Visualizing Variation: Continuous EELS Mapping of Metal Nanoparticle Ensemble Phase Transformations

Electron energy loss spectroscopy (EELS) is a powerful technique for characterizing nanomaterials, providing insights into various material properties. One application is nano-thermometry, where the local temperature of individual metal nanoparticles is measured by determining the plasmon peak position, which depends on conduction electron density [1]. This method can also detect melting and crystallization, which significantly alter the density of metals. Continuous EELS mapping while heating above and below the melting point allows precise determination of melting and crystallization temperatures of individual particles. Utilizing the new in-situ EELS spectrum imaging features of the GIF Continuum®, we can acquire a continuous series of drift-corrected spectrum images over an ensemble of particles, independently monitoring each particle's melting and crystallization behaviors, which were found to be heterogeneous.

In this work, a series of EELS spectrum images were acquired and processed, along with temperature data from a MEMS-based heating holder from DENSsolutions. With modern fast detectors and spectrometers, spectrum images containing thousands of spectra can be acquired in less than a second, making continuous in-situ spectrum imaging feasible. The holder temperature data was automatically synchronized and correlated with the EELS spectrum image data during acquisition. The entire series of in-situ EELS spectrum images was then rapidly fit using the built-in NLLS tools in DigitalMicrograph^®, yielding several series of synchronized fit maps. After summing the EELS spectra over a single nanoparticle, plots of the peak position over time were generated and plotted against the nominal temperature from the holder using scatterplots in DigitalMicrograph. The plasmon peak position within each particle indicated whether the Sn was melted or crystallized at that time.

The new in-situ EELS spectrum imaging capability was applied to a Sn nanoparticle sample oscillated above and below its melting temperature with varying ramp rates. Spectrum images were recorded at a rate of 1 frame every 1.54 seconds (2000 spectra/second). Observations of the maps of plasmon position over time revealed that while most Sn particles crystallized during every cycle, some particles occasionally did not crystallize even though surrounding particles did. Images and 4D STEM maps of the particles after the in-situ recording did not indicate a clear difference between a particle that always crystallized and those that sometimes did not. Some variation could be attributed to random chance, as the nucleation and growth of a crystalline nanoparticle from a nano-droplet are governed by both thermodynamics and kinetics. However, the consistent non-crystallization of some particles suggests other factors at play. However, there are only 3 particles (of 11 in the field of view) that sometimes stay melted, and one of them remains melted during 4 cycles. This is highly unlikely to be the result of purely random statistical fluctuations in nucleation time.

This heterogeneous and stochastic behavior at the nanoscale can only be observed with high spatial and temporal resolution. In-situ electron microscopy, particularly in-situ EELS spectrum imaging, is an excellent technique for exploring these dynamics.

[1] Mecklenburg, M. et al. Nanoscale temperature mapping in operating microelectronic devices. Science 347, 629–632 (2015).