Advanced Plating Chemistry for 65 nm Copper Interconnects
Mike Rousseau, Shipley Co. LLC, Marlborough, Mass. -- Semiconductor International, 5/1/2003
|
As 65 nm node technology unfolds, myriad process-related challenges associated with electroplated (EP) copper interconnect fabrication arise. Some of these challenges relate to the actual plated film and copper plating bath control. Others relate to the physical limitations of nanostructure seed layer deposition and subsequent copper film plating kinetics, as well as issues presented by the post-processing of the deposited film.
This article explores how advances in copper plating chemistry can reduce or eliminate many of the critical plating issues as device fabricators move from the 90 to 65 nm node. New plating chemistries, using a leveler component in addition to the standard accelerator and suppressor components, allow more precise control over plating results than two-component systems used at the 130 nm node. The leveler addresses a key limitation of a two-component system — a phenomenon called mounding, which refers to copper overplating on top of dense structures.
Two-component plating
Figure 1 shows the plating dynamics of a two-component EP copper process. When a wafer is immersed in a plating bath, the polymer/chloride complex fully suppresses the entire wafer surface. With a bias applied to the immersed wafer, conformal filling of the structures begins. Filling kinetics (Fig. 1, step 2) is controlled by the suppressive characteristics of the polymer/chloride complex.
Next, the transition from conformal fill to bottom-up filling occurs as the autocatalytic species of the accelerator reaches a critical concentration. This catalytic accelerator species in the feature bottom controls the process. As plating proceeds, momentum deposition takes over. Because the catalytic accelerator species is concentrated in the area where the maximum density of small structures is being filled, mounding occurs over dense structures.
Figure 2 shows examples of real-world mounding in FIB micrographs. The subsequent planarization step of the highly non-uniform copper requires differential CMP activity in the areas of the mounded topography. The added complexity of non-uniform copper deposits presents many undesirable effects that are considered unacceptable in most 90 nm and all 65 nm interconnect process flows.
The introduction of a third organic additive component allows for controlled plating of dense structures and the elimination of mounding. We have designed novel organic molecules to impart additional benefits, such as minimization of terminal effects, to greatly reduce feature-fill bias and accomplish smaller feature/high-aspect-ratio (AR) filling capability. The plating characteristics of a three-component organic additive system begin with the same dynamics as the two-component system (Fig. 3). However, the addition of a leveler component affords advantages in the control of copper overburden. The leveler is absorbed at the feature openings, minimizing the copper deposit at those openings.
Once the features are filled, the leveler minimizes the adsorption and catalytic effects of the accelerator additive, thereby controlling deposition (Fig. 4). The additional benefits of a properly designed leveler contribute to improved CMP processing. The as-plated results (Fig. 5) demonstrate the minimized dishing of leveled copper over large trench structures. An increase in copper removal rate of ~4.3:1 (flat-field to trench area) has resulted from the introduction of the leveler. The improvements noted are critical for a high-yielding, 65 nm device production process.
A careful investigation of acid strength vs. tool
performance is necessary for optimized plating performance. Figure 6 shows the
relative gap-fill kinetics for various plating tools from three vendors, with
optimized acid concentrations at two different points in the plating cycle. The
examination of gap-fill performance can be seen at 350 and 550 Å thick copper
films in both low-AR and high-AR design rules.
 |  |
| 4. As-deposited plated copper with a leveler produces smoother surfaces over 0.2 µm (left) and 2 µm (right) trenches. | |
In this study, it was extremely important to modify the acid concentration for these three advanced plating systems to properly match the electrolyte plating performance to the differences in tool design. This optimization resulted in smooth film surfaces over dense and isolated nanostructures, aiding the integration with the subsequent CMP step. AFM analysis of the copper revealed films of average roughness of 4.3, 2.5 and 2.5 nm for the low-, medium- and high-acid toolsets, respectively. The highly reflective copper surfaces (reflectivity at 480 nm of 152, 159 and 160, normalized to silicon) reduce the background noise for inspection tools, allowing better detection of killer defects.
 |
 |
| 5. As-plated wide features (25 µm) show some dishing (left), but CMP removes copper on the flat field 4.3× faster than over the dished area, producing a planar film (right). | |
We further optimized plating performance by adjusting the additive chemistry prescribed for each of the three plating tool manufacturers. These necessary optimizations allow for precise control of the gap-fill quality (void-free), gap-fill rate leveling performance and elimination of feature-fill bias.
As device manufacturers make the transition from 90 to 65 nm, the need for seed layer enhancement (SLE) techniques becomes critical. The physical shadowing effects from the PVD seed deposition tool into the nanostructures required in 65 nm device fabrication cause discontinuity in seed layer thickness and uniformity. In addition, seed layer discontinuity and oxidation of thin copper seed layers contribute to the formation of micro-voids in the resultant plated film (Fig. 7a). Copper interconnects containing micro-voids are prone to short- and long-term electrical failure.
 |
| 7. A via filled without SLE chemistry contains voids (a). The single-step SLE process repairs a discontinuous or damaged seed layer (b) while ensuring continuous seed coverage (c). |
The issues of mounding during overburden deposition, gap-fill kinetics and seed layer discontinuity have been explored and addressed. We have shown how these plating challenges can be minimized or eliminated through optimized chemical approaches. These approaches are being incorporated in the most challenging plating applications. Such changes are requiring a heightened level of sophistication from the materials supplier. Copper plating chemistry is moving from the realm of "legacy" commodity product offering to that of a customized chemical system designed for specific plating tool configurations and structure-filling applications.
Further investigation of chemical and electrochemical interactions is necessary for a complete understanding of plating bath dynamics and control. These include additive-to-additive interactions, additive-to-acid interactions, tool-to-additive interactions, tool-to-additive-to-electrolyte interactions, barrier/seed combinations and thicknesses, and accelerator breakdown byproducts, to name a few. It is important to understand the effects of these interactions on resultant defect reduction, wafer throughput and minimization of terminal effects as customers make the transition from 200 to 300 mm wafers. The scope of these studies will be the subject of future articles.
| Author Information |
| Michael Rousseau has worked at Shipley for the past 17 years. He has held key positions in the electroplated copper program, including eight years in R&D, four years in sales and five years in marketing and business development. He has a B.S. in chemistry from Hartwick College. |
| Acknowledgments | ||
| Several people contributed to this article, including Elie Najjar, Janet Wu, Deyan Wang, Osnat Younes-Metzler, James Rychwalski, David Valeri, Rob Binstead and Bob Mikkola of Shipley. | ||
