The Schottky barrier height of W on Si1-xGex/ Si has been investigated as a function of composition and strain retained in the alloy for a given composition. The barrier height to ntype films does not vary significantly while that to p-type films follows the same trends than the band gap: it decreases with x and the strain. These results suggest that the Fermi level at the interface is pinned relative to the conduction band.

Amorphous and polycrystalline silicon films are commonly used in a wide range of microelectronic and optical devices. Polycrystalline silicon is conventionally deposited by chemical vapor deposition (CVD) at temperatures in excess of 600°C. At these high deposition temperatures, thermal diffusion of dopants and thermally induced chemical reactions may occur within the substrate or device. Also, substrates with low melting temperatures such as germanium, may undergo irreversible deformation. In the present study, ion beam sputtering has been shown to enable the deposition of a stable polycrystalline silicon film on germanium as well as on silicon and glass substrates at temperatures as low as 350-400°C. The crystallization properties of silicon on the different substrate surfaces is reported. Crystallinity of the ion beam sputtered silicon films as a function of deposition temperature and substrate type is measured by X-Ray diffraction. These polysilicon films are shown to be thermally stable, have randomly oriented crystals, and have good adhesion to the substrates despite high compressive deposition stresses ranging from 700MPa to 1000MPa. Magnetron sputtered silicon films deposited on substrates in the same temperature range produced only completely amorphous films, with lower stresses and which are also thermally stable. This study demonstrated the feasibility of depositing extremely hard polycrystalline silicon films on germanium and other substrates by means of physical vapor deposition at temperatures as low as 350°C.

In lead–on–chip (LOC) packaging technology, the lead fingers are attached directly to the surface of the chip using a double–side adhesive tape. This method of chip attachment naturally concerns with regards to stress on the polyimide coated chip surface. Device failure related to fracture in the passivation layers and the Al–Si metal has been observed in temperature cycle tests. The dominant issue is polyimide cracking due to mechanical stress caused by internal stress of LOC packages. To investigate the effect of polyimide characterization, devices were fabricated with different polyimides. The degree of polyimide cracking was determined with scanning electron microscope (SEM) technique. The polyimides were characterized with chemicals, tensile measurement tester, thermal analyzer, fourier transform infrared spectroscope (FTIR), computer simulation and various package-level reliability tests. The results of the characterization and an explanation of the primary factors affecting chip surface damage are presented in this paper.

In this paper we report for the first time theoretical and experimental study on smallpulse DLTS measurements of deep levels in semiconductor heterostructures. A theoretical model has been developed on the basis of the Schodinger and Poisson's electrostatic equation. Distribution of charge density in the superlattice has been considered, especially transferred charges in the “narrow gap” sublayers. The calculated results indicate that tinder the 1017/cm3 doping condition, a 30mV small pulse corresponds to a 2nm “sampling space window”, it is enough to detect special signal of deep levels in each sublayer in the semiconductor heterostructures. A SiGe/Si sample has been measured by the small-pulse DLTS. The experimental results agree well with the theoretical prediction and show that the small-pulse DLTS is a good method to study deep levels in the semiconductor heterostructures.

In order to better understand the mechanics of metallization structures having very fine features, we performed in-situ, temperature-dependent X-ray strain measurements on Al-Cu films patterned into gratings with linewidths of 0.2-0.8μm, with and without a dielectric covering In comparison with earlier work on coarser structures, we find larger strains and less hysteresis, indicating an increased yield stress. The stress in unpassivated lines is almost independent of the linewidth. The slope of the stress/temperature curves for passivated lines increases with decreasing linewidth, so that the compressive stress at 500°C reaches 900MPa for the narrowest lines. The tensile stress at RT is smaller for the narrower lines, suggesting that these lines may not stress-void as much as wider ones. The significance of these results for stress voiding and electromigration will be discussed.

A fluorescent and diffraction X-ray spectrometer with a 0.8 μmϕ Xray beam has been developed. It allows simultaneous measurement of local strains and minute amounts of metal contaminants in fine ULSI devices. The minimum sample size for X-ray diffraction measurement with it is a 0.3 μm diameter by 0.2 μm deep volume. It is applied to analyze strain in Al lines, showing that strain in Al single-layer lines (no barrier) increases significantly as the line width is reduced below 1.5 μm. Introducing barrier metals reduces this dependence of strain on line width. It is also found that hillock and void formation has a very strong correlation to strain.

A study was conducted towards fundamentally understanding the nature and magnitude of thermal stresses in aluminum thin films on silicon and correlating them with deformation mechanisms that serve to relax the stresses imposed by the thermal strain. A number of experiments and theoretical considerations are described in order to describe the sources of strengthening and stress relaxation in such films. It is shown that the flow stress of an aluminum film corresponds to the critical stress required to drive dislocation glide events. This mechanism entails an inverse flow stress dependence on the film thickness and grain size. A novel set of experiments in which the effects of strengthening due to finite film thickness and the grain size were separated and quantified is described. The variation of these strengthening components with temperature during heating and cooling were noted. Theoretical considerations of diffusional relaxation combined with the results of grazing incidence X-ray scattering (GIXS) experiments suggest that grain boundary diffusion is much less effective in limiting film flow stress than would be expected from the strain rate relation derived for a free-standing film. Line broadening observations in the GIXS experiments suggested large changes in dislocation density in the tested films during thermal cycling. The nature of these variations are qualitatively consistent with those of the film flow stress, suggesting that strain hardening could be an additional strengthening mechanism in Al films during thermal cycling.

Existing work in the mechanical behavior of thin films focuses mainly on measurement of macroscopic properties without strong correlation to microstructural features. We used transmission electron microscopy (TEM) to characterize the microstructures of free-standing copper thin films both before and after monotonic tensile deformation in an ex-situ thin film tensile testing system, as well as during in-situ loading in the TEM. The defect structures contributing to plastic deformation were investigated with an emphasis on comparison to mechanisms known to operate in bulk copper. The thin film exhibited much lower ductility (approximately 1%) than that normally observed in bulk form (greater than 40%). The predominant plastic deformation mechanisms did not include the typical dislocation activity that occurs in bulk copper, but rather greatly inhibited dislocation interactions typical of stages I and II hardening only. The absence of those structures normally found in tensile-deformed bulk copper is attributed to the differences in characteristic sizes of features within the microstructure available for deformation in bulk versus evaporated thin film material, that is, grain size and film thickness. The thin film ductility is an order of magnitude lower than what has been observed in bulk, ultrafine-grained copper, implying that a pure thin film effect on ductility exists and is significant. Microstructural features both near to and far from the fracture surface regions will be presented.

The stresses in Al-0.75w%Si-0.5w%Cu unpatterned metallization on silicon wafers have been measured using substrate curvature and x-ray diffraction techniques after quenching in liquid nitrogen. Stresses were measured with and without phospho-silicate glass overlayers and SiO2 underlayers, and thermal cycling followed by relaxation at room temperature. It was found that cooling the substrates to 77 K and warming to room temperature caused the metallization stress to go from tensile to compressive. Subsequent heating of the substrates to above ∼70°C followed by cooling to room temperature caused the stress to become tensile. Both compressive and tensile stresses were found to relax at room temperature with a time constant of 2.3 ± 0.2 hours. The magnitude of stress relaxation was a function of temperature, being about 20 MPa after heating to 240°C. The metallization exhibited both compressive and tensile flow stresses of ∼100 MPa near room temperature.

The determination by X-ray diffraction of the elastic strain tensors and the corresponding stress tensors in patterned lines of tungsten has been performed to investigate the effect of various passivation. For unpassivated lines, the stresses are biaxial and decrease with decreasing line width. Passivation over patterned lines results in triaxial tensile stress. The stress along the line is not changed by the passivation. The stress across the line decreases as the line width decreases. The third component of stress, perpendicular to the surface, which appears with the passivation, increases with decreasing line width. High compressive intrinsic stress in the passivation does not result in high tensile stress in the metal line.

A bending beam technique was used to measure the mechanical stresses in AlSi(l%)Cu(0.5%) blanket films as well as in patterned lines (aspect ratio: 0.8) - unpassivated and passivated with SiNx - during thermal cycling from –170°C or room temperature to 450°C.

Main results are:

a) No significant differences in unpassivated and passivated blanket films with thickness ranging from 0.2 µm to 3.2 µm.

b) In unpassivated patterned lines of 0.8 µm thickness the stresses across the lines are very small, while parallel to the lines they show nearly elastic behaviour, except at high temperatures.

c) In passivated patterned lines the stresses are much higher than in blanket films, very similar parallel and across the line and nearly elastic. The stress relaxation is small compared with blanket films and depends strongly on the temperature.

The areas, perimeters, lengths, and widths of 998 electromigration induced voids on 38 test stripes have been measured by SEM and digital image analysis. Virtually all of the voids occurred along the passivation-conductor interface on the side of the stripe. Plots of number of voids or total void area versus time to failure do not extrapolate to zero voids at zero time to failure which suggests there is a certain number of active nucleation sites predisposed to voiding.

We have obtained theoretical stress-temperature curves for passivated Al lines undergoing thermal cycling. A finite element plane-strain cross-sectional analysis with a time-dependent constitutive property for Al, based on equations for discreteobstacle controlled plasticity, was performed. The parameters for this Al constitutive relation were obtained by fitting with experimentally obtained stress-temperature curves for Al blanket films on silicon. Theoretical results agree well with the x-ray diffraction experimental data of Besser et al.1 Using a time-dependent property for Al helps match the data better than a time-independent property. Theoretical stress-temperature curves were also obtained for the longitudinal, transverse, and normal stress components in aluminum lines for line-widths ranging from 0.5 to 10 µm. The hysteresis of the stress-temperature curve of Al gets less as the line-width gets smaller. All stress components in the Al line change substantially with linewidth for the same oxide thickness.

Electromigration and stress - induced voiding are two of the most important metallization failure mechanisms for integrated circuits. These mechanisms are not independent since stress voiding may affect electromigration. In this paper we examine the impact of stress voiding on electromigration failure of narrow Al alloy stripes, and show that the major effects of stress voiding are significant reductions in failure times, and a non-Arrenhius temperature dependence of the lifetime. We propose a model to explain these effects based on the formation of flux divergences at pre-existing voids due to stress gradients.

X-ray diffraction was used to measure the strain relaxation in passivated Al-0.5%Cu lines at 200°C after cooling directly from an anneal at the passivation deposition temperature. The strain decay is fit with a stress-assisted, thermally-activated relaxation model. A high-voltage scanning electron microscope was used to image the presence and growth of voids through the passivation. The time scale of the void growth is not the same as that of the hydrostatic relaxation, indicating that voiding is not solely responsible for the observed relaxation. The relaxation of the line is modeled using a von Mises yield criterion and allowing elastic compliance of the passivation. The magnitude of the calculated relaxation agrees with the measurements. It is suggested that a combination of voiding and passivation compliance is responsible for the measured hydrostatic strain relaxation.

Stress relaxation behavior of thermally induced stresses in passivated line structures is strongly influenced by the metal yield strength. For some line geometries, stress relaxation can lead to void formation. In this study, bending beam measurements have been carried out to measure the thermal stress and stress relaxation behavior of passivated Al(l wt.% Cu) line structures with 3, 1, and 0.5 µm line widths. Our results reveal that stress relaxation in Al(Cu) films and lines shows log(time) kinetics consistent with a thermally activated dislocation glide mechanism. The kinetics of stress relaxation depend on line geometry and temperature, which can be explained by a combined effect of temperature (mass transport) and shear stress (driving force).

Stresses in passivated copper lines were determined from strains measured using x-ray methods and compared to stresses calculated using finite element methods. Copper lines, 2 µm wide by 0.8 µm thick, were fabricated by DC magnetron sputtering and lift off patterning. The cross section of these lines was trapezoidal, with approximately 50° sidewalls. One set of samples was left bare, the others passivated with Si3N4 in several different thicknesses. Strain was determined using x-rays. Stress was calculated using elastic moduli adjusted for predominantly (111) fiber texture. Elastic-plastic finite element models provided comparisons of the effects of passivation thickness, rectangular and trapezoidal cross section, and mechanical anisotropy due to preferred texture on the hydrostatic stress in a line. A similar model for aluminum also provided a comparison case of stress as a function of passivation thickness.

Grain growth in thin-film strips is important to interconnect reliability because grain boundary structures strongly effect the rate and mechanism of electromigration-induced failure. Previous simulations of this process have indicated that the transformation to the fully bamboo structure proceeds at a rate which decreases exponentially with time, and which is inversely proportional to the square of the strip width. We have also reported that grain boundary pinning due to surface grooving implies that there exists a maximum strip width to thickness ratio beyond which the transformation to the bamboo structure does not proceed to completion. In this work we have extended our simulation of grain growth in thin films and thin film strips to consider the effects of variations in grain boundary energy. Boundary energy is taken to depend on the misorientation between the two neighboring grain and the resulting variations in grain boundary energy mean that dihedral angles at triple junctions deviate from 120°. The proportionality between boundary velocities and local curvatures, and the critical curvature for boundary pinning due to surface grooving also both depend on boundary energy. In the case of thin-film strips, the effect of boundary energy variability is to impede the transformation to the bamboo structure, and reduce the width above which the complete bamboo structure is never reached. Those boundaries which do remain upon stagnation tend to be of low energy (low misorientation angle) and are therefore probably of low diffusivity, so that their impact on reliability is probably reduced.

Electromigration lifetimes of multilayered metallizations of Al alloy (Al-0.5%Cu-l.0%Si) sandwiched with refractory metal films have been correlated to the interconnect microstructure and interactions with the underlying layer. In this study, Median Time to Failure (t50%, MTTF) data was obtained for multilayered Al alloy films by varying the Al deposition temperature from 200°C to 400° and the composition of the underlying substrate films. The Al alloys examined in this study were deposited over substrate films of sputtered Ti, reactively sputtered TiN and CVD TEOS deposited SiO2. All Al alloys were capped with a reactively sputtered TiN layer and covered with passivation.

Physical characterization using TEM, XRD, SIMS, and AFM, was performed to determine the microstructural and chemical properties of the films and correlate them to the MTTF of the packaged interconnects. The role of material properties such as the grain size and interactions of the metallization with the underlying layers were found to substantially impact the electromigration lifetime.

Based on data obtained from these studies, a number of interconnect metallization options will be discussed with respect to reliability and associated integration issues.

Grain size distribution and texture of thin films of sputtered copper were characterized as a function of underlayer choice, deposition temperature and annealing temperature. For Cu deposited on Ta at room temperature and annealed at 100, 150, 200 and 250 °C, abnormal growth of the (100) orientation was observed for annealing temperatures above 150 °C, resulting in a bimodal grain size distribution. This orientation was also seen to grow abnormally for films deposited and annealed under identical conditions, but with a W underlayer. (100) growth for this case was observed at annealing temperatures as low as 100 °C. Abnormal grain growth was suppressed for Cu deposited on Ta at 150 °C and annealed to 250 °C. A bimodal grain size distribution was observed in a similar sample deposited on W under identical conditions, however, the abnormally growing orientation for this case was observed to be the (111) orientation, not the (100) orientation.