Transcript
TVIPS CMOS cameras, an essential and affordable tool for biologists and material scientists R. Ghadimi, I. Daberkow, P. Sparlinek, Y. Maniette, C. Kofler, and H. Tietz TVIPS GmbH, Eremitenweg 1, 82131 Gauting, Germany Introduction In the last two decades, digital cameras became an important imaging tool for Transmission Electron Microscopy (TEM) in both, life sciences and materials science. However, the development of digital cameras, which are applicable for a wide acceleration energy range, and which simultaneously satisfy different criteria, including large detection area, high Signal-to-Noise Ratio (SNR), high spatial resolution, high dynamic range and fast read out, turn out to be a challenge. TVIPS cameras based on CMOS (Complementary Metal Oxide Semiconductor) technology with active pixel sensor have been introduced to the market in 2006. Meanwhile, three different cameras are available: TemCam-F816 (8k×8k), TemCam-F416 (4k×4k) and TemCam-F216 (2k×2k). These cameras fulfill all criteria mentioned above due to the large pixel size of 15.6 µm with high fill factor, innovative fast readout, done simultaneously on 8, 4, or 2 channels at 10 MHz each, and the usage of lowest noise sensors resulting in high dynamic range (> 10.000:1). These detectors are able to detect single electrons, opening a new door to measure the performance parameters (resolution and sensitivity) without additional hardware [1,2]. This novel method, developed by TVIPS, based on calculation of the Point Spread Function (PSF) using Single Electron Events (SEE) [3]. The influence of the High Tension (HT) on the sensitivity, describing the conversion rate, demonstrates the ability of such a camera to be used in a wide range of electron energies.
Fig. 1: TemCam-F816, TemCam-F416 and TemCam-F216. A new generation of 16 bit cameras based on CMOS technology for high-end scientific imaging in transmission electron microscopy.
Detection of Single Electron Events (SEE) Procedure steps [3]: At very low intensities (~ 1 e-/300 pixel) SEEs are clearly visible and well separated (Fig. 2 (a)). All separated SEE can be automatically localized by a special developed software using the following criteria: Upper threshold: 50 counts Lower threshold: 2 counts, corresponding to camera readout noise A mask with a kernel size of 7×7 pixels is used to find pixels with the maximum count Pixelmax. Hot pixels and x-ray events can be localized and discarded by the software, if the average value of the next neighbor pixels of Pixelmax is lower than 3 counts. (a)
Determination of the Modulation Transfer Function (MTF) The resolution of TEM cameras can be described by the MTF which is a frequency response of a detector describing the modulation of contrast, measured at different spatial frequencies. To quantify the MTF, the point spread function (PSF) of the camera should be determined and then the MTF is generated by the Fourier Transformation of the PSF. All single events are expanded in a 16× larger image by a bi-cubic spline interpolation and the centers of gravity (COG) are calculated. The events are averaged by using the individual COG. This averaged SEE is the PSFSEE (Fig. 2 (b), inset). The modulus of the Fourier transform of the PSFSEE gives the MTF (Fig. 2 (c)). Due to the interpolation the MTF is already aliasing corrected.
(b)
(c)
Fig. 2: (a) TEM image at ultra-low electron dose condition imaging the beam stop at 200 kV recorded with TemCam-F416 CMOS camera. SEEs are clearly visible. The pseudo color of two SEEs is shown in inset. (b) Histogram of integrated signals generated by 12169 detected SEE. Inset: Averaged 3D plot of the point spread function using 12169 SEEs which have been 16× over-sampled by bi-cubic spline interpolation before. The integrated number of counts in the averaged Single Electron Point Spread Function (inset) represents the average counts per electron. (c). Comparison of the MTF curves calculated by SEE method (red) and standard edge method (blue). The reasons for the differences between these two MTF curves in low frequency region are the diffused scattered electrons from the edge which can be clearly observed underneath the edge (not shown). Even at Nyquist frequency (fNyq) the MTF of 5% is sufficient to resolve contrast at high frequencies close to the fNyq.
Resolution
Sensitivity
As an alternative method to the MTF we have used a direct way to demonstrate the spatial resolution of cameras under different low dose conditions. Cross grating replica is an ideal specimen for comparison of resolution versus dose. The magnification was chosen so that the 2.3 Å gold lines appear close to Nyquist frequency f Nyq.
The sensitivity describing the conversion factor in terms of counts per incident electron has been measured by the screen current amplifier. This is equivalent to the integrated number of counts in the averaged PSFSEE (Fig. 2 (b)) which represents the average counts per electron.
(a)
(b)
TemCam-F416
TemCam-F416
(c)
(d)
Fig. 5: Sensitivity as function of high tension. Even at 8 kV we observed a sensitivity of about 5 counts. With a noise level of about 2.5 counts the SNR is still about 2:1. These experimental results demonstrate the ability of such a camera to be used in a wide range of electron energies.
Summary & Examples TVIPS TemCam cameras are able to detect SEEs and demonstrate a high resolution at Nyquist frequency even under extreme low dose conditions. Therefore these cameras are an ideal recording device for life sciences (cryo-microscopy) and material science (recording diffraction patterns, energy filter imaging and spectroscopy) applications. These cameras can be used for a wide range of accelerating electron energies. Unlike CCDs, CMOS sensors do not show any blooming effects or smearing artefacts owing to intensive illumination or overexposure. This inherent ability renders them entirely suitable for all types of applications using high intensities such as recording diffraction patterns. TemCam-F416
TemCam-F816
Fig. 3: 4k×4k power spectrum of: (a) a gold shadowed cross grating replica at an electron dose of 288 e-/Ų showing clearly the 2.3 Å gold lines close to Nyquist frequency fNyq. Reflexes beyond fNyq can be observed (indicated by circles). (b) Even under extreme low dose conditions the gold reflexes close to the fNyq are clearly visible in the power spectra (indicated by arrows). Images were taken on a JEM-2010 (LaB6 , total mag. 130.000×, 200 kV). (c) The back convoluted reflexes containing frequencies higher than fNyq can be observed using the TemCam-F416 (Tecnai F20, total mag. ~ 119.000×, 200 kV), and (d) using the TemCam-F816 (Titan Krios, total mag. 113.000×, 300 kV). The 4k×4k power spectrum has been calculated from an 8k×8k image with 2× binning.
(b)
(a)
(c)
Dose: 36 e-/Å2
TemCam-F416
Dose: 12 e-/Å2
TemCam-F816
Fig. 4: In addition to cross grating (mentioned above), a thin amorphous carbon film is also a good test sample to demonstrate the resolution of a camera system. Power spectrum of a thin amorphous carbon film under low electron dose condition acquired by the (a) TemCam-F416 (4k×4k image, 1× binning) and (b) TemCam-F816 (8k×8k image, 2× binning).
Fig. 6: (a) Zero loss energy filtered diffraction pattern of BiOCl super-structure acquired by the TemCam-F416 . Pseudo colour map of central reflex (inset) demonstrates the saturation without any blooming effect due to CMOS technology. (b) Tobacco mosaic virus on carbon film demonstrates an excellent contrast even at 4 kV recorded by the TemCam-F416 (c) Cryo-EM reconstruction of the TcdA1 prepare toxin complex Photorhab-dus at 6.3 Å resolution. Data was acquired on the TemCam-F816 [4].
References: [1] I. Daberkow, K.-H. Herrmann, L. Liu and W. D. Rau: “Performance of electron image converters with YAG single-crystal and CCD sensor“, UM 38 (1991) 215-223 [2] W. J. de Ruijter and J.K. Weiss: “Methods to measure properties of slow-scan CCD cameras for electron detection“, Rev. Sci. Instr. 63 (1992) 4314-4321 [3] H. R. Tietz, R. Ghadimi and I. Daberkow: “Single Electron Events in TEM - Measuring the Resolution of Fiber Optic Coupled CMOS Cameras“, Imaging & Microscopy 4 (2012) 46 - 48 [4] C. Gatsogiannis, A. E. Lang, D. Meusch, V. Pfaumann, O. Hofnagel, R. Benz, K. Aktories and S. Raunser: ” A syringe-like injectionmechanismin Photorhabdus luminescens toxins”, Nature, 495 (2013) 520–523
