Reliability of Tantalum Based Diffusion Barriers between Cu and Si
- PDF / 879,737 Bytes
- 6 Pages / 612 x 792 pts (letter) Page_size
- 94 Downloads / 213 Views
subsequently sputtered without breaking the vacuum. The samples were annealed in the vacuum of 10-4 Pa or better at temperatures ranging from 500 to 800 °C for 30 min. The sheet resistance measurements at room temperature using a four-point probe were used to detect interfacial reactions after the each annealing step. The reaction products in the Si|Ta|Cu and Si|TaC|Cu metallization schemes were characterized by x-ray diffraction (XRD), Rutherford backscattering spectroscopy (RBS), and transmission electron microscope (TEM). Surfaces of the samples were also examined with an optical microscope and scanning electron microscope (SEM). The results of the investigations were compared with the assessed phase diagrams and to the results obtained from the literature. Although the complete thermodynamic equilibria are never met in thin film systems because the materials in the contact regions are under continuous microstructural evolution - the local equilibrium is, however, generally attained at interfaces. Therefore, the phase diagrams provide us an efficient method for designing diffusion barrier layer between various metallizations, especially when they can be combined with kinetic information. Hence, ternary Ta-Si-Cu, Si-Ta-C, Ta-C-Cu, Si-Ta-N and Ta-N-Cu phase diagrams were calculated from the assessed binary thermodynamic data by using the CALPHAD method [12] and compared with the experimental results obtained. RESULTS AND DISCUSSION Since the detailed investigation of the Si|Ta|Cu metallization system is described elsewhere[13], only a summary of the results obtained is presented here. Three Ta layer thicknesses of 10, 50 and 100 nm were used to investigate the failure mechanism(s) of the metallization system. The thin (10 and 50 nm) Ta diffusion barriers failed at 550 °C and 600 °C, respectively, due to the diffusion of Cu through the Ta layer and the resulting formation of Cu3Si. This was followed by the formation of TaSi2 in the 50 nm sample, but in the 10 nm sample Cu3Si was the only reaction product. The Cu overlayer enhanced the formation of TaSi2, which was not expected to take place before 650 °C[14]. The thicker (100 nm) samples did not fail completely until at 685 °C when the formation of the large Cu3Si "precipitates" took place (figure1). TaSi2
Amorphous layer (Cu,Si,O)
TaSi2
TaSi2 Cu3Si
TaSi2 Si
Amorphous layer SiO2(a) (Cu,Si,O)
Figure 1. Bright field TEM image from the sample annealed at 685 °C for 30 min. However, the formation of TaSi2 had already started at 650 °C as revealed by the XTEM investigations (figure 2). It was concluded that the failure mechanisms of thin (10-50 nm) and thick (100 nm) Ta layers were different: as long as Ta layer was thick enough to prevent the diffusion of Cu up to the formation temperature of TaSi2, tantalum silicide was the first phase to form. If the Ta layer was so thin that Cu penetrated through it before the formation temperature D7.4.2
of TaSi2 was reached, the Cu3Si was the first phase to form. The resulting reaction structure Si|Cu3Si|TaSi2 was supposed
Data Loading...
