**Combining Thermal Spraying and Magnetron Sputtering for the Development of Ni**/**Ni-20Cr Thin Film Thermocouples for Plastic Flat Film Extrusion Processes**

#### **Wolfgang Tillmann 1, David Kokalj 1,\*, Dominic Stangier 1, Volker Schöppner <sup>2</sup> and Hatice Malatyali <sup>2</sup>**


Received: 26 August 2019; Accepted: 20 September 2019; Published: 24 September 2019

**Abstract:** In the digitalization of production, temperature determination is playing an increasingly important role. Thermal spraying and magnetron sputtering were combined for the development of Ni/Ni-20Cr thin film thermocouples for plastic flat film extrusion processes. On the thermally sprayed insulation layer, AlN and BCN thin films were deposited and analyzed regarding their structural properties and the interaction between the plastic melt and the surfaces using Ball-on-Disc experiments and High-Pressure Capillary Rheometer. A modular tool, containing the deposited Ni/Ni-20Cr thin film thermocouple, was developed and analyzed in a real flat film extrusion process. When calibrating the thin film thermocouple, an accurate temperature determination of the flowing melt was achieved. Industrial type K sensors were used as reference. In addition, PP foils were produced without affecting the surface quality by using thin film thermocouples.

**Keywords:** nickel–chromium; thin film thermocouples; physical vapor deposition; flat film extrusion; foil quality

#### **1. Introduction**

Self-learning machines play a crucial role in today's and future production. In the 4th Generation Industrial Revolution (Industry 4.0), machines are digitally upgraded and merged to a big data environment [1]. The overall goal is the improvement of product quality and production scheduling by building up and using prediction tools [1]. However, self-learning machines are still far from being implemented in many industries [1]. Data required for these prediction tools can only be generated if the machines are equipped with corresponding sensors, such as temperature-, wear-, distance-, or pressure-sensors, that work exactly. In many areas, the determination of the temperature, in particular, plays an important role. For this purpose, especially thin film thermocouples are suitable, since they reveal a lower mass and thus a faster reaction time in contrast to bulk thermocouples [2]. Multi-functional PVD coatings are successfully used for the temperature measurement in combustion engines [3] or gas turbine engines [4], for example. However, there is also a need for thin film thermocouples in production processes. Therefore, different thin film thermocouples are currently developed for several applications, such as welding [5] or metal [6–9] and polymer processing [10,11]. However, the thin film thermocouples must be adapted to the respective application fields. On the one hand, the materials of the thermocouples need to be selected according to the desired temperature range. According to DIN EN 60584, eight different thermocouple pairs are available, whereas Fe–CuNi [12],

Cu–CuNi [13], NiCr–Ni [6–9] and Pt-10%Rh/Pt [14] are the most reported thin film thermocouples deposited by means of PVD. Apart from the standardized thermocouples, TiC–TaC [15], Al–Au [11], Ni–Cu [12,16], Ni–Fe [12], Cu–Fe [12], Chromel–Alumel [5,12], and Pt–Pd [4] thin film thermocouples are synthesized. On the other hand, depending on the application, the thin film thermocouples need to be protected against wear or oxidation processes by means of a suitable tribological cover layer. To enhance the wear resistance of thin film sensors, protection layers consisting of Al2O3 [7,14,16,17], AlN [17], or TiN [8,18] are mainly used. For electrical isolation purposes, HfO2 [6] and Al2O3 [16] layers are utilized.

In plastic processing, thermoplastics are extruded into semi-finished products that can be processed further (e.g., films). The plastic is plasticized and formed in the die [19]. For a good product quality, the melt needs to flow through the die at constant pressure, throughput and temperature [20]. In particular, an accurate measurement of the melt temperature is preferred and helps to increase the productivity. Temperature sensors are part of every extruder system, so that the time stability and the absolute level of the melt temperature can be monitored. Malfunctions can be detected quickly and the reaction time to undesired temperature changes increases. Nevertheless, state-of-the-art touching sensors cannot be used in the die itself because of the negative impact of the sensors on the product quality. The measurement tip protrudes at different depths into the melt and can disturb the homogeneity in the process itself. Thin film thermocouples offer an alternative measuring method to conventional sensors due to their flat construction and low net weight. In addition, they offer the possibility to measure the temperature even when embedded in the wall. Especially, in the case of embedded sensors, it is important to measure the temperature of the melt and not of the die wall. Therefore, thin film thermocouples are to be thermally insulated from the die walls. Thermal sprayed barrier coatings, such as ZrO2 and Al2O3, are particularly suitable for thermal insulation applications since they reveal pores and a higher thickness compared to sputtered coatings. Therefore, these coatings were applied by means of atmospheric plasma spraying and serve as substrate for the thin film thermocouples [19,20]. In particular, in film extrusion, it is important that the melt flow is not impaired to ensure a homogeneous film quality. Accordingly, the influence of different PVD top coatings on the wall friction and the adhesion of the melt is investigated. Since AlN coatings reveal high electrical resistance and high thermal conductivity, three AlN thin films deposited with different sputter parameters are selected. As an alternative, BCN is used as additional top coat. Finally, the application of the Ni/Ni-20Cr thin film thermocouple is proved in a real operation test und compared to a reference bulk thermocouple.

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

#### *2.1. Deposition Process*

For the deposition of the nitride thin films and the Ni/NiCr thin film thermocouples, the industrial scale magnetron coating system CC800/9Custom (CemeCon AG, Würselen, Germany) was utilized. The AlN and BCN thin films were synthesized on AISI 1045 steel substrates (Ø40 mm), which were prior coated with Al2O3 using atmospheric plasma spraying. Detailed spray parameters of the Al2O3 coating are reported in [21]. The nitride thin films were deposited in medium frequency (MF) mode using one B4C (99.50% purity, Sindlhauser Materials GmbH, Kempten, Germany) and two Al (99.50% purity, CemeCon AG, Würselen, Germany) targets operating with 3150 W and 2× 1875 W at a total chamber pressure of 330 mPa. The frequency was set to 50 kHz with a duty cycle of 50%. Three AlN thin films and one BCN thin film were synthesized using different heating powers and Ar/N2 gas flow rates according to Table 1. The heating power was changed to influence the crystallinity and the topography of the AlN thin films. Changing the Ar/N2 gas flow ratio results in a change of the chemical composition of the coating. Therefore, it was possible to analyze which properties influence the melt flow, respectively, the friction behavior. The two-fold rotation system was biased with –150 V in MF mode during all deposition processes. The deposition time of all coatings was adjusted to achieve

comparable thin film thicknesses of approximately 1200 nm. The thickness was measured analyzing the cross section of the coatings by means of scanning electron microscopy.

**Table 1.** Deposition parameters of the AlN and BCN thin films. **Coating AlN-1 AlN-2 AlN-3 BCN-1**


In addition to the reference nitride layers, Ni/Ni-Cr thin film thermocouples were synthesized for the temperature measurement of the plastic melt in flat film extrusion processes. Therefore, a modular tool system was developed. This system consisted of a screw-in bush (AISI 4140), a ceramic insert, and a locking nut. The bush can be integrated into the die by means of a drilling with a thread (M24 × 1.5). The ceramic insert was coated with the thin film thermocouples. By means of a step on the one side and a locking nut on the other side, the thin film thermocouple was positioned and fixed in the bush. The ceramic insert was composed of a steel core (AISI 1045), which was coated by atmospheric plasma spraying from all sides. For a sensitive temperature measurement, the insert had to be thermally insulated from the die walls, since the temperature of the on-rushing plastic melt was measured. The thermal insulation was achieved by a thermal sprayed multilayer design consisting of a 120-μm-thick NiCoCrAlTaY (AMDRY 997, Sulzer Metco, Pfäffikon, Switzerland) bond coat, a 530-μm-thick 7Y2O3-ZrO2 (AMPERIT 817.7, HC Starck, München, Germany), and a 680-μm-thick Al2O3 (Metco 6062, Sulzer Metco, Pfäffikon, Switzerland) ceramic coating. After the coating processes, the Al2O3 coating was entirely ground to a thickness of 350 μm, resulting in a total coating thickness of 1000 μm. After the grinding process, the outsides were polished in several steps to a roughness of *R*<sup>a</sup> = 0.212 ± 0.104 μm.

The step of the ceramic insert was manufactured using a micro milling machine HSPC 2522 (Kern, Eschenlohe, Germany) equipped with a solid carbide end mill, type 910 Marlin, with a corner diameter of 1 mm (Zecha, Königsbach-Stein, Germany). The milled surfaces reveal a roughness of *R*<sup>a</sup> = 0.811 ± 0.116 μm.

After processing the final dimensions of the ceramic inserts, they were cleaned of contaminations by means of ethanol in an ultrasound bath. Thereafter, the Ni/Ni-Cr thin film thermocouple was deposited on the ceramic inserts using a steel masking system reported in [22] to synthesize the individual Ni and Ni-20Cr conducting paths. The conducting paths were sputtered in DC mode using a heating power of 1200 W and a Ar/Kr gas flow rate of 2.7 at a chamber pressure of 250 mPa. For the Ni conducting path, two Ni targets (99.99% purity, Sindlhauser Materials GmbH, Germany) were operated at 1875 W in DC mode, whereas for the Ni-20Cr path, a Cr target (99.95% purity, Sindlhauser Materials GmbH, Kempten, Germany) was operated at 1035 W, additionally. The deposition time of both conducting paths was adjusted to reveal similar thicknesses of 1050 ± 35 nm.

#### *2.2. Characterization*

The structural properties of the as-deposited nitride thin films were analyzed by means of X-Ray Diffraction (XRD) utilizing the diffractometer D8 Advance (Bruker, Madison, WI, USA). The thin films were investigated using Cr-radiation (2.29106 Å), whereas the current and voltage were set to 40 mA and 35 kV. To avoid an overlapping of the reflexes from the Al2O3 substrate, detector scans were performed, whereby the tube was set to an incident angle of 5◦. The topography and morphology of the AlN and BCN thin films were investigated using the Scanning Electron Microscope (SEM) JSM 7001F (Jeol, Tokyo, Japan). The roughness *R*<sup>a</sup> of the Al2O3 substrate, as well as the thin films, was measured by the confocal white-light microscope μSurf (NanoFocus, Germany). The mechanical properties hardness (H) and Young's modulus (E) were examined by means of a nanoindentation test using the nanoindenter G200 (Agilent Technology, Santa Clara, CA, USA). The evaluation was performed in a depth between 100 and 400 nm using the equations in accordance to Oliver and Pharr [23]. Thereby, a constant Poisson's ratio of 0.25 was assumed. The adhesion of the nitride thin films under plastic deformation was tested by a Rockwell indentation test in accordance with DIN EN ISO 26443 [24], using a force of 60 kgf and a dwell time of 4 s. Tribological investigations were carried out using a high-temperature Ball-on-Disc tribometer (CSM-Instruments, Peseux, Switzerland). The velocity was set to 0.1 m/s, the normal force to 10 N, and the radius to 8 mm with a total distance of 20 m per experiment. Counter body balls (Ø6 mm) made of Polypropylene (PP) were used, since this material was used for the operation test of the thin film thermocouples. The tribological investigations were carried out at 85 ◦C (Vicat Softening Temperature) and 153 ◦C (Heat Deflection Temperature) to simulate a non-steady-state operating point of the plastic flat film extrusion process during heating up or cooling down. The adhesion of the PP to the coating surface was investigated by means of SEM. Additionally, the wear of the PP balls was analyzed using a confocal light microscope type InfiniteFocus (Alicona, Raaba/Graz, Austria). By utilizing a High-Pressure Capillary Rheometer (HPCR), the influence of the different coatings on the melt flow was investigated. The measurement was carried out according to DIN 54811. When measuring with the HPCR, first, a purely thermal melting takes place in a pre-chamber. In order to melt the material completely, the definition of a residence time is important. For the investigations with the Moplen HP420 M, a duration of 4 min was specified. After melting, a piston pushed the melt through the tempered capillary. The capillary was preheated to a certain temperature by two thermal sensors. The operating temperature for the measurements was set to 230 and 250 ◦C. The capillary was preheated to the desired temperature by means of two sensors. The piston speed and the pressure loss of the melt flow were recorded simultaneously by means of a pressure sensor.

For the operational test, a wide-slot die was used, which allows to measure the melt temperature. The positioning of the thermocouples was important for the accuracy and reproducibility of measuring points. The thin film and industrial thermocouples were inserted near to the melt outlet from the die. The investigations were carried out using an extruder with a diameter of 45 mm and a typical three-zone screw for plasticizing (Battenfeld-Cincinnati Gmbh, Germany). The mass temperature measurements were carried out using a high density Polyethylen, while for the film production, a high flow polypropylene homopolymer (Moplen HP420M, lyondellbasell, Rotterdam, The Netherlands) was selected. Furthermore, for producing films, a chill roll unit was used (Collin GmbH, Maitenbeth, Germany). To compare the thin film sensors with conventional sensors of type K (Gneuss GmbH, Bad Oeynhausen, Germany), sensors with measuring tips (depth length) of 0, 0.5, and 1.5 mm were also utilized to measure the melt temperature and serve as reference. Besides the operational tests, the quality of the produced film was examined for evaluating the influence of the measuring tip of the industrial sensors. The optical properties of the film were determined by measuring the reflection and the transmission. An optical spectroscope HR2000+ (Ocean optics, EW Duiven, The Netherlands) enabled the measurement of the film quality. The measuring setup was in accordance with DIN 5036-3. Optical spectroscopy enabled a fast quality criterion for the evaluation of samples. In this method, a light beam with an intensity I0 irradiated the sample, while the exit intensity was recorded. The radiation sources covered a selected wavelength range, respectively, a certain spectrum. The intensity of the light depends on the wavelength λ and, accordingly, a function I(λ) of the light intensity [25]. The spectrum of the incident intensity was compared over the wavelength with the spectrum of the incident intensity. In the present study, the visual spectrum (VIS), which reveals wavelengths of approximately 380–780 nm, was used. When the light beam hits the matter, the light is distributed to different parts and the following output intensities are generated: The reflected IR, the transmitted IT, the scattered IS, and the absorbed intensity IA. These intensities form in sum I0 [25].

#### **3. Results**

#### *3.1. AlN*/*BCN Thin Films*

#### 3.1.1. Structure and Morphology

XRD patterns of the three AlN and the one BCN coatings synthesized with different deposition parameters are presented in Figure 1. For the AlN thin films, the positions of the reflections are 2θ ≈ 50.2◦, 2θ ≈ 54.7◦, 2θ ≈ 57.7◦, 2θ ≈ 77.4◦, 2θ ≈ 94.8◦, and 2θ ≈ 106.6◦, which correspond to the hexagonal AlN phase (JCPDS 25-1133). The reflections of the three AlN thin films reveal comparable intensities, showing a similar crystallinity. Nevertheless, slight differences can be detected. The AlN-3 thin film shows the highest crystallinity, which is favored by the higher nitrogen flow during the coating process. At constant gas flow, the AlN-2 film shows a slightly higher crystallinity compared to the AlN-1 film, which is due to the higher deposition temperature. Compared to the AlN-3 thin film, the AlN-1 und AlN-2 thin films reveal small additional peaks at 2θ ≈ 69.0◦ and 2θ ≈ 106.4◦, which correspond to the cubic AlN phase (JCPDS 46-1200). However, none of the AlN thin films show a high crystallinity, which is also shown by the broad peaks. Some of these also overlap with the comparable structure of the substrate, which makes accurate evaluation difficult. The wurtzite-type AlN was already obtained by Aissa et al. for a dc and high power impulse magnetron sputtering process [26]. The remaining reflections, marked with a black circle, belong to the cubic (α) and trigonal (γ) Al2O3 phase of the thermally sprayed substrate. Concerning the BCN coating, basically, only peaks of the substrate are visible, which indicates a nearly amorphous state of the thin film. Only slight reflections at 2θ ≈ 52.8◦ and 2θ ≈ 57.4◦ are observed, which coincide with the position of rhombohedral B4C.

**Figure 1.** XRD patterns of the AlN-1, AlN-2, AlN-3, and BCN thin films deposited with various parameters.

Additionally, the nitride thin films were analyzed regarding topography and morphology by means of SEM, as shown in Figure 2. It can be seen that the thermally sprayed Al2O3 coating, which was used as substrate for the thin films, reveals a fine topography accompanied by a small crack pattern in the polished state. The AlN-1 and AlN-2 thin films possess a coarse cauliflower-like topography. By contrast, the AlN-3 thin film shows a fine structured topography. Accordingly, a high Ar/N2 gas flow ratio of 6.0 and high heating powers of 5000 and 7000 W, as is the case for the AlN-1 and AlN-2

coatings, lead to a coarse topography. The structure of the topography can be correlated with the XRD pattern shown in Figure 1. The AlN-1 and AlN-2 thin films reveal a lower crystallinity compared to the AlN-3 thin film and a small amount of a second phase, which results in the coarser growth behavior. The BCN-1 coating reveals an even finer structure, which is comparable to the polished Al2O3 surface. The cross-sections of all thin films show a glass-like featureless structure.

**Figure 2.** SEM images of the topography (top) and morphology (bottom) of the AlN-1, AlN-2, AlN-3, and BCN thin films, as well as the thermally sprayed Al2O3 coating.

#### 3.1.2. Mechanical Properties

Using nanoindentation, the mechanical properties—hardness and Young's modulus—of the thin films, as well as the Al2O3 coating, were analyzed. As listed in Table 2, the Al2O3 coating reveals a hardness of 13.8 ± 3.9 GPa and a Young's modulus of 155.2 ± 31.9 GPa. The AlN-1 and AlN-2 thin films reveal the lowest hardness of 6.0 ± 1.6 GPa and 6.2 ± 2.0 GPa, respectively. A higher hardness of 13.6 ± 3.0 GPa is analyzed for the AlN-3 thin films, which reveals a finer structure compared to the AlN-1 and AlN-2 thin films. The highest hardness of 19.1 ± 2.3 GPa is obtained for the BCN-1 coating. With increasing hardness, the Young's modulus increases from 98.6 ± 21.2 GPa (AlN-1) to 187.7 ± 20.2 GPa (BCN-1). AlN thin films deposited on stainless steel substrates by means of magnetron sputtering reveal similar hardness values between 6 and 14 GPa in dependence on the bias-voltage as reported by Choudhary et al. [27]. The H/E ratio, which demonstrates the resistance to plastic deformation, is between 0.048 (AlN-2) and 0.102 (BCN-1). Since a higher value demonstrates a higher resistance to plastic deformation, the BCN-1 and AlN-3 thin films, as well as the uncoated Al2O3 coating, are the best choices as top layer for the application in plastic melt extrusion dies.

**Table 2.** Roughness and mechanical properties of the nitride thin films, as well as the Al2O3 coating.


Hence, the thin films were exposed to high pressures in the nozzles without abrasive particles occurring. The adhesion of the thin films was analyzed by means of a Rockwell indentation test, which represents the thin film behavior under plastic deformation. The corresponding light microscope images of the indents are shown in Figure 3. According to DIN EN ISO 26443, the adhesion of the thin films can be classified into four different classes. Thereby, class 0 represents a coating without cracks and local delamination, and class 3, full delamination of the coating after indentation. The thin films AlN-2 and BCN-1 are classified as class 1, since they reveal only slight cracks around the indent. AlN-1 and AlN-3 correspond to class 2, showing local delamination. Accordingly, the adhesion strength of the thin films decreases with decreasing heating power, respectively, deposition temperature. The

lowest adhesion is shown by the AlN-3 thin film, deposited at the lowest temperature and with the highest nitrogen gas flow.

**Figure 3.** Light microscope images of the Rockwell indents of the AlN-1, AlN-2, AlN-3, and BCN thin films.

#### 3.1.3. Tribological Properties

Polymer processing mainly TiN, CrN, TiAlN, Cr1−*x*Al*x*N, CrAlON, and DLC coatings deposited by means of PVD were investigated and used [28–30]. It was shown that the addition of Al to CrN increases the contact angle between the surface and Polycarbonate melt, resulting in a lower adhesion. Therefore, the AlN and BCN thin films were investigated regarding their tribological properties. Tribological experiments were conducted at 85 and 153 ◦C using PP counter balls, whereby the temperatures are the Vicat Softening Temperature and the Heat Deflection Temperature. These temperatures were selected to simulate the adhesion behavior of the PP melt to the different top layer-coated thin film thermocouples during heating up or cooling down processes of the thin film extrusion process. Different friction coefficients between the melt and the thin film thermocouple, as well as the uncoated extrusion die, can lead to inhomogeneous film qualities due to different friction forces. Since the extrusion die was made of AISI H11 (1.2343) steel, this steel was used as reference for the tribological experiments. In Figure 4, the friction coefficients of the AlN-1, AlN-2, AlN-3, and BCN-1 thin films, as well as the steel substrate and the Al2O3 coating, are shown for temperatures of 85 and 153 ◦C using PP counter balls. It was observed that there is no temperature dependency of the friction coefficient, except for the reference steel and the Al2O3 coating. At 85 ◦C, a friction coefficient of 0.52 ± 0.11 is measured for the steel substrate, which drops down to 0.27 ± 0.05 at 153 ◦C. A contrary behavior is observed for the Al2O3 coating; the friction coefficient increases from 0.42 ± 0.10 at 85 ◦C to 0.66 ± 0.16 at 153 ◦C. Concerning the thin films, they basically reveal comparable friction coefficients independent from the used deposition parameters. The lowest friction coefficient of 0.28 ± 0.05 is obtained for the AlN-2 thin film, which slightly increases in the order AlN-1, AlN-3 up to 0.39 ± 0.07 for the BCN-1 thin film at 85 ◦C. In a different study, Ball-on-Disc experiments were also used to investigate the friction coefficient between CrAlN, respectively, DLC thin films and Polycarbonate counter balls [31]. It was shown that the friction coefficient depends on the temperature, and the adhesion behavior is related to the surface energies of the deposited thin films and surfaces. Moreover, it was reported that the friction coefficient depends on the roughness of the surfaces. Smoother surfaces lead to a higher fraction of the adhesive effect on the friction, which in turn leads to a higher coefficient of friction [32]. In fact, the obtained friction coefficients at 85 ◦C can be tendentially correlated with the roughness. The polished steel surface reveals the smoothest surface with a roughness of *R*<sup>a</sup> = 0.006 ± 0.001 μm and the highest friction coefficient. The second lowest roughness is shown by the thermally sprayed Al2O3 coating, which shows the second highest roughness. Higher roughness and lower friction coefficients were analyzed for the thin films. At 153 ◦C, the influence of the roughness on the friction coefficient is not observed. The melting point of PP is between 160 and 165 ◦C, and therefore the mechanical properties of PP are already decreased at 153 ◦C. The surface of the Al2O3 coating contains defects such as pores, as shown in Figure 2, which affect the friction behavior, especially in contact with a soft material as also shown by the large deviation bar of the friction coefficient. Therefore, the Al2O3 coating reveals a higher friction coefficient at 153 ◦C compared to the other thin films.

**Figure 4.** Friction coefficient of the AlN-1, AlN-2, AlN-3, and BCN-1 thin films, as well as the steel substrate and the Al2O3 coating, at 85 and 153 ◦C using PP counter balls.

In addition to the frictional forces, a comparable adhesion of the plastic to the thin film thermocouple and the extrusion die is to be ensured. Therefore, the wear of the PP balls after the Ball-on-Disc experiments was analyzed, which is an indirect hint for the adhesion behavior to the used coatings. As visualized in Figure 5, the wear coefficients of the PP balls are between 16.9 <sup>±</sup> 1.3 <sup>×</sup> 10−<sup>5</sup> mm3/Nm (AlN-3) and 34.3 <sup>±</sup> 5.1 <sup>×</sup> 10−<sup>5</sup> mm3/Nm (AlN-1) for the PVD coated counter parts, whereas a wear coefficient of 26.7 <sup>±</sup> 3.4 <sup>×</sup> <sup>10</sup>−<sup>5</sup> mm3/Nm is analyzed for the uncoated steel counterpart at 85 ◦C. Using the thermally sprayed Al2O3 coating as a counter body, a wear coefficient of 10.6 <sup>±</sup> 0.6 <sup>×</sup> <sup>10</sup>−<sup>5</sup> mm3/Nm is noticed. Accordingly, at 85 ◦C, all counterparts cause a comparable wear coefficient of the PP balls, since the maximum difference of the wear coefficient is 23.7 <sup>×</sup> 10−<sup>5</sup> mm3/Nm within the analyzed counter parts.

**Figure 5.** Wear coefficient of the PP counter balls after the sliding against the AlN-1, AlN-2, AlN-3, and BCN-1 thin films, as well as the steel substrate and the Al2O3 coating, at 85 and 153 ◦C.

At 153 ◦C, an increased wear coefficient of the PP balls is detected, compared to the tribological tests at 85 ◦C. The wear coefficient of the PP balls is increased by the factor 10 to 20, since the mechanical properties of the PP balls are reduced at elevated temperatures. Analyzing the wear coefficient at 153 ◦C, the differences between the counter parts become more visible compared to 85 ◦C. The PP balls slid against the steel surface reveal a wear coefficient of 253.7 <sup>±</sup> 48.7<sup>×</sup> <sup>10</sup>−<sup>5</sup> mm3/Nm, which is comparable to a wear coefficient of 250.3 <sup>±</sup> 41.0 <sup>×</sup> <sup>10</sup>−<sup>5</sup> mm3/Nm for the Al2O3 coating. A lower coefficient of 149.9 <sup>±</sup> 22.9 <sup>×</sup> 10−<sup>5</sup> mm3/Nm is obtained for the BCN-1 counterpart, and higher wear coefficients between 467.0 <sup>±</sup> 70.5 <sup>×</sup> 10−<sup>5</sup> mm3/Nm (AlN-3) and 594.8 <sup>±</sup> 175.2 <sup>×</sup> 10−<sup>5</sup> mm3/Nm (AlN-1) for the AlN thin films. The used PP counter ball is a non-polar plastic, revealing a low fraction of polar surface energy. Accordingly, only the disperse fraction of the free surface energy influences the adhesion behavior to the PP counterpart [33]. Theiss et al. reported that the work of adhesion to PP, which is calculated based on the surface energies, is higher for AlN-rich thin films compared to steel [34]. Especially, for thin films grown in the hexagonal structure, a high work of adhesion is observed [34]. Accordingly, the higher wear of the PP balls slid against the AlN thin films compared to the steel counter body can be related to the surface energies. Moreover, this mechanism is overlaid with the surface roughness. As listed in Table 2, the used deposition parameters of the AlN thin films lead to a change in the roughness. The roughness Ra varies between 0.26 ± 0.06 μm (AlN-3) and 0.36 ± 0.06 μm (AlN-1). The wear of the balls decreases with decreasing roughness in the order AlN-1, AlN-2, AlN-3.

Summarized, within the AlN thin films, the deposition parameters influence the friction coefficient and the wear coefficient of the PP balls at 85 ◦C, as well as 153 ◦C. At 153 ◦C, the influence of the deposition parameters of the AlN thin films on the friction coefficient is lower compared to the wear coefficient. Concerning the wear coefficient at 153 ◦C, the wear of the PP balls is more influenced by the counter body material than the used deposition parameters of a coating, since the difference between the different materials is higher than the difference within the AlN thin films.

Corresponding SEM images of the adhesions from the PP balls to the different coatings after the test at 85 ◦C are shown in Figure 6. No adhesion of PP to the surface of the uncoated 1.2343 substrate can be observed. In addition, no adhesions to the Al2O3, AlN-3, and BCN-1 surfaces have occurred. In contrast to that, local PP accumulations are observed on the AlN-1 and AlN-2 thin films, whereas the amount adhered to the AlN-1 thin film is slightly higher. On the one hand, the adhesion to these thin films can be explained by the cauliflower-like surface, resulting in a higher surface roughness in contrast to the other coatings, as shown in Figure 2. On the other hand, the adhesive layer is only observed for the AlN thin films featuring a small fraction of a second phase, as discussed in Section 3.1.1.

**Figure 6.** SEM images of the different surfaces after the Ball-on-Disc experiments at 85 ◦C.

SEM images of the different surfaces after the Ball-on-Disc experiments at 153 ◦C are visualized in Figure 7. In contrast to the surfaces analyzed at 85 ◦C, at 153 ◦C for all coatings, no rubbing of the PP balls on the surfaces can be noted, which could be related to the change in properties of the PP material with increasing temperature. At 85 and 153 ◦C, there is no general correlation between the ball wear coefficients and the formation of an adhesive layer. However, the formation of the adhesive layer on the AlN-1 and AlN-2 thin films at 85 ◦C correlates with the highest ball wear rates within the three AlN thin films.

**Figure 7.** SEM images of the different surfaces after the Ball-on-Disc experiments at 153 ◦C.

Furthermore, the influence of the different coating surfaces on the melt flow was investigated at higher temperatures corresponding to the extrusion process. Therefore, the coatings were analyzed in a High-Pressure Capillary Rheometer (HPCR). By means of coated surfaces of specially designed exchangeable inserts, the influence on the flow properties of the melt was investigated. Figure 8 shows the basic setup of the capillary and the coated inserts on the example of the AlN-3 thin film.

**Figure 8.** (**a**) Setup High-Pressure Capillary Rheometer and (**b**) coated exchangeable inserts with AlN-3.

Flow anomalies, which are due to the different surface qualities based on the compositions of the coatings, can be detected with a so called "critical" wall shear stress. Until the critical wall shear stress is reached, the melt flow behavior is like "wall sticking". Therefore, flow anomalies cannot be detected. The onset of the throughput jump depends on, among other things, the melt temperature and the

polymer type. In the transition area above the critical wall shear stress, the throughput multiplies many times over, while the wall shear stress remains constant. This range is also referred to as the limiting wall shear stress. Above a certain shear rate, the wall shear stress increases again and the effect of sliding occurs.

Figure 9 shows for the two operating temperatures of 230 and 250 ◦C and the obtained values of the shear rate in dependence of the shear stress using HPCR. Both diagrams show a steady increase of the shear rate over the wall shear stress. A comparison of the coated measurement curves with the reference measurement (grey curve with squares) shows no significant differences. Minimal deviations between the curves are found and can be attributed to the used setup. From the obtained values, it is concluded that no particular flow anomalies can be detected with the surface coatings present. A critical wall shear stress cannot be detected at the typical shear rates in the extrusion process . <sup>γ</sup> = 100 <sup>−</sup> <sup>10</sup>−<sup>4</sup> <sup>1</sup> *s* . Consequently, all investigated coatings can be used as top layers for the thermocouples, since there is no difference in the melt flow behavior between the uncoated and coated surfaces of the wide-slot nozzle.

**Figure 9.** Critical wall shear stress at (**a**) 230 ◦C and (**b**) 250 ◦C, independent of the thin film.

#### *3.2. Insertable Thin Film Thermocouple*

The insertable thin film thermocouple is designed to be exchangeable, applicable to plastic film extrusion dies. As shown in Figure 10, the unity consists of a bush, which can be screwed into any die revealing an appropriate drilling (M24). The ceramic insert contains the thin film thermocouple and can be inserted into the bush, and is fixed by a step on the bottom side and a cap on the top side. The total length of the screw-in bush, including the top cap, is 63.5 mm.

**Figure 10.** Structure of the exchangeable thin film thermocouple.

The Ni and Ni-20Cr conducting paths (width of 0.5 mm), deposited on the ceramic insert using a masking system, are shown in Figure 11. On the bottom side (step), the conducting paths are connected to form the hot junction length of 1.5 mm. The conducting paths run over the lateral surface to the top side. On the top side, the conducting paths end with contact points (2.5 <sup>×</sup> 2.5 mm2), as visualized in Figure 10. Thermocouple balancing lines can be connected to these points, either by brazing or spring probe pins, to conduct the generated thermovoltage to the measuring system.

**Figure 11.** Ceramic insert showing the conducting paths of the Ni/Ni-20Cr thin film thermocouple.

The multilayer design of the thermally coated insert is shown in Figure 12. Primarily, the steel core is coated with a bond coat (NiCoCrAlTaY) to enhance the adhesion of the ceramic layers to the steel substrate, especially at higher temperatures. The bond coat is followed by a coarse 7Y2O3-ZrO2 coating to ensure the thermal insulation between the thin film thermocouple and the die, as well as the steel core of the insert, for an accurate measurement of the temperature of the plastic melt variations. Especially for wall mounted thermocouples, it is reported that the measurement of temperature fluctuations is difficult when the thermal insulation is not adequate [10]. The 7Y2O3-ZrO2 coating is layered by a dense Al2O3 coating, which improves the adhesion of the thin film thermocouples due to a lower content of defects, such as pores. Moreover, Al2O3 reveals a higher electrical resistance compared to ZrO2 to guarantee no electrical short of the thermocouple.

**Figure 12.** Cross-section of the multilayer design of the thermally sprayed insulating coatings of the ceramic insert.

The extrusion nozzle is made of steel and, therefore, the inserted tool with the thin film thermocouple should reveal the same tribological properties. As shown by the Ball-on-Disc experiments, all coatings reveal no adhesions of the PP counter ball on the surface at 153 ◦C. Concerning the wear of

the PP ball, similar values are obtained for the steel surface and the thermally sprayed Al2O3 coating. High-Pressure Capillary Rheometer experiments also demonstrate no significant differences in the melt flow for the different coatings. Therefore, the thin film thermocouple was not covered by an additional nitride layer for the operation test.

#### *3.3. Operation Test*

#### 3.3.1. Validation of the Temperature Measurement

Before the operational testing, the calibration curves for the thermocouples have to determined. In a previous study [20], the procedure was described and the dependence of the thermovoltage and temperature for the thermocouples was found. In the present study, the mass temperature was measured shortly before the outlet of the die. The used experimental setup is shown in Figure 13.

**Figure 13.** Experimental setup showing the array of the thermocouples in the wide-slit nozzle.

After the extrusion line reached a steady state for the cylinder heating and the used material, the embedded thin film thermocouple (TE) was inserted in the die and the recording of the measurement signal began. All four elements were inserted at the same time. Figure 14 shows the heating process of the thermocouple (grey line) until a static state was reached. All other temperature sensors were installed in advance to check the tool temperature. In total, the thermocouple required 130 s until 90% of the final value was reached, and approximately 1000 s until the stationary state was reached.

**Figure 14.** Heating curve of the thin film thermocouple in the plastic flat film extrusion process.

After the steady state of the thermocouple was reached, the measurement with all sensors was recorded over a time of 1.5 h. The measurement of the melt temperature at the mold outlet was additionally determined in 5 min intervals by means of a thermocouple as a guideline value. The surface temperature was also measured with a laser to verify the values. Figure 15 shows the temperature curve over time.

**Figure 15.** (**a**) Monitor conditioning of the melt temperature and (**b**) influence of the calibration on the measured mass temperature.

During the measuring process, no temperature peaks were detected (Figure 15a). All sensors displayed a constant measuring signal within the time interval. Due to the measuring tip of the T1.5, the sensor was placed deeper in the melt and measured the melt temperature significantly better. These measured values and the mathematical equations for the calibration curves were used to adjust accuracy of the measured melt temperatures. Figure 15b shows the adjusted temperatures. All thermocouples show relatively similar measured values after calibration. Compared to the mass temperature of 200.6 ◦C obtained by a laser at the nozzle exit, the sensors T0 and T0.5 measure with the lowest deviation after calibration. The deviation of the thin film thermocouple is 2.52%. Especially in polymer extrusion, the determination of the temperature is important, since even small deviations lead to a high change of the shear viscosity of the melt [35]. However, all used thermocouples, including the thin film thermocouple, can be used for the temperature monitoring of the mass in the extrusion die process after recalibration regarding the mounting place in the nozzle. Before the recalibration, the accuracy of the measured mass temperature increased with extent into the melt. For obtaining die melt temperature profiles in extrusion processes, generally, thermocouple mesh arrangements are inserted in the melt flow [10]. However, this technique can only be used in test conditions, since the melt flow is disturbed. Using many thin film thermocouples, a temperature profile can be obtained without affecting the melt flow.

#### 3.3.2. Foil Quality

The optical film properties were determined by measuring gloss and transmission. Ten samples (dimensions 50 <sup>×</sup> 100 mm2) were taken for evaluation during flat film production. The transmission and the gloss of the films were tested based on the described experimental setup. For the transmission, the determination of the light intensity *I*<sup>0</sup> is necessary as reference. The transmitted light intensity *I*<sup>T</sup> was measured by clamping the samples in the sample holder. The difference between *I*<sup>0</sup> and IT is the transmission spectrum. For the reflection, a white comparison sample is required, which completely reflects the entire light intensity *I*0. Ocean Optics provided white specimens for the comparison sample. Then, the samples were irradiated with the light intensity *I*<sup>0</sup> at the angle of 2◦ defined in DIN 5036-3, and the reflected part IR was measured.

Table 3 shows the reflection results for settings 1 and 2 as reflectance [%]. The velocity of the trigger unit was changed between the two settings. While at setting 1, the velocity reaches 1.5 mm/s, at setting 2 the velocity is 1.7 mm/s. For both settings, the values are between 29.91% and 31.88% and, accordingly, there is no significant difference between the different sensors. The wall-flush thermal sensors and the thin film thermocouples have a slightly lower value than the others.


**Table 3.** Results of the reflection results in dependence of the used thermocouple.

Table 4 shows the transmission results for settings 1 and 2 as transmittance [%]. There are no significant differences in the settings. Therefore, the geometry of the thermocouples does not influence the transmittance. Only for setting 2, the measured values are 5% higher than setting 1. This effect is to be explained due to the fact that setting 2 produces thinner films, and an increased light intensity penetrates the sample.

**Table 4.** Results of the transmission results in dependence of the used thermocouple.


#### **4. Conclusions**

In this study, AlN and BCN thin films were magnetron sputtered on thermally sprayed Al2O3 coatings using different deposition parameters. In the case of the AlN thin films, a high Ar/N2 gas flow ratio of 6.0 and a high heating power led to a coarse structure with a reduced hardness from about 14 to 6 GPa. Since these films are proposed to be used in flat film extrusion nozzles, tribological experiments were conducted. Using a high deposition temperature and high Ar/N2 gas flow ratio for the AlN coatings favor the adhesion of Polypropylen to the thin film surfaces after Ball-on-disc experiments at 85 ◦C. The formation of the adhesive layer is caused by the higher roughness of the AlN thin films, which can be related to the physical structure, and the work of adhesion. In contrast, no adhesions to the steel surface, the thermally sprayed Al2O3 coating, the BCN thin film, or a AlN thin deposited with lower temperature and higher Ar/N2 gas flow ratio were observed. At 153 ◦C, for all coatings, no adhesions emerged, since the mechanical properties of the PP ball were changed with the temperature. The interaction of the thin film thermocouples with the plastic melt was investigated in application tests. Compared to the uncoated tool surface, the thin films reveal no negative effects on the wall shear stress and the wall shear speed. Accordingly, an effect of the thermocouple on the foil quality could not be detected in the samples examined under the spectroscope. Thus, flat films with constant reflection properties could be produced by measuring the melt temperature using thin film thermocouples.

Ni/Ni-20Cr thin film thermocouples were deposited on newly developed exchangeable tool inserts and deployed in a flat film extrusion process. The measurement signal of the thin film thermocouple and industrial reference thermocouples were tested using a specially designed die. No temperature peaks were detected during the measuring process, demonstrating a stable measurement behavior. At a melt temperature value of 200.57 ◦C, the obtained value by the help of the thin film thermocouple was 205.75 ◦C, which is slightly higher compared to the industrial sensors. The deviation of the thin film thermocouple did not exceed 2.52%, which is within a typical tolerance range of a type K sensor. Within the proposed attempt, PVD thermocouples are a promising approach which can be used for online monitoring. However, the used design can be used to further enhance the performance of the wide-slit nozzle by using wear resistant top layers.

**Author Contributions:** Conceptualization, D.K., H.M. and D.S.; Methodology, D.K. and H.M.; Validation, D.K., H.M. and D.S.; Investigation, D.K. and H.M.; Writing—Original Draft Preparation, D.K. and H.M.; Writing—Review and Editing, W.T., D.S. and V.S.; Supervision, W.T. and V.S.; Funding Acquisition, W.T. and V.S.

**Funding:** The authors gratefully acknowledge the financial support of project 18627 N of the German Federation of Industrial Research Associations (AiF) and the Forschungskuratorium Maschinenbau e.V. (FKM).

**Acknowledgments:** We acknowledge financial support by Deutsche Forschungsgemeinschaft and Technische Universität Dortmund/TU Dortmund University within the funding programme Open Access Publishing. In addition, the authors thank Dirk Biermann and Eugen Krebs from the Institute of Machining Technology (TU Dortmund University) for machining the ceramic inserts and providing the confocal 3D microscope.

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

#### **References**


© 2019 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/).

## **Influences of Oxygen Ion Beam on the Properties of Magnesium Fluoride Thin Film Deposited Using Electron Beam Evaporation Deposition**

#### **Gong Zhang 1, Xiuhua Fu 1,\*, Shigeng Song 2,\*, Kai Guo <sup>3</sup> and Jing Zhang <sup>1</sup>**


Received: 30 October 2019; Accepted: 3 December 2019; Published: 7 December 2019

**Abstract:** Magnesium fluoride (MgF2) materials are commonly used for near/medium infrared anti-reflection optical coatings due to their low refractive index and wide-range transparency. Ion assistant deposition is an important technique to increase the density of MgF2 and reduce absorption around 2.94 μm caused by high porosity and moisture adsorption. However, the excessive energy of Argon ion will induce a color center and; therefore, lead to UV/Visible absorption. In this paper, oxygen ion was introduced to reduce the color center effect in MgF2 thin film deposited using electron beam evaporation with ion assistant. The films were deposited on Bk7 and single crystal silicon substrates under various oxygen ion beam currents. The microstructure, optical constant, film density, stress, and adhesion are investigated, including the absorption properties at the infrared hydroxyl (–OH) vibration peak. The results show that as the oxygen ion beam current increases, the absorption value at the position of the infrared OH vibration, defects, and stress of the film decrease, while the refractive index increases. However, MgF2 has poor adhesion using oxygen ion-assisted deposition. Thin MgF2 film without ion beam assistant was used as adhesive layer, high density, and low absorption MgF2 film with good adhesion was obtained.

**Keywords:** MgF2; color center absorption; density; crystal frequency; stress; adhesion

#### **1. Introduction**

Near/medium infrared detectors can work in both low light night vision and medium bands simultaneously, which can be widely used in military and civil fields [1–3]. The MgF2 is a suitable material due to its low refractive index and transparency over the ultraviolet range to the far-infrared region [4,5]. However, the films easily absorb moisture in the atmosphere and relatively large tensile stress because of poor film structure, resulting in spectrum shift. The results show that increasing the deposition temperature of the substrate can increase the density of the film and reduce the instability caused by moisture absorption of MgF2. When the deposition temperature of the substrate reaches 0.6 Tm [6] (where Tm is the melting temperature of the material), the aggregation density reaches a stable value of 0.9. Thereafter, the density will not increase with the increase of deposition temperature. Raising the temperature cannot completely solve the problem of the aggregation density [7,8]. Therefore, ion beam-assisted deposition is used to improve the density of the film. However, there are limitations to be considered in ion beam-assisted deposition. Both Matin Bischoff and Targove researches showed a technology which can produce high packing density and a low extinction coefficient by

plasma(ion)-assisted deposition, but the pure fluorine introduced in the vacuum chamber will influence the environment and also reduce the life of the vacuum chamber. M. Kennedy research showed that xenon ion bombardment has a better result than Ar ion backfill oxygen at ultraviolet–visible wavelengths, but the infrared property of films has not yet been considered in this paper [9–11]. Dumasp's research showed that the use of Ar as an ion source to assist gas deposition increased the bulk density of MgF2 films, but excessive particle energy led to a change in the stoichiometric ratio of Mg and F, resulting in F atom vacancies, which led to the absorption of ultraviolet and visible parts [12,13]. These studies further show that excessive Ar<sup>+</sup> ion energy can lead to more serious moisture absorption problems when MgF2 thin films are evaporated by an electron gun [14].

The working band of near/mid infrared anti-reflective film contains the characteristic peak of water absorption within the band. Moisture absorption will greatly affect the performance of anti-reflective film. Therefore, it should be considered to increase the aggregation density of film while filling F− ion vacancies. Sun discussed the effects of MgF2 prepared at different temperatures on the spectral transmittance and absorption of the deep ultraviolet band produced by the MgO content in the film layer placed in air, which reduced the absorption of the color center [15]. It should be noted, the oxygen can reduce the F− vacancy. In the paper the effect of oxygen ion source-assisted deposition on the properties of MgF2 is studied. In order to obtain MgF2 thin films with high density and low absorption by O2<sup>−</sup>assisted deposition. The problem of poor bonding between MgF2 thin films and Si substrates in the experimental process was also analyzed, and the ion source-assisted deposition technology was optimized to improve the adhesion of MgF2 thin films on Si substrates assisted by O2 ion source.

#### **2. Experiment**

#### *2.1. Thin Film Preparation*

The films were prepared by electron beam evaporation with O2 ion beam-assisted deposition. The vacuum chamber was evacuated to a base pressure of less than 1.0 <sup>×</sup> 10−<sup>3</sup> Pa. With a substrate temperature of 300 ◦C, thin films were deposited on crystal silicon (ϕ20 mm × 1 mm) and float glass (ϕ25 mm × 1 mm) with a deposition rate of 0.8nm/s with different ion beams. The substrates were cleaned by ultrasonication. The experiment used a quartz crystal to monitor the physical thickness of the film. The monitor was SQC-310 produced by Inficon Co., Ltd (Inficon, Shanghai, China). To investigate oxygen ion effects on MgF2 film properties, samples were deposited under various oxygen ion currents (Kaufman Ion beam assistant) of 0, 50, and 80 mA at deposition temperature of 300 ◦C and 0.8 nm/s deposition rate. Sample thicknesses were controlled at 700 nm.

#### *2.2. Film Characterization*

The properties of MgF2 samples deposited under various oxygen ion currents, such as the crystal frequency, optical transmittance, chemical phase of the film, surface roughness, and adhesion, were analyzed using the techniques listed in Table 1.



#### *2.3. Some Background for Characterizations*

#### 2.3.1. Quartz Frequency Measurement of Density

The refractive index of the film will change with the absorption of moisture by the pores within the film. Usually, the packing density or porosity of thin films can be calculated using linear interpolation formula according to the change of refractive index [16]. The refractive index of dense MgF2 film is about 1.38, which is very close to the refractive index of water 1.33. The refractive index change of the film layer after moisture absorption is very small, and it is difficult to calculate the density of the film through the change of refractive index. Therefore, in this study, the density of porous MgF2 was calculated based on the measurement of the mass change of film before and after moisture absorption. A quartz crystal plate is very sensitive to the change of mass. Therefore, the density or porosity of the film layer can be obtained accurately through the change of frequency of quartz crystal plate before and after the absorption of water. The derivation process of the specific calculation is as follows [17].

Its natural frequency is inversely proportional to the thickness t and is proportional to the frequency constant *N*. The relationship is shown in Equation (1)

$$f = \mathcal{N}/t\tag{1}$$

$$
\Delta f = -\frac{N\Delta t}{t^2} \tag{2}
$$

Differentiating Equation (1) with respect to the thickness to get the crystal frequency:

After MgF2 film is deposited, its density is close to that of quartz and the thickness of the coated film during the experiment is much smaller than that of the quartz crystal oscillator, the increase in film thickness can be approximated as the change of the thickness of the quartz crystal.

$$
\Delta \mathbf{m} = A \bullet \rho\_M \bullet \Delta t\_M = A \bullet \rho\_Q \bullet \Delta t \tag{3}
$$

where Δ*m* represents the mass change, *A* is the crystal coating area, and ρ*<sup>M</sup>* and ρ*<sup>Q</sup>* are the film density and the quartz density, respectively. By substituting into Equation (2), the relationship between the change in the quartz crystal frequency and the film thickness can be obtained by Equation (4).

$$
\Delta f = -\frac{\rho\_M}{\rho\_Q} \bullet \frac{f^2}{N} \Delta t\_M \tag{4}
$$

It can be seen from the formula that the decrease in the quartz crystal frequency is proportional to the thickness and density of the coated film.

Assume that the initial frequency of the crystal is *f*1, the frequency after water absorption is *f*<sup>∗</sup> 1 , the bulk density of the film is ρ*s*, and the density is shown in Equation (5),

$$p = \frac{\Delta f\_1}{\Delta f\_1 + \rho\_s \Delta f\_1^\*} \tag{5}$$

According to Equation (5), the density under various conditions can be easily calculated.

#### 2.3.2. Zygo Interferometry for Measuring Film Stress

Stoney proposed calculating the stress of a thin film by measuring the radius of the deformation curvature of the film and the substrate surface. When the actual thickness of the film is considered to be infinitesimal relative to the substrate, both the film and the substrate can be considered as homogeneous and isotropic materials [18]. The Stoney formulate is shown below:

$$
\sigma\_f = \left(\frac{E\_s}{1 - \nu\_s}\right) \frac{t\_s^2}{6Rt\_f} \tag{6}
$$

where, *E*s and *V*<sup>s</sup> are the elastic modulus and Poisson's ratio of the substrate, respectively. *T*<sup>s</sup> and *t*<sup>f</sup> are the thickness of the substrate and the film, respectively. *R* is the radius of curvature of the substrate. Based on the interference principle, the Zygo interferometer calculates the curvature change of the measured object by comparing the interference fringe generated by the optical path difference between the measured light and the reference light reflected from the measurement plane [19]. By processing the test results of the Zygo interferometer, the power value indicating the apparent degree of surface deformation can be obtained. The power is showed in Equation (7):

$$power = \frac{D\_s^{\,^2}}{8R} \tag{7}$$

where *Ds* is the diameter of the substrate, and the radius of curvature before and after coating can be expressed by Equation (8),

$$\frac{1}{R\_2} - \frac{1}{R\_1} = \frac{8}{D\_s^2} (power\_2 - power\_1) \tag{8}$$

Combining Equation (6), the surface stress of the thin film can be expressed by Equation (9),

$$
\sigma = \frac{4Es}{3(1-\nu\_s)} \frac{t\_s^2}{t\_f D\_s^2} \Delta power
\tag{9}
$$

#### **3. Result and Analysis**

#### *3.1. Packing Density Character*

The variation of crystal oscillator frequency with time after coating is shown in the Figure 1.

**Figure 1.** Crystal frequency changed with time in different ion beams (the black, red and blue lines represent the frequency with ion beam assistant condition in ion beams 0, 50, and 80 mA, respectively).

As can be seen from the figure, during the venting of the vacuum chamber to the atmospheric state, the frequency of the crystal vibrator is significantly decreased. As the immersion time in the deionized water increases, the frequency of the crystal vibrator decreases continuously, and finally reaches a stable value. The value in the crystal vibrator frequency decreases as the ion beam density increases. According to the formula listed in Section 2.1, the crystal oscillator frequency variation value is used to calculate the packing density of the film. The aggregation densities under different processing conditions are shown in Table 2.

It can be seen from the table that as the oxygen ion beam increases, the density of the film increases. Therefore the oxygen ion-assisted deposition can be an effective technique to reduce the porosity of MgF2 film.

As discussed above, the deposited MgF2 films on quartz crystal were immersed in DI water for a certain time to allow the void of film to be filled by water, and oscillation frequencies of the crystal were recorded. Then the samples were placed back in a chamber under high vacuum and at high temperature to remove the water trapped in the void. The oscillation frequencies of crystal were also recorded. Table 3 below shows the frequencies of crystal with MgF2 film with and without adsorbed moisture, and the frequency differences.


**Table 2.** The density of the MgF2 films produced at various oxygen ion currents.



As the ion beam current increases, the frequency difference between the crystal oscillators before and after water immersion decreases (C–A in Table 4). When the ion beam density is 80 mA, the frequency of the crystal oscillator before (A) and under vacuum after the immersion (C) is very close. It shows that under ion beam condition of 80 mA, the main reason for the change of the crystal vibration frequency is that water vapor penetrates into the pores of the film, and only physical adsorption occurs mainly. Water molecules trapped in the pores with strong bonding are almost negligible. When oxygen ion current is 0 mA, the difference between the crystal oscillators frequency under vacuum at 300 ◦C (C in Table 4), and the frequency just after the completion of MgF2, is much higher compared to the one under 80 mA oxygen ion current, which means there is a much stronger bonding force between water molecule and MgF2. This indicates that the water vapor is not only physically adsorbed within the deposited film on the crystal oscillator, but also undergoes a chemical reaction to form a strong bond. Through the above tests, it was shown that oxygen ion bombardment introduces modifications of MgF2 film when the MgF2 film was deposited using electron beam evaporation under oxygen ion assistant. The participation of oxygen ion would supplement the crystal defects, improve the crystal structure, and cause the modification of the MgF2 film surface state, including the surface of pores in film (e.g., Mg suspending bond). Oxygen ion also improves the density of the film. From a macroscopic point of view, the refractive index of the film increases, absorption and scattering decrease, and tensile stress decreases.

**Table 4.** Transmittance change.


#### *3.2. Optical Properties*

Visible/near infrared and mid-infrared transmittance were measured. Visible/near infrared transmittance was used to fit the optical constant and the mid-infrared transmittance was used to analyze the absorption of hydroxyl near 2940 nm.

#### 3.2.1. The Visible/Near Infrared Analysis

The Visible-near infrared band film transmittance curve is shown in the Figure 2.

**Figure 2.** The transmittance of different ion beams in the Vis-infrared band (the black, red, blue lines represent transmittance with the ion beam assistant condition in ion beams 0, 50, and 80 mA, respectively).

It can be seen from the figure that there are significant differences in the refractive index and absorption of MgF2 films under different ion beam flow conditions. The optical constant dispersion curve of the film obtained by the full spectrum fitting method is shown in Figure 3.

**Figure 3.** Optical constant dispersion curve (**a**) is the refractive index, (**b**) is the extinction coefficient the black, red and blue lines represent optical constant with the ion beam assistant condition in ion beams 0, 50, and 80 mA, respectively.

It can be seen that the O2 ion beam causes the refractive index to increase. It was because of the presence of MgO compound in MgF2 film. Using the Wiener bounds model, the amount of MgO roughness was estimated as 13.1%.

#### 3.2.2. Hydroxyl Vibration Absorption

The infrared transmittance curve of the single crystal Si substrate [20] and the transmittance of the MgF2 film prepared under different ion source conditions are shown in Figure 4.

**Figure 4.** The transmittance of MgF2 and substrate (the black, red, blue lines represent mid-infrared transmittance with the ion beam assistant condition in ion beams 0, 50, 80 mA, respectively. The dashed line is Si substrate).

It can be seen that the spectral transmittance of the monocrystalline Si substrate at the hydroxyl vibration position of 2770~3200 nm also fluctuates, indicating that the compounds containing –OH exist on the surface of the Si substrate [21]. The spectral transmittance of MgF2 prepared under different ion source conditions at the band of 2770~3200 nm is significantly decreased, and the characteristic absorption peak of –OH appears at 2940 nm. When the substrate was immersed in deionized water, the spectral transmittance decreases in accordance with the frequency change of the crystal oscillator of Section 3.1. As the immersion time increases, the spectral transmittance decreases continuously and is substantially stable after 32 h. After baking under vacuum for two hours, the infrared spectrum of 2500~5000 nm was retested. The average reduction values of the spectral transmittance before and after baking in the range of 2770~3200 nm are shown in Table 4.

From the table it can be seen that ion beam-assisted deposition can effectively reduce the absorption of MgF2 in the range of 2770 to 3200 nm, and the spectral transmittance decreases the least when the ion beam is 80 mA. After baking under vacuum, the decrease of the spectral transmittance of MgF2 with an ion beam density of 80 mA at 2770 to 3200 nm is basically the same as that before immersion, indicating that the moisture physically adsorbed in the film causes the decrease in the spectral transmittance of the film. While, without using ion bombardment, the spectral transmittance of the film is only changed by 0.4% after baking under vacuum conditions. After baking in vacuum, the film still contains compounds with hydroxyl groups.

This indicates that the decrease of absorption of MgF2 thin film prepared by oxygen ion source-assisted deposition near 2940 nm is not only due to the increase of concentration density of the film layer, but also due to the recombination of oxygen ion and F− vacancy caused by Mg suspension bond during deposition, preventing the suspension bond from forming other compounds with hydroxyl in water vapor.

#### *3.3. XRD*

According to the result of spectrum test, the oxygen ion and Mg suspending bond fill the film defects, reduce the colour center absorption, and reduce the hydroxyl absorption of film by filling the F− vacancy with O particles. To analyze the phase change of film, an XRD measurement was taken, the result are shown in Figure 5.

The XRD results show that MgF2 film exhibits a polycrystalline state, and the film has a remarkable crystal orientation, with the maximum intensity of the diffraction peak in the <110> direction. It can be seen that the location of the diffraction peak without assisted deposition by oxygen ion source is consistent with that of the standard PDF (powder diffraction file) card, as the oxygen ion beam flow increases, the diffraction peak positions in each direction shifts to the right, the intensity of the diffraction peak decreases, and the diffraction peak gradually widens as the ion beam flow increases. When the ion beam was increased to 80 mA, the MgF2 diffraction peak in the <220> direction and significantly broadened. The diffraction peak became inconspicuous. In addition, the characteristic diffraction peak of MgO appears at positions where the diffraction angle (2θ) values are 36.703◦ and 62.219◦, indicating that under the beam condition the O ion compounds with Mg in the film to form a MgO crystal.

**Figure 5.** XRD measurement of different ion beams (the blue, red, and black represent ion beams 80, 50, and 0 mA, respectively; the vertical lines represent the theoretical positions of the peaks for MgF2 and MgO; F represents MgF2, O represents MgO).

The comparison parameters of the crystal plane spacing of MgF2 thin film and PDF standard card are as per Table 5.


**Table 5.** The crystal plane spacing of sample and PDF standard value.

The lattice constants of the MgF2 film and the PDF reference card prepared under different process conditions were calculated by Jade software and are shown in Table 6.


**Table 6.** The crystal character calculates by MDI Jade.

As can be seen, with the increase of oxygen ion beam, the crystal axis cell becomes smaller and the spacing between the crystal planes decreases. The oxygen ion source-assisted deposition indicates that the crystal cells gather more closely during the growth of the film.

#### *3.4. Adhesion*

The tape peeling method was used for adhesion test, where tape is attached to the sample and then peeled off. The adhesion is then judged on the extent of the film delamination. The measurement process is simple and has good repeatability. The test was done using 3M scotch 610 test tape: Stickiness of (10 ± 1) N/25mm, and keeping tape pull angle at 90◦ to the film surface [22].

The film without O2 ion assistant deposition can endure being peeled off 20 times. When the O2 ion beam was 50 mA, the film delaminated after being peeled off 15 times. When the ion beam increased to 80 mA the film delaminates after being peeled off 12 times. Indicating the film adhesion decreases with increasing ion beam. The adhesion of the film is mainly affected by stress and inter-material adsorption [23]. Thus, the stress was discussed as follows.

The changes in curvature of the MgF2 film before and after deposition obtained by a Zygo interferometer are shown in Table 7. According to the formula in Section 2.2, the stress of the film is 986, 436, and 357 MPa, respectively. It can be seen that the tensile stress of the film decreases as the ion beam density increases.

**Table 7.** Curvature of substrate before and after coating.


The stress test results show that when the ion beam is 80 mA, the tensile stress of the film is at the minimum, and the stress of the film should not be the main cause for poor adhesion. It can be seen from the infrared spectrum test results that at 2770~3200 nm the Si substrate has obvious absorption, indicating that the surface of the Si substrate contained the OH root compound. When the MgF2 film was modified without using oxygen ions, there were obvious F− vacancies in the film layer. The free Mg2<sup>+</sup> ions were combined with the hydroxyl of the Si substrate. The film and the substrate were combined by the bonding force. The MgF2 film substrate, in which the F ionic vacancies were filled by the O particles, was combined by van der Waals force. Therefore, the film–substrate adhesion is much smaller than the bonding force [24,25]. In order to obtain a MgF2 film with good adhesion to the substrate, and effectively reducing the absorption of water molecules, ion beam-assisted deposition was not used at the beginning of up to 50 nm thickness of deposition, and a film (transition layer film) having F− vacancies was obtained. The –OH compounds formed by bonding with the Si substrate made the film adhere well to the substrate. Next, O ion bombardment by increasing the beam density to 80 mA was used to fill the F− ion vacancies in the film during deposition. The comparison of 3M adhesive experiments before and after adjusting the ion source is shown in Figure 6.

**Figure 6.** Sample condition after tape pull test: (**a**) Without transition layer which was produced by electron beam vapor deposition and O2 IAD; (**b**) the first transition 50 nm deposited using only electron beam vapor deposition, then deposition completed with IAD.

In addition, the comparison of transmittance before and after adjusting the ion source is shown in Figure 7.

**Figure 7.** The transmittance of MgF2 (the black and red line represent mid-infrared transmittance with 50 nm pre-layer and without pre-layer, respectively.)

#### **4. Conclusions**

The influence of oxygen ion source on MgF2 film's visible near-infrared band and mid-wave infrared absorption was studied. As the oxygen ion beam flow increases, the film aggregation density increases. Due to the oxygen ions filling into the colour center defects, generated by F− vacancies, an MgO compound is formed in the film, so that the absorption value of the film in the visible range is reduced, and the refractive index is increased. In the infrared light portion, as the oxygen ion beam increases, the F<sup>−</sup> ionic vacancies are filled with O, preventing the Mg2<sup>+</sup> ions from recombining with the –OH in the air, and reducing the absorption of the film in the infrared. However, the stress and adhesion test results show that as the oxygen ion beam flow increases, the bonding force between the film and the substrate changes from chemical bonding force to van der Waals force, and the tensile stress of the MgF2 film decreases, which lead to a bad adhesion for MgF2 film on Si substrate. A solution of two steps of deposition was proposed to solve the problem of poor adhesion: At the early stage of deposition of the MgF2 film, a very thin adhesion layer without ion beam-assisted was formed and allowed further MgF2 film to bond to the substrate with good adhesion, then the O<sup>+</sup> ion current for further MgF2 deposition was increased to allow O to fill the F anion vacancies efficiently, and reduce the absorption. The experimental results demonstrate that MgF2 film with low absorption, high stability, and good adhesion was achieved. The effect of the thickness of the thin adhesive MgF2 layer deposited without ion beam assisted requires further investigation.

**Author Contributions:** Conceptualization, G.Z. and X.F.; methodology, G.Z.; software, G.Z.; validation, S.S., K.G. and G.Z.; formal analysis, S.S.; investigation, J.Z.; resources, J.Z.; data curation, S.S.; writing—original draft preparation, G.Z.; writing—review and editing, S.S.; visualization, K.G.; supervision, X.F.; project administration, X.F.

**Funding:** This research received no external funding.

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

#### **References**


© 2019 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* **Polarization Controlled Dual Functional Reflective Planar Metalens in Near Infrared Regime**

**Yuhui Zhang 1, Bowei Yang 1, Zhiying Liu 1,2,3 and Yuegang Fu 1,2,3,\***


Received: 12 March 2020; Accepted: 11 April 2020; Published: 15 April 2020

**Abstract:** The metalens has been a hotspot in scientific communications in recent years. The polarization-controlled functional metalens is appealing in metalens investigation. We propose a metalens with dual functions that are controlled by polarization states. In the first design, when applied with x- and y-polarized light, two focal spots with different focal lengths are acquired, respectively. The proposed metalens performs well when illuminated with adjacent wavelengths. In the second design, the reflected light is focused when applied with x-polarized light, and when applied with y-polarized light, the reflected light is split into two oblique paths. We believe that the results will provide a new method in light manipulation.

**Keywords:** polarization controlling; dual functional-metalens; focusing; splitting

#### **1. Introduction**

Metamaterials are novel artificial structures that are designed for specified functions like negative refractive index [1–3], polarization conversion [4–6] and perfect absorption [7–9]. Metasurfaces possess the advantages of small losses and are easy to manufacture [10,11], and have been attractive in recent years. Among various metasurfaces, metalens with its unit cell interacting with incident light and introducing abrupt phase shift along the interface of the metasurface have been widely presented to achieve the function of focusing [12,13] and anomalous reflection [14–16]. Lots of models like the nanoslit [17], nanohole [18,19], and graphene ribbons [20–24] have been introduced in metalens design, where 2π phase shift resulting from the designed antennas is needed in controlling the wavefront. Although a polarization-independent metalens [12,25] and polarization-conversion metalens [26] have been reported in former works, and the designed metalenses are always adjusted to manipulate different kinds of incident waves with various structures [27–30], the metalens, with its function controlled by incident polarization state [31,32], has not been fully investigated.

In this paper, a metal-insulator-metal (MIM) structure is proposed to achieve the polarization controlled dual functional two-dimentional (2D) cylindrical metalens, which consists of rectangle gold antennas and a gold mirror spaced by a dielectric layer. The proposed metalens possesses two functions that can be realized when illuminated with x- and y-polarized light. The x-direction and y-direction length of the rectangle gold antenna are adjusted to control the phase and reflectance of the reflected light, because the length of the dipole resonances along x- and y-directions are determined by the x-direction and y-direction lengths, respectively. We first investigate the influence of the y-direction length or x-direction length on the dipole along the x or y-direction. By presenting the phase and reflectance of the reflected light for x-polarization incidence with different x-direction and y-direction

lengths of the rectangle antenna, we find that the y-direction length of the rectangle has little influence on the phase and reflectance in most of its length range. Then we apply the phase approach to design the polarization-controlled metalens, in which the y-direction or x-direction length is fixed as 100 nm when varying the x-direction or y-direction length. We design a metalens with the focal length of *F* = 5 μm and *F* = 5 μm when applied with x- and y-polarized light, which focus well as desired. Then we validate the approach by presenting two metalenses with focal length of 5 and 15 μm for x- and y-polarization incidence only. The y-direction and x-direction lengths are fixed as 100 nm. The focusing effects agree well with that of the proposed polarization-controlled metalens. The proposed metalens works well within a broadband of wavelengths that range from 750 to 850 nm. We also design the metalens to split y-polarized normal incident light and focus the x-polarized light. Therefore, the dual functional metalens can be achieved by tuning the incident polarization state.

#### **2. Design and Simulation Method**

Figure 1 illustrates the schematic of the proposed 2D cylindrical lens, which is a metal-insulator-metal structure to form a Fabry–Perot cavity to enhance the interaction between light and resonance antenna. The resonance antenna and the bottom mirror are selected as Au with data acquired from [33]. The dielectric spacer is chosen as MgF2 with a refractive index of 1.892 [14]. The thicknesses of Au antenna, MgF2 spacer and Au mirror shown in Figure 1b are set as *ta* = 30 nm, *td* = 50 nm and *ts* = 130 nm, respectively. The top view of the unit cell of the metalens is shown in Figure 1c. The period in x- and y-directions are *px* = *py* = 200 nm. The resonance antenna is a rectangle, which possesses dipole resonances along both the x- and y-directions. The rectangle lengths along x- and y-directions are variable to control the phase and reflectance of the reflected light for x- and y-polarization incidences, respectively. Thus, we can separately control the wavefront of the reflected x- and y-polarized light by configuring the rectangle lengths along x- and y-directions. The working mechanism of the proposed metalens is shown in Figure 1a, where different polarization state leads to different function. For numerical analysis, all simulations are carried out by using the finite-difference time-domain software (Lumerical FDTD solutions 8.15.736.0). PML boundary condition is applied in zand x- directions. Periodic boundary condition is applied in the y-direction. The minimal mesh size is 4 nm.

**Figure 1.** Schematic of the proposed metalens. (**a**) The fragment, (**b**) cross section view and (**c**) top view of the proposed metalens.

Generally, the influence of the y-direction length *b* or x-direction length *a* on the dipole along the x or y-direction cannot be ignored, which results in amplitude and phase deviations for the x- or y-polarization incidences. To explore this, we choose the working wavelength as 800 nm. We show the phase and reflectance of the reflected light for x-polarization incidence in Figure 2 with different lengths *b* and *a* of the rectangle antenna. A near 2π phase shift can be acquired when the length *a* increases from 10 to 190 nm, and the length *b* has little influence on it in most of the length range as shown in Figure 2a,b. Due to the symmetry property of the structure, an identical performance can be acquired for y-polarization incidence.

**Figure 2.** (**a**) Phase and (**b**) reflectance of the reflected light for x-polarization incidence with different lengths *b* and *a*. Phase shift and reflectance of reflected wave for (**c**) x- and (**d**) y-polarization incidences. Insets are z-component electric field distributions corresponding to (**c**) horizontal dipole and (**d**) vertical dipole.

Figure 2c shows the phase and reflectance of the reflected light for x-polarization incidence when x-direction length *a* increases from 10 to 190 nm and the y-direction length *b* is fixed as 100 nm. Figure 2d shows the phase and reflectance of the reflected light for y-polarization incidence when y-direction length *b* increases from 10 to 190 nm and the x-direction length *a* is fixed as 100 nm. The insets show the *z* component electric field distributions at the resonant length. Two dipole resonances along x- and y-directions are acquired, together with the Fabry–Perot resonance, a near 2π phase shift can be achieved. We use this approach to design the metalens that possesses different functions controlled by the polarization states of the incident light.

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

As illustrated above, we utilize the phase shift in Figure 2c,d to design the focusing metalens, which possesses the focal length of *F* = 5 μm for x-polarization incidence and *F* = 15 μm for y-polarization incidence. To design a metalens to focus incident light, the phase profile of the metalens should follow the expression [34]:

$$\varphi(\mathbf{x}) = \frac{2\pi}{\lambda\_0} \left( \sqrt{\mathbf{x}^2 + F^2} - F \right) \tag{1}$$

where *x* is the horizontal position from the center of the metalens, Δ*x* is the horizontal shift of the focal point, λ<sup>0</sup> is the incident wavelength, and F is the focal length. Based on Equation (1), we calculate the phase distribution curves for *F* = 5 μm and *F* = 15 μm, and show them in Figure 3a,b. The corresponding length distributions of x-direction and y-direction lengths *a* and *b* are shown in Figure 3c,d, according to which we design and simulate the planar metalens with 60 unit cells with x- and y-polarization incidences, respectively. The electric field intensity distributions for x- and y-polarization incidences are shown in Figure 3e,f, respectively. The reflected lights are well focused in the air side except for a little deviation in the focal length (4.95 and 13.32 mm for x- and y-polarization incidences) from the theoretical values. Because the phase shift resulting from the length change cannot cover the full 2π range, we select the adjacent unit cell instead as an approach. The longitudinal and transverse sizes of the focal spots are represented by full width at half maximum (FWHM) of the focal spots along x- and z- directions. For x-polarization incidence, the FWHM values along xand z- directions are 0.78 and 2.72 mm. For y-polarization incidence, The FWHM values are 1.1 and 9.2 mm, respectively. The focusing efficiencies, defined as the proportion of the incident light energy going to the central focal spot, are calculated to be 41% and 45% for x- and y-polarization incidences. Thus, the metalens with two polarization-controlled focal points is achieved. In practice, a minor inclination of polarization plane is inevitable, which results in a negligible focusing effect for another polarization state.

**Figure 3.** Phase profile to focus incident light for (**a**) x- and (**b**) y-polarization incidences. Corresponding lengths of (**c**) x-direction length *a* and (**d**) y-direction length *b* of the rectangle antenna in metalens design. The electric field intensities for (**e**) x- and (**f**) y-polarization incidence, respectively.

To investigate the influence of the x-direction length *a* and y-direction length *b* on the focusing effect of y- and x-polarization incidences, we design two metalenses that focus x-polarized light and y-polarized light only. For x-polarization incidence, only the x-direction length *a* is designed for focusing with a focal length of *F* = 5 μm as shown in Figure 4a, while the y-direction length *b* is fixed as 100 nm as shown in Figure 4c. For y-polarized light focusing, x-direction length a is fixed as 100 nm as shown in Figure 4b, while the y-direction length b is designed for focusing with a focal length of *F* = 15 μm as shown in Figure 4d. We simulate the two metalenses with x- and y-polarization incidences, and the results are shown in Figure 4e,f, from which we can see that the results are nearly the same as that shown in Figure 3e,f. Therefore, the influence of the x-direction length *a* and y-direction length *b* on the focusing effects of y- and x-polarization incidences are negligible.

**Figure 4.** (**a**) X-direction length *a* and (**c**) y-direction length *b* of the metalens for x-focusing only. (**b**) X-direction length *a* and (**d**) y-direction length b of the metalens for y-focusing only. The electric field intensities for (**e**) x- and (**f**) y-polarization incidence, respectively.

In addition, we also investigate the cases of the proposed metalens working in other incident wavelengths. The simulated results are shown in Figure 5. Figure 5a,b show the electric field distributions when illuminated with 750 nm wavelength x- and y-polarized light, respectively. Figure 5c,d show that of 850 nm wavelength incident light. From Figure 5 we can conclude that the proposed metalens can work well within a broadband wavelength that ranges from 750 to 850 nm. In addition, we investigate the focusing effects for other incident wavelengths. The focal lengths for different incident wavelengths are listed in Figure 5e,f, from which we can see that the focal length changes when incident light wavelength is changing, and the focal length is inversely proportional to the wavelength of light.

**Figure 5.** The electric field intensities for (**a**) x- and (**b**) y-polarization incidences with the incident wavelength of 750 nm. The electric field intensities for (**c**) x- and (**d**) y-polarization incidences with the incident wavelength of 850 nm. (**e**,**f**) The focal lengths with different incident wavelengths for x- and y-polarization incidences, respectively.

Furthermore, another function called the beam splitter can be considered in the dual functional metalens design, which manipulates the normal incident light into two opposite oblique paths [35]. To bend the normal incident light to an oblique reflected light, a phase gradient should be introduced along the interface of the structure following the generalized Snell's law [36]. For normal incidence, the reflected angle can be expressed by:

$$\sin(\theta\_l) = \frac{\lambda\_0}{2\pi n\_l} \frac{d\varphi}{d\mathbf{x}} \tag{2}$$

where λ0, *x*, ϕ and *ni* are the incident wavelength, surface length, phase shift and refractive index of the incident side material, respectively. In our design, the reflection angle of the oblique beams are ±30◦, then the calculated phase shift between two adjacent unit is 45◦. The phase profile calculated from Equation (2) is shown in Figure 6b. The corresponding y-direction length *b* distribution is shown in Figure 6d, based on which the normal incident y-polarized light can be split. However, the reflection angle is less than 30◦, because the phase shift produced by the gold antennas cannot cover the full 2π range, and we take some approximations in the designing process. Then for x-polarization incidence, the x-direction length *a* is designed according to Figure 6a,c, which focus the x-polarized incident light with the focal length of *F* = 5 μm. The simulated results of both x- and y-polarization incidences are shown in Figure 6e,f. For x-polarization incidence, the reflected light is well focused at the point of (*x* = 0 *z* = 5 μm), and the y-polarized light is split.

**Figure 6.** Phase profile (**a**) to focus incident light for x-polarization incidences and (**b**) to split incident light y-polarization incidences. Corresponding lengths of (**c**) x-direction length *a* and (**d**) y-direction length *b* of the rectangle antenna in metalens design. (**e**) The electric field intensities for x-polarization incidence. (**f**) The real part of y-component electric distribution for y-polarization incidence.

The advantage of our results is listed in Table 1 to compare with other works.



#### **4. Conclusions**

In summary, we proposed a dual functional metalens controlled by polarization state of the incident light. Two functions can be achieved by adjusting the incident light with two polarization states. The proposed metalens is designed according to the approach that the dipole resonance is not influenced by the width of the rectangle gold antenna. A metalens working at 800 nm wavelength with focal lengths of *F* = 5 μm for x-polarization and *F* = 15 μm for y-polarization incidence is designed, which agree well with the two exact metalens that focus x- and y-polarized light only. The proposed metalens works well within a broadband wavelength that ranges from 750 to 850 nm. We also designed a metalens with two functions of focusing and splitting controlled by incident polarization states. Therefore, the proposed metalens can achieve two functions by applying two orthogonal polarized lights.

**Author Contributions:** Conceptualization, Y.Z. and Y.F.; Methodology, Y.Z.; Software, Y.Z. and Y.F.; Validation, Y.Z. and Y.F.; Formal analysis, Y.Z.; Investigation, B.Y.; Resources, Y.Z. and B.Y.; Data Curation, Y.Z. and B.Y.; Writing-Original Draft Preparation, Y.Z.; Writing-Review and Editing, Z.L. and Y.F.; Supervision, Y.F.; Project Administration, Y.F.; Funding Acquisition, Y.F. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work is funded by National Natural Science Foundation of China (NSFC) (11474041); National Natural Science Foundation of China (NSFC) (61805025); "111" Project of China (D17017).

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

#### **References**


© 2020 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* **The Analysis of Resistance to Brittle Cracking of Tungsten Doped TiB2 Coatings Obtained by Magnetron Sputtering**

#### **Jerzy Smolik, Joanna Kacprzy ´nska-Gołacka, Sylwia Sowa \* and Artur Piasek**

Łukasiewicz Research Networks–Institute for Sustainable Technology, 6/10 Pułaskiego St., 26-600 Radom, Poland; jerzy.smolik@itee.lukasiewicz.gov.pl (J.S.); joanna.kacprzynska-golacka@itee.lukasiewicz.gov.pl (J.K.-G.); artur.piasek@itee.lukasiewicz.gov.pl (A.P.)

**\*** Correspondence: sylwia.sowa@itee.lukasiewicz.gov.pl; Tel.: +48-48-364-44-332

Received: 27 July 2020; Accepted: 19 August 2020; Published: 20 August 2020

**Abstract:** In this work, the authors present the possibility of characterization of the fracture toughness in mode I (*K*IC) for TiB2 and TiB2 coatings doped with different concentration of W (3%, 6% and 10%). The Young's modulus, hardness and fracture toughness of this coatings are extracted from nanoindentation experiments. The fracture toughness was evaluated using calculation of crack length measurement. An important observation is that increasing tungsten concentration in the range 0–10% changes the microstructure of the investigated coatings: from columnar structure for TiB2 coating to nano-composite structure for Ti-B-W (10%) coating. It can be concluded that doping with concentration 10 at.% W causes an increase of the fracture toughness for the tested coatings.

**Keywords:** PVD coatings; nanoindentation; brittle cracking; fracture toughness

#### **1. Introduction**

Brittle cracking resistance is one of the most important material features. The fracture toughness in mode I (*K*IC) is the parameter considered to be a material constant, which determined the material's resistance to brittle cracking. It is independent of the thickness of the material being tested because it refers to a constant state of stress. This makes it possible to assess how much load a structural element containing a crack can carry. The method for determining the *K*IC factor, which is currently the basic material constant in fracture mechanics, is standardized for solid materials [1,2] and consists of analyzing the process of cracking samples with a properly prepared notch in the three-point bending test. The use of the penetration method to study the fracture toughness *K*IC was initiated in the 1970s by the Evans and Charles, who observed the relation between the crack lengths, which was generated in the corners of Vickers indenter during the hardness test and the value of *K*IC [3]. This made it possible to use the microindentation and nanoindentation methods for the mechanical characterization of micro-volume systems, including layers and coatings.

The mechanical characteristics of thin coatings are of great importance in the processes of optimization and development of material solutions for coatings with high tribological efficiency. For mechanical characterization of thin coatings, including determination of their hardness (*H*), Young's modulus (*E*) and fracture toughness as the *K*IC coefficient, nanoindentation has proved to be a very effective technique. Hardness and Young's modulus are determined on the basis of load and displacement curves, while the *K*IC coefficient can be estimated based on the crack lengths initiated by the indenter.

It should be remembered that, in the case of thin coatings, the results of the brittleness index tests obtained by the nanoindentation method cannot be verified by the impact method, as it is possible for solid materials. Therefore, in the case of solid materials, we can more precisely select the model and range of indenter loads in nanoindentation tests to determine the brittleness of specific materials, e.g., SiC [4], glass, and Al2O3 [5]. In the case of thin coatings, we additionally encounter a very large number of material combinations, i.e., chemical composition, phase structure, state of internal stresses resulting from, e.g., the type of substrate or the method of coating deposition. Therefore, it seems very difficult to indicate one model for the determination of the *K*IC coefficient for all coatings in the current state of knowledge. It has been shown that methods using the nanoindentation method and crack length analysis to determine the brittleness of materials lead to comparable results [6]. This can be a good tool for comparing the brittleness of materials. Importantly, in the case of thin coatings, it is currently difficult to find another method that has as much potential and capabilities in determining the fracture toughness of thin coatings as the nanoindentation method.

The literature analyses show different models for determining the *K*IC coefficient of thin coatings [7,8], of which two are most often used, as proposed by Anstis's [9] and Laugier [10].

The aim of the research, the results of which are presented in this article, was to propose a method for determining the *K*IC coefficient of thin PVD coatings using the nanoindentation method. In the article, the Laugier model was used to analyze changes in fracture toughness (*K*IC) for TiB2 ceramic coatings doped with tungsten. The coatings were produced by magnetron sputtering in a Direct Current (DC) system. The problem to be solved was the methodology of selecting loads for the Berkovich indenter in the indentation process and the methodology of measuring the crack lengths generated in the coating.

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

#### *2.1. Coating Deposition*

Tungsten doped TiB2 coatings were prepared by DC magnetron sputtering method using original magnetron systems made by Łukasiewicz Research Network-Institute for Sustainable Technology in Radom (Ł-ITeE Radom) with a Balzers pump system (Radom, Poland) with two circular magnetrons placed at an angle of 120◦ to each other. In the deposition process was using two targets made of TiB2 (99.50% purity) and pure-W (99.95% purity) according the procedure present in paper [11]. The diameter of targets was *d* = 100 mm and thickness *g* = 7 mm. The TiB2 and Ti-B-W coatings were deposited in an atmosphere of pure argon (Ar 100%). The parameters of the Ti-B-W coating process are shown in Table 1.


**Table 1.** Deposition parameters of reference TiB2 coating and TiB2 coatings doped with tungsten.

The tested coatings were deposited on to samples made of high-speed steels SW7M with a diameter of 25 mm, thickness 6 mm and surface roughness *R*<sup>a</sup> ≤ 0.05 μm and on samples of monocrystalline Si (100). The samples were washed with 99.9% pure alcohol before being placed in the process chamber. Before the coating process, the samples were ion-etched in the Ar + plasma. During the coating process, the sample temperature was stabilized at 300 ◦C using resistance heaters. The Ti-B-W coatings were deposited by changes of the source power in the range 25–75 W. The deposition time for each coating was 1 h. The scheme and view of the magnetron system that we used are shown in Figure 1a–c.

**Figure 1.** Dual DC magnetron system used to apply TiB2 coatings doped with tungsten: (**a**) scheme; (**b**,**c**) process chamber view.

#### *2.2. Coating Characterization*

Samples of monocrystalline Si (100) were used for chemical composition analysis by Wavelength-Dispersive X-Ray Spectroscopy–WDS (Nova NanoSEM 450 with WDS IbeX, Thermo Fisher Scientific, Waltham, MA, USA) localized in AGH (Kraków, Poland). The coatings thickness of brittle fracture cross-section measurements was performed using SEM-Hitachi TM3000 scanning electron microscopy (Radom, Poland).

Samples of SW7M steel with the deposited coatings were subjected to hardness and Young's modulus testing using the Nanoindentation Tester NHT manufactured by Anton Paar with Berkovich diamond indenter (Anton Paar, Ł-ITeE Radom, Poland). For each of the tested coatings, 20 indentations were made in the regime of a contact-depth. The contact-depth of the indenter does not exceed 10% of the coating thickness. The correct indentations were selected and the average hardness values—*H* and Young's modulus—*E*, were calculated based on the results obtained, as well as the corresponding standard deviations. The measurement results made it possible to determine the *H*/*E*—as the plasticity index or load factor which is responsible for the maximum elastic deflection, when the coating is not destroyed and the *H*3/*E*2—as a resistance to plastic deformation, which determines the load capacity of the coating.

The surface roughness for all tested coatings were measured by Hommel Tester T1000 produced by JENOPTIK (Ł-ITeE Radom, Poland) by contact method. The mean values of *R*a, *R*z and *R*t parameters were calculated.

#### *2.3. Berkovich Indentation Fracture Toughness Test*

SW7M steel samples were also used to study the resistance of brittle cracking using the nanohardness tester CSM with Berkovich diamond indenter. The analysis was carried out in two steps. In the first step, indenter load was selected, when radial cracks were generated in the tested TiB2 and Ti-B-W coatings. For this purpose, 5 indentations were made for each of the tested coatings at different values of the intender normal force, i.e., 50, 100, 200, 300 and 400 mN. The indenter load, for which generated cracks were measurable, was selected for each coating. In the second step, 20 indentations were made for each coating. For each indentation, individual crack length measurements, i.e., (*a*n) and (*l*n) were made according to the scheme shown in Figure 2. For each coating, based on a group of 20 indentations, the mean values *a* and *l* were determined. Then, according to the Laugier model formula (Equation (1)) [10,12], the value of the fracture toughness *K*IC was determined. Observation and crack lengths measurement in the area of the indentations were carried out using SEM Hitachi TM3000 scanning electron microscopy.

$$\mathbf{K\_{IC}} = \mathbf{x}\_{\mathbf{V}} \cdot \left(\frac{a}{l}\right)^{\frac{1}{2}} \cdot \left(\frac{E}{H}\right)^{\frac{2}{3}} \cdot \frac{P}{c^{\frac{3}{2}}} \tag{1}$$

where: *K*IC—fracture toughness; *x*v–indenter geometry factor (for Laugier Equation *x*<sup>v</sup> = 0.016); *E*—Young modulus of coating (GPa); *H*—hardness of coating (GPa); *P*–the indentation load (mN); *a*—the length from the center of the indent to the corner of the indent (μm); *l*1,2,3—the length of the cracks; *l* = (*l*<sup>1</sup> + *l*<sup>2</sup> + *l*3)/3; *c* = *l* + *a*—the total length from the center of the indent to the end of crack.

**Figure 2.** Illustrations of the measurement method of fracture toughness for all investigated indentations.

#### **3. Results**

#### *3.1. Coating Characterization*

The pure TiB2 coatings and Ti-B-W coatings doped with tungsten were analyzed after the deposition processes, obtained by the DC magnetron sputtering method, according the parameters presented in Table 1**.** First, the surface observations and coating thickness measurements were made using the scanning electron microscopy (SEM-Hitachi TM3000) and the surface roughness tests with using the Roughness Hommel Tester. The results of these measurements were necessary to the fracture toughness test, which was planned in the next step. Figure 3 shows the results of surface and cross-sections observations for the tested TiB2 and Ti-B-W coatings. All tested coatings are characterized by high surface smoothness and good coherence, which are free of cracks and defects. The thickness of all investigated coatings was in the range 1.2–1.3 μm.

The prepared coatings were subjected to chemical composition analysis using the WDS method. The obtained results (Table 2) showed that the tungsten concentration doped into the TiB2 coating increased with increasing tungsten power and it is range 3%–10%. The Ti/B or B/(Ti+W) ratios were determined adequately in the TiB2 and Ti-B-W coatings.


**Table 2.** Parameters for investigated TiB2 and Ti-B-W coatings.

The hardness and Young's modulus measurements were carried out without exceeding the depth of 10% of the coating thickness, i.e., max 100 nm. Examples of changes in the force acting on the Berkovich indenter in the case of testing the hardness and Young's modulus of the TiB2 coating and the view of the indentation. Based on the obtained results, the plasticity index as *H*/*E* and the resistance to plastic deformation as *H*3/*E*<sup>2</sup> of investigated coatings were determined. All parameters of the investigated TiB2 and Ti-B-W coatings are shown in Table 3.


**Table 3.** The results of investigations of the chemical composition for TiB2 and Ti-W-B coatings, which were made using the WDS method.

**Figure 3.** SEM micrographs from the surface (on the left section) and from the brittle fracture cross-section (on the right section) with coating thickness measurements for coatings obtained with different power of tungsten sputtering (*P*<sup>W</sup> in Table 1): (**a**) TiB2: *P*<sup>W</sup> = 0 W; (**b**) Ti-B-W(1): *P*<sup>W</sup> = 25 W; (**c**) Ti-B-W(2): *P*<sup>W</sup> = 50 W; (**d**) Ti-B-W(3): *P*<sup>W</sup> = 75 W.

#### *3.2. Determination of the Fracture Toughness KIC*

For measured the sizes of the crack lengths emerging from their corners after indentation with different forces 50, 100, 200, 300 and 400 mN, all the generated cracks were imaged using a scanning electron microscopy (SEM-Hitachi TM3000). The images of indentations for different types of coatings are shown in Figure 4. Observations for the indentations made with the different loads showed, that the cracks for coatings such as TiB2, Ti-B-W (3%), Ti-B-W (6%), which were generated in the corners of the indentation, were clearly visible and well measurable already at the indenter load of 200 mN (Figure 4a–c). On the other hand, the cracks for coating Ti-B-W (10%) were visible and well measurable only with indenter load of 400 mN (Figure 4d). Such intender loading values were accepted for testing of the fracture toughness *K*IC for selected coatings.

**Figure 4.** The indentations for selected TiB2 and Ti-B-W coatings after testing with different values of indenter loading force: (**a**) TiB2: 200 Mn; (**b**) Ti-B-W (3%): 200 mN; (**c**) Ti-B-W (6%): 200 mN; (**d**) Ti-B-W (10%): 400 mN.

According to the adopted methodology (present in Section 2.1), 20 indentations were made for each coating with the selected indenter load (*P*), i.e., for the coatings: TiB2, Ti-B-W (3%), Ti-B-W (6%) was *P* = 200 mN, while for the coating Ti-B-W (10%) was *P* = 400 mN. The penetration depth was respectively *h*200mN = 1000 nm and *h*400mN = 1600 nm. For each indention, we measured the edge lengths for the indentation *a*<sup>n</sup> and the crack lengths *l*n1, *l*n2, *l*n3, where n is the number of indentation. For each indentation, we determined the average values of *l*<sup>n</sup> according with equation *l*<sup>n</sup> = (*l*n1 + *l*n2 + *l*n3)/3. Then, for each tested coating, the average values of *l* and *a* were determined for the whole series of 20 indentations in accordance with relation Equations (2) and (3). Figure 5a,b shows an example of series of indentations made for the Ti-B-W (6%) coating and the results of the indentation *n* = 1.

$$a = (a\_{n=1} + a\_{n=2} + \dots + a\_{n=20}) / 20 \tag{2}$$

$$l = (l\_{n=1} + l\_{n=2} + \dots + l\_{n=20}) / 20\tag{3}$$

**Figure 5.** SEM images of groups indentations used for fracture toughness analysis (*K*IC) for Ti-B-W (6%) coating: (**a**) series of 20 indentations with a Berkovich indenter at a load of *P* = 200 mN; (**b**) dimensional analysis of crack lengths for different indentations where *n* = 1, 2, 5, 6.

Based on the material tests and measurements of the crack lengths in the corners of the indentations made of the Berkovich indenter, Table 4 summarizes the parameters that are needed to determine the *K*IC for the tested coatings, i.e., TiB2, Ti-B-W (3%), Ti-B-W (6%) and Ti-B-W (10%). The value of the coefficient (*x*v) in the Lougier model depends on the geometry of the indenter, which was used. According to the analysis carried out by N. Cuadrado et al. [13], for materials such as SiC, Si and soda-lime glass, in the case of Berkovich's geometry the value of coefficient (*x*v) = 0.022 ± 0.001.

**Table 4.** Results of calculation of fracture toughness (*K*IC), hardness and Young modulus for all investigated coatings.


The analysis carried out according to the Laugier model showed that, as a result of doping the TiB2 coating with tungsten, there is a significant increase in its fracture toughness. When the tungsten concentration increases up to 10%, the cracks that were generated in the corners of the indentation made with Berkovich indenter become shorter. For concentrations of 3%–6% W, the fracture toughness of Ti-B-W coatings achieves a value comparable to *K*IC suitable for TiN [14] and CrN [15] coatings. The fracture toughness of Ti-B-W with 10 at.% W is *K*IC [TiBW (10%)] = 4.98 and is nearly 7.5 times higher than for the TiB2 coating *K*IC [TiB2] = 0.67.

#### **4. Discussion**

In the article, the authors presented how the changes in the fracture toughness *K*IC for TiB2 and Ti-B-W coatings depend on the atomic % tungsten concentration (at.% W), as shown in Figure 6a. Observations of the brittle fracture cross-section showed that the TiB2 coating has a column structure, where the grain diameter is ≈100 nm (Figure 3a). The brittle fracture cross-section in the TiB2 coating has a clear intergranular character. In Figure 6b, the authors presented a scheme of the microstructure of the TiB2 coating, where the direction of cracking is parallel to the direction of growth of pillar grains (perpendicular to the surface of the coating). The presented diagram explains the high brittleness of the TiB2 coating and the low value of the fracture toughness *K*IC (*K*IC [TiB2] = 0.67).

Doping the TiB2 coating with 3% tungsten reduces column grains to a diameter of ≈30 nm (Figure 3b). Figure 6c proposes a diagram of the microstructure of the Ti-B-W (3%) coating, where grain refinement does not change the cracking mechanism, which still works mainly in a direction perpendicular to the surface. However, increasing the hardness and Young's modulus of the Ti-B-W (3%) coating, and consequently also increasing the plasticity index of coating *H*/*E* and resistance to plastic deformation of coating *H*3/*E*2, results in increased resistance to brittle cracking (*K*IC [TiBW (3%)] = 1.84).

For Ti-B-W coatings, which contain 6% tungsten, the column structure with diameter of grains 30 nm also dominates (Figure 3c). The cracking mechanism is noticeably changed, where the directions of parallel and perpendicular cracking to the surface of the coating are equivalent. At the brittle fracture cross-section, the columnar grains were observed, which cracked in a direction parallel to the surface of the coating. In Figure 6d, the authors proposed a scheme of the Ti-B-W (6%) coating microstructure, where columnar grains change their structure and growth directions due to tungsten segregation and the possibility of new phases appearing, e.g., WB4. The occurrence of local differences in the microstructure and phase structure justifies the increase in the hardness of the TiBW (6%) coating (*H*TiBW (6%) = 37 GPa) and the possibility of cracking energy dissipation by changing the direction of cracking and its separation into several others.

In the Ti-B-W coating, obtained by the TiB2 coating doped with 10% tungsten, one can observe a compact columnar structure with the individual columns, which are agglomerates of equiaxed grains with a diameter ≈100 nm. This allows us to make conclusions regarding the possible occurrence of nano-composite microstructure in Ti-B-W (10%) coatings, which was confirmed by the results of the research presented by Sobol et al. [16]. In Figure 6e, the authors presented a scheme of the Ti-B-W (10%) coating microstructure, where the cracking process is not oriented, but indicates the possibility of energy dissipation during cracking. In addition, the doping 10% of tungsten in the structure of the TiB2 coating results in a large increase in hardness and Young's modulus, i.e., *H*TiB2 = 34 GPa → *H*TiBW (10%) = 38GPa and *E*TiB2 = 405 GPa → *E*TiBW (10%) = 435 GPa, and the increase plasticity index *H*/*E* of coating and resistance to plastic deformation *H*3/*E*<sup>2</sup> of Ti-B-W (10%) coatings, which are respectively: *H*/*E*TiBW (10%) = 0.087 and *H*3/*E*<sup>2</sup> TiBW (10%) = 0.289. As a result, the fracture toughness is significantly increased: *K*IC TiBW (10%) = 4.98.

**Figure 6.** The results of the fracture toughness *K*IC analysis for TiB2 and Ti-B-W coatings: (**a**) changes of values of *K*IC for TiB2 and Ti-B-W coatings depending on the tungsten concentration (at. % W); (**b**) scheme of microstructure TiB2 coating; (**c**) scheme of microstructure Ti-B-W (3%) coating; (**d**) scheme of microstructure Ti-B-W(6%) coating; (**e**) scheme of microstructure Ti-B-W (10%) coating.

#### **5. Conclusions**

The article demonstrated that the brittleness of thin TiB2 ceramic coatings can be effectively improved by tungsten doping. Tungsten concentration in the range of 0–10% can significantly changes the microstructure of coatings, from a typical columnar structure with a grain diameter of about 100 nm for TiB2, to a clear nano-composite structure for Ti-B-W (10%). Analysis of the tested TiB2 and Ti-B-W coatings, including: brittle fracture cross-section, changes in hardness (*H*) and Young's modulus (*E*) and plasticity index *H*/*E* and resistance to plastic deformation *H*3/*E*2, enabled the development of microstructure diagrams for TiB2 coatings and Ti-B-W coatings with different tungsten concentration. Observations of brittle fracture cross-section show that, in the case of doping TiB2 coatings with tungsten at an amount of 10%, the microstructure is fragile and the nano-composite structure is created. Then it is possible to change the direction of cracking many times and to disperse it into several other directions. The energy of a single crack is reduced, which often results in the disappearance of cracks. As a result, in such a coating we observed a significant increase in values of fracture toughness *K*IC.

To assess the fragility of the tested coatings, the authors used the Berkovich indenter induction method and the calculation of the fracture toughness *K*IC according to Laugier model. Doping of TiB2 coating show that 10% tungsten causes a more than 7-times increase in the fracture toughness KIC from *K*IC [TiB2] = 0.67 to *K*IC [TiBW (10%)] = 4.98.

At the same time, Ti-B-W (10%) coatings are characterized by greater hardness and comparable surface roughness compared to TiB2 coating, which makes them coatings with a very high application potential.

**Author Contributions:** Conceptualization, J.S.; methodology, J.S., J.K.-G., and S.S.; investigation, S.S. and A.P.; data curation, J.S, J.K.-G., and S.S.; writing—original draft preparation, J.S.; writing—review and editing, S.S. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

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

#### **References**


© 2020 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*
