Effects of Cu and Si Dopants On Electromigration Mass Transport in Al Interconnects for VLSI

  • PDF / 1,357,667 Bytes
  • 6 Pages / 414.72 x 648 pts Page_size
  • 97 Downloads / 233 Views



then performed in a single wafer LAM 4620 RIE system using standard BC13 and Cl chemistries. Approximately 1000A of the oxide sublayer was removed during the overetch of the metal lines. To minimize any corrosion potential, the LAM 4620 system utilized a separate chamber for ashing the wafers prior to exposing them to atmosphere, followed by hot N 2 and a hot deionized water rinse. Wafers at this point were removed from the etch system and processed in a Semitool spray acid etch tool for an ethylene glycol/ammonium fluoride, hydrofluoric acid etch targeted to remove between 300-400A of oxide. The oxide etch selectively removed any further "veils" or chlorine residues from the etched Al lines. A final Semitool amine based solvent rinse was used to remove any remaining organic residues. The x-ray data was acquired using a split-ring HUBER 4-circle diffractometer equipped with an offset Phi-box and 6-inch x-y wafer scanning stage. The diffractometer was coupled to a 18kW rotating anode using a multi-capillary x-ray lens with a variable aperture. The x-ray lens collect approximately 8 degrees of solid angle (at full beam size) from a 0.5 mm X 10 mm point source, and redirected the x-rays into a 30 mm diameter parallel beam with a beam divergence of 0.22 degree FWHM. The x-ray lens provided a factor of approximately 8X more x-ray flux to the sample than did a standard Bragg-Brentano system, reducing acquisition time accordingly. Strain data was acquired using the standard sin 2,q method. The amplitude of the sin 2 q oscillations were measured and compared against strain-broadening analysis (Klug-Alexander method) to determine average microcrystalline strain variations. TEM, using a JEOL200CX transmission electron microscope, operating at an incident electron beam energy of 200 KeV, together with energy dispersive x-ray spectroscopy (EDS), was used to study the grain size distribution and the second phase precipitate identification. TEM samples were prepared from the wafers by dimpling from the substrate side and then ion milling to achieve electron transparency in the metal films. NIST structures with line widths of 1.0, 1.8, 3.0, 5.0 and 10.0 jim were stressed in the Sienna system at 200°C with a current density of 2 MA/cm 2. The failure criterion depended on the total current applied. For currents less than 50mA, the test was terminated if the resistance variation between two measurements exceeded 100%. For currents greater than 50mA, the test was terminated at a resistance variation greater than 250%. Sample size is 16 for all the tests in this study. RESULTS AND DISCUSSIONS Figure 1 plots the electromigration resistance MTF (median time to failure) vs. line widths of 1.0, 1.8, 3.0, 5.0, 10.0 gim for four different materials; Al-1.5% Cu, AI-1.0%Cu-l.0%Si, A]1.5%Cu-1.5%Si, Al-0.5%Cu-l.0%Si. For all line widths tested, the electromigration results ranged from best to worst in the following order: Al-1.5%Cu, Al-L.0%Cu-l.0%Si, AI-1.5%Cu1.5%Si, and Al-0.5%Cu-l.0%Si. TEM was used to evaluate the grain sizes as shown in Figu

Data Loading...