It is important to understand the distribution of recoil-implanted atoms and the impact on device performance when ion implantation is performed at a high dose through surface materials into single crystalline silicon. For example, in ultralarge scale integration impurity ions are often implanted through a thin layer of screen oxide and some of the oxygen atoms are inevitably recoil implanted into single-crystalline silicon. Theoretical and experimental studies have been performed to investigate this phenomenon. We have modified the Monte Carlo ion implant simulator, UT-Marlowe (B. Obradovic, G. Wang, Y. Chen, D. Li, C. Snell, and A. F. Tasch, UT-MARLOWE Manual, 1999), which is based on the binary collision approximation, to follow the full cascade and to dynamically modify the stoichiometry of the Si layer as oxygen atoms are knocked into it. CPU reduction techniques are used to relieve the demand on computational power when such a full cascade simulation is involved. Secondary ion mass spectrometry (SIMS) profiles of oxygen have been carefully obtained for high dose As and BF2 implants at different energies through oxide layers of various thicknesses, and the simulated oxygen profiles are found to agree very well with the SIMS data.

No abstract

Ion implantation is a critical technology in semiconductor Ultra Large Scale Integration (ULSI). Binary collision approximation (BCA)-based Monte Carlo (MC) ion implantation simulators are commonly used to predict the impurity and damage profiles. A deterministic propagator is needed in these simulators to simulate the propagation of ions in crystalline materials. A search-for-target algorithm is frequently called to determine the collision partners and collision parameters in a deterministic propagator, and this is usually the computational bottleneck of MC ion implantation simulators. The standard search-for-target algorithm has been redesigned for computational efficiency and for economic usage of memory. Instead of searching for collision partners in a standard 29-atom crystal neighborhood identical to all ions, narrowed-down potential target lists are pre-computed based on the ion's relative position to a reference point as well as its direction of motion. The American National Standards Institution (ANSI) C++ standard container class bitset [1] is used to store such potential target lists, and the memory usage is very efficient. Combined with a quasi-simultaneous collision algorithm, the CPU times for MeV P and B implantation simulations are found to be reduced by more than a factor of two, rendering very reasonable computation times for MeV ion implantation simulations on standard workstations.

No abstract