© R.F. Egerton

The scanning electron microscope (SEM) was invented soon after the TEM but took longer to be developed into a practical tool. Nowadays, SEM's number TEM's and are used extensively in medical and materials research, in the semiconductor industry, in forensic-science laboratories, etc.

Fig. 1. Hitachi S-4000 scanning electron microscope, which uses a field-emission source and achieves an image resolution of 1.5 nm.

---------------------------------------

Figure 1 shows a modern high-resolution SEM. Although smaller (and less expensive) than a TEM, the SEM makes use of many of the same operating principles as the TEM. The electron source may be a tungsten filament, a LaB₆ emitter or (increasingly often) a tungsten field-emission tip. The electron gun and the lens column are smaller than for a TEM, mainly because the maximum accelerating voltage is lower (typically 40 kV). Axially-symmetric magnetic lenses are used, as in the TEM, but there are fewer of them: in fact, none below the specimen plane. There are typically two or three lenses above the specimen, which can be likened to the condenser-lens system of the TEM. But whereas the TEM normally uses illumination of diameter ~ 1 m m or greater, the diameter of the incident beam (known as the electron probe) in an SEM is made as small as possible: 10 nm is typical and 1 nm is possible with a field-emission source. The final lens which forms this very small probe is called the objective lens, since its performance (including aberrations) largely determines the image resolution, just as in the case of the objective in a TEM or a light-optical microscope. In fact, the spatial resolution of the SEM can never be better than the incident-beam diameter, a consequence of the method which is used to obtain the image.

Fig. 2. Schematic diagram of a scanning electron microscope.

---------------------------------------

Whereas the TEM uses a stationary incident beam, the electron probe of an SEM is scanned across the specimen in two perpendicular (x and y) directions. The x-scan is relatively fast, and is generated by feeding the current output of a sawtooth-wave generator (at a line frequency f_x) to beam-scanning coils located just above the objective lens; see Fig. 2. The y-scan is much slower, and is generated by a second sawtooth-wave generator running at a frame frequency f_y=f_x/n where n is an integer. This procedure is known as raster scanning and results in the beam covering a rectangular area on the specimen (Fig. 3c). As a result of the x-deflection signal, the electron probe moves in a straight line, from A to B in Fig. 3c, which constitutes a single line of the raster. After reaching B, the beam is deflected back along the x-axis as quickly as possible (during the flyback portion of the x-waveform). But because the y-scan generator has increased its output during the line scan, it returns not to A but to point C, slightly displaced in the y-direction. A second line scan takes the probe to point D, at which point it flies back to E and the process is repeated until n lines have been scanned and the beam arrives at point Z. This entire sequence constitutes a single frame of the raster scan. From Z, the probe quickly returns to A, as a result of the rapid flyback or both the line and frame generators, and the next frame is executed. This process may run continuously for many frames.

Fig. 3. (a) Line-scan waveform (deflection current versus time), (b) frame-scan waveform, and (c) its digital equivalent. (d) Elements of a single-frame raster scan.

---------------------------------------

The output of the two scan generators is also applied to the deflection coils of a display device, typically a TV-type cathode-ray tube (CRT) on which the SEM image will appear. Since the electron beams in the SEM column and in the CRT are scanning in synchronism, for every point on the specimen (within the raster-scanned area) there is a corresponding point on the display screen, in conformity with Maxwell's first rule for image formation. In order to introduce contrast into this image, a voltage signal is applied to the electron gun of the CRT (to vary the brightness of the scanning spot). This voltage is derived from a detector which responds to some change in the specimen, induced by the SEM incident probe, so the image records some property of the specimen in response to electron bombardment.

Image magnification (M) is achieved by ensuring that the x- and y- scan distances on the specimen are a small fraction of the image size, since:

M = (scan distance in the image) / (scan distance on the specimen)

Since it is convenient to make the image a fixed size (just filling the display screen), increasing the magnification involves reducing the x- and y- scan currents (each in the same proportion, so as to avoid image distortion).

In a modern SEM, the scan signals are generated digitally (by computer-controlled circuitry) and the x- and y- scan waveforms are staircase functions with m and n levels respectively; see Fig. 3c. This procedure divides the image into a total of mn picture elements (pixels). The SEM probe and image-display spots remain stationary for a certain dwell time before jumping to the next pixel. One advantage of digital scanning is that the SEM computer (which generates the scan) "knows" the (x,y) location of each pixel and can store the intensity value (also digitized as a number) alongside this information. The digital image, in the form of position and intensity information, can therefore be stored in computer memory or on a magnetic or optical disk, and can be transmitted over data lines (e.g. the Internet) over short or long distances. Older SEM's have a second CRT on which the image is recorded photographically, usually on instant (Polaroid) film.

The scanning is sometimes done at video rate (about 60 frames/second) and generates a "live" TV-type image that is useful for focussing the specimen and for viewing it at low magnification. At higher magnification or to permanently record an image, slow scanning (10 - 100 seconds per frame) is usually preferred; the additional recording time results in a higher-quality image containing less electronic noise.

The signal which modulates the image brightness can be derived from any property of the specimen which is caused by (or changes in response to) electron bombardment. Most commonly, the emission of secondary electrons (atomic electrons ejected from the specimen as a result of inelastic scattering) is employed. However, a signal derived from the backscattered electrons (incident electrons which are elastically scattered through more than 90 degrees) is also useful. In order to understand these (and other) possibilities, we need to discuss what happens when an electron beam enters a thick specimen.

When 30keV (or lower-energy) primary electrons first enter the SEM specimen, they are scattered elastically (by Coulomb interaction with atomic nuclei) and inelastically (by interaction with atomic electrons), just as in the case of a TEM specimen. Most of this scattering is forward scattering, involving deflection angles less than 90 degrees. But a small fraction of the primaries are elastically backscattered (q > 90 deg) with only a small loss of energy; these electrons have a high probability of re-entering the vacuum above the specimen, in which case they may be collected as a backscattered signal. Electron scattering being a statistical process, it can be simulated in a computer by running a Monte-Carlo program (containing a random-number generator and information about the angular distibutions of scattering), as illustrated in Fig. 4.

Fig. 4. Monte-Carlo calculations of electron trajectories (Curgenven and Duncumb, 1971). (a) Effect of the electron accelerating voltage, for a copper specimen.

2. Results for Al (Z=13) and Au (Z=79) with 20kV accelerating voltage.

---------------------------------------

Although inelastic scattering does not contribute to the backscattered signal, it reduces the kinetic energy of the primary electrons until they are eventually brought almost to rest and are absorbed into the solid (in a metal specimen they would become one of the conduction electrons). The depth (below the surface) at which this occurs is called the penetration depth or the electron range. The volume of sample containing (most of) the scattered electrons is called the interaction volume, and is usually represented as pear-shaped in cross-section (Fig. 5) since scattering causes the beam to spread laterally as the electrons continue to penetrate and slow down.

The higher the primary-electron energy, the more inelastic collisions it takes to slow them to rest and greater the penetration depth and lateral spreading (Fig. 4a). Also, the inelastic-scattering probability decreases with increasing atomic number (as discussed in the TEM-specimen section), so the penetration depth and spreading are less for heavier elements (Fig. 4b).

Fig. 5. Interaction volume as a function of incident energy E₀ and atomic number Z of the incident (primary) electrons. (**NEED)

---------------------------------------

From conservation of energy, we can expect that the energy lost by a primary electron in an inelastic-scattering event is gained by one or more atomic electrons. If these electrons are outer-shell (valence or conduction) electrons, weakly bound (electrostatically) to a nucleus, this acquired energy may enable them to escape from the confines of a particular atom and travel through the solid. In doing so, these excited electrons will also be scattered inelastically, gradually losing their excess energy. Most of the atomic electrons acquire a kinetic energy of less than 100 eV, and since the probability of inelastic scattering increases with decreasing electron energy, the distance the low-energy electrons can travel in the solid is very small, typically one or two nm (on average). Most of them are therefore brought to rest well within the excitation volume. However, any electrons which receive their excess energy in a scattering event which takes place very close to the surface, and which are travelling in the right direction (momentum towards the surface), may escape into the vacuum as secondary electrons. These electrons are generated within very small depth (< 2 nm) below the surface, known as the escape depth.

The average number of secondaries produced per primary electron is called the secondary-electron yield d , and is typically in the range 0.1 to 10 (varying between different materials). For a given sample material, d decreases with increase in incident energy E₀ since the probability of inelastic scattering of a primary electron within the escape depth decreases.

Fig. 6. Dependence of SE yield on the angle of tilt f of the specimen relative to the primary-electron beam (f = 0 corresponds to perpendicular incidence). Data points represent experimental measurements for copper; the curve represents a 1/cosf function.

---------------------------------------

As Fig. 6 indicates, the secondary-electron yield also depends on the angle between the incoming primary electron and the surface. Its value is lowest for normal (perpendicular) incidence and increases with increasing angle between the primary beam and the surface-normal. The reason for this is illustrated in Fig. 7a, which shows a focused parallel beam of primary electrons incident at two locations on a specimen, where the surface is normal (at A) and inclined (at B) to the incident beam. The volume from which secondary electrons can escape is that which lies within the escape depth l of the surface, measured perpendicular to the surface. This escape volume, and therefore the SE yield d , is greater at point B by a factor of 1/cosf , where f is the inclination of the surface (relative to the case of normal incidence).

Fig. 7. (a) Electron beam incident normal to a surface (at A) and inclined to the surface (at B). The volume from which secondaries can escape is proportional to the shaded cross-sectional area, which is l d for case A and l d/cosq for case B.

(b) Specimen surface containing triangular and square protrusions, a square-shaped well and a round particle.

(c) Corresponding secondary-electron signal (from a line-scan along the surface), assuming a detector located to the right of the specimen.

---------------------------------------

In the case of a surface with topographical (height) variations, as shown in Fig. 7b, this orientation dependence of d results in protruding or recessed features appearing bright in outline in the SE image (similar to thickness-gradient contrast from a TEM replica. In practice, there is usually some asymmetry due to the fact that the SE detector is located to one side of the column (Fig. 2) rather than directly above. Surface features that are tilted towards the detector appear particularly bright because electrons emitted from these regions have a greater chance of being collected; see Fig. 8. This fact can be used to distinguish raised features and depressions (see Fig. 7c). ). It gives a characteristic three-dimensional appearance to the SE image, and makes the topographical contrast relatively easy to interpret, from analogy with a rough surface which is obliquely illuminated by light.

Fig. 8. (a) SEM image of the ball from a ball-point pen; the SE detector is located towards the top of the image. The scale bar is of length 0.5 mm.

(b) SE image of a purslane seed; the bar length is 0.25 mm. From Postek et al. p.10.

---------------------------------------

Taking advantage of the orientation dependence of d , the whole sample is often tilted towards the detector (as in Fig. 2) in order to increase the SE signal from all regions of the sample.

Secondary electrons are attracted towards any positively-biased electrode in the vacuum surrounding the specimen, and would generate a current which could be amplified and used as a secondary-electron signal. However, such a signal would be weak and noisy (influenced by induced voltages from ac magnetic fields, for example).

A stronger signal is obtained by first accelerating the secondaries towards a wire-mesh electrode which is biased positively by a few hundred volts; see Fig. 9. Most of the electrons pass through the grid and are accelerated further towards a scintillator which is biased positive by several thousand volts. The scintillator may be a layer of phosphor (similar to the coating on a TEM screen), a light-emitting plastic or a garnet (oxide) material, each made conducting by a thin metallic surface coating. It has the property (called cathodoluminescence) of emitting visible-light photons when bombarded by charged particles such as electrons. The number of photons emitted is roughly proportional to the kinetic energy of the electrons, and may be in the range 30 - 100 for the accelerated secondaries. The emitted light travels from the scintillator through a light pipe, a solid plastic or glass rod passing through a sealed port in the specimen chamber, to a photomultiplier tube (PMT) which is located outside the vacuum.

Fig. 9. A typical scintillator/PMT secondary-electron detector, often named an Everhart-Thornley (ET) detector after its two inventors.

---------------------------------------

The PMT is a highly sensitive detector of visible (or ultraviolet) photons and consists of a sealed glass tube containing a hard (good-quality) vacuum. The light-entrance surface is coated internally with a thin layer of a material of low work function, which acts as a photocathode. When photons are absorbed in the photocathode, they supply sufficient energy to liberate conduction electrons, which escape from the internal surface as photoelectrons. These low-energy electrons are accelerated towards the first of a series of dynode electrodes, each biased positive with respect to the photocathode. At the first dynode, biased at 100 - 200 volts, the accelerated photoelectrons generate secondary electrons within a thin surface layer, just like primary electrons striking an SEM specimen. However, the dynodes are coated with a material with high SE yield (d ) so at least two (sometimes as many as ten) secondary electrons are emitted for each photoelectron. These secondaries are accelerated towards a second dynode, biased at least 100 volts positive with respect to the first, and each of them produces at least two new secondaries. The process is repeated at each of the n (typically eight) dynodes, resulting in a current amplification of (d )ⁿ of typically 10⁶. Since each secondary produced at the SEM specimen generated 30 - 100 photoelectrons, the effective amplification (gain) of the PMT/scintillator combination may reach 10⁸ if necessary (depending on the accelerating voltage applied to the dynodes). Thus, although this conversion of SEM secondary electrons into photons, then into photoelectrons and finally back into secondary electrons seems complicated, it is justified in terms of the eventual signal level and signal/noise ratio.

A backscattered electron (BSE) is a primary electron which has been turned around (by elastic scattering) through an angle greater than 90 degrees. Such deflection could occur as a result of several collisions, some or all of which might involve a scattering angle of less than 90 degrees; however, a single elastic event with q > 90 deg is more probable. Since this elastic scattering involves only a small energy exchange, most BSE's escape from the sample with energies not too far below the primary-beam energy; see Fig. 10. The secondary and backscattered electrons can therefore be distinguished on the basis of their kinetic energy.

Fig. 10. Number of electrons emitted from the SEM specimen, as a function of their kinetic energy E₀ .

---------------------------------------

Because the cross section for elastic scattering is proportional to Z² (neglecting nuclear screening), we might expect strong atomic-number contrast if we use backscattered electrons as the signal which controls the image intensity. In practice, the backscattering coefficient h (the fraction of primary electrons which escape as BSE) does increase with atomic number, almost linearly for low Z . BSE images therefore tend to portray the chemistry of the specimen whereas SE images reflect mainly its surface topography. Another difference between the two kinds of image is the depth from which the information comes. In the case of the BSE images, the signal comes from a depth of up to half the penetration depth (some hundreds of nm). In the case of SE images, the signal comes from within the SE escape depth, which is of the order of 1 nm.

Backscattered electrons can be collected by an ET (scintillator/PMT) detector if the bias on the first grid is made negative, to repel secondary electrons. BSE's are recorded if they impinge directly on the scintillator (causing light emission) or if they strike a surface beyond the grid and create secondaries which are then accelerated to the scintillator (see Fig. 9). However, the efficiency of such a detector is very low, since it subtends a very low (solid) angle at the specimen. In other words, BSE's are emitted over a broad angular range and travel in almost a straight line (their high energy makes them relatively unresponsive to electrostatic fields) so that only those within a small angle (defined by the detector opening) reach the scintillator. Such a weak BSE signal would result in a very noisy image, so a more efficient BSE detector is required.

In the so-called Robinson detector, an annular (ring-shaped) scintillator is mounted just below the objective lens, immediately above the specimen; see Fig. 11a. The scintillator subtends a large angle and collects a large fraction of the backscattered electrons. Light is channeled (by internal reflection) through a light pipe and into a PMT, as in the case of the ET detector.

Fig. 11. Backscattered-electron detectors installed below the SEM objective lens:

(a) Annular-scintillator (Robinson) design and (b) solid-state (semiconductor) detector.

---------------------------------------

The solid-state detector shown in Fig. 11b consists of a large area (several cm²) of a silicon wafer mounted just below the objective lens. Impurity atoms (arsenic, phosphorus) are added to make it electrically conducting. There is an n-type layer, in which conduction is by electrons, and a p-type layer in which conduction is by holes (absence of electrons in an otherwise full valence band). Between these p- and n-type layers is a narrow transition region, within which the current carriers (electrons and holes) are absent because they have diffused to the other side. A voltage applied between opposite surfaces of the silicon (via metal conducting layers) results in an internal electric field across this high-resistivity transition region. When a backscattered electron arrives at the detector and penetrates as far as the transition region (assumed within the electron range), it creates extra current carriers (electron and holes) which move under the influence of the internal field, causing a current pulse to flow through the circuit. The BSE signal is measured by counting pulses, or more usually by measuring a current which is proportional to the number of backscattered electrons arriving per second. Note that secondary electrons do not have enough energy to reach the transition region, so are not registered by the detector.

A specimen-current image is obtained by using a specimen holder that is insulated from ground and connected to the input terminal of a sensitive current amplifier. Conservation of charge implies that the electron current I_s which flows to ground (through the amplifier) is equal to the primary-beam current I_p minus the rate of loss of electrons from secondary emission and backscattering:

I_s = I_p - I_BSE - I_SE = I_p (1 - h - d ) ……………………….. (1)

The specimen-current image therefore contains a mixture of Z-contrast and topographical information. To ensure that the signal is not too noisy, the current amplifier must respond to a limited frequency range, therefore the specimen must be scanned slowly (frame time of many seconds).

Electron-beam induced conductivity (EBIC) occurs when a voltage is applied to a semiconductor specimen containing p-n junctions, such as a silicon integrated circuit (IC) containing diodes and transistors. When the primary-electron probe passes near a p-n transition region, a current is induced in the specimen, as in the case of a solid-state detector responding to backscattered electrons. These regions therefore show up bright in the EBIC image. The p-n junctions in IC's are buried below the surface, but provided they lie within the penetration depth of the primary electrons, an EBIC signal will be generated. It is even possible to utilize the dependence of penetration depth on primary energy E₀ to image junctions at different depths; see Fig.12.

Fig. 12. Imaging of perpendicular p-n junctions in a MOSFET transistor (Reimer, p.265). (a) SE image, (b - f) EBIC images for increasing primary-electron energy E₀ and increasing penetration depth.

---------------------------------------

Voltage contrast arises when voltages are applied to surface regions of a specimen, usually a semiconductor integrated circuit. The secondary-electron yield is reduced in regions which are biased positive, since some lower-energy secondaries are attracted back to the specimen. Conversely, negative regions exhibit a higher SE yield because secondaries are repelled and have a greater probability of reaching the detector. The voltage-contrast image is useful way for checking whether the supply voltages applied to an integrated circuit are reaching the appropriate locations. It can also be used to test whether a circuit is operating correctly, with signal voltages appearing in the right sequence. Although most IC's (such as microprocessors) operate at too high a frequency for their voltage cycles to be observed directly, this sequence can be slowed down and viewed in a TV-rate SEM image by use of a stroboscopic technique. By means of a square-wave current applied to deflection coils installed in the SEM column, the electron beam is periodically deflected and intercepted by a suitably-placed aperture. If this chopping of the beam is performed at a frequency slightly different from the operational frequency of the IC, the voltage cycle appears in the SE image at the beat frequency (the difference between the chopping and IC frequencies), which may be as low as one cycle per second.

Fig. 13. MOSFET device (Reimer, p.300) showing voltage contrast with (a) the gate electrode (G) at -6 volt, source (S) and drain (D) electrodes grounded; (b) gate at -6 V, source and drain at -15 V.

---------------------------------------

As an alternative to detecting electrons coming from the SEM specimen, photons can be detected in certain cases. As discussed in connection with a scintillator detector, some materials emit visible light in response to electron bombardment (cathodoluminescence). Besides phosphors (Fig. 14), certain semiconductors fall into this category and may emit light uniformly except in regions containing crystal defects. Cathololuminescence (CL) images have been used to show the locations of dislocations in such specimens. The light can be detected by a photomultiplier tube, sometimes preceded by a colour filter or a wavelength-dispersive device (lass prism or diffraction grating) so that a limited range of photon wavelength is selected. Cathodoluminescence is more efficient at low temperatures, so some types of specimen are cooled (using liquid helium as the refrigerant) to below 20 K.

Fig. 14. Red, green and blue phosphor dots in a colour TV screen , imaged using secondary electrons (on the left) and in CL mode (on the right) without wavelength filtering. The lower images show the granular structure of the phosphor at higher magnification (Reimer, p. 285).

---------------------------------------

Visible light is emitted when a primary electron transfers a few eV of energy (in an inelastic collision) to an outer-shell (valence) electron, which then emits a photon while returning to its lowest-energy state. If the primary electron collides with an inner-shell electron, more energy must be transferred to excite the atomic electron to a higher energy state, and a photon of higher energy (hundreds or thousands of eV) may be emitted as a characteristic x-ray photon. The x-ray energy can be measured and used to identify the atomic number of the participating atom, as we will discuss in the section on analytical electron microscopy. If the characteristic x-ray signal is used to control the image intensity, the result is an elemental map showing the distribution of a particular chemical element within the SEM specimen.

The SEM operator can control several parameters of the SEM, such as the electron-accelerating voltage, the distance of the specimen below the objective lens, known as the working distance (WD), and sometimes the diameter of the aperture which is used in the objective lens to control spherical aberration. The choice of these variables in turn influences the performance obtained from the SEM.

The accelerating voltage determines the kinetic energy E₀ the primary electrons, and therefore their penetration depth and the information depth of the BSE image. Since secondary electrons are generated within a very narrow escape depth, the SE image might be expected to be independent of the choice of E₀ . However, only some of the SE signal (the so-called SE1 component) comes from secondaries generated by primary electrons close to the surface. Other components (named SE2 and SE3 respectively) arise from the generation of secondaries by backscattered electrons as they exit the specimen and when they strike an internal surface of the specimen chamber; see Fig. 15a. As a result, SE images can show contrast from structure present well below the surface (but within the primary-electron penetration depth) if this structure results in a change in backscattering coefficient (e.g. due to difference in local atomic number). This effect increases with increasing penetration depth, so at higher accelerating voltage the sample appears more "transparent" in the SE image; see Fig. 16. Conversely, if the accelerating voltage is reduced to 1 kV or less, the penetration depth becomes very small (even comparable with the secondary-electron escape depth) and only surface features are seen in the SE and BSE images.

Fig. 15. (a) Generation of SE1 and SE2 electrons by primary electrons and by backscattered electrons, respectively. SE3 electrons are generated outside the specimen when a BSE strikes the bottom of the objective lens (for example).

(b) Resolution function, showing the relative contributions from secondaries generated at different distances from the centre of the electron probe.

---------------------------------------

Fig. 16. SE image of a TEM grid, imaged with an electron-accelerating voltage of 40 kV (on the left) and 100 kV (on the right). Roughness on the bottom surface of the grid becomes visible in the 100kV image.

---------------------------------------

Since the primary beam spreads laterally as it penetrates the specimen, and since backscattering occurs over a broad angular range, the spatial resolution of the (SE2+SE3) signal is worse than for the SE1 component. In other words, SE2 and SE3 electrons contribute a tail or skirt to the image-resolution function; see Fig. 15b. This fact complicates any definition of spatial resolution; however, a sharp central peak in the resolution function ensures that some high-resolution information will be present in the SE image.

The width of this central peak (Fig. 15b) is approximately equal to the diameter d of the electron probe, which depends on the electron optics of the SEM column. To obtain high demagnification of the electron source, the objective lens is strongly excited and the image distance (the working distance WD) between the objective and specimen becomes correspondingly small (below 2 cm). A strong lens (small focal length, therefore small C_s and C_c) also helps to minimise broadening of the probe due to spherical and chromatic aberration. The smallest available values of d (below 1 nm) are achieved by employing a field-emission source, which has a small electron-emitting diameter, and a working distance of only a few mm.

Of course, the optimum image resolution is obtained only if the SEM is properly focused, which is done by carefully adjusting the objective-lens current. Small particles on the specimen form a convenient feature for focussing; the objective excitation is adjusted until their image is as small and sharp as possible. Astigmatism of the SEM lenses can be corrected at the same time; the two stigmator controls are adjusted so that there is no streaking of image features as the image goes through focus. This is similar to the procedure used in the TEM; in the SEM, zero astigmatism corresponds to a round (rather than elliptical) electron probe, and an axially-symmetric resolution function.

Compared to a light-optical microscope, one advantage of the SEM (besides higher resolution) is the large depth of focus. The latter can be defined as the change D v in specimen height (or working distance) which produces a just-observable loss (D r) in image resolution. As seen from Fig. 17b, D r ~ (2a )(D v) where 2a is the total convergence angle of the probe. Taking D r equal to the probe diameter d , so that the resolution is degraded by a factor ~ 2 (similar to the TEM argument),

D v ~ d / (2a ) …………………………………… (2)

The large depth of focus is seen to be a direct result of the relatively small convergence angle a of the electron probe, which in turn is dictated by the need to limit probe broadening due to spherical and chromatic aberration. But as shown in Fig. 17a,

a ~ D/(2.WD) ………………..………………… (3)

so the depth of focus can be controlled by varying the diameter D of the objective aperture or the working distance, as illustrated in Fig. 18.

Fig. 17. (a) Formation of a focused probe of diameter d by the SEM objective lens.

(b) Increase D r in electron-probe diameter for a plane located a distance D v above or below the plane of focus.

---------------------------------------

Fig. 18. Demonstration of the depth of focus of an SEM image, for different values of the objective-aperture diameter D and the working distance WD. White lettering indicates the probe semi-angle a , calculated from Eq. (3). Note the large depth of field which is possible with small a .

---------------------------------------

A major advantage of the SEM (in comparison to a TEM) is the ease of specimen preparation, a result of the fact that the specimen is not required to be thin. In fact, many conducting specimens require no special preparation (other than cleaning, sometimes) before use in the SEM. However, insulating specimens tend to charge up electrostatically when exposed to the scanning electron probe, since they provide no path to ground for the specimen current I_s. As shown by Eq. (1), this current could be of either sign, depending on the values of the backscattering coefficient h and secondary-electron yield d , so the specimen could charge up positively or negatively. Negative charge is the more serious problem, since it tends to repel (deflect) the primary-electron beam, resulting in image distortion or fluctuations in image intensity.

One solution to the problem of charging is to coat the surface of the SEM specimen with a thin film of a metal or graphitic carbon This is done in vacuum, using the evaporation or sublimation technique already discussed in connection with TEM specimen preparation. Films of 10-50 nm thickness conduct sufficiently well to prevent charging at the surface. Since this thickness is greater than the SE escape depth, secondaries are recorded from the coating rather than from the original specimen. However, the external contour of such a thin film closely follows that of the specimen, so a faithful topographical image is usually obtained in SE mode. Gold and chromium are common as coating materials; graphitic carbon is also used, which has a low SE yield but an extremely small grain size so that granularity of the coating does not appear (as an artifact) in a high-magnification SE image and mask real specimen features.

Where coating is undesirable or else difficult (e.g. very rough or convoluted surfaces), specimen charging can often be avoided by choice of the SEM accelerating voltage. This possibility arises because the values of backscattering coefficient h and secondary-electron yield d for a given material depend on the primary-electron energy E₀. For high E₀, the penetration depth is large and only a small fraction of the excited atomic electrons escape into the vacuum as secondaries. Also, BSE which are generated deep within the specimen will not have enough energy to escape, so h will be low. Low total yield (h + d ) means that the specimen will charge negatively; see Eq.(1). As E₀ is reduced, d increases and the specimen current I_s required to maintain charge neutrality eventually falls to zero at some incident energy E₂ , where (h + d ) = 1. Further reduction in E₀ could result in a positive charge, but this would attract secondaries back to the specimen, thereby neutralising the charge, so it is not usually a problem.

For E₀ less than some energy E₁ , the total yield again falls below unity because the primary electrons do not have enough energy to create secondaries; since h < 1, negative charging occurs. As shown in Fig. 20, there is therefore a range of primary energy (E₁ to E₂) over which charging does not occur, even for an insulating specimen. Typically E₂ is in the range 1 - 10 keV; although there are tables giving this value for common materials, E₂ is usually found experimentally, by reducing the voltage until charging effects in the image disappear.

Fig. 20. Total yield (h + d ) as a function of primary-electron energy, showing the range (E₁ to E₂) over which specimen charging is not a problem.

---------------------------------------

The use of low-voltage SEM is therefore an attractive option for imaging insulating specimens. However, there is a downside to low E₀ : increased chromatic-aberration broadening of the electron probe, by approximately r_c = C_ca (D E/E₀). This problem can be minimised by using a field-emission source (which has a low energy spread: D E < 0.5 eV) and by design of objective lenses with a low chromatic-aberration coefficient C_c . This is an area of recent and continuing research.

An alternative approach to overcoming the specimen-charging problem is to surround the specimen with a low pressure of gas, rather than a high vacuum. Before reaching the specimen, primary electrons will ionize gas molecules and the positive ions will be attracted to the specimen (if it charges up negatively), neutralizing the charge.

To allow operation of a thermionic or field-emission source, the use of high voltage to accelerate the electrons and the focussing of electrons (without scattering from gas molecules) there must still be a good vacuum inside the SEM column. Primary electrons only encounter gas molecules during the last few mm of their journey, after being focused by the objective lens. A small-diameter aperture in the bore of the objective allows the electrons to pass through but prevents most gas molecules from travelling up the SEM column. Those that do are removed by continuous pumping. In some designs, a second pressure-differential aperture is placed just below the electron gun, allowing a good vacuum to be maintained in the gun, which has its own vacuum pump.

The pressure in the sample chamber may be as high as 5000 Pa (0.05 atmosphere) but is more typically a few hundred Pascal. The environmental gas is frequently water vapour, since this arrangement allows wet specimens to be examined without dehydration, provided the specimen-chamber pressure exceeds the saturated vapour pressure (SVP) of water at the temperature of the specimen. At 25° C, the SVP of water is about 3000 Pa, but the required pressure can be reduced by a factor of 5 or more by cooling the specimen.

Examples of specimens which have been successfully imaged in the environment al SEM include include plant and animal tissue (see Fig. 21), textile specimens (which charge easily in a regular SEM), rubber and ceramics. Oily specimens can also be examined without contaminating the entire SEM; hydrocarbon molecules that escape through the differential aperture are removed by pumping.

Fig. 21. Mouse intenstine imaged in an environmental SEM. The width of the image is 0.7 mm. Courtesy of ISI / Akashi Beam Technology Corporation.

---------------------------------------

Although the environmental SEM extends the range of materials which can be examined and avoids the need for coating the specimen to make it conducting, it has one drawback: electrons are deflected through scattering by gas molecules during the final phase of their journey. This effect adds an additional skirt (tail) to the current density distribution in the electron probe and therefore reduces the image resolution and contrast.

A solid-state detector can be used to generate a BSE image, as described previously. The Everhart-Thornley detector cannot be used because the voltage it uses to accelerate secondary electrons would cause an arc discharge in the environmental specimen chamber. Instead, a potential of a few hundred volts is applied to a ring-shaped electrode just below the objective lens; secondary electrons initiate a controlled discharge between this electrode and the specimen, resulting in a current which is amplified and used the SE signal.