Sematech: Lack of Litho Reference Measurements a Concern

Alexander E. Braun, Senior Editor -- Semiconductor International, 12/14/2007 11:30:00 AM

With the industry inexorably headed toward the 32 nm node, lithography increasingly faces demands on CD and overlay metrology, and in producing the necessary models required to work at this node. Among other industry organizations, Sematech (Austin, Texas) is devoting efforts and resources to meet this important need.

According to John Allgair, Sematech’s metrology program manager, one of the areas that the organization is concerned about and focused on is accuracy as it applies to CD and overlay metrology. “For CD metrology — particularly in the application of CD-SEM and, to some extent, scatterometry — the data from measurement tools is actually being used to build optical proximity correction and reticle enhancement technology models,” he said. “It’s because this is the way in which the data is being used that I see accuracy becoming more of a concern than it has been in the past — it is directly affecting the quality of those models.”

Both for optical proximity correction (OPC) and reticle enhancement technology (RET), data obtained from metrology platforms, such as the CD-SEM, is actually being used along with physical principles to build these necessary models. “These models then have a profound effect on how the reticle itself is generated, both for OPC features and for sizing the features on the reticle,” Allgair said. He then added that if there is any accuracy issue with the data, it then manifests itself in the writing of the reticle. This, in turn, will show up on the features that are being produced on the wafer. The additional accuracy required by increasingly shrinking features exacerbates the situation.

CD-SEM accuracy can be affected by several factors present in the features being measured on the wafer. These can include doping of the material, roughness and granularity, as well as proximity effects arising from features that are close together when measured on the wafer. “These can all cause a bias from what the actual measurement of the feature is,” Allgair said, “and then these biases or inaccuracies can end up becoming part of the OPC model, which, of course, directly affects how the reticle is built.”

According to Allgair, there are several things that need to be done. “One is the development of reference metrology that will allow characterizing the features that are being measured in a way that is traceable, then those reference measurements can be applied to keep everything else in line,” he said. As the features are measured by a CD-SEM, the reference measurements can be used to determine how well the CD-SEM or any other metrology tool is performing.

There are ways in which CD-SEM manufacturers can correct for some of their measurements to make them more accurate. Some of these include the modeling of the electron signal and the development of models based on the materials and proximity of features that are being measured. “All these can be applied to correct the CD-SEM signal and obtain the desired level of accuracy,” he said. “However, in order to do this assessment, one first needs to have some form of reference metrology applicable to the features being measured, something that will provide reliable traceability to a known standard.”

Both OEMs and standards producing organizations should be involved in this; some already have work underway that should eventually provide the requisite traceability back to known standards. Allgair thinks that the need for close cooperation between standards producers and OEMs cannot be exaggerated. “It makes sense to ensure that those originating the reference resources and the OEMs understand the measurements that are being taken, and how samples are characterized.”

Traceability to established standards is necessary more than ever for the advancement of lithography. (Source: ASML)

According to Allgair, another piece of the puzzle that needs addressing, particularly for the 32 nm node, is the development of overlay metrology. “Accuracy is just as critical here,” he said. “Right now, our concern is that most overlay metrology today has been image-based and has scribe line-type features. As we progress into the 32 nm node, there are other technologies available that will assist in getting more sensitive overlay targets, such as scatterometry-based targets or CD-SEM targets. The assessment of those targets has been an important effort for us. Another area that we are also looking into is measurements at the 32 nm node. We think that there will be a need to move to in-die overlay targets.”

At present, it is still unclear what the correct versions of that in-die target should be. There are several optical — as well as some CD-SEM — targets available. “Trying to assess and determine which are the best techniques for doing in-die overlay measurements at the 32 nm node is a serious concern for us,” Allgair said. “And, of course, there’s the matter of accuracy as it applies to those targets.”