Copper-induced dielectric breakdown in silicon oxide deposited by plasma-enhanced chemical vapor deposition using trimethoxysilane

Abstract

The barrier mechanism against copper-ion diffusion in silicon-oxide films deposited by plasma-enhanced chemical vapor deposition (PECVD) using trimethoxysilane (TMS) and nitrous oxide (N2O) chemistry (PE-TMS oxide) was studied. It was found that the flow ratio of TMS gas to N2O gas during deposition strongly affects a time-dependent dielectric-breakdown lifetime of PE-TMS oxide with a copper electrode as well as other PE-TMS oxide film properties such as electrical properties (leakage current and dielectric constant), a physical property (atomic composition), and chemical properties (chemical bonding states and wet-etching rate). The dielectric-breakdown lifetime of PE-TMS oxide film with a copper anode is a maximum at a source-gas ratio ranging from 1.7% to 3.3%. On the other hand, leakage current density, wet-etch rate, and dielectric-breakdown lifetime of PE-TMS oxide film with an aluminum electrode are degraded by increasing the source-gas flow ratio (0.83% to 12%). These results suggest that two types of degradation mode exist in the dielectric breakdown of PE-TMS oxide with a copper electrode. Namely, at low flow ratio (<1.7%), copper-induced degradation is dominant, but at high flow ratio (>3.3%), the dielectric degradation is probably not caused by copper contamination but by low-quality dielectric material. The dielectric-breakdown lifetimes of a PE-TMS oxide film (flow ratio: 3.3%) with a copper anode show an Arrhenius-type temperature dependence. That is, the activation energy of the dielectric-breakdown lifetime depends on the applied electric field and decreases from 1.8 to 0.55 eV when the applied field is increased from 0 to 5 MV/cm. As a simple kinetic model of the copper injection reaction at the anode surface, a thermally activated reaction process between two energy states—copper atom state on the anode surface and copper ion state in the dielectric material—is proposed.

Bias-temperature stress analysis of Cu∕ultrathin Ta∕SiO2∕Si interconnect structure

“…from 1.0 and 1.6 eV at 3 MV cm −1 , respectively. 22 The reported thermal oxide TDDB values are comparable with our estimated value, while the reported PECVD oxide failure value involving copper contamination is much lower than our estimated value. This is comparable to the normal requirement of 10 years for gate oxide reliability.…”

Electrical Breakdown in Advanced Interconnect Dielectrics

“…Of course, TDDB is a type of CVS test. BTS can be run on either metal-to-metal or MOS structures and can be used as a means to assess dielectric diffusion barrier effectiveness [270]. BTS is actually pretty much a TDDB experiment [26,235,306], except that it is generally run at a much higher temperature (typically ≥ 200 °C) and maybe a moderate electric field strength (∼ 0.5 to 2 MV/cm).…”

“…The barrier thickness is generally thicker at the bottom of the feature than on the sidewalls, but the sidewall may be thickened at the expense of the bottom by a barrier resputtering method. Cu is known to be a ready diffuser into such dielectrics under electrical bias and greatly reduces dielectric reliability [26,27], although injection of Cu into dielectric should be an activated process [28]. The trench/via is then filled completely with Cu metal using electrochemical deposition (ECD) until an overburden layer lies on top.…”

“…Because Cu metallization is involved, any free Cu from either corrosion defects or weak barrier confinement will be activated at higher temperatures. Dielectric breakdown is generally expected to have low activation energy (E a ≤ 0.5 eV) [221,[267][268][269] as opposed to Cu drift-diffusion controlled breakdown (E a ≥ 1.0 eV) [259] so that their individual contributions to dielectric breakdown can be determined (rather painfully) by testing over a wide temperature range [270]. Too high a testing temperature cannot be used [265,266], unless one is testing to directly assess the impact of such free Cu.…”

“…In a follow-up discussion [373], Suzumura et al noted that high-field stress significantly affects the capping SiC x N y reliability (which should not be ignored [374][375][376]), but at lower field, the IMD reliability can be substantially different. It is also clear that Cu ingress into the ULK dielectric is quite complicated and depends upon the surface chemistry of the ULK, including the pore surfaces [28,380]. In their most recent update, Suzumura et al [377] assert that an √E-model fits their data best.…”