Cu Resistivity in Narrow lines: Dedicated Experiments for Model Optimization

  • PDF / 775,764 Bytes
  • 6 Pages / 612 x 792 pts (letter) Page_size
  • 9 Downloads / 191 Views

DOWNLOAD

REPORT


0914-F09-07

Cu Resistivity in Narrow lines: Dedicated Experiments for Model Optimization Sylvain Maitrejean1, Roland Gers1, Thierry Mourier1, Alain Toffoli1, and Gérard Passemard2 1 CEA-LETI, 17, Rue des Martyrs, Grenoble, 38000, France 2 STMicroelectronics, 850, rue J. Monet, Crolles, 38920, France

ABSTRACT As the dimensions of interconnects shrink, the confinement effects associated with a very small grain size lead to a significant increase in the resistivity of the metal. Cu resistivity models have been proposed using the classical Fuch and Sondheimer approach for surface effect, and the Mayadas and Shatzkes approach for the grain boundary effect. In these models, three adjustable parameters must be used. Good agreement between experimental data and models can easily be obtained. However, numerous fitting parameter sets can be used with equivalent fitting quality and opposite physical meaning. In this work, experiments dedicated to model parameter extraction are proposed and released. They are based on the used of Cu lines with various line widths and heights. Classical resistivity increase with line width and line height decrease is observed. The resistivity behaviour is modelled. In this case, limited fitting parameter options are obtained. For Cu narrow lines confined with Ta, these parameters suggest maximum surface effect, medium grain boundary effect and low impurity content inside the lines. INTRODUCTION Because of the increasing integration level of modern IC’s, metallic interconnects have reached nanometer-scale dimensions. The leap from Al to Cu based technologies have partially resolved the resistance and heat problems linked to the Joule effect in microelectronic components. However, at the nanometer scale, the electron mean free path becomes similar to the characteristic dimensions of the structure of the metallic line. Indeed, at 303 K, the mean free path of pure Cu is equal to 38nm whereas typical line width and grain size are below 100nm. The charge carriers are then scattered additionally at the interfaces. Thus, as experimentally observed for thin films at the beginning of the 20th century [1], large metal resistivity increase occurs. This increase has a negative impact on circuit performance and, for this reason, has been widely studied in recent years [2-8]. A precise Cu resistivity model shall be a useful tool for interconnect process improvement and interconnect design optimization. The ultimate goal shall be to predict resistivity as a function of metallization schemes and line geometries. The effect of wire surface is usually described using the Fuch – Sondheimer (FS) theory [9, 10]. In this theory, a phenomenological parameter p is used to describe the probability of elastic reflection of the electrons at the surface: a p equal to 1 induces total elastic reflection of the electrons and no surface effect. Concerning grain boundary effect, Mayadas and Shatzkes (MS) approach [11] is frequently followed. Another phenomenological parameter R is used. It represents the occurrence of charge c