The binding energies of an atom's core-state electrons are generally changed by chemical bonds, an effect called the chemical shift. A well-known example is the L 2,3 edge in silicon, which changes from 99 eV in pure silicon to 102.9 eV in silicon nitride to 104 eV in silicon dioxide [1] . When aluminum is oxidized, both its L 2,3 and its K-edges shift up by 2-4 eV from the 73 eV characteristic of the pure metal [2, 3] . With sub-1 eV energy resolution, electron energy loss spectroscopy (EELS)

more »

... s commonly employed to detect these shifts, which are small compared to the binding energy. Here, we seek to develop a technique for measuring chemical shifts that is based on energy dispersive spectroscopy (EDS) instead. When the full width at half-maximum (FWHM) of an x-ray peak is used as the metric, EDS has an energy resolution of order 100 eV [4] , which would make it useless for chemical shift analysis. But curve fitting can locate the center of a peak to much better precision than the FWHM. We perform EDS mapping on a lithographically defined, 100 nm-thick aluminum nanowire supported by a 20-nm thick silicon nitride membrane (Fig. 1A) . The aluminum is coated with the usual native oxide. For each EDS spectrum in the datacube, we fit the aluminum K-alpha peak with a Gaussian (Fig. 1B) . We are able to locate the EDS peak with a precision of 1 eV or better, which is similar to the performance that can be achieved with nonmonochromated EELS. We generate a map showing the x-ray energy at each real-space position and apply a threshold to remove noise where no x-rays from the aluminum are detected ( Fig. 2A) . Averaging along the vertical (spatial) axis produces a line profile (Fig. 2B) . Although no significant features are evident near the edge, a linear fit shows that the peak energy decreases by 0.8 eV between the beginning and the end of the profile. While smaller than the 2 eV expected based on EELS results from the literature, the data show energy resolution that should be sufficient to detect chemical shifts. Though still ongoing, our work so far implies detectability; we are working to develop a sample optimized for demonstrating EDS-detectable chemical shifts. https://www.cambridge.org/core/terms. https://doi.