Very Low Energy Scanning Electron Microscopy - Formatex

Modern Research and Educational Topics in Microscopy. ©FORMATEX 2007 A. Méndez-Vilas and J. Díaz (Eds.) _______________________________________________________________________________________________

Very Low Energy Scanning Electron Microscopy I. Müllerová*, and L. Frank Institute of Scientific Instruments, Academy of Sciences of the Czech Republic, Královopolská 147, CZ61264 Brno, Czech Republic Very low energy scanning electron microscopy is introduced as a scanning version of the low energy electron microscopy, i.e. the emission electron microscopy with the sample excited by means of a parallel wave of electrons. The incident primary electrons are retarded in the cathode lens consisting of a negatively biased sample (cathode) and earthed anode. Interaction of slow electrons with solid targets is discussed and principles of the instrument are described together with selected applications demonstrating image contrasts not available in conventional scanning electron microscopy devices. Keywords electron optics; scanning electron microscopy; low energy electrons; cathode lens mode; surfaces

1. Introduction The Emission Electron Microscope (EEM) is one of the oldest versions of electron microscopes – review see [1]. In this type of direct imaging microscope the specimen itself is the source of electrons released under some of the scope of possible excitations. Sixty years ago Recknagel [2] showed that the immersion objective lens, which is a crucial part of the EEM that accelerates the electrons emitted at quite low energy E0 to some final energy E and forms the first image of the emitting surface, has surprisingly good properties. Its basic aberration coefficients are proportional to the ratio E0/E so that they keep decreasing down to lowest emission energies. One of the versions of EEM is the Low Energy Electron Microscope (LEEM), in which the specimen excitation is made via a parallel coherent wave of incident slow electrons [3]. The lowest energies (usually below 50 eV) are obtained in the decelerating field of the Cathode Lens (CL), in which the specimen itself is on a high negative potential with respect to the anode on the ground potential. The same CL field then accelerates the reflected electrons or electrons emitted under the bombardment. The brilliant results obtained with this instrument [4] encourage us to test the CL also in the Scanning Electron Microscope (SEM). In the following the design of a simple adaptation of a commercially available SEM to the SLEEM method (Scanning LEEM = SEM equipped by the CL) is presented together with short summary of our over 10 years of practical experience with this system [5].

2. Interaction of slow electrons with solids At electron energies conventional for the SEM, i.e. at 5-30 keV, the scattering theory can be briefly summarized as follows [6]. Fast electrons are elastically scattered by atom cores and hence the scattering rate is proportional to the atomic number of the target. Generation of the continuous x-ray emission (the Bremsstrahlung) connected with the non-uniform motion of electrons in the nucleus field can be well neglected. The inelastic scattering phenomena, primarily concerning the target electrons, are multiple and include excitation or damping of collective vibrations of the electron gas (plasmons), as well as the excitation of atomic electrons within a partially occupied band or into an unoccupied band, and finally the ionization of target atoms by striping them of electrons. The incident electrons diffuse into an interaction volume of a diameter between tens of nm for heavy elements at 5 keV, and tens of µm for lightest elements at tens of keV. From the upper half of this *

Corresponding author: e-mail: [email protected], phone: +420 541514300

795


volume the incident electrons can rebound after single or multiple scattering events and can even be emitted as the backscattered electron (BSE) with energy at or below the energy of impact. Secondary electrons (SE) released in the ionization process start at energies in tens of eV only so that their diffusion prior to recapture is much shorter. Internal absorption of SE once released, caused by the scattering on phonons and also connected with intraband transitions in metals and electron-hole pair generation in semiconductors, restricts the escape depth to 1 or 2 nm for conductors and, say, 10 to 20 nm for insulators [7]. When emitted, SE’s kinetic energy distribution peaks at 1 to 3 eV and extends to tens of eV. In the range of 5 to 30 keV, the average yield of BSE per one incident electron is between 0.1 and 0.55, depending on the atomic number but not varying with energy. Across the given energy interval the SE yield decreases about 4 times, dropping in this way well below the BSE yield. One important feature is the penetration depth of incident electrons being much larger than the escape depth of SE. Inclined surface facets then offer an extended emitting area as the main contrast source in the SE images. One of consequences is the so called edge effect consisting of the overbrightening of any facets steeply inclined with respect to the overall surface plane, like grain boundaries, etch pit walls, or artificial structures of rectangular profiles. Moreover, BSE returning toward the surface release an additional emission of SE, called SE2, and also SE3 from exposed internal surfaces of the microscope chamber. The faster BSE bear the material information, and also some traces of the structural information, i.e. the backscattering rate influenced by the crystal orientation, can be caught mainly at large emission angles. With few exceptions the total emission yield remains below one and some negative charge dissipates in the sample, causing localized charging in insulators. Let us now move along the energy scale to the range from 5 keV down to, say, 100 eV. The penetration depth of incident electrons shortens proportionally to En with n between 5/3 and 4/3, therefore faster than linearly, and the spatial density grows not only in regard to the scattering events but also in terms of the dissipated energy. The SE signal grows and the total emitted signal approaches the unit level, in this way reducing the local charging in non-conductors. The lateral diffusion of electrons is also shorter and the complete interaction volume shrinks, making the acquired information better localized. Small “inclusions” differing in the signal electron yield become observable. The penetration depth progressively shortens towards the escape depth, so the edge effect weakens and disappears. In the range of units of keV various materials pass through their critical energy of electron impact at which the total emission equals to one, no charge is dissipated, and we have a non-charging mode of a high-vacuum SEM as an alternative to low-vacuum or environmental devices. Below the critical energy the local charging of an insulator changes its sign and a positive charge is effective as regards to the above-surface field. This field recaptures a portion of the slowest SE, restoring in this way the charge balance at only moderate positive surface potential. Finally, let us touch upon the question of carbonaceous contamination created on samples by the decomposition of adsorbed hydrocarbons under electron impact and with the problem of radiation damage in general. Down to about 100 eV the spatial density of dissipated energy delivered by incident electrons keeps growing so around this impact energy the strongest contamination and damage is met. Only below this threshold the contamination and damage rates drop. Now we enter the very low energy range of electrons below 100 or 50 eV. The boundary of this energy range is a bit fuzzy but still based on a crucial feature following from Fig. 2, namely the minimum of the inelastic mean free path appearing at just 50÷100 eV for all materials. Electrons injected at energy below 50 eV behave like excited (hot) electrons, i.e. similar to SE liberated from the target atoms. Practically, we deal here with a slow electron (SE-like) emission without issues connected with the action of fast BSE, which include mainly the SE2 and SE3 emissions, normally deteriorating the image resolution and complicating interpretation of the contrast observed. At higher electron energies, at which the scattering can be well divided into individual events taking place on atomic scatterers, the whole process can be simulated by Monte Carlo methods based on the density of scatterers and their cross-sections. Random number generators choose here the characteristic parameters of individual events, i.e. changes in energy and direction of motion in an event, so that their known statistics are respected. The simulation tools have to reflect that while at high energies the

796


forward elastic scattering prevails, at lower energies the elastic backscattering and inelastic scattering dominate. The atomic potential, i.e. potential of the nucleus screened by atomic electrons, becomes truncated with neighboring atoms so it loses its spherical symmetry and effectively the assembly of individual scatterers is replaced with the whole scattering crystal. In the energy range discussed here, also correlation and exchange interactions, including spin-spin ones, between incoming and target electrons have to be incorporated.

Fig. 1 Elastic mean paths calculated using the Mott cross-sections [8].

Fig. 2 Elastic (el.) and inelastic (inel.) mean free paths and attenuation lengths (atten.) calculated by means of a MC program incorporating the Mott cross-sections and the dielectric loss function [9].

When determining the cross-sections very carefully, we can extend the use of the Monte Carlo type simulation tool down to very low energies. Figs. 1 and 2 show the inelastic and elastic mean free paths (IMFP and EMFP) down to a level just above the vacuum one, and confirm that while IMFP steeply grows in the very low energy range, EMFP only loses its inverse proportionality to the atomic number. The inelastic scattering data is usually calculated from the complex dielectric permittivity based on an experimental electron energy loss spectrum. When the elastic scattering is described by the quantum mechanical Mott cross-sections, the scattering rate is overestimated at very low energies, particularly in semiconductors and insulators. Instead, the incident electron is proposed to be treated as a Bloch electron within the given energy band structure of the target and the quasi-elastic scattering on acoustic phonons is incorporated. Longitudinal optical phonons play an important role in the scattering, too [10]. At very low energies the concept of scattering events has to be extended to effects inherent to the spatial distribution of electron states, which are connected to complex properties of the whole crystal. The energy band structure exists even above the vacuum level and the allowed bands are separated by forbidden gaps, e.g. on margins of the Brillouin zones. Generally, the incident electron enters the crystal in dependence whether or not a non-negligible density of empty Bloch states couple to the incident wave. The condition is that the electron enters with a sufficiently low energy at which the inelastic phenomena are already reduced enough to preserve the energy and wave vector of the incident electron for sufficient time; this takes place below about 20 to 25 eV. The “band structure region” is known from the I/V characteristic of the specular (00) spot of the low energy electron diffraction (LEED) pattern below the threshold at which the first non-specular spot emerges [11]. This phenomenon makes the very low energy elastic BSE yield dependent on the crystal orientation and also on other local variations in the density of states like those connected with a doping in semiconductors. When moving near to zero impact energy, we can map any local fields appearing on and above the specimen surface with sensitivity. Examples include lateral fields appearing between areas of different densities of surface dipoles, i.e. different inner potentials, like the fields at p/n junctions or metalsemiconductor junctions. These fields are concentrated in tiny volumes so very slow electrons can only change their trajectories significantly within this volume in order to visualize the fields. Electron imaging at near-to-zero energy not only reveals these fields but the processing of images can even provide quantitative data [12]. Specific type of observation of the above-surface fields works with the threshold between very low energy backscattering and total reflection at or slightly above the surface. Suppose we

797


have a part of the sample at a negative potential with respect to the neighborhood. When using an incident beam of energy near to this bias, we get from the sample a BSE flux of a still quite broad energy and angular distributions but from the biased area we may acquire a narrow, totally reflected quasimonochromatic pencil. As in observations of the lateral fields, what is crucial is which part from all emitted fluxes is captured by the detector [13]. The mirror electron microscopy (MEM) mode [14,15], which is closely related to our topic, is not treated here but any user of the SLEEM instrument would profit from familiarity with the basics of the MEM. Mutual interaction of the signal electrons, taking place either after they leave the sample surface or even before this, is normally taken into account in the transmission electron microscopy (TEM) where the image wave results from interference between the scattered wave and the primary electron wave passing through the sample without interaction. In the SEM “reflection-type” experiment the scattered waves seem so smashed that any coherent imaging mode is impossible. Really, the exit wave angles are between 90° and 180° with respect to the input wave so that the path differences are by orders of magnitude longer than the wavelength at energies usually used. However, at tens and units of eV the wavelength of electrons approaches the interatomic distances (λ[nm] = 1.226/{E[eV]}1/2) so that partial electron waves scattered from atoms can interfere to form diffraction patterns revealing the crystal structure. In addition, waves scattered on both sides of the surface atomic steps or on both surfaces of ultrathin overlayers can interfere visualizing so these features. When a sufficient portion of the illuminated spot produces mutually coherent partial waves with path differences equal to integer multiples of λ/2, we get very intensive bright and dark stripes or rings in the emitted flux. And again, what type of detector we have and how its collection efficiency is distributed over emission energies and angles is important for the final image. For appearance of the wave-optical contrasts, the beam coherence within the single primary spot and the convergent-beam geometry of the experiment is what is significant. The coherence problems are similar to those faced in the scanning TEM (STEM) experiment [16]. The beam coherence is governed by the energy and angular spreads of the incident beam. The energy spread projects itself into a finite coherence length limiting the maximum path difference between waves to be amplitude-added. When considering a cold field emission tip as the electron source of a spread, say, 0.25 eV, we get the coherence length equal to 30 nm at 10 eV and to 10 nm at 1 eV [5]. Coherence is further restricted by the finite size of the electron source. When accepting a decrease in the complex degree of coherence on margins of the pupil not below 88% and supposing the entrance pupil of the objective lens illuminated from a source of an angular radius 1 mrad, we get the tolerable source size in hundreds of nm [17]. In crystals, the reciprocal lattice points have some effective dimensions given by size of the crystal area from which the amplitude adding takes place, and therefore the diffracted beams also have a finite angular width given by the Bragg condition for diffraction at a particular set of atomic planes. When inverting this relation we get some limiting size of the crystal area for any spread in the incident wave vector, i.e. also for spreads resulting from the energy and angular width of the beam. Altogether, in an available SLEEM configuration, the coherently illuminated area sizes to 10÷20 nm [18] so it comprises well the primary spot. The beam aperture angle in the specimen plane in a SLEEM is, say, around 30 mrad, so the diffracted beams are broadened and might even be mutually overlapped. In this case the neighboring diffracted beams further enhance their intensity through mutual amplitudeaddition.

3. Instruments Modern SEMs are equipped with a field emission gun and the specimen is usually immersed in a relatively strong magnetic field in order to obtain high resolution of 1 nm or even less at 15 keV of the primary beam energy. Nearly the same resolution could be preserved down to 1 keV if the retarding field optics is used [19]. Resolution of about 4 nm at 100 eV was obtained experimentally in a commercially available SEM with a compound electrostatic/magnetic objective lens utilizing the retarding field principle [20], but the spot size steeply increased at lower energies.

798


There is only way to keep high resolution down to zero landing energy and that is to immerse the specimen in a strong electrostatic field in a similar fashion as is done in the LEEM. When calculating the basic aberrations of the immersion lens formed by two flat electrodes held on different potentials and focusing the primary beam on the nearby specimen surface, we get minimum aberrations for zero working distance of this lens, i.e. for the specimen immersed in a strong electrostatic field [21]. This type of immersion lens is called the cathode lens. The next step is to estimate the spot sizes in the SEM and SLEEM modes along the energy scale and to compare them for the optimum beam aperture separately adjusted at each energy value, and for a certain fixed aperture defined by a diaphragm. The summation rule based on defined limits of encircled current [22] was used in the calculation. The total aberration coefficients were calculated for sequential arrangement of the electrostatic field of the CL and the focusing field of the objective lens, taking into account aberrations of the anode lens (i.e. electrostatic field in the anode opening) [23], and of the magnetic focusing lens [24]. The spot size was determined for two categories of the SEM: "TEG SEM", representing medium class instruments for everyday work, and "FEG SEM" for high quality modern devices. The parameters were chosen as β=105 Acm –2sr –1, I=5 pA, ∆E=2 eV, CSm=10 mm, CCm=-12 mm for the TEG SEM, and β=109 Acm –2sr –1, I=100 pA, ∆E=0.2 eV, CSm=1.9 mm, CCm=-2.5 mm for the FEG SEM (β is the brightness of the electron gun, I is the beam current in the specimen plane, ∆E is the energy spread of the beam, and CSm and CCm are aberrations of the objective lens). In Fig. 3 we have the calculated spot sizes plotted in dependence upon the landing energy EL (EL = EP – eUSP with EP for the primary beam energy and USP for the specimen potential). The measured image resolution was determined as a 25/75 edge width on the standard Au/C testing sample. In the FEG SEM the ultimate resolution of 9 nm at 10 eV was experimentally obtained [25]. Fig. 3 Calculated (lines) and measured (points) spot sizes in dependence of the landing energy for (a) TEG SEM and (b) FEG SEM; the fixed beam apertures in front of the cathode lens are (a) 8 mrad, and (b) 6 mrad.

Our simple and most successful arrangement of the SEM adaptation to the SLEEM mode is based on the bored YAG:Ce3+ single-crystal disc with outer/inner diameters of 10/0.3 mm, which forms the CL anode at the ground potential, situated closely above the sample, while the sample serves as the cathode on a negative potential [26]. The crystal is side-attached to a light guide and can be precisely aligned to the optical axis. The bore size was tuned to a balance between reasonable dimensions of the field of view and successful acquisition of very-low-energy electrons collimated towards close vicinity of the optical axis and partly escaping detection through the crystal bore. More sophisticated SLEEM arrangements can be capable of detecting all emitted electrons but a complicated electron optical system has to be employed in order to split the primary and signal beams and deflect the latter toward a side attached detector [27]. Through-the-lens detector configurations can be combined with the cathode lens as well. Cleanliness of surfaces under observation is important, of course, but not so much as one might expect. In the lowest energy range the penetration/information depth increases again (see Fig. 2) so the atomic cleanliness is required only where it forms the necessary condition for the phenomena to be observed. Otherwise the standard SEM vacuum is sufficient. More critical is the specimen flatness and roughness. Experience has showed that protrusions, steps, or ridges can be tolerated up to heights in the units of µm range down to 1 eV of electron energy. They are, however, observed at a much reduced depth of focus. Otherwise, the specimen has to be flat at least within a diameter comparable to the anodecathode distance. The optical power of the cathode lens influences the primary beam before its impact on the specimen and modifies characteristics like the image magnification, working distance, and impact angle connected

799


with the beam rocking around the pivot point of the scanning system. Moreover, the field of view is reduced because of the small opening in the detector. Our experience showed none of these restrictions too serious for practical use. Nevertheless, let us briefly summarize the factors mentioned. Let us suppose the primary beam is deflected in the scanning system, passed the centre of the objective lens, and directed to a certain radial coordinate r0 in the specimen plane. The diverging anode lens will change the trajectory slope and shift the impact point further off the axis. Finally the homogeneous field will reshape the linear trajectory to a parabolic one and move the impact to its final position r1. Thus, a magnification correction factor ∆M=r0/r1 has to be incorporated in the control system of the SEM [28]. At very low energies (i.e. with a large immersion ratio k = EP/EL) we get ∆M ∈ (1/2, 2/3) for the full range of positions of the pivot point of the scanning system with the most often met value ∆M ≈ 0.6. In a similar way we can derive a refocusing factor correcting the axial shift of the sharp image plane owing to action of the cathode lens. Effectively, the objective lens has to be underfocused by about one third of the anode-cathode distance L, in order to compensate for the optical powers of the anode lens and the homogeneous field. The calculated refocusing factor [28] amounts, for example, to 7 µm/eV at 10 eV for L=5mm and EP=10 keV. When in the CL, the specimen serves as the electrode of an electrostatic lens and so any specimen tilt introduces a lateral field component, which shifts the primary beam in the tilt direction, unidirectionally blurs the primary spot proportionally to the off-axis distance, and enlarges the impact angle. When tilting the specimen, we have to correct the position of the field of view. Blurring of the spot can be expressed by the number N of spot sizes just filling the field of view when the most blurred spot at the field margin is the measure. When considering values typical for the very-low-energy range, i.e. the immersion ratio k=1000, the angular aperture in the anode plane 2 mrad and the specimen tilt ω=1°, we get N=407, i.e., a number on the edge of acceptability [18]. The action of the lateral field strongly enlarges the final impact angle of the beam with respect to the specimen tilt; except when in the near-glancing position, the impact angle is well approximated by ωk1/2. Obviously, at very low energies even slight tilts by tenths of a degree can secure the full scale of impact angles, which is important for applications requiring acquisition of non-specular diffracted beams or energy-band contrasts. Moreover, the mechanical tilts necessary for large impact angles are only tolerably destructive as regards to the image quality.

4. Image contrasts at very low energies When mapping applications of the SLEEM, we should best start with experiments performed at the upper margin of the energy range chosen, i.e. around 50 eV. Here some traces can still be found of the contrast types known from the conventional SEM devices, which should facilitate our start on the new method. However, it is always highly recommendable to acquire not a single image but series of micrographs of the identical field of view, starting from standard keV energies and continuing down to very slow electrons. If we refrain from doing this we usually fall into trouble when interpreting what we see in terms of what we already know from the SEM practice. Variations in image contrasts throughout the full energy scale are surprisingly large and so with the microscope column not well aligned together with the cathode lens and therefore with the field of view drifting at high specimen biases it may be difficult to identify the area originally chosen. At around 50÷100 eV the carbon contamination in standard vacuum devices grows so much that acquisition of an image undamaged by black traces of previous frames requires great care. It is especially recommended to focus and correct the primary spot in a neighboring field of view and to jump on the right one only for recording the final frame. If any adjustments have to be made in the proper field, then a bit smaller magnification is recommended in order to get then the deposited frame behind margins of the snap saved afterwards. Below about 20 eV the contamination rate steeply decreases and also the previous traces stop being visible (Fig. 4). When approaching the energy range around 50 eV and below, in most cases we first notice one or more of the significant features following from the scattering characteristics mentioned above. These are enhanced surface sensitivity, increased influence of the overall crystal potential, and phenomena

800


connected with the electronic structure of the target. For example, when comparing the 10 eV micrograph with the standard 5 keV one in Fig. 5, we recognize disappearance of the edge effect and appearance of contamination islands not visible before, both owing to diminished penetration depth of primary electrons. Without glaring edges the surface relief appears richer in tiny details so technologically important surfaces can be well examined. In addition, we see the grains of the metal polycrystal distinguished by intensity, as corresponds to differences in the scattering potential along different crystal axes. These effects allow for observation of crystalline grains (offering an alternative to the Electron BackScatter Diffraction method) as well as for distinguishing between amorphous and crystalline phases.

Fig. 4 Islands of a 300 nm thick Cu layer deposited through a mask, exposed by e-beam lithography, onto the Si substrate, period of squares is 10 µm; electron energies from the left: 5000, 100, and 10 eV. (Specimen provided by F. Matĕjka, ISI Brno.)

Fig. 5 Surface of a polycrystalline Cu sheet etched in nitric acid, landing electron energies 5 keV (left) and 10 eV (right), the width of the field of view is 100 µm.

The shortened penetration depth and lateral diffusion range also provide a tool for diagnostics of surface bound artificial structures or cross sections of such structures. For example, in an integrated circuit we see local defects at the very surface, which are in-depth “averaged” and hence invisible in the standard mode micrograph. Small structure features not protruding above the surface are normally “dissolved” in the neighborhood signal but when the full 3D interaction volume diminishes, these details become observable much better than is normal in an SEM. This concerns various types of inclusions, fillings of polymer blends, precipitates in alloys [29] or ceramic particles in composites, etc. Moreover, the material contrast between components even near to each other in the mean atomic numbers can appreciably grow at suitable energies. Remember that we acquire the full emitted spectrum of energies, which, as experience has shown, only rarely suppresses image contrasts with background enhanced relative to standard SEM observation but mostly adds new contrasts enabling one to resolve more details. On the other hand, well known and “reliable” contrasts, like that between gold and carbon in commercial samples for resolution testing, can disappear or even invert (Fig. 6). This phenomenon again consists in the scattering on individual atomic scatterers being predominated by the scattering on the complete crystal potential. Possible applications span plenty of material science or even life science tasks but in order to understand the image, tailoring of the electron energy is needed with respect to the type of observable details. In Fig. 6 some instrument dependence of the contrast is even demonstrated. The key is in the difference between energy and angular distribution of emission from gold and carbon. When the original material contrast of the BSE yields disappears, the more peaky angular distribution of both BSE and SE from Au becomes most important, causing more electrons to be lost through the detector bore. Then the contrast inverts and the gold areas are darker. However, if the sample is immersed in the magnetic field of the objective lens, the signal trajectory bundle is broadened and acquisition of the gold emission is less suppressed. Energy distributions contribute to the effect as well [25].

801


Fig. 6 The gold-on-carbon resolution testing specimen in two microscopes equipped with the cathode lens: a high resolution SEM with the sample immersed into magnetic field (upper row), an allelectrostatic ultrahigh-vacuum SEM (lower row); electron energy 10 keV (left) and 20 (eV) (right); field of view 250 µm (top) and 150 µm (bottom), respectively.

A new world of image contrasts opens when employing the wave-optical phenomena. The interference of partial waves is simply governed by their path differences but intensities of waves and their angular distributions are complex functions of the specimen including its intentional or spontaneously grown coatings. Contrasts in Fig. 7 are caused by the signal-enhancing addition of amplitudes, taking place where the kinematic Bragg condition is fulfilled and some of the non-specular diffracted beams impinge on the dark field detector. Dynamic diffraction phenomena appear here as well [18]. Oblique incidence of primary electrons, secured by only a small sample tilt, adds one degree of freedom and complicates interpretation. However, when combining the tilt and energy series of images, the complete mapping of I/V data for individual crystal orientations would be possible. The beam aperture angle in the specimen plane was around 30 mrad so the diffracted beams are broadened and likely overlapped. However, these effects do not demonstrate themselves when using the integral annular dark field detector (the bored scintillator disc above the sample) except that the overlap of diffracted spots can increase the contrast level. Let us notice that the diffraction and interference contrasts are observable solely under very clean conditions, when the crystallinic arrangement of the target reaches the very surface and no damaged layer or contamination coverage are present. This kind of application task is extremely attractive, especially because of the historical success of the non-scanning LEEM version acquiring the LEED-like structural information at large spatial resolution [4]. The SLEEM method can potentially go behind the LEEM results thanks to the possibility of the parallel acquisition of multiple diffracted beams, not yet utilized in the present instrumentation. As mentioned above, below 20 or 25 eV, the electron reflectivity becomes inversely proportional to the local density of states coupled to the incident wave. The effect should be observable on grains in polycrystals (see Fig. 8) and on doped patterns in semiconductors [13]. Concerning the doped semiconductor areas, a specific dynamic effect has been observed, consisting in the p-type silicon negatively charged to about 1 V, which might be enough for total reflection of 1.5 or 2 eV electrons. The local charge is obviously due to ionized acceptors that are not balanced any more with holes, now recombined by incident electrons. Very high contrast generated in this way between backscattered and totally reflected signals is shown in Fig. 9 together with its explanation based on the impact of a narrow totally reflected electron beam on or outside the active detector surface. This example concerns imaging of local electric fields, i.e. the fields generated by surface charges and combined with the cathode lens field and its lateral component due to the sample tilt.

802


Fig. 7 Diffraction contrasts acquired with flat Pb crystals in-situ deposited onto a clean Si (100) surface; upper row: normal electron impact, energies are (a) 5 eV, (b) 12.5 eV, (c) 42.5 eV, field of view 60 µm; lower row: specimen tilted to 1.3° in the direction inclined at 55° with respect to the horizontal line, energies are (d) 6.5 eV, (e) 10.5eV, and (f) 22 eV, field of view 50 µm.

Fig. 8 Micrographs of the surface of pure aluminium extruded to a rod, cut into discs, annealed above 500°C, quenched in cold water and electrolytically polished; electron energies in eV as labelled, width of the field of view 100 µm.

Fig. 9 P-type doped pattern in n-type Si substrate observed in the induced-charge contrast at 4×10-14 C/pixel, electron energy 1.5 eV, specimen tilt/impact angle: (a) 0°/0°, (b) 0.45°/31°, (c) 0.72°/80°, length of the scale bar is 20 µm. (d) Scheme of the signal trajectories influenced by fields indicated by arrows, explaining the total reflection of the primary beam above the sample surface toward the detector or into its bore.

The principle of the cathode lens with negatively biased specimen enables one to put another detector below the specimen as in an STEM, and to collect the transmitted signal of very slow electrons. Electron transmissivity of ultrathin films and layers is expected to increase due to the increasing of the inelastic mean free path at lowest electron energies. However, the elastic mean free path does not exhibit the same general behavior so the actual penetration through a sample remains an issue and the first experiments have not answered the question unambiguously. Foils penetrable with very slow electrons would, among other things, find applications in correcting the aberrations of conventional electron optical lenses.

803


5. Conclusions Since its appearance on the instrument market at the end of the 1980’s, the LEEM method initializes excellent progress in understanding crystallinic surfaces, including real time observations of processes changing the surface status. Only a couple of tens of these sophisticated and expensive instruments were necessary to do the work. Our data, collected since 1992, revealed the adaptation of the easily accessible conventional SEM device to the Scanning LEEM mode quite feasible and producing a surprisingly rich scope of results. The adaptation has not been introduced to the market yet so the above list of already collected application examples is rather short (see also the brief review [30]). The recent signals from manufacturers of electron microscopes may indicate that things are looking up. Acknowledgements The authors acknowledge support of the Czech Science Foundation under grants no. 102/05/2327 (I.M.) and 202/04/0281 (L.F.). Ms. Z. Pokorná, ISI Brno, made the micrographs in Fig. 8.

References [1] R.A. Schwarzer, Microscopica Acta 84, 51 (1981). [2] A. Recknagel, Zeitschrift für Physik 117, 689 (1941). [3] E. Bauer, Proceedings of the 5th International Congress on Electron Microscopy, S.S. Bresse, jr. ed., Academic Press, New York, 1962, Vol. 1, pp. D11-D12. [4] E. Bauer, Reports on Progress in Physics 57, 895 (1994). [5] I. Müllerová and L. Frank, Advances in Imaging and Electron Physics 128, 309 (2003). [6] L. Reimer, Scanning electron microscopy: Physics of image formation and microanalysis, Springer Series in Optical Sciences Vol. 45 (Springer, Berlin, 1998), chap. no. 3 and 4. [7] H. Seiler, Zeitschrift für Angewandte Physik 22, 249 (1967). [8] Z.J. Ding, Fundamental studies on the interactions of kV electrons with solids for application to electron spectroscopies, Ph.D. Thesis (Osaka University, 1990). [9] H.J. Fitting, E. Schreiber, J.Ch. Kuhr and A. von Czarnowski, Journal of Electron Spectroscopy and Related Phenomena 119, 35 (2001). [10] E. Schreiber, H.J. Fitting, Journal of Electron Spectroscopy and Related Phenomena 124, 25 (2002). [11] R.C. Jaklevic and L.C. Davis, Physical Review B 26, 5391 (1982). [12] S.A. Nepijko, N.N. Sedov and G. Schönhense, Advances in Imaging and Electron Physics 136, 227 (2005). [13] L. Frank and I. Müllerová, Ultramicroscopy 106, 28 (2005). [14] C. Guittard, M. Babout, S. Guittard and M. Bujor, Scanning Electron Microscopy 1981/I, 199. [15] G.F. Rempfer and O.H. Griffith, Ultramicroscopy 47, 35 (1992). [16] P. Buseck, J. Cowley and L. Eyring, High-resolution transmission electron microscopy and associated techniques (Oxford University Press, London, 1988), pp. 269-275. [17] M. Born and E. Wolf, Principles of optics, 7th ed. (Pergamon Press, Oxford, 1999), pp. 590-595. [18] L. Frank, I. Müllerová, K. Faulian and E. Bauer, Scanning 21, 1 (1999). [19] I. Müllerová and M. Lenc, Ultramicroscopy 41, 399 (1992). [20] http://www.smt.zeiss.com. [21] L. Frank and I. Müllerová, Journal of Electron Microscopy 48, 205 (1999). [22] J.E. Barth and P. Kruit, Optik 101, 101 (1996). [23] M. Lenc and I. Müllerová, Ultramicroscopy 41, 411 (1992). [24] M. Lenc and I. Müllerová, Ultramicroscopy 45, 159 (1992). [25] I. Müllerová and L. Frank, Scanning 26, 18 (2004). [26] I. Müllerová and L. Frank, Scanning 15, 193 (1993). [27] I. Vlček, B. Lencová and M. Horáček, Proceedings of the 13th European Microscopy Congress, D. van Dyck and P. Oostveld, eds. (Belgian Society for Microscopy, Liege, 2004) Vol. 1, pp. 337-338. [28] J. Zobačová, M. Zobač, M. Oral, I. Müllerová and L. Frank, Scanning 28, 155 (2006). [29] K. Matsuda, S. Ikeno, I. Müllerová and L. Frank, Journal of Electron Microscopy 54, 109 (2005). [30] L. Frank, I. Müllerová, K. Matsuda and S. Ikeno, Materials Transactions 48, 944 (2007).

804

Very Low Energy Scanning Electron Microscopy - Formatex

Recommend Documents