UCLA

As devices are scaled to ever smaller dimensions, their performance becomes increasingly governed by nanoscale heterogeneity in the materials they made of. Rather than the bulk average, local variations in composition, structure, and electrostatic environment start to dominate the materials’ properties. Understanding the behavior, locating the failure point, and predicting the scaling limits of next-generation microelectronic components therefore require quantitative, spatially resolved characterization techniques capable of measuring chemical bonding, temperature, and electrical properties at the relevant length scales. This work advances three such techniques, each leveraging the unique capabilities of scanning (transmission) electron microscopy—STEM and SEM—to push measurement precision and characterization capability beyond what is conventionally considered possible.First, we demonstrate that energy-dispersive x-ray spectroscopy (EDS), despite its ∼140 eV energy resolution, can detect sub-eV chemical shifts in transition energies through statistical averaging in the shot-noise-limited regime. Systematic measurements across compounds of tungsten, titanium, and aluminum reveal that EDS chemical shifts can be either positive or negative—in contrast to EELS binding energy shifts—refecting the diferential perturbation of the two orbitals involved in each transition. We introduce a corrected paralyzable detector model accounting for the fnite response time of the event discriminator, which enables accurate dead-time characterization and quantitative evaluation of of pile-up artifacts. Built on top of the corrected paralyzable detector model, we develop a post-acquisition pile-up correction algorithm that improves data acquisition throughput by a factor of ten while preserving sub-eV energy precision.Second, we introduce Kikuchi Difraction Thermometry (KDTh), a non-contact nanoscale thermometry technique that operates on a standard SEM platform. By expanding the parameter space of cross-correlation-based pattern matching to include lattice constant variations, KDTh achieves 2.2 K/√Hz temperature sensitivity and sub-20 nm spatial resolution— comparable to TEM-based techniques and surpassing all existing SEM-based thermometries. Dynamic temperature tracking under both stepped and continuous power ramps on Joule-heated graphite devices demonstrates linear thermal response consistent with Fourier heat conduction theory.Third, we develop a methodology for predicting ferroelectric memory scalability through in situ STEM electron-beam-induced current (EBIC) imaging. An interleaved PUNDp/NDPUn pulse protocol enables pixel-wise hysteresis loop construction and extraction of local coercive and remanent felds across a functioning TiN/HZO/W capacitor. Statistical modeling of the distribution of device averaged polarization allows for predicting the manufacturing yield as a function of operating voltage and capacitor area. The analysis demonstrates that FeRAM cells with sub-1 V operation and sub-50 nm footprints can achieve >99.9% yield with the characterized 5-nm HZO flm.Together, these studies trace an arc from fundamental physical measurements to practical device metrology, forming a general toolkit for quantitative characterization of functional materials with electron microscopy.