Novel XPS Technique for Fluorocarbon Layer Evaluation on Nano-Scale Sicoh Sidewalls

Abstract

Motivation: The Back End of Line (BEoL) in semiconductor industry uses carbon doped silicon oxides (SiCOH), also called ultra-low-k (ULK) or low-k dielectrics as insulating materials. During reactive ion etching (RIE) of those materials, fluorocarbon passivation layer is formed which controls the etch profile and the plasma induced damage of the material. The etch front of the dielectric is dynamically removed during etching. Due to anisotropic plasma the FC layer builds up differently on the trench sidewalls. Therefore the FC layer on trench sidewalls and the trench bottom should be quantified [i]. It is challenging to study the vertical sidewalls in a patterned film. This work aims to implement X-Ray Photoelectron Spectroscopy (XPS) technique to analyze the sidewalls of the etched film and quantify the buildup of passivating FC layer on the sidewalls in different etch recipes. Methodology: Dense SiCOH (k=2.75, open porosity=7%) with initial thickness of 160nm was deposited on 300mm Si wafers. The wafers were patterned using E-beam lithography. The pattern consists of parallel trench arrays of varying trench widths (50nm to 1500nm) and line widths (75nm to 125nm). Wafers were etched with C4F8/CF4/N2/Ar (Recipe_1) and CHF3/CF4 (Recipe_2) plasma in a Dual Frequency Capacitive Coupled Plasma (CCP) etch chamber. Post etch trench depth and critical dimension (CD) bias of the etched structures were measured using spectroscopic ellipsometry and critical dimension scanning electron microscopy (CDSEM) respectively. Each structures were then measured under XPS for the surface chemical composition.The XPS methodology is based on the mathematical separation of the photoelectron signal intensities of the three different surfaces of a trench, namely the Line Top, Trench Sidewalls and Trench Bottom[ii]. In a model-based approach, electron shadowing by the opposite trench sidewall is calculated using the measured geometry from ellipsometry and CDSEM. It assumes that every point surface on the sidewall emits electrons and only the electrons falling in the acceptance cone of XPS electron detector are counted which limits the conical opening to 60º inclined at 50º to sample surface normal (Fig.1a). Analysis depth of the sidewall varies with the trench width and the trench angle. A physical model is developed to calculate the sidewall analysis depth which is not shadowed by the opposite sidewall (Fig.1b).The layout contains various pitch size which constitute different area ratios of Line Top, Trench Sidewalls and Trench Bottom. In order to find the chemical composition of each surfaces independent of their visible areas, the system of linear equations is solved using Multiple Linear Regression method to find out the signal intensity per unit area from all three surfaces, assuming that each surface composition remains constant across every pitch size. Comparison of the signal intensity per unit area indicates amount of sidewall passivation w.r.t trench bottom. As an additional verification to the methodology and electron shadowing, parallel angle resolved XPS (pAR-XPS) have been carried out at the same structure to find out the variation in sidewall contributions with different angles. Observation: Recipe_1 is highly passivating (C4F8 presence) but directional (high Ar content), therefore shows less deposition of FC layer on the sidewalls compared to line top and trench bottom, shown by the least F1s F-C and C1s C-Fx (x=1:3) intensity but high FC layer deposition on flat surfaces with remaining photoresist adding more carbon and nitrogen to line top than trench bottom (Fig.1c).Recipe_2 shows lower passivation (CHF3 presence) and anisotropy (No Ar) than Recipe_1, therefore lower carbon is seen on line top and trench bottom compared to Recipe_1, but a thicker FC layer on sidewalls. In both the recipes sidewalls and trench bottom shows similar level of Si2p Si-O and O1s O-Si signal intensities but only differ in their FC layer buildup due to different plasma chemistry. Line top with higher carbon content has lesser intensity of Si2p Si-O and O1s O-Si than the other with thinner FC layer buildup. Conclusion: The results confirm that the physical model is capable of separating the signal intensities from the three surfaces through which the FC layer on each surfaces have been accurately predicted. Therefore the normal mode XPS technique can be utilized to gain insight into the chemical composition of sidewalls. Acknowledgement: This work was funded via subcontract from Globalfoundries Dresden within the framework Important Project of Common European Interest (IPCEI) by the Federal Ministry for Economics and Energy and by the State of Saxony.