How Low-k Porosity Affects Interconnect Reliability
Laura Peters, Senior Editor -- Semiconductor International, 6/1/2003
A traditional approach to lowering the dielectric constant (k) of films is to add porosity to an SiO2-containing matrix. Full testing of the reliability performance of interconnects with low-k SiOC and porous films is relatively new territory. In addition to the change in dielectric material, the reliability behavior is impacted further by the shrinking space between metal lines, which approaches 100 nm in advanced devices.
In a recent report titled "Leakage, breakdown and TDDB Characteristics of Porous Low-k Silica-Based Interconnect Dielectrics," Ennis Ogawa and coworkers at Texas Instruments (Dallas), investigated the impact of pores in silica-based dielectrics (Table) on leakage, breakdown strength (Edb) and time-dependent dielectric breakdown (TDDB). Not surprisingly, their tests revealed that the reliability performance of low-k films definitely degrades with increased porosity.
A more interesting finding was the similar behavior in failure kinetics between SiO2, SiOF, SiOC and porous MSQ films. The failure behavior, described in the field acceleration parameter and activation energy, proved to be independent of the level of porosity. The group reported their finding at the 41st Annual International Reliability Physics Symposium in March.
Material |
Deposition method |
keff |
Porosity (%) |
SiO 2 |
Thermal |
3.9 |
0 |
TEOS |
PECVD |
4.2 |
0 |
SiOF |
PECVD |
3.6 |
0 |
SiOC |
PECVD |
2.9 |
20-30 |
Porous MSQ |
Spin-on |
2.3 |
40-50 |
The engineers used an interdigitated comb-serpent test structure formed using standard damascene processing, to measure leakage, Ebd and TDDB. The typical serpent length and width were ~0.32 m and 0.18 µm, respectively, with a metal-to-metal spacing of ~0.17 µm. The comb height was 1665 µm. TDDB measurements were taken at 105°C but activation energy measurements were taken at 105-120°C. Effective capacitor area was 1 × 10-3 cm2. All measurements were performed at the wafer level with hot chuck temperature control.
Leakage and breakdown measurements were conducted at 25-250°C. Anode and cathode currents were separately monitored for at least 1 hour to look for evidence of copper drift. None was observed. During testing, the serpent was typically grounded while the comb-metal was positively biased. Breakdown was reached when leakage current rapidly increased by 2× or more. The electric field was reported as the voltage drop across the dielectric divided by nominal spacing (170 nm). Weibull statistics were used to analyze TDDB distributions.
The study showed that leakage current densities were similar at low fields but the porous MSQ film broke down much earlier than SiOF and SiOC films, even though the current density at breakdown was relatively low (~2 MV/cm). The TDDB results fit reasonably well with Weibull distributions. The data showed that films with a higher degree of porosity tend to fail more quickly. Interestingly, a comparison of time to failure between SiO2, PETEOS, SiOF, SiOC and p-MSQ showed that the field acceleration parameter is quite similar among all the films and approaches that of thick oxide-based films. TDDB activation energies were also similar, indicating common failure kinetics. The TDDB results suggest that porosity in a silicon matrix has little or no impact on failure physics, but the pores have a major impact on the breakdown strength and time to failure.
Finally, the group from Texas Instruments developed a model for dielectric breakdown using percolation theory, which has previously been applied to gate oxide breakdown in MOS capacitors. They modeled the porous dielectric breakdown as an array of three-dimensional normal cells between the copper lines. Under electrical stress, normal cells degrade to defective cells — possibly caused by a bond-breaking mechanism. The electronic wavefunction describing defective sites is assumed to be sufficient to permit wavefunction overlap with an adjacent defective cell and in this way electrical communication is possible between adjacent defective cells. Electrical breakdown occurs when a single column of defective cells connect the metal electrodes of the capacitor.
The model assumed that the addition of porous cells was equivalent to adding defective cells. The relationship between the Weibull slopes showed a linear relationship between breakdown strength and effective dielectric constant. The results were checked against Monte Carlo simulations to test that key assumption. The models agree well with the observed data, showing that Ebd and TDDB degradation directly relate to the addition of porosity to low-k films.
For additional information on yield management, go to www.semiconductor.net/yield
