RequestLink
MICRO
Advertiser and
Product
Information

Buyer's Guide
Buyers Guide

tom
Chip Shots blog

Greatest Hits of 2005
Greatest Hits of 2005

Featured Series
Featured Series


Web Sightings

Media Kit

Comments? Suggestions? Send us your feedback.

 

MicroMagazine.com

Behind the Mask

Performing 3-D metrology and advanced repairs on leading-edge photomasks

Troy Morrison and Chris Marotta, FEI

As chipmakers continue toextend photolithographic processes to create subwavelength-sized features, photomasks have evolved from simple, binary, 2-D patterns of chrome on glass to complex 3-D structures. In addition to containing the circuit pattern, advanced photomasks include specialized structures that enhance the contrast and resolution of the image that will be transferred to the wafer. Resolution-enhancement techniques add or subtract small features from the pattern to improve the fidelity of the transferred image. Phase shifters manipulate light to improve image contrast and sharpen edges. While these advances have allowed manufacturers to extend the use of expensive equipment across device generations, they have also increased the need for advanced mask metrology and repair capability.

Driven by increased device complexity, decreasing feature size, and the growing number of photomasks per circuit, mask costs have risen dramatically in recent generations. As a result, the need for improved mask-repair capability has increased commensurately. This article describes the use of an integrated measurement-and-repair system that uses stylus nanoprofilometry (SNP) for 3-D metrology, a focused ion beam (FIB) for material removal and deposition, and an electron beam (E-beam) for high-resolution imaging and repair. In most of the work described in this article, an SNP XT from FEI (Hillsboro, OR) was used to perform 3-D metrology, while the company’s Accura XT+ was used to perform mask-repair steps. The integration of these techniques into a single system permits users to evaluate and control mask-repair processes quickly and accurately.

Mask Metrology

The use of phase-shifting masks has added a new dimension to metrology requirements. Because it must retard the incident light in the direction of propagation, phase shifting necessarily operates in the third dimension, perpendicular to the plane of the pattern. Although there are different approaches to phase shifting, all of them depend on manipulating the thickness and/or composition of material in the light’s path, requiring accurate 3-D metrology.

Nondestructive critical dimension (CD) metrology is typically performed using CD scanning electron microscopy (CD-SEM), optical scatterometry, or scanning probe microscopy. Because of its speed and precision, CD-SEM dominates CD measurements in the production environment. However, because of its top-down perspective, it is generally not capable of performing 3-D measurements. In mask applications, CD-SEM is further hindered by charging artifacts on nonconductive mask materials. The second method, scatterometry, is fast and precise, but its requirement for large regular arrays of identical features restricts its ability to provide in-circuit measurements. Finally, atomic force microscopy, a form of scanning probe microscopy, provides accurate 3-D characterization but can be relatively slow.

The SNP XT is a form of scanning probe microscopy that is optimized for CD metrology. Like all forms of scanning probe microscopy, it uses a physical probe to interrogate the sample, but it differs from most other forms in several significant ways. In most scanning probe microscopy technologies, the probe is attached to a flexible cantilever. Movements of the tip are detected by tracking the movements of a laser beam reflected from the back of the cantilever. Often the scanning probe microscope vibrates the tip to improve sensitivity—at the cost of reduced precision. The flexibility of the cantilever also interferes with fine position measurements because it allows the probe to respond to forces that develop between the tip and the sample.

In contrast, SNP’s probe is attached to a noncompliant rocker beam and a precision bearing. The capacitive actuator/sensor provides direct information about the tip location and helps control applied force. In CD metrology applications, eliminating tip oscillations and cantilever flexibility improves edge-detection accuracy and measurement precision. Direct control of the tip position also enables users to perform direct edge-to-edge measurements, significantly improving throughput. Figure 1 contains a schematic diagram of an SNP.

Figure 1: Schematic diagram of the SNP technique.

Current-generation masks include geometries such as vertical and undercut sidewalls that cannot be characterized easily using the cone-shaped probes typically used in scanning probe microscopy. In contrast, SNP provides shaped probes that are designed for these geometries. Images of two different probe-tip shapes are presented in Figure 2. For features that are not severely undercut, a slightly flared tip can measure height, width, and wall angle. Large undercuts (up to 150 nm), such as those found in rim-shifter phase-shift masks, require a hammer-shaped probe. Probes are made from single-crystal diamond and can be fabricated in virtually any shape using FIB milling. A simple magnetic attachment mechanism makes it easy to exchange probe tips to meet a variety of analytical applications, and the exchange procedure is completely automated. The SNP system maintains an onboard library of precharacterized tips.

Figure 2: SNP requires different probe tip shapes for different feature shapes: (a) a flared tip for height, width, and positive-sidewall-angle measurements, and (b) a hammerhead-shaped tip for severly undercut features.

Measurement of feature shapes requires a precise and detailed knowledge of probe-tip shapes, which can be acquired by scanning each tip over a series of calibration artifacts containing features that are designed for shape characterization. Once a shape is scanned, software extracts the true feature shape from the measured data points using a mathematical process known as erosion. SNP has demonstrated subnanometer static precision on photomask CDs and angstrom-level precision on z-height measurements.1

Most forms of scanning probe microscopy maintain close proximity between the probe and the sample surface throughout the scanning procedure. While that procedure works well in imaging applications, it is not ideal for CD measurements since it spends much time acquiring data points that do not contribute to CD measurement quality. Using pattern recognition, SNP has an efficient point-to-point measurement algorithm that acquires data only from the critical points required for each measurement. In addition to improving throughput, reducing the time required for each measurement also improves repeatability by minimizing errors associated with system drift during data acquisition.

Mask Repair

The latest generation of particle-beam-based mask-repair systems incorporate both E-beam and ion-beam capabilities. Either beam may be used for imaging or for adding or removing material from the mask. In imaging mode, the system scans the beam over the sample surface and maps various signals that are generated by interactions between beam particles and sample atoms. These signals include secondary electrons, which provide good image resolution. In the case of ion beams, the signals also include secondary ions, which carry information about sample composition. Although the beams operate similarly, their imaging capability differs. In general, E-beams can focus on a smaller spot than ion beams, providing better spatial resolution. E-beam imaging can also be used for extended periods without damaging the sample.

In repair mode, the Accura XT+ mask-repair system uses a FIB to deposit and remove material from the sample. The relatively massive FIB ions erode material directly through physical sputtering processes. While FIB-based milling is a well-developed technology that generally offers higher removal rates on a broad range of materials than other methods, it can damage the sample or leave residual implanted beam ions.

Gas-assisted etch can enhance FIB milling rates and specificity by injecting small amounts of gas through a fine needle near the sample surface in the vicinity of the beam. The injected gas combines with sputtered sample molecules to form volatile compounds that are then extracted by a vacuum system. Gas-assisted etch permits users to select gases that preferentially enhance the removal of certain materials.

In contrast to FIB milling, E-beam milling is a relatively new technology that offers tighter spatial control of the milling step. A purely chemical process, it does not involve physical sputtering, thus avoiding beam-induced damage and contamination. While the E-beam milling process is chemically specific to the composition of the sample, rendering it less than suitable for general milling applications such as sample cross-sectioning for subsequent imaging, its specificity can be an advantage in repair applications in which it is often desirable to preferentially mill certain materials while leaving others undisturbed. In addition, E-beam milling offers higher spatial resolution than FIB milling. An example of an E-beam repair is presented in Figure 3.

Figure 3: SEM images showing E-beam repair of a MoSi defect: (a) mask with extraneous material before repair, and (b) mask after the material has been removed.

Employing fragile new materials and smaller feature sizes than current mask technologies, next-generation technologies will favor the use of E-beam imaging and repair. At the 65-nm node, the industry is generally moving toward the use of E-beam technology for many repair processes, a trend that will only increase at the 45-nm node. However, since certain next-generation mask materials have proven difficult to repair using E-beam technology, a flexible dual-beam approach will be required to repair all types of masks.

Repairing Different Types of Masks

Chrome Binary Masks. Chrome-on-quartz binary masks constitute the majority of masks used in the semiconductor industry. In these masks, chrome forms an opaque pattern on a transparent quartz substrate. Opaque defects occur when chrome remains in a region that should be transparent. Using a FIB, the defective chrome can be removed precisely.

Figure 4 shows a programmed test defect extending from a chrome line in a 280-nm half-pitch array (70-nm printed CD). The primary criterion for mask repairs is the quality of the printed features in the vicinity of the repair. Feature quality can be measured using an Aerial Image Measurement System (AIMS) from Carl Zeiss (Oberkochen, Germany). The SNP’s 3-D measurement capability provides the means to relate the physical characteristics of the repair to the optical quality of the printed features. Precise edge placement and minimal quartz damage are the most important physical factors determining repair performance.2

Unintended milling of the underlying or surrounding quartz and the deposition of process by-products during chrome removal can reduce transmission through a repaired opaque defect. While edge biasing—reducing the width of the chrome line adjacent to the repair, as shown in Figure 4—can restore transmission through the repaired area, it begins to reduce the CDs of the printed lines at some level. Gas-assisted etch and chrome-only milling can help to maintain transmission by reducing both damage and redeposition.

Figure 4: Binary chrome-on-glass mask (a) before and (b) after the removal of an opaque defect that was created to test a FIB-based repair approach. The reduced linewidths after the repair are an intentional edge bias that was introduced to restore transmission in the vicinity of the repair. 2

Bromine preferentially enhances chrome removal rates and water vapor inhibits quartz removal. Used alone, bromine can reduce overall milling time significantly and, consequently, the amount of time the quartz is exposed to potential beam damage. While the combination of bromine and a second gas further reduces quartz damage, that approach may increase the redeposition of milling by-products in
some cases.

Figure 5 compares the residual quartz damage and residual process by-products resulting from the use of bromine and bromine plus another gas. The figures overlay a series of transverse SNP scans along the length of the line encompassing the repair area. In a perfect repair, all of the scans would very nearly coincide. The bromine-only repair (Figure 5a) shows little evidence of residual process by-products, but it does show some quartz removal below the defect. The repair using bromine and a second gas (Figure 5b) shows an increase in residual by-products but little or no quartz removal. While both processes can result in successful repairs, the best choice of gases depends on the particular repair requirements. Further research continues to find gases that offer better etch rates and higher selectivity.

Figure 5: Residual quartz damage and residual process by-products resulting from using (a) bromine alone and (b) bromine plus a second gas to enhance the removal of a chrome opaque defect. 2

The repair tool includes a chrome-only milling mode. In this mode, the system acquires a low-dose image of the repair area as defined by the operator. Within the repair area, it automatically distinguishes chrome from quartz, scanning the ion beam only over the chrome. As the bulk of the chrome is removed, the repair begins to break through to the quartz. As this occurs, the system repeats the low-dose imaging with increasing frequency to update its chrome/quartz map and maintain chrome-only milling.

Attenuating Phase-Shift Masks. As feature sizes have decreased, diffraction at the edges of patterns has limited the resolution and contrast of the image projected by a simple binary mask. Phase-shift masks use interference between light transmitted through various regions of the mask to improve the performance of the pattern-transfer process. Interference occurs when two “beams” of coherent light have different phases. In the absence of phase difference, the two beams interfere constructively, and their intensities add up normally. If the beams are 180° out of phase, they interfere destructively and their intensities cancel each other.

In attenuating phase-shift masks, the opaque pattern material is replaced by a partially transparent material, such as molybdenum silicide (MoSiON), which is deposited in a thickness that attenuates the incident light by approximately 94% and, at the same time, shifts its phase by 180°. Light passing through the clear areas of the mask diffracts at the edges of the pattern, so that some of it falls in areas that are intended to be dark. Phase-shifted light passing through the attenuator interferes destructively with the diffracted light, resulting in an overall improvement in contrast and resolution.

Defects in attenuating phase-shift masks can be either opaque (containing attenuating material where there should be none) or clear (lacking attenuating material). Opaque defects are repaired using a gas-assisted-etch E-beam or FIB with xenon difluoride to remove the unwanted material. Clear defects are repaired using the Accura system’s phase-and-transmission-matching software (PaTMatch), which controls the injection of a proprietary gas in the vicinity of the beam to deposit a transmissive quartzlike film that matches the transmission of the bulk MoSiON absorber material. That film is then deposited in a thickness that yields the correct 180° phase shift. This process results in repairs that offer better optical performance than those that simply replace missing absorber with opaque styrene-based material.

As with repairs of chrome binary masks, the primary concerns with repairing attenuating phase-shift masks are edge placement, quartz damage, and the presence of residual material. And as with binary masks, the primary measure of success is transmission, which is measured using AIMS. FIB-based repairs of attenuating phase-shift masks typically achieve transmission values that are within ±3% of the values of nondefective sites.3

Using SNP, operators can perform 3-D characterization of repairs such as riverbeds and halos. Riverbeds are areas near the edges of opaque repairs where quartz removal should not have occurred. They appear because defect material sputters faster from the exposed edges of defects than from flat interior surfaces. With care, FIB repair of attenuating phase-shift masks can keep riverbeds well within acceptable limits.3 Halos are potential artifacts of clear repairs resulting from the deposition of carbon where repairs abut a clear region. They are caused by beam tails, the blurred edges of beams (which have a Gaussian profile) that possess enough energy to deposit a thin layer of material from the residual process gas outside the intended repair area. Typically, halos can be limited to less than 3 nm in height and have little or no effect on optical transmission.

Alternating-Aperture Phase-Shift Masks. Alternating-aperture phase-shift masks also manipulate the phase of illuminating light to improve optical performance. They are used for regions of the circuit that include densely spaced line arrays. In alternating spaces, quartz is etched to a level that shifts the phase of the transmitted light by 180°. Thus, light diffracted from one edge of an opaque line is cancelled by light diffracted from the opposite edge, as illustrated in Figure 6.

Figure 6: In an alternating-aperture phase-shift mask, the quartz in alternating spaces between lines is etched to a level that shifts transmitted light 180°.

In alternating-aperture phase-shift masks, quartz bumps occur when unwanted quartz remains in one of the phase-shifting trenches, as shown in Figure 7a. (The repaired bump is shown in Figure 7b.) Often these bumps are caused by a particle or defect in the overlying chrome pattern that protects the quartz during the trench etch step. While the chrome defect is subsequently removed, the quartz remains. Quartz bumps pose a difficult problem for FIB repair because they are composed of the same material as the underlying substrate and can be difficult to image.

One of the difficulties with any subtractive FIB repair is knowing when to stop. When the material to be removed has a different composition than the material to be left in place, the endpoint can often be determined by watching for signal changes that indicate that the interface between the two materials has been reached. This may be as simple as a change in signal intensity that is associated with the change in composition. With quartz bumps, no such interface or endpoint signal exists. The material to be removed is the same as the material to be left in place. Thus, quartz-bump repair requires careful characterization of the instrumental parameters that affect quartz removal and an equally careful characterization of the 3-D size, shape, and location of the bump.

In typical mask-manufacturing operations, quartz bumps are first detected and localized by optical inspection tools, after which the defect coordinates are passed to the mask metrology system. The mask is then loaded into the SNP, and the system navigates to the defect location and scans the defect to record its 3-D topography. The topographic data are transferred to the repair tool, which automatically defines the volume of quartz to be removed. Finally, the repair tool safely etches the quartz bump with a minimal impact on the mask substrate. Figures 8a and 8b present FIB images of a mask substrate before and after the removal of a quartz bump, respectively.

Figure 7: SNP images showing a quartz bump (a) before and (b) after repair4. Quartz bumps are residual quartz left in the 180 ° phase-shifting space of an alternating-aperture phase-shift mask.
Figure 8: Images of an alternating-aperture phase-shift mask (a) before and (b) after being repaired using a FIB tool. The quartz defect's faint outline can be seen in the center of the first image.

In all mask repairs, AIMS transmission is the primary criterion of success, and the topography of the repaired area is the primary determinant of transmission. Several aspects of topography are relevant: final quartz height at the repair relative to the surrounding quartz level, the extension of the defect into the phase-shifting space as a percentage of the space width, and the flatness of the repaired quartz.

Figure 9 shows the relationship between final quartz height and AIMS transmission. The data demonstrate a purely empirical relationship, with maximum transmission occurring when there is no difference in height between the repair and the surrounding quartz. For reference purposes, the figure also shows a plot of the transmission that could be expected as a result of interference shifts introduced by the final quartz height.

Figure 9: Final quartz height after bump repair plays a primary role in determining transmission. The lower curve, with a maximum percentage of transmission at 6 nm, is a purely empirical fit to the data, but it shows a close resemblance to the transmission that would be expected based on interference.5 (y = -0.002x2 + 0.002x + 0.8656, R2 = 0.95.)

The distance that a defect extends into the 180° phase-shift space is a good predictor of transmission after repair. Because the extension of a defect has a much stronger effect on transmission than its length, it provides a means for determining at an early stage in the repair process whether edge bias will be required to restore adequate transmission.5 Figure 10 plots the percentage of transmission for 500- and 1000-nm-long defects (in the direction of the line). The two curves largely overlap, showing that the extension of a defect in the transverse direction (into a space) has a greater influence on transmission than the length of the defect. Minimum transmission occurs for defects that fully bridge the space.

Figure 10: Relationship between extension (as a percentage of space width) and transmission as measured by AIMS.5 (The fitted dashed line corresponds to y = -0.2004x + 0.9651, R2 = 0.7028, and the fitted solid line corresponds to y = -0.249x + 1.0113, R2 = 0.8862.)

As might be expected, flatness was found to attenuate transmission in proportion to the magnitude of the residual topography. Moreover, if the quartz volume of the nonflat region is large enough relative to the quartz volume of the overall defect region, flatness also enters into the calculation of over- or underetch.

Conclusion

The number, complexity, and cost of photomasks required by advanced semiconductor manufacturing processes have escalated dramatically over recent technology generations. As mask costs have increased, mask repair has become critical for both device and mask manufacturers. Device manufacturers benefit from mask repairs by enjoying direct savings on recovered masks and faster time to market. Mask manufacturers save money and improve performance to schedule by eliminating the need to repeat the entire mask fabrication process in order to correct minor defects.

This article has described the use of an integrated metrology/FIB/E-beam system that measures features and repairs defects on binary chrome-on-quartz, attenuating phase-shift, and alternating-aperture phase-shift masks. In all cases, the integration of metrology and FIB/E-beam milling and deposition resulted in successful repairs. The utility of the system was shown particularly in the case of quartz bump repairs, where conventional endpoint detection cannot be used and the success of the repair depends on a combination of predictable etch performance and intelligent scan control.

The forces driving the rise in mask costs and complexity will certainly persist. In coming device generations, new mask structures, substrate and absorber materials, and lithography processes such as extreme ultraviolet and immersion will continue to proliferate. Mask repair will play an increasingly important role in both mask and device manufacturing. An integrated method, such as that described in this article, will be able to support advanced lithography processes now and in the device generations to come.

Acknowledgments

The authors offer special thanks to Marcus Ramstein and Stefan Burges of AMTC (Dresden, Germany) for their valuable assistance in preparing this article.

References

1. R Kneedler et al., “3D Metrology Solution for the 65 nm Node,” in Proceedings of the BACUS Symposium on Photomask Technology, SPIE vol. 5567 (Bellingham, WA: SPIE, 2004), 905–910.

2. D Ferranti et al., “Focused Ion Beam Repair of Binary Chrome Defects for the 65 nm Node,” in Proceedings of the BACUS Symposium on Photomask Technology, SPIE vol. 5567 (Bellingham, WA: SPIE, 2004), 486–496.

3. C Marotta et al., “Repair and Imaging of 193 nm MoSiON Phase Shift Photomasks,” in Proceedings of the BACUS Symposium on Photomask Technology, SPIE vol. 4562 (Bellingham, WA: SPIE, 2002), 1161–1171.

4. J Lessing et al., “New Advancements in Focused Ion Beam Repair of Alternating Phase-Shift Masks,” in Proceedings of Photomask and Next-Generation Lithography Mask Technology X, SPIE vol. 5130 (Bellingham, WA: SPIE, 2003), 496–509.

5. T Robinson et al., “Nanoscale Dimensional Focused Ion Beam Repair of Quartz Defects on 90 nm Node Alternating Aperture Phase Shift Masks,” in Proceedings of Photomask and Next-Generation Lithography Mask Technology XI, SPIE vol. 5446 (Bellingham, WA: SPIE, 2004), 384–401.


Troy Morrison is senior product manager at FEI in Hillsboro, OR. He joined the company in 2001 and has more than 15 years of experience in the semiconductor and mask industry. Previously, he was SNP applications manager at Surface Interface (Sunnyvale, CA) and product manager for CD-SEM at KLA-Tencor (San Jose). Morrison has also held various process engineering positions at Motorola and Intel. He received BS and MS degrees in materials science and engineering from the Massachusetts Institute of Technology in Cambridge. (Morrison can be reached at 503/726-7724 or tmorrison@feico.com.)

Chris Marotta is product marketing engineer for the Accura mask-repair product line at FEI in Peabody, MA. He joined the company in 2000 and has more than seven years of experience in the semiconductor and mask industry. Before assuming his present position, Marotta was an applications engineer for the Accura mask-repair tool. Before joining the company, he worked as an applications engineer in the metal etch area at Drytek (Wilmington, MA). He received a BA in chemistry and physics from Gettysburg College in Pennsylvania. (Marotta can be reached at 978/538-6736 or cmarotta@feico.com.)


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

Questions/comments about MICRO Magazine? E-mail us at cheynman@gmail.com.

© 2007 Tom Cheyney
All rights reserved.