Interconnect Bonding
(1272 daisy chained connections per channel). This channel yield was much lower than the yields achieved with Cu/Sn-Cu bonding primarily due to non planarity of the Cu pillars as a result of dishing during the CMP planarization process. The amount of non planarity of the Cu pillars was measured after CMP. For the corner pillar, the average dishing was 0.35µm below the die center. The minimum value observed was 0.20µm and the maximum was 0.60µm. The corner pillar exhibited a larger amount of dishing because it is the most exposed pillar in the die during CMP. The amount of edge pillar and corner pillar dishing did not depend on wafer location to a significant degree. The non-planarity reduced bonding yield by
preventing physical contact of edge pillars during the thermocompression process. The effect of the nonuniformity can be seen in Fig. 4 which shows CMP Cu to CMP Cu at 10µm pitch at the center of the die (left image) and the edge (right image).
16 At the center the Cu to Cu
thermocompression bond has been made, but at the edge the nonuniformity in the copper pillar created a gap between the Cu surfaces. Prior to bonding, the oxide on the bottom
die was recessed by 3µm to expose the copper pillars, and the oxide on the top die was recessed to expose the copper pillars by 0.2µm. During the bond process at 325°C, copper oxide forms on the exposed surfaces, visible in the image.
The recess (or dishing) occurred during the
Table I. Contact resistance comparison of three metal to metal bonding types
CMP process and may be reduced by adding dummy features or pillars outside the active die area. The lack of dummy features caused the outer rows of Cu pillars to recess more than the inner rows, resulting in a non planar bonding surface. The majority of the bonds formed were good in each channel, as confirmed by cross- sectional SEM, but the links at the die edge
were electrically open due to the edge dishing. This was confirmed by optical inspection after shear testing (see Shear Testing section).
Cu-Cu Bonding (CMP’d Cu to as-plated Cu) In part to try to achieve higher channel yield with Cu-Cu, devices with CMP’d Cu on the top die were bonded to devices with plated Cu and BCB mechanical key on the bottom die. This process is shown schematically and with SEM in Fig. 5.
This approach was taken because by mating the CMP Cu die with the plated Cu die, the non- uniformities in thickness were expected to offset partially. The CMP Cu pillars are lower at the die corners due to CMP dishing, whereas the plated Cu pillars are higher at the corners. Typical bonding pressure was 10.5 kg/m2 (30 kgf) for the 15µm pitch samples. Samples were bonded at 325°C for 900s. Electrical measurements gave 94.8% channel yield with median resistance of 62Ω, for a resistance of 73 mΩ per bond and routing segment.
Underfill
Cu/Sn-Cu parts were underfilled with epoxy to prevent oxidation of exposed copper surfaces and increase bond strength. The epoxy (LORD Exp A) was unfilled, with low viscosity (5600 cPs at 25°C) and low Young’s Modulus (0.2 GPa at 100°C). Seven Cu/Sn-Cu bonded samples were underfilled. The underfill process consisted of a vacuum-assisted application of the epoxy at 60°C to promote capillary flow, followed by oven cure (1hr at 180°C).
Reliability Testing
The Cu/Sn-Cu devices with underfill were subjected to 100 thermal cycles (+125°C to - 40°C) and re-probed, and then 100 hours 85% RH / 85°C stress testing and reprobed. There were no significant changes in the electrical yield or channel. The average yield was 93% before and after stress testing and the average resistance was 156Ω before and after stress testing. After stress testing, samples were cross- sectioned for SEM and EDS analysis. EDS was used to quantify the elements in the IMC phase
at the bonding interface. The IMC is the Cu3Sn phase. There was no Cu6Sn5 or unreacted Sn. There was no visible corrosion on underfilled
www.euroasiasemiconductor.com Issue III 2011
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32