by Michael A. Fury, Techcet Group
August 10, 2010 – For the 15th consecutive year, the Clarkson University Center for Advanced Materials Processing (CAMP) opened its International Symposium on CMP in Lake Placid, NY. The number of papers submitted (82) was high enough that several Clarkson student papers were relegated to a poster session to make more room for the outside attendees. A total of 93 attendees will review 33 presentations and 15 posters.
Chris Hobbs, manager of nonplanar integration at Albany SEMATECH, brought us up to date on the latest uses of chemical mechanical planarization (CMP) in device integration. FinFET, graphene, and III-V hetero-integration are among the topics that will figure prominently in the future of FEOL CMP.
Tricia Burroughs, CMP process owner at Albany CNSE, showed a new serpentine comb test structure for electrically measuring post-CMP poly gate height and detecting underpolish. This provides a quick turn e-test method for optimizing their high-k metal gate last process.
Rob Rhoades, CTO at Entrepix, continued his series of presentations on applications of CMP in non-CMOS domains. This time, he discussed a direct wafer bonding process for SOI. Flatness specs of Ra ~0.12nm, comparable to prime Si wafer polishing, were achieved while removing only 70nm of material from the transfer layer.
Lukasz Hupka from IBM Research showed some joint work with JSR on controlling the selectivity between SiO2 and SiNx through a better understanding of the steric hindrance behavior of slurry surfactants on both the abrasive particles and the wafer surface. Cationic surfactants tend to collect on both surfaces, while anionic surfactants only act on the wafer surface.
Jin Goo Park of Hanyang University practiced the dark art of intentionally adding particles to the polishing pad surface to characterize the types of scratch defects caused by each. Defect-causing categories included diamond grit, polishing pad debris, and slurry agglomerates.
Taesung Kim from Sungkyunkwan University took a close look at the effect of pad surface texture on the flow of slurry in the gap between wafer and pad, in a joint project with Samsung. The study spanned three scale domains: macroscopic (pad grooves), intermediate (pad pores), and microscopic (surface texture). While grooves are important for slurry mixing, the other two domains are critical to defect levels. In another Samsung-Sungkyunkwan collaboration, Ji Chul Yang used ceria polishing as a vehicle for another study of the correlation between polishing debris and wafer defects.
Some unresolved detective work was presented by Anita Natarajan of IBM on the nature of Cu dendrites in 45nm interconnect patterns, which appear immediately after polishing and exhibit specific geometric patterns as a function of position on the wafer. This orientation phenomenon is not yet understood. The dendrites contain Ti & O, but no Cu. The Q&A proved to be a clever technique to generate some group-think on potential causes and solutions.
Duane Boning of MIT showed his latest work with the U. of Arizona and Intel on the role of pad asperities, and thus the importance of pad conditioning, in the observed polishing results. Pad modulus measured at very shallow depths is significantly different from the modulus measured more deeply; the deep measurements approach the expected polymer bulk modulus. Pad wear uniformity is more highly variable than the local modulus measurements would predict, suggesting other contributors to pad wear.
Yun Zhang from the U. of Arizona showed a tribological and kinetic study of pads and conditioning for Cu processing, including a direct comparison between the IC1000 and D100 pads, and two different diamond disk designs. A five-parameter model was developed to better understand the observed behavior. The IC1000 pad appears to suffer from rapid local transitions between boundary lubrication and partial lubrication regimes.
Masahiro Takei of Nihon U. demonstrated an electrical capacitance tomography microchannel fluidic device used to measure particle concentration in CMP slurry. The technique is adapted from circulating fluidized bed and fluid catalytic cracking concepts used in large-scale plants. The active measurement area comprises 60 layer electrodes in an 800μm channel to provide the requisite sensitivity.
Eiichi Kondoh, U. Yamanishi, showed the efficacy of scCO2 for pore cleaning in porous low-k dielectrics, with implications for PCMP cleaning. The parameter that makes scCO2 especially interesting for such applications is its high mass diffusion flux, which is the product of its density and diffusivity.
Igor Sokolov, Clarkson CAMP, made an unscheduled presentation on a comparison of AFM and nanoindentation methods for characterizing pad polymer properties. He purports the superiority of AFM for soft polymers, because nanoindenter tips have a contact area of ~45μm2 with the sample, while AFM tips are ~150nm2. Also, vertical deformation of the sample is 0.2-2μm for nanoindentation compared to <10nm for AFM.
Michael A. Fury, Ph.D, is senior technology analyst at Techcet Group, LLC, P.O. Box 29, Del Mar, CA 92014; e-mail [email protected]