A diagnostic which combines a quartz crystal microbalance ͑QCM͒ and a gridded energy analyzer has been developed to measure the metal flux ionization fraction in a modified commercial dc magnetron sputtering device. The sensor is mounted on a linear motion feedthrough and embedded in a slot in the substrate plane to allow for measuring the uniformity in deposition and ionization throughout the plane of the wafer. Radio-frequency ͑rf͒ power is introduced through a coil to ionize the Al atoms. The metal flux ionization fraction at the QCM is determined by comparing the total deposition rate with and without a bias that screens out the ions, but that leaves the plasma undisturbed. By varying the voltage applied to the grids, the plasma potential is determined. At a pressure of 35 mTorr, a magnetron power of 2 kW, and a net rf power of 310Ϯ5 W, 78Ϯ5% ionization was found.

A rotating magnet dc planar magnetron with a 33-cm diameter aluminum target is coupled with a secondary plasma source to ionize the sputtered metal neutral flux to control the angular distribution of the flux arriving at the surface of the substrate. For this purpose, a radio-frequency ͑rf͒ plasma is created between the sputtering target and substrate by a three-turn coil located in the vacuum chamber. The rf plasma increases the electron temperature and density, which results in significant ionization of the neutral metal flux from the sputtering target. By applying a small negative bias to the substrate, metal ions are drawn to the substrate at normal incidence. A gridded energy analyzer and a quartz crystal microbalance ͑QCM͒ were used to determine the ion and neutral deposition rates. From this, the ionization fraction of the flux incident onto the QCM is determined.

A computer simulation was created to model the transport of sputtered atoms through an ionized physical vapor deposition ͑IPVD͒ system. The simulation combines Monte Carlo and fluid methods to track the metal atoms that are emitted from the target, interact with the IPVD plasma, and are eventually deposited somewhere in the system. Ground-state neutral, excited, and ionized metal atoms are tracked. The simulation requires plasma conditions to be specified by the user. Langmuir probe measurements were used to determine these parameters in an experimental system in order to compare simulation results with experiment. The primary product of the simulation is a prediction of the ionization fraction of the sputtered atom flux at the substrate under various conditions. This quantity was experimentally measured and the results compared to the simulation. Experiment and simulation differ significantly. It is hypothesized that heating of the background gas due to the intense sputtered atom flux at the target is primarily responsible for this difference. Heating of the background gas is not accounted for in the simulation. Difficulties in accurately measuring plasma parameters, especially electron temperature, are also significant

No abstract

No abstract