**2. Master mold fabrication**

The whole replication process of the nanostructures by Soft UV NIL is composed of three separate steps: the fabrication of the master mold, the replication by this master of the polymeric stamp and the imprinting in the UV curable resist by using this replicated polymeric stamp. Together, these steps affect the quality of the final replication in terms of resolution and line edge roughness of the nanostructures.

The first challenge is obtaining an accurate fabrication of the master mold; particular care has to be placed in order to achieve high resolution with low roughness features. Compared to other lithography techniques, the quality of the master mold is very important since the PDMS stamp will replicate any eventual imperfections with precision. These constraints can be relaxed, for instance, in the fabrication of the UV photomask for projection optical lithography, in which a demagnified image can be projected onto the photoresist (4:1). This allows for the patterning of smaller features as well as the reduction of mask damages and imperfections (Cui (2008)). For this reason, the fabrication process of such a master mold is more often based on electron beam lithography (EBL), which provides flexibility and high resolution.

The resolution limit of EBL is determined by proximity effects inside the resist, which are due to the scattering of electrons within the substrate. Working with a high energy electron beam (above 50 KeV) and with a thin resist layer reduces these proximity effects. We will compare standard processes based on a PMMA resist with direct etching, a lift-off process and a process based on a Hydrogen-silsesquioxane (HSQ) resist. Once the template is fabricated, it can be used many times, either as a mold for direct nanoimprint in T-NIL or UV-NIL, or as a master mold for the replication of polymeric stamps. In practice, both silicon and quartz templates can be damaged during these processes and it is thus highly desirable to develop a low cost solution for large area replications. Conventional methods such as EBL are time consuming and unsuitable for mass production. Other high throughput lithography techniques developed for the semiconductor industry, like immersion optical lithography and extreme UV lithography, include complicated processes and expensive equipment. To overcome this issue, we will describe a master mold replication process based on the combination of T-NIL and reactive ion etching.

#### **2.1 EBL processes for high resolution patterns**

Among the high resolution resists available for EBL, PMMA (polymethyl methacrylate) was one of the first polymeric materials to be studied for this purpose. The most prominent features of the PMMA resist are its high resolution, its wide process latitude, and its high contrast associated with its low sensitivity. Master molds have been fabricated by two EBL processes based on the PMMA resist: a conventional single-step process with PMMA exposure and direct Silicon etching (Figure 2 - Process 1(A)), and a multi-step process with PMMA exposure combined with metallic lift-off and Si reactive ion etching (Figure 2 - Process 1(B)).

In the first process, EBL exposure at 100 KeV is performed in a 430 nm thick layer of PMMA (950PMMA A5, MicroChem Corp.), spin coated on a Silicon wafer and baked at 180 ◦C for 30 minutes. The sample is developed in a methylisobutylketone (MIBK)/isopropanol (IPA) 4 Will-be-set-by-IN-TECH

experimental conditions. To conclude with real application, we will present the fabrication by Soft UV NIL of plasmonic nanostructures for biosensing applications with superior optical performances directly related to their high quality and uniformity on the whole pattern area.

The whole replication process of the nanostructures by Soft UV NIL is composed of three separate steps: the fabrication of the master mold, the replication by this master of the polymeric stamp and the imprinting in the UV curable resist by using this replicated polymeric stamp. Together, these steps affect the quality of the final replication in terms of

The first challenge is obtaining an accurate fabrication of the master mold; particular care has to be placed in order to achieve high resolution with low roughness features. Compared to other lithography techniques, the quality of the master mold is very important since the PDMS stamp will replicate any eventual imperfections with precision. These constraints can be relaxed, for instance, in the fabrication of the UV photomask for projection optical lithography, in which a demagnified image can be projected onto the photoresist (4:1). This allows for the patterning of smaller features as well as the reduction of mask damages and imperfections (Cui (2008)). For this reason, the fabrication process of such a master mold is more often based on electron beam lithography (EBL), which provides flexibility and high

The resolution limit of EBL is determined by proximity effects inside the resist, which are due to the scattering of electrons within the substrate. Working with a high energy electron beam (above 50 KeV) and with a thin resist layer reduces these proximity effects. We will compare standard processes based on a PMMA resist with direct etching, a lift-off process and a process based on a Hydrogen-silsesquioxane (HSQ) resist. Once the template is fabricated, it can be used many times, either as a mold for direct nanoimprint in T-NIL or UV-NIL, or as a master mold for the replication of polymeric stamps. In practice, both silicon and quartz templates can be damaged during these processes and it is thus highly desirable to develop a low cost solution for large area replications. Conventional methods such as EBL are time consuming and unsuitable for mass production. Other high throughput lithography techniques developed for the semiconductor industry, like immersion optical lithography and extreme UV lithography, include complicated processes and expensive equipment. To overcome this issue, we will describe a master mold replication process based on the

Among the high resolution resists available for EBL, PMMA (polymethyl methacrylate) was one of the first polymeric materials to be studied for this purpose. The most prominent features of the PMMA resist are its high resolution, its wide process latitude, and its high contrast associated with its low sensitivity. Master molds have been fabricated by two EBL processes based on the PMMA resist: a conventional single-step process with PMMA exposure and direct Silicon etching (Figure 2 - Process 1(A)), and a multi-step process with PMMA exposure combined with metallic lift-off and Si reactive ion etching (Figure 2 - Process

In the first process, EBL exposure at 100 KeV is performed in a 430 nm thick layer of PMMA (950PMMA A5, MicroChem Corp.), spin coated on a Silicon wafer and baked at 180 ◦C for 30 minutes. The sample is developed in a methylisobutylketone (MIBK)/isopropanol (IPA)

**2. Master mold fabrication**

resolution.

1(B)).

resolution and line edge roughness of the nanostructures.

combination of T-NIL and reactive ion etching.

**2.1 EBL processes for high resolution patterns**

Fig. 2. Schematic of the Process 1, based on positive-tone PMMA resist and of the Process 2, based on negative-tone HSQ resist.

solution at 20 ◦C for 35 sec, rinsed in IPA for 10 s and then dried with a nitrogen gun. Pattern transfer in the Silicon substrate is then performed with a highly anisotropic RIE process based on CHF3/SF6/O2, and the PMMA resist is finally removed with acetone. This process allows for the fabrication of high resolution and low sidewall roughness structures, as shown in the Figure 3 (a and b).

Since PMMA is a positive tone resist, an inverted pattern can be obtained either by exposing the complementary area, which is time consuming, or by a multi-step process with metal lift-off. In order to test the PMMA performance and the feasibility of this second process, we have carried out a challenge exposure of nanodots in a hexagonal pattern with a nominal diameter of 20 nm and a pitch of 60 nm, in a 80 nm thick PMMA resist (950PMMA A2, MicroChem Corp.). After the sample was developed as described in process 1 and lifted off of the 15 nm Ni mask, the dot patterns were transferred into the silicon wafer by reactive ion etching. The Ni mask was then removed by chemical wet etching in a HNO3 acid solution.

The Figure 3 (c) shows a SEM image of the Si master molds after the Ni mask was removed. The inset shows a distribution graph of the area of the dots, including the standard deviation and the relative average diameter of the dots. In general, this process is suitable for the fabrication of structures up to the 100 nm scale, but it is not reproducible on sub-20 nm scale structures. This is due to the multiple intermediate steps which introduce broadening of the structures, defects and sidewall roughness. In order to improve shape uniformity and sidewall roughness, we have developed an alternative process based on negative-tone resist, Hydrogen Silsesquioxan (HSQ). HSQ is a new non-chemically amplified resist for EBL

data storage is one of the most important applications of this technology, and it was actually one of the first applications for which NIL was proposed and studied (Chou (1997)). Patterned media by means of Nanoimprinting is one of the most promising approaches pursued by global companies like Hitachi and Toshiba in order to overcome the superparamagnetic limit of current perpendicular magnetic storage media. Our group has been working within the framework of the European project "'Teramagstor"' [www.teramagstor.com], in which the main goal of the consortium is the demonstration of a Bit pattern media prototype for the next generation of disk drivers with an areal density greater than 1 Tbit/in2 fabricated by EBL and Soft UV NIL. In contrast to writing on conventional media, where the bits can be placed everywhere on the medium, a key challenge of writing on nanodot arrays of a bit patterned media is that the write pulses must be synchronized to the dot period and a perfect displacement of the dots is compulsory. Figure 4 shows a Silicon master mold fabricated utilizing the HSQ process with 15 nm dots and a 60 nm pitch (corresponding to an areal density of 180 Gbit/in2), and 15 nm dots and a 30 nm pitch (corresponding to an areal density of 720 Gbit/in2), with perfect dots displacement on the whole area. At the moment, we are working on further process optimization in order to achieve 1 Tbit/in2 which corresponds to

<sup>145</sup> Soft UV Nanoimprint Lithography:

(b)

(c)

Fig. 4. SEM images recorded on Si master molds fabricated by HSQ Process 2, after annealing at 420 ◦C of (a-b) 15 nm dots, 60 nm pitch (corresponding to an areal density of 180 Gbit/in2)

and 15 nm dots, 30 nm pitch (corresponding to an areal density of 720 Gbit/in2). The

a 27 nm pitch and will eventually demonstrate higher areal density.

A Versatile Tool for Nanostructuration at the 20nm Scale

(a)

(c)

thickness of the resist is 25-30 nm.

**<A>: 277 nm2 ( ї D** a**19 nm) StdDev: 15 nm2**

Fig. 3. SEM images recorded on Si master molds that have been processed with the PMMA process 1 (a-c) and with the HSQ process 2 (d-f). In particular, (c) and (d) show the results obtained for the processes 1(B) and the processes 2 respectively, by exposing nanodots with the same nominal diameter of 20 nm (pitch of 60 nm) in a hexagonal geometry. The correspondent dot area (A) distributions in the insets, shows that the HSQ master displays smaller dots with a better shape and area uniformity with a mean diameter (D) of 19 nm compare to the broadened mean diameter of 27 nm obtained with the multiple step process 1 (B).

which has been widely investigated in recent years (Chen et al. (2006); Falco & van Delft (20); Grigorescu & Hagen (2009); Namatsu et al. (1998)). It offers ultra-high resolution with linewidth fluctuations lower than 2 nm (Word & Adesida (2003)) and high stability and strength that can be improved by a thermal post-baking (Yang & Chen (2002)). In this second process, EBL exposure at 100 KeV is carried out on 40 nm thick HSQ (XR 1541, Dow Corning, 2% solids) spin coated on a silicon wafer. In order to avoid any change in the structure of the HSQ and possibly reduce the resist contrast, no pre-baking was conducted and the wafer was simply left for one day at room temperature to remove the solvents. The wafer is then developed in a potassium hydroxide (KOH) based solution at 20◦C for 30 sec, rinsed for 2 min in deionised water, and gently dried with pure nitrogen gas. Finally, the post thermal curing of HSQ was performed at 400◦C. By comparison with the multi-step fabrication process 1(B), this single step process results in a superior resolution and uniformity of the nanodots (Figure 3 (d)). The post-baking treatment induces a silica-like network redistribution of the HSQ which can be directly used, after anti-sticking treatment, as a master mold for replication of a bi-layer hard-PDMS/PDMS stamps.

#### **2.2 Performances of the HSQ EBL process at the 10 nm scale**

Even though a sub-10 nm resolution has already been achieved on a HSQ resist for isolated structures (Grigorescu & Hagen (2009)), one of the biggest challenges in nanofabrication is the patterning of high resolution and high density structures. Bit patterned media for magnetic 6 Will-be-set-by-IN-TECH

**Process 1(B)**

**(e)**

**Process 1(B)**

**(f)**

**(c)**

**(d)**

**<A>: 602 nm2 ( ї D** a**27 nm) StdDev: 80 nm2**

**Process 2**

**<A>: 277 nm2 ( ї D** a**19 nm) StdDev: 15 nm2**

which has been widely investigated in recent years (Chen et al. (2006); Falco & van Delft (20); Grigorescu & Hagen (2009); Namatsu et al. (1998)). It offers ultra-high resolution with linewidth fluctuations lower than 2 nm (Word & Adesida (2003)) and high stability and strength that can be improved by a thermal post-baking (Yang & Chen (2002)). In this second process, EBL exposure at 100 KeV is carried out on 40 nm thick HSQ (XR 1541, Dow Corning, 2% solids) spin coated on a silicon wafer. In order to avoid any change in the structure of the HSQ and possibly reduce the resist contrast, no pre-baking was conducted and the wafer was simply left for one day at room temperature to remove the solvents. The wafer is then developed in a potassium hydroxide (KOH) based solution at 20◦C for 30 sec, rinsed for 2 min in deionised water, and gently dried with pure nitrogen gas. Finally, the post thermal curing of HSQ was performed at 400◦C. By comparison with the multi-step fabrication process 1(B), this single step process results in a superior resolution and uniformity of the nanodots (Figure 3 (d)). The post-baking treatment induces a silica-like network redistribution of the HSQ which can be directly used, after anti-sticking treatment, as a master mold for replication of a bi-layer

Even though a sub-10 nm resolution has already been achieved on a HSQ resist for isolated structures (Grigorescu & Hagen (2009)), one of the biggest challenges in nanofabrication is the patterning of high resolution and high density structures. Bit patterned media for magnetic

Fig. 3. SEM images recorded on Si master molds that have been processed with the PMMA process 1 (a-c) and with the HSQ process 2 (d-f). In particular, (c) and (d) show the results obtained for the processes 1(B) and the processes 2 respectively, by exposing nanodots with

the same nominal diameter of 20 nm (pitch of 60 nm) in a hexagonal geometry. The correspondent dot area (A) distributions in the insets, shows that the HSQ master displays smaller dots with a better shape and area uniformity with a mean diameter (D) of 19 nm compare to the broadened mean diameter of 27 nm obtained with the multiple step process 1

**Process 1(A)**

**(a)**

**(b)**

(B).

hard-PDMS/PDMS stamps.

**2.2 Performances of the HSQ EBL process at the 10 nm scale**

data storage is one of the most important applications of this technology, and it was actually one of the first applications for which NIL was proposed and studied (Chou (1997)). Patterned media by means of Nanoimprinting is one of the most promising approaches pursued by global companies like Hitachi and Toshiba in order to overcome the superparamagnetic limit of current perpendicular magnetic storage media. Our group has been working within the framework of the European project "'Teramagstor"' [www.teramagstor.com], in which the main goal of the consortium is the demonstration of a Bit pattern media prototype for the next generation of disk drivers with an areal density greater than 1 Tbit/in2 fabricated by EBL and Soft UV NIL. In contrast to writing on conventional media, where the bits can be placed everywhere on the medium, a key challenge of writing on nanodot arrays of a bit patterned media is that the write pulses must be synchronized to the dot period and a perfect displacement of the dots is compulsory. Figure 4 shows a Silicon master mold fabricated utilizing the HSQ process with 15 nm dots and a 60 nm pitch (corresponding to an areal density of 180 Gbit/in2), and 15 nm dots and a 30 nm pitch (corresponding to an areal density of 720 Gbit/in2), with perfect dots displacement on the whole area. At the moment, we are working on further process optimization in order to achieve 1 Tbit/in2 which corresponds to a 27 nm pitch and will eventually demonstrate higher areal density.

Fig. 4. SEM images recorded on Si master molds fabricated by HSQ Process 2, after annealing at 420 ◦C of (a-b) 15 nm dots, 60 nm pitch (corresponding to an areal density of 180 Gbit/in2) and 15 nm dots, 30 nm pitch (corresponding to an areal density of 720 Gbit/in2). The thickness of the resist is 25-30 nm.

For feature sizes smaller than 100 nm, a small broadening of the pattern size is observed, even though the general shape is maintained (Chen et al. (2010)). An average broadening of 20 nm is observed for patterns (nanoline and nanohole) with feature sizes above 70 nm, while the broadening of feature sizes smaller than 50 nm become bigger (30 nm). This broadening originates from a loss of etch resistance of the NXR1020 mask resist during Ar IBE at high energy. Possible solutions to improve accuracy will be explored by using a thinner metallic

<sup>147</sup> Soft UV Nanoimprint Lithography:

The first unquestionable advantage of using a polymeric replica of the original master mold in the nanoimprint process is the cost reduction. As previously stated, from a single and expensive master mold, it is possible replicate a large number of polymeric stamps to use in the nanoimprint process. This prevents damages to the original master mold if it is directly used in the nanoimprint process. Moreover, the long range flexibility of the elastomeric material used for the stamp ensures the contact between the stamp and the substrate on large surfaces at low pressures (tens of bars) and on curved or flexible substrates. Several kind of polymeric materials have been tested as candidate for stamp replication including polycarbonate resins (Posognano et al. (2004)), cross-linked novolak based epoxy resin (Pfeiffer et al. (2002)), fluoropolymer materials and tetrafluoroethylene(PTFE) (Kang et al. (2006)). Nevertheless, poly(dimethylsiloxanes) (PDMS) still offers numerous attractive properties as a stamp elastomer. First of all, PDMS ensures a conformal adhesion of the stamp with the substrate on large areas without applying any pressure. To stress this point, Soft UV NIL performed with PDMS based stamps was recently renamed as UV enhanced Substrate Conformal Imprint Lithography (UV-SCIL) by Philips Research and SUSS MicroTec (Ji et al.

Additionally, PDMS offers numerous other attractive properties: (1) its flexible backbone enables accurate replication of relief shapes in the fabrication of the patterning elements, (2) its low Young's modulus (∼750 KPa) and its low surface energy enable conformal contact with the substrate without applied pressure and nondestructive release from patterned structures, (3) its good optical transparency down to a light wavelength of approximately 256 nm, (4) its commercial availability in bulk quantities at low cost. While PDMS offers some advantages, there are a number of properties inherent to PDMS which severely limit its capabilities in soft UV-NIL. First, its low Young's modulus limits the fabrication of features with high aspect ratios due to collapse, merging and buckling of the relief structures. Second, its surface energy (<sup>≈</sup> 22-25 mNm−1) isn't low enough to replicate profiles with high accuracy. Third, its high elasticity and thermal expansion can lead to deformation and distortions during the fabrication. Finally, it shrinks by ∼1% after curing and can be readily swelled by some organic solvents. In general, long range deformations can be avoided by the introduction of a thin glass backplane which preserves a global flexibility. On the other hand, short range deformations can be avoided only by increasing the elastic modulus of PDMS, as we will discuss in deepth in the next paragraph. Therefore, compared to hot-embossing lithography (HEL) and UV-NIL which use rigid molds, it is particularly important to control the mold

underlayer or a T-NIL resist with a larger IBE resistance.

deformation as much as possible for the soft UV-NIL.

**3. Fabrication of the replicated soft stamp**

A Versatile Tool for Nanostructuration at the 20nm Scale

**3.1 Polymeric stamp materials**

(2010)).

#### **2.3 The whole wafer inversion based on T-NIL**

We have recently proposed and developed a T-NIL based process to replicate and invert an EBL master mold (Chen et al. (2010)). The inversion by means of T-NIL allows for the fabrication of a silicon mold with inverted features, so that pillars in the original master mold become holes in the daughter mold (Figure 5). We have studied how each pattern transfer process can affect feature size, pattern shape and homogeneity. Sub-200nm nanostructures have been inverted with good reproducibility and homogeneity on field as large as 1cm2.

Fig. 5. (a) Schematic of master mold inversion based on T-NIL and SEM images of daughters molds

A schematic of this inversion process is presented in Figure 5. A thin gold layer (20nm) is first deposited on the silicon wafer. The thermal imprints are then performed on the NXR-1020 resist (Nanonex), spin coated at a thickness of 180 nm on Au/Si wafers. These resist films are annealed at 120◦C for 10 minutes to remove the solvents. The Si master mold is gently placed in contact with the resist substrate and sandwiched between the two membranes of the NXR-2500 Nanonex, which provides optimal uniformity over the whole imprinted field. Imprinting is carried out at 130◦C in two successive steps: 10 sec at 120 PSI (8 Bars) followed by 30 sec at 200 PSI (14 Bars). The imprinting tool is then cooled down below the estimated glass transition temperature of the resist, before carefully demolding the master. After removing the residual NXR1020 resist layer with O2 plasma at a rate of 30 nm/min, the thin Au mask layer is etched by ion beam etching (IBE). The resist layer is then simply removed with acetone. Finally, the patterns are transferred into the silicon wafer with a standard Si reactive ion etching process using the Au mask, obtaining hole arrays. A trichloromethylsilane (TMCS) anti-sticking treatment is then applied to the daughter Si mold. With this process, we demonstrated that inversion is feasibile at the whole wafer scale. Figure 5 shows the SEM images of a 2 inch silicon wafer inverted daughter mold. It was patterned on a surface of 1 cm2, with 200 nm-wide hole arrays. A good homogeneity was observed over the whole imprint field. These results clearly show the ability of the process to reach a resolution of 100 nm.

For feature sizes smaller than 100 nm, a small broadening of the pattern size is observed, even though the general shape is maintained (Chen et al. (2010)). An average broadening of 20 nm is observed for patterns (nanoline and nanohole) with feature sizes above 70 nm, while the broadening of feature sizes smaller than 50 nm become bigger (30 nm). This broadening originates from a loss of etch resistance of the NXR1020 mask resist during Ar IBE at high energy. Possible solutions to improve accuracy will be explored by using a thinner metallic underlayer or a T-NIL resist with a larger IBE resistance.
