Growth orientation and type of internal structures are both observed to change abruptly as a function of growth temperature in catalyst free growth of gallium nitride nanowires. In the present work, corresponding temperature-dependent changes in the growth matrix substrate that can affect the availability of nucleation sites and influence the reactivity of constituent adatom materials in catalyst-free nanowire growth are investigated. The influence of Ga vapor pressure and an abrupt change in the availability of single versus molecular adatom constituents is identified as a possible controlling parameter.
Gallium nitride (GaN) nanowires have been under extensive investigation in recent years due to their unique optical  and electrical properties  as well as their potential advantages in enabling new device applications [3, 4] . For device applications, control of crystalline quality and orientation is essential. To develop controlled nanowire growth processes, it is critical to understand nanowire nucleation and growth mechanisms.
The nanowires examined in this study were grown using a catalyst-free growth method  at furnace growth temperatures of 850°C, 950°C and 1000°C. Catalyst free nanowire growth [5, 6, 7, 8] results in no metal impurities in the nanowires and can be used in applications where device fabrication steps are incompatible with the presence of a metal catalyst [9, 10] . In previously reported work [11, 12, 13] , we demonstrated that the nanowire growth orientation was influenced by growth temperature. Between 950°C to 1000°C, an abrupt change in wurtzite growth orientation from € 112 0 to ⟨0001⟩ is observed. We further reported that the GaN nanowires had internal structures that continued along the entire length of the nanowires. The majority of nanowires grown at 850°C and 950°C had multiphase wurtzite and zinc-blende crystalline domains that persisted along the entire nanowire length. The growth direction was ⟨110⟩ for the zinc-blende domains and € 112 0 for the wurtzite domains as stated. Very recently, zinc-blende/wurtzite multi-domain GaN nanowires have been reported by another group  . In contrast, the majority of nanowires and rods grown at 1000˚C were single-phase wurtzite with a ⟨0001⟩ growth orientation and a different internal structure consisting of a single nanopipe  . A nanopipe can develop from relaxation of a large Burgers vector screw dislocation  . While ours is the first reported observation of nanopipe formation in the GaN nanowire system, evidence for formation of an internal screw dislocation in the related PbS nanowire system has been reported by another group  .
In the present work we identify corresponding temperature-dependent changes in the growth matrix substrate. These changes in the growth matrix substrate have a direct effect on the availability of nucleation sites and subsequent catalyst-free growth of nanowires with the observed orientations and internal structures.
Preparation Of Samples
The GaN nanowires nucleated on a polycrystalline growth matrix, whose formation preceded the onset of nanowire growth. Details of the growth conditions in a quartz tube furnace correlated with flow rate and temperature are given in Ref.  . TEM  and X-ray diffraction  to date have found only wurtzite growth matrix with no evidence of zinc-blende.
The top, bottom and side (fracture cross section) surfaces of 850°C, 950°C and 1000°C growth matrix samples were characterized using scanning electron microscopy (SEM) using a Hitachi S-4700-II FESEM. The growth matrix morphology was observed to change as a function of growth temperature. The 850°C and 950°C growth matrix crystallite formations were similar. The top surface consisted of approximately 1µm-sized crystallites. Nanowires with triangular cross-section widths ranging from 50nm to 250nm were observed to grow from the top surfaces of the 850°C and 950°C growth matrix samples. Figure 1 (a) is a side view image that also shows the sub-surface of the growth matrix near the top. A hexagonal platelet morphology was observed (close-up shown in inset)
At 1000°C, three different crystallite formations were typically observed at the top surface of the growth matrix. Type 1 (top of figure 1 (b)) consisted of relatively large 1-20µm crystallites of GaN. A hexagonal block morphology was typically observed for the large crystallites (close-up shown in inset). Large hexagonal rods, with widths ranging from 0.5µm to 30µm, grew from the Type 1 matrix. TEM confirmed that the rods had ⟨0001⟩ growth orientation. Type 2 (bottom left) consisted of medium sized 0.1-1µm crystallites. The Type 2 growth matrix at 1000°C appeared to be comparable to the 850°C and 950°C Figure 1 . SEM images of different morphologies of (a) 850˚C GaN growth matrix (side view and top surface) and (b) 1000˚C GaN growth matrix (top view). Arrows mark areas shown at higher magnification in insets. growth matrix and also nucleated similar nanowires, with no rods observed. Type 3 (bottom right) consisted of small 50-300 nm crystallites typically clustered in spherical mounds. Nanowire growth was also observed from Type 3 growth matrix. The nanowires that nucleated from Type 2 and 3 matrixes at 1000°C had triangular cross sections associated with € 112 0 wurtzite-phase orientation  in nanowires that nucleated from 850°C and 950°C matrix samples. The bottom of the growth matrix (the side that formed on the quartz tube in the growth furnace) was also imaged using SEM to study the evolution of the matrix. In figure 2 , the top and bottom sides of two pieces of matrix from the 850°C growth are shown side-by-side. In figure 2 (a), the morphologies of the matrix on both surfaces appeared similar, however, nanowire nucleation was only observed on the top surface of the matrix, as shown in figure 2 (b) . The top and bottom of the 950°C growth matrix were similar, with nanowire growth observed only on the top surfaces. In contrast, the bottom surface of the 1000°C growth matrix was as active in nanowire nucleation as the top surface. A low magnification image of the bottom surface of a 1000°C growth matrix sample is shown in figure 3 (a) with rod growth apparent. A close-up of the area at the base of one rod (solid box) in figure 3 (b) shows that nanowire nucleation was also present. Nanowires were observed to nucleate from areas of the matrix with smaller crystallites, while the rod nucleation sites were ‘buried’ in a growth of smaller crystallites. A close-up of a different point on the same sample (dotted box) in figure 3 (c) shows that some nanowires display nodules and branching. This morphology indicated the onset of dendritic growth.