Measurement of the phase defect size using a scanning probe microscope and at-wavelength inspection tool

Abstract. Predicting the lithography impact of a phase defect embedded in a mask used for extreme ultraviolet lithography on the printed image on wafer is a challenging task. In this study, two types of measurement tools were employed to characterize the phase defects. The prior measurement tool was a scanning probe microscope used for measuring the surface topography of phase defects, and the second was an at-wavelength dark-field inspection tool capable of capturing a phase defect and then calculating the defect detection signal intensity (DSI) from those images. A programmed phase defect mask with various lateral sizes and depths was prepared. The sizes and DSIs were then measured. The measured data indicated that the DSIs did not directly correlate with the phase defect volumes. The influence of the phase defects on the printed image on a wafer was also calculated using a lithography simulator. The simulation results indicated that the printed critical dimensions (CDs) were strongly correlated with the DSIs rather than with the phase defect volumes. As a result, the influence of the phase defect on the printed CD can be predicted from the values of the DSIs.


Introduction
Extreme ultraviolet lithography (EUVL) is considered the most promising next-generation lithography after the point at which 193-nm immersion lithography will not be able to deliver smaller features.However, the path to establish the EUVL is not without technical difficulties.For example, a lack of sufficient light-source power, particle-free mask handling, defect-free and flat mask blanks, [1][2][3][4][5] and resist material development 6,7 all need to be addressed.From the viewpoint of EUV mask fabrication, mask pattern defect inspection [8][9][10][11] and repair [12][13][14] are some of the most demanding tasks to be dealt with.6][17] Regarding the types of the defects, the nature of the pattern defects in the EUV mask is mostly the same as in the case of optical masks except for those defects that are classified as reflective multilayer defects, such as bump or pit phase defects.These propagate through the multilayer mirror during its deposition on to the substrate surface and are hard to repair. 18Therefore, to reduce the effect of a phase defect on the printed image, two elimination methods are suggested.0][21] The other is to eliminate the influence of the phase error by removing the absorber at the close proximity of the phase defects after fabricating the device pattern. 22n order to make these methods successful, it is necessary to measure the size or volume of the phase defects.This can be done using a scanning probe microscope (SPM).

Preparation of a Programmed Phase Defect Extreme Ultraviolet Mask
A programmed pit phase defect EUV blank was prepared.
The EUV mask used in this study consisted of a reflective multilayer deposited on a quartz substrate.The multilayer consisted of 40-bilayer Mo (2.2-nm thick)/Si (4.8-nm thick) with Ru (2.5-nm thick) as a capping layer.For this purpose, seeds for phase defects on a quartz substrate were fabricated through resist patterning and quartz etching processes before coating the multilayer.The quartz substrate was coated with a chrome layer as a hard mask in order to obtain enough etching selectivity against the quartz.A photoresist layer was coated on the chrome layer and a set of seeds to serve as phase defects, and guide patterns were drawn on the photoresist layer by an electron-beam pattern writing tool (JBX-9300FS, JEOL Ltd, Tokyo, Japan).After developing the photoresist layer, the chrome layer was etched.Then the quartz substrate was etched followed by photoresist stripping.To obtain different depths of the phase defects, resist patterning and quartz etching processes were carried out three times under different quartz etching process conditions.Figure 1 shows the design of the programmed phase After removing the chrome layer, the patterned quartz substrate was coated with the multilayer.

Phase Defect Size Measurement Using Scanning Probe Microscope
All phase defects had their images taken with a combination of SPM (L-Trace II, Hitachi High-Tech Science Corp., Tokyo, Japan) and a triangular pyramid shaped probe (SI-DF40P2, Hitachi High-Tech Science Corp., Tokyo, Japan), as shown in Figs.2(a) and 2(b), respectively.The typical values of their resonant frequency and spring constant were 300 kHz and 26 N∕m.The scan area and speed were set to 600 × 600 nm 2 and 1.7 Hz, respectively.The numbers of the image pixels were 512 points for the X direction and 256 points for the Y direction.
After capturing an SPM image, the defect size was calculated as illustrated in Fig. 3. Figure 3(a) shows the SPM image (after the image tilt correction) of the multilayer surface that includes the phase defect.Figure 3(b) shows line profiles of the SPM image.The depth of the phase defect was defined as a distance between the surface of the multilayer and the bottom of the phase defect.The width of the phase defect was defined as a full-width at half-maximum (FWHM), as shown in Fig. 3(b).The shaded areas in Fig. 3(b) show the profiles of a pit-type phase defect.The volume of the phase defect was defined as a sum of the shaded areas of all line profiles.

Phase Defect Inspection Tool
To evaluate the relationship for phase defects between size and detection intensity, their images were taken using an actinic blank inspection (ABI) high-volume manufacturing (HVM) model (Lasertec Corporation, Yokohama, Japan).Figure 4 shows a schematic model of the ABI HVM optics.EUV light is focused on the EUV blank.The scattered light due to the existence of the phase defect is collected by Schwarzschild optics with a magnification of 26× and detected by a charge-coupled device camera.The inner and outer numerical apertures (NAs) of the Schwarzschild optics are 0.1 and 0.27, respectively. 23,25g. 1 Design of the programmed phase defect mask and an arrangement of the programmed phase defects.

Simulation Condition for Defect Printability
To evaluate the impact of the phase defect on the printed image on wafer, a lithography simulator LAIPH EUV defect printability simulator (KLA-Tencor Corporation, California) was utilized.An NA of 0.33 with a dipole illumination (sigma ¼ 0.4∕0.8,pole opening angle ¼ 90 deg) was used to calculate the areal image of 22 nm L/S patterns that corresponded to an 88-nm half pitch on mask.The mask L/S patterns were set in a direction parallel to the projected incoming EUV light.The EUV mask structure was set to be the same as the prepared EUV blank structure with a 66-nmthick Ta-based absorber layer.The location of the phase defect was defined in terms of its relative distance to the center of a space pattern, as illustrated in Fig. 5(a).A calculated areal image with contour is shown in Fig. 5(b).The critical dimension (CD) was defined as the minimum space width affected by the phase defect.Figure 5(c) shows an example of the calculated result of the printed CD error as a function of phase defect position.In this work, the maximum impacts on the printed CD error were observed from the phase defects located at 0 (zero) nm in position.
3 Results and Discussions

Measurement Results of the Programmed Phase Defect Size and Volume
In this section, the measured size of the programmed phase defect using SPM is described.Each phase defect, as explained in Sec.2.1, was measured five times for its depth and height.Figure 6(a) shows the measurement result of the phase defect sizes.Each plot indicates an average value of the depth and width of the 10 phase defects.Since the blocks A, B, and C had different etching processing conditions applied, the phase defect sizes became different as was intended, even if the designed lateral sizes were the same.In the same block, the quartz etch depth decreased with decreasing lateral size.This nonlinearity of depth is mainly caused by a proportional relationship between the lateral size of the etch area and the quartz etching speed.At larger lateral sizes of more than 1 × 1 μm 2 areas, the etch depths of blocks A, B, and C reached the saturated values of 5.6, 2.9, and 2.1 nm, respectively.The average volumes of the 10 phase defects are also shown in Fig. 6(b).Apparently, each phase defect shows  different volumes within the same block.However, there are some combinations of phase defects, such as IDs A2 and B1 or IDs A4 and B3, that have different depths and widths even when their volumes happen to be almost the same.These combinations are useful in analyzing the impact of the defect size or volume on defect detection signal intensity.

Defect Detection Signal Intensity
The influence of the phase defect volume on the ABI HVM defect signal intensity (DSI) was evaluated.The image of each phase defect was acquired five times with an exposure time of 40.5 ms, and the DSIs were calculated from those images.Figure 7 shows the normalized DSI as a function of the phase defect volume obtained by SPM (see Fig. 3).As shown in Fig. 1, there are 10 phase defects for each design size.Plots represent the average value of the phase defect volume and DSI.Due to the measurement repeatability of the phase defect volume and DSI, all the plots show error bars along the horizontal and vertical axes, respectively.As shown in Fig. 6(b), there are some combinations of phase defect IDs with almost the same volume but different depths and widths.Comparing between A6 and B5, A4 and B3, B2 and C1, each combination had almost the same phase defect volume but a different DSI.From a view point of the defect shape, the higher DSI case, specifically, A6, A4, and B2, shared a common trend of deeper phase defects than their respective counterparts.This tendency was also observed at the small-size combination of A2 and B1, where the volumes were almost the same; the volume of A2 was 1.01 times larger than that of B1, but the average DSI value of A2 was 1.2 times larger than that of B1.Although the ABI HVM can detect the phase defect on the EUV blank, these results indicated that the volume of the phase defect is hard to predict with only the information of its DSI.

Influence of the Phase Defect on Wafer Printability
Although there is a trend-like relationship between the phase defect volume and the DSI, the important point is to know which measured value can be seen as reliable to predict the impact of the phase defect on a wafer printed CD.To calculate the wafer printed CD impact for various shapes or volumes of phase defects, a lithographic simulator was employed.Figure 8 shows the calculated results of the printed CD errors as a function of the phase defect positions.
The depths and FWHMs of the phase defects, shown in Fig. 6(a), were used to calculate the printed images.The phase defects of A4 and B3 had almost the same volumes of 11450 and 11329 nm 3 , respectively.However, the influences of the phase defects on the printed CDs were quite different.A4 caused more than 12 nm of CD error even though B3 caused only 8 nm of CD error.The same tendency was observed in other combinations of B2 and C1, and of A2 and B1, as shown in Figs.8(b) and 8(c), respectively.These results indicate that the wafer printability cannot be predicted only from the information of the phase defect volume.Figures 9(a) and 9(b) represent the CD error to print 22 nm L/S on wafer as functions of the phase defect volume and width, respectively.A4, A5, and A6 are not shown because these phase defects caused lines bridging defects on the wafer, as shown in Fig. 8(a).Judging the influence of different size phase defects in the same block on a printed CD, the larger phase defects caused greater CD errors.However, compared with phase defects between groups A,  B, and C, there are relatively weak correlations between the phase defect size and CD error.For example, among the phase defects in blocks A and C, the phase defect in block C was the largest in volume and width but had the least influence on the wafer printed CD.On the other hand, as shown in Figs.9(c) and 9(d), the CD errors caused by the phase defects in blocks A, B, and C were almost linear with respect to the defect depths and DSIs.Linear regression analysis showed strong correlations between the defect depth and CD error, and DSI and CD error.The coefficient of correlation values were 0.89 and 0.97, respectively, whereas the correlation factors between the phase defect volume and CD error, and the defect width and CD error were only 0.75 and 0.44, respectively.Such a strong correlation, especially between the DSI and CD error, allowed distinguishing the lithographic impacts caused by the phase defects A2 and B1 that had DSIs of 0.21 and 0.25, respectively, although the difference in the volume was negligibly small.These phase defects caused CD errors of −3.9 and −5.3 nm, respectively.As a result, the printed CD error had a strong correlation with DSI and defect depth rather than the phase defect volume and width.

Summary and Conclusion
In order to predict the printing impact on a wafer for phase defects embedded in the multilayer mirror of an EUVL mask, two types of phase defect measurement methods were applied to characterize the phase defects and calculate their impact on the printed CD on a wafer.A programmed phase defect mask was fabricated that comprised pit-type phase defects of different depths in the range of 30, 40, 50, 60, 70, 80, and 100 nm in diameter.The volumes of the phase defects were measured with SPM, and their DSIs were calculated from the images acquired by ABI HVM.It was found that the DSIs did not correlate directly with the phase defect volumes.Next, the influence of the phase defects on a wafer printed image of 22 nm L/S patterns was calculated using a lithography simulator.The calculated results indicated that the printed CD error had a strong correlation with DSI and defect depth rather than with phase defect volume and width.The correlation factors were 0.97, 0.89, 0.75, and 0.44, respectively.As a result, the influence of a phase defect on a printed CD error can be predicted from the value of its DSI.
defect mask.Blocks A, B, and C indicate the same design of the phase defects but have different etch depths as mentioned earlier.The lateral sizes of the phase defects were 30, 40, 50, 60, 70, 80, and 100 nm in diameter.The phase defects were arrayed in 5 × 2 matrices with their pitch values of 5 μm surrounded by guide patterns of 100-nm lines and spaces (L/S).

Fig. 2
Fig. 2 Phase defect imaging tool and probe: (a) photograph of L-Trace II and (b) schematic model of the cantilever and tip used.

Fig. 3
Fig. 3 Calculation methods of the phase defect size and volume: (a) a post-tilt correction SPM image and (b) line profiles of the SPM image along the lines between the arrows indicated in the image (a).

Fig. 4
Fig. 4 Schematic model of the optics of the actinic blank inspection tool.

Fig. 5
Fig. 5 Calculation method of the CD affected by the phase defect: (a) Top view of a phase defect in the 88 nm L/S pattern.The scale indicates the phase defect position relative to the center of the space pattern.(b) Calculated areal image and contour.(c) Calculated CD error as a function of the phase defect position.The target space CD on the wafer is 22 nm.

Fig. 6
Fig. 6 Summary of the obtained sizes for the programmed phase defects: (a) depth and width.(b) volume, together with average depth and width.

Fig. 7
Fig. 7 Normalized defect signal intensities as a function of the phase defect volume.IDs from A1 to C1 correspond to the same IDs in Fig. 6(b).