No abstract

In many cases, electron counting with direct detection sensors offers improved resolution, lower noise, and higher pixel density compared to conventional, indirect detection sensors for electron microscopy applications. Direct detection technology has previously been utilized, with great success, for imaging and diffraction, but potential advantages for spectroscopy remain unexplored. Here we compare the performance of a direct detection sensor operated in counting mode and an indirect detection sensor (scintillator/fiber-optic/CCD) for electron energy-loss spectroscopy. Clear improvements in measured detective quantum efficiency and combined energy resolution/energy field-of-view are offered by counting mode direct detection, showing promise for efficient spectrum imaging, low-dose mapping of beam-sensitive specimens, trace element analysis, and time-resolved spectroscopy. Despite the limited counting rate imposed by the readout electronics, we show that both core-loss and low-loss spectral acquisition are practical. These developments will benefit biologists, chemists, physicists, and materials scientists alike.

The serial detectors traditionally used in Electron Energy Loss Spectroscopy (EELS) have mostly been replaced by parallel detection systems, for reasons of efficiency and ease of use. So far most parallel detection systems have been based on fiber or lens coupled Photo Diode Array (PDA). Although these systems have proven satisfactory they have some limitations: 1) high readout noise, 2) limited correction for gain variations and 3) point spread function or cross talk between channels all of which effects the ultimate sensitivity and resolution. This paper discusses a newly developed parallel detector for EELS based on a CCD and a new type of scintillator and fiber technology. The new detector shows large improvements in all areas of performance.The sensitivity of a detector is limited by the readout noise at low incident electron dose and the gain correction at high dose. Both these areas have been addressed in the new design. The readout noise is equivalent to approximately 1/2 a primary electron, about 40 times better than a typical PDA.

Transmission electron microscopes (TEMs) conventionally employ indirect detection cameras (IDC) for electron imaging. Such IDCs consist of a scintillator and a digital imaging device with a lens or fiber optic network coupling photons from the scintillator to the camera. Alternatively direct detection cameras (DDC) directly image electrons. Compared to IDCs, DDCs offer an improved point spread function (PSF), lower read-out noise, and potential for higher frame rates [1,2]. DDCs have been successfully utilized by the cryo-TEM community [3] and, more recently, for in-situ TEM applications [4]. Here we evaluate a DDC for the application of electron energy-loss spectroscopy (EELS). We compared the performance of a Gatan K2 Summit (DDC) with a Gatan US1000FTXP (IDC). Both detectors were mounted to a Gatan GIF Quantum energy filter. Our results show that the narrow PSF of the DDC improves measured resolution given a fixed beam energy spread and spectrometer dispersion. Additionally, the low read-out noise of the DDC increases spectrum signal to noise (SNR) for short acquisition times. These results indicate DDCs will enable efficient acquisition of low-noise spectra for applications ranging from in-situ EELS to low-dose chemical mapping. Figure 1 shows a zero loss peak (ZLP) captured with each detector. In both cases the spectrometer was set to 1 eV/channel dispersion. The DDC was mounted on a FEI T20 (LaB 6 ), which has an energy spread full width at half maximum (FWHM) ≈ 1 eV. The detected ZLP has a FWHM of 1 eV, demonstrating that the DDC is able to accurately detect the ZLP without significant signal spreading, a result of the DDC's narrow PSF. The IDC, mounted on a FEI TF20 (Schottkey) with a FWHM ≈ 0.8 eV, recorded a ZLP with a FWHM of 3 eV. The comparatively broad IDC PSF prevents accurate detection of the ZLP at this low dispersion.A SrTiO 3 sample was investigated with each detector. Figure 2 shows Ti L edge spectra acquired with three different acquisition times. All spectra acquired with the DDC display splitting of the Ti L 2,3 edges, however, spectra acquired with the IDC detector do not show splitting. Again, this result is attributed to the differing PSFs of the detectors. The narrow PSF of the DDC is able to resolved the Ti L 2,3 edge splitting while to broader PSF of the IDC cannot resolve the splitting at this dispersion.For each spectra shown in Figure 2, pre-edge SNR values were calculated (Table 1). A 50 eV pre-edge window was defined and SNR was calculated as SNR = N / σ, where N is the average number of counts and σ is the signal standard deviation. For spectra acquired with 10.0 and 1.0 second acquisition times, the IDC provided higher SNR while the DDC provided higher SNR for 0.1 second acquisition time. This transition is due to the interplay between shot noise, read-out noise, detector PSF, and electron dosage. For short acquisition times, read-out noise contributes significantly to total spectrum noise. In this regime, the DDC provides higher SNR because of its low read-out noise. For lon...

The emergence of commercially available Cs aberration corrected transmission electron microscopes [1] in combination with monochromators [2,3] raises the question how spherical aberration correction may affect the energy resolution of a high resolution sub-50 meV electron energy loss spectrometer [4]. Spherical aberrations of the objective lens are not only detrimental to spatial resolution, but can also significantly degrade the spectral energy resolution of an electron energy loss spectrum or limit the slit width used for energy-filtered diffraction imaging.The filter's electron optical energy resolution is a function of 1.) filter aberrations and 2.) the optical coupling of the specimen with the spectrometer, called image-spectrum mixing. An energy filter can be coupled to the microscope optics either in diffraction mode or in image mode. A postcolumn filter focuses the projector lens crossover into an energy dispersed spectrum. In diffraction mode the projector crossover contains an image of the specimen. Conversely in image mode the projector crossover contains a diffraction pattern of the specimen. Most electron energy loss spectroscopy (EELS) measurements, STEM spectrum imaging and energy filtered diffraction patterns are acquired in the image-coupled filter mode. In comparison to diffraction-spectrum mixing much less is known about image-spectrum mixing. The effect of diffraction-spectrum mixing has been shown for both in-column [5,6] and post-column filters [7,8]. It has also been mathematically presented in terms of transmissivity [9], which is the imaged area as a function of solid scattering angle that can be passed through the filter for a given energy resolution defined by the filter aberrations. Calculating the size of the image in the projector crossover using the approach in [9], then for a camera length of 15 mm and a probe size of 5 nm the image size in the spectrum of the post-column filter is only 0.08 eV. However, this analysis neglects the effect of spherical aberrations of the objective lens. Accounting for spherical aberrations, it can be shown that the size of the projector crossover in the spectrum at the disc of minimum confusion with an objective lens Cs of 1.1 mm and a 100 mrad collection angle amounts to 9.3 eV. This is a much more realistic value of the projector crossover size in the spectrum of an image-coupled filtered mode. If most of the image magnification is obtained with the objective lens, the aberrations of all the other electron microscope lenses are expected to be far less in shaping the projector crossover waist. This condition may vary for different microscopes.A post-column filter with partial 3 rd order aberration correction and a 5 mm entrance aperture was used to acquire an energy filtered CBED pattern of Si <111> at 200 kV on a conventional TEM with a LaB6 emitter. The filter aberrations are shown in FIG 1. The non-isochromaticity is 1.5 eV. The Cs of the objective lens is 1.1 mm. FIG 2 shows an energy filtered diffraction pattern, which is an image-coupled filt...