Growth mechanisms of hBN crystalline nanostructures with rf sputtering deposition: challenges, opportunities, and future perspectives

Most hBN nanostructures were fabricated using the chemical method. However, growing by the physical method also has many advantages, they are easy to synthesize this material on a large area with up- scaling setups. Even two-dimensional hexagonal boron nitride is similar to graphene structure, however there is a little work referring to the fabrication process of this material. Hence, a sufficiently detailed report on physically fabricated hBN materials is essential. This review analyzes the results that we have studied over the past ten years with the synthesis and fabrication of this material using physical vapor deposition - RF sputtering, incorporation with other techniques, strongly emphasized on growth mechanisms of this material.


Introduction
The chemical composition of the boron nitride (BN) compound consists of equal numbers of atoms B and N, (B 3 N 3 )n [1][2][3][4]. This compound exists in various crystal structures such as hexagonal BN (hBN), rhombic BN (rBN), wurtzite BN (wBN) and cubic BN (cBN), depending on the conditions of the crystal structure processing. Each type of structure has certain advantages and disadvantages. The hBN structure is the softest and most stable, while cBN is the hardest material structure among the said BN phases [5][6][7][8]. With such diverse properties, BN can be used in various applications such as a lubricant in equipment requiring high chemical and thermal stability [9]. By adjusting the fabrication conditions, the structure of BN can be obtained in many forms such as nanotubes, nanosheets, nanowalls and nano-cocoons [1,3,10,11]. Hence, explicitly understanding the formation of each BN phase for a typical fabrication method, will open new possibilities for their applications, which is one of fundamental research tasks.
BN is a binary compound synthesized from elements in columns 13 (group III) and 15 (group V) of the periodic table. It is isoelectronic where B − and N + ions have the same number of electrons (1s 2 2s 2 2p 2 ). BN is a light binary compound at which the B and N atomic numbers (Z) are 5 and 7, respectively, leading to similarities in physical and chemical properties of C-based compounds. For example, cBN and diamond have quite similar crystal structures such as lattice constant, hardness, wide band gap, and high thermal conductivity [6,9,11]. Similar to graphite, hBN has B-N and B=N bonds, which have similar properties to the C-C and C=C bonds in graphite, especially since both materials contain the van der Waals bonds between the sheets. However, one significant difference is that hBN has mixed covalent-ionic bonds, while graphite has only covalent bonds in the plane of each sheet. Because of its higher ionization, the hardness of hBN is lower than that of graphite. In addition, BN has a higher antioxidant capacity than carbonaceous compounds due to the formation of nonvolatile boron oxide. Similar to BN, the growth of cBN and hBN mono-crystals is difficult to perform [11,12], while the graphite and diamond structures can be tuned for easier control [13][14][15][16][17][18][19][20][21].
Four BN crystal structures are depicted in figure 1. There, hBN is structurally similar to graphite, figure 1(a), with B and N atoms interspersed on the corners of a hexagon. The B and N atoms in the same plane are linked together by strong covalent bonds to form the hBN sheet, the different hBN sheets have weak interactions governed by van der Waals forces [1,3,4,12]. The distance between the B and N atoms is 1.46 Å in the hBN covalent plane, denoted by the A or B plane in figure 1(a). Between the two A and B planes there is a distance of 3.33 Å. The position of each B or N atom on one hBN plane plane can be found in another hBN one when it rotates 60°. This arrangement is the so-called ABAB stack [22]. In fact, point/line defects due to the lack of one or more N/B atoms can occur on those planes from any fabrication technique. The hBN phase structure and its properties will be discussed in more detail in the following sections.
A similar layered structure can also be stacked in a rhombohedral (rBN), figure 1(b), where the covalent plane BN slides by a 3 where a is the lattice constant of the rBN structure, and rotates 60°with respect to the original hBN plane. The rBN structure type can be synthesized using high-pressure and temperature (HPHT) methods or ion beam-assisted physical vapor deposition (PVD) techniques [23][24][25]. The third form of BN is the cubic BN (cBN or βN) in which the N atoms are located at the corners and faces of a cubic lattice unit cell, the B atoms bond to the N at one of the four corners of the BN cube, the other three N atoms bond to B at the nearest faces such that B lies at the center of a tetrahedron, as depicted in figure 1(c). This cBN structure is identical to the zinc blende (ZnS) crystal structure [3]. The fourth form of BN is the wurtzite BN or γBN where the unit cell is a superposition of a B and an N atom in a tetrahedron, figure 1(d). The bond of each atom has sp 3 hybridization [3,[26][27][28][29][30].
The transformation of those BN structural types that can occur depending on the specific synthesis conditions such as temperature (T), pressure (p) and the composition of the reactive gas present in the synthesizing process such as H 2 , N 2 , Cl 2 . The BN phases are summarized in the pressure and temperature (p − T) dependence or the p − T phase diagram as shown in figure 2. Based on crystallographic symmetry, polymorphism, phase transitions can occur in the following sequence: hBN → γN and rBN → βN [26][27][28][29][30][31]. As shown in the phase diagram, figure 2, the cBN phase is usually produced at high temperature (T > 1800 K) and high pressure (p > 4 GPa), while the hBN phase can be fabricated at high T and low p (p 100 kPa). This means that the hBN phase is thermodynamically more stable at low T and p, cBN is stable only at high p, depicted by the Bundy-Wentorf transform curve on the phase diagram. In this region, the cBN phase will form spontaneously [ [29][30][31][32]. Such a thermodynamic diagram as shown in figure 2 was obtained from equilibrium processes. In fact, we used an unbalanced RF sputtering system with the plasma creation processes, thus the above phase diagram is not completely fitted with our sample making process. However, based on this phase diagram, BN can be classified into two subtypes according to the hardness [3,[27][28][29][30][31][32][33][34]. The soft phases of BN have low density and are characterized by sp 2 bonding, figures 1(a) and (b). The hBN and rBN phases are in this soft phase. Besides, turbostratic BN (tBN) and amorphous BN (aBN) also belong to this soft category. There, the aBN has no periodic order of the B and N atoms, and tBN is partially crystallized. The formation of these phases is largely dependent on the density of defects that may be created during fabrication. The tBN and aBN phases have also been studied quite meticulously in our studies [35][36][37][38][39]. Meanwhile, the hard phases of BN are formed by sp 3 bonding and have a higher density compared to the soft ones. The wBN and cBN phases belong to this type of hard phase, figures 1(c), (d). We do not discuss much about wBN and cBN phases in this mini-review, because many systematic studies for the hard phases of BN have been published previously [3,6]. The fundamental parameters of the BN phases are listed in table 1, besides some parameters of graphite and diamond are given for comparison.
1.1. The fundamental properties of the hBN phase As mentioned above, the hBN structure has an ABAB stack type as shown in figure 1(a), the lattice of this phase is formed from the chemical composition (B 3 N 3 ) n [22,[40][41][42][43][44][45]. There, the c-axis of the ABAB stack is  Bundy-Wentorf. Therein, the boundaries of hBN and cNB phase creations were defined from experimental data, denoted as hBN ↔ cBN.Reprinted from, [33] with the permission of AIP Publishing. perpendicular to the (B 3 N 3 ) n plane. In the bulk structure, each B atom is bonded to three N atoms where the B plane is rotated 60°with respect to the A plane and moved along the [0002] direction, perpendicular to the A and B planes. Therefore, hBN is considered as a highly crystallized anisotropic layered compound with strong bonds -covalent bonds in the layered plane and weak bonds in the third dimension, parallel to the c-axis. The a, b and c parameters of the hBN unit cell are shown in figure 1(a). Due to the lack of free electrons in the hBN crystal, the perfect hBN crystal is an insulator [1,3,23]. Moreover, the layers of materials can easily slide on their planes by weak bonds, thus hBN has very high compressive resistance and can be used in lubricating technology [1,23]. Furthermore, the π-electrons around the N atom partially create ionization from the B-N bond. To take advantage of the superior properties of hBN, various types of heterostructures have been created when combined with hBN [1,12,46]. However, impurities have always been an issue in the fabrication of hBN materials. When impurities are involved in hBN crystal formation, it leads to creating other undesirable phases such as tBN or aBN [36]. These sub-phases have a local structure similar to the hBN phase but have a long-range disorder of the crystal structure. The results of our research over the past ten years on hBN materials are mainly fabricated at small sizes such as hBN thin films containing hBN nanowalls (hBN-NWs). Here, an hBN-NW is defined as a group of about a few dozen to several hundred hBN nanosheets oriented vertically relative to the substrate [47]. BN compounds are almost all produced synthetically, with the exception of the cBN structure found in nature by Dr. Q. S. Fang (2009). In most cases, hBN is synthesized using chemical reactions based on boron trioxide-B 2 O 3 or boric acid-B(OH) 3 and ammonia (NH 3 ) or urea-CO(NH 2 ) 2 in nitrogen gas [47,48]. BN can be produced in a variety of ways such as hot pressing. There, BN powder together with boron oxide are compressed at high temperature. The thermal properties of the obtained hBN largely depend on the crystallinity of the hBN during compression and heating. The thermal properties of hBN materials are significantly enhanced by annealing under pressure, which further demonstrates that annealing increases the crystallinity of the hBN phase. In addition, the chemical vapor deposition (CVD) method is commonly used with boron trichloride (BCl 3 ) and N 2 precursors [3]. In particular, boron powder reacts with nitrogen plasma at very high temperatures >5000°C, resulting in ultrafine BN structure that can be used in lubricating technology [23].
Using chemical or physical deposition methods, low-dimensional hBN structures such as nanosheets, nanowalls, nanotubes and nanoshells can be fabricated [1,12]. The structural properties of the grown hBN films depend on the density of defects existing in such films. The researchers have found that hBN thin films are generally more defected than graphite or graphene consists of sp 2 monolayers. The BN structure is a binary compound with B and N components, while graphene is composed of only C atoms [1,12,[49][50][51][52][53]. Density and composition of defects can significantly alter the physical and chemical properties of hBN layers.
For applications using hBN as the coating material, it is necessary to fabricate an hBN film with few defects. In contrast, using defects as quantum photonic centers, we need to use defects purposefully [54][55][56]. Therefore, scientists continue to exploit the interesting properties of 2D-hBN such as: (i) The chemical and mechanical stability of hBN will be useful for a number of industrial applications requiring high temperature thermal stability and flexible coating [1]. (ii) The hBN has a wide energy bandgap of 5.97eV which can be varied to emit ultraviolet light from deep levels [2]. There, a deep ultraviolet (DUV) emitter was designed and worked successfully when the hBN powder was heated to a certain temperature. This material emits photons with wavelength λ = 225 nm at a steady state of operation. The authors also show that the defect density existing in the hBN material plays an important role in the wavelength emitted from the device. This results in the wavelengths of the photons being in the DUV range of 225-400 nm [2,57]. (iii) In addition, the large area lowdimensional hBN layer is useful for electronic devices which are integrated with graphene (G) [58][59][60]. In such a device, the combination of the hBN dielectric and the high conductive G materials was realized. The 2D hBN film can be used as an insulating layer in heterostructures: metal/insulator/semiconductors (MIS) used in fieldeffect transistors (MIS-FETs) [61][62][63][64]. (iv) The hBN film has a wide band gap, this material is thus used to develop next generations of imaging detectors in the DUV-far-IR region. The hBN layer is also used in MIM tunneling structures [65][66][67]. (v) In particular, the porous hBN material containing many defects is useful for water purification applications [12,68]. Therein, porous hBN nanosheets can be produced by various techniques where they are able to absorb oils, organic solvents and dyes with an absorption weight of 33 times higher than that of its own weight, while the material is hydrophobic [61].

Emerging features of hBN structures at the nanoscale
Two-dimensional (2D) materials with large band gaps have recently been of great interest because of their emerging properties, especially 2D-hBN materials [69]. Taking advantage of the intrinsic properties of those materials, which can be used for a variety of applications such as quantum technology, quantum computing, quantum communications or highly sensitive sensing devices [70]. Various fabrication methods recently used to fabricate 2D-hBN nanostructures with high purity or intentionally induced lattice defects through the fabrication processes such as bombarding the hBN crystal structures by a certain dose of high energy electrons or ions [71]. The creation of foreign elements implanted in the crystal or forming local lattice defects will cause a change of the system energy, especially a polarization could be formed due to electron deficiency or excess. However, the photon emission mechanism of defects in wide bandgap semiconductor materials such as diamond, SiC and hBN is still unclear [72,73]. Because there are many factors affecting the emission process at which the properties of defects are highly affected by the fabrication process on the optical and magnetic behaviors of those materials and/or the interference of noises occurring around the material surfaces under study. Hence, the factors affecting luminescence at defected sites/vacancies of hBN nanostructures still need to be exploited. Herein, we analyze some experimental results obtained when studying the nucleation and growth mechanisms of hBN-NWs films deposited on different substrate materials, i.e. Si, SiN and diamond using a physical vapor deposition technique, we have created hBN-NWs films with a high crystallinity [36,39]. However, the crystallization process of hBN-NWs still has many defects in their crystal lattice due to the bombarding events inside the chamber of the physical deposition method. The crystallization of hBN-NWs has produced many defects or vacancies, the orientation of the grown hBN-NWs still perpendicular to the substrate surface leading the luminescence detection is difficult because the c-axis of the hBN crystal layers is oriented relatively parallel to the substrate plane. This makes it very difficult to probe the luminescence of defect centers using any optical methods. For example, the optical detected magnetic resonance (ODMR) method will face many challenges because the hBN-NWs films are very rough, noise will be thus the main factor to saturate the signals of ODMR [74].
Due to the given reason, we evaluated the concentration of defects indirectly through measuring qualitatively the concentration of N-H bonds with the FTIR measurement [36]. There, we assumed that the N-H bonds were generated during the crystal formation process. At boron vacancy (VB) sites, we assumed that the free H atoms would bond to the N terminated edges. This assignment was also approved by many recent simulation research papers [75][76][77]. Those data also suggested that the formation energy with N atoms is greater than that of formation with B ones. Therefore, the crystal formation through bombarding of B and N ions from the BN target of the precursor material using a physical radio frequency (RF) sputtering technique will generate more B vacancies than N ones. Moreover, the outcome data from several research groups indicate that, some impurities are believed to interact or intercalate into the lattice defected sites atV B to enhance the emission brightness of the defected centers, i.e. C and O [76,77]. Based on the simulation data recently obtained in combination with our experimental results, we will therefore discuss in detail some experimental factors affecting the growth behaviors of the deposited hBN-NWs on different substrate surfaces.
In fact, various semiconductor materials can be used as substrates such as Si, GaAs and InP [78][79][80] to enhance the crystallinity of the depositing materials at the substrate surface. Therefore, the surface properties of the substrate materials have a certain influence on the quality of the grown hBN film. During more than fifteen years of research on hBN materials using RF sputtering technique, we have used a variety of substrates to investigate the impact of substrate surface properties on the structural and optical characteristics of grown hBN films [36,39]. We temporarily divided the surface properties of the substrate into three categories. The first is, the Si substrate has a neutral surface characteristic profile, that is, Si atoms in the substrate structure are indirectly bonded to the B and N atoms in the fabrication process. The second is the diamond film grown by a CVD technique. The third is, a metallic bilayer substrate that fabricated with a combination of transition metals to lower the vaporization temperature of the given substrate relative to the higher vaporization temperature of each metal component, and also aiming at using transition metal atoms as catalysts in hBN crystal formation at the early stages of thin film deposition with our RF sputtering system [36,39,81]. Therefore, we will briefly describe in turn the types of substrate surfaces we have used in our studies in the following section: 1.3. Types of substrates used to fabricate hBN films Silicon (Si) is the main material used as the substrate for most of our studies. Herein, we can describe the two types of substrates as Si and diamond materials, as shown in figure 3. We only emphasized on those substrate surfaces because the Si is a neutral surface whilst the C edges are terminated at the NCD one. Such edges are facilitated to bond with free H atoms in our CVD/RF sputtering plasma. Moreover, the NCD substrate layer often consists of both sp 3 (cubic) and sp 2 (graphite) phases of the carbon. Therein, a small amount of the sp 2 phase is located at the boundaries of the NCD particles, while the sp 3 phase is in the cores of those NCD particles. We also used the Cr/Au substrate with its role as a catalyst agent in the crystallization process, however the Cr/ Au crystal structures are not shown in figure 3. In figure 3(a), we can see that the Si(100) surface has two atoms per cell contact to the (100) plane. Those Si atoms can be bonded indirectly to ionized particles such as H − , N + or B − . This binding is highly dependent on various parameters of atoms or ions in contact with the substrate surface and Si surface characteristics [36,39,47]. In most cases, the Si surface is treated as a neutral plane at which the ionized particles will stick after a number of physical collisions of those ionized particles and the chemical kinetics of those ions/atoms at the sole surface. The Si substrate is also capable of forming terminated H edges.
In addition to Si(100) which was used as the substrate, the artificial diamond surface is also used to grow hBN films. Diamond is an allotrope of C where the C atoms are covalently bonded to each other [13][14][15][16][17][18][19][20][21]. The C atoms are arranged in a variant of the radial settable crystal structure, known as the diamond lattice, figure 3(b). Herein, eight C atoms are located at the corner of the unit cell with bond distance of 3.56 Å. Each C atom is symmetrically bonded tetrahedrally to four other nearest atoms. The bond distance from this atom for each corner of the tetrahedron is 1.54 Å. This very short bond distance creates a very strong bonding within those C atoms, this makes diamond the hardest material in nature. The crystal faces of diamond are terminated with C edges. At those terminated edges, the C atom can bond with either a free H or O atom or different atoms from other functionalized groups to minimize the total energy of the system [6,26,32]. Therefore, the matching possibility of hBN to diamond is also a possible feature for future applications, because diamond is applicable in many different fields such as chemistry and biology. In particular, it can be used as functionalization surfaces [13], biochemical platforms [19][20][21], high charge carrier mobility devices [82]. The high thermal conductivity of diamond is also an advantage [15]. The intrinsic properties of diamond can be modified by adding impurities to the diamond lattice, i.e. turning the diamond into dielectrics, metals or superconductors [18,83]. Even so, in our studies, we have only focused on exploiting the surface terminated C edges of diamond that can bond with free H atoms/ions in our CVD deposition systems. If H atoms are temporarily bonded to C, they can form a virtual buffer at the diamond surface. This buffer layer acts as a spring during the elastic collision between B and N ions when bombarded from the BN target of the RF sputtering. Hence, this buffer layer will temporarily reduce the acceleration of B and N ions, and at the same time limit the ions in the plasma to elastically interact with the neutral substrate surface as the case of Si. This will make the crystallization process faster, resulting in a better crystallized order than using a neutral Si surface substrate [39].
In fact, crystalline diamonds include a wide variety of defect concentrations, sizes and carbon phases [13,21,84]. An important factor commonly used to determine the quality of a CVD diamond sample is the sp 2 :sp 3 ratio [6,85,86]. Single crystal diamond (SCD) exists in various types and usually concerns with the impurity of N atoms [87]. Polycrystalline CVD diamonds can be grown on a non-diamond substrate such as Si, quark or Mo with a different crystal structure. An example of a CVD nanocrystalline diamond (NCD) growth is schematically shown in figure 4. The diamond seed particles are seeded on Si, they are grown step by step as a function of time (t). The growth development is often referred to as the column model [88]. There, each diamond is considered as an individual entity and is grown in the 3D space. In the process of development, the competition of those enlarged entities occurs, and the boundaries of those particles are subsequently defined. The sp 3 phase usually comes from the core of the diamond particle, whilst a large content of sp 2 comes from the boundaries of those crystal diamond particles. Depending on their average grain sizes, one can classify CVD diamonds into three categories: microcrystalline diamonds (MCDs), nano-crystalline diamonds (NCDs) and ultra-nano-crystalline diamond (UNCDs) with average particle sizes in the range of thousands, hundreds, and several nanometers, respectively [83,84]. In our work, NCD thin films are mostly used, while the two types of MCD and UNCD are not discussed in depth. In the experiment, the particle size and sp 2 :sp 3 ratio of the NCD film can be controlled by changing the deposition parameters such as seeding density, temperature, doping . The unit cells of crystalline (a) silicon (Si) and (b) diamond (nanocrystalline diamond, NCD) materials at which the surfaces of these materials were used in most of our studies over the past ten years or more when hBN films were grown on the given substrate surfaces using a homebuilt radio frequency (RF) sputtering technique.
In addition, transition metals such as Ni, Cr, Fe, Pt, can also be used as catalysts in the crystallization process of hBN [46,89]. There, these transition metal atoms participate in the chemical formation of the first lattice cells in the hBN layer structures. However, most of the studies use transition metals in the form of a single element, in order to work out the role of each transition metal in the formation of the hBN phase with a CVD technique [1,12]. Our studies not only use transition metals as the catalyst for hBN phase formation, but transition metals were combined together into a bilayer aiming to reduce melting temperature of the bilayer substrate. This will facilitate the evaporation process of hBN films in our RF sputtering system with a low working temperature (<600°C).
The above substrates were commonly used in our studies over the past fifteen years. In addition, we emphasize on analyzing the formation of hBN phase during the growth process by PVD-RF sputtering method affected by various physical parameters such as distance from target-to-substrate (d), angle created by substrate plane to the center target-substrate axis (α), substrate temperature (T sub ) and substrate surface behaviors when using different materials (Si, NCD, Cr/Au) as just discussed. For each change, we have systematically studied the structural characteristics of the material that is evaporating at the surface of the substrates, the degree of lattice defects is also systematically exploited by means of advanced techniques such as transmission electron microscopy (TEM), Raman and FTIR spectroscopy. To support the experimental TEM investigation, an amorphous substrate of Si 3 N 4 , was used [36]. The conclusions based on the data obtained from TEM have also been confirmed by Raman and FTIR spectroscopy techniques, the amorphous Si 3 N 4 substrate is only for the TEM investigation and not affecting the obtained data and the conclusions drawn from the analyzed results.
2. RF sputtering and experimental parameters 2.1. Unbalanced radio frequency sputtering Unbalanced RF sputtering is a physical deposition method based on the interaction between ionized gases (Ar, N 2 , H 2 , CH 4 , CO 2 ) and a solid target material (BN). In principle, the target material acting as the cathode can be bombarded with inert reactive gas ions or a mixture of both inert (Ar) and non-inert (N 2 , H 2 ) gases [47]. In conventional sputtering techniques, direct current (DC) is used, resulting in a positive charge being generated on the front surface of the target, this process can be prevented by bombarding the insulating target with fluxes of both ions and electrons [90]. Therefore, an RF potential is applied to a metal electrode placed behind the target. With this RF potential, the electrons are oscillated in the alternating field of the applied RF voltage. Particles with enough energy will cause collisions to ionize, resulting in the discharge being self-sustaining without being extinguished. For this reason, high voltage is no longer required to maintain plasma. Since electrons are easier to move than ions, therefore more electrons will reach the target BN surface in the positive half-cycle and similarly there are more ions in the negative cycle, resulting in a negative charge on the BN target surface. Hence, if a DC potential is used, a negative DC potential generated on the target BN surface will repel electrons from the BN surface. This creates a shell with a higher density of ions in front of the BN target. These ions bombard the BN target and sputtering is realized. If the frequency is less than 5 kHz, sputtering does not occur. Therefore, the actual RF frequency is usually in the range of 5-30 MHz, and the choice of 13.56 MHz frequency is widely used for plasma technology. The plasma forming criteria can be satisfied with this frequency. When electric charges appear within the two electrodes, the electrons will no longer oscillate in this regime because they do not receive Figure 4. A simplified model of the polycrystalline diamond growth process at which a few nanometers in size of single-crystal diamond particles are used as nucleation seeds (a). When the grown diamond particles are large enough and the spaces among them are narrowed, the interfaces are created when those particles initially come into contact (b). Those interfaces were created with increasing complexity as the diamond particles overlapped. However, some priority directions are continuously developed for further growth, and the other ones will be eliminated (c). enough energy to create plasma, the plasma between the electrodes can thus be extinguished. Hence, if a magnetic field is placed in the same direction as the static electric field, the kinetic energy of the electrons can be enhanced and the plasma will not be extinguished. Subsequently, the performance of RF plasma can be improved by controlling the strength of the magnetic field between the two electrodes. Various dielectric and amorphous materials are commonly fabricated by RF sputtering. Metals can also be fabricated with RF sputtering if the RF power supply is capacitively coupled to a metal electrode. This prevents DC current in the circuit, and prevents the accumulation of negative charge on the metal target. However, we only used one type of target BN throughout the course of our studies.
The main difference between RF sputtering and other PVD techniques is the need for an impedance matching between the power source and the discharge chamber. Therefore, grounding of the substrate is important to avoid unwanted RF voltage fluctuations on the substrate surface. In our unbalanced RF sputtering technique, the electrons gain energy directly from the RF source to maintain the plasma. The oscillations of electrons are necessary to ionize the gas molecules present in the plasma chamber, so RF technology can work at low pressures. This means that plasma can be created with little collisions of gas ions [90][91][92][93].
The interaction of the active gas ions (Ar, N 2 , CO 2 , H 2 ) and the target material (BN) is quite complex. This process is largely influenced by a combination of gas composition, RF power, working pressure, and magnetic field strength behind the BN target [93][94][95][96]. These parameters directly affect the nucleation and growth of BN materials on the substrate surface. Because of that we fixed the above parameters, in order to stabilize the hBN sample fabrication conditions [35][36][37][38][39]. At the same time, we changed the external conditions such as the distance from the substrate surface to the BN target (d), substrate temperature (T sub ), tilting substrate surface in respect of the virtual line from the BN target center to the substrate (α) and the substrate materials (Si, NCD, Cr/ Au) for the purpose of creating hBN has the desired quality and orientation.
To generate plasma in our RF sputtering system, the sample chamber is always maintained at a high vacuum to avoid the contamination of the air into the chamber. There, the vacuum level always remained at 9.8 × 10 −8 mbar. The mixed gases that we used have been optimized for the composition of Ar(51%), N 2 (44%) and H 2 (5%) with the condition d = 3 cm, working pressure of 2.1 × 10 −2 mbar, RF source power of 75 W will give optimal hBN sample quality [36,47]. With different gas compositions, we can observe the color of the plasma to be different. The plasma produced by the gas composition of only Ar/N 2 is blue-purple, while in the presence of H 2 , the plasma turns pale yellow, as shown in figure 5(a). Over time, the erosion of BN material is generated on the BN target, the erosion forms a race-track groove on the target surface. We can see this effect before and after a certain period of use in figures 5(b) and (c). There, the BN target was unused and used after four years. The shape of the groove is largely induced by the design of the magnetic field behind the target [97][98][99].
The trajectories of electrons from the BN target to the substrate surface depend largely on the magnitude of the magnetic field (B) and the magnetic configuration behind the BN target. The trajectory of an electron or ion satisfies the Lorentz equation [98], Figure 5. An example of the plasma environment in the sample chamber of our unbalanced RF sputtering machine where the evaporation process uses reactive gases is argon (Ar), nitrogen (N 2 ) and hydrogen (H 2 ). When these three gases are present, the color of the plasma bubble is pale yellow as shown in (a). A high purity of BN ceramic compound target was used. With an evaporation rate of a few tens to several hundred nanometers per hour, the erosion rate of the BN target is relatively low. The surfaces of the BN target are in a comparison when their surface images were recorded at the zero-year and four-year used, as shown in (b) and (c), respectively. Therein, a racetrack created resulting from the erosion process is seen. The morphological shapes of the racetrack were highly induced by the magnetic configuration behind the BN target. The dark bands at the edges of the racetrack were the result from the redeposition processes at the BN target surface.
Here, v, m and q are the velocity, mass and charge of the electron or ion. E and B are the electric and magnetic field vectors generated by the magnet behind the BN target. The Lorentz force is directly determined by the components of E and B, where the two vectors are always perpendicular to each other in the 3D-space. Thus, the trajectory of an electron/ion tends to be a spiral around the magnetic flux with the radius of the helix expressed by the Larmor equation [98].
Herein, r Larmor is the Larmor radius, and v ⊥ is the velocity component perpendicular to the surface of the BN target, and the velocity component parallel to the BN target (v ∥ ) is unchanged. The vector sum of [E + (v ⊥ ×B)] is the third component of motion, which creates a drift in the direction perpendicular to both E and B directions, this is the so-called Hall drift. This affects significantly in an unbalanced RF sputtering system and negligible in a balanced one [98,100]. In most magnetron deposition systems, the drift effect always exists because the magnetic field design is imperfect. Resulting in a rather complicated trajectory of electrons because of several effects related to its path. This affects the condensation process of BN phases onto the substrates. Many research groups have been studying different sputtering techniques to build a unique model for both theory and experiment. However, quantitative results on such effects are yet to be observed [98,101,102]. The effect of B varies as a function of the deposition time or the thickness of the BN target. Such an effect is assumed to be constant during the sputtering process at which the depth of the erosion groove is small compared to the entire BN target thickness for a deposition time of several tens of hours. One of the technical solutions is used to control B locally by placing solenoid coils rolling around the sample chamber [103,104]. However, this solution is not applicable for our RF sputtering, because the magnetizing process during deposition can affect the condensation processes on the substrate surface.

Physical parameters affecting the RF growing process
As discussed, several parameters influence the nucleation and growth of thin films. This has raised many questions for researchers when designing a new generation of sputtering techniques [98,105,106]. The reactive gases can be used in the deposition process to combine with the target material. The ions of those reactive gases and the target material can react and condense on a substrate surface [107]. As described, the gas composition Ar/N 2 /H 2 was used for our research purposes. Therefore, the interaction between the ions of the said reactive gases and the BN target is much more complex than in the case of a single gas. During a deposition, various factors such as scattering, trapping, vibration, rotation and bouncing of ions or atoms/molecules on a substrate surface are often of interest [83,108,109]. Those parameters can participate in the dynamics of the chemical and physical processes of the reactive gases in the plasma environment and at the substrate surface. Here, we investigate the properties of deposited hBN films depending on several fundamental physical parameters which are highly related to our homebuilt RF sputtering system. During a deposition process, the mean free path of a particle, i.e. electron, ion, atom, molecule, from the BN target depends on the magnetic field at the backside of the target. Because the magnetic field strength changes as a function of d. This effect slightly affects the deposition rate of BN on any substrate material [36,104]. For each sputtering system, the dependence of the deposition rate as a function of d is different. Our RF sputtering system allows the farthest position of the substrate surface to the BN target is around 7cm. Therefore, we have selected d in the range of 3-6 cm in most of our results, as described in figure 6. This means that if each point on the BN target will have a different solid angle in respect of the substrate if d is changed. That is, the solid angle for a stage of d = 3 cm will be larger than the d = 6 cm case. The decrease in solid angle means a decrease in the efficiency of plasma per a substrate area unit as d increases from 3 cm to 6 cm. Because of this, the growth rate (R G ) of the BN material that deposited onto the substrate will decrease, and as a result, the material properties obtained on the substrate surface will be different. In addition to investigating the change in d, we also studied the effect when rotating the substrate plane at different angles (α) to exploit the adhesion of the BN material to the substrate surface and exploring the chemical role of the BN material in respect of the substrate materials [36]. We assumed that if the substrate plane is tilted at an angle of α with respect to the virtual line connecting the BN target and the substrate centers, the electrons/ions from the BN target when they collide with the substrate will behave differently in a deposition. This can reveal different chemical properties of the deposited BN material in respect of the substrate surface. We will discuss the results in detail in section 5.1.
When changing d from 3 cm to 6 cm, this means that the acceleration of the particles present in the plasma environment is changed. This leads to the physical collisions of those mentioned particles and the substrate surface are altered. Therefore, the temperature of the substrate decreases naturally when the substrate position is moved away from the BN target [98], for example d = 3 cm (T sub = 125°C) and d = 6 cm (T sub = 78°C), the temperature difference of 47°C is the result of the 3 cm displacement. Hence, if we compare the quality of the BN thin films deposited at two positions, d = 3 cm and d = 6 cm, we have to provide the heat energy to the substrate to compensate the different temperature. However, to ensure comparison of the properties of the BN films deposited on a substrate, e.g. Si, NCD or Cr/Au, the d value is fixed at 3 cm and T sub is varied from 125°C to 550°C, depending on the specific research purpose [35][36][37][38][39]. Our RF sputtering deposition system is only capable of raising the substrate temperature up to 550°C, thermal energy is supplied by the DC source and is monitored by a thermocouple probe, as depicted in figure 6.
The above parameters affect the properties of the deposited hBN films on the substrate, the obtained results will be discussed in more detail in section 5. Even so, we can easily see a number of possible effects in the grown hBN thin films such as lattice deformation and defects in the hBN nanostructures, a change in the composition of BN, the heterogeneity of structures, creation of new phases (e.g. aBN, tBN) and adsorbing impurities during the deposition process if a rather complex reactive gas composition of Ar/N 2 /H 2 or Ar/N 2 /CH 4 is used. This will alter the physical and chemical properties of the deposited hBN films [36]. Varying in the properties of the hBN films deposited on different substrates, leading to the structural and optical properties of those hBN films will change. In particular, the role of intercalated elements in the defect sites of the hBN lattice has many interesting properties, which many research groups around the world are currently interested in [53][54][55]. Hence, the results that we discuss in this mini-review only concern the results we have achieved, some challenges we are facing and few solutions that we are addressing. These aspects will provide the necessary information for a PVD approach to grow the hBN phase.

Substrate material surfaces
During the course of our studies, we used most of the commercial Si(100) substrates. There, the Si(100) substrate is considered as a neutral surface, easy to handle during measurements with both high (ρ = 10-22 kΩ.m) and low (ρ = 7-10 Ω.m) resistivity, which are suitable for many common measurements such as Raman and FTIR spectroscopy when measuring in both reflectance and transmittance modes. In addition, the NCD substrate is also used to take advantage of its surface containing free H-bonds. Using this substrate, we assume that the electrons and ions when landing on the substrate surface will be different from the case of Si substrate surface. Hence, we carefully prepare the Si(100) and NCD substrates before using them for our studies. The surface profiles of the two substrates that we used are shown in figure 7.

Si surface
The surface of a Si (100) substrate is shown at the atomic level, the image taken by a scanning probe microscope (SPM). There, the surface roughness of the Si substrate is characterized through an important factor, ρ RMS , which is calculated as follows [110]: Figure 6. The simplified diagram of a deposition setup in the sample chamber of our unbalanced RF sputtering system. Herein, some peripheral physical parameters are defined such as the distance between the target and substrate surfaces (d), the angle inclined from the substrate surface compared to the vertical direction (α) and the heating supplied to the substrate (T sub ) read out by a thermocouple. The reactive gases used for our studies such as argon (Ar), nitrogen (N 2 ) and hydrogen (H 2 ) were separately rectified by mass flow controllers, which were mixed together before injecting into the sample chamber.Reprinted with permission from, [36] Copyright (2016) American Chemical Society.
Here, ρ RMS is the surface roughness, h i and h avg are the actual and the average heights of all points in the scanned Si(100) substrate region, an area of A = 512 × 512 pixels 2 was commonly scanned with a scanning speed of 4 lines/sec, the exposed time of the SPM image recording is about 3-5 min. There, mean height (〈h〉), mean roughness (ρ RMS ), mean grain size (〈g s 〉), standard deviation (σ gs ) and mean grain radius can be calculated from an area of 1 × 1 μm 2 . The ρ RMS value for the Si(100) substrate surface was calculated around 0.2 nm, as shown in figure 7(a). With a surface roughness value of about 2 Å, it is an acceptable condition to use for our depositions. We assume that the Si substrate surface is not oxidized to become SiO 2 . However, it is still possible to exist a rather thin layer of SiO x at the Si substrate surface.

Nanocrystalline nanodiamond surface
In addition to using the Si substrate as mentioned above, we used the NCD buffer as a substrate for hBN deposition. Therein, the surface roughness of the NCD buffer changes randomly depending on the conditions of the CVD fabrication method [39]. However, we used the NCD substrate with the same fabrication process and the fixed thickness, so the surface profile of the NCD substrate may differ only in terms of individual NCD grains at the NCD substrate surface. As an example, the surface profile of a 300nm thick NCD buffer is shown in figure 7(b), where we can see that the NCD grain faces have very different shapes and they expose their faces outward in very random directions. As discussed in section 1.3, in addition to diamonds that are created in nature under high pressure and temperature, diamonds can also be artificially synthesized through various methods [15,16,19,20]. However, we use diamonds made by chemical vapor deposition (CVD) technique in the laboratory, which requires relatively low temperatures of a few hundred to a thousand°C and pressures 40 kPa [20]. Using this technique, sp 2 and sp 3 phases of carbon are both formed during the NCD deposition, the percentage between the two phases largely depends on the fabrication conditions such as pressure, temperature, gas precursors [13]. Various CVD techniques are being used with different plasma generators, i.e. hot filament, microwave, RF, combustion flame [21]. Here, we use a microwave plasma generator with the gas precursors being two H 2 and CH 4 gases. Some other reactive gases can be added during NCD growth such as O 2 , Ar, N 2 , but we do not use them for these research purposes. The plasma generation during NCD fabrication was performed with a microwave frequency of 915 MHz [13][14][15][16]. The NCD deposition process realized in our ASTeX reactor is depicted as figure 8, and the deposition process is described as follows: The precursor gases were initially mixed and introduced into the reaction chamber before being diffused towards a substrate surface, e.g. Si. The microwave power is set to 2500 W. Microwaves interact with electrons in the gas phase and transfer energy to them through collisions. This leads to the dissociation of gas molecules and the formation of active molecules and ions in the plasma environment. The activation of the microwave breaks gas molecules into reactive radicals and atoms [21]. The ions and electrons of those gas molecules are initially created, and the temperature inside the reactor chamber increases to hundreds or thousands of°C. Those molecules, atoms, ions of the reactive gases can be absorbed, diffused, reacted or etched on the substrate surface until a suitable site for the NCD nuclei to grow is found, resulting in the NCD deposition is initiated. There are  (100) and (b) diamond. The image of the Si substrate surface was recorded by scanning probe microscope (SPM) and the diamond surface was imaged by scanning electron microscope (SEM). Herein, a nanocrystalline diamond (NCD) thin film with a thickness of 300 nm was produced using the microwave plasma enhanced chemical vapor deposition (ASTeX-MW PE CVD) technique of Hasselt University. many ways to explain the nucleation of NCD particles during a CVD deposition. However, there is no unique explanation for all approaches. This means, the physical and chemical processes that occur during diamond creation and growth do not have a complete explanation. One of the outstanding research groups involved in diamond development, leading by Prof. Peter K. Bachmann, proposed a suitable picture for the diamond growth at the initial stage [111]. There, a C-H-O composition triangle was established based on their experimental results obtained from different reactors with different initial deposition gases. They concluded that the H atom is the most important element in the gas mixture and governs the entire chemical system that builds the diamond structure. A possible reaction process at the diamond surface and its growth is described in figure 8. Therein, the diamond surface is terminated by H radicals because a high density of H-radicals exists in the plasma, figure 8(a). During diamond growth, some H radicals can be removed and replaced by hydrocarbon radicals-CH 3 (methyl), figure 8(b). As a result there is an extra C added to the lattice, figure 8(c). The same process can be observed on the site adjacent to the attached methyl, figure 8(c) and (d). A further H abstraction process on one of the CH 3 groups and produces a radical on it, figure 8(e). This leads to C atoms in the neighboring positions attracting each other to complete the ring structure of diamond, figure 8(f). Resulting in locking the two C atoms into the diamond lattice, figure 8(g). When the two CH 3 reactants bond together, the H atoms are then released. Therefore, the stepwise NCD growth process is to add C atoms to the diamond lattice on suitable surface sites [21,111].
In addition to the above two substrate materials, we also used a bilayer substrate -heterostructure when combining two transition metal elements, Cr/Au [35,[37][38][39]81]. The aim is that the Cr and Au atoms will facilitate the deposition of the hBN films and the combination of those two metals will reduce the melting temperature of the bilayer substrate when compared with the melting temperature of each metal. The results of which will be discussed in section 5.8.

Specimen preparations
Regarding sample preparation, the procedure of cleaning the Si substrate surface before each prototyping is briefly described in this section. Herein, the Si substrate is commonly used for both PVD and CVD deposition methods. This was followed by the cross-sectional sample processing of SEM (X-SEM) and TEM (X-TEM) measurements. We use X-SEM measurement mainly to determine the thickness of hBN thin films in the range 0.5-4.0 μm. While X-TEM can accurately measure the thickness of thin films <500 nm. Determining the thickness of each hBN thin film, we can work out the growth rate of the thin film at a specific deposition condition, i.e. d = 3 cm or 6 cm and α = 0°or 90°, if the deposition time is known. We also used two X-TEM methods [112], however we highly analyze the results for the case with X-TEM specimens using focused ion beam (FIB) technique. One type of direct TEM sample used an amorphous Si 3 N 4 membrane to grow an hBN film on top, and then directly measure the projection of hBN nanowalls (hBN-NWs) or hBN particles on it. This allows us to determine how the hBN-NWs or particles initiate growth on the substrate. Finally, a metal bilayer of Cr/Au is used as a buffer on the Si substrate, with the aim of taking advantage of the bilayer with transition metals as catalysts to improve the quality of the deposited hBN film. Therefore, the following sections will briefly describe the preparation work for the above research process, which will assist in the interpretation of experimental results in the following sections of this mini-review.

Si substrate cleaning
Both high and low resistivity Si substrates are cleaned with the substrate cleaning procedure, which is cleaned prior to direct or indirect deposition of hBN films on top. The standard RCA sample cleaning procedure was used [113]. The composition of the solution used to clean the sample is in the ratio 5:1:1 for H 2 O:H 2 O 2 :NH 3 (RCA1) and H 2 O:H 2 O 2 :HCl (RCA2). The mixed solution heated up to 70°C. The Si substrates were immersed in the solution for 30 min. Those Si substrates were then taken out and washed with deionized (DI) water and dried with a pure gas flow of nitrogen [110,113]. The Si substrate surface was examined by SPM as seen in figure 7(a). With such surface quality, Si substrate continues to be used for further purposes. If the Si substrate surface is contaminated by a lot of dirt, then they can continue to use the process as described above.

NCD seeding
As discussed above, we use the NCD film as a substrate for the hBN-NWs to grow on it. To create a thin NCD film on Si substrate by CVD method, we prepared such samples following the procedure with some typical steps. In order to deposit diamonds on the Si substrate surface, this Si surface needs to be seeded with the initial diamond nanoparticles. There, a water-based colloidal suspension of nanodiamond particles with a concentration of 0.33 g.l −1 , was used. Such diamond powder was provided by the NanoCarbon Institute Co., Ltd., Japan [114]. The average size of diamond nanoparticles is estimated in the range of 5 -10nm. To seed diamond particles on the Si surface uniformly, a solution containing the nucleated diamonds was dropped to the Si substrate surface, and the Si substrate was then rotated at 4000 rpm. Therein, the Si substrate containing the seeded diamond particles was washed with DI water for the first 20 s of the 40 s spinning time [115][116][117]. After the Si substrate is seeded with NCD nano-diamond particles, the Si substrate is put into our CVD machine to deposit a NCD layer on it and the deposition process is described in section 4.3. Herein, the principle of diamond crystal nucleation was already described in section 3.2 (figure 8).

Si/NCD/hBN heterostructure
Prior to growing hBN films on 300 nm-thick NCD substrates, such a 300 nm-thick NCD film was realized with an ASTeX 6500 MWPE CVD [118][119][120]. There, the nanodiamond particles were seeded on the cleaned Si substrate, as described in section 4.2. The Si substrate with the seeded nanodiamond particles was loaded into the given ASTeX machine using a gas composition of H 2 and CH 4 . Working pressure and microwave power were maintained at 25 Torr and 2500 W. The deposition temperature was about 680°C. The Si/NCD bilayer film was then transferred to our unbalanced RF sputtering system for hBN thin film deposition. The transfer between CVD and RF systems is only 2 min, to ensure that contamination from the air is negligible for each NCD film surface. The growth mechanisms of the NCD and hBN nanostructures will be discussed in section 5.7.

Cross-sectional SEM and TEM specimens
The X-SEM and X-TEM images of the hBN films grown on both NCD and Cr/Au bilayer substrates were recorded. Therein, the samples with the thickness in the range of 0.5-4.0 μm are suitable to characterize by X-SEM image. The hBN thin films grown on the Si substrates were broken, and their cross-sections were exposed and attached to an L-shape substrate holder. The holder is capable of holding the samples upright, to perform SEM imaging from the cross-section of the sample with the normal scanning mode of SEM.
Moreover, if the film thickness of samples is less than 500 nm, X-TEM samples were prepared to measure their thickness accurately by a TEM. Here, we used both the gentle ion milling (GentleMill TM ) and FIB techniques [112,[121][122][123][124][125][126][127] for sample cross-sectioning. These two methods were well described in our previous work [112]. Therefore, we skip describing it in this work. In order to exploit the structural properties of the interfaces between the substrate and hBN film, the FIB X-TEM samples are used most of all, this will be convenient in comparison of results obtained from samples with and without the NCD film. Because if the NCD layer is present, it is not feasible to use the GentleMillTM technique, as the NCD layer has a hardness much greater than that of the hBN one. Therefore, in this mini-review we mainly discuss the results of FIB X-TEM images. The FIB X-TEM samples are all about 100nm thick which is suitable for the characterization of structural and chemical properties, realized by Dr. Svetlana Kyerschuk at EMAT, Belgium. The detailed steps of FIB specimen preparation were described in [112].

Cr/Au bilayer buffer
A DC sputter deposition technique was used to deposit metal materials to create a Cr/Au bilayer structure, performed at the Department of Materials Science and Engineering, National Tsing Hua University, Taiwan, realized by Dr. K. J. Sankaran and Prof. Nyan-Hwa Tai and Dr. Ping-Yen Hsieh. The Cr/Au bilayer was deposited on the Si substrate before growing the hBN thin film. The operation mechanism of the DC sputter is based on the Ar + gas glowing discharge [104,107]. The sputtering chamber contains a metal target (Cr, Au, Ni, Fe) as the cathode and a substrate holder as the anode, maintained in a high vacuum. The metal conducting target is bombarded by the high energy of Ar + ions, resulting in glow discharge plasma. The Si substrate is in contact with the plasma, a thin metal film is thus deposited. The thickness of the metal thin film depends on the deposition time and the distance from the metal target to the Si surface. Herein, the Cr and Au layers are deposited with their thicknesses of 10 nm and 100 nm, respectively. This aim at the thin Cr layer will help create the initial adhesion at the Si surface so that the Cr/Au bilayer will stick well on the Si surface. Moreover, the melting temperature of the Cr/Au bilayer will be much lower than the melting temperature of each individual metal [81].

Si 3 N 4 -TEM membrane
To understand the nucleation of hBN-NWs on any substrate surface, we observed the hBN-NWs growing from the initial stage of thin film growth with TEM. With this approach, TEM images of hBN-NWs were viewed from the top of those hBN-NWs, perpendicular to the substrate surface. As mentioned above, the Si 3 N 4 -TEM membrane was chosen for this purpose. Using such a TEM membrane is suitable for direct structural analysis of TEM images, while being unaffected by sample preparation by FIB or GentleMill TM . The Si 3 N 4 thin film is not only transparent to the 200 keV electron beam of the TEM, but it is also an amorphous material where the interaction of the TEM electron beam with the Si 3 N 4 substrate is almost isotropic. The Si 3 N 4 membranes are typically around 35 nm-thick, supported on a 500 μm-thick silicon frame with an electron transparent window of 100 × 100 μm 2 , the Si 3 N 4 membranes were provided by TED PELLA, INC [39].
5. hBN nanowalls growing on tilted substrates and varying in d 5.1. 0°and 90°-tilted substrates As discussed in section 2.2, growing hBN thin films on Si substrates rotated with different angles (α) to determine the role of Si neutral substrate surface. Moreover, the reactive gas components of RF sputtering contain a very strong reducing agent of H 2 . Hence, we want to see the role physical and chemical processes induce the process of hBN deposition [36,47]. To determine such effects, a series of samples were deposited in varying d in the range of 3-6cm. The samples were grown under the same deposition conditions such as gas precursors, working pressure, and substrate temperature. After the hBN-NWs were grown, then X-SEM images were recorded to measure the film thicknesses. The thickness value of each sample corresponds to its deposition time, which will let us calculate the R G values (nm/h). In addition to fabricating hBN films at the normal case (α = 0°), we also deposited hBN films at the α = 90°stage. Particularly, when the Si substrate rotated at an angle of 90°, we designed a Si substrate about four times larger than the size of a regular Si one (1×1 cm 2 ). This Si substrate was erected so that the plane of the Si substrate passes through the center of the BN target. The Si substrate was glued to an aluminum plate and attached it to the Mo substrate holder of the RF sputtering machine. Hence, we assume that the temperature variation of the sample positions is close to the BN target (d = 3 cm) similar to that at locations far from the BN target (d = 6 cm). Because we assumed that the thermal conductivity of the aluminum plate is good enough, then there is not too much temperature change between the near and far positions of the hBN film. The R G values are calculated for each position of d as the means of the films, and the error of each measurement is calculated from five different measurements with the same d value. The correlation of the RG values is a function of d, shown in figure 9(a). In the R G (d) correlation, we found that the trends of the R G (d) functions described for the two rotated Si substrates is similar in terms of the linear form [36]. A small difference is the slope of these two linear functions deviated from the R G values taken from the two α values (! GR ) at two positions of the sample in respect of the BN target, e.g. ! GR = 75nm/h (d = 3 cm) and ! GR = 40 nm/h (d = 3 cm). The decrease in R G value for the two α values is understandable because the solid angle from the d positions of the α = 0°case is reduced by 50% when the Si substrate is rotated 90°. This means that the flux of particles deposited on the Si substrate surface is physically reduced by 50%. In addition, the strong decrease of R G value at the d = 3 cm position indicates that the role of the kinetic process occurring at the Si surface is important. Therein, the kinetic processes of the ions occurring at the Si surface with a small d value are quite fast, so the efficiency of hBN material adhering to the substrate surface is low. Resulting in the degree of adhesion of hBN materials at d = 3 cm compared to the case d = 6 cm is smaller, leading to higher porosity of hBN deposited films at locations closer to the BN target. This result is confirmed by two X-SEM images of two hBN films with their thickness of 2.86 μm, deposited at d = 3 cm for the two different angles, figure 9(b) and (c). Based on the contrast of the two X-SEM images, we can see that the size of the hBN-NWs existing in the deposited hBN film with α = 0°is larger than that of the sample with α = 90°. That is, the density of nucleated hBN-NWs is low in the α = 0°case, compared with the α = 90°one. This results in the porosity of hBN films fabricated with α = 0°is lower than that of the α = 90°case. Therefore, the fabricated sample with α = 0°a ssisted hBN-NWs grows linearly when the spacing between hBN-NWs is large enough, figure 9(b). Therein, the hBN-NWs will grow spontaneously and do not compete for spaces at the early stage of thin film growth, resulting in less overlapping of the NW-branches than in the case of samples fabricated at α = 90°. This assessment is clearly seen through the amplitude contrast of the X-SEM images in figure 9(b) and (c). The early overlap of NW-branches can be observed from figure 9(c), where a higher density of hBN-NWs deposited on the 90°rotated Si substrate is observed at one-third of the film thickness from the Si surface (R 1 ). The hBN particles continue to grow and enlarge in size, resulting in competing in their spacing. Some hBN-NW branches were suppressed under the competition, while others were further grown, resulting in a higher porosity of the hBN film in the mid-film thickness region due to the presence of bigger hBN-NWs (R 2 ). When the grown hBN-NWs are large enough, they continue to nucleate small branches and compete with each other for space. Those hBN-NW branches compete with each other to increase the density of material or decrease porosity of material, as seen in the R 3 region of figure 9(c). In particular, even though the Si substrate is rotated 90°, the orientation of the hBN-NWs does not change where all the hBN-NWs tend to grow perpendicular to the Si substrate surface. Therein, the third axis of the hBN-NWs (c-axis) tends to be parallel or at a very small angle to the Si substrate surface plane. This can be visualized directly from our two X-SEM images or this was also studied in detail by our group member [128]. The c-axis of the hBN-NWs can be twisted at an angle of < 20°from the horizontal. Such a conclusion is quite consistent with our results that we observed directly from those X-SEM images.
To further explore the growth of the deposited BN material near the Si substrate surface, X-TEM images of the hBN film deposited on the Si substrate with a thickness of 300 nm, d = 3 cm and α = 0° [36], as shown in figure 10. The 300 nm-thick hBN film is aimed at seeing the entire film thickness of the X-TEM specimen using TEM technique. Since this specimen has only a single layer of hBN that grows onto the Si substrate, hence we prepared the X-TEM specimen by the GentleMill TM method [112] to reduce a certain damage from the highenergy ion beam like FIB. From the X-TEM image, we can see very clearly that the two hBN films are facing each other because of using GentleMill TM preparation method, leading the two Si substrates appearing on opposite Figure 9. (a) The growth rate (R G ) of the hBN films deposited at different distances (d) from the target BN to the plane of the tilting substrates with different angles (α = 0°and 90°). The dashed lines serve as guides for the eyes. Each data point in the graph was calculated from a single deposited film, each error bar was calculated independently for each hBN film as the standard deviation for the thickness measurement of ten different points in the cross-sectional SEM (X-SEM) image. The measurement error comes from a single film, which does not represent repeatability for different films. (b) and (c) are typical X-SEM images of two hBN films of approximately the same thickness deposited at d = 3 cm with α = 0°and α = 90°, respectively. The SEM images of these samples are also seen on the right panels where the density of hBN-NWs is low in (b), while it is higher in (c) at which the wall branches of hBN-NWs emerged and participated in the growth process. In particular, the three thickness regions denoted as R 1−3 in the X-SEM image at (c) are determined based on the amplitude contrast of the X-SEM image, the explanations of these regions are detailed in the content of this paper. Reprinted with permission from [36], Copyright (2016) American Chemical Society.
sides of the head-to-head hBN film border, figure 10(a). From the contrast of the X-TEM image, the thickness of the deposited hBN film was estimated around (300± 20) nm, as shown in figure 10(a). This thickness value is consistent with the expected growth thickness controlled by the deposition time, calculated from R G as seen in figure 9(a). The change of the structural properties of the deposited hBN film is also detected, based on the contrast of the TEM image. The results show that the BN film is nonuniform with the increase of thickness, closer to the Si substrate surface, the region with lower contrast in the TEM image, indicating the presence of less crystalline structures, this conclusion is consistent with the data shown from figure 9.
To clarify the intrinsic structure near the Si/hBN interface, a high-resolution X-TEM (HRTEM) image was recorded at a location close to the Si substrate surface and the deposited BN film, as shown in figure 10(b). From the HRTEM image, we found that the material near the Si surface is a mixture of the imperfect hBN phase (tBN) and a part of the aBN phase. That is, the short range ordered atoms in the hBN film dominated at the early stage of thin film growth. Farther from the Si substrate to the mid-area of the hBN film, the hBN phase begins to appear, as shown in figure 10(c). The hBN phase grown as hBN-NWs is partly in the visible region, but two undesirable phases remain, tBN and aBN. The two unwanted phases are constrained in the spaces between hBN-NWs. The hBN phase in hBN-NWs is clearly seen in figure 10(c) where the hBN lattice planes are not perfectly aligned. This indicates a change in the lattice spacing between the hBN nanosheets, resulting in the lattice distortion possibly caused by crystal defects in the grown hBN-NWs. The dominant growth direction of the hBN-NWs is in the vertical direction relative to the Si surface.

hBN nanowalls grown in differing film thicknesses
As discussed in section 5.1, all hBN films were grown and characterized with X-SEM and X-TEM images. To get a comprehensive view of the growth of hBN-NWs on the Si substrate surface, we measured plane-view SEM images of those hBN films in differing thicknesses (t d ). This aims to clarify the growth process of many hBN-NWs on a large scale and investigate the interactions between hBN-NWs during their deposition. Hence, planeview SEM images of hBN films were recorded in the variations of td. Some typical SEM images are shown in figure 11. The SEM images shown on the left panel of figure 11 is for the films deposited at d = 3 cm, and the SEM images on the right panel is for the d = 6 cm samples. Figure 11 indicates that the growth characteristics of the hBN-NWs vary with the td of the hBN films. In particular, at the same thickness t d ∼300 nm, the lower density of hBN-NWs grown on a Si unit area at d = 3 cm is observed as compared to the case of d = 6 cm. With increasing t d , the size of hBN-NWs is increased linearly for the films with t d 710 nm. If the thickness continues to increase up to 1430nm, the NW-branches have started to grow and dominate over the growth of the cores of hBN-NWs. The growth of the hBN films is also related to the growth of NW-branches, leading to the growth rate of the thin films is low. There, the NW-branches tend to fill in the spaces between the hBN-NWs. When the NWbranches are large enough (t d ∼3050 nm), we found that the morphological characteristics of the hBN films are quite close for both cases of d = 3 cm and d = 6 cm. These results obtained from the given SEM images are consistent with the conclusions extracted from figure 9. Furthermore, the results are more extensive when we directly see the shapes of those hBN-NWs and their branches which were grown in the hBN films with different td. These results will support the arguments assigned from the data of figures 9 and 10.
To further investigate the crystal quality as a function of the film thickness, Raman and FTIR spectra for the thin films given in figure 11, were recorded. In principle, when an infrared or laser beam illuminates an hBN film, the lattice of hBN will vibrate. Leading the change in the dipole moment during an oscillation period [129], which is constrained by the phonon modes of the hBN lattice [61]. Therefore, the reactive IR modes absorb energy at individual frequencies which are related to the vibration modes [129,130]. Those modes will appear in the interferometer after being transmitted or reflected with the grown hBN film. After a Fourier transform (FT) of the given signal, the absorption spectrum of the hBN film can be recorded. Since hBN-NWs in hBN films are oriented in a direction perpendicular to the Si surface. Hence, FTIR and Raman measurements were preferably realized vertical to the hBN film surface, as described in figure 12. Here, the FTIR spectra were mainly performed in the transmission mode in order to exploit the intrinsic properties of hBN films. As discussed above, the c-axis of the hBN-NW lattice is oriented relatively parallel to the Si substrate surface, the directions of the incident IR and laser beams in both FTIR and Raman measurements are described in figure 12(a). The vibration modes of the hBN lattice when IR and laser beams interacting with hBN films are simplified, as shown in figure 12(b) and (c). Characterizing the optical vibrational modes of hBN crystals indirectly exploits the intrinsic properties of the material, i.e. crystal anisotropy, thin film thickness, crystallinity, and elemental composition present in the crystal of the sample. To prevent some of the contamination from adsorbing from the H 2 O and CO 2 in the air, a continuous flow of dry nitrogen gas was supplied into the chamber of the FTIR instrument, which ensures that unwanted absorption from the air is negligible. Each measurement takes about 30min with a resolution of 2.0 cm −1 in the frequency range of 400-4000 cm −1 . There are two IR-active optical phonon modes for the hBN crystal with frequencies at 783 cm −1 and 1376 cm −1 [3,131,132]. The two IR modes are denoted as A 2u and E 1u correspond to the out-of-plane and in-plane B-N bonds as described in figure 12(b). In addition, there is a Raman active optical phonon mode at 1367 cm −1 , denoted A 2g [3], which is described in figure 12(c). Depending on the relative orientation between those phonon modes and the incident beam direction, the crystal and phase characteristics of the grown hBN film can be determined.
Raman spectra of hBN films was recorded in the backscattering configuration using a blue Ar + gas laser beam with wavelength of λ = 488 nm and power of 10 mW, integration time in the range of 5-300 s, chromatic slit size of 0.1 mm [110]. We not only did Raman spectroscopy measurements with the sole hBN films, but also measured Raman spectra of NCD/hBN heterostructures, as shown in figure 13. Figure 13 shows the highfrequency Raman active peak of the hBN film (E 2g hBN ) is of 1367 cm −1 [132], while the NCD film has two vibration peaks of sp NCD 3 and sp NCD 2 [133,134]. The first peak is related to the vibration of the NCD cubic lattice, it is a sharp peak located at the frequency of ν = 1333 cm −1 . The second peak is related to the graphitic carbon phase existing with two broadened peaks, denoted as 'G' and 'D'. These peaks are contributed by the random vibrations of carbon atoms in the covalent plane. The 'G'-peak at ν = 1575 cm −1 resulting from the movements of neighboring carbon atoms in opposite directions and perpendicular to the sp NCD 2 plane. The 'D' peak results from dislocations in the NCD lattice, it is also known as disorder − induced mode. The relative intensity of the 'D' and 'G' peaks can be used as a rapid approach of determining the degree of disorder in the NCD sample [6]. The sp 2 carbon phase has an overlapping peak of those two 'D' and 'G' one, occurring at the frequency of ν = 1487 Figure 12. Optical properties of grown hBN films were characterized by FTIR and Raman spectroscopy techniques. In both measurements, the incident beams are always perpendicular to the Si substrate surface, as described in (a). The infrared (IR) and laser beams of those techniques penetrating into the hBN films in either transmittance or reflection modes will interact with the hBN lattice within the hBN films. Therein, we assume that the hBN nanosheets are vertical to the Si substrate surface. Therefore, the simple phonon modes of the crystalline hBN films in respect of the IR and laser beams are described in (b) and (c). Figure 13. Raman spectra of a heterostructure containing the grown NCD and hBN films, which have the same thickness of 300 nm. Therein, the NCD film was firstly grown on the Si substrate surface, then the hBN layer was continuously deposited on the given NCD film to form a Si/NCD/hBN structure. Raman spectra show some typical vibration modes of diamond (sp 3 ), hBN and graphite (sp 2 ), these phases are present at the core of the NCD particles, hBN-NWs and the boundaries of the NCD particles, respectively.Reprinted from [39], Copyright (2017), with permission from Elsevier. cm −1 . We did not exploit the NCD film in detail, but such information is necessary to recognize the presence of the NCD phase, the quality of the NCD film needs to be determined in case the hBN film was deposited on top of the NCD one.
For Raman spectra, the inverse of the crystal size is proportional to the full width half of the maximum value (Γ 1/2 ) at the high frequency Raman peak (E 2g ) [132]. That is, if the value of Γ 1/2 decreases at the above frequency position, then we can conclude that the crystallization of the material increases. Raman spectra and Γ 1/2 values at the E 2g frequency were measured and calculated for the hBN films deposited at d = 3cm and d = 6 cm, as given in figure 11, results of which are shown in figure 14.
The Raman peaks were fitted with a Gaussian distribution function to estimate the crystal size via their Γ 1/2 value. In fact, the Raman spectrum of deposited BN film is a combination of effects from three hBN, tBN and aBN phases. The hBN and tBN phases contributed mainly to the peak intensity and Γ 1/2 value, while the aBN phase had no contribution. Figures 14(a) and (b) show that the intensity of the Raman spectrum increases as the thickness of the hBN film increases. This means that the summation of the laser beam and the hBN lattice interactions are proportional to their film thickness. In addition, the width of the spectrum is inversely proportional to the thickness of the hBN thin film, which indicates that the signal caused by unwanted phases in the thin samples predominantly contributes in the entire Raman spectrum leading the Raman signal blurred. Herein, the peak of the Raman spectrum corresponding to the E 2g mode at ν = 1367 cm −1 is quite uniform for most of the samples, which indicates that the hBN phase is evident in the hBN films. Figure 14(c) shows that the hBN films with thickness t d > 1 μm deposited at d = 6 cm, the obtained hBN phase is quite homogeneous and govern all quality of the samples. In other worlds, the Raman signal is not affected by the BN layer close to the Si substrate surface. For the hBN films with thickness < 1 μm, the unwanted phases created at the Si substrate surface strongly induce the Raman signal, resulting in a wider Raman spectral width corresponding to a large Γ 1/2 value. This result is in good agreement with the data obtained from figures 10 and 11. As for the hBN films fabricated with d = 3 cm, the quality of the hBN phase is quite uniform, and is less affected by the material layer close to the Si surface. This leads to low and stable values of Γ 1/2 as a function of thickness. Figure 14. Raman spectra of the hBN films with different film thicknesses deposited at d = 3 cm (a) and d = 6 cm (b), which are described in figure 11. The mean Γ 1/2 values of the E 2g peak calculated from those Raman spectra measured on five different locations on each film, as shown in (c). The error bars represent the standard deviation of the Γ 1/2 mean value for each hBN film measured at five different locations.Reprinted with permission from [36]. Copyright (2016) American Chemical Society.

hBN nanowalls grown in differing T sub
The above results did not take into account the effect of temperature, because the hBN films depositing at d = 3 cm and d = 6 cm also differ by only about 47°C. This temperature difference is naturally created resulting from interactions between atoms or ions inside the plasma chamber of the RF sputtering system. Therefore, we investigated the influence of temperature supplied to the sample substrate during deposition. Therein, we predict that some kinetic processes occurring at the Si substrate surface during the deposition will be directly influenced by the substrate temperature (T sub ). Therefore, a series of hBN thin films were deposited with different T sub , X-SEM images of those samples were recorded, and the deposited rate (R G ) was then calculated for each specific value of T sub . The R G values as a function of T sub is plotted in figure 15(a). Furthermore, since the reactive gas component of the deposition process contains H 2 , the H ions have very strong reducing properties, so it is possible to act as an acid that can corrode the hBN material during the deposition process. This process is in contrast to the growth of the hBN film over time. Although we cannot directly measure the kinetic effect of the said corrosion process, we can estimate this effect at the static state. Hence, the fabricated hBN thin films were reloaded into the sample chamber and exposed to the H 2 plasma and the Ar and N 2 active gases were removed [36]. The removal of two Ar and N 2 gases will cause the working pressure in the sample chamber to change. Therefore, we added an extra amount of H 2 to the sample chamber to compensate, which means an increase in the corrosion concentration. At this stage, we assume that the corrosion process due to increased H 2 concentration balances with the bombardment produced by Ar and N ions during normal deposition. The hBN film thicknesses before and after etching by H 2 plasma were measured, resulting in the etching rate (R E ) as a function of T sub was found, as plotted in figure 15(b).
As shown in figure 15 where the hBN films deposited far from the BN target (d = 6 cm), T sub slightly induces the growth rate of the hBN films. When T sub increases from 78°C to 250°C, R G tends to increase slightly. This shows that, in the temperature region < 250°C, the deposition process is enhanced by T sub , which will make the H ions more active, leading to the adhesion of the hBN material at the early stage of thin film deposition. When the temperature rises above 250°C, the etching process plays a key role, leading to a decrease in the R G value. For the hBN films deposited close to the BN target (d = 3 cm), R G was less affected by T sub < 250°C, but did play a key role for T sub > 250°C as the R G value highly decreased when the temperature increased. We temporarily conclude that the substrate temperature greatly affects the growth rate of hBN films. However, the question is, why this effect is so large in the films deposited closer to the BN target. We did assume that this effect is caused by the etching of the strongly reducing H ions at high temperature. Therefore, we measured this effect with the R E -T sub curve, as shown in figure 15(b). The obtained results are in agreement with our initial predictions. There, the etching rate of depositing samples at T sub > 250°C and d = 3 cm is increased as a function of T sub . Besides, the effect was not obvious in the hBN films deposited at d = 6 cm. The trend of R E -T sub curve (d = 3 Each data point represents a single deposition or etching experiment, error bars were calculated independently for each hBN film, each mean value was calculated from ten different locations on the X-SEM image of each sample used to determine each thickness measurement. The substrate temperature was firstly raised by the external heating source to the temperature naturally generated by the interaction of the reactive gas ions and the substrate surface at positions d = 3 cm (T sub = 125°C) and d = 6 cm (T sub = 78°C). The later measurement points were then started from T sub = 125°C to 500°C with an increment step of 125°C. For the case of etching, the samples after be deposited, those grown films were put into the RF sputtering chamber and removed the reactive gases of Ar and N 2 , leaving only H 2 in the RF chamber, to carry out the etching measurements. The etching rate was calculated as a thickness change of the grown film before and after etching divided by the etching time. Reprinted with permission from [36]. Copyright (2016) American Chemical Society. cm) seems to be inversely to the curve of R E -T sub , our initial prediction was thus correct. Therefore, we can conclude that hBN crystal growth is strongly dependent on T sub at d = 3 cm. In the case of d = 6 cm, the density of hBN-NWs is high where the hBN particles adhere to the Si substrate surface at the early stage of thin film growth, leading to the growth and etching rates of the hBN films are low. As the density of hBN-NWs is related to the growth of NW-branches. Hence, the etching process is expected to affect the NW-branches directly rather than the cores of the hBN-NWs. This effect could also be seen in X-SEM images, recorded from the surface of the deposited hBN films at two different T sub and d values, figure 16. Figure 16 shows that two hBN films were grown at two different T sub , the morphology of those two films are clearly different. The morphological properties of the two films grown at d = 3 cm and d = 6 cm with T sub = 125°C are quite similar. When the temperature increased up to 500°C, those film surfaces changed significantly. Therein, the film deposited at d = 3 cm and T sub = 500°C did not show the existence of hBN-NWs on its surface, and the morphological behavior was replaced by distinguishable clusters with the sizes of BN particles being very small. These clusters can be distinguished among each other by the amplitude contrast of the SEM images, denoted by the yellow dashed lines in figure 16(b). In contrast, the film fabricated at d = 6 cm and T sub = 500°C still exhibited the existence of hBN-NWs on its surface. These hBN-NWs are different from those that existed on the film grown with T sub = 125°C at which the NW-branches have been truncated and the cores of hBN-NWs remained, figure 16(d). Those hBN-NWs and NW-branches are denoted by yellow and cyan arrows in those SEM images. Clusters surrounded by yellow dashed lines have an average diameter of about 0.5-1.0 μm, figure 16(b). The formation of these BN cluster boundaries might follow the growth model of diffusion-limited aggregation during deposition with the mechanism of chemical rather than physical diffusion [135][136][137][138]. This hints that chemical processes strongly influenced the deposition process where chemical and temperature effects strongly influenced the intrinsic structure of the hBN film.

Thermal stability and defects of hBN-NWs
From the data discussed above, it leads us to conclude that the quality of the hBN-NWs changes as a function of the experimental parameters. Therein, optimal conditions can be selected for the deposition of the best quality hBN films. Even so, the intrinsic properties of the hBN-NWs reveal that they have many lattice defects, resulting in lattice distortion as shown in figure 10(c). Furthermore, when studying the surface properties of hBN films with their SEM images, we found that the material etching caused by H ions is very significant. Based on the obtained results of other research groups, we predict that the H ions will be temporarily bound during the deposition of hBN films. Therein, the H atoms will bind temporarily at N and B vacancies in the hBN lattice. H ions can be intercalated on those vacancies during the deposition of hBN films, H is thus an important factor to support the formation of hBN lattice cells. However, it is also the main factor causing defects in the hBN crystal lattice. Due to the fact that the random and complex arrangement of atoms in the physical and chemical kinetic processes that occur at the Si substrate surface, resulting in the hBN film contains many defects. If the defects exist, the residual bonds of B or N will directly connect to the free H ions. This means that there are N-H or B-H bonds in the grown hBN films [139][140][141][142]. Hence, the given bonds should be the clue to determine the degree of defects in the grown films [36]. Therefore, FTIR spectra of the hBN films deposited at different temperatures were recorded. We have (i) investigated the possibility of the existence of N-H or B-H bonds in the grown samples and (ii) if the given bonds exist, how stable are they under Tsub. We have examined most of the grown samples carefully, for the B-H bonds with their IR vibration frequency of 2400 cm −1 , and we did not find the said vibration signal despite increasing FTIR spectrum acquisition time. Even so, our results are also consistent with the experimental and simulation data of other research groups [139,140]. Therein, other research groups also concluded that the B site in hBN lattice is often the preferred position in the crystallization process of hBN materials. Thus, we also postulated that, if there is a lattice defect at the B site, then the H ions will bind to the N-terminated edges. Because of that, we assumed that our hBN films mostly contain N-H bonds. FTIR spectroscopy was used to quantify the strength of H-N bonding vibrations, as a measure of the relative density of H present in various hBN films. A series of hBN films with a thickness of 3.5μm, deposited at d = 3 cm and d = 6 cm in differing T sub . The FTIR transmission spectra of the hBN films were recorded in the H-N vibrational mode at ν = 3450 cm −1 [143,144], the results of which we reported in [36], as recited in figures 17(a) and (b). FTIR measurements were taken perpendicular to the substrate surface or quasi-parallel to the planes of the hBN-NWs, as described in figure 12. Sum IR absorption at the typical phonon peak of the FTIR spectrum was used to estimate the IR absorption band area at the N-H vibration mode (S IR−absorption−band ). This factor is calculated from the product of the absorbance and the bandwidth at half-absorbance, expressed by the following equation  subtracting the given spectra. With the said spectral subtraction, the spectra of the individual hBN films will be obtained, and we used 128 averaged scans to obtain the final spectrum for each measurement. The peak area of the FTIR spectra calculated using equation (4), as described in figure 17(c) and the obtained results are shown in figure 17(d). Therein, the temperature difference between at d = 3 cm and d = 6 cm of 47°C was corrected by having samples measured at those d values heated to 125°C. This value is equal to the temperature naturally generated during the physical interaction of ions in the plasma environment in respect of the substrate surface at d = 3 cm. However, the FTIR spectrum of the unheated hBN film grown at d = 6 cm was still included for comparison. The data shown in figure 17 shows that the peak position of the H-N bond may differ slightly, resulting from the difference in the H-N binding environment. This could be explained that the defects and impurities present in the hBN films are not uniform. In addition, the presence of undesirable phases such as aBN and tBN affects the order of the N-H bonds/orientations. The absorbance areas obtained from the FTIR adsorption spectra tended to decrease as T sub increased from 125°C to 250°C for the samples deposited at d = 3 cm. This suggests that the H-N bond density is high in the sample deposited at low T sub . The density of H-N is kept stable and does not decrease when T sub 250°C. This means that T sub = 250°C is the critical temperature at which the N-H bonds are stable in the films deposited at d = 3 cm. The number of N-H bonds decreases with increasing temperature. This also means that the dangling bonds at the N-terminated edges are reduced, resulting in a low probability of NW-branches being created. The creation of NW-branches is low, which means the creation of unwanted phases is also low. Therefore, T sub clearly plays an important role in enhancing hBN phase formation. In other words, the chance of making NW-branches at low T sub is higher than that of high T sub .
A significant decrease of S IR−absorption−band was seen when T sub increased from 78°C to 125°C in the film deposited at d = 6 cm. The density of N-H bonds remained at this absorption level when T sub was increased to 250°C. The density of N-H bonds present in the films deposited at d = 6 cm is higher than that of the films grown at d = 3 cm. The desorption mechanism for the d = 6 cm case is similar to that of the case of d = 3 cm. The N-H bond density in the films grown at T sub > 250°C also remained stable and lower than that of the samples deposited at d = 3 cm. This hints that the N-H bond density present in the hBN phase sample is higher than in the sample containing undesirable phases such as aBN and tBN. This is consistent with the conclusions obtained from the data discussed above. Many research groups have been delving deeper into this issue [81,[144][145][146]. There, they are trying to find a correlation between the N-H bonds and the N-H concentration. The change of the N-H content as a function of temperature can also be quantified by a thermal gravimetric analysis (TGA). The results show that the H-N bonds are usually broken in the range of 250-350°C. However, the samples used with TGA are suitable for a large amount, which is different from our approach when using a small volume of hBN film. The difference between the desorption temperature values of our N-H bonds and those studied using the TGA method could come from differences in sample mass or porosity of the measured films.
To determine a difference in the quality of hBN films were deposited at different T sub at which the desorption process affects the quality of those deposited films. Therefore, a series of Raman spectra of the hBN films with a thickness of 700 nm, deposited at d = 3 cm and d = 6 cm were recorded, as shown in figures 18(a) and (b). The Γ 1/2 values of those Raman spectra recorded around the E 2g vibration mode were calculated, as described in figure 14(c), results of which are shown in figure 18(c). Since we only deposited hBN films with a thickness of t d = 700 nm in order to reduce the evaporation time. Using this thickness, the Raman signal is also strong enough for analysis and less affected by unwanted phases created by overlapping of the NW-branches when the thickness of the samples is large. The results reveal that the Raman spectra recorded around the E 2g vibration mode are clear, showing that the hBN phase quality is dominant in hBN films, figures 18(a) and (b). There is only a large change in the Γ 1/2 -T sub curve for the films when increasing T sub from 78°C to 125°C, the crystallinity of hBN markedly increases, figure 18(c). The results also show that T sub plays a role in the deposition process at the early stage of thin film growth. At the T sub > 125°C and d = 3 cm, the crystallinity of the hBN films changed insignificantly, but kept a little better than that of the samples deposited at d = 6 cm. While the hBN films grew at d = 6 cm, the crystallization quality of the hBN phase gradually decreased with the increase of T sub . This is assigned that the desorption of N-H bonds strongly affects the samples grown at d = 6 cm, while having little effect on the films deposited at d = 3 cm. This allows us to conclude that the amount of N-H bonds desorbed from the NW-branches is more than from the cores of hBN-NWs.
To investigate the stability of the hBN phase at the high annealing temperature, the as-deposited films were annealed at temperatures up to 1000°C. We have measured the FTIR spectra of the samples that have not been annealed, and annealed at 750°C and 1000°C. The FTIR spectra of those samples are shown in figure 19. This approach allows us to investigate the thermal stability of the N-H bonds within those samples. Similar to what described in figure 17, the FTIR spectrum is recorded around the vibration mode of the H-N bond at ν = 3436.6 cm −1 and the optical phonon modes of the hBN can be observed as B-N bending A 2u (LO) at ν = 817 cm −1 and B-N stretching E 1u (TO) at ν = 1376 cm −1 . The FTIR spectra of the vibration modes of A 2u , E 1u and N-H are given in figure 19. These spectra were recorded from three hBN films with a thickness of 1.2 μm, which grew at d = 3 cm. One sample was selected for comparison without annealing or as-deposited sample, the other two samples, one annealed at 750°C and the other annealed at 1000°C. Both samples were incubated in a high vacuum (10 −6 mbar) for 5.5 hours to avoid contamination by the atmosphere.
For the H-N vibration mode, figure 19(c), the intensity of the H-N peak decreases with increasing annealing temperature, indicating that desorption of the N-H bond occurs during annealing. Desorption is a temperature dependent process, the degree of desorption is greater as the temperature is higher. In particular, the crystal structure is suppressed when the annealing temperature reaches 1000°C, as noted the annihilation of the A 2u and E 1u phonon modes in figures 19(a) and (b), possibly resulting from the desorption of the N-H bonds. This is also evidenced by the results obtained from other research groups that the desorption of H atoms in BN is nearly   E 1u (b), were recorded to investigate the stability of the hBN crystalline structures at high temperatures. In addition, the stability of H-N bonds existing in those hBN films was also exploited, as shown in (c).Reprinted with permission from [36]. Copyright (2016) American Chemical Society.
completed when annealed at 1000°C [140,144,147]. If the H atoms leave the lattice, some vacancies appear in the hBN crystal structure. As a result, the hBN films showed more defects than the film annealed at 1000°C. Our data are quite consistent with those of other research groups [140,146]. The loss of phonon modes in the hBN film annealed at 1000°C can be caused by the formation of non-crystalline phases in the sample. Therein, the migration of the B and N vacancies can form states with more stable bonds such as BN, resulting in the existence of di-vacancies in the film at high annealing temperatures. The formation of di-vacancies will cause a lattice distortion of the hBN film or warping relative to the hBN-NW planes [140]. The hBN phase can be transformed to aBN and/or tBN. Although most of the hBN phase still remains, however the presence of new BN phases within the annealed hBN film can disrupt the long-range crystalline order that maintains the LO and TO phonon modes. Moreover, to investigate the degradation of hBN crystalline phase for different annealing temperatures, we recorded the SEM images of the above films, results of which are given in figure 20. Therein, the results show that the hBN-NWs appearing at the hBN film surface is quite stable and similar to the asdeposited sample in figure 20(a) or the annealed at 700°C in figure 20(b). Increasing the temperature further, the hBN-NWs and their NW-branches were degraded by the desorption processes that occurred strongly and caused the hBN lattice to be distorted, deformed or transformed to other phases such as tBN and aBN [36].

hBN-NWs grown on Si 3 N 4 membrane
To have a better understanding how the hBN-NWs grow on any substrate surface, a thin layer of hBN film (< 100 nm) was grown onto a Si 3 N 4 membrane, which is suitable for structural and chemical TEM characterization without using a sample preparation process. Therein, the preparation is easy to damage the sample [36]. With the given thickness, the grown hBN film does not cover the substrate surface, figure 21(a) at which hBN-NWs were randomly nucleated and grown on the Si 3 N 4 substrate surface. Based on the imaging contrast, the hBN-NWs appear as black-gray fringes in the BF-TEM image. Those structures followed the configurations of the initially created hBN-NWs. Hence, the thickness of those hBN-NWs is higher than that of the regions without hBN-NWs. We can see that the hBN-NWs grow as they stack on each other and are randomly oriented in different directions. The selected area electron diffraction (SAED) TEM image, figure 21(b), shows that the hBN-NWs are oriented perpendicular to the Si 3 N 4 substrate surface. There, the halo intensity ring of electron diffraction from the (0002) plane from the hBN-NWs is dominant. The circular and uniform halo shape indicates that the hBN-NW planes contribute equally in any direction or the hBN-NWs are randomly orientated when growing on the Si 3 N 4 surface. In other words, the planes of hBN-NWs parallel to the electron beam of TEM. Furthermore, if hBN-NWs are considered as particles, and projecting them on the Si 3 N 4 substrate surface, we can estimate their sizes using an image analysis tool, e.g. ImagJ, by modeling the projection image of the hBN-NWs as ellipses. Therein, the minor axis length of the ellipse is the width of the hBN-NW. This analysis assumed that the hBN-NWs are single entities with a common shape, oriented perpendicular to the substrate surface. Several effects such as curling, overlapping, curving [1] can affect the size of hBN-NWs. However, we assume that these effects are insignificant compared to the majority of hBN-NWs.
The process of calculating mean hBN particle sizes was realized in three steps. (i) A specific area is selected from the BF-TEM image. Therein, the hBN-NWs appear as dark areas, we adjusted through the electron intensity of the image, and then cropped a favorable area, as typically shown in figure 22(a). (ii) The hBN particles are selected when adjusting the electron intensity level of the TEM image, those particles will be covered with different color pixels from the background image, figure 22  limit of minimum and maximum gray level to estimate the mean hBN particle size from more than 1300 particles with the ImageJ software [148]. Based on those distributions of hBN-width and -length, the mean values of widths and lengths were estimated to be (6.2± 1.4) nm and (133.6 ± 4.4) nm, respectively. The statistical distribution of length values is usually estimated by fitting the data with probabilistic functions such as Gauss, Lorentz, Voigt. However, we used the log-normal distribution function [149,150], for the width  . The BF-TEM image of the 100 nm-thick hBN film grown on the Si 3 N 4 membrane was recorded at which it was unmapped (a) and mapped (b) with the given mapping procedure of hBN-NW features, this image area was selected from the BF-TEM image given figure 21(a). The grain-size width and length distributions of about 1300 wall-features were projected onto the substrate surface plane. This data was calculated based on the distributions as shown in equation (5). From these distributions, the mean width and length values of those hBN-NWs were estimated. Reprinted from [39], Copyright (2017), with permission from Elsevier. calculation because it is the best fit for this distribution. The log-normal distribution function is described as equation (5). Meanwhile, the probability length distribution of the hBN particles is consistent with the Voigt function, which is a conjugate function of the Gaussian and Lorentzian ones.
D is the width or length of the hBN particle, μ is the mean of ln(D), δ is the standard deviation of ln(D). The peak of the hBN width distribution is narrow, figure 22(c), confirming that the good homogeneity of the hBN-NWs grows on the Si 3 N 4 membrane. While the length distribution of hBN particles is wider, figure 22(d), implying that the hBN particles grew more freely in their lengths than in the widths during a deposition. Although we use two different functions to describe two similar physical properties, the physical significance of the distributions is not different.
A detailed study of some specific regions on figure 22(a), aiming at investigating the intrinsic properties of the crystallinity of a single hBN-NW and the superposition could be occurred during the growth of hBN-NWs, HRTEM images were taken and shown in figure 23. Figure 23(a) shows that a single hBN-NW has a projection onto the Si 3 N 4 substrate surface with its shape similar to an ellipse, bounded by a red dashed curve. Based on the electron intensity histogram of the HRTEM image, the imaging histogram was scanned from A to B of the hBN particle, the electron intensity plot is also included as the insert of figure 23(a). This histogram morphology tells us that the height of hBN-NW is not uniform even within the grain. Therein, the core of hBN-NW has a higher height, the height decreases towards the either side of the hBN-NW, i.e. the two sides of the A and B points, or the top of the hBN-NW is tapered-end. Another HRTEM image recorded at a region with more complex hBN-NW structure, as shown in figure 23(b). This shows us that the degree of overlapping is greater between neighboring hBN-NWs. Both selected HRTEM regions show wrinkles, as indicated by the green arrows in figure 23. This wrinkling behavior may be due to distortion induced by defects of hBN-NWs [151].
The EELS spectrum of the hNW film deposited on the said Si 3 N 4 membrane was also recorded, as shown in figure 24. The EELS spectrum shows that the intensities of B-and N-edges are high enough to recognize the hBN phase. The sharp peak of the K-edge of sp 2 B is dominant and the very small peak of the C K-edge is also detected. The composition with C signals might come from contamination from sample handling processes. The C component present in the sample is not significant compared with the peak intensity of the C signal in other X-TEM specimens prepared with FIB, which will be discussed in section 5.7.

The growth mechanism of hBN-NWs on Si substrate
Based on the obtained results, the creation and growth of hBN-NWs on the Si surface were worked out. Those results are described through a growth model as shown in figure 25. Therein, a layer of BN material containing disordered phases such as aBN and tBN was coated onto the Si substrate surface at the early stage of the hBN film growth. The creation of these phases is due to the physical and chemical kinetic processes occurring at the substrate surface at which interacting of the ions present in the plasma environment and the Si substrate surface is dominantly driven. This disordered layer of BN material is about 10-30 nm thick. This thickness also depends on the concentration of H 2 present in the reactive gas mixture of RF sputtering. Herein, the initial etching of H ions will facilitate the creation of hBN-NWs earlier. This hBN-NWs created early, the hBN phase is enhanced. Figure 23. The two HRTEM images magnified from two typical areas in figure 22(a). One area consists of a well-defined hBN-NW configuration (a) and the other has a more complex configuration of hBN-NWs (b). In addition, an attached picture in (a) is an electron intensity profile extracted from the AB segment at which the A and B sites are at either side of the hBN-NW defined in the dashed ellipse.Reprinted from [39], Copyright (2017), with permission from Elsevier.
Further to the Si substrate surface, the deposited BN film dominantly contains the hBN phase. The thickness of such an hBN phase layer is around 50-100 nm, which means that the hBN phase in this layer is the largest. Therein, the hBN-NWs are well defined in the hBN phase. The porosity of this layer is usually high due to the NW-branches not yet created. The thickness of this material layer depends on T sub and d.
We found that the hBN-NWs are structurally heterogeneous with better hBN phase qualities in their cores than in either sides of the NWs. Moreover, the crystal planes of hBN-NWs are curled and distorted during deposition due to defects. These defects mainly come from the B sites rather than the N ones. When such Figure 24. An EELS spectrum was recorded from the hBN film grown on the Si 3 N 4 membrane as analyzed in figure 23. Therein, the Band N-K edge features were significantly exposed. This means that the hBN phase is highly existent. Moreover, the signal of the C-K edge is relatively low which means that the contribution of contaminated components is small due to the unprocessed TEM specimen. Figure 25. The growth mechanism of an imperfect hBN-NW at which it is nucleated from the Si substrate surface. Therein, the Si substrate is considered as a neutral surface, a thin BN layer deposited at the substrate surface is mostly a/tBN phase. Continuing to grow beyond the a/tBN layer, a block of hBN-nanosheets is grown perpendicular to the substrate surface and each hBN nanosheet contains many defects. Those defected sites are assigned to occur at the B sites rather than the N ones. The existence of such defects in the hBN-NWs structures will probably lead to a distortion in those NWs. Reprinted with permission from [36]. Copyright (2016) American Chemical Society.
vacancies are created during deposition, H free atoms are directly bound to the N terminated edges. The random N-H bonds thus exist in the grown hBN film. The density of N-H bonds was also investigated, this might correlate relatively with the density of defects in the hBN crystal films. Such N-H bonds are weak, they can therefore be affected by temperature. If Tsub increases, it is possible to make the movement of the N-H bond higher, leading H atoms can be broken and released out of the hBN film. The as-deposited hBN films can be stable at around 750°C, the N-H bonds are completely destroyed if the temperature is increased to 1000°C. If the N-H bonds are absent, the vacancies existing in the hBN film can migrate and form di-vacancies in the hBN film. Those vacancies and displacements make the hBN phase transform to the disordered ones such as aBN and tBN. If the thickness of hBN films is greater than 100-150 nm, the NW-branches highly create, those branches interact/overlap with each other and produce more defects than the previous layer of deposited hBN film. The creation and growth of those NW-branches result in the porosity of this hBN layer decreasing. The whole growth of hBN-NWs is quasi-perpendicular to the Si substrate surface. We emphasized that the unsaturated N edges in the hBN-NWs can also interact with the terminated edges of other adjacent hBN-NWs. Those N-edges also act as nucleation sites for the other adjacent hBN-NWs, resulting in hBN-NWs extending outward in different directions. Therefore, the dangling bonds of N edges can act as a nucleation agent for the growth of NWbranches or sub-hBN-NWs from the primary hBN-NWs.
Hence, the growth behaviors of the hBN films deposited with our unbalanced magnetron RF sputtering have been studied by changing in various parameters of the home-built RF machines such as d, T sub and α. A growth model of the hBN films deposited on the Si substrate surface described in figure 25. The quality of deposited hBN films depends on those studied factors. For example, hBN films are deposited at smaller d values, giving a higher R G , leading to a higher porosity and a higher crystallinity of the hBN phase, as compared to results obtained at larger d values. The orientation of the hBN-NWs does not change for the hBN films deposited at α = 90°and 0°, and increasing the α values only plays a role in reducing R G , the crystallinity of the deposited films decreases. This hints that the growth of hBN-NWs is dominantly governed by chemical processes rather than physical ones. Even so, undesirable phases such as aBN and tBN need to be reduced or suppressed. This is the main reason that we are looking for different methods to reduce unwanted phases to improve the quality of the deposited hBN films. Ultimately, the ability to change the quality or defect density of hBN films fabricated by our RF sputtering is possible by changing one of the d, T sub , α and content of H 2 in the reactive gas composition. 5.7. The growth mechanism of hBN-NWs on NCD film As discussed, it is beneficial to use the NCD surface for hBN deposition. Since there, the NCD surface has many free bonds of H at the terminated C edges. These free and weak bonds will act as springs at the NCD substrate surface. As a result, the ions present in the plasma environment of RF sputtering when interacting with the NCD surface will reduce the scattering and backscattering at the substrate surface much more than that of the Si one. Reducing scattering events leads to some effects such as bouncing, curling and chemical diffusion at the NCD surface are also reduced. This leads to faster binding between hBN crystal structures and the NCD substrate surface. Due to such initial predictions, hBN films were grown onto NCD substrates with a thickness of 300 nm. Using this NCD layer, the NCD particles at the NCD surface are also large enough and provide a flat surface for the hBN films to grow upwards. Figure 26 shows some surfaces of the thin films. Therein, the surface of the hBN film with a thickness of 300 nm deposited on the bare Si substrate is shown in figure 26(a). As seen, the hBN-NWs have fairly regular sizes which are oriented perpendicular to the Si surface, and the dimensions of the random hBN-NWs created from the Si surface have been discussed extensively in the previous sections. Figure 26(b) is an NCD thin film with a thickness of 300 nm grown on a bare Si substrate. Herein, the NCD particles with different shapes are seen clearly from the SEM image. This NCD film was then used as a substrate and deposited the hBN film on top, the morphology of the hBN film deposited on the NCD substrate is given in figure 26(c). A comparison of those surface morphologies of the two hBN films deposited on Si and NCD substrates is seen in figure 26(a) and figure 26(c). We found that the hBN-NWs tended to grow and follow the morphological structure of the NCD particles, denoted by the dashed yellow circle in figure 26(c). Those crystal faces of the NCD grains contain many C-H bonds, which will help accelerate the evaporation and adhesion of hBN phase. As a result, the size of hBN-NWs grown on the NCD substrate will be larger than that of the hBN-NWs deposited on the Si one [39]. In addition, the NCD particle size can also be estimated using the method described in section 5.5. A threshold image is generated by choosing the limit of minimum -maximum gray levels to estimate the grain size automatically, figure 26(e), cropped from an SEM image with an appropriate magnification, figure 26(d). As a result, more than 800 grains were selected and the particle size distribution function was then estimated, as plotted in figure 26(f). The average particle size was estimated to be (74± 8) nm, which is about nine times larger than the average width of hBN-NWs deposited on the Si substrate with the similar thickness of 300 nm. As we discussed in figure 8 that the diamond surface consists of many NCD particles with their facets oriented in different directions to the entire diamond substrate surface. These NCD facets are particularly terminated with H bonds. Moreover, the width of the 300 nm-thick hBN-NWs grown on Si(100) was averaged around 8.4 nm. Hence, hBN-NWs could be grown and localized directly from the facets of those NCD particles. This leads to the conclusion that the role of the facets of the diamond grains is very important, which localizes the hBN-NWs on each NCD grain.
The hBN films deposited onto the Si and NCD substrates were also examined for their crystallinity with Raman and FTIR spectroscopy analyses. Raman spectra of the hBN film and the NCD/hBN heterostructure on the Si substrate are shown in figures 27(a) and (b). The Γ 1/2 value of Raman spectrum corresponding to the E 2g vibration mode at ν = 1367 cm −1 was also estimated to determine the sp 2 hBN phase [10,134,152]. The calculated Γ 1/2 value of 26 cm −1 at the E 2g vibration peak for the hBN film, while this value is lower (Γ 1/2 = 18 cm −1 ) calculated at the E 2g mode for the hBN film deposited onto the NCD substrate. This further confirmed that the crystal quality of hBN phase for the hBN film grown onto the NCD substrate is higher than that deposited on the Si one. In addition, the Raman spectra of the hBN film deposited onto the NCD buffer also Figure 26. Top-view SEM images of several thin films used to compare their surface behaviors. Therein, a 300 nm-thick hBN film was deposited on the Si substrate surface at which the hBN-NWs are well-defined on the Si(100) surface. These hBN-NWs are randomly and uniformly distributed on the Si surface and their sizes and shapes are quite similar (a). Moreover, a 300 nm-thick NCD layer was evaporated onto another Si substrate where the grown NCD particles are clearly seen on the substrate surface (b), as its surface properties discussed in figure 7(b). Furthermore, another 300 nm-thick hBN film was deposited onto the given NCD film in (b), the surface of the resulting hBN film is shown in (c). Therein, the hBN-NWs are localized to the faces of the exposed NCD grains at which they formed clusters of hBN-NWs following the NCD particle shapes. The low magnification SEM image of the NCD film in (b) was recorded to use for the NCD particle size distribution calculation as was similarly realized for the case of hBN-NWs in figure 22, is seen in (d), and its selected NCD particle map shows in (e). The mean NCD size distribution function calculated from the SEM image in (e) using equation (5) is given in (f). phases are clearly seen. Since the sp NCD 2 phase contributed from the NCD grain boundaries, they are therefore scattered over a wide spectrum and the relatively low signals correlate with the signals of the remaining peaks.Reprinted from [39], Copyright (2017), with permission from Elsevier.
show that the characteristic properties of diamond are clearly revealed. Therein, the first-order Raman peak of diamond (sp 3 ) at ν = 1333 cm −1 and a non-diamond carbon peak (sp 2 ) at ν = 1487 cm −1 were also detected. The sp 3 bonds are present in the cores of the NCD grains, while sp 2 results from the carbonaceous structures present in the NCD grain boundaries [153].
FTIR spectroscopy is also used to detect vibrational modes at defects commonly occurred in hBN films, i.e. N-H at ν = 3437 cm −1 [144], B-C at ν = 1100 cm −1 [154,155] and sp 3 -BN at ν = 1085-1110 cm −1 [156,157]. FTIR spectra of the hBN films reported in figure 27, are shown in figure 28. Therein, a small peak was found corresponding to ν ∼1110 cm −1 , indicated by purple arrows, which means the presence of sp 3 BN phase or B-C bond. In fact, the sp 3 BN phase is difficult to form at low temperature (T sub = 125°C) and RF sputtering power is 75 W. Hence, the presence of sp 3 BN might come from disordered BN phases existing in hBN-NWs, this leads to a weak FTIR signal [156]. It is most likely that the vibration peak at ν = 1110 cm −1 comes from the B-C binding, where contamination has been observed in the samples and particularly at the interface of the NCD/hBN layers [155]. In addition, both samples have fundamental peaks corresponding to B-N bending (A 2u : ν = 817 cm −1 ), B-N stretching (E 1u : ν = 1376 cm −1 ) modes and optical phonon mode of sp 2 hBN [1,156]. Based on the given data, we realize that the hBN films integrated well onto the NCD substrate, the role of the NCD particles at the NCD surface is shown to be outstanding. In particular, the quality of the hBN film on the NCD buffer has been significantly increased.
To investigate in detail the interfacial properties of the NCD/hBN heterostructure, the X-SEM image of the heterostructure shows that the hBN layer grew directly onto the rough NCD surface. Since the NCD particles are much harder than the hBN-NWs, the X-SEM imaging was facing a difficulty when the magnitude contrast of the SEM image is highly dominated by the NCD particles as compared to that of the hBN-NWs. This leads to those hBN-NWs being blurred, as seen in figure 29(a). Therefore, the X-TEM image of the said sample was then recorded at which the X-TEM specimen was prepared by FIB, figure 29(b). Based on the contrast of the TEM image, the NCD particles are still clearly visible throughout the entire thickness of the NCD buffer and dominantly contributed onto the imaging contrast. However, the hBN-NWs in the grown film were strongly eroded by the high-energy ion beam of the FIB technique [123]. The NW-branches are more susceptible to erosion, and the rest of the hBN film mainly consists of the hBN-NW cores. To visualize in detail at the interface of the Si/NCD layer, BF-STEM and ADF-STEM images recorded in the area bounded by a white dashed square are given in figures 29(c) and (d), respectively. In particular, an interface area of the NCD/hBN heterostructure, marked by a yellow solid line rectangular was also characterized by the STEM image, as shown in figure 30.
The NCD layer is clearly visible to each NCD particle with the (100) and (111) NCD faces based on the STEM imaging contrast. The existence of ∼2.5 nm SiO 2 layer created during the substrate preparation is also visible. Therein, NCD grain boundaries are clearly seen where the growth process of NCD grains over time follows the growth mechanism of CVD diamond thin films [21]. In addition, the orientation of the sp 3 NCD and sp 2 hBN phases is clearly visualized in figure 30. The sp 2 hBN phase is directly grown from the NCD faces, the purity of this phase is high where the mixed a/tBN phase layer appearing in the case of the Si substrate has disappeared. The quality of hBN film is therefore increased in the case of depositing onto the NCD substrate surface, this result is in good agreement with the data obtained by calculating the value of Γ 1/2 from the Raman spectra in figure 27. In particular, the orientation of the hBN-NWs is always perpendicular to the NCD faces, as indicated by the arrows in figure 30(b). This caused a localization effect during the deposition of the hBN film onto the NCD buffer, which is also consistent with the data obtained from the SEM images, figure 26(c). A few bright spots are particularly seen in the ADF-STEM image, figure 30(b), these spots could be produced by adsorption of free C atoms generated during the FIB specimen cutting process with high-energy ion beams, denoted by yellow circles. This leads to the hBN-NWs that can be turned into a single or multi-photon emitting source at which many research groups are currently focusing on [54][55][56]. Even so this is one of the issues our research groups have been working on. In fact, we have not yet to fabricate hBN monolayers in which the defect sites must be controlled by doping elements such as carbon or some other foreign elements, which makes hBN monolayers capable of being a single or multi-photon emitter. Ultimately, the crystal structure of NCD particles located close to the Si surface are clearly seen where the crystal planes are clearly defined in the BF-STEM image, figure 30(c). A SiO 2 layer of ∼2.5 nm is also seen which is sandwiched between the NCD layer and the Si surface.
To exploit the distribution of composition in the interface of the NCD/hBN-NW heterostructure, as already seen in figure 30(a), the EELS spectra of the two areas located on the NCD and hBN-NW film were recorded, as seen in figure 31(a). These two EELS spectra are recorded in two regions denoted by blue and purple circles in the spectrum image, defined as a green square in figure 31(b). The sp 2 hBN and diamond phases are clearly shown in the above spectra. Moreover, a small amount of amorphous carbon was found with the bonding energy of π * = 285 eV, while the bonding energy states of both carbon symmetries at π * = 285 eV and σ * = 291 eV were also  worked out from the EELS spectrum recorded from the hBN layer. Although the X-TEM was not deposited with the carbon environment, the received carbon signal is however quite large. These carbon elements could be created during the FIB X-TEM specimen preparation [112]. This carbon signal is much stronger than the case of EELS spectra recorded from the TEM specimen without using FIB (figure 24). In particular, the broadened peak of σ * might result from the presence of an amorphous-like carbon phase [6]. This phase could be related to the white spots denoted in figure 30(b). However, some other researchers have suggested that carbon elements can be adsorbed or intercalated into the defected sites of hBN lattice, and it is a main factor that leads the hBN structures to emit single photons [76]. It is possible that our sample also has the ability to emit photons appearing as such white spots. Even so, whether single-photon or multi-photon emission needs to be further investigated in the future. Similarly, the energy states of boron symmetry at π * = 192 eV and σ * = 198 eV are also detected.
An image showing the distribution of B, N and C elements recorded from the EELS spectrum image is given in figure 32. Therein, the elemental compositions of B and N are represented in dark color, while the elemental composition of C is shown in white one. The element distribution map of B shows the existence of B elements is less in the hBN lattice regions or the B content in the high hBN crystallinity is less. This means that B content will be large in regions consisting of disordered BN phases. This conclusion is consistent with the data that we obtained from the hBN film growing on the bare Si substrate. There, we showed that N-terminated edges frequently occurred in the main-hBN-NWs ( figure 25). This also reconfirms in the map of the N elemental distribution, figure 32(b). The distribution of N elements appears randomly where the N atoms are scattered on the map with dark pixels and appears over the entire hBN film thickness. This means that more N atoms exist on the surface of the hBN specimen and they have a greater contrast than the B elements [53]. Finally, the map of C elements is given in figure 32(c), indicating that C atoms are derived from NCD particle regions and mainly belong to the sp 3 diamond phase. This is also consistent with the sharp peak of σ * observed in the EELS spectrum of the NCD layer, figure 31(a). A small amount of carbon distributed near the hBN/NCD interface was also detected which may be due to C re-deposition during FIB X-TEM specimen preparation using a high-energy Figure 31. (a) EELS spectra were recorded from two different areas at the hBN and NCD layers, indicated as blue and pink circles in the spectrum image (b). Therein, the signal level of the C-K edge in the NCD EELS spectrum is very large as compared to the case of the hBN EELS one. However, the C-K edge signal in the hBN EELS spectrum is assigned to originate from the FIB cross-sectioning process which is significant compared to that of the sample without the FIB preparation, as seen in figure 24. In addition, the mixed sp 3 and sp 2 NCD phase caused the signal at the peak of the C-K edge to split in the energy range of 280-340 eV. Ga + ion beam [125,126]. Both B and N elements have small Z numbers, resulting in a lower Z imaging contrast than in the case of C. Herein, the thickness of the X-TEM specimen also plays an important role at which the mirror contrast was also overwhelmed by the heterogeneity of the specimen thickness, especially a difference in the thickness of the NCD grains and the BN film.
The mechanism of nucleation and growth of hBN films on the NCD surface is clearly different from that of Si. Especially, the ability of the hBN phase film crystallized on an NCD substrate occurs earlier than the case of Si. There, we assume that the role of C-H bonds at the C-terminated edges of the NCD particle surface plays a dominant role. Due to such an early crystallization, the condensation process occurs rapidly at the early stage of hBN film formation. As a result, the growth rate of hBN films (R G ) will be greater than the case on Si substrate. Therefore, we verified this effect by comparing the RG values obtained from the growing hBN films on Si and NCD substrates with d = 3 cm and three different deposition times, t 1 = 66 min, t 2 = 150 min and t 3 = 240 min. The R G values were calculated for the samples depositing on the bare Si substrates, R G1−Si = 228 nm, R G2−Si = 280 nm, R G3−Si = 270 nm. The R G values were calculated for the samples growing on the NCD substrates, R G1−NCD = 336 nm, R G2−NCD = 320 nm, R G3−NCD = 242 nm. Based on the correlation of the above R G values, we can conclude that the hBN phase was crystallized faster on the NCD surface at the early stage of thin film growth with a short growing time (< 150 min) or a thin hBN layer. This conclusion is consistent with the results discussed above. Even so, when hBN-NWs are rapidly grown and those NWs are localized following NCD facets, the interaction among those hBN-NWs has occurred, leading to the creation of early NW-branches. The overlapping of those NW-branches will reduce the growth rate of the hBN films when using the NCD substrate. While the BN film deposited on the bare Si substrate does not have these limitations, the thickness continues to increase rapidly. This conclusion is truly reflected in our data, R G3−NCD < R G1−Si at t 3 = 240 min.
Based on the above data, we propose a growth mechanism of hBN-NWs on the NCD substrate, as shown in figure 33. Therein, the formation of hBN-NWs is mainly a chemically oriented rather than a physical process, resulting in hBN-NWs oriented perpendicular to the NCD substrate surface, regardless of whether the substrate is neutral (Si) or C-terminated edges (NCD). As seen in the interface of NCD/hBN heterostructure, the presence of disordered BN phases is mostly absent where the hBN phase seems to grow directly onto the NCD surface. Even so, the hBN-NWs still oriented perpendicular to the substrate surface in the entire film, the hBN-NWs typically changed their orientation slightly during the first growth stage, figures 30(a) and (b). This implies that the flow of incoming ions, denoted by the green arrow in figure 30(b), means that physical effects do not affect too much on the growth direction of hBN-NWs. Hence, what is the fundamental process that promotes directly the nucleation of hBN on diamond to obtain hBN-NWs at which the addition of H 2 to the reactive gases of the RF sputtering is thought to be a prerequisite [128]. Using the NCD substrate where NCD particles are terminated by H-dangling bonds when a NCD film takes off from the CVD reactor. The growth mechanism of hBN-NWs deposited on the NCD substrate surface was already proposed [39], which can be described as the followings: The proposed growth model starts with the removal of an H atom and the substitution of a sputtered B ion at a C-H transient bond on the NCD surface [21,158], to initiate an hBN ring, figures 33(a) and (b). The sputtered N ions can connect to the B site to form a third of the hBN ring with two N-terminated edges, figures 33(c) and (d). The other two sputtered B ions are further connected to the given N sites, figure 33(e). Finally, one of the sputtering N ions locks the two bonds of the residual B atoms together, forming the first hBN ring in the lattice of hBN-NWs, figure 33(f). The processes as mentioned above occur in parallel, and continuously to form hBN-NWs. Since the C-H bonds are very random on the facets of the NCD particles. This leads the growing process asynchronously and can cause defects in the hBN lattice. The growth mechanism of hBN-NWs deposited on the NCD substrate herein is a hypothetical model obtained from our experimental results. In principle, both C-B and C-N bonds can be formed on the NCD surface at which the binding energies of C-B and C-N are 356 kJ/mol and 305 kJ/mol, respectively. Hence, we can create another model of the growth of hBN-NWs on the NCD substrate with an initial bond of C-N, similar to the growth mechanism with a starting bond of C-B. However, the results obtained in literature showed that C-B bonds are more stable than C-N. This means the B atom easily incorporates into the diamond surface during CVD growth rather than N. Those bonds can be estimated theoretically with their EELS spectra, while experimental results are more difficult to realize with FIB-TEM specimen preparation, and TEM alignment is also an issue.

hBN-NWs grown on Cr/Au heterostructure buffer
As discussed in sections 5.6 and 5.7, two different growth mechanisms are involved in the development of hBN-NWs depositing onto Si and NCD substrates. The quality of hBN phase is enhanced when growing the hBN material onto the NCD surface instead of the Si one. Therein, a layer of disordered BN phases is eliminated when the hBN film is deposited onto the NCD substrate. However, the entire hBN films still tend to grow perpendicular to the substrate surface whether it is Si or NCD at which the chemical role of the interface layer between the grown material and its substrate is assigned to be important affecting the quality of the hBN crystalline phase. Due to the given situations, we continued to carry out the deposition of hBN films onto the Cr/ Au bilayer substrates. We hoped to use the benefits of transition metal elements in this heterostructure as catalysts in the deposition process, to orient the material layers parallel to the substrate surface instead of the perpendicular orientation as in the two cases above. Herein, the characteristics of the substrate surface is a key factor in the nucleation of BN phases [83,159]. To minimize the kinetic energy of the free ionized particles at the substrate surface, some weakly chemical reactive metals such as Au, Pt, Pd, Ir, are often used as a buffer layer [1]. However, if we use the single metal layer mentioned above, the required temperature is high during the deposition process. Hence, we have coupled an additional layer of Cr (t Cr = 10 nm) below to create good adhesion of the Au layer (t Au = 100 nm) with Si substrate surface. The Cr/Au heterostructure is capable of reducing their vaporization temperature to a certain temperature within the working temperature range of our RF sputtering, < 600°C.
A 300 nm-thick hBN film was grown onto the Si/Cr/Au substrate at d = 3 cm and T sub = 450°C. The reason for using this temperature is due to the melting temperature of the Si/Au composited structure as we predicted that it is in the range of 350-480°C. Since the eutectic point of Au-Si compound is 363°C, the melting point of this heterostructure is much lower than the melting point of each structure composition, i.e. Au (1063°C) and Si (1414°C). A low magnification HAADF-STEM image of the given X-TEM specimen recorded is shown in figure 34(a). The Cr/Au substrate with their Z numbers are of 24 (Cr) and 79 (Au), which are much larger than the Z numbers of N, B and C. At T sub = 450°C, the Cr/Au structure could be partially melted [152,[160][161][162]. Based on the Z imaging contrast of the STEM images, we can see that the Au layer has the brightest contrast and formed discontinuously as material droplets with a diameter of (145± 15) nm. The crystallization of the hBN layer seems to be improved at the Au/hBN interface, where the hBN-NWs appear to initiate and grow above the Au droplet regions. Bundles of hBN-NWs were formed, denoted by white U-shaped curves. Those bundles of hBN-NWs formed close to the Au droplets. This may give us an initial view of the enhanced hBN phase formation in the presence of Au. The elemental distribution through the cross-sectional TEM specimen measured with the energy dispersive x-ray (EDX) image is given in figure 34(b). Therein, the Cr/Au substrate appears heterogeneous when the Au layer is formed as islands or droplets. In fact, the Cr layer was only 10 nm, so it was barely detected at the expected location, it was possibly buried by a thin layer of Si redeposited from the Si substrate during FIB milling [163][164][165][166][167]. Some residual Cr accumulated as a droplet, located between the Si substrate and the hBN film, distinguished by the purple region in figure 34(d). This Cr droplet may be due to the migration of Cr elements during deposition at T sub = 450°C, where chemical diffusion effects can occur [167]. All three elements are scattered throughout the protective Pt layer and the hBN film, which was possibly resulted from the X-TEM FIB specimen preparation [126,127].The region containing the Cr component, indicated by the yellow arrow in figure 34(b) which was magnified and shown as a BF-STEM image in figure 35. Therein, some Au islands formed at the Au/hBN interface are clearly seen, and the Cr/Au bilayer thickness is not uniform, as seen in figure 35(a). The diffraction fringes of the hBN crystal structure at the Au/hBN interface are also recognized, as indicated by colored arrows in figure 35(b), this image was recorded in an area as marked by the cyan rectangular box in figure 35(a). It is clear that the hBN phase forms directly in the early stage of hBN film growth. The orientation of the grown hBN sheets seems to be locally dependent on the geometry of each Au island. This situation is similar to the case when hBN-NWs grow following NCD particle shapes, as already discussed in section 5.7. Therefore, hBN-NWs will grow competitively after a certain film thickness and the shape of Au droplets directly induces the orientation of the hBN-NWs, figure 35(b).
BF STEM images recorded at three different locations near the Au island are shown in figure 36. In the first area, as seen in figure 36(a), the hBN sheets constituting hBN-NWs appear to be parallel to the Au island surface, denoted as blue arrows. The growth direction of hBN-NWs tends to be vertical to the substrate surface.  Moreover, the diffraction fringes of Au crystals are also indicated by white arrows. A very thin layer of aBN is however observed at the Au/BN interface which might result from a random sputtering of N and B ions condensing on the Au surface or the contaminated Au surface during sample preparation prior to deposition of hBN film. In the other locations, as seen in figures 36(b) and (c), the mixed growth directions of the different bundles of hBN-NWs, where their lattice fringes are clearly seen, are annotated with different colored arrows. Thus, it is clear that the tBN and hBN phases already exist. Even the tBN phase remained, however the crystallinity of hBN film was significantly improved, as compared to the case where hBN films were grown onto the Si substrate at which the two films were deposited under the same conditions of RF sputtering.
The above results show that Au layer has the ability to increase crystallization of hBN film, especially at the interface of Au/hBN layers. However, the morphological behavior of Au as islands or drops influenced the growth direction of hBN-NWs at the Cr/Au substrate surface. Because of the inhomogeneity of the Cr/Au substrate surface that leads to the competition of hBN-NWs occurring earlier than in the case of hBN-NWs grown on the flat Au surface. Moreover, the C signals were assumed to originate from the X-TEM specimen preparation. Therefore, the EELS spectrum of the X-TEM specimen was also recorded at the interface of the Au/ hBN structure, as shown in figure 36(a), the EELS spectrum was analyzed as figure 37. Therein, the EELS spectrum of the specimen shows that the C peak intensity is significant as compared to the main peaks of the Band N-K edges at which the spectrum was recorded at the hBN film. Such significant C K-edge signals come from the FIB sample preparation, as already discussed in figure 31. This conclusion is reconfirmed by the EELS elemental maps of B, N and C, as given in figure 38. The distribution of B and N elements in those maps is less obvious due to the low Z contrast, while the distribution of C atoms is mostly present on both sides of the  Herein, the B and N signals are dominant shown, while the C signal is also clearly seen, this confirms the C signal generated from the X-TEM specimen preparation using the FIB technique.
interface of the Au/hBN regime. This data is consistent with the results obtained from the EELS spectrum as discussed in figure 37(a).
It is clear that the crystallinity of hBN films deposited on the Cr/Au buffer substrate is significantly enhanced. Similar to the other substrates, the overall orientation of the deposited hBN-NWs remains vertically oriented relative to the substrate surface. This may result from competition of hBN-NWs nucleated on those substrate surfaces to minimize the total free energy for the whole nucleation process [159]. The results obtained on the Cr/Au substrate also show that chemical processes greatly induce the formation of hBN-NWs at the early stages of thin film development. Using a Cr/Au bilayer substrate will improve the quality of the hNB phase at the Cr-Au/hBN interface and particularly the hBN-NWs oriented parallel to the Au droplet surface. However, it is difficult to control Au droplet size and substrate surface morphology when increasing Tsubin our current experimental setup. Despite this, the enhanced crystallization of the hBN phase would be beneficial for future applications using such optical properties of this material [168][169][170]. The structural characteristics of grown hBN films with less a/tBN phase and controllable orientated hBN-NWs can be particularly of interest in photonic devices where fundamental understanding the orientation of hBN-NWs with respect to the conducting characteristics of hBN film is important [1,13,[168][169][170][171][172].

Wetting and other properties of hBN film
As discussed in the previous sections, we have succeeded in fabricating the highest possible crystallinity hBN-NWs with our homebuilt RF sputtering. Therein, the hBN crystalline phase was improved when the hBN film was directly bound to the NCD substrate. Furthermore, the ability to orient the hBN monolayers or nanosheets is possible when using the impact of the Cr/Au bilayer substrate at T sub = 450°C which reacted as a catalyst for the crystal growth process at the substrate surface, resulting in the reduction of unwanted phases such as a/tBN. Even so, such grown hBN films are still porous, which contain many defects and especially high surface roughness. In order to use such characteristics for the application purposes, we also measured the wettability of hBN films. Therein, we wanted to analyze the dependence of the wettability of grown hBN films as a function of surface roughness or the size of the hBN-NWs projected on their substrate surface. In other words, we investigated the wettability of the material with respect to the change of hBN film thicknesses.
Two series of hBN films were deposited at d = 3 cm and d = 6 cm, with their thicknesses ranging from 100 to 900 nm. SEM images of those films were recorded, as shown in figure 39. Similar to the results discussed in figure 11. There the size of the hBN-NWs projected onto the substrate surface increases with the thickness of the hBN films for both d-values. Based on the SEM imaging contrast, the main difference in the morphology of the two hBN film series is that the widths and lengths of the hBN-NWs grown at d = 3 cm are larger than that of the case for the hBN-NWs deposited at d = 6 cm at the same hBN film thickness.
We used the contact angle (CA) method to measure the wettability of the given films in respect of the water droplets. This is a useful method [173][174][175] in determining the wettability of any solid plane interacting with a liquid, indicating indirectly the degree of wetting of the solid material to water or another solvent. In fact, if a liquid droplet drops on a solid surface, the wetting properties of the given surface depend on several typical factors such as the solid surface tension (g ), wetting behaviors of a solid material. Hence, experimentally measuring the contact angle (θ Y ) value is an indirect route to understanding the surface tension of a solid material [173,174]. Therein, the contact angle is defined as the angle between the intersection of the liquid-solid interface (e.g. H 2 O-hBN) and the liquid-vapor interface (e.g. H 2 O-air). Herein, we used the droplet size in all our measurements as 15μl. In principle, the contact angle of a liquid drop on an ideal solid surface was first described by T. Young [176], where θ Y is determined by the interaction of three interfaces in equilibrium, which is  figure 39, we conclude that if the roughness of hBN films is high, the water droplet-air interaction plays a dominant role in the CA value. This result is quite consistent with the results of the original theoretical work realized by Wenzel and Cassie-Baxter [173,177]. In this case, each hBN-NWs can be considered as a hierarchical structure or a groove shaped on a smooth surface. The shape and height of the hBN-NWs are strongly related to the surface roughness factor, which is directly seen in the SEM images in the upper panels of figures 40(b) and (c). In addition, interstitial spaces between hBN-NWs also play an important role, where air pockets exist in those interstitial spaces that repel water droplets from the surface of hBN films, as described in the lower panels of figures 40(b) and (c).
Moreover, there are many external forces that affect the value of CA such as gravity at which droplet size and droplet type are also important factors. Therefore, the wetting characteristics of hBN films in respect of some other liquid droplets such as oils or dyes which can largely depend on the intrinsic properties of hBN films, i.e. defect density, impurities, surface roughness and the porosity or the hBN-NW shapes projected on the the hBN surface plane. To determine quantitatively the impact of the above parameters on the final CA value, it is still open questions for us, and we are also working on. The intrinsic properties of hBN materials are of interest, for example, (i) defects in hBN films are related to the ability of hBN films adsorb pollutants which can be used in Figure 39. SEM images of two series of the hBN thin films with different thicknesses deposited at d = 3 cm (a)-(d) and d = 6 cm (e)-(h), sputtered at the substrate temperature of 125 • (T sub ). Their surface behaviors expose that, the hBN-NWs have different sizes, especially the roughness of those films is different at which the surface of these films is considered as an idea plane and that was etched with hBN-NWs in differing shapes and different spaces between those hBN-NWs. This can result in wetting properties of the given films to water, oils and dyes. water purification technology; (ii) defects capable of absorbing foreign elements at their vacancies and having the ability to emit single/multi-photons, which can be used in quantum information encoding technology; (iii) the temporary adsorption mechanism of the H atom during the deposition of hBN-NWs is very important, if another gas such as CO 2 or CH 4 is substituted for H 2 , where they both react as reactive gases and they have the ability to intercalate into the lattice of hBN during hBN creation. Such properties of hBN materials with those conditions remain challenging and are hot topics for many research groups.

Some features in porous hBN nanostructures
The porous hBN nanostructures as we have been investigating, the ability to control the nucleating and growth of hBN-NWs on a substrate material array needs to be investigated more further and deeply. As an advantage, hBN films deposited by an unbalanced RF sputtering will be useful for a large-scale material coating purpose. The quality of the grown hBN phase films is highly influenced by experimental parameters. We have concluded that the hBN material is more porous when depositing at positions closer to the BN target or d is small. At the positions far away from the BN target or d is large, the density of grown material increases, disordered BN phases easily exist. On the other hand, increasing substrate temperature also improves the quality of the hBN film, T sub = 250°C was assigned to be the most suitable deposition temperature to obtain the best quality of the grown hBN films. The porous hBN film is thermally stable below 1000°C in high vacuum, if higher than that temperature the desorption of N-H bonds at the defected sites leads to many chemical processes occurring inside the hBN film. Due to the di-vacancy creating process and this breaks the hBN crystal structure, the hBN material turns into poorly ordered phases. The process of rotating the substrate only reduces the growth rate per unit area of the substrate plane for the same deposition time, as compared to the case of the un-tilting substrate. This situation is the same as we place the substrate plane away from the BN target, resulting in greater material density and more unwanted phases present. The hBN-NWs in all cases tended to grow perpendicular to the substrate surfaces. Hence, it is also an advantage to use hBN-NWs as field electron-emission sources in some potential applications.
Altering the surface characteristics of the substrate materials, the quality of the hBN phase is enhanced for the two cases of deposition onto the NCD and 450°C-heated Cr/Au substrate surfaces. The growth mechanisms of the hBN films deposited onto typical substrates were proposed based on the detailed experimental data. Therein, hBN-NWs were directly nucleated on both above substrates without a thin layer of disordered aBN/tBN BN phase. There is little difficulty in controlling the surface roughness of the Cr/Au buffer at the melting temperature of the Cr/Au bilayer. Therein, the hBN-NWs were localized on Au droplets which we need to further investigate by reducing T sub to a certain point in the range of 363-450°C. The Cr/Au bilayer buffer not only suppress the aBN phase created at the interface of the hBN film and the Cr/Au substrate, but also  figure 39. (b), (c) The schematics of the two thin films deposited at d = 3 cm with thicknesses of 100 nm and 900 nm, respectively. Herein, the film with a thickness of 100 nm which has small hBN-NWs, uniform surface and density of hBN-NWs is higher than that of the 900 nm-thick film containing large hBN-NWs, non-uniform surface and the density of hBN-NWs is low. Differences in surface behaviors of hBN films lead to different wettability in respect of some liquids, i.e. water, oils and dyes. produces a thin hBN layer close to the Cr/Au surface which tends to orient the hBN nanosheets parallel to the substrate surface. The lattice spacing between the covalent planes of the hBN monolayers in an hBN-NW was twisted, and this hints that a certain level of defects exists in those hBN monolayers, leading to change the conducting properties of those grown hBN films. The N-H bonds are abundant in the porous hBN films at which H dangling bonds are highly available in the reactive gas composition of our RF sputtering. Our data is consistent with the results reported in the literature [53] where B vacancies are preferable in such hBN monolayers. Therefore, the presence of H atoms cooperating in the hBN lattice, indirectly indicates the existence of B vacancies in hBN-NWs. During the growth of hBN-NWs, they grow unevenly, leading to hBN wrinkles that persist in the very thin hBN layer, resulting from defects or lattice distortion. Furthermore, an advantage of using NCD substrates is that the morphology and grain size of NCD particles can be controlled by the addition of N 2 during the growth of the NCD films with the CVD technique [162]. NCD layers with higher sp 3 content are more favorable for hBN films growing with uniform widths of hBN-NWs. This highlights the feasibility of this deposition approach for future coating purposes with negative electron affinity or field emission applications [1,13].

Concluding remarks
As a highlight for this review, we have analyzed in detail the growth mechanisms of the hBN films containing hBN-NWs deposited on three different substrate surfaces, e.g. Si, NCD and Cr/Au. Therein, the Si surface is considered as a neutral material in which the terminated Si atoms at the substrate surface are very inert during the growth process of the hBN film. Besides, the NCD surface is more active at which the H dangling bonds can connect to the C-terminated edges at the facets of the NCD-sp 3 particles and their sp 2 boundaries. The weakly C-H bonds at such surfaces acting as springs, and the interaction of reactive gas ions in the plasma environment of RF sputtering with those at those facets will alter the crystallographic properties of grown hBN films. As a result, the hBN films grown on the NCD substrate have better crystallinity than the hBN films deposited on the Si one. In particular, the grown hBN-NWs are localized following the morphology of each NCD particle, leading the hBN-NWs deposited on the NCD substrate being more porous than on the Si. Those hBN-NWs were directly grown and oriented vertically to the facets of each NCD particle, this makes a competition in the growth orientations of hBN-NWs occurring early. This means that the disordered BN phases are compressed at the interface of NCD/hBN films, but they are more likely to exist between hBN-NWs due to the overlapping of those NW-branches. As for using the Cr/Au substrate, the crystallinity of the hBN film is enhanced, the hBN-NWs are however localized following the Au droplet surfaces, thus controlling the fineness of the Cr/Au surface is still an issue that needs further explorations. In particular, the hBN-NWs grown on the Cr/Au substrate changed the orientation of the deposited hBN-NWs parallel to the surface of the Au droplets with a very thin layer. That is, we can fabricate hBN films with crystal orientation parallel to the Cr/Au substrate at a local scale and at a relatively low working temperature (T sub = 450°C) using our homebuilt RF sputtering.
Based on the obtained data analyzed above, some issues need to be solved and need further studies. For example, it is necessary to optimize the suitable Cr/Au substrate temperature to control the roughness of the Cr/ Au surface where Au droplets could be present, which will affect the growth orientation of hBN material relatively parallel to the substrate surface on a large scale. Moreover, highly crystallized hBN-NWs can be produced at an optimal condition, but defects still exist, so it is necessary to control these defects using other reactive gases such as CO 2 or CH 4 . There, the role of C, O, H impurities should be exploited systematically. Assuming that if the said impurities coexist in hBN-NW lattices, how the intrinsic properties of hBN-NWs will change. In fact, one research group has found that by absorbing C atoms into the vacancies of hBN monolayers, such monolayer is able to be a single-photon emitter source [55,56]. Particularly, the temporal adsorption properties of foreign atoms or molecules present in the plasma environment such as CO, CH 3 , SO 2 at which such gas molecules could be adsorbed into the defect site of hBN monolayer need to be studied systematically with a theoretical approach to save time and effort. Finally, as we discussed, transition metal elements act as catalysts for the nucleation of the hBN phase, the interaction between some transition metals such as Au, Ni, Pt, Pd, in respect of the hBN crystalline monolayers needs to be elucidated.
(DQH). This is also supported by Duy Tan University (XHC) & the Military Institute of Mechanical Engineering (DKP), Vietnam.

Data availability statement
The data that support the findings of this study are available upon reasonable request from the authors.
Author contribution statement DQH conceived the research, carried out the deposition of hBN films and realized some fundamental measurements such as Raman and FTIR spectroscopy, analyzed the obtained data, and drafted the manuscript. NHV, TQN, TDH, XHC and DKP planned for the main tasks of the mini-review, arranged references for the manuscript, searched and compared our results with data from other research groups during the writing/ editing process of the manuscript and co-supervised the project. Some advanced measurements such as FIB and S/TEM were supported by Hasselt University and University of Antwerp, Belgium.