Machine Learning Data Augmentation Strategy for Electron Energy Loss Spectroscopy: Generative Adversarial Networks.

Recent advances in machine learning (ML) have highlighted a novel challenge concerning the quality and quantity of data required to effectively train algorithms in supervised ML procedures. This article introduces a data augmentation (DA) strategy for electron energy loss spectroscopy (EELS) data, employing generative adversarial networks (GANs). We present an innovative approach, called the data augmentation generative adversarial network (DAG), which facilitates data generation from a very limited number of spectra, around 100. Throughout this study, we explore the optimal configuration for GANs to produce realistic spectra. Notably, our DAG generates realistic spectra, and the spectra produced by the generator are successfully used in real-world applications to train classifiers based on artificial neural networks (ANNs) and support vector machines (SVMs) that have been successful in classifying experimental EEL spectra.

Automated Image Analysis for Single-Atom Detection in Catalytic Materials by Transmission Electron Microscopy.

Single-atom catalytic sites may have existed in all supported transition metal catalysts since their first application. Yet, interest in the design of single-atom heterogeneous catalysts (SACs) only really grew when advances in transmission electron microscopy (TEM) permitted direct confirmation of metal site isolation. While atomic-resolution imaging remains a central characterization tool, poor statistical significance, reproducibility, and interoperability limit its scope for deriving robust characteristics about these frontier catalytic materials. Here, we introduce a customized deep-learning method for automated atom detection in image analysis, a rate-limiting step toward high-throughput TEM. Platinum atoms stabilized on a functionalized carbon support with a challenging irregular three-dimensional morphology serve as a practically relevant test system with promising scope in thermo- and electrochemical applications. The model detects over 20,000 atomic positions for the statistical analysis of important properties for establishing structure-performance relations over nanostructured catalysts, like the surface density, proximity, clustering extent, and dispersion uniformity of supported metal species. Good performance obtained on direct application of the model to an iron SAC based on carbon nitride demonstrates its generalizability for single-atom detection on carbon-related materials. The approach establishes a route to integrate artificial intelligence into routine TEM workflows. It accelerates image processing times by orders of magnitude and reduces human bias by providing an uncertainty analysis that is not readily quantifiable in manual atom identification, improving standardization and scalability.

Elucidation of Metal Local Environments in Single-Atom Catalysts Based on Carbon Nitrides.

The ability to tailor the properties of metal centers in single-atom heterogeneous catalysts depends on the availability of advanced approaches for characterization of their structure. Except for specific host materials with well-defined metal adsorption sites, determining the local atomic environment remains a crucial challenge, often relying heavily on simulations. This article reports an advanced analysis of platinum atoms stabilized on poly(triazine imide), a nanocrystalline form of carbon nitride. The approach discriminates the distribution of surface coordination sites in the host, the evolution of metal coordination at different stages during the synthesis of the material, and the potential locations of metal atoms within the lattice. Consistent with density functional theory predictions, simultaneous high-resolution imaging in high-angle annular dark field and bright field modes experimentally confirms the preferred localization of platinum in-plane in the corners of the triangular cavities. X-ray absorption spectroscopy (XAS), X-ray photoelectron spectroscopy (XPS), and dynamic nuclear polarization enhanced 15 N nuclear magnetic resonance (DNP-NMR) spectroscopies coupled with density functional theory (DFT) simulations reveal that the predominant metal species comprise Pt(II) bound to three nitrogen atoms and one chlorine atom inside the coordination sites. The findings, which narrow the gap between experimental and theoretical elucidation, contribute to the improved structural understanding and provide a benchmark for exploring the speciation of single-atom catalysts based on carbon nitrides.

Functional Group Mapping by Electron Beam Vibrational Spectroscopy from Nanoscale Volumes.

Vibrational spectroscopies directly record details of bonding in materials, but spatially resolved methods have been limited to surface techniques for mapping functional groups at the nanoscale. Electron energy loss spectroscopy (EELS) in the scanning transmission electron microscope presents a route to functional group analysis from nanoscale volumes using transmitted subnanometer electron probes. Here, we now use vibrational EELS to map distinct carboxylate and imidazolate linkers in a metal-organic framework (MOF) crystal-glass composite material. Domains <100 nm in size are observed using vibrational EELS, with recorded spatial resolution <15 nm at interfaces in the composite. This nanoscale functional group mapping is confirmed by correlated EELS at core ionization edges as well as X-ray energy dispersive spectroscopy for elemental mapping of the metal centers of the two constituent MOFs. These results present a complete nanoscale analysis of the building blocks of the MOF composite and establish spatially resolved functional group analysis using electron beam spectroscopy for crystalline and amorphous organic and metal-organic solids.

Scalable Patterning of Encapsulated Black Phosphorus.

Atomically thin black phosphorus (BP) has attracted considerable interest due to its unique properties, such as an infrared band gap that depends on the number of layers and excellent electronic transport characteristics. This material is known to be sensitive to light and oxygen and degrades in air unless protected with an encapsulation barrier, limiting its exploitation in electrical devices. We present a new scalable technique for nanopatterning few layered BP by direct electron beam exposure of encapsulated crystals, achieving a spatial resolution down to 6 nm. By encapsulating the BP with single layer graphene or hexagonal boron nitride (hBN), we show that a focused electron probe can be used to produce controllable local oxidation of BP through nanometre size defects created in the encapsulation layer by the electron impact. We have tested the approach in the scanning transmission electron microscope (STEM) and using industry standard electron beam lithography (EBL). Etched regions of the BP are stabilized by a thin passivation layer and demonstrate typical insulating behavior as measured at 300 and 4.3 K. This new scalable approach to nanopatterning of thin air sensitive crystals has the potential to facilitate their wider use for a variety of sensing and electronics applications.

Subwavelength Spatially Resolved Coordination Chemistry of Metal-Organic Framework Glass Blends.

Microstructured metal-organic framework (MOF) glasses have been produced by combining two amorphous MOFs. However, the electronic structure of these materials has not been interrogated at the length scales of the chemical domains formed in these glasses. Here, we report a subwavelength spatially resolved physicochemical analysis of the electronic states at visible and UV energies in a blend of two zeolitic imidazolate frameworks (ZIFs), ZIF-4-Co and ZIF-62-Zn. By combining spectroscopy at visible and UV energies as well as at core ionization energies in electron energy loss spectroscopy in the scanning transmission electron microscope with density functional theory calculations, we show that domains less than 200 nm in size retain the electronic structure of the precursor crystalline ZIF phases. Prototypical signatures of coordination chemistry including d- d transitions in ZIF-4-Co are assigned and mapped with nanoscale precision.

Prospects for Engineering Thermoelectric Properties in La1/3NbO3 Ceramics Revealed via Atomic-Level Characterization and Modeling.

A combination of experimental and computational techniques has been employed to explore the crystal structure and thermoelectric properties of A-site-deficient perovskite La1/3NbO3 ceramics. Crystallographic data from X-ray and electron diffraction confirmed that the room temperature structure is orthorhombic with Cmmm as a space group. Atomically resolved imaging and analysis showed that there are two distinct A sites: one is occupied with La and vacancies, and the second site is fully unoccupied. The diffuse superstructure reflections observed through diffraction techniques are shown to originate from La vacancy ordering. La1/3NbO3 ceramics sintered in air showed promising high-temperature thermoelectric properties with a high Seebeck coefficient of S1 = -650 to -700 µV/K and a low and temperature-stable thermal conductivity of k = 2-2.2 W/m·K in the temperature range of 300-1000 K. First-principles electronic structure calculations are used to link the temperature dependence of the Seebeck coefficient measured experimentally to the evolution of the density of states with temperature and indicate possible avenues for further optimization through electron doping and control of the A-site occupancies. Moreover, lattice thermal conductivity calculations give insights into the dependence of the thermal conductivity on specific crystallographic directions of the material, which could be exploited via nanostructuring to create high-efficiency compound thermoelectrics.

Fluid-induced organic synthesis in the solar nebula recorded in extraterrestrial dust from meteorites.

Isotopically anomalous carbonaceous grains in extraterrestrial samples represent the most pristine organics that were delivered to the early Earth. Here we report on gentle aberration-corrected scanning transmission electron microscopy investigations of eight (15)N-rich or D-rich organic grains within two carbonaceous Renazzo-type (CR) chondrites and two interplanetary dust particles (IDPs) originating from comets. Organic matter in the IDP samples is less aromatic than that in the CR chondrites, and its functional group chemistry is mainly characterized by C-O bonding and aliphatic C. Organic grains in CR chondrites are associated with carbonates and elemental Ca, which originate either from aqueous fluids or possibly an indigenous organic source. One distinct grain from the CR chondrite NWA 852 exhibits a rim structure only visible in chemical maps. The outer part is nanoglobular in shape, highly aromatic, and enriched in anomalous nitrogen. Functional group chemistry of the inner part is similar to spectra from IDP organic grains and less aromatic with nitrogen below the detection limit. The boundary between these two areas is very sharp. The direct association of both IDP-like organic matter with dominant C-O bonding environments and nanoglobular organics with dominant aromatic and C-N functionality within one unique grain provides for the first time to our knowledge strong evidence for organic synthesis in the early solar system activated by an anomalous nitrogen-containing parent body fluid.

Local A-Site Layering in Rare-Earth Orthochromite Perovskites by Solution Synthesis.

Cation size effects were examined in the mixed A-site perovskites La0.5 Sm0.5 CrO3 and La0.5 Tb0.5 CrO3 prepared through both hydrothermal and solid-state methods. Atomically resolved electron energy loss spectroscopy (EELS) in the transmission electron microscope shows that while the La and Sm cations are randomly distributed, increased cation-radius variance in La0.5 Tb0.5 CrO3 results in regions of localised La and Tb layers, an atomic arrangement exclusive to the hydrothermally prepared material. Solid-state preparation gives lower homogeneity resulting in separate nanoscale regions rich in La3+ and Tb3+ . The A-site layering in hydrothermal La0.5 Tb0.5 CrO3 is randomised upon annealing at high temperature, resulting in magnetic behaviour that is dependent on synthesis route.

Characterization of Ordering in A-Site Deficient Perovskite Ca1-xLa2x/3TiO3 Using STEM/EELS.

The vacancy ordering behavior of an A-site deficient perovskite system, Ca1-xLa2x/3TiO3, was studied using atomic resolution scanning transmission electron microscopy (STEM) in conjunction with electron energy-loss spectroscopy (EELS), with the aim of determining the role of A-site composition changes. At low La content (x = 0.2), adopting Pbnm symmetry, there was no indication of long-range ordering. Domains, with clear boundaries, were observed in bright-field (BF) imaging, but were not immediately visible in the corresponding high-angle annular dark-field (HAADF) image. These boundaries, with the aid of displacement maps from A-site cations in the HAADF signal, are shown to be tilt boundaries. At the La-rich end of the composition (x = 0.9), adopting Cmmm symmetry, long-range ordering of vacancies and La3+ ions was observed, with alternating La-rich and La-poor layers on (001)p planes, creating a double perovskite lattice along the c axis. These highly ordered domains can be found isolated within a random distribution of vacancies/La3+, or within a large population, encompassing a large volume. In regions with a high number density of double perovskite domains, these highly ordered domains were separated by twin boundaries, with 90° or 180° lattice rotations across boundaries. The occurrence and characteristics of these ordered structures are discussed and compared with similar perovskite systems.

Tungsten Bronze Barium Neodymium Titanate (Ba(6-3n)Nd(8+2n)Ti(18)O(54)): An Intrinsic Nanostructured Material and Its Defect Distribution.

We investigated the structure of the tungsten bronze barium neodymium titanates Ba(6-3n)Nd(8+2n)Ti(18)O(54), which are exploited as microwave dielectric ceramics. They form a complex nanostructure, which resembles a nanofilm with stacking layers of ∼12 Šthickness. The synthesized samples of Ba(6-3n)Nd(8+2n)Ti(18)O(54) (n = 0, 0.3, 0.4, 0.5) are characterized by pentagonal and tetragonal columns, where the A cations are distributed in three symmetrically inequivalent sites. Synchrotron X-ray diffraction and electron energy loss spectroscopy allowed for quantitative analysis of the site occupancy, which determines the defect distribution. This is corroborated by density functional theory calculations. Pentagonal columns are dominated by Ba, and tetragonal columns are dominated by Nd, although specific Nd sites exhibit significant concentrations of Ba. The data indicated significant elongation of the Ba columns in the pentagonal positions and of the Nd columns in tetragonal positions involving a zigzag arrangement of atoms along the b lattice direction. We found that the preferred Ba substitution occurs at Nd[3]/[4] followed by Nd[2] and Nd[1]/[5] sites, which is significantly different to that proposed in earlier studies. Our results on the Ba(6-3n)Nd(8+2n)Ti(18)O(54) "perovskite" superstructure and its defect distribution are particularly valuable in those applications where the optimization of material properties of oxides is imperative; these include not only microwave ceramics but also thermoelectric materials, where the nanostructure and the distribution of the dopants will reduce the thermal conductivity.

Tuning the thermoelectric properties of A-site deficient SrTiO3 ceramics by vacancies and carrier concentration.

Ceramics based on Sr0.8La0.067Ti0.8Nb0.2O3-δ have been prepared by the mixed oxide route. The La1/3NbO3 component generates ∼13.4% A-site vacancies; this was fixed for all samples. Powders were sintered under air and reducing conditions at 1450 to 1700 K; products were of high density (>90% theoretical). Processing under reducing conditions led to the formation of a Ti1-xNbxO2-y second phase, core-shell structures and oxygen deficiency. X-ray diffraction (XRD) confirmed a simple cubic structure with space group Pm3[combining macron]m. Transmission electron microscopy revealed a high density of dislocations while analytical scanning transmission electron microscopy at atomic resolution demonstrated a uniform distribution of La, Nb and vacancies in the lattice. X-ray photoemission spectroscopy and thermogravimetry showed the oxygen deficiency (δ value) to be ∼0.08 in reduced samples with enhanced carrier concentrations ∼2 × 1021 cm-3. Both carrier concentration and carrier mobility increased with sintering time, giving a maximum figure of merit (ZT) of 0.25. Selective additional doping by La or Nb, with no additional A site vacancies, led to the creation of additional carriers and reduced electrical resistivity. Together these led to enhanced ZT values of 0.345 at 1000 K. The contributions from oxygen vacancies and charge carriers have been investigated independently.

Silicon-carbon bond inversions driven by 60-keV electrons in graphene.

We demonstrate that 60-keV electron irradiation drives the diffusion of threefold-coordinated Si dopants in graphene by one lattice site at a time. First principles simulations reveal that each step is caused by an electron impact on a C atom next to the dopant. Although the atomic motion happens below our experimental time resolution, stochastic analysis of 38 such lattice jumps reveals a probability for their occurrence in a good agreement with the simulations. Conversions from three- to fourfold coordinated dopant structures and the subsequent reverse process are significantly less likely than the direct bond inversion. Our results thus provide a model of nondestructive and atomically precise structural modification and detection for two-dimensional materials.

Amorphous 1-D nanowires of calcium phosphate/pyrophosphate: A demonstration of oriented self-growth of amorphous minerals.

Amorphous inorganic solids are traditionally isotropic, thus, it is believed that they only grow in a non-preferential way without the assistance of regulators, leading to the morphologies of nanospheres or irregular aggregates of nanoparticles. However, in the presence of (ortho)phosphate (Pi) and pyrophosphate ions (PPi) which have synergistic roles in biomineralization, the highly elongated amorphous nanowires (denoted ACPPNs) form in a regulator-free aqueous solution (without templates, additives, organics, etc). Based on thorough characterization and tracking of the formation process (e.g., Cryo-TEM, spherical aberration correction high resolution TEM, solid state NMR, high energy resolution monochromated STEM-EELS), the microstructure and its preferential growth behavior are elucidated. In ACPPNs, amorphous calcium orthophosphate and amorphous calcium pyrophosphate are distributed at separated but close sites. The ACPPNs grow via either the preferential attachment of ∼2 nm nanoclusters in a 1-dimension way, or the transformation of bigger nanoparticles, indicating an inherent driving force-governed process. We propose that the anisotropy of ACPPNs microstructure, which is corroborated experimentally, causes their oriented growth. This study proves that, unlike the conventional view, amorphous minerals can form via oriented growth without external regulation, demonstrating a novel insight into the structures and growth behaviors of amorphous minerals.

High-spatial resolution functional chemistry of nitrogen compounds in the observed UK meteorite fall Winchcombe.

Organic matter in extraterrestrial samples is a complex material that might have played an important role in the delivery of prebiotic molecules to the early Earth. We report here on the identification of nitrogen-containing compounds such as amino acids and N-heterocycles within the recent observed meteorite fall Winchcombe by high-spatial resolution spectroscopy techniques. Although nitrogen contents of Winchcombe organic matter are low (N/C ~ 1-3%), we were able to detect the presence of these compounds using a low-noise direct electron detector. These biologically relevant molecules have therefore been tentatively found within a fresh, minimally processed meteorite sample by high spatial resolution techniques conserving the overall petrographic context. Carbon functional chemistry investigations show that sizes of aromatic domains are small and that abundances of carboxylic functional groups are low. Our observations demonstrate that Winchcombe represents an important addition to the collection of carbonaceous chondrites and still preserves pristine extraterrestrial organic matter.

Comparative of machine learning classification strategies for electron energy loss spectroscopy: Support vector machines and artificial neural networks.

Machine Learning (ML) strategies applied to Scanning and conventional Transmission Electron Microscopy have become a valuable tool for analyzing the large volumes of data generated by various S/TEM techniques. In this work, we focus on Electron Energy Loss Spectroscopy (EELS) and study two ML techniques for classifying spectra in detail: Support Vector Machines (SVM) and Artificial Neural Networks (ANN). Firstly, we systematically analyze the optimal configurations and architectures for ANN classifiers using random search and the tree-structured Parzen estimator methods. Secondly, a new kernel strategy is introduced for the soft-margin SVMs, the cosine kernel, which offers a significant advantage over the previously studied kernels and other ML classification strategies. This kernel allows us to bypass the normalization of EEL spectra, achieving accurate classification. This result is highly relevant for the EELS community since we also assess the impact of common normalization techniques on our spectra using Uniform Manifold Approximation and Projection (UMAP), revealing a strong bias introduced in the spectra once normalized. In order to evaluate and study both classification strategies, we focus on determining the oxidation state of transition metals through their EEL spectra, examining which feature is more suitable for oxidation state classification: the oxygen K peak or the transition metal white lines. Subsequently, we compare the resistance to energy loss shifts for both classifiers and present a strategy to improve their resistance. The results of this study suggest the use of soft-margin SVMs for simpler EELS classification tasks with a limited number of spectra, as they provide performance comparable to ANNs while requiring lower computational resources and reduced training times. Conversely, ANNs are better suited for handling complex classification problems with extensive training data.

Dislocation and strain mapping in metamorphic parabolic-graded InGaAs buffers on GaAs.

We investigate different architectures for parabolic-graded InGaAs metamorphic buffers grown on GaAs using transmission electron microscopy techniques. The different architectures include InGaP and AlInGaAs/InGaP superlattices with different GaAs substrate misorientations and the inclusion of a strain balancing layer. Our results correlate: (i) the density and distribution of dislocations in the metamorphic buffer and (ii) the strain in the next layer preceding the metamorphic buffer, which varies for each type of architecture. Our findings indicate that the dislocation density in the lower region of the metamorphic layer ranges between 108 and 1010 cm-2, with AlInGaAs/InGaP superlattice samples exhibiting higher values compared to samples with InGaP films. We have identified two waves of dislocations, with threading dislocations typically located lower in the metamorphic buffer (~ 200-300 nm) in comparison to misfit dislocations. The measured localised strain values are in good agreement with theoretical predications. Overall, our results provide a systematic insight into the strain relaxation across different architectures, highlighting the various approaches that can be used to tailor strain in the active region of a metamorphic laser. Supplementary Information: The online version contains supplementary material available at 10.1007/s10853-023-08597-y.

High Power Factor Nb-Doped TiO2 Thermoelectric Thick Films: Toward Atomic Scale Defect Engineering of Crystallographic Shear Structures.

Donor-doped TiO2-based materials are promising thermoelectrics (TEs) due to their low cost and high stability at elevated temperatures. Herein, high-performance Nb-doped TiO2 thick films are fabricated by facile and scalable screen-printing techniques. Enhanced TE performance has been achieved by forming high-density crystallographic shear (CS) structures. All films exhibit the same matrix rutile structure but contain different nano-sized defect structures. Typically, in films with low Nb content, high concentrations of oxygen-deficient {121} CS planes are formed, while in films with high Nb content, a high density of twin boundaries are found. Through the use of strongly reducing atmospheres, a novel Al-segregated {210} CS structure is formed in films with higher Nb content. By advanced aberration-corrected scanning transmission electron microscopy techniques, we reveal the nature of the {210} CS structure at the nano-scale. These CS structures contain abundant oxygen vacancies and are believed to enable energy-filtering effects, leading to simultaneous enhancement of both the electrical conductivity and Seebeck coefficients. The optimized films exhibit a maximum power factor of 4.3 × 10-4 W m-1 K-2 at 673 K, the highest value for TiO2-based TE films at elevated temperatures. Our modulation strategy based on microstructure modification provides a novel route for atomic-level defect engineering which should guide the development of other TE materials.

Controlling the Thermoelectric Behavior of La-Doped SrTiO3 through Processing and Addition of Graphene Oxide.

The addition of graphene has been reported as a potential route to enhance the thermoelectric performance of SrTiO3. However, the interplay between processing parameters and graphene addition complicates understanding this enhancement. Herein, we examine the effects of processing parameters and graphene addition on the thermoelectric performance of La-doped SrTiO3 (LSTO). Briefly, two types of graphene oxide (GO) at different oxidation degrees were used, while the LSTO pellets were densified under two conditions with different reducing strengths (with/without using oxygen-scavenging carbon powder bed muffling). Raman imaging of the LSTO green body and sintered pellets suggests that the added GO sacrificially reacts with the lattice oxygen, which creates more oxygen vacancies and improves electrical conductivity regardless of the processing conditions. The addition of mildly oxidized electrochemical GO (EGO) yields better performance than the conventional heavily oxidized chemical GO (CGO). Moreover, we found that muffling the green body with an oxygen-scavenging carbon powder bed during sintering is vital to achieving a single-crystal-like temperature dependence of electrical conductivity, implying that a highly reducing environment is critical for eliminating the grain boundary barriers. Combining 1.0 wt % EGO addition with a highly reducing environment leads to the highest electrical conductivity of 2395 S cm-1 and power factor of 2525µW m-1 K-2 at 300 K, with an improved average zT value across the operating temperature range of 300-867 K. STEM-EELS maps of the optimized sample show a pronounced depletion of Sr and evident deficiency of O and La at the grain boundary region. Theoretical modeling using a two-phase model implies that the addition of GO can effectively improve carrier mobility in the grain boundary phase. This work provides guidance for the development of high-performance thermoelectric ceramic oxides.