Abstract text (incl. figure legends and references)

Introduction

Spatially resolved STEM-EELS is widely used to investigate local elemental and electronic structures at an atomic scale. However, acquiring spectra with both a high signal-to-noise ratio (SNR) and a high spatial resolution is difficult even if we use advanced direct counting CMOS camera because of the limitations imposed by practical experimental conditions (e.g., instrument instability, electron damage to the sample, limited dynamic range and low probe current). Hence, a multiple acquisition technique is sometimes useful and can improve the SNR, avoiding both sample damage and image drift.

Objectives

Previously, we reported the detection of extremely low count in EELS multiple spectrum imaging (SI) experiments by randomizing the CCD dark reference noise [1]. Although a direct-detection CMOS camera is useful, it does not always provide superior performance compared with a CCD camera. Thus, understanding the advantages and disadvantages of direct-detection and CCD cameras is important for obtaining optimal experimental data. Here, we compare the performance and detection limits for a CMOS K2 camera operated in counting mode and a conventional CCD camera in EELS experiments. We investigate the detection limit by focusing on the electron dose entering the camera [2].

Materials & methods

STEM–EELS experiments were carried out on a Themis Z (300 kV) equipped with a GIF Quantum970 spectrometer. The spectra were recorded using a direct-detection K2-IS camera and an UltraScan CCD camera (Gatan). The test spectra of Ti L_2,3-edge were measured from SrTiO₃.

Results

In the case of a single spectrum acquired at the shortest dwell times (2.5 ms for K2 and 1 μs for CCD), the detection limit, defined as three times the standard deviation of the spectral noise (3σ), was very low (1 e⁻/channel) in the K2 camera compared with that acquired with the CCD camera (5 e⁻/channel). By contrast, the spectral noise of the K2 camera changed depending on the dwell time because of the multiple read-outs related to its fixed frame rate (400 fps). The spectral noise of the K2 camera was larger than that of the CCD camera when the dwell time was longer than ~30 ms. Thus, the CCD camera was found to be still useful when detecting a very small number of electrons with a long acquisition time. However, in the case of an accumulated spectrum obtained by acquiring 10,000 spectra after subtracting the average dark reference signal, the detection limits per read-out were ~0.016 and ~0.025 e⁻/channel/read-out for the K2 and CCD cameras, respectively.

Conclusion

Because both cameras have advantages and disadvantages with respect to their detection limit, speed, and dynamic range, their proper use is important.

Reference

[1] M. Haruta, Y. Fujiyoshi, T. Nemoto, A. Ishizuka, K. Ishizuka and H. Kurata, Ultramicroscopy, 207 (2019) 112827.

[1] M. Haruta, K. Kikkawa, K. Kimoto and H. Kurata, Ultramicroscopy, 240 (2019) 113577.

This work was supported by Kakenhi Grants-in-Aid No. 22H01956 from the Japan Society for the Promotion of Science (JSPS).