A deep learning approach for semantic segmentation of unbalanced data in electron tomography of catalytic materials.

In computed TEM tomography, image segmentation represents one of the most basic tasks with implications not only for 3D volume visualization, but more importantly for quantitative 3D analysis. In case of large and complex 3D data sets, segmentation can be an extremely difficult and laborious task, and thus has been one of the biggest hurdles for comprehensive 3D analysis. Heterogeneous catalysts have complex surface and bulk structures, and often sparse distribution of catalytic particles with relatively poor intrinsic contrast, which possess a unique challenge for image segmentation, including the current state-of-the-art deep learning methods. To tackle this problem, we apply a deep learning-based approach for the multi-class semantic segmentation of a γ-Alumina/Pt catalytic material in a class imbalance situation. Specifically, we used the weighted focal loss as a loss function and attached it to the U-Net's fully convolutional network architecture. We assessed the accuracy of our results using Dice similarity coefficient (DSC), recall, precision, and Hausdorff distance (HD) metrics on the overlap between the ground-truth and predicted segmentations. Our adopted U-Net model with the weighted focal loss function achieved an average DSC score of 0.96 ± 0.003 in the γ-Alumina support material and 0.84 ± 0.03 in the Pt NPs segmentation tasks. We report an average boundary-overlap error of less than 2 nm at the 90th percentile of HD for γ-Alumina and Pt NPs segmentations. The complex surface morphology of γ-Alumina and its relation to the Pt NPs were visualized in 3D by the deep learning-assisted automatic segmentation of a large data set of high-angle annular dark-field (HAADF) scanning transmission electron microscopy (STEM) tomography reconstructions.

Progressive growth of the solid-electrolyte interphase towards the Si anode interior causes capacity fading.

The solid-electrolyte interphase (SEI), a layer formed on the electrode surface, is essential for electrochemical reactions in batteries and critically governs the battery stability. Active materials, especially those with extremely high energy density, such as silicon (Si), often inevitably undergo a large volume swing upon ion insertion and extraction, raising a critical question as to how the SEI interactively responds to and evolves with the material and consequently controls the cycling stability of the battery. Here, by integrating sensitive elemental tomography, an advanced algorithm and cryogenic scanning transmission electron microscopy, we unveil, in three dimensions, a correlated structural and chemical evolution of Si and SEI. Corroborated with a chemomechanical model, we demonstrate progressive electrolyte permeation and SEI growth along the percolation channel of the nanovoids due to vacancy injection and condensation during the delithiation process. Consequently, the Si-SEI spatial configuration evolves from the classic 'core-shell' structure in the first few cycles to a 'plum-pudding' structure following extended cycling, featuring the engulfing of Si domains by the SEI, which leads to the disruption of electron conduction pathways and formation of dead Si, contributing to capacity loss. The spatially coupled interactive evolution model of SEI and active materials, in principle, applies to a broad class of high-capacity electrode materials, leading to a critical insight for remedying the fading of high-capacity electrodes.

Engineering of Ruthenium-Iron Oxide Colloidal Heterostructures: Improved Yields in CO2 Hydrogenation to Hydrocarbons.

Catalytic CO2 reduction to fuels and chemicals is a major pursuit in reducing greenhouse gas emissions. One approach utilizes the reverse water-gas shift reaction, followed by Fischer-Tropsch synthesis, and iron is a well-known candidate for this process. Some attempts have been made to modify and improve its reactivity, but resulted in limited success. Now, using ruthenium-iron oxide colloidal heterodimers, close contact between the two phases promotes the reduction of iron oxide via a proximal hydrogen spillover effect, leading to the formation of ruthenium-iron core-shell structures active for the reaction at significantly lower temperatures than in bare iron catalysts. Furthermore, by engineering the iron oxide shell thickness, a fourfold increase in hydrocarbon yield is achieved compared to the heterodimers. This work shows how rational design of colloidal heterostructures can result in materials with significantly improved catalytic performance in CO2 conversion processes.

Experimental and Theoretical Structural Investigation of AuPt Nanoparticles Synthesized Using a Direct Electrochemical Method.

In this report, we examine the structure of bimetallic nanomaterials prepared by an electrochemical approach known as hydride-terminated (HT) electrodeposition. It has been shown previously that this method can lead to deposition of a single Pt monolayer on bulk-phase Au surfaces. Specifically, under appropriate electrochemical conditions and using a solution containing PtCl42-, a monolayer of Pt atoms electrodeposits onto bulk-phase Au immediately followed by a monolayer of H atoms. The H atom capping layer prevents deposition of Pt multilayers. We applied this method to ∼1.6 nm Au nanoparticles (AuNPs) immobilized on an inert electrode surface. In contrast to the well-defined, segregated Au/Pt structure of the bulk-phase surface, we observe that HT electrodeposition leads to the formation of AuPt quasi-random alloy NPs rather than the core@shell structure anticipated from earlier reports relating to deposition onto bulk phases. The results provide a good example of how the phase behavior of macro materials does not always translate to the nano world. A key component of this study was the structure determination of the AuPt NPs, which required a combination of electrochemical methods, electron microscopy, X-ray absorption spectroscopy, and theory (DFT and MD).

A Small Spot, Inert Gas, Ion Milling Process as a Complementary Technique to Focused Ion Beam Specimen Preparation.

This paper reports on the substantial improvement of specimen quality by use of a low voltage (0.05 to ~1 keV), small diameter (~1 µm), argon ion beam following initial preparation using conventional broad-beam ion milling or focused ion beam. The specimens show significant reductions in the amorphous layer thickness and implanted artifacts. The targeted ion milling controls the specimen thickness according to the needs of advanced aberration-corrected and/or analytical transmission electron microscopy applications.

In situ transmission electron microscopy probing of native oxide and artificial layers on silicon nanoparticles for lithium ion batteries.

Surface modification of silicon nanoparticles via molecular layer deposition (MLD) has been recently proved to be an effective way for dramatically enhancing the cyclic performance in lithium ion batteries. However, the fundamental mechanism of how this thin layer of coating functions is not known, which is complicated by the inevitable presence of native oxide of several nanometers on the silicon nanoparticle. Using in situ TEM, we probed in detail the structural and chemical evolution of both uncoated and coated silicon particles upon cyclic lithiation/delithation. We discovered that upon initial lithiation, the native oxide layer converts to crystalline Li2O islands, which essentially increases the impedance on the particle, resulting in ineffective lithiation/delithiation and therefore low Coulombic efficiency. In contrast, the alucone MLD-coated particles show extremely fast, thorough, and highly reversible lithiation behaviors, which are clarified to be associated with the mechanical flexibility and fast Li(+)/e(-) conductivity of the alucone coating. Surprisingly, the alucone MLD coating process chemically changes the silicon surface, essentially removing the native oxide layer, and therefore mitigates side reactions and detrimental effects of the native oxide. This study provides a vivid picture of how the MLD coating works to enhance the Coulombic efficiency, preserves capacity, and clarifies the role of the native oxide on silicon nanoparticles during cyclic lithiation and delithiation. More broadly, this work also demonstrates that the effect of the subtle chemical modification of the surface during the coating process may be of equal importance to the coating layer itself.

Revealing the atomic restructuring of Pt-Co nanoparticles.

We studied Pt-Co bimetallic nanoparticles during oxidation in O2 and reduction in H2 atmospheres using an aberration corrected environmental transmission electron microscope. During oxidation Co migrates to the nanoparticle surface forming a strained epitaxial CoO film. It subsequently forms islands via strain relaxation. The atomic restructuring is captured as a function of time. During reduction cobalt migrates back to the bulk, leaving a monolayer of platinum on the surface.

Mitigating voltage fade in cathode materials by improving the atomic level uniformity of elemental distribution.

Lithium- and manganese-rich (LMR) layered-structure materials are very promising cathodes for high energy density lithium-ion batteries. However, their voltage fading mechanism and its relationships with fundamental structural changes are far from being well understood. Here we report for the first time the mitigation of voltage and energy fade of LMR cathodes by improving the atomic level spatial uniformity of the chemical species. The results reveal that LMR cathodes (Li[Li0.2Ni0.2M0.6]O2) prepared by coprecipitation and sol-gel methods, which are dominated by a LiMO2 type R3̅m structure, show significant nonuniform Ni distribution at particle surfaces. In contrast, the LMR cathode prepared by a hydrothermal assisted method is dominated by a Li2MO3 type C2/m structure with minimal Ni-rich surfaces. The samples with uniform atomic level spatial distribution demonstrate much better capacity retention and much smaller voltage fade as compared to those with significant nonuniform Ni distribution. The fundamental findings on the direct correlation between the atomic level spatial distribution of the chemical species and the functional stability of the materials may also guide the design of other energy storage materials with enhanced stabilities.

In-situ TEM visualization of vacancy injection and chemical partition during oxidation of Ni-Cr nanoparticles.

Oxidation of alloy often involves chemical partition and injection of vacancies. Chemical partition is the consequence of selective oxidation, while injection of vacancies is associated with the differences of diffusivity of cations and anions. It is far from clear as how the injected vacancies behave during oxidation of metal. Using in-situ transmission electron microscopy, we captured unprecedented details on the collective behavior of injected vacancies during oxidation of metal, featuring an initial multi-site oxide nucleation, vacancy supersaturation, nucleation of a single cavity, sinking of vacancies into the cavity and accelerated oxidation of the particle. High sensitive energy dispersive x-ray spectroscopy mapping reveals that Cr is preferentially oxidized even at the initial oxidation, leading to a structure that Cr oxide is sandwiched near the inner wall of the hollow particle. The work provides a general guidance on tailoring of nanostructured materials involving multi-ion exchange such as core-shell structured composite nanoparticles.

XEDS STEM tomography for 3D chemical characterization of nanoscale particles.

We present a tomography technique which couples scanning transmission electron microscopy (STEM) and X-ray energy dispersive spectrometry (XEDS) to resolve 3D distribution of elements in nanoscale materials. STEM imaging when combined with XEDS mapping using a symmetrically arranged XEDS detector design around the specimen overcomes many of the obstacles in 3D chemical imaging of nanoscale materials and successfully elucidates the 3D chemical information in a large field of view of the transmission electron microscopy (TEM) sample. We employed this technique to investigate 3D distribution of Nickel (Ni), Manganese (Mn) and Oxygen (O) in a Li1.2Ni0.2Mn0.6O2 (LNMO) nanoparticle used as a cathode material in Lithium (Li) ion batteries. For this purpose, 2D elemental maps were acquired for a range of tilt angles and reconstructed to obtain 3D elemental distribution in an isolated LNMO nanoparticle. The results highlight the strength of this technique in 3D chemical analysis of nanoscale materials by successfully resolving Ni, Mn and O elemental distributions in 3D and discovering the new phenomenon of Ni surface segregation in this material. Furthermore, the comparison of simultaneously acquired high angle annular dark field (HAADF) STEM and XEDS STEM tomography results shows that XEDS STEM tomography provides additional 3D chemical information of the material especially when there is low atomic number (Z) contrast in the material of interest.

Formation of the spinel phase in the layered composite cathode used in Li-ion batteries.

Pristine Li-rich layered cathodes, such as Li(1.2)Ni(0.2)Mn(0.6)O(2) and Li(1.2)Ni(0.1)Mn(0.525)Co(0.175)O(2), were identified to exist in two different structures: LiMO(2)R3[overline]m and Li(2)MO(3)C2/m phases. Upon 300 cycles of charge/discharge, both phases gradually transform to the spinel structure. The transition from LiMO(2)R3[overline]m to spinel is accomplished through the migration of transition metal ions to the Li site without breaking down the lattice, leading to the formation of mosaic structured spinel grains within the parent particle. In contrast, transition from Li(2)MO(3)C2/m to spinel involves removal of Li(+) and O(2-), which produces large lattice strain and leads to the breakdown of the parent lattice. The newly formed spinel grains show random orientation within the same particle. Cracks and pores were also noticed within some layered nanoparticles after cycling, which is believed to be the consequence of the lattice breakdown and vacancy condensation upon removal of lithium ions. The AlF(3)-coating can partially relieve the spinel formation in the layered structure during cycling, resulting in a slower capacity decay. However, the AlF(3)-coating on the layered structure cannot ultimately stop the spinel formation. The observation of structure transition characteristics discussed in this paper provides direct explanation for the observed gradual capacity loss and poor rate performance of the layered composite. It also provides clues about how to improve the materials structure in order to improve electrochemical performance.

Conflicting roles of nickel in controlling cathode performance in lithium ion batteries.

A variety of approaches are being made to enhance the performance of lithium ion batteries. Incorporating multivalence transition-metal ions into metal oxide cathodes has been identified as an essential approach to achieve the necessary high voltage and high capacity. However, the fundamental mechanism that limits their power rate and cycling stability remains unclear. The power rate strongly depends on the lithium ion drift speed in the cathode. Crystallographically, these transition-metal-based cathodes frequently have a layered structure. In the classic wisdom, it is accepted that lithium ion travels swiftly within the layers moving out/in of the cathode during the charge/discharge. Here, we report the unexpected discovery of a thermodynamically driven, yet kinetically controlled, surface modification in the widely explored lithium nickel manganese oxide cathode material, which may inhibit the battery charge/discharge rate. We found that during cathode synthesis and processing before electrochemical cycling in the cell nickel can preferentially move along the fast diffusion channels and selectively segregate at the surface facets terminated with a mix of anions and cations. This segregation essentially can lead to a higher lithium diffusion barrier near the surface region of the particle. Therefore, it appears that the transition-metal dopant may help to provide high capacity and/or high voltage but can be located in a "wrong" location that may slow down lithium diffusion, limiting battery performance. In this circumstance, limitations in the properties of lithium ion batteries using these cathode materials can be determined more by the materials synthesis issues than by the operation within the battery itself.

Complementary techniques for the characterization of thin film Ti/Nb multilayers.

An aberration corrector on the probe-forming lens of a scanning TEM (STEM) equipped with an electron energy-loss spectrometer (EELS) and X-ray energy-dispersive spectrometer (XEDS) has been employed to investigate the compositional variations as a function of length scale in nanoscale Ti/Nb metallic multilayers. The composition profiles of EELS and XEDS were compared with the profiles obtained from the complementary technique of 3D atom probe tomography. At large layer widths (h > or = 7 nm, where h is the layer width) of Ti and Nb, XEDS composition profiles of Ti/Nb metallic multilayers are in good agreement with the EELS results. However, at reduced layer widths (h approximately 2 nm), profiles of EELS and atom probe exhibited similar compositional variations, whereas XEDS results have shown a marked difference. This difference in the composition profiling of the layers has been addressed with reference to the effects of beam broadening and the origin of the signals collected in these techniques. The advantage of using EELS over XEDS for these nanoscaled multilayered materials is demonstrated.