The redistribution of implanted dopant atoms during silicide formation has attracted much interest recently because of its important implications for shallow junction device technology. Ion channeling and electrical measurements have shown that dopant atoms are pushed ahead in front of the moving silicide-silicon interface during the growth of near-noble metal silicides. However, dopant redistribution has not been observed with refractory metal silicides. This unique feature of near-noble metal silicides is discussed in conjunction with the growth kinetics of these silicides.

We investigated the interaction of extremely thin (less than 10 nm) crystalline gold and rhodium films with amorphous silicon by transmission electron microscope in situ annealing. In thin Au/Si bilayers an amorphous phase with a diffraction peak at d ≂ 0.226 nm is formed by thermal annealing between 150 and 200 °C. Depending on the thickness and composition, silicon sputtered onto thin gold films leads to the formation of a layer of amorphous silicon and a partially amorphous Au-Si layer during deposition. The silicon layer crystallizes by itself at temperatures as low as 150 °C, and at 300 °C the amorphous Au–Si layer crystallizes into a metastable gold silicide (for silicon-rich compositions). In Rh/Si bilayers an amorphous Rh–Si phase is formed by annealing to 300 °C and can be detected by electron diffraction for a rhodium thickness of less than 5 nm and compositions with more than 50% Si if completely reacted. Above 300 °C the amorphous Rh-Si crystallizes preferentially in the cubic form of RhSi for intermediate silicon compositions and in the orthorhombic form of RhSi for high silicon compositions. Excess amorphous silicon is not found to have a lowered crystallization temperature when in contact with the amorphous Rh-Si alloy, and crystalline silicon is only observed above 730 °C together with the cubic and/or orthorhombic RhSi. In Rh/Si bilayers with a thicker rhodium layer, no formation of an amorphous phase was observed on annealing; instead crystalline Rh2Si forms during annealing above 300 °C.

Lateral diffusion couples formed by depositing platinum islands on silicon layers on Al2O3 were used in conjunction with scanning electron microprobe measurements to investigate the growth of platinum silicides in the temperature range 400–700 °C. The phase Pt2Si grows over a length of 4–30 μm with a rate proportional to the square root of time and an activation energy of approximately 1.3 eV. With samples containing 7 at.% Rh in the platinum, the growth rate of Pt2Si is reduced and the activation energy is increased to about 2.0 eV. In these Pt–7at.% Rh samples, electron-induced X-ray measurements indicate that rhodium remains in the original deposited region while both platinum and silicon diffuse in the formed Pt2Si region.

The redistribution of p-type dopants (boron and indium) during PtxSi growth was studied as a function of annealing time using secondary ion mass spectrometry. It was found that both boron and indium atoms redistribute into the silicide during a thermal anneal. Initially, the dopant becomes trapped at the PtxSi–Si interface and concomitantly getters to defect sites located at the moving Pt-PtxSi interface. Even after silicide formation is complete, the dopant continues to out-diffuse to the surface via an enhanced mechanism due to defects and/or grain boundaries. Finally, no evidence for a “snowplow” effect was observed for either boron or indium at the implant dose of 2 × 1013 ions cm−2.

The room temperature oxidation of PdSi, Pd2Si and Pd4Si has been studied using X-ray photoelectron spectroscopy (X-ray photoemission spectroscopy or electron spectroscopy for chemical analysis). We find that only silicon atoms in these silicides are oxidized and the oxidation of Pd4Si surfaces is enhanced compared with that of Pd2Si and PdSi, as is evidenced by both a higher silicon oxidation state and thicker oxide films. This behavior is discussed in terms of silicide stability and a spillover effect where palladium atoms catalyze molecular oxygen dissociation.

We investigated the effects of thin Si16O2 and Si18O2 interfacial layers on Pd2Si formation using He+ backscattering spectrometry and the16O(d,α)14N and 18O(p,α)15N nuclear reactions. SiO2 films from 10 to 50 Å thick were formed on Si<100> substrates by 16O2 or 18O2 plasma oxidation. Nickel, palladium or platinum films from 1000 to 1400 Å thick were then evaporated onto this SiO2 layer and the samples were annealed in vacuum at temperatures ranging from 250 to 750 °C. The minimum temperature necessary for Pd2Si formation is approximately 400 °C for about 24 Å of SiO2 compared with 600 °C for Pt2Si and 650 °C for Ni2Si formation. The ability of Pd2Si to form in the presence of an SiO2 interfacial layer is compared with that of Ni2Si and Pt2Si and is interpreted using the relative mobilities of the metals in SiO2 and the redistribution of the 18O during silicide formation.

We investigated the redistribution of oxygen after a thermally induced reaction of an evaporated chromium film deposited on an Si<100> substrate. To monitor the redistribution, 18O was implanted in the metal film or in the silicon substrate. Depth profiles were determined by the isotope-specific reaction 18O (p,α) 15N before and after thermal annealing. Silicide formation was monitored by 2 MeV 4He+ backscattering spectrometry.

It was found that oxygen in the chromium film remains stationary with respect to the chromium during silicide formation, but oxygen in the silicon substrate accumulates at the silicon-silicide interface. The phase formed is CrSi2, as for the unimplanted reference samples. The effect of oxygen on the rate of silicide formation is much more pronounced when the oxygen is in the silicon substrate rather than in the chromium film. This asymmetry is attributed to the fact that silicon is the dominant diffusing species during CrSi2 formation. These results are contrasted with those for nickel and platinum, where the metal moves in the silicide formed by these metals.

In the present study we investigate the influence of nitrogen on the silicide formation of thin nickel and platinum films. Nitrogen is introduced by implantation either in the metal films or in the silicon substrates. We use the rare stable isotope 15N for the implantation and the nuclear reaction 15N(p,α)12C for the detection and profiling of the impurity. For nitrogen in nickel, we find that at 350 °C nitrogen is mobile. It accumulates at the bottom nickel interface and, for a dose exceeding about 0.5 × 1016 N atoms cm−2, forms a barrier to silicide formation. When nitrogen is initially in the silicon substrate, the nitrogen profile is broadened in the same proportion as the dilution of the silicon sublattice when forming the silicide. Similar experiments with platinum films showed partially different behaviors. All results are explained in terms of a model based on the moving species during the silicide formation and the chemical affinity of nitrogen to the metal and to silicon.

Molybdenum thin films were deposited with an electron beam gun onto (100)- oriented silicon substrates. The samples were then annealed in vacuum, and the internal stresses in the molybdenum thin films were studied as functions both of the annealing temperature and of the substrate temperature during deposition. Silicide formation and the film thickness after annealing were monitored by the Rutherford backscattering spectrometry technique, and the stress was determined from the substrate curvature which was measured from Newton's ring interference fringes. It was found that, when the substrate temperature was kept at 400°C during deposition, MoSi2 was formed after annealing at temperatures above 500 °C. This MoSi2 exhibited large tensile stresses of about 2 × 1010 dyn cm−2 for annealing above 700 °C, whereas at the lower annealing temperature of 500 °C the stresses were compressive. No detectable silicides were observed when the substrates were kept at temperatures below 150 °C. The Mo-Si film stresses were tensile for substrates kept at room temperature during deposition and compressive for substrates kept at 150 °C.

A short review of characteristic electronic properties of heterojunction interfaces is given. Band edge discontinuities and interface band structures for lattice-matched junctions are discussed in detail. The examples presented include non-polar and polar junctions as well as overlayer systems. The results of involved calculations are interpreted in terms of simple physically appealing pictures by directly relating the changes in bonds across an interface to the resulting bands in the interface band structure. The meaning of the results for the transport properties of semiconductor heterojunctions is briefly assessed.

New materials preparation techniques such as molecular beam epitaxy (MBE) provide dramatically improved possibilities for interface control and novel device structures. However, compared with our knowledge of Schottky barriers or semiconductor-insulator interfaces, our microscopic information on the basic science underlying semiconductor heterojunctions is very limited. Insight into the growth, extent and composition of such interfaces is needed to illuminate the origins of the electronic and optoelectronic behavior. Most recent experimental studies have not properly monitored the detailed structural characteristics of the heterojunctions being measured.

In this paper we discuss the scientific exploration of the interrelationship between growth and electronic properties of hetero-interfaces prepared by MBE. In situ low energy electron diffraction and synchrotron radiation photoemission were performed on germanium interfaces with GaAs and AlAs polar and non-polar surfaces. The individual surface and subsurface bonding interactions (rather than bulk thermodynamic properties of the semiconductors) are shown to control the formation and electrical properties of hetero-interfaces of either germanium on the compound or the inverse deposition sequence. Simple arguments based on bulk semiconductor heats of formation incorrectly predict the relative stability of interfaces in the lattice-matched Ge/GaAs/AlAs system. Rather the formation of a GeAsx surface layer provides a lower energy free surface during MBE growth and causes arsenic diffusion from the interface layer to lower the free energy when a pure germanium surface begins to form. Significant variations in the band edge offsets of heterostructures are observed when the semiconductor-semiconductor interface microstructure changes with substrate orientation and overlayer crystallinity. Neither the common electron affinity difference “rule” nor recent tight-binding models for determining heterojunction barriers allows for effects of interface microstructure. Measurement and analysis of such “materials-related” aspects of semiconductor interfaces are central to the understanding and control of device characteristics.

These studies are partially funded by the U.S. Office of Naval Research (L. R. Cooper), Contract N00014-81-C-0696. Part of this work was performed at SSRL and Stanford Linear Accelerator Center, supported by National Science Foundation Grant DMR 77–27489 and the U.S. Department of Energy.

Quantum effects occurring at interfaces between GaAs and AlxGa1−x As are reviewed. Electrons and holes can be confined to two-dimensional quantum states in thin semiconductor layers or at semiconductor heterointerfaces. The twodimensional states in GaAs/AlxGa1−xAs structures show strongly modified electrical and optical behavior relative to unconfined material, with the properties strongly dependent on the structural design.

Electrons confined in high mobility structures at selectively doped GaAs-AlxGa1−xAs heterointerfaces show a quantized Hall effect with an associated vanishing parallel resistance and, in high magnetic fields, an apparent electron crystallization.

Optical transitions between confined electron and hole states in quantum well thin layers form the basis of quantum well luminescence and for new non-linear optical logic structures.

All the quantum effects are sensitive to interfacial properties which are closely dependent on the conditions of epitaxial growth.

Metal-semiconductor contacts and semiconductor heterojunctions are wellestablished concepts in semiconductor device technology. The key parameters characterizing such junctions are the barrier height and valence band discontinuity at the interface and the electronic interface states. Clean cleaved GaAs(110) surfaces exhibit no intrinsic surface states in the bulk band gap but do exhibit extrinsic cleavage-induced states. Furthermore, local segregates of arsenic were detected. The chemisorption of metals, semiconductors, hydrogen and oxygen causes depletion layers to form on both n-and p-type crystals. The surface states responsible for these band bendings, which persist even under thick layers of metals and semiconductors, are thought to be related to chemisorption-induced defects. Possible candidates are discussed. Chemical trends are also considered.

Recent theoretical ideas on the origin of the Schottky barrier on the cleavage plane of III–V semiconductors are reviewed, using the Al/GaAs(l10) system as a benchmark.

The crystal structure of aluminum layers grown by molecular beam epitaxy (MBE) on GaAs(100) was studied using high resolution transmission electron microscopy. We obtained cross-sectional images with nearly atomic resolution of the lattice structure at Al-GaAs interfaces. GaAs epitaxial layers were grown at 585 °C and cooled in the ultrahigh vacuum MBE growth chamber prior to aluminum deposition. After the desired substrate temperature had been achieved, aluminum was immediately deposited onto arsenic-rich surfaces at a rate of 1000 Å h−1.

Aluminum layers deposited onto substrates at 80 °C exhibit an Al(110)/GaAs(100) epitaxial relationship along an abrupt interface. In some regions of the interface we observed isolated arsenic crystallites which were epitaxial with the GaAs substrate. The current-voltage (I-V) and capacitance-voltage (C–V) characteristics of this material are well behaved. Barrier heights of 0.79 eV and ideality factors of 1.04 were determined from room temperature I–V measurements.

For aluminum deposited onto GaAs substrates at 400 °C, evidence of twodimensional nucleation and growth was observed. The resulting pyramid-shaped aluminum crystallites had facets along crystallographic planes. Two epitaxial relationships were observed: Al(100)/GaAs(100) and Al(110)/GaAs(100). We observed 12 Å steps along the Al(100)–GaAs(100) interface. C–V and scanning electron microscopy measurements indicate that the individual aluminum crystallites are electrically and physically separated.

Unimplanted n-type GaAs epitaxial layers have been annealed under the same conditions as those required to activate ion implants. If the duration of the anneal is sufficiently long (e.g. 10 min at 800°C) acceptor ions accumulate at the surface with a concentration of about 2×1022 m−3. Schottky contacts to annealed samples always show increased current densities up to six orders of magnitude greater than those in unannealed layers. This has the effect of lowering the Q values of varactor diodes made out of this material, up to frequencies of 1 GHz.

New results are presented on the dependence on energy hv of the photoemission cross sections from d states of d atoms interacting with silicon (Si-Pd and Si-Ag). It is shown that the Cooper minimum is influenced by the atomic environment and that the Cooper effect is more pronounced when the interaction between the d atom and silicon becomes weaker. This approach together with an analysis of the shape of photoelectron spectra at the Cooper minimum is used in a discussion of the electron states in the reacted region of Si-Pd and Si-Mo interfaces.

The binding energy of the bulk Si 2p levels observed with soft X-ray photoemission can be used to monitor the band bending in the silicon space charge region when a metal is deposited onto the silicon surface. Changes in the 2p binding energy with metal coverage can then be used to determine the change in the Schottky barrier height as the metal-silicon contact is formed. By tuning the photon energy and therefore the photoemitted electron escape depth, chemical shifts (atomic environment effects) at the interface can be separated from the bulk band bending effects. When combined with annealing to produce in-depth atomic intermixing, the result may reveal information on the distribution of metal atoms at the interface and its effect on the barrier height.

Measurements of the Schottky barrier height change as a function of palladium deposition were made on (2 × 1) p-type and (7 × 7) n-type Si(111) surfaces by monitoring Si 2p core level shifts in a bulk sensitive mode. The barrier height change reached 1/e of its final value at a palladium coverage of 2.9 Å. Several experiments have made it possible to relate the measured Si 2p core level monitor of band bending to absolute Schottky barrier heights in a fully consistent fashion. Therefore, these results provide a means to measure the barrier height in the initial stages of Schottky barrier formation (i.e. at low metal coverage) and to compare these observations with the chemical behavior of the interface at low coverage and with electrical measurements on bulk contacts. Since the Pd/Si Schottky barrier height is established at the bulk value within a coverage of 3–5 Å (even before the overlayer is metallic), the role of interface chemical bonds in determining the barrier height is paramount.

This work was supported in part by the U.S. Office of Naval Research.

The introduction of defect levels in the band gap of silicon (ND = 2.5×1015 cm−3), after ion beam sputtering of molybdenum contacts, was investigated with deep level transient spectroscopy and scanning electron microscopy (electron-beam-induced current mode). Molybdenum contacts were fabricated with beam voltages of 700, 1100 and 1500 V at a beam current density of 2.5 mA cm−2. Defect levels were observed at energies ranging from 0.18 to 0.55 eV below the conduction band. It was established that the introduction of these defect levels depends on the processing conditions and that they reside very close to the Mo-Si interface (less than 0.4 μtm). A correlation was observed between the fabrication conditions, the current-voltage and capacitance-voltage characteristics of the contacts, and the concentrations of the defects present. Charge collection micrographs confirmed the presence of a defective layer close to the surface.

X-ray photoelectron spectroscopy, Auger electron spectroscopy and ellipsometry measurements were performed at Cr-SiOx, and Yb-SiOx interfaces, which are being considered for metal/insulator/semiconductor (MIS) solar cell applications. These measurements have led to a knowledge of instability mechanisms associated with MIS solar cells. The dynamics of the metal-oxide interface reduce the effective thickness of the insulating SiOx layer. Typically, in a 3 week period after fabrication the oxide thickness decreases by about 2 Å for ytterbium MIS structures and by less than 1 Å for chromium MIS structures. A structural equilibrium is reached thereafter. To explore further the dependence of MIS Schottky barrier heights on oxide thickness variations (from about 5 to 20 Å) a comprehensive charge balance analysis was performed at a general MIS interface. This theoretical study supports experimental data on barrier heights and also on instabilities. In addition, the application of the theoretical model to the experimental barrier heights has yielded quantitative information regarding the density of interface states at the SiOx,–Si interface.