Grain Boundary Electromigration Research Articles

Electromigration failure in ultra-fine copper interconnects

This paper presents experimental evidence suggesting that electromigration (EM) can be a serious reliability threat when the dimension of Cu interconnects approaches the nanoscale range. To understand the failure mechanism prevailing in nanoscale Cu interconnects, single-level, 400-µm long interconnects with various effective widths, ranging from 750 nm to 80 nm, were made, EM tested, and characterized in this investigation. The results indicate that interface EM (Cu/barrier) may be the predominant EM mechanism in all line widths. The evidence supporting the active Cu/barrier interface EM includes the fact that the EM lifetime is inversely proportional to the interface area fraction. Microscopic analysis of the failure sites also supports the conclusion of interface EM because voids and hillocks are found at the ends of the test strip, which is not possible if lines fail by grain-boundary EM in the test structure used in this study. In addition, our study finds evidence that failure is assisted by a secondary mechanism. The influence of this factor is particularly significant when the feature size is small, resulting in more uniform distribution of failure time in narrower lines. Although limited, evidence suggests that the secondary factor is probably attributed to pre-existing defects or grain boundaries.

Microtexture and electromigration-induced drift in electroplated damascene Cu

In this work, the electromigration (EM) performance of electroplated damascene Cu is investigated by drift experiments on Blech-type test structures in both polycrystalline and bamboo microstructures. For the first, microtexture data were obtained from electron backscatter diffraction as well. While both bonding areas and 10 μm wide lines were found to have a predominantly random grain orientation, the drift studies indicated the importance of strongly segregating impurities in controlling Cu grain-boundary EM. For the bamboo lines, the impact of different barrier layers has been investigated, comparing Ta, TaN, and TiN. Drift was shown to proceed in all cases at the metallic Cu barrier interface, but faster for the Ta as compared to the TaN and TiN barriers. Cu drift data were finally compared to available literature results and to our previous drift studies on Al(Cu).

On the effect of grain boundary segregation on electromigration driving force in thin metal films

Electromigration (EM) describes flux of atomic diffusion induced by electron and hole “winds”, i.e. by fluxes of electrical charge carriers. In use conditions of thin metal film lines referred to as interconnects, electric current induced diffusion proceeds mainly along grain boundaries (GB), and fast grain boundary electromigration (GBEM) has been recognized as a major failure mechanism of modern integrated circuits. Despite intensive studies concentrated mainly on the kinetic (diffusivity) aspects of the problem, control and prevention of GBEM remains a challenge, both scientific and technological. We explore a novel line of attack on the problem. The approach outlined addresses to the driving force of GBEM and considers how it can be reduced, and even eliminated, through equilibrium GB segregation (GBS) of specially selected impurities. We consider this possibility by the example of copper, that is the most promising substitute for the traditional aluminum interconnects, though it may suffer from GBEM problems. The estimates based on simple uniformly-charged-slab GB model suggest that a submonolayer coverage with electronegative impurities, e.g. sulfur in copper, may result in hole conductance at GB, and hence, in partial or full compensation of the electron-wind, as the grain-boundary electromigration driving force, by the hole-wind.

Analysis of grain-boundary electromigration

A general analysis for grain-boundary electromigration applicable for type-A, -B, and -C diffusion kinetics is presented. Following Whipple’s approach in analyzing grain-boundary diffusion, solutions for composition profiles have been derived for isolated boundaries (for type-B and -C kinetics) and for interacting boundaries (for type-A kinetics). Asymptotic expressions for the exact solution have been obtained using the method of steepest descents and conditions for their validity are specified. The asymptotic solutions with opposite driving forces can be linearly combined into two relationships which relate the slopes of the composition profiles directly to the grain-boundary diffusivity and the effective charge. These relationships provide useful formulas to analyze the edge profiles measured in cross-stripe experiments. Martin’s solution was found to provide a reasonable approximation for the composition profile within the valid range of the asymptotic solutions.

Grain-boundary electromigration in thin films. I. Low-temperature theory

A Macroscopic mathematical theory is presented accounting for grain-boundary diffusion and electromigration in the presence of a simultaneous flux of atoms into the surrounding bulk lattice. The model employs a semi-infinite bicrystal geomettry with a constant source at the origin, and both integral and numerical solutions to the subsequent non-steady-state transport equations are given. ONly the case where V2/4Db-DL/δ2≳0, which is ttrue at low temperatures, is considered in detail. (V, Db, and aδ are the grain-boundary ion drift velocity, diffusivity, and width, respectively, while DL is the lattice diffusivity.). It is predicted that the time dependence of the displacement of a front of fixed concentration is linear only at short times and parabolic at longer times or with increasing amounts of atoms leakage into the bulk. In addition, the concentration profiles are predicted to have a plateau which becomes less apparent at higher temperatures. A comparison between the present theory and a previous treatment based on an extension of the Fisch analysis will be made. Application to recent results in thin films will be discussed.

Grain-boundary electromigration in thin films II. Tracer measurements in pure Au

The first direct measurement of the grain-boundary ion drift velocity in thin Au films over the temperature range 120–250 °C is reported. Central regions on the narrow stripe conductors were selectively embedded with 195Au tracer atoms and the extent of the subsequent transport was evaluated by a high-resolution autoradiography technique employing the scanning electron microscope. In one operating mode, the secondary electron image of the exposed Ag grains in the photographic emulsion was revealed. The second mode of operation involved detection of fluorescent Ag-Lα x rays, generated upon impingement of the finely focused electron beam on the Ag grains. The resultant spatial dependence of the x-ray intensity was ultimately converted into tracer activity profiles. It was found (1) that Ion transport in grain boundaries of Au is anode directed (this is the case whether or not a Mo glue layer is present), (2) that the ion drift velocity in grain boundaries can be expressed as V=[(5.7±0.6) ×10−2j/T] exp [−(0.80±0.03 eV)/kT] (cm/sec), with j given in units of A/cm2, (3) that the effective valence of ions migrating in grain boundaries ranged from −89 at 120 °C to −20 at 250 °C (for a diffusion correlation factor of unity), and (4) that a plateau was observed in the low temperature (<180 °C) profiles. (At higher temperatures, the plateau was less apparent and the integrated intensity was higher). This last result is in semiquantitative agreement with the theory developed by the authors to account for grain-boundary diffusion and electromigration, in the presence of simultaneous atom leakage into the lattice.

Grain-boundary solute electromigration in polycrystalline films

It is demonstrated that with proper analysis of the diffusion profiles, the effective charge and diffusivity for grain-boundary electromigration can be separately measured. Electromigration profiles are analyzed based on Fisher-type approximations for an equiaxial and perfectly textured grain structure. Electromigration profiles were obtained for Cu in Al at 255°C by measuring the spreading of a Cu cross-stripe subject to a direct current; the results were checked by annealing experiments performed under similar conditions. Cu was observed to migrate in the direction of electron flow. Results obtained from analysis of the anode and annealing profiles are −16.8±1.2 for the effective charge and (1.75±0.12) × 10−14 cm3/sec for the diffusivity parameter αDbδ. The over-all consistency of the results is better than 5%. Whipple analysis gives a more exact value of (2.74±0.16) × 10−14 cm3/sec for αDbδ.

Electromigration and metalization lifetimes

A model has been developed to predict the lifetime τf of integrated circuit metalizations which operate at high dc densities Je. Grain-boundary electromigration, internal heat generation, and current crowding at growing voids dominate the rate processes that lead to failure. Joule heating of the stripe causes an initial temperature rise ΔT0 and heat flow into the substrate. If this rise is appreciable an instability exists in the stripe. When vacancies electromigrate down grain boundaries and precipitate on a suitable boundary, forming an elongated void, the electric current will be diverted. This is serious if the crack has a substantial length component perpendicular to the current flow which increases vacancy currents to the crack tip and the local heating. An analytical model considering these effects and the time for a crack to propagate across the stripe width yields a stripe lifetime integral which fits the form Pn =(1/2)(1+0.265γΔT0). Here γ=ΔH/kT2 is the temperature coefficient of the diffusion constant. Pn, the crack width, and the initial electromigration grain boundary flux then largely determine the stripe lifetime. The self-heating contribution ΔT0 is shown to be an important term in the interpretation of accelerated test data and for proper extrapolation to lower temperature-and-current stress levels. The stripe temperature coefficient of resistance and melting point are shown to have only secondary effects on lifetime. Lifetimes have a Je−n dependence with n varying from unity at low ΔT0 levels to 15+ for high ΔT0 levels and are determined by the stripe and heat-sink temperatures. For maximum stripe lifetimes, wide stripes with good thermal coupling to the heat sink are desirable.