Electromigration refers to the motion of atoms induced by the flow of electric current. It increasingly affects the reliability of integrated circuits as the dimensions of the metal lines that connect the transistors on a chip become ever smaller with each new technology generation, so that the current density in these interconnects can be enormous (around a million amperes per square centimeter). The resulting atomic transport leads to the formation of voids or metal extrusions and, eventually, circuit failure due to breaks in the lines or short circuits with neighboring metal areas. Insight into the details of electromigration has awaited techniques able to measure local stresses with micron spatial resolution.

Enter x-ray microbeam techniques and microdiffraction, in particular. X rays make an ideal probe because they can be focused to submicron spot sizes to probe individual grains within the patterned polycrystalline metal films that represent interconnect lines on a silicon chip. X rays can also penetrate through passivating layers, such as silicon dioxide, that overlie the metal lines. The ALS has been one of the centers of microdiffraction activity with the development of Beamline 7.3.3, a bend-magnet beamline that provides a white-light beam spanning the photon-energy range from 6 to 14 keV for Laue diffraction measurements. A Kirkpatrick-Baez pair of bendable elliptical mirrors focuses the beam to a spot 0.8 µm by 0.8 µm, and a CCD area detector records the diffraction patterns.

Laue diffraction pattern of grains in an aluminum line on a silicon substrate.

Caught in the Act

Anybody who has contemplated a boulder-strewn beach after a big storm implicitly understands the power of large numbers of little things (the water molecules) acting in concert to push around much bigger things (the boulders). The same phenomenon plagues the metal conductors that connect the hundreds of millions of transistors on a state-of-the-art computer chip. Here the small things are electrons and the big things are the atoms that make up the conductor. Owing to the small cross-section of the connectors, nowadays only a fraction of a micrometer wide and much thinner than that, the electric current density is so immense that the raging electrons dislodge the atoms and carry them away. It is easy to imagine that over time, such atomic transport can lead to breaks in a connector and, hence, failure of the chip to operate correctly. The first step to solving this increasingly urgent problem, known as electromigration, is understanding exactly how it occurs. At the ALS, researchers have developed and now put to use an x-ray technique (microdiffraction) that is able to look with sub-microscopic resolution at local stresses in metal conducting lines on test microchips, thereby catching the early stages of electromigration in the act.

In their electromigration experiments, the researchers studied a test line consisting of an aluminum (plus 0.5-weight-percent copper) strip 30 µm long by 4.1 µm wide by 0.75 µm thick that was sputtered onto a silicon substrate and covered by a 0.7-µm-thick passivation layer of silicon dioxide. The first set of measurements made with no current applied yielded a map showing the orientation of each of the grains in the aluminum line and the diagonal components (i.e., along the length, across the width, and through the thickness) of the distortional (deviatoric) stress tensor for each of the grains. The changing values of these components from grain to grain demonstrated that the stress state was far from homogeneous and that appreciable local stress gradients existed even without an applied current.

Next, the experimenters increased the current to 30 mA in steps of 10 mA; after 24 hours, they turned off the current for 12 hours; then they reversed the current to -30 mA for 18 more hours. At the 30-hour point, they observed gradients from the anode to the cathode in both the width of the diffraction peaks and the changing angular positions of the diffraction peaks. During the 54-hour experiment, they saw that the distortional stress components averaged over all the grains increased while the current was on, relaxed when it was off, and increased again when the current was reversed. Taken together, these findings demonstrate the existence of electromigration-induced plasticity, most likely due to local shear stresses as metal is removed from the cathode end and deposited at the anode end. Such plastic deformation, which results in rotation and concave bowing of the grains, occurs before formation of failure-causing voids or hillocks.

Research conducted by B.C. Valek and J.C. Bravman (Stanford University); N. Tamura, A.A. MacDowell, R.S. Celestre, and H.A. Padmore (ALS); R. Spolenak and W.L. Brown (Lucent Technologies); T. Marieb and H. Fujimoto (Intel Corporation); and B.W. Batterman and J.R. Patel (ALS and Stanford Synchrotron Radiation Laboratory).

Research funding: U.S. Department of Energy, Office of Basic Energy Sciences (BES), and Intel Corporation. Operation of the ALS is supported by BES.

Publication about this research: N. Tamura et al., "High Spatial Resolution Grain Orientation and Strain Mapping in Thin Films Using Polychromatic Submicron X-Ray Diffraction," Appl. Phys. Lett. 80, 3724 (2002), and B.C. Valek et al., "Electromigration-Induced Plastic Deformation in Passivated Metal Lines," Appl. Phys. Lett. 81, 4168 (2002).

ALSNews Vol. 214, January 22, 2003