnetometry where the entire image is acquired simultaneously.