Adsorption and Removal of Corrosion Inhibitors during Co CMP Process

Abstract

In logic devices, the resistivity of W and Cu becomes worse for the smaller dimensions, and conformal deposition in that small volume is becoming impossible due to the physical limit. For advanced technology nodes at 10 nm and below, cobalt (Co) has emerged as a leading candidate to replace tungsten (W) in the middle of line (MOL) for the contact/interconnect formation, and also replace copper (Cu) for the formation of metal lines and vias in the back end of line (BEOL) for the first few metal layers. Co is an attractive alternative with its unique properties, such as better electro-migration performance, shorter electron mean free path, conformal coverage in high-aspect-ratio features, and good adhesion to Cu. In addition, replacing W and Cu with Co in MOL and BEOL processing has demonstrated lower resistivity values at smaller dimensions. For the integration of Co in MOL and BEOL processing, chemical mechanical planarization (CMP) is necessary to remove Co overburden and planarize the surface. Recently, many studies have been carried out to develop Co CMP process and its slurries. Most of the studies were about controlling CMP removal rate, etch rate and corrosion defect.

In this dissertation, the adsorption behavior of corrosion inhibitors on Co surface, which is closely related to corrosion inhibition and organic defectivity, was investigated. First of all, the fundamental corrosion and passivation behavior of Co at different pH conditions was investigated. Co surface showed a severe corrosion at acidic pH, a weak passivation at neutral pH and a strong passivation at alkaline pH. The passivation layers were Co(II) oxide and Co(III) oxide at neutral and alkaline pH, respectively. Benzotriazole (BTA), which is a well-known corrosion inhibitor for Cu CMP process, was not effective to inhibit corrosion at acidic pH and it was effective at neutral and alkaline pH showing reduced etch rate and a high degree of adsorption. The sequential electrochemical impedance spectroscopy (EIS) was developed to characterize the inhibitor adsorption at in-situ and ex-situ conditions. Interestingly, most of adsorbed BTA was removed from Co surface by de-ionized (DI) water for all pH conditions which might indicate lower organic defectivity. To understand the stability of Co-BTA complex on Co surface, the Eh-pH diagram of Co-BTA-water system was constructed through a calculation of stability constants and free energy for BTA related species for the first time. The experimental results were compared with the Eh-pH diagram. The oxidation-reduction potential (ORP) of BTA solution and DIW was measured and used to understand the stability of adsorbed BTA at different environment by plotting the ORP in the Eh-pH diagram.

During Co CMP process, various Co surfaces can be formed by CMP slurry and post-CMP cleaning solution and they are naturally exposed to corrosion inhibitors. Thus, possible Co surfaces, metallic Co and Co oxide, were prepared by chemical treatment. The adsorption behavior of BTA on Co surfaces was characterized. From sequential EIS results, all Co surface showed similar BTA adsorption level at in-situ measurements, but most of adsorbed BTA was removed from Co surfaces at ex-situ measurements in DI water. The stability of Co-BTA complex on Co surface in DI water was further investigated by controlling the dissolved oxygen (DO) concentration which is closely related to ORP value. When DO in DI water was reduced by Ar aeration, ORP also decreased and much less removal of adsorbed BTA on Co surface was observed. It might be also explained by using the Eh-pH diagram of Co-BTA-water system. When the ORP of solution, which is exposed to adsorbed BTA on Co surface, was located in the stability region of Co-BTA complex, adsorbed BTA was maintained. However, ORP of solution is far from the stability region of Co-BTA complex, adsorbed BTA was removed from Co surface.

Alternative corrosion inhibitors, 5-Methylbenzotriazole (MBTA) and 1,2,4-Triazole (TAZ), were compared with BTA. Among three corrosion inhibitors, MBTA showed the highest adsorption degree on Co surface due to the positive inductive effect from the methyl group. It was also found that BTA could form a more passive layer than TAZ by forming a polymeric film with Co ions due to the difference in N positions in heterocyclic ring. The rinse ability of corrosion inhibitors was also evaluated. They all showed a good rinse ability with DI water of high DO. From the sequential EIS measurement, polarization resistance change in sequential EIS (ΔRp) could be suggested as a new quantitative parameter of corrosion inhibitor for CMP application. MBTA showed the highest ΔRp and could be suggested as a good corrosion inhibitor for Co CMP application.

Lastly, corrosion inhibitors were applied to Co CMP process with slurry composition at neutral pH. When slurry contains only colloidal silica abrasives and hydrogen peroxide (H2O2), severe pitting was observed on Co surface with almost zero Co removal rate which is due to the radical generation by catalytic decomposition of H2O2 with free Co ions. Addition of BTA could suppress the pitting corrosion by forming a complex with Co ions. When a citric acid was added into the slurry, Co removal rate increased with its concentration and pitting corrosion was not observed. It was due to the chelating effect of citric acid. The corrosion inhibitors (BTA, MBTA and TAZ) were added in the slurry containing colloidal silica, H2O2 and citric acid, Co removal rate was slightly increased which is an opposite result from Cu CMP. The adsorption of BTA on Co surface at slurry composition was found to be lower than BTA solution (corrosion inhibitor only) using in-situ EIS measurement. In case of galvanic corrosion with Cu, BTA could reduce the difference of corrosion potential between Co and Cu by increasing its concentration and MBTA showed higher efficiency among three corrosion inhibitors. Thus, it was concluded that adsorption of azole type corrosion inhibitor on Co surface during CMP process (dynamic condition) is weak enough to slightly increase Co removal rate, but they could reduce the galvanic corrosion after CMP process (static condition).