**Time Evolution Characterization of Atmospheric-Pressure Plasma Jet (APPJ)-Synthesized Pt-SnO***x* **Catalysts**

#### **Chia-Chun Lee 1, Tzu-Ming Huang 1,2, I-Chun Cheng 3,4,\*, Cheng-Che Hsu 5,\* and Jian-Zhang Chen 1,2,\***


Received: 17 August 2018; Accepted: 29 August 2018; Published: 1 September 2018

**Abstract:** We characterize the time evolution (≤120 s) of atmospheric-pressure plasma jet (APPJ)-synthesized Pt-SnO*<sup>x</sup>* catalysts. A mixture precursor solution consisting of chloroplatinic acid and tin(II) chloride is spin-coated on fluorine-doped tin oxide (FTO) glass substrates, following which APPJ is used for converting the spin-coated precursors. X-ray photoelectron spectroscopy (XPS) indicates the conversion of a large portion of metallic Pt and a small portion of metallic Sn (most Sn is in oxidation states) from the precursors with 120 s APPJ processing. The dye-sensitized solar cell (DSSC) efficiency with APPJ-synthesized Pt-SnO*x* CEs is improved greatly with only 5 s of APPJ processing. Electrochemical impedance spectroscopy (EIS) and Tafel experiments confirm the catalytic activities of Pt-SnO*<sup>x</sup>* catalysts. The DSSC performance can be improved with a short APPJ processing time, suggesting that a DC-pulse nitrogen APPJ can be an efficient tool for rapidly synthesizing catalytic Pt-SnO*x* counter electrodes (CEs) for DSSCs.

**Keywords:** atmospheric pressure plasma jet; platinum; tin oxide; dye-sensitized solar cells; chloroplatinic acid; tin chloride

#### **1. Introduction**

Atmospheric-pressure plasma (APP) technology is operated without using a vacuum chamber and associated pumping system. It is therefore considered a cost-effective manufacturing tool. Recent developments have resolved stability and arcing problems, making APP technology promising for industrial applications. Traditional APP sources include transferred arc, corona discharge, dielectric barrier discharge, and atmospheric pressure plasma jet (APPJs) [1,2]. APPs with various heavy particle temperatures and charge densities can be produced by using different excitation methods and electrode configuration designs. The synergy between the reactive plasma species and heat can promote rapid chemical reactions during material processing [3–6]. APPs have been used for processing various types of materials, such as carbon nanotubes [3,7,8] and reduced graphene oxides [9–11]. Applications of APPs for surface cleaning or modification [12–14], deposition of metal oxides [15,16], and syntheses of metal compounds from liquid precursors [6,17,18] have been extensively investigated. Metals and metal oxides are common catalysts [19–24]. APPs also have been used for syntheses and post-treatments of catalysts [25].

In 1991, Grätzel et al. reported a great breakthrough of DSSCs [26], and since then, dye-sensitized solar cells (DSSCs) have been extensively investigated. A conventional DSSC consists of a dye-adsorbed photoanode, an electrolyte, and a counter electrode (CE). A catalytic CE is used for reducing triiodide into iodine in the electrolyte. Generally, Pt is the most commonly used CE material in DSSC, owing to its high catalytic activity and stability [27]. Various alternative CE materials such as carbon-based materials, metal oxides or chalcogenides, and alloys or intermetallics have been studied extensively [3,5,28–36]. Composites containing Pt and Sn have been used as electrocatalysts for CEs of DSSCs [36,37], methanol or ethanol oxidation [38–44], aqueous phase oxidation [45], and gas sensing [46]. The addition of metal oxides has been reported to improve the catalytic activity [40,47]. Pt:SnO2 electrocatalytic films were used as CEs of DSSCs [48]. Dao et al. fabricated DSSCs with a PtSn alloy supported by reduced graphene oxides via dry plasma reduction [36]. In the present study, Pt-SnO*<sup>x</sup>* composites were synthesized by mixing chloroplatinic acid and tin(II) chloride that were processed using a DC-pulse nitrogen APPJ. X-ray photoelectron spectroscopy (XPS) results showed that the majority of Sn was in the oxidation state. The DSSC efficiency can be improved rapidly through 5 s APPJ processing of the chloroplatinic acid and tin(II) chloride mixture precursor; no metallic Pt was converted within such a short processing time. This suggests the catalytic effect of oxidized Pt and Sn compounds. A DSSC with a 120 s APPJ-processed Pt-SnO*x* CE shows efficiency comparable to that of a cell with a furnace-processed Pt CE.

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

#### *2.1. Preparation of Pt-SnOx CEs*

25-mM chloroplatinic acid (H2PtCl6) (purity: 99.95%, Uniregion Biotech, Taipei, Taiwan) and 25-mM tin(II) chloride (SnCl2) isopropanol solutions were separately stirred for 24 h. These two solutions were mixed with the same volume ratios and were stirred using a magnetic stirrer (PC-420D, Corning Inc., Corning, NY, USA)for another 24 h. Next, 60 μL of the mixture precursor was spin-coated onto fluorine-doped tin oxide (FTO) substrates with an area of 1.5 cm × 1.5 cm at a speed of 1000 rpm for 15 s. The spin-coated precursors were then processed by a nitrogen APPJ for 5, 15, 30, 60, and 120 s. Figure 1a shows the APPJ setup. The operation parameters are as follows: nitrogen flow of 46 standard liter per minute (slm), power supply voltage of 275 V, and ON/OFF duty cycle of 7/33 μs. The temperature evolution of the substrates, shown in Figure 1b, was measured using a K-type thermocouple (OMEGA Engineering, Norwalk, CT, USA). The temperature rapidly increased to ~510 ◦C, and it dramatically decreased after the APPJ was turned off. Because our process is conducted at ~510 ◦C, we use FTO glass substrates (Sigma-Aldrich, St. Louis, MO, USA) which can tolerate a higher processing temperature.

**Figure 1.** (**a**) Schematic of APPJ setup; (**b**) Temperature evolution of substrate during APPJ treatment.

#### *2.2. Preparation of TiO2 Photoanode and Assembly of DSSCs*

The photoanode consists of a TiO2 compact layer and a TiO2 nanoporous layer for dye adsorption. First, a 0.23-M titanium isopropoxide solution (Fluka, St. Louis, MO, USA) was spin-coated on a FTO substrate and then baked at 200 ◦C for 10 min to form a TiO2 compact layer to prevent electron recombination. Then, 1.6 g of TiO2 nanoparticles (diameter: ~21 nm), 8 mL of ethanol, 6.49 g of terpineol (anhydrous, #86480, Fluka, St. Louis, MO, USA), 4.5 g of 10 wt % ethyl cellulose ethanolic solution (5–15 mPa·s, #46070, Fluka, St. Louis, MO, USA), and 3.5 g of 10 wt % ethyl cellulose ethanolic solution (30–50 mPa·s, #46080, Fluka, St. Louis, MO, USA) were mixed together. Next, a 0.4 g mixture containing TiO2 was mixed with 500 μL of ethanol and stirred using a magnetic stirrer for 24 h. The mixed solution was baked at 53 ◦C until its weight became 0.175 g, thus completing the preparation of the TiO2 pastes. The TiO2 pastes were screen-printed onto the TiO2 compact layer coated FTO substrate with a printed area of 0.5 cm × 0.5 cm. The screen-printed pastes were calcined at 510 ◦C for 15 min in a conventional furnace to form the TiO2 photoanode. Next, the TiO2 photoanode was immersed in a 0.3-mM N719 solution, which is mixed with acetonitrile and tertbutyl alcohol in a 1:1 volume ratio for 24 h. This completed the preparation of the dye-anchored nanoporous TiO2 photoanodes.

The Pt-SnO*<sup>x</sup>* CEs and dye-anchored TiO2 photoanodes were assembled with a 25-μm-thick spacer to form sandwich-structure DSSCs. Then, a commercial electrolyte (E-Solar EL 200, Everlight Chemical Industrial Co., Taipei, Taiwan) was injected into the solar cells.

Counterpart DSSC with furnace-processed Pt CE was fabricated for comparison. In this case, 60 μL of 25-mM H2PtCl6 isopropanol solution was spin-coated on the FTO substrate and calcined at 400 ◦C for 15 min using a tube furnace. The assembly procedure of DSSC with furnace-processed Pt CE is the same as that of DSSC with APPJ-processed Pt-SnO*x* CE.

#### *2.3. Characterization of Materials and DSSCs*

During the APPJ reduction processes, a spectrometer (USB4000, Ocean Optics, Largo, FL, USA) was used for monitoring the plasma optical emission spectra (OES). Pt-SnO*<sup>x</sup>* nanoparticles were inspected using a scanning electron microscope (SEM, JSM-7800F Prime, JEOL, Tokyo, Japan) with an energy-dispersive spectroscopy (EDS) attachment. To investigate the chemical configuration of Pt-SnO*x*, XPS (Thermo K-Alpha, VGS, Waltham, MA, USA was used for analyzing the binding status. The C1s core level was centered at 284.6 eV to calibrate the binding energy scale. XPSPEAK 4.1 software (was used for fitting binding energy positions. XPS samples were prepared with Corning glass substrates instead of FTO glass ones to avoid the interference of Sn signals emitted from FTO substrates. To examine the electrochemical catalytic activities of Pt-SnO*<sup>x</sup>* CEs, electrochemical impedance spectroscopy (EIS) and Tafel measurements were performed using an electrochemical workstation (PGSTAT204, Metrohm Autolab, Herisau, Switzerland). EIS measurements were performed with a sinusoidal amplitude of 10 mV with frequencies of 0.1-10<sup>5</sup> Hz, and the data were fitted using Z-view 3.1 software. Tafel curves were recorded from −0.6 V to 0.6 V at a scan rate of 50 mV/s. Both measurements were performed on a symmetrical cell with two equal Pt-SnO*x* CEs. A solar simulator (WXS-155S-L2, WACOM, Saitama, Japan) with an AM 1.5 filter equipped with an electrometer (Keithley 2440, Tektronix, Beaverton, OR, USA) was used for measuring the photocurrent-voltage characteristics of the DSSCs.

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

Figure 2a shows the plasma OES evolution during APPJ processing of the mixed H2PtCl6/SnCl2 precursor. NOγ, NOβ, N2 1st positive, and N2 2nd positive emissions were observed clearly during 120 s APPJ processes. Figure 2b shows the plasma spectra when processing H2PtCl6, SnCl2, and mixed H2PtCl6/SnCl2 precursors on the FTO substrates. The NO<sup>γ</sup> system (A2Σ+-X2Π) is located at wavelengths lower than 280 nm. The NO<sup>β</sup> system (B2Π-X2Π) is located from around 260 to 500 nm, and it partially overlaps the NOγsystem. The other emissions at 357, 385, and 389 nm were attributed to the N2 2nd positive system (C3Πu-B3Πg); these overlap with the NO<sup>β</sup> system. The N2 1st positive system (B3Πg-A3Σ<sup>u</sup> +) was located at wavelengths higher than 530 nm.

**Figure 2.** (**a**) OES evolution during APPJ processing of mixed H2PtCl6/SnCl2 precursors. (**b**) OES when processing H2PtCl6, SnCl2, and mixed H2PtCl6/SnCl2 precursors using nitrogen APPJ.

Figure 3a–e shows the SEM images of Pt-SnO*<sup>x</sup>* nanoparticles converted from mixed H2PtCl6/SnCl2 precursors on the FTO glass substrates using various APPJ processing times. The nanoparticle size and morphology remained similar for APPJ processing times of 5-120 s. Figure 3f shows EDS results for the 120 s and APPJ-processed sample. Pt and Sn signals indicate the presence of two elements in the nanoparticles. Both of Sn and O signals could result from the nanoparticles and the FTO substrates.

**Figure 3.** Scanning electron microscope (SEM) images of samples processed by APPJ for various durations: (**a**) 5 s, (**b**) 15 s, (**c**) 30 s, (**d**) 60 s, and (**e**) 120 s. (**f**) Energy-dispersive spectroscopy (EDS) spectrum of nanoparticles converted from H2PtCl6/SnCl2 precursors on FTO glass substrates using 120 s APPJ processing.

*Metals* **2018**, *8*, 690

To identify the chemical states of Pt-SnO*<sup>x</sup>* compounds, Figure 4a,b shows the XPS spectra of Pt4f and Sn3d for samples. The Pt4f spectrum can be deconvoluted into three components including Pt, Pt2+, and Pt4+. The metallic peaks of Pt are located at 71.30 and 74.65 eV, Pt(II) components are located at 72.70 and 76.50 eV, and Pt(IV) components are located at 73.80 and 77.15 eV [49,50]. In Figure 4a, the major peaks belong to Pt2+ and Pt4+ for as-deposited and 5 and 15 s APPJ-processed samples. These results indicate that most of the H2PtCl6/SnCl2 precursor was not converted to metallic Pt by APPJ processing for less than 15 s. As the APPJ processing time increases, increased conversion of precursors into metallic Pt was clearly observed. The Pt2+ signal is noted as the oxidation state of Pt, and it could indicate PtO [51,52] or Pt(OH)2 [53]. The presence of Pt oxidation states, due to the interaction with the Pt-support, is attributed to an electronic effect or oxygen absorption from air [54,55]. Figure 4b shows the oxidation state of Sn3d under various APPJ processing times. The binding energy of Sn3d can be deconvoluted into two categories: one at 485.8 and 494.2 eV for the zero-valent state of Sn, and the other at 487.3 and 495.7 eV for Sn(II/IV) components [56]. The major peak is attributed to the oxidation state of Sn for up to 120 s, and the percentage of metallic Sn increased only slightly increased with the APPJ processing time. Sn(II) and Sn(IV) species are difficult to distinguish from XPS measurements because of the small difference between their binding energies [57,58]. Tables 1 and 2 show the percentages of Pt and Sn species, respectively. The Pt-support interaction may influence charge transfer from Pt to oxygen species on the surface and improve the electrochemical catalytic abilities and catalyst stability [47].

**Figure 4.** X-ray photoelectron spectroscopy (XPS) spectra of (**a**) Pt4f and (**b**) Sn3d for samples processed by APPJ for various durations.


**Table 1.** Percentage of Pt species obtained from XPS analysis.



Figure 5a,b shows the EIS Nyquist and Bode phase plots to evaluate the catalytic activities of APPJ-processed Pt-SnO*<sup>x</sup>* CEs. The inset of Figure 5a shows the equivalent circuit for Nyquist curve fitting [59]. The series resistance (*R*s) and charge-transfer resistance (*R*ct) can be described as the resistance of substrates and the catalytic effect of the electrode-reducing triiodide ions, respectively. *R*<sup>s</sup> can be obtained from the high-frequency intercept on the real axis and *R*ct, from the radius of the real semi-circle [60]. Table 3 shows the EIS parameters including *R*s, *R*ct, and constant phase element (CPE1) [29]. A higher catalytic effect and lower charge-transfer resistance would improve the DSSC performance. For all cases, *R*<sup>s</sup> of Pt-SnO*<sup>x</sup>* CEs remained similar. *R*ct generally decreased (i.e., semi-circle became smaller) as the APPJ processing time increased, indicating that APPJ processing can enhance the catalytic activity. *R*ct was comparable for APPJ processing times of 60 s (4.72 Ω) and 120 s (4.69 Ω). Lower *R*ct results in a higher electrocatalytic activity at the interface between the CEs and the electrolytes [61]. CPE1, which represents the interfacial capacitance between the electrode and the electrolyte, is also a good indicator of the surface activity of CEs [62–64]. The 120 s APPJ-processed CEs had a higher CPE1-T (105.5 μF/cm2), indicating larger surface reaction between the CE and the electrolyte. Bode phase plots show the electron lifetime for recombination in devices; the electron lifetime is expressed as τ<sup>e</sup> = 1/(2π*f* peak), where *f* peak is the frequency of the highest peak. Shorter electron lifetime indicates faster charge transfer at the interface between the CE and the electrolyte [64,65]. In Figure 5b, the trend of the electron lifetime follows the EIS results. The 5 s APPJ-processed CE has the smallest peak frequency, indicating the largest electron lifetime with slower charge transfer. Furthermore, electron lifetimes are comparable in 60 s and 120 s APPJ-processed CEs, and this is consistent with the results for *R*ct.

To further clarify the catalytic activities of Pt-SnO*<sup>x</sup>* CEs, Tafel polarization experiments were conducted and the results are shown in Figure 6. The exchange current density (*J*0) was measured by the intercept of the Y-axis (zero voltage) from the tangential line of the curve [66,67]. The 120 s APPJ-processed CEs had a large *J*0, indicating better electrocatalyic activity and lower charge-transfer resistance at the interface of the CE and the electrolyte. Table 3 shows that *J*<sup>0</sup> increases with the APPJ processing time. APPJ processes enhanced the triiodide reduction reaction [60]. The exchange current density is also proportional to *R*ct obtained from the EIS measurement. It can be described as *J*<sup>0</sup> = *RT*/*nFR*ct, where *R* is a gas constant; *T* is temperature; *n* is the number of electrons involved in the redox reaction; and *F* is the Faraday's constant [68]. EIS and Tafel measurements both indicate that APPJ-processed Pt-SnO*<sup>x</sup>* elecrodes show suitable catalytic performance for use as the CEs of DSSCs. *J*<sup>0</sup> increases with the APPJ processing time, indicating that APPJ processing can enhance the catalyst activity of Pt-SnO*x*.

**Figure 5.** (**a**) Nyquist curves of symmetric cells with two Pt-SnO*x* CEs. The inset shows the equivalent circuit diagram. (**b**) Bode phase plots of symmetric cells with two Pt-SnO*x* CEs.


**Table 3.** EIS parameters of Pt-SnO*x* CEs.

<sup>a</sup> *J*0: Exchange current density is calculated from *R*ct. <sup>b</sup> *J*0: Exchange current density is calculated from Tafel curve.

**Figure 6.** Tafel curves of symmetric cells with various Pt-SnO*x* CEs.

Figure 7 shows the IV curves of DSSCs with APPJ-processed Pt-SnO*<sup>x</sup>* CEs. Table 4 shows the photovoltaic parameters, including the open-circuit voltage (*V*oc), short-circuit current (*J*sc), fill factor (FF), and efficiency (EFF) with their standard deviations. The power conversion efficiencies (PCEs) of DSSCs with 5 s and 15 s APPJ-processed Pt-SnO*<sup>x</sup>* CEs are 3.87 ± 0.58% and 3.86 ± 0.28%, respectively, indicating that APPJ processing for a short duration can improve the DSSC performance. XPS results show that almost no metallic Pt was converted with 5 s and 15 s APPJ processing, indicating the catalytic effect of oxidized Pt and Sn compound CEs in DSSCs, and this agrees with previous reported findings [30,32]. As the APPJ treatment time increases, the PCE of DSSCs with 30 s, 60 s, and 120 s APPJ-processed CEs reaches 4.01 ± 0.34%, 4.20 ± 0.41%, and 4.46 ± 0.29%, respectively. The performance of DSSC with a 120 s APPJ-processed Pt-SnO*<sup>x</sup>* CE was comparable to that with a

conventional furnace-processed Pt CE (4.42 ± 0.26%). Figure 8 shows the statistics of the DSSC parameters. APPJ processing gradually increased the FFs and PCEs of DSSCs, consistent with the results obtained from EIS and Tafel measurement. The improved FF and efficiency with APPJ processing time could result from the better conversion of metallic Pt from the precursor solution.

**Figure 7.** Photocurrent density-voltage curves of DSSCs with various CEs.


**Figure 8.** Statistics of DSSC parameters based on APPJ-processed Pt-SnO*x* CEs and furnace-processed Pt CE (reference).

#### **4. Conclusions**

We analyze the time evolution of Pt-SnO*<sup>x</sup>* nanoparticle catalysts that are converted from a mixture of chloroplatinic acid and tin(II) chloride using DC-pulse nitrogen APPJ. XPS analyses indicate the conversion of a large portion of the metallic Pt and tin oxide. EIS and Tafel measurements indicate improved electrochemical catalytic effects. The synthesized Pt-SnO*<sup>x</sup>* nanoparticles on FTO glass substrates are used as the CEs of DSSCs. The I-V curve shows that the performance of DSSCs with APPJ-processed Pt-SnO*x* CEs is comparable to that of DSSCs with conventional furnace-processed Pt CEs. As the APPJ processing time is increased, the FF and efficiency of DSSCs gradually increase. Our results show that a DC-pulse nitrogen APPJ is an efficient tool for synthesizing Pt-SnOx catalysts from a mixture precursor solution consisting of chloroplatinic acid and tin(II) chloride.

**Author Contributions:** C.-C.L. performed the experiments, analyzed the data, and wrote the paper draft. T.-M.H. assisted in conducting experiments. I.-C.C. and C.-C.H. assisted in instructing the research and analyzing the data. J.-Z.C. directed the research direction, analyzed the data, and revised the paper. All authors commented on the manuscript.

**Funding:** This work is supported by the "Advanced Research Center for Green Materials Science and Technology" from The Featured Area Research Center Program of the Higher Education Sprout Project by the Ministry of Education (107L9006) and the Ministry of Science and Technology in Taiwan (MOST 105-2221-E-002-047-MY3, MOST 106-2221-E-002-193-MY2 & MOST 107-3017-F-002-001).

**Acknowledgments:** The cleanroom facility is provided by the Nano-Electro-Mechanical-Systems (NEMS) Research Center at National Taiwan University. Yuan-Tzu Lee of the Instrumentation Center at National Taiwan University helps with the SEM operation.

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

#### **References**


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

## *Article* **Mechanical and Microstructural Features of Plasma Cut Edges in a 15 mm Thick S460M Steel Plate**

**Javier Aldazabal 1,\*, Antonio Martín-Meizoso 1, Andrzej Klimpel 2, Adam Bannister <sup>3</sup> and Sergio Cicero <sup>4</sup>**


Received: 4 May 2018; Accepted: 8 June 2018; Published: 11 June 2018

**Abstract:** In general, the thermal cutting processes of steel plates are considered to have an influence on microstructures and residual stress distribution, which determines the mechanical properties and performance of cut edges. They also affect the quality of the surface cut edges, which further complicates the problem, because in most cases the surface is subjected to the largest stresses. This paper studies the influence of plasma cutting processes on the mechanical behavior of the cut edges of steel and presents the characterization results of straight plasma arc cut edges of steel plate grade S460M, 15 mm thick. The cutting conditions used are the standard ones for industrial plasma cutting. The metallography of CHAZ (Cut Heat Affected Zones) and hardness profiles versus distance from plasma cut edge surface are tested; the mechanical behavior of different CHAZ layers under the cut edge surface were obtained by testing of instrumented mini-tensile 300 μm thick specimens. Also, the residual stress distribution in the CHAZ was measured by X-ray diffraction. The results for the mechanical properties, microstructure, hardness, and residual stresses are finally compared and discussed. This work concludes that the CHAZ resulting from the plasma cutting process is narrow (about 700 μm) and homogeneous in plate thickness.

**Keywords:** plasma cutting; cut heat affected zone; mini-tensile test; steel plate; residual stress

#### **1. Introduction**

The use of steel plates in construction elements, structures, parts of machinery, etc. requires, in practically all cases, the cutting of these metallic sheets into smaller parts. These parts will later be connected to other elements using mechanical joints or welds. Nowadays we have at our disposal a plethora of cutting techniques: shear, oxy-cut, laser, plasma, water jet (with or without abrasive particles), thermal lance, etc., but all these cutting techniques introduce modifications in the regions close to the cut surface: they modify their surface roughness and, when they provide enough heat, they introduce modifications in the microstructure [1]. These changes result in local variations in the mechanical properties; in many cases, they also introduce or modify the profiles of the residual stresses in the areas close to the cutting surface [2–5]. A question arises here as to whether it is preferable to leave the cutting edge as it is or, on the contrary, if it is preferable to eliminate it, for example, by grinding (as specified or recommended in some construction standards [6,7]) to optimize the use of the cut pieces and, in particular, their subsequent performance in applications under alternating loads (fatigue) [8–11].

Criteria to establish whether it is better to keep the original edge or to remove it was one of the objectives proposed within the European project HIPERCUT ("High Performance Cut Edges in Structural Steel Plates for Demanding Applications", RSFR-CT-2012-00027 [12]). In this project, the results obtained using different cutting techniques were analyzed. Steel plates with thicknesses ranging from 8 mm up to 25 mm were studied, while the grades and mechanical strength of the plates varied between S355M and S890Q [13–16]. The cutting techniques that were analyzed and compared were plasma jet, laser beam, and oxyfuel (oxy-acetylene). This article presents the results obtained with the plasma cutting technique in a 15 mm S460M structural steel plate.

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

The characterization of the cutting edge obtained in a 15 mm thick steel plate with nominal yield stress of 460 MPa is summarized. The chemical composition of the steel was as follows (wt. %): 0.12 C, 0.45 Si, 1.49 Mn, 0.012 P, 0.001 S, 0.062 Cr, 0.001 Mo, 0.016 Ni, 0.048 Al, 0.011 Cu, 0.036 Nb, 0.005 N, 0.002 Sn, 0.003 Ti, 0.066 V. This plate was cut with a plasma jet, in the standard industrial conditions for the cutting of this thickness. The cutting parameters for the cutting equipment used (Hypertherm 260) and for the referred thickness were: Plasma arc current 200 A, arc voltage 131 V, cutting speed 2200 mm/min., torch standoff 4.1 mm, O2 plasma gas flow rate 69 L/min. and shielding gas flow rate 28 L/min. Here, it should be noted that the objective of the research is not to optimize the cutting parameters, but to determine how well-defined industrial parameters affect the cut edge properties and characteristics.

After the cutting process, samples were obtained (from the cut edge area) for metallographic study, hardness measurements and machining of mini-specimens for tensile tests. The metallographic samples were fixed in a conductive acrylic resin (Condufast™, Struers, Cleveland, OH, USA). Then, the surface under observation was polished with SiC papers up to grade 1200 and finally polished with 0.6 μm diamond paste, on velvet, until a specular finish is achieved. The polished samples were etched with 2% Nital for 15 s, rinsed with ethyl alcohol and dried under a hot air stream, before being observed in an optical microscope (Leica MEF-4, Leica, Wetzlar, Germany).

The hardness profiles were made using a LECO hardness tester (model M-400-G2, Leco, Saint Joseph, MI, USA) equipped with a Vickers pyramidal tip. The indentations were carried out with a load of 4.93 N (0.5 kg).

The mini-tensile samples were cut from the plate using a wire electro-discharge machine (WEDM). Four dog bone shaped prisms were cut from the surface of the cut edge, with their longitudinal axes in the direction of the cut. These prisms were in the mid plane of the plate thickness (or the cut edge), as shown in Figure 1.

**Figure 1.** (**a**) General geometry of the cut plate including the orientation of tensile specimens, zones where hardness measurements were made, and directions used for measuring stresses. (**b**) Extraction of four bone shaped blocks (for later slicing and extraction of the mini-tensile samples), by electro-erosion of the central zone of the cut edge. Scale shown is in millimeters.

These four prisms were sliced into 300 μm-thick specimens. The distance between two consecutive specimens obtained from the same block was also 300 μm, due to the material removed by the wire during the machining process (see Figure 2). In other words, between two consecutive mini-tensile specimens obtained from the same block, there were 300 μm of material that were lost during the machining process that, consequently, could not be characterized. To better characterize the tensile properties all along the depth from the cut edge, the initial cut of each block or prism was moved (in depth) 150 μm from one another. With this shift between blocks, it was possible to obtain mini-tensile samples each of 150 μm in depth (or distance to the cutting edge).

**Figure 2.** Slicing of one of the prisms to obtain tensile mini-samples.

As can be seen from Figures 1 and 2, the mini-tensile samples have a longitudinal orientation (same as the cut edge) and their faces are parallel to the cut edge.

According to the bibliography [17], wire EDM cutting (WEDM) introduces residual stresses up to a depth of approximately 80 μm. To eliminate, as much as possible, the effects of the WEDM, 50 μm were eliminated on each side of the mini-tensile samples by polishing (with SiC grade 1200 sandpaper and a subsequent polishing with 1 μm diamond paste in a velvet cloth) [18]. To remove the material, an automatic Struers polishing machine (Struers, Cleveland, OH, USA) was used. During the polishing, the orientation of the samples was fixed to provide a longitudinal polishing pattern, parallel to the future direction of loading. The final thickness of the mini-tensile samples was nominally 200 μm (the real thickness was measured and recorded for each sample. Average sample thickness was 203 μm with a standard deviation of 25 μm).

Exceptionally, in the sample containing the surface of the cutting edge, this surface was not removed by polishing, but it was preserved. This first mini-tensile specimen was polished only on the inner side (polishing was carried out removing 100 μm on the inner side, so that the final thickness was also approximately 177 μm).

Figure 3 shows a mini-tensile sample. Its nominal dimensions are 20 mm in length, 5 mm in total width (the gauge length having 2.5 mm width and 3 mm length) and 0.2 mm in thickness.

To obtain accurate strain measurements, each specimen was instrumented with a strain gauge (HBM 1-LY11-3/120, HBM, Darmstadt, Germany, with 5% maximum strain), as shown in Figure 3. A San-Ei (San-Ei Electric Co., Ltd., Osaka, Japan) amplifier is used to record the lengthening of the strain gauge. The tensile tests were carried out in a universal testing machine with a crosshead displacement speed of 0.1 mm/minute. For deformations greater than 5%, the crosshead position records were used (based on the correlation between the position of the actuator and the previous measurements of the strain gauge). The strain gauge stiffness is not negligible when compared to that of the specimen, and it was measured in an independent test. Thus, the contribution of the strain gauge to the force measured by the load cell was considered when defining the actual load applied to the mini-tensile specimens during tests.

**Figure 3.** Mini-tensile specimen with the strain gauge for measuring deformations up to 5%. Scale shown is in millimeters.

The tensile tests were carried out in an electro-mechanical machine (Instron Mini 44, Instron, High Wycombe, UK). This test machine is equipped with a load cell of ±500 N. For fixing the samples, their ends were inserted in two narrow channels machined at the end of two bolts. These channels were 300 μm wide. The mini-tensile samples shoulders were fixed into the channels with a cyanoacrylate adhesive (Loctite); the capillarity of glue guarantees the complete fixation of the sample.

Finally, X-ray diffraction equipment was used to measure longitudinal (along the cutting direction) and transversal (thickness direction) residual stresses at different depths from the cut edge. Prismatic samples of 8 × 10 × 35 mm were machined (using WEDM) from half the thickness of the plate to avoid the influence of the top/bottom surfaces. These samples were cleaned in a solution of 500 mm3 of HCl and 500 mm<sup>3</sup> of distilled water, for 20 min at room temperature. The measurements were made on a X-Ray diffractometer (X'Pert, Philips, Amsterdam, The Netherlands), with the following parameters: anode material Cr (K-*α*<sup>2</sup> = 2.2936663 Å), voltage 40 kV, current 40 mA, angle 2*θ* scanning range 144.1~166.0◦ (0.3◦/step), *ψ* scan range 60.00~60.00◦ (7.76◦/step), time per step 12.05 s. The lattice equivalent planes considered for measuring the residual stresses were the {211}.

For the measurement of stresses at different depths, the cut surface material was eliminated in a controlled manner. To remove these thin layers of steel, an electrolytic polish setup was used. This technique allowed us to remove material without introducing additional stresses in the samples. The material was etched applying a potential of 14 V on blocks in an electrolytic medium made of 90% perchloric acid and 10% ethanol. This procedure (electro-polishing and X-ray diffraction) was repeated four times, until reaching a depth of 700 μm from the original cut surface. The stresses were always measured on the polished surface. The measured residual stress does not correspond to the original stress at that depth because the removal of material induces a relaxation of internal stresses. It is possible to deduce the original stresses at a certain depth, *σ*, using expression [19]:

$$
\sigma(z\_1) = \sigma\_m(z\_1) + 2 \int\_{z\_1}^H \frac{\sigma\_m(z\_1)}{z} dz - 6z\_1 \int\_{z\_1}^H \frac{\sigma\_m(z\_1)}{z^2} dz
$$

where *H* corresponds to the initial sample thickness, *z*<sup>1</sup> to the current thickness and *σ<sup>m</sup>* to the measured stress obtained after eliminating the material. An in-house developed code was used to deduce the stress profile that exists on the original plate from measurements and material thickness removed.

To validate this measurement methodology, a set of 5 tests were carried out on a pre-stressed sample. These 5 samples were stressed up to 390 MPa. Results from X-Ray measurements provided stress values of 388 ± 7 MPa.

#### **3. Results**

#### *3.1. Metallography*

Figure 4 shows the metallographic section (already etched) of the zone affected by the cut (CHAZ). In the figure, the left side corresponds to the cut edge, whereas the right side corresponds to the bulk material, or remaining plate. Moreover, the top side of the picture corresponds to the plate surface where the plasma nozzle was located, and the bottom side corresponds to the zone that rested on the cutting table (slag side). Rounding and a thickness reduction of around 1.0 mm are observed in the upper-left zone (jet entry) of the cut. The CHAZ (in principle, the darker material) is very thin, around 0.4 mm deep, especially when compared to that obtained by other cutting procedures with lower energy densities, such as oxyfuel. The latter generates a CHAZ whose depth varies between 1 mm (nozzle side) and 4 mm (slag side) [12,20]. The CHAZ provided by the plasma cut is, however, deeper than that obtained in the same material when cutting with laser, which varies between 0.1 mm (nozzle side) and 0.4 mm (slag side) [12,20]. Moreover, the depth of the CHAZ obtained when performing plasma cuts is nearly constant, whereas oxyfuel and laser cuts generate HAZs with variable depths along the plate thickness.

The heat generated during the cutting process produces phase transformations and the grain growth of the underlying matrix material, as shown in Figure 5. The microstructure shown in this figure corresponds to the area highlighted in the frame in the middle section of Figure 4. Grain size reduces drastically in areas close to the cut edge. At distances from the cut edge greater than approximately 600 μm, there are no changes in gran size.

Just below the cutting edge, layers of martensite and bainite are observed. At a depth of about 200 μm from the cutting edge, polygonal ferrite is observed. At approximately 400–500 μm, the ferrite grains are larger and beyond the 700 μm pearlite and (even larger) polygonal ferrite grains are observed; this last microstructure corresponds to the base material, not affected by the cut (a hypoeutectoid steel, with bands of perlite and ferrite).

#### *3.2. Microhardness*

Figure 6 shows the Vickers hardness profiles (0.5 kg, HV05) of the cutting edge measured at the upper part of the plate (at 0.5 mm and 2.5 mm from the plate upper surface, or nozzle side), at half the thickness plate, and at the lower part of the plate (at 0.5 mm and 2.5 mm from the plate lower surface). The measurements are presented as a function of the distance to the cutting edge (see also Figures 1a and 4 to clarify the position of the measurements). To obtain detailed hardness profiles, indentations of 0.5 kg (4.91 N) were made instead of 1 kg (9.81 N). It is known that non-standard hardness measurements can differ from standard ones but the indentations of half a kilogram are smaller and can be placed closer to one to another and to the cutting edge itself. At each height, three lines of indentations were made, with a small shift between them, with the purpose of obtaining in greater detail the evolution of hardness in the CHAZ versus depth. For each depth a single measurement was made.

It can be observed that the microhardness measurements provide very similar results in the upper, middle, and lower part of the cut section. This is related with the uniform thickness of the CHAZ, as revealed in the images included in Figure 4. The effect of plasma cutting vanishes at a distance of approximately 700 μm from the cut edge, along all the cut thickness. The European standard EN 1090-2 [21] sets a limit of 380 kg/mm<sup>2</sup> for the Vickers hardness after cutting. The microhardness measured near the surface slightly exceeds this limit in a thin layer, with a depth of about 350 μm (EN 1090-2 [21] specifies Vickers with 1 kg load and here, the results are presented with only 0.5 kg). The problem with a very hard cut surface is that it is prone to cracking in subsequent bending processes. Several bend tests of the cut edges analyzed in the HIPERCUT project were carried out (e.g., [12,22]). All samples tested were bent 180◦ without cracks, including those specimens that had the entire surface of the cutting edge on the tensile side, revealing that the observed hardness measurements, which are

slightly above the limits provided in [21], do not negatively affect the bending behavior of the plasma cut edges analyzed.

**Figure 4.** Optical micrograph of the CHAZ, ethed with 2% Nital. The Vickers indentations made for measuring hardness profiles are visible. The upper-left part corresponds to the plasma inlet (see also Figure 1a).

**Figure 5.** Detail corresponding to the framed area on Figure 4 close to the middle section. The microstructures shown in the second row correspond to the zone closer to the cut edge and its microstructure is finer in regions closer to the cut edge (left).

**Figure 6.** Hardness profiles against the distance to the cut edge. Each color corresponds to a region of the sample, blue corresponding to the one closest to the plasma nozzle and red to the region further from the plasma nozzle. Images included in the graph show the microstructure and indentations distribution used to make the plot in three regions (**top**, **middle** and **bottom**).

#### *3.3. Mini-Tensile Tests*

Figure 7 shows the (ductile) fracture observed in a tensile test, typical of a mini-tensile specimen after necking. Figure 8 summarizes the results obtained in the mini-tensile tests. The engineering stress is plotted against the (engineering) strain, as a function of the distance to the original surface produced by the plasma cut. As mentioned above, the mini-tensile specimens were obtained from four blocks extracted from the center of the thickness of the cut edge. For each depth, a single tensile test was carried out. It can be observed that the closer to the cut edge, the larger the resistance parameters and the lower the ductility. Differences in stress-strain curves tend to stabilize at a distance of approximately 750–900 μm. This is basically consistent with the microhardness measurements made at the center of the cutting edge.

**Figure 7.** Mini-tensile test specimen during tensile test. The top and bottom screws have the channel parallel to the image plane where sample shoulders were glued. The necking and later fracture can be easily observed.

**Figure 8.** Stress-strain curves obtained at different depths from the cut edge.

Figure 9 represents the change in the mechanical behavior (yield stress, Ultimate tensile strength, etc.) versus distance to plasma cut, for mini-specimens extracted from the central section of the cut. From this figure, it is clear that the material strength decreases as the distance from cut edge increases. For distances of over 800 μm there are no changes in the ultimate tensile strength, UTS, and the mechanical properties of the base material are reached. Something similar happens with the yield stress, which decreases from 1060 MPa at 88.5 μm to 526 MPa at 750 μm. For distances greater than 750 μm there are no changes in UTS, so it could be assumed that base material has been reached.

**Figure 9.** Evolution of mechanical properties as a function of distance to cut edge.

Figure 10 shows the evolution of the uniform strain (up to necking), fracture strain, hardening index, and critical strain for Voce's fitting, versus distance to plasma cut (Appendix A).

**Figure 10.** Evolution of the uniform strain, fracture strain, hardening index, and critical strain as a function of the distance to edge cut by plasma.

From this figure, it is possible to appreciate that the material gets softer with the distance from the cut edge. There are no significant changes in the ductility over distance but the fracture strain increases, making the material tougher. From the fitting parameters shown in this figure, it is not possible to say that at 1 mm the base material has been reached, but with the information of Figures 8 and 9 it could be assumed that, from a practical point of view, CHAZ reaches 700 μm. This deduced value is smaller than the one observed for other thermal cutting processes [23].

#### *3.4. Residual Stresses*

Figure 11 shows the residual stresses measured in the longitudinal direction (L) and thickness direction (T) obtained by X-ray diffraction, after its deconvolution. Error in depth was obtained by measuring the removed material in different locations of the sample. The error in residual stress was obtained from the error given by the X-Ray equipment used and the error previously measured in depth. Plasma cutting generates at the surface of the cut a great compression in both directions; L and T (see Figure 1a). The residual compression extends to a distance of approximately 700 μm in the underlying material, located under the cut surface.

The values of these compressive residual stresses measured on the surface are in the order of the material yield stress. These big stresses are probably produced by the great cooling gradients that appear in the plate during the cutting process and they can play a key role in the fatigue behavior of these cut edges (for example, delaying the initiation of cracks and, therefore, increasing fatigue life). However, this behavior also depends on other parameters such as the surface roughness and the microstructural characteristics at those areas underlying the cut [14–17]. Whether or not the removal of the plasma cut surface would be beneficial for the material fatigue performance requires further research.

**Figure 11.** Distribution of residual stresses according to distance to cut edge.

#### **4. Conclusions**


**Author Contributions:** J.A. wrote the paper, performed the measurements of residual stresses and analyzed X-Ray results. A.M.M. designed mini-tensile tests, prepared the samples, and performed the experiments, analyzing their results. He also contributed with the manuscript. S.C. contributed with the writing of the paper and designed, programmed and analyzed all indentation experiments. A.K. supplied all the samples and performed the bending tests. He also optimized conditions for performing plasma the cuts. A.B. performed the metallography preparation and analysis of the cut samples.

**Acknowledgments:** The authors of this work would like to express their gratitude to the European Union for the financial support of the project HIPERCUT: "High Performance Cut Edges in Structural Steel Plates for Demanding Applications" (RSFR-CT-2012-00027), on the results of which this paper is based.

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

#### **Appendix A**

The simplest fit to tensile behavior (stress vs. strain curve) of a material is given by Hollomon's expression [24]

$$
\sigma = \sigma\_0 \varepsilon\_p^n \tag{A1}
$$

where *σ* represents the stress as a function of the plastic strain, *εp*. Only two parameters are used: *σ*<sup>0</sup> and *n*. *n* is known as strain hardening index.

A little more sophisticated and realistic is Voce's relation [25];

$$
\sigma = \sigma\_{\infty} - (\sigma\_{\infty} - \sigma\_0)e^{-\frac{\varepsilon\_p}{ic}} \tag{A2}
$$

This expression has three parameters: *σ*∞ represents the saturation stress (the stress that ideally will be reached for infinite strain), *σ*<sup>0</sup> is the stress for a negligible plastic deformation and *σ<sup>c</sup>* is a critical strain (as shown in Figure A1, at intercept with the abscissa).

**Figure A1.** Interpretation of Voce's parameters.

#### **References**


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

## *Article* **Crystal Structures of GaN Nanodots by Nitrogen Plasma Treatment on Ga Metal Droplets**

#### **Yang-Zhe Su and Ing-Song Yu \***

Department of Materials Science and Engineering, National Dong Hwa University, Hualien 97401, Taiwan; s1992811227@yahoo.com.tw

**\*** Correspondence: isyu@gms.ndhu.edu.tw; Tel.: +886-3-890-3219

Received: 20 April 2018; Accepted: 30 May 2018; Published: 4 June 2018

**Abstract:** Gallium nitride (GaN) is one of important functional materials for optoelectronics and electronics. GaN exists both in equilibrium wurtzite and metastable zinc-blende structural phases. The zinc-blende GaN has superior electronic and optical properties over wurtzite one. In this report, GaN nanodots can be fabricated by Ga metal droplets in ultra-high vacuum and then nitridation by nitrogen plasma. The size, shape, density, and crystal structure of GaN nanodots can be characterized by transmission electron microscopy. The growth parameters, such as pre-nitridation treatment on Si surface, substrate temperature, and plasma nitridation time, affect the crystal structure of GaN nanodots. Higher thermal energy could provide the driving force for the phase transformation of GaN nanodots from zinc-blende to wurtzite structures. Metastable zinc-blende GaN nanodots can be synthesized by the surface modification of Si (111) by nitrogen plasma, i.e., the pre-nitridation treatment is done at a lower growth temperature. This is because the pre-nitridation process can provide a nitrogen-terminal surface for the following Ga droplet formation and a nitrogen-rich condition for the formation of GaN nanodots during droplet epitaxy. The pre-nitridation of Si substrates, the formation of a thin SiN*x* layer, could inhibit the phase transformation of GaN nanodots from zinc-blende to wurtzite phases. The pre-nitridation treatment also affects the dot size, density, and surface roughness of samples.

**Keywords:** nitrogen plasma; Ga droplet; GaN nanodot; transmission electron microscopy; wurtzite; Zinc-blende

#### **1. Introduction**

Group-III nitride-based semiconductors (InN, GaN, and AlN) have very wide bandgaps from 0.64 to 6.2 eV for various applications in optoelectronics and electronics [1]. Gallium nitride, with a direct bandgap of 3.4 eV, which shows high carrier mobility and high thermal conductivity, has been successful applied in commercialized devices, such as light emitting diodes (LED) and high electron mobility transistors (HEMT) [2,3]. Moreover, for the quantum computing applications, nanostructures of GaN have large exciton binding energy and confinement potential for devices, such as single-photon emitters [4] and single-electron transistors [5]. In particular, GaN quantum dots are promising materials in these devices [6].

For the fabrication techniques of GaN, GaN thin films can be epitaxially grown by molecular beam epitaxy (MBE), metal-organic chemical vapor deposition (MOCVD), pulsed laser deposition (PLD), and Hydride vapor-phase epitaxy (HVPE) [7–11]. Once the epi-layers have the lattice mismatch with two-dimensional wetting layers or substrates, self-assembled GaN nanodots can form due to the strain relaxation; this method is called the Stranski-Krastanov (SK) mode of epitaxial growth [12,13]. Among these techniques, MBE procedure is performed in an ultra-high vacuum chamber in order to minimize contamination and in a lower-temperature growth condition. Other advantages of MBE are its capability to create heterostructures with sharp interfaces, and also to form

metastable phase as zinc-blende structure of GaN. For another growth method of semiconductor nanostructures, droplet epitaxy mode using an MBE system was first proposed by Koguchi in 1990 [14,15]. Therefore, the study of droplet epitaxy technique for GaN nanodots was initiated, which is the method of first forming Ga metal droplets in ultra-high vacuum, followed by the treatment of a nitrogen plasma source [16]. There are some advantages of droplet epitaxy by plasma-assisted MBE systems. For instance, self-organized GaN nanodots can be grown directly on various substrates. Density of GaN nanodots can be controlled by the growth parameters. Metastable zinc-blende structure can be performed in the GaN nanodots on sapphires by droplet epitaxy [17–20].

For the crystal structure of GaN nanostructures, wurtzite GaN is the thermodynamically stable phase, which suffers from the presence of a large built-in electric field, called the quantum confined Stark effect, that may degrade the device performance [21,22]. On the other hand, the metastable zinc-blende (cubic) phase of GaN has no polarization fields. The polarization field effect, called the piezoelectric field effect, significantly affects the band structures and optical gain of optoelectronic devices. The radiative recombination time of cubic GaN quantum dots is two orders of magnitude higher than the one for wurtzite GaN quantum dots. The mobility of electrons and holes in a zinc-blende GaN is also intrinsically higher than in a wurtzite GaN due to lower phonon scattering in cubic crystals [23]. Up to now, most of the reports on the growth of cubic GaN nanodots have been proposed by the method of SK mode [24–26]. For the growth mode of droplet epitaxy, Wang et al. reported the mixture phases of wurtzite and zinc-blende on a sapphire substrate at the substrate temperature of 710 ◦C [18]. As's group reported cubic GaN quantum dots grown on 3C-AlN (001) substrates [27]. Studies on the phase transformation of GaN nanodots on silicon by droplet epitaxy technique are still rare. In this work, we focus on the investigation the crystal structures of GaN nanodots on Si (111) substrates fabricated by nitrogen plasma nitridation on the Ga metal droplets. The microstructure and crystal structures of GaN nanodots were characterized by transmission electron microscopy (TEM). The growth parameters of GaN nanodots, such as substrate temperatures and surface pre-nitridation treatment, influence on the crystal phases of GaN nanodots. Nonpolar GaN nanodots with cubic crystal structure can be performed on Si substrates by droplet epitaxy for the future applications in solid state quantum devices.

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

GaN nanodots grown by the method of droplet epitaxy were carried out in our ULVAC MBE system with a radio frequency (RF) nitrogen plasma source [19]. Two inch Si (111) wafers were cleaned using acetone to remove organic impurities and using 10% HF solution to remove the native oxide. After the chemical cleaning, one Si wafer was immediately put into MBE chamber with the base pressure of 1.0 × <sup>10</sup>−<sup>7</sup> Pa. Then, Si substrates were heated to the temperature 850 ◦C for 20 min. After the thermal cleaning, the temperature of Si substrate was set to 500 or 550 ◦C for droplet epitaxy. For the droplet epitaxy process, nanoscale Ga metal droplets were initially formed on Si substrate at beam equivalent pressure 1.9 × <sup>10</sup>−<sup>4</sup> Pa for the duration of 1 min by a Knudsen cell with 99.999999% Ga metal. Ga adatoms had high sticking coefficients on the substrate at lower growth temperature. At elevated temperatures, surface Ga accumulation was less effective since Ga atoms can evaporate from the surface [28]. The accumulation of Ga atoms had high surface free energy to form nanoscale Ga droplets according to the Volmer-Weber (VW) growth model [29]. Subsequently, the droplets were converted into GaN nanodots by using a nitrogen plasma source, called nitridation process. The parameters of nitridation were operated at an RF forward power of 500 W and N2 flux of 2 sccm for 5 or 10 min. In addition, the pre-nitridation treatment at temperature 600 ◦C for 60 min on Si wafers was optionally conducted before the thermal evaporation of Ga metal droplets. The growth parameters of four samples are summarized in Table 1. MBE system was equipped with an in-situ reflection high-energy electron diffraction (RHEED), which was employed for the observation of surface condition during the process of droplet epitaxy. RHEED patterns of samples C1 and C2 are shown in Figure 1. The flat and reconstructed Si (111) surfaces can be obtained after thermal

cleaning in Figure 1a,d, which showed long-streak patterns. After the pre-nitridation treatment on Si, the RHEED patterns changed from striped to foggy due to the formation of amorphous nitride layer on Si, shown in Figure 1b. After the formation of Ga droplets, the pattern of surface was still foggy (not shown) which indicated an amorphous structure on the surface. Then, nitridation process led to the formation of GaN nanodots which showed the ring-centered pattern in Figure 1c. For the sample without the pre-nitridation process (C2), the RHEED pattern became foggy after the formation of Ga droplets, shown in Figure 1e. After the nitridation process, the formation of GaN nanodots also showed the ring-centered pattern in Figure 1f, indicating the polycrystalline GaN nanodots on Si [30].


**Table 1.** The growth parameters of four samples.

**Figure 1.** In-situ reflection high-energy electron diffraction (RHEED) observations: (**a**) C1 sample after thermal cleaning; (**b**) C1 sample after pre-nitridation on Si surface; (**c**) C1 sample after nitridation process (i.e., the formation of GaN nanodots); (**d**) C2 sample after thermal cleaning; (**e**) C2 sample after Ga droplets; and (**f**) C2 sample after nitridation process (i.e., the formation of GaN nanodots).

After the formation of GaN nanodots, surface morphology of the samples was examined using the images generated by field-emission scanning electron microscopy (SEM, JSE-7000F, JEOL, Tokyo, Japan) with accelerating voltage 15 KV. According to the observation of SEM, densities of GaN nanodots could be obtained. The surface roughness of samples was investigated by atomic force microscopy (AFM, C3000, Nanosurf, Liestal, Switzerland). The surface chemical composition of samples was studied by X-ray photoelectron spectroscopy (XPS, K-Alpha, Thermo Scientific, Waltham, MA, USA). For the observation of GaN nanodots microstructure and crystal structures, high-resolution TEM images, and selective-area diffraction patterns were performed by using JEOL JEM-3010 with

accelerating voltage 300 KV. The TEM specimen were prepared by Ar ion milling with an accelerating voltage of 5 KeV for the cross-section and plane-view TEM observations.

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

In the left column of Figure 2, SEM images showed GaN nanodots on the surface of Si (111) in the magnification of 50,000, and their densities in the area of 5.65 μm2 were calculated from the SEM images as 5.72 × <sup>10</sup><sup>10</sup> cm−2, 2.57 × 1010 cm−2, 5.61 × 1010 cm−<sup>2</sup> , and 5.01 × 1010 cm−<sup>2</sup> for samples C1, C2, C3, and C4. To compare C1 with C2 or C3 with C4, the pre-nitridation treatment of Si surface made the density of GaN nanodots increase. C1 and C3 samples with pre-nitridation had higher density than the C2 and C4 samples without pre-nitridation, respectively. Moreover, the cross-section TEM images of GaN nanodots in the range of 500 nm are shown in the right column of Figure 2. The average diameters of GaN nanodots were calculated from the TEM images as 21.9 nm, 24.5 nm, 15.1 nm, and 18.3 nm for C1, C2, C3, and C4, respectively. The pre-nitridation treatment made the size of GaN nanodots decrease slightly. In brief, the surface conditions of Si (111) can be modified by the substrate pre-nitridaiton process, as shown in the observation of in-situ RHEED, the nitrogen-terminal surface could influence the sticking coefficient of Ga atoms or the surface diffusion of Ga atoms. The surface-sticking coefficients of Ga atoms over N atoms can be higher than Ga atoms over Si atoms according to the chemical surface stability. The stronger bonding could reduce the surface diffusion length of Ga atoms during the formation of Ga droplets [31,32]. Therefore, we found an increase of density and a decrease of average diameter of GaN nanodots for the samples with nitrogen plasma pre-nitridation treatment (C1 and C3).

**Figure 2.** SEM images (**left column**) and cross-section TEM images (**right column**) of GaN nanodots for samples C1, C2, C3 and C4.

To further analyze their surface roughness, AFM images in an area of 5 μm × 5 μm of four samples are shown in Figure 3. The average roughness of samples C1, C2, C3, and C4 are 2.56, 3.04, 1.34, and 1.65 nm, respectively. The samples without pre-nitridation treatment (C2 or C4) had higher surface roughness value than the ones with pre-nitridation (C1 or C3). This is because the larger size and lower density of GaN nanodots on Si which can perform higher surface roughness of samples. The observation of AFM was consistent to the results of SEM and TEM in Figure 2.

**Figure 3.** Atomic force microscopy (AFM) images and average roughness of four samples: C1, C2, C3, and C4.

Since the pre-nitirdation treatment of Si (111) influenced the growth of GaN nanodots, we went forward to investigate the surface chemical composition using the measurements of XPS. Figure 4 shows the de-convoluted Ga-3d XPS spectra of GaN nanodots on Si (111) for samples C1, C2, C3, and C4. The XPS spectra were divided into three major components: Ga–O bonding with peak energy around 20.7 eV, Ga–N bonding with peak energy around 19.7 eV, and O–O bonding with the peak energy around 25.0 eV, obtained using the software *Thermo Avantage* (version 4, Thermo Scientific, Waltham, MA, USA). The samples with pre-nitridation treatment (C1 and C3) had relatively stronger peak intensities of Ga–N bonding due to more nitrogen provided for the formation of GaN nanodots. Meanwhile, we found high amount of oxygen chemisorptions on the surface of samples [33]. The strong Ga–O and weak O–O bonds came from the oxidation of nanodots on exposure to the air due to strong difference the electro negativity between Ga and O atoms [21]. This was because the uncompleted crystallization of Ga metal droplets could have existed after nitrogen plasma nitridation. When the nitridation time increased for samples C3 and C4, the peak intensity of Ga–N bonding was stronger due to the formation of the GaN crystal being more complete.

**Figure 4.** De-convoluted Ga-3d X-ray photoelectron spectroscopy (XPS) spectra of GaN nanodots on Si (111) for samples C1, C2, C3, and C4.

In order to identify the crystal structures of GaN nanodots, we prepared the specimens for the cross-section images of high-resolution transmission electron microscopy (HRTEM) and analyzed the crystal planes using the software *Digital Micrograpy* (version 1, Gatan, Pleasanton, CA, USA). The fast Fourier transformation (FFT) of the original HRTEM images was conducted on a single GaN nanodot, which can provide the simulated diffraction patterns. And then, the inverse fast Fourier transformation (IFFT) of the simulated diffraction patterns provided a clearer crystal structure of a single GaN nanodot. Finally, crystal planes can be identified according to the database of software [34]. Moreover, for the TEM observations of multiple GaN nanodots, we prepared the specimens for plane-view TEM images as shown in Figure 5 and analyzed the selective area diffraction (SAD) patterns of GaN nanodots at a magnification of 500,000. The d-space of ring-like patterns can provide the planes for each ring using the software *CSpot* (version 1, CrystOrient, Zabierzów, Poland) with the aid of a crystallography open database [35]. Figure 5 shows the plane-view TEM image of sample C1 and its ring-like diffraction pattern in the inset. The ring-like pattern came from the polycrystalline GaN nanodots, which was consistent with the results of RHEED observations. In the TEM images, we could also find that the individual GaN dot could contain different misoriented nano-crystals.

**Figure 5.** Plane-view TEM images of sample C1 and its diffraction pattern in the inset.

Figure 6a,b shows cross-section HRTEM images of a single GaN nanodot, and Figure 6c,d shows the SAD patterns of GaN nanodots for samples C1 and C2, respectively. According to the analysis using TEM software, the crystal planes of a single GaN nanodot could be identified, and the phase of GaN nanodots could be investigated by the polycrystalline ring-like patterns. For the sample C1, the (200) plane of zinc-blende GaN is shown in Figure 6a, and the cubic structure of GaN nanodots with planes (111), (200), and (220) is shown in Figure 6c. Cubic GaN nanodots could be obtained on Si (111) by the growth parameters of sample C1. On the contrary, we verified the TEM images of sample C2. The (101) and (002) planes of wurtzite structure are shown in Figure 6b, and the ring patterns with planes (100), (002) and (101) for the wurtzite crystal structure of GaN nanodots are shown in Figure 6d. According to the TEM results of samples C1 and C2, the pre-nitridation treatment on Si (111) could inhibit the phase transformation from metastable zinc-blende to stable wurtzite structures. For the microstructures of GaN films grown by MBE, the nitrogen-rich growth condition showed the existence of zinc-blende GaN phase [36]. The pre-nitridation treatment could serve as a larger nitrogen source for the formation of GaN nanodots. Therefore, the cubic GaN nanodots dominated in the sample C1.

**Figure 6.** Cross-section HRTEM images of a single GaN nanodot: Sample with pre-nitridation C1 (**a**) and sample without pre-nitridation C2 (**b**); Selective-area diffraction patterns of GaN nanodots: sample C1 (**c**), and sample C2 (**d**).

To further investigate the effect of thermal budget on the crystal structures of GaN nanodots, Figure 7a,b shows cross-section HRTEM images of a single GaN nanodot, and Figure 7c,d shows the SAD patterns of GaN nanodots for samples C3 and C4, respectively. The (002) and (102) planes of wurtzite structure are shown in Figure 7a,b, and the ring-like patterns with planes (100), (002) and (101) both show the wurtzite crystal structure of GaN nanodots in Figure 7c,d. The higher substrate temperature of droplet epitaxy could lead to GaN nanodots phase transformation to wurtzite crystal structure. When the substrate temperature was raised during droplet epitaxy, high surface migration occurred, which improved the quality of the growth front of GaN and it became favorable to form the more thermodynamically equilibrated wurtzite phase [31,36].

**Figure 7.** Cross-section HRTEM images of a single GaN nanodot: sample C3 (**a**) and sample C4 (**b**); Selective-area diffraction patterns of GaN nanodots: sample C3 (**c**) and sample C4 (**d**).

#### **4. Conclusions**

GaN nanodots were fabricated on Si (111) by droplet epitaxy using a nitrogen plasma-assisted MBE system. Polycrystalline GaN nanodots were shown by the characterizations of in-situ RDEED and TEM. Uncompleted crystallization of the Ga metal droplets were observed from the measurement using XPS. The phase transformation of GaN nanodots from metastable zinc-blende crystal structure to stable wurtzite crystal structure could be inhibited by the surface modification of nitrogen plasma (i.e., pre-nitridation treatment). The pre-nitridation treatment also influenced the size and density of GaN nanodots. The nitrogen-terminal Si surface could not only change the sticking coefficient or the surface diffusion of Ga atoms, but also could provide the nitrogen-rich condition for the existence of zinc-blende phase. On the other hand, the stable wurtzite structure of GaN nandots could be found at higher growth temperatures. This was because the phase transformation of crystal structures depended on the activation free energy barrier. The higher temperature growth could decrease the energy barrier of phase transformation. In this report, we focused on the investigation of GaN-nanodot crystal structures. Nonpolar GaN nanodots with cubic crystal structure could be synthesized during the process of droplet epitaxy, which could have future applications of GaN nanodots for solid state quantum devices.

**Author Contributions:** I.-S.Y. designed the experiments and wrote the manuscript; Y.-Z.S. performed the experiments and analyzed the data.

**Funding:** This research was funded by Ministry of Science and Technology Taiwan (MOST 106-2221-E-259-010) and ULVAC Taiwan (NDHU 106A605).

**Acknowledgments:** The authors acknowledge Ministry of Science and Technology Taiwan and ULVAC Taiwan for financially supporting this study and publication. The authors also would like to acknowledge Micheal Chen and Stanley Wu of ULVAC Taiwan for the maintenance of PA-MBE system.

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

#### **References**


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