2.1.

Samples

The samples used for the STEM measurements included 45-nm half-pitch lines and spaces of ArF photoresists. We measured three samples with different material combinations, as shown in Fig. 1. They are positive tone developed (PTD) PR lines on a ${\mathrm{SiO}}_2$ underlying layer, PTD PR lines on a bottom antireflective coating (BARC) underlying layer, and negative tone developed (NTD) PR lines on a ${\mathrm{SiO}}_2$ underlying layer. The ${\mathrm{SiO}}_2$ layers were spin-on-glass (SOG) layers which were fabricated by spin-coating of the mixture of ${\mathrm{SiO}}_2$ and organic solvent. Hereafter, these three samples are referred to as the PTD/SOG sample, PTD/BARC sample, and NTD/SOG sample.

Fig. 1

Examples of top-down scanning electron microscope (SEM) images of (a) positive tone developed (PTD) photoresist (PR) lines on spin-on-glass (SOG) underlying layer (PTD/SOG sample), (b) PTD PR lines on bottom antireflective coating underlying layer (PTD/BARC sample), and (c) negative tone developed PR lines on SOG underlying layer (NTD/SOG sample).

The elastic deformation of the PR patterns associated with BARC shrinkage can be examined because SOG is more rigid than BARC and is expected to be more durable against EB irradiation. In addition, since the pristine profile would be more angular for the NTD/SOG sample than for the PTD/SOG sample, the elastic deformation effect would be different between these two samples.

2.2.

Scanning Transmission Electron Microscope Measurements

The process flow of the proposed cross-sectional measurement on EB induced PR-pattern shrinkage is shown in Fig. 2. First, a SEM is used to irradiate the EB on part of the PR line patterns to intentionally cause shrinkage. This SEM EB irradiation is denoted simply as “EB irradiation” hereafter. This does not mean the EB irradiation during the STEM observation performed afterward. Second, ALD is used to deposit a 2-nm thick ${\mathrm{HfO}}_2$ thin film on the sample. The sample surface is thus conformally coated with a ${\mathrm{HfO}}_2$ thin film. In the process, tetrakis(ethlmetylamino)hafnium and ${\mathrm{H}}_2\mathrm{O}$ were used as precursors, the number of cycles was 14 and the process temperature was 80°C. Third, the sample is coated with amorphous carbon and tungsten layers to prevent pattern collapse during the following focused ion beam (FIB) process. Also during the process, the temperature was kept below 80°C in order to avoid thermal damage to the PR patterns. Fourth, a thin-film sample for the STEM observation is cut out by using a microsampling technique with a FIB. Finally, a cross-sectional Z-contrast image of the patterns is obtained by using a STEM with a 200-kV accelerating voltage. The ${\mathrm{HfO}}_2$ thin film improves the Z-contrast at the pattern edge. As a result, the cross-sectional feature with a bright thin band, which corresponds to the PR pattern surface, is clearly shown in the image. The contour of the pattern is defined as the center of the bright thin band to further quantitatively evaluate the pattern shape.

Fig. 2

Process flow of cross-sectional measurement: (a) electron-beam (EB) irradiation on part of PR sample using SEM; (b) ${\mathrm{HfO}}_2$ thin-film deposition; (c) coating with amorphous carbon and tungsten; (d) microsampling using focused ion beam; and (e) scanning transmission electron microscope (STEM) observation.

We confirmed that the above described sample preparation itself barely influences the measurement accuracy of the resist pattern profiles. An isolated PR line was measured using 3-D AFM²² before and after the ALD process. Figure 3(a) shows the obtained AFM cross-sectional profiles. The bold and dashed lines are the profiles obtained before and after the ${\mathrm{HfO}}_2$ ALD process. Figure 3(b) shows the distance, or residuals, between these two profiles, which was plotted as a function of the position measured along the profile. The distance is about 2 nm except for the pattern’s foot region. This indicates that the 2-nm ${\mathrm{HfO}}_2$ film was conformally deposited onto the resist surface and that the pattern shape was conserved during the ALD process. Dispersion of the distance, which is the nonuniformity of the ALD thickness, is larger in the pattern’s foot region, but is 2 nm at most. Considering that the profile is defined at the center of the bright band of the ${\mathrm{HfO}}_2$ film, we can thus conclude that the profile error induced in the ALD process is at most 1 nm. Moreover, the obtained images do not change during a continuous STEM observation for more than 10 min. These results suggest the EB-induced damage formed during the STEM observation is negligible. This seems against the expectation that irradiation with the higher acceleration voltage causes the more serious shrinkage. It can be considered as an effect of a thin-film specimen. Qualitatively, such high-energy electrons have a rather small scattering cross section so that they possibly lead to almost no interaction event in a very thin film. For a quantitative confirmation, detailed experiments and simulations are needed that consider a film much thinner than a mean free path of the electrons.

Fig. 3

(a) Profiles measured by three-dimensional atomic force microscope before and after atomic-layer deposition (ALD) process. (b) Distance between profiles as function of position on profile before ALD process measured along the profile. The distance is almost constant, which confirms that the ALD film was conformally deposited and the pattern shape was conserved during the ALD process. The profile error due to the ALD process was 1 nm at most because the center of the ALD film was extracted as a profile.

An averaging approach was adopted to reduce the local pattern-shape variation in order to evaluate the change in the cross-sectional shape due to shrinkage. Since a STEM measurement is a destructive one, it is impossible to compare the cross-sectional profiles of the same pattern before and after the shrinkage. The shrinkage is instead evaluated from a comparison of the results from different patterns in a same matrix of lines and spaces, effectively averaging the results of multiple but nominally equivalent cross-sections to suppress error due to roughness and finite local CD uniformity of the features under study. The procedure for averaging the profiles is shown in Fig. 4. A continuous contour was extracted from the bright thin band in the Z-contrast STEM image. The one contour obtained, which consists of multiple line pattern profiles, is divided into discrete profiles that correspond to each line. Then the profiles are aligned, folded and averaged.

Fig. 4

Procedure to evaluate averaged profile: (a) acquisition of Z-contrast image using STEM; (b) extraction of continuous contour; (c) averaging obtained multiple profiles in contour.

2.3.

Electron-Beam Irradiation Conditions

We performed three series of STEM measurements under different SEM EB irradiation conditions to control the shrinkage shown in Fig. 2(a). A summary of the conditions is given in Table 1.

Table 1

Scanning electron microscope electron-beam irradiation conditions.

The first series was the measurement using dose-dependent irradiations. All three samples were evaluated in this examination. The EB accelerating voltage was 500 V, the irradiation area was $675{\mathrm{nm}}^2$, the number of pixels was 512 pix, and the dose density was varied from 0 to $3.2\mathrm{mC}/{\mathrm{cm}}^2$. This dose density of $3.2\mathrm{mC}/{\mathrm{cm}}^2$ corresponds to 64 times the SEM image acquisitions using an 8-pA probe current.

The second series was the measurement using area-restricted irradiations. The PTD/SOG sample was used and the EB irradiation area was restricted to a very narrow region at the center of the line patterns or spaces, as shown in the painted part of Fig. 5. Each narrow scan area was 22-nm wide and 675-nm long. The dose density was fixed at $1.6\mathrm{mC}/{\mathrm{cm}}^2$. This measurement was taken to clarify the effect of the scattered electrons from the spaces around the patterns.

Fig. 5

Schematic view of area-restricted EB-irradiation, where painted regions are irradiated areas. The lower diagram is a cross-section of the PR sample and the irradiated area.

The last series was the measurement using accelerating voltage dependent irradiations. The PTD/SOG sample was evaluated. The EB accelerating voltages were 100 and 500 V, and the dose densities were 0.2 and 1.6 nm. The irradiation area was fixed at 675 nm square. This measurement was taken to clarify the effect of the electron’s penetration depth, which depends on the accelerating voltage.

2.4.

Electron Scattering Simulation

Electron scattering simulations were conducted to obtain the spatial distribution of the energy transferred from the incident electrons to the PR patterns. A Monte Carlo simulator, Chariot,²³ was used to calculate the trajectories and the energy loss of the incident electrons. In the Chariot simulations, the EB conditions were set to be equivalent to those of the experiments of area-restricted irradiations. The accelerating voltage of the incident electrons was 500 V. The irradiated region was 22-nm wide and located at the center of the lines or the center of the spaces. The incident dose was $1.6\mathrm{mC}/{\mathrm{cm}}^2$. The Mott model and the discrete loose approximation model were selected for the models describing elastic and inelastic scattering. Regarding the materials, polymethyl methacrylate (${\mathrm{C}}_5{\mathrm{H}}_8{\mathrm{O}}_2$) and ${\mathrm{SiO}}_2$ were used to represent the PR and SOG.

3.1.

Result of Dose-Dependent Irradiations

Figure 6 shows cross-sectional STEM images for various EB preirradiation doses for all three samples. Each image contains the region with and without EB irradiation. These images show that the change in the cross-sectional shape due to the EB irradiation was clearly captured. The average cross-sectional profiles extracted from these STEM images are shown in Fig. 7, where the height origin (0 nm) is set to the bottom level of the nonirradiated region. This figure shows the detailed profile change, including the rapid shrinkage for a small dose, and how it slows down for a large dose, indicating the expected saturating tendency of the shrinkage (as observed in typical CD-SEM shrinkage measurements).

Fig. 6

Examples of acquired cross-sectional STEM images of: (a) PTD/SOG sample, (b) PTD/BARC sample, and (c) NTD/SPG sample.

Fig. 7

Extracted profiles of: (a) PTD/SOG sample, (b) PTD/BARC sample, and (C) NTD/SPG sample. Three interesting features associated with PR shrinkage were identified: the rounding of the top of the patterns, the regression of the BARC layer, and a necking profile in the pattern shape, namely, the maximum sidewall recession at the middle height.

Three interesting features associated with the PR shrinkage can be identified from these results: the rounding of the top of the patterns, the regression of the BARC layer, and the necking profile in the pattern shape, namely, the maximum sidewall recession at the middle height.

The top rounding feature was commonly observed for all three samples. The pattern top was similarly round-shaped after sufficient shrinkage, even though the NTD/SOG and PTD/SOG samples have quite different pristine cross-sectional profiles. The comparison of the results of these two samples clearly showed that the protuberant region, which is the pattern center for the PTD/SOG sample and the pattern’s shoulder for the NTD/SOG sample, retreated due to the compressive stress induced by the shrinkage. This is evidence of the elastic effects in shrinkage.

The results from the PTD/BARC sample served as further evidence of the elastic effects. This sample showed a rounding of the foot region associated with the regression of the BARC layer. This behavior can be interpreted as follows. The foot region of the line patterns was pulled toward the outside oblique direction by the tensile stress induced by the underlying material in the spaces.

The necking profile is rather complicated to interpret. This feature cannot be described in terms of elasticity. This issue will be discussed in Sec. 3.2 by referring to the results from the area-restricted irradiations and the Monte Carlo simulations.

3.2.

Area-Restricted Irradiations

Figure 8 shows the cross-sectional STEM images and cross-sectional profiles of the PTD/SOG sample with the area-restricted EB irradiations. Figures 8(a) and 8(c) clearly show that the EB irradiation on the top of the PR line patterns causes a significant decrease in height and that there is no necking profile. On the other hand, Figs. 8(b) and 8(d) show that the EB irradiation only on the space region causes less height shrinkage and measurable sidewall recession and a necking profile. These contrasting behaviors suggest that the shrinkage processes in the pattern top and sidewall are different. The pattern-top shrinkage is mainly caused by the direct irradiation onto the pattern top. On the other hand, the sidewall shrinkage, including the necking profile, at least to some extent is caused by the electrons scattered from the spaces. This observation is consistent with the models presented elsewhere.¹⁷

Fig. 8

(a) and (b) Acquired cross-sectional STEM images. The EB-preirradiated areas are indicated by arrows. (c) and (d) Extracted profiles for EB irradiation only on tops of patterns and only on spaces between patterns, respectively. The dashed lines are the pristine profiles. The EB irradiation on the top region causes a significant decrease in the height. On the other hand, the EB irradiation on the spaces causes a measureable sidewall shrinkage recession and a necking of the profile.

Figure 9 shows further evidence of the impact of the scattered electrons on the shrinkage phenomena. This is a STEM image of the PTD/SOG sample with the EB irradiated onto one side of the adjacent spaces, as shown by the arrows in the figure. The PR line patterns look as if they are leaning toward the irradiated side. In other words, the profile of the irradiated side shows a necking profile. This result clearly indicates that the sidewalls’ shrinkage and the necking profile stem from the scattered electrons.

Fig. 9

Acquired cross-sectional STEM image of area-restricted irradiations. The EB-irradiated areas are indicated by the arrows. The PR line patterns look as if they are leaning toward the irradiated side due to the shrinkage.

Figures 10(a) and 10(b) show the results of the electron scattering simulations for the area-restricted EB irradiation. The calculated energy distribution deposited on the PR lines was mapped using the experimentally obtained profiles before and after EB irradiation. In this calculation, all of the electrons were taken into account, including the secondary electrons (SEs) and back-scattered electrons (BSEs) from the spaces around the patterns. For the irradiation on the top, almost all the incident energy is deposited on the vicinity of the irradiated area at the top. Meanwhile, for the irradiation on the spaces, measurable energy is deposited on the sidewall via the scattered electrons. These results indicate that the height shrinkage in the SEM observation is mainly caused by the direct irradiation on the top of the pattern and that the electrons scattered from the spaces onto the sidewall contribute to the sidewall shrinkage. Quantitatively, the energy deposition density per unit surface area of the sidewall is about 5% of that on the top region in the case of irradiation on the top.

Fig. 10

Calculated distributions of energy deposition caused by EB irradiations on: (a) top of PR lines and (b) spaces between lines. The solid lines are the measured pristine profiles and the dashed lines are the measured profiles after shrinkage. (c) Height dependence of total energy deposition density on sidewall for EB irradiation on spaces.

Figure 10(c) shows the height dependence of the energy-deposition density per unit surface of the sidewall for the irradiation on the spaces. The region with large energy deposition was found to be coincident with the necking location region (which was observed at around 20 nm above the bottom). It is, therefore, confirmed that the electrons scattered from the spaces cause the necking profile.

3.3.

Electron-Beam Accelerating Voltage Dependent Irradiations

Figure 11 shows the averaged cross-sectional profiles of the PTD/SOG sample after EB irradiation using 100- and 500-V accelerating voltages with small and large doses.

Fig. 11

Extracted profiles of the PTD/SOG sample with 100 V and 500 V preirradiation (solid lines) shown together with the profiles without EB preirradiation (dashed lines): (a) small irradiation case and (b) large irradiation case. The difference of the shrinkage between 100 V case and 500 V is specifically small for the sidewall at large dose. It suggests that major cause of the sidewall shrinkage is the SEs emitted from the spaces because penetration depth of SEs is irrespective to the energy of incident electrons in contrast to that of back-scattered electrons.

The decrease in height and the sidewall recession are much smaller for the 100-V EB-irradiation than for the 500-V EB-irradiation for a small dose, as shown in Fig 11(a). This is well expected from a naive description that states that an EB with a higher energy causes larger shrinkage. For a large dose, however, the sidewall recession is almost equivalent at both 100 and 500 V, although the decrease in height is still smaller for 100 V.

This anomalous behavior can be interpreted by considering the difference in nature of the electrons inducing the shrinkage. Discussion in Sec. 2.2 revealed that the sidewall recession is mainly induced by the scattered electrons from the spaces, as reported in the literature.¹⁷ In the literature, it was regarded that the BSEs are the major cause of the sidewall recession because the energy of BSEs is much higher than the SEs. However, the observed equivalent sidewall recession does not meet this interpretation because the energy of BSEs is higher and the expected shrinkage is larger for the 500-V EB irradiation than for the 100-V EB-irradiation. On the other hand, the observed phenomenon can be interpreted by considering the SEs as the major cause of sidewall recession. The shrinkage amount should be controlled by the penetration depth of the incident electrons with an EB dose large enough for the saturation of shrinkage. Because the typical energy of SEs is universally several tens of eV, the typical penetration depth of SEs is generally irrespective of the incident EB energy. Thus, a similar amount of sidewall recession between the 100-V and the 500-V EB irradiations can be explained by the effect of SEs emitted from the spaces.

Based on this interpretation, the result of the sidewall recession for a small dose implies that the dose of the SEs emitted from the spaces is larger for the 500-V case. This means that the SE yield of the SOG is larger for a 500-V EB irradiation than for a 100 V one. Although the SE yield of ${\mathrm{SiO}}_2$ is difficult to determine by experiment due to charging, a simulation study which eliminates charging effect reported that the yield is larger for 500 V than for 100 V.²⁴ Detailed investigations on the SE yield of ${\mathrm{SiO}}_2$ are needed for a comprehensive interpretation of the observed accelerating-voltage dependency.