Behind the mask

Investigating defect inspection and sensitivity analysis for MoSi-based PSMs

Jerry X. Chen, Robert K. Henderson, and Franklin Kalk, DuPont Photomasks Reticle Technology Center

Because MoSi's optical properties are significantly different from conventional chromium (Cr)-based materials, special MoSi defect inspection and characterization are necessary.

Attenuated embedded phase shift masks (PSMs) based on molybdenum silicon (MoSi) are being increasingly used in semiconductor lithography to create high-density layers in memory chips. However, because MoSi's optical properties are significantly different from conventional chromium (Cr)-based materials, special MoSi defect inspection and characterization are necessary. Since the partial transmittance of PSMs may also affect inspection tool sensitivity, it is important to characterize the sensitivity of the tool as well.¹

Inspection tool sensitivity is characterized and verified by measuring the capture rates of intentionally placed (i.e., "programmed") defects in representative mask patterns. Since feature size dictates the inspection tool setup (and, therefore, sensitivity), it is also important to characterize sensitivity as a function of feature size. The accurate measurement of programmed defect size is critical to proper sensitivity characterization. This article discusses research on defect inspection and sensitivity analysis for MoSi-based phase shift masks.

Defect Size Calibration

Inspection tool sensitivity characterization requires accurate defect size measurement (defect sizing). The most commonly used programmed defect mask to achieve this characterization is the Verimask (DuPont Photomasks, Round Rock, TX). The MoSi phase shift masks investigated in this study were inspected with a KLA 351 patterned defect inspection system (KLA-Tencor, San Jose) running the advanced performance algorithm (APA). Both i-line and deep ultraviolet (DUV) Verithoro programmed defect masks from DuPont Photomasks were investigated.

Although engineers have traditionally used shearing microscopes to measure programmed defect sizes, the sizing errors associated with these instruments can be as large as 100 nm, making it difficult to characterize the inspection tool's sensitivity accurately, especially when dealing with defect sizes <0.5 µm. Fortunately, as scanning electron microscopes dedicated to mask measurement have become available, SEM-based defect review and sizing have come to be considered essential for advanced mask development and qualification. The tests described here, which paid particular attention to defects 0.4 µm, were performed with a JWS-7815 SEM from JEOL (Akishima, Japan) that had been calibrated with polystyrene latex (PSL) spheres. Seven PSL spherical particle size standards were employed, ranging from 50 to 404 nm in diameter. The size uncertainty for all of these NIST-traceable PSL spheres, which were obtained from Duke Scientific (Palo Alto, CA), was 6 nm.

A dilute particle mixture was formed with deionized water (filtered to 0.05 µm) and dispersed onto masks with a spray gun. One mask for each sphere size was used. While depositing many particles on a mask made it possible to locate the particles with the SEM, care had to be taken to avoid particle clustering. A clean spray gun and careful dilution were essential to optimize particle density on the masks. Figure 1 is a photomicrograph of 240-nm PSL spheres on a mask structure. Electrostatic effects led to particle congregation along line edges of certain structures and more even distributions in other areas. Because particles along line edges tended to cause distortion, SEM particle measurements were taken only at a distance from line edges. Figure 2 displays an SEM image of a 240-nm PSL sphere on the mask shown in Figure 1 with a measurement box indicating the particle size measurement.

 

Figure 1: Incident light photomicrograph of 240-nm PSL spheres on a photomask. Only spheres in open areas were used to calibrate the SEM.

Figure 2: SEM image of a 240-nm PSL sphere on the mask shown in Figure 1.

After various SEM beam conditions were tested, settings for MoSi-on-quartz imaging were chosen (MoSi has a higher electron density than chrome). The probe current was kept low and the voltage setting was chosen to provide high resolution without risking the loss of surface definition that is common in low-electron-density materials at higher voltages.

The SEM-based PSL sphere measurement data and the error analysis were derived from 10 particles located away from the edges of each plate. The bias results indicate that the SEM measurements were fairly close to the stated sphere sizes, while no bias estimate was larger than 10 nm. Given the errors in defining the PSL standards according to a procedure provided by Duke Scientific as well as sampling and SEM measurement errors, a significant positive bias at 404 nm and a significant negative bias at 155 nm was observed. As shown in Figure 3, if a linear interpolation of the bias values is used to adjust the SEM measurements of MoSi defects, the accuracy error should be less than ±7 nm. However, this analysis is applicable only for defects within the 50­404-nm size range. Outside this range it is not clear what adjustments would be reasonable or what the relative accuracy of the analysis would be. Figure 3 demonstrates that because of its resolution and repeatability, an SEM permits the measurement of small artifacts with much greater accuracy than does a shearing microscope.

MoSi Mask Fabrication and SEM-Based Defect Measurement

One goal of the research described in this article was to characterize inspection sensitivity as a function of feature size, especially for feature sizes below the inspection system's specification. To accomplish this, two MoSi plates were built, one of which was composed of 8% i-line material and the other of 5% DUV material. Each plate consisted of five programmed defect patterns, each with a different minimum feature size. The labels for the patterns were 890EX for a 0.8-µm minimum feature size, 790EX for a 0.7-µm minimum feature size, 690EX for a 0.6-µm minimum feature size, 590EX for a 0.5-µm minimum feature size, and 490EX for a 0.4-µm minimum feature size. Patterns were written in 895i resist from Arch Chemicals (Norwalk, CT) on an Alta 3000 laser pattern generator from Etec Systems (Hayward, CA), after which a fluorine-based gas etch of the MoSi was performed in a high-density plasma reactor. Figure 4 shows the test mask layout and Figure 5 shows pinhole and pinspot design examples from the test mask.

 

Figure 3: SEM measurement bias curve (99% confidence interval) with estimated standard, sampling, and measurement errors.

Figures 6 through 10 illustrate the methods used to measure edge, corner, pinhole, and pinspot defects. The SEM image in Figure 6 shows a clear extension defect in the MoSi i-line 890EX pattern. Figure 6a highlights the defect in the pattern while the rectangular box in Figure 6b indicates the defect size measurement with a 45° SEM stage rotation. The size of the defect is given as the y value printed at the bottom of the SEM image, reflecting the height of the rectangular box. The box is aligned with the correct feature edge and with the maximum excursion of the defect away from the intended line edge.

 

Figure 4: Test mask layout showing the five programmed defect patterns on the mask (left) and a magnified image of some of the test cells (right).

Figure 5: Pinhole design examples (a) and pinspot design examples (b) from the test mask.

Figure 6: SEM measurement of clear extension defect in the MoSi i-line 890EX pattern: (a) defect in the pattern and (b) defect size measurement with a 45° SEM stage rotation.

Figure 7 illustrates an external corner clear intrusion defect measurement on MoSi DUV 890EX. The upper left corner of the measurement box in Figure 7a defines the ideal corner and is locked in place. In Figure 7b the lower right corner of the measurement box is moved to identify the maximum extent of the defect from the ideal corner. The linear defect size is then defined as (x² + y²)^1/2, where x and y are the dimensions of the box in Figure 7b.

Figure 8 illustrates an external corner dark extension defect measurement on MoSi DUV 890EX. The ideal corner is defined by the lower right corner of the measurement box shown in Figure 8a. With the upper left corner of the measurement box locked in place, the lower right corner is moved to define the maximum extent of the defect from the ideal corner, as seen in Figure 8b. The linear defect size is then defined as ((x^₂ ­ x^₁)² + (y^₂ ­ y^₁)²)^1/2, where (x^₁, y^₁) and (x^₂, y^₂) define the horizontal and vertical extents of the boxes in Figures 8a and 8b, respectively.

 

Figure 7: External corner clear intrusion defect measurement on MoSi DUV 890EX: (a) the ideal corner and (b) maximum extent of the defect from the ideal corner.

Figure 8: External corner dark extension defect measurement on MoSi DUV 890EX: (a) the ideal corner as defined by the lower right corner of the measurement box and (b) maximum extent of the defect from the ideal corner.

Figure 9a shows a pinspot defect on MoSi i-line 890EX and Figure 9b shows a magnified version of this defect. The defect size, in this case 162 nm, is determined by the larger of the x or y measurement box sides. Figure 10 illustrates a pinhole defect measurement on MoSi i-line 890EX, where the defect size is the larger of the x or y measurement box sides, in this case 583 nm.

 

Figure 9: Pinspot defect measurement on MoSi i-line 890EX (a) and magnified version of this defect (b).

The beam conditions used for these SEM-based defect measurements were the same as those used for SEM calibration, unless the magnification was changed (to measure a large defect, for example). SEM measurement techniques mimic those employed by optical microscopes, since in both cases defect sizes are defined as the largest distance away from the correct geometry.² However, the optical microscope method accounts for corner rounding while SEM measurements assume a perfectly square corner. Thus, corner defects with Cr excursions toward the interior of the feature on the MoSi DUV mask, as in Figure 7, appear larger in SEM images than in optical microscope images. On the other hand, corner defects with Cr excursions toward the exterior of the feature on the mask, as in Figure 8, appear smaller in SEM images than in optical microscope images. The measured defect sizes were generally within a few nanometers of the targeted size. Similar results were observed for the MoSi i-line mask.

Sensitivity of the Inspection Tool

The pattern inspection system scanned the masks at a wavelength of 488 nm. While binary photoblank materials at this wavelength transmit about 0.1% of the light, i-line MoSi material transmits about 20% and DUV MoSi about 40%. Thus, the dynamic range of the inspection tool transmission is lower for MoSi than for binary material, making it essential to assess the effect's impact on tool sensitivity.

Since the design of PSMs tends to be aggressive because of their small features, the effect of main feature size must be studied as well. To inspect masks with main feature sizes below the tool's 0.75-µm minimum feature size specification, the tool must be desensitized properly. For feature sizes meeting the inspection tool's requirements, the 890EX pattern was used. For feature sizes below the tool's minimum specification, however, the 790EX, 690EX, 590EX, and 490EX patterns were used. The sensitivity studies conducted with these patterns involved 10 identical inspections of the subject pattern, in which the capture rate for each defect was calculated in 10% increments. All inspections were performed with a 0.25-µm pixel size. The inspections yielded the following results:

• On the i-line MoSi 890EX pattern, the inspection system exhibited slightly better sensitivity to dark-edge defects and pinholes in the die-to-die (D-D) mode than in the die-to-database (D-B) mode. However, the D-B mode was more sensitive to corner defects than the D-D mode.
• On the DUV MoSi 890EX pattern, the tool exhibited better sensitivity in the D-D mode than in the D-B mode for every defect category. However, for sizes 0.18 µm, the tool performed better for five of the eight defect types studied than for the other three.
• Except for the DUV D-B case, the tool's performance met the binary APA specification for edge defects, which is 0.20 µm.
• The tool did not meet the binary APA specification for pinholes, which is 0.30 µm. Because opaque corner intrusions are undersized on SEM images, the situation is less clear for this defect type.
• Some of the results of the feature inspections below the minimum specification of 0.75 µm are listed in Tables I through IV. Each table shows the sensitivity settings employed for the range of test patterns as well as the minimum defect size with a 100% capture rate for three types of defects (line edge, corner, and pinhole). For each set of runs, tool sensitivity was set at the most sensitive value that yielded acceptable false defects. The data indicate that:
• The tool can inspect feature sizes as small as 0.5 µm on the MoSi i-line material and 0.6 µm on the MoSi DUV material. On the MoSi DUV material, the tool must be completely desensitized to run in the D-B mode, even for 0.8-µm features.
• Inspecting features smaller than the 0.75-µm specification requires desensitizing the tool and imposes a sensitivity penalty. Pinhole sensitivity in particular degrades rapidly with desensitization.

The images in Figure 11 depict a pinspot and a pinhole defect detected on a MoSi i-line product plate. The defects were captured in D-D mode on the KLA 351 and reviewed on the JWS-7815 SEM. Images from both tools are shown. X-ray emission spectroscopy revealed that the pinspot was MoSi and that the pinhole contained no foreign material.

Conclusion

This study demonstrates that programmed defects on MoSi i-line and DUV Verimasks can be measured on an SEM with better than 10-nm accuracy. This measurement error is significantly better than that obtained with a shearing microscope and provides the means to qualify inspection systems with sensitivities <150 nm.

 

Figure 10: Pinhole defect measurement on MoSi i-line 890EX.

Figure 11: Defect inspection system and SEM images of pinspot and pinhole defects on a MoSi i-line mask.

Sensitivity studies on MoSi PSMs indicate that the KLA 351/APA inspection system captures edge defects and corner defects on i-line MoSi masks with sensitivity similar to that achieved on binary masks. Pinhole defects, on the other hand, are captured at reduced sensitivity. DUV MoSi, with its relatively high transmission rate, impairs tool sensitivity significantly more than the sensitivity on binary masks. There is little difference in sensitivity between the D-D and D-B modes on i-line MoSi, but the difference is substantial on DUV MoSi. For both phase shift materials, the inspection tool must be desensitized as the feature size decreases, thus impairing capture rates. At reduced sensitivity, the tool can inspect features as small as 0.5 µm on the MoSi i-line material and 0.6 µm on the MoSi DUV material.

Acknowledgments

We would like to thank John Riddick at DuPont Photomasks Reticle Technology Center (Round Rock, TX) for performing some of the inspection runs. We also wish to acknowledge Scott Pomeroy at KLA-Tencor in Austin, TX, for engaging in helpful discussions and providing applications support. Finally, we wish to thank Adam Pudnos, pilot line manager at DuPont Photomasks Reticle Technology Center, for offering inspection tool time.

References

1. M Ushida et al., "Development of Deep-UV MoSi-Based Embedded Phase-Shifting Mask (EPSM) Blanks," in Proceedings of the SPIE: 16th Annual BACUS Symposium on Photomask Technology and Management (Bellingham, WA: SPIE, 1996), 403­411.
2. H Villa, "Verimask Measurement Procedure" and "Verithoro Measurement Procedure" (Santa Clara, CA: DuPont Photomasks).

Jerry X. Chen, PhD, is a staff inspection engineer and development project leader at the DuPont Photomasks Reticle Technology Center (Round Rock, TX). He has more than 10 years of experience in semiconductor process and inspection technology, including at Motorola, Semiconductor Product Sectors, and Microlithograph. He received a BS in laser optics from Zhongshan University (Guangzhou, China) in 1982, an MS in physics from Florida Atlantic University (Boca Raton) in 1992, and a PhD in physics from the University of Texas (Austin) in 1999. (Chen can be reached at 512/310-6419 or jerry.chen@photomask.com.)

Robert K. Henderson, PhD, holds a metrology engineering position at the DuPont Photomasks Reticle Technology Center. Before this assignment, he held a variety of management staff positions at the company, working on various internal consulting assignments for the parent DuPont company. He has BA degrees in mathematics and history from Trinity University in San Antonio, TX, a PhD in mathematical statistics from Southern Methodist University in Dallas, and an MBA from the University of Delaware in Newark. (Henderson can be reached at 512/310-6409 or robert.henderson@photomask.com.)

Franklin Kalk, PhD, is development manager at the DuPont Photomasks Reticle Technology Center. His group develops photomask materials, processes, and tools. Kalk holds many patents and has published numerous articles and papers. He received a PhD in optics from the University of Rochester in Rochester, NY. (Kalk can be reached at franklin.kalk@photomask.com.)

MicroHome | Search | Current Issue | MicroArchives
Buyers Guide | Media Kit