Fabrication of periodic arrays of metallic nanoparticles by block copolymer templates on HfO₂ substrates

HfO₂ films of 3 and 6 nm were deposited by atomic layer deposition (ALD) in a Savannah reactor (Cambridge Nanotech) [19, 20] on TiN films and on Si substrates with native oxide. The as deposited HfO₂ films of 3 nm were found to be amorphous by x-ray diffraction measurements, while the 6 nm films are partially crystalline. In order to fully crystallize the 6 nm films, samples were annealed in a rapid thermal processing (RTP) machine at 500 °C for 60 s in N₂ ambient.

The HfO₂ surface was cleaned either in piranha solution (H₂SO₄/H₂O₂ with 3/1 volume ration at 80 °C for 40 min) or with oxygen plasma (40 W for 2 min). Samples were then sonicated in an isopropanol bath prior to polymer solutions spinning for cleaning from particles contamination.

A brush layer of hydroxyl-terminated P(S-r-MMA) RCP with 0.62 styrene fraction, 14.5 Kg mol⁻¹ molecular weight and polydispersivity index (PDI) 1.25 was deposited on the HfO₂ surface by spinning a solution of 18 mg in 2 mL of toluene at 3000 rpm for 30 s. The surface neutralization was achieved by grafting the P(S-r-MMA) polymer chains to the hydroxyl groups present on the oxide surface by thermal treatment in an RTP machine in N₂ atmosphere for 10 min at temperatures between 230 and 310 °C. The non-grafted polymer chains were afterward removed by washing in toluene for 5 min. The thickness of the grafted RCP was measured by means of spectroscopic ellipsometry (M-200U, J A Wollam Co. Inc.).

Asymmetric PS-b-PMMA BCPs with styrene fraction ∼0.7 (Polymer Source Inc.) in toluene solution (18 mg in 2 mL) were spun on the RCP-treated surface at 3000 rpm for 30 s, obtaining a polymer thickness of ∼30 nm. The nanoscale phase separation was induced with an RTP treatment at 250 °C in N₂ atmosphere, with process time varying from 5 to 15 min. After annealing, a regular array of hexagonally packed PMMA cylinders perpendicular to the surface embedded in a PS matrix was obtained. In order to vary the size and spacing of the nanodomains, various M_w were adopted: 54 Kg mol⁻¹ (PDI 1.07), 67 Kg mol⁻¹ (PDI 1.09) and 102 Kg mol⁻¹ (PDI 1.08). The final nanoporous template used as a soft mask for the BCP-based lithography was obtained by selectively removing the PMMA component by degrading the polymer chains under UV light exposure (λ = 253.7 nm; 5 mW cm⁻²), followed by soaking in acetic acid for 8 min, rinsing in deionized water and drying under N₂ flow. The pores were finally opened and the RCP on the bottom was removed by exposure to oxygen plasma (40 W), obtaining a nanoporous PS template.

Thin films of 5 nm of Pt or 6 nm Pt/4 nm Ti were physically evaporated through the nanoporous PS templates. Electron-beam evaporation was the deposition method of choice since it allows a directional covering of the template walls, giving better results for pattern transfer if compared to other techniques (i.e. sputtering, CVD, etc). The PS template together with the overlying excess metal were afterward removed either in piranha solution [21] or in a sonicated bath of warm toluene [12, 13, 22].

Images of the nanoporous PS templates and metal dot arrays were acquired using a field emission scanning electron microscope (Supra 40, Carl Zeiss AG) operated at the acceleration voltage of 15 KV and 4 mm working distance using the in-lens detector.

The AFM images were acquired in tapping mode using a Dimension Edge instrument (Bruker) equipped with a topography probe (PPP-NCHR, Nanosensors).

The determination of the size of the characteristic features in the polymeric templates and in the patterned samples was carried out by software analysis applying a threshold mask to the SEM images, then the equivalent radius was extracted using the grain distribution function of Gwyddion software. The reported mean values and standard deviations were extracted by a Gaussian fit to the data. At least four SEM images of 1.125 × 0.774 μm were used for statistical data analysis for each sample.

The first step in the lithographic process is the substrate cleaning. This step is required to remove contaminants like carbon impurities and particles, however it can have an impact also on the surface energy (i.e. change the wettability of the surface) [23] and occasionally affect the substrate properties. Piranha solution is often applied for the removal of carbon contaminants from SiO₂ substrates and is widely applied prior to RCP and BCP spinning [21, 22, 24, 25]. The effect of the piranha surface treatment in producing a highly hydrophilic surface was reported for silica and borofloat glass substrates [26–28], greatly improving the grafting of the RCP chains. This wet cleaning is however highly aggressive and entails a very poor materials compatibility, limiting its applicability for transition metal oxides. For the HfO₂ material, we found that the as deposited amorphous film is completely removed by a few minutes of immersion in piranha solution. If however the sample is annealed at a high enough temperature, the HfO₂ film crystallizes and only 0.5 nm of the original layer are removed after 40 min in piranha solution, a thickness compatible with the removal of the carbon contaminants. A deeper inspection however reveals the appearance of holes and cracks in the treated film, indicating that a surface damage occurred. These modifications are clearly visible in the SEM pictures of the HfO₂ surface and can be identified even in AFM topographic maps of the treated surface (figure 1).

Zoom In Zoom Out Reset image size

In order to avoid any damage of the HfO₂ film, an alternative route for the surface cleaning from carbon contaminants can be a selective dry etch like oxygen plasma [29]. For an oxygen plasma treatment of 120 s at 40 W, no detectable HfO₂ film thickness variation was found by spectroscopic ellipsometry, while from SEM and AFM analysis the appearance of major surface topography modifications was excluded for both crystalline and amorphous films.

Furthermore, a relevant question is how the plasma treatment affects the wettability of the HfO₂ oxide and creates bonding sites for the RCP chains grafting to the oxide surface, i.e. how the hydroxyl groups are modified during the plasma treatment. In figure 2, the grafted RCP thickness is reported as a function of the annealing temperature for the O₂ plasma-treated samples, with the uncleaned samples inserted for reference. It is worth noting that the thickness variation among repeated experiments lies within 0.5 nm. For temperatures lower than 290 °C, the O₂ plasma highly enhances the grafting behaviour, resulting in a flat trend as a function of temperature even at 230 °C, allowing to obtain an optimal surface neutralization at lower temperature. On the contrary, the untreated surface requires a temperature higher than 290 °C for the complete neutralization to occur, showing a monotonous increase of the grafted thickness for increasing annealing temperatures, meaning that a higher kinetic energy is required for the RCP chains to find available OH sites on the surface. At 310 °C, both treated and untreated samples reach the same saturation thickness, indicating that all the available OH sites are grafted to RCP chains. Since the saturation level is equal for both sample series, the different dependence on temperature is to be related to more reactive hydroxyl groups formed after O₂ plasma exposure, rather than to an increase of the number of hydroxyl groups.

Zoom In Zoom Out Reset image size

In summary, for the ALD-grown HfO₂, improved results in terms of RCP grafting are obtained using the dry O₂ plasma cleaning of the surface, which in turn offers the advantage of a broader process compatibility in comparison with the wet piranha cleaning.

In order to investigate the effect of the grafted RCP thickness on the effective surface neutralization, we analysed the orientation evolution of the cylindrical nanodomains in BCP layers deposited over various RCP thicknesses (figure 3). For a thickness ≤4 nm, cylinders completely parallel to the surface appear after nanophase separation throughout the sample surface. Increasing the RCP thickness, the perpendicular orientation starts to nucleate and above 5 nm areas with perpendicular orientation appear. The relative coverage of the areas with perpendicular orientation increases with increasing RCP thickness and for a thickness ≥6 nm the cylinders are perfectly oriented perpendicularly to the surface. This effective surface neutralization can be achieved for a RCP annealing temperature ≥290 °C for the uncleaned samples or throughout the whole tested temperature range for the O₂ plasma-cleaned samples.

Zoom In Zoom Out Reset image size

The maximum RCP thickness achievable, together with the value above which a perfect neutralization is obtained, is highly dependent on the RCP molecular weight. In this case, a M_w of 14.5 Kg mol⁻¹ was adopted, obtaining a saturation thickness slightly lower than what is obtained on SiO₂ [25, 30], suggesting that a lower density of available hydroxyl groups is present on the HfO₂ surface, even after surface treatment. These results however demonstrate the possibility to obtain a complete surface neutralization and thus a perpendicular BCP nanodomains orientation on HfO₂ substrates. The same RCP grafted thickness and an effective surface neutralization above 6 nm was obtained on both amorphous (as deposited) and crystalline (annealed) HfO₂ films of 6 nm and on amorphous films of 3 nm (as deposited), indicating that the hydroxyl groups type and content do not vary significantly during the ALD growth and are not changed by the oxide post-deposition annealing treatment.

After effective surface neutralization by RCP grafting to the HfO₂ surface, BCP thin films of ∼30 nm were spun on the neutralized surface. Different M_w were adopted, allowing the formation of templates with different pores density and diameter [25, 31]. The nanoscale phase separation was promoted by thermal annealing in an RTP machine, which allows to reduce the processing time to a few minutes [32]. An annealing time of 5 min was sufficient to obtain a well ordered periodic structure for all the considered BCPs, while the correlation length, defined as the mean distance over which the periodic distribution is conserved, greatly varies for the different M_w and also depends on the annealing time and temperature [33]. The pores density attainable with the different M_w used is reported in table 1, while in figure 4 an SEM plane view of the three nanoporous templates is reported.

Zoom In Zoom Out Reset image size

While the density of the pores is fixed for a particular M_w, the mean pore diameter strongly depends on the O₂ plasma time that is necessary to completely open the pores and remove the RCP from the bottom, uncovering the oxide surface. Indeed, after the PMMA component has been degraded and removed, an exposure to O₂ plasma isotropically erodes the remaining PS matrix, enlarging the pores in a mostly linear fashion as a function of time (figure 5). The error bars of figure 5 depict the standard deviation of the pores diameter as derived from multiple SEM images. Even if the pores are enlarged by nearly 10 nm after 150 s, the diameters dispersion only slightly increases and σ remains below ±1.3 nm, meaning that the control over the pores size distribution is not worsened by the O₂ plasma exposure.

Zoom In Zoom Out Reset image size

Figure 5 also reports a substantial diameter overlap for different M_w obtained by adjusting the O₂ plasma time. This property potentially allows a decoupling between the defined features density and their dimension. However, for an RCP thickness of ∼6 nm, a minimum plasma time of 150 s is required to remove the bottom-lying RCP and completely open the pores in order to perform a pattern transfer using a lift-off process. As an example, in figure 6 an incomplete pattern transfer is reported for samples exposed to 120 s of O₂ plasma.

Zoom In Zoom Out Reset image size

Platinum is widely employed as electrode material in nanoscaled systems, as well as active material in sensors and catalytic applications [34], where its nanostructuration is required. Here we employ an electron-beam physical deposition method to deposit Pt thin films with a thickness of 5 nm over the BCP template, in order to obtain ordered arrays of nanosized Pt dots over the HfO₂ surface. The deposited thickness respects the rule of thumb of ≤1/3 of the polymer template thickness, necessary for the fulfilment of the lift-off process. If a higher metal thickness is required, thicker template masks can be deposited by either lowering the spinning velocity or depositing a denser polymer solution [33].

The electron beam evaporation method offers the advantage of a low chamber pressure (∼10⁻⁶ mbar) and a less energetic deposition if compared for example with the sputtering method, allowing a good separation between the metal covering the template and the metal within the pores, resulting in an easier template removal.

After the metal deposition, the template mask should be dissolved and removed together with the excess metal on top. Two wet methods (acid or solvent) were applied. The immersion of the sample in piranha solution rapidly removes the polymeric component, resulting in big areas developed in a short time. However, the etch rate is so high that the process can be difficultly controlled. Additionally, only a few materials have a sufficiently high etch selectivity to be patterned with this etchant [35]. Another concern is the surface damage that can be produced even on a crystallized HfO₂ substrate, as reported in figure 1.

Platinum is among the few metals compatible with piranha. However, several flaws manifested with increasing processing time. In figure 7, the same sample analysed after 5 and 25 min of piranha soak shows a clear loss of the patterned particles and a deterioration of the periodic distribution for increasing processing time. The type of the generated flaws, including repositioning of the metallic dots and tilting, indicates a poor Pt surface adhesion.

Zoom In Zoom Out Reset image size

An alternative route to the piranha etch can be a selective solvent for the polymeric template, an approach that brings a clear advantage in terms of materials compatibility. Toluene is a PS selective solvent, however during the O₂ plasma treatment a partial cross-linking of the polymer chains occurs. Starting from 30 s of plasma time, the solubility is greatly decreased. The consequence is that long baths in warm toluene are required, while a low power sonication should be added to break the metallic film, aiding its removal. With this method, areas of the sample with lateral dimension of the order of hundreds of micrometers were effectively patterned, but a whole template removal is difficult to achieve unless very long times are applied. The lift-off performed in toluene allowed a better pattern transfer, with the disappearing of flaws caused by repositioning of the metal dots (figure 8). Nevertheless, the long wet process and the addition of sonication, even at low power, caused the loss of some of the patterned features in the periodic matrix.

Zoom In Zoom Out Reset image size

For application purposes, platinum is known to have a poor surface adhesion on many materials, including transition metal oxides. The addition of an interlayer of titanium greatly improves the pattern adhesion [36, 37]. In figure 9, the pattern transfer of Pt/Ti metal dots by lift-off in toluene is reported for the different BCP templates adopted.

Zoom In Zoom Out Reset image size

Comparing the PS nanoporous template with the obtained metal NPs, a perfect pattern transfer can be found for the 67 and 102 Kg mol⁻¹ BCPs, with flaws only coming from the original template, mainly in correspondence with grain boundaries. The 54 Kg mol⁻¹ BCP constitutes a particular case. The small dimension of the pores and their close vicinity establishes a nearly continuous metallic film after deposition. This adds a difficulty in the template removal, requiring longer processing times and a higher sonication power, leading to a complete loss of the patterned features. In order to degrade the polymer in a faster time, we adopted a short dip in piranha solution (5 s) before soaking in toluene. It is worth to note that this very short dip does not impact significantly on the underlying HfO₂ surface. The result is reported in figure 9(a). Contrary to the templates with higher M_w, more deficiencies can be identified, which can be either related to a not complete opening of the pores or to the fast piranha dip required in this case.

The pores diameters for the three adopted templates and the relative dots diameters after pattern transfer are summarized in table 2. For the 67 and 102 Kg mol⁻¹, the resulting metal dots are about 2 nm smaller than the original template holes, a value still compatible within the uncertainties. The slightly smaller dimension can be related to a partial covering of the template walls. For the 54 Kg mol⁻¹, the discrepancy is higher, about 6 nm. Metal NPs with a diameter of only 12 nm were obtained in this case, while the width of the Gaussian distribution is similar to that obtained with the wider templates.

The experimental results suggest that the process of pattern transfer operated with the smallest template reaches a limit beyond which the final result changes significantly. A plausible explanation is that when the dimension of the pores is approaching the dimension of the grains composing the Pt film, the deposited film can no longer be seen as continuous. At this size, the granularity of the patterned film could influence the result of the pattern transfer technique [38]. Indeed, SEM images of the deposited Pt/Ti film evidence the presence of grains with an average dimension of 6–7 nm (figure 10). The result is that NPs of only 12 nm were produced, without losing control over the particles size distribution.

Zoom In Zoom Out Reset image size

The proposed fabrication method on top of HfO₂ films can be easily extended to the patterning of other materials with the same control over the defined features size and density, which are mainly determined by the self-assembled polymeric templates adopted. The final dimension of the NPs obtained with the smaller templates is however likely to depend also on the nanostructure characteristics of the deposited film, which can result in a particle size smaller than the original template as in the present case.

Different fabrication methods can be applied in order to obtain noble metal NPs with controlled size and density. For instance, the deposition of thin (<50 nm) Pt or Au films followed by heat treatment leads to the agglomeration of the metal film in NPs with controllable size by solid-state dewetting [39–41]. In comparison with the BCP-based process, this method often requires higher processing temperatures and results in a worse control over the NPs size distribution. Most importantly, if a periodic pattern is required for the NPs matrix, pre-patterned substrates are needed, dramatically increasing the complexity of the process [42]. Conversely, the metal deposition assisted by colloidal PS particles [5] and the block copolymer micelle lithography [4] allow the generation of periodically arranged noble metal NPs over large areas, but they allow a lower control over the quality of the obtained metal NPs and the adoption of an aqueous solution strongly limits the fields of application. Alternatively, by tailoring the interaction between the NPs and the BCP blocks in block copolymer–NP composites, it is possible to direct the location of pre-synthesized NPs exploiting BCP self-assembly [43–45]. In conclusion, the advantage of our method consists in a very uniform control over the NPs size distribution for different dimensions, together with the possibility to reach extremely high densities.