Micron 38 (2007) 390–401 www.elsevier.com/locate/micron A history of scanning electron microscopy developments: Towards ‘‘wet-STEM’’ imaging A. Bogner a,b,*, P.-H. Jouneau a,c, G. Thollet a, D. Basset b, C. Gauthier a a Groupe d’Etudes de Métallurgie Physique et de Physique des Matériaux, UMR CNRS 5510, INSA de Lyon, Bâtiment B. Pascal, 7 Avenue Jean Capelle, 69621 Villeurbanne Cedex, France b Total France, Centre de Recherche de Solaize, BP 22, 69360 Solaize Cedex, France c CEA Grenoble, DRFMC/SP2M, Laboratoire d’Etudes des Matériaux par Microscopie Avancée, 15 Rue des Martyrs, 38054 Grenoble, France Abstract A recently developed imaging mode called ‘‘wet-STEM’’ and new developments in environmental scanning electron microscopy (ESEM) allows the observation of nano-objects suspended in a liquid phase, with a few manometers resolution and a good signal to noise ratio. The idea behind this technique is simply to perform STEM-in-SEM, that is SEM in transmission mode, in an environmental SEM. The purpose of the present contribution is to highlight the main advances that contributed to development of the wet-STEM technique. Although simple in principle, the wet-STEM imaging mode would have been limited before high brightness electron sources became available, and needed some progresses and improvements in ESEM. This new technique extends the scope of SEM as a high-resolution microscope, relatively cheap and widely available imaging tool, for a wider variety of samples. # 2006 Elsevier Ltd. All rights reserved. Keywords: Electron microscopy; STEM-in-SEM; Transmission mode; Scattered electrons; Environmental scanning electron microscopy; ESEM 1. First steps in scanning electron microscopy In scanning electron microscopy (SEM), a fine probe of electrons with energies typically up to 40 keV is focused on a specimen, and scanned along a pattern of parallel lines. Various signals are generated as a result of the impact of the incident electrons, which are collected to form an image or to analyse the sample surface. These are mainly secondary electrons, with energies of a few tens of eV, high-energy electrons backscattered from the primary beam and characteristic X-rays. This section reviews the most important steps that have allowed using such rich physical interaction in a practical tool, making the SEM the powerful instrument it is today in materials and life-science. The history of electron microscopy began with the development of electron optics. In 1926, Busch studied the trajectories of charged particles in axially symmetric electric and magnetic fields, and showed that such fields could act as particle lenses, laying the foundations of geometrical electron optics (Oatley, 1982 and references therein). Nearly at the same * Corresponding author. Tel.: +33 4 72 43 61 30; fax: +33 4 72 43 85 28. E-mail address: Agnes.Bogner@insa-lyon.fr (A. Bogner). 0968-4328/$ – see front matter # 2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.micron.2006.06.008 time, the French physicist de Broglie introduced the concept of corpuscule waves. A frequency and hence a wavelength was associated with charged particles: wave electron optics began (Hawkes, 2004 and references therein). Following these two discoveries in electron optics, the idea of an electron microscope began to take shape. In 1931, independently of the ‘‘material wave’’ hypothesis put forward by de Broglie several years earlier (1925), Ruska and his research group in Berlin, were working on electron microscopy. They were disappointed learning that even with electrons a wavelength would limit the resolution. But they found using de Broglie equation that electron wavelengths were almost five orders of magnitude smaller than the wavelength of light used in optical microscopy. It was thus considered that electron microscopes still could prove a better resolution than light instruments, and no reason existed to abandon this aim. In 1932, Knoll and Ruska tried to estimate the resolution limit of the electron microscope. Assuming the resolution limit formula of the light microscope was still valid for material waves, they replaced the light wavelength by the electrons wavelength at an accelerating voltage of 75 kV. A theoretical limit of 0.22 nm resulted, a value experimentally reached only 40 years later. Although these calculations proved it was possible to reach a better-than-light-microscope resolution when working at high A. Bogner et al. / Micron 38 (2007) 390–401 magnifications (Ruska, 1986), many technical limits had to be to overcome. Ruska and Knoll tried to implement Busch’s lens formula experimentally. Their work resulted in the construction of the first transmission electron microscope (TEM) in 1931, with a magnification of 16 (Haguenau et al., 2003). Knoll built a first ‘‘scanning microscope’’ in 1935. However, as he was not using demagnifying lenses to produce a fine probe, the resolution limit was around 100 mm because of the diameter of the focused beam on the specimen. In 1938, von Ardenne clearly expressed the theoretical principles underlying the scanning microscope, as we know it today. Because it was difficult to compete with TEM in resolution achieved for thin specimens, the scanning microscopy development was oriented more toward observing the surface of samples. The first true SEM was described and developed in 1942 by Zworykin, who showed that secondary electrons provided topographic contrast by biasing the collector positively relative to the specimen. One of his main improvements was using an electron multiplier tube as a preamplifier of the secondary electrons emission current. He reached a resolution of 50 nm, which was still considered low in comparison with the performance obtainable in TEM (Goldstein et al., 2003). Indeed, many microscopists in the TEM community were seeing only limited applications for an instrument having a lower resolution than a TEM. Fortunately many scientists and technologists quickly recognised the SEM ability to yield threedimensional information from the surfaces of bulk specimens over a large range of length-scales. In 1948, Oatley began to build an SEM based on Zworykin’s microscope. Following this development, Smith (1956) shown that signal processing could be used to improve micrographs. He introduced nonlinear signal amplification, and improved the scanning system. Besides, he was also the first to insert a stigmator in the SEM to correct lens cylindrical imperfections. In 1960, Everhart and Thornley greatly improved the secondary electron detection. A new detector was created with a positively biased grid to collect electrons, a scintillator to convert them to light, and a light-pipe to transfer the light directly to a photomultiplier tube (Goldstein et al., 2003). In 1963, Pease and Nixon combined all of these improvements in one instrument: the SEM V with three magnetic lenses and an Everhart–Thornley detector (ETD). This was the prototype for the first commercial SEM, developed in 1965— the Cambridge Scientific Instruments Mark I ‘‘Stereoscan’’ (Breton, 1999). The SEM we are using today are not very different from this first instrument. 2. What makes the SEM a high-resolution technique? 2.1. Lens aberrations and source brightness limit the resolution in SEM In the SEM, electron optics is used to demagnify the size of the electron source, usually a small tungsten tip, to form the smallest possible probe. The demagnification is achieved using a series of ‘‘condenser lenses’’, and a final ‘‘objective lens’’ also known as the ‘‘probe-forming’’ lens. This last lens 391 provides the final demagnification and focuses the electron beam on the surface. At high magnification, the image resolution is roughly equal to the width of this probe. It is limited by a few key parameters: aberrations of the lenses, especially the objective lens because it works at large convergence angles, the brightness of the electron source (Hawkes, 2004), and the interaction volume, especially when the samples are not very thin. Aberrations, that are lens imperfections, limit the ability of focusing the beam, and therefore blur the image. The difficulty arises from two main aberrations. Due to spherical aberration, rays travelling far from the optical axis are focused more strongly than those close to the axis. To reduce this effect, an objective lens aperture is used to limit the angle of the outer rays through the lens. Chromatic aberration also blurs the image since electrons with slightly different wavelengths are focused more or less strongly. Electron beams with narrow energy distribution are used to limit this effect. Today, the energy spread of the beam in a microscope is typically less than 2 eV for a thermal source and less than 1 eV for field-emission sources. The second key parameter limiting the resolution is the conservation of brightness throughout the microscope column. It means it is impossible to decrease the size of the electron probe (by additional optical demagnification) without decreasing the current at the same time. When the electron flux becomes to low in the probe, the poor signal to noise ratio limits the resolution. Two main approaches exist to improve the SEM resolution: decrease the lens aberrations, or increase the source brightness. Optical aberrations may be reduced by improving lens design, and by developing electro-optical correcting devices. Much work is currently being carried out on this subject. But in recent years, most of the progresses in SEM resolution have resulted from the development of high brightness sources. The early thermionic sources, still in use in low cost instruments, are now superseded by field emission and Schottky electron guns, which exhibits a higher brightness and a better monochromaticity, as shown in Table 1. Their use leads to the emergence of the ‘‘high-resolution SEMs’’ in the 1980s (Joy, 1991). 2.2. Different technologies of electron sources in the electron microscope The first and most widespread electron source in electron microscopy was the thermionic gun. It consists of a tungsten filament bent into a V-shaped hairpin, with a tip radius of about 100 mm. This tip spontaneously emits thermionic electrons when electrically heated to a temperature around 2700 K. The thermionic electron source presents the advantages to be relatively inexpensive, and to require a relatively low vacuum. However, the lifetime of W-cathodes is limited to about 100 h by evaporation of the cathode material, which result in a failure when part of the wire becomes too thin. In thermionic LaB6 electron guns developed later, the emitting material has a lower work function, meaning the same amount of electrons can be 392 A. Bogner et al. / Micron 38 (2007) 390–401 Table 1 Comparison of different electron sources at 20 kV (Goldstein et al., 2003) Source Brightness (A/cm2 sr) Lifetime (h) Virtual source size Energy spread DE (eV) Beam current stability (%/h) Tungsten hairpin LaB6 Cold field emission Thermal field emission Schottky field emission 105 106 109 108 108 40–100 200–1000 >1000 >1000 >1000 30–100 mm 5–50 mm <5 nm <5 nm 15–30 nm 1–3 1–2 0.3 1 0.3–1.0 1 1 2 2 1 emitted at a lower heating temperature. The emitter is a tiny block of LaB6 single-crystal, about 100 mm in diameter and about 0.5 mm long, polished to present a 1 mm tip radius. Depending on the sharpness of the tip, this electron gun exhibits 5–10 times higher brightness and a 10 times longer lifetime than a tungsten filament. The second main technology of electron sources, originally developed in the 1930s, is based on field emission. A fieldemission gun is a wire with a sharp point, supported by a hairpin. The tip emits electrons by quantum mechanical tunnelling process when it is brought into close proximity with a positively biased extraction electrode. Two types of fieldemission sources exist: cold field emission (CFE), and thermal field emission (TFE), including Schottky field emission (SFE). In CFE, the high electric field causes electrons to tunnel through the potential barrier and leave the cathode wire at room temperature. Electrons are emitted from a virtual source with an area of only a few nanometers in diameter. TFE are similar in principle but operate at a higher temperature. This helps to keep a clean tip and to reduce noise and instability. Finally, in SFE, a thin ZrO2 layer is deposited on the flattened tip from a small dispenser to further reduce the work function. SFE is in fact a thermionic source but with a brightness comparable to CFE (cf. Table 1). The main benefit of these field-emission sources is a higher brightness, of the order of 109 A/cm2 sr at 20 keV. As previously said, brightness is a fundamental limit for the resolution in the SEM, as it determines the current available at a given probe size, and sets the recording time per pixel of the image. Despite this obvious advantage, the development of field-emission sources took time. Many technical barriers had to be overcome increasing the instrument cost. An additional delay may have also been caused by the great success of the early thermionic SEM instruments (Pawley, 1997). Besides the high brightness, fieldemission guns show another important advantage: the virtual source is small, usually in the 5–25 nm range. The optics required to demagnify this probe for high-resolution observations is therefore simplified in comparison with the case of thermionic emitters, where the effective source size is of the order of 5 mm. Lower demagnification contribute to reduce lens aberrations. Moreover, the FEG has a small energy spread, as low as 0.35 eV for a thermal or Schottky FEG, to compared with 1.5 eV or more for a thermionic emitter. While chromatic aberration does not really limit the resolution at high energies (>10 keV), it is a major factor of improvement for low-energy imaging (Joy, 1991). Today, thanks to these new field-emission sources, resolutions on the order of 1–2 nm are routinely achieved for commercial instruments operating between 1 and 30 keV. 3. The use of STEM mode in SEM 3.1. STEM mode The STEM concept was described by von Ardenne in the late 1930s: he was the first to perform a STEM mode experiment in 1938, by adding scan coils to a transmission electron microscope (Goldstein et al., 2003). However, the STEM did not develop at that time due to a lack of electronics and adequate electron sources. In 1960s, interest in STEM was revived by Crewe and coworkers with the development of the cold field-emission electron source and the optimization of electron-optical components, culminating in the first visualization of single heavy atoms in the electron microscope in 1971. In a dedicated STEM, the electron optics are designed to produce an atomic-size beam of electrons that illuminates a small area on the surface of the specimen. Images are formed by rastering the subnanometer probe over the surface and collecting electrons that were transmitted through the sample (unlike in SEM where secondary electrons are used). A STEM image may be considered as a collection of individual scattering experiments. The STEM can be considered a low dose technique, in comparison with fixed-beam TEM. Various types of signals discriminated in scattering angle and/or energy loss yield different structural and chemical information, and may be captured simultaneously. Quantitative analysis may be performed, with no limitation of the solid angle and the energy loss interval over which scattered electrons may be collected. In fact, nearly all electrons coming through the specimen can be collected by at least one of the detectors. Biological specimens are difficult to see in the transmission electron microscope because of their low contrast. A solution is to increase contrast by staining, i.e. heavy atoms addition, however a question arises whether the observed image features are actually artifacts. Dark-field STEM imaging is a direct approach that has been developed to circumvent the need for staining. It is for this reason, and for low dose techniques requirements, that STEM was deliberately aimed at biological applications (Colliex and Mory, 1994). Although STEM can be performed in a dedicated instrument specifically designed for the technique, it is more often developed as a hybrid technique performed on modified SEMs A. Bogner et al. / Micron 38 (2007) 390–401 or TEMs, microscopes more user friendly and widely available (Tracy and Alberi, 2004). A STEM system added to a standard SEM is often designated as ‘‘low-voltage STEM’’, term referring to the 20–30 kV regime, i.e. low relative to typical TEM operating energies. It is near the lower limit of energies that will provide sufficient transmission through the sample. As in a SEM, the beam focuses on a small spot that scans over the sample. The image is formed by mapping some signal intensity synchronously with the scan. As in a TEM, image information is extracted from electrons that have passed through a thin sample. Using the transmission mode in a scanning electron microscope, both contrast and resolution are improved due to the lower accelerating voltages, which increase the crosssections and reduce the interaction volume of the incident electron beam (Tracy and Alberi, 2004; Golla-Schindler, 2004). Indeed in transmission electron microscopy, high electron energies result in a scattering contrast produced by the narrow cone of transmitted electrons scattered through small angles and limited by the objective aperture. For contrast enhancement, the electron energy must be reduced and/or in the case of low atomic number elements the sections stained with heavy metal compounds. However, low-voltage fixedbeam TEM is limited by the strong decrease in transmission with decreasing electron energy and by chromatic aberrations from more frequent energy losses. Low-voltage scanning transmission electron microscopy can be an interesting alternative. The transmission mode in SEM has an advantage of avoiding chromatic aberration. As there is no projection lens, no image deterioration occurs due to chromatic aberrations, even in the case of inelastic interactions at low voltages. As already suggested by von Ardenne, the potential advantage of transmission-SEM indeed lays in the fact that electrons are not required to pass through a lens after crossing the specimen. The spread of velocities caused by absorption in 393 the specimen therefore do not give rise to chromatic aberration induced loss of resolution, as it does in a conventional transmission microscope: it is thus possible to examine a thicker specimen (Oatley, 1982). It is an argument also cited by Merli et al. (2004): due to the absence of image-forming lenses, the energy loss and the large scattering angles do not affect the resolution as in transmission electron microscopy. In low-energy STEM, the specimen response will be proportional to the average scattering power of atoms which will give rise to mass–thickness contrast, or absorption contrast. For a good Z-contrast, it is necessary to use large scattering angles in transmission. In the transmission mode of SEM, the aperture can be increased to obtain a high transmission and signal-to-noise ratio. SEM remains one of the most flexible tools for highresolution imaging (resolution down to 1 nm), however 1 nm resolution is sometimes not achieved when characterizing ordinary samples, due to limitations arising from the nature of beam–sample interactions when secondary electrons signal is used. STEM-in-SEM is one method that can overcome these limitations and enable imaging with high resolution (Vanderlinde, 2005). As the samples investigated are thin in comparison with SEM, the interaction volume is rather small. For this reason, in STEM mode, the resolution is only limited by the spatial broadening of the electron probe at the exit surface of the sample, whereas structures at the entrance surface are resolved with a resolution of approximately same as the diameter of the electron probe. This beam broadening phenomenon is called the ‘‘top–bottom effect’’ (Golla-Schindler, 2004). For illustration, Fig. 1 shows the investigation of small indium cluster on a formvar film with a polystyrene sphere placed on the indium layer. This test structure can be investigated with the indium cluster above the polystyrene sphere (top object) or below (bottom object) in STEM-in-SEM, mode called ‘‘transmissionSEM’’ by the authors (Fig. 1a and b) or in TEM (Fig. 1c and d). Fig. 1. Illustration of the top–bottom effect according to Golla-Schindler et al. (Golla-Schindler, 2004). 394 A. Bogner et al. / Micron 38 (2007) 390–401 Improved sharpness of the indium cluster can be obtained by using the top object and the transmission-SEM mode (Fig. 1a). Due to the parallel illumination in TEM the best object is the bottom object, but the resolution of the indium cluster is limited by the chromatic aberration of the post specimen objective lens (Fig. 1c). Therefore, for TEM investigations thinner sections and increased accelerating voltages are necessary to improve the attainable resolution (Fig. 1d). Today, every microscope manufacturer offers a STEM detector for SEM. In addition with the success related to SEM ease of use and availability when compared to TEM, it is indeed recognised that STEM-in-SEM extends the usefulness of the SEM (Woolf et al., 1972). On one hand, thicker samples can be observed thanks to large collection angles, and high contrast is results from the use of low voltages. On the other hand, high resolution is available thanks to limited interaction volume and chromatic aberration. 3.2. STEM-in-SEM detection methods In the present article, we are interested in the transmission mode performed in SEM. Several detection strategies for transmitted 10–30 keV electrons exist, using several types of STEM detectors. 3.2.1. Transmitted electrons/secondary electrons converter and Everhart–Thornley detector (Golla et al., 1994; GollaSchindler, 2004; Vanderlinde, 2005) Various devices have been developed using the following idea, that does not require the development of a new detector: transmitted electrons (TE) are converted to secondary electrons (SE) using a plate below the sample: the converter. SEs can then be detected by using a conventional Everhart–Thornley detector (ETD). This is one of the configurations exhibiting the advantage of TV-rate image. This principle is applied in the configuration used by Golla et al. presented in Fig. 2. In their device, a cylindrical shield stops the ‘‘reflected’’ secondary electrons, so that only Fig. 2. Scheme of a STEM-in-SEM system: TE/SE converter and ETD configuration (Golla-Schindler, 2004). transmitted or scattered electrons can be collected by the ETD. An electron trap also prevents large angle scattered electrons from reaching the ETD. An adjustable aperture determines the collection angle for the bright-field mode or the angle segment for the dark-field mode. 3.2.2. Direct collection of transmitted electrons by a solidstate detector Transmitted electrons can also be directly collected by a semiconductor detector. In this case, electrons reaching the detector are physically detected as they create electron–hole pairs in the silicon diode (2700 pairs by an 10 keV electron incident on the detector), resulting electrical charge is collected from the biased p–n junction of the detector. The configuration described in Fig. 3 has been used by Merli et al. Transmitted electrons are collected using the detector normally used for backscattered electrons (BSE) but placed just below the sample (Merli and Morandi, 2005). There are several positions for the detector: (1) by placing the center of the detector on the optical axis and varying the specimen-detector distance, it is possible to select the angular distribution for darkfield imaging (Fig. 3a); (2) by removing the center of the annular detector from the optical axis and defining with an aperture a portion of the detector itself, it is possible to define the angular range of transmitted electrons producing the brightfield image (Fig. 3b). It should be noted that in the case of a dipolar detector, two diodes are placed on the optical axis. In this way, bright- or darkfield images can be acquired through the collection of transmitted Fig. 3. Merli et al. STEM-in-SEM detection conditions: (a) BSE annular detector centered on the optical axis for dark-field imaging; (b) off-axis annular detector with an aperture on the cover to define the collection angle a for brightfield imaging. Scheme courtesy of Merli et al. (Merli and Morandi, 2005) and Microscopy and Microanalysis from Cambridge University Press. A. Bogner et al. / Micron 38 (2007) 390–401 Fig. 4. Scheme of Woolf et al. transmission stage for the SEM, including a scintillator. Scheme courtesy of David Joy (Woolf et al., 1972) and IOP Publishing. or scattered electrons, respectively. This setup is implemented for example in the FEI high vacuum STEM holder. 3.2.3. Transmitted electrons/photons conversion via a scintillator STEM-in-SEM has become an established technique because it extends the capabilities of the SEM and offers considerable advantages in signal processing and analysis when compared to the TEM (Woolf et al., 1972). The detection strategy of Woolf et al. in their ‘‘transmission stage for the scanning electron microscope’’ is presented in Fig. 4. It used a scintillator/light-pipe combination. After their path through the sample, electrons are converted to photons by a scintillator and pass through a lightguide. The signal is amplified afterwards by a photomultiplier. This is a relatively expensive standard detector system, but it allows TV-rate imaging, and the selection of diffracted beams. Each of the three precedent detection strategies presents advantages, but practically a type of detector is often adopted because of financial or availability reasons. 4. Environmental SEM development The origin of environmental SEM is directly linked to the high vacuum needed in electron microscopes, that introduce restrictions on the way that certain specimens are prepared and imaged. Very early in the history of electron microscopy, studies were related to the possibilities of imaging specimens in a more ‘‘natural’’ state. In the 1950s, experiments concerned differentially pumped, aperture-limited TEMs, or creation of ‘‘environmental chambers’’. The separation of high vacuum electron gun chamber from a gaseous specimen chamber via open diaphragms (PLA for pressure limiting apertures) was first used in several high-voltage TEM. Research by Danilatos and Robinson in the 1970s led to the first SEM capable of maintaining a relatively high pressure, removing the need to dry and coat the specimens (Danilatos, 1991). The term ‘‘environmental’’ SEM was introduced in 1980. As mentioned in Section 1, during the first development steps of SEM, the established TEM community was very suspicious. In the same way, the SEM community was not convinced when the first environmental SEMs appeared (Stokes, 2003). The introduction of gas in an electron microscope is completely opposed to 395 the generally established ideas of a clean high vacuum. It was generally perceived that a gaseous environment would deteriorate the resolution by an assumed broadening of the electron beam and exclude the SE detection mode of imaging because highly positive bias of the ETD induces arcing in the presence of gas. To convince the microscopy community, these two main problems – i.e. reduced resolution, and exclusion of SE detection – had to be resolved with the: (1) determination that the electron probe diameter remained small at elevated pressure; (2) invention of the gaseous detection device (GDD), where the principle of gaseous ionization produced by the signal-gas interaction is used for imaging; this GDD was called GSED (for gaseous secondary electron detector) to insist on the fact that it is possible to detect secondary electrons in ESEM (Danilatos, 1991). Over more than two decades, Danilatos showed tenacity and developed the ESEM instrument to a point where the rest of the scientific community and manufacturing world could further benefit. ESEM became a stable instrument. By the late 1980s, the first commercial environmental scanning electron microscopes (ESEMs by Electroscan, a company created for the purpose of gaseous microscope manufacturing) were being produced, opening up a world of possibilities for observing untreated specimens. A few years later in 1996, a traditional electron microscope manufacturer Philips FEI took over the operation to further commercialize the instrument. Other manufacturers have in the meantime followed with low-vacuum systems reaching 100 Pa, the level of pre-ESEM technology, probably due to patent restrictions (ESEM Research Laboratory, Australia). The ESEM has rapidly gained acceptance by the scientific, technical and industrial community as shown by the large number of publications arising from its use. To summarize, ESEM differs from conventional SEM mainly by the presence of a gas in the specimen chamber. Samples are thus not viewed under high vacuum but under a deteriorated or ‘‘low’’ vacuum. This is possible thanks to a special design of electron optics column that allows differential pumping: the column is divided into different pressure zones separated by pressure limiting apertures (Danilatos, 1993). The presence of gas, or environment around the sample that inspired the term ‘‘environmental’’ SEM, can play two main roles. The first, common for environmental as well as low-vacuum SEMs,1 is electronic. The gas acts as an electrical charge conductor avoiding sample charging and facilitates signal detection. This first role is described in Fig. 5: collisions between electrons and gas molecules create positive ions that can balance the accumulation of negative charges on the surface of specimens, and ‘‘cascade’’ electrons that help to amplify the signal collected by the gaseous secondar electrons detector (GSED). The second role, more specific to environmental SEMs, is thermodynamic, i.e. the gas is a conditioning medium, preventing evaporation of liquids from a sample (Thiel and Toth, 2005). 1 ESEM is a trademark from FEI Company, but other constructors have developed their low-vacuum SEM too. 396 A. Bogner et al. / Micron 38 (2007) 390–401 Fig. 5. A schematic of the amplification process due to collisions between the secondary electrons and gaseous molecules. Image courtesy of D.J. Stokes (Stokes, 2003) and the Royal Society of U.K. With this specialized form of electron microscopy a wide range of insulating materials, oils, liquids, etc., can be observed at equilibrium or during in situ evolutions. The potential applications are described in numerous recently published articles, in particular by Donald (2003). In this paper, we discuss liquid specimens, most of the time aqueous. Water vapour is the most common gas used in ESEM, due to its efficiency in signal amplification, and its useful thermodynamic properties leading to good quality images and specimen stability. Fig. 6 is an example concerning the amplification signal of water, compared with four other gases: carbone dioxide, nitrous oxide, nitrogen, helium (Fletcher et al., 1997). The authors conclude that water vapour gives quantitatively the largest maximum amplification of all the cases considered in their study (variable accelerating voltage and partial pressure). It should also be noted that the maximum of these amplification curves is around 4–5 Torr for water. Concerning aqueous specimen stability, it is important to consider the thermodynamic saturated vapour pressure (SVP) curve for water as reproduced in Fig. 7 (Stokes, 2003). It can be seen that the pressure range up to around 10 Torr meets the SVP when the specimen temperature is lowered to a few degrees Celsius from ambient. Typically, a water-cooled Peltier stage is employed to keep the specimen at around 3 8C, within a chamber environment of water vapour at a pressure in the region of 4.5 Torr, leading to improved Fig. 6. Dependences of the amplification on the pressure for various primary-beam energies and for five gases: (a) water vapour, (b) carbon dioxide, (c) nitrous oxide, (d) nitrogen, and (e) helium. Figures courtesy of Bradley Thiel (Fletcher et al., 1997) and IOP Publishing. A. Bogner et al. / Micron 38 (2007) 390–401 397 Fig. 7. Saturated vapour pressure curve for water. Image courtesy of D.J. Stokes (Stokes, 2001) and Wiley. specimen stability and good quality images. It is also essential to ensure that the specimens’ natural moisture is preserved during the initial pumpdown of the chamber. It is therefore usual to perform a sequential pumpdown such that the air in the chamber is successively replaced with water vapour, and that evaporation from and condensation onto the sample are minimized (Cameron and Donald, 1994). ESEM also extends the possibilities of SEM in term of the wide variety of samples and of states that can be characterized. However, in the case of liquids, classical wet mode in ESEM only allows the observation of the surface, which is often very smooth so that very little information about the sample can be collected. For this purpose, the wet-STEM imaging mode described hereafter may be suitable for aqueous samples, or other liquids. 5. Wet-STEM imaging: observing a thin liquid film in transmission mode It actually refers to the STEM-in-SEM applied to environmental SEM, benefiting from the improved FE SEM performance in STEM mode and low-vacuum techniques. 5.1. History Bultreys and Thollet were the pioneers of the ‘‘wetSTEM’’ imaging mode a few years ago. During one of their experimental collaborations in the GEMPPM laboratory in Lyon, they placed a TEM grid on a Peltier stage of an FEI XL 30 FEG ESEM, and relocated the dipolar detector usually used for backscattered electrons detection below the grid: the same configuration used in the high vacuum STEM detector, already commercially available in FEI XL series SEMs, but adapted for the use on a cooling stage. A series of investigations were conducted by Bogner et al., examining the experimental procedure and signal detection conditions as well as the influence of gas purge, type of grids, sample-detector distance, annular dark-field detection (Bogner et al., 2005). The wet-STEM instrument in our laboratory is described as follows. Fig. 8. The wet-STEM device at the GEMPPM laboratory: (a) Peltier stage; (i) incident convergent electron beam; (b) SEM mount maintaining the TEM grid; (c) solid-state annular detector. 5.2. Experimental setup of wet-STEM As the presented imaging mode allows the observation of wet samples in transmission mode, the term ‘‘wet-STEM’’ is self-explanatory. The device, described in Fig. 8, is placed in an FEI XL 30 FEG ESEM. A copper grid is placed on a TEM sample holder, and positioned on a Peltier cooling stage. A small amount of liquid containing particles or floating objects (organic, inorganic, liquid or solid) is dropped on the grid with a micropipette. As in the conventional wet mode of ESEM and in relation with the work described in Cameron and Donald (1994), we use an optimized pump down sequence in order to prevent evaporation from, and condensation onto the sample droplet. Classical ESEM detectors are also available enabling to control of the sample surface in SE mode and in BSE mode, using the gaseous SE detector (GSED), and the gaseous backscattered electron detector (GAD), respectively. This is very helpful for example to control the presence of liquid until the thickness is adequate for transmission imaging to perform transmission observations and to detect whether objects are submerged in water. When the required partial pressure of water is reached, pressure and temperature can be adjusted to evaporate a small amount of water – if the considered sample is aqueous – from the droplet. It allows obtaining a water layer thin enough such that the incident electrons can pass through it, and can be collected to form a STEM image. Films of wet samples are thinned in situ in the ESEM chamber, their thickness depends on the quantity of water evaporated from the initial droplet. As evaporation is an endothermic reaction, it is then possible to follow it by checking the difference between the setting temperature and the measured one. Then, the thickness of the film is kept constant thanks to an equilibrium water pressure 398 A. Bogner et al. / Micron 38 (2007) 390–401 using the (P, T) water diagram (presented in the precedent section). For instance, a water pressure of 5.3 Torr is required at a sample temperature of 2 8C, so that objects remain in a water layer with constant thickness. By controlling the sample temperature through the Peltier stage, and using water vapour as the imaging gas at a controlled pressure, samples can be kept above their saturated vapour pressure during all the experiment. When the required thickness is obtained, the incident electron beam is focused on the droplet, and passed through the liquid layer and floating nano-objects. The signal is then collected by a detector, usually used for the collection of backscattered electrons, but in the present configuration located below the sample. Holey carbon coated TEM copper grids placed with the carbon layer faced pointing down enable copper squares to play the role of retention basins. In the carbon layer, holes of typical diameter ranging from less than 1 to 20 mm allow maintaining overhanging liquid films on very small areas. It is important to note that adequate initial droplet parameters – i.e. volume and solid content – enable to control the amount of nano-objects on the grid. It is possible, for instance in the case of a latex emulsion sample, to choose to image only a monolayer of particles if required. Due to the initial concentration of the suspension, when the droplet of smallest volume that can be dropped contains more objects than for a monolayer on the grid, a dilution step with distilled water can be performed before placing the liquid on the grid. As discussed above, several detection strategies can be implemented for ‘‘transmitted’’ electrons detection in an SEM. The present device is based on the direct collection of electrons passed through the sample by a solid-state detector (two semiannular detectors A and B). For the setup usually used in STEM detectors for FEI SEMs, the incident electron beam arrives on the border of diodes A and B, and bright- or dark-field images can be produced with the collection of transmitted or scattered electrons, respectively. In the latter case, only a small area of the sample is above the border of diodes A and B; so that we do not have the choice of the imaging mode (bright- or dark-field) for the other areas. In the present study, we choose another possibility: annular dark-field imaging conditions can be obtained if the dipolar detector is placed on the optical axis so that the direct transmitted electron beam is not collected and only scattered beams are detected on a ring constituted by both of the diodes A and B. Using this method, a more important part of the scattered electrons available is used to form an image and higher-contrast images can be obtained. Consequently, all the ESEM micrographs presented in the following section have been obtained using the annular dark-field configuration. The second advantage of this configuration, using the summed signal, is that imaging conditions are not linked to the area of the sample imaged. Moreover, in our experimental setup, the distance between the sample and the detector has been experimentally investigated in order to optimize the contrast. Best results have been obtained with a distance of about 7 mm, corresponding to collection angles between 349 and 820 mrad for the dark-field mode. Nevertheless, the optimal distance between the grid and the detector depends on the sample composition and thickness, and should be a variable parameter. 5.3. Benefits of STEM-in-SEM mode By using low voltages, which lead to improved contrasts, and thanks to the absence of a projection lens when compared to TEM, thicker samples can be imaged using the present STEM mode. For example, an estimation of the thickness of water that can be passed through with an incident electron voltage of 30 kV is presented in Fig. 9. It has been estimated using an evoluted Monte Carlo Model applied by SAMx1 in their Hurricane software. This powerful tool has been specially adapted so that it is possible to consider transmitted electrons and to store them as a function of their energy and their scattering angle. Here, transmission is defined as the ratio between the number of electrons scattered with angles from 349 to 820 mrad and the number of the incident electrons. It can be shown that transmission occurs in as much as several microns of water. Although it is not within the scope of the present article, it should be pointed out that this special extension of Hurricane is also very helpful to progress in wet-STEM images contrast interpretation. Theoretically, the resolution of wet-STEM mode should be better than seen in ESEM wet mode, due to the properties of the thin sample (STEM-in-SEM a high-resolution mode in SEM because there is no interaction volume, only top–bottom effect: see Section 3), and that contrary to TEM, electrons do not have to pass through a lens after their path through the sample (see Section 3 also). Unfortunately in both cases (wet mode in classical reflection ESEM, and wet-STEM), we have observed that the resolution is limited by the mechanical vibrations partially due to the cooling flow of the Peltier stage and to the vacuum system. We are limited to relatively high beam energies (25–30 kV) when compared to the range typically used in SEM (a few eV to 30 kV). Indeed, in ‘‘environmental’’ or ‘‘low-vacuum’’ SEMs, the electrons leaving the specimen are significantly retarded by their collisions with gas molecules reducing their energies. The use of a TE/photons conversion via a scintillator, as proposed by KE company in the Centaurus detector, could be used to explore other imaging conditions in wet-STEM, Fig. 9. Evaluation of the scattering transmission through a water layer at 30 kV with Hurricane (SAMx1) Monte Carlo simulation: scattered electrons are collected from 349 to 820 mrad. A. Bogner et al. / Micron 38 (2007) 390–401 Fig. 10. Au particles suspended in water, and imaged in the wet-STEM annular dark-field mode. such as using lower. This detector is very sensitive at low energies (0.5 keV) and offers the advantage of TV-rate imaging. As a feature of STEM-in-SEM imaging mode, wet-STEM mode improves contrast and resolution. This technique also improves volume information in comparison with classical wet mode in ESEM, where only the surface of liquids can be observed. 6. Applications of wet-STEM imaging Using the wet-STEM imaging mode described previously, a wide variety of nm- to mm-scale objects suspended in a liquid layer (not only water) can be investigated (Bogner et al., 2005). The present imaging conditions correspond to annular darkfield mode, using very large collection angles. An acceleration voltage of 30 kV has been chosen to optimize resolution and contrast. An image of aqueous suspension of gold nanoparticles is presented in Fig. 10. This image highlights the high resolution of wet-STEM imaging: particles with diameter of 20 nm are 399 well resolved; the resolution is estimated as low as 5 nm. Au particles exhibit a very high bright contrast, as expected when considering their high atomic number. Even objects constituted of lighter elements induce an observable contrast in wet-STEM using annular dark-field detection. Fig. 11 presents images of colloidal clay called ‘‘chocolate’’ referring to the colour of the aqueous solution. Clay platelets typical sizes are 0.1 mm 0.5 mm 2 mm. These objects can be observed using a cryo-TEM, but exhibit little contrast when they are disposed as to be imaged through their thinner thickness. Wet-STEM imaging mode corresponds to a scattering contrast: whether the flattest face of a platelet is parallel or perpendicular to the grid, the scattering of electrons will occur in a different local thickness. In wetSTEM, perpendicular platelets show brighter contrast than others because of the important local thickness acting for the scattering of electrons, and some lamellar structures are observed; even ‘‘parallel’’ platelets exhibit a good contrast. Contrast interpretation is not always trivial, as suspensions often contain several additives, that may modify the scattering of electrons, or the sample sensitivity to the electron beam. Fig. 12 presents carbon nanotubes in two different solutions: (1) water with two different concentrations of sodiumdodecylsulfate (SDS) surfactant; (2) pure ethanol. Micrometric holes present in the carbon layer of the grid are observed as dark regions. In Fig. 12a, carbon nanotubes are well distinguished, and some brighter contrasts correspond to superpositions, or heavy atomic number particles resulting from the carbon nanotubes synthesis. In Fig. 12b, carbon nanotubes are also well distinguished but the presence of water introduces a fuzziness, and some bright nm-scale features are present. The latter objects are understood as surfactant clusters, and are found to be larger and more numerous in a solution containing more surfactant: Fig. 12c. Biological specimens, typically low in atomic number and sensitive to beam damage and dehydration, have also been successfully observed in wet-STEM. Fig. 13 presents an image of bacteria of the species Pseudomonas syringae, acquired at a magnification of 50 000. The contrast of the bacteria is very Fig. 11. ‘‘Chocolate’’ colloidal clay in water imaged in wet-STEM annular dark-field imaging mode. 400 A. Bogner et al. / Micron 38 (2007) 390–401 Fig. 12. Carbon nanotubes: (a) dispersed in ethanol without surfactant; (b) in water with surfactant at the concentration C1; (c) in water with surfactant at the concentration C2 > C1. good without preliminary staining. Moreover, volumic information is obtained as double membrane structure is distinguished. Others types of samples have been imaged in wet-STEM (Bogner et al., 2005) such as mini-emulsions, latices, bacteriophages, particles in oil, etc. In fact, the term ‘‘wet’’ is restrictive in comparison with the effective imaging possibilities of the wet-STEM imaging mode: provided that the liquid is compatible with the microscope, a thin layer of nonaqueous liquids can also be studied, its stability only depending on its saturated vapour pressure. 7. Summary and outlook Fig. 13. Bacteria Pseudomonas syringae, imaged in wet-STEM annular darkfield imaging mode. The history of electron microscopy presented in this article highlights the extent of SEM applications. SEM is not in competition with TEM as it allows different imaging modes. Wet-STEM, i.e. STEM-in-SEM performed in environmental SEM, has been presented as an powerful imaging technique developed thanks to general progress in electron microscopy. It allows straightforward transmission observations of wet samples constituted of nano-scale objects in a liquid layer. With the benefits of field emission and STEM mode applied in SEM, an excellent resolution can be achieved. For example 5 nm was achieved on a resolution test sample of gold nanoparticles in colloidal suspensions. Thanks to the low operating voltage of an SEM and large scattering angle collection, the contrast is enhanced, this result is especially A. Bogner et al. / Micron 38 (2007) 390–401 interesting for low atomic number materials. Low-vacuum technology allows imaging samples in their native state. The scanning mode allows to image samples at a ‘‘low dose’’, that is important when examining polymers and biological samples. Finally, the transmission mode gives access to volume information since that we do not image only the surface of the liquid drop. Particularly adapted for suspension-type samples, and others delicate objects with nanometric features, wet-STEM allows characterizations in materials science as well as life-science at the nano-scale. Acknowledgments We are very grateful to D. Bultreys from FEI Company (Brussels) for shared experimental sessions and discussions. We also would like to acknowledge authors for figures reprint permissions. References Bogner, A., Thollet, G., Basset, D., Jouneau, P.-H., Gauthier, C., 2005. Wet STEM: a new development in environmental SEM for imaging nano-objects included in a liquid phase. Ultramicroscopy 104, 290–301. Breton, P.J., 1999. From microns to nanometers: early landmarks in the science of scanning electron microscope imaging. Scanning Microsc. 13-1, 1–6. Cameron, R.E., Donald, A.M., 1994. Minimizing sample evaporation in the environmental scanning electron microscope. J. Microsc. 173, 227–237. Colliex, C., Mory, C., 1994. Scanning transmission electron microscopy of biological structures. Biol. Cell 80, 175–180. Danilatos, G.D., 1991. Review and outline of environmental SEM at present. J. Microsc. 162, 391–402. Danilatos, G.D., 1993. Introduction to the ESEM instrument. Microsc. Res. Tech. 25, 354–361. Donald, A.M., 2003. The use of environmental scanning electron microscopy for imaging wet and insulating materials. Nat. Mater. 2, 511–516. Fletcher, A.L., Thiel, B.L., Donald, A.M., 1997. Amplification measurements of potential imaging gases in environmental SEM. J. Phys. D 30, 2249–2257. 401 Goldstein, J., et al., 2003. Scanning Electron Microscopy and X-ray Microanalysis, 3rd ed. Plenum Press, New York. Golla, U., Schindler, B., Reimer, L., 1994. Contrast in the transmission mode of a low-voltage scanning electron microscope. J. Microsc. 173-3, 219– 225. Golla-Schindler, U., 2004. STEM-Unit measurements in a scanning electron microscope. In: Proceedings of the European Microscopy Congress in Antwerpen. Haguenau, F., Hawkes, P.W., Hutchison, J.L., Satiat-Jeunemaı̂tre, B., Simon, G.T., Williams, D.B., 2003. Key events in the history of electron microscopy. Microsc. Microanal. 9, 96–138. Hawkes, P., 2004. Recent advances in electron optics and electron microscopy. Ann. Fondation Louis de Broglie 29-1 . Joy, D.C., 1991. The theory and practice of high-resolution scanning electron microscopy. Ultramicroscopy 37, 216–233. Merli, P.G., Morandi, V., Corticelli, F., 2004. Low energy scanning transmission electron microscopy: influence of the electron detection mode. In: Proceedings of the European Microscopy Congress in Antwerpen. Merli, P.G., Morandi, V., 2005. Low-Energy STEM of multilayers and dopant profiles. Microsc. Microanal. 11, 97–104. Oatley, C.W., 1982. The early history of the scanning electron microscope. J. Appl. Phys. 53 (2). Pawley, J., 1997. The development of field-emission scanning electron microscopy for imaging biological surfaces. Scanning 19, 324–336. Ruska, E., 1986. The development of the electron microscope and of electron microscopy. Nobel Lecture. Stokes, D.J., 2001. Characterisation of soft condensed matter and delicate materials using environmental scanning electron microscopy (ESEM). Adv. Eng. Mater. 3-3, 126–130. Stokes, D.J., 2003. Recent advances in electron imaging, image interpretation and applications: environmental scanning electron microscopy. Phil. Trans. R. Soc. Lond. A 361, 2771–2787. Thiel, B.L., Toth, M., 2005. Secondary electron contrast in low-vacuum/ environmental scanning electron microscopy of dielectrics. J. Appl. Phys. 97 (051101), 1–18. Tracy, B., Alberi, K., 2004. Adopting low-voltage STEM and automated sample prep to perform IC failure analysis. Micromagazine 87–93. Vanderlinde, W.E., November 2005. Ultra-high resolution in the scanning electron microscope (SEM). In: ISTFA Proceedings. Woolf, R.J., Joy, D.C., Tansley, D.W., 1972. A transmission stage for the scanning electron microscope. J. Phys. E 5, 230–233.