No abstract

Degradation and breakdown mechanisms of a SiOCH low-k material with k=2.3 (25% porosity) and thicknesses ranging from 90nm to 20nm were investigated. By combining time dependent dielectric breakdown (TDDB) data at positive/negative bias stress with thickness scaling results, dielectric failure is proven to be intrinsic and not influenced by copper drift or barrier deposition induced dielectric damage, which is further demonstrated by a very low TDDB thermal activation energy. During dielectric degradation, low field leakage current increase is suggested to be caused by donor type trap generation. It is shown that stress induced leakage current (SILC) can be used as a measure of dielectric degradation. Therefore, by monitoring SILC, low field lifetime can be safely estimated using extrapolation. Based on our experimental results, it is suggested that the impact damage model has a better accuracy for low field lifetime prediction.

No abstract

The time-dependent-dielectric-breakdown of eight different low-K films with porosities between 3% and 50% and thicknesses between 15 and 120 nm were investigated using imec planar capacitors. The analogous OSG-films show all power-law field acceleration factors of-21 independently of porosity. Weibull slopes decrease linearly with thickness and porosity. The maximum allowed electrical fields to meet 10-years lifetime are discussed and data show that for 20-nm spacing remedial measures are required for porosities >30%.

We summarize the current understanding of BEOL TDDB lifetimes models. We first review long-term TDDB data studying the intrinsic reliability behavior of low-k materials, where imec's so-called p-cap test vehicle was employed. Then, damascene data, where copper lines are integrated in the low-k materials, are discussed. When simply assuming that the electric field scales inversely proportional with spacing, not taking into account the impact of process variability like line-edge-roughness, line-to-line overlay errors and via-to-line misalignment, the impact damage model and the power law fit the available data in the best way over the wide range of applied fields. Finally, we discuss the eventual impact of this process variability on the assessment of life-time models and make recommendations for future work. Time dependent dielectric breakdown (TDDB) of porous interor intra-level low-k dielectrics used in advanced back-end-of-line (BEOL) interconnects 1-3 is a serious reliability concern where a severe degradation with porosity increase and spacing scaling is reported, 4,5 Current leading edge CMOS technology development focusses on 10 nm and 7 nm nodes, where line-to-line/via spacings below 20 nm and dielectrics with a k-value below 2.4 are being integrated.6 With such aggressive dimensions and porosity, meeting reliability specifications becomes difficult. Recent literature proposes data analysis methods where process variability like line-edge-roughness (LER), line-to-line overlay errors and via-to-line misalignment are taken into account, 7,8 where a proper data analysis leads to more optimistic lifetime predictions, mainly because of correct estimates of the Weibull slope β. Besides process variability, another key-element for such predictions is the assumption of a TDDB life-time model that is used to predict life-time data from high voltage/field conditions to operating conditions. Current reliability test standards 9 make use of both the E-and √ E-model, where the relation between failure time TTF and field E is assumed to be TTF∼exp(-γE) and the TTF∼exp(-α √ E), respectively. γ and α are so-called acceleration factors. The physics behind these models have been widely discussed in the literature. The E-model has been proposed both for gate oxides and BEOL dielectrics.10-13 Theoretical justifications were based on the assumption that weak bonds in the dielectric can be broken by applying an external electric field 13 or that Cu-ion diffusion and drift through the dielectric forms leakage passages and results in dielectric breakdown. 11 The √ E-model was first introduced for Si 3 N 4 -based capacitors in GaAs MMICs 14 and has been proposed more recently for BEOL dielectrics. 15,16 This model was motivated by assuming dielectric degradation due to Cu-ion drift through the dielectric and on the observation that the two main conduction mechanisms (Poole-Frenkel and Schottky Emission) have a √ E-dependency of the leakage current. Note that there is no general agreement on the role of copper during TDDB in inte...