Spectral imaging of substrates

Abstract

Spectral imaging systems and methods are provided for monitoring a substrate during a chemical-mechanical planarization process. An example system includes a carrier configured to receive a substrate, and a platen configured to receive a polishing pad. The platen includes an aperture configured to pass light. The system also includes a frame that disposes the platen in any number of positions relative to the carrier. An optoelectronic system is coupled to the aperture, and the aperture passes light of the optoelectronic system to illuminate the substrate and passes reflected light from the substrate to the optoelectronic system. A processing system is coupled to the optoelectronic system and uses the reflected light to image the substrate as the polishing pad is polishing the substrate.

Description

RELATED APPLICATIONS [0001] This application claims the benefit of U.S. Patent Application No. 60 / 638,660, filed Dec. 23, 2004, and is a continuation-in-part application of U.S. patent application Ser. No. 10 / 815,555, filed Apr. 1, 2004.FIELD OF THE INVENTION [0002] The embodiments described herein relate to in situ metrology during chemical-mechanical planarization (CMP) and, more particularly, to monitoring a film on a substrate as it is planarized. BACKGROUND OF THE INVENTION [0003] Chemical-mechanical planarization (CMP) has emerged as a dominant technology for minimizing surface topology during the manufacture of integrated circuits. By minimizing surface topology, the entire surface can be arranged to be within the depth of field of lithography tools, which results in significantly reduced feature dimensions and a dramatic rise in the value of devices made with such features. [0004] The basic concept of CMP is generally to press a substrate against a polish pad in the presence...

AI Technical Summary

Problems solved by technology

However, if there are structures present in the substrate, such as metal lines embedded in a dielectric film (formed using a so-called damascene process), exposure of such metal lines during CMP can lead to significant variation in etch rate depending on the exposed materials.

If polishing a metal film, incomplete polishing results in regions of residual metal, which causes electrical shorts and device failures.

Excessive polishing of metal films causes erosion of underlying dielectric layers, which can dramatically degrade device performance due to increased circuit capacitance.

Likewise, incomplete or excessive dielectric planarization can also cause problems.

Incomplete planarization results in excessive residual topology, which causes poor feature definition during lithographic exposures and therefore results in poor yield.

Excessive planarization causes dishing and increased capacitance, which also degrades circuit performance.

Differential material removal rates across a substrate being polished causes non-uniformity that also contributes to poor performance.

Finally, the inability to compensate for overall substrate polish non-uniformity prevents optimum performance.

However, this technique suffers from being susceptible to variations in the etch rate as the polish pad ages and wears out.

Global endpoint detection also suffers from the lack of information about individual sites on the surface of the substrate during the polish process.

Typically, such measurements involve multiple such measurements and result in some limited information about the condition of the substrate at the time of the measurement.

However, this technique suffers from being unable to accurately infer the condition of the substrate across the entire surface being polished, i.e., where measurements are not being made.

However, typical systems fail to offer the benefits of both types of systems.

This approach suffers from a dependence on polishing materials having very distinct friction against the polish pad being used.

It has also involved a very poor signal to noise ratio, and provides no local polish process information.

These patents generally describe approaches that assess the overall surface of the substrate being polished, but suffer from being unable to provide locations specific process information.

Thus, these approaches generally are not production worthy and require sophisticated expertise to use.

Although this approach can provide a diameter scan, it suffers from providing no information about areas away from the diameter line being measured.

This approach further suffers from an inability to detect whether a substrate has slipped during polish, which can be a serious process issue.

However, all of these techniques provide limited information about the planarization process across the entire substrate.

An additional limitation of these optical techniques is that they are sensitive to signal noise created by light scattering from feature edges.

However, this technique does not seem to address the issue of assessing light scattered and / or diffracted from patterned features.

Sites that do not pass over the fixed-location sensor cannot be measured.

Furthermore, incidental substrate rotation during the polish process would cause measurements to be erroneous.

An additional challenge of optical endpoint detection systems relates to the multiplicity of film stacks likely to be present during CMP.

The motion of the substrate causes light to scan across multiple film stacks, which significantly affects the measured spectral response and complicates signal analysis.

However, slurries suitable for planarizing copper do not work well on the barrier metals, so processes using multiple slurries have been developed.

Images