Epitaxial patterning of thin-films: conventional lithographies and beyond

Photolithgraphy and EBL are the most widely used techniques in most scientific laboratories with established recipes for Si-wafer-based processing, but their application to epitaxial patterning may not be straightforward due to the issues related to both the substrates and the resists. First, the e-beam and photon sources may have different behaviors on the single crystal substrates required for epitaxial growth, for instance, whose conductivity is often not ideal (e.g. insulators) for efficient charge dissipation. The difficulty with patterning on insulating substrates using EBL is the substrate heating and charging effect. Charging produces a large electric field at the surface, which can deflect the incoming electron beam and result in pattern distortion and positioning error [21, 22]. To overcome such a problem, the most widely used anti-charging method is to coat the resist with a light metal (such as Al, Cr, Cu [23]) or a conducting polymer layer (PEDOT/PSS, AquaSAVE [24–27]) to dissipate the charge. The latter is usually favored due to its uncomplicated application and removal process. Figure 2 compares scanning electron microscopy (SEM) images of nanostructures obtained by EBL on insulating quartz substrates with and without the top discharging layer—PEDOT/PSS (poly (3,4-Etylenedioxythiophene)/poly (styrenesulfonate [24])). The characteristic bright area in figure 2(a) clearly indicates the charge build-up; however, no charging effect was observed as shown by the confined dot at the resist surface in figure 2(b). Figure 2(c)is a schematic diagram of the electron path that leads to secondary electron emission and built-up charges. In comparison, figure 2(d) shows the electron path with the existence of a conducting layer. Most trapped and built-up charges are grounded through the layer. Nevertheless, even with the conducting layer, the resolution may still be affected due to electron scattering. In addition to the conducting layer coating, other types of solutions have also been demonstrated, such as the variable pressure EBL [28] (balancing the negative charges by the positive ions created by electron-gas molecule collision) and the critical energy EBL [29] (balancing the injected and ejected electrons). All these methods, described above, bring in either additional sample fabrication steps or extra variable parameters.

Zoom In Zoom Out Reset image size

Another issue with epitaxial patterning is the incompatibility between the high deposition temperature of epitaxial growth and the low glass-transition temperature of the resist. This problem exists for both photolithography and EBL. To overcome this issue, metallic or ceramic resist materials, with much higher melting points, may be used to replace conventional organic resist. Recently, Banerjee et al demonstrated epitaxial patterning of PbZr_0.52Ti_0.48O₃ (PZT) structures on insulating substrates, with AlO_x being the resist material for high temperature processing (figure 3 [30, 31]). Apart from the ability to withstand high temperatures, AlO_x can also be directly patterned with ultra-violet (UV) light, and also subsequently developed by a basic solution that reacts with exposed AlO_x to form water-soluble alkali-metal aluminates. After deposition of desired material, the AlO_x mask is subjected to a liftoff process by using aqueous NaOH solution, in which the AlO_x dissolves as water soluble sodium aluminate. This process is compatible with both photolithography and EBL. Epitaxial patterning using AlO_x masks at the mesoscale (100–1000 nm) has been successfully demonstrated [30, 31]. For EBL epitaxial patterning, the use of AlO_x mask was combined with a Au sandwich layer between AlO_x and PMMA to enhance discharging [31]. However, in principle, other types or combinations of light metal and conducting polymer should also be applicable for this purpose.

Zoom In Zoom Out Reset image size

These processes demonstrated the key methodology of combining epitaxial growth with lithographic patterning using conventional methods —via the combination of a discharging layer (for insulating substrates) and a ceramic resist (for high temperature deposition). The extension of such processes to the patterning of various other materials is thus expected. It is to be noted that AlO_x is not the only candidate as resist material for high temperature deposition. Some metallic 'resist', such as Mo, Cr, and Ti may be favored due to their higher resolution and easier removal, as will be discussed in the following sections. In general, experiments have shown that the epitaxial patterning can be done with photolithography and EBL, but the long-standing issues with photolithography (resolution limited to the photon wavelength) and EBL (low throughput) still limit their application for sub-100 nm and wafer-scale patterning purposes. On the other hand, unconventional lithography methods have experienced significant development in recent years with unique advantages over photolithography and EBL. These novel techniques also offer new opportunities and perspectives for epitaxial patterning.

In addition to its well-developed imaging and characterizing capabilities, AFM has also demonstrated good potential for nanolithography. The various capabilities of AFM lithography include modification, deposition, removal, and manipulation of materials for nanofabrication. Depending on the basis of the operating mechanisms, AFM lithography can be classified [32] into two main groups, i.e. force-assisted and bias-assisted nanolithography, where a pure-mechanical-process (force) and an electric-assisted-process (bias) are involved, respectively. Force-assisted lithography relies on direct mechanical process, including indentation, plowing and writing; while bias-assisted lithography encompasses probe anodic oxidation and electrochemical deposition/modification. Owing to the rapid development in AFM hardware, other energy sources in addition to electrical, such as thermal or chemical can also be incorporated on the AFM tip for enhanced fabrication. Interested readers are referred to two review articles for a comprehensive understanding of AFM-based nanofabrication [33, 34].

The electric-bias-assisted AFM lithography has been widely explored due to its simplicity and stability. The manipulation of electric source and components are also relatively straightforward compared to thermal and chemical sources. The experimental setup can be realized by applying a voltage-bias between the AFM tip and the conducting substrate. One of the most extensively-studied mechanisms of bias-assisted lithography is the local anodic oxidation (LAO), where the water meniscus (due to capillary condensation) in the tip-sample gap is dissociated by the negative tip bias, and the O⁻ and OH⁻ oxidative ions react with the substrate to form localized oxide nanostructures. LAO has been studied for various materials including semiconductors [35–37], oxides [38–41], metals [42–45], molecular surfaces [46–49], and for the purpose of material deposition, removal, and modification. LAO-induced material removal is probably the most relevant process for general lithography and patterning: the locally-removed material serves as a template for subsequent pattern transfer of desired material. In this sense, LAO of typical lithographic resist, such as PMMA and SU8 have been reported [50–53]. These resist templates can be used for pattern transfer through the general 'deposition-liftoff' process. However, they are usually not thermally robust enough for epitaxial patterning through deposition, which takes place at elevated temperatures. In this sense, certain types of metallic 'resist' are favorable, that are suitable for LAO process and with much higher melting points (T_melt), such as Ti (T_melt = 1668 ºC), Cr (T_melt  = 1907 ºC), and Mo (T_melt = 2623 ºC). LAO processes on metallic thin films have been reported. For example, Snow et al and Held et al showed AFM-induced LAO on Ti films and its selective removal for fabrication of quantum point-contact in the nanoscale and mesoscale, respectively [54, 55]; Sugimura et al demonstrated nanolithography using LAO on Ti metallic resist. They found that the composition of the anodized pattern is TiO₂, which can be further etched by HF during the final liftoff [56]. Similar studies have been performed also on Cr [57] and Al [58]. However, one of the most promising candidates is Mo, with a much higher melting point than Ti, Cr, and Al. Rolandi et al first introduced the LAO of Mo to form MoO₃ patterns that can be subsequently etched by simply using water [59]. The patterned Mo were then used either as an etch mask to pattern underlying material/substrate, or as a resist mask to fabricate nanostructures through deposition and liftoff.

Kawai et al extended the application of LAO-induced Mo patterning to the fabrication of epitaxial nanostructures [60–62]. Their process takes advantage of the high melting point of Mo, which allows the deposition (with the Mo template) to be done at elevated temperatures suitable for epitaxial growth. The strong conductivity of Mo also enables decent resolution, ~100 nm, achievable for the AFM lithography. In their experiments, as shown in figure 4, a Mo film was first deposited onto the Al₂O₃(0001) single crystal substrate. MoO₃ lines were then defined using the AFM tip through LAO, which were subsequently removed by immersion in water. The Mo pattern thus obtained was used as a resist for deposition of epitaxial Fe_2.5Mn_0.5O₄ at 340 ºC. The Mo mask is stable and therefore retains its shape during deposition. After cooling down to room temperature, the Mo layer was removed in 10 vol% H₂O₂ solution. Fe_2.5Mn_0.5O₄ nanoscale lines were successfully defined in the designated area on the substrate. Goto et al further expanded the scope of this technique to the fabrication of single-nanowire devices [63] and nanoconstrained structures [64], in which the magneto-transport measurements are made possible for these epitaxial nanostructures.

Zoom In Zoom Out Reset image size

The epitaxial patterning using AFM lithography and Mo liftoff can be expanded to a wide range of materials not limited to metals and oxides. The advantages of using AFM lithography include: (1) a mask-free process, sharing the same advantage with EBL; (2) possibility of multi-step lithography by computer programming; (3) positioning of the pattern with greater accuracy without complex alignment steps as in photolithography; (4) performing simultaneous in-situ imaging after the lithography without additional sample transfer; (5) ultrahigh resolution (sub-10 nm); and (6) the good LAO performance of Mo makes it very suitable for epitaxial patterning. However, currently the biggest limitation for this technique is its low speed and low throughput, which make it unfavorable for industrial mass production. This issue may be solved by operating multiple tips in parallel, or by increasing the tip writing or scan speed. Future work should emphasize this aspect of the technique.

Nanostencil lithography (NSL) describes the direct deposition, etching, or ion implantation of materials through a pre-patterned mask—often called a stencil—to achieve the patterning of desired materials. Stencils are just shadow-masks in a general sense; however, modern stencils have evolved such that they are suitable for many different applications. It has been widely used in microelectronics, bio-technology, and nanotechnology, and can allow fabrication of structures at sub-100 nm resolution. The most attractive feature of NSL is perhaps the fact that it is a resist-free lithography. On one hand, since the material is deposited directly onto the substrate, any pre-treatment for the substrate and heat load and/or chemical interactions arising from resist processing are absent; this allows simple fabrication of high-quality structures that are atomically clean. On the other hand, the deposition condition can be kept the same as that for continuous thin films, which ensures a minimal deviation of the chemical structures and properties from the thin-film samples. These features make NSL quite favorable for epitaxial patterning. Recently, different operating modes for NSL have been also demonstrated: apart from the conventional static mode where the stencil is aligned and fixed onto substrate, the multistep and dynamic modes enable nanostencils to move relative to the substrate in between and during depositions, respectively [65–72]. These additional modes also bring new capabilities for NSL that include: (1) fabrication of multi-mask and multi-material structures; (2) fabrication of arbitrary 2D geometries; (3) varied material thicknesses and/or structures using the same stencil.

A number of different materials can be used for nanostencils, including metal, Si, Si_xN_y, and polymer. The most widely used material is probably Si₃N₄ membranes [73]; there are also increasing reports of NSL using inexpensive, easy to fabricate anodic aluminum oxide (AAO) nanostencils [74–77]. Nanostencil apertures can be defined by laser interference lithography [78], electron beam lithography [79], focused ion beam lithography [80], and local anodic oxidation [81, 82]. Nanostencils are usually made very thin, with a thickness around hundreds of micrometers. They are also very fragile and need to be handled carefully. Figure 5 shows an example of a commercial Si₃N₄ nanostencil membrane featuring 5 µm circular aperture arrays.

Zoom In Zoom Out Reset image size

Since nanostencil fabrication is a non-trivial process, the reusability of a nanostencil is quite important to make this technique truly cost-effective. However, one intrinsic issue that affects the lifetime of a stencil is the clogging effect: during deposition through the stencil, material is deposited not only on the substrate through the apertures, but also on the backside of the stencil, including around and inside the apertures. This effect gradually reduces the effective aperture size of the stencil and leads ultimately to aperture clogging. For example, Takano et al showed that about 50% of the intended thickness of the deposited metals was deposited on the sidewalls of the apertures [83]. The clogged stencil apertures are clearly illustrated in figure 6. In addition, heavily deposited metal also induces mechanical stress to the stencil and further causes the membrane to deform around the aperture edges. These effects related to clogging are probably acceptable for patterning microstructures and for low-throughput fabrication (Laboratory R&D); however, in order to apply this technique to large-scale industrial production, cleaning treatments to recycle clogged stencils are necessary. Significant effort has been put to increase the lifetime of nanostencils. This can be done by either the selective removal of deposited material from the stencil, or the functional coating on the stencil prior to deposition. For example, Vazquez-Mena et al demonstrated ex-situ cleaning of nanostencils using wet-etching. In the case of Al deposition through a Si₃N₄ membrane, the clogged aperture can be cleaned in a solution of CH₃COOH (100%), HNO₃(70%) and H₃PO₄(85%) in proportions 5:3:75, which has a good etching selectivity of Al to silicon nitride [84]. Yan et al demonstrated functionalized coating for reduced clogging from Pt deposition. Specifically, the stencil surface is functionalized by tridecafluoro-1,1,2,2-tetrahydrooctyl-1-trichlorosilane (FTS) prior to use to form a fluorine-terminated self-assembled monolayer. The low-surface-energy FTS helps to reduce the incident Pt metal from sticking to the stencil. It also serves as a protective layer during stencil cleaning after the stencil has been used for a few Pt depositions [85].

Zoom In Zoom Out Reset image size

Another challenge with NSL is known as 'blurring', i.e. the unfaithful pattern transfer (usually an enlargement of the initial pattern) from the stencil to the substrate. This could come from the geometry of deposition, including the directionality of deposition flux and the physical gap between the stencil and substrate [86]. The influence of the stencil-substrate gap on the blurring effect is clearly illustrated in the nice experiment done by Takano et al [83], where a SU-8 layer is used to tune the gap between a Si substrate and a microstencil. The deposited structures with and without the gap is compared in figure 7. A quantitative, geometrical analysis for the blurring effect can also be found in [83]. Figure 8(a) is a simple schematic illustration of the blurring effect of deposited structures when a gap is present between a stencil and a substrate. The evaporated metal flux from a typical point source is modeled as a straight line from the source to the substrate, since the mean free path of evaporated metal particles is usually much longer than the distance between the source and the substrate. Depending on the deposition geometry, evolution of the blurring effect can be quite significant over the lateral dimension of the substrate. Gradually enhanced blurring effect can be even observed over several micrometers as shown in figure 8(b). Last but not least, the blurring effect can be also due to the surface diffusion of the arriving species on the non-restricted substrate surface [87]. This effect becomes more significant during epitaxial patterning since the substrate temperature is usually quite high, giving higher mobility to the arriving adatoms. The high temperature may also induce thermal stress to the stencil, resulting in the membrane deformation similar to the situation with the mechanical stress, such as membrane stretching and shrinking. To reduce the blurring effect, the most effective approach is to minimize the gap between the substrate and the stencil. This can be done by engineering special substrate-stencil holder [88] with precise position control of the stencil relative to the substrate. Such holders also help to avoid tilting of the stencil with respect to the substrate surface to ensure a constant gap distance across the whole substrate. Alternatively, one could also develop the stencil mask directly from the substrate by chemical synthesis [74]; however, such stencil may only be used once and can be subsequently disposed by selective etching (lift-off). In addition to the gap control, increasing the distance between substrate and source, as well as reducing the thickness of the stencil membrane also helps to mitigate the blurring effect. Figure 9 provides a visual summary of all the parameters that determine a faithful pattern replication using NSL.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Since NSL requires neither resist coating nor substrate-pretreatment, it can be applied directly for high-temperature, epitaxial patterning on single-crystal and/or functionalized substrates. High temperature, epitaxial patterning using NSL was demonstrated by Shin et al, where epitaxial Pb(Zr_0.2,Ti_0.8)O₃ was deposited on a SrRuO₃ film on a SrTiO₃(001) substrate, at 760 ºC through a Si₃N₄ stencil mask [89]. The Si₃N₄ membrane was used as a stencil because it can withstand temperatures as high as ~800 ºC. The nano-pore arrays (~100 nm dia.) on the Si₃N₄ mask was defined by e-beam lithography and RIE process. Epitaxial Pb(Zr_0.2,Ti_0.8)O₃ nanodot arrays were successfully demonstrated; however, the parameters influencing the lithography process were not discussed in detail. Cojocaru et al investigated NSL patterning combined with PLD of various materials including metals (Pt, Cr), semiconductors (Ge), and complex oxides (BaTiO₃ [90–92]), in which the desired materials are directly deposited through the stencil mask on the substrate. In order to avoid the substrate-stencil gap, Morelli et al deposited an array of Al dots through a stencil on a BiFeO₃ film and used the Al dot array as an etch mask for epitaxial patterning BiFeO₃ nanostructures through FIB [93]. Such etch masks can also be defined by stencils obtained from self-assembly, such as monolayer of ordered microspheres [94] or nanoparticles [95].

Nechache et al showed controlled epitaxial patterning of Bi₂FeCrO₆ nanostructure arrays on Nb-STO(100) substrate using PLD at a substrate temperature ~650 ºC [96, 97] (figure 10). The nanostencil used in the experiment, featuring 400 nm circular apertures in a hexagonal array, was defined by laser interference lithography. An intimate contact between the stencil and substrate was achieved by using a proper mechanical fixture. Thus, the blurring effect due to deposition geometry is minimized. Nevertheless, a broadening of structure sizes was still observed and was attributed to both geometrical broadening and, more importantly, to the lateral surface diffusion of species impinging with relatively high kinetic energies at high temperature (figure 9(c)). The existence of the surface diffusion of species was confirmed by the equilibrium-shapes of the resulting Bi₂FeCrO₆ nanodots. Specifically, the nanodots all display a rectangular shape even with circular apertures of the nanostencil. This is because the kinetic energy and the surface mobility of the absorbed species are high enough to induce nucleation and crystallization for the epitaxial nanodots to favor their lowest surface energies on the {100} facets. As a result, for epitaxial patterning using nanostencil lithography, the shape of the structures may be determined primarily by their crystalline preferences instead of the nanostencil aperture. This may lead to certain limitations if one wants to study a specific shape and at small scales. Nevertheless, NSL is one of the most promising techniques for epitaxial patterning due to its resist-free implementation, simplicity, and cleanliness, plus its uncomplicated deposition of various other metals, oxides [98], and multi-layer structures [99].

Zoom In Zoom Out Reset image size

Nanoimprint lithography (NIL) is an unconventional lithography technique for high-throughput, high-speed patterning of polymer nanostructures with great precision and low cost [100, 101]. NIL relies on direct mechanical deformation of the resist material under a 'rigid' stamp and therefore avoids expensive electron or photon sources required in EBL or photolithography, respectively. In addition, it can achieve resolutions beyond the limitations set by light diffraction or beam scattering that are encountered in the conventional techniques. NIL has experienced numerous developments in the past decade due to its wide range of applications in electronics, photonics, data storage, and biotechnology. Interested readers are referred to the following reviews for the recent developments of NIL in various research fields [102–107].

Unlike EBL and photolithography, NIL uses a mechanical approach to generate resist patterns with the advantage of using a process independent of the substrate electrical conductivity. Therefore, NIL is compatible with process on many unconventional substrates, including insulating substrates that are usually encountered in epitaxial growth. In order to withstand the high temperatures during deposition, metallic resist with high melting point, such as Mo, is recommended. An appropriate metallic negative-mask (e.g. ordered hole array) can be achieved by direct deposition through a resist mask with an opposite tone (e.g. ordered pillar array). For example, Suzuki et al reported a pioneering NIL-based epitaxial patterning of Fe_2.5Mn_0.5O₄ nanodot arrays, using a metallic Mo mask that was created via depositing Mo onto a pre-fabricated polymer mask [108]. The polymer mask was generated by typical UV-NIL process and etching. The process is illustrated in figure 11. The Fe_2.5Mn_0.5O₄ deposition is conducted at 340 ºC to result in an epitaxial film. The Mo liftoff was achieved using 10–30 vol% H₂O₂ solution under a brief ultrasonication. The chemical etching mechanism based on general reactions between amorphous Mo and H₂O₂ is:

$$\begin{eqnarray}&&\text{Mo}\left(\text{amorphous}\right)+{{\text{H}}_{2}}{{\text{O}}_{2}}\to \text{MoO}_{4}^{2-}+~2{{\text{H}}^{+}}+\text{}\text{}{{\text{H}}_{2}}~\left(\text{gas}\right).\end{eqnarray}$$

Zoom In Zoom Out Reset image size

This reaction usually yields very clean liftoff for the final epitaxial nanostructures. The method was quickly extended to epitaxial patterning of other oxides [109, 110] and metallic structures [111]. Such process demonstrated a basic protocol for epitaxial patterning using NIL. However, the imprinting yield of the stamp and the properties of the Mo mask were not discussed in detail, which are important aspects for large-area, high quality epitaxial patterning. We address these two factors in the following sections.

2.4.1. Soft stamps for defect-free epitaxial patterning.

2.4.2. Ethylene tetrafluoroethylene (ETFE).

ETFE belongs to the family of fluorinated polymer materials. In general, fluorinated polymers have a number of outstanding advantages for soft stamps including very low surface energy, suitable Young's modulus, good mechanical strength, and good chemical stability. Fluorinated polymers were first used to replace PDMS-based polymers for cleaner and finer soft- and imprint-lithography [112, 113]. However, these stamps were used for low pressure and room temperature NIL process, and may not be suitable for thermal-NIL resist and/or inorganic resist. To solve this issue, Barbero et al demonstrated ETFE stamps for high-temperature thermal NIL [114]. ETFE is a copolymer of ethylene and tetrafluoroethylene. It has an exceptional toughness and flexibility, high thermal stability, and superior mechanical properties compared to Teflon. Since ETFE has a high melting point in the range of 255–280 °C, it can be readily applied for imprinting most thermal NIL resist, such as Nanonex series (Nanonex Corporation), mr-I series (Microresist Technology), Poly(methyl methacrylate) (PMMA), polystyrene (PS), and inorganic resist such as hydrogen silsesquioxane (HSQ) or TiO₂. In addition, the low surface energy adds benefit of low-adhesion and easy-demolding. The anti-sticking layer that is required for many stamp materials is unnecessary for ETFE. Figure 12 compares the NIL processes using a conventional rigid stamp and an ETFE working stamp. The ETFE working stamps can be made by hot embossing ETFE sheets with a rigid master stamp such as a Si master mold, at a temperature close to the melting point of ETFE. For example, Weiss et al demonstrated ETFE replication of a Si master stamp at a temperature ~250 ºC and a pressure ~450 psi using a commercial nanoimprinter [115].

Zoom In Zoom Out Reset image size

We first applied the ETFE stamp for epitaxial patterning with NIL [111]. The semi-transparency and flexibility of as-fabricated ETFE working stamp is illustrated by the photograph in figure 13(a). The imprinting results using conventional-Si and soft-ETFE stamps are compared in figures 13(b) and (c). Direct application of Si stamp on the resist yields significant defect generation, as shown in figure 13(b) [116]. Even tiny defects between the mold and wafer lead to visibly large unimprintable areas, due to the rigidness of both the wafer and the stamp. Unpredictable but significant defects in the form of pinholes and thin strips were observed. These defects can be transferred to the final deposition and liftoff, resulting in unwanted thin-film objects (large particles, irregular film pieces, etc). In addition, the issue with defects gets more significant as the stamp is used again and again. This disadvantage casts substantial limitation on potential process scaling. On the other hand, by using ETFE working stamp, the local defects can be accommodated due to the flexibility of the stamp. After imprinting, the demolding is also much easier for the ETFE due to its low surface energy. Defect-free imprints were achieved nearly on the entire stamping area (>1 cm²) (figure 13(c)), which is ideal for epitaxial patterning purposes. High quality imprints on single-crystal MgO substrate (after metallization) is demonstrated in figure 13(d) [111].

Zoom In Zoom Out Reset image size

ETFE working stamps have been experimentally proven to be efficient for mesoscale and microscale patterning (figure 14(a)). However, at the sub-100 nm scale the intrinsic ETFE crystallization may prevent a faithful replication of the ETFE from the original master stamp and this should be taken into consideration. Weiss et al revealed a thermal-driven ETFE crystallization by inspecting AFM topographies of ETFE stamps that were fabricated using different Si masters [115]. They found that the ETFE sheet embossed with a flat Si wafer exhibits rod-shaped crystalline domains spaced 20–40 nm apart, with a height between 2.5 nm and 5 nm (figure 14(c)). These needle-like grains are also likely accountable for the sidewall roughness of the ETFE features observed in figure 14(b). Such crystallization of ETFE under hot-embossing conditions limits the spatial resolution and therefore the performance of ETFE working stamps. To work with sub-100 nm features, one may have to turn to other types of polymers such as Ormostamp as proper working stamp materials.

Zoom In Zoom Out Reset image size

2.4.3. Ormostamp^®.

Ormostamp^® (Microresist Technology) is a promising working stamp material for sub-100 nm epitaxial patterning. Ormostamp is made of UV-curable inorganic-organic hybrid polymers, and its UV transparency and thermal stability makes it suitable for both UV and thermal NIL practices. Ormostamp is gel-like before curing, with relatively low viscosity that enables efficient filling of the master template cavities. After UV curing, it acquires sufficient hardness and good elasticity, which allows for patterning high quality, sub-100 nm structures without inducing cracks and fractures. In addition, unlike ETFE, a cured Ormostamp can be used as a master stamp for making new working stamps; therefore the tone of the stamp can be changed from positive to negative and vice versa. The fabrication process of the Ormostamp is shown in figure 15. The gel-like Ormostamp is cast either on the master stamp or a clean substrate (glass, quartz, or PET) pre-coated with Ormoprime08 (Microresist Technology)—an adhesion promoter for Ormostamp (figure 15(a)). The Ormoprime08 can be applied by conventional spin coating. The 'master stamp-Ormostamp-substrate' stack is then UV cured and thermally annealed (figure 15(b)). To perform NIL with Ormostamp, the anti-sticking layer, trichloro(1H,1H,2H,2H-perfluorooctyl)silane, is subsequently applied by vacuum evaporation (figure 15(c)). Excellent pattern transfer fidelity has been demonstrated for sub-100 nm structures [117–119].

Zoom In Zoom Out Reset image size

We studied a number of properties of Ormostamp in the context of patterning epitaxial films. For defect-free process, transparent and flexible plastics such as PET, rather than glass or quartz, are preferred as host substrates for the Ormostamp features. We demonstrated Ormostamp copies on PET substrate and their corresponding imprints, as shown in figure 16. An advantage of Ormostamp is that the feature size can be controllably reduced by oxygen plasma etching. For example, we replicated an Ormostamp from a quartz master stamp featuring 100 nm square dot array. After UV and thermal curing, the Ormostamp were cut into four pieces and subjected to O₂ plasma RIE for 1 min, at 1 Torr, and at different rf powers (no-etch, 25 W, 50 W, and 75 W), as shown in figures 16(a)–(d). Figures 16(e)–(h) show the corresponding imprints using the above working stamps. We found that the feature size of working stamps and therefore the corresponding imprints gradually reduces as the rf power increases. As a result, the RIE process allows tunable feature size of the Ormostamp made from a single master stamp, and can therefore even reduce the cost of the master stamp fabrication. Nevertheless, a reliable aspect-ratio of the Ormostamp features needs to be maintained during the RIE. In our case, if a power >75 W is used, Ormostamp pillars with relatively high aspect-ratio were created and the resultant imprinting process was not efficient due to the plastic nature of Ormostamp (figure 16(h)). However, this effect is not limited to Ormostamp [120] and is a feature of all high-aspect-ratio pillar imprinting.

Zoom In Zoom Out Reset image size

In NIL practice, a successful demolding process is the final key factor for obtaining high-quality patterns. This raises another issue with Ormostamp involving appropriate adhesive property of the three key interfaces: (1) good adhesion between Ormostamp and its host substrate (e.g. PET); this is achieved by the adhesion promoter—Ormoprime08; (2) good adhesion between resist and substrate (e.g. single-crystal oxides); this can be optimized by thorough surface cleaning and hydrophilic treatment of the substrate; and (3) low adhesion between Ormostamp and resist; this is achieved by coating the release layer, trichloro(1H,1H,2H,2H-perfluorooctyl)silane, on the Ormostamp. All the three interfaces are critical for a good demolding process. By selectively eliminating good adhesive qualities of (1) or (2), we observed that either the resist peels-off and sticks to the Ormostamp (figure 17(a)), or the Ormostamp peels off and sticks to the resist (figure 17(b)). Both these scenarios should be avoided in the demolding process. A thorough cleaning of both the stamp host and the substrate is recommended. For some materials with good water absorption properties, long time baking, and/or UV-Ozone and/or low-power oxygen plasma etching might be necessary. Large area nanostructures can be obtained after resolving the demolding issues, as illustrated in figure 17(c).

Zoom In Zoom Out Reset image size

2.4.4. Crystallization of Mo mask.

For epitaxial patterning it is also important to optimize the Mo mask that is widely used due to its high melting point and convenient liftoff. It has been experimentally found that Mo tends to crystallize on single crystal substrates, such as on MgO or SrTiO₃, before it reaches its melting point. Such crystallization may or may not hinder its subsequent liftoff process, depending on the different crystallization orientations [121]. Guilloux-Viry et al showed that Mo can grow epitaxially along (200) on MgO(100) at a substrate temperature above 450 °C [122]. For mask application, Mo is deposited at room temperature but heated to elevated temperatures for epitaxial growth. Mo thin films, deposited on MgO at room temperature but post-annealed at different high temperatures (figure 18), showed the onset of a Mo(200) peak in x-ray θ-2θ scans, indicating that Mo tends to crystallize along (200) on MgO substrate upon annealing. Such crystallization is initiated at ~310 °C and is fully established at ~420 °C. However, this crystallization—along (200)—does not seem to affect the subsequent liftoff process. On the other hand, similar crystallization effect along a different orientation, Mo(110), deposited on SrTiO₃ substrate affect the liftoff as reported by Suzuki et al. To solve this problem, a Mo/SiO_x bilayer resist was used to avoid the Mo crystallization by simply depositing the amorphous SiO_x at the bottom (figure 19) [121].

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

2.4.5. Reliable liftoff through mask undercutting.

In any resist-based lithography techniques, the liftoff step plays a most important role in determining the quality of the final patterned structures. For single nanostructure devices made by AFM-lithography or EBL, the quality of the final patterns can be examined by in-situ AFM or SEM imaging. However, the process is time-consuming for large-area, parallel patterning techniques such as NIL. Therefore, a predictable and reliable liftoff process that yields high-quality patterns becomes particularly important for techniques such as NIL.

Single-layer resist template sometimes fails to provide a good liftoff due to the unfavorable depositions along the side-walls of the template. Such effect becomes more significant for meso- and nano-scale patterns compared with micro-scale ones. It is also a universal phenomenon observed for almost all kinds of deposition techniques, especially for those with poor directionality of the depositing flux. To overcome this issue, an additional undercut resist layer is often used (e.g. LOR resist series, MicroChem LLC). This layer is coated prior to the lithography-resist (e.g. PMMA), and then selectively developed after the lithography. The idea is to create an undercut profile of the bilayer-resist for subsequent metallization. Such undercut has proven to be efficient for liftoff [123]. In addition, the bilayer-resist also offers the possibility of varying the feature size by controlling the degree of undercutting [124].

In the epitaxial patterning process using Mo mask, two liftoff processes are required: (1) resist liftoff after deposition of Mo layer, and (2) Mo liftoff after deposition of epitaxial material layer. Suzuki et al used a bilayer-resist undercut for Mo mask deposition, however, the undercut of Mo mask for subsequent Fe_2.5Mn_0.5O₄ deposition was not described [108]. We showed that the cross-section profile of the Mo mask is also critical for the final liftoff [111]. We further demonstrated a modified process that can enable a natural undercut profile of the Mo mask (figure 20) [111, 125]. Specifically, a bilayer of polymeric undercut resist and imprint resist was thermally imprinted with an appropriate stamp. Then, a brief anisotropic de-scum etch step removes the residue of the imprint resist and a certain amount of the undercut resist (figure 20(a)). The undercut resist was then wet-etched appropriately to develop a wedge-shaped resist profile. Since the wet-etch proceeds in a nearly isotropic manner, the etched resist profile therefore can be approximated by a circle, with the center located at the start point of the etch (figure 20(b)). The Mo layer is then deposited (figure 20(c)), followed by the polymer-resist liftoff (figure 20(d)). The wedge-shaped resist produced a natural undercut profile of the Mo mask. Epitaxial material is then deposited at high temperature onto the Mo template (figure 20(e)), followed by the Mo liftoff in the H₂O₂ solution (figure 20(f)). Owing to the Mo undercut, very clean liftoff can be obtained and high quality epitaxial patterned structures can be produced. The duration of the de-scum etch step in figure 20(a) is the key to obtain the desired wedge-shaped polymer resist profile. Figure 20(g)–(l) illustrates a process with a de-scum etch duration increased by a factor 2. In brief, the start point of the wet-etch was moved downward, deep into the undercut resist layer by the longer anisotropic de-scum etch. After the isotropic wet-etch, as approximated by the circle, a conventional bilayer resist undercut was created, figure 20(h). The different resist profiles significantly affected the cross-sectional profiles of the deposited Mo structures, which further lead to a failure of liftoff.

Zoom In Zoom Out Reset image size

Material deposition along the sidewall of the template is not always a nuisance that one wants to eliminate. Such deposition can be used for fine structure patterning beyond the size limit of the stamp [126–128]. For example, Cha et al reported a size-reducible, high-aspect-ratio, hollow Mo nanomask that was obtained from intentional side-wall deposition on the template, and demonstrated the fabrication of epitaxial 3D ferromagnetic nanostructures with high controllability [129]. The process of creating the hollow Mo mask is illustrated in figure 21. In their experiment, a bilayer resist is imprinted by a stamp with 200 nm feature size, and subsequently developed to form a typical undercut. Amorphous Mo was deposited on the patterned polymer substrate that covered all over the resist pillar (with a big head) including its sidewalls. The top head region of the pillar is then cleaved by ultrasonication. After acetone cleaning and spin-drying, a hollow Mo mask with a reduced window size ~100 nm was created for subsequent epitaxial deposition. It was noted that the size of the final structure is determined by the window size of the hollow mask, which is further controlled by the duration of the selective etching of the bottom resist layer during the undercut development. This result provided a simple method for the fabrication of integrated epitaxial nanostructure arrays with potential to overcome the present resolution limit.

Zoom In Zoom Out Reset image size

2.4.6. Nanoseeding and heterostructures integration.

In this section, we briefly introduce a developing technique named nanoseeding growth. It is a large-area and position-specific epitaxial growth that can be viewed as an extension of NIL-based epitaxial patterning. Different from the processes discussed so far, nanoseeding aims to fabricate nanocomposites such as ferromagnetic or ferroelectric nanodots array embedded in another function material matrix. A variety of functionalities can arise by coupling the functionalities of the nanophase and a matrix through hetero-epitaxial interfaces [130–133].

The nanoseeding process involves, as a first step, epitaxial patterning of nanoseeds in an ordered array on the substrate. This is done by NIL and Mo liftoff, combined with dry deposition of the seed material. Next, the matrix materials are deposited and they are selectively grown on the substrate surface but avoiding the nanoseeds, due to minimization of total free energy required for the formation of the interface between the two phases and the nanoseeded substrate. Applying this strategy, Sakamoto et al demonstrated fully aligned epitaxial (Fe,Zn)₃O₄ nanodots on SrTiO₃(001) substrate, that were perfectly surrounded by an insulating BiFeO₃ matrix (figure 22). They found that a perfect selectively-growth of BiFeO₃ can be achieved for (Fe,Zn)₃O₄ small dots (<800 nm).This method was also extended to the fabrication of (Fe,Zn)₃O₄/BiFeO₃ core-shell type structure by using a co-deposition method [131]. As shown in figure 22, it is obvious that BiFeO₃ grew epitaxially along the side of the (Fe,Zn)₃O₄ dots, passing over the area of the (Fe,Zn)₃O₄ terrace and forming the shell/matrix, while (Fe,Zn)₃O₄ grew epitaxially on top of the (Fe,Zn)₃O₄ dots (seeds). In general, the nanoseeding technique achieves epitaxial fabrication of functional nanocomposites and resolves the issue of precise positional alignment of materials with high density nanostructure integration. The position alignment relies on pre-patterned nanoseed arrays, which are achieved by epitaxial patterning using NIL and Mo liftoff. This method can be extended to other material combinations and may lead to advanced integration of hetero-junctions composed of oxide materials with precise position, size and shape controllability over a large area.

Zoom In Zoom Out Reset image size

In standard lithography, the micro- and nano-structures are fabricated on a certain rigid substrate (e.g. Si wafer) and are bound to the substrate permanently. This rigidness impedes the fabrication and incorporation of micro- and nano-devices on flexible substrates, which is a critical step toward future soft-electronics [134–138]. Besides, it is difficult to translate the favorable economic scaling to industrial manufacturing since the wafer volume is still comparatively small. To solve the above issues, a roll-to-roll process has been developed that can achieve patterned nanostructures directly on flexible substrate with high-throughput and high-speed [139]. However, certain materials and mature devices are incompatible with the roll-to-roll process. An alternative approach is therefore to fabricate the structures on conventional substrate using standard lithography, but then transport the products to another substrate using contact-transfer. The transferred objects can be preliminary resist patterns, intermediate device components, or mature devices [140].

Contact-transfer can enable epitaxial patterning in such a way that it allows transfer of resist patterns to 'hard-to-tackle' substrates, such as ones with very small dimension, irregular shape, non-flat surface, and/or very thin/soft material, where the conventional spin-coating is unachievable. Cheng et al demonstrated transfer bonding of PMMA resist from NIL stamps to both flat and pre-patterned substrates, using a technique call reverse imprinting [141–143]. Furthermore, additional resist layers can be continuously piled up to form 3D polymer structures. In some cases such transfer bonding can be made easier under elevated temperatures and/or pressures. In this sense, a typical nanoimprinter, being a temperature and pressure source, serves a good platform for such transfer bonding process. For example, we used a nanoimprinter (Nanonex NXB-100) and demonstrated successful transfer bonding of AZ1512 photoresist from Si wafer to MgO substrate for epitaxial patterning (figure 23(a)). To reduce the adhesion between the photoresist and the Si wafer, we directly spin-coated AZ1512 on clean Si/SiO₂ surface without the HDMS evaporation. The MgO(001) substrate was pre-cleaned and baked on a hotplate at 200 °C for 10 min to remove the absorbed water. Micrometer AZ1512 wire array were directly transferred onto MgO substrates for subsequent metallization (figure 23(b)). The resist transfer process was done in a nanoimprinter, which provides the ideal pressure and temperature environment (200 psi and 80 °C) needed for the transfer. For epitaxial patterning, the AZ1512 mask was then transferred to Mo mask (figure 23(c)). Finally, figure 23(d) shows a photograph of the epitaxial Fe micrometer wire patterns obtained on MgO substrate using the Mo liftoff process introduced in the earlier section.

Zoom In Zoom Out Reset image size

Another process related to this subject is the transfer and/or release of the epitaxially patterned structures through substrate etching [144–147]. This method allows the structures to be transported to another substrate (such as a plastic film) or their release to a free-standing form (e.g. in solution). The process enables the epitaxial structures to be incorporated in devices with more flexibility, but not bound to the original substrate required for epitaxial growth. For example, Qi et al reported an approach to transfer epitaxial PZT ribbons from the host MgO substrate to PDMS rubber substrate using selective etching of MgO in 20% phosphoric acid [148]. The exposed MgO between the PZT ribbons provided an avenue for the acid to undercut and loosen the ribbons without completely dislocating them. Transfer printing of PZT ribbons was done by bringing a piece of PDMS into conformal contact with the wafer and quickly peeling it back to retrieve the ordered ribbon arrays (figure 24(a)). Similarly, such substrate etching can also be realized via dry etching methods such as RIE [116]. On the other hand, if the epitaxial growth does not require a specific host single-crystal substrate, the material can be simply patterned on a 'sacrificial layer' that is then selectively etched after the lithography to release the patterned structures [149–154].Such approach has been widely adopted in high-volume production of lithographically fabricated free-standing nanostructures.

Zoom In Zoom Out Reset image size