**Aluminum Coated Micro Glass Spheres to Increase the Infrared Reflectance**

**Laura Schwinger 1,\*, Sebastian Lehmann 1, Lukas Zielbauer 1, Benedikt Scharfe <sup>2</sup> and Thorsten Gerdes <sup>1</sup>**


Received: 4 February 2019; Accepted: 8 March 2019; Published: 12 March 2019

**Abstract:** The reflective properties of micro glass spheres (MGS) such as Solid Micro Glass Spheres (SMGS, "glass beads") and Micro Hollow Glass Spheres (MHGS, "glass bubbles") are utilized in various applications, for example, as retro-reflector for traffic road stripe paints or facade paints. The reflection behavior of the spheres can be further adapted by coating the surfaces of the spheres, e.g., by titanium dioxide or a metallic coating. Such coated spheres can be employed as, e.g., mid infrared (MIR)-reflective additives in wall paints to increase the thermal comfort in rooms. As a result, the demand of heating energy can be reduced. In this paper, the increase of the MIR-reflectance by applying an aluminum coating on MGS is discussed. Aluminum coatings are normally produced via the well-known Physical Vapor Deposition (PVD) or Chemical Vapor Deposition (CVD). In our work, the Liquid Phase Deposition (LPD) method, as a new, non-vacuum method for aluminum coating on spherical spheres, is investigated as an alternative, scalable, and simple coating process. The LPD-coating is characterized by X-ray diffraction (XRD), energy dispersive X-ray spectroscopy (EDX), scanning electron microscopy (SEM), and reflection measurements. The results are compared to a reference PVD-coating. It is shown that both sphere types, SMGS and MHGS, can be homogeneously coated with metallic aluminum using the LPD method but the surface morphology plays an important role concerning the reflection properties. With the SMGS, a smooth surface morphology and a reflectance increase to a value of 30% can be obtained. Due to a structured surface morphology, a reflection of only 5% could be achieved with the MHGS. However, post-treatments showed that a further increase is possible.

**Keywords:** micro hollow glass spheres (MHGS); solid micro glass spheres (SMGS); liquid phase deposition (LPD); aluminum coating

#### **1. Introduction**

Micro glass spheres (MGS), such as micro solid glass spheres and micro hollow glass spheres, are utilized in various technical fields due to their excellent physical and chemical properties. Solid Micro Glass Spheres (SMGS) exhibit high strength, as well as smooth surfaces, and are excellently suited as grinding and dispersing balls [1]. They are also utilized as fillers in thermoplastics and thermosets to enhance the physical properties of the matrix, like increasing the Young Modulus and hardness [2–4]. Micro Hollow Glass Spheres (MHGS) are characterized by a low density, as well as by low thermal conductivity. Applied as fillers, e.g., in polymers or building materials, a decrease in weight and minimization of thermal conductivity is possible and could lead to energy savings, especially in the automotive and building industry [5–11].

Possible applications of MGS as additives in the building industry are exterior and interior paints or plasters. MHGS, for example, are processed in facade paints and plasters. SMGS, on the other hand, are used as reflection pigments in marking strips on motorways for improved night visibility in rain and fog as a result of their reflective properties [10,12]. The application as reflection pigment can also be advantageous for wall paints. In the exterior area, reflectance in the visual and near infrared wavelength range (VIS/NIR, 0.4–2.5 μm) plays an important role due to solar radiation, whereas in the interior area, reflection in the mid infrared wavelength range (MIR, 3–50 μm) is important. An increased MIR-reflectance in the range of human black body thermal radiation (~5–30 μm), for example, can increase the thermal comfort in rooms, and thus contribute to saving thermal energy. A beneficial aspect is that the reflection behavior of the spheres can be further modified, e.g., by coating the spheres. For increasing the reflectance behavior in the VIS/NIR range, titanium dioxide coatings are applicable. To improve the reflectance in the MIR-range, metallic coatings, like silver coatings can be applied [13–16].

Due to its properties as a noble metal, silver coatings can be applied rather easily on many materials via electroless deposition [17,18]. Yet there are disadvantages, especially in the application as reflective coating for fillers in paints. First, silver is one of the more expensive precious metals, which would lead to high costs when large amounts of materials are required. Secondly, the high oxidation potential of silver makes silver coatings prone to chemical degradation, which gradually causes an intolerable visible change in the color of the spheres and so of the paint. This also may influence the reflectance [19]. Additionally, silver has a poor recyclability as a color pigment.

In the given case, aluminum is a suitable alternative coating material. It also exhibits a high reflectance in the mid-wave infrared range but is better recyclable and also cheaper than silver. In addition to a low oxidation potential, aluminum films have a good thermal stability and a good adherence to substrates [20]. Thus, none or less influence on the reflection behavior by degradation of aluminum is expected. However, the coating process with aluminum is much more complex than the process with silver. Because of its properties as not noble metal, aluminum cannot simply be deposited by means of electroless deposition. Instead, thin films of aluminum on surfaces are mainly applied via Physical Vapor Deposition (PVD) or Chemical Vapor Deposition (CVD) methods [21–23]. PVD advantageously produces high quality coatings, but is limited in batch size and materials to be coated. For an application as filler on the painting industry, large amounts of coated material are required. However, an upscaling of the PVD process quickly becomes expensive and very complex, not only because of the necessary vacuum process. With the CVD method a wide variety of materials can be deposited with a very high purity and without high vacuum conditions. A disadvantage is that high temperatures (200–1600 ◦C) are necessary, which results in higher process and energy costs and a limitation of processable substrates. In addition, to generate metallic coatings, metallic-organic precursors are necessary, most of which are highly toxic, explosive, or corrosive, as well as quite costly [23].

A possible alternative to achieve a scalable, more cost-effective process is provided by a chemical coating method, where a precursor is applied in a solvent and deposited on a substrate by means of a decay reaction. Lee at al. [24] have already shown that this Liquid Phase Deposition (LPD) method can be used to produce thin, highly conductive aluminum layers on soda-lime glass and polyethylene terephtalate.

In this paper, the practicability of the LPD method to coat MGS is examined. Applied to hollow glass spheres (MHGS, made of borosilicate glass) and micro solid glass spheres (SMGS, made of soda-lime glass), the achieved coatings are compared to coatings resulting from a PVD process concerning coating quality and homogeneity. Furthermore, the reflection behavior of the LPD-coated spheres is compared with uncoated and PVD-coated spheres in the wavelength range of 7.5 to 18 μm, which in part corresponds to the wavelength range of thermal radiation of a human, as well as to the MIR atmospheric window [25]. The modification of reflection as well as the influence of the different coating techniques and layer qualities on the reflective properties are discussed.

#### **2. Experimental**

#### *2.1. Materials*

#### 2.1.1. Micro Glass Spheres

Three types of spheres were investigated: Micro Hollow Glass Spheres (MHGS) iM16K-ZF and S38HS from 3M/Dyneon Germany and Micro Solid Glass Spheres (SMGS) Type S from Sigmund Lindner GmbH (SiLi). Relevant properties of these glass spheres are listed in Table 1. The MHGS possess low density and high compressive strength, which can guarantee processability and applicability, e.g., as light filler in wall paints. MHGS type S38HS bear an additional anti-agglomeration agent on their surfaces. To ensure a good comparability between spheres of different consistence, hollow, and solid spheres with a similar reference value D 50 (median of particle size, 50 % of the particles are smaller than the declared value) were chosen.

**Table 1.** Relevant properties of the examined Micro Solid Glass Spheres (SMGS) and Micro Hollow Glass Spheres (MHGS).


\* laboratory product, not commercially available; \*\* median of particle size, 50 % of the particles are smaller than the declared value; \*\*\* according to the supplier.

#### 2.1.2. PVD-Coated Reference Spheres

PVD-coated spheres are used as a reference material with regard to layer morphology and reflectance. As listed in Table 2, batches of all types of the investigated spheres were coated with different layer thicknesses of aluminum at the University of Vienna. For this purpose, a PVD process especially designed for spherical particles was utilized [21,22]. During the coating process, a strong agglomeration behavior especially for sphere type iM16K-ZF, without anti-agglomeration agent on the surface, was observed. However, these agglomerates could easily be broken up by vibratory plate and sieving without damaging the MHGS.


**Table 2.** Coating parameters of the aluminum coated glass spheres via Physical Vapor Deposition (PVD) at the University of Vienna.

Figure 1 exemplarily shows SEM images of PVD-coated spheres with the highest layer thickness for each type. The entire surface of the spheres is uniformly coated with aluminum and no free spots are present. The surface morphology of the PVD-coated spheres is almost homogeneous with only small imperfections that may be caused either by impurities or by the anti-agglomeration-agent on the surface of the commercial spheres.

**Figure 1.** Exemplary SEM-images of the reference micro glass spheres coated via PVD ((**a**): iM16K-ZF: 19 nm; (**b**): S38HS: 34 nm; and (**c**): Type S: 36 nm). An almost homogeneous aluminum coating with some small imperfections on all three types of spheres is achieved on all three types of spheres.

#### *2.2. Liquid Phase Deposition (LPD)*

#### 2.2.1. Pre-Conditioning

In contrast to the PVD-coating process, the anti-agglomeration agent as well as possible impurities on the surface of the spheres may act as single seed particles during the LPD-coating process, which favor growth on single spots and thus lead to inhomogeneous layers. To create equal starting conditions for the LPD process and obtain preferably homogeneous aluminum layers, a two-step pre-conditioning of the spheres was performed. First, the MGS were washed three times with ethanol and distilled water, and then dried in a drying cabinet for 24 h at 80 ◦C to clean the surface of the spheres from impurities. To further support a homogenous coating with aluminum, a calcium silicate nanoparticle layer was created. Following Jin et al. [13], 10 g of the cleaned MGS were then dispersed into a saturated calcium hydroxide solution (Ca(OH)2, Sigma Aldrich) for 4 h at 90 ◦C. Afterwards, the spheres were filtered and again stored in the drying cabinet for 24 h at 80 ◦C before coating.

A comparison of untreated MGS to the pre-conditioned MGS is shown in SEM pictures in Figure 2. The untreated surface of the MGS is nearly smooth. The visible small agglomerates (indicated by the white arrows) are either impurities resulting of the manufacturing process, or in the case of the sphere type S38HS, the additional anti-agglomeration agent. After pre-conditioning, calcium silicate nanoparticles are deposited on the surface of the three sphere types. Nevertheless, it appears that more calcium silicate nanoparticles are deposited on the MHGS than on the surface of Type S. This may be due to the different glass types. It was additionally noticed that, for sphere type iM16K-ZF, it agglomerates with sizes of 10–20 spheres that occur after the pre-conditioning. This is explained by the absence of the anti-agglomeration agent. It is presumed that these small clusters are broken up by mechanical means in the later coating process, similar to the reference spheres.

**Figure 2.** Comparison of untreated Micro Glass Spheres (MGS) and pre-conditioned MGS with Ca(OH)2. The white arrows indicate either impurities resulting from the manufacturing process, or in the case of the sphere type S38HS, the additional anti-agglomeration agent. After pre-conditioning, calcium silicate nanoparticles are deposited on the surface of each sphere type. In contrast to the PVD-coating, the spheres were pre-conditioned to reduce the amount of single seed particles, which would lead to an inhomogeneous aluminum growth. Untreated MGS: (**a**) iM16K-ZF, (**c**) S38HS, (**e**) Type S; pre-conditioned MGS: (**b**) iM16K-ZF, (**d**) S38HS, (**f**) Type S.

#### 2.2.2. LPD of Aluminum

The theoretical reaction step for the LPD of aluminum, as proposed by Lee et al [24] and Brower et al. [26], is shown in Equation (1), the individual reaction steps in Equations (2)–(4).

$$\text{3. LiAlH}\_4 + AlCl\_3 \overset{O(C\_4H\_9)\_2}{\rightarrow} \text{3LiCl} + 6\text{ }H\_2 + 4\text{ }Al\text{ }\tag{1}$$

$$3\text{ }LiAlH\_4 + AlCl\_3 + 4\text{ }O(\text{C}\_4H\_9)\_2 \rightarrow 3\text{ }LiCl + 4\text{ }[OAlH\_3(\text{C}\_4H\_9)\_2].\tag{2}$$

$$4\left[OAlH\_3(\mathbb{C}\_4H\_9)\_2\right] \to 4\left[OAl(\mathbb{C}\_4H\_9)\_2\right] + 6\,H\_{2\prime}\tag{3}$$

$$4\left[OAl(\mathbb{C}\_4H\_9)\_2\right] \to 4\left[Al\downarrow + 4\left[O(\mathbb{C}\_4H\_9)\_2\right]\right] \tag{4}$$

First the precursor aluminum trihydride dibutyl etherate is produced by reacting 3 M lithium aluminum hydride (LiAlH4, Sigma Aldrich) with 1 M aluminum chloride (AlCl3, Sigma Aldrich) in 2 × 50 mL of anhydrous dibutyl ether (Sigma Aldrich) for 120 min at room temperature (2) [21,22]. The generated precursor is then separated from the byproduct lithium chloride (LiCl) by filtration and subsequently mixed with the Micro Glass Spheres (MGS). Afterwards the precursor is decomposed at a reaction temperature of 130 ◦C (3,4), where, according to Equation (4), the aluminum supply is related to reaction time as well as the amount of the precursor.

The coating process itself was carried out with two different standard laboratory setups, which also seem appropriate for a technical scaling. In Setup 1, a round flask with magnetic stir bar and reflux condenser under nitrogen atmosphere was employed. Setup 2 comprises a rotating round flask also under a nitrogen atmosphere. For testing and optimizing the coating process of the MGS, several experiments were carried out (Table 3). First, SMGS were coated with different reaction times (120, 240, and 300 min) in both setups to determine a reasonable coating time. Due to the preprocessing step, 10 g of SMGS are employed in each attempt. By transferring this amount of spheres in a graduated cylinder, a bulk volume of 6.5 mL was determined, which has to be taken into account due to the different bulk densities of the investigated sphere types. In addition, experiments with the MHGS were also carried out in both setups.


**Table 3.** Parameters of the individual coating experiments.

#### 2.2.3. Post-Treatment

As for the PVD-coated spheres, agglomerations were also observed in the PVD process. As mentioned before, the agglomerated spheres can be broken up by mechanical means. Therefore, different post-treatments were applied to the LPD-coated spheres to further investigate the disaggregation of clustered spheres. Simultaneously, a possible influence on the surface morphology and the stability of the aluminum coating on the sphere surface can be tested.

#### Turbula®Mixer:

Agglomerates of the PVD-coated spheres could be broken up by vibrational movement, so the LPD-coated spheres were first treated with a Turbula®Mixer (Willy A. Bachofen AG, Muttenz, Switzerland). Different mixing times (2, 4, 6, 8, 12, and 24 h) were investigated.

#### Magnetic Stir Bar Treatment:

To further examine the mechanical influence of the magnetic stir bar utilized in Setup 1, the coated spheres were placed in water and stirred for 12 h. Water was chosen as solvent as it allowed the simultaneous investigation of how the coated spheres behave in a slightly alkaline milieu. This behavior is interesting for a later possible use as additive in water-based wall paints.

#### Furnace Treatment:

In addition to the mechanical influences, a thermal influence on the coated spheres was also investigated.

A possible homogenization or smoothing of the coating layer by sintering or melting is pursued by heating up coated spheres in a furnace to different temperatures with a gradient of 10 ◦C/min in a forming gas (Ar/H2 95/5) atmosphere. Two temperatures below (400 and 500 ◦C) and one temperature above (750 ◦C) the melting temperature Tm of aluminum (approx. 660 ◦C) were chosen. The holding time was set to 2.5 h.

#### *2.3. Experimental Characterization Methods*

To determine coating quality and reflective properties, each batch of spheres was subjected to a detailed experimental investigation by the following characterization methods.

#### 2.3.1. Qualitative Phase Composition

The coating was analyzed by X-Ray structure analysis (XRD) to determine if pure aluminum was deposited on the surfaces of the spheres. It was carried out with an X'-Pert MPD PW 3040 X-ray diffractometer from Philips (Philips Co., Netherlands). The individual samples were measured using Cu Kα radiation in a range from 10–90◦ 2Θ. The step speed was 0.03◦ with a holding time of 30 sec. The obtained spectra were analyzed with the software X' Pert High Score Plus 4.1 (Malvern Panalytical, Malvern, UK).

To analyze the elemental phase composition of the coating, an Energy Dispersive X-Ray analysis (EDX) was conducted. The measurements were carried out with a Thermo Fisher Ultra Dry SDD Silicium Drift Detector (Thermo Fisher, Darmstadt, Germany) with an excitation voltage of 20 kV.

#### 2.3.2. Morphology Characterization

The morphology of the coated spheres was observed using a Zeiss LEO 1530 FESEM scanning electron microscope (SEM) (Zeiss, Oberkochen, Germany) at the Bavarian Polymer Institute (BPI) of the University of Bayreuth. The excitation voltage was set to 3 kV. In advance, the samples were sputtered with a 1.3 nm thick platinum layer by a 208HR sputter coater (Cressington Scientific Instruments, Dortmund, Germany).

#### 2.3.3. Reflection Measurement

MIR-range reflection measurements were performed with a FTIR instrument (Vertex70 from Bruker Optic GmbH, Ettlingen, Germany) and an integrating sphere (U-Cricket TM-BR4 from Harrick Scientific Products, Darmstadt, Germany) under a nitrogen atmosphere against a gold standard. The measuring range of the Vertex 70 is approx. 600 to 7000 cm<sup>−</sup>1, which corresponds to a wavelength range of approx. 1.5 to 18 μm. The samples were applied as a monolayer to an adhesive strip and then placed on the opening of the integrating sphere. The reproducibility was confirmed by repeated preparation and measurements of the powder samples. An error of less than 0.01% was determined.

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

#### *3.1. Morphological Examination*

#### 3.1.1. Setup 1—Round Flask with Magnetic Stir Bar

Experiment 1 was carried out in a round flask with a magnetic stir bar and reflux condenser under a nitrogen atmosphere. The color change from white to gray after each coating test qualitatively demonstrates the metallization of the spheres (Figure 3)

**Figure 3.** Coating of Type S spheres after a coating time of 240 min: The original spheres have a white color, the Al-coated spheres show a gray coloration.

In Experiment 1 A coating tests with 10 g of spheres of Type S and coating times of 120, 240, and 300 min were performed. As indicated in Figure 4, a coating is achieved for all investigated times. However, an increased coating time leads to smoother surfaces. The interaction of the friction of the magnetic stir bar, which is used for mixing of the coating solution during the coating process, and the necessary reaction time for a complete decomposition are identified as possible reasons.

**Figure 4.** SEM images of Type S spheres coated via LPD-process (Experiment 1, 10 g) at different coating times: (**a**): 120 min; (**b**): 240 min; and (**c**): 300 min. As the coating time increase, the aluminum coating morphology becomes smoother due to the reaction process and the influence of the magnetic stir bar; (**d**) EDX-Mapping of a coated sphere Type S.

After 120 min, the reaction does not seem to be completed, and therefore, additional aluminum was deposited on the surface of the spheres despite the friction of the stir bar. Between 120 and 240 min, however, the reaction was completed and the friction of the stir bar smoothed the surface. It also becomes clear that the aluminum must be firmly fixed to the sphere surface, since no abrasion of the coating takes place. This will be examined more closely in the section of the post-treatment (3.1.3). As only little differences are obtained between coating times of 240 and 300 min, the coating time of 240 min was chosen for further experiments.

Compared to the PVD-coated spheres, coating times of 240 and 300 min results in a homogeneous coating and smooth surface morphology. The uniform application of the aluminum coating is also confirmed by the EDX mapping. The results of X-Ray diffraction analysis are plotted in Figure 5. The uncoated spheres show an amorphous halo in the range of 20–40 Θ, which is characteristic for amorphous materials like glass. Besides the halo, the aluminum coated samples deliver characteristic peaks of metallic aluminum at 38◦ (111), 45◦ (200), 65◦ (220), 78◦ (211), and 82◦(222). Thus, it can be verified that pure metallic aluminum is deposited on the sphere surfaces with the liquid phase deposition (LPD) method.

Exemplary EDX-Mapping-Measurements (d) of a coated sphere of Type S (Experiment 1) shows a uniform aluminum coating (red) on the surface of the spheres after a coating time of 240 min.

**Figure 5.** Qualitative comparison of X-Ray Diffraction of uncoated and LPD-coated Type S spheres. Besides the characteristic halo of amorphous materials, all characteristic peaks of metallic aluminum are detected on the coated spheres.

While for the spheres of Type S a rather homogenous and smooth coating was possible, attempts to coat the MHGS spheres type S38HS with Setup 1 was not successful. Contrary to the SMGS, the MHGS are floating on top of the reaction solution due to their significant lower density. To enforce a complete mixing, the speed of the magnetic stir bar had to be increased. The higher shear forces caused an increased breakage of the spheres, especially for type S38HS. This is justified by the low compressive strength of this sphere type (Table 1). Due to the higher compressive strength, type iM16K-ZF did not show an increased sphere breakage under these conditions, but a strong agglomeration and a very inhomogeneous coating were observed.

#### 3.1.2. Setup 2—Rotating Flask

To eliminate the damage of the spheres by the magnetic stir bar and the influence of floating, especially for the MHGS, Experiment 2 A and 2 B were carried out in a rotating flask under nitrogen atmosphere. Again, 10 g or 6.5 mL of spheres of Type S were coated to obtain a comparison to the results attained with round flask and magnetic stir bar. Figure 6 left exemplarily shows a SEM image of a coated sphere of Type S. Compared to Setup 1 partly bigger agglomerations stick on the surface (indicated by white arrows). In addition, clumping of the spheres take place. This is explained with the absence of the magnetic stir bar, which smooths the surfaces of the spheres by friction and probably is breaking up agglomerates. This will be examined in more detail in Chapter 3.1.3. Nevertheless, also uniform coated spheres could be obtained with Setup 2, which is confirmed by the EDX mapping (Figure 6b).

**Figure 6.** (**a**): SEM image of Type S coated via LPD-process in a rotating flask (Experiment 2 A). A complete aluminum coating with bigger agglomerates was achieved; and (**b**): EDX-Mapping of a coated sphere Type S (Experiment 2 A). A homogenous aluminum coating (red) could be proven on the surface of the sphere.

In Experiment 2 B a defined volume of 6.5 mL of each type of MHGS was coated. The color change of the spheres from white to gray after each coating test once again also indicates a successful metallization of the spheres (Figure 7).

**Figure 7.** Qualitative verification of the coating of the MHGS with Setup 2: The original spheres have a white color, the coated spheres show a gray coloration.

SEM images and EDX mappings in Figure 8 confirm that a homogenous coating on both types of MHGS can be reached in Experiment 2B. No ball breakage occurs. The XRD results again indicate a metallic aluminum coating (Figure 9).

In comparison with the reference spheres and the coated spheres of Type S of Experiment 1 A, however, a more structured morphology results. This is explained by the reduced friction in the rotary flask compared to the stirring, which led to a smoother surface structure in Experiment 1. In addition, both types of MHGS now show a partly agglomeration (10–20 spheres). This, nonetheless, is not a problematic result as the post-treatment is intended to break up these agglomerates.

**Figure 8.** SEM images and EDX-Measurements of MHGS ((**a,b**): iM16K-ZF and (**c,d**): S38HS) coated via LPD-process in a rotating flask (defined volume: 6.5 mL, Experiment 2 B). In comparison to the reference spheres, a more structured aluminum coating could be obtained on both types of spheres.

**Figure 9.** Qualitative comparison of X-Ray Diffraction of uncoated and LPD-coated MHGS. Besides the characteristic halo of amorphous materials, all characteristic peaks of metallic aluminum are detected on the coated spheres.

#### 3.1.3. Post-Treatment

In the previous chapters, it was shown that, with the LPD-method, it is possible to metallize a rather uniform aluminum layer on micro hollow and micro solid glass spheres. Different results regarding layer morphology were achieved and partly agglomeration was observed. As agglomerates of PVD-coated spheres could be broken up by mechanical means, additional post-treatments were carried out to investigate whether the small agglomerates of the LPD-coated spheres can also be broken up by mechanical means. In addition, the stability of the coating as well as the surface morphology after the treatments were examined. Coated sphere types iM16K-ZF and Type S of Experiment 2 were used for the post-treatments.

As already investigated, friction by the magnetic stir bar effects a smoothing of the aluminum coating of the SMGS, whereas MHGS show increased ball breakage. Friction in the rotating flask caused no breakage of the MHGS but is not sufficient to smooth the aluminum coating. Therefore, the rotating friction, simulated in a Turbula®Mixer, is increased by raising the time in order to achieve a smoothing effect. In addition, agglomerated spheres could be broken up. Hence, the coated spheres were first treated in a Turbula®Mixer for different times (2, 4, 6, 8, 12, and 24 h).

Figure 10 exemplarily shows SEM-images of spheres of Type S treated for 24 h. The agglomerates present in the untreated sample are almost completely broken up by the treatment. This can also be seen with sphere type iM16K-ZF. Thus, with Setup 1 and Setup 2, single coated spheres can be obtained. This is advantageous, because single coated MGS can be processed more easily. Further, the overall surface is increased, which is positive for the reflective behavior. In addition, there is also a small change in the surface morphology. The previously structured surface can be smoothed slightly, which in turn again can have a positive effect on the reflective behavior. The aluminum coating is firmly bonded to the surface of both spheres, and no break off of the coating results by friction. This can have a positive influence on the potential use in wall paints, where fillers are also included in paints by mixing.

**Figure 10.** SEM images of the untreated ((**a,c**) and post-treated (Turbula®Mixer; (**b,d**)) SMGS Type S (Experiment 2 A). Breaking off of the agglomerates and no change in surface morphology is evident.

A second post-treatment method was carried out to investigate the influence of the magnetic stir bar in Setup 1. It was already shown that the magnetic stir bar breaks up the agglomerates during the coating process and smooths the surface of the spheres. It was tested whether this is also possible afterwards for coated spheres of Setup 2. Therefore, coated spheres of Type S (Experiment 2A) were transferred into distilled water and stirred for 12 h. As explained above, distilled water was used to equally test the influence of a slightly basic milieu. No color change of the post-treated spheres occurs, which means that no influence on the aluminum layer results from the slightly basic milieu. Again, the agglomerates could be broken up by the post-treatment with magnetic stir bar. Additionally, it can be recognized in Figure 11 that there is a partially significant change of the surface structure. On some spheres, a very smooth surface, like in Experiment 1, results. The fact that it only occurs partially may be due to different sphere-solvent ratios (2 g/100 mL) compared to the coating test (10 g/100 ml). Thus, it is possible that not all spheres get the same level of friction. This treatment, however, could only be carried out with the spheres of Type S, as the problem of floatation occurred with the sphere type iM16K-ZF.

Considering the stability of the coating, no flaking or destruction of the aluminum layer took place even after 12 h. Thus, the aluminum coating sticks firmly on the surface of the spheres of Type S and can also be smoothed by a post-treatment with a magnetic stir bar.

**Figure 11.** SEM images of the post-treated (with magnetic stir bar, after 12 h) SMGS Type S (Experiment 2 A). Partially, a significant change in morphology takes place.

Since the agglomerates can be broken up and the surface morphology can be changed by friction, a third treatment was carried out to investigate whether a thermal treatment has an influence on the morphology of the coating, too. As already shown in Experiment 2 the coated spheres showed a more structured surface morphology. The furnace treatment is intended to show whether smoothing the coating is possible by sintering or melting. Therefore, the post-treatment method consists of heating up the coated spheres to different temperatures (400, 500 and 750 ◦C, 10 ◦C/min, no holding time, Ar/H2 95/5 atmosphere) in a furnace.

Exemplarily SEM images of coated Type S spheres are shown in Figure 12. For furnace temperatures of 400 and 500 ◦C no significant change of the aluminum layer occurs. Agglomerates of single spheres were formed, indicating that a sintering of the layer took place. The sticking together could possibly be prevented by a rotating chamber. However, the temperature is not sufficient to completely smooth the aluminum layer on the surface of the spheres. For a furnace temperature of 750 ◦C, in contrast, a noticeable change in the aluminum layer can be detected. Smaller crystal-like and worm-like structures were formed, which makes the sphere surface much more inhomogeneous than it was before. This structural change is caused by the recrystallization of aluminum, as was also described from K. D. Vanyukhin et al. [27]. A melting and thus a smoothing of the coating does not take place.

**Figure 12.** SEM images of the post-treated (furnace) SMGS Type S (Experiment 2 A). No significant change in surface morphology at a temperature of 400 ◦C (**a**) and 500 ◦C (**b**) is apparent. For a temperature of 750 ◦C (**c,d**) a more inhomogeneous surface occurs due to the recrystallization of aluminum.

Since neither friction nor heat causes a break off or flaking of the aluminum layer, it can be assumed that a stable aluminum coating results from the LPD method. A positive effect of the post-treatment with the Turbula®Mixer and the magnetic stir bar, that agglomerates can be broken, which makes the spheres more easily processable and increases in reflective surface. Additionally, a partially significantly change in surface morphology occurred for the coated SMGS. Thus, through sufficient friction, a smooth coating morphology, similar to that of the reference spheres, can be achieved on solid glass spheres.

#### *3.2. Reflection Measurements*

In Chapter 3.1, it was shown that it is possible to get a rather homogeneous aluminum coating on micro hollow and solid glass spheres with the LPD method. Since the aim of the work is not only to investigate the possibility to metallize the MGS, but also to explore the increase of the reflection in the MIR wavelength range, reflection measurements were carried out in addition to the optical and morphological evaluation using SEM images and as well as XRD and EDX. The question whether the reflection can be increased due to the properties of aluminum in contrast to the uncoated MGS was investigated. Further, the reflectance of the LPD-coated spheres and PVD-coated spheres was compared to examine the influence of different surface morphologies.

As an adhesive strip is utilized to apply the powdery samples to the opening of the integrating sphere of the measurement device, a possible influence of the strip was first examined. As can be seen in Figure 13 (black curve), the adhesive strip has a nearly 0% reflectance in the chosen measurement range between 7.5 and 18 μm and thus can be neglected. The samples were sieved beforehand to ensure that only coated spheres and no Al dust were measured. Furthermore, the influence of the pre-conditioning with Ca(OH)2 was investigated by comparing the reflectance of pre-conditioned MHGS and SMGS with untreated spheres in the same range (Figure 13). Below 10 μm, soda-lime and borosilicate glasses have absorptions bands, above reflectance values between 3% and 10% arise.

The influence of the pre-conditioning can be neglected for the sphere type Type S, since the pre-conditioned micro spheres exhibit nearly the same reflectance in the wavelength range of interest as the original spheres (Figure 13, blue curves). For the sphere types iM16K-ZF (Figure 13, red curves) and S38HS (Figure 13, green curves), a slight difference in the reflection can be observed between 9 and 11 μm and for wavelengths greater than 17 μm. The latter one is explained by the resolution limit of the device, the differences in the smaller wavelength range may be due to an inhomogeneous coating.

**Figure 13.** Reflectance behavior of the adhesive strip, the original spheres and the pre-conditioned spheres in the wavelength range of 1.6 to 25 μm. Less or no influence of the adhesive strip and pre-conditioning in the wavelength range of interest (10 to 18 μm) is assumed.

#### 3.2.1. PVD-Coated Spheres

First, the reflectance values of the PVD-coated reference spheres were determined with the described setup and a correlation between different layer thicknesses and reflection was evaluated. At lower layer thicknesses, the curve still resembles the reflection curve progression of uncoated spheres. With increasing layer thickness, almost constant reflection values emerge, which is typical for aluminum (Figure 14a–c). Thus, the almost constant curve-shape indicates a sufficient deposition of aluminum on the surface of the spheres, as no features of glass reflection are observed any more [23]. The decrease in reflectance observed for wavelength values greater than 17 μm is again related to the resolution limit of the measurement device.

**Figure 14.** Results of the reflection measurements of the PVD-coated MGS (reference spheres, (**a**): iM16K-ZF, (**b**): S38HS, (**c**): Type S) as well as the reflection at a wavelength of 15 μm of each sphere types plotted over the layer thickness (**d**). An increased reflection with increasing layer thickness can be observed.

It becomes apparent that, with the increasing layer thickness, an increase in reflection also takes place for all three sphere types. To show this more clearly, in Figure 14d the reflectance values of each sphere type at a wavelength of 15 μm is plotted against the layer thickness. A similar behavior was observed by Lugolole et al. [20]. Interestingly, a similar layer thicknesses of approx. 36 nm lead to different reflectance values. A higher reflection degree of the SMGS in comparison to the MHGS is achieved.

#### 3.2.2. LPD-Coated Spheres—Setup 1

In Figure 4, it was already shown that a very homogeneous, smooth aluminum coating can be achieved with Setup 1 and a coating time of 240 min for the spheres of Type S. The results of the reflection measurements of these LPD-coated spheres and the comparison with the PVD-coated spheres are represented in Figure 15.

Considering the reflectance measurements, it can be seen that a significant increase in reflectance is observed as a function of the coating time from approx. 5% to 35%. However, after a coating time of 300 min, no more increase of the reflectance occurs in comparison to a coating time of 240 min. It was already examined in Chapter 3.1.1, how this is in the line with no change in surface morphology of the two samples, as the available aluminum has already been completely deposited.

Compared to the reflectance of the reference spheres (layer thickness 36 nm) similar values are achieved for the LPD-coated spheres. Thus, using the LPD method for Type S, both a similar surface morphology and reflectance in comparison to the PVD-coated Type S can be obtained.

**Figure 15.** (**a**): Results of the reflection measurements of the LPD-coated Type S spheres (Experiment 1) with different coating times. Reflection increases with increasing coating time. (**b**): Comparison of the reflectance of LPD-coated (Experiment 1 A, 240 min) and PVD-coated (36 nm) spheres of Type S.

#### 3.2.3. LPD-Coated Spheres—Setup 2

The results of the reflectance measurements of the coated spheres from Experiment 2 are shown in Figure 16. It was already established by SEM images Figures 6 and 8 that all three types of spheres could be coated with aluminum, but in comparison to the reference spheres, a partly rougher surface morphology occurs. Despite the coated layer, for both types of spheres, no or less reflectance increase can be observed. This behavior was already established by Lugolole et al. [20], where the lowest reflectance was achieved with the sample, which has a rough surface. Das et al. [28], Moushumy et al [29], and Sharma et al [30] also show numerical simulations of different nano-structured gratings on GaAs substrates, that the surface structure has a significant influence on the reflection behavior. With a flat substrate, a reflection of about 28% can be achieved, whereas a structured surface shows only a reflection of about 2% [28].

Nevertheless, the reflectance curves of the coated spheres show an almost constant progress over the wavelength range and compensate at least partially the reflectance curve of the uncoated glass. This indicates that at least a sufficient amount of aluminum is coated on the spheres. Therefore, the uneven layer structure of the aluminum coating must be the determining factor.

**Figure 16.** (**a**): Results of the reflection measurements of the LPD-coated SMGS (Experiment 2 A) with an amount of 10 g. No increase in reflectance occurs. (**b**): Results of the reflection measurements of the LPD-coated MHGS (Experiment 2 B) with a volume of 6.5 mL. Further, no increase in reflection occurs.

#### 3.2.4. Post-Treated Spheres

As can be seen in Figure 17, the reflectance can generally be increased by all three post-treatments. They led to a partial change in surface morphology towards smoother surfaces. Only in the case of the furnace treatment over the melting temperature of aluminum, a surface deterioration was observed. Here the dependence of reflectivity surface can again be clearly determined. The very inhomogeneous surface structure (Figure 12a,b) which is reflected in a clear decrease in reflectance to almost 1%. On the contrary, the furnace treatment below the melting point is associated with a reflection increase of approx. 3% (Type S) and approx. 2% (iM16K-ZF), despite no change in the SEM images was observed. The partially smoothed surfaces after magnetic stir bar treatment lead to a slightly better increase of 5% (Figure 17c). Besides surface smoothening, agglomerates of spheres could be broken up by mixing. This also should increase the reflection of the spheres due to the enlargement of the total surface. After 24 h in the Turbula®Mixer, an actual increase in reflection of approx. 3% occurred for the spheres of Type S (Figure 17d). For type iM16K-ZF a similar increase of 2% could be achieved. Thus, the dissembling of the agglomerates not only facilitates the processability, but also leads to an increase in reflectance.

**Figure 17.** Measured reflectance of the post treated MGS from Experiment 2 A and 2 B, (**a,b**): Furnace treatment; (**c**): Turbula®Mixer, and (**d**): Stir bar treatment, an increase in reflection can be determined for all three post-treatments.

In conclusion, the reflection measurements show that a change in morphology and thus an increase in reflection can be achieved through post-treatments. Despite the post-treatments, however, the result from Experiment 1 could not be tied up. Therefore, further works will be required to find a post-treatment that adapts the surface morphology for the coated MGS from Setup 2.

#### **4. Summary**

In this paper, the Liquid Phase Deposition (LPD) method was examined as an additional coating method to the Physical Vapour Deposition (PVD) and the Chemical Vapor Deposition (CVD) method for metallization of micro glass spheres. Two different micro hollow glass sphere types (MHGS, iM16K-ZF and S38HS) and one micro solid glass sphere type (SMGS, Type S) were coated with aluminum by the hydridolyse of AlCl3. Measurements of X-Ray Diffraction (XRD), Energy Dispersive X-Ray Spectroscopy (EDX) and Scanning Electron Microscope (SEM) show, that both sphere types (MHGS and SMGS) can be coated with metallic aluminum using the Liquid Phase Deposition (LPD) method. In comparison to reference spheres, coated via the Physical Vapor Deposition (PVD) method, a similar surface can be achieved for the SMGS, whereas a more structured morphology occurred for the MHGS. Post-treatments with Turbula®Mixer, magnetic stir bar and furnace exhibit that the aluminum layer is stable and firmly bonded to the sphere surface. In addition, the post-treatments were able to break up agglomerates, as well as slightly smooth the surface morphology.

The influence of the coating and the post-treatments on the reflective properties in the MIR-range of the spheres was examined by additional reflection measurements in a wavelength range of 7.5 to 18 μm. It was found that in addition to a sufficient amount of aluminum, surface properties play an important role. With a rather rough surface morphology, only an increase in reflectance of approx. 5% was observed for SMGS. In contrast, a very homogenous smooth surface resulted in an increase in reflectance of approx. 30 % compared to the uncoated spheres, which corresponds to the reflectance of the PVD-coated spheres. However, this cannot be obtained with the MHGS. However, the post-treatments showed that an adaptation of the surface morphology is possible, leading to double the reflectance compared to the untreated coated spheres.

The paper showed that the LPD method, as a non-vacuum process, is suitable for an aluminum coating of spherical particles. It was even possible to achieve coating results comparable to the PVD process, which was shown by the similar reflection values of the LPD-coated spheres and PVD-coated spheres. Besides reflection, aluminum coated (hollow) glass spheres can also be interesting for further application, e.g., the modification of thermal conductivity or dielectric properties, where, unlike in our case, the surface condition will have less influence.

**Author Contributions:** L.S. performed the experiments and wrote the paper. L.Z. and B.S. helped with the experiments. S.L. und T.G. provided guidance and helped in manuscript preparation.

**Funding:** Financial support by the BMBF project EcoSphere (BMBF: 13N13188) is gratefully acknowledged.

**Acknowledgments:** The presented work is dedicated to Monika Willert-Porada, who passed away on 11 December 2016. She initiated the above-mentioned project. In addition, the authors would like to thank Angelika Kreis and Ingrid Otto for laboratory support. The authors also want to thank Andreas Eder from the University of Vienna for producing the PVD-coated spheres. This publication was funded by the German Research Foundation (DFG) and the University of Bayreuth in the funding program Open Access Publishing.

**Conflicts of Interest:** The authors declare no conflict 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* **Study on** β**-Ga2O3 Films Grown with Various VI**/**III Ratios by MOCVD**

**Zeming Li 1, Teng Jiao 1, Daqiang Hu 1, Yuanjie Lv 2, Wancheng Li 1, Xin Dong 1,\*, Yuantao Zhang 1, Zhihong Feng <sup>2</sup> and Baolin Zhang 1,\***


Received: 20 March 2019; Accepted: 23 April 2019; Published: 26 April 2019

**Abstract:** β-Ga2O3 films were grown on sapphire (0001) substrates with various O/Ga (VI/III) ratios by metal organic chemical vapor deposition. The effects of VI/III ratio on growth rate, structural, morphological, and Raman properties of the films were systematically studied. By varying the VI/III ratio, the crystalline quality obviously changed. By decreasing the VI/III ratio from 66.9 <sup>×</sup> 103 to 11.2 <sup>×</sup> <sup>10</sup>3, the crystalline quality improved gradually, which was attributed to low nuclei density in the initial stage. However, crystalline quality degraded with further decrease of the VI/III ratio, which was attributed to excessive nucleation rate.

**Keywords:** β-Ga2O3; MOCVD; VI/III ratio

#### **1. Introduction**

β-Ga2O3, the most stable phase of Ga2O3, shows great potential because of its excellent material properties. It is a wide bandgap (WBG) semiconductor with band gap of ~4.9 eV, breakdown field of 8 MV cm–1 and Baliga's figure of merit of 3444 at room temperature, which offers more advantages in high-efficiency power device application than SiC and GaN [1]. Moreover, its high transparency in UV wavelength range, and excellent thermal and chemical stability also have great application potential in flat panel displays, UV detectors, and high-temperature gas sensors [2–6]. There are several ways to produce a β-Ga2O3 film, which include molecule beam epilayer (MBE) [7], metal organic chemical vapor deposition (MOCVD) [8], halide vapor phase epitaxy (HVPE) [4], chemical vapor deposition (CVD) [9], magnetron sputtering [10], and thermal oxidation [11]. Conventional CVD methods [12–14], especially MOCVD have several advantages, including excellent reproducibility and capability for scale-up to high-volume production [15]. Impressive studies on the growth of β-Ga2O3 by MOCVD have been recently reported. Lv et al. investigated the epitaxial relationship between β-Ga2O3 and sapphire substrates [16]. Zhuo et al. studied the control of the crystal phase composition of the Ga2O3 thin film [17]. Sbrockeyet al. demonstrated the large-area growth of β-Ga2O3 films using rotating disc MOCVD reactor technology [15]. Alema et al. studied the growth rates of β-Ga2O3 epitaxial films by close coupled showerhead MOCVD [18]. Takiguchi et al. studied β-Ga2O3 epitaxial films obtained by low temperature MOCVD [19]. Chen et al. investigated the effect of growth pressure on the characteristics of β-Ga2O3 films grown on GaAs (100) substrates [20]. However, the crystalline quality of heteroepitaxial β-Ga2O3 films has not been able to meet the requirements of device fabrication so far.

In this paper, β-Ga2O3 films were grown by MOCVD on sapphire (0001) substrates with various VI/III ratios. In addition, the effects of VI/III ratio on growth rate, structural, morphological, and Raman properties were systematically studied. By varying the ratio, the crystalline quality of the films was effectively improved.

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

#### *2.1. Materials*

High purity O2 (purity, 5 N) and trimethylgallium (TMGa, 6 N in purity, Nata Opto-electronic Material Co., Nanjing, China) and were used as oxidant and organometallic source, respectively. High purity Ar (purity, 6 N) worked as a carrier gas.

#### *2.2. Preparation*

The β-Ga2O3 films were grown on sapphire (0001) substrates by MOCVD. The equipment was modified from an Emcore D180 MOCVD (Emcore, Alhambra, CA, USA). The close coupled showerhead method is used; the highest growth temperature of the MOCVD was 1150 ◦C. Before the growth process, the substrates were cleaned sequentially by acetone, ethanol, deionized water in an ultrasonic bath, and then dried with N2. The growth pressure and substrate temperature were kept at 20 mbar and 750 ◦C during the whole growth process, respectively. High purity O2 was injected into the reaction chamber with a fixed flow rate of 1200 sccm. TMGa was stored in a stainless steel bubbler, maintained at 1 ◦C. The pressure inside the bubbler were kept at 900 Torr. Ar carrier gas passed through the TMGa bubbler and delivered the TMGa vapor to the reactor. To obtain β-Ga2O3 films grown with various VI/III ratios, the flow rates of Ar carrier gas were varied from 5 sccm to 60 sccm (5 sccm, 15 sccm, 30 sccm, 45 sccm, 60 sccm). The growth time was 30 min.

#### *2.3. Characterization*

The structural properties of β-Ga2O3 films were investigated by X-ray diffractometer (XRD, Rigaku, Ultima IV, Tokyo, Japan, λ = 0.15406 nm, graphite filter). The morphological properties of the β-Ga2O3 films were studied by field emission scanning electron microscopy (FESEM, JSM-7610F, JEOL, Tokyo, Japan) and atomic force microscopy (AFM, Veeco, Plainview, NY, USA). Raman properties of the films was analyzed by a Raman spectrometer (HORIBA, LABRAM HR EVO, Kyoto, Japan) using a wavelength of λ = 633 nm laser. The thicknesses of the films were measured by a thin film analyzer (F40, Filmetrics, San Diego, CA, USA).

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

The molar flow rates in the experiments can be calculated by Equations (1)–(3) [21,22]:

$$\ln\left(P\_{\rm MO}\right) = a - b\gamma T\tag{1}$$

where *P*MO is the vapor pressure of TMGa, *a* = 8.07, *b* = 1703, *T* is the thermodynamic temperature of TMGa,

$$n\_{\rm MO} = F \times P\_{\rm MO} / [V\_{\rm m} \times (P\_{\rm bulb} - P\_{\rm MO})]\_{\prime} \tag{2}$$

where *n*MO is the molar flow rate of TMGa, *F* is the flow rate of carrier gas, *V*<sup>m</sup> = 22414 cm3/mol, *P*bub is the pressure inside the bubbler,

$$m\_{\rm O} = F\_{\rm O} / V\_{\rm m\nu} \tag{3}$$

where *n*<sup>O</sup> is the molar flow rate of O2, *F*<sup>O</sup> is the flow rate of O2. The VI/III ratios in the experiments are shown in Table 1.


**Table 1.** The VI/III ratios at various flow rates of Ar carrier gas.

#### *3.1. Growth Rate Analysis*

To investigate the growth rates, the thicknesses of the samples were measured by a thin film analyzer. The sample obtained with VI/III ratio of 5.6 <sup>×</sup> 10<sup>3</sup> is unsuitable for such analysis due to its excessively rough surface [18]. The growth rate showed a strong dependence on the VI/III ratio (Figure 1). Because the flow rate of oxygen was a constant, the growth rate was mainly limited by the flow rate of organometallic source. By increasing the flow rates of Ar carrier gas from 5 sccm to 45 sccm, the VI/III ratio decreased from 66.9 <sup>×</sup> <sup>10</sup><sup>3</sup> to 7.4 <sup>×</sup> <sup>10</sup>3, and the growth rate improved from 0.26 to 1.98 μm/h.

**Figure 1.** Growth rates of the samples obtained with various VI/III ratios.

#### *3.2. XRD Analysis*

Figure 2 shows the XRD θ–2θ scan patterns of β-Ga2O3 films grown with various VI/III ratios. For the film grown with VI/III ratio of 66.9 <sup>×</sup> 103, except the diffraction peaks of Al2O3 substrate, only three peaks located at 18.76◦, 38.10◦ and 58.84◦ could be observed, which related to β-Ga2O3 (-201), (-402), and (-603). It indicated that the thin film consisted of pure β-Ga2O3. By decreasing the VI/III ratio from 66.9 <sup>×</sup> 103 to 11.2 <sup>×</sup> 103, the three peaks of <sup>β</sup>-Ga2O3 were strengthened and sharpened. The crystallite sizes along the direction vertical to (-201) plane of the samples obtained with the VI/III ratios of 66.9 <sup>×</sup> 103, 22.3 <sup>×</sup> <sup>10</sup>3, and 11.2 <sup>×</sup> 103 were calculated to be 11.2, 12.2, and 17.5 nm, respectively (by Scherrer equation). Larger crystallite sizes indicated lower defect density and an improvement of crystalline quality. Lower VI/III ratio was helpful to reduce the nuclei density in the initial stage of deposition process and enlarge the size of islands in the subsequent stage, which indicated that less defects occurred in island coalescence [23,24]. However, further decreasing the VI/III ratio caused crystalline quality degradation. For the film grown with VI/III ratio of 7.4 <sup>×</sup> 103, the intensities of the three β-Ga2O3 peaks declined, and peaks related to β-Ga2O3 (401), (-601), (601), and (-801) were observed, indicating the polycrystalline structure of the film. The change in crystalline structure

is caused by excessive nucleation rate with this VI/III ratio. At this nucleation rate, the deposited particles were unable to migrate to the appropriate lattice positions, and the films grew and oriented in unsuitable directions, which caused random growth. As for the sample obtained with VI/III ratio of 5.6 <sup>×</sup> 103, the change in crystalline structure was obvious—15 peaks of <sup>β</sup>-Ga2O3 showed up. The crystallite sizes of the films grown with VI/III ratio of 7.4 <sup>×</sup> 103 and 5.6 <sup>×</sup> <sup>10</sup><sup>3</sup> were calculated to be 14.2 and 21.3 nm, respectively.

**Figure 2.** XRD <sup>θ</sup>–2<sup>θ</sup> scan patterns of <sup>β</sup>-Ga2O3 films grown with various VI/III ratios: (**a**) 66.9 <sup>×</sup> 103; (**b**) 22.3 <sup>×</sup> <sup>10</sup>3; (**c**) 11.2 <sup>×</sup> 103; (**d**) 7.4 <sup>×</sup> <sup>10</sup>3; (**e**) 5.6 <sup>×</sup> 103. <sup>Δ</sup> the peaks of the sapphire substrates.

#### *3.3. AFM Analysis*

To investigate the effects of VI/III ratios on the surface morphology of β-Ga2O3 films, AFM was carried out; the images are shown in Figure 3. The surface roughness of the films depended highly on the VI/III ratios. For the film grown with the VI/III ratios from 66.9 <sup>×</sup> 103 to 11.2 <sup>×</sup> <sup>10</sup>3, root-mean-square (RMS) surface roughness increased from 3.71 to 7.83 nm. The hillocks on the surfaces enlarged and decreased in density, in good agreement with the XRD analysis. By decreasing the VI/III ratio to 7.4 <sup>×</sup> <sup>10</sup>3, the surface roughness had little change, while the morphology changed greatly. Many wheat-like structures were observed, which means that excessive nucleation rate hindered particle migration and caused random growth. For the film grown with VI/III ratio of 5.6 <sup>×</sup> <sup>10</sup>3, the roughness increased greatly, even reaching 56.3 nm (seven times higher than that of any other film), in accordance with its XRD pattern (Figure 2).

**Figure 3.** AFM images (5 <sup>μ</sup><sup>m</sup> <sup>×</sup> <sup>5</sup> <sup>μ</sup>m) of <sup>β</sup>-Ga2O3 films grown with various VI/III ratios: (**a**) 66.9 <sup>×</sup> <sup>10</sup>3; (**b**) 22.3 <sup>×</sup> 103; (**c**) 11.2 <sup>×</sup> <sup>10</sup>3; (**d**) 7.4 <sup>×</sup> <sup>10</sup>3; (**e**) 5.6 <sup>×</sup> <sup>10</sup>3.

#### *3.4. FESEM Analysis*

According to XRD analysis of all the films, the sample obtained with VI/III ratio of 11.2 <sup>×</sup> 103 was measured by FESEM. Figure 4 shows the top and cross-sectional views of FESEM images of the sample. The surface with minor defects is in accordance with the AFM image in Figure 3. The relatively smooth cross-sectional image indicates high film quality. In addition, the thickness measured by the cross-sectional view images is about 0.68 μm.

**Figure 4.** FESEM images of <sup>β</sup>-Ga2O3 films grown with VI/III ratio of 11.2 <sup>×</sup> 103: (**a**) Top-view; (**b**) cross-sectional view.

#### *3.5. Raman Analysis*

Figure 5 presents the Raman spectra of β-Ga2O3 films grown with various VI/III ratios. For comparison, the Raman spectra of the sapphire substrates is also shown in this figure. Except for the peaks related to the substrates, only one Raman peak related to β-Ga2O3 was observed. For the film grown with VI/III ratio of 66.9 <sup>×</sup> 103, due to poor crystalline quality and a smooth surface, only one peak related to β-Ga2O3 was clearly observed. By decreasing the VI/III ratio, due to the change in crystalline quality and roughness, more peaks related to β-Ga2O3 showed up, which were gradually enhanced. However, when the VI/III ratio was decreased to 5.6 <sup>×</sup> 103, owing to the excessively rough surface of the obtained sample, its surface area increased and its Raman spectrum changed greatly. Ten peaks related to β-Ga2O3 showed up and the peaks were enhanced greatly. The 10 peaks were divided into three categories [25–27]—the peaks located at 115, 147, 171, and 201 cm−<sup>1</sup> were attributed to libration and translation of tetrahedral-octahedra chains; the peaks located at 322, 349, and 476 cm−<sup>1</sup> were attributed to deformation of GaO6 octahedra, and the peaks located at 631, 655, and 768 cm−<sup>1</sup> were attributed to stretching and bending of GaO4 tetrahedra. The Raman results confirmed that all the obtained films consisted of pure β-Ga2O3.

**Figure 5.** Raman spectra of β-Ga2O3 films grown with various VI/III ratios: (**a**) sapphire substrate; (**b**) 66.9 <sup>×</sup> 103; (**c**) 22.3 <sup>×</sup> 103; (**d**) 11.2 <sup>×</sup> 103; (**e**) 7.4 <sup>×</sup> 103; (**f**) 5.6 <sup>×</sup> 103. <sup>∗</sup>, the Raman peaks related to sapphire. Δ, the Raman peaks related to tetrahedral-octahedra chains. #, the Raman peaks related to GaO6 octahedra. -, the Raman peaks related to GaO4 tetrahedra.

#### **4. Conclusions**

In summary, β-Ga2O3 films were grown on sapphire (0001) substrates with various VI/III ratios by MOCVD. By varying the VI/III ratio, the crystalline quality obviously changed. For the film grown with VI/III ratios from 66.9 <sup>×</sup> 103 to 11.2 <sup>×</sup> 103, the crystalline quality improved gradually, attributed to low nuclei density in the initial stage. However, further decreasing the VI/III ratio caused degradation of crystalline quality, and the morphological and Raman properties changed greatly, which was attributed to excessive nucleation rate. This work offers a feasible way to improve the crystalline quality of heteroepitaxial β-Ga2O3 films and is beneficial for device fabrication.

**Author Contributions:** Conceptualization, Z.L.; Methodology, W.L.; Validation, D.H.; Formal Analysis, D.H.; Investigation, T.J.; resources, B.Z.; Data Curation, Y.Z.; Writing—Original Draft Preparation, Z.L.; Writing—Review and Editing, X.D.; Visualization, Y.L.; Supervision, Z.F.; Project Administration, X.D.; Funding Acquisition, X.D.

**Funding:** This research was funded by the National Natural Science Foundation of China, Grant Numbers 61774072, 61376046, 61674068 and 61404070; the Science and Technology Developing Project of Jilin Province, Grant Number 20170204045GX, 20150519004JH, 20160101309JC; the National Key Research and Development Program, Grant Number 2016YFB0401801; and the Program for New Century Excellent Talents in University, Grant Number NCET-13-0254.

**Conflicts of Interest:** The authors declare no conflict 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* **Structure and Conductivity Studies of Scandia and Alumina Doped Zirconia Thin Films**

**Mantas Sriubas 1, Nursultan Kainbayev 1,2, Darius Virbukas 1, Kristina Boˇckute˙ 1,\*, Živile Rutk ˙ unien ¯ e˙ <sup>1</sup> and Giedrius Laukaitis <sup>1</sup>**


Received: 22 March 2019; Accepted: 8 May 2019; Published: 12 May 2019

**Abstract:** In this work, scandia-doped zirconia (ScSZ) and scandia–alumina co-doped zirconia (ScSZAl) thin films were prepared by electron beam vapor deposition. X-ray diffraction (XRD) results indicated a presence of ZrO2 cubic phase structure, yet Raman analysis revealed the existence of secondary tetragonal and rhombohedral phases. Thus, XRD measurements were supported by Raman spectroscopy in order to comprehensively analyze the structure of formed ScSZ and ScSZAl thin films. It was also found that Al dopant slows down the formation of the cubic phase. The impedance measurements affirmed the correlation of the amount of secondary phases with the conductivity results and nonlinear crystallite size dependence.

**Keywords:** scandium stabilized zirconia thin films; e-beam physical vapor deposition; thin films ceramics; Raman spectroscopy; X-ray diffraction

#### **1. Introduction**

Zirconium oxide-based materials have been extensively studied for a few decades as electrolytes and fulfill almost all requirements: high oxide ion conductivity (0.1–0.01 S/cm), low electronic conductivity, chemical stability, mechanical strength, and low cost [1].

Pure ZrO2, due to its low number of vacancies, has a low specific ionic conductivity and has a monoclinic structure, and up to 1170 ◦C does not exhibit high ionic conductivity. The highly-conductive cubic phase forms only at 2370 ◦C [2]. A tetragonal phase exists in the middle-temperature range (1170–2370 ◦C). The monoclinic–tetragonal (m → t) phase transition is produced by a condensation of two phonons at point M of the Brillouin zone of the tetragonal phase (P42/nmc) and the tetragonal–cubic (t → c) phase transition is produced by a condensation of phonons at point X of the Brillouin zone of the cubic phase (Fm3m) [3,4].

To improve ionic conductivity and stabilize the cubic lattice at low temperatures, zirconia is doped with trivalent or bivalent elements, such as Ca, Mg, Y, and Sc, etc. [5,6]. The stabilization process can be explained by crystal chemistry, i.e., oxygen vacancies associated with Zr can provide stability for cubic zirconia. Vacancies introduced by the oversized dopants are located as the nearest neighbors to the Zr atoms, leaving the eightfold coordination to dopant cations. Undersized dopants compete with Zr ions for oxygen vacancies in zirconia, resulting in six-fold oxygen coordination and a large disturbance to the surrounding next nearest neighbors [7–9].

Scandia-doped zirconia (ScSZ) is considered the most suitable material at intermediate temperatures because it has much higher ionic conductivity than yttria-doped zirconia [10]. However, four different phases can be observed below 600 ◦C for ScSZ systems depending on the concentration

of Sc2O3, i.e., cubic-tetragonal (5–7 mol %), cubic (8–9 mol %), and rhombohedral (10–15 mol %) [7,11]. Transitions between the highly conductive cubic phase and tetragonal or rhombohedral phases cause an abrupt decrease in ionic conductivity and suspend the application of ScSZ for intermediate temperature solid oxide fuel cells (IT-SOFC) [7,12]. Co-doping can be a possible solution. Adding dopants like CeO2, Bi2O3, Al2O3, and etc., has already been attempted [13–15]. Results vary for different dopants. Mostly, the cubic phase has been mixed with other phases (rhombohedral, monoclinic, or tetragonal) except for co-doping 0.5 mol % of Al2O3. It has been demonstrated that 0.5 mol % of Al2O3 stabilizes the cubic phase of ScSZ to room temperature [13]. The mechanism of cubic phase stabilization is not well known. It has been suggested that the strain in the crystalline, which is induced by adding a secondary dopant with a different radius to the Sc3+, is involved [13,16]. However, this research was based only on the X-ray diffraction (XRD) method, which cannot fully prove the existence of a fully stabilized cubic structure. In a mixture of phases, XRD peaks of the main cubic phase and secondary phases (tetragonal and rhombohedral phases) can overlap and the presence of secondary phases is not reflected ine XRD analysis [17], indicating that XRD analysis should be supported by an additional characterization method, e.g., Raman spectroscopy. Raman spectroscopy is a powerful analysis tool used to study the fundamental vibrational characteristics of molecules and is very sensitive to phase changes in the material. It can provide information about the phase, the chemical composition of the material, and oxygen vacancies [18,19]. Light scattering occurs and a small percentage of the scattered light may be shifted in frequency when monochromatic light is incident on the material. This frequency shift of the Raman scattered light is directly related to the structural properties of the material. The change in the phonon frequency of the vibrational mode will be produced in the presence of discontinuation of translational symmetry in the crystalline material due to the doping and secondary phases. It is known that the wavenumber of the vibrational modes follows a linear relationship with chemical composition as well as with strain induced in the crystalline lattice [20]. In the case of phase transformation of ZrO2, Raman spectroscopy can determine the change of the bond length and the angle between cation and anion [21]. Based on the reports of other authors, monoclinic [12,18,22], tetragonal [15,21–24], rhombohedral [12,24], and cubic [12,15,18,24] phases can be identified from Raman spectra for ZrO2. Moreover, the great majority of investigations analyzing co-doping of ScSZ have been focused on the powders and pellets. The properties and phase transitions of co-doped ScSZ thin films have not been studied enough. Thin films can be formed using a variety of deposition techniques, such as screen printing [25], wet powder spraying [26], and electrostatic spray deposition [27], etc. Electron beam vapor deposition allows producing a dense and homogenous thin film with a strictly controllable thin film growing process during the deposition [28]. Furthermore, it is a particularly appropriate formation method for ceramic thin films, which are distinguished by their high melting temperature. In e-beam evaporation, Sc-doped and Sc and Al co-doped ZrO2 evaporate by partial dissociation [29–31], causing the different structure and crystallographic phases of the formed thin films than powders. In this paper, the structure and ionic conductivity of thin ScSZ and scandium alumina stabilized zirconia films (ScAlSZ) are analyzed using XRD, energy dispersive X-ray spectrsoscopy (EDS), Raman, and electrochemical impedance spectroscopy (EIS) methods.

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

ScSZ and ScAlSZ thin films were formed using an e-beam physical vapor deposition system (Kurt J. Lesker EB-PVD 75, Hastings, UK). The formation was carried out on crystalline Alloy 600 (Fe-Ni-Cr) and polycrystalline Al2O3 substrates using deposition rates from 0.2 to 1.6 nm/s in steps of 0.2 nm/s. The thickness (~1500 nm) and deposition rate were controlled with an INFICON (Inficon, Bad Ragaz, Switzerland) crystal sensor. The temperature of substrates was changed from room temperature (20 ◦C) to 600 ◦C temperature. In order to get a homogenous thin film, substrates were rotated at 8 rpm speed. The acceleration voltage of the electron gun was kept at a constant of 7.9 kV and the required deposition rate was achieved by adjusting the e-beam current in the range of 60–100 mA. An initial pressure of ~2.0 <sup>×</sup> <sup>10</sup>−<sup>4</sup> Pa and working pressure of ~2.0 <sup>×</sup> <sup>10</sup>−<sup>2</sup> Pa was used in the vacuum

chamber during the experiments. The substrates were cleaned in an ultrasonic bath and treated in Ar<sup>+</sup> ion plasma for 10 min before deposition. The powders of (Sc2O3)0.10(ZrO2)0.90 (ScSZ) and (Sc2O3)0.10(Al2O3)0.01(ZrO2)0.89 (ScSZAl) (Nexceris, LLC, Fuelcellmaterials, Lewis Center, OH, USA) were pressed into pellets and used as evaporating material. Pellets were evaporated using a single e-beam configuration. Elemental analysis was performed using energy-dispersive X-ray spectroscopy (EDS, BrukerXFlash QUAD 5040, Bruker AXS GmbH, Karlsruhe, Germany) and the atomic ratios of Sc, Al, and Zr are presented in Table 1.


**Table 1.** Atomic ratios of Sc, Al, and Zr in the initial powders.

The crystallographic nature of the formed thin films was determined using XRD (D8 Discover (Bruker AXS GmbH, Billerica, MA, USA) standard Bragg focusing geometry with Cu Kα<sup>1</sup> (λ = 0.154059 nm) radiation, 0.01◦ step, and Lynx eye PSD detector). Rietveld analysis was performed to calculate ratios of cubic and rhombohedral phases in the ScSZAl thin films. The refinement was conducted with isotropic atomic thermal parameters [32]. The size of the crystallites and lattice constants were estimated with the software Topas 4-1. The Pawley method was used to fit XRD patterns and the crystallite size was estimated using the Scherrer equation [33].

Raman scattering measurements were performed using the Raman microscope inVia (Renishaw, Gloucestershire, UK). The excitation beam from a diode laser of 532 nm wavelength was focused on the sample using a 50× objective (NA = 0.75, Leica, Wetzlar, Germany) and the diameter of the laser spot was 4 μm. Laser power at the sample surface varied from 1.75 to 3.5 mW. The integration time was 15–30 s and the signal was accumulated five times and then averaged. The Raman Stokes signal was dispersed with a diffraction grating (2400 grooves/mm) and data was recorded using a Peltier cooled charge-coupled device (CCD) detector (Renishaw, Gloucestershire, UK) (1024 × 256 pixels). This system yields a spectral resolution of about 1 cm<sup>−</sup>1. Silicon was used to calibrate the Raman setup for both the Raman wavenumber and spectral intensity. Positions of Raman peaks were determined by fitting the data to the Lorentz line shape using a peak fit option in OriginPro 2016 software. The phase ratio was calculated using the formula [18]

$$\mathfrak{G}\_{\mathbb{C}} = \frac{I\_{\mathbb{C}}}{I\_{\mathbb{C}} + I\_{\mathbb{H}} + I\_{\mathbb{B}}} \tag{1}$$

where *I*c, *I*t, and *I*<sup>β</sup> are the scattering intensity of cubic (~620 cm−1), tetragonal (~473 cm−1), and rhombohedral (~557 cm<sup>−</sup>1) Raman modes, respectively.

Electrical characterization and impedance spectroscopy measurements were performed using a Probostat® (NorECs AS, Oslo, Norway) measurement cell in the frequency range 1–10<sup>6</sup> Hz and a 200–1000 ◦C temperature interval. Electrodes of the geometry 3 mm × 10 mm (*L* × *B*) were made of Pt ink and applied on top of the thin films using a mask reproducing the geometry of the electrodes (the two-probe method). The total conductivity was calculated according to:

$$
\sigma = \frac{L\_{\text{e}}}{R\_{\text{s}}A} = \frac{L\_{\text{e}}}{R\_{\text{s}}h l\_{\text{e}}} \tag{2}
$$

where *L*<sup>e</sup> is the distance between the Pt electrodes, *R*<sup>s</sup> is the resistance obtained from the impedance spectra, *A* is the cross-sectional area, *h* is the thickness of the thin films, and *l*<sup>e</sup> is the length of the electrodes.

#### **3. Results**

XRD measurements revealed that ScSZ and ScSZAl powders have a mixture of ZrO2 rhombohedral (JCPSD No. 01-089-5482) and monoclinic (JCPSD No. 01-089-5474) phases (Figure 1a). It can be seen that the prevailing phase is rhombohedral, showing sharp (003), (101), (012), (104), (110), (015), (11-3), (021), (006), and (202) peaks. Raman spectra of investigated powders are in good agreement with the XRD results (Figure 1b). The peaks detected at 147, 148, 176, 190, 191, 259, 260, 315, 321, 384, 386, 422, 471, 473, 507, 517, 552, 554, 585, 597, and 640 cm<sup>−</sup>1, and around 800–1100 cm−<sup>1</sup> showing a polymorphism of ZrO2, can be assigned to monoclinic, tetragonal, and rhombohedral phases. The peaks at 507–517, 552–554, 585, and 597 cm−<sup>1</sup> belong to the β-rhombohedral phase [12], the peaks at 147–148, 259–260, 315–321, 384–386, 422, 547, 551, and 640 cm−<sup>1</sup> are indicative of the tetragonal phase and the peaks at 176, 190, 191, and 469–473 cm−<sup>1</sup> belong to the monoclinic phase [18,22]. Peaks observed above 800 cm−<sup>1</sup> correspond to the second-order active Raman modes wave numbers combination [34]. The Raman spectra of cubic zirconia should consist of a weak broad line peak assigned to a single Raman mode F2g symmetry centered between 605 and 630 cm<sup>−</sup>1. Peaks corresponding to the cubic phase of ZrO2 were not observed in Raman spectra of the powders [12,15,18,21,24]. The obtained results agree with the results of other authors [24,35]. The material should have a rhombohedral structure if the Sc2O3 concentration is above 9% [5,11]. Raman peaks corresponding to the tetragonal phase (Figure 1b) indicate that the tetragonal structure is deformed and is predominant mostly at the grain boundary [15].

**Figure 1.** ScSZ and ScSZAl powders: (**a**) X-ray diffraction (XRD) patterns and (**b**) Raman shifts.

It is known that ZrO2 evaporates by partial decomposition during e-beam evaporation process [29,31]:

$$\text{ZrOO}\_{2(s)} = \text{ZrOO}\_{2(g)}\tag{3}$$

$$\text{ZrOO}\_{2(s)} = \text{ZrOO}\_{(g)} + \text{O}\_{(g)} \tag{4}$$

$$\text{ZrO}\_{2(s)} = \text{Zr}\_{(g)} + \text{O}\_{2(g)}\tag{5}$$

Therefore, the vapor phase of ScSZ and ScSZAl powders could consist of ZrO2, ZrO, O2, O, Sc2O3, Sc2O2, ScO, Sc, Al2O3, Al2O2, AlO, Al atoms, molecules, and molecule fragments. The atoms and molecule fragments landed on the surface of the substrate migrate and form grains. The grains with the densest planes are selected, e.g., (111) for face-centered cubic. However, surface diffusivity plays an important role in the formation of textured thin films. Adatoms deposited near grain boundaries have a higher probability of becoming incorporated at a low diffusivity surface in comparison to adatoms at high diffusivity planes having longer mean free paths with correspondingly higher probabilities of moving off the plane and becoming trapped on the adjacent grains [36]. Therefore, grains with low surface diffusivities grow faster. A similar growth mechanism was observed in the ScSZ and ScSZAl thin films. The thin films have a cubic structure (Figure 2a) and the positions of (111), (200), (220), (311), and (222) orientation peaks in the XRD patterns (measured at room temperature) confirm this (JCPSD No. 01-089-5483). The preferential orientation is (200) at low temperature (up to 450 ◦C) and preferential orientation is (111) at high temperatures because adatoms have high enough diffusion energy to move on the surface and become trapped on the high diffusivity surfaces. Moreover, the positions of the peaks of the ScSZAl thin films are shifted to higher angles by 0.2◦ in comparison to the positions of the peaks of the ScSZ thin films (Figure 2b) due to distortions in the crystal lattice. The ion of Al (~0.53 Å) is smaller than the ion of zirconium (0.84 Å).

**Figure 2.** XRD patterns (measured at room temperature) of (**a**) ScSZAl thin films and (**b**) ScSZ and ScSZAl thin films deposited on Alloy 600 substrates using a 0.4 nm/s deposition rate.

However, Rietveld analysis showed that a mixture of cubic (87.8%) and rhombohedral (12.2%) phases can be found in the ScSZAl thin films (Figure 3) with equal probability as a pure cubic phase. The weighted profile R-factors (Rwp) are almost the same for pure cubic (1.45) and the mixture of cubic and rhombohedral (1.35). Therefore, the peaks of cubic and rhombohedral phases can be overlapped [17,35]. In this case, XRD patterns require a complimentary analysis and Raman spectroscopy can be a quick solution [15,21].

**Figure 3.** Rietveld analysis of XRD patterns of ScSZAl thin films deposited on Alloy 600 substrates using a 0.4 nm/s deposition rate.

The Raman spectra of ScSZ and ScSZAl thin films indicate mixed phases (Figure 4). Raman peaks are detected around 140, 262, 354, 382, 475, 540, 618, 726, 954, and 1000 cm−1. The broad peak observed between 100 and 200 cm−<sup>1</sup> consists of several peaks, indicating different ZrO2 phases. The peaks at 354 and 382 cm−<sup>1</sup> belong to the monoclinic phase [12,18,22], peaks detected at 147, 260, and 475 cm−<sup>1</sup> belong to the tetragonal phase [15,21–24], and peaks detected around 540 cm−<sup>1</sup> belong to the rhombohedral phase [12,24]. The broad peak expressed around 620 cm−<sup>1</sup> consists of several peaks indicating the presence of cubic (c), tetragonal (t), and monoclinic (m) phases. On the other hand, Raman peaks near 620 cm−<sup>1</sup> can be shifted due to a disordered oxygen sublattice after doping and co-doping by Sc and Al. Substitution of Zr4<sup>+</sup> by Sc3<sup>+</sup> results in the formation of high quantities of oxygen vacancies [37] and such a high defect concentration can lead to a violation of the selection rules and the appearance of additional modes that are forbidden for the cubic fluorite structure [38].

**Figure 4.** Raman spectra of (**a**) ScSZ and (**b**) ScSZAl thin films deposited on Alloy 600 substrates using a 0.4 nm/s deposition rate.

The Raman spectra of ScSZ and ScSZAl are of a similar shape; however, the intensities of peaks at 620 cm−<sup>1</sup> are lower for the ScSZAl thin films (Figure 4). In the case of thin films with an Al dopant (Figure 4b), the cubic phase begins to form at a temperature higher than 50 ◦C and further increase of the temperature does not significantly influence the amount of the cubic phase. This means that Al slows down the cubic phase formation and stabilizes it at higher than 300 ◦C substrate temperatures. Quantitative calculations give the same substantiation (Table 2). The ratios of the cubic phase to the tetragonal and rhombohedral phases are from 42% to 53% for ScSZ and around 42% for ScSZAl. Calculations reveal that the amount of cubic phase increases by 10% for ScSZ and by 2% for ScSZAl with increasing substrate temperature. Finally, Raman analysis shows that thin films have a mixture of cubic (~44%), tetragonal (~18%), and rhombohedral (~38%) phases (Table 2). The obtained results demonstrat that a polymorphous transition from rhombohedral to cubic phase occurred in the formed thin films, indicating a typical behavior for the doped ZrO2 [39]. According to the XRD spectra, the ZrO2 rhombohedral phase is not significantly expressed and the tetragonal phase is not observed, while Raman spectra show large amounts of these phases, meaning that both phases could be located in the grain boundaries [15,39].


**Table 2.** The ratio of cubic, tetragonal, and rhombohedral phases in the ScSZ and ScSZAl thin films.

The average crystallite size dependence on the substrate temperature shows a nonlinear behavior (Figure 5). Crystallites grow larger (17.4–69.9 nm) with increasing substrate temperature (50–300 ◦C) during deposition. At 450 ◦C temperature, a sudden decrease of crystallite size occurs (~45–30 nm). This can be related to the changes in preferential orientation and to the changes in the ratio of phases. The preferential orientation changes from (200) at the temperatures up to 450 ◦C to (111) at high temperatures due to the higher diffusion energy of adatoms which allows them to move on the surface and become trapped on the high diffusivity surfaces.

**Figure 5.** Crystallite size dependence on substrate temperature of ScSZAl thin films deposited on Alloy 600 substrates using different deposition rates.

Arrhenius plots show linear dependences (Figure 6a). No obvious breaking or bending points were observed. This means that no phase transitions occurred during the measurements. Moreover, ionic conductivity (600 ◦C measurement temperature) is observed to be related to the substrate temperature (Table 3). Ionic conductivity is higher for the thin films deposited on higher temperature substrates. The highest value of ionic conductivity of 4.2 <sup>×</sup> <sup>10</sup>−<sup>3</sup> <sup>S</sup>/cm (substrate temperature 600 ◦C and deposition rate 0.4 nm/s) is similar to other authors' results (~7 <sup>×</sup> <sup>10</sup>−<sup>3</sup> <sup>S</sup>/cm) [13,16,24]. Crystallite size has a similar dependence on temperature. Crystallites were observed to grow larger (17.4–69.9 nm) with increasing substrate temperature (50–300 ◦C) during deposition. At 450 ◦C, a sudden decrease of crystallite size occurs (~45–30 nm). This may be related to the changes in preferential orientation and to the changes in the ratio of phases. It is known that ionic conductivity is strongly related to crystallite size. According to the brick layer model, materials consisting of larger crystallites exhibit higher ionic conductivity because grain boundaries slow down oxygen ion diffusion [40–42].

**Figure 6.** (**a**) Arrhenius plots of ScSZAl thin films deposited on Al2O3 substrates using a 0.4 nm/s deposition rate and (**b**) vacancy activation energy dependence on the substrate temperature of ScSZAl thin films deposited on Al2O3 substrates using a 0.4 nm/s deposition rate.


**Table 3.** Ionic conductivity (S/cm) of ScSZ and ScSZAl thin films deposited using different substrate temperature and a 0.4 nm/s deposition rate.

Co-doping of aluminum was observed to have a minor effect on the ionic conductivity of thin films, although conductivity is slightly lower for the ScSZAl thin films deposited on higher temperature substrates. The lower ionic conductivity is a result of a lower amount of cubic phase and a higher amount of tetragonal phase (Table 2).

It was noticed that vacancy activation energy increased from 0.91 to 1.22 eV using higher temperature substrates (Figure 6b). This increase also occurred due to the higher amount of tetragonal phase in those thin films (Table 3).

#### **4. Conclusions**

In this work, thin films of ScSZ and ScSZAl were deposited using electron beam vapor deposition, which allows the production of a dense and homogenous thin film. Structure and conductivity studies of the formed thin films were performed. It was found that the structure of the formed thin films does not repeat the structure of the initial evaporated material. Analysis of XRD patterns and Raman spectra of the initial evaporated powders of ScSZ and ScSZAl shows that they exhibit a polymorphism of ZrO2 monoclinic, tetragonal (in Raman spectroscopy), and rhombohedral phases. Contrarily to the structure of evaporating material, XRD of ScSZ and ScSZAl thin films depict only a pure ZrO2 face-centered cubic phase of a preferential orientation (200) at temperatures up to 450 ◦C, with a change in preferential orientation (100) at higher temperatures. In order to investigate the influence of an Al dopant, Rietveld analysis was performed, demonstrating that a pure cubic phase can be found in the ScSZAl thin films with equal probability as a mixture of cubic (87.8%) and rhombohedral (12.2%) phases. Raman spectra of ScSZ and ScSZAl thin films also indicate a polymorphism of ZrO2 phases. Raman peaks detected around 140, 262, 354, 382, 475, 540, 618, 726, 954, and 1000 cm−<sup>1</sup> indicate the presence of cubic (~44%), tetragonal (~18%) and rhombohedral (~38%) phases, showing a transition from a rhombohedral to a cubic phase. In ScSZAl thin films, co-doping with Al delays and slows down the formation of a cubic phase and stabilizes it at higher than 300 ◦C substrate temperatures. The ratios of the cubic phase to the tetragonal and rhombohedral phases are from 42% to 53% for ScSZ and around 42% for ScSZAl. It was found that the crystallite size depends on substrate temperature, demonstrating a nonlinear behavior. Crystallites grew larger (17.4–69.9 nm) with increasing substrate temperature (50–300 ◦C) and a sudden decrease of their size occurred (~45–30 nm) at a 450 ◦C deposition temperature due to the changes in preferential orientation and phase ratio. On the other hand, Arrhenius plots showed linear dependences and exhibited the highest value of ionic conductivity of 4.2 <sup>×</sup> 10−<sup>3</sup> S/cm for thin films deposited on 600 ◦C temperature substrates using a 0.4 nm/s deposition rate. The increase in the amount of tetragonal phase in the formed thin films influences the vacancy activation energy which increases from 0.91 to 1.22 eV using higher temperature substrates.

**Author Contributions:** Conceptualization, M.S., G.L, and N.K.; methodology, K.B. and Ž.R.; formal analysis, D.V., M.S., and K.B.; investigation, D.V., N.K., M.S., and Ž.R.; writing—original draft preparation, M.S., N.K., K.B., Ž.R., D.V., and G.L.; writing—review and editing, M.S., N.K., K.B., Ž.R., D.V., and G.L.; visualization, N.K.; supervision, G.L.; project administration, K.B.; funding acquisition, G.L.

**Funding:** This research was funded by the European Regional Development Fund according to the supported activity "Research Projects Implemented by World-class Researcher Groups" under Measure No. 01.2.2-LMT-K-718. **Acknowledgments:** Authors would like to express their gratitude for the following individuals for their expertise and contribution to the manuscript: Arvaidas Galdikas, Teresa Moskalioviene, Gediminas Kairaitis, and Matas Galdikas.

**Conflicts of Interest:** The authors declare no conflict 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* **Fabrication of a Conjugated Fluoropolymer Film Using One-Step iCVD Process and its Mechanical Durability**

#### **Hyo Seong Lee 1, Hayeong Kim 2, Jeong Heon Lee <sup>1</sup> and Jae B. Kwak 1,\***


Received: 12 June 2019; Accepted: 4 July 2019; Published: 8 July 2019

**Abstract:** Most superhydrophobic surface fabrication techniques involve precise manufacturing process. We suggest initiated chemical vapor deposition (iCVD) as a novel CVD method to fabricate sufficiently durable superhydrophobic coating layers. The proposed method proceeds with the coating process at mild temperature (40 ◦C) with no need of pretreatment of the substrate surface; the pressure and temperature are optimized as process parameters. To obtain a durable superhydrophobic film, two polymeric layers are conjugated in a sequential deposition process. Specifically, 1,3,5,7-tetravinyl-1,3,5,7-tetramethylcyclotetrasiloxane (V4D4) monomer is introduced to form an organosilicon layer (pV4D4) followed by fluoropolymer formation by introducing 1H, 1H, 2H, 2H-Perfluorodecyl methacrylate (PFDMA). There is a high probability of covalent bond formation at the interface between the two layers. Accordingly, the mechanical durability of the conjugated fluoropolymer film (pV4D4-PFDMA) is reinforced because of cross-linking. The superhydrophobic coating on soft substrates, such as tissue paper and cotton fabric, was successfully demonstrated, and its durability was assessed against the mechanical stress such as tensile loading and abrasion. The results from both tests confirm the improvement of mechanical durability of the obtained film.

**Keywords:** initiated chemical vapor deposition (iCVD); superhydrophobic; fluoropolymer

#### **1. Introduction**

Superhydrophobic surfaces attracts considerable research attention as they are used in many industrial applications, including water repellents, antifouling, and self-cleaning surfaces [1–12]. There are many methods and techniques to produce superhydrophobic surfaces such as metal etching [2,3], sol-gel processing [4], dip-spraying [5], electro-spray and spinning [6,11], combining solution and phase inversion process [7,8], solution method [9], ultrasound-assisted method [10], and plasma-based method [12]. However, most of these techniques are only suitable for nanostructures with fine surface roughness and are further limited to specific substrates such as metals. A hierarchical structure is always better for a superhydrophobic surface because it can reduce the contact area between the water and the surface. This surface can be obtained using a combination of the topographical properties of the surface texture and the chemical properties of low surface energy. To demonstrate this, Zhou et al. have synthesized nanostructured ZnO film with super-hydrophobicity using a chemical vapor deposition (CVD) method [13].

In comparison with other methods, the CVD process can produce large area of homogeneous thin films. However, it is not readily available for flexible substrates such as cotton fabrics and papers because it requires the use of high temperature and high vacuum pressure. For commercial purposes, a superhydrophobic surface needs to be mechanically and chemically durable. For instance, Zhou et al. fabricated a durable superhydrophobic polyester fabric with fluoroalkyl silane-modified silicone rubber, which is a nanoparticle composite, using the dip coating method [14]. Yan et al. also fabricated hydrophobic and hydrophilic surface of cotton fabric using a plasma-induced graft method [15]. Recently, Heydari Gharahcheshmeh and Gleason reported the fabrication of an antifouling surface on conductive polymer film via the oxidative chemical vapor deposition (oCVD) process that improves signal-to-noise ratio of the neural recording electrodes against blood or tissue contamination [16]. In this study, we use an iCVD process to generate fluoropolymer film for super hydrophobicity because this is relatively simple process in comparison with the previous methods. Unlike the conventional CVD process, the iCVD is achievable at a low temperature and vacuum pressure and shows excellent step coverage at high aspect ratio structure [17].

Figure 1 shows the iCVD process. The process begins with canisters containing an initiator (I2) and one or more monomers (M), which are the building blocks of the desired polymer coating. These materials are vaporized, either by heating or reducing the air pressure, and are introduced simultaneously into a vacuum chamber with the substrate to be coated. Once vaporized, the initiator molecules are thermally decomposed upon the contact with a hot filament to become radicals (I\*). The radicals activate the vaporized monomer to link in chains that form polymers on the surface of the substrate kept at mild temperature (25–40 ◦C). This is a one-step, solvent-free process to grow polymer films uniformly onto complex substrates structures, regardless of the substrate material. The functional performance of the polymer thin film is attributed to the properties of the monomers used [18].

**Figure 1.** A schematic of iCVD process.

To obtain the fluoropolymer film with the iCVD process, tert-butyl peroxide (TBPO) and 1H, 1H, 2H, 2H-perfluorodecyl acrylate (PFDA) are typically used as the initiator and the monomer, respectively [19–22]. In particular, 1H, 1H, 2H, 2H-perfluorodecyl methacrylate (PFDMA) is of great interest because of its exceptionally low surface energy (3.5 mN·m<sup>−</sup>1), similar to PFDA and more efficient commercialization possibilities [23,24]. These monomers are synthesized in a one-step iCVD process and show more than 150◦ of water contact angle. Previously, mechanical and chemical robustness

of the iCVD superhydrophobic coating had not been studied. Therefore, 1,3,5,7-tetravinyl-1,3,5,7 tetramethylcyclotetrasiloxane (V4D4) was introduced as a cross linker, which improves mechanical strength. Finally, a conjugated polymer film was obtained with a successive deposition of an organosilicon polymer, pV4D4 (poly-tetramethylcyclotetrasiloxane) and a fluoropolymer, pPFDMA (poly-perfluorodecyl methacrylate) using the customized iCVD reactor. Because pPFDMA is strongly bound on the pV4D4 layer, the durability of the conjugated film (pV4D4-PFDMA) improves due to increasing adhesion strength between the substrate and the fluoropolymer film. In this study, only pPFDMA, pV4D4-PFDA, and pV4D4-PFDMA were synthesized with iCVD, after which the durability of the each film was evaluated using the rubbing test. In addition, infiltration capability was examined after coating the folded copper sheet.

#### **2. Experiment**

#### *2.1. iCVD Process Basics*

The iCVD reaction mechanism is initiated by pyrolysis of the initiator to radicalize non-covalent electrons to bond with the vinyl group of the monomer and to grow the polymer thin film through the sequential chain bonding [25,26]. Notably, the initiator TBPO is radicalized through the pyrolysis and polymerized with the vinyl groups of monomers such as V4D4 and PFDMA, resulting in thin film growth. As shown in Equations (1)–(3), the parameter of iCVD process is defined as an optimal value of *P*M/*P*Sat. The *P*<sup>M</sup> and *P*Sat values are derived from the Antoine Equations (1) and (2), where *P*<sup>M</sup> is the ratio of the incoming gas amount controlled by the vaporization temperature of the monomer and the initiator, and the *P*Sat value is determined by monomer's characterization and the temperature of the substrate (*T*substrate). The optimized *P*M/*P*Sat is between 0 and 1, which is the rate of polymerization. When *P*M/*P*Sat is close to 0, the growth rate is very slow. When it is close to 1, condensation occurs instead of polymerization. Also, the duration of the deposition affects the thickness of the polymerized film. Optimized *P*M/*P*Sat varies depending on monomers.

$$P\_{\rm M} = P\_{\rm Chamber} \times \frac{F\_{\rm M}}{F\_{\rm M} + F\_{\rm I}} \tag{1}$$

$$\log P\_{\text{Sat}} = A - \frac{B}{T\_{\text{substrate}}} \tag{2}$$

$$0 < \frac{P\_{\rm M}}{P\_{\rm Sat}} < 1\tag{3}$$

where, *F*<sup>M</sup> (sccm) is a gaseous monomer input flow, *F*<sup>I</sup> (sccm) is a gaseous initiator input flow, *P*<sup>M</sup> (mTorr) is the partial pressure of the monomer in the chamber, *P*Sat (mTorr) is the saturated vapor pressure at the *T*Substrate, *P*Chamber (mTorr) is the total pressure of the chamber, and *T*Substrate ( ◦C) is the temperature of substrate.

The customized iCVD reactor system uses canisters for monomers and the initiator. The canisters are heated to efficiently vaporize the materials into the vacuum chamber. Also, there are filaments arrayed in the vacuum chamber to instantly apply the elevated temperature for decomposing the initiator into radicals.

#### *2.2. Fabrication of Superhydrophobic Film and its Characteristics*

Table 1 shows themonomers and theinitiator examinedin this study. The deposition of pV4D4-PFDMA was performed in the iCVD reactor. V4D4 and PFDMA monomers were heated to 55 and 75 ◦C, respectively, and TBPO was heated to 35 ◦C through each canister. Each flow rate of the vapor was controlled by needle valves, and the flow was monitored by a pressure gauge installed at the inlet of the reactor. In this study, for all monomers and initiator, the vapor saturation ratio (*F*M:*F*I) was maintained to be 1:1 as empirically optimized *P*M/*P*sat. First, V4D4 and TBPO were delivered into the reactor at 0.6 sccm while maintaining 270 mTorr vacuum pressure in the reactor. The substrate and the filaments

temperature were held at 40 and 180 ◦C, respectively. Then, a homopolymer of pV4D4 was grown from the surface of substrate. Once the desired thickness of pV4D4 was obtained, the next vaporized PFDMA was introduced into the reactor together with V4D4 for 2–3 min. This period ensures strong adhesion at the film interface between the top pV4D4 layer and the bottom pPFDMA layer. Then, the flow of V4D4 was stopped, and the deposition of pPFDMA continued to the top layer. With PFDMA, only the vacuum pressure in the reactor was changed to 100 mTorr to maintain the optimal *P*<sup>M</sup> value; other temperature settings remained the same. As a result, a conjugated fluoropolymer film was obtained as a superhydrophobic surface, as shown in Figure 2a. The film was sequentially stacked pV4D4 and pV4D4-PFDMA on a silicon wafer substrate. Figure 2b shows the morphology of the fluoropolymer film; one can see about 5 μm size fluorocarbon structures tangled from the top view. To confirm the super-hydrophobicity of film, the contact angle was measured with 50 μL of deionized water droplets using a contact angle meter (SmartDrop, FEMTOBIOMED, Seongnam, Korea). As a result, the water contact angle of 150.1◦ ± 3.6◦ was obtained while the sliding angle was only 8.5◦. This means that the pV4D4-PFDMA film exhibits a highly non-stick super-hydrophobic surface.


**Table 1.** Chemical structures and functions of the initiator and the monomers.

Because the temperature of the substrate was kept at 40 ◦C, relatively flexible substrates were considered such as paper and fabric. Accordingly, we choose tissue paper and cotton fabric as substrates and proceeded with the pV4D4-PFDMA coating process in the iCVD reactor. At this time, the total thickness of the film was 200 nm so there were no significant changes in dimensions of the fiber structures. Both tissue and fabric are known to easily get wet. After coating, however, blue ink was dropped on the surface, and excellent repellency to the droplet was observed, as shown in Figure 3a,b. In addition, we performed SEM analysis (FE-SEM, S-4800, Hitachi, Tokyo, Japan), on both substrates before and after coating as shown in Figure 3b–f, respectively. No noticeable changes were observed.

Another advantage of the iCVD process is that it shows ultra-step coverage on high aspect ratio of up to 1:40 for the opened area [27]. Therefore, we made a quantitative analysis of the capability of coating with shadow area. A manifolded copper sheet was subjected to the iCVD process and coated with 200 nm of the pV4D4-PFDMA, as seen in Figure 4a. The thickness and the width of the copper sheet were 0.02 and 25 mm, respectively, and it was firmly folded seven times, as seen in Figure 4b. Next, the gap between each surface was measured about 0.1 mm. The pure copper sheet had a hydrophilic surface, as seen in Figure 4c; therefore, water droplets on various places of the unfolded sheet indicated the successful coating, as seen in Figure 4d. Additionally, Fourier transform infrared spectra (Nicolet 6700 FT-IR Spectrometer, Thermo Scientific, Waltham, MA, USA) analysis has proven an infiltration capability, as shown in Figure 4e. Although the peaks are not so obvious due to an extremely high aspect ratio for the shadow area, the result verifies the presence of the fluorocarbon functional group.

**Figure 2.** (**a**) A schematic of the fluoropolymer (pV4D4-PFDMA) coating; (**b**) surface morphology using scanning electron microscope (FE-SEM, S-4800, Hitachi, Tokyo, Japan) imaging before and after coating on wafer substrate; (**c**) static and dynamic water contact angles of the pV4D4-PFDMA surface.

**Figure 3.** (**a**) Blue ink droplet on tissue paper before and after pV4D4-PFDMA coating; (**b**) SEM image of the tissue paper; (**c**) SEM image of the tissue structure after coating; (**d**) blue ink droplet on cotton fabric before and after pV4D4-PFDMA coating; (**e**) SEM image of the cotton fabric; and (**f**) SEM image of the cotton fabric after coating.

**Figure 4.** (**a**) A manifolded copper sheet in the iCVD reactor; (**b**) schematic of the folded copper sheet and FT-IR Spectrometer measurement locations; (**c**) water droplets on bare copper sheet; (**d**) water droplets on unfolded copper sheet after coating; and (**e**) FT-IR analysis before and after coating.

#### *2.3. Examination of Durability*

The development of a thin film that is resistant to various mechanical stresses is the key for successful commercialization of the superhydrophobic surfaces. The proposed film was obtained using the iCVD process, with a mechanically robust organosilicon polymer and a fluoropolymer layer. The two layers were covalently adhered to form a conjugated polymer structure. We evaluated the durability of the proposed coating against tensile and abrasion tests.

To compare the mechanical durability against deformation under external force between pPFDMA and pV4D4-PFDMA, we examined optical microscopic images of the cracks developed when the tensile load is applied. The pPFDMA (500 nm) and pV4D4 (100 nm)-PFDMA (500 nm) were deposited on the elastomer film (50 μm thickness) and subjected to tensile tests using Instron tester (5960 Series, INSTRON, Chicago, IL, USA) according to ASTM standard [28], as shown in Figure 5a. Tensile tests for each thin film were proceeded with the strain rate of 2.5 mm/min, and the obtained stress-strain plots were compared with the bare elastomer film, pPFDMA coated, and pV4D4-PFDMA coated, as seen in Figure 5d. According to the result, the pPFDMA coating shows the tensile strength was similar to that of the bare elastomer film; surface cracking was observed when the strain reached about 60%, as seen in Figure 5b,c. However, for pV4D4-PFDMA coating, crack occurrence was observed at 110% of strain, indicating an improvement of the adhesion strength between the elastomer film and the fluoropolymer film, resulting in eventual enhancement of the global tensile strength.

**Figure 5.** Tensile test of the elastomer film with superhydrophobic coating: (**a**) tensile test set-up; (**b**) surface cracks occurred during tensile loading; (**c**) microscopic image of the surface cracks; and (**d**) stress-strain curves.

The combined pV4D4 and pPFDMA film showed better stability and the super-hydrophobicity because of the covalent bonding between the interlayers. In this structure, benzene ring of the V4D4 acts as a cross linker, protecting the conjugated pV4D4–PFDMA from external stresses that normally occur in daily life. Accordingly, we evaluated the resistance in the abrasion test (CT-RB Series, CORETECH, Uiwang, Korea) and compared pPFDMA (500 nm) only, pV4D4 (100 nm)-PFDA (500 nm), and pV4D4 (100 nm)-PFDMA (500 nm). For the abrasion test, these films were deposited onto the SUS304 plate; the comparison was made using a contact angle meter. Figure 6a shows the abrasion test set-up. The coated sample was placed and firmly fixed in the sample holder. Then, the cotton-covered ball-shape tip was released to contact the top surface of the coated sample. 1 kgf weight was loaded on the tip, which was repeatedly rubbed on the surface left- and right-hand side throughout 3000 cycles. The contact angle was measured at every five cycles in the beginning and intermittently after hundreds of cycles. In abrasion test, the coated layer appeared to be scratched on the surface, as seen in Figure 6c and contact angle was gradually decreased. Figure 6d shows the comparison of the contact angle change for each film. According to the results, pPFDMA only showed the lowest abrasion resistance indicating drastic reduction of the contact angle within less than 10 cycles. However, the other two fluoropolymer films with pV4D4 conjugation withstood 200 cycles of rubbing and retained contact angle of 120◦, which is considered as sufficiently high hydrophobicity. The proposed pV4D4-PFDMA showed the highest abrasion resistance, retaining 120◦ of the contact angle after 3000 cycles.

**Figure 6.** (**a**) Abrasion test set-up; (**b**) microscopic image of the initial surface; (**c**) microscopic image of the tested surface; and (**d**) contact angle measurements.

#### **3. Conclusions**

A fluoropolymer fabricated using the iCVD process has great advantages such as cost efficiency and high functionality when compared with other surface modification solutions to obtain super hydrophobicity. In this study, we examined in detail parametric optimization of the iCVD process and successfully fabricated a new super-hydrophobic film conjugated with V4D4 and PFDMA. The proposed superhydrophobic film was applied to tissue paper and cotton fabric and demonstrated great liquid repellency. In addition, high infiltration capability of the iCVD process was discussed using manifolded copper sheet. These results provide enough insight for industrial applications in which superhydrophobic surfaces are needed.

A conjugated film (pV4D4-PFDMA) was achieved by adding the pV4D4 layer before the introduction of the pPFDMA and showed exceptional stability and durability. The pV4D4 significantly enhances mechanical stability of the pPFDMA as it allowed for both monomers to flow into the reactor. Therefore, the fluoropolymer was reinforced by binding the organosilicon layer. We evaluated the mechanical and chemical robustness of the proposed film. First, the tensile test was performed using the deposition on the elastomer film; it showed improvement both in terms of the tensile strength and

delay in surface cracking. Second, the mechanical abrasion test was performed and the proposed film showed better rubbing resistance when compared with other films.

The deposition process is applicable to various types and complex shapes of the substrates without the need of surface pretreatment; is also allows for improved adhesion between the coated film and the substrate. We found that the proposed superhydrophobic film obtained in the iCVD process provides an industrial grade of low surface energy with sufficient durability against various mechanical stresses. In addition, optimization of iCVD process can be further studied to enhance mechanical properties.

**Author Contributions:** Conceptualization, J.B.K.; Data Curation, H.K.; Formal Analysis, J.H.L.; Investigation, H.S.L. and J.B.K; Methodology, H.K.; Writing—Original Draft, H.S.L.; Writing—Review and Editing, J.B.K.

**Funding:** This research was funded by National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIT) (No. NRF-2018R1C1B5045726) and Korea Institute for Advancement of Technology (KIAT) grant funded by the Korea Government (MOTIE) (P0002092, The Competency Development Program for Industry Specialist).

**Conflicts of Interest:** The authors declare no conflict 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/).

## **Simple Non-Destructive Method of Ultrathin Film Material Properties and Generated Internal Stress Determination Using Microcantilevers Immersed in Air**

**Ivo Stachiv 1,2,3,\* and Lifeng Gan <sup>1</sup>**


Received: 5 June 2019; Accepted: 27 July 2019; Published: 1 August 2019

**Abstract:** Recent progress in nanotechnology has enabled to design the advanced functional micro-/nanostructures utilizing the unique properties of ultrathin films. To ensure these structures can reach the expected functionality, it is necessary to know the density, generated internal stress and the material properties of prepared films. Since these films have thicknesses of several tens of nm, their material properties, including density, significantly deviate from the known bulk values. As such, determination of ultrathin film material properties requires usage of highly sophisticated devices that are often expensive, difficult to operate, and time consuming. Here, we demonstrate the extraordinary capability of a microcantilever commonly used in a conventional atomic force microscope to simultaneously measure multiple material properties and internal stress of ultrathin films. This procedure is based on detecting changes in the static deflection, flexural and torsional resonant frequencies, and the corresponding quality factors of the microcantilever vibrating in air before and after film deposition. In contrast to a microcantilever in vacuum, where the quality factor depends on the combination of multiple different mechanical energy losses, in air the quality factor is dominated just by known air damping, which can be precisely controlled by changing the air pressure. Easily accessible expressions required to calculate the ultrathin film density, the Poisson's ratio, and the Young's and shear moduli from measured changes in the microcantilever resonant frequencies, and quality factors are derived. We also show that the impact of uncertainties on determined material properties is only minor. The validity and potential of the present procedure in material testing is demonstrated by (i) extracting the Young's modulus of atomic-layer-deposited TiO2 films coated on a SU-8 microcantilever from observed changes in frequency response and without requirement of knowing the film density, and (ii) comparing the shear modulus and density of Si3N4 films coated on the silicon microcantilever obtained numerically and by present method.

**Keywords:** thin film; atomic layer deposition; nanomechanics; Young's modulus; shear modulus; resonant frequency; *Q*-factor; microcantilevers; internal stress

#### **1. Introduction**

Functional micro-/nanostructures made of substrate and one or multiple ultrathin films are widely used in applications like photovoltaics [1], micro-electronics [2], optics [3,4], tunable resonators [5,6], and various sensors [7–12]. Preparation of these structures involves repeated usage of multiple fabrication processes such as deposition, lithography, etching, and cleaning. In order to prevent the mechanical failure or to guarantee that the structures would reach the desired operating conditions, the material properties of prepared ultrathin films must be known. During film deposition the internal

stress that often originates from a coefficient of thermal expansion mismatch can be generated [13]. For films of thicknesses ranging from hundreds of nm to about tens of μm, the nanoindentation [14], bulge test [15], and the resonant methods [16] are the most common techniques used to determine the film material properties. Noticing that for micro-sized samples the density is estimated based on the known bulk values. However, once film thickness shrinks from micro- to nanoscale (i.e., tens of nm), the density of particularly polymer, organic, composite, or porous films can start to deviate from the bulk values (e.g., changes in deposition parameters affect notably prepared film density) [17,18]. In addition, usage of different film preparation processes can also cause significant variations in the material properties and density of designed nanostructure. As a result, current procedures of ultrathin film material properties determination require either simultaneous measurements on multiple sophisticated devices [19–21] or the specially designed micro-/nanomechanical resonators or experimental setup with the advanced computational tools [22–26]. For instance, the high-resolution transmission electron microscope is used to precisely control the force loading/unloading during the nanoindentation of a nanoscale sample [20,21]. Drawbacks of these procedures are that combined measurements on either several sophisticated devices or one specially-designed device are difficult to perform, time consuming, often expensive, and each developed procedure is usually limited to only a specific class of materials. The resonant methods can be also integrated in situ into the nanomaterial deposition systems [27,28].

In response, here we demonstrate the outstanding capability of common microcantilever to determine the density, generated internal stress, the Poisson's ratio, and the elastic properties of solid and polymer ultrathin films, from measured static and dynamic responses of the microcantilever, before and after depositing a thin layer film on its surface. Sketch of the microcantilever with the deposited film is given in Figure 1. We emphasize here that the present procedure utilizes, in addition to well-established measurements of the cantilever static deflection, the flexural and torsional resonant frequencies of the cantilever operating in air; also monitoring often neglected changes in the corresponding quality factors (*Q*-factors). As a direct consequence, no additional experimental setup or specially designed microcantilevers are required, enabling non-destructive and easily accessible material characterization and testing of ultrathin films. Noticing that *Q*-factor is a dimensionless parameter describing the efficiency of the designed resonator (i.e., higher *Q*-factor values stand for lower dissipation and higher efficiency). Importantly, in air the *Q*-factor is proportional to the material properties and dimensions of the designed microcantilever, and a known air damping that can be precisely controlled by changing the air pressure [29]. Other energy losses, such as the support, surface or the thermo-elastic loss, have only a negligibly small impact on the *Q*-factor of microcantilever submerged in air. When film is sputtered on the resonator surface, it alters the material properties and dimensions of the microcantilever resonator, yielding changes in *Q*-factor. As such, *Q*-factor provides an additional source of information on prepared ultrathin film(s).

**Figure 1.** Sketch of the two-layered microcantilever made of an elastic substrate and coated ultrathin film performing (**a**) flexural and (**b**) torsional oscillations.

We first derive easily accessible expressions needed to calculate the material properties of an ultrathin film from observed changes in flexural and torsional resonant frequencies and *Q*-factors of the microcantilever operating in air. Then, we analyze the sensitivity of calculated material properties of ultrathin film on uncertainties in frequency (*Q*-factor) and dimensions measurements and, afterwards, we validate our theoretical findings by comparing theoretical predictions with experimental results and numerical computations. Despite the fact that analysis is carried out on rectangular microcantilevers, the obtained results and developed procedure of thin film material characterization are valid for other cantilever shapes. In this case, just the flexural and torsional rigidities and hydrodynamic functions used in the model must be recalculated.

#### **2. Theory**

#### *2.1. Flexural Oscillations of Two-Layered (Multi-Layered) Microcantilever Operating in Air*

To begin, we recall a known fact that once a thin layer film is sputtered on an elastic substrate, it generates in-plane stresses and also alters the overall cantilever resonator elastic properties, particularly in near vicinity of its clamped end [6,30–33]. These effects, that originate from mismatches in strains and the coefficient of thermal expansion between substrate and film, have been proven to notably affect the resonant frequencies of ultrathin cantilever resonators (i.e., thin sheets) [32,33]. Nevertheless, for relatively thick microcantilevers (e.g., ultrathin film sputtered on a thick elastic substrate), of which are considered in the present work, the cantilever free end allows the generated internal stress to be relaxed [31]. Hence, for out-plane flexural vibrational modes, the governing equation for the dynamic deflection function *u*(*x*,*t*) of the microcantilever consisting of substrate and coated film (see Figure 1a) is given by

$$\left(\rho\_1 \mathbf{S}\_1 + \rho\_2 \mathbf{S}\_2\right) \frac{\partial^2 u(\mathbf{x}, t)}{\partial t^2} + D\_\mathbf{F} \frac{\partial^4 u(\mathbf{x}, t)}{\partial \mathbf{x}^4} = F\_{\text{drive}}(\mathbf{x}, t) + F\_{\text{hydro}}(\mathbf{x}, t), \tag{1}$$

where *DF* = <sup>1</sup> 12*E*1*WT*<sup>3</sup> <sup>1</sup>*r*(ξ*F*, η); subscript 1 and 2 stand for substrate and film, respectively; ρ, *S*, *W*, *T* are the density, cross sectional area, width, and thickness, respectively; *r*(ξ*F*,η) = [ξ*<sup>F</sup>* <sup>2</sup>η<sup>4</sup> + 4ξ*F*η(1 + 1.5η + η2) + 1]/(1 + ξ*F*η), ξ*<sup>F</sup>* = *E*2/*E*1, η = *T*2/*T*1, *F*drive(*x*, *t*) is the external driving force per unit length of an arbitrary form that set the microcantilever into motion; *F*hydro(*x*, *t*) is the hydrodynamic force of surrounding air.

It shall be pointed out that detailed theoretical analysis of a homogeneous microcantilever (i.e., made of one material layer) performing flexural oscillations in air can be found in [34,35]. In present work, we extend these theoretical results to account for two material layers required to characterize the material properties of ultrathin films. Moreover, our results can be also directly applied to the microcantilever consisting of *N* material layers just by recalculating linear density, ρ*S* (where ρ can be viewed as the effective density and *S* is the cantilever cross-sectional area), and flexural rigidity using the following general relationships:

$$\rho \text{S} = \sum\_{i=1}^{N} \rho\_i \text{S}\_{i\prime} D\_{\text{F}} = \sum\_{i=1}^{N} E\_i \int\_{\prod i} u^{\*2} \text{dS} - \frac{\left(\sum\_{i=1}^{N} E\_i \int\_{\prod i} u^{\*} \text{dS}\right)^2}{\sum\_{i=1}^{N} E\_i \text{S}\_i} \,\tag{2}$$

where *u*\* is the local coordinate in the lateral direction, *<sup>i</sup>* is the *i*-th region of cantilever beam cross section [36].

The general form of the hydrodynamic force obtained by solving the Fourier-transformed continuity and Navier–Stokes equations (i.e., computations are in the time domain Fourier-transform), for an incompressible fluid as *<sup>F</sup>*hydro(*x*|ω) = <sup>κ</sup>*F*ρairω2*W*2Γ*F*(ω)*U*(*x*|ω), where *<sup>U</sup>*(*x*|ω) is the Fourier-transformed deflection function, κ*<sup>F</sup>* = <sup>π</sup> <sup>4</sup> , ρair is the air density, and Γ*F*(ω) is the hydrodynamic

*Coatings* **2019**, *9*, 486

function for flexural vibration mode. Then, taking the Fourier transform of Equation (1) and rearranging terms yields:

$$\frac{d^4lI(\mathbf{x}|\boldsymbol{\omega})}{d\mathbf{x}^4} - \left(\gamma\_{\left(n\right)}^2 \frac{\boldsymbol{\omega}}{\boldsymbol{\omega}\_{\rm V\_F}^{\left(n\right)}}\right)^2 \left[1 + \frac{\boldsymbol{\kappa}\_F \rho\_{\rm air} \mathcal{W}}{\rho\_1 \mathcal{T}\_1 (1 + \mu \eta)} \boldsymbol{\Gamma}\_F(\boldsymbol{\omega})\right] lI(\mathbf{x}|\boldsymbol{\omega}) = \widetilde{\mathcal{F}}\_{\rm drive}(\mathbf{x}|\boldsymbol{\omega}),\tag{3}$$

where <sup>μ</sup> <sup>=</sup> <sup>ρ</sup>2/ρ1, *F*drive(*x*|ω) <sup>=</sup> *<sup>F</sup>*drive(*x*|ω)*L*4/*DF*, <sup>ω</sup>(*n*) <sup>v</sup>*<sup>F</sup>* <sup>=</sup> <sup>γ</sup>(*n*) *L* <sup>2</sup> *DF*/[ρ1T1(1 + μη)] is the cantilever angular resonant frequency in vacuum of the *n*-th vibrational mode, *n* = 1, 2, 3, ... stands for the considered vibrational mode, *L* is the cantilever length, ω is a characteristic angular frequency of the microcantilever oscillations and γ(*n*) is obtained as the positive root(s) of the following characteristic transcendental equation:

$$
\cosh \gamma \,\cos \gamma + 1 = 0.\tag{4}
$$

For an arbitrary form of the driving force the general solution of Equation (3) can be found by the eigenfunction expansion method. In this case, the dynamic deflection function can be obtained as a linear combination of the microcantilever mode shapes, <sup>θ</sup>*F*(*n*)(*x*) = sin h γ(*n*)*x* <sup>−</sup> sin γ(*n*)*x* − sin h(γ(*n*))+sin(γ(*n*)) cos h(γ(*n*))+cos(γ(*n*)) × cos h γ(*n*)*x* <sup>−</sup> cos γ(*n*)*x* (see [37]) and the one reads:

$$\mathcal{U}I(\mathbf{x}|\boldsymbol{\omega}) = \sum\_{n=1}^{\infty} \mathcal{B}\_{F(n)}(\boldsymbol{\omega}) \boldsymbol{\theta}\_{F(n)}(\mathbf{x}),\tag{5}$$

where *BF*(*n*)(ω) is found using the orthonormal properties of θ*F*(*n*)(*x*) as

$$B\_{F(n)}(\omega) = \frac{\int\_0^L \widetilde{F}\_{\rm dr}(\widetilde{\bf x}|\omega) \theta\_{F(n)}(\widetilde{\bf x}) d\widetilde{\bf x}}{\gamma\_{(n)}^4 - \left(\gamma\_{(n)}^2 \frac{\omega}{\omega\_{\rm vir}^{(n)}}\right)^2 \left[1 + \frac{\kappa \overline{\rho} \rho\_{\rm vir} W}{\rho\_1 \overline{\bf T}\_1 (1 + \mu \eta)} \Gamma\_F(\omega)\right]}.\tag{6}$$

The dissipative effect of air is small compared to viscous fluid (i.e., *Q*(*n*) >> 1 [34,35,38]); therefore, in a vicinity of the resonance peaks Γ*F*(ω) ≈ Γ*F*\_*r*(ω) + *i*Γ*F*\_*im*(ω), where Γ*F*\_*r*(ω) and Γ*F*\_*im*(ω) are the real and imaginary components of the dimensionless hydrodynamic function for cantilever performing flexural oscillations. Then, the resonant frequency and *Q*-factor of *n*-th microcantilever vibrational mode in air can be obtained with an analogy to a simple harmonic oscillator as:

$$\alpha\_F^{(n)} = \frac{\alpha\_{\rm V\_F}^{(n)}}{\sqrt{1 + \frac{\kappa\_F \rho\_{\rm air} W}{\rho\_1 T\_1 (1 + \mu \eta)} \Gamma\_{F\_{-}} \left(\alpha\_F^{(n)}\right)}} \tag{7}$$

$$Q\_F^{(n)} = \frac{\frac{\rho\_1 T\_1 (1 + \mu \eta)}{\kappa\_F \rho\_{\text{air}} W} + \Gamma\_{F\_-} \left(\omega\_F^{(n)}\right)}{\Gamma\_{F\_- \text{im}} \left(\omega\_F^{(n)}\right)}. \tag{8}$$

For reader's convenience, we present dependencies of the real Γ*F*\_*<sup>r</sup>* ω(1) *F* and imaginary <sup>Γ</sup>*F*\_*im* ω(1) *F* components of the hydrodynamic function on the fundamental mode frequency represented through the Reynolds number, Re = <sup>π</sup> <sup>4</sup> <sup>ω</sup>*W*2ρair/μair, where <sup>μ</sup>air is the viscosity of air, in Figure 2a. Noticing only that the Reynolds number is a dimensionless parameter used to predict the flow pattern. For a microcantilever consisting of *N* material layers, the normalized "effective" density ρ1*T*1(1 + μη) in Equations (7) and (8) is replaced by *<sup>N</sup> <sup>i</sup>*=<sup>1</sup> ρ*iTi*. Furthermore, for higher vibrational modes the hydrodynamic function depends on the following two dimensionless parameters: (i) The Reynolds number defined now as Re = ω*W*2ρair/μair; and (ii) the normalized mode shape given by κ =

γ(*n*)*W*/*L* [39]. The exact form of the hydrodynamic function for an arbitrary mode number can be found in [39,40]. The general solution for the dynamic deflection has been obtained using the eigenfunction expansion method (see structure of Equation (5)); therefore, the obtained expressions for the resonant frequency and *Q*-factor represented by Equations (7) and (8) are valid for an arbitrary vibrational mode (i.e., for higher modes only the hydrodynamic function must be recalculated).

**Figure 2.** Calculated dependencies of the real and imaginary components of the hydrodynamic function on the fundamental resonant frequency of a rectangular two-layered microcantilever (see Figure 1) immersed in air, represented through the Reynolds number for (**a**) flexural and (**b**) torsional oscillations.

#### *2.2. Torsional Oscillations of Two-Layered (Multi-Layered) Microcantilever Operating in Air*

The resonant frequency and *Q*-factor of a two(multi)-layered microcantilever vibrating in air can be obtained in the same way as done in the previous section for flexural oscillations. Briefly, by accounting for the membrane analogy proposed by Prandtl in 1903, similarities with the theoretical model for flexural oscillations of a multilayered beam in vacuum given in Zapomel et al. [36] and theory of Green and Sader [41], the general governing equation and boundary conditions for torsional oscillations of a two-layered microcantilever (see Figure 1b) operating in air takes the following form:

$$D\_{Tr}\frac{\partial^2 \phi(\mathbf{x},t)}{\partial \mathbf{x}^2} - \frac{\rho\_1 \mathcal{W}^3 T\_1}{12} \left[1 + \varepsilon\_1^2 + \mu \eta \left(1 + \varepsilon\_1^2 \eta^2\right)\right] \frac{\partial^2 \phi(\mathbf{x},t)}{\partial t^2} = M\_{\text{drive}}(\mathbf{x},t) + M\_{\text{hydro}}(\mathbf{x},t), \tag{9}$$

$$
\phi(0,t) = 0, \ \frac{\partial \phi(L,t)}{\partial x} = 0. \tag{10}
$$

Here φ(*x*, *t*) is the deflection angle about the cantilever major axis, ε<sup>1</sup> = *T*1/*W* is the characteristic dimensional scale, *DTr* = <sup>1</sup> 3*G*1*WT*<sup>3</sup> <sup>1</sup>*r*(ξ*Tr*, <sup>η</sup>) is the torsional rigidity, *<sup>r</sup>*(ξ*Tr*, <sup>η</sup>) <sup>=</sup> [ξ*Tr*2η<sup>4</sup> <sup>+</sup> <sup>4</sup>ξ*Tr*η(1 <sup>+</sup> 1.5<sup>η</sup> + η2) + 1]/(1 + ξ*Tr*η), ξ*Tr* = *G*2/*G*1, *G* is the shear modulus; *M*drive(*x*, *t*) is the driving moment per unit length and *M*hydro(*x*, *t*) is the hydrodynamic torque per unit length, which the general form is obtained by solving the equation of motion of fluid in complex space [39,41], given by:

$$M\_{\rm hydro}(\mathbf{x}|\omega) = -\frac{\pi}{8} \rho\_{\rm air} \omega^2 \mathcal{W}^4 \Gamma\_T(\omega) \Phi(\mathbf{x}|\omega),\tag{11}$$

*Coatings* **2019**, *9*, 486

where Γ*T*(ω) is the torsional dimensionless hydrodynamic function, Φ(*x*|ω) is the deflection angle in complex space. Plugging Equation (11) into the Fourier-transformed Equation (9) and rearranging terms yields:

$$\frac{d^2\Phi(\mathbf{x}|\boldsymbol{\omega})}{d\mathbf{x}^2} - \left(\lambda\_{\text{(n)}}\frac{\boldsymbol{\omega}}{\boldsymbol{\omega}^{(\boldsymbol{n})}\_{\text{vTr}}}\right)^2 \left[1 + \frac{\boldsymbol{\kappa}\_T \rho\_{\text{air}} \mathcal{W}}{\rho\_1 T\_1 \left[1 + \boldsymbol{\varepsilon}\_1^2 + \mu \eta \left(1 + \boldsymbol{\varepsilon}\_1^2 \boldsymbol{\eta}^2\right)\right]} \Gamma\_T(\boldsymbol{\omega})\right] \Phi(\mathbf{x}|\boldsymbol{\omega}) = \tilde{M}\_{\text{drive}}(\mathbf{x}|\boldsymbol{\omega}), \tag{12}$$

where <sup>κ</sup>*<sup>T</sup>* <sup>=</sup> <sup>3</sup>π/2, *<sup>M</sup>*drive(*x*|ω) <sup>=</sup> *<sup>M</sup>*drive(*x*|ω)/*DTr*, <sup>ω</sup>(*n*) <sup>v</sup>*Tr* <sup>=</sup> <sup>λ</sup>(*n*) *L* <sup>4</sup>*G*1*T*<sup>2</sup> <sup>1</sup>*r*(ξ*Tr*,η) ρ1*W*<sup>2</sup>[1+ε<sup>2</sup> 1+μη(1+ε<sup>2</sup> <sup>1</sup>η<sup>2</sup>)] is the angular resonant frequency in vacuum of the *n*-th torsional mode and λ(*n*) = π(2*n* − 1)/2, *n* = 1, 2, 3, ... .

The general solution for torsional oscillations of the microcantilever driven by an arbitrary form torque can be again obtained by the eigenfunction expansion method:

$$\Phi(\mathbf{x}|\omega) = \sum\_{n=1}^{\infty} B\_{Tr(n)}(\omega) \theta\_{Tr(n)}(\mathbf{x}),\tag{13}$$

where θ*Tr*(*n*)(*x*) = sin[(2*n* − 1)π*x*/2], *n* = 1, 2, 3, ... and *BTr*(*n*)(ω) is given by

$$B\_{Tr(n)}(\omega) = \frac{2\int\_0^L \widetilde{M}\_{\text{dr}}(\overline{\mathbf{x}}|\omega)\theta\_{Tr(n)}(\overline{\mathbf{x}})d\overline{\mathbf{x}}}{\left(\lambda\_{(n)}\frac{\omega}{\omega\_{Tr}^{(n)}}\right)^2 \left[1 + \frac{\kappa\_T \rho\_{\text{air}} W}{\rho\_1 T\_1 \left[1 + \kappa\_1^2 + \mu \eta \left(1 + \kappa\_1^2 \eta^2\right)\right]}\Gamma\_T(\omega)\right] - \lambda\_{(n)}^2}.\tag{14}$$

Then, in analogy with flexural motion, for small dissipative effects the desired expressions that enable to accurately predict the torsional resonant frequency and *Q*-factor of the *n*-th vibrational mode of microcantilever consisting of substrate and ultrathin film (i.e., η << 1 and ε<sup>1</sup> < 1) operating in air are:

$$
\omega\_{Tr}^{(n)} \approx \frac{\omega\_{\rm vir}^{(n)}}{\sqrt{1 + \frac{\kappa\_T \rho\_{\rm air} \mathcal{W}}{\rho\_1 T\_1 (1 + \mu \eta)} \Gamma\_T \left(\alpha\_{Tr}^{(n)}\right)}} \tag{15}
$$

and

$$Q\_{Tr}^{(n)} \approx \frac{\frac{\rho\_1 T\_1 (1 + \mu \eta)}{\kappa\_T \rho\_{\rm air} W} + \Gamma\_{T\_J} \left(\alpha\_{Tr}^{(n)}\right)}{\Gamma\_{T\_{\rm im}} (\alpha\_{Tr}^{(n)})},\tag{16}$$

where Γ*T*\_*<sup>r</sup>* ω(*n*) *Tr* and <sup>Γ</sup>*T*\_*im* ω(*n*) *Tr* are the real and imaginary components of the dimensionless hydrodynamic function that, for fundamental torsional vibrational mode, are given in Figure 2b.

For a multilayered beam, just coefficients for torsional mass and hydrodynamic effect of fluid must be recalculated. In general cases, ρ1*T*1(1 + μη) is replaced by *<sup>N</sup> <sup>i</sup>*=<sup>1</sup> ρ*iI*<sup>P</sup>*i*, where *I*<sup>p</sup> is the polar moment of inertia and the hydrodynamic effect of air is now represented by π 8 ρair*W*<sup>4</sup> [see Equation (11)].

#### **3. Results**

#### *3.1. Method of Determining Material Properties, Density, and Generated Internal Stress of Ultrathin Film(s)*

We first evaluate impact of coated film on the microcantilever resonant frequency and *Q*-factor and, afterwards, we derive easily accessible expressions enabling calculation of the density, the Young's and shear moduli, the Poisson's ratio and stress of the solid and polymer ultrathin films from experimentally observed changes in the microcantilever resonant frequencies, *Q*-factors, and the cantilever static deflection. Using Equations (7), (8), (15), and (16), the desired changes in resonant frequency and

*Q*-factor of the *n*-th vibrational mode, represented by ratio of the microcantilever made of substrate and coated film (i.e., ω and *Q*) to the one of without film (i.e., ω<sup>0</sup> and *Q*0), are obtained as

$$\frac{\alpha\_j^{(n)}}{\alpha\_{j0}^{(n)}} = \sqrt{r(\xi\_j, \eta) \times \left[\frac{1 + \mathbb{C}\_j \Gamma\_{j, \mathcal{I}}\left(\alpha\_{j0}^{(n)}\right)}{1 + \mu \eta + \mathbb{C}\_j \Gamma\_{j, \mathcal{I}}\left(\alpha\_j^{(n)}\right)}\right]},\tag{17}$$

$$\frac{\mathcal{Q}\_{j}^{(n)}}{\mathcal{Q}\_{j0}^{(n)}} = \frac{\Gamma\_{j\\_jm}(\boldsymbol{\omega}\_{j0}^{(n)})}{\Gamma\_{j\\_jm}(\boldsymbol{\omega}\_{j}^{(n)})} \times \left[ \frac{1 + \mu \eta + \mathcal{C}\_{\boldsymbol{j}} \Gamma\_{\boldsymbol{j}\\_\mathcal{I}} \left(\boldsymbol{\omega}\_{\boldsymbol{j}}^{(n)}\right)}{1 + \mathcal{C}\_{\boldsymbol{j}} \Gamma\_{\boldsymbol{j}\\_\mathcal{I}} \left(\boldsymbol{\omega}\_{\boldsymbol{j}0}^{(n)}\right)} \right],\tag{18}$$

where subscript *j* = *F* (flexural) and *Tr* (torsional), and *Cj* = κ*<sup>j</sup>* ρair *W* <sup>ρ</sup><sup>1</sup> *<sup>T</sup>*<sup>1</sup> .

Once film is coated on the elastic substrate, it alters the microcantilever flexural and torsional rigidity represented by the dimensionless parameter *r*(ξ*j*,η) and increases its "effective" linear density through the coefficient μη. According to Equations (17) and (18), changes in the microcantilever resonant frequency caused by the film differ essentially from those obtained for *Q*-factor of the same vibrational mode. Since dissipative effect of air is small [38], changes in the resonant frequency depend just on interplay between the rigidity and effective linear density of the prepared sample as *r*(ξ*j*,η)/(1 + μη). As a result, an increase, decrease, or even non-monotonic dependency of ω/ω<sup>0</sup> on film thickness can be observed depending on the exact film and substrate material properties and density values. Quality factor is; however, proportional to combination of the linear density and known hydrodynamic load represented by Γ*j*\_*<sup>r</sup>* and Γ*j*\_*im*. We note that rigidity affects the *Q*-factor only indirectly through the resonant frequency used to calculate both components of the hydrodynamic function (see Equation (18)). As such, with an increase of film thickness only an increase in *Q*-factor can be observed. It immediately implies that combined measurements of the resonant frequency and *Q*-factor changes enable evaluation of the material properties of ultrathin film, even when no shift in the resonant frequency can be observed [24]. For example, dependencies of the fundamental resonant frequency and *Q*-factor changes of the silicon microcantilever of length *L* = 300 μm, *W* = 30 μm, and *T*<sup>1</sup> = 1 μm (ρ<sup>1</sup> = 2.33 g/cm3, *E*<sup>1</sup> = 169 GPa, and *G*<sup>1</sup> = 42 GPa) on thickness of film made of gold (ρ<sup>2</sup> = 19.3 g/cm3, *E*<sup>2</sup> = 79 GPa, and *G*<sup>2</sup> = 27 GPa), platinum (ρ<sup>2</sup> = 21.45 g/cm3, *E*<sup>2</sup> = 168 GPa, and *G*<sup>2</sup> = 61 GPa), and silicon nitride (ρ<sup>2</sup> = 3.2 g/cm3, *E*<sup>2</sup> = 350 GPa, and *G*<sup>2</sup> = 100 GPa) are given in Figure 3. As expected, an increase and/or decrease in the frequency ratio can be observed depending on the density and material properties of coated film (Figure 3a,b). For gold and platinum (silicon nitride), film density (rigidity) dominates the frequency response, thus with an increase of film thickness the resonance shifts to lower (higher) values. For a given film thickness, the higher *Q*/*Q*<sup>0</sup> values can be achieved for heavier films (i.e., platinum and gold; Figure 3c,d). Importantly, obtained theoretical predictions are in a good agreement with published experimental observations carried out on the microcantilever resonator-based biosensor [42]. In these experiments, the antibody and antigen formed thin layer films on the cantilever surface, yielding both an increase and decrease in the resonant frequency depending on the interplay between stiffness and stress effects, and just an increase in *Q*-factor as predicted by the present model.

**Figure 3.** Dependencies of (**a**,**b**) the fundamental resonant frequencies and, correspondingly, (**c**,**d**) *Q*-factors of the silicon cantilever on film thickness for flexural and torsional vibrational modes.

Rearranging terms in Equations (17) and (18), the density, and the Young's and shear moduli of prepared ultrathin film can be determined from the following equations:

$$\mu = \frac{1}{\eta} \frac{Q\_j^{(n)}}{Q\_{j0}^{(n)}} \frac{\Gamma\_{j,jm}(\boldsymbol{\omega}\_j^{(n)})}{\Gamma\_{j,jm}(\boldsymbol{\omega}\_{j0}^{(n)})} \Big(1 + \mathbb{C}\_j \Gamma\_{j,J} \Big(\boldsymbol{\omega}\_{j0}^{(n)}\Big)\Big) - \frac{1}{\eta} \Big(1 + \mathbb{C}\_j \Gamma\_{j,J} \Big(\boldsymbol{\omega}\_j^{(n)}\Big)\Big),\tag{19}$$

$$\sigma(\xi\_{j},\eta) = \left(\frac{\alpha\_{j}^{(n)}}{\alpha\_{j0}^{(n)}}\right)^{2} \frac{Q\_{j}^{(n)}}{Q\_{j0}^{(n)}} \frac{\Gamma\_{j\dots jn}(\alpha\_{j}^{(n)})}{\Gamma\_{j\dots jn}(\alpha\_{j0}^{(n)})}.\tag{20}$$

The Young's modulus is related to the shear modulus as *E* = 2*G*(1 + υ), where υ is the Poisson's ratio. Hence, accounting for Equation (20), the Poisson's ratio of coated film is then calculated by:

$$
\omega\_2 \approx \frac{(R\_F - 1)B\_T}{(R\_T - 1)B\_F} (1 + \upsilon\_1) - 1,\tag{21}
$$

where *Rj* = ⎛ ⎜⎜⎜⎜⎝ <sup>ω</sup>(*n*) *j* <sup>ω</sup>(*n*) *j*0 ⎞ ⎟⎟⎟⎟⎠ 2 *<sup>Q</sup>*(*n*) *j <sup>Q</sup>*(*n*) *j*0 <sup>Γ</sup>*j*\_*im* <sup>ω</sup>(*n*) *j* <sup>Γ</sup>*j*\_*im* <sup>ω</sup>(*n*) *j*0 and *Bj* <sup>=</sup> <sup>4</sup> <sup>+</sup> <sup>6</sup><sup>η</sup> <sup>+</sup> <sup>4</sup>η<sup>2</sup> <sup>−</sup> *Rj*. And, finally, the generated internal stress can be obtained from the microcantilever static deflection measurement [7]:

$$\sigma = \frac{\left[ (1 - \nu\_2) + (1 - \nu\_1)\xi\_F \eta^3 \right] E\_1 T\_1}{3(1 - \nu\_2)(1 - \nu\_1)(1 + \eta)\eta L^2} z\_\prime \tag{22}$$

where *z* is the detected microcantilever static deflection caused by stress due to film coating. Equation (22) is obtained using a plate approximation and without accounting for the clamped end effect. Hence, this equation is strictly valid for microcantilevers with a large aspect ratio *L*/*W* >> 5. If the aspect ratio is small, then the effect of clamping region must be accounted and the microcantilever deflection can be obtained by following approach given in work of Tamayo et al. [43]. In present work we assume *L* >> *W* >> *T* and, consequently, Equation (22) describes accurately the relationship between bending of the cantilever free end and the internal stress [44].

Equations (19) and (20) reveal that the Young's and shear moduli of prepared film can be determined even without a requirement for knowing its density and vice versa. Similarly, the Poisson's ratio of ultrathin film calculated by Equation (21) does not require previous calculation of the Young's and shear moduli of film and substrate. These findings are particularly of value in testing of micro-/nano-electronic devices, where in order to prevent their mechanical failure, it is of emergent importance to know the material properties of designed ultrathin films [45]. It shall be pointed out that the present method can be also extended to determine, in addition to elastic properties and density, the ultrathin film thickness. In this case, the rarely-measured in-plane flexural resonant frequencies must be taken into account [24]. Then, for in-plane flexural mode *r*(ξ*F*,η) = 1 + ξ*F*η and, consequently, the density, elastic properties and thickness of prepared film can be determined using Equations (19) and (20).

#### *3.2. Impact of Errors in Dimensions, Frequency, and Quality Factor Measurements on the Accuracy of the Present Method*

To ensure the proposed procedure of material properties determination is practical, we now examine impact of the dimensional discrepancy and uncertainties in frequency measurements on the determined material properties. It is worth noting that thickness of film can be measured by the ellipsometry with a typical measurement error of sub-nanometer, whereas the cantilever length and width are often determined by a scanning electron microscope with the common uncertainties ranging from few nm to tens of nm. In general, the uncertainties in dimensions, frequency, and *Q*-factor yield inaccuracies in the determined properties of designed ultrathin films. These inaccuracies expressed through errors in the dimensionless thickness, Δη, density, Δμ, modulus parameters, Δξ*i*, and the Poisson's ratio, Δυ, can be viewed as a perturbed term in a given quantity. The relative errors calculated from Equations (19)–(21) for η << 1 (i.e., *r*(ξ*j,* η) ≈ [4ξ*j*η(1+ 1.5η) + 1]/(1 + ξ*j*η)) read:

$$\frac{\Delta\mu}{\mu} = \frac{1}{1 + \frac{\Delta\eta}{\eta}} - 1\_{\prime} \tag{23}$$

$$\frac{\Delta\xi\_i}{\xi\_i} = \frac{1}{1 + \frac{6\Delta\eta}{6\eta + 4 - R\_i^\*} + \frac{\Delta\eta}{\eta}} - 1\_\prime \tag{24}$$

$$\frac{\Delta\overline{\upsilon}}{\overline{\upsilon}} = \frac{1 + 4\Delta\eta (1.5 + 2\eta) / B\_T^\*}{1 + 4\Delta\eta (1.5 + 2\eta) / B\_F^\*} - 1,\tag{25}$$

where υ = (1 + υ2)/(1 + υ1), *R*<sup>∗</sup> *<sup>i</sup>* and *B*<sup>∗</sup> *<sup>i</sup>* are the measured and calculated properties with due account for uncertainties in the frequency, *Q*-factor, and dimensions. The achievable relative sensitivity in determined material properties and density of gold film sputtered on the silicon substrate, with dimensions 300 μm (*L*), 30 μm (*W*), and 1 μm (*T*1), are given in Figure 4. For example, for 40 nm thick gold film and the uncertainties in frequency, thickness, and width measurements of 0.5 kHz, 1 and 10 nm, the following properties of gold film are obtained: The Young's modulus of 78.9 ± 3.3 GPa, the shear modulus of 27.2 <sup>±</sup> 0.4 GPa, the density of 20.3 <sup>±</sup> 1.2 g/cm3, and the Poisson's ratio of 0.45 ± 0.1. These results demonstrate that the present procedure of material properties measurement is accurate, even for the relatively high uncertainties in thickness and resonant frequencies (*Q*-factors) measurements. Importantly, for given measurement errors, the accuracy in determined material properties can be easily improved just by detecting changes in the resonant frequency and *Q*-factor of

the higher vibrational modes (i.e., higher resonant frequencies yield a significant increase in *Q*-factor values) [38]. For an illustration, we again extract the material properties of 40 nm thick gold film with errors in measurements given in the previous example by considering the second vibrational mode. The calculated properties of gold film using the second vibrational modes are as follows: The Young's modulus of 78.8 <sup>±</sup> 1.4 GPa, the shear modulus of 27.2 <sup>±</sup> 0.2 GPa, the density of 20.1 <sup>±</sup> 0.9 g/cm3, and the Poisson's ratio of 0.43 ± 0.1. Different values of film's Young's modulus, density and the Poisson's ratio mainly originate from the uncertainties in calculated Γ*j*\_*<sup>r</sup>* ω(*n*) *j* and <sup>Γ</sup>*j*\_*im* ω(*n*) *j* [34,39].

**Figure 4.** The achievable relative sensitivity of (**a**) the Young's modulus (Equation (24)); (**b**) the shear modulus (Equation (24)); (**c**) density (Equation (23)); and (**d**) the Poisson's ratio (Equation (25)) of gold film of η = 0.01, 0.04, and 0.1 coated on the silicon substrate with dimensions 300 μm (*L*), 30 μm (*W*), and 1 μm (*T*1).

#### **4. Discussion**

We now assess the validity and versatility of the proposed procedure of material characterization by extracting the Young's modulus of the atomic-layer-deposited TiO2 ultrathin film, of thickness 20 and 50 nm, sputtered on the microcantilever made of SU-8 photoresist polymer substrate [46]. In contrast to data presented in [46], where the density of TiO2 film was estimated based on the X-ray reflectometry measurements and, then, the Young's modulus of TiO2 film was calculated from changes in the resonant frequencies before and after conformal coating of the film, we determine the TiO2 film Young's modulus without requirement for the density measurement. We remind the reader that the effective linear density of microcantilever vibrating in air can be expressed through measured *Q*-factor values and, consequently, the simple flexural resonant frequencies of the *n*-th vibrational mode can be accurately predicted by:

$$f\_{F0}^{(n)} = \frac{\mathcal{V}\_{(n)}^2}{2\pi L^2} \sqrt{\frac{1}{\kappa\_F \rho\_{\text{air}} \, \text{W}^2} \frac{D\_F}{Q\_F^{(n)} \Gamma\_{F, \text{im}}(\omega)}},\tag{26}$$

where *f* = ω/(2π) and the flexural rigidity of a microcantilever with conformal coating (i.e., the TiO2 film covers entire surface area of the microcantilever) is given by:

$$D\_F = \frac{1}{12} E\_1 W T\_1^3 + \frac{1}{6} E\_2 \left[ T\_1^3 T\_2 + T\_2^3 (\mathcal{W} + 2T\_2) + 3T\_2 (\mathcal{W} + 2T\_2) (T\_1 + T\_2)^2 \right]. \tag{27}$$

For a microcantilever of length, width, and substrate thickness of 300 ± 1 μm, 100 ± 1 μm, and 5.6 <sup>±</sup> 0.05 <sup>μ</sup>m, substrate density of 1.2 <sup>±</sup> 0.01 g/cm3, and the measured fundamental resonant frequency in air of *f* (1) *<sup>F</sup>*<sup>0</sup> = 18.2 ± 0.32 kHz, the Young's modulus of SU-8 substrate of 4.04 ± 0.3 GPa is estimated by Equation (26). Importantly, published experimental data from [46] shows that, in accordance with present theoretical predictions (see Figure 3a,c), coated TiO2 film shifts the microcantilever resonant frequencies to the higher values (i.e., for TiO2 film, stiffness dominates and causes an increase in *Q*-factor). Accounting for the flexural rigidity given by Equation (27) and structure of Equation (20), the expression enabling calculation of the Young's modulus of TiO2 film from the experimentally-detected changes in frequency response reads:

$$\xi\_F = \frac{1}{2r\_0} \left[ \left( \frac{f\_F^{(n)}}{f\_{F0}^{(n)}} \right)^2 \frac{Q\_F^{(n)}}{Q\_{F0}^{(n)}} \frac{\Gamma\_{F\text{-}in}\left(\omega\_F^{(n)}\right)}{\Gamma\_{F\text{-}in}\left(\omega\_{F0}^{(n)}\right)} - 1 \right],\tag{28}$$

where *r*<sup>0</sup> = ε<sup>2</sup> + η3(1 + 2ε2) + 3η(1 + 2ε2)(1 + η) <sup>2</sup> and ε<sup>2</sup> = *T*2/*W*. Table 1 presents a comparison of the Young's modulus of 20 and 50 nm thick TiO2 films calculated using Equation (28) and determined previously by Colombi et al. [46].

**Table 1.** Comparison of the Young's modulus of atomic layer deposition (ALD) TiO2 films calculated by Equation (28) and determined previously in [46].


In addition, comparisons of the shear moduli and densities of 20 and 50 nm thick silicon nitride (Si3N4) films, coated on the silicon microcantilevers of length 300 μm, width 30 μm, and thickness of 1 μm, calculated by the present method (i.e., Equations (19) and (20)) and numerically by using COMSOL Multiphysics, are given in Table 2. The uncertainties in frequency, thickness, and width measurements are: 0.5 kHz, 1, and 10 nm, respectively.

**Table 2.** Comparisons of the shear modulus and density of silicon nitride film coated on the silicon microcantilever obtained by the proposed procedure and numerically. Considered errors in frequency, thickness, and width measurements are: 0.5 kHz, 1, and 10 nm, respectively.


The results given in Tables 1 and 2 show that here derived expressions enabling calculation of the ultrathin film material properties and density are valid and, in addition, the proposed procedure of material properties determination is relatively simple, practical, universal, and accurate, even for low accuracies in dimensions and frequency measurements.

#### **5. Conclusions**

We have proposed and demonstrated the non-destructive and easily accessible method of material characterization utilizing the well-established measurement of static and dynamic modes of the microcantilevers operating in air. Expressions needed to calculate the film properties from measured frequency and *Q*-factor changes are derived. We have showed that by monitoring changes in resonant frequency and, correspondingly, *Q*-factor, the Young's (shear) modulus of film can be determined without the requirement of knowing the film density. This finding would be of great value in material testing of ultrathin films of which the density deviates from known bulk values. The usual discrepancies in dimensions and errors in frequency (*Q*-factor) measurements were proven to have only a small impact on the calculated material properties of ultrathin film. In addition, for given errors in dimensions and frequency measurements, the accuracy in extracted film properties can be easily improved by using the higher vibrational modes. A good agreement between the Young's modulus (the shear modulus and density) determined by the present procedure and previous experimental measurements (numerical computations) carried out on the microcantilever, consisting of an elastic substrate and coated ultrathin film(s), has allowed us to confirm the validity of: (a) derived expressions and (b) the present procedure of the ultrathin film material properties determination.

**Author Contributions:** I.S. proposed the topic, developed the models, and wrote the manuscript. I.S. and L.G. performed the systematic investigations and analysis of data.

**Funding:** This work was supported by Harbin Institute of Technology, Shenzhen, Shenzhen, China, the project of Czech Republic SOLID 21 (CZ.02.1.01/0.0/0.0/16\_019/0000760) and Drážní revize s.r.o.

**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/).
