2.1.

Preparation of Substrates

Silicon wafers were purchased from WRS materials. Wafers were hot piranha cleaned (DANGER!) utilizing 70:30 sulfuric acid:hydrogen peroxide (30%) at 130°C. The wafers were placed into the piranha bath for a minimum of 5 min before being cleaned with house DI for several minutes to create a pristine ultrahydrophilic surface. Wafers were used within days of cleaning to prevent reabsorption of organic material. Metal substrates were deposited at a rate of one angstrom per second onto cleaned wafers using the following methodologies: Copper substrates were prepared by thermal evaporation utilizing an Angstrom Nexdep at a base pressure of at least $3\times {10}^{-6}\mathrm{Torr}$. Gold substrates were prepared by an e-beam evaporating a 10-nm chromium adhesion layer and subsequently at least 20 nm of gold using an AJA ATC-Orion 8E with a base pressure of at least $10\times {10}^{-7}\mathrm{Torr}$. Tungsten, cobalt, and aluminum substrates were sputtered utilizing an AJA Orion 5 UHV Sputtering system with a base pressure of at least $10\times {10}^{-8}\mathrm{Torr}$ in an argon plasma. All substrates used have whatever native oxide may be present after an hour in an air environment. When it was desired to use the underlying copper, the native oxide was stripped by a 5-min soak in a five-weight percent citric acid solution after which the sample was DI rinsed, N2 dried, and immediately used.

2.2.

Determination of Substrate Wetting

Substrate preferentiality was determined using the hole island test on unpatterned metal substrates. The hole island test consists of spincoating a lamella-forming BCP to a thickness incommensurate with its natural periodicity, L0. At these incommensurate thicknesses, the BCP will terrace to form features that maximize favorable interactions at both the substrate and free interface. In the case of asymmetric wetting, both 0.5 and 1.5 L0 thicknesses maximize the favorable interactions, and thus an incommensurate thickness of 1.25 L0 will result in either one L0 deep holes. Similarly, in the case of symmetric wetting, both 1 and 2 L0 thicknesses are favorable resulting in the formation of one L0 high islands. One can then use optical microscopy to determine the presence of bright “holes” or darker “islands.” Our analysis is aided by the fact that at annealing temperatures below 200°C PS prefers the free interface. This means symmetric wetting indicates that PS wets both the top and bottom interfaces, whereas asymmetric wetting indicates that PMMA wets the bottom interface with PS still at the top interface. The hole island test is, therefore, a definitive probe of the substrate polymer interaction. Pristine substrates water contact angle (WCA) was also measured using millipore DI water in a goniometer. The angle reported was measured 100 ms after the $1\mu \mathrm{L}$ drop detached from syringe tip.

2.3.

Preparation of Metal Dielectric Patterns

Gl-2000 was spincoated onto a piranha-cleaned silicon wafer to a thickness of 55 nm as determined by single-angle ellipsometry on a J.A. Woollam Alpha SE, using a Cauchy model. The resist was softbaked at 150°C for 3 min. Patterning was performed using a JEOL 9300 at a voltage of 100 KeV and a current of 1 nA. Exposed wafers were developed for 15 s with gentle swirling in N-amyl acetate and then rinsed in IPA for a minimum of 1 min. Metal was then deposited onto the patterns as described previously. Metal on top of the resist was then lifted off by soaking upside down in NMP at 80°C for 15 min, after which time the sample was placed in fresh NMP and sonicated for a further 10 min at 80°C. Metal dielectric patterns were then rinsed successively with IPA and DI water before being spun dry (Fig. 1).

Fig. 1

A PS-b-PMMA film is spuncoat to a thickness of 1.25 L0. After annealing at a temperature where PS prefers the free interface, the preference of the substrate can be determined by the presence of holes or islands.

2.4.

Self-Aligned Directed Self-Assembly

PS-b-PMMA with a molecular weight of 22 to 22k and an L0 of 25 nm was spincoated to a thickness of 1.5 L0 as determined by single-angle ellipsometry. The sample was then annealed at 250°C for 5 min inside a glovebox. Post annealing samples were imaged on a Zeiss Merlin through an in-lens detector with an accelerating voltage of 1 kV and a working distance of $\sim 3\mathrm{mm}$. The PMMA block is subsequently converted to alumina oxide using the sequential infiltration synthesis technique well described elsewhere.^15–17 Post SIS an oxygen plasma etch was conducted in a Plasma-Therm ICP to remove the PS block as well as any un-infiltrated PMMA. This creates a free-standing alumina oxide template, which can be used for the purposes of pattern transfer.

3.1.

Hole Island Tests Show Metals Can be Nonpreferential for Polymers Despite Large Differences in Surface Energy

Hole island tests were performed on bulk samples as described previously, and optical micrographs were collected as shown in Fig. 2. Piranha-cleaned silicon oxide is known to be PMMA preferential and thus serves as an internal standard. PS-b-PMMA was found to form holes on chromium, cobalt, tungsten, copper, and piranha-cleaned silicon oxide. Copper oxide was found to be weakly PMMA preferential, almost to point of nonpreferentiality as subsequent SEM analysis found large areas of perpendicular assembly. Gold and aluminum had no holes or island indicating the nonpreferential nature of the substrate. SEM inspection revealed extant fingerprint on the gold substrate, whereas the aluminum substrate had far poorer assembly.

Fig. 2

Optical micrographs of a 300 micron wide region captured of hole island tests conducted on substrates of interest. For Cu the native oxide was specifically stripped using citric acid. Inset shows the drop profile used to calculate WCA. Inset SEM image is half micron square and highlights the assembly quality.

3.2.

Self-Aligned Directed Self-Assembly was Successfully Implemented on Copper Oxide/Silicon Oxide Patterns

Metal dielectric lines were successfully patterned at varying pitches and width across 100 by 100 micron areas utilizing e-beam lithography, thermal evaporation, and lift off as previously described. Copper oxide/silicon oxide alternating patterns were chosen because e-beam evaporation or sputtering deposition techniques alter the e-beam resist making lift off difficult at nanometer length scales. A metal line height of 10 nm was chosen as recent work within our group, which has shown how combining chemoepitaxy with half L0 topography to create a hybrid epitaxial flow can result in orders of magnitude speedup for the DSA process by prohibiting the formation of some defective states.¹⁸ This patterns also differs from more pure graphoepitaxial DSA strategies as implemented elsewhere,¹⁹ in that the PS-b-PMMA film thickness of 37.5 nm is far above that of the topography, and the bottom of the trench is preferential (silicon oxide) instead of nonpreferential. These patterns were then used to implement a successful self-aligned DSA process. In this implementation a pattern with a copper oxide line width of 1.5 L0 (37.5 nm) was used to the direct the self-assembly of PS-b-PMMA with a natural periodicity of 25 nm. The orientation of the domains relative to the substrate is confirmed via the SIS process. After infiltrating the PMMA domains with alumina and etching out the PS domains, two domains remain on top of the copper stripe. As we show schematically, the resulting structure thus consists of a PS domain centered on top of the metal stripe with two PMMA domains also on top of the stripe. The remaining free-standing alumina stripes can either be used in conjunction with the metal stripe as a hard mask to create self-aligned features or could be utilized as is (Fig. 3).

Fig. 3

(a) SEM image and (b) AFM height image of metal dielectric pattern utilized for self-aligned DSA. Extracted profiles (c) show line height and line height roughness. The self-aligned process was inspected via SEM (d) after assembly, and after the creation of alumina lines for pattern transfer. A schematic (e) showing the process at each inspection step is below for clarity. The pitch of the patterns (100 nm) serves as an internal scale bar.

4.1.

Metal–Polymer Interactions Are Not Explained by Surface Energy or Water Contact Angle Approaches

It is customary to think of substrate polymer interactions in terms of surface energies or solubility parameters,^6,20 in which substrates that interact equally with the two blocks are termed nonpreferential. As a demonstration of how sensitive BCPs are to surface free energy, in our hole island test utilized earlier we exploit a 1% difference in surface free energy to ensure that the lower surface free energy polystyrene preferentially wets the free interface. Despite this sensitivity, nonpreferential substrates can be made by creating a random copolymer of the two blocks of the BCP of interest and grafting it to the substrate.^10,21,22 This random copolymer naturally has a surface energy between the two blocks and is thus nonpreferential; however, it is well known that metals and their oxides have surface energies at a minimum of 20 times that of many common polymers.^23,24 As a result, previous work studying metal deposition into PS-b-PMMA found the equilibrium metal structure formed to be metal spheres in which the metal aggregated to minimize metal–polymer interactions.²⁵ It is, thus, an unexpected result that copper oxide ($\gamma \cong 1\mathrm{J}/{\mathrm{m}}^2$) or gold ($\gamma \cong 1.6\mathrm{J}/{\mathrm{m}}^2$)²⁶ would be nonpreferential for PS ($\gamma \cong 40.7\mathrm{mJ}/{\mathrm{m}}^2$) and PMMA ($\gamma \cong 41.1\mathrm{mJ}/{\mathrm{m}}^2$).²⁰

Another approach often used to analyze the preferentiality of a substrate is that of WCA. This technique allows for the probing of the hydrogen-bonding component of the surface free energy, and trends in a roughly linear manner for random copolymers of PS-b-PMMA. Previously, this technique was used to probe the nature of the LiNe flow chemical pattern after processing, where it was found the preferential XPS guide stripe (AZEMBLY NLD-128) and the nonpreferential backfill brush (AZEMBLY NLD-127, 51% PS) had WCAs of 83.7 deg and 76.0 deg, respectively.⁶ While we found that all PMMA preferential substrates had WCA below 74 deg, we do note that copper oxide with a WCA of 74 deg is close enough to nonpreferential to be used in DSA. Chromium with a WCA of 78 deg which, by WCA, should be more PS preferential than the nonpreferential backfill brush is instead PMMA preferential.

WCA and surface energy analyses only include information about dispersive, polar, and hydrogen-bonding interactions, which may adequately capture the polymer–polymer interactions as have been previously studied, but are not immediately translatable to metal–polymer interactions. As one example, the apparent nonpreferentiality but poor assembly of the aluminum substrate can be understood through recent work that has demonstrated the ability of PMMA to graft strongly to aluminum through acid—base interactions.²⁷ These strong interactions compete with the assembly kinetics, which traps a poorly assembled morphology. Still they can guide us to potential nonpreferential substrates that we can confirm via the hole-island test. All three metal surfaces, which were nonpreferential, had WCAs above 70 deg and surface free energies lower than $1500\mathrm{mJ}/{\mathrm{m}}^2$ (Table 1).

Table 1

The preferentiality of each substrate as determined via the hole island test, as well as the pristine substrates WCA and surface free energy. The surface free energy metals presented is those previously summarized by Vitos et al.26 *The water contact of piranha-cleaned SiOx though 10 deg after 100 ms, quickly approaches 0 deg as the water drop wets the entire surface of the wafer.

4.2.

Self-Aligned DSA is Complementary to Other Self-Aligned Strategies and Can Correct for Placement Error

Self-aligned DSA is complementary to the family of self-aligned strategies already developed by industry including self-aligned double patterning²⁸ and self-aligned contacts.²⁹ Self-aligned double patterning produces density multiplied features that are self-aligned to a lithographically defined template yielding intralayer aligned features. Self-aligned contacts are created by covering transistor gates in an insulating cap allowing for source/drain contacts to be made without shorting to the gate. In contrast to these strategies, our work utilizes self-alignment to align one device layer with lithographically defined features on the layer below it. Our work further serves as an expansion of recently published work by IBM and Tel.³⁰ Their work utilizes an organic planarization layer, which allows for similar creation of sparse features. They also highlight the importance of DSA’s thermodynamically driven CDU, where our work differs is in the use of materials. We utilize dielectric and metal patterns without the need for brush grafting. Through the use of metal patterns, we also allow for the possibility of self-alignment of a set of features to an existing patterned feature. With many processes now potentially requiring metal traces even below the active layer, self-aligned DSA could present a more tolerant path towards patterning transistors. These self-aligned strategies can also be utilized to create cut masks, which must align to metal dielectric patterns.