1090 ELECTRON TECHNIQUES Reimer, L. (ed). 1995. Enorgy-Filtering Transmission Electron ‘Microscopy. Springer-Verlag, Berlin, Rioja, R. J, and Laughlin, D, E, 1977. The early stages of GP zone formation in naturally aged Al-t wt pet Cu alloys. Metall ‘Trans, 8A:1257-1261, Sawyer, L. C. and Grubb, D. T, 1987. Polymer Microscopy. Chap- ‘man and Hall, London. Schwartz, L, H. and Cohen, J.B, 1987. Diffraction from Materials, ‘2nd ed. Springer-Verlag, New York. Smith, F. G. and Thomeon, J. H. 1988. Optics, 2nd ed. John Wiley ‘& Sons, Chichester Suenitzky, D.W. and Johnson, K.D. 1998, Focused ion beam (FIB) ‘milling damage formed during TEM sample preparation of sli- con. In Microscopy and Mieroanalysis 1998 (G. W. Bailey, K.B. Alexander, W. G. Jerome, M.G. Bond, and J.J. McCarthy, eds.) pp. 656-867. Springer-Verlag, New York. ‘Thomas, G. and Goringe, M. J. 1979. Transmision Electron ‘Microscopy of Metals, John Wiley & Sone, New York, ‘Voolll,E., Alexander, K.B., Mabon, J. C., O'Keofe, M.A., Postel, ‘M. J, Wright, M.C, and Zaluzee, N. J, 1998, ‘The DOE2000 Materials MicroCharacterization Collaboratory. In Electron Microscopy 1998, Proceedings of the 14th International Con- ‘gress on Electron Microscopy (HI. A. Calderon Benavides and ‘M. Jose Yacaman, eds.) pp. 289-299, Institute of Physica Pub- lishing, Bristol, UK ‘Weatherly G. C. and Nicholson, R. B, 1968, An electron microscope investigation of the interfacial structure of semi-coberent pre- cipitates. Philos. Mag. 17:801-831. Williams, D. B. and Carter, C. B. 1998, Transmission Microscopy: A Textbook for Materials Science. Plenum Press, New York KEY REFERENCES Edington, 1974, See above Reprinted edition available from Techbooks, Fairfax, Va. Filled with examples of diffraction and imaging analyses. Fultz and Howe, 2000, See above ‘An integrated treatment of microscopy and diffraction, with ‘emphasis on principles. Hirsch et al, 1977, See above. For many years, the essential text on CTEM, Reimer, 1993. See above. Excellent reference with emphasis on physics of electron scattering and TEM Shindo, D_ and Hiraga, K. 1998. High Resolution Electron Micro- scopy for Materials Science, Springer-Verlag, Tokyo. Provides numerous high-resolution TEM images of materials, Williams and Carter, 1996, See above, A.current and most comprehensive text on modern TEM tech: niques. INTERNET RESOURCES Ittp:www ame.enl gov ‘An excellent source for TEM information on the Web in the United Sates, Provides access to the Microscopy ListServer and a Soft- ‘ware Library as well asa connection t0 the Microscopy & Micro- analysis FTP Site and Libraries plus connections to many other, useful sites. Ittplicimewww.epfl.chiwelcometext. html A similar site based at the Beole Polytechnique Federale de Lau- ‘sanne in Switzerland that contains software and a variety of ‘electron microscopy information Inetp:vww.mea microscopy cor Provides access to up-to-date information about the Microscopy Society of America, affiliated societies, and microscopy resources that are sponsored by the society Inttpzireb info nih. gov/nib-image Public domain software developed by National Institutes of Health (US) for general image processing and manipulation. Avail able from the Internet by anonymous fp from zippy.nimh.mih. _g0v or on floppy disk from National Technical Information Ser vice, 5285 Port Royal Ra, Sprangyield, VA 22161, part number PB9S-504868, Janes M. Howe, University of Virginia Charlottesville, Virginia Brewr Fuvrz California Institute of Technology’ Pasadena, California SCANNING TRANSMISSION ELECTRON MICROSCOPY: Z-CONTRAST IMAGING INTRODUCTION As its name suggests, the scanning transmission electron microscope is a combination of the scanning electron microscope and the transmission electron microscope. ‘Thin specimens are viewed in transmission, while images are formed serially by the scanning of an electron probe. In recent years, electron probes have become available with atomic dimensions, and, as a result, atomic resolution images may now be achieved in this instrument. The nat- ure of the images obtained in scanning transmission elec- tron microscopy (STEM) can differ in significant ways from those formed by the more widespread conventional trans- mission electron microscopy (CTEM). The key difference lies in their modes of image formation; the STEM instru- ‘ment can be configured for almost perfect incoherent ima- ging whereas CTEM provides almost perfect coherent imaging. The latter technique is generally referred to as high-resolution electron microscopy (HREM), though both methods now provide atomie resolution. ‘The difference between coherent and incoherent ima- ging was first discussed over one hundred years ago in the context of light microscopy by Lord Rayleigh (1896) ‘The difference depends on whether or not permanent phase relationships exist between rays emerging from dif- ferent parts of the object. A self-luminous object results in perfect incoherent imaging, as every atom emits indepen- dently, whereas perfect coherent imaging occurs if the entire object is illuminated by a plane wave, eg., a point source at infinity. Lord Rayleigh noted the factor of 2 improvement in resolution available with incoherentSCANNING TRANSMISSION ELECTRON MICROSCOPY: Z:CONTRAST IMAGING 1091 Figure 1, Schematic showing Z-contrast imaging and atomic retolution electron energy lose spectroscopy with STEM. The image is of GaAs taken on the 300-V STEM instrument at Oak Ridge National Laboratory, which directly resolves and distin guishes the sublattice, as shown in the line trace, imaging and also the lack of artifacts caused hy interfer- ence phenomena that could be mistaken for real detail in the object. Of particular importance in the present context, he appreciated the role of the condenser lens (1896, p. 175): “Tt seems fair to conclude that the function of the eonden- ser in microscopic practice is to cause the object to behave, at any rate in some degree, as if it wore self-luminous, and thus to obviate the sharply-marked interference bands which arise when permanent and definite phase relation- ships are permitted to exist between the radiations which issue from various points of the object.” A large condenser lens provides a close approximation to perfect incoherent imaging by ensuring a range of optical paths to neighbor- ing points in the specimen, ‘A bundred years later, STEM can provide these same advantages for the imaging of materials with electrons. A probe of atomic dimensions illuminates the sample (see Fig. 1), and a large annular detector is used to detect elec- trons scattered by the atomic nuclei, The large angular range of this detector performs the same function as Lord Rayleigh's condenser Jens in averaging over many optical paths from each point inside the sample. This ren- ders the sample effectively self-luminous; ie., each atom in the specimen scatters the incident probe in proportion to its atomic scattering cross-section. With a large central hole in the annular detector, only high-angle Rutherford scattering is detected, for which the cross-section depends on the square of the atomic number (Z); hence this kind of microscopy is referred to as Z-contrast imaging. ‘The concept of the annular detector was introduced by Crewe ot al. (1970), and spectacular images of single heavy {onorent phase) prorat out \ ofprase) Figure 2. Image intensity for two point objects, P1 and P2, ilu- ‘inated coherently in phase, 180° out of phase, and with incoher- ent illumination. (After Lord Rayleigh, 1896.) atoms wore obtained (seo, e.g., Isaacson ot al, 1979). In the field of materials, despite annular detector images show- ing improved resolution (Cowley, 1986a) and theoretical predictions of the lack of contrast reversals (Engel et al., 1974), it was generally thought impossible to achieve an incoherent image at atomic resolution (Cowley, 1976; Ade, 1977). Incoherent images of thick crystalline materi- als wore first reported by Pennycook and Boatner (1988), and the reason for the preservation of incoherent charac- teristics despite the strong dynamical diffraction of the crystal was soon explained (Pennycook et al., 1990), as described below. An incoherent image provides the most direct represen- tation of a material's structure and, at the same time, improved resolution. Figure 2 shows Lord Rayleigh’s clas- sic result comparing the observation of two point objects with coherent and incoherent illumination. Each object gives rise to an Airy disc intensity distribution in the image plane, with a spatial extent that depends on the aperture of the imaging system. ‘The two point objects are separated so that the first zero in the Airy dise of one coincides with the central maximum of the other, a condi- tion that has become known as the Rayleigh resolution cri= terion, With incoherent illumination, there are clearly two peaks in the intensity distribution and a distinet dip in between; the two objects are just resolved, and the peaks in the image intensity correspond closely with the posi- tions of the two objects. With coherent illumination by a plane-wave source (identical phases at the two objects), there is no dip, and the objects are unresolved. Interest ingly, however, if the two objects are illuminated 180° out of phase, then they are always resolved, with the inten- sity dropping to zero half-way between the two. Unfortu- nately, this desirable result can only be achieved in practice for one particular image spacing (e.g, hy illumi- nating from a particular angle), and other spatial frequen- cies will have different phase relationships and therefore show different contrast, a characteristic that is generie to coherent imaging, Note also that the two peaks are signif- icantly displaced from their true positions, In an incoherent image, there are no fixed phase rela- tionships, and the intensity is given by a straightforward1092 ELECTRON TECHNIQUES Object _ oO oy —— bol os jpg bao = ' Object Funetion O1R) c LA nage Lay 1) = 008)" 4R) ae IL Figure 8. Schematic showing incoherent imaging of a thin speci- ‘men with STEM: (A) monolayer raft of Si{110); (B) the Z-contrast jobject function represents the high-angle scattering power localized at the atomic sites; (C) illumination of the sites for a 1.26-A probe located over one atom in the central dumbbell. As the probe scans, it maps out the objet function, producing an ineo- herent image. convolution of the electron probe intensity profile with a real and positive specimen object function, as shown sche- matically in Figure 3, With Rutherford scattering from the nuclei dominating the high-angle scattering, the object function is sharply peaked at the atomic positions and pro portional to the square of the atomic number. In the figure, a monolayer raft of atoms is scanned by the probe, and. ‘each atom scatters according to the intensity in the vici- nity of the nucleus and its high-angle cross-section. This gives a direct image with a resolution determined by the probe intensity profile. Later it will be shown how crystal- line samples in a zone axis orientation can also be imaged incoherently. These atomic resolution images show similar characteristics to the incoherent images familiar from optical instruments such as the camera; in a Z-contrast image, atomic columns do not reverse contrast with focus or sample thickness. In Figure 1, columns of Ga can be dis- tinguished directly from columns of As simply by inspect» ing the image intensity. Detailed image simulations are therefore not necessary, ‘This unit focuses on Z-contrast imaging of materials at atomic resolution. Many reviews are available covering other aspects of STEM, .g., microanalysis and microdiff- raction (Brown, 1981; Pennycook, 1982; Colliox and ‘Mory, 1984; Cowley, 1997). Unless otherwise stated, all images were obtained with a 200-kV scanning transmis- sion electron, microscope, a VG Microscopes HB 603U with a 1.26-A probe size, and all spectroscopy was per- formed on a 100-kV VG Microscopes HB 501UX STEM Competitive and Related Techniques ‘The absence of phase information in an incoherent image is an important advance; it allows the direct inversion from the image to the object. In coherent imaging, the structural relationships between different parts of the object are encoded in the phase relationships between the various diffracted beams. These are lost when the intensity is recorded, giving rise to the well-mown phase problem. Despite this, much effort has been expended in attempts to measure the phase relationships between dif ferent diffracted beams in order to reconstruct the object (eg, Coene et al., 1992; Orchowski et al,, 1995; Mabus, 1996; Mobus and Dehm, 1996) However, electrons interact with the specimen poten- tial, which is a real quantity. There is no necessity to involve complex quantities; in an incoherent image, there are no phases and so none can be lost. Information about the object is encoded in the image intensities, and images may be directly inverted to recover the object. As seen in Figure 2, intensity maxima in an incoherent image are strongly correlated with atomic positions, so that often this inversion can be done simply by eye, exactly as we interpret what we see around us in everyday life. With the additional benefit of strong Z contrast, structures of defects and interfaces in complex materials may often be determined directly from the image. Ina phase-contrast image, the contrast changes drama- tically as the phases of the various diffracted beams change with specimen thickness or objective lens focus, which means that it is much more difficult to determine the object uniquely. It is often necessary to simulate many trial objects to find the hest fit. The situation is espe- cially difficult in regions such as interfaces or grain bound- aries, where atomic spacings deviate from those in the perfect crystal, which ean also cause atomic columns to reverse contrast. If one does not think of the correct struc- ture to simulate, obviously the fit will be spurious. Unex- pected phenomena such as the formation of new interfacial phases or unexpected atomic bonding are easily missed. ‘Such phenomena are much more obvious in an incoherent image, which is essentially a direct image of atomic struc- ture. As an example, Figure 4 shows the detection of van der Waals bonding between an epitaxial film and its sub- strate from the measurement of film-gubstrate atomic separation (Wallis et al., 1997). At 3.2 A, this is signifi- cantly larger than bond lengths in covalently bonded semi conductors, which are typically in the range of 2.3 to 2.6 A. Closely related to STEM, scanning electron microscopy (SEM) givos an image of the surface (or near-surfaco) Figure 4. Z-contrast image of an incommencurate CéTe(111)/ 'Si(100) interface showing a 3.2-A spacing between film and sub strate indicating van der Waals bonding.