Behind the Mask

Integrating a CD SEM into an optical system for photomask metrology operations

Bryan S. Kasprowicz and Benjamin G. Eynon, Photronics

The limitations of optical tools for small-dimension and phase-shift mask measurement dictate that photomask manufacturers consider a dual-tool metrology strategy.

In semiconductor wafer dimensional metrology, the critical dimension scanning electron microscope (CD SEM) has long been the instrument of choice for measuring submicron features because of its nanometer-scale resolution and ever-improving throughput. Because photomask features historically have been at least four or five times larger than wafer features, nanometer-scale resolution has not been necessary and optical CD measurement techniques have sufficed to provide lithography and process engineers with information on the effects of certain modes of operation.

The advantages of using optical tools to accomplish this task have been their high throughput and the fact that they do not damage or alter the photomask in any measurable way. On the other hand, the ultimate resolution of such tools is subject to diffraction limitations, and the linearity between measured and actual values is typically lost as a result of interference, resonance, and shadowing effects well before this limit is reached. Although a typical optical metrology tool has a resolution well into the submicron regime, the lower end of its usable range actually stops at 700 to 800 nm. The current 180-nm technology generation of semiconductor products has features of approximately that size, but as features shrink in succeeding generations, it will become impossible for photomask manufacturers to perform metrology-related product development activities using only optical tools.

The recent implementation of subresolution optical proximity correction (OPC) features and phase-shift mask (PSM) materials and strategies has also contributed to the need for new photomask metrology practices. Used extensively on devices at the 250-nm node and below, OPC features are much smaller than 500 nm and can be isolated or attached to the main features on the photomask. An optical metrology tool is not capable of performing dimensional measurements or collecting information on the integrity of such small features. Thus, without a new metrology strategy, photomask makers would have no way of getting any dimensional feedback for OPC process development.

Phase-shifting masks present different kinds of problems because the act of shifting the phase of stepper illumination is an optical phenomenon. When these masks are measured using optical tools, there is constructive and destructive interference that differs from the type of interference caused by pure diffraction associated with standard, binary chromium photomasks. In addition, it is not yet clear whether PSM materials remain optically or physically stable over time. If their optical properties change appreciably, an optical metrology tool would provide inaccurate measurements. Since the stoichiometry of PSM materials from various photomask blank suppliers differs, multiple standards would have to be fabricated and maintained as well. And if the optical properties of each material type also changed at different rates, the challenges facing metrologists using optical tools in high-volume photomask operations would be daunting.

Another issue is that because no NIST standard exists for chromium photomask dimensions, CD metrology tools must be calibrated to some other trusted standard or measurement method. This situation has led to the use of such physical measurement techniques as atomic force microscopy (AFM) and surface profilometry. While atomic force microscopes provide high resolution, their low throughput prevents them from being viable process tools in manufacturing environments. Now coming to the forefront, surface profilometry techniques may have the potential to solve many material-related issues, but throughput is also not yet at the level required by high-volume manufacturing. However, since a small reference sampling of various features (e.g., isolated and dense lines) on a photomask is all that is necessary at the outset, AFM or profilometry tool measurements could serve as a calibration standard. The logical choice for small-feature measurement in high-volume photomask manufacturing applications would thus seem to be the CD SEM, provided that its throughput and metrology capabilities are robust.

Implementing CD SEM

An SEM operates like an optical microscope, but instead of focusing waves of light emitted from a partially coherent broadband source through fused-silica lenses, an SEM focuses waves of electrons emitted from an electron gun with electromagnetic lenses. In all cases, the shorter the source wavelength is, the greater the resolution will be. Since electron waves can easily be made much shorter than the wavelength of visible light, the resolution limit of a CD SEM can be extended to several hundred times that of an optical system. In addition, electrons do not suffer from diffraction the way optical waves do, which makes a CD SEM much more capable than an optical tool for measuring deep submicron features.

Electron-beam systems have one drawback: the problem of charging. Once an incoming beam impinges on a sample, the sample begins to charge up, causing subsequent imaging electrons to be repelled or deflected. The result is image distortion that translates directly into errors in feature size measurement. While charging can occur even on fairly conductive materials, a photomask is extremely likely to charge up because it is made of isolated conductors (chromium lines, for example) floating on a sea of fused silica, which is an insulator. The imaging electrons impinging on the chromium lines have nowhere to go and continue to accumulate until the built-up charge causes the SEM image to be unrecognizable. Recent improvements to CD SEM designs, however, have the potential to overcome this problem by suppressing the charging associated with imaging on nonconducting substrates such as fused silica.

The issue thus becomes how to incorporate a CD SEM with existing optical metrology tools in photomask production lines. It is expected that a CD SEM will be used to measure the smallest photomask features required by semiconductor customers, but it should be capable of measuring very large features as well. Because errors arise in both CD SEM and optical metrology tools when magnification is changed to accommodate various feature sizes, the dual-tool deployment strategy must also consider how to minimize this error response and maximize precision and accuracy over a wide range of products and requirements.

Decisions on when and when not to use a CD SEM rather than an optical tool also should be based on a cost-of-ownership study. Generally, SEM metrology equipment is appreciably more expensive than optical metrology equipment. Operating and other associated costs are also higher for SEMs because of their larger footprint, greater electrical and mechanical complexity, and higher resource requirements. There is a point at which it stops making financial sense to use a CD SEM on features that can just as easily (and accurately) be measured with a less-costly optical metrology tool.

To realize both technical and business success, users of both types of CD metrology equipment must understand the relationship between the tools' capabilities, throughput, and cost. The research described here compares the capabilities of a CD SEM and an optical metrology tool. The objective was to define the most useful range of the CD SEM--where it should be used because of its unique capability and where it may be used interchangeably with an optical CD tool. The overriding motivation was to determine a strategy for implementing CD SEM metrology in flexible-volume photomask manufacturing environments.

Experimental Methods and Equipment

In this study, an AFM was used to measure a representative sample of lines, ranging in size from 0.4 to 1.2 µm, on a test vehicle. The resulting data set was then used as the standard by which an optical tool and a CD SEM were compared. The comparable data set for the full series of lines run on the optical metrology tool using white-light illumination represented the current metrology practice in high-volume photomask manufacturing. With the mainstreaming of phase-shifting masks and binary intensity OPC features, many absorber materials and feature types must be characterized and calibrated. However, this analysis concentrated on non-OPC patterned binary photomasks on fused silica fabricated using NanoRange IID, a Photronics process that provides extremely tight CD uniformity and pattern placement.

The test vehicle was developed by IMEC, a research consortium in Leuven, Belgium, with which Photronics is affiliated. The specific patterns and feature sizes used in the analysis are similar to those found in semiconductor devices in the 100­250-nm nodes, and feature-proximity conditions included isolated and 1:1 dense lines. Other feature sizes and orientations were also available, but the study focused on x-axis features for simplicity. The patterns were laid out so that electrical CD measurements could also be made later using a probing device. Of particular interest were the differences between isolated and dense feature measurements made by the different metrology tools.

The AFM measurements that served as the reference standard were performed by a Sierra AFM with a 0.18-µm tip from Zygo (Middlefield, CT). Optical CD measurements were made using an LWM-250 from Leica (Deerfield, IL). Although this tool can perform monochromatic illumination (g-line and i-line), only white-light illumination was employed during this study to determine the tool's capabilities under current production conditions. Using monochromatic illumination would result in an approximately 400% reduction in overall throughput. The CD SEM measurements were made with an 8100XP-R CD SEM from KLA-Tencor (San Jose). This tool leverages both back-scattered and secondary electron imaging to provide a high-contrast, hence very precise, image of the target feature. It was run using a measurement macro employing pattern recognition, which makes it possible to run a multihundreds-of-points study unattended.

Before the CD SEM measurements were taken, beam operating conditions were customized for chromium. These parameters, typically found by minimizing sample charging and maximizing image contrast, were fundamental to ensuring precise, repeatable measurements both for this study and for daily manufacturing use in the future. The system magnification was then pitch-calibrated to eliminate errors in width determination. Once the CD SEM was optimized for chromium measurements, the AFM standard was used to further tune the tool. After measurements were taken using several different edge detection algorithms, the algorithm that best reproduced the measurement of the reference AFM tool was chosen by fitting curves over several feature sizes. Finally, after the beam conditions had been set up, the system magnification pitch-calibrated, and the edge detection algorithm chosen, the CD-SEM measurements of the series of dense and isolated lines described above were taken and the results compared with the AFM and optical tool data sets.

Results and Discussion

A major point of interest to both photomask and semiconductor manufacturers is the CD variation caused by proximity effects, which manifest themselves as significant deviations from the AFM standard on dense versus isolated features. As shown in Figure 1, which depicts data for isolated and dense lines plotted against the AFM reference data, the measurement values from the CD SEM and the optical tool agreed fairly well over the CD size range selected for this study--even down to 400 nm. As expected, however, the linearity curves do not directly coincide, indicating that there were some differences among the values. The differences appear minimal in the figure because of the scale used to accommodate the wide range of CD sizes, but every nanometer is important as devices become smaller and smaller. Comparing Figures 1a and 1b, it is clear that the values for the dense lines exhibited a higher degree of variance than those for the isolated lines.

Figure 1: Measurement linearity for chromium features on fused silica: (a) isolated lines and (b) dense lines.

The tool reference offsets, or differences between the tool measurement and the standard at each feature size, are shown for isolated and dense lines in Figure 2. As has been reported elsewhere, AFM performance is limited by the inability to accurately characterize and control tip width and shape.¹ Evidence of this phenomenon is seen in Figure 2b in the extreme valley at the 1000-nm feature size; there was a fairly significant error (about 20 nm) in the AFM tip-width calibration when the dense-line measurements were taken at that site. The curve shapes for features larger than 800 nm are similar for both tools, with differences attributed to the measurement noise or precision capabilities of each system. This similarity indicates that at these larger feature sizes, the CD SEM and the optical metrology tool performed equally well compared to the AFM reference. Thus, it makes sense to use only the less-costly optical tool for such measurements in the photomask environment.

However, the curves for features smaller than 800 nm, particularly dense features, diverge. As Figure 2b indicates, to obtain optical tool­to­SEM system matching for 700-nm dense lines, an offset of roughly ­2 nm would need to be applied to the SEM and an offset of +6 nm would be required for the optical metrology tool. This can be done, and probably will be in the real world of manufacturing, but the preferred tool for features 700 nm is the CD SEM.

Figure 2: Measurement deltas for chromium features on fused silica: (a) isolated lines and (b) dense lines.

In addition to revealing differences in tool capability at various feature sizes, the study showed that there are differences related to the proximity effects of the pattern being measured. This pattern-type dependency results largely from light scattering in the case of optical tools and sample charging in the case of an SEM. The charging phenomenon can be seen in Figure 3, which includes a three-dimensionally rendered AFM dense-line image (3a) and corresponding SEM images of dense (3b) and isolated (3c) lines. The dense lines seem to experience more charging, which becomes visible as a noticeable intensity increase, than the isolated line.

 

Figure 3: Examples of micrographic images: (a) AFM image of a dense line, (b) SEM image of a dense line, and (c) SEM image of an isolated line.

Figure 4 highlights the pattern-dependency issue and illustrates the need for separate offset requirements for dense and isolated features. The problem with charging typical of SEMs is seen in Figure 4a as the difference in curve shapes between the dense and isolated features. Measurements for all sizes of isolated features were consistent with respect to the AFM measurements, but the results for dense features varied substantially. However, the large variance at 1000 nm was at least partially attributable to the AFM tip-calibration error. Whenever the charging effect is consistent across feature size and type, offsets can be used to maintain CD SEM accuracy and precision.

As shown in Figure 4b, the optical tool experienced an interference effect that caused the divergence from the standard at the smaller feature sizes. The same effect would occur when measuring dense periodic features with any optical tool. The closeness of such features to each other determines the angular position of an optical tool's scatter components and manifests itself as an error source in the accuracy of measurements 700 nm. It would appear that the magnitude of the error changes as the pitch or proximity effect changes. The divergence increases as feature size decreases, which makes optical tools unsuitable for 700-nm measurements in volume manufacturing applications.

Figure 4: Measurement deltas for isolated and dense chromium lines on fused silica: (a) SEM-AFM, (b) LWM-AFM, and (c) SEM-LWM.

Figure 4c compares the CD SEM values with those from the optical tool. Results for the two tools tracked fairly well under both dense and isolated conditions for feature sizes >800 nm, indicating that either technique can be used for such large measurements. Divergence occurred at the smaller features, however, reinforcing the need for a dual-tool measurement strategy.

Conclusion

When SEM and optical CD measurements were correlated to an AFM reference standard, it became apparent that while the differences from the standard were relatively constant throughout the linewidths studied for the CD SEM, there was a feature-width dependence in the case of the optical tool. For linewidths 800 nm, results from the CD SEM and the optical tool correlated to within tool-precision error, but for linewidths 700 nm, the optical tool measurements diverged from the standard. The study also showed that there is a pattern dependency that results in either intensity blooming within CD SEM images or interference effects from optical tools when measuring dense features. For feature sizes 700 nm, the optical interference effects become so severe that resolution, precision, and accuracy degrade. Thus, users of both SEM and optical CD tools must establish multiple offsets based on both feature size and proximity.

Based on the study results, it also is recommended that photomask manufacturers implement a dual-tool measurement strategy by using a white-light optical CD metrology tool to measure both isolated and dense features 700 nm and a CD SEM to measure smaller features down to the smallest semiconductor device dimensions, including subresolution OPC features. CD SEM calibrations also could be maintained for larger feature sizes, so that the tool could serve as a backup to an optical tool in hard-down situations. In addition, the CD SEM could be used for PSM material measurements because it is immune to the optical effects such materials produce. The degree of surface charging on various PSM materials will differ from that on chromium, but once these differences are understood, these different materials can be imaged accurately.

Given the growing importance of CD SEM metrology, more work must be done to address the issues raised in this article. The push for consensus approval of more-accurate chromium and phase-shift mask CD standards must continue. Few, if any, PSM standards exist other than those used internally by photomask manufacturers. Some standards development efforts are under way in different parts of the world, but unless all interested parties come together and reach agreement, there will never be true international standards.

Within the bounds of the experiment described here, more analysis can be done by looking at the x versus y effects that may occur with both types of metrology tools. The effects of varying the pitch must be investigated to determine whether calibration offsets should be sensitive to pitch as well as size. Looking farther ahead, comparing photomask CD results for dense and isolated features with final wafer results for the same features can provide mask lithography and process engineers with valuable process development input. Also, correlating mask CDs to device electrical CDs would provide the true feedback needed to modify design data to enhance device performance and bin sort yields.

Acknowledgments

The authors would like to thank the following people for their assistance in the development of this article: Susan Ericshrud of Photronics for her LWM-250 expertise, Mohan Ananth of KLA-Tencor for his CD SEM expertise, Brad Anderson and Michael Eilers of Zygo for providing the AFM measurements, and Mireille Maenhoudt and Kurt Ronse of IMEC for allowing us to use their reticle design.

Reference

1. H Marchman, "Scanning Electron Microscope Matching and Calibration for Critical Dimensional Metrology," Future Fab International 1, no. 3 (1997): 345­354.

Bryan S. Kasprowicz joined the Photronics Technology Group in 1999 and is a senior engineer based at the company's Allen, TX, facility. After receiving his BS in microelectronic engineering from the Rochester Institute of Technology (Rochester, NY) in 1996, he worked at Texas Instruments in Dallas, where his projects included evaluating and characterizing exposure systems and SEMs for both capability and cost of ownership. (Kasprowicz can be reached at 972/889-6351 or bkasprowicz@dallas.photronics.com.)

Benjamin G. Eynon is director of back-end-of-line technology development at the Photronics Technology Group in Austin, TX. Before assuming that position in 1998, he worked as a process engineer, technical marketing engineer, and manufacturing manager for semiconductor and photomask manufacturers. He received his BS in microelectronic engineering from the Rochester Institute of Technology in 1987. (Eynon can be reached at 512/248-6169 or beynon@austin.photronics.com.)