Spectromicroscopy and internal photoemission spectroscopy of semiconductor interfaces

Pergamon

Progress in Surface Science, Vol. 56, No. 4, pp.31 l-346, 1997 0 1998 Elsevier Science Ltd All rights reserved. Printed in Great Britain 0079-6816198 $19.00

PII: SOO79-6816(98)00003-3

SPECTROMICROSCOPY AND INTERNAL PHOTOEMISSION SPECTROSCOPY OF SEMICONDUCTOR INTERFACES G. MARGARITONDO Institut de Physique AppliquCe, Ecole Polytechnique F&d&ale, CH-1015 Lausanne, Switzerland and Sincrotrone Trieste SCpA, Trieste, Italy

Abstract For decades, techniques based on syncbrotron light sources have played a central role in solid interface research. This role has been recently enhanced by two factors: the commissioning of the third generation of sources, characterized by unprecedented levels of brighmess, and the fist utilization cases of another class of photon sources related to syncbrotron facilities, the free electron lasers @EL’s). This review will first present some relevant examples of how the new facilities are changing the scene of interface research, most notably in the domain of spectromicroscopy. We will specifically illustrate how the crucial problem of the lateral fluctuations of interface properties is being attacked with both syncbrotron-light and FEL techniques. Then, we will argue that the present applications are only marginally exploiting the amazing capabilities of the new sources. The main case to illustrate this point is coherence-sharpened x-ray imaging, a very promising and spectacular technique developed for medical radiology, which could find extremely interesting applications in interface research.

Contents 1. Introduction: New Photon Sources A. New performance levels 2. Surface and Interface Spectromicroscopy A. Scanning photoelectron spectromicroscopy B. Interface chemistry lateral fluctuations

312

3. FEL Techniques A. SNOM-FEL techniques 4. Future: Coherent Radiation and Other Avenues A. Coherence-enhanced imaging B. Other avenues: ultrahigh resolution, time resolution Acknowledgments References

323

311

312 315 315 320 326 333 336 342 343 343

G. Margaritondo

312

Acronyms BEEM ESCA ESRF EIPE MAXIMUM MEPHISTO

SNOM

1. Introduction:

Advanced Light Source Ballistic Electron Emission Microscope Electron Spectroscopy for Chemical Analysis European Synchrotron Radiation Facility Free Electron Laser FEL Internal PhotoEmission Multiple Application X-ray IMaging Undulator Microscope Microscope a Emission de PHotoelectrons par Illumination Synchrotronique Type Onduleur (Photoelectron Emission Microscope by Undulator Synchrotron Illumination) Scanning Near-field Optical Microscope

New Photon

Sources

Semiconductor interface research is often based on photon probes such as photoemission and optical spectroscopy.[ 1,2] The progress in this field is strongly correlated to the development of new and advanced photon sources.[2,3] In that respect, the past five years have been quite important. First of all, there was the commissioning of the third generation of synchrotron sources, which is quite powerful for interface studies.[2] Furthermore, the first systematic studies using free-electron lasers @EL’s) were initiated, opening up new domains of semiconductor interface research.[4] Our review briefly describes the present status of this rapidly evolving field. First of all, we discuss the novel characteristics of the most recent photon sources for interface research. Then, we review some of their applications, beginning with laterally-resolved photoemission with high-brightness synchrotron light. The review continues with a discussion of FEL spectroscopy and spectromicroscopy, including very recent applications to interface research. Finally, we briefly discuss one of the most exciting new avenues in synchrotron-based research. This concerns techniques exploiting the high spatial coherence of the new sources. We will specifically discuss the technique of coherence-enhancement of x-ray imaging, and its possible applications to interface research.[5,6]

A. New performance levels (i) Synchrotron light sources. The keyword for describing the third generation of synchrotron sources - whose commissioning initiated in 1992-93 and is still continuing - is “brightness”.[2,3] As we shall, see, this keyword does not give a complete picture of the new machines, and a more complete keyword should be “brightness and coherence”. Let us see then what the keyword “brightness” means, whereas “coherence” will be discussed in the last section.. In a typical semiconductor interface experiment using synchrotron light, the main requirement is to concentrate as much photon flux as possible into a small area. To achieve this objective, one must take

IPES of Semiconductor

Interfaces

313

into account the brightness of the photon beam, which is - roughly speaking - related to the flux divided by the beam transverse size and divided by its angular divergence.[3] The phase-space volume conservation requires the brightness to be conserved along and ideal (lossless) optical system.[3] Therefore, focusing a large flux into a small area is easier - and often only possible - if the source itself has high brightness. Up until the first planning for the third-generation synchrotron sources, the merit figure of choice was the flux.[3] Not much attention was devoted to geometric characteristics such as size and angular divergence. On the contrary, a major effort was dedicated to increasing the electron beam current in the storage ring at the heart of each synchrotron facility, thereby increasing the flux. The flux improvement more or less saturated in the late 1980’s, and the main attention shifted to brightness. Figure 1 shows the historical evolution of this parameter. Its increase since the advent of synchrotron light has been nothing less than spectacular. This is mainly due [3] to two factors: first, the improvement in the control of the geometry of the circulating electron beam in the ring; second, the use of specialized photon-emitting insertion devices called “undulators”. The very rapid performance improvement of Fig. 1 has one clear implication: the creativity of the experimentalists is strained in order to fully exploit the new sources. Veterans of the field like this author feel that the new sources are still underutilized. Nevertheless, their impact on interface research is already spectacular, as the examples of the present review will try to prove.

(ii) Relativity at work. The secret [2,3] of the third-generation synchrotron sources is special relativity. This can be schematically understood with Fig. 2. Seen from the reference frame of the electron circulating in the storage ring at a speed cp = c, the periodic transverse magnetic field of the undulator becomes the combination of a transverse magnetic field plus a transverse electric field -- and electromagnetic wave. The wavelength of this wave is the Lorentz-coneacted period of the undulator, =L/y along the undulator axis. The wave is scattered by the electron, giving rise to a secondary emission of the same wavelength. Seen in the laboratory frame, this wavelength is Doppler-shifted by -2y, becoming =L/2+. Since the typical y-factor for a storage ring is of the order of several thousands, the macroscopic undulator period is thus converted into a microscopic-wavelength x-ray radiation. Furthermore, the emitted bandpass is quite narrow because of the diffraction-grating-like effect of the multiple undulator periods.[3] Thus, the emitted brightness is concentrated into a small spectral range rather than over the conventional broad band of synchrotron light.[3] The result is a very high spectral brightness as shown in Fig. 1.

(iii) Free electron lasers. The mechanism described in the previous section is a type of spontaneous stochastic emission of photons. One can also have a stimulated-emission phenomenon, which is exploited by the FEL’s.[3,7] The emitted wavelength is still =L/2+, and the “lasing” action exploits as the optical amplifying medium the bunch of circulating electrons interacting with the insertion device. The electron acceleration can be provided by a storage ring or by a different type of device.

G. Margaritondo

314

A synchrotron light source is typically used as an emitter of x-rays. On the contrary, almost all of the operating FEZ’s emit in the infrared. This is mainly due to two factors. First of all, the optical gain decreases as the wavelength becomes shorter.[3] Second, for ultraviolet and even more for x-ray radiation it is difficult or impossible to build an optical cavity because of the lack of mirrors. Thus, the FEL must work as the equivalent of a super-radiant laser, achieving high enough optical gain to produce a lasing effect without the multiple paths given by an optical cavity.

1018

I 00

I

1940

I

I

1980

Year Fig. 1. Schematic historical trend towards increasing brightness of the x-ray sources.

I

IPES of Semiconductor

Interfaces

315

These limitations notwithstanding, the EEL’s are very interesting devices. They produce highly bright infrared radiation that can be easily tuned over a broad spectral range.[7] This range includes the forbidden gaps of many of the important semiconductors, and the semiconductor interface energy ban&s that are related to such gaps. Therefore, FEL’s are potentially very interesting for semiconductor interface research, as the results presented in this review clearly demonstrate.

2. Surface and Interface

Spectromicroscopy

HOW can the new performance levels discussed in the previous section be exploited for

semiconductor interface research? The first response is quite obvious: by attacking the crucial and quite unexplored problem of the lateral variations of interface properties. Consider for example a “Schottky barrier”. Even after eliminating the old, simplistic ideas that attributed the barrier formation process to purely electrostatic phenomena with no interface chemistry, no interdiffusion etc., the notion of “Schottky barrier” may still be a pseudoconcept. In fact, this notion implies no barrier fluctuations in the plane of the metal-semiconductor interface. However, if fluctuations do occur, the notion must be completely revised. This implies that the weakest-barrier interface areas dominate the transport properties, determining the dynamic response of the corresponding devices, the forward resistivity etc. It is quite clear, therefore, that possible barrier fluctuations in the plane of a metal-semiconductor are a key issue for semiconductor interface science. This issue is very difficult to attack from an experimental point of view. Transport measurements are typically blind to local barrier fluctuations. A similar limitation is present for conventional spectroscopic measurements of barrier heights. In recent years, three techniques emerged as solutions of this problem:

BEEM

(Ballistic

Electron Emission Microscope),[8]

scanning photoelectron

spectromicroscopy [9] and near-field internal photoemission spectromicroscopy.[7,10,1 l] BEEM is outside the scope of this review; we will discuss scanning photoelectron spectromicroscopy in this section and near-field internal photoemission spectromicroscopy with an FEL in Sect. 3. A. Scanning photoelectron spectromicroscopy High-brightness x-ray sources can be exploited to remove one of the major limitations of photoemission spectroscopy: the lack of lateral resolution.[2,3,9]

Before seeing how this can be

accomplished, let us briefly review the conventional photoemission spectroscopy way to measure interface barriers such as Schottky barriers or heterojunction band discontinuities.[3] The photoelectric effect extracts photoelectrons from a solid system, increasing the ground-state energy of electrons by an amount equal to the photon energy. Therefore, measurements of the photoelectron kinetic energy can be converted into measurementsof the electron energies in the solid. A p-type Schottky barrier is by definition the energy distance between the Fermi level and the top of the semiconductor valence band at the inferface. One can thus measure with photoelectron spectroscopy the position of the valence-band edge for a clean semiconductor surface, then deposit a thin metal film on top of it, correct the previous

316

G. Margaritondo

measurement for possible metal-induced changes in the semiconductor band bending, and then measure the Fermi edge position, also with photoemission. Over the years, this basic method has been refined becoming quite sophisticated and reliable.[3]

v=c -e L + L/y

Fig. 2. Scheme [2,3] of an undulator: (a) the undulator consists of a periodic series of magnets; (b) the period seen by the electrons moving in the storage ring at almost the speed of light is Lorentzcontracted; (c) the electrons are forced by the undulator to small oscillations, and as a consequence they emit electromagnetic waves synchrotron light - of wavelength equal to the Lorentz-contracted undulator period. Seen from the laboratory frame, this wavelength is Doppler shifted to the region of the x-rays.

IPES of Semiconductor

Interfaces

317

The lack of lateral resolution, however, still constitutes a crucial limitation. In a typical measurement, the probed area is of the order of a few tenths of a millimeter. All effects inside this area are averaged, and one cannot detect barrier fluctuations over a shorter scale. HOW can one overcome this limitation? The answer is: by focusing x-rays.[9] This is not an easy task, since no conventional transmission optics exists for x-rays and reflection is quite ineffective. However, the opportunities opened up by the new, high-brightness synchrotron sources have stimulated a major effort in this domain, and led to the development of several x-ray-focusing devices. We would like to mention, in particular, the Schwarzschild objective, which is the coaxial combination of a convex and of a concave mirror.[ 121 Its operation in the soft-x-rays requires high reflectivity at near-normal incidence. This was achieved by extending to spherical surfaces the technique of reflection enhancement by multilayer coating. Another effective device for soft-x-ray focusing is the transmission Fresnel zone plate.[2,3,9] This type of device is fabricated with advanced electron-beam lithography, and can reach focusing levels of a few hundred angstroms. Both Schwarzschild objectives and Fresnel zone plates have been successfully used for photoelectron spectromicroscopy, including laterally-resolved investigations of semiconductor interface barriers.[l3-151 Figure 3 shows the rough scheme of a Schwarzschild-objective spectromicroscope. The latest device of this kind mounted in the Scanning Spectromicroscopy beamline [ 161recently commissioned on the Elettra storage ring in Trieste by Michele Bertolo, Stefano Fontana and their co-workers, in collaboration with Cerrina and coworkers of the University of Wisconsin-Madison and with the Ecole Polytechnique Fed&ale de Lausanne. The instrument is the most recent improvement of the first scanning spectromicroscope, the “MAXIMUM” device jointly developed at the Wisconsin Synchrotron Radiation Center by Cerrina and co-workers and by scientists from several other institutions, currently operating on the Berkeley Advanced Light Source (ALS) (the Elettra instrument is also known as “SuperMAXIMUM”). As one can see in Fig. 3, the device can take spectra from microscopic areas (corresponding to the focused photon beam). It can also work at constant photoelectron energy, and take intensity maps by scanning the sample with respect to the photon beam. Such images carry information on the electronic and chemical structure of the surface. The Scanning Spectromicroscopy-SuperMAXIMUM system will be used to expand early results produced by MAXIMUM, which clearly revealed interface barrier fluctuations.[l3,14] Consider for example the results of Fig. 4, concerning the heterojunction between a GaSe substrate and a thin Ge overlayer.[ 141The explored issue here is that of lateral fluctuations of the band lineup between the two sides of the interface -- and the corresponding fluctuations in the valence and conduction band d&continuities that accommodate the difference between the two forbidden gaps.1171 The results of Fig. 4 are quite straightforward: first of all, photoelectrons from the substrate Ga and Se core-levels are found at different energies when the measured spot position is changed on the interface plane. On the contrary, no change is observed for the overlayer Ge core-level peak. Thus, the

318

G. Margaritondo

band lineup must change from place to place, and the same conclusion is valid for the band discontinuities. The same approach enabled MAXIMUM to discover similar phenomena for the clean-surface band bending [ 181 and for metal-semiconductor barriers.[ 131As a matter of fact, barrier fluctuations appear from these experiments to be more a rule than an exceptional event.

Scanning Stage

Display Fig. 3. Scheme [12] of a Schwarzschild-objective scanning spectromicroscope.

photoemission

IPES of Semiconductor

GaSe+O.J ML Ge Ga3d

I 5

a

I

1

I

70 70.5 71 71.5 72 7 5 Kinetic energy (eV)

58

Interfaces

GaSe + 0.5 ML Ge Se3d

-AL O.l2(5)eV

I

I

I

I

I

35 35.5 36 36.5 37 3' Kinetic energy (eV)

59 60 61 62 63 Kinetic energy (eV)

64

Fig. 4. Spatially-resolved photoemission spectra from two different small spots of a GaSe-Ge interface (Ref. [ 143), revealing fluctuations in the band lineup of the two sides of the heterojunction.

320

G. Margaritondo

B. Interface chemistry lateral fluctuations Interface barrier are quite important, but not the only interface features whose fluctuations could play a crucial role in the corresponding devices. Interface chemistry and its fluctuations are also very important. Devices like MAXIMUM and Scanning Spectromicroscopy-SuperMAXIMUM and others in the same class can probe the local chemistry by detecting core levels and their chemical shifts. In fact, they have been used to detect chemistry fluctuations for semiconductor interfaces. An example is shown in Fig. 5, which illustrates core-level spectra taken [15] on the GaSe-Ge interface with the Fresnel zone plate spectromicroscope ESCA Microscopy, developed on Elettra by Maya Kiskinova and her coworkers. The two spectra were taken on two different small areas of the same interface. Quite clearly, the core-level manifold changes from place to place, revealing different chemically reacted phases. This result is quite relevant since it virtually destroys the notion of “Schottky interface”. One must be careful with semantics: a “Schottky barrier” is the energy barrier at a metal-semiconductor interface, no matter what is its nature. A “Schottky interface” is a metal-semiconductor interface whose Schottky barrier height can be justified by the simple model developed in the 1930’s by Schottky.[l9,20] This model excludes chemical interactions, thus it is limited to nonreactive interfaces. The term “Schottky interface” is also applied to semiconductor heterojunctions [ 19,201 whose interface barrier heights are justified by the Anderson model, which is the heterojunction-equivalent of the Schottky model.]171

Over the years, the examples of true “Schottky interfaces” have become scarcer and scarcer, as more sophisticated probes were able to detect previously unobserved interface chemical interactions. However, several GaSe-based Schottky barriers and heterojunctions - including GaSe-Ge - were still believed to be true “Schottky interfaces”.[l9,20] The results of Fig. 5 contradict this belief: when chemical analysis is performed on a microscopic scale, chemical reactions become quite evident.] 1S] It is thus likely that also this family will be removed from the list of examples of “Schottky interfaces”, leaving the list empty. Some degree of chemical interaction seems to always occur at the interface, and is likely to influence the interface barriers as well as their lateral fluctuations. (i) Electron-optics imaging techniques. Scanning-focusing instruments like those previously discussed are powerful and rather sophisticated. For a bit less sophisticated but much quicker chemical analysis one can use instead devices in the general class of electron-optics imaging.[21,22] Basically, these devices do not rely on photon-beam focusing to achieve high lateral resolution: the unfocused or partially focused beam covers a rather large sample area. On the other hand, the emitted photoelectrons are processed by an electron-microscope-like optical system, thereby achieving high lateral resolution. These instruments can yield photoelectron intensity maps (images) at a fixed photon energy. They are also typically used to take photoelectron intensity scans as a function of the photon energy, which correspond to the local x-ray absorption spectra.[21,22] Core-level absorption edges in these spectra can be used to reveal the presence of the corresponding elements and their chemical status. This last analysis does not reach the ultimate

IPES of Semiconductor

321

interfaces

sophistication levels of core-level photoemission, but it can quickly yield very valuable chemical information.[2 1,221 This approach is widely used to investigate the chemical properties of life-science specimens, but can also be used for chemical analysis. Figure 6 shows the scheme of one of the best instruments of this class, Gelsomina De Stasio’s MEPHISTO spectromicroscope. Figure 7 shows the same instrument in a transmission spectroscopy configuration, also suitable for microchemical analysis.[23] Figure 8 shows some interesting results of this last approach.[23]

GaSe-Ge+Ge Ge3d

l

l

I i

I

I

462 460 458 Kinetic Energy (eV)

Fig. 5. Spatially-resolved Ge3d photoemission spectra from two different small spots of a GaSe-Ge interface (Ref. [ 15]), revealing fluctuations in the local chemical composition and the presence of reacted species.

322

G. Margaritondo

computer and mmonitor

onodwomator

Fig. 6 Scheme of the electron-optics imaging photoelectron spectromicroscope MEPHI!3TO, developed by G.De St&o et al$23].

Specifically, we see a series of images taken at several different photon energies in the region of the Si2p absorption edge. The change in contrast reveals the silicon-related portions of the specimen. Such & experimental approach is worth considering not only because of its speed (the detection can occur in real time, with images observed with a video system) but also for other interesting features. For example, its ultimate lateral resolution. Quite recently, an instrument of this class developed by Ernst Bauer and coworkers and recently re-commissioned on Elettra [24] was able to reach a resolution level of the order of 200 A, quite suitable for microchemical analysis of mesoscopic semiconductorrelated structures.

IPES of Semiconductor

Interfaces

323

3. FEL Techniques We will now move from high-photon energy synchrotron light sources to infrared FEL sources. Sources of this kind have been operated for many years, but their practical applications are still somewhat scarce. The fiit positive tests [4,7,10] show that this is certainly not due to lack of interest or to some great difficulty of implementation. The most likely explanation is insufficient knowledge of their great potentiality. This is unfortunate, since the few existing examples of FEL-based experiments are very interesting and encouraging. This is true, in particular, for experiments on semiconductor interfaces. The oldest program in this field was implemented at the Vanderbilt Free Electron Laser center,[4,7,10] and resulted in the development of the internal-photoemission technique known as “FELIPE” (FEL-Internal PhotoEmission).

Si wafer frame

ob@ive

phosphor screen /

projective lens

monochromatic x-ray beam Si micro- / structures

video camera

E

Si nit&e membrane

‘cathode

Fig. 7. Scheme of MEPHISTO

\

microchannel plates

in the transmission mode of

spectromicroscopy.

The basic philosophy of this technique [lo] is quite simple: the semiconductor interface barrier to be measured is inserted in a photocurrent detecting circuit. Infrared radiation can excite carriers over the barrier thereby producing photocurrent. The barrier height is then simply given by the photocurrent threshold. The approach can be applied both to Schottky barriers and to heterojunction band discontinuities.[4,7,10]

324

92teV

104 eV Fig. 8. Series of micrographs taken (Ref. [23]) with MEPHISTO

in the configuration of the previous figure, showing silicon microstructures over a silicon nitride membrane, taken at different photon energies in the spectral region including the Si2p edges of silicon and silicon nitride. The diameter of each image is 180 micron. Note the change in contrast as a function of the photon energy.

IPES of Semiconductor

Interfaces

325

The role of an FEL source in this approach is to enhance its effectiveness and flexibility by providing tunable and intense radiation over a broad spectral range. In order to appreciate the importance of this role, one must consider the practical problems in the implementation of the FELIPE approach.[4,7,10] In addition to barrier-related threshold, one normally detects several other photocurrent spectral features, and it is essential to distinguish these spurious contributions to those actually related to barriers. This is possible by performing experiments over a rather broad spectral range, and have an overall picture of the different types of spectral contributions to analyze and identify them. A narrow-band approach, on the other hand, can be potentially misleading: one could detect a spurious threshold and attribute it to the barrier of interest, whereas the real barrier-related threshold could be just out of reach. The superior performances of the Vanderbilt University FEL stimulated a few years ago the development of the first experimental program in this area. The program has been in operation for several years, producing routine experimental measurements of semiconductor interface barriers with a few meV accuracy. Figure 9 shows [lo] an example of results from the FELIPE approach.

GaMAdGaAs

1.68

1.7

1.72 1.74 1.76 1.78

1.8

1.82 1.84

Photon Energy [eV] Fig. 9. Example of early FELIPE spectrum (Ref.[lO]) taken with the Vanderbilt FEL, with features corresponding to the GaAlAs energy gap and to one of the photocurrent barriers of a GaAsAslGaAs interface.

326

G. Margaritondo

A. SNOM-FEL techniques. This approach, however, was negatively affected by the common problem of many energy barrier measurements: lack of lateral resolution and consequent blindness to lateral barrier fluctuations. This limitation is being removed, and there are some important very recent step in this direction. The general idea is to implement the FELIPE technique in a mode related to the scanning near-field optics microscope (SNOM),[25-271 typically by using an optics fiber with a narrow end aperture to reduce the FEL spot size on the sample. The first element of progress was the implementation of a SNOM-like FELIPE approach without using an FEL.[ 1l] This approach enabled us to measure Pt/n-GaP(OO1)interface photocurrents with high lateral resolution, We observed photocurrent intensity lateral variations that do not match topographic features of shear-force microimages (within their maximum sensitivity of 1 nm), which indicate fluctuations in the electron-hole recombination rate, We also detected lateral variations of the PtGap Schottky barrier of 3-8 meV on a lateral scale of less than 1 pm. This approach can be considered as complementary to the BEEM technique.[8] We achieved [ 1l] high lateral resolution by illuminating the sample with a small light spot, based on the SNOM technique. [25-271 The light source was a tunable At+ pumped Ti:sapphire laser or a solid state laser diode (fixed photon energy at 1,518 eV). The Ti:sapphire laser provided tunable light in the interval 1.3 19-1.53 eV. The overall photon energy resolution was =l meV, as derived from full width at half maximum of luminescence lines in test samples. The maximum power output of the cw Ti:sapphire laser was 100-300 mW. The effective power in the fiber was about 30 mW. For the laser diode, we measured a power of 1 mW at the end of the optical fiber. The light beam was chopper modulated and split in two; one part was detected with a Si avalanche photodiode the other was focused over the metal-covered side of the sample. The tip of the optical fiber was stretched and aluminum coated to concentrate the light beam into a 50-nm-wide pin-hole. Piezoelectric translators implemented the vertical approach up to the near-field condition and the x-y scanning over a rectangular area of size ranging from 8 to 30 pm The experiments were performed with a constant 10 nm tip-to-sample distance: this resulted in a topographic vertical resolution of about 1 nm. The tip-sample distance was controlled by a shear-force feedback system, described in Ref. [26]. The Ti:sapphire laser high power induced partial damage of the metallic coating of the fiber tip, as was observed by scanning electron microscopy. ~hotocurrent microimages were obtained by detecting the photocurrent signal while scanning the beam over the sample. The photocurrent was revealed by a standard lock-in technique. Topographic images were obtained by measuring the shear-force signal with a synchronous detection including a piezoelectric oscillator and a lock-in amplifier.[26-281 From the smallest visible features, we estimated that the lateral resolution was 40-60 nm in topographic images, and 100-200 nm in photocurrent images. These values are better than the classical limit of 3L/2= 400 nm.[29] We note, however, that the photocurrent-image value of 100-200 nm could

IPES of Semiconductor

interfaces

be an overestimate since it was performed in areas with rapidly

varying

327 surface

profiles

which

could

affect the tip-surface optical coupling. We believe that the resolution was also limited by the tip damage and the ultimate resolution achievable with this approach is not yet determined. Also undetermined at the present time is the lateral resolution achievable for deep interfaces which could be affected for example by the spreading of the beam. Figure 10 shows [ll] typical microimages taken at a photon energy of 1.518 eV (laser diode source) In Fig. 10a we see the topographic image over a 16x16 pm area. The analyzed region presents lateral topographic variations on the scale of less than 1 pm. Dark-bright zones of the micrograph reveal profile variations of the sample surface of about 140 nm. Figure lob shows the scanning internalphotoemission micrograph of the same area. A comparison of Figs. 10a and lob shows no one-to-one correspondence between their features: the feature in the marked square is present in both figures, whereas those emphasized by the arrows are only seen in Fig. lob. We first consider the feature in the 6x6 pm marked square: since it is seen in Fig. IOa, this feature must be of topographic origin (and not, for example, due to compositional variations) Its presence in Fig. lob suggests that it is due to a scratch on the substrate. We can, in fact, hypothesize that the scratch results in local variations in the Pt overlayer thickness, which in turn create the contrast in the photocurrent micrograph, Fig lob. Figures 1Oc and 10d show three-dimensional reconstructions of this feature. The comparison between the lateral spread shown here was used to evaluate the ratio between resolution in topographic and photocurrent measurements, obtaining =1:1.5. The resolution is worsened in the photocurrent images by the lateral diffusion of the photoexcited electrons, but for our wavelengths photoexcitation occurs only in the 80 A Pt overlayer resulting in a limited lateral spread. We now consider the features marked by arrows in Fig. lob that are not visible in Fig. 10a. The calibration of the shear-force measurements corresponds to a maximum sensitivity of 1 nm. Thus the absence of the arrow-marked features seemsreal rather than an artifact due to limited sensitivity -- and indicates a non-topographic nature of the features. We explored the nature of the features by means of photoelectron spectromicroscopy and found that they are due to non-topographic chemical inhomogeneities.[30] We hypothesize that the corresponding internal photoemission intensity variations are due to local fluctuations in the electronhole recombination rate. A strong interface recombination was independently confirmed by currentvoltage measurements: the ideality factor was larger than 2 and the saturation current was much higher that the expected value for a barrier of 1.41 eV. Images like Fig. 10 can then be used to detect the “weak” (high recombination rate) areas of the interface and to link them to other local factors, such as microchemistry. Evidence for microchemistry changes was recently obtained on this kind of structures, again by photoemission spectromicroscopy.

328

G. Margaritondo

SNOM images hv = 1.512 eV shear force

Fig. 10. SNOM images from Ref. [ 111, for a Pt-GaP interface. (a) Shear-force topography and (b) corresponding photocurrent microimage of a 16x16 pm2 area in constant distance mode. (c) and (d) three-dimensional reconstructions of the marked areas in Figs. 10a and lob.

IPES of Semiconductor

Interfaces

329

The approach dis~ussd here is also capable of locally exploring the interface energy barriers -Schottky harrier heights in the present case. Typical results are reported in Fig 11. Figure 1la shows the topographic and photocurrent intensity images taken at hv = 1.465 eV, Figure lob shows the square root of the photocurrent versus photon energy (Fowler plot) taken in points A and B of the photocurrent image. The photocurrent threshold position corresponds to the Schottky barrier height. From a linear fit, we obtain Schottky barrier heights of 1.409 and 1.413 eV. The comparison of the thresholds in Fig. 11 clearly indicates that this difference is beyond the experimental uncertainty, which we estimate from the fit not to exceed +l meV. We note, however, that these Schottky barrier lateral variations cannot be the main cause of the large photocurrent-intensity fluctuations on the right-hand side of Fig. 1la. Assuming that thermionic emission is dominant in the current-voltage characteristics, a variation of 4 meV of the Schottky barrier height should induce a photocurrent intensity variation of 15%, whereas the observed variations are much larger, of the order of 250-300%. As already mentioned, these large intensity variations can be explained instead by lateral changes of the recombination rate probably related to microchemical inhomogeneities like those that were recently detected by photoemission spectromicroscopy. Note that inhomogeneities are expected for chemically cleaned Ill-V substrates, as was revealed for example by the BEEM studies of Ref. 1311. In summary, a combination of internal photoemission and SNOM was successfully tested to study the lateral variations of different properties of solid interfaces. We specifically detected topographic effects, energy barrier height variations, and photocurrent intensity variations of nontopographic origin which are primarily attributed to recombination-rate fluctuations. A second, important step was the even more recent implementation of an atomic-force-microscope module for multiple experiments in conjunction with an FEL. The same module will be used in the future for energy barrier measurements with a laterally-resolved FELIPE technique similar to that described in the previous paragraph. The successful tests were conducted by Cricenti et al.1321 at the Vanderbilt FEL, demonstrating that FEL spectroscopy can be implemented in SNOM conditions, by detecting the reflected infrared radiation of the FEL with a SNOM. The tests specifically succeeded in measuring the local optical properties of a buried PtSi/Si interface with lateral resolution well beyond the diffraction limit. Lateral resolution was achieved by using a stretched optical fiber with a very small open edge aperture to collect the reflected radiation. The local FEL-SNOM reflectivity measurements were complemented by shear-force SNOM measurementsrevealing the local topology of the same small area. This made it possible to distinguish topological features from true lateral variations of the optical properties. Figure 12 shows the experimental set-up. The SNOM is a two-pieces cylinder, similar to the atomic-force microscope (AFM) described in Ref. [33]. The detection system is on the upper part and the sample scanning devices in the lower part. The piezoelectric scanning range is 17x17~5 l.trn3. The

G. Margaritondo

330

tip-sample distance is controlled by a shear-force feedback system and the topographic vertical resolution is 21 nm. The tip of the optical fiber was stretched obtaining a few tens of nanometer wide pinhole.

(

a

)

shear force photocurrent

2cwn

0 253nm 2p.m

8

24nA

photon energy (eV) Fig. 11. SNOM Results related to Fig. 10: (a) Shear-force topography and photocurrent microimages of a 8x8 Frn2 area taken with a photon energy of 1.465 eV. (b) Square root of the internal photoemission yield taken as a function of the photon energy for the points A and B; the difference in threshold reveals a smaII lateral fluctuation of the Pt-GaP Schottky barrier . Data from Ref. [ 111.

IPES of Semiconductor

331

Interfaces

Infrared reflectivity images were obtained by detecting the reflected intensity while scanning the fiber position relative to the sample. Topographic images were obtained by measuring the shear-force signal with a synchronous detection including a piezoelectric oscillator and an AC/DC converter.

InGaAs

lock-in amplifier reflectance t signal oscillator

convr

tip

laser

lizzz shear-force signal

x. scan computer

Fig. 12. Scheme of the experimental setup for the first SNOMFEL experiment of Cricenti et al., Ref. [32].

Figure 13a is a 12x12 pm2 shear force image. Brighter areas correspond to higher topography values (or to higher reflectivity in the optical images shown later). A rather big structure (corrugation ~1500 nm) is visible in the upper left corner. Figure 13b is a reflectivity image of the same area of Fig. 13a for 2.4 pm wavelength. The strongly corrugated feature is still visible, but in addition we see a ~1000 nm wide, 0.2 mV deep transversal “valley”. Due to the deep penetration of the infrared light, its origin must be well below the surface, and attributed to a Si groove.

332

Fig. 13. (a) 12x12 pm2 shear-force topography image of a “buried” PtSi/Si interface taken by Cricenti et a1.[32] with the apparatus of Fig. 12. (b) Reflectivity image corresponding to Fig. 13a, taken with a photon wavelength of 1.2 pm. (c) Reflectivity image corresponding to Fig. 13b, taken with a photon wavelength of 0.65 w. Note the different intensity profiles along the line A-A’

IPES of Semiconductor

Interfaces

333

The transverse “valley” is no longer visible in the image of Fig. 13c, taken at 0.65 m wavelength, whereas the strongly corrugated upper-left-comer structure is still visible. This indicates that the “valley” in the 2.4 image is deeper than the maximum penetration of 0.65 pm photons at the incidence angle of the experiment (~75 degrees from the normal direction). The estimated upper limit for the lateral resolution was 100 nm for topographic images, and 200400 nm for reflectivity images. Although these values are most likely overestimates, they are already well below the h/2 level -- demonstrating that the main objective of near-field conditions was achieved. This opens the way to future interface barrier measurements using an EEL, and to a number of other interface-exploring technique with high lateral resolution.

4. Future:

Coherent Radiation

and Other Avenues

What is in the future of synchrotron and FEL studies of solid interfaces? Of course, an easy prediction is that with even better sources coming on line - such as the record-brightness Swiss Light Source (SLS) - the previously discussed techniques will become even more effective and useful. But there is more: besides being very bright, the new synchrotron sources are also becoming very coherent, and this feature can be exploited for some spectacular new techniques, of potential interest for interface science. Such techniques are still unknown to most scientists. Until now, researchers using x-rays never dealt with coherence: there were simply no coherent sources available, Many of them do not yet realize that coherent x-rays have already become reality with the third-generation synchrotron sources . In fact, space and time coherence [2,3] are related to (1) the same geometric factors that also determine the source brightness, and (2) the spectral bandwidth, which can be made extremely narrow with third-generation sources. Considering their potential importance, it is reasonable to spend a few words for an elementary discussion of these aspects. The best simple definition of coherence is “the source capability to produce interference effects”. This concept can be translated into equation by using, as shown in Fig. 14, one of the simplest interference effects: a two-slit (sl and ~2) Young interference. A point source (zero size) with infinitely narrow wavelength bandwidth (Ah = 0) would of course give a well-defined interference pattern. But what happens when these conditions are relaxed? Is the pattern still visible? Consider the situation of Fig. 14, where the source linear size is of the order of x. Different points of the source emit rays that reach the two slits with different optical paths. Each point would produce an interference pattern, but the patterns are not in the same position, so there may not be an overall, visible pattern. The condition for having an overall visible pattern produced by all emitting points is that the phase shift between any two points is much smaller than 2p. Figure 14 shows the geometry for maximum phase shift between two emitting points: (1) one considers the two extreme points a and b of the source, and (2) one takes the slits at the edge of the source emission, whose angular aperture is 8.

334

G. Margaritondo

The difference in optical path between a and b is of the order of &.sin(8/2) = &/2 for small source divergence, and the corresponding phase shift is 58/2h. Thus, the condition for seeing the interference pattern is, roughly speaking, {e/21
Fig. 14. Young’s two-slit interference experiments can be used to illustrate the notions of space and time coherence.

This condition for space coherence can be expressed as gf3 < constant x h. A less approximated approach would yield a similar condition but with a different value for the constant, depending on the specific criterion for defining a “visible” overall pattern -- for example, the Rayleigh criterion. In reality, the source is two-dimensional, therefore one must define the condition for space coherence in different directions. In the case of a symmetric source like an (approximately) elliptic

IPES of Semiconductor

Interfaces

335

synchrotron source, one can take the two high-symmetry source sizes 51,52 and angular divergences 01, @: the conditions for space coherence are $l&(

constant x h and 5282
One can also define an overall degree of space coherence (coherent fraction) as C = constant x (h2/515201fk2).Using the Airy half-disk criterion, the constant becomes approximately (1.22/4rc2), and: c = (1.22 x2)(4X2 c&+@#

.

Note that the same geometric factors determining the space coherence of the source - size and angular spread - also determine [3,4] the source brightness. Thus, by improving the source geometric properties to increase the brightness, one also increases the spacecoherence. The remarkable point is that third-generation sources like Elettra are highly coherent (C = 1) up to photon energies of the order of tens of eV. So, they are as coherent for such photon energies as a laser would ever be. Furthermore, their coherence degree is quite high up to photon energies in the soft-x-ray domain. When full space coherence is achieved, the source geometry cannot be further improved. The reason is quite simple: suppose that one tries to improve the source coherence and geometry by putting a small pinhole in front of it. When the pinhole size becomes comparable with the wavelength, the diffraction contributes significantly to the angular spread. This prevents any further improvement of the product (source size)x(angular spread): one has reached the diSfraction limit. This limit is linked to (1): roughly speaking, a source which reaches its diffraction limit also reaches full spatial coherence. This has a fundamental implication: a third-generation machine like Elettra is already a perfect source up to tens of eV, except for an unlikely increase in flux. Thus, the already envisioned fourth generation of synchrotrons will not improve the brightness at these photon energies. And the third generation is also pretty good at higher photon energies, thus it will not be made useless by the future fourth generation. We must now consider the effects of the finite source bandwidth. Consider again Fig. 14, assuming that the source has full spacecoherence but emits wavelengths over a bandwidth A& Different wavelengths give different interference patterns, possibly making it impossible to observe an overall pattern. The condition for observing an overall pattern is that the phase difference for any two different wavelengths is ~27~.The maximum phase difference occurs for two wavelengths differing by A.h. Taking an optical path L, such phase difference is = 2~ Id(LA)/dhlAh = 27t (L/k)(Ah/h). Thus, one must have: (LhL)(Ahn) <<1 . This is, roughly speaking, the condition for time coherence or longitudinal coherence. One usually expresses this condition by defining the coherence length of the source, which is the value of L giving CL/X>(AUN = 1, or:

336

G. Margaritondo

Lc = ‘h (h/AX) .

(3)

Note that the time coherence depends on the capability to narrow the bandwidth and enhance the so-called resolving power (h/Ah). This is possible with advanced monochromators, whose performances depend in turn on the source quality. Third-generation sources make it possible to reach amazingly high values of the resolving power, and similarly impressive value of the coherence length. In summary, x-ray coherence is already here, both space coherence and time coherence. But it is not used, because most x-ray scientists are unaware of coherence. This paradoxical situation will hopefully change with creative new experiments exploiting coherence. Consider, for example, the holographic techniques. Roughly speaking, one cannot (with standard holographic techniques) produce a holographic image with details smaller than the wavelength. A highly coherent source up to photon energies of 100 eV could produce very high quality holograms with details as small as 100 A. Quite amazingly, many scientists are struggling to produce atomic-scale holograms with techniques like photoelectron holography -- which are affected by severe controversies. On the other hand, there are wonderful and less controversial opportunities offered by straight x-ray holography to explore mesoscopic and nanoscopic structures, including those relevant to life-science specimens.

A. Coherence-enhanced imaging Similar opportunities are offered by coherence in the general domain of microimaging. But they are still widely unknown: in a sense, we are using laser-like sources as if they were blowtorches. This means that the source capabilities are still quite unexplored, whereas these capabilities this can lead to outstanding new classes of experiments. Third-generation synchrotron light sources like Elettra in Trieste [6,34] and the European Synchrotron Radiation Facility @RF) in Grenoble [5] recently produced very spectacular and sharp xray images of different weakly-absorbing objects. Similar results were also obtained with other sources. The Elettra results demonstrate that a remarkable improvement in image quality can be achieved whit a reduced dose, suitable for medical radiology.[6,34] Quite remarkably, the conditions for sharpening are not as stringent as one might think. In particular, very limited time coherence is necessary. And the required spatial coherence, more specifically the small source size, was already available from second-generation sources. These conclusions can be reached with a simple analytical model [35] of the coherence-sharpening mechanism. The main point in our analytical model is that Fresnel diffraction by the object edges is responsible for image sharpening. Consider first the classic Fresnel edge diffraction of a perfectly opaque object, using the coordinate system of Fig. 15. The standard textbook treatment (including all relevant approximations) gives:

IVES of Semiconductor

337

Interfaces

I cc IT(-) - C(-u,>12+ [S(~) - S(-u,>P ,

(4)

where I is the intensity at the detector D; C(u) and S(u) are the Fresnel integrals; u is the reduced variable: u(z) = [2(ro + ro)/rnroh]1/2 2 ,

(5)

(h = wavelength), and u. = U(Q).

tz

D ~--,po-~-~-i-~------~o.--source ‘ZO

Fig. 15. Geometry for the Fresnel edge diffraction analysis.

Using the Cornu-spiral method, it is easy to see that (4) gives the well-known series of Fresnel diffraction maxima and minima as a function of uo. In x-ray imaging, however, most objects are weak absorbers (which is the root of the difficulties in obtaining good contrast) whereas they can cause phase shifts. The wave field at the detector is thus: Wave field = [C(m) - C(-uo)] + i[S(aJ) - S(-uo)l + + exp(-a + i$) { [C(-u,) - C(-m)] + i[S(-uo) - S(--)I) ,

(6)

G. Margaritondo

338

where a is the object attenuation coefficient and f is the phase shift. Assuming small values of a and 9, (6) approximately gives: I = 1 - a + I$ [C(-u,) - S(-ud)] - a [C(-u,) + S(-uo)]

.

(7)

This produces again Fresnel diffraction maxima and minima, depending on the interplay of the two quantities [C(-u,) - S(-uo)] and - [C(-u,) + S(-uc)] -- see the plots of Fig. 16. The most realistic case is a <(41,thus Eq. 7 approximately becomes: I = 1 + Q,[C(-u,) - S(-u,)]

.

From either Fig. 16 or the Comu spiral, one derives a strong intensity maximum at u. = 0.7, with a symmetric minimum at u. = -0.7. This maximum is the main cause of the image sharpening. But what are the conditions for detecting it? First of all, as argued for example in Ref. [34], the detector (for example a photographic plate) must be placed at a sufficiently large distance r, so that the intensity maximum is not blurred by the detector spatial resolution (e.g., the granule size for photographic plates). Note that for a given geometry the z-value ztr, of the main maximum, corresponding to the u-value un, = 0.7, increases with rO: Zm = (PoroWro

+ro))lpum

,

Consider now time coherence, which one might suspect to be the second condition for observing the main diffraction maximum. But this is not really true; in fact, Eq. 9 gives: 6zmJzm = dhi2h , thus the condition 6zm/zm <<1 corresponds to &L/2X (<2, which requires very limited resolving power and time coherence. Finally, we consider spacecoherence. Among the space-coherencefactors, the only one relevant to the present case is the source size. From Fig. 15 we see that a source size 5 in the z-direction would cause a spread =(rc/po)~ in the diffraction pattern. Thus, the approximate condition for seeing the main diffraction maximum is that (r&& remains smaller than the distance 22, between the maximum and its symmetric minimum: hJpo)5

6 2Zrn = 2 [fX&W(p0

+ ro)11’2Urn .

(10)

Taking um = 0.7 and typical values (for Elettra) p. = 2 m, r. = 20 m, h = I A, (10) gives approximately:

IPES of Semiconductor

Interfaces

339

5
(11)

a not-so-stringent condition certainly met by third-generation synchrotrons, but also by some second generation sources. Furthermore, a condition that could be reached with a suitable pinhole and further weakened by increasing the distance ratio (p&J.

1.5

1

0.5

0

-0.5

-1

-1.5 -2

-1.5

-1

-0.5

0

0.5

1

1.5

2

U Fig. 16. Plots of the quantities [C(-u,) - S(-uO)l and - [C(-u,) + S(-uO)] relevant to Eqs. 7 and 8.

We note that the rather forgiving condition of (11) also explain why image sharpening is achieved even with bending-magnet radiation, which is of limited size only in the vertical direction. Edges along

340

G. Margaritondo

other direction appear in fact sharpened to some extent, contributing to the overall object visibility and image quality. The simple model adopted in our analysis is fully adequate to reveal the essential points in the sharpening conditions and the conditions for its achievement. We note, by the way, that the commonly used name of “phase contrast” [5] for this technique is somewhat misleading, since it is reminiscent of phase-contrast microscopy -- whose implementation is entirely different. In particular, “phase contrast” might suggest that a non-absorbed reference wave is required, which would make radiological applications all but impossible. Figure 17 shows a simulated image sharpening effect.[35] Figure 18 shows and example of real experiment, performed by Giuliana Tromba and coworkers on Elettra in Trieste.[6,34] In both cases, the improvement in image quality is nothing less than spectacular.

Fig. 17. Simulation of the effect of edge sharpening by a coherent x-ray sources based on Eq. 8. On the right-hand-side, the conventional absorption-based contrast is almost invisible.

The coherence-sharpening technique has been developed having in mind medical radiology as the first application. However, other applications are certainly possible, most notably those on interfaces. Basically, the technique enhances the contrast between two regions with different values of the refractive index. Even small differences can produce a large phase shift, because of the small value of the wavelength. Thus, it appears possible to sharpen the interface boundaries as well as those of different portions of the interface. In summary, image sharpening exploiting the source coherence is a technique with widespread potential applications and quite simple to use. The analysis of buried interfaces is an obvious candidate for its exploitation, but other domains of materials science are also quite promising.

IPES of Semiconductor

Interfaces

Fig. 18. A recent real example of image sharpening by exploitin :rent source, obtained on ELETTRA (Refs. [6] and [34]).

341

342

G. Margaritondo

B. Other avenues: ultrahigh resolution, time resolution Coherence is, in our opinion, the most revolutionary new aspect of x-ray-based interface research. However, other development areas that are opened up by the new source performances. We would like to mention as examples time-resolved photoemission and photoemission at very high energy resolution. In the first case, the source brightness is exploited to decreasethe data taking time per spectrum in standard photoelectron spectroscopy with synchrotron light.[36,37] The time is reduced to 0. l- 1 s, and this makes it possible to explore dynamic processesrather than just their final products. This could also enable us to intervene in the intermediate steps and optimize the entire process.

Ag Fermi level

meV

Energy (meV) Fig. 19. High-resolution Fermi edge photoemission spectrum. After taking into account thermal broadening, the instrumentalbroadening contribution does not exceed 3-4 meV (Ref. [38]).

IPES of Semiconductor

interfaces

343

The few examples of this approach are quite exciting. There is no need to emphasize the important role that these new technique could play in the study of semiconductor interfaces and of their fabrication. Similarly important is the potential role of very high resolution. How high? Figure 19 shows a state-of-the-art example, a metallic Fermi edge taken at the Lausanne Center of Spectromicroscopy by Grioni et a1.[38] without using synchrotron light. The thermal broadening correction shows that the instrumental broadening contribution is no more than 3-4 meV. Performances at this level are potentially very important for interface studies, since they reveal what happens in the crucially important region near the Fermi level -- which is responsible for some if not most of the important interface properties. Studies in other systems have clearly demonstrated that the properties of this region can depart from the standard framework of solid-state physics, e.g., Fermiliquid theories. This conclusion is specifically valid for low-dimensionality systems. Therefore, high-resolution studies of non-conventional phenomena are potentially very important for intrinsically lowdimensionality systems like semiconductor interfaces. And this corroborates the central point that we are argue: the new photon sources for interface research are only marginally exploited at the present time. New and creative ideas must be stimulated, and the entire field remains in a state of rapid evolution.

Acknowledgments The results used as examples in this overview were produced in collaboration with many scientists, including all authors of Refs. [4,6,7,10-16,18-21,23,30,32,35,38]. Support was provided by the Fonds National Suisse de la recherche Scientifique, by the Ecole Polytechnique F&l&ale de Lausanne, by the US Office of Naval Research, by the Sincrotrone Trieste SCpA, and by the European Commission.

References [l] [2] [3] [4]

D. Wolf and S. Yip, (Eds.), Materials Interfaces Atomic-Level Structure and Properties, Chapman & Hall, London (1992). See, for example: G. Margaritondo, Riv. Nuovo Cimento 18, 1 (1995). G. Margaritondo, Introduction to Synchrotron Radiation, Oxford, New York, (1988), and the references therein. J. T. McKinley, R. G. Albridge, A. V. Barnes, P. A. Baudat, G. C. Chen, C. Coluzza, J. L. Davidson, C. Dupuy, F. Gozzo, M. Ilegems, M. L. Languell, D. Martin, F. Morier-Genoud, P. L. Polavarapu, A. Rudra, J. F. Smith. E. Tuncel, X. Yang, A. Ueda, G. Margaritondo and N. H. Tolk, Nucl. Instrum. Meth. A341, 156 (1994); J. T. McKinley, R. G. Albridge, A. V. Barnes, A. Ueda, N. H. Tolk, C. Coluzza, F. Gozzo, G. Margaritondo, D. Martin, F. MorierGenoud, C. Dupuy, A. Rudra and M. Ilegems, J. Vat. Sci. Technol. Bll, 1614 (1993); G. Margaritondo, C. Coluzza, J.-L. Staehli, E. Tuncel, J. T. McKinley, R. G. Albridge, A. V.

344

G. Margaritondo

Barnes, A. Ueda, X. Yang and N. H. Tolk, J. Physique IV 4,C9-357 (1994); T. McKinley, R. G. Albridge, A. V. Barnes, G. C. Chen, J. L. Davidson, M. L. Languell, P. L. Polavarapu, J. F. Smith, X. Yang, A. Ueda, N. Tolk, C, Coluzza, P. A. Baudat, C. Dupuy, F. GOZZO,M. Ilegems, D. Martin, F. Morier-Genoud, A. Rudra, E, Tuncel and G. Margaritondo, J. Vat. Sci. Technol. A12, 2323 (1994); N. H. Tolk, J. T. McKinley and G. Margaritondo, Surface Rev. and Lett. 2, 501 (1995); N. H. Tolk, R. G. Albridge, A. V. Barnes, B. M. Barnes, J. L. Davidson, V. D. Gordon, G. Margaritondo, J. T. McKinley, G. A. Mensing and J. Sturmann, Appl. Surf. Sci. 106, 205 (1996); N. H. Talk, C. A. Brau, G. S. Edwards, G. Margaritondo and J. T. McKinley, Proc. Conf. on Short-Wavelength Radiation Sources, San Diego, California 1991 (SPIE Proceedings Series, Bellingham, Washington, USA, 1991) vol. 1552, p. 7; C. Coluzza, G. Almeida, E. Tuncel, J.-L. Staehli, P. A. Baudat, G. Margaritondo, J. T. McKinley, A. Ueda, A. V. Barnes, R. G. Albridge, N. H. Tolk, D. Martin, F. Morier-Genoud, C. Dupuy, A, Rudra and M. Ilegems, Proc. Internat. Symposium on Physical Concepts and Materials for Novel Optoelectronic Device Applications II, F. Beltram and E. Gomik eds. (SPIE Proc. Series 1993), vol. 1985, p. 442; N. Tolk, R. G. Albridge, A. V. Barnes, J. T. McKinley, H. B. Nielsen, A. Ueda, J. F. Smith, J. L. Davidson, M. L. Languell, C, Coluzza, E. Tuncel and G. Margaritondo, SPIE 1854,60 (1996). 151 A. Snigirev, I. Snigireva , V. Kohn, S. Kuznetsov, I. Schelokov, Rev. Sci. Instrum. 66, 5486 (1995); Nucl. Instrum. Methods A370,634 (1996), and the references therein. [61 M. Di Michiel, A. Olivo, G. Tromba, F. Arfelli, W. Bonvicini, A. Bravin, G. Cantatore, E. Castelli, L. Dalla Palma, R. Longo, S. Pani, D. Pontoni, P. Poropat, M. Prest, A. Rashevsky, A. Vacchi and E. Vallazza, unpublished. [71 G. Margaritondo, C. Coluzza, J.-L. Staehli, E. Tuncel, J. T. McKinley, R. G. Albridge, A. V. Barnes, A. Ueda, X. Yang and N. H. Tolk, J. Physique IV 4, C9-357 (1994), and the references therein.

i31 H. Palm, M. Arbes and W. Schulz Phys. Rev. Lett. 71, 2224 (1993); R. H. Williams, Appl. Surf. Sci. 70/71, 386 (1993); A. A. Talin, R. S. Williams, B. A. Morgan, K. M. Ring and K. L. Kavanagh, Phys. Rev. B 49, 16474 (1994); A. Bauer, M. T. Cuberes, M. Prietsch. and G.

PI

Kaindl, J. Vat. Sci. Techn Bll, 4 (1993): R. Ludeke, M. Prietsch and A. Samsavar, J. Vat. Sci. Technol. B9, 2342 (1991). G. Margaritondo, Scanning Spectromicroscopy 9,949 (1995); G. Margaritondo, Surface Rev.

Lett. 2, 591 (1995); G. Margaritondo, Surface Rev. and Lett. 2, 305 (1995); G, Margaritondo and Y. Hwu, Appl. Surf. Sci. 92, 273 (1996). UOI C. Coluzza, E. Tuncel, J.-L. Staehli, P. A. Baudat, G. Margaritondo, J. T. McKinley, A. &da, A. V. Barnes, R. G. Albridge, N. H. Tolk, D. Martin, F. Morier-Genoud, C. Dupuy, A. Rudra and M. Ilegems, Phys. Rev. (Rapid Commun.) B46, 12834 (1992). El11 J. Almeida, Tiziana Dell’Orto, C. Coluzza, G. Margaritondo, 0. Bergoss, M. SpajeT and D. Coujon, Appl. Phys. Lett. 69, 2361 (1996).

IPES of Semiconductor

Interfaces

345

[12] F. Cerrina, B. Lai, C. Gong, A. Ray-Chaudhuri, G. Margaritondo, M. A. Green, H. Hochst, R. Cole, D. Crossley, S. Collier, J. Underwood, L. J. Brillson

and A. Franciosi, Rev. Sci.

Instrum. 60,2249 (1989). [13] F. GOZZO, M. Marsi, H. Berger, G. Margaritondo, A. Ottolenghi, A. K. Ray-Chau&w-i, W. Ng, S. Liang, S. Singh, J. T. Welnak, J. P. Wallace, C. Capasso and F. Cerrina, Phys. Rev. B48, 17163 (1993). [14] F. Gozzo, H. Berger, I. R. Collins, G. Margaritondo, W. Ng, A. K. Ray-Chaudhuri, S. Liang, S. Singh and F. Cerrina, Phys. Rev. B51, 5024 (1995). [15] J. Almeida, I. Vobornik, H. Berger, M. Kiskinova, A. Kolmakov, M. Marsi and G. Margaritondo, Phys. Rev. B (Rapid Commun.) 55,4899 (1997). [ 161 M. Bertolo, S. Fontana and S. La Rosa, unpublished. [17] F. Capasso and G. Margaritondo, (Eds.), Heterojunction Band Discontinuities: Device Applications, North-Holland, Amsterdam (1987).

Physics and

[18] F. Cerrina, A. K. Ray-Chaudhuri, W. Ng, S. Liang, S. Singh, J. T. Welnak, J. P. Wallace, C.

[19] [20] [21] [22]

Capasso, J. H. Underwood, J. B. Kortright, R. C. C. Perera, and G. Margaritondo, Appl. Phys. Letters 63, 63 (1993). R. R. Daniels, G. Margaritondo, C. Quaresima, M. Capozi, P. Perfetti and F. Levy, Sol. State Commun. 51, 495 (1984). R. R. Daniels, G. Margaritondo, P. Perfetti, C. Quaresima and F. Levy, J. Vat. Sci. Technol. A3, 979 (1985). Gelsomina De Stasio and G. Margaritondo, in Spectomicroscopy with WV Photons and XRays, H. Ade, (Ed.), Elsevier, Amsterdam (1997), p. 137, and the references therein. G. Margaritondo, in Chemical, Structural and Electronic Analysis of Heterogeneous Surfaces on Nanometer Scale, R. Rosei, (Ed.), Kluwer, Dordrecht (1997), p. 43.

[23] Gelsomina De Stasio, G. F. Lorusso T. Droubay M. Kohli P. Muralt, P. Perfetti, G. Margaritondo T. F. Kelly and B. P. Tonner, Rev. Sci. Instrum. 67, 737 (1996). [24] Th. Schmidt, J. Slezak, S. Heun, J. Diaz, K. C. Prince and E. Bauer unpublished. [25] E. Betzig and J. K. Trautmann, Science 257, 189 (1992). [26] E. Betzig, P. L. Finn and J. S. Wiener, Appl. Phys. Lett. 60, 2484 (1994). [27] S, K. Buratto, J. W. P: Hsu, E. Betzig, J. K. Trautmann, R. B. ByJlsma, C. C. Bahr and M. J. Cardillo, Appl. Phys. Lett. 65, 21 (1994). [28] 0. Bergossi and M. Spajer, Proc. SPIE 2341, 239 (1994). [29] M. Spajer and A. Jalocha, in Near Field Optics, D. W. Pohl and D. Courjon, (I%.), Kluwer, Dordrecht (1993). p. 80. [30] J. Almeida, C. Coluzza, T. dell’Orto, F. Barbo, M. Bertolo. A. Bianco, S. Cerasari, S. Fontana and G. Margaritondo, J. Appl. Phys. 80, 1460 (1996). [31] L. D. Bell , W. J. Kaiser, M. H. Hecht and L. C. Davies, in Scanning Tunneling Microscope, A. Stroscio and W. J. Kaiser (Eds.), Academic, Boston (1993), p. 307.

346

G. Margaritondo

[32] A. Cricenti, R. Generosi. P. Perfetti, J. M Gilligan, N. H. Tolk, C. Coluzza and G. Margaritondo, unpublished. [33] A. Cricenti, R. Generosi, Rev. Sci. Instrum. 66,2843 (1995). [34] F. Arfelli, W. Bonvicini, A. Bravin, G. Cantatore, E. Castelli, L. Dalla Palma, M. Di Michiel, R. Longo, A. Olivo, S. Pani, D. Pontoni, P. Poropat, M. Prest, A. Rashevsky, G. Tromba, A. [35] [36] [37] [38]

Vacchi, E. Vallazza, unpublished. G. Margaritondo, G. Tromba, M. Di Michiel and A. Olivo, unpublished. A. Baraldi, M. Barnaba, B. Brena, D. Cocco, G. Comelli, S. Lizzit, G. Paolucci and R. Rosei, J. Electron Spectr. 76, 145 (1995). G. Comelli, A. Baraldi, S. Lizzit, G. Paolucci and R. Rosei, to be published. M. Grioni, I. Bobomik, M. Zacchigna, F. Zwick and G. Margaritondo, unpublished.