Jonathan C. Doan

The evolution of the resistance of aluminum interconnects during electromigration

Abstract Resistance monitoring is a traditional method to investigate electromigration failure. It is important to understand how much information can be extracted from the data generated by these experiments. To this end, precision resistance measurements were included as part of accelerated electromigration tests performed inside of a high voltage scanning electron microscope (HVSEM). Twenty-two passivated Al interconnects were tested at 30 mA/μm 2 and at two temperatures, half at 212°C and half at 269°C. During every test, our automated apparatus stored images of each 300 μm long structure several times per hour. The resistance of each line was also precisely measured and recorded. Changing the temperature affected only the time scale of the resistance evolution. There were resistance changes before voids formed that were neither due to temperature fluctuations nor solute effects. In most cases, the nucleation of the first void to form in a line was signaled by an increase in the time derivative of the resistance. Due to the strong effect of void shape, the void volume could not be determined by the magnitude of the resistance change. The width of a void (transverse to the line) rather than the volume largely determined the resistance change.

Effects of dielectric materials on electromigration failure

The effects of dielectric layers on electromigration failure were studied in situ using a high-voltage scanning electron microscope and at the wafer level using conventional accelerated testing. Several different passivation layers were deposited on wafers with A1 interconnect test structures. Prior to the deposition of the final dielectric, the wafers were processed identically and, whenever possible, simultaneously. Interconnects encapsulated with compliant polymer and very thin (0.1 μm) SiO2 layers demonstrated substantial lifetime extensions over those with more rigid (1 μm thick) SiO2 layers. Unpassivated lines behaved dramatically differently and failed much sooner than those covered with only 0.1 μm of SiO2. As expected, increasing the passivation thickness from 0.5 to 4 μm increased the electromigration lifetime for SiO2 covered specimens. The fabrication of silicon dioxide dielectrics using electron-cyclotron-resonance chemical-vapor deposition (CVD) and silicon nitride dielectrics via plasma-enh...

Stress-induced and electromigration voiding in aluminum interconnects passivated with silicon nitride

Void nucleation in aluminum interconnects passivated with silicon nitride was studied using a high voltage scanning electron microscope. Extensive stress-induced voiding was observed in these interconnects independent of the passivation thickness. Some of the stress-induced and electromigration-induced voids formed away from the interconnect sidewall in contrast to results from other studies of void nucleation in passivated aluminum lines. Nuclear reaction analysis measured large amounts of hydrogen in the aluminum films passivated with silicon nitride. Transmission electron microscopy showed a high density of nanometer-sized bubbles in the aluminum. These bubbles, which are thought to have formed from hydrogen that evolved from the silicon nitride layer during processing, served as nucleation sites for stress- and electromigration-induced voids.

Void nucleation on intentionally added defects in Al interconnects

Void nucleation in passivated aluminum interconnects was studied using high voltage scanning electron microscopy. To test theories about stress-induced and electromigration void nucleation, Ar ions were implanted into Al specimens. The Ar atoms precipitated and formed bubbles that served as nucleation sites with high surface energy. In the implanted samples, voids formed away from the interconnect sidewalls, in contrast to voids in ordinary passivated Al interconnects. The evolution of the void volume was also affected by the reduction in the nucleation barrier. These results strongly support the theory of void nucleation on interface flaws in Al interconnects.

Void phenomena in passivated metal lines: Recent observations and interpretation

Since the pioneering TEM work of Blech and Meieran thirty years ago, it has been evident that direct in-situ observations of electromigration in progress can provide information essential to the understanding of the phenomenon. Unfortunately, the practical difficulties of carrying out such observations have been so great that almost all experimental work on electromigration in passivated metal has been limited to lifetime measurements and post-mortem examination of failed specimens. Without an adequate qualitative picture, theoretical efforts at detailed modeling of the phenomenon have had limited success. In particular, the extremely large variance observed in life tests of virtually identical lines has not been satisfactorily explained. To make possible the necessary in-situ observations, Intel and Stanford have, for some years, been engaged in the development of specialized SEM technology leading to a High Voltage Scanning Electron Microscope (HVSEM) with complete computer control. A special computer g...

The relationship between resistance changes and void volume changes in passivated aluminum interconnects

We performed electromigration tests on 11 passivated Al lines at 212/spl deg/C and 3 MA/cm/sup 2/ inside a high voltage scanning electron microscope (HVSEM). Our automated experimental apparatus saved an image of the entire test structure every six minutes while it continuously measured and recorded the resistance of each line. We determined the void volume as a function of time using digital image processing. While it is straightforward to determine resistance changes knowing the changes in the void geometry, it remains to be seen whether one can perform the reverse calculation. From our data, the amount of information that can be extracted solely from the resistance versus time curves appears limited. The nucleation time of the first void can usually be found, but it is not possible to detect the nucleation of subsequent voids. The resistance change depends much more critically on void shape than void size, making an estimation of the void volume from the resistance impossible.

Model for localized buildup and thinning of metal due to electromigration

We present a model to account for the highly localized buildup and thinning of metal that has been observed during electromigration testing. It assumes variations in mobility along the line, rather than complete blockage. Using reasonable estimates for the parameters needed, it is possible to obtain a quantitative fit to previously observed data.

Comparison of electromigration behavior in passivated aluminum interconnects

Electromigration void behavior in Al-0.1%Sc and Al-0.5%Cu alloys was compared by conducting tests in a High Voltage Scanning Electron Microscope (HVSEM). Wafer level test results showed that the median time to failure for the Al-Cu lines was ten times longer than that of the Al-Sc lines. In-situ tests of the two alloys showed a similarity in void nucleation times but differences in void movement between the two sample sets; the interactions of voids with Al2Cu theta precipitates in Al-Cu lines were especially interesting. From our studies, it appeared that the addition of scandium to Al does not have the proposed benefits for electromigration resistance, as suggested by previous studies.