Metrology: A case of multiples
03/01/2005
Solid State Technology asked industry experts to comment on the proliferation of metrology tools and what, if anything, can be done to mitigate the drawbacks.
Process complexity drives proliferation of metrology methods
Tomorrow’s ICs will incorporate new, nonclassical CMOS devices and more complex interconnection. In turn, these devices and structures will use new materials and processes and controlling them will require new metrology methods. The properties of today’s transistors are different from the transistors of five years ago, however, so controlling key properties, such as transistor drive current, often requires additional measurements. The same is true for interconnect, and the possibility of 3D interconnect can only add complexity.
The transistor’s evolution will continue as long as possible. Today, IC manufacturers are using a variety of processes to induce stress in the transistor channel and thus increase carrier mobility. The success of this approach can be found in its extendibility from the 90nm to the 65nm nodes. The impact of carrier mobility enhancement on transistor drive current is described in material available on a number of IC manufacturers’ websites (e.g., Intel’s: www.intel.com) or from conference proceedings such as the 2004 International Electron Devices Meeting.
New or available metrology methods that measure stress must infer the stress in the channel because it is buried. In the near term, metal gates and high-k gate dielectric will move into manufacturing. In addition to traditional optical and electric metrology, other measurement methods are now commercially available, such as x-ray reflectivity and (in-line) x-ray photoelectron spectroscopy. These new methods often will be used in addition to the existing metrology; thus, one can see a trend toward the “metrology farm.”
As on-chip interconnect changes the low-k materials and barrier layers, new issues face the viability of the final IC. This will continue as porous low-k materials move from R&D into production. Pore size distribution measurement is used during R&D and may be needed during manufacturing. Therefore, both small-angle x-ray scattering and ellipsometric porosimetry may be required in the future.
The coefficient of thermal expansion (CTE) mismatches between the various metal layers and dielectric stacks is driving CTE measurement capability. Copper void and low-k killer pore detection remain important challenges. Thus, new interconnect materials also are driving the need for additional measurement capability. There are other examples of adding metrology methods to meet critical measurement needs beyond just the addition of new materials.
Over the past several years, the difficulties associated with measuring nanosized features have driven implementation of more than one method, such as using both scatterometry and CD-SEM for critical dimension (CD) control. These additions have resulted in the term “metrology farm,” as noted previously. The issue with the farm is that, while new measurement needs must be met, adding other methods means more cost. The industry is just beginning to address this issue. It may be possible that as new materials and processes become more mature, some of these measurement methods will either move to the off-line laboratory, or remain in the cleanroom. Pore size distribution and CTE are two examples of measurements that may not be needed in-line when a process reaches maturity.
 |
Alain C. Diebold is a Senior Fellow at Sematech, 2706 Montopolis Dr., Austin, TX 78741; e-mail [email protected].
Reducing metrology tool proliferation
To a certain degree, the question of how to reduce the proliferation of metrology tools underestimates the importance of controlling the variety and numbers of metrology and defect-detection tools in advanced wafer fabs.
The spread of niche metrology solutions has occurred as a natural result of ever-smaller transistor structures fabricated to tighter specifications with more complex lithography processes; new materials and stacked thin-film combinations; and a greater need to control defects, particles, and other factors. Today, a typical 130nm wafer fab starting 20,000 wafers/month might contain more than 100 individual tools for defect detection and measurement, with up to 15 different underlying detection technologies from multiple suppliers. While an added metrology technique is valuable and helps achieve greater yield and IC performance, the added cost is the most significant factor. Operations managers tell us they must consolidate metrology to have more efficient and productive advanced IC manufacturing fabs and to combine metrology data for process monitoring and control.
Today, when looking at tools for just metrology in an advanced wafer fab, you might see different tools for thin-film measurement, lithography CDs, pattern overlay, gate-thickness analysis, metal thin films, ion-implant evaluation, and resistivity. An application such as CD measurement might involve different tools, perhaps from various suppliers, in different process bays (i.e., one in lithography, another in etch). The industry needs to focus on metrology solutions that provide multiple functionalities - a suite of complete, complementary solutions integrated onto a common platform. However, metrology for advanced processing is complicated by the need for greater measurement sensitivity and accuracy.
Scatterometry-based optical CD has attributes that will help reverse the niche metrology trend. This technique can deliver multiple measurements from one tool nondestructively, including CDs, feature shapes (i.e., CD profiles), overlay, and film thickness on patterned structures including stacked films. Work is underway to characterize linewidth roughness. Scatterometry involves flood exposure of a field, the measurement of changes in the polarization state of the reflected light, and calculation of IC-structure characteristics via real-time or library-based regression software algorithms, all based on the optical average of hundreds of features.
Optical CD combines the information-rich content of scatterometry with spectroscopic ellipsometry or reflectometry. Spectroscopic ellipsometry enables rooting thin-film measurements with solid n and k indices. This provides means for fundamental CD metrology-tool calibration, which leads to better long-term tool stability and matching. One of the primary attributes of optical CD with scatterometry includes a sampling frequency much higher than anything possible with conventional techniques resulting in real-time data - a true fingerprint of a fab process.
The strength of using optical CD to provide CDs and line profile data is that the lithography bay is the best place to address the high cost of metrology proliferation because CD and thin-film measurement tools are among the largest groups of individual metrology tools. Plus, CD metrology tools have the highest average selling prices (ASP).
Using optical CD tools capable of measuring 30 wafers/hour in a 20,000 wafer-starts/month fab, running a process with five critical CD layers, can reduce metrology investment by 63% and metrology cost/wafer by 68%. Optical CD tools are half the ASP of conventional CD-SEMs and need only one-quarter of the maintenance.
Beyond tool and maintenance costs, in fabs where scatterometry-based metrology is used, it is showing production-worthiness through tool matching, measurement traceability or correlation, accuracy, and the robustness needed for advanced manufacturability. While some of these improvements are difficult to quantify, they all can reduce the cost of manufacturing. It also is significant that the current acceptance of optical CD metrology in wafer fabs is coming from the ability to meet 45nm-node precision requirements (solid data shows extendibility to <20nm features), correlate these tools to legacy systems, calibrate CD measurements to standards, and match stand-alone and integrated metrology tools in all process modules. Finally, scatterometry’s unprecedented measurement speed improves throughput, which always improves manufacturing, and addresses the mandate for increased sampling frequency.
Optical CD and thin-film metrology combined on a single platform certainly will not reverse the entire niche metrology trend. It is, however, an example of the multifunctional path that metrology suppliers need to pursue while incrementally improving measurement precision to stay ahead of fab requirements.
 |
Paul Terbeek is program manager at Therma-Wave Inc., 1250 Reliance Way, Fremont, CA 94539; e-mail [email protected].