7.1 Summary and Conclusions
In this chapter, the key results of this study will be summarized and key contributions of this research will be reviewed. Potential directions for investigations are also suggested. The major contribution of each theme presented in this work can be summarized as follows.
First, the characteristics and failure mechanisms involved in Cu electromigration have been investigated. In chapter 3, following the successful integration of Cu/low-k BEOL processes, package level electromigration tests have been performed using various low-k materials in order to compare the performance and verify the stability of the Cu dual-damascene process. Various stress conditions for a number of structures were studied enable an understanding of the failure modes and to identify the weak links in an interconnect system. In addition to a well-understood failure mechanism, the multimodality distributions in various Cu/low-k processes are fitted using various bimodal methods to obtain precise lifetime.
Second, methods of optimizing Cu electromigration performance through Cap/dielectric interface re-engineering have been reported. Chapter 4 describes the correlation between electromigration lifetime and Cu surface cap-layer process. The dual via structure not only extends the EM lifetime but also changes the major failure mechanism from instantaneous to long-lasting mode. The effects of a pre-clean and the cap-layer material on the MTF are significant. A SiCN+pre-clean A can improve the lifetime by 10x compared to a PESiN+pre-clean B process. The adhesion of the Cu/cap interface can be directly correlated to the electromigration MTF and activation energy. A Cu-silicide formation mechanism before cap-layer deposition was proposed to explain the enhancement of electromigration lifetime. A significant improvement of
electromigration (EM) lifetime is achieved through modification of the pre-clean step prior to cap-layer deposition and by changing Cu cap/dielectric materials.
Next, Chapter 5 outlines the effects of width scaling and layout variation on dual-damascene copper interconnect electromigration. There are two scenarios that cover the impact of width scaling on electromigration. One is the width <1 µm region, in which the MTF shows a weak width dependence, except for the via-limited condition. The other is the width >1 µm region, in which the MTF shows a strong width dependence. A theory was proposed to explain the observed behavior. For polycrystalline lines (width >1 µm), the dominant diffusion paths are a mixture of grain boundary and surface diffusion. The activation energy for the dominant grain boundary transport (width >1 µm) is approximately 0.2 eV higher than that of the surface and grain-boundary transport (width ~ 1 µm). The derived activation energies for grain-boundary and surface diffusion are obtained from the Cu drift velocity under EM stressing.
Finally, in Chapter 6, the Blech effect or the short-length effect on a dual-damascene Cu process and its temperature dependence is investigated by using a technologically realistic interconnect structure. A lower jL2 product shows large sigma values as a result of back-stress-induced TTF dispersion. In order to avoid the Blech effect, the EM test structure length should be longer than 200 µm at moderate stress currents. The difference in the relationship between MTF and jL can be seen to be separated by a distinct boundary. One region (i) is the electromigration behavior which follows Black’s empirical equation and has no significant back-stress. The other (ii) is the electromigration behavior that can be better represented by a modified Black’s empirical equation, and includes significant back-stress. As an alternate means of determining the threshold–length product (jL)C value, a model that relates it to the failure volume and atomic flux was proposed. The extracted values are consistent with the regression value from the modified Black’s empirical equation, and these (jL)C value were verified during an extended stress test period. The resulting threshold–length product (jL)C value
appears to be temperature dependent, decreasing with an increase in temperature in a range of 250oC to 300oC. Much insight has been gained through electromigration experiments with Cu dual-damascene technology to identify its distinctive behaviors.
7.2 Future Work
Almost all electromigration lifetime measurements have been carried out under a steady current (DC). However, in most applications, the current signals in the device are often applied under pulsed conditions, as shown in Fig.7.1. So the electromigration characteristics under pulsed conditions are of practical interest. This can be examined using a square current to study the effect on the lifetime of the on/off time. In practice, the actual waveform of the pulse current contains a transient part due to the charging and discharging of the circuit, so nearly square waveform is obtained at high frequencies, as shown in Fig. 7.2.
A number of investigations have reported the effect of frequency and duty cycle on the Al electromigration lifetime under pulsed current conditions [7.1-7.7]. Several models have been proposed to correlate the electromigration lifetime under pulse conditions to that of DC conditions. One assumes that the damage accumulates only when the current is on, and the current–off periods have no effect. The MTF under pulsed DC conditions is inversely proportional to the duty cycle and can be expressed in terms of a steady DC condition as:
MTF(pulse DC) = r
1MTF(steady DC) (7.1)
where r is the duty cycle defined as the percentage of time while the current is on. A second model assumes that the migrating ions experience an average current density (rj) rather than the individual pulse. Then, the MTF under pulse conditions can be expressed as:
MTF(pulse DC) = A( ) exp( ) kT
rj −n Ea = n r
1 MTF(steady DC) (7.2)
However, few studies have reported details of Cu electromigration under time-varying
current stress. A question remains whether Cu dual-damascene interconnects exhibit a similar behavior. The dual-damascene Cu via is connected to the metal (M2) trench above and separated from the metal (M1) trench below by a thin diffusion barrier to form a flux divergence. A very small mass material depletion at the cathode via may cause an open circuit. A complication arises in the experiment exists because of potential electromigration damage in the dual-damascene Cu via/line interface. In addition, the relaxation time for Cu (0.2 µs) has been reported to be about 100 times smaller that that for Al-2%Si (20 µs) [7.6]. The characteristics of Cu electromigration under a pulsed DC current are expected to be different from Al, which gives room for further investigation.
Many of the problems that result in Cu/low-k interconnect reliability issues will become more severe as feature sizes scale down, as current density rises, and as new materials are introduced. Continuing research is needed to fully understand the multi-variable nature of Cu/low-k interconnect reliability and provide more accurate models to ensure design-in reliability.
In the past 30 years, it has been found that shorter Al wires are substantially less susceptible
Fig.7.1 A schematic AC pulse current, the current is applied every P period and remains on for a time ton, so that the duty cycle r = ton/P.