Poly(styrene)-block-Maltoheptaose Films for Sub-10 nm Pattern Transfer: Implications for Transistor Fabrication

Sequential infiltration synthesis (SIS) into poly(styrene)-block-maltoheptaose (PS-b-MH) block copolymer using vapors of trimethyl aluminum and water was used to prepare nanostructured surface layers. Prior to the infiltration, the PS-b-MH had been self-assembled into 12 nm pattern periodicity. Scanning electron microscopy indicated that horizontal alumina-like cylinders of 4.9 nm diameter were formed after eight infiltration cycles, while vertical cylinders were 1.3 nm larger. Using homopolymer hydroxyl-terminated poly(styrene) (PS–OH) and MH films, specular neutron reflectometry revealed a preferential reaction of precursors in the MH compared to PS–OH. The infiltration depth into the maltoheptaose homopolymer film was found to be 2.0 nm after the first couple of cycles. It reached 2.5 nm after eight infiltration cycles, and the alumina incorporation within this infiltrated layer corresponded to 23 vol % Al2O3. The alumina-like material, resulting from PS-b-MH infiltration, was used as an etch mask to transfer the sub-10 nm pattern into the underlying silicon substrate, to an aspect ratio of approximately 2:1. These results demonstrate the potential of exploiting SIS into carbohydrate-based polymers for nanofabrication and high pattern density applications, such as transistor devices.


S-2
Synthesis of propargyl-maltoheptaose (propargyl-MH) N-maltoheptaosyl-3-acetamido-1-propyne (propargyl-MH) was synthesized according to previous report. 1 Synthesis of hydroxyl-terminated poly(styrene) (PS-OH) Hydroxyl-terminated poly(styrene) was prepared by anionic polymerization of styrene accompanied by the termination with ethylene oxide. Toluene (300 mL) was introduced in a 1 L flamed dried round bottom two-necked flask equipped with magnetic stirrer and specially designed joint with roto-flow, under vacuum. Styrene (40 g, 44.15 mL) was then added and the flask filled with argon. Sec-butyllithium (~1.4 M in cyclohexane, 10 mM, 7.14 mL) was then introduced in order to initiate the polymerization. The color of the reaction mixture turned red. The reaction flask was placed in an oil bath at 35°C for 3h. The polymerization reaction was finally terminated by the addition of ethylene oxide (5.0 mL, ~3M solution in THF) in the reaction mixture, accompanied by the addition of excess of degassed methanol. The solvent was removed under vacuum using rotary evaporator at 40°C. The polymer was redissolved in appropriate amount of THF and precipitated twice in methanol (1L). The white precipitate of hydroxyl terminated poly(styrene) was filtered using a sintered glass funnel under vacuum and then dried in a vacuum oven at 40°C overnight. The outcome was 39 g of solid product, 95% yield. The sample was characterized using 1 H NMR and SEC, which resulted in Mn ( 1 H NMR) ~ 4500 g/mol, and Mn (SEC, DMF) = 3800 g/mol (see Figure S1 and Figure S3).

Synthesis of azido-functionalized poly(styrene) (PS-N3)
The azido-functionalized poly(styrene) was prepared in following two steps: In the first step, poly(styrene) (10.00 g, 2.22 mM, Mn = 4500 g mol -1 ) was dissolved in dried dichloromethane (100 mL) in a two-necked round bottom flame-dried flask equipped with magnetic stirrer, followed by addition of trimethylamine (9.3 mL, 66.7 mM). The temperature of the reaction was reduced to 0 °C by putting the flask in an ice bath. Finally, p-toluenesulfonyl chloride (4.24 g, 22.2 mM) was added in small portions under argon flow. The temperature of the system was allowed to raise slowly to room temperature and the reaction mixture was allowed to react for overnight under stirring. The reaction mixture was diluted by the addition of CH2Cl2 (100 mL) and transferred into separating funnel (500 mL) where residual salts were removed by extraction with water (3×100 mL). The organic layer was then dried using MgSO4 and the solvent was removed by rotary evaporator. The polymer was redissolved in appropriate amount of THF and precipitated twice in methanol (500 mL). The white precipitate of tosyl terminated poly(styrene) (PS-OTs) was filtered using a sintered glass funnel under vacuum and dried in a vacuum oven at 40°C overnight. The outcome was 9.3 g of solid product, 90% yield. The sample was analysed by 1 H NMR (see Figure S1).
In the second step, ω-tosyl polystyrene (9 g, 1.94 mM) prepared in above step was charged in a two-necked round bottom flask containing DMF (60 mL) and equipped with magnetic stirrer. NaN3 (2.50 g, 38.7 mM) was then added under stirring and the reaction mixture was placed in an oil bath at 60°C overnight. The system was then let to coole down to room temperature, diluted with CH2Cl2 (200 mL) and transferred into a separating funnel where it was repeatedly washed with water to remove the residual tosylate salt, excess of NaN3 and DMF. The organic S-3 layer was then dried by adding anhydrous MgSO4. CH2Cl2 was removed by evaporation using rotary evaporator. The polymer was redissolved in an appropriate amount of THF and precipitated twice in methanol (500 mL). The white precipitate of azido-terminated poly(styrene) was filtered using a sintered glass funnel under vacuum and dried in a vacuum oven at 40°C overnight. The outcome was 8.0 g of solid product, ~89% yield. The sample was characterized by 1 H NMR and SEC (see Figure S1 and Figure S3).
Synthesis of poly(styrene)-block-maltoheptaose (PS-b-MH) block copolymer Poly(styrene)-b-maltoheptaose was prepared by click chemistry of azido-functionalized PS and alkynyl-functionalized maltoheptaose. In a round bottom, one-necked flask equipped with rotoflow and magnetic stirrer was charged with ω-azido poly(styrene) (1 eq, 6.0 g, ~1.33 mM), propargyl-maltoheptaose (1.2 eq, 2.0 g, 1.6 mM), and DMF (40 mL) and degassed by three freeze-pump-thaw cycles. Thereafter, copper nanopowder (2 eq vs acetylene group, 205 mg, 3.20 mM) was added to the solution under argon flow and subjected to another freeze-pumpthaw cycle. The solution was stirred under argon atmosphere at 65°C for 3 days. At the end of the reaction, the crude heterogeneous solution was diluted with THF and filtered through diatomaceous earth. The obtained filtrate was stirred with 5.0 g cuprisorb resin at 40°C overnight. The solution was filtered to remove the cuprisorb resin and the solvent was removed by distillation using roto-evaporator. The crude product was redissolved in appropriate amount of THF and precipitated in methanol to remove excess of maltoheptaose. The unreacted poly(styrene) was removed by re-precipitation of block copolymer in cyclohexane/heptane (60/40, v/v) mixture. The resulting white solid was dried in vacuum at 40°C overnight and characterized by 1 H NMR and SEC (see Figure S2 and Figure S3). The outcome was 6.5g solid product, ~85% yield.
Polymer characterization 1 H NMR spectra of polymer samples were recorded on a Bruker Avance 400 MHz spectrometer with a frequency of 400.13 MHz and calibrated with the signal of deuterated solvent (see Figure  S1 and Figure S2). The size exclusion chromatography (SEC) was performed at 40°C using an Agilent 390 MDS system (290 LC pump injector, ProStar 510 column oven, 390 MDS refractive index detector) equipped with Knauer Smartline UV detector 2500 and two Agilent Poly Pore PL1113−6500 columns (linear, 7.5 × 300 mm; particle size, 5 μm; exclusion limit, 200−2,000,000) in DMF containing lithium chloride (0.01 M) at the flow rate of 1.0 mL min -1 (see Figure S3).

Self-assembly
Characterization of self-assembled, horizontal cylinder oriented, PS-b-MH using AFM in tapping mode in an Icon (Bruker, US), can be seen in Figure S1. The height difference can be interpreted as an existence of two self-assembled layers, in some areas. The height difference between the two self-assembled layers of vertical cylinders was AFM characterized in a Dimension 3100 (Bruker, US), and measured to be 7.7 nm (see Figure S5). In SEM, the secondary electron contrast between the pristine blocks, PS and MH, was insufficient for a reliable detection of self-assembly. Therefore, the BCP self-assembly structures were imaged by SEM after an 8-cycle dynamic SIS process and polymer removal. Characterization of self-assembled, horizontal and vertical cylinder oriented, PS-b-MH using SEM (SU8010, Hitachi, Ltd., Japan), can be seen in Figure S6.

Polymer removal
Optimizing the reactive ion etching (RIE) process to etch the polymer from the infiltrated BCP samples was not straight forward. Polymer removal was first explored on 8-cycle infiltrated PS-b-MH with vertically oriented cylinders, using RIE in an Oxford Instruments Plasmalab System 100 in 50 sccm O 2 at 10 mTorr, 10 W RF and 600 W ICP for 15 s. SEM inspection showed that the alumina-like features were disrupted from their positions (see Figure S7). When instead a low-pressure polymer removal was performed using reactive ion etching in the same system in Cl2/Ar (20:5) sccm at 5 mTorr and 100 W RF for 10 s, the features remained in a b S-7 their positions. The latter process was therefore used for evaluation of lateral size of the alumina-like features of the vertical cylinders. The etch rate of alumina should be slower in an oxygen plasma than in a chorine-based plasma. As there was a concern that the chlorine would also etch the formed alumina-like mask, the process was further optimized, for the horizontal cylinder samples, in an Apex SLR ICP-RIE (Plasma-Therm, US), allowing lower pressure oxygen plasma processes. Since the etch time in chlorine-based plasma is short, and the measured lateral sizes of vertical cylinders are actually larger than for horizontal cylinders, it is here considered valid to compare the measured values after polymer removal, using the two methods. However, indications are that the lateral feature size is approximately 0.9 nm, or 15%, smaller after the chlorine-based plasma process, than after the initial oxygen-based plasma process (see Figure S7).

Neutron reflectometry (NR)
Above the critical Qz for total reflection, the neutron specular reflectivity profile can provide information about the scattering length density (SLD) profile perpendicular to the surface, and the layer thickness of the material. Typically, two graphs are here shown for each simulation/data set: (1) the reflectivity, R, as a function of the scattering vector for neutron momentum transfer, Qz, where the data is shown with error bars, and a model fitted to the experimental data is shown as a solid, red curve, and (2) the fitted simulated neutron scattering length density as a function of depth position, z, where zero is defined at the interface between silicon and its native oxide. A schematic illustration of the model can be seen in Figure S8. Snell's law describes how a beam at incident angle q1 (to the interface) is refracted to an angle q2 by an interface of material with different refractive indices ni according to The refractive index n for neutrons can, by ignoring the absorption coefficient, be approximated as Theoretical neutron scattering length densities (SLDs) The theoretical neutron SLDs for different materials can be expressed as where NA is the Avogadro constant, the density, bc the bound coherent scattering length, 2,3 and M the molar mass of the substance, summarized for all constituent elements. 4 Calculated SLDs for relevant substances using this formula can be found in Table S1. S-9
Alumina An atomic layer deposition, using trimethyl aluminium and water, was made on top of a silicon substrate and measured in NR. The deposited layer thickness was measured to be 17.4 nm using ellipsometry. On top of the model for the silicon substrate, including the native oxide, one layer was added to represent alumina. NR data analysis of this alumina layer resulted in 171 Å thickness, SLD 4.49•10 -6 /Å 2 , roughness 5 Å (see Figure S10). This would correspond to a 21 vol% air inclusion into a pure Al2O3 layer (SLD 5.67•10 -6 /Å 2 ). The NR analysis results are summarized in Table S3. Figure S10. NR data, simulated model and SLD for AlOx on SiO2 on Si.

PS-OH infiltration
Hydroxyl terminated polystyrene (PS-OH) was spin-coated upon a silicon substrate and measured in NR. The deposited layer thickness was measured to be 17.2 nm using ellipsometry.
On top of the model for the silicon substrate, including the native oxide, one layer was added to represent PS-OH. NR data analysis of the PS-OH layer resulted in 178±2.5 Å thickness, SLD of (1.43±0.03)•10 -6 /Å 2 , roughness of 8 Å (see Figure S11).

Dynamic PS-OH infiltration
After 2 dynamic SIS cycles into hydroxyl terminated polystyrene (PS-OH) on a silicon substrate, the sample was measured in NR. On top of the model for the silicon substrate, including the native oxide, one layer was added to represent the infiltrated PS-OH. NR data analysis of the 2-cycle infiltrated PS-OH layer resulted in 178±2.1 Å thickness, SLD of (1.43±0.03)•10 -6 /Å 2 , roughness of 8 Å (see Figure S12). Thus, very similar SLD as the pristine PS-OH film.
After 8 dynamic SIS cycles into hydroxyl terminated polystyrene (PS-OH) on a silicon substrate, the sample was measured in NR. On top of the model for the silicon substrate, including the native oxide, one layer was added to represent the infiltrated PS-OH. NR data analysis of the 8-cycle infiltrated PS-OH layer resulted in 189±2.9 Å thickness, SLD of (1.46±0.04)•10 -6 /Å 2 , roughness of 8 Å (see Figure S13). The difference in layer thickness, comparing to the pristine PS-OH film, might be explained by variations from spin-coating, where 1 nm thickness difference is to be expected. There might be a slight addition of alumina in the hydroxyl terminated PS after 8 infiltration cycles. It would then be corresponding to an inclusion of 0.8 vol% pure Al2O3, or 1.1 vol% ALD Al2O3.

Semi-static PS-OH infiltration
After 2 semi-static SIS cycles into hydroxyl terminated polystyrene (PS-OH) on a silicon substrate, the sample was measured in NR. On top of the model for the silicon substrate, including the native oxide, one layer was added to represent the infiltrated PS-OH. NR data analysis of the semi-statically infiltrated PS-OH layer resulted in 186±2.3 Å thickness, SLD of (1.42±0.03)•10 -6 /Å 2 , roughness of 8 Å (see Figure S14). Thus, very similar SLD as the pristine PS-OH film.  A summary of the NR analysis of TMA/H2O infiltration into PS-OH can be seen in Table S4.

MH infiltration
Maltoheptaose (MH) was spin-coated upon a silicon substrate and measured in NR. The deposited layer thickness was measured to be 12.1 nm using ellipsometry. On top of the model for the silicon substrate, including the native oxide, one layer was added to represent the MH. NR data analysis of the MH layer resulted in 134±2.9 Å thickness, SLD of (1.63±0.04)•10 -6 /Å 2 , roughness of 5 Å (see Figure S15). Figure S15. NR data, simulated model and SLD for MH on SiO2 on Si.
When modelling the data for the infiltrated MH samples, a better fit was obtained when splitting the MH into two layers, where the top layer (having an interface to air) represents the infiltrated MH, and the lower layer (having an interface to the substrate) represents the unmodified MH. Regardless if a model with fix roughness of the lower layer and the top layer was used, or a model with varying roughness, the trend of increasing top layer SLD with number of cycles remained unchanged. For each infiltration cycle, the SLD increases, and after the 2 nd cycle, the infiltration depth also increases. The model with varying roughness is, however, assumed to better describe the real situation.

Dynamic MH infiltration
After 1 dynamic SIS cycle into MH on a silicon substrate, the sample was measured in NR. On top of the model for the silicon substrate, including the native oxide, two layers were added, since a 1-layer model on top of the substrate was unsuccessfully fitted. The lower layer represents the unmodified MH, whereas the top layer represents the alumina enriched MH. NR data analysis of the 1-cycle infiltrated MH resulted in a top layer of 20±2.4 Å thickness, SLD of (1.80±0.12)•10 -6 /Å 2 , roughness of 5 Å, and underneath a layer of 118 Å thickness, SLD of 1.63•10 -6 /Å 2 , roughness of 19 Å (see Figure S16). This would correspond to 4 vol% pure Al2O3 mixed into our top MH layer, or to 6 vol% of ALD Al2O3. Figure S16. NR data, simulated model and SLD for 1-cycle infiltrated MH on SiO2 on Si, with varying roughness.
After 2 dynamic SIS cycles into MH on a silicon substrate, the sample was measured in NR.
On top of the model for the silicon substrate, including the native oxide, two layers were added, since a 1-layer model on top of the substrate was unsuccessfully fitted. The lower layer represents MH, whereas the top layer represents the alumina enriched MH. NR data analysis of the 2-cycle infiltrated MH resulted in a top layer of 19±2.5 Å thickness, SLD of (2.05±0.16)•10 -6 /Å 2 , roughness of 8 Å, and underneath a layer of 119 Å thickness, SLD of 1.63•10 -6 /Å 2 , roughness of 9 Å (see Figure S17). This would correspond to 10 vol% pure Al2O3 mixed into the top MH layer, or to 14 vol% of ALD Al2O3.
After 4 dynamic SIS cycles into MH on a silicon substrate, the sample was measured in NR.
On top of the model for the silicon substrate, including the native oxide, two layers were added, since a 1-layer model on top of the substrate was unsuccessfully fitted. The lower layer represents MH, whereas the top layer represents the alumina enriched MH. NR data analysis of the 4-cycle infiltrated MH resulted in a top layer of 22±2.9 Å thickness, SLD of (2.16±0.17)•10 -6 /Å 2 , roughness of 8 Å, and underneath a layer of 120 Å thickness, SLD of 1.63•10 -6 /Å 2 , roughness of 7 Å (see Figure S18). This would correspond to 13 vol% pure Al2O3 mixed into the top MH layer, or to 19 vol% of ALD Al2O3. Qz / Å -1

S-19
After 8 dynamic SIS cycles into MH on a silicon substrate, the sample was measured in NR.
On top of the model for the silicon substrate, including the native oxide, two layers were added, since a 1-layer model on top of the substrate was unsuccessfully fitted. The lower layer represents MH, whereas the top layer represents the alumina enriched MH. NR data analysis of the 8-cycle infiltrated MH resulted in a top layer of 25±2.6 Å thickness, SLD of (2.54±0.17)•10 -6 /Å 2 , roughness of 6 Å, and underneath a layer of 128 Å thickness, SLD of 1.63•10 -6 /Å 2 , roughness of 5 Å (see Figure S19). This would correspond to 23 vol% pure Al2O3 mixed into the top MH layer, or to 32 vol% of ALD Al2O3. Figure S19. NR data, simulated model and SLD for 8-cycle infiltrated MH on SiO2 on Si, with varying roughness.
Using the MH infiltration model with varying roughness of both lower layer and infiltrated layer, the SLD increases with number of infiltration cycles. After the first cycle, the infiltration depth is 20 Å, after 2 cycles, it is 19 Å. Thereafter, the infiltration depth increases with number of cycles.

Semi-static MH infiltration
After 2 semi-static SIS cycles into MH on a silicon substrate, the sample was measured in NR.
On top of the model for the silicon substrate, including the native oxide, two layers were added, since a 1-layer model on top of the substrate was unsuccessfully fitted. The lower layer represents MH, whereas the top layer represents the alumina enriched MH. NR data analysis of the 2-cycle semi-statically infiltrated MH resulted in a top layer of 22±2.4 Å thickness, SLD of (1.92±0.16)•10 -6 /Å 2 , roughness of 8 Å, and underneath a layer of 118 Å thickness, SLD of 1.63•10 -6 /Å 2 , roughness of 11 Å (see Figure S20). This would correspond to 7 vol% pure Al2O3 mixed into the top MH layer, or to 10 vol% of ALD Al2O3. These results indicate that the semi-static infiltration depth after 2 cycles was 3 Å deeper than after 2 dynamic infiltration cycles, and equal to the depth after 4 dynamic infiltration cycles. However, the SLD after 2 semi-static cycles was 0.13•10 -6 /Å 2 lower than after 2 dynamic cycles, and 0.12•10 -6 /Å 2 higher than after 1 dynamic cycle.  A summary of the NR analysis of TMA/H2O infiltration into MH can be seen in Table S5.

Solubility
To evaluate solubility of trimethyl aluminium (TMA) and water precursors in PS and MH, tabulated values of the solubility parameter d for PS of 18.7 (MPa) 1/2 and water of 48.0 (MPa) 1/2 were used, whereas estimations were made for MH and TMA using Hoy group contributions of molar attraction constants F (see Table S6 and Table S7) where r is the density, and M0 the molecular weight of the repeating unit 5,6 . The experimentally found density of 1.42 g/cm 3 for MH results in a solubility parameter of 26.3 (MPa) 1/2 , whereas the tabulated density if 1.85 g/cm 3 gives 34.2 (MPa) 1/2 . Using a density of 0.752 g/cm 3 for TMA, and ignoring any possible F contribution from Al, gives a solubility parameter of 9.5 (MPa) 1/2 . These solubility estimations indicate that TMA should be more soluble in PS than in MH, whereas water should be more soluble in MH than in PS.