▪ Abstract  The increasingly rapid transition of the electronics industry to high-density, high-performance multifunctional microprocessor Si technology has precipitated migration to new materials alternatives that can satisfy stringent requirements. One of the recent innovations has been the substitution of copper for the standard aluminum-copper metal wiring in order to decrease resistance and tailor RC delay losses in the various hierarchies of the wiring network. This has been accomplished and the product shipped only since the fall of 1998, after more than a decade of intensive development. Critical fabrication innovations include the development of an electroplating process for the copper network, dual-damascence chem-mech polishing (CMP), and effective liner material for copper diffusion barrier and adhesion promotion. The present copper technology provides improved current-carrying capability by higher resistance to electromigration, no device contamination by copper migration, and the performance enhancement analytically predicted. This success of the shift to copper will accelerate the industry movement to finer features and more complex interconnect structures with sufficient device density and connectivity to integrate full systems on chips. The next innovation will be the introduction of low-dielectric constant material that, in combination with copper, will create added excitement as the industry learns how to utilize this new capability.

To address the future use of alloying elements for Cu interconnect applications in integrated circuits, first, available bulk experimental data such as residual resistivity per at. % solute and binary phase diagrams are used to arrive at a set of 24 potential elements. Next, experimental results in thin films and lines allow the authors to arrive at a smaller set that includes ten elements, namely, Pd, Au, Al, Ag, Nb, Cr, B, Ti, In, and Mn, with higher priority and six, namely, Zn, V, C, Mg, P, and Sn with lower priority for further studies. These additional studies are needed before a strong case for or against alloying additions to Cu can be made. The available thin film and line data are summarized in a series of tables that should prove useful for the readers. In particular, the thin film data allow the authors to obtain an effective average residual resistivity (EARR) per at. % solute that combines the effects of impurity scattering, second phase precipitates, and grain size refinement resulting from solute additions.

In situ depth-resolved positron annihilation spectroscopy (PAS) is used to show dynamic formation of vacancies in 1 μm×1 μm Al-0.5 wt % Cu lines under current flow. We show that the number of vacancies in these lines increases when a dc current (8×104 A/cm2) is applied. This increase in vacancy concentration is substantially greater than that due to thermal vacancy generation alone (4×1018 cm−3 versus 3×1017 cm−3). Isothermal measurements (with no current flow) yield a vacancy formation energy of 0.60±0.02 eV. These results show that PAS can be used to examine the initial stages of interconnect damage due to electromigration.

Articles you may be interested inThermal stress and plastic deformation of Al fine line structures: Effects of oxide confinement and line geometry AIP Conf. Proc. 305, 62 (1994); 10.1063/1.45701 Xray determination of strains, stress, and relaxation in interconnect metallizations AIP Conf. Proc. 305, 46 (1994); 10.1063/1.45700 Stress evolution during stress migration and electromigration in passivated interconnect lines AIP Conf. Proc. 305, 231 (1994); 10.1063/1.45694 Characterization of thin films by a pulsed positron beam AIP Conf. Proc. 303, 84 (1994); 10.1063/1.45549The effects of stress on thin film aluminum metallization AIP Conf.The salient features of the deformation mechanisms that can be obtained from stress-temperature plots of thin films deposited on substrates are discussed in this article. It is shown that the slope of the cooling portion of these plots at low temperatures indicates when dislocation-glide-dominated mechanisms are important and whether these occur without changes in microstructure. Furthermore, the behavior of the films during and after isothermal holds gives important information about the relaxation processes. Some of these general principles have been demonstrated by experimental observations on the shape of the stress-temperature curves for 1-m-thick copper, silver, and aluminum films deposited on silicon. Two specific observations can be made based on these results. First, for the materials and conditions examined, isothermal holds do not affect the residual stress state upon subsequent cooling. Second, while the behavior of the copper and silver films can be qualitatively described by constitutive models appropriate for bulk forms of those materials, the behavior of aluminum is anomalous. It is believed that the properties of the aluminum film are dictated by the presence of the native oxide which suppresses relaxation mechanisms associated with diffusion and dislocation climb. To support this contention, copper, silver, and aluminum films were passivated by SiO 2 layers. The shape of the stress-temperature plots for copper and silver changed qualitatively from one consistent with bulk relaxation mechanisms to one consistent with dislocation-glide mechanisms. The behavior of the aluminum film was unaffected by passivation.

Electroplated Cu was found to have a fine as-plated microstructure, 0.05 ±0.03 μm, with multiple grains through the film thickness and evidence of twins and dislocations within grains. Over time at room temperature, the grains grew to greater than 1 μm in size. Studied as a function of annealing temperature, the recrystallized grains were shown to be 1.6 ± 1.0 μm in size, columnar and highly twinned. The grain growth was directly related to the time dependent decrease in sheet resistance. The initial grain structure was characterized using scanning transmission electron microscopy (STEM) from a cross-section sample prepared by a novel focused ion beam (FIB) and lift-out technique. The recrystallized grain structures were imaged using FIB secondary electron imaging. From these micrographs, the grain boundary structures were traced, and an image analysis program was used to measure the grain areas. A Gaussian fit of the log-normal distribution of grain areas was used to calculate the mean area and standard deviation. These values were converted to grain size diameters by assuming a circular grain geometry.