X-ray characterization of contact holes for block copolymer lithography

Abstract

The directed self-assembly (DSA) of block copolymers (BCPs) is a promising low-cost approach to patterning structures with critical dimensions (CDs) which are smaller than can be achieved by traditional photolithography. The CD of contact holes can be reduced by assembling a cylindrical BCP inside a patterned template and utilizing the native size of the cylinder to dictate the reduced dimensions of the hole. This is a particularly promising application of the DSA technique, but in order for this technology to be realized there is a need for three-dimensional metrology of the internal structure of the patterned BCP in order to understand how template properties and processing conditions impact BCP assembly. This is a particularly challenging problem for traditional metrologies owing to the three-dimensional nature of the structure and the buried features. By utilizing small-angle X-ray scattering and changing the angle between the incident beam and sample we can reconstruct the three-dimensional shape profile of the empty template and the residual polymer after self-assembly and removal of one of the phases. A two-dimensional square grid pattern of the holes results in scattering in both in-plane directions, which is simplified by converting to a radial geometry. The shape is then determined by simulating the scattering from a model and iterating that model until the simulated and experimental scattering profiles show a satisfactory match. Samples with two different processing conditions are characterized in order to demonstrate the ability of the technique to evaluate critical features such as residual layer thickness and sidewall height. It was found that the samples had residual layer thicknesses of 15.9 ± 3.2 nm and 4.5 ± 2.2 nm, which were clearly distinguished between the two different DSA processes and in good agreement with focused ion beam scanning transmission electron microscopy (FIBSTEM) observations. The advantage of the X-ray measurements is that FIBSTEM characterizes around ten holes, while there are of the order of 800 000 holes illuminated by the X-ray beam.

Graphical Abstract

1. Introduction

Block copolymer (BCP) lithography is being explored as a low-cost option for patterning features at length scales below what is directly achievable with current photolithography techniques (Tavakkoli et al., 2012; Ruiz et al., 2008; Stoykovich et al., 2005; Doerk et al., 2014). There are two applications of this technology which have received considerable attention: density multiplication of line-space features (Liu et al., 2011, 2013; Lane et al., 2017; Bates et al., 2014) and shrinking the critical dimensions (CDs) of contact holes (Bao et al., 2011; Tiron et al., 2013; Yi et al., 2012). Contact holes form vias for electrical interconnects between layers in integrated circuits. The hole density can be increased through multiple patterning steps, but continued reduction of their CD is a critical challenge for future technology nodes. The directed self-assembly (DSA) of block copolymers can utilize current lithographic technologies for patterning a guiding template and tailor the pitch of the hole to that of the BCP used to fill the template. Characterizing the structure of a BCP patterned inside a cylindrical contact hole is extremely challenging, even after one of the blocks has been etched away, and is typically done via cross sectioning with transmission electron microscopy (TEM). An alternative is to use X-ray measurements, which can characterize the shape of the BCP in the template on the basis of differences in the electron density. In this manuscript, we will detail the use of small-angle X-ray scattering (SAXS) with multiple incident angles (referred to by the semiconductor industry as critical-dimension small-angle X-ray scattering; Sunday, List et al., 2015; Sunday et al., 2014; Jones et al., 2006) to characterize DSA-patterned contact holes.

BCP lithography utilizes the native shape and pitch of a block copolymer to reduce the feature size in a pattern below what can be obtained through conventional high-volume manufacturing techniques. In the case of line-space patterns, a BCP with lamellar morphology is assembled on a pattern with periodicity commensurate to a multiple of the BCP pitch (Liu et al., 2013; Blachut et al., 2016; Bai et al., 2015). This approach has yielded patterned feature sizes as low as 5 nm with defect densities below 1 cm⁻² (Lane et al., 2017; Suh et al., 2017; Muramatsu et al., 2017). Contact-hole shrink with DSAutilizes a graphoepitaxy approach where the template is patterned via traditional lithographic techniques and then typically modified with a polymer brush to control the surface chemistry. A combination of the template CD, pitch of the BCP and choice of surface modification dictates the process window which leads to successful assembly (Gharbi et al., 2016; Delachat et al., 2017; Graves et al., 2015). Polystyrene-b-poly(methyl methacrylate) (PS-b-PMMA) is the polymer system most commonly utilized for this process, owing to the ease of processing and nearly equivalent surface energies which enable perpendicular orientation through thermal annealing. The guiding template is designed such that the PMMA portion of the BCP forms a cylinder at the center. The PMMA is selectively removed, leaving the PS to function as an etch mask for transfer to the underlying layers (Freychet et al., 2016). The ratio between the template CD and BCP pitch which yields good assembly shifts depending on the choice of surface modification. PS-modified surfaces yield a window of good assembly when the template CD is of the order of the BCP pitch, and PMMA-modified surfaces have a window of good assembly at template CDs of the order of twice the BCP pitch (Gharbi et al., 2016). The assembly quality is generally monitored via top-down scanning electron microscopy (SEM) images after PMMA removal, which can determine CD uniformity and the fraction of holes which form openings at the surface but provides limited information on the depth profile. As a result, sub-surface defects and process variations can go undetected. Simulations have suggested that the internal structure of the BCP assembled in the template can be complex, including variation of the depth of the PMMA cylinder and discontinuities of the PMMA phase (Iwama, Laachi, Kim et al., 2015; Laachi et al., 2015). The extent to which these morphologies occur has not yet been explored experimentally, and it is not fully understood how those morphologies will impact the etch process. One of the additional challenges for DSA contact shrink is that the guiding templates are not necessarily uniformly distributed over the wafer surface, and this has led to difficulty obtaining uniform features in regions with heterogenous template density. This is primarily due to mismatch of over- versus underfilling of the polymer in the template, which complicates the etch process. Recently it has been shown that by significantly overfilling, to the point where the BCP surface is flat across the wafer, these issues can be minimized and a variety of contact-hole densities can be handled on the same sample (Barros et al., 2017). One of the outstanding challenges for DSA in contact holes is density multiplication within a single template, which would provide cost-efficient patterning of extremely dense features. Initial demonstrations have proven the potential of this approach, although controlling the position and uniformity of the cylinders in those cases remains a considerable challenge (Iwama, Laachi, Kim et al., 2015; Iwama, Laachi, Delaney et al., 2015; Graves et al., 2015; Gharbi et al., 2016).

The ultimate goal of the hole-shrink process is to transfer the DSA contact-hole pattern into the underlying layers. For this approach to be successful the residual height of polymer at the bottom of the template must be minimized while retaining sufficient polymer around the edge of the template to act as an etch mask. While these features are generally characterized via SEM and TEM cross sections, X-rays can interrogate this type of buried structure without damaging the sample. X-ray measurements can be done with either transmission or grazing scattering geometries. Grazing-incidence small-angle X-ray scattering (GISAXS) is implemented in a reflection geometry where the incident beam is near the critical edge of the sample. As a result, the beam footprint is large and measurement times are generally fast; the drawback is that data analysis is more difficult than in a transmission geometry owing to multiple scattering effects. There are several efforts underway to develop improved methods of dealing with this challenge (Babonneau, 2010; Chourou et al., 2013). GISAXS has been used to characterize several different types of periodic nanostructures (Suh et al., 2016; Soccio et al., 2014; Hlaing et al., 2011) and thin-film BCPs, where it is particularly useful for evaluating BCP orientation (Kim et al., 2010, 2009). In a similar system to the contact holes, GISAXS was used to monitor the structure of vertically aligned PMMA cylinders under several different etching conditions (Freychet et al., 2016). Transmission scattering measurements offer simpler data analysis and the ability to measure smaller sample sizes, but do require longer measurement times compared to the grazing geometry. Reconstruction of three-dimensional profiles using SAXS works by rotating the sample in the incident beam to obtain information on the vertical composition profile of the periodic structure; greater rotation away from normal incidence results in more information about the out-of-plane structure compared to the in-plane structure. Scattering from the full range of collected incident angles is compiled in a reciprocal-space map, which is analyzed using an inverse-iterative approach to characterize the shape of the measured nanostructure. Line-space features have been investigated most frequently. In one example, a multiple-patterning approach was used to produce a line grating with asymmetric, ‘shark-fin’ like features and the measurement was able to characterize both the line shape and the interline spacing to sub-nanometre accuracy (Sunday, List et al., 2015; Sunday et al., 2016). SAXS has also been applied to characterize the buried structure inside line-space BCPs patterned on a chemical template by using low-energy soft X-rays to enhance the scattering contrast (Sunday et al., 2014, 2017) and to measure the line shape after etch transfer (Sunday, Ashley et al., 2015).

In order to demonstrate the ability of X-ray measurements to aid in the development of structure-processing relationships for DSA contact-hole-shrink samples, variable-angle transmission SAXS measurements were performed on two different sets of samples which varied by both template diameter and processing conditions. This is a well established technique for one-dimensional gratings. This work will extend the technique to two-dimensional arrays of cones. DSA-patterned contact holes and guiding templates with CDs within the known process window for good assembly conditions were prepared. The DSA samples were processed using different surface affinity treatments and different etching times, as detailed in previous work (Barros et al., 2017; Gharbi et al., 2016). Both the empty template and the result of the DSA process after removal of the PMMA were characterized. This result demonstrates the ability of X-ray scattering methods to accurately evaluate key parameters such as residual layer thickness and polymer fill height.

2. Materials and methods

2.1. Sample preparation

The DSA hole-shrink process is described in detail in several other publications and will be overviewed here only briefly (Barros et al., 2017; Gharbi et al., 2016). PS-b-PMMA with a period of 35 nm and mass ratio of 70:30 was purchased from Arkema. The samples were prepared on the 300 mm DSA pilot line at the Laboratory of Electronics, Technology and Instrumentation (Grenoble, France) using a process described in Fig. 1. 193 nm immersion lithography is used to pattern the initial template, after which a dry plasma etch is used to remove the anti-reflective coating (ARC), spin on carbon (SOC) and residual photoresist. The surface of the template is then modified by either a PS or PMMA brush. PS-b-PMMA is subsequently spin coated at a thickness that overfills the template holes and results in a smooth top surface. The resulting sample is annealed at 523 K for 5 min, and annealing is followed by removal of excess polymer and PMMA inside the template using an O₂-based dry plasma etch.

Figure 1.

Process for preparing 1:1 density-multiplication DSA in cylindrical templates. The initial template is patterned using 193 nm lithography, followed by a dry plasma etching to remove the ARC/SOC. The surface of the template is then modified with either a PS or PMMA brush. This is followed by spin coating with a BCP film and thermal annealing. A dry plasma etch is used to remove excess polymer and PMMA from within the template, leaving only the PS layer.

2.2. SAXS measurements

Measurements were conducted at the 5ID beamline at the Advanced Photon Source at Argonne National Laboratory. The incident beam had an energy of 17 keV and a spot size of 70 × 120 μm. A CCD detector at a sample–detector distance of 8.5 m with a pixel size of 88.6 μm was used to collect the scattered beam (Weigand & Keane, 2011). The measurement geometry is shown in Fig. 2(a), and the corresponding scattering vectors along the BCP are designated in Fig. 2(b). The sample was rotated from −60 to 60° in 1° steps. Between 30 and −30° the sample was exposed for 10 s (empty templates) or 30 s (filled BCP) and between ±60 and ±30 the sample was exposed for 20 s (empty template) or 60 s (filled BCP) to improve the signal–noise ratio. Given that the cylinders are symmetric, the geometry can be simplified by converting to a cylindrical coordinate system, which uses q_R in the sample plane and maintains q_z as the out-of-plane vector. Conversion of q_y and q_x to q_R is done according to equations (1) and (2):

$$\alpha =\arctan \left(q_y/q_x\right),$$ (1)

$$q_R=q_x/\cos (\alpha ).$$ (2)

An example of how the individual images are recombined into a set of reciprocal-space maps is shown in Fig. 3. The basic structure used to approximate the shape of the empty template or the BCP is a conic section, which is described by an upper (R₂) and lower radius (R₁) and a height (Z₂ − Z₁). The definitions of the real- and reciprocal-space vectors for these structures are described in Fig. 2. The scattering intensity [I(q)] is calculated according to equations (3) and (4):

$$I_{\text{o}}(\mathbf{q})={\left|\underset{V}{\int }\rho (\mathbf{r})\ast \sum _{n_x}\sum _{n_y}\delta \left(x-n_{x}P,y-n_{y}P\right)\exp (-i\mathbf{q}\cdot \mathbf{r})\text{d}\mathbf{r}\right|}^2,$$ (3)

$$I(\mathbf{q})=I_{\text{o}}(\mathbf{q})\exp \left(-q^2{\text{DW}}^2\right),$$ (4)

Where ρ(r) is the shape function, which describes the scattering length density distribution, the summation represents the structure factor with n_x and n_y referencing individual lattice positions, * represents a convolution, P is the pitch, and DW is the Debye–Waller factor, which accounts for the decay in scattering intensity caused by edge roughness. The DW factor is generally used to account for thermally induced positional disorder of atoms. It assumes the thermal vibrations result in a Gaussian distribution centered around the ideal atomic lattice position with motions of individual atoms being independent (Guiner, 1963). This time-averaged position distribution results in a decay in the intensity as a function of scattering vector. In our case we have process-induced roughness which results in a variation in the edge position around the average contact-hole position. This static distribution can be assumed to be Gaussian, and the resulting scattering intensity decay can be modeled using the DW factor to extract the root-mean-square edge-displacement error. Any deviations in the position distribution from Gaussian will have a negligible effect on the DW factor extracted. The scattering peaks are very sharp, so contributions from paracrystallinity are small. Equation (3) contains both the contributions from the form factor, whose calculation is described in detail in equation (5), and the structure factor, which is described by the summation of the delta functions. The Fourier transform of a conic section is described by equation (5):

$$\begin{gathered} F_{\text{Cyl}}\left(q_z,q_R\right)=\underset{V}{\int }\rho (\mathbf{r})\exp (-i\mathbf{q}\cdot \mathbf{r})\text{d}\mathbf{r} \\ =\underset{z_1}{\overset{z_2}{\int }}\frac{2\pi }{q_R}\left(\frac{Z-Z_1}{m}+R_1\right)J_1 \\ \times \left[q_R\left(\frac{Z-Z_1}{m}+R_1\right)\right]\exp \left(i{q}_{z}Z\right)\text{d}z, \end{gathered}$$ (5)

where m represents the sidewall slope of a given section and J₁ is a Bessel function of the first kind. This equation cannot be solved analytically and must be integrated via numerical methods. In this case the integration was carried out with the trapezoid rule, discretizing each step such that it was no larger than 2 nm. The amplitudes calculated from each of the conic sections are coherently combined prior to calculating the scattering intensity.

Figure 2.

(a) Measurement geometry, including definitions of the reciprocal- (q) and real-space (x, y, z) sample coordinates. (b) Definition of real-space (x, y, z) coordinates and model parameters for the conic sections which were used to odel the contact hole and BCP shape.

Figure 3.

(a) Normal-incidence scattering from T-50. The peak orders along the q_xz axis are labeled in white numerals in the upper-right-hand corner, and the peak orders on the q_y axis are labeled at the top of the image. The rotation axis in indicated on the left-hand side of the image. (b) Distribution of the scattering intensity for the y0 peaks after conversion of the full range of scattering measurements (60 to −60°) to q_R space. The distribution of data from this peak set is equivalent to previous work with one-dimensional gratings. (c) Addition of the y1 peak orders to the reciprocal-space map, demonstrating their position relative to the y0 peaks. Beyond the fourth order the positions of the two peaks are nearly identical. Both (b) and (c) show only the position of the data in reciprocal space, not the intensity.

The experimental data are fitted using an inverse approach. The scattering from a structural model is calculated using equations (3)–(5) and then compared with the experimental fit and iterated until the best possible fit is achieved. The fit is optimized using a Monte Carlo Markov chain (MCMC) algorithm. The sampling process in the MCMC algorithm is also used to provide a model library weighted by fit quality, which can be used to evaluate the uncertainty. Briefly, the model is initialized with a given parameter set, random perturbations to these parameters are generated, and the new model is either accepted into the population or rejected on the basis of the change in the goodness of fit. Details of the algorithm, such as specific acceptance conditions, are presented elsewhere (Mosegaard & Sambridge, 2002; Sunday et al., 2016; Hannon et al., 2016). Given that the inverse fitting methods can produce non-unique solutions, large numbers of initial conditions were tested to ensure that no alternative structures produced simulated scattering profiles with equivalent fit quality. In all cases there were no other structures which resulted in comparable fits.

3. Results and discussion

Measurements were conducted on two different types of DSA contact-hole patterns and their corresponding guiding templates: a PS-modified template (center diameter of 50 nm) and a PMMA-modified template (center diameter of 65 nm). These will be described by the following nomenclature: T,P-center diameter (Trefers to an empty template, P refers to the filled sample after PMMA removal), e.g. T-50 for the empty template with a 50 nm CD which will be modified with a PS affine brush. A false-color image of the normal-incidence scattering from T-50 is shown in Fig. 3(a). Scattering occurs at uniform intervals along both the q_xz and q_y directions with Δq_xz, Δq_y ≃ 0.052 nm⁻¹, which converts to a pitch of ~121 nm. The peak orders along the q_xz axis are defined by the numerals on the upper-right-hand side of the image. The zeroth orders are excluded as they were not utilized in the data analysis. The peak orders along the q_y direction are labeled along the top of the image. The y0 peaks are equivalent to the peaks that are obtained when scattering from a one-dimensional grating. The distribution of those peaks after conversion of the full range of rotation measurements to q_R space is shown in Fig. 3(b). The spacing between the peak orders in the q_R direction is uniform; they do not have any contribution from the q_y component. The contributions from the y1 peaks are shown in Fig. 3(c), where because of the addition of the q_y component to the scattering vector the distribution along q_R is no longer uniform. As a result, the low-order y1 q_R positions vary from the y0 peaks, filling in more of the reciprocal-space map than was covered in Fig. 3(b). Beyond the fourth-order peak the positions of the y0 and y1 peak orders are nearly equivalent. More space in the I(q_z, q_R) map can be filled with the addition of higher-order y series peaks.

The fits to the experimental data for T-50 including y peak orders y0–y3 are shown in Fig. 4(a), with the structure corresponding to the best fit shown in Fig. 4(b). Five conic sections were used to model the structure, three for the SOC and two for the ARC; the progression in the goodness of fit for the structure as a function of the number of conic sections is shown in the supporting information (Fig. S1). A single conic section describes the majority of the etched portion of the SOC, from the base to the template to a height of ~123 nm. The remaining sections are needed to describe the curvature near the top of the template arising from the difference in the etch contrast between the ARC and the SOC, which results in the overhang of the ARC over the template. The width at the base of the template was found to be 43.4 ± 0.1 nm, and that at the top to be 87.1 ± 1.7 nm, while the overall height of the template was 152.2 ± 0.4 nm. The larger uncertainty in the width near the top of the structure resulted from the small volume of the conic section used to fit that portion of the structure relative to the large volume of the conic section which describes the base of the structure.

Figure 4.

(a) Simulated fit (solid line) to experimental data (colored circles) from T-50 for peak orders y0–y3; intensities are arbitrarily scaled for visual clarity. (b) Structure of best fit to T-50, showing two unit cells. The white space represents the empty template, the lighter gray the SOC and the darker gray the ARC. The best fit to the bottom and top width is shown in the left unit cell. Lines indicating the interface between the individual conic sections are shown in the right unit cell. Dashed lines represent 95% confidence intervals in the shape as determined by the MCMC algorithm. (c) Base, middle and top diameters as a function of which y peak orders were included in the optimization. Error bars indicate 95% confidence intervals. (d) Total height of the template as a function of which y peak orders are included in the analysis. Error bars indicate 95% confidence intervals.

The addition of the data from the y1+ peak orders is potentially beneficial as it fills additional space on the I(q_z, q_R) map, but there is a tradeoff with the additional computation time needed to analyze the results, particularly given the lack of an analytical solution for the Fourier transform of a conic section. Therefore, the impact of the higher-order y peaks was analyzed by evaluating the change in the fit and uncertainties upon the addition of each peak set to the data analysis, i.e. first the structure was fitted using only the y0 peaks, and then the y1 set was added in and analyzed up to the y3 set. The results of this analysis for the template diameter at the base, middle and top of the structure, and for the overall height, are shown in Figs. 4(c) and 4(d), respectively. The addition of the y1 peaks does result in a small reduction in the uncertainty for all parameters; for example, with only the y0 peaks the diameter at the base of the structure was found to be 43.4 ± 0.2 nm and it is reduced to 43.4 ± 0.1 nm with addition of the y1 series. Addition of further data sets results in an even smaller reduction in the uncertainty. A compromise approach which could provide additional data at reduced computation cost would involve taking only those higher-order peaks which are sufficiently differentiated in q_R space from the initial data series.

An identical sample to T-50 was used to direct the assembly of PS-b-PMMA. After removal of the PMMA portion the sample was measured and fitted using the same process as described for the empty template. This sample is labeled P-50. The resulting scattering from the y0 and y1 peak orders is shown in Fig. 5(a). In contrast to the scattering from the empty template, which had relatively well defined fringes in the q_z direction, many of the fringe minima are significantly less well resolved. This is due to the added complexity in the structure from the polymer fill. In order to reduce the potential for multiple solutions the results of the fit to T-50 were used to provide an initial condition for the empty template, although the template structure was allowed to vary slightly to account for changes during the annealing/etching portion of the process. Three conic sections were used to model the polymer fill inside the template: one to describe the majority of the polymer between a height of ~16 and −105 nm, the other two to describe the residual polymer near the top of the SOC. For the polymer to act as an etch mask for transfer to the underlying substrate the residual height (RH) at the bottom of the template needs to be minimized and the sidewall height (SH) of the polymer on the sidewalls needs to be maximized. For this sample RH was found to be 15.9 ± 3.2 nm and SH was found to be 106.5 ± 2.1 nm. There remain small deviations between the simulated and experimental intensities beyond the fourth-order peak in q_R. The specific origin of the discrepancy between the simulated and experimental data is unclear, but additional conic sections to the polymer model do not improve the fit. Additionally, these more complex models reduce to a nearly equivalent shape to the model with only three sections, suggesting that the three-cone model succeeds in capturing the major features of the target structure. A comparison with a TEM cross section (Fig. 6) obtained from an identical sample shows that the key features of the structure are captured in the scattering model. The same general features are all observed, including the presence of a significant residual layer, polymer fill near to the top of the SOC and the overhang around the ARC/SOC interface. Owing to potential artifacts from the TEM sample preparation and statistical hole-to-hole variations, the TEM results are not quantitatively compared with the X-ray characterization. They should be viewed as a qualitative guide to the features which are expected to be observed.

Figure 5.

(a) Fits to the y0 and y1 peak orders for P-50, showing both experimental data (colored circles) and simulated scattering from the best fit (solid lines). (b) Best fit structure to the data shown in part (a). The empty template fit for T-50 (shown in Fig. 4) was used to seed the initial condition for the template structure. The polymer fill used a model which had three conic sections. RH indicates the height of the residual layer and SH indicates the sidewall height of the polymer remaining on the sidewalls.

Figure 6.

Cross section TEM image of a sample prepared under identical conditions to P-50, showing the presence of a residual layer at the bottom of the template.

In addition to the sample prepared with the PS affine process, a sample prepared via the PMMA affine process, using a wider template diameter, was investigated. Scattering from the empty template was measured and fitted. The results are shown in the supporting information (Fig. S2). Scattering from P-65 is shown in Fig. 7(a), for both the y0 and y1 peak orders. The initial structure of the template was based on the scattering from T-65, and two conic sections were used to describe the BCP fill. The RH here was found to be 4.5 ± 2.2 nm, and the SH was 84.3 ±1.6 nm. The clear difference in the RH and SH determined by the SAXS method for these two processes demonstrates the ability to utilize X-ray-based metrology to interrogate changes in these critical process parameters.

Figure 7.

(a) Experimental (colored circles) and simulated scattering (black line) from the best fit to the y0 and y1 peak orders for P-65. (b) Best fit structure to the experimental data shown in part (a), RH indicates the height of the residual layer and SH indicates the sidewall height of the BCP at the edge of the template. Corresponding scattering from the empty template is shown in the supporting information.

4. Conclusions

Contact-hole shrink through DSA is a potential approach for patterning smaller CD contact holes for next-generation technology nodes. Characterization of the internal structure of DSA contact holes under different process conditions is critical for understanding how to minimize defects and optimize etch transfer. In this manuscript, we demonstrate that transmission X-ray scattering is an effective approach for characterizing the three-dimensional shape of DSA BCPs patterned in graphoepitaxial templates. The square array pattern and cylindrical shape of the holes enable additional space to be filled in a reciprocal-space map beyond what can be achieved with scattering from a sample with periodicity along a single dimension, providing additional data that can be used to evaluate uniqueness and reduce uncertainty. Two samples were prepared under different processing conditions and characterized to determine the resulting DSA-patterned hole shape. We observed significant differences in the polymer residual height at the bottom of the template and sidewall polymer thickness between the two different processes. These results are consistent with FIBSTEM observations. The residual layer is a particularly difficult feature to measure, and our results demonstrate the utility of X-ray metrology in characterizing the DSA process for contact-hole fabrication.