PVD Liners Extend to 45 nm Interconnects

Like optical lithography, physical vapor deposition (PVD) has proven time and again that its performance will last yet another technology node. A recent study by IBM (East Fishkill, N.Y.), Infineon (East Fishkill, N.Y.) and AMD (Sunnyvale, Calif.) engineers demonstrated that very thin PVD barriers can exhibit good reliability at the 45 nm node, particularly if the liner process is properly managed. Among other things, this is a function of the PVD resputter process, which has become a common means of distributing the TaN material along the sides of high-aspect-ratio (HAR) vias. Engineers Armin Fischer of Infineon Technologies and colleagues presented their findings at the recent International Reliability Physics Symposium (IRPS) in Phoenix.

To fill smaller vias of increasing HAR, sputtered liners have evolved from the collimated and ionized PVD techniques used at the 180 nm node to the resputtered PVD liners of the 130 and 90 nm generations. From a tooling standpoint, there are at least two options: Perform the argon resputter ex situ — in a separate tool, after TaN deposition, before tantalum deposition; or perform argon resputtering in situ — during TaN deposition. This study examined the electromigration (EM) and stress migration (SM) behavior of test structures with respect to liner thickness, resputter sequence and resputter intensity. The test structures consisted of single vias and narrow lines (1:1 width:diameter), with current flowing from the via to the line. Tests were carried out at 30 mA/μm² and 300°C. The growth of stress-induced voids was monitored by the resistance drift during high-temperature storage (1000 hr at 225°C) using chains of large metal reservoirs with various shapes and several micron-squared sizes, which are linked over single vias by short line segments in reservoir-above and reservoir-below configurations (Fig. 1 ).

1. Stress migration (SM) test structures are configured with reservoir-above (top) and reservoir-below (bottom) configurations. (Source: Infineon/IBM/AMD)

The engineers observed a different failure mode for both early EM failures (upstream stressed) and SM failures on single vias connected to a metal reservoir above for the 65 nm node relative to 130 nm node failures. In both of the EM and SM cases, slit voids formed in the mid-section of the vias. The same failures occurred for both the ex situ and in situ liners. In previous technology nodes, the failures were located at the via bottom. The engineers attributed the change in via void location to better penetration of the via into the underlying metal, transferring the weak spot from the via bottom to the via sidewall.

For single vias stressed in a downstream direction or vias connected to a large reservoir below, the failure mode for 65 nm EM and SM matched that of the previous generation: Voids formed in the metal line directly below the via.

Tests of the ex situ resputter time showed that penetration of the via into the metal level underneath increases with higher resputter intensity. The SM failure rate with vias connected to a reservoir below turned out to be more susceptible to intensity than those with a reservoir above, which has to do with a change in location of the maximum stress points (Fig. 2 ). Therefore, the stress gradient between the low and high stress points becomes reduced, which plays a bigger role when a large reservoir of metal lies below the via. Since the weak point is mid-via when a large reservoir exists above the via, the degree of via penetration has a lesser effect on failure rate.

2. Increasing the distance between the via bottom and the liner cap/copper interface yields lower stress gradients on vias with deeper anchoring. (Source: Infineon/IBM/AMD)

With the in situ sputter processes, the effect of resputter sequence on SM behavior followed the same trend, but the failure rate effects were even more pronounced than in the ex situ cases, giving a smaller process window. When the resputter intensity was doubled, the SM failure rate dropped by a factor of almost eight.

The liner thickness studies showed that SM and EM generally degraded with decreasing liner thickness using the ex situ approach. With these tests, the thickness and resputter rates were equally scaled. However, with the in situ liners, the engineers determined that if they did not scale the resputtered amount as aggressively as the liner thickness, greater reliability would be maintained. The results suggest that, regardless of deposition and resputter sequence, the resputter intensity primarily defines the degree of via anchoring and enhances SM and EM behavior for a given liner thickness. Higher resputter intensity leads to better via penetration into the line below and, therefore, longer lifetimes.

Find more information on yield management.