- Email: [email protected]
Lensless electron reflection microscopy using a coaxial point-source structure
ARTICLE IN PRESS
Ultramicroscopy 106 (2006) 480–485 www.elsevier.com/locate/ultramic
Lensless electron reflection microscopy using a coaxial point-source structure Zoubida Hammadi, Roger Morin CRMC-N1 - CNRS Campus de Luminy, Case 913, F-13288 Marseille cedex 09, France Received 21 September 2005; received in revised form 21 December 2005; accepted 12 January 2006
Abstract A lensless image of the surface of a crystal is obtained by the reflection on this surface of a low-energy electron beam originated from a point source integrated in a coaxial structure. The point source is a sharp field emission tip and a free propagation of reflected electrons results from the shielding of the tip voltage provided by the coaxial structure. Images are obtained for an incidence angle between 3 and 451 and for nA incident currents with a kinetic energy down to 40 V. On silicon surfaces a magnification up to a few thousands and a spatial resolution of 100 nm are demonstrated. r 2006 Elsevier B.V. All rights reserved. PACS: 07.78.+s; 68.65.k Keywords: Electron point source; Field emission; Lensless electron microscopy; Reflection electron microscopy
1. Introduction Surface imaging methods based on elastic scattering of electron beams use various approaches from high-energy beams in grazing-incidence to low-energy beams in normalincidence. Associated instruments, except in the case of simple low-energy electron diffraction designs, use lenses which lead to involved products. Recent experiments [1–3] were carried out in order to get rid of these lenses. The basic idea is to illuminate the surface by the divergent beam originated from an electron point source placed at a microscopic distance from the surface of a sample and to detect the elastically reflected beam on a screen placed at a macroscopic distance from the source (Fig. 1). Essentially this approach is the reflection analogue of the transmission method previously developed in the low-energy point projection microscope [4]. In that later microscope design, the point source is a field emission tip negatively biased and the grounding of the conductive sample provides at one Corresponding author. Tel.: +33 6 62902849; fax: +33 4 91418916. 1
E-mail address: [email protected] (R. Morin). Laboratoire associe´ aux Universite´s d’Aix-Marseille II et III.
0304-3991/$ - see front matter r 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.ultramic.2006.01.001
and the same time the electron extraction field on the tip apex and a field free region for the transmitted electrons. In any attempt to use a similar design for reflection, a problem has to be tided over: reflected electrons come back towards the tip and their trajectories are bent towards the sample surface due to the negative voltage of the emission tip. If only secondary electron detection associated with scanning technique were used to image the surface, the effect of this negative voltage could be limited to a decrease of the detection efficiency [5,6]. But as long as elastic scattered electrons imaging is concerned, the problem cannot be reduced to a detection loss. Previous works already mention this effect without solving the problem. An approach taken by Mizuno is, by assuming a tip-sample geometry, to take into account the field effect on the image. Our approach is to provide an electrostatic shielding of the tip voltage for the reflected electron beam. We first discuss why a micrometer scale control of this shielding is the main experimental challenge for lensless reflection microscopy. First a high enough resolution requires a short tip-to-sample distance and thus electrostatic shielding on this scale. Placed at 10 cm from the tip, an image detector with a 0.1 mm resolution (resolution of a
ARTICLE IN PRESS Z. Hammadi, R. Morin / Ultramicroscopy 106 (2006) 480–485
481
Fig. 2. Reflection from a stepped surface. A disoriented area is a succession of equal width s terraces limited by equal height h steps. This provides overlaps in the reflected images from different part of the surface.
Fig. 1. Principle of projection microscopy using an electron point source. The edge of the sample can be seen either in transmission or in reflection. The arrows shows (not at scale) the electrostatic force on the electrons in the vicinity of the emitter apex and of the sample surface.
channelplate-screen assembly) limits the resolution on the surface to 1 mm for a 1 mm tip-to-sample distance. Thus sub-micrometer resolution on the surface requires 100 mm shielding. But if a high resolution is desirable for any microscope one has to realize that, in lensless reflection microscopy, this is a requirement for any correct imaging for most samples. The reason is that on a millimeter scale, the roughness of a crystal surface raises two imaging problems for this microscopy which aims to image flat surfaces. First are the reflections from disoriented areas in the field of view which overlap on the screen. Second is the intensity decrease which results from the addition of out of phase waves reflected from areas of parallel but random height terraces inside the first Fresnel zone defined on the surface from the tip. A direct consequence from the decrease of the source-to-sample distance is the decrease of the number of terraces in the illuminated area which is defined by the emission cone angle of sharp field emission tips [7] and is about 0.1 rd. Thus, a 10 mm diameter area is illuminated by a tip placed at 100 mm from the surface. For a disoriented area (Fig. 2) the decrease of the tip-to-sample distance d increases the angle a, this area is seen from the source. This reduces the overlap of its image with that of an adjacent flat area. A quantitative criteria is a42 m where m is the disorientation angle (relative to the terrace orientation). Introducing the distance between steps s, the step
height h and the incidence angle i this criteria is dos2 sin2 i=ð2 hÞ for small a values. Therefore for 451 incidence angle and atomic step size (0:3 nm) submicrometer terraces are imaged for do100 mm. Concerning the out of phase wave addition problem, decreasing the tipto-sample distance d decreases the size b of the region of the surface which contributes to constructive interferences at a given point of the screen. b is (ld)1/2/sin i (i.e. the intercept by the surface at an incidence angle i of the beam limited by the first Fresnel zone) where l is the electron wavelength. This means that for a 100 eV electron beam at an incidence i ¼ 451, sub-micrometer resolution is obtained (b0:2 mm) for do100 mm. These considerations show that the use of lensless reflection microscopy requires to operate the tip at a short distance (o100 mm) from the surface and therefore to provide shielding of the tip voltage on this scale. In addition appropriate surface preparation techniques to increase the flatness of the sample up to a distance between steps equal to a fraction of micrometer distance are required. 2. Experimental Our approach to provide such a short distance shielding is in the use of a coaxial structure which integrates the field emission tip. This structure, described in a previous study [8] is made of a sharp etched W tip positioned inside a glass capillary. A metal layer deposited on the outer wall of the capillary constitutes an electrical sheath. The total diameter of the structure can be made as small as 60 mm. By applying a negative voltage between 60 and 200 V to the tip relative to the sheath, nA electron currents can be extracted from the tip. Using this structure as the source of a transmission projection microscope, we showed that the sheath shields the electrostatic influence of a third conductor placed
ARTICLE IN PRESS 482
Z. Hammadi, R. Morin / Ultramicroscopy 106 (2006) 480–485
further than about 100 mm (i.e. the overall diameter of the structure) in front of the tip. In the present study, we take advantage of this electrostatic shielding in order to get reflection images from flat samples. Typical scanning electron micrographs of the gun are shown in Fig. 3. The previously reported fabrication process is somewhat changed. Quartz instead of glass capillaries are used. While more difficult to pull than glass, quartz is a better insulator: a stable 200 GO resistance is measured between the tip and the sheath for 10 mm long structures with 10 mm thick capillary walls. Thus using quartz improves the electron emission stability. A high inner to outer diameter ratio of tubes results from the quartz pulling process. Changing from glass to quartz thus leads to smaller diameter (down to 50 mm) structures. Another change is the bevelling of the exit face of the capillary by polishing as shown in Fig. 3 and its plating with a metal layer which improves the electrostatic shielding by decreasing the tip-to-sheath distance. Together with the reduction of the structure diameter this makes possible a closer approach of the tip towards the sample surface for grazing incidence illumination. The experimental arrangement is shown in Fig. 4. The structure and the sample are fixed on a round plate which rotates in the horizontal plane. This rotation brings either the transmitted beam or the reflected beam onto a channel plate screen assembly placed 10 cm from the tip. In order to change the electron beam incidence on the sample surface, an electrical motor rotates the sample also in the horizontal plane but relative to the plate. In this way, the mean incidence angle can be controlled from a few to 451. The tip-to-sample distance is controlled using a scanning tunnelling microscope head (‘‘Micro STM’’ from Omicron) moving on a plane made of a Si wafer fixed on the plate. The original tip holder is modified in order to hold the coaxial structure which is kept in the horizontal plane. CCD video camera and recorder are used for video imaging and recording of the images formed on the screen. High dynamics and resolution reflection images are acquired using a digital 3 Mpixel 12 bits camera. This dynamics is necessary in order to distinguish details in the
intense background. The whole projector system is placed inside an ultra-high-vacuum ion-pumped chamber. The samples are cleaned by Joule heating but no in situ characterization of the sample surface is made. Different samples were used for various incidence angles i ranging from a few to 451 and for different tip-to-sample distances i.e. for different magnifications. Along the vertical axis the magnification in reflection is identical to that one in transmission and is given by the ratio of the tipto-screen distance D divided by the tip-to-sample distance d. The horizontal magnification is dependent on the incidence. Small magnification reflection images of any part of the sample can be obtained after the tip-to-sample distance is adjusted through the projection image of an edge of the sample followed by a displacement of the tip roughly parallel to the sample surface. High magnification
Fig. 4. Tip and sample experimental arrangement. The sample holder rotates relative to the plate using the motor. The tip moves on the plate relative to the sample using the STM head. The plate rotates around the feedthrough axis in order to observe either transmitted or reflected patterns on the channel plate (not shown).
Fig. 3. Scanning electron micrograph of the gun. The capillary is made of quartz and its outer wall is covered by a conductive gold layer. The exit face of the capillary is eventually bevelled by polishing and covered by a gold layer. This makes possible to approach the tilted tip close to the surface.
ARTICLE IN PRESS Z. Hammadi, R. Morin / Ultramicroscopy 106 (2006) 480–485
images corresponding to d smaller than 1 mm cannot be easily obtained by this method except in the vicinity of the sample edges. For tip positions far from the edges, the blind displacement of the tip often leads to a mechanical contact between the capillary and the surface which may lead to the destruction of the source structure. 3. Results and discussion For small incidence angles the reflected intensity is high. Projection and reflection patterns can be simultaneously observed for the same voltage as shown in Fig. 5. This image which is obtained with a (1 0 0) oriented Si sample illustrates some behaviours of the technique. The trapezoidal shape of the projection pattern on the left comes from the incidence angle i of the sample which is rectangular. A simple calculation shows that i is directly related to the
Fig. 5. Transmission and reflection images of a (1 0 0) oriented Si rectangular sample. The grazing incidence makes it possible to see both images on the screen at the same intensity. The nonrectangular transmission image ðba0Þ is related to the incidence which is found to be 4.71. The bright region on the right is the reflection pattern. Its left border corresponds to reflection parallel to the surface sample.
483
opening angle b (Fig. 5) of this trapezoidal projection by: tg i ¼ H=ð2D tg bÞ, where H is the length of the image of the vertical edge of the sample (Fig. 5). Experimental measurements of H and tg b are, respectively, 0.83 cm and 0.5 and a projection distance D of 10 cm gives tg i ¼ 0:083 i.e. an incidence angle of 4.71. It might be thought that the dark-to-bright transition line on the right is the reflection image of the vertical edge of the sample. However the angular distance s between the vertical edge and this line is obtained from the distance L (Fig. 5) by: tg s ¼ L=D L is measured to be 0.84 cm which gives tg s # tg i ¼ 0:084. Therefore this line is the intercept of the sample surface plane with the screen and does not provide a reflected image of the edge of the sample. This corresponds to a strong scattering parallel to the surface. Such a scattering was already reported by Eves et al. [6] for normal incidence but was attributed to the repulsion of the tip, an argument which is not supported by the present experiment. Two features also appear in Fig. 5: (i) the reflected intensity pattern presents a much wider opening angle than the opening angle b of the transmission pattern, (ii) two small cusps appear in this reflection pattern at the same heights than those of the corners of the sample. Such a pattern is observed on this Si(1 0 0) rectangular sample as well as for similar incidence angles on other samples like a Si(1 1 1) rectangular sample with mm size grooves parallel to the incidence plane. For large incidence angles the projection pattern is observed at a much lower current than the reflection pattern. Around 301 incidence a typical 100 current change is necessary, which corresponding to about 20% tip voltage. This reflects the small elastic scattering for electrons at large scattering angles [9]. Moreover the reflection pattern is localized at a large angular distance from the projection pattern and therefore both patterns cannot be observed together on the screen as for grazing incidence conditions. The reflection pattern is always composed of vertical stripes i.e. perpendicular to the incidence plane as shown in Fig. 6. A slight curvature
Fig. 6. Stripes normal to the incidence plane observed at large tip to surface distances and large incidence angles (301). The illumination angle is 81. From a–d the stripe width is reduced by rotating the sample relative to the tip. The angular opening is a few degrees wide and often presents a curvature turned towards the sample.
ARTICLE IN PRESS 484
Z. Hammadi, R. Morin / Ultramicroscopy 106 (2006) 480–485
towards the sample side of these stripes is often observed. A reliable measurement of the angular position and angular width of these stripes is difficult because of the limited reliability of the geometrical parameters. Up to three stripes can be observed.
Decreasing the tip to sample distance leads to the transformation of stripes into blurred intermediate images and finally to the emergence of some well defined details. As shown in Fig. 7 the size of these details increases with the decrease of the tip-to-sample distance without any
Fig. 7. From a to d, successive images of a detail inside a stripe when approaching the tip towards the sample. The illumination angle is 121 and the mean incidence angle is 201. From a to d the respective vertical magnifications are estimated to 225, 675, 2700 and 5400 corresponding to a tip to sample distance of 440, 150, 37 and 19 mm.
Fig. 8. Reflection patterns obtained with light (He–Ne laser l ¼ 632 nm) on a piece of channel plate 1 mm size, the incidence plane is horizontal. In b,c and d the divergence of the beam is 81. This a–d sequence is obtained in a wide range of incidence angle. (a) using a parallel beam, Laue zones are along circles. The specular beam is the brightest one; (b) divergent beam, images of the sample are observed in each Bragg spot. The overlap of the reflected images does not occur because of the small size of the sample. With a slightly larger sample, this overlap produces vertical stripes; (c) approaching the source towards the sample. An overlap of the images in the same Laue zone occurs; (d) the source is very close to the sample, an image of the hexagonal lattice (12 mm) of the sample is observed.
ARTICLE IN PRESS Z. Hammadi, R. Morin / Ultramicroscopy 106 (2006) 480–485
strong distortion. This supports real space imaging and a free propagation of reflected electrons. In order to understand these patterns we carried out reflection experiments with light. A point source of light is produced at the focus point of the objective of an optical microscope illuminated by a parallel laser beam. The sample is a 1 mm2 piece of a channel-plate which behaves as a mirror except in holes organized in a two dimensional hexagonal lattice (lattice parameter ¼ 12 mm). The reflection pattern (Fig. 8) shows vertical stripes for large sourceto-sample distances resulting from the overlap of reflection images of the sample from the diffracted beams of each Laue zone. The overlap increases with the sample dimension normal to the incidence plane which is indeed the case in the electron experiments. The precise extension of the stripes depends on the reflection area which is either defined by the emission cone for a large sample or by the sample size for a small sample or by a combination of both parameters for a partially illuminated sample. After intermediate blurred images, approaching the source towards the sample makes it possible to observe the hexagonal hole lattice. This corresponds to the observation shown in Fig. 7 with electrons. We never observed electron images for different Bragg reflections as reported by Zhang et al. These images were obtained for a relatively small magnification ( 200) and high energy (450 eV) on a freshly cleaved GaAs sample. This requires a small illuminated area in order to minimize the overlap of the different Bragg images. Such conditions can only result from a weakly divergent beam and a sample with very large terraces as discussed in the introduction of the present paper. This likely is the origin of their report that ‘‘reasonably good reflection contrast could be obtained for about 30 min after the cleavage’’. In our experiments shown in Fig. 7 we use a sample made of a
485
piece of a Si(1 1 1) wafer flashed at 1150 1C and annealed at 800 1C for 12 h in ultra-high vacuum. It is known that this treatment produces flat surfaces limited by regions where bunching of steps takes place. Therefore for less stringent surface preparation methods than freshly cleaved surfaces, a reflection image of the surface can however be obtained if the tip is close enough to the surface. In such a situation the magnification is a few thousand and the distance between the tip and the sample is less than 100 mm. 4. Conclusion Using a point source of electrons made of a sharp field emission tip integrated in a coaxial structure of an overall diameter of 60 mm, it is possible to obtain a reflection image from the surface of a crystal. The coaxial structure provides a free propagation of elastically reflected electrons and the imaging process does not use any lens. It is thought that the diameter of the structure is the main limitation to routine imaging. The fabrication of similar structures with a diameter of a few micrometer should overcome this limitation. References [1] Xu Zhang, U. Weierstall, J.C.H. Spence, Ultramicroscopy 72 (1998) 67. [2] S. Mizuno, JVST B 19 (5) (2001) 1874. [3] D.C. Joy, B.G. Frost, e-J. Surf. Sci. Nanotech. 2 (2004) 81. [4] W. Stocker, H.-W. Fink, R. Morin, Ultramicroscopy 31 (1989) 379. [5] H.-W. Fink, Phys. Scripta 38 (1988) 260. [6] B.J. Eves, F. Festy, K. Svensson, R.E. Palmer, Appl. Phys. Lett. 77 (2000) 4223. [7] S. Horch, R. Morin, J. Appl. Phys. 74 (1993) 3652. [8] Z. Hammadi, M. Gauch, R. Morin, JVST B 17 (4) (1999) 1390. [9] E. Bauer, Surf. Rev. Lett. 5 (6) (1998) 1275.
Ultramicroscopy 106 (2006) 480–485 www.elsevier.com/locate/ultramic
Lensless electron reflection microscopy using a coaxial point-source structure Zoubida Hammadi, Roger Morin CRMC-N1 - CNRS Campus de Luminy, Case 913, F-13288 Marseille cedex 09, France Received 21 September 2005; received in revised form 21 December 2005; accepted 12 January 2006
Abstract A lensless image of the surface of a crystal is obtained by the reflection on this surface of a low-energy electron beam originated from a point source integrated in a coaxial structure. The point source is a sharp field emission tip and a free propagation of reflected electrons results from the shielding of the tip voltage provided by the coaxial structure. Images are obtained for an incidence angle between 3 and 451 and for nA incident currents with a kinetic energy down to 40 V. On silicon surfaces a magnification up to a few thousands and a spatial resolution of 100 nm are demonstrated. r 2006 Elsevier B.V. All rights reserved. PACS: 07.78.+s; 68.65.k Keywords: Electron point source; Field emission; Lensless electron microscopy; Reflection electron microscopy
1. Introduction Surface imaging methods based on elastic scattering of electron beams use various approaches from high-energy beams in grazing-incidence to low-energy beams in normalincidence. Associated instruments, except in the case of simple low-energy electron diffraction designs, use lenses which lead to involved products. Recent experiments [1–3] were carried out in order to get rid of these lenses. The basic idea is to illuminate the surface by the divergent beam originated from an electron point source placed at a microscopic distance from the surface of a sample and to detect the elastically reflected beam on a screen placed at a macroscopic distance from the source (Fig. 1). Essentially this approach is the reflection analogue of the transmission method previously developed in the low-energy point projection microscope [4]. In that later microscope design, the point source is a field emission tip negatively biased and the grounding of the conductive sample provides at one Corresponding author. Tel.: +33 6 62902849; fax: +33 4 91418916. 1
E-mail address: [email protected] (R. Morin). Laboratoire associe´ aux Universite´s d’Aix-Marseille II et III.
0304-3991/$ - see front matter r 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.ultramic.2006.01.001
and the same time the electron extraction field on the tip apex and a field free region for the transmitted electrons. In any attempt to use a similar design for reflection, a problem has to be tided over: reflected electrons come back towards the tip and their trajectories are bent towards the sample surface due to the negative voltage of the emission tip. If only secondary electron detection associated with scanning technique were used to image the surface, the effect of this negative voltage could be limited to a decrease of the detection efficiency [5,6]. But as long as elastic scattered electrons imaging is concerned, the problem cannot be reduced to a detection loss. Previous works already mention this effect without solving the problem. An approach taken by Mizuno is, by assuming a tip-sample geometry, to take into account the field effect on the image. Our approach is to provide an electrostatic shielding of the tip voltage for the reflected electron beam. We first discuss why a micrometer scale control of this shielding is the main experimental challenge for lensless reflection microscopy. First a high enough resolution requires a short tip-to-sample distance and thus electrostatic shielding on this scale. Placed at 10 cm from the tip, an image detector with a 0.1 mm resolution (resolution of a
ARTICLE IN PRESS Z. Hammadi, R. Morin / Ultramicroscopy 106 (2006) 480–485
481
Fig. 2. Reflection from a stepped surface. A disoriented area is a succession of equal width s terraces limited by equal height h steps. This provides overlaps in the reflected images from different part of the surface.
Fig. 1. Principle of projection microscopy using an electron point source. The edge of the sample can be seen either in transmission or in reflection. The arrows shows (not at scale) the electrostatic force on the electrons in the vicinity of the emitter apex and of the sample surface.
channelplate-screen assembly) limits the resolution on the surface to 1 mm for a 1 mm tip-to-sample distance. Thus sub-micrometer resolution on the surface requires 100 mm shielding. But if a high resolution is desirable for any microscope one has to realize that, in lensless reflection microscopy, this is a requirement for any correct imaging for most samples. The reason is that on a millimeter scale, the roughness of a crystal surface raises two imaging problems for this microscopy which aims to image flat surfaces. First are the reflections from disoriented areas in the field of view which overlap on the screen. Second is the intensity decrease which results from the addition of out of phase waves reflected from areas of parallel but random height terraces inside the first Fresnel zone defined on the surface from the tip. A direct consequence from the decrease of the source-to-sample distance is the decrease of the number of terraces in the illuminated area which is defined by the emission cone angle of sharp field emission tips [7] and is about 0.1 rd. Thus, a 10 mm diameter area is illuminated by a tip placed at 100 mm from the surface. For a disoriented area (Fig. 2) the decrease of the tip-to-sample distance d increases the angle a, this area is seen from the source. This reduces the overlap of its image with that of an adjacent flat area. A quantitative criteria is a42 m where m is the disorientation angle (relative to the terrace orientation). Introducing the distance between steps s, the step
height h and the incidence angle i this criteria is dos2 sin2 i=ð2 hÞ for small a values. Therefore for 451 incidence angle and atomic step size (0:3 nm) submicrometer terraces are imaged for do100 mm. Concerning the out of phase wave addition problem, decreasing the tipto-sample distance d decreases the size b of the region of the surface which contributes to constructive interferences at a given point of the screen. b is (ld)1/2/sin i (i.e. the intercept by the surface at an incidence angle i of the beam limited by the first Fresnel zone) where l is the electron wavelength. This means that for a 100 eV electron beam at an incidence i ¼ 451, sub-micrometer resolution is obtained (b0:2 mm) for do100 mm. These considerations show that the use of lensless reflection microscopy requires to operate the tip at a short distance (o100 mm) from the surface and therefore to provide shielding of the tip voltage on this scale. In addition appropriate surface preparation techniques to increase the flatness of the sample up to a distance between steps equal to a fraction of micrometer distance are required. 2. Experimental Our approach to provide such a short distance shielding is in the use of a coaxial structure which integrates the field emission tip. This structure, described in a previous study [8] is made of a sharp etched W tip positioned inside a glass capillary. A metal layer deposited on the outer wall of the capillary constitutes an electrical sheath. The total diameter of the structure can be made as small as 60 mm. By applying a negative voltage between 60 and 200 V to the tip relative to the sheath, nA electron currents can be extracted from the tip. Using this structure as the source of a transmission projection microscope, we showed that the sheath shields the electrostatic influence of a third conductor placed
ARTICLE IN PRESS 482
Z. Hammadi, R. Morin / Ultramicroscopy 106 (2006) 480–485
further than about 100 mm (i.e. the overall diameter of the structure) in front of the tip. In the present study, we take advantage of this electrostatic shielding in order to get reflection images from flat samples. Typical scanning electron micrographs of the gun are shown in Fig. 3. The previously reported fabrication process is somewhat changed. Quartz instead of glass capillaries are used. While more difficult to pull than glass, quartz is a better insulator: a stable 200 GO resistance is measured between the tip and the sheath for 10 mm long structures with 10 mm thick capillary walls. Thus using quartz improves the electron emission stability. A high inner to outer diameter ratio of tubes results from the quartz pulling process. Changing from glass to quartz thus leads to smaller diameter (down to 50 mm) structures. Another change is the bevelling of the exit face of the capillary by polishing as shown in Fig. 3 and its plating with a metal layer which improves the electrostatic shielding by decreasing the tip-to-sheath distance. Together with the reduction of the structure diameter this makes possible a closer approach of the tip towards the sample surface for grazing incidence illumination. The experimental arrangement is shown in Fig. 4. The structure and the sample are fixed on a round plate which rotates in the horizontal plane. This rotation brings either the transmitted beam or the reflected beam onto a channel plate screen assembly placed 10 cm from the tip. In order to change the electron beam incidence on the sample surface, an electrical motor rotates the sample also in the horizontal plane but relative to the plate. In this way, the mean incidence angle can be controlled from a few to 451. The tip-to-sample distance is controlled using a scanning tunnelling microscope head (‘‘Micro STM’’ from Omicron) moving on a plane made of a Si wafer fixed on the plate. The original tip holder is modified in order to hold the coaxial structure which is kept in the horizontal plane. CCD video camera and recorder are used for video imaging and recording of the images formed on the screen. High dynamics and resolution reflection images are acquired using a digital 3 Mpixel 12 bits camera. This dynamics is necessary in order to distinguish details in the
intense background. The whole projector system is placed inside an ultra-high-vacuum ion-pumped chamber. The samples are cleaned by Joule heating but no in situ characterization of the sample surface is made. Different samples were used for various incidence angles i ranging from a few to 451 and for different tip-to-sample distances i.e. for different magnifications. Along the vertical axis the magnification in reflection is identical to that one in transmission and is given by the ratio of the tipto-screen distance D divided by the tip-to-sample distance d. The horizontal magnification is dependent on the incidence. Small magnification reflection images of any part of the sample can be obtained after the tip-to-sample distance is adjusted through the projection image of an edge of the sample followed by a displacement of the tip roughly parallel to the sample surface. High magnification
Fig. 4. Tip and sample experimental arrangement. The sample holder rotates relative to the plate using the motor. The tip moves on the plate relative to the sample using the STM head. The plate rotates around the feedthrough axis in order to observe either transmitted or reflected patterns on the channel plate (not shown).
Fig. 3. Scanning electron micrograph of the gun. The capillary is made of quartz and its outer wall is covered by a conductive gold layer. The exit face of the capillary is eventually bevelled by polishing and covered by a gold layer. This makes possible to approach the tilted tip close to the surface.
ARTICLE IN PRESS Z. Hammadi, R. Morin / Ultramicroscopy 106 (2006) 480–485
images corresponding to d smaller than 1 mm cannot be easily obtained by this method except in the vicinity of the sample edges. For tip positions far from the edges, the blind displacement of the tip often leads to a mechanical contact between the capillary and the surface which may lead to the destruction of the source structure. 3. Results and discussion For small incidence angles the reflected intensity is high. Projection and reflection patterns can be simultaneously observed for the same voltage as shown in Fig. 5. This image which is obtained with a (1 0 0) oriented Si sample illustrates some behaviours of the technique. The trapezoidal shape of the projection pattern on the left comes from the incidence angle i of the sample which is rectangular. A simple calculation shows that i is directly related to the
Fig. 5. Transmission and reflection images of a (1 0 0) oriented Si rectangular sample. The grazing incidence makes it possible to see both images on the screen at the same intensity. The nonrectangular transmission image ðba0Þ is related to the incidence which is found to be 4.71. The bright region on the right is the reflection pattern. Its left border corresponds to reflection parallel to the surface sample.
483
opening angle b (Fig. 5) of this trapezoidal projection by: tg i ¼ H=ð2D tg bÞ, where H is the length of the image of the vertical edge of the sample (Fig. 5). Experimental measurements of H and tg b are, respectively, 0.83 cm and 0.5 and a projection distance D of 10 cm gives tg i ¼ 0:083 i.e. an incidence angle of 4.71. It might be thought that the dark-to-bright transition line on the right is the reflection image of the vertical edge of the sample. However the angular distance s between the vertical edge and this line is obtained from the distance L (Fig. 5) by: tg s ¼ L=D L is measured to be 0.84 cm which gives tg s # tg i ¼ 0:084. Therefore this line is the intercept of the sample surface plane with the screen and does not provide a reflected image of the edge of the sample. This corresponds to a strong scattering parallel to the surface. Such a scattering was already reported by Eves et al. [6] for normal incidence but was attributed to the repulsion of the tip, an argument which is not supported by the present experiment. Two features also appear in Fig. 5: (i) the reflected intensity pattern presents a much wider opening angle than the opening angle b of the transmission pattern, (ii) two small cusps appear in this reflection pattern at the same heights than those of the corners of the sample. Such a pattern is observed on this Si(1 0 0) rectangular sample as well as for similar incidence angles on other samples like a Si(1 1 1) rectangular sample with mm size grooves parallel to the incidence plane. For large incidence angles the projection pattern is observed at a much lower current than the reflection pattern. Around 301 incidence a typical 100 current change is necessary, which corresponding to about 20% tip voltage. This reflects the small elastic scattering for electrons at large scattering angles [9]. Moreover the reflection pattern is localized at a large angular distance from the projection pattern and therefore both patterns cannot be observed together on the screen as for grazing incidence conditions. The reflection pattern is always composed of vertical stripes i.e. perpendicular to the incidence plane as shown in Fig. 6. A slight curvature
Fig. 6. Stripes normal to the incidence plane observed at large tip to surface distances and large incidence angles (301). The illumination angle is 81. From a–d the stripe width is reduced by rotating the sample relative to the tip. The angular opening is a few degrees wide and often presents a curvature turned towards the sample.
ARTICLE IN PRESS 484
Z. Hammadi, R. Morin / Ultramicroscopy 106 (2006) 480–485
towards the sample side of these stripes is often observed. A reliable measurement of the angular position and angular width of these stripes is difficult because of the limited reliability of the geometrical parameters. Up to three stripes can be observed.
Decreasing the tip to sample distance leads to the transformation of stripes into blurred intermediate images and finally to the emergence of some well defined details. As shown in Fig. 7 the size of these details increases with the decrease of the tip-to-sample distance without any
Fig. 7. From a to d, successive images of a detail inside a stripe when approaching the tip towards the sample. The illumination angle is 121 and the mean incidence angle is 201. From a to d the respective vertical magnifications are estimated to 225, 675, 2700 and 5400 corresponding to a tip to sample distance of 440, 150, 37 and 19 mm.
Fig. 8. Reflection patterns obtained with light (He–Ne laser l ¼ 632 nm) on a piece of channel plate 1 mm size, the incidence plane is horizontal. In b,c and d the divergence of the beam is 81. This a–d sequence is obtained in a wide range of incidence angle. (a) using a parallel beam, Laue zones are along circles. The specular beam is the brightest one; (b) divergent beam, images of the sample are observed in each Bragg spot. The overlap of the reflected images does not occur because of the small size of the sample. With a slightly larger sample, this overlap produces vertical stripes; (c) approaching the source towards the sample. An overlap of the images in the same Laue zone occurs; (d) the source is very close to the sample, an image of the hexagonal lattice (12 mm) of the sample is observed.
ARTICLE IN PRESS Z. Hammadi, R. Morin / Ultramicroscopy 106 (2006) 480–485
strong distortion. This supports real space imaging and a free propagation of reflected electrons. In order to understand these patterns we carried out reflection experiments with light. A point source of light is produced at the focus point of the objective of an optical microscope illuminated by a parallel laser beam. The sample is a 1 mm2 piece of a channel-plate which behaves as a mirror except in holes organized in a two dimensional hexagonal lattice (lattice parameter ¼ 12 mm). The reflection pattern (Fig. 8) shows vertical stripes for large sourceto-sample distances resulting from the overlap of reflection images of the sample from the diffracted beams of each Laue zone. The overlap increases with the sample dimension normal to the incidence plane which is indeed the case in the electron experiments. The precise extension of the stripes depends on the reflection area which is either defined by the emission cone for a large sample or by the sample size for a small sample or by a combination of both parameters for a partially illuminated sample. After intermediate blurred images, approaching the source towards the sample makes it possible to observe the hexagonal hole lattice. This corresponds to the observation shown in Fig. 7 with electrons. We never observed electron images for different Bragg reflections as reported by Zhang et al. These images were obtained for a relatively small magnification ( 200) and high energy (450 eV) on a freshly cleaved GaAs sample. This requires a small illuminated area in order to minimize the overlap of the different Bragg images. Such conditions can only result from a weakly divergent beam and a sample with very large terraces as discussed in the introduction of the present paper. This likely is the origin of their report that ‘‘reasonably good reflection contrast could be obtained for about 30 min after the cleavage’’. In our experiments shown in Fig. 7 we use a sample made of a
485
piece of a Si(1 1 1) wafer flashed at 1150 1C and annealed at 800 1C for 12 h in ultra-high vacuum. It is known that this treatment produces flat surfaces limited by regions where bunching of steps takes place. Therefore for less stringent surface preparation methods than freshly cleaved surfaces, a reflection image of the surface can however be obtained if the tip is close enough to the surface. In such a situation the magnification is a few thousand and the distance between the tip and the sample is less than 100 mm. 4. Conclusion Using a point source of electrons made of a sharp field emission tip integrated in a coaxial structure of an overall diameter of 60 mm, it is possible to obtain a reflection image from the surface of a crystal. The coaxial structure provides a free propagation of elastically reflected electrons and the imaging process does not use any lens. It is thought that the diameter of the structure is the main limitation to routine imaging. The fabrication of similar structures with a diameter of a few micrometer should overcome this limitation. References [1] Xu Zhang, U. Weierstall, J.C.H. Spence, Ultramicroscopy 72 (1998) 67. [2] S. Mizuno, JVST B 19 (5) (2001) 1874. [3] D.C. Joy, B.G. Frost, e-J. Surf. Sci. Nanotech. 2 (2004) 81. [4] W. Stocker, H.-W. Fink, R. Morin, Ultramicroscopy 31 (1989) 379. [5] H.-W. Fink, Phys. Scripta 38 (1988) 260. [6] B.J. Eves, F. Festy, K. Svensson, R.E. Palmer, Appl. Phys. Lett. 77 (2000) 4223. [7] S. Horch, R. Morin, J. Appl. Phys. 74 (1993) 3652. [8] Z. Hammadi, M. Gauch, R. Morin, JVST B 17 (4) (1999) 1390. [9] E. Bauer, Surf. Rev. Lett. 5 (6) (1998) 1275.