Reference metrology and standards: staying ahead of the ITRS
10/01/2008
EXECUTIVE OVERVIEW
The International SEMATECH Manufacturing Initiative’s (ISMI’s) Metrology Program, as a result of its charter to evaluate leading-edge metrology, spends a great deal of time building and evaluating testing structures and techniques. ISMI’s goal is to evaluate new metrology equipment on structures that are relevant for current and future ITRS nodes. Tests must also conform to latest definitions outlined in the ITRS text. Some of the materials and methods used to stay ahead of the ITRS are described.
The International SEMATECH Manufacturing Initiative’s (ISMI’s) Metrology Program is chartered to evaluate the capabilities of leading-edge metrology equipment and report findings to its members. To accomplish this, ISMI is tasked with generating relevant metrology test structures and evaluation techniques that provide value to its consortium member companies. Many member companies have processes either on par with, or ahead of, the ITRS for parameters such as critical dimensions (CDs), use of advanced gate stack materials, etc. Therefore, the ISMI metrology group is always looking for ways to build these advanced structures and material types to ensure these evaluations are relevant for the group’s members.
Defect metrology standards
ISMI’s defect metrology group has three focus areas for equipment evaluations and standards generation: patterned wafer defect detection, unpatterned defect detection, and a new effort in bevel-edge defect detection. Each of these areas has different needs for evaluation samples.
Intentional defect arrays. Patterned wafer defect detection systems typically compare adjacent die on a production wafer and highlight the pattern differences from die to die as defects. Over the years, ISMI has evolved intentional defect arrays (IDAs) that metrology suppliers find useful in assessing the capability of their detection systems.
Requirements for defect detection parallel the ITRS critical dimension (CD) half-pitches. Defects of interest are generally considered to be some fraction of the half-pitch, depending upon the film and location on the wafer. For the front-side patterned area, ISMI develops defect standards for IDAs. These arrays contain SRAM and logic layouts typical of the particular half-pitch. ISMI attempts to create IDA patterns with half-pitches that are two or three nodes ahead of current production. Embedded in these structures are patterned co-planar defects varying in size from 25 to 400% of the half-pitch. These are various types of defects including line-extensions, corner extensions, shorts, and opens. The IDA wafers are used to evaluate current defect-detection platforms to establish their relative sensitivity and identify any shortcomings for the coming nodes. Figure 1 shows some examples of programmed defect types that are written into the IDA wafers.
 Figure 1. Examples of programmed defects in a metal 1 trench logic array. |
To create IDA wafers relevant to the 32nm and 22nm ITRS nodes, ISMI has been forced to use direct-write e-beam patterning. The e-beam generated pattern sizing is limited only by the shape of the beam itself. However, creating an intentional defect that is 25% of the 22nm node or 5.5nm is not without difficulty. Additionally, because the cost of e-beam write time is high, a limited number of IDA wafers can be created with current funding.
The defect metrology group continually investigates methods that would allow for the creation of a larger number of these IDA wafers. Recently, chromeless phaseshift lithography (CPL) reticle technology and imprint lithography have been researched, but show little promise in creating the smallest IDA defects.
Unpatterned wafer defect standards. Unpatterned wafer defect detection systems have historically been evaluated and calibrated using wafers that have had polystyrene latex (PSL) spheres deposited on them with a deposition system built for this purpose. Several of ISMI’s member companies support ongoing work to expand the material types deposited on wafers for evaluating unpatterned wafer defect detection systems. This year, a proposal was submitted to ISMI to generate some bare, and several film-deposited, standard wafers that have a variety of materials, such as silicon and copper, with some known dimensions on and under the films. These efforts are for the study of detection sensitivity and sizing effects from real particles and formation of particles in the process.
Bevel-edge defect detection standards. Until recently, no standard test wafers have been available for testing the performance of bevel-edge inspection and review tools. Creating programmed defects similar to those used for testing frontside inspection and review tools is more challenging, primarily because of the complexity of using lithography to pattern structures in all five zones of the beveled area (Fig. 2). A lithography approach suffers from issues such as a stepper’s limited DOF, elimination of edge bead removal (EBR) that may cause cross-contamination to other process tools, test structures cannot be created at the wafer’s apex, etc. To overcome these difficulties, ISMI designed a new bevel-edge test standard by using scanning electron microscopy/focused ion beam (SEM/FIB) for milling the test structures in all five bevel-zones. By creating different shapes and sizes, it is now possible to test the performance of the inspection and review tools for repeatability, detection capture-rate, false alarm rate, and targeting accuracy. These wafers should be available later this year.
 Figure 2. Graphic of cross-section of wafer depicting the 5 zones of interest for bevel-edge defect detection systems. |
null
Litho metrology evaluations and standards
ISMI’s lithography metrology group is focused on evaluation techniques and standards for controlling CD and overlay metrology to correspond with the ITRS lithography requirements. As such, this group focuses on generating reference standards and techniques for evaluating CD, overlay, and the newer design-based metrology (DBM) systems that are commercially available. There is also a large effort in reticle design to create artifacts for testing. ISMI is now completing a test reticle design that will provide test structures relevant to the 32nm and 22nm ITRS nodes. The reticle design itself will be used to benchmark and develop CD-SEM, scatterometry, X-ray diffraction, optical overlay, and diffraction-based overlay. The lithographic technologies to be applied to the design will include optical proximity correction, image modeling, phase-shift reticle enhancement techniques (RETs), 193nm immersion exposure, double patterning, and double exposure.
NIST also assists with the design of testing structures that can be calibrated at NIST, and have been incorporated in this latest reticle revision. One example is a newer single crystal CD reference material (SCCDRM) pattern that incorporates learning from NIST’s last revision of the SCCDRM. Once the reticle is fabricated, wafers will be processed for NIST in their continued reference material development efforts.
The ISMI lithography metrology group continuously refines its testing methods based upon input from its member company representatives. More recently, accuracy studies have been added to metrology equipment benchmarking activities. There is now a widely accepted use of Mandel correlation statistics when assessing the accuracy of metrology equipment [2]. In this year’s ITRS update, a method for defining measurement uncertainty has been added to the text. ISMI and its members determined that the classical approach to defining metrology measurement error, primarily precision to tolerance (P/T) ratios, may not have given a clear picture of metrology equipment’s capabilities to meet the ITRS allowed error requirements. This uncertainty definition includes precision, matching, accuracy, and sampling components.
New definition of metrology uncertainty
The 2007 ITRS Metrology chapter [3] includes a new framework for uncertainty, where uncertainty replaces the precision metric used in pre-2007 versions. This framework eliminates past confusion about matching and accuracy as they pertained, through vague footnotes in the text, to the old precision definition, and it makes key metrics compatible with SEMI standard definitions. Likewise, it shows that accuracy, matching, and sampling are major error components in many measurements. The ITRS should be applicable to any research, development or manufacturing measurement, and thus must include flexible metrics that can be applied to any case. The uncertainty definition includes precision, matching, accuracy, and sampling components; it is left to the user to determine which components are important to a given case for defining uncertainty. Which terms to use are extremely application-specific. The uncertainty definition is as follows:
 |
null
Where:
- Ucombined is the combined uncertainty
- σP is Precision, which is single tool random variation (PR&R)
- variation caused by internal and external noise sources, drifts, mechanical motions, thermal effects, etc. - σM is Matching, which is multiple tool variation
- Same random sources as with sP but differing among tools, plus systematic differences in probe angle, probe size, navigation, etc. - σS is Sample variation, which is uncertainty due to product variabily
- Estimated uncertainty due to sample variation
- Observed by different measurement instances of different locations - σother is single tool systematic variation or inaccuracy
- sample-to-sample bias variationfrom secondary, often uncontrolled process variations- profile shape
- measurement changes from tool/sample interaction, such as resist shrinkage, charging, contamination, bleaching, etc.
- cross correlations among the sources of variation above
In this new ITRS scheme, Ucombined acts exactly as precision did in the pre-2007 versions; all specified metrology tool performances are in terms of the same 20% P/T ratio except that Ucombined is used for the P.
Accuracy determination for litho metrology
To drive the supplier community to comply with ITRS CD and overlay tool requirements, the ISMI lithography metrology project includes a continual regimen of rigorous and leading-edge tool evaluations. For years, the well-known Advanced Metrology Advisory Group (AMAG), a group of leading metrologists from member companies, NIST and tool suppliers, has formulated, written, and regularly updated unified specifications for CD-SEM, optical CD (OCD), and overlay metrologies. These are consensus specifications of best-known tool evaluation practices, metrics, and methodologies. The specifications cover many different aspects of equipment performance, including precision, accuracy, probe properties, interaction with the sample, navigation, throughput, etc. They are a successful long-term undertaking, resulting in much strategic improvement and applications learning for metrology tools over the years.
For many of these tool properties, artifacts are sufficient, and standards are not necessary. However, accuracy does indeed require standards, and the accuracy of CD measurement has become extremely important when characterizing and validating optical proximity correction (OPC) and in validating model-based scatterometry measurements; consequently, this tool property is a high priority in an evaluation. ISMI and NIST jointly operate a CD-AFM reference measurement system (RMS), which is carefully calibrated to NIST scale and tip characterization standards, allowing artifacts to be non-destructively measured and validated to become traceable standards (to a 1nm error bar for CDs). To rate accuracy, data from a tool under test (TuT) is correlated to the RMS values, and the widely-known total measurement uncertainty (TMU) metric is calculated, which is related to the net residual error of the data points around the line of best fit, based upon Mandel statistics.
A well-executed example of such a calibration exercise is already published [4,5] describing ISMI evaluation of OCD tools. In these evaluations, different models for linewidth and profile are tried, and the model that exhibits an optimal TMU as compared to the reference metrologies is the one that is ultimately used for rating tool precision and accuracy. This iterative technique is powerful, as it helps decide which model is correct by including accuracy in the decision process. Figure 3 shows the results of such an activity. Note that statistical approximations of the RMS and the TuT are broken out separately.
 Figure 3. Mandel correlations for the case of optimum TMU???trapezoidal model using the feed-forward approach, optimized optical constants, with a thin oxide spacer layer. |
When implementing a quality metrology regimen, this exercise should be performed to select the best model for scatterometry implementation in a production environment for a given production process.
For the past few years, NIST and ISMI have collaborated on methods to determine the accuracy of overlay measurements. ISMI has investigated the possibility of using CD-SEM and atomic force microscopy (AFM) for overlay reference measurements. Although agreement between the standards and the TuT has been shown, the primary focus has been to reduce the uncertainty of the calibrated standard or reduce the uncertainty of the RMS. Significantly reducing the uncertainty of reference measurements and standards is a large part of the research within the ISMI lithography metrology group. This is especially true in light of the new ITRS definition of uncertainty.
Reducing uncertainty
Several techniques are being researched within ISMI to provide better reference data in both defect and lithography applications. As discussed previously, the defect group is continually in the process of shrinking the IDA pattern to reflect the current ITRS process nodes. New methods for bevel-edge and unpatterned wafer standards are under development.
The ISMI lithography group is also investigating several promising reference improvements. Next-generation AFM development addresses key error components of current-generation systems. The NIST reference SEM that has been under development for a few years is almost ready for external samples. This SEM, along with NIST-provided model-based linewidth algorithms, promises to provide accurate SEM-based reference measurements. NIST is also developing a He-ion based reference system that has little of the edge bloom that plagues traditional secondary electron systems.
In overlay metrology, ISMI and NIST have developed sensitive structures that could be used as reference structures for calibration activities. Because NIST and ISMI share a common goal in standards and testing development, there are many opportunities for collaboration. ISMI will aslo investigate disruptive technologies such as the previously discussed He-ion SEM and current development of X-ray based CD and overlay measurements. Error propagation models of X-ray based CD and overlay measurements show that the technique may be suitable for providing reference measurements.
Conclusion
All of the work in reference material development, RMS development, and testing methodologies is designed to keep ISMI’s metrology testing relevant to the current ITRS technology node. ISMI and its members help guide the metrology community towards meeting future challenges. This promotes quality metrology tooling and methodologies for successful manufacturing.
Acknowledgment
John Allgair, AMD assignee to ISMI, is a co-author.
References
- The International Roadmap for Semiconductors, 2007, www.itrs.net.
- W. Banke, C. Archie, “Characteristics of Accuracy for CD Metrology,”Proc. of SPIE, v3677, pp 291???308, 1999.
- B. Bunday, B. Rijpers, W. Banke, C. Archie, I. Peterson, V. Ukraintsev, et al., “Impact of Sampling on Uncertainty: Semiconductor Dimensional Metrology Applications,”Proc. SPIE 6922, 6922-0X, pp 0X-1 to 0X-22, (2008).
- B. Bunday, O. Sorkhabi, Y. Wen, A. Paranjpe, P. Terbeek, J. Allgair, et al., “Improvement in Total Measurement Uncertainty for Gate CD Control,”Proc. SPIE 5878, 5878-0M, pp 0M-1 to 0M-12, (2005).
- G. Orji, R. Dixson, A. Martinez, B. Bunday, J. Allgair, T. Vorburger, “Progress on Implementation of a Reference Measurement System Based on a Critical-Dimension Atomic Force Microscope,”Jour. Micro/Nanolith. MEMS MOEMS, 6(2), 023002, Apr???Jun 2007.
Pete Lipscomb received his BS in mathematics from Texas State U. and is a litho-metrology project manager at the International SEMATECH Manufacturing Initiative, 2706 Montopolis, Austin, TX 78741 USA, (512)356-3500; email [email protected].
Mike Bishop is a metrology project engineer at the International SEMATECH Manufacturing Initiative.
Ben Bunday received his BS in engineering physics from the U. of Tulsa, and his MS/ABD in materials science and engineering from Cornell U. and is a litho-metrology project manager and senior member of the technical staff at the International SEMATECH Manufacturing Initiative.
Milt Godwin received his BSEE and BS in physics from Southern Methodist U. and is a defect metrology project manager at the International SEMATECH Manufacturing Initiative.
John Allgair received his PhD in electrical engineering with an emphasis in semiconductor physics and processing from Arizona State U. and is an AMD assignee and the metrology program manager at the International SEMATECH Manufacturing Initiative.
Doron Arazi is a Spansion assignee and a defect metrology project manager at the International SEMATECH Manufacturing Initiative.
Kye-weon Kim is a Samsung assignee and a defect metrology project manager at the International SEMATECH Manufacturing Initiative.