Copper diffusion in amorphous thin films of 4% phosphorus-silcate glass and hydrogenated silicon nitride

Abstract

We have carried out direct Cu diffusion measurements in amorphous thin films of: (a) 4% phosphorus silicate glass and (b) hydrogenated silicon nitride. Thin films of 0.5 μm thickness were grown onto oxidized Si wafers by a chemical vapor deposition technique. 67Cu radiotracer diffusion techniques combined with microdepth profiling by sputtering with neutralized Ar atoms of 400–500 eV energy have been used. The 67Cu diffusion coefficients in 4% P-Si glass and in SiN:16 at. % H films in the temperature range of 227–550 °C are described by: (a) 5.3 ×10−11 exp (−0.5 eV/kT) and (b) 4×10−6 exp (−1.1 eV/kT) cm2/s, respectively. To assure reliability, the extent of Cu diffusion should be considered in device design and subsequent processing.

Dielectric Barrier, Etch Stop, and Metal Capping Materials for State of the Art and beyond Metal Interconnects

“…Based on these various methods, PECVD a-SiN:H has been shown to be an excellent Cu diffusion barrier. 36,37,73,75 Similar tests have also shown that a-SiC:H 59,60,64 and a-SiCN:H 44,45 can perform equivalently to a-SiN:H as Cu diffusion barriers with reduced values of dielectric permittivity as low as 4.2-4.9. 276,277 Recent studies have also indicated that dense a-SiCO:H dielectrics can also serve as Cu diffusion barriers with k values as low as 3.5.…”

“…This is a particularly germane question as the 2013 ITRS projects the thickness of the DB in a low-k/Cu interconnect will need to be only 1-2 nm thick in the next decade. 107 Unfortunately, early studies of the Cu diffusion barrier performance of a-SiN:H DB materials utilized relatively thick (>=100 nm) films, 37,73,75 and more recent studies of low-k DB materials have focused on evaluating Cu and H 2 O barrier performance at thicknesses of 30-100 nm. Only a few studies have attempted to investigate the ultimate limits in terms of DB thickness scaling.…”

“…The low-k ILD also provides some mechanical support to metal lines during downstream processing of multiple metal layers and packaging. The low-k DB/CCL/ES, however, serves these and many other additional integrated functionalities such as: preventing Cu out-diffusion into overlying metal layers, 36,37,[72][73][74][75][76] preventing in-diffusion of moisture [77][78][79] and other aqueous cleaning chemicals [80][81][82][83] that can corrode lower metal and ILD layers (see Figure 2), passivating the top surface of the Cu line to reduce/prevent electromigration, [84][85][86][87][88] halting/stopping the trench or via ILD etch to prevent over etching into the underlying metal layer and to facilitate unlanded vias (see Figure 3), 89,90 and prohibiting resist poisoning during next layer metal patterning (see Figure 2). 91,92 As these functionality requirements all relate to the material properties of the DB/CCL/ES, we briefly outline and review the basic electrical, mechanical, thermal and optical material property requirements for these layers in a general nano-electronic metal interconnect.…”

The influence of incorporated nitrogen on the thermal stability of amorphous HfO2 and Hf silicate

“…This is because previous experimental data indicated that the diffusion of metallic atoms in SiO 2 and SiN x follows Fick's law of [17][18][19] The Gaussian distribution of the chemical composition at the interface in our unannealed samples should be attributed to the roughness at the HfO 2 / SiO 2 ͑SiON͒ interfaces and Hf diffusion that occurred during sample fabrication, such as during the Hf deposition. 12 Because the shape of the electron probe also affects the measured chemical distribution, the distribution is a convolution of the electron probe shape and the chemical distribution profile due to the roughness and the Hf diffusion at the interface.…”

Material design of plasma-enhanced chemical vapour deposition SiCH films for low-kcap layers in the further scaling of ultra-large-scale integrated devices-Cu interconnects

“…Neither a Cu atom nor a Cu + ion interacts with lone pairs in nitrogen or oxygen, because all of the 3d-orbitals in Cu atoms and Cu + ions are occupied by electrons. In addition, the calculated activation energies of Cu + ions were much lower than those previously reported (SiO 2 : 13.6-32.6 kcal mol −1 , SiN: 29.9 kcal mol −1 ) [27, 28]. Therefore, the activated barrier of migration, namely the chemical interaction, such as the coordination bond between Cu–N or Cu–O, was not the dominant factor in Cu diffusion.…”