Atom probe analysis of electrode materials for Li-ion batteries: challenges and ways forward

The worldwide development of electric vehicles as well as large-scale or grid-scale energy storage to compensate for the intermittent nature of renewable energy generation has led to a surge of interest in battery technology. Understanding the factors controlling battery capacity and, critically, their degradation mechanisms to ensure long-term, sustainable and safe operation requires detailed knowledge of their microstructure and chemistry, and their evolution under operating conditions, on the nanoscale. Atom probe tomography (APT) provides compositional mapping of materials in three dimensions with sub-nanometre resolution, and is poised to play a key role in battery research. However, APT is underpinned by an intense electric field that can drive lithium migration, and many battery materials are reactive oxides, requiring careful handling and sample transfer. Here, we report on the analysis of both anode and cathode materials and show that electric-field driven migration can be suppressed by using shielding by embedding powder particles in a metallic matrix or by using a thin conducting surface layer. We demonstrate that for a typical cathode material, cryogenic specimen preparation and transport under ultra-high vacuum leads to major delithiation of the specimen during the analysis. In contrast, the transport of specimens through air enables the analysis of the material. Finally, we discuss the possible physical underpinnings and discuss ways forward to enable shielding from the electric field, which helps address the challenges inherent to the APT analysis of battery materials.

. EDX spectrum of as-received Li 4 Ti 5 O 12 nanoparticles. Inset image shows the corresponding nanoparticles. Cu and C peaks are originated from a commercial TEM grid. No Li peak detected due to high background signal at low keV range.      . Voltage (black) and noise level (red) curves vs. ions sequences on each representative APT measurement of the Li battery materials. Note that the LTO bulk was measured with green-laser assisted 3000 HR. In the LTO particle plot, only Ni matrix ions were detected for first 7M ions until the LTO embedded particle appeared.
For the LTO, since the majority of the data acquired is form the metallic matrix, we also ranged an arbitrary peak at 5 Da (no peak appeared) for both bulk and particle (extracted) LTO datasets at same sampled-ion counts. The LTO bulk dataset had 0.1427 at.% background concentration and for LTO particle, it showed 0.038 at.%. For another an arbitrary peak at 150 Da, the LTO bulk shows 0.6086 at.% and the LTO particle: 0.007 at.%. Overall, this indicates a much higher level of background in the LTO data from the bulk acquired on the LEAP 3000X Si, compared to LEAP 5000 XS.     S12. The three NMC811 specimens were prepared using Ga-FIB with different ion beam cleaning conditions to investigate the (remaining) Ga effect on the APT measurement. The low Ga ion beam at 5 kV and at 8 pA was used for (a) 0, (b) 5, and (c) 45 sec. All these specimens were transferred in air. Three distinguish regions were detected in the sample (a) and (b). #1 region contains high amount of Ga whereas #2 region showed some Li clusters region. But after 2M ions (region #3), the elements in the NMC811 specimens evaporated homogeneously without any indication of Li hotspots. In the case of the cleaned specimen (c), less Ga ions was detected. The measurement started with region #2 but it got improved to region #3 after <0.5 M ions. All measurement was stopped manually after 5M ions collection.  Figure S14. Voltage history curves of the NMC811 specimens fabricated using Ga-and Xe-FIBs: Ga-FIB (a) with and (b) without cleaning process. Xe-FIB (c) with air-transfer and (d) with UHV transfer. Note that there were many micro-fractures after UHV transfer and also note that there were three distinguished regions in the curve (b): Ga-concentrated region, Li-clustered region, and a pristine NMC811 region (see Figure S12a). All APT measurement was done at a base temperature of 60 K, a detection rate of 0.5 %, a pulse frequency of 125 kHz, and a laser energy of 5 pJ.

Oxygen flux calculation
The number of oxygen molecules (O 2 ) impinging on a sharpened specimen at different pressure can be calculated using the classical gas kinetic theory. The flux of O 2 (molecules m -2 sec -1 ) is given by: , where P is the gas pressure in Pa; here we assumed all remaining gases are O to yield an upper bound.
Another assumption was made that all O gases will be adsorbed on the specimen surface (sticking factor = 1). M, T, and k B are the molecular weight of the gas in a.m.u., gas temperature in K, and the Boltzmann constant, respectively.
For the case of the (non-cryo) UHV transfer, the overall average pressure of the FIB chamber, intermediate chamber, and the suitcase, was ~10 -5 Pa. This value yields the flux of 14605 O 2 molecules m -2 sec -1 . A typical APT specimen has a size of ϕ100 x 100 nm 3 . Assuming all O 2 molecules strike only onto a APT specimen, the O 2 adsorption rate will be 4.59 x 10 -10 molecules sec -1 and it requires 69 days to achieve a monolayer coverage of O 2 on the specimen. Unlikely, when the ambient pressure value (e.g. the case of the air transferring) was inputted in the equation, the approximated time for monolayer development of O 2 on the battery specimen is 0.29 sec. This big difference in O 2 monolayer development time suggests that effect of reactive O 2 should be considered. (e) Mass spectrum from acquired dataset. Note that no Li was detected indicating that Li did not electronically diffuse towards to the APT specimen apex.