rwigham - Apparently they've come a long way since June last year:
reed-electronics.com 
Why Low-k Dielectrics Fail the Way They Do
Laura Peters, Senior Editor -- Semiconductor International, 6/1/2005
To date, the mechanism responsible for the time-dependent-dielectric-breakdown (TDDB) deterioration of low-k dielectrics in fully integrated interconnects has not been fully understood. This is partly because of the complexity of back-end-of-line integration schemes. However, an extensive study using over 60 process split lots and 100 wafers revealed interesting findings:
* SiCOH TDDB is sensitive to all aspects of integration.
* SiCOH breakdown appears to follow a three-step, electrochemically based model.
* Field, temperature, moisture and oxidized copper drive copper ionization.
* Copper likely migrates along the SiCOH/cap interface.
Researchers from IBM Microelectronics, Sony Electronics, Chartered Semiconductor Manufacturing, AMD, Infineon Technologies and Toshiba America Electronic Components studied the effects of process variations on CVD carbon-doped oxides (SiCOH) TDDB based on extensive experimental data. They reported results at the 43rd Annual IEEE International Reliability Physics Symposium in San Jose in April.
Low-k dielectrics are more conductive and have much worse TDDB performance than silicon dioxide for a given electric field. The SiCOH films in this work were fabricated at the 65 nm CMOS process technology node on 300 mm wafers. Damascene copper features were plated on thin Ta/TaN barriers and a thin copper seed. CMP was used; then a capping layer was deposited.
Wafer-level and module-level constant voltage stresses were conducted using test structures of interdigitated comb-serpentine configurations with 1 and 10 m serpentine lengths. To ensure a uniform field at high stress bias, the test structures were equipped with multiple taps on the serpentines and combs to minimize series resistance effects. During testing, the serpentines were grounded while the combs were positively biased. Stress-induced leakage current (SILC) was monitored using multiple time steps of low sensing voltage. Hard breakdown occurred at 100× SILC. TDDB was conducted at 100-175°C.
Both leakage current and dielectric breakdown strength were found to depend on the tantalum liner process, particularly the post-CMP liner profile at the trench top. Liner thickness and quality affected SiCOH TDDB.
The engineers found queue time between copper CMP and capping to be a problem for copper regardless of interlevel dielectric (ILD), liner and line space, as similar trends were observed with SiCOH films on FSG and with different split lots on 65 and 90 nm features. TEM photos before and after bias temperature stress (BTS) testing indicated particle formation at the low-k/capping interface, confirmed by electron energy loss spectroscopy as copper. Unstressed samples showed no particle formation. Copper exposed to ambient forms CuO and Cu2 O, which could lead to copper ionization under BTS. Ionized copper could migrate into SiCOH along the SiCOH CMP surface and recombine with electrons to form particles at the surface.
The post-TDDB fast diffusion path along the SiCOH/cap interface allowed copper ion diffusion. The metal bridge resulted in a resistive short. (Source: IBM)
"In general, two competing low-k insulation failure mechanisms could coexist during BTS: intrinsic low-k dielectric molecular bond breakage due to a thermochemical reaction process, which has a small temperature dependence; and percolated copper metal bridge formation due to copper out-diffusion, which has a relatively large temperature dependence," explained Fen Chen, technology reliability engineer for IBM Microelectronics (Essex Junction, Vt.). He said that, based on SiCOH's breakdown field (8 MV/cm) and extensive TDDB results, at the 65 nm node, copper out-diffusion (Figure ) dominates the observed SiCOH breakdown, and it always occurs earlier than thermochemical breakdown.
Monitoring of SILC changes indicated three distinct steps in degradation: initial current decay after the start of stress, a gradual current increase, and an abrupt current jump. The first stage depends on the liner process; in the second, line-to-line insulation begins to degrade and leakage increases; finally, the hard breakdown corresponds with insulation failure.
Copper can oxidize in contact with interfacial oxygen and moisture. Trapped residual moisture, plus an applied electric field and temperature, creates the driving force for copper ionization and migration through the dielectric:
CunO + H20 ? Cu(OH)n Cun+ + OH- (1)
Cu ? Cun+ + ne - (2)
where n=1 or 2. Copper ions move through or over the liner into the SiCOH in a diffusion-limited process. Leakage increases, and more particles form at the dielectric interface to deform the electric field. When the copper concentration in SiCOH reaches a critical level, catastrophic failure occurs because of the potential joule energy created by an electrical resistive short under the electric field.
Since most of the TDDB failures analyzed were observed at the top interface, the SiCOH/cap interface is critical. SiCOH could be damaged during cap deposition, cap plasma preclean and CMP. Control of copper corrosion and surface moisture levels can prevent the electrochemical reaction.
Finally, lifetime prediction using a conservative model, aggressive test structures, and a field acceleration factor based on the E-field model exceeded reliability targets.
Additional article re low-k, copper and 45,32nm:
reed-electronics.com |
