4 Microstructure and Mechanical Reliability Issues of TSV
77
4.2.2.4 Stress Metrology Challenges
Although the synchrotron X-ray micro-diffraction is by far the most powerful technique for stress analysis, it is the most expensive and requires access to a synchrotron
radiation source [28]. None of these aforementioned techniques can directly measure
stress in Si or Cu. In Raman spectroscopy phonon frequency shifts are converted into
stress values [28]; in diffraction-based techniques strain tensors are mathematically
converted to stress tensor using linear elastic properties of bulk single crystalline
samples [28]. Although μRS can provide stress profiles in the depth direction, the
penetration depth is within a submicrometer range. Therefore, the complete picture
of the stress distribution in TSVs is still unavailable. How- ever, this information is
critical to elucidate the root cause for the metal pumping phenomenon as discussed in
Sect. 4.3.1.3. The modeling techniques discussed in Sect. 4.4 may provide a solution
towards this direction.
4.3 Reliability Issues Associated with TSVs
Reliability complications in TSVs typically arise either from stress-related or electrical sources, and are often convoluted with microstructural effects. In the following,
we discuss the cause and effects of each of these reliability issues, along with potential
approaches to circumvent or minimize these.
4.3.1 Stresses in TSVs
4.3.1.1 Origin and Effect of Stresses
Stresses in TSVs, and therefore in the neighboring Si, arise from two sources: (i)
growth stresses, that arise as a result of via-filling by electroplating, and (ii) thermomechanical stresses that arise due to thermal expansion mismatch between Cu in the
via and the surrounding Si.
Growth stresses arise when the electrodeposited Cu grows radially inwards on the
Cu-seed layer from the via side-walls, and abut itself, often resulting in a seam along
the TSV axis. Such a seam may leave a thin void-line along the TSV axis (Fig. 4.5),
which can grow during post-deposition annealing, and result in mechanical and
electrical performance loss. Electroplated copper undergoes significant grain growth
by self-annealing at room temperature, which results in volume shrinkage of the
Cu TSVs by elimination of grain boundaries [32]. This may allow any existing
void to grow during both self-annealing, as well as during the pre-CMP (chemicalmechanical polishing) annealing step. During annealing, hydrostatic stress gradients
generated around pre-existing defects in the Cu may be relieved by diffusion of lattice
vacancies towards the existing void near the axis, causing the void to grow [33].
77
4.2.2.4 Stress Metrology Challenges
Although the synchrotron X-ray micro-diffraction is by far the most powerful technique for stress analysis, it is the most expensive and requires access to a synchrotron
radiation source [28]. None of these aforementioned techniques can directly measure
stress in Si or Cu. In Raman spectroscopy phonon frequency shifts are converted into
stress values [28]; in diffraction-based techniques strain tensors are mathematically
converted to stress tensor using linear elastic properties of bulk single crystalline
samples [28]. Although μRS can provide stress profiles in the depth direction, the
penetration depth is within a submicrometer range. Therefore, the complete picture
of the stress distribution in TSVs is still unavailable. How- ever, this information is
critical to elucidate the root cause for the metal pumping phenomenon as discussed in
Sect. 4.3.1.3. The modeling techniques discussed in Sect. 4.4 may provide a solution
towards this direction.
4.3 Reliability Issues Associated with TSVs
Reliability complications in TSVs typically arise either from stress-related or electrical sources, and are often convoluted with microstructural effects. In the following,
we discuss the cause and effects of each of these reliability issues, along with potential
approaches to circumvent or minimize these.
4.3.1 Stresses in TSVs
4.3.1.1 Origin and Effect of Stresses
Stresses in TSVs, and therefore in the neighboring Si, arise from two sources: (i)
growth stresses, that arise as a result of via-filling by electroplating, and (ii) thermomechanical stresses that arise due to thermal expansion mismatch between Cu in the
via and the surrounding Si.
Growth stresses arise when the electrodeposited Cu grows radially inwards on the
Cu-seed layer from the via side-walls, and abut itself, often resulting in a seam along
the TSV axis. Such a seam may leave a thin void-line along the TSV axis (Fig. 4.5),
which can grow during post-deposition annealing, and result in mechanical and
electrical performance loss. Electroplated copper undergoes significant grain growth
by self-annealing at room temperature, which results in volume shrinkage of the
Cu TSVs by elimination of grain boundaries [32]. This may allow any existing
void to grow during both self-annealing, as well as during the pre-CMP (chemicalmechanical polishing) annealing step. During annealing, hydrostatic stress gradients
generated around pre-existing defects in the Cu may be relieved by diffusion of lattice
vacancies towards the existing void near the axis, causing the void to grow [33].