Ultralow k films by using a plasma-enhanced chemical vapor deposition porogen approach: Study of the precursor reaction mechanisms

Abstract

As interconnects are scaled down, much effort is made to achieve ultralow k material with a dielectric constant lower than 2.5. Thus, many new precursors are investigated in plasma-enhanced chemical vapor deposition. This is particularly true with the porogen approach where two molecules are used: an organosilicon to create the silicon matrix and an organic molecule “porogen” that creates material porosity during a post-treatment such as annealing. In this article, the influence of the organosilicon molecular structure is investigated. Two “matrix precursors” with different structures are therefore compared. The first one, referred to as D5, has a ring structure (decamethyl pentacyclosiloxane); the second one, referred to as DEOMS, has a star structure (diethoxymethyl silane). The porogen organic molecule, referred to as CHO, is cyclohexen oxide. The fragmentation paths of the precursor molecules in the plasma are investigated by quadrupole mass spectroscopy and the film structure is studied by Fourier transform infrared spectroscopy. The mass spectroscopy analysis shows that the fragmentation in plasma is highest for DEOMS, intermediate for CHO, and lowest for D5 in comparable process conditions. At the maximum plasma power setting, the loss rate, which yields molecule consumption, is 43%–81% for the D5-CHO mixture, respectively, and 73%–37% for the DEOMS-CHO mixture, respectively. This is related to higher bond-dissociation energy for the siloxane (SiOSi) link in D5 than silane (SiH), silylethoxyde (SiOC2H5) in DEOMS, or CC and epoxy cycle in CHO. Indeed, a higher electron-energy relative threshold for dissociation under electron impact is measured for D5 (around 7eV) than for DEOMS and CHO (around 4eV). Moreover, the fragment structures differ from one precursor to another. Methyl groups are abstracted from D5 and a few polysiloxane chains are produced from pentacycle opening and fragmentation. In the case of DEOMS, many single silicon-atom-bearing species are produced. Consequently, the D5-based films have significant retention of siloxane cycles and a less diverse silicon environment than DEOMS-based films. The porogen incoporation (organic phase) was evidenced through alkyl group absorption and is more important with DEOMS than D5 as a matrix precursor. Moreover, the epoxy moiety of the porogen seems scavenged by the plasma and is not retained in the films. These results confirm other studies that discarded D5-CHO chemistry for porous dielectric achievement in an industrial reactor, whereas DEOMS-CHO leads to porous films with an ultralow dielectric constant. Eventually, this study shows that the usefulness of cyclosiloxane precursors is not straightforward.