Thin Film and Interconnect Reliability

Our goal is to develop new ways to evaluate the reliability of thin films and interconnects in their as-manufactured states. Such tests are particularly important for nanoscale structures, where extrapolation of bulk properties is largely inaccurate. To generate fatigue in situ, we apply controlled alternating currents which, in turn, generate thermo-mechanical stress. This method can be applied to advanced integrated circuits, discrete devices, optoelectronic components, and sensors.

High-performance devices, such as microprocessors and memory chips, are typically composed of structures of specific dimensions and geometries and are made up of various materials, all in close proximity in a three-dimensional system. Material interactions lead to reliability issues such as fracture, delamination, fatigue cracking, and void formation. To measure material response in such complicated systems, we have developed an electrical method to determine the properties of these materials without the need for special specimens. Our test applies controlled joule heating via a 4-point probe system. Conditions are such that electromigration does not take place; rather, heat is generated and dissipated in each power cycle. Cyclic thermal strain is induced, allowing for determinations of fatigue lifetime and estimates of strength. Whereas tests that directly measure force-displacement response are restricted to certain specimen geometries or limited to surfaces, the electrical approach can evaluate extremely narrow lines, including those buried beneath other materials. We have validated our method using aluminum, gold, and state-of-the-art damascene copper lines and vias.

Application of low-frequency alternating current to a film on a substrate results in cyclic thermal straining of the film. We have made significant strides in understanding the materials science associated with this phenomenon in order to confirm its validity as a reliability test. We have also demonstrated its applicability to industrially relevant structures.

Materials Science
Rapid thermal cycling due to ac stressing can cause severe damage to patterned films, as seen in the figure, obtained from a cycled nonpassivated aluminum line. Electron microscopy showed that damage arises from dislocation activity, specifically the development of surface topography by formation of intrusions and extrusions. Continued cycling eventually leads to open circuit failure, as selected regions become so seriously damaged that they can no longer sustain current flow. This result showed that an electrical method can indeed be used to perform thermal cycling, which is a common product qualification test used by the electronics industry. Advantages of the more localized test include easier isolation of root failure causes, significantly faster testing, and lower energy consumption.

Reliability Tests for Industry
These processes result in overall cyclic lifetime behavior strongly resembling classic S-N curves (stress vs. cycles to failure) in metal fatigue. As cyclic deformation depends heavily on interactions with any neighboring material, such as those in a complex interconnect architecture, we initiated studies on oxide-passivated damascene copper lines supplied by a semiconductor tool supplier. We found that these fully constrained lines also exhibited S-N-like behavior, as shown in the figure, but withstood much higher temperatures and stresses than those shown by nonpassivated lines. This result showed that not only is an electrical method for measuring reliability feasible for industrial structures, it can distinguish material quality differences arising from different processing methods.Our work on cyclic behavior also led to a new method for estimating strength of patterned thin films on substrates. We directly compared this new method with a more established method, microtensile testing, to further validate the utility of the electrical approach.By extrapolating S-N data back to a single load reversal, we estimated the ultimate strength of a film while it remained attached to its substrate. A comparison between this method and microtensile testing resulted in values for ultimate strength that agreed within experimental uncertainty. This result suggested that multiple forms of mechanical reliability can be measured electrically through a single series of tests.