Copper plating for 3D interconnects A. Radisic * , O. Lühn, H.G.G. Philipsen, Z. El-Mekki, M. Honore, S. Rodet, S. Armini, C. Drijbooms, H. Bender, W. Ruythooren IMEC vzw, Kapeldreef 75, B-3001, Leuven, Belgium article info Article history: Available online 23 June 2010 Keywords: Copper Cu Plating TSV abstract In this paper we report on Cu plating of through-silicon-vias (TSV-s) using in-house made acidic Cu bath with model additives (SPS, PEG, and JGB). Although the model additives might not be as potent as com- mercial additives, they have been studied in detail, and their role in Cu plating has been described exten- sively in scientific literature. This in turn allows deeper insight into how changes in bath composition affect the plating mechanism and Cu via-fill. Ó 2010 Elsevier B.V. All rights reserved. 1. Introduction Establishing a cost-effective technology for the metallization of through-silicon-vias is an important factor in the realization and volume manufacturing of 3D-stacked integrated circuits (3D-SIC). Cu electroplating should provide not only a void-free TSV fill, but also short plating time and small overburden. The filling results de- pend on factors such as via dimensions, current density, current waveform, Cu bath composition, mass transport phenomena, and properties of the seed [1–11]. In the work presented here, we focus on Cu deposition from the bath with model additives and explore the effects of bath compo- sition, current waveforms, die-layout, and feature size on the Cu fill profile, deposition time, and overburden. At the time we started our studies, there was a number of commercial additives available on the market that were more potent than our model additives (for example, they could provide stronger inhibition for Cu deposition on Cu). However, model additives, which are sometimes also referred to as ‘‘public domain” additives, offer several advantages over commercial plating baths. The model/‘‘public domain” addi- tives have been studied for years and there is an abundance of sci- entific literature detailing their properties and role in Cu plating [1–11]. Since there are still questions remaining, they are revisited periodically and explored with new, more sophisticated instru- ments and experimental techniques, and then shared with the sci- entific community. The chemical formulas of model additives are known; the model additives are readily available, and can be pur- chased from chemical suppliers. Thus, one has direct control over the bath composition. On the other hand, commercial additives are typically ready-to-use, pre-mixed solutions, with very little (or no) supplemental information on additive properties. However, the most important reason for choosing the model additives is that they can provide super-conformal fill of TSV-s [1–11], which is cru- cial for fabrication of defect-free structures. Scanning electron microscopy (SEM) and focused ion beam (FIB) analysis are used to characterize Cu ‘‘nails” inside the vias. 2. Experimental The plating experiments were performed using in-house made virgin make-up solution or VMS (aqueous acidic Cu-sulfate solu- tion with Cl  ), and model additives such as poly-ethylene glycol (PEG), bis(3-sulfopropyl) disulfide (SPS), and Janus Green B (JGB). Coupon-level experiments were performed using three-electrode rotating disk electrode (RDE) setup and computer controlled Auto- lab/Metrohm PGSTAT30 potentiostat. Working electrode (WE) was the coupon with TSV-s. The sample area (i.e. the apparent area of the WE) was 1.5 cm 2 . The true area of the WE, including the area of TSV sidewalls, could be up to three times larger than the appar- ent area, depending on the via density. Deposition current density was calculated using the apparent WE area. A Pt mesh was used as a counter electrode (CE), and was separated from the WE by a dia- phragm, which prevents anodic additive breakdown products from affecting the deposition process. Ag/AgCl was used as a reference electrode (RE). Volume of the electrolyte in the plating cell was between 100 and 300 ml. The same bath was used for plating of two coupons, and then replaced by a fresh Cu bath. The additive consumption was not monitored in coupon-level experiments. Three-hundred millimeter wafer-level experiments were per- formed on Stratus 300 (NEXX Systems) and SlimCell (AMAT) highly automated plating tools. In a typical wafer-level experiment, four wafers were plated in a sequence. Cu bath samples were collected before, during, and after plating of each wafer for composition analysis. Accelerator (SPS) and suppressor (PEG) organic additives 0167-9317/$ - see front matter Ó 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.mee.2010.06.030 * Corresponding author. E-mail address: radisic@imec.be (A. Radisic). Microelectronic Engineering 88 (2011) 701–704 Contents lists available at ScienceDirect Microelectronic Engineering journal homepage: www.elsevier.com/locate/mee