**Antireflection Coatings for Strongly Curved Glass Lenses by Atomic Layer Deposition**

#### **Kristin Pfeiffer 1,\*, Ulrike Schulz 2, Andreas Tünnermann 1,2 and Adriana Szeghalmi 1,2,\***


Received: 12 June 2017; Accepted: 3 August 2017; Published: 9 August 2017

**Abstract:** Antireflection (AR) coatings are indispensable in numerous optical applications and are increasingly demanded on highly curved optical components. In this work, optical thin films of SiO2, Al2O3, TiO2 and Ta2O5 were prepared by atomic layer deposition (ALD), which is based on self-limiting surface reactions leading to a uniform film thickness on arbitrarily shaped surfaces. Al2O3/TiO2/SiO2 and Al2O3/Ta2O5/SiO2 AR coatings were successfully applied in the 400–750 nm and 400–700 nm spectral range, respectively. Less than 0.6% reflectance with an average of 0.3% has been measured on a fused silica hemispherical (half-ball) lens with 4 mm diameter along the entire lens surface at 0◦ angle of incidence. The reflectance on a large B270 aspherical lens with height of 25 mm and diameter of 50 mm decreased to less than 1% with an average reflectance < 0.3%. The results demonstrate that ALD is a promising technology for deposition of uniform optical layers on strongly curved lenses without complex in situ thickness monitoring.

**Keywords:** atomic layer deposition; optical coatings; antireflection; strongly curved surface

#### **1. Introduction**

Most optical systems contain a large number of lenses or other optical elements. Reflections at each interface reduce the intensity of the transmitted light and thus the overall efficiency of the systems. Reflection losses can be greatly reduced by applying antireflection (AR) coatings to the optical surfaces [1–4]. In addition, AR coatings attenuate the effect of ghost images that are caused by multiple reflection of light from lens surfaces. Optical interference coatings that are typically thin film multilayers of high-refractive and low-refractive index materials demand precise thickness control of each layer. Commonly, thin films applied in precision optics are produced by physical vapor deposition (PVD) [5,6]. Due to the line-of-sight nature of PVD, the surface of a convex lens that is normal to the deposition flux receives a higher amount of material than the edges of the lens. As indicated in Figure 1a, significant thickness gradients might occur on highly curved lenses. Consequently, the required film thickness might not be met over the entire surface of the lens, leading to a distortion of the resulting transmittance spectra. To achieve a better uniformity on curved substrates, complex technical modifications are necessary when using PVD methods, that includes f.e. the constant rotating and tilting of the lens during the deposition, with or without the usage of complicated shadowing masks [7–10]. Antireflection nanostructures are another approach to reduce reflection losses at curved surfaces [11–13]. Nevertheless, for outer lenses of optical systems multilayer AR coatings are preferably used due to their better cleanability and mechanical stability.

Atomic layer deposition (ALD) is an alternative and promising technology for uniform multilayer optical coatings [14–18]. We have previously shown a broadband AR coating on flat high refractive

index glasses using SiO2/HfO2 multilayers [19]. Atomic layer deposition is also being considered for more complex interference coatings such as dichroic mirrors and narrow bandpass filters [16,20,21]. Atomic layer deposition is a modified form of chemical vapor deposition, where the precursors are sequentially exposed to the surface until saturation is reached [22]. Precursor pulses are separated by inert gas purging; as a result, no gas-phase reactions can take place and the chemical reactions are limited to the surface, see Figure 1b. A typical ALD cycle for the deposition of metal oxides contains four steps: precursor pulse, inert gas purge, oxidizing pulse and inert gas purge. In the case of precursors with low chemical reactivity, often a hold step is introduced after the precursor pulse. Hence, the precursor is trapped in the reactor to entirely react with the surface active groups. Due to this cyclic surface-controlled growth, ALD inherently offers precise thickness control, good thickness uniformity and high reproducibility. It is well known for its conformal film growth on complex nanostructures with high aspect ratios [23,24]. In this work, the capability of ALD for deposition of antireflection coatings on highly curved lenses has been analyzed.

**Figure 1.** Illustration of (**a**) physical vapor deposition (PVD) deposition and (**b**) atomic layer deposition (ALD) on a hemispherical lens.

This paper first discusses single-layer properties and thickness uniformity of the SiO2, Al2O3, TiO2, and Ta2O5 coatings, then an AR design and its adjustment to the ALD coating is presented on flat glass substrates. Finally, ALD antireflection coatings are demonstrated on curved lenses, firstly on a half-ball lens and secondly on an asphere.

#### **2. Materials and Methods**

ALD-deposited SiO2, Al2O3, Ta2O5 and TiO2 thin films were used for the antireflection coatings. Depositions were carried out in an Oxford Instruments (Bristol, United Kingdom) OpAL™ ALD reactor and a Picosun Oy (Espoo, Finland) Sunale™ R200 ALD reactor with a showerhead setup for single-wafer processing. In the OpAL tool, thin films have been grown by plasma-enhanced ALD (PEALD) at substrate temperatures of 100 ◦C. In the Sunale tool, thermal ALD processes were performed at 300 ◦C.

All metal oxide films were grown from commercially available precursors. The low-index material SiO2 was deposited using tris[dimethylamino]silane (3DMAS). Trimethylaluminium (TMA) was applied to deposit the mid-refractive index material Al2O3. The high-index materials TiO2 and Ta2O5 were deposited from titanium(IV)isopropoxide (TTIP) and tantalum(V)ethoxide (Ta(OEt)5), respectively. Process parameters are summarized in Table 1.


**Table 1.** Process parameter for depositing SiO2, Al2O3, TiO2 and Ta2O5 ALD thin films.

The growth rates and the optical properties of the ALD thin films are determined from single-layer experiments on flat substrates. The growth rates (growth per cycles, GPC) were determined on Si samples by measuring the film thickness with a J.A. Woollam Co. (Lincoln, NV, USA) M-2000® spectroscopic ellipsometer. A Sentech Instruments GmbH (Berlin, Germany) SE850 spectroscopic ellipsometer was used for uniformity mapping of the film thickness on an 8 inch (200 mm) silicon wafer over 180 mm central area on the wafer.

Refractive indices were determined by spectrophotometry of 200 to 300 nm thin films coated on fused silica samples. The reflectance and transmittance spectra were measured with a PerkinElmer, Inc. (Waltham, MA, USA) Lambda 950 spectrophotometer equipped with a home-build accessory for absolute reflectance measurements [25].

For demonstration purposes, antireflection (AR) coatings were applied to a half-ball lens with a diameter of 4 mm and to an aspheric lens with a diameter of 50 mm and a center thickness of 25 mm. An Olympus K. K. USPM-RU-W NIR micro-spectrophotometer (Tokio, Japan) was used to measure the reflectance from a minute spot on different positions of the lens, whereas the lens is placed on a tilt stage and tilted to angles up to 60◦. The tilted lens is then moved in the *x*-, *y*- and *z*-direction so that the light from a fixed source is focused on the lens surface and the light rays are perpendicular to the surface (AOI = 0◦).

#### **3. Results and Discussion**

#### *3.1. Characterization of ALD Thin Films*

ALD processes for dielectric thin films have frequently been reported, whereas Al2O3 is the most investigated ALD material [26,27]. Al2O3 has been applied in ALD antireflection coatings in combination with TiO2 [17,20] or Ta2O5 [28]. Next to this, SiO2 is a very important low-index material that we recently applied in ALD optical coatings [19,21,28,29]. The properties of the single-layer films resulting from the ALD processes used in this work are summarized in Table 2. The listed GPC values have been used to calculate the necessary ALD cycles to reach the thicknesses of each layer of the following AR coatings.

**Table 2.** Growth rate on silicon substrates (growth per cycles (GPC) in Å/cycle), refractive index and thickness non-uniformity over a 200 mm area of deposited ALD thin films. The corresponding deposition temperatures are given in brackets.


<sup>1</sup> thickness non-uniformity, defined as NU% = (*d*max − *<sup>d</sup>*min)/2*d*average × 100.

Growth rates and refractive index of SiO2 thin films are similar to films grown from other commercially available precursors, as BDEAS, BTBAS and AP-LTO®330 [30,31]. Alumina ALD thin films show a lower refractive index at lower deposition temperature [32] owing to a lower density at lower deposition temperatures [26]. The lower GPC of Al2O3 at higher deposition temperatures

is attributed to less OH groups on the surface. Determined growth rates of Ta2O5 are comparable to growth rates reported for Ta2O5 thin films deposited using H2O and Ta(OEt)5, Ta(NEt2)3 or Ta(NEt)(NEt2)3 [33–35]. The reported GPC for PEALD TiO2 using TTIP in the range of 0.3–0.6 Å/cycle are relatively low, whereas thin films grown from TDMAT, Ti-Prime or Ti-Star have slightly higher growth rates than films grown from TTIP [36–38].

Very good lateral film thickness uniformity in the reactor is a prerequisite to ensure a uniform coating on a lens surface. However, non-uniformity in ALD processes is not explicitly analyzed in most research articles. Most tool providers guarantee a standard deviation of the ALD coatings of ca. 1%–3% depending on the material and process conditions. Noteworthy, the upscaling of ALD processes in batch reactors with similar non-uniformity distribution on larger-area batches is feasible [16]. The ALD coatings deposited in the OpAL research tool have thickness non-uniformities (NU%), defined as *(d*max − *d*min*)/2d*average × 100, of about ±1.5% (Al2O3, SiO2) and ±2.0% (TiO2). The processes in the Sunale R200 ALD reactor result in a thickness non-uniformity of about ±2.1% for Al2O3 and ±4.0% for Ta2O5, see Table 2. Elers et al. [39] discussed the sources of non-uniformities in ALD processes including overlapping precursor pulses due to short purge times, death pockets, etc., but also non-uniform gas and temperature distributions in the reactor chamber.

Figure 2a shows the surface mapping of a 200 mm wafer after thermal Al2O3 ALD process using 1156 cycles (TMA + H2O). The alumina film thickness on the wafer in this thermal process does not show a statistical random distribution, but a specific and well-reproducible lower film thickness on the right side of the reactor chamber than on the left side. Interestingly, the precursor and purge gas inlet is on the side where lower film thickness is measured indicating that the precursor dose is sufficient. There might be a temperature gradient on the wafer due to the gases entering the reactor on the right side or the purge time and gas flow might be not sufficient due to inadequate inert gas distribution. In PEALD processes, rather concentric thickness contour lines have been observed (not shown here), whereby the maximum position can be adjusted by the flow rates of the precursor and purge gas. We have demonstrated the possibility to improve the film thickness uniformities by rotating the substrate. Figure 2b depicts a thickness mapping of a wafer where the thermal Al2O3 ALD process was stopped after 500 cycles, the wafer manually rotated by 180◦ and the process continued for another 500 cycles. The wafer rotation could significantly improve the thickness non-uniformity from 2.4% to 0.6%, calculated from 392 mapping points on a 180 mm wafer area.

**Figure 2.** Thickness uniformity mapping of Al2O3 (thermal ALD) on a 200 mm wafer measured after (**a**) 1156 ALD cycles without rotation, (**b**) 1000 ALD cycles with manual sample rotation by 180◦ after 500 ALD cycles.

#### *3.2. Antireflection Coatings on Plane Glass Substrates*

An AR design consisting of seven layers has been calculated using the thin film software OptiLayer (version 11.65e, OptiLayer GmbH, Garching, Germany) to reduce the residual reflectance of a fused silica substrate from approximately 3.5% to less than 0.5% in the visible spectral range from 400 to 750 nm. Silicon dioxide was chosen as final layer, as its low refractive index will significantly improve the performance of the AR coating. ALD oxide films are typically amorphous, especially when deposited at low temperatures [40]. However, TiO2 ALD thin films tend to crystallize at moderate deposition temperatures. The growth of crystallites leads to high surface roughness and, as a result, strong scattering of light. The surface roughness significantly increases for film thicknesses greater than about 40 nm [41]. The crystallization can be inhibited by inserting a thin Al2O3 interlayer [42]. In the first design AR-D1 (Table 3) this interlayer was not included into the design, whereas experimentally, the thick 63.9 nm TiO2 has been split in two thinner TiO2 layers by introducing a 1.5 nm thin Al2O3 interlayer to inhibit the crystallization.

The AR coating was first tested on a plane substrate. By applying the AR-D1 coating to a fused silica glass sample, the reflectance could be reduced to an average reflectance of 0.3% in the visible spectral range from 400 to 750 nm, see Figure 3a. Comparing the reflectance spectra, the AR-D1 coating shows a deviation from the AR-D1 design. It was found that the misfit between design and coating has two origins. First, the 1.5 nm thin Al2O3 layer needs to be taken into account when designing the AR coating. This presumption is based on the good agreement of the measured spectrum to calculated expected one that includes the interlayer, see Figure 3b. Thin ALD layers are well known to be very dense and pinhole-free and are intensively investigated for barrier coating [43]. Therefore, the reflections at the interfaces of this ultra-thin layer must be considered in the optical design.


**Table 3.** Designed layer thickness and necessary ALD cycles of AR coating on fused silica.

Second, a recalculation of the actual thicknesses from the measured spectra using the Film Wizard™ software (version 8.5.0, Scientific Computing International, Carlsbad, CA, USA) showed that Al2O3 layers on TiO2 are thinner as expected. The GPC on the underlying TiO2 is only 1.17 Å/cycle instead of 1.21 Å/cycle on Si or fused silica. Also, SiO2 thin films have a lower GPC on the underlying TiO2 films of only 1.17 Å/cycle instead of the expected 1.20 Å/cycle. Altered growth rates on different substrates have been repeatedly observed and are possibly a reason of different OH group concentrations or irregular OH group distributions on the underlying surface [17].

The film thickness deviation has been 0.4 and 0.6 nm for the alumina layers and approximately 2 nm for silica. This deviation in film thicknesses results in slight deviation of the measured curve (coating AR-D1) and the corrected design curve in Figure 3b. Note that no in situ control of the film thicknesses has been applied during the ALD process. In situ monitoring might be necessary for more complex AR coatings or interference coatings such as narrow bandpass filters or dichroic mirrors [20,21].

*Coatings* **2017**, *7*, 118

A second AR coating AR-D2 was designed including the Al2O3 interlayer. Furthermore, the adapted GPC values were used for calculating the necessary ALD cycles of Al2O3 and SiO2 layers on TiO2, see Table 2. By applying these two corrections, the reflectance of the design and the coating are in an excellent agreement for a sample that was placed in the center of the substrate table, see Figure 4a.

**Figure 3.** Reflectance spectra of AR-D1 on fused silica reference glass substrate: (**a**) Design and coating; (**b**) Corrected design (with interlayer) and recalculation from measured spectra (taking the Al2O3 interlayer into account).

**Figure 4.** Measured and expected reflectance spectra of AR-D2 on fused silica reference glass substrate positioned: (**a**) at the center of the substrate table; (**b**) at approximately 75 mm from the center table.

As the thickness non-uniformity was expected to be the main origin of errors, a worst-case analysis was performed, whereas the maximum allowed thickness deviation was specified as the expected NU% of each material, see Table 2. The area between the dotted lines in Figure 4b indicates the worst-case error corridor of the calculated maximum possible deviations from the theoretical reflectance spectra. To estimate the influence of the NU experimentally, next to the fused silica substrate (sample 1) that was placed in the middle of the substrate table, a second substrate (sample 2) was positioned at approximately 75 mm from the center of the table during the deposition. The measured reflectance spectra of sample 2 lies within the worst-case error corridor, indicating that the small deviations to the AR design are most likely a consequence of the lateral film thickness non-uniformity on the substrate table.

#### *3.3. Antireflection Coatings on a Half-Ball Lens*

The antireflection coating AR-D2 was applied to a hemispherical lens. The refractive index of the lens was calculated from the measured reflectance spectra of the uncoated half-ball lens, which is slightly higher than the reflectance of the fused silica glass substrate, see Figure 5a. Due to the higher effective refractive index of the bare lens, the appearance of the expected and measured AR spectra on the lens differs from the spectra on the coated glass slab (compare Figures 4a and 5b). The measured spectra of the AR coating on the lens is in good agreement with the adapted AR design (Figure 5b).

**Figure 5.** (**a**) Measured reflectance spectra of uncoated fused silica reference glass substrate and uncoated fused silica half-ball lens; (**b**) Measured reflectance spectra (AOI = 0◦) of AR-coated fused silica half-ball lens at different positions of the lens.

It should be emphasized that the reflectance spectra are consistent at all positions on the lens. Hence, the AR coating was deposited uniformly on the hemispherical lens without any complex equipment to control the layer thickness.

An upright-positioned glass sample was used as reference sample for the edge of the glass plate since it is not possible to measure the reflectance at the very edge of the lens. As shown in Figure 6, the measured reflectance is in very good agreement with the design. The deposition occurs simultaneously on both sides of the glass sample and the measured spectra are identical on both sides of the substrates. The results show that the ALD-technology is not restricted to the radius of curvature.

**Figure 6.** Measured and expected reflectance spectra (AOI = 0◦) of AR-D2 deposited simultaneously on both sides of an upright-positioned flat fused silica glass substrate.

The AR performance of the coated lens depends on the position in the chamber due to the lateral thickness non-uniformity. Hence, it has been possible to obtain an excellent AR coating on a curved lens matching very well the design curve, see Figure 7.

**Figure 7.** Measured and expected reflectance spectra (AOI = 0◦) of AR-coated fused silica half-ball lens at different positions of the lens.

#### *3.4. Antireflection Coating for an Aspheric Lens*

To confirm that ALD AR-coatings can be also used to reduce reflection losses of larger lenses, a second antireflection coating was applied to a steeply curved aspheric lens with center height of 25 mm and a diameter of 50 mm. Ta2O5 was used as high-index material for the AR coating, since the grown Ta2O5 ALD thin films are amorphous and hence no additional Al2O3 interlayer is needed to inhibit crystal growth. TEM and SEM images of about 5 nm, 35 nm and 200 nm Ta2O5 thin films show an amorphous structure [33,44,45]. X-ray diffraction (XRD) measurements also confirmed the amorphous nature of 200 nm tanatala thin films grown from Ta(OEt)5 at 300 ◦C. These spectra are not shown here.

The glass lens has a refractive index that is similar to that of B270. An AR-D3 coating (see Table 4) was designed to reduce the reflectance of a B270 substrate from approximately 4.0% to less than 0.5% in the visible spectral range from 400 nm to 700 nm. The first part of the coating design is based on the patented AR-hard® (Jena, Germany). A thin high-index layer is sandwiched by two thicker lower-index layers forming a symmetrical stack of three-quarter-wave optical thickness [46]. Silicon dioxide was chosen as final layer to attain a low residual reflectance. After completion of the Al2O3/Ta2O5 sequences in the Sunale R200 tool at a deposition temperature of 300 ◦C, the samples were unloaded to atmosphere and transferred to the OpAL tool for further processing of the top SiO2 layer at 100 ◦C.

**Table 4.** Designed layer thickness and necessary ALD cycles of AR coating on B270.


Figure 8 depicts the reflectance of the AR-D3 design and the AR-D3-coated lens. The reflectance spectra of the lens show a good match to the design. Minor deviations between design and the measured reflectance at the inclined surface of the lens (position A and E) may be attributed to a temperature gradient of the lens during deposition and to lateral thickness non-uniformity across the chamber.

**Figure 8.** Measured and expected reflectance spectra (AOI = 0◦) of the AR-coated steeply curved B270 aspherical lens at different positions of the lens.

#### **4. Conclusions**

Atomic layer deposition successfully applies to deposit antireflection coatings on strongly curved lenses. In particular, the average reflectance could be minimized to 0.3% for a fused silica half-ball lens with 4 mm diameter and a steeply curved B270 aspherical lens in the visible spectral range from 400 to 750 nm and 400 to 700 nm, respectively. Similar reflectance spectra across the entire lens surface at normal light incidence are a result of the very good conformality of ALD coatings. The good agreement between design and coatings confirms the precise thickness control of ALD thin films. Thickness monitoring was not necessary to reach the desired film thicknesses, but only the counting of ALD cycles. Moreover, it was demonstrated that the conformal deposition is not restricted to the radius of curvature of a lens, as an AR coating that was deposited simultaneously on both sides of a flat glass substrate showed identical spectra on both sides. Noteworthy, these antireflection coatings are demonstrated in two commercially available tools with significantly different configurations, indicating that ALD can become highly attractive for production purposes.

Further development of ALD coating equipment such as spatial ALD, atmospheric pressure ALD, and batch tools will increase the applicability of this technology for high volume applications. The slow deposition rate is considered as the main disadvantage of ALD. The long deposition times are generally the consequence of the required purge times between the precursor pulses. The possibility to perform double-sided coatings increases the throughput of this coating technology. Spatial ALD [47] is a promising approach to shorten the purge times, in that the substrate is moved to different precursor zones, hence precursor pulses are spatially separated and purge steps become dispensable. The use of batch coaters is another possibility to increase the throughput [16]. However, the lateral thickness uniformity needs to be improved for scale up to large-area. For a better uniformity, both the chamber design and the precursor chemistry needs to be improved. The development of precursors that are highly reactive and volatile, but at the same time thermally stable and non-corrosive, as well as the design of a tool, that comprises a uniform gas distribution, a homogeneous temperature and the absence of dead volumes remains a future challenge [39].

Although further research and developments are needed, ALD is a promising method to deposit optical thin films that can be prospectively applied for optical coatings on complex formed optical components due to the very good conformality of ALD coatings (convex and concave lenses, cylinders, ball lenses, etc.).

*Coatings* **2017**, *7*, 118

**Acknowledgments:** The research was supported by the Deutsche Forschungsgemeinschaft (DFG) (Emmy-Noether-Project SZ253/1-1) and the European Space Agency (ESA) (Contract No. 4000109161/13/NL/RA). This work was partially supported by the FhG Internal Programs under Grant No. Attract 066-601020. Kristin Pfeiffer thanks the Carl Zeiss Foundation for promoting her doctoral research studies. The authors gratefully acknowledge David Kästner for the micro-spectrophotometer measurements.

**Author Contributions:** Adriana Szeghalmi and Kristin Pfeiffer conceived and designed the experiments; Ulrike Schulz supported the design of the coatings; Kristin Pfeiffer performed the experiments, analyzed the data and wrote the paper. All authors critically revised the article.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2017 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Continuous Tip Widening Technique for Roll-to-Roll Fabrication of Dry Adhesives**

#### **Sung Ho Lee 1, Hoon Yi 2, Cheol Woo Park 1, Hoon Eui Jeong 2,\* and Moonkyu Kwak 1,\***


Received: 23 August 2018; Accepted: 28 September 2018; Published: 30 September 2018

**Abstract:** In this study, we reported continuous partial curing and tip-shaped modification methods for continuous production of dry adhesive with microscale mushroom-shaped structures. Typical fabrication methods of dry adhesive with mushroom-shaped structures are less productive due to the failure of large tips on pillar during demolding. To solve this problem, a typical pillar structure was fabricated through partial curing, and tip widening was realized through applying the proper pressure. Polyurethane acrylate was used in making the mushroom structure using two-step UV-assisted capillary force lithography (CFL). To make the mushroom structure, partial curing was performed on the micropillar, followed by tip widening. Dry adhesives with properties similar to those of typical mushroom-shaped dry adhesives were fabricated with reasonable adhesion force using the two-step UV-assisted CFL. This production technology was applied to the roll-to-roll process to improve productivity, thereby realizing continuous production without any defects. Such a technology is expected to be applied to various fields by achieving the productivity improvement of dry adhesives, which is essential for various applications.

**Keywords:** dry adhesive; biomimetics; continuous process; partial curing

#### **1. Introduction**

Microstructure-based dry adhesives, which were inspired by the feet of gecko lizards and beetles, have attracted attention in various applications due to their strong adhesion, repeatability, reversibility, and self-cleaning properties [1–13]. They are applied to a transfer or fixing device, such as a glass substrate and silicon wafer, and an attempt has been made to replace the existing electrostatic or vacuum chuck [8,14]. These dry adhesives are typically made from polymeric pillars of the size of a few hundred nanometers to tens of micrometers, and the tip of the pillar is shaped like a spatula or mushroom to maximize its adhesive properties [15–17]. Generally, dry adhesive production method is a molding technique, such as soft lithography; thus, manufacturing the aforementioned size structure is easy [4]. However, constructing a precisely designed tip shape on a master can be a complex process [4,8,18]. For example, in the case of producing mushroom-like microstructure dry adhesives on a master, the tip must be formed inside the silicon wafer to form a wide tip on the surface of the dry adhesive when replicated. This tip plays an important role in securing the adhesion; this role has been confirmed in studies that handle the comparison of adhesive strengths according to various tip shapes [15]. Based on the literature, a dry adhesive with a thin and wide area tip has a high adhesive strength, and a complicated process is required to produce a master for such a dry adhesive. An example of fabricating a master with a wide tip is to use the footing effect that occurs in the etch stop layer during deep RIE (Reactive Ion Etching) process [19]. If etching is performed to a

thickness over the working depth simply by using the SOI (Silicon on insulator) wafer, then the etching ion does not proceed in the depth direction in the SiO2 layer but is etched laterally. This process is called the footing effect, in which forming a wide tip on a master is relatively easy. However, this method has a disadvantage; it cannot precisely control the thickness or size of the tip. To solve this problem, the tip precisely fabricated on the wafer surface through the surface micromachining and the tip buried masters are manufactured by fusion bonding between the surface-machined wafers and bare wafer [5,8,18]. The fabricated master mold can be used to produce a dry adhesive after a simple release layer treatment; however, it is difficult to apply in a continuous production using a roll-to-roll process due to poor demolding (e.g., tip tearing depending on the polymer used) [20]. Particularly, in the roll-to-roll process, which can only use a flexible polymer mold, the breakage of the tip during demolding becomes evident. Therefore, the productivity should be improved through new methods. In the present work, we introduce an advanced UV molding technique called two-step UV-assisted capillary force lithography (CFL) to produce a mushroom-shaped dry adhesive without special demolding failure. This method can be applied to the roll-to-roll-based apparatus without any changes; thus, it is applied to the prototype roll-to-roll apparatus to test the mass production possibility. Moreover, the pull-off strength of the fabricated dry adhesive is measured. The measured adhesive strength of approximately 9.5 N/cm2 is slightly lower than that of dry adhesives made by typical one-step molding [8]; nevertheless, this value is appropriate for application. In addition, higher adhesion can be achieved if the tip size can be widened in a future optimization process. The study of tip widening using partial curing has been experimentally proven in previous studies [21]. In this paper, we intend to implement this process continuously.

#### **2. Materials and Methods**

Polyurethane acrylate (PUA): PUA is an ultraviolet ray curable material composed of prepolymer with acrylate group, a photoinitiator, a crosslinker and a release agent. The PUA was dropped onto a master mold with pillar shaped microstructures, which were fabricated by surface micromachining. A PET (polyethylene terephthalate) film as supporting layer was covered on the dropped PUA, followed by UV exposure for 40 s (wavelength = 250–400 nm, dose = 100 mJ/cm2). After UV exposure, the PET film was peeled off from the master with cured PUA structure.

Partially cured micropillar: Our fabrication method was based on the two-step process of capillary UV molding and additional tip shape modification. For the fabrication of PUA partially cured micropillars, the polydimethylsiloxane (PDMS) mold with micro holes was used with enough window for partial curing time. Then, the PUA resin was partially cured by UV exposure for 15–20 s (wavelength = 365 nm, intensity = 100 mW/cm2).

Tip widening: The film was placed on the top of the partially cured pillar, and further curing was performed for 30 s while applying the appropriate pressure of 20–30 kPa (wavelength = 365 nm, intensity = 500 mW/cm2). For a uniform pressure distribution, a thin PDMS block was placed as a buffer on top of the PET film prior to the application of pressure.

Measurement of pull-off force: We measured pull-off force and durability with custom built equipment [8]. To measure normal adhesion force between fabricated dry adhesive and target smooth substrate, the smooth substrate was made with a glass sample and was moved vertically with speed of 3 mm/s by a step motor connected crank. The device is equipped with a load cell capable of measuring the load in the z-axis direction, allowing measurement of the pull-off force including the preload. Pull-off forces were measured for various preloads, and marathon tests were conducted at the speed of 50 cycles per minute and 40 N/cm2 preload.

#### **3. Results and Discussion**

Figure 1 shows the schematic of the two-step UV-assisted CFL and that of a continuous production apparatus. PDMS is a known porous material. In this study, PDMS was used to induce partial curing and fabricate hierarchical structures. In the case of the PDMS mold, the surface of the molded polymer

film was exposed to air that was permeated through a porous PDMS mold. The trapped or permeated oxygen inhibited UV curing by scavenging radicals generated from the photoinitiator by UV [22]. Thus, the surface of the PUA resin in contact with the air remained tacky, whereas the resin beneath the surface cured completely. Some pre-dissolved oxygen existed in the liquid resin. This oxygen was rapidly consumed under UV exposure due to high mobility and reactivity of oxygen with a large number of initial radicals. Therefore, the formation of a tacky surface could be attributed to the diffusion of oxygen that was trapped in the mold cavity or permeated out of the mold. In a previous research on two-step CFL, micro/nano hierarchical structure was fabricated using a nano-structured mold after the microstructure fabrication [22]. A microstructure was fabricated, and a wide tip was then formed on the partially cured microstructure by pressurization using a flat substrate. The tip of the microstructure was widened by simply using a glass substrate, and the low adhesion between glass and PUA enabled the production of a mushroom structure without any surface treatment. After the process proof at the wafer level, this principle was used to realize continuous production using a roll-to-roll apparatus. The roll-to-roll apparatus was divided into microstructure and tip production sections. In the microstructure production section, a PDMS negative mold was attached to the roll, whereas the tip production section is composed of a simple quartz cylinder and a urethane roll. The partial curing phenomenon in the microstructure fabrication was theoretically predictable. On the basis of the literature, the oxygen saturation inside the polymer resin, which depends on the depth of the resin from the surface, can be expressed as [22]:

$$\frac{\mathbb{C}\_{\text{O}\_2}(\text{x})}{\mathbb{C}\_{\text{O}\_2, \text{surface}}} = \cosh\left(k / D\_{\text{O}\_2/\text{polymer}}\right)^{1/2} \text{x} - \tanh\left(k / D\_{\text{O}\_2/\text{polymer}}\right)^{1/2} L \cdot \sinh\left(k / D\_{\text{O}\_2/\text{polymer}}\right)^{1/2} \text{x}$$

where *C*O2 is the oxygen concentration in the cavities, *D*O2/polymer is the oxygen diffusivity coefficient in the polymer layer, *C*O2\_surface is the surface concentration of oxygen, *L* is the depth of the cured polymer resin, *x* is distance from free-surface between air and resin. When the appropriate constants were substituted for this equation (*<sup>k</sup>* = 1, *<sup>D</sup>*O2/polymer = 10−<sup>12</sup> <sup>m</sup>2·s−1), the penetration of oxygen during the microstructure fabrication of PUA resin proceeded from the resin surface to approximately 4–5 μm. That is, the upper 4 μm of the microstructure remained in a state capable of subsequent patterning while still maintaining a non-cured tacky surface. The tip structure to be fabricated through this study was expected to express sufficiently by this partial curing process, in which the diameter was approximately 1 μm larger than the micropillar and the thickness was approximately 500 nm. To fabricate an appropriate partially cured micropillar, the intensity of UV LED at 365 nm wavelength was adjusted to 100 mW/cm2 and exposed for a suitable time (15–20 s) to fabricate a microstructure whose surface remained tacky.

Thereafter, various pressures were applied to the tip widening. As a result, suitable pressure (25 kPa) was identified, and the mushroom structure was constructed. As shown in Figure 2, an extremely low pressure (10 kPa) did not sufficiently expand the tip, and an extremely high pressure (40 kPa) caused columnar deformation or collapse. At lower pressures, the not fully cured resin is irregularly wetted, clumped like contaminants, or wicked down the pillar to create an unusual pillar shape. (Figure 2a) To solve this problem, a structure with a wide tip is successfully fabricated by making perfect contact with high pressure. The column part is already cured and fixed to its own shape differently from the tacky surface of the tip part, but since it is not over-cured, it is possible to deform a little by strong pressure and exposure so that the upper part becomes a little wider column structure after the pressure applying process. The production speed is determined by the type of material, the time to produce partial curing, and the time required to derive the structural change through pressure. In this work, production speed was about 100 cm2/min. We expect this result to be improved through future optimization studies.

**Figure 1.** Schematic of the continuous fabrication process for mushroom-shaped dry adhesives and procedure for mushroom-shaped structures via two-step UV molding process.

**Figure 2.** SEM images of mushroom structures fabricated with various pressures: (**a**) 0 kPa, (**b**) 10 kPa, (**c**) 25 kPa.

After the appropriate pressurization, the mushroom structure was fabricated through second exposure, and the pull-off strength of the sample was measured by a typical adhesion measurement. The preload was fixed to 40 N, and the repeated adhesion test was performed using a load cell. We predicted the life of fabricated dry adhesive through the analysis of the tendency of decrease in adhesive force by using a marathon test equipment made from a simple crank system. The marathon test was conducted in a normal laboratory environment and not in a clean room; thus, various contaminants may be found.

Figure 3 shows the pull-off strength measurement results of the prepared dry adhesive samples. The measured adhesive strength was lower than that of the mushroom-shaped wide-tip microstructures (~20 N/cm2) produced by conventional methods because forming a wide and thin tip necessary for strong adhesion was impossible, which may be solved by optimization of the subsequent process.

**Figure 3.** Measurement data of the pull-off strength of fabricated dry adhesives as a function of preload.

Selection of the resin, optimization of the primary curing time and UV intensity, the amount of pressure applied during the secondary curing, and the curing time should be considered for the production of a wide tip. Figure 4 shows the results of the marathon test for the durability of dry adhesives. The adhesives initially showed a maximum adhesive strength of 9.5 N/cm<sup>2</sup> and a 10% decrease in adhesion after repeated use of 5000 times. As shown in Figure 4, numerous contaminants were found on the dry adhesive surface observed after 5000 uses, and several defects were also found. Although not many, defects were present where the tip portion was torn out or the microstructures were matted together. Usually, the tip is torn rather than the column is broken. This can be seen as fatigue failure and breakage at the tip connection part where the greatest stress concentration occurs. After using the test for 5000 cycles, the surfaces of the dry adhesive and the substrate were cleaned using a commercial pressure sensitive adhesive tape. Subsequently, the recovered adhesive strength was 94% of the original adhesive strength. As shown in Figure 4, all of the contaminants, except for defects, were removed through the cleaning process, and some of the matting was recovered during the cleaning process using the adhesive tape, resulting in a slight increase in the adhesion. The subsequent 5000 additional adhesion tests resulted in the same level of adhesion of the contaminants and additional defects. However, the degradation tendency was considerably better than that of dry adhesives with conventional wide tips. Adhesive strength and lifetime are inversely related to each other; thus, the fabricated dry adhesives are likely to be fully utilized in applications where they have to be used for a long period of time with proper adhesive strength. The marathon test result shows reasonable adhesive strength of up to 10,000 times, and the test is expected to be used in various fields.

**Figure 4.** Durability test of the fabricated dry adhesives over 10,000 cycles of attachment and detachment. Inset microscopic images for dry adhesive surfaces at initial state, 5000 cycles before cleaning, 5000 cycles after cleaning, and 10,000 times usage.

#### **4. Conclusions**

We developed a continuous production technology of mushroom-shaped dry adhesives by utilizing the two-step UV-assisted CFL. The mushroom shape was produced by continuously producing the micropillars and by applying pressure to the partially cured micropillars, the tip portion was widened, and the fabricated dry adhesives exhibited an adhesive strength of 9.5 N/cm2 on the glass substrate. In addition, the dry adhesives were excellent in terms of durability due to their shape characteristics and exhibited an adhesive strength of approximately 80% even in a marathon test of 10,000 times. The complicated mold-making process was necessary in the conventional continuous production technique. However, this process could be omitted to reduce the production cost and increase production efficiency. Furthermore, if the productivity is secured with the performance of a reasonable dry adhesive, then it can be applied to various application fields.

**Author Contributions:** Conceptualization, M.K.; Methodology, H.E.J. and M.K.; Experiments, H.Y. and S.H.L.; Writing-Original Draft Preparation, S.H.L. and M.K.; Writing-Review & Editing, M.K., H.E.J. and C.W.P.; Supervision, M.K. and H.E.J.; Funding Acquisition, M.K.

**Funding:** This work was supported by the National Research Foundation of Korea (2016R1A2B4007858) funded by the Korean Government (MSIP).

**Acknowledgments:** The authors thank Jeong Hyeon Lee for his technical support.

**Conflicts of Interest:** The authors declare no conflicts of interest.

#### **References**


© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*
