Review of GaN Thin Film and Nanorod Growth Using Magnetron Sputter Epitaxy

Magnetron sputter epitaxy (MSE) offers several advantages compared to alternative GaN epitaxy growth methods, including mature sputtering technology, the possibility for very large area deposition, and low-temperature growth of high-quality electronic-grade GaN. In this article, we review the basics of reactive sputtering for MSE growth of GaN using a liquid Ga target. Various target biasing schemes are discussed, including direct current (DC), radio frequency (RF), pulsed DC, and high-power impulse magnetron sputtering (HiPIMS). Examples are given for MSE-grown GaN thin films with material quality comparable to those grown using alternative methods such as molecular-beam epitaxy (MBE), metal–organic chemical vapor deposition (MOCVD), and hydride vapor phase epitaxy (HVPE). In addition, successful GaN doping and the fabrication of practical devices have been demonstrated. Beyond the planar thin film form, MSE-grown GaN nanorods have also been demonstrated through self-assembled and selective area growth (SAG) method. With better understanding in process physics and improvements in material quality, MSE is expected to become an important technology for the growth of GaN.


Introduction
For three decades, III-nitride semiconductors including GaN, InN, AlN, and their alloys have been a subject of intensive research and development activity, resulting in the maturing of the material system and widespread adoption of III-nitride-based devices. III-nitrides can thus be found in a wide range of devices, including high-power and high-frequency field-effect transistors (FET) [1], visible and ultraviolet (UV) light-emitting diodes (LEDs) [2], and laser diodes (LDs) [3]. The tunable band-gap of ternary III-nitride enables the material system to cover a wide range of spectrum, from the UV regime to the near infrared.
For commercial large-scale application, the metal-organic chemical vapor deposition (MOCVD) method is still the most widely used method to grow GaN layers due to its high throughput capabilities. However, MOCVD requires high growth temperature in order to enable the pyrolysis of the ammonia gas into active nitrogen species. This requirement results in limited substrate selection, an upper limit for material doping, and low indium incorporation due to the phase separation in ternary and quaternary systems [4]. Furthermore, the chemical vapor deposition (CVD) system requires the use of toxic gases which can be harmful to the environment and require careful handling procedures.
In contrast to the MOCVD reactor, molecular-beam epitaxy (MBE) is a physical vapor deposition (PVD) process where growth proceeds in a layer-by-layer manner. The ultra-high vacuum (UHV) condition promotes the growth of high-purity GaN and the low working pressure of~10 −5 Torr enables long mean free path for the atoms inside the chamber. The metal atoms are supplied through thermal effusion using Knudsen cells, while the nitrogen plasma source in plasma-assisted MBE allows N 2 bonds to be broken, producing the reactive nitrogen species without the need for high growth temperature. However, the high operation cost and difficulty in scaling up the growth process makes MBE impractical for large-scale industrial application, limiting it to lab scale research.
Magnetron sputter epitaxy (MSE) is a more recent addition to the PVD method that is based on the reactive sputtering process widely adopted in both engineering and semiconductor industries. The growth process during MSE takes place in a non-equilibrium condition, and the kinetic diffusion of surface adatom plays an important role during growth [5,6]. The main advantages of MSE include mature technology, low equipment cost, and the possibility for large area deposition. Using MSE, high flux and low energy (~20 eV) ion bombardment can be employed to increase the crystallinity of the growing film even at low growth temperatures. In terms of environmental impact, the MSE is more environmentally friendly compared to the MOCVD as the process uses non-toxic gases, such as argon, nitrogen, and oxygen. The comparison between MOCVD, MBE, and MSE for the growth of GaN film is given in Table 1. Growth at moderate temperature (~800 • C) Growth at moderate temperature (~700 • C). High-quality film growth at lower temperature is possible.
Over the past decade, significant progress has been made in the growth of semiconductor-grade GaN and its compounds using MSE [7]. GaN thin films with material quality comparable to films grown using alternative methods such as hydride vapor phase epitaxy (HVPE) and MBE has been demonstrated [8,9]. Furthermore, a novel growth GaN film method using high-power impulse magnetron sputtering (HiPIMS) has been demonstrated, with the potential to improve GaN material quality [10]. Beyond material growth, electronic and optoelectronic devices grown using MSE have been demonstrated [11][12][13][14]. A more recent progress shows the possibility of using MSE to produce high-quality GaN nanorods on various cost-effective and functional substrates [15][16][17], allowing for the fabrication of novel devices. This review paper is organized as follows. Section 2 provides an introduction to the reactive sputtering mechanism and gives a brief review of GaN sputtering using alternative targets. Afterwards, a detailed explanation of GaN MSE system configuration using liquid Ga target is given. Section 3 focuses on the growth mechanism of GaN thin film, discussing how the various growth parameters of the MSE system influence the properties of the GaN film. We also discuss the use of MSE system based on pulsed sputter deposition (PSD) and HiPIMS for GaN growth. Finally, we review the possibility of using MSE for both n-type and p-type GaN, which are essential for device application. Section 4 focuses on the growth of both self-assembled and selective-area growth (SAG) of GaN nanorods using MSE in terms of process conditions and patterning methods.

An Overview of MSE Growth of GaN
Following Birch et al. [18], magnetron sputter epitaxy is defined as epitaxial growth by magnetron sputter deposition of GaN under the same stringent vacuum and sample handling conditions as in practice in MBE. The background pressure of the main sputtering chamber is maintained below 1 × 10 −8 Torr to maintain a high material purity. This section begins with a review of the basic mechanism of reactive sputtering, which is the foundation of MSE, and an account of GaN deposition using alternative sputtering targets. Finally, the design considerations of an MSE chamber equipped with a liquid Ga target is discussed.

Basic Principles of Reactive Sputtering
Reactive sputtering and the details of its underlying mechanisms have been extensively described [19,20]. Therefore, this part will only focus on the sputter mechanism to help understand the sputter epitaxy process of GaN films.
The sputtering mechanism takes place through the ejection of atoms from a sputter target under the bombardment of ions. A working gas, usually inert Ar and N 2 gas in the case of III-nitride growth, is introduced into a vacuum sputtering chamber. Negative voltage bias is applied to the sputtering target to start the gas ionization process. As the gas is ionized, plasma consisting of electrons and positively charged ions is formed. Due to the target's negative bias, the ions accelerate and hit the target with kinetic energy proportional to the target bias. Above an energy threshold, the target's atoms will be ejected, travel across the plasma, and eventually condense on the surface of the substrate. In addition to ejecting the target materials, this ion bombardment process also generates secondary electrons which are required for maintaining the plasma. For magnetron sputtering system, permanent magnets are installed behind the target in order to elongate the trajectory of electrons in a helical loop, helping to increase the plasma density with increasing scattering probability to gases in the vicinity of the target.
Under reactive sputtering, reactions take place between the metal adatoms and ions sputtered from a target and the reactive gas on the substrate surface, forming a compound thin film. A schematic of the reactive sputtering process for GaN deposition is shown in Figure 1. The Ga target is connected to the cathode and is under a negative bias. The target is bombarded with positive nitrogen and argon ions, and sputtered Ga atoms and ions travel through the plasma onto the substrate surface. Reaction takes place between nitrogen and Ga reactive species, including atoms and ions, on the substrate surface, forming a nitride film.

Figure 1.
Simplified schematic of reactive sputtering process for the deposition of GaN [21].
Reproduced with permission from Linköping University Electronic Press.

Alternative Sputtering Target as Ga Source
Using pure Ga as the sputtering target for the MSE process is challenging due to several conditions. The Ga target needs to have a proper container due the low melting point of metallic Ga (29 °C), and the formation of GaN layer on the target surface results in a drift of process condition over time. In magnetron sputter epitaxy for III-nitride growth, the sputtered group III metal normally interacts with the reactive gas particles, forming a compound material at the substrate surface. However, this reaction can occur at the target surface and deeper into the target depending on the ion energy [22]. The formation of a compound at the surface of the metal is referred to as target poisoning [23], and can be problematic if the formation rate is faster than the sputtering rate. The target poisoning can lead to a drift in the process condition due to the reduced sputtering yield and reduced yield of secondary electron. To avoid this problem, several alternative sputtering targets have been used as a Ga source.
It is possible to grow GaN by using a GaAs target as the metal source. The incorporation of nitrogen within the process gas mixture is proven to be effective in suppressing the formation of GaAs [24]. Elkashef et al. has grown single-phase GaN with a negligible incorporation of As in the film by using 100% nitrogen process condition at 550 °C. As the negative enthalpy of GaN formation is higher than that of GaAs, the formation of GaN is preferred even with GaAs as the sputtering target. Yadav et al. [25] controlled the formation of GaAs inside the GaN film by increasing the substrate temperature, where the film showed better crystallinity together with incorporation of As at growth above 700 °C.
Alloying the Ga metal with other group III-metal is shown to be another method to stabilize the Ga target and prevent melting during sputtering. Shinoda et al. have demonstrated the growth of AlGaN and InGaN ternary compounds using AlxGa1-x and InxGa1-x metal alloy targets, respectively. By adjusting the substrate temperature, the composition of the ternary alloy can be tuned [26]. A "pure" GaN was claimed to be grown using an Al0.1Ga0.9 alloy target, where the substrate temperature and partial gas pressure can be tweaked to modify the composition of the ternary compound in the direction of GaN. Alloyed target such as AlxGa1-x, however, shows strong phase separation at room temperature, making it challenging to control their composition.
Cermet targets have also been shown to be suitable for growing ternary and quaternary IIInitride compounds with an adjustable composition [27][28][29][30][31]. For III-nitride cermet targets, the targets are typically fabricated by pressing together a mixture of group III metals and GaN. The amount of metals within the target mixture should be less than 50% to have a solid support within the target. By changing the composition of the cermet target, ternary films with varying composition can be obtained. In addition to modifying the target itself, the sputter power and substrate temperature can  [21]. Reproduced with permission from Linköping University Electronic Press.

Alternative Sputtering Target as Ga Source
Using pure Ga as the sputtering target for the MSE process is challenging due to several conditions. The Ga target needs to have a proper container due the low melting point of metallic Ga (29 • C), and the formation of GaN layer on the target surface results in a drift of process condition over time. In magnetron sputter epitaxy for III-nitride growth, the sputtered group III metal normally interacts with the reactive gas particles, forming a compound material at the substrate surface. However, this reaction can occur at the target surface and deeper into the target depending on the ion energy [22]. The formation of a compound at the surface of the metal is referred to as target poisoning [23], and can be problematic if the formation rate is faster than the sputtering rate. The target poisoning can lead to a drift in the process condition due to the reduced sputtering yield and reduced yield of secondary electron. To avoid this problem, several alternative sputtering targets have been used as a Ga source.
It is possible to grow GaN by using a GaAs target as the metal source. The incorporation of nitrogen within the process gas mixture is proven to be effective in suppressing the formation of GaAs [24]. Elkashef et al. has grown single-phase GaN with a negligible incorporation of As in the film by using 100% nitrogen process condition at 550 • C. As the negative enthalpy of GaN formation is higher than that of GaAs, the formation of GaN is preferred even with GaAs as the sputtering target. Yadav et al. [25] controlled the formation of GaAs inside the GaN film by increasing the substrate temperature, where the film showed better crystallinity together with incorporation of As at growth above 700 • C.
Alloying the Ga metal with other group III-metal is shown to be another method to stabilize the Ga target and prevent melting during sputtering. Shinoda et al. have demonstrated the growth of AlGaN and InGaN ternary compounds using Al x Ga 1-x and In x Ga 1-x metal alloy targets, respectively. By adjusting the substrate temperature, the composition of the ternary alloy can be tuned [26]. A "pure" GaN was claimed to be grown using an Al 0.1 Ga 0.9 alloy target, where the substrate temperature and partial gas pressure can be tweaked to modify the composition of the ternary compound in the direction of GaN. Alloyed target such as Al x Ga 1-x , however, shows strong phase separation at room temperature, making it challenging to control their composition.
Cermet targets have also been shown to be suitable for growing ternary and quaternary III-nitride compounds with an adjustable composition [27][28][29][30][31]. For III-nitride cermet targets, the targets are typically fabricated by pressing together a mixture of group III metals and GaN. The amount of metals within the target mixture should be less than 50% to have a solid support within the target. By changing the composition of the cermet target, ternary films with varying composition can be obtained. In addition to modifying the target itself, the sputter power and substrate temperature can be used in order to tune the composition of the ternary and quaternary compounds [30]. However, the composition of the cermet target may vary over time after each sputtering process, which may affect the repeatability of experiments.

MSE System Configuration Based on Liquid Ga Target
Although alternative targets as Ga sources exist and using a liquid Ga target is challenging, a liquid Ga-based sputtering system is still preferable due to its potential for a scalable and industrially relevant system. The use of liquid Ga target gives multiple advantages, including high growth rates, elimination of target erosion, and the possibility for continuous target supply [15]. In addition, precise control of the film composition can be achieved by using liquid Ga target. This can be done by co-sputtering and controlling the bias power delivered to each target, separately.
In this section, most of the MSE chamber design considerations will be based on the MSE chamber used in the Thin Film Physics division of Linköping University, although some design variations may exist. Variations may include the power supply choice (DC, RF, or pulsed), the magnetic configuration, the container material for liquid Ga, addition of effusion cell for doping, and chamber cooling system.
A simplified diagram for a typical MSE system is shown in Figure 2a. To facilitate the liquid Ga target, the sputtering is done in a sputter-up configuration with the liquid Ga contained inside a 50 mm diameter water-cooled trough. The substrate is clamped by the substrate holder, facing downwards toward the target. During deposition, substrate rotation can be introduced to enhance uniformity across the substrate, and growth kinetics can be modified by heating the substrate and further enhanced by ion assistance through a negative substrate bias attracting ions from the sputtering plasma. The sputtering target is a very high purity liquid Ga (99.99999% pure) placed in a stainless steel cup clamped to the magnetron. The chamber pressure is controlled by adjusting a butterfly valve, which determines the total pressure during deposition. For MSE growth, ultra-pure process gas including Ar (99.999999% pure) and N 2 (99.999999% pure) are introduced during the sputtering process with the partial pressure of each gas controlled by mass flow controllers and a capacitance manometer.
The ionization process during the sputtering is enhanced by using a combination of magnets on the back side of the target (called magnetron). The fast moving secondary electrons, generated during ion bombardment of the target, are trapped by magnetic field leading to an increase in electron density near the target surface.
According to their magnetic field configuration, magnetrons can be classified as balanced, unbalanced type I, and unbalanced type II. In our system, we are using an unbalanced type II magnetron configuration, as shown in Figure 2b. In this configuration, the magnetic flux of the center pole is much weaker than that of the outer magnet, resulting in magnetic field lines extending from the magnetron toward the substrate. Secondary electrons, emitted from the sputtering target, are magnetically guided toward the substrate while ionizing the sputtering gas and, because of charge neutrality requirements they also pull the positive ions along with them. When the substrate is floating (without ground), a negative self-bias is formed due to the accumulation of electrons at the substrate surface. More ions can be attracted towards to the substrate, which can provide additional kinetic energy to the film growth through momentum transfer from the ions to the adatoms [32,33]. By applying a proper external substrate bias (typically less than −100 V), the as-grown film quality can be improved with the assistance of low-energy ions.
An important aspect for sputtering, using liquid Ga target, is selecting the trough material and the design for containing liquid Ga. Due to the low melting point of Ga and the significant heat generated during the sputtering process, the Ga target will preferably be in a liquid phase during the deposition process, proper care must be taken to ensure a stable deposition condition and prevent the liquid Ga from spilling. Pure Ga does not wet any surface and has a very high surface tension (7.35 × 10 −3 N/cm) [34]. However, the formation of native oxide layer due to reaction between Ga and oxygen may also change the wetting properties of liquid Ga [35]. Therefore, the trough material should be selected so that it can be wet by the liquid Ga while also UHV-compatible at the same time. In the Appl. Sci. 2020, 10, 3050 6 of 26 literature, stainless steel [8,36] and molybdenum [37] have been used to contain liquid Ga. In addition to the material selection, the shape of the trough must also be optimized to prevent the formation of air pockets in the target and maintain a flat surface, as shown in Figure 3a-d. From our experiments, we found that a stainless steel trough with concave-shaped surface is the best option to minimize the formation of Ga droplets and voids in the target during the deposition process [21].
Appl. Sci. 2020, 9, x FOR PEER REVIEW 6 of 28 In the literature, stainless steel [8,36] and molybdenum [37] have been used to contain liquid Ga. In addition to the material selection, the shape of the trough must also be optimized to prevent the formation of air pockets in the target and maintain a flat surface, as shown in Figure 3a-d. From our experiments, we found that a stainless steel trough with concave-shaped surface is the best option to minimize the formation of Ga droplets and voids in the target during the deposition process [21].  Schematic of a liquid Ga target equipped with a type II unbalanced magnetron. This configuration results in magnetic field lines extending toward the target [21]. Reproduced with permission from Linköping University Electronic Press.
As mentioned above, one of the challenges of using a Ga target is the target poisoning, caused by the formation of compound GaN on the target surface, which leads to a reduced sputtering rate. The change of Ga target from fresh Ga metal to partially-poisoned mode is shown in Figure 3e,f respectively. In addition to a poisoning of the target surface, nitrogen ion-bombardment can also cause nucleation of microscopic solid GaN particles inside the liquid target which irreversibly leads to a solid-liquid two-phase bulk target composition, which is not wanted for maintaining a stable process.
Another issue related to liquid Ga target for sputtering is the formation of gas bubbles inside the target. The gas bubbles are formed by the Ar and N 2 gas ion bombardments, resulting in trapped gas within the target. During deposition, there is a possibility for these bubbles to "burst" out of the target surface, resulting in splashing of Ga droplets within the chamber. From our experiments, the bubble formation can be suppressed by using a lower chamber base pressure and better cooling mechanism of the magnetrons.  [21]. Reproduced with permission from Linköping University Electronic Press.
As mentioned above, one of the challenges of using a Ga target is the target poisoning, caused by the formation of compound GaN on the target surface, which leads to a reduced sputtering rate. The change of Ga target from fresh Ga metal to partially-poisoned mode is shown in Figure 3e,f respectively. In addition to a poisoning of the target surface, nitrogen ion-bombardment can also cause nucleation of microscopic solid GaN particles inside the liquid target which irreversibly leads to a solid-liquid two-phase bulk target composition, which is not wanted for maintaining a stable process.
Another issue related to liquid Ga target for sputtering is the formation of gas bubbles inside the target. The gas bubbles are formed by the Ar and N2 gas ion bombardments, resulting in trapped gas within the target. During deposition, there is a possibility for these bubbles to "burst" out of the target surface, resulting in splashing of Ga droplets within the chamber. From our experiments, the bubble formation can be suppressed by using a lower chamber base pressure and better cooling mechanism of the magnetrons.

Effect of Growth Parameters on GaN Thin Films
The deposition process of GaN thin film using sputtering relies on a complex interplay of various process conditions. In general, the parameters that affect the growth include substrate temperature, total working pressure, partial pressure of gas composition, substrate bias, and the amount of power delivered into the sputtering target. The effects of these parameters are highlighted in Table 2.

Effect of Growth Parameters on GaN Thin Films
The deposition process of GaN thin film using sputtering relies on a complex interplay of various process conditions. In general, the parameters that affect the growth include substrate temperature, total working pressure, partial pressure of gas composition, substrate bias, and the amount of power delivered into the sputtering target. The effects of these parameters are highlighted in Table 2. Table 2. Effect of sputtering parameter on the growth of GaN.

Parameter Effects
Substrate temperature At higher temperature, Ga desorbs from the substrate surface, which limits the film growth rate. Higher temperature leads to better adatom mobility and overall better crystal quality [9].

Chamber pressure
Changes the mean free path of reactive species and concentration of gas particle. Lower pressure results in more active nitrogen species. Affects the directionality of the metal flux [38].
Partial pressure between Ar and N 2 Determines whether the growth occurs under Ga-or N-rich conditions. Determines the sputter yield of GaN and Ga within the gas [26,37].
Sputter target bias Changes ion current and sputter yield. Affects the amount of ion damage induced on the film [39].

Substrate bias
Changes the ion energy and affects the interaction between the ions and surface. May enhance the adatom mobility through ion-adatom momentum transfer, but may also induce ion damage. Can be employed to selectively attract metal ions for HiPIMS system [10].
By taking advantage of the high kinetic energy of the sputtered atoms to stimulate the adatom mobility, MSE growth of GaN has been demonstrated at near room temperature. However, the film is generally polycrystalline with [0001] preferred orientation [38,40]. An increase in growth temperature typically results in better film quality in terms of reduced domain boundaries and lattice defect density and smoother film surface [9,36,[41][42][43][44]. In general, increased substrate temperature helps to increase the surface adatom mobility, which in turn results in better crystal quality. As the substrate temperature is increased, the growth mode shifts from island growth to layer by layer growth mode [42]. However, a higher growth temperature also means a higher desorption rate of Ga adatom from the surface, which would put a limit on the deposition rate for the given temperature.
The growth of GaN using MSE is typically done using a mixture of Ar and N 2 as the process gas. Parallel to the MBE process, changing the ratio of partial gas pressure between Ar and N 2 will directly change the ratio of Ga and nitrogen (N+ N 2 ) arriving on the substrate surface [26,37]. This would then determine whether the growth process occurs under metal-rich or nitrogen-rich condition. Under nitrogen-rich condition, Ga adatom mobility is reduced, resulting in three-dimensional growth and the formation of pyramidal structures [26,37,41,45]. On the other hand, growth in a metal-rich condition results in a smoother surface morphology. The improvement in surface morphology is caused by the formation of Ga ad-layer which increase the Ga surface mobility and promote layer growth mode [46].
Another consideration in determining the mixture of the process gas is to control the deposition rate and target poisoning. Target poisoning can occur on the Ga target during reactive sputtering due to the formation of GaN compound on the target. This results in a reduction in sputtering deposition rate over time as the target shifts to poisoned mode. Vasquez et al. [47] has simulated the process of Ga target poisoning under a purely nitrogen gas ambient and confirms that under low nitrogen ion energy, a smooth GaN surface forms at the surface of the target. Introducing Ar gas into the mixture can assist in preventing target poisoning. The deposition rate of Ga is shown to increase proportional to the Ar partial pressure due to the partial removal of the GaN surface through Ar + ion bombardment on the target surface [39].
Increasing the power applied to the target (cathode) is the main way to increase the deposition rate in sputter deposition processes. However, this is also associated with an increased potential drop between the plasma and the cathode which leads to higher kinetic energy of the sputtering gas ions. Upon impact, the ions transfer a majority of their momentum to the atoms in the target, causing, e.g., sputtering and subsurface collision cascades, but a fraction are kinetically reflected as neutrals, back into the chamber. Back reflected neutrals, which may retain a part of their initial kinetic energy, may result in the formation of crystal defects, gas implantation, re-sputtering, etc., if allowed to reach the growing film. For example, Knox-Davies et al. [39] discovered an increase in residual compressive stress in nanocrystalline GaN films with increasing negative target voltage.
The total chamber pressure affects the concentration, mean free path, and energy of the incoming sputtered atoms as well as the back-reflected neutrals. Song et al. [38] has demonstrated that the crystallinity of the GaN layer can be improved by decreasing the total pressure of the chamber. At a lower chamber pressure, the number of activated nitrogen radicals is increased, and more Ga atoms with higher kinetic energy arrive on the substrate due to the increased mean free path length; thus, implying a delicate balance in process parameters to achieve a high-quality GaN thin film.

MSE Growth of Single-Crystal GaN Films Using Liquid Ga Target
Although the growth of GaN film using sputtering has been reported from as early as 1972 [48], early works typically report the growth of polycrystalline GaN films [49,50]. In order for MSE to compete with established epitaxial growth techniques such as MOCVD and MBE, the GaN film must have high crystallinity, an acceptable level of impurities, and contain a low number of threading dislocations. This section describes the progress for the epitaxial growth of single crystal GaN films using MSE. Singh et al. [51] has epitaxially grown single-crystal GaN on (0001) sapphire substrate using DC MSE with Ar and 99.9995% pure ammonia as the working gas. Under this configuration, the growth at higher temperature (>850 • C) is limited by the Ga flux due to additional thermal cracking of ammonia. The growth was performed at 4 mTorr with two growth steps consisting of the initial GaN buffer layer growth and the GaN epilayer at 900 • C. X-ray diffraction (XRD) measurement using rocking curve and ω-2θ scan gives an FWHM of 620 and 32 arcsec, respectively, indicating high degree of crystal quality, but with low angle tilt between the crystal grains. Low-temperature PL measurement on a 1.7 µm thick GaN epilayer gives a strong band-edge emission at 3.48 eV, with an FWHM of 8 meV. In this ammonia-based technique, the supply of active nitrogen species is sensitive to the substrate temperature as the temperature may affect the thermal cracking of ammonia.
Daigo et al. [41] utilized RF MSE to produce single crystalline c-plane GaN film on (0001) sapphire. In this work, only pure N 2 gas is used as the working gas. By adjusting the growth temperature and total working pressure, the crystalline quality of the GaN can be improved. By using an optimized growth condition, 1.5 µm thick GaN film was grown at 800 • C substrate temperature, 100 W RF power, and 2 mTorr total pressure. XRD rocking curve measurement gives an FWHM of 2106 arcsec while Raman spectroscopy measurement gives a E 2 (high) phonon mode peak width that is close to bulk single crystal GaN (3.5 cm −1 ). The film surface shows a faceted pyramid structure with six {10-11}, which can be attributed to the limited Ga adatom diffusion under N-rich growth conditions. Junaid et al. [8] have demonstrated the possibility of using MSE to produce an electronic grade GaN thin film grown by DC power source and liquid Ga metal source. By using an ultra-high vacuum chamber condition coupled with ultra-pure materials, thin film growth with a growth rate of~1.5 Å/s is obtained. The growth was performed at a substrate temperature of 700 • C with the substrate at floating potential (−20 V) under 2.5 mTorr Ar partial pressure (P Ar ) and 2 mTorr N 2 partial pressure (P N 2 ). Ga was sputtered at 20 W for 2 min before the power was reduced to 10 W. Figure 4a shows the time of flight elastic recoil detection (ToF-ERDA) analysis of the grown GaN. From the relative amount of Ga and N, the film grows under a stoichiometric condition. Selective-area electron diffraction (SAED) pattern result in Figure 4b gives an epitaxial relationship of [11][12][13][14][15][16][17][18][19][20]GaN//[1-100]Al 2 O 3 and (0001)GaN//(0001)Al 2 O 3 . The rocking curve FWHM of GaN(0002) peak measured using high-resolution XRD is 1054 arcsec, comparable to GaN layer grown directly on sapphire by HVPE and MBE [52,53]. The low-temperature µ-PL result shown in Figure 4c gives a near band-edge emission at 3.474 eV with an FWHM of 6.3 meV, indicating good crystal quality. The threading dislocation density of the GaN film is estimated to be~8 × 10 9 cm −2 . KOH etching shows that the film grows in the 0001 direction with nitrogen polarity.
Appl. Sci. 2020, 9, x FOR PEER REVIEW 10 of 28 Shinoda et al. [9] has demonstrated the MSE-growth of GaN thin film on a (0001) sapphire substrate and GaN on sapphire template using RF target biasing. From a series of experiment, they found that higher substrate temperature results in a more stable crystalline quality and smoother GaN surface. However, growth performed at 1050 °C results in the formation of pits, caused by the decomposition of GaN under ultra-high vacuum environment. At a growth temperature of 890 °C, they managed to perform homoepitaxial growth of GaN thin film on GaN template with threading dislocation density and defect density comparable with MOCVD-grown films. The film on GaN  [11][12][13][14][15][16][17][18][19][20] in the GaN crystal. (c) Low-temperature µ-PL spectrum of the sample obtained at 4 K [8]. Reprinted with permission from AIP publishing. Shinoda et al. [9] has demonstrated the MSE-growth of GaN thin film on a (0001) sapphire substrate and GaN on sapphire template using RF target biasing. From a series of experiment, they found that higher substrate temperature results in a more stable crystalline quality and smoother GaN surface. However, growth performed at 1050 • C results in the formation of pits, caused by the decomposition of GaN under ultra-high vacuum environment. At a growth temperature of 890 • C, they managed to perform homoepitaxial growth of GaN thin film on GaN template with threading dislocation density and defect density comparable with MOCVD-grown films. The film on GaN template shows excellent quality, with a threading dislocation density of 1.7 × 10 8 cm −2 . XRD rocking curves from the (0002) and (10)(11)(12) plane gives an FWHM of 300 and 310 arcsec, respectively. This result shows that the choice of substrate template or interlayer is also crucial in reducing the number of structural defects for MSE-grown GaN films.

Pulsed DC Sputtering and High-Power Impulse Magnetron Sputtering (HiPIMS)
Pulsed DC sputtering, also known as pulsed sputter deposition (PSD), is a technique where the sputtering target is biased with a pulsed DC voltage, usually in the range of hundreds of volts [54]. During the pulse on period, the target is biased with negative voltage, initiating the sputtering process by the ions. During the off period, the target is typically biased with low positive voltage in order to clear up any charge buildup. Pulsed sputtering enables the supply of high-energy precursors onto large-area substrates, enabling low-temperature growth and significant reduction in threading dislocations. DC pulsed sputtering has been used extensively for the growth of III-nitrides [11][12][13][14]37,[55][56][57][58][59][60][61][62][63][64] resulting in various III-nitride based devices, which is discussed in Section 3.4.
HiPIMS is a pulsed sputtering method characterized by very high peak power (>10 MW) in short pulse on period in the range of 50 to 200 µs. Research in HiPIMS gained traction after the publication of the seminal paper by Kouznetsov in 1999 [65]. HiPIMS has the advantage of producing film with better adhesion, high density, low roughness, and no droplet formation. The high-density plasma produced in HiPIMS results in the ionization of the sputtered metal atom, up to 90% depending on the metal and applied power per pulse. Using metal ions is more advantageous compared to using inert gas ions for enhancing adatom mobility. Inert gas ions can reside at the interstitial sites of the film, resulting in residual compressive stress, whereas in the case of metal ions, the metal is incorporated into the lattice of the growing film [66].
The growth of GaN using a HiPIMS system was first demonstrated by Junaid et al. [10], who grew a GaN(0001) film directly on c-plane sapphire. The growth was performed using a liquid Ga target under a mixture of N 2 /Ar gas. TEM results show that the thin film is composed of two domains with different strains, referred to as the less strained (A) and more strained (B) domains, shown in Figure 5. The less strained domain is almost fully relaxed, with superior structural and optical properties, with a rocking curve FWHM of 885 arcsec and low temperature band edge luminescence at 3.37 eV with an FWHM of 10 meV. On the other hand, the more strained domain has~14 times higher isotropic strain components, caused by the higher number of point defects density caused by ion bombardment during growth. The more strained domain shows a band edge luminescence at 3.48 eV with a broader FWHM of~14 meV due to the lower material quality. Based on these results, a good control of stable HiPIMS growth condition is necessary in order to produce high-quality GaN film. For example, a substrate bias can be used to slow down the incoming ions, or by adjusting timing of the bias pulse to filter out which ions are allowed to reach the substrate.
higher isotropic strain components, caused by the higher number of point defects density caused by ion bombardment during growth. The more strained domain shows a band edge luminescence at 3.48 eV with a broader FWHM of ~14 meV due to the lower material quality. Based on these results, a good control of stable HiPIMS growth condition is necessary in order to produce high-quality GaN film. For example, a substrate bias can be used to slow down the incoming ions, or by adjusting timing of the bias pulse to filter out which ions are allowed to reach the substrate.

Doping and Fabrication of III-nitride Devices Grown Using MSE
In order for MSE to realize a functional GaN-based device, an efficient p-type and n-type doping scheme must be used. The growth of doped GaN using MSE has several advantages compared to using MOCVD. Due to the nonequilibrium growth condition and the relatively lower process temperature, very high carrier concentration can be achieved. As the MSE growth chamber is maintained under ultra-high vacuum environment, the concentration of carbon and hydrogen can be suppressed. Thus, p-type doping is only compensated lightly, eliminating the need for post-growth annealing for dopant activation [67].
Typically, Si and Ge are used as the dopant to achieve n-type conductivity within MSE-grown GaN [55,57,58]. Ueno et al. [55] has grown a degenerate n-GaN thin film with a carrier density as high as 5.1 × 10 20 cm −3 by using Ge as the dopant using a PSD system. They have also demonstrated a Sidoped n-type GaN film which exhibits a room temperature resistivity value of as low as 0.16 mΩ cm with electron concentration of 3.9 × 10 20 cm −3 and mobility as high as 100 cm 2 V −1 s −1 , which is almost as low as those of transparent conductive oxide (TCO) such as indium tin oxide (ITO). Under a nitrogen-rich condition, the compensation ratio is lower as the incorporation of Si or Ge at the N sites is suppressed. In addition, the concentration of Ga vacancies decreases at lower growth temperature due to the lack of energy for defect formation. P-type doping using Mg as the dopant has also been demonstrated by PSD [13,56,63]. Fudetani et al. [56] obtained a highly tunable p-type carrier concentration over two orders of magnitude. A hole concentration from 2.8 × 10 16 cm −3 to 2.7 × 10 18 cm −3 was achieved, comparable to p-GaN grown using MOCVD. Optical characterization of the GaN shows negligible green luminescence and blue luminescence band associated with p-dopant compensation and hydrogen defects. The minimum room temperature resistivity achieved was as low as 0.6 Ω cm, with compensation ratio as low as 5%.

Doping and Fabrication of III-Nitride Devices Grown Using MSE
In order for MSE to realize a functional GaN-based device, an efficient p-type and n-type doping scheme must be used. The growth of doped GaN using MSE has several advantages compared to using MOCVD. Due to the nonequilibrium growth condition and the relatively lower process temperature, very high carrier concentration can be achieved. As the MSE growth chamber is maintained under ultra-high vacuum environment, the concentration of carbon and hydrogen can be suppressed. Thus, p-type doping is only compensated lightly, eliminating the need for post-growth annealing for dopant activation [67].
Typically, Si and Ge are used as the dopant to achieve n-type conductivity within MSE-grown GaN [55,57,58]. Ueno et al. [55] has grown a degenerate n-GaN thin film with a carrier density as high as 5.1 × 10 20 cm −3 by using Ge as the dopant using a PSD system. They have also demonstrated a Si-doped n-type GaN film which exhibits a room temperature resistivity value of as low as 0.16 mΩ cm with electron concentration of 3.9 × 10 20 cm −3 and mobility as high as 100 cm 2 V −1 s −1 , which is almost as low as those of transparent conductive oxide (TCO) such as indium tin oxide (ITO). Under a nitrogen-rich condition, the compensation ratio is lower as the incorporation of Si or Ge at the N sites is suppressed. In addition, the concentration of Ga vacancies decreases at lower growth temperature due to the lack of energy for defect formation. P-type doping using Mg as the dopant has also been demonstrated by PSD [13,56,63]. Fudetani et al. [56] obtained a highly tunable p-type carrier concentration over two orders of magnitude. A hole concentration from 2.8 × 10 16 cm −3 to 2.7 × 10 18 cm −3 was achieved, comparable to p-GaN grown using MOCVD. Optical characterization of the GaN shows negligible green luminescence and blue luminescence band associated with p-dopant compensation and hydrogen defects. The minimum room temperature resistivity achieved was as low as 0.6 Ω cm, with compensation ratio as low as 5%. The superior quality is thought to arise from the suppression of nitrogen vacancies formation due to the dense nitrogen radicals contained within the plasma and the low growth temperature.
In addition to MSE-growth of doped GaN using separate dopant sources, it is also possible to fabricate cermet sputtering target with the dopant directly incorporated within the target. Both GaN n-doping using Sn [68] and p-doping [69][70][71][72][73] using Mg and Zn have been demonstrated. The dopant concentration can be controlled either by varying the dopant composition inside the cermet target, or by adjusting the substrate temperature during deposition. Although this method is advantageous in terms of cost, as it does not require the installation of additional effusion cell or magnetron to introduce dopants, its capability for dopant concentration control is not as flexible.
The low process temperature in sputtering is also beneficial for the growth and fabrication of novel devices. It allows for the deposition of III-nitride on temperature-sensitive substrates, such as polyimide [74]. At lower temperature, more indium can be incorporated, allowing for the growth of full color LED structure [11]. An example of full color LEDs fabricated from III-nitride deposited on a flexible metal foil is shown in Figure 6. Another advantage of the low growth temperature in PSD is the reduced strain and increased critical thickness for GaN films. Watanabe et al. [60] has demonstrated an AlGaN/GaN HEMT structure on a Si(110) substrate with a much higher critical thickness compared to MOCVD-grown structure. The heterostructure exhibits a two dimensional electron gas at the interface, with a mobility of 1360 cm 2 /Vs and a sheet carrier density of 1.3 × 10 13 cm −2 . The fabricated device and its output characteristics are shown in Figure 7.
The superior quality is thought to arise from the suppression of nitrogen vacancies formation due to the dense nitrogen radicals contained within the plasma and the low growth temperature.
In addition to MSE-growth of doped GaN using separate dopant sources, it is also possible to fabricate cermet sputtering target with the dopant directly incorporated within the target. Both GaN n-doping using Sn [68] and p-doping [69][70][71][72][73] using Mg and Zn have been demonstrated. The dopant concentration can be controlled either by varying the dopant composition inside the cermet target, or by adjusting the substrate temperature during deposition. Although this method is advantageous in terms of cost, as it does not require the installation of additional effusion cell or magnetron to introduce dopants, its capability for dopant concentration control is not as flexible.
The low process temperature in sputtering is also beneficial for the growth and fabrication of novel devices. It allows for the deposition of III-nitride on temperature-sensitive substrates, such as polyimide [74]. At lower temperature, more indium can be incorporated, allowing for the growth of full color LED structure [11]. An example of full color LEDs fabricated from III-nitride deposited on a flexible metal foil is shown in Figure 6. Another advantage of the low growth temperature in PSD is the reduced strain and increased critical thickness for GaN films. Watanabe et al. [60] has demonstrated an AlGaN/GaN HEMT structure on a Si(110) substrate with a much higher critical thickness compared to MOCVD-grown structure. The heterostructure exhibits a two dimensional electron gas at the interface, with a mobility of 1360 cm 2 /Vs and a sheet carrier density of 1.3 × 10 13 cm −2 . The fabricated device and its output characteristics are shown in Figure 7.

Growth of GaN Nanorods Using MSE
GaN nanorods hold several advantages compared to their thin film counterpart. Due to their high surface to volume ratio, GaN nanorods can grow on substrates with lattice and thermal mismatch while maintaining high crystal quality. The threading dislocations commonly associated

Growth of GaN Nanorods Using MSE
GaN nanorods hold several advantages compared to their thin film counterpart. Due to their high surface to volume ratio, GaN nanorods can grow on substrates with lattice and thermal mismatch while maintaining high crystal quality. The threading dislocations commonly associated with mismatch between the GaN and substrate can be eliminated via the nanorod's sidewalls [77]. In addition, the three-dimensional structure allows the nanorod to relieve strain, resulting in better carrier recombination efficiency for quantum confined structures [78].
In contrast to the vapor-liquid-solid (VLS) mechanism of catalyst-driven GaN nanorod growth in CVD systems [79,80], nanorods can grow without the need for catalysts in MSE. MSE-grown nanorods can grow either through a self-assembled growth mode or in a selective-area growth (SAG) mode. The former is driven by the anisotropy in GaN surface facet, and is analogous to self-assembled nanorods grown using MBE. On the other hand, the latter is driven by anisotropy of sticking coefficient between the substrate and SAG mask material, and has been observed in both MBE and MOCVD-based systems. Both methods have been subjects of many research activities, resulting in functional devices. The mechanism behind both the self-assembled and SAG growth mode is elaborated in the following section.

Self-Assembled GaN Nanorods Grown Using MSE
The self-assembled growth of GaN nanorods is attractive as it is capable of producing a large number of nanorods with high crystal quality without the need of complex growth preparation. Interfacial studies of GaN nanorods grown on silicon shows that the nanorods grow on a thin layer of amorphous silicon nitride (SiN x ), indicating that the nanorods can grow even with no epitaxial relationship with the substrate [81,82].
The main driving force for spontaneous formation of GaN nanorods is the anisotropy of sticking coefficient and chemical potential between the various GaN facets during the nanorod formation. This results in the self-assembled nanorods growing in the 0001 direction. The model for spontaneous nanorod growth under nitrogen-rich condition has been proposed by Ristic et al. and Bertness et al. [83,84] for MBE-grown GaN nanorods. The spontaneous growth of nanorods follows the Volmer-Weber growth mechanism, where the nanorods grow spontaneously without the formation of wetting layer prior to 3D growth. The nanorod growth begins with the nucleation of GaN islands by the impinging adatoms. During the nucleation process, the number of islands on the substrate surface increases until the saturation of nucleation sites occurs. To be energetically stable, the dimension of the islands must be bigger than a certain critical size. Islands with dimension below the critical size will decay either through desorption or diffusion to other islands. On the other hand, island above the critical size are considered stable and will continue growing by incorporating diffusion atoms. Under heavily nitrogen-rich condition, the coalescence of the islands is blocked due to the preferential incorporation of metal atoms on the top side of the island, leading to vertical growth. The 3D growth of nanorod then occurs through two main mechanisms: (i) direct incorporation of Ga atoms impinging on the apex and (ii) diffusion of Ga atoms arriving on either the substrate surface or the nanorod's sidewalls towards the top. As the (0001) top surface is more energetically favorable compared to the nonpolar sidewalls, growth in the axial direction is faster than in the lateral direction.
Junaid et al. [15] have demonstrated the first growth of self-assembled GaN nanorods using MSE. The GaN nanorods were grown on a Si(111) substrate. Under MSE growth condition, the N 2 -containing plasma is enough to supply the nitrogen-rich condition required to drive vertical nanorod growth without the need of other nitrogen sources such as ammonia or atomic nitrogen. From the XRD pole figure measurement, a GaN (0001)//Si(111) and GaN [11][12][13][14][15][16][17][18][19][20]//Si[110] epitaxial relationship can be observed. The density, length, and diameter of MSE-grown self-assembled nanorods are controlled by the growth conditions, including the substrate temperature, chamber pressure, and partial gas pressures P Ar and P N 2 [15,85].
The effect of the MSE working pressure on the nanorod growth is investigated by changing the pressure from 5 mTorr to 20 mTorr under a purely N 2 gas ambient. When operating at a low pressure of 5 mTorr, the Ga flux is relatively high and unidirectional. The sputtered Ga atoms have longer mean free path, allowing them to retain their kinetic energy, which is typically a few eV. This leads to growth of nanorods with a relatively larger diameter from the nucleation stage. As the pressure is increased, gas phase scattering results in a more diffuse Ga flux, with part of the Ga atoms lost to the chamber walls. The reduced Ga flux and increased nitrogen saturation results in smaller nucleation size, and eventually smaller nanorod diameters.
The low-temperature micro-photoluminescence (µ-PL) spectra of the nanorods grown at different pressure are shown in Figure 8. The sample grown at 5 mTorr shows a broad spectrum with a band edge luminescence at 3.475 eV and a line width of~35 meV. At increased pressure, several lines associated with free A exciton recombination (X A ), free B exciton recombination (X B ), and neutral donor bound exciton transitions (D 0 X A ) and (D 0 X B ) can be seen. The excitonic lines become more visible with increasing N 2 pressure, indicating an increase in crystal quality and enhanced optical properties. The full-width at half maximum (FWHM) of the D 0 X A line of the GaN nanorods grown at 20 mTorr is as narrow as~1.7 meV, which is comparable with a high-quality CVD-grown 2 µm thick GaN layer (~0.5 meV) [86], further demonstrating the high crystal quality of MSE-grown nanorods.
The effect of the MSE working pressure on the nanorod growth is investigated by changing the pressure from 5 mTorr to 20 mTorr under a purely N2 gas ambient. When operating at a low pressure of 5 mTorr, the Ga flux is relatively high and unidirectional. The sputtered Ga atoms have longer mean free path, allowing them to retain their kinetic energy, which is typically a few eV. This leads to growth of nanorods with a relatively larger diameter from the nucleation stage. As the pressure is increased, gas phase scattering results in a more diffuse Ga flux, with part of the Ga atoms lost to the chamber walls. The reduced Ga flux and increased nitrogen saturation results in smaller nucleation size, and eventually smaller nanorod diameters.
The low-temperature micro-photoluminescence (µ-PL) spectra of the nanorods grown at different pressure are shown in Figure 8. The sample grown at 5 mTorr shows a broad spectrum with a band edge luminescence at 3.475 eV and a line width of ~35 meV. At increased pressure, several lines associated with free A exciton recombination (XA), free B exciton recombination (XB), and neutral donor bound exciton transitions (D 0 XA) and (D 0 XB) can be seen. The excitonic lines become more visible with increasing N2 pressure, indicating an increase in crystal quality and enhanced optical properties. The full-width at half maximum (FWHM) of the D 0 XA line of the GaN nanorods grown at 20 mTorr is as narrow as ~1.7 meV, which is comparable with a high-quality CVD-grown 2 µm thick GaN layer (~0.5 meV) [86], further demonstrating the high crystal quality of MSE-grown nanorods. Junaid et al. [85] has also shown that the dilution of the N2 gas with Ar affects the growth condition. In their experiment, nanorods are grown on a Si(111) substrate at 1000 °C while PN2 and PAr is varied from PN2 = 5 mTorr and PAr = 0 mTorr to PN2 = 1 mTorr and PAr = 4 mTorr. As shown in Figure 8. PL spectra from MSE-grown GaN nanorods on Si(111) under purely N 2 ambient with varying chamber pressures [15]. Reproduced with permission from Elsevier Publishing.
Junaid et al. [85] has also shown that the dilution of the N 2 gas with Ar affects the growth condition. In their experiment, nanorods are grown on a Si(111) substrate at 1000 • C while P N 2 and P Ar is varied from P N 2 = 5 mTorr and P Ar = 0 mTorr to P N 2 = 1 mTorr and P Ar = 4 mTorr. As shown in Figure 9, the introduction of Ar (P N 2 = 4.5 mTorr and P Ar = 0.5 mTorr) leads to an increase of nanorod nucleation density and increased growth rate. This is caused by improved metal sputtering yield by enabling momentum transfer from the sputter gas and also by preventing target poisoning. However, further dilution by Ar will result in the transition from nitrogen-rich to metal-rich condition, reducing the diffusion driven growth mechanism. This is indicated by the reduction of average rod length and increase of average rod diameter. Increasing P Ar beyond 2.5 mTorr will result in the formation of continuous GaN film with large columnar grains instead of well-separated nanorods.  . SEM micrographs showing cross section (left) and plan views (right) from self-assembled GaN nanorod samples grown by MSE at (a,b) P N 2 = 5 mTorr, (c,d) P N 2 = 4.5 mTorr, (e,f) P N 2 = 3.5 mTorr, (g,h) P N 2 = 2.5 mTorr, and (i,j) P N 2 = 1 mTorr pressure on Si(111) substrates [85]. Reproduced with permission from MDPI publishing. The ratio of P N 2 and P Ar also affects the formation of structural imperfections within the nanorods. Under different partial pressures, a PL peak at~3.43 eV associated with basal plane stacking faults (BSFs) can be observed [85,87]. BSFs can be understood as the insertion of cubic zincblende structure inside the wurtzite matrix of GaN nanorods. Currently, the mechanism that results in the formation of BSFs is still debatable. Zhou et al. [88] attributed their presence in MBE-grown nanorods to process instabilities during the growth process. BSFs were also observed in CVD-grown GaN nanowires using Au as catalyst, where their formation is attributed to variations in the supersaturation during growth [89]. Generally, the formation of BSFs is not desirable, as it reduces the quantum efficiency by introducing type II multiple quantum well (MQW)-like structure within the nanorod structure. The BSF-related PL emission spectra are shown in Figure 10. This emission shows linear polarization, with almost 50% degree of polarization [90]. By using an optimized growth condition (P N 2 = 2.5 mTorr and P Ar = 2.5 mTorr), Junaid et al. has grown nanorods free of BSFs, which confirms that BSFs can be eliminated under proper growth condition. Figure 10. (a) Cross-sectional high-resolution transmission electron microscope (HRTEM) images from a self-assembled GaN nanorod grown by MSE showing BSFs. Near band-edge PL spectra measured at various temperatures for a single GaN NR with (b) and without (c) BSFs. The measurement geometry is shown in the inset. Polarization angle vs. PL intensity for a single NR measured at 5 K (d) and at room temperature (e). Inset shows the excitation and detection geometry. The BSF emission (3.43 eV) and donor bound exciton (3.48 eV) are shown by circles and squares, respectively. The PL intensity at room temperature is taken at 3.42 eV [90]. Reprinted with permission from AIP publishing.

Selective-Area Growth of Nanorods Using MSE
As self-assembled nanorods nucleate and grow in a random manner, each nanorod can have variations in properties such as of length, diameter, alignment, composition, and doping concentration. This can lead to undesirable effects, such as the dispersion of electrical and optical properties. Furthermore, the variation in size would lead to reduced injection efficiency when the nanorods are fabricated into a single device [91,92]. Selective-area growth (SAG) of GaN nanorods addresses this issue by controlling the diameter and spatial location of the nanorods, resulting in better spatial uniformity. This opens up various novel applications, such as the fabrication of photonics crystal [93], monolithic integration of multicolor LED [94][95][96], and controlled coalescence of nanorods [97].
In SAG mode, the growth substrate is covered with a mask material (typically Ti, SiN, or SiO2) with an array of nanoholes opening. In order to achieve successful SAG mode, there must be a good selectivity between the mask and the growth substrate, i.e., the nanorods only nucleate within the mask opening. This selectivity is driven by various factors, including the sticking coefficient, mask material, Ga adatom diffusion length, desorption rates, and the geometry of the nanoholes [98]. The selectivity between the mask and the substrate can be improved by increasing the growth process temperature, with Ga desorption rate setting the upper temperature limit.
One of the most commonly used techniques for the fabrication of nanohole arrays is e-beam lithography (EBL) [99]. The technique has achieved sub-10 nm resolution for resist patterning, which is enough for most nanoscale lithography feature demands. EBL has been used for SAG patterning of III-nitride nanorods array, due to its capability to fully control the geometric features in the Figure 10. (a) Cross-sectional high-resolution transmission electron microscope (HRTEM) images from a self-assembled GaN nanorod grown by MSE showing BSFs. Near band-edge PL spectra measured at various temperatures for a single GaN NR with (b) and without (c) BSFs. The measurement geometry is shown in the inset. Polarization angle vs. PL intensity for a single NR measured at 5 K (d) and at room temperature (e). Inset shows the excitation and detection geometry. The BSF emission (3.43 eV) and donor bound exciton (3.48 eV) are shown by circles and squares, respectively. The PL intensity at room temperature is taken at 3.42 eV [90]. Reprinted with permission from AIP publishing.
Taking advantage of the capability of GaN nanorods to relief strain, Serban et al. [16] have grown self-assembled nanorods on multiple functional and cost effective templates and substrates, including Si, SiC, TiN/Si, ZrB 2 /Si, ZrB 2 /SiC, Mo, and Ti. The nanorods show high purity and quality, as evidenced by the strong band-edge emission from room-temperature cathodoluminescence spectroscopy. The work demonstrates the capability of MSE-grown self-assembled nanorods as a simple and affordable fabrication option, making it attractive for future device fabrications.

Selective-Area Growth of Nanorods Using MSE
As self-assembled nanorods nucleate and grow in a random manner, each nanorod can have variations in properties such as of length, diameter, alignment, composition, and doping concentration. This can lead to undesirable effects, such as the dispersion of electrical and optical properties. Furthermore, the variation in size would lead to reduced injection efficiency when the nanorods are fabricated into a single device [91,92]. Selective-area growth (SAG) of GaN nanorods addresses this issue by controlling the diameter and spatial location of the nanorods, resulting in better spatial uniformity. This opens up various novel applications, such as the fabrication of photonics crystal [93], monolithic integration of multicolor LED [94][95][96], and controlled coalescence of nanorods [97].
In SAG mode, the growth substrate is covered with a mask material (typically Ti, SiN, or SiO 2 ) with an array of nanoholes opening. In order to achieve successful SAG mode, there must be a good selectivity between the mask and the growth substrate, i.e., the nanorods only nucleate within the mask opening. This selectivity is driven by various factors, including the sticking coefficient, mask material, Ga adatom diffusion length, desorption rates, and the geometry of the nanoholes [98]. The selectivity between the mask and the substrate can be improved by increasing the growth process temperature, with Ga desorption rate setting the upper temperature limit.
One of the most commonly used techniques for the fabrication of nanohole arrays is e-beam lithography (EBL) [99]. The technique has achieved sub-10 nm resolution for resist patterning, which is enough for most nanoscale lithography feature demands. EBL has been used for SAG patterning of III-nitride nanorods array, due to its capability to fully control the geometric features in the nanometer scale [94,97,[100][101][102][103]. However, EBL is not suitable for large-scale production due the very low throughput compared to conventional photolithography techniques. As an alternative to EBL, we have demonstrated SAG patterning and GaN nanorod growth using nanosphere lithography (NSL) and focused ion beam lithography (FIBL).
NSL is a technique where a monolayer of nanospheres with diameters ranging from sub-100 nm to 1 µm is deposited on a substrate surface. The nanospheres can be made of silica or polystyrene, and nanosphere deposition techniques may include spin-coating [104] or scooping [105]. After nanosphere deposition, a mask material is deposited on the substrate. Because of the shadowing effect by the nanospheres, an array of nanohole is formed. The final step is the removal of the nanospheres, either by chemical etching or by physical removal. NSL-based SAG growth of III-nitride nanorods has been demonstrated by using MOCVD, showing very high uniformity and density of nanorods [106,107].
The first MSE-based SAG growth of GaN nanorods was demonstrated by Serban et al. using a combination of 150 nm wide nanoholes array and TiN x mask formed using NSL [17]. As no oxide removal was performed prior to GaN deposition, the nanorods grow directly on a layer of native oxide. Selectivity between TiN x and the native oxide was achieved by performing the growth at 950 • C with an N 2 partial pressure of 20 mTorr.
By performing a series of growth periods with varying deposition time, a time lapse of MSE-SAG nanorod growth mechanism is obtained, as shown in Figure 11. The growth stage of MSE SAG-nanorods is as follows. (1) The formation of a polycrystalline GaN wetting layer within the nanohole; (2) formation of 3D GaN nuclei inside the nanohole; (3) coarsening and coalescence of the nuclei into a single nanorod; (4) growth of a single nanorod driven by directly impinging adatoms and surface diffusion under catalyst-free conditions; and (5) quasi-equilibrium growth, where adatom diffusion is suppressed due to self-shadowing effects.
By taking advantage of the well-formed facets of the nanorods during the SAG process, Serban et al. has also demonstrated lasing characteristics from the nanorods, shown in Figure 12. The sample was excited using a µ-PL setup with a 50 µW 266 nm Nd:YAG laser as the excitation source. The lasing cavity was confirmed to be based on a Fabry-Perot type, with a Q factor of 548. Calculation based on wavelength-dependent refractive index and Fabry-Perot cavity in the bottom part of Figure 12 confirms that the peak of the lasing mechanism indeed comes from the Fabry-Perot lasing mode instead of random lasing or whispering gallery mode.
Despite the advantages of NSL for fabricating a large array of nanoholes with uniform diameter, the method is unable to fabricate nanoholes with precise geometric placement for applications such as photonic crystals. FIBL is an alternative nanoscale lithography technique that is faster and less expensive compared to EBL, but is capable of forming patterns with accurate control over nanoholes opening size and positions. Serban et al. [108] demonstrated the first use of FIBL for MSE growth of nanorods on a Si(001) substrate with a TiN x mask. An example of a 14 × 14 GaN nanorod array grown using a combination of MSE and FIBL-based SAG process is shown in Figure 13.
SAG nanorod growth mechanism is obtained, as shown in Figure 11. The growth stage of MSE SAGnanorods is as follows. (1) The formation of a polycrystalline GaN wetting layer within the nanohole; (2) formation of 3D GaN nuclei inside the nanohole; (3) coarsening and coalescence of the nuclei into a single nanorod; (4) growth of a single nanorod driven by directly impinging adatoms and surface diffusion under catalyst-free conditions; and (5) quasi-equilibrium growth, where adatom diffusion is suppressed due to self-shadowing effects.   [17]. Reproduced with permission from Nature Publishing Group.
Appl. Sci. 2020, 9, x FOR PEER REVIEW 19 of 28 By taking advantage of the well-formed facets of the nanorods during the SAG process, Serban et al. has also demonstrated lasing characteristics from the nanorods, shown in Figure 12. The sample was excited using a µ-PL setup with a 50 µW 266 nm Nd:YAG laser as the excitation source. The lasing cavity was confirmed to be based on a Fabry-Perot type, with a Q factor of 548. Calculation based on wavelength-dependent refractive index and Fabry-Perot cavity in the bottom part of Figure  12 confirms that the peak of the lasing mechanism indeed comes from the Fabry-Perot lasing mode instead of random lasing or whispering gallery mode. Bottom plot shows the dispersion of wavelength dependent refractive index [17]. Reproduced with permission from Nature Publishing Group.
Despite the advantages of NSL for fabricating a large array of nanoholes with uniform diameter, the method is unable to fabricate nanoholes with precise geometric placement for applications such as photonic crystals. FIBL is an alternative nanoscale lithography technique that is faster and less expensive compared to EBL, but is capable of forming patterns with accurate control over nanoholes opening size and positions. Serban et al.
[108] demonstrated the first use of FIBL for MSE growth of nanorods on a Si(001) substrate with a TiNx mask. An example of a 14 × 14 GaN nanorod array grown using a combination of MSE and FIBL-based SAG process is shown in Figure 13. In order to achieve a proper nanorod growth using FIBL, the patterning parameters, including ion beam milling current and milling time, must be optimized to minimize substrate damage. The effect of milling current dose and duration on the nanorod growth is summarized in Figure 14. It is found that high milling current generally results in the removal of the mask layer, resulting in a selfassembled growth mode of nanorods directly on the substrate [108]. The high milling current also results in the roughening of the substrate surface which creates a high density of tilted nanorods. Ion beam-induced damage will result in the growth of multiple and tilted nanorods inside a single nanohole opening. This results in the formation of irregular and nonuniform nanostructures. Therefore, the minimization of mask and substrate damage through proper beam control is vital in growing uniform and well-defined nanorod arrays using MSE.

Conclusions and Future Outlook
In this paper, we have reviewed the progress made in MSE growth of GaN thin film and nanorods. By using a liquid Ga target combined with strict vacuum condition, MSE is able to produce high-quality, semiconductor-grade GaN films. The capability of the sputtering system for large-scale production makes MSE highly relevant for the semiconductor industry. By taking advantage of high- In order to achieve a proper nanorod growth using FIBL, the patterning parameters, including ion beam milling current and milling time, must be optimized to minimize substrate damage. The effect of milling current dose and duration on the nanorod growth is summarized in Figure 14. It is found that high milling current generally results in the removal of the mask layer, resulting in a self-assembled growth mode of nanorods directly on the substrate [108]. The high milling current also results in the roughening of the substrate surface which creates a high density of tilted nanorods. Ion beam-induced damage will result in the growth of multiple and tilted nanorods inside a single nanohole opening. This results in the formation of irregular and nonuniform nanostructures. Therefore, the minimization of mask and substrate damage through proper beam control is vital in growing uniform and well-defined nanorod arrays using MSE.
Appl. Sci. 2020, 9, x FOR PEER REVIEW 20 of 28 Figure 13. An array of GaN nanorods grown by MSE using FIBL-based SAG process on Si(001) substrate and TiNx mask. Author's work, unpublished.
In order to achieve a proper nanorod growth using FIBL, the patterning parameters, including ion beam milling current and milling time, must be optimized to minimize substrate damage. The effect of milling current dose and duration on the nanorod growth is summarized in Figure 14. It is found that high milling current generally results in the removal of the mask layer, resulting in a selfassembled growth mode of nanorods directly on the substrate [108]. The high milling current also results in the roughening of the substrate surface which creates a high density of tilted nanorods. Ion beam-induced damage will result in the growth of multiple and tilted nanorods inside a single nanohole opening. This results in the formation of irregular and nonuniform nanostructures. Therefore, the minimization of mask and substrate damage through proper beam control is vital in growing uniform and well-defined nanorod arrays using MSE.

Conclusions and Future Outlook
In this paper, we have reviewed the progress made in MSE growth of GaN thin film and nanorods. By using a liquid Ga target combined with strict vacuum condition, MSE is able to produce high-quality, semiconductor-grade GaN films. The capability of the sputtering system for large-scale production makes MSE highly relevant for the semiconductor industry. By taking advantage of high-

Conclusions and Future Outlook
In this paper, we have reviewed the progress made in MSE growth of GaN thin film and nanorods. By using a liquid Ga target combined with strict vacuum condition, MSE is able to produce high-quality, semiconductor-grade GaN films. The capability of the sputtering system for large-scale production makes MSE highly relevant for the semiconductor industry. By taking advantage of high-flux low-energy ion bombardment, adatom mobility on the substrate can be enhanced, leading to improved crystallinity. A GaN film with x-ray rocking curve (0002) reflection FWHM of 1054 arcsec and threading dislocation density of the order of ≤ 10 10 cm −2 grown on sapphire using DC-MSE has been demonstrated. The material quality can be improved further by using a GaN template as an MSE growth substrate, resulting in a GaN film with threading dislocation density of~1.7 × 10 8 cm −2 . Furthermore, effective material doping using PSD has been demonstrated, with electron concentration as high as 5.1 × 10 20 cm −3 for n-GaN and hole concentration of 2.7 × 10 18 cm −3 for p-GaN. This doping capability enables the fabrication of various functioning devices, including LEDs and transistors.
The use of HiPIMS system was also discussed. Compared to PSD, the high energy pulse used in HiPIMS enables the ionization of metal ions, resulting in momentum transfer to the adatoms without the formation of interstitials within the film. To date, the research of GaN material growth using HiPIMS is still in its infancy. The high energy ions produced during the process is shown to induce damage in the growing GaN film, resulting in the formation of GaN domains with high residual stress. Future work is needed to improve the growth condition, for example, by applying a substrate bias to act as a retardant to the ions, or adjusting the timing of the substrate bias as a filter to select which ions are allowed to reach the substrate.
Beyond the growth of thin films, the growth of self-assembled and SAG GaN nanorods on various functional substrates using MSE has also been demonstrated. A range of process conditions, including substrate temperature, process gas composition, and total chamber pressure, and their effects on the growth of the nanorods, have been investigated. Under an optimized growth condition (T s = 1000 • C, P N 2 = 2.5 mTorr, and P Ar = 2.5 mTorr), the nanorods can grow on Si(111) free from structural defects. In order to achieve structural uniformity between the nanorods, we employed multiple SAG methods using Si(001) substrate and TiN x mask. The first method, based on NSL, enables the growth of identical nanorods within a large area. It is found that there is a minimum substrate temperature required in order to achieve selective growth between the mask and the substrate. The second method, based on FIBL, enables the exact geometrical arrangement of nanorods. However, the ion beam current and exposure time must be tuned in order to prevent substrate and mask damage.
Despite the progress in MSE-grown GaN, several issues need to be addressed. The shift of target condition from pure metal to the poisoned mode under high nitrogen partial pressure can change the deposition condition over time. Therefore, a pre-sputtering would be beneficial in order to stabilize the target condition before deposition begins. Another important issue is the possibility of material damage due to high-energy ions. Currently, the flux and energy distribution of arriving ions and sputtered species from a Ga target is still not explored yet. Furthermore, these numbers may change over the course of the deposition run depending on the condition of the Ga target (solid, liquid, fresh, or poisoned). Detailed characterization of plasma using liquid Ga is important in order to set the proper growth condition and minimize point defects, as this will ultimately affect the electrical and properties of GaN.
The capability of MSE to produce high quality GaN thin films and nanorods within a large substrate area makes it an interesting alternative for GaN epitaxy. With better understanding in process physics and improvements in the material quality in the future, MSE can become one of the main technologies for the growth GaN-based materials and devices, moving it from an academic research area to practical industrial applications.