Complementary ADF Analysis in Microscopy

• Complementary ADF analysis is a suite of techniques combining ADF imaging with auxiliary modalities to reveal atomic-scale structures and address signal ambiguities.
• It employs correlative multi-modal and differential processing approaches, as well as virtual signal synthesis, to enhance light-atom and defect detection.
• Advanced instrumentation and algorithmic frameworks enable rapid, high-throughput quantification of material properties with nanoscale resolution.

Complementary ADF (Annular Dark Field) analysis refers to a suite of methodologies, experiments, and algorithms whereby ADF image or signal data—often in conjunction with other modalities—are leveraged to extract information that would be inaccessible, or far less precise, using any single ADF (or related) technique alone. Across disciplines, “complementary ADF analysis” commonly denotes (i) correlative multi-modal imaging in microscopy, (ii) enhanced atomic-scale quantification using joint ADF–spectroscopy datasets, and (iii) the exploitation of ADF’s mathematical or physical properties (e.g., Z or demerit factor contrasts) for rigorous discrimination among otherwise ambiguous signals or structures. The following sections detail foundational principles, device schemes, algorithmic frameworks, experimental metrics, limits, and applications as manifest in recent arXiv research.

1. Physical Principles and Mathematical Foundations

Annular Dark Field detection in scanning transmission electron microscopy (ADF-STEM) is predicated on capturing electrons elastically scattered to intermediate or high angles, yielding contrast that is characteristically monotonic in atomic number Z and (to leading order) in specimen thickness t. The signal can typically be modeled as

$I_{\rm ADF} \propto Z^n t^m$

with empirically determined exponents (e.g., $n \simeq 2/3$, $m \simeq 1$ for sub-30 kV FEG-SEM-based ADF devices (Lagos et al., 2015)). This simple dependency is distorted in crystalline materials, where dynamical channelling gives rise to non-linearities, which have been formalized via lensing or superposition models for multi-element columns (Zhang et al., 2022).

Complementary ADF analysis is distinct from conventional ADF in that it either:

• Integrates ADF with auxiliary signals (e.g., energy-dispersive X-ray spectroscopy, EDX (Zhang et al., 2022); electron energy-loss spectroscopy, EELS (Oh et al., 27 Sep 2025)); or
• Extracts subtle features by differential or correlative processing (e.g., light-atom detection by column-type subtraction (Mostaed et al., 2020)); or
• Synthesizes virtual ADF signals from high-dimensional datasets (e.g., complementary ADF in 4D-STEM (Esser et al., 2022)).

ADF contrast is rendered nearly incoherent when the inner detector angle excludes all Bragg diffraction (phonon scattering domination), which underpins linear ADF–EDX relationships and the analytic tractability of atomic lensing algorithms.

2. High-Throughput Instrumentation and Device Design

Instrumental advances have democratized ADF contrast for routine, high-throughput analysis. An example is the multi-sample TE-ADF “carrousel” holder, which fits standard FEG-SEM airlock systems and permits up to sixteen TEM grids to be screened in a single mounting (Lagos et al., 2015). High-angle transmitted electrons strike a concave, Au/MgO conversion layer, generating secondary electrons (SE) collected by the native Everhart–Thornley detector. This design provides Z- and t-sensitive imaging at nanometric resolution (FWHM ~0.9 nm edge) without major SEM hardware modifications or specialist operator intervention.

In 4D-STEM approaches, complementary ADF (cADF) analysis is realized by computing, from pixelated diffraction patterns, the complement of low-angle scattering relative to the total measured beam current, thereby circumventing dynamic range and angular-collection trade-offs inherent to physical annular detectors (Esser et al., 2022). The cADF signal is

$$I_{\rm cADF}(r) = 1 - \int_{0}^{\theta_0} I(\theta, r) d\theta$$

where θ₀ is the user-defined inner semi-angle.

3. Correlative and Computational Multi-modal Approaches

Complementary ADF analysis often refers to pipelines that integrate ADF with elemental (EDS/EDX) and electronic (EELS) signals, reconstructing co-registered three-dimensional maps of structure, composition, and valence (Oh et al., 27 Sep 2025). Tilt-series tomography with simultaneous ADF, EDS, and EELS acquisition and algorithmic co-registration (e.g., via GENFIRE or fused joint-regularization) enables direct voxel-by-voxel correspondence among morphology, Z-contrast, chemical identity, and electronic state at sub-10 nm resolution. The ADF provides mass–thickness and Z-contrast for overall structure, EDS reconstructs atomic percent of transition metals (e.g., Mn, Co, Ni, O in battery materials), and EELS encodes oxidation-state gradients and local electronic environments. Such integrated pipelines are fundamental for probing nanoscale inhomogeneities, surface–bulk segregation, and three-dimensional valence-state changes under operando conditions (Oh et al., 27 Sep 2025).

In the context of atom-resolved spectroscopy, fast analytical models (atomic lensing/superposition) exploit the linearity of incoherent ADF–EDX signals to scan millions of possible column occupancy/orderings rapidly, with deviations from multislice simulation predictions <10% for realistic thicknesses (Zhang et al., 2022). Explicitly, the cross-sections for mixed columns are built from precomputed pure-element libraries, drastically reducing computational overhead.

4. Complementary ADF Analysis for Light-Atom and Defect Detection

Traditional ADF-STEM fails to directly image low-Z species due to weak scattering. Complementary ADF analysis circumvents this by exploiting subtle probe de-channelling and scattering-coupling effects: when two types of adjacent heavy-atom columns differ only by their neighboring light-atom (e.g., O) occupancy, subtraction of their radially averaged ADF profiles cancels the dominant Z-contrast and isolates the light-atom-induced intensity (and shape) differences (Mostaed et al., 2020). For perovskite oxides (LaFeO₃), 2D Gaussian fitting to the cation columns, followed by Fe1–Fe2 subtraction and thresholding, provides direct mapping of O columns at tens of picometer precision. Radial difference curves in pyrochlores (A₂B₂O₇) pinpoint O-vacancy sites and are sensitive to site-specific occupancy, even in the presence of complex magnetic phenomena.

5. Quantitative Performance Metrics and Algorithmic Validation

Performance metrics in complementary ADF analysis span spatial resolution, signal-to-noise ratio (SNR), throughput, and calibration quality. FEG-SEM ADF holders achieve sub-nanometer edge resolution and resolve nanoparticles as small as 2 nm diameter (Lagos et al., 2015). In multi-modal tomographic workflows, in-plane resolution is fixed by pixel step ($\sim$5 nm), depth resolution by the tilt-range and algorithmic regularization ($\sim$8–10 nm with fused GENFIRE), and currency in cross-modality alignment (<1 voxel) by transfer of ADF-based tilt and shift parameters to EDS/EELS reconstructions (Oh et al., 27 Sep 2025). Analytical ADF–EDX predictions using the atomic lensing model match full multislice simulations to within 5–10% for core-shell and complex alloys, establishing the feasibility of fully ab initio, element-specific atom counting on the scale of millions of structural candidates in seconds (Zhang et al., 2022).

6. Application Space and Limitations

Complementary ADF analysis is foundational in:

• High-throughput nanoparticle screening: multi-sample ADF selectors rapidly discard suboptimal batches before costly full-TEM or analytical runs (Lagos et al., 2015).
• Catalysis research: high-Z nanoparticles on light-matrix supports are highlighted, enabling automated morphological and compositional assays.
• Battery and complex oxide studies: 3D mapping of transition-metal valence fluctuations, mass–thickness changes, and phase segregation under cycling or environmental stimuli (Oh et al., 27 Sep 2025).
• Atom-by-atom chemical mapping in heterogeneous nanomaterials: fast, lensing-model-driven assignment of species and sequence in atomic columns, critical for high-entropy alloys or core-shell nanostructures (Zhang et al., 2022).
• Direct determination of light-element (e.g., O) site occupancy, site-specific defect mapping, and quantification of oxygen vacancies with picometric sensitivity (Mostaed et al., 2020).

Limitations are primarily dictated by signal and noise characteristics, angular/detector coverage (for cADF and 4D-STEM), beam broadening in low-kV regimes, and loss of coherence for thin or low-Z samples. Nonlinearities from beam absorption and neighbor-column crosstalk can confound strict Z- and thickness quantitation, necessitating hybrid (analytical plus simulation) approaches for thick/complex cases (Lagos et al., 2015, Zhang et al., 2022). The indirect nature of light-atom detection via heavy-atom channeling requires rigorous reference averaging and subtraction to ensure validity (Mostaed et al., 2020).

7. Perspectives and Generalizations

The general philosophy of complementary ADF analysis—to extract maximal, otherwise hidden, structural and compositional insight by strategic integration of ADF contrast with auxiliary modalities, physical modeling, or computational subtraction—has wide ramifications. While initially rooted in physical electron microscopy, the underlying mathematical strategies (differential contrast, incoherent addition, deconvolution, and analytic modeling) are adaptable to multi-signal data fusion problems elsewhere in the physical sciences. In atom-resolved spectroscopy and tomography, joint or fused reconstructions that exploit the mutual constraints of ADF and chemical/electronic channels will continue to expand the frontier of quantitative, mechanism-resolved materials analysis (Esser et al., 2022, Oh et al., 27 Sep 2025, Zhang et al., 2022). The essential limitations imposed by detector geometry, sample preparation, beam dose, and computational tractability frame ongoing methodological innovation in this domain.