45nm node integration of low-k and ULK porous dielectrics
11/01/2005
As the semiconductor industry approaches the 45nm process-technology node, the capacitance of the copper interconnect must be reduced to lower the RC delay and increase chip performance. The International Technology Roadmap for Semiconductors (ITRS) predicts that the industry will require a k value ≤2.2 at 45nm for interlevel dielectric materials. The primary method of lowering the dielectric constant is to make the dielectric film less dense by introducing porosity, which causes the film to become more fragile - leading to reliability and yield issues when these low-k dielectrics are incorporated into the Cu dual-damascene process. These reliability issues also greatly affect the packaging of completed devices.
To date, the challenges of integrating low-k dielectrics have been largely evolutionary, as engineers have developed methods of dealing with mechanically weak films within current manufacturing processes.
With the introduction of porous low-k dielectrics, engineers have been charged with developing additional characterization methods besides modulus, crack propagation velocity, cracking limit, and cohesive strength, because the materials have important new attributes - such as pore size, pore density, and pore interconnectivity - that affect the film’s mechanical, electrical, and thermal properties. As the industry transitions to porous dielectrics, a dramatic degradation in the breakdown strength of the materials may need to be accommodated [1].
Traditionally, dense and porous low-k dielectrics have been characterized by how readily gases and liquids penetrate the films; for example, dense SiOC films allow some gaseous diffusion and porous ultralow-k (ULK) spin-on films allow ready penetration to liquids and gases. To prevent contaminants from degrading the film and increasing the dielectric constant, these porous spin-on films need to have their surfaces sealed. For example, if the film has 2nm pores, a 4nm or 5nm-thick coating is expected to be required to seal the pores [2]. Therefore, at the 45nm technology node with 60nm spacing between metal lines, roughly one-sixth (10nm) of the film will be this “high-k” sealant - offsetting the low-k benefit of the 50nm of bulk dielectric that remains. In fact, a dense dielectric with k = 2.6 provides better electrical performance than a porous dielectric with k = 2.2 that requires a sealant. Thus, a ULK dielectric material is needed that does not require a sealing process so it can maintain the benefits of the ULK dielectric.
Analysis of Novellus’ Coral low-k and ULK plasma-enhanced chemical vapor deposition (PECVD) SiOC films using positron annihilation lifetime spectroscopy (PALS) shows that there is no clear distinction between “dense” (i.e., k>2.6) and “porous” (i.e., k<2.6) low-k PECVD films. Depending on the k value, the pore size increases from 1.2nm for a k = 3.0 film to 1.2-1.5nm for a k = 2.5 film. These porous films are created by growing a dielectric backbone around removable porogens (polymeric pore generators). When the porogens are removed, voids are left that form the pore structure of the dielectric. The closed-cell nature (i.e., zero pore interconnectivity, a parameter measured using PALS) of these films eliminates the need for conventional pore sealing.
There are three primary methods for removing the porogen from these films. All three methods involve cracking the porogen into volatile by-products. Because of the porous films’ closed-cell nature, each method requires an elevated temperature to allow the cracking by-products of the porogen to diffuse out of the film.
One method is to thermally decompose the porogen so that gaseous by-products diffuse out. Because porogens are typically organic molecules, the heating process can leave behind carbon - a conductive material - in the pores. This contamination can dramatically reduce the breakdown voltage of the dielectric.
A second method is to use electron beams to crack the porogen molecules. However, electron beams tend to be indiscriminate when breaking bonds so the technique will create volatile organic side-groups that diffuse out. It is also likely to leave carbon residue, damage the backbone, and leave a fixed charge in the film.
The third method, ultraviolet (UV) curing, uses wavelengths with the correct energy for breaking given bonds. Consequently, UV curing can decompose the porogen into small, specific organic molecules that can diffuse out. By offering the ability to discriminately break bonds, the UV technique has the lowest likelihood of leaving carbon residue without producing damage or fixed charge in the dielectric film or underlying transistors.
Plot of modulus vs. dielectric constant for porous and dense low-k materials. |
Applying the UV-assisted thermal processing (UVTP) step to a ULK SiOC film yielded a film with a k value of ≤2.5, a dielectric breakdown strength of 7-8MV/cm as measured using a single-damascene comb structure with an HCM (hollow cathode magnetron) TaN/Ta Cu barrier film, and a Young’s Modulus of 9GPa. This film is nearly 15% lower in k value, but similar in mechanical and electrical properties to the first-generation Coral film (k = 2.85) introduced five years ago (see figure). As demonstrated, it can be integrated with the existing PVD TaN/Ta bilayer Cu barrier films, or with ion-induced atomic layer deposition (i-ALD) TaN Cu barrier films.
The porogen approach of creating ULK dielectrics introduces new challenges associated with the incomplete removal of porogens. However, these films, coupled with novel annealing techniques, show promise for lowering dielectric constants while providing high mechanical stability, good electrical properties, and low permeability - all crucial requirements for successful Cu/ULK interconnect integration.
Acknowledgment
Coral is a registered trademark of Novellus Systems Inc.
References
- E.T. Ogawa, et al., Proc. Intl. Reliability Physics Symp., pp. 166-172, 2003.
- J.C. Lin, et al., Proc. IITC, pp. 21-23, 2002.
Wilbert G. M. van den Hoek received his Drs. degree from Rijks Universiteit Utrecht, The Netherlands, and is CTO and EVP at Novellus Systems Inc., 4000 N. First St., San Jose, CA 95134; ph 408/943-9700, e-mail [email protected].