ABSTRACT
CONTENTS 8.1 Introduction ........................................................................................................................ 259 8.2 Scanning Transmission Electron Microscopy ................................................................ 261 8.3 Theoretical Microscope ..................................................................................................... 267 8.4 Structure and Properties of Si=SiO2=HfO2 Gate Stacks
by Aberration-Corrected STEM and Theory ................................................................. 271 8.5 Future Directions ............................................................................................................... 276 Acknowledgments ..................................................................................................................... 277 References.................................................................................................................................... 278
The quest for smaller, faster semiconductor devices is rapidly approaching the so-called ‘‘end of the roadmap’’ for silicon-based devices, in which the thickness of the gate oxide is becoming too small to maintain the macroscopic properties of that mainstay of the semiconductor industry, SiO2. As the thickness approaches the size of the intrinsic ring structure of SiO2, the electronic integrity of the gate oxide becomes compromised, and undesirable characteristics such as leakage and dielectric breakdown cannot be managed effectively. This situation is the driver for the search for alternative, high dielectric constant (high-k) gate structures. The introduction of new materials to augment a thin SiO2 interlayer raises the possibility of new kinds of impurities entering either SiO2 or the Si substrate, and that dopant atoms may enter new configurations, adversely affecting macroscopic device properties. It therefore becomes very valuable to be able to characterize these nanometer scale device structures with probes that are sensitive to individual atoms, ideally in three dimensions. At the same time, the ability to determine local electronic structure with similar resolution and sensitivity would reveal any local defect states introduced into the band gap as a result of these impurity or dopant atoms. In conjunction with theoretical calculations of electronic structure, it would then become possible to link the atomic-scale characterization of actual device structures to their macroscopic
current or low mobility. In this chapter, we describe the latest advances in aberration-corrected scanning trans-
mission electron microscopy (STEM), which have in the last few years brought this vision to a practical possibility. The ability to correct for the primary aberrations of electron lenses represents a revolution in the field of electron microscopy. The rate of instrumental advance today is faster than at any time since the invention of the electron microscope in the 1930s, as shown in Figure 8.1 [1]. Aberrations occur inevitably in round lenses, as a result of physics, and not only from imperfections, as was recognized by Scherzer over 70 years ago [2]. The spherical aberration of the simple round lens has limited electron microscope resolution for most of its history, as was appreciated by Feynman in his 1959 lecture, [3] ‘‘There’s Plenty of Room at the Bottom,’’ where he explicitly called for a 100-fold improvement in resolution by overcoming spherical aberration, ‘‘why must the field be symmetric?’’ Designs for aberration correctors have existed for over 60 years [4], but they require multipole lenses, which makes such schemes enormously complex. It was not until the era of the fast computer and the efficient detection of the chargecoupled device that it has become practically possible to measure aberrations with the necessary sensitivity, 1 part in 107, and to control all the multipoles to the required accuracy. Focusing, or tuning, the aberration corrector involves optimization in a space of 40 or more dimensions, which has proved to be beyond human capability. Therefore, it is not so surprising that the successful correction of aberrations in electron microscopy has had to wait for the computer age. Now, aberration correction has been successfully achieved both in the conventional
transmission electronmicroscope (TEM) [5] and in the STEM [6]. In the STEM, direct imaging of a crystal lattice has been achieved at subangstrom resolution; see Figure 8.2 [7]. But aberration correction brings more than just a factor of 2 or 3 in resolution. As shown, in the STEM it also brings a greatly improved signal-to-noise ratio, resulting in the ability to image individual Hf atoms inside the nanometer-wide gate oxide of an advanced dielectric device structure [8]. Furthermore, since the lens aperture is much wider after
0.1
105 1800 1840 1880 1920 1960 2000 2040
Year
R es
ol ut
io n
(Å )
aberration correction, the depth of field is much smaller, and it becomes possible to locate atoms in three dimensions to a precision better than 1 nm in depth and 0.1 nm in lateral position [9,10]. Three-dimensional mapping of heavy atoms inside SiO2 has become possible, and in principle, three-dimensional spectroscopy should be feasible. With this level of sensitivity, new insights are available into the microscopic defects in device structures and their influence on device characteristics. From such knowledge, properties can be calculated from first principles, enabling the determination of the atomic-scale defects responsible for macroscopic device properties such as leakage and mobility, leading to improved device design.
