Elsevier

Ultramicroscopy

Volume 108, Issue 12, November 2008, Pages 1653-1658
Ultramicroscopy

Experimental quantification of annular dark-field images in scanning transmission electron microscopy

https://doi.org/10.1016/j.ultramic.2008.07.001Get rights and content

Abstract

This paper reports on a method to obtain atomic resolution Z-contrast (high-angle annular dark-field) images with intensities normalized to the incident beam. The procedure bypasses the built-in signal processing hardware of the microscope to obtain the large dynamic range necessary for consecutive measurements of the incident beam and the intensities in the Z-contrast image. The method is also used to characterize the response of the annular dark-field detector output, including conditions that avoid saturation and result in a linear relationship between the electron flux reaching the detector and its output. We also characterize the uniformity of the detector response across its entire area and determine its size and shape, which are needed as input for image simulations. We present normalized intensity images of a SrTiO3 single crystal as a function of thickness. Averaged, normalized atom column intensities and the background intensity are extracted from these images. The results from the approach developed here can be used for direct, quantitative comparisons with image simulations without any need for scaling.

Introduction

High-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM or Z-contrast) is remarkably sensitive to the atomic number (Z) [1], [2], [3]. Quantitative HAADF imaging thus holds enormous potential as chemical information could be extracted in parallel with information on the atomic structure with sub-Ångström resolution. To date, most comparisons between experimental and theoretical HAADF images have been based on image contrast or scaling by an arbitrary amount [4], [5]. Such comparisons are only semiquantitative and place severe limitations on identifying the origins of the contrast mismatch between experiments and simulations [5]. In a previous study, Singhal et al. have reported absolute intensity measurements of small Re particles, using a calibrated HAADF detector [6]. Placing HAADF images on an absolute scale, with image intensities that are quantified relative to the incident beam intensity, will enable direct comparison of experiment with theory [6], [7], [8] and compositional mapping [4].

A prerequisite for quantitative HAADF imaging is that the detector output is directly proportional, over the entire range of image intensities, to the time-averaged flux of electrons reaching the detector. Modern annular dark-field (ADF) detectors are fabricated using YAlO3 perovskite (YAP) doped with Ce as the scintillator material [9]. These detectors exhibit the necessary properties for quantifying HAADF images. In particular, YAP is resistant to radiation damage and the decay lifetime for YAP is ∼30 ns [10], [11]. The short lifetime ensures that adjacent image pixel intensities do not affect one another. This is in contrast to the charged-coupled device (CCD) detectors often used in conventional high-resolution transmission electron microscopy (HRTEM) for which scattering of electrons and photons within the scintillator and optical coupling fibers tend to blur the image. These blurring effects must be accounted for in quantitative HRTEM [3], [12], [13]. Furthermore, as will be shown in this paper, the incident beam intensity can be quantified with the ADF detector, without saturation, so that image intensities can be normalized to the incident beam.

To ensure linearity of the imaging process, the steps involved from when the electrons strike the detector to when an image is finally displayed on the screen must also be considered. The voltage across the anode and cathode of the photomultiplier tube (PMT), which converts the time-averaged photon flux to a photocurrent, is used to control the gain (image contrast). The output from the PMT is then passed to a preamplifier that converts the photocurrent to a measurable voltage. Adding a constant voltage to the output, the offset of the preamplifier, changes the brightness of the image. The preamplifier output is then read by the imaging hardware that digitizes the signal via an analog-to-digital (A/D) converter circuit. With commercial microscope software, the resulting image is 16-bit, meaning that only 65,536 possible values of gray can be associated with each image pixel.

The objectives of the present study are (i) to report on an experimental method for truly quantitative HAADF imaging in which image intensities are placed on a scale normalized to the incident beam intensity and (ii) to establish the experimental conditions for a linear imaging process. The dynamic range of standard microscope signal processing hardware is too limited for these experiments. Therefore, a 24-bit dynamic signal analyzer (DSA) was used to directly measure the detector preamplifier output. We develop a procedure to verify that the detector output is sufficiently linear over the intensity range of interest for quantitative HAADF imaging. We construct normalized images from the data and show that image intensities can be obtained on an absolute scale for a wide range of sample thicknesses and atomic numbers.

Section snippets

Sample preparation

A SrTiO3 single crystal was chosen as a model material. Transmission electron microscopy (TEM) samples were prepared along 〈1 0 0〉 using a MultiPrep system (Allied High Tech Products Inc., Rancho Dominguez, California) and a procedure described in the literature [14]. Final thinning was achieved by Ar-ion milling (Fischione Model 1000 Ion Mill) at voltages between 0.8 and 2 kV. The samples were etched in buffered HF to remove surface damage and plasma cleaned with a 25/75% O2/Ar mixture.

Instrumentation

A

ADF detector characterization and measurements of the incident beam intensity

Changing the brightness via the TEM control software varies the detector preamplifier offset. The range of values for which the preamplifier showed linear behavior was determined by measuring the detector output voltage as a function of brightness with the microscope column values closed. As shown in Fig. 3, the output voltage changed linearly over a wide range and saturated below ∼–2.0 and above ∼2.10 V. In principle the entire linear range could be used; however, to provide a way to focus the

Conclusions

We have shown that current generation ADF detectors can be used to measure the incident beam intensity and to place atomic column and background intensities in atomic resolution HAADF images on an absolute scale. We have developed a method to characterize the ADF detector and imaging system performance. The procedure is straightforward and does not require specialized equipment. The detector output is sufficiently linear for intensities ranging from the incident beam to image intensities

Acknowledgements

The authors gratefully acknowledge Prof. Les Allen and Dr. Scott Findlay for many stimulating discussions. We also thank Mr. Zahid Chishty, Mr. Bill Anderson and Dr. Fred Kiewiet of FEI for discussions regarding the ADF detector hardware. The research was supported by the US DOE (DE-FG02-06ER45994) and NSF (DMR-0804631). J.M. L. also thanks the US Department of Education for a fellowship under the GAANN program (Grant no. P200A07044). The work made use of the MRL Central facilities supported by

References (23)

  • S.C. Anderson et al.

    Ultramicroscopy

    (1997)
  • D.O. Klenov et al.

    Ultramicroscopy

    (2006)
  • A. Singhal et al.

    Ultramicroscopy

    (1997)
  • A. Howie

    Ultramicroscopy

    (2004)
  • E.J. Kirkland et al.

    Ultramicroscopy

    (1996)
  • V.G. Baryshevsky et al.

    Nucl. Instrum. Methods Phys. Res. B

    (1991)
  • M.V. Korzhik et al.

    Nucl. Instrum. Methods Phys. Res. B

    (1992)
  • K. Du et al.

    Ultramicroscopy

    (2007)
  • Z.H. Yu et al.

    Ultramicroscopy

    (2003)
  • M.J. Fransen et al.

    Appl. Surf. Sci.

    (1999)
  • A.V. Crewe et al.

    Science

    (1970)
  • Cited by (212)

    View all citing articles on Scopus
    View full text