Nanoindentation was used to examine the impact of impurities and grain boundaries on the mechanical properties of a “model” (110)/(100) grain boundary (GB) interface prepared using direct silicon bonding via the hybrid orientation technique of (110) and (100) p-type silicon wafers. Remarkable differences were found between the mechanical behavior of Fe- and Cu-contaminated samples. The direct silicon bonded wafers contaminated with either Fe or Cu showed opposite effects on mechanical properties, with Fe enhancing the silicon hardness, while Cu contamination induces a gradual weakening. High-resolution transmission electron microscopy was used to verify that the abrupt hardness changes observed during increasing nanoindentation loading is attributed to local deformation induced by the GB interface, Cu precipitate colony induced dislocations, and the abrupt crystallographic orientation change across the GB. The resulting dislocation loop generation facilitated the deformation process during nanoindentation and therefore softened the material.

Since the initiation and propagation of a micro-crack in a silicon wafer introduces local variations in stress, it is critical to the understanding of wafer breakage that accurate profiling of stress be performed in the vicinity of the micro-crack. In this study, nanoindentation has been used to investigate the stress-relaxation during crack initiation and propagation in material of particular interest to the photovoltaic (PV) industry. The low load (<1 mN) capability of a Hysitron Triboindenter® was used to accurately profile the extent of plastic deformation and resulting amorphization. Measurements were made on Si samples extracted from top, middle and bottom of a (100) oriented single crystal ingot to evaluate the impact of different carbon, oxygen and metallic impurity concentrations. A gradual but significant drop in hardness from 10.2 to 6.9 GPa occurred as indents were made closer to the micro-crack and was attributed to local amorphization. Electron back scattered diffraction (EBSD) and Raman spectroscopy confirmed the amorphization, respectively, at nano- and micro-scale.

Examination of dislocations in the as-grown and annealed SiGe/Si heterostructures by DLTS indicates the three strong DLTS bands from 70 to 270K in the as-grown sample are likely related to intrinsic point defects or dislocation trails. A small amount of Fe at dislocations dramatically increases their electrical activity, and the trap concentration due to Fe-decorated dislocations can well exceed the total Fe impurities presented along dislocations. Through examining the competitive trapping of Fe at boron and dislocations, it is suggested that Fe trapping only happens at disordered sites along dislocations, such as kinks.

The strain in the strained Si layer on a blanket strained Si/SiGe structure could not be determined with only convergent beam electron diffraction to high order Laue zone (HOLZ) line splitting. Combined with CBED and finite element calculations, we quantified the deformation field from HOLZ line splitting and demonstrated a procedure to determine the initial strain in the strained Si layer. Our results also gave us insights in strain relaxation in a TEM sample. The CBED technique combined with FE modeling has the potential for initial strain measurements on new generation short channel CMOS technology nodes.

Nitrogen doped silicon samples were irradiated with 200 kV electrons in a transmission electron microscope (TEM). The resulting room temperature point defect creation, bonding and segregation were studied by in situ conventional and Z Contrast TEM imaging and electron energy loss spectroscopy (EELS). Energy loss spectra from areas attributed to be rich in vacancies or silicon self-interstitials are found to be significantly different from the bulk in the near-edge structure of their Si-L2,3 edges. The experimental results are compared with ab initio density of states calculations for electronically excited atoms near relaxed point defect structures and plans are outlined to extend this technique to individual point defect characterization.

The lateral motion of iron impurities was observed and studied in ptype iron contaminated silicon. The lateral diffusion was induced by and then measured using Schottky diodes with a special interdigitated fingers design. Capture of the impurities was done by diffusing to laterally placed dislocation loops formed by a self aligned ion implantation. Lateral changes in Fe concentration were determined using capacitance-voltage and deep level transient spectroscopy.

Capillary instability poses a serious concern for applicability of thin layers in next generation of microelectronics devices. Two completely independent models, Mullins' Grain Boundary Grooving model and the Geometrical Agglomeration model have been used to explain capillary instabilities in CoSi2 layer formed on polycrystalline and single crystal substrates, respectively. Inadequacies of these separate models have been highlighted in our experimental work using cross-sectional transmission electron microscopy. A new empirical model, which incorporates both of these models, will be presented which shows the dependence of groove depth on annealing duration, diffusion coefficient, annealing temperature and grain diameter to film thickness ratio, D. The new model not only explains the various peculiarities of silicide grooving on both single and polycrystalline substrates, it also predicts the severe agglomeration of nanoscale layers experimentally observed during silicide formation by Co/Si reaction at 700 °C, 10 sec rapid thermal anneal. Also, experimental verification of agglomeration suppression schemes which are based on the above empirical model will be presented.

Co silicide formed via selective diffusion of Co through a Ti interfacial layer has been reported by several groups. In this report, phase identification of the silicide has been further studied for both Co/Ti and Ti/Co multi- and bilayers deposited on p(100)-Si substrates. The samples were either vacuum furnace or RTA annealed from 550°C to 900°C. The Co silicide formation sequence in the Co/Ti-Si systems follows CoSi2→Co2Si→CoSi→CoSi2 with the formation temperature increasing for each phase. The Co/Ti bilayer CoSi to CoSi2 transformation temperature was lower than that for the six layer Co/Ti system. For the multilayer sample with Co as the first layer in contact with the Si substrate, CoSi2 formed at 550°C and then CoSi was observed at higher temperatures due to the effect of Co supply on disilicide phase instability. Epitaxial CoSi2 growth occured at higher temperatures after the removal of the unreacted upper layers. A 15 μΩ-cm film resistivity was obtained from 50 nm epitaxial CoSi2.

The formation of a 12nm thick, continuous and thermally stable COSi2 layer was described in our previous work [MRS Proc. 238, 587 (1992)]. Interdiffusion in the Co/Ti-Si multilayer system has been further studied and the initial Ti(O) thickness is shown to be a critical parameter in controlling its effectiveness as a diffusion barrier, and in modulating the Co-Si and Ti-Si compctctive reactions. Three Ti(O) and three Co layers with thickness from ∼5nm 20nm were deposited sequentially, with Ti(O) as the first layer, on Si-(100) substrates by dual source thermal evaporation. The morphology of the CoSix/Si interface was strongly influenced by Ti(O) thickness from ∼5nm to ∼10nm, and a 12nm thick uniform CoSi2 layer with ∼28μΩ-cm resistivity was produced as decribed previously. When the initial Ti(O) thickness was increased to ∼20nm and the Co thickness set at -10nm, Co diffusion was suppressed and Ti reacted with Si yielding an ∼10nm amorphous TiSix layer at 550°C. This amorphous layer transformed to a 15nm thick uniform C-54 TiSi2 layer after selective removal of upper layers and a 750°C plus 800°C annealing. A flat silicide/Si interface and a ∼58μΩ-cm resistivity were obtained. The significance of both thermodynamic and kinetic factors in the compctetive reactions is discussed.

Cobalt silicide formed by diffusion of Co through a Ti compound has been reported for both bilayer and multilayer Co/Ti-Si structurest[2-4]. To compare the bilayer and multilayer systems in terms of the Co silicide interfacial morphology, both bilayers and six layers of 20nm Co and lOnm Ti were sputter deposited on (100)Si in a system where background. oxygen and carbon were gettered by each Ti layer. The samples were annealed from 550°C to 800°C for 60 sec by lamp RTA in N2 ambient. XTEM micrographs revealed that significant differences in interfacial morphology existed between bilayer and multilayer samples. The interfacial amorphous Ti(O.C) diffusion barrier layer was found to be more effective in the multilayer system producing uniform CoSix layers as thin as ∼5nm after 550°C RTA, whereas the silicide formed in the bilaver samples under the same condition was rough. RBS analysis showed that the transformation temperature from CoSi to CoSi2 was 800°C in bilayers and even higher for multilayers. The higher transformation temperature is attributed to the additional Co available in the multilayer system and its effect on Co monosilicide phase stability as previously reported[10].

A study of gettering and electrical activity of metallic impurities Ni, Au and Cu has been carried out on epitaxial Si/Si(2%Ge)/Si wafers containing interfacial misfit dislocations. The impurities were intentionally introduced from a backside deposited thin metal followed by rapid thermal annealing (RTA). Transmission Electron Microscopy (TEM) results indicate that the impurities were gettered along the misfit dislocations in near-surface regions either as Au precipitate colonies, or as NiSi2 and CuSi silicide precipitates. Data from Scanning Electron Microscopy (SEM) in the Electron Beam Induced Current (EBIC) mode revealed that these precipitates dominate the recombination properties of the initially inactive misfit dislocation.

In this work, Co/Ti multilayers were deposited on Si-(100) substrates by dual source thermal evaporation. Oxygen was found to be selectively incorporated into the Ti layers during the deposition. The samples were then treated by a two step annealing process. A —6nm Co2Si + CoSi2 layer was formed at the original Ti/Si interface after a 550°C, 2hr. annealing. The Ti(O) layers acted as selective diffusion membrane for Co and Si during this interaction. The morphology of the suicide layer was dependent on the thickness of the Ti(O) barrier layer. Selective removal via wet etching of the upper layers and a 750°C annealing produced stoichiometric and uniform epitaxial cobalt disilicide with a reduced film resistivity. A very flat top surface with no sign of groving or agglomeration and a much improved interface with the Si substrates were obtained.

This paper discusses the temperature dependence of recombination lifetime in a variety of silicon materials using energy level as a parameter. A theoretical approach based on the Shockley-Read-Hall theory for energy level calculations has been used. Various types of defects created by introducing impurities, dislocations and grain boundaries into silicon waferswere studied. Results are presented for Czochralski grown Si wafers intentionally contaminated with gold and chromium, EFG ribbon with varying concentration of oxygen, web ribbons with extended defects and contaminants, large grain polycrystalline material, and Si/Si-Ge/Si heterostructures with varying misfit and threading dislocation density.

The delineation of n-p junctions is one of the techniques routinely required in the analysis of electronic devices. The delineation of the junction profile is typically done either in a scanning electron microscope (SEM) or by angle lapping / staining/ optical microscopy. Recently TEM foils have been selectively etched to simultaneously provide an interface demarcation and an image of processed induced dislocations. A case study is presented in this report on concentration dependent delineation techniques with emphasis on the special problems associated with cross sectional transmission electron microscopy (XTEM) samples. The etchants used in delineating junctions for the SEM or with angle lapping / optical microscopy of transistors are too aggressive for the extremely thin TEM foils. This paper discusses etchants which deliver reproducible, high contrast delineations in XTEM samples with known concentration profiles.

A silicon epitaxial structure containing spatially confined arrays of misfit dislocations has been used in order to investigate the interaction between hydrogen and individual extended defects. Hydrogen was introduced using a Kaufman plasma ion beam source. A characteristic Si-H peak at 2100 cm-1 was obtained using multiple internal reflection infrared spectrophotometry. Microdefects such as gas bubbles and {111} planar defects were found near the surface, as well as at the misfit dislocation interfaces, using transmission electron microscopy. A heavily damaged region was obtained on the top Si surface and an extended area SEM/EBIC contrast was obtained due to a surface electrical field.

We describe how a simple qualitative understanding of the interfacial reactions occurring during typical ULSI processes for junction formation, dopant activation, and contact silicidation can be used to eliminate end-of-range interstitial dislocation loops and beneficially impact the diffusion of dopants. Following a brief discussion of the well-documented effects of oxidation and nitridation on extended defects and dopant diffusion, conditions for elimination of implantation-induced defects are specified. Cross-section and plan-view TEM along with angle lapping and chemical etching of implanted and diffused junctions are presented to illustrate the application of point defect engineering to process technology.