Publications

Abstract not provided.

Abstract not provided.

This document is the final SAND Report for the LDRD Project 105877 - 'Novel Diagnostic for Advanced Measurements of Semiconductor Devices Exposed to Adverse Environments' - funded through the Nanoscience to Microsystems investment area. Along with the continuous decrease in the feature size of semiconductor device structures comes a growing need for inspection tools with high spatial resolution and high sample throughput. Ideally, such tools should be able to characterize both the surface morphology and local conductivity associated with the structures. The imaging capabilities and wide availability of scanning electron microscopes (SEMs) make them an obvious choice for imaging device structures. Dopant contrast from pn junctions using secondary electrons in the SEM was first reported in 1967 and more recently starting in the mid-1990s. However, the serial acquisition process associated with scanning techniques places limits on the sample throughput. Significantly improved throughput is possible with the use of a parallel imaging scheme such as that found in photoelectron emission microscopy (PEEM) and low energy electron microscopy (LEEM). The application of PEEM and LEEM to device structures relies on contrast mechanisms that distinguish differences in dopant type and concentration. Interestingly, one of the first applications of PEEM was a study of the doping of semiconductors, which showed that the PEEM contrast was very sensitive to the doping level and that dopant concentrations as low as 10{sup 16} cm{sup -3} could be detected. More recent PEEM investigations of Schottky contacts were reported in the late 1990s by Giesen et al., followed by a series of papers in the early 2000s addressing doping contrast in PEEM by Ballarotto and co-workers and Frank and co-workers. In contrast to PEEM, comparatively little has been done to identify contrast mechanisms and assess the capabilities of LEEM for imaging semiconductor device strictures. The one exception is the work of Mankos et al., who evaluated the impact of high-throughput requirements on the LEEM designs and demonstrated new applications of imaging modes with a tilted electron beam. To assess its potential as a semiconductor device imaging tool and to identify contrast mechanisms, we used LEEM to investigate doped Si test structures. In section 2, Imaging Oxide-Covered Doped Si Structures Using LEEM, we show that the LEEM technique is able to provide reasonably high contrast images across lateral pn junctions. The observed contrast is attributed to a work function difference ({Delta}{phi}) between the p- and n-type regions. However, because the doped regions were buried under a thermal oxide ({approx}3.5 nm thick), e-beam charging during imaging prevented quantitative measurements of {Delta}{phi}. As part of this project, we also investigated a series of similar test structures in which the thermal oxide was removed by a chemical etch. With the oxide removed, we obtained intensity-versus-voltage (I-V) curves through the transition from mirror to LEEM mode and determined the relative positions of the vacuum cutoffs for the differently doped regions. Although the details are not discussed in this report, the relative position in voltage of the vacuum cutoffs are a direct measure of the work function difference ({Delta}{phi}) between the p- and n-doped regions.

Using low-energy electron microscopy, we measure the diffusion of Pd into bulk Cu at the Cu(100) surface. Interdiffusion is tracked by measuring the dissolution of the Cu(100)-c(2 x 2)-Pd surface alloy during annealing (T > 240 C). The activation barrier for Pd diffusion from the surface alloy into the bulk is determined to be (1.8 {+-} 0.6) eV. During annealing, we observe the growth of a new layer of Cu near step edges. Under this new Cu layer, dilute Pd remaining near the surface develops a layered structure similar to the Cu{sub 3}Pd L 1{sub 2} bulk alloy phase.

Abstract not provided.

Abstract not provided.

Abstract not provided.

Abstract not provided.

Abstract not provided.

Abstract not provided.

We have used scanning tunneling microscopy and low-energy electron microscopy to measure the thermal decay of two-dimensional Cu, Pb-overlayer, and Pb-Cu alloy islands on Pb-Cu(1 1 1) surface alloys. Decay rates covering 6-7 orders of magnitude are accessible by applying the two techniques to the same system. We find that Cu adatom diffusion across the surface alloy is rate-limiting for the decay of both Pb and Pb-Cu islands on the surface alloy and that this rate decreases monotonically with increasing Pb concentration in the alloy. The decrease is attributed to repulsive interactions between Cu adatoms and embedded Pb atoms in the surface alloy. The measured temperature dependences of island decay rates are consistent with first-principles calculations of the Cu binding and diffusion energies related to this 'site-blocking' effect.

Abstract not provided.

The length scale of stress domain patterns formed at solid surfaces is usually calculated using isotropic elasticity theory. Because this length depends exponentially on elastic constants; deviations between isotropic and anisotropic elasticity can lead to large errors. Another inaccuracy of isotropic elasticity theory is that it neglects the dependence of elastic relaxations on stripe orientation. To remove these inaccuracies; we calculate the energy of striped domain patterns using anisotropic elasticity theory for an extensive set of surfaces encountered in experimental studies of self-assembly. We present experimental and theoretical evidence that elastic anisotropy is large enough to determine the stripe orientation when Pb is deposited on Cu(111). Our analytical and numerical results should be useful for analysis of a broad range of experimental systems.

Pb deposition on Cu(111) causes the surface to self-assemble into periodically arranged domains of a Pb-rich phase and a Pb-poor phase. Using low-energy electron microscopy (LEEM) we provide evidence that the observed temperature-dependent periodicity of these self-assembled domain patterns is the result of changing domain-boundary free energy. We determine the free energy of boundaries at different temperatures from a capillary wave analysis of the thermal fluctuations of the boundaries and find that it varies from 22 meV/nm at 600 K to 8 meV/nm at 650 K. Combining this result with previous measurements of the surface stress difference between the two phases we find that the theory of surface-stress-induced domain formation can quantitatively account for the observed periodicities.

We use low-energy electron microscopy to study the mechanisms of thermal smoothing on Rh(001) surfaces at high temperature. By examining the change of areas of two-dimensional islands as a function of time and temperature, we find that smoothing from 1210 K to 1450 K is limited by the rate of surface diffusion on terraces and not by bulk vacancy diffusion as observed in other systems in the same temperature range. However, the activation energy we measure for island decay is inconsistent with previous measurements and calculations of the activation energy of surface diffusion and the adatom formation energy. This inconsistency combined with an unexpectedly large activation entropy suggests a surface transport mechanism other than simple hopping of adatoms across the surface.

The energetics and thermal motion of the self-assembled domain structures of lead on copper were discussed. It was found that the self-assembled patterns arose from a temperature-independent surface stress difference of approximately 1.2 N/m. The domain patterns evolved in a manner consistent with models, when the lead coverage was increased.

Using low energy electron microscopy (LEEM), we have followed Cu(100) surface morphology changes during Pb deposition at different temperatures. Surface steps advance and two-dimensional (2-D) islands nucleate and grow as deposited Pb first alloys, and then dealloys, on a 125°C Cu(100) surface. From LEEM images, we determine how much Cu is being displaced at each stage and find that the amount of material added to the top layer for a complete Pb/Cu(100) c(4 × 4) reconstruction (a surface alloy) is consistent with the expected c(4 × 4) Cu content of 0.5 monolayer. However, as the surface changes to the Pb/Cu(100) c(2×2) overlayer, we find that the displaced material from surface dealloying, 0.22 ML, is more than a factor of two lower than expected based on a pure Pb c(2 × 2) overlayer. Further, we find that in the 70-130°C range the amount of Cu remaining in c(2 × 2) increases with increasing substrate temperature during the deposition, showing that surface Cu is alloyed with Pb in the c(2 × 2) structure at these temperatures. When holding the sample at 125°C, the transformation from the c(2 × 2) structure to the higher coverage c(5 √2 × √2)R45° overlayer structure displaces more Cu, confirming the c(2 × 2) surface alloy model. We also find the c(2 × 2) surface has characteristically square 2-D islands with step edges parallel to the (100) type crystallographic directions, whereas the c(5√2 × √2)R45° structure has line-like features which run parallel to the dislocation double rows of this surface's atomic structure and which expand into 2-D islands upon coarsening. © 2000 Elsevier Science B.V. All rights reserved.

Low energy electron microscopy (LEEM) is used to investigate the dynamics of Pb overlayer growth on Cu(100). By following changes in surface morphology during Pb deposition, we measure the amount of Cu transported to the surface as the Pb first alloys into the surface during formation of the c(4 × 4) phase and subsequently dealloys during conversion to the c(2 × 2) phase. We find that the added coverage of Cu during alloying is consistent with the proposed model for the c(4 × 4) alloy phase, but the added coverage during dealloying is not consistent with the accepted model for the c(2 × 2) phase. To account for the discrepancy, we propose that Cu atoms are incorporated in the c(2 × 2) structure. Island growth and step advancement during the transition from the c(2 × 2) to c(5√2 × √2) R 45° structure agrees with this model. We also use LEEM to identify the order and temperature of the two-dimensional melting phase transitions for the three Pb/Cu(100) surface structures. Phase transitions for the c(5√2 × √2) R 45° and c(4 × 4) structures are first-order, but the c(2 × 2) transition is second-order. We determine that rotational domains of the c(5√2 × √2) R 45° structure coarsen from nanometer- to micron-sized dimensions with relatively mild heating (∼ 120°C), whereas coarsening of c(4 × 4) domain requires considerably higher temperatures (∼ 400°C). In studies of three-dimensional island formation, we find that the islands grow asymmetrically with an orientational dependence that is directly correlated with the domain structure of the underlying c(5√2 × √2) R 45° phase.

Abstract not provided.

The authors use low energy electron microscopy to identify a correlation between the growth shape of three-dimensional Pb islands on Cu(100)and the domain structure of the underlying Pb overlayer. Deposition of 0.6 monolayer Pb on Cu(100) produces a compressed c(2x2) overlayer, designated c(5{radical}2x{radical}2)R45{degree}, with periodic rows of anti-phase boundaries. They found that heating the surface to temperatures above 100 C coarsens the orientational domains of this structure to sizes that are easily resolved in the low energy electron microscope. Three-dimensional Pb islands, grown on the coarsened domains, are found to be asymmetric with orientations that correlate with the domain structure. Once nucleated with a preferred growth orientation, islands continue to grow with the same preferred orientation, even across domain boundaries.

Wanting to convert surface impurities from a nuisance to a systematically applicable nano-fabrication tool, the authors have sought to understand how such impurities affect self-diffusion on transition-metal surfaces. Their field-ion microscope experiments reveal that in the presence of surface hydrogen, self-diffusion on Rh(100) is promoted, while on Pt(100), not only is it inhibited, but its mechanism changes. First-principles calculations aimed at learning how oxygen fosters perfect layerwise growth on a growing Pt(111) crystal contradict the idea in the literature that it does so by directly promoting transport over Pt island boundaries. The discovery that its real effect is to burn off adventitious adsorbed carbon monoxide demonstrates the predictive value of state-of-the-art calculation methods.