Phase enhanced DUV inspection of alternating phase shift reticles

Previous work [1, 2] described early phase contrast enhanced inspection results using altPSMs manufactured for the 248nm lithographic wavelength with inspection wavelengths at 364nm (UV) and 257nm (DUV). Since that time, DUV inspection tools have been shipped to advanced lithographic companies whose primary interest is applying altPSM techniques at the 193nm lithographic wavelength. Work has also been carried out to extend characterisation to reticles for the 193nm lithographic wavelength and targeting sub-90nm node critical dimensions.

A programmed defect test mask has been designed based upon a darkfield alternating phase (DAP) mask design using the poly gate layer of a 6-transistor static random access memory layout as the background pattern. The "as-drawn" poly gate length target critical dimension in this design is 70nm at 1X. Since the programmed defects were placed on-edges and isolated from edges, the name "Dione" was chosen for this test mask as an acronym for "Defects Isolated and On-edge". The design included programmed phase bump and divot defects at 60¡, 120¡ and 180¡ with target design sizes of 14-280nm incremented by 14nm (20 programmed defect sizes).

One of the challenges in executing the CAD layout was in the pre-compensating for the expected mask making process bias. A detailed understanding of the mask making process bias characteristics, writer overlay errors, writer minimum address grid, and writer resolution is required to produce sufficiently small (<<100nm) programmed defects of the desired size and shape. Further complicating this task is the issue that different writing tools are used for patterning the chrome and phase layers and that the associated processes for the chrome and phase layers have different inherent process biases. An understanding of these parameters was considered crucial since they were key to producing sufficiently small on-edge programmed phase defects. On-edge defects were considered as one of the most critical defect types since they are most likely to affect the printed gate critical dimension. Fortunately, a number of the process bias values had previously been experimentally estimated during prior 248nm test mask fabrication. The test mask vendor provided this information. It had also been demonstrated from the 248nm test mask results that overlay registration between the chrome and phase layers was sufficient (<30nm, 3s) to produce sub-100nm on-edge programmed defects. In order to resolve the small-programmed phase defects, the phase layers were exposed using a 50kV shaped e-beam exposure system instead of a raster scanned laser-based system, which is typically used for second-level exposure of altPSMs.

Programmed defect images were captured using a KLA-Tencor 8100XP-R CD-SEM at 75KX magnification. The defect sizes were then measured from these images using the maximum inscribed circle diameter (MICD) methodology [3]. Due to the finite resolution of the CD-SEM and quartz edge slope, the edge of the quartz defects was defined as the 50% midpoint of the white edge "halo". The chrome line width critical dimension was also estimated from SEM images and estimated to be 260nm. Several Dione programmed phase defect CD-SEM images are shown in Figures 1 and 2.

Fig.1a: 180¡ programmed on-edge phase bump defects

Fig.1b: 120¡ programmed on-edge phase bump defects

Fig.1c: 60¡ programmed on-edge phase bump defects

Fig.2a: 180¡ programmed on-edge phase divot defects

Fig.2b: 120¡ programmed on-edge phase divot defects

Fig.2c: 60¡ programmed on-edge phase divot defects

A comparison of phase defect detection performance was performed for the KLA-Tencor TeraScan 525 DUV inspection system. All inspections were performed at the 125nm pixel size. A comparison was made of the standard transmitted light configuration versus the phase contrast configuration, called TeraPhase 501. Sets of ten consecutive inspections were performed using the TeraPhase 501 configuration in order to estimate the defect sensitivity performance. A set of five consecutive runs for the standard transmitted light configuration was made to estimate defect sensitivity. Focus offsets were used for the TeraPhase 501 inspections since this is the recommended normal operating mode. The recommended positive (objective closer to the object plane) focus offset was empirically determined to be at a system setting of +458nm for best phase bump sensitivity. Phase bumps are the most commonly produced phase defect type by most altPSM quartz etch processes and hence were given detection priority. Separate sets of optimisation runs were performed at negative defocus (objective further from the object plane) in order to maximise phase divot sensitivity. The results of these runs showed that an increase in phase divot sensitivity was realised at a system setting of -300nm. The phase defect sensitivity results are shown in Figures 3 and 4.

The Dione test mask CD-SEM images showed that programmed phase bump defects smaller than 100nm were successfully resolved (Figure 1a-c). Small phase divot defects were also resolved, but it was apparent from the CD-SEM images that defects smaller than ~200nm were not etched to the target etch depth. For the on-edge phase divots, the slope of the phase edge tended to decrease and "fill-in" the defect at approximately 200nm for the 180¡ phase divots and at approximately 220nm or larger for the 120¡ and 60¡ phase divots (Figures 2a-c). The phase bump defect heights appeared to be less affected by this problem. Without taking this behaviour into account, it would appear from the phase defect detection performance characterisation that the TeraScan 525 system is less sensitive to phase divots. It is believed that this apparent sensitivity difference between phase bumps and divots results from the test plate phase defect etch depths.

During the course of test mask manufacturing, a system of checks and balances was initiated with the vendor in order to minimise risk. Since making a programmed phase defect test mask involves multiple writing and processing steps (a minimum of four write and processing steps for a three phase defect height test mask), test masks of this type have a long cycle time and high value associated with them. The basic premise of the checks and balances put in place was to detect any defect issues as early in the process as possible before more value was added. To accomplish this, additional defect inspections were put in place after several of the process steps throughout the mask manufacturing cycle. The inspections at the vendor's site were performed with a TeraStar SLF27 using the TeraPhase 400 inspection algorithm using the maximum sensitivity setting of this system. These inspection results were made available to KLA-Tencor via ftp transfer over the internet. After reviewing inspection results at KLA-Tencor, a mutual agreement to proceed with the next set of processing steps was reached.

Fig.3: Comparison of TeraScan 525 inspection modes for on-edge phase defects

Figures 3 and 4 show the sensitivity characterisation results. Figure 3 shows results for the on-edge phase defects and Figure 4 shows results for the isolated phase defects. The tables are arranged with defect types in the major row headings and defect size arranged in columns. The defect sizes are expressed in nanometres and are to the right of the defect labels. An empty cell in the table to the right of the defect label indicates defects that did not resolve on the test mask. Under each defect type is listed the three test conditions. These are "TeraScan 525" indicating the normal transmitted light mode of operation, "TeraScan 525 + TeraPhase 501 FO +458" indicating the phase contrast mode of operation with a +458nm focus offset and "TeraScan 525 + TeraPhase 501 FO -300" indicating the phase contrast mode of operation with a -300nm focus offset. The shaded bars to the right of the operating mode indicate 100% detection of the stated defect size. Detection less than 100% is indicated by the percentage detection rate in the appropriate box.

Fig.4: Comparison of TeraScan 525 inspection modes for isolated phase defects

On-edge phase bump sensitivity results in Figure 3 show an increasing improvement with TeraPhase 501 over the standard transmitted light configuration as the programmed phase defect height decreases. At 180¡ bump height, an improvement from 83nm to 63nm with TeraPhase 501 using a +458nm focus offset can be seen. TeraPhase 501 with a -300nm focus offset appears to perform worse than standard transmitted light for the 180¡ on-edge phase bump (113nm vs. 83nm). However, at 120¡ and 60¡ bump heights, both positive and negative focus offsets show improved performance as compared to the standard transmitted light configuration. The positive focus offset shows greater sensitivity at all phase bump defect heights.

On-edge phase divot sensitivity results in Figure 3 show improved sensitivity with TeraPhase 501 for both focus offsets as compared to the standard transmitted light configuration. The improvement appears to be less dramatic than the on-edge bump defect improvements in terms of numerical defect size improvements. As noted above, the phase divot defects appear to be less than full height for defect sizes less than approximately 220nm. Furthermore, fewer of the small (<100nm) on-edge divot defects resolved as compared to on-edge bump defects.

Fig.5: Proposed inspection methodology for hypothetical altPSM manufacturing process

The isolated phase defect sensitivity shows the general trend of improved phase defect detection with TeraPhase 501 as compared to the standard transmitted light mode. However, the detection behaviour is dissimilar for the positive focus offset and isolated phase divots as compared to the on-edge phase divots. The data shows a decrease in sensitivity for isolated divots using a positive focus offset as compared to the standard transmitted light mode. For on-edge defects, TeraPhase 501 defect sensitivity equals or exceeds the standard transmitted light mode. This behaviour was a result of optimising the positive focus offset to find on-edge phase bumps. All of the resolved isolated phase defects were detected using TeraPhase 501 at negative focus offset with the exception of the two smallest 60¡ phase bumps (13 and 42nm). "Resolved" means the smallest defect the vendor was capable of manufacturing.

A two-pass inspection at a positive and negative focus offset can be used to maximise phase bump and divot defect sensitivity. The user interface in the TeraScan system allows an operator to clone and then edit a previously defined inspection pass. To define a second inspection pass, an operator would copy the defined first pass setup and then simply edit the focus offset in the user interface during the normal plate setup. The TeraScan system would then automatically perform both inspection passes when the inspection was started. Another possible implementation methodology is to perform single focus offset TeraPhase 501 inspections at intermediate process steps where bump or divot defects are most likely to be generated and optimising the focus offset to enhance the detection of the most likely defect type to be produced. The exact implementation of this methodology would be determined by the specific altPSM manufacturing processes used. As an example, phase divot defects are considered "killers" since effective repair is virtually impossible. Therefore, it is desirable to detect phase divot defects as early as possible in manufacturing before more value is added by additional write and process cycles. If a divot defect were intrinsic to the quartz substrate, it would be better to detect this defect as early as possible. That is, after the initial chrome write and process step and not at the final quality assurance (QA) inspection step. As such, TeraPhase 501 can be used while the mask still exists as a "binary" mask to detect intrinsic phase defects thereby improving cycle time and reducing scrap costs if a "killer" divot defect is detected early in the manufacturing process. Figure 5 shows a table of a possible best known method (BKM) implementation strategy for a hypothetical two patterning step process using a phase balance etch step. The best BKM implementation will depend upon the exact altPSM manufacturing process flow.

After a defect is detected it is desirable to determine the nature of the defect during defect review. One method available with TeraPhase 501 is to manually adjust the focus offset above and below nominal focus from the defect review screen and observe the behaviour of the defect's image. When focus is adjusted for a chrome defect it appears to be of constant polarity. A dark extension appears to be dark on both sides of nominal focus. Bump and divot defects show a different behaviour than chrome defects in that they change polarity when the focus offset is adjusted above and below nominal focus. Furthermore, the change in polarity behaviour is the opposite for bumps and divots, which allows for identification and classification. Phase bump defects appear darker than the quartz background under positive focus offset conditions whereas phase divots appear brighter than the quartz background. Under negative focus offset conditions, phase bumps appear brighter than the quartz background and phase divots appear darker than the quartz background. Figure 6 illustrates this behaviour using the captured inspection defect images. It should be noted that the same bright/dark phase defect behaviour is also apparent in the difference images (lower images in Figure 6) when the defect is place into the "Test" image location. TeraStar and TeraScan defect review screens show "Reference" and "Test" images where the image containing the defect is located in the "Test" image location. The defect review operator classifies the defects so that the defect appears in the "Test" location.

Fig.6: TeraPhase 501 defect review. Determining whether a defect is a chrome, bump or divot defect using defect review images from a multi-pass inspection. Bump and divot defects change polarity in opposite fashion whereas chrome defects behave in a similar manner on both sides of focus

Authors:

Larry Zurbrick, Maciej Rudzinski, KLA-Tencor.

Long He, Kurt Kimmel, Alvina Williams, International Sematech.

Acknowledgement:

The development of phase contrast enhanced inspection hardware and software was partially funded under International SEMATECH contract LITG-336.

REFERENCES:

1. L Zurbrick, M Rudzinski, S Stokowski, L He, K Kimmel and N Kashwala, "Alternating Phase Shift Mask Inspection Through the Use of Phase Contrast Enhancement Techniques", 22nd Annual BACUS Symposium on Photomask Technology, Brian J Grenon, Kurt R Kimmel, Editors, Proceedings of SPIE vol.4889, pp.241-246, (2002).

2. L Zurbrick, M Rudzinski, S Stokowski, L He, K Kimmel and N Kashwala, "Inspection of alternating phase shift masks through the use of phase contrast techniques", 19th European Mask Conference on Mask Technology for Integrated Circuits and Micro-Components, pp.167, VDE Verlag, 2003.

3. L Zurbrick, S Khanna, J Lee, J Greed, E Laird, R Blanquies, "Reticle defect size calibration using low voltage SEM and pattern recognition techniques for sub-200nm defects", in 19th Annual BACUS Symposium on Photomask Technology, Frank E Abboud, Brian J Grenon, Editors, Proceedings of SPIE vol.3873, pp.651-58 (1999).