We have shown that electron energy loss spectroscopic imaging (EELS) in the scanning transmission electron microscope (STEM) can provide single atom sensitivity for elements such as calcium and iron bound to proteins adsorbed onto thin carbon films. A spectrum-imaging system incorporating a 100x1340 pixel CCD detector has enabled the parallel collection of EELS data at each image pixel. Advantages of this system are (i) the low noise level, which provides higher analytical sensitivity, (ii) the short read-out time of around 10 ms, which enables faster acquisition, (iii) the two-dimensional array of square pixels, which allows exact correction for channel gain variations, and (iv) the improved point-spread function of the optical coupling of the detector array to the scintillator. Compensation for specimen drift is achieved in real-time by cross-correlating annular dark-field images recorded at the end of each scan line. For example, we have applied spectroscopic imaging to study a model neurodegenerative disorder in an IRP2 knockout mouse. In these experiments, it has been possible to map the distributions of individual ferritin molecules in oligodendrocytes and neurons in unstained sections of brain. These data from selected small regions of tissue are correlated with electron tomographic measurements performed in NCI. We are currently implementing new software to improve the accuracy of spectral analysis in order to reduce the elemental detection limits further.
