Nanopipes in Gallium Nitride Nanowires and Rods


Gallium nitride nanowires and rods synthesized by a catalyst-free vapor-solid growth method were analyzed with cross section high-resolution transmission electron microscopy. The cross section studies revealed hollow core screw dislocations, or nanopipes, in the nanowires and rods. The hollow cores were located at or near the center of the nanowires and rods, along the axis of a screw dislocation. The formation of the hollow cores is consistent with effect of screw dislocations with giant Burgers vector predicted by Frank.


Semiconducting nanowires represent a new class of device building blocks with properties enhanced by their small size and large aspect ratios. Gallium nitride (GaN) nanowires in particular have received attention due to their unique material properties. GaN nanowire-based blue/UV lasers are an especially desirable application and are under investigation by several groups. [1] [2] [3] In these reports, optical pumping of GaN nanowire lasers has been successfully demonstrated; nevertheless electronic pumping for a compact all-solid-state device has not yet been clearly realized. Electronic pumping requires well-defined energy states in a highly crystalline material, which is an essential characteristic for efficient lasing action.

Nanowires are known for their high crystallinity. However, very recent results indicate that nanowires may possess an internal structure 4-9 that is not readily apparent even with investigation by plan-view high-resolution transmission electron microscopy (HRTEM). In this paper, we report evidence of hollow core screw dislocations with giant Burgers vectors in GaN nanowires and rods. We refer to the hollow core screw dislocations as nanopipes. This is the first direct evidence of this structure in nanowires and rods similar to the micropipes that are known to form in thin films.

Micropipes have been extensively investigated in silicon carbide (SiC) thin films 10, 11 and recently in gallium nitride (GaN) thin films, [12] [13] [14] driven by their use in blue-UV quantum well lasers. Theories developed by Eshelby and Stroh, Frank, and Tunstall et al. have been successfully applied to describe these formations. 15 However, early research by Frank and Eshelby also explored the possibility of nanopipes in thin cylinders as well as thin films. Eshelby first described axial screw dislocations in thin cylinders with isotropic elastic theory. 16 Frank then explained the formation of a hollow core screw dislocation through a total energy minimization approach. 17, 18 He proposed that the strain around a screw dislocation with a Burgers vector that exceeds a critical size will remove material next to the dislocation to reduce strain energy, resulting in an additional surface or a hollow pipe. Frank-Eshelby theory for nanopipes in thin cylinders now takes on new significance as these conditions can be physically realized in nanowires and rods.

The GaN nanowires and rods examined in this study were synthesized by way of a catalyst-free direct reaction of gallium vapor and ammonia. The details of this growth method are described elsewhere. 19, 20 Nanowires and rods with two orientations were obtained using this growth method and depended on the furnace growth temperature. 21, 22 At furnace growth temperatures between 850 and 950°C, multiphase zinc blende/wurtzite GaN nanowires with triangular cross sections ranging between 60 and 150 nm in width, were commonly obtained. 4, 5 The growth orientation was along the 〈011〉 direction for zinc blende and the 〈21 j 1 j 0〉 direction for wurtzite. At 1000°C, single-phase wurtzite GaN nanowires and rods with hexagonal cross sections, ranging between 200 and 5000 nm in width, were commonly obtained. The singlephase wurtzite growth orientation was along the [0001] direction.

Cross-sectional investigations of the nanowires grown at 850°C identified the internal arrangements which render the multiphase configuration energetically feasible. 4 These results motivated a similar cross-sectional investigation of the nanowires grown at 1000°C. The cross-section studies for nanowires and rods grown at 1000°C revealed hollow core screw dislocations, or nanopipes, in the nanowires and rods. The hollow core was located at or near the center of the nanowires and rods, along the axis of the screw dislocation.

Recently, independent research identified an axial screw dislocation mechanism in PbS nanowires using TEM analysis perpendicular to the axis of the nanowire. 9 In the investigation presented here, the TEM analysis was parallel to the axis of the GaN nanowires and rods by way of cross sections. Therefore, screw dislocation mechanisms have been reported in two different nanowire systems, using complementary perpendicular and parallel TEM investigations. Figure 1a shows a scanning electron microscope (SEM, Hitachi 4700-II) image of the matrix that formed at a growth temperature of 1000°C. The base of a GaN rod in the center is contrasted with a much thinner GaN nanowire in the foreground, indicated by the arrow. This image emphasizes the different structures that can form with this growth method at this furnace temperature. A lower magnification view of the entire ∼100 µm long rod is shown in Figure 1b . A typical plan-view TEM (JEOL 2200FS, 200 kV) image of another GaN rod is shown in Figure 1c . The HRTEM image, upper inset, shows a well-resolved lattice, indicating high crystal quality of the rod. A selected area electron diffraction (SAED) pattern of the rod, lower inset, was solved for the wurtzite crystal structure along the [011 j 0] zone axis with the growth axis oriented along the [0001] direction.

Cross sections for HRTEM (JEOL 2200FS, 200 kV) were fabricated using a focused ion beam (FIB, FEI Quanta 200 3D) system. The FIB was used to extract and mill nanowire/ rod samples until they were transparent to a TEM electron beam. For cross section preparation, the nanowires/rods and growth matrix were placed in ethanol and sonicated to free the nanowires/rods from the matrix. The suspensions were then deposited on silicon wafers that had a layer of native oxide. A layer of gold was thermally evaporated on the surface to provide initial protection from ion beam damage. A thick layer of platinum, ∼4 µm, was deposited using reactive gas/ion interaction so that the nanowires would not be damaged during the ion milling process. The Pt/Au/ nanowire/rod layers were ion milled and extracted from the substrate using a micromanipulator and placed and secured on TEM grids using reactive gas/ion Pt deposition. The samples were milled further using the FIB so that they were thin enough to be transparent to the TEM electron beam.

An SEM image of a cross section taken from a group of nanowires and rods is shown in Figure 2a . The rods, roughly 1 µm in width, are labeled 1-4. The nanowires, roughly 200 nm in width, are labeled 5 and 6 and are shown in the insets. Of note are the small dark contrast features observed near the centers of rods 1, 2, and 3 and nanowire 5 consistent with topographical contrast from small pores or voids, discussed further below. The schematic drawing in the upper right corner of Figure 2a shows the growth direction, [0001], and side facet planes, {101 j 0}, of the nanowires and rods. An SEM image of the nanowire and rod group before the cross-sectioning process began is shown in Figure 2b . A dotted line indicates where the cross section was taken. A higher magnification image of the base of the nanowires and rods prior to cross sectioning, with the substrate tilted 52°r elative to the electron beam, is shown in Figure 2c . An ion beam image of the nanowires and rods midway through the extraction process embedded within the Pt/Au layer is shown in Figure 2d , with the sample tilted 52°relative to the ion beam. The circle marks the location of the four rods embedded within the Pt/Au layer, which were deposited to protect the sample from ion beam damage.

Experimental results of the TEM investigations of the rod and nanowire cross sections are shown in Figures 3 and 4 . Figure 3a shows a bright-field TEM image of rods 1 and 2. Spiral bend contour patterns 30 were observed in both rods. The arrows in this figure indicate bend contour pairs, which will be discussed below. SAED patterns of rods 1 and 2, shown in panels b and c of Figure 3 , were solved for the wurtzite crystal structure along the [0001] zone axis. Light contrast hexagonal features were observed near the center axis of both rods 1 and 2. TEM images (not shown) identified the light contrast hexagonal features as having {101 j 0} side facets, parallel to the external rod facets. Panels d and e of Figure 3 show HRTEM images of the hexagonal features. Contrast variations in both images implied that these areas were thinner than the surrounding areas. The lightest contrast areas were completely transparent to the electron beam, indicative of holes. It was noted that the open hole edges were at different heights, as confirmed by adjusting the microscope focus. A few nanometers of amorphous material, possibly an oxide layer, was also observed. In Figure 3f , a scanning TEM (STEM) image of rods 1 and 2 shows the interface between the two rods as a thin line of dark contrast. The fringe contrast at the interface is consistent with stacking faults along the {101 j 0} planes as shown in Figure 3g . Figure 4 shows bright-field TEM images of rods 3 and 4 and nanowire 5. The orientations of rods 3 and 4 were determined using SAED shown in the insets. The orientation of nanowire 5 was determined from an FFT shown in the inset, as this nanowire cross-sectional area was too small for reliable SAED imaging. All three were solved for the wurtzite crystal structure along the [0001] zone axis.

The TEM image of rod 3 in Figure 4a did not exhibit the spiral bend contour patterns of rods 1 and 2. Furthermore, HRTEM could not be obtained from the rod 3 cross section. This implies that the cross section was too thick at the rod 3 location. A hexagonal feature was observed near the rod center, boxed area. An HRTEM image of the central hexagonal feature, shown in Figure 4b , indicates that it had {101 j 0} side facets parallel to the external rod facets and a light contrast area completely transparent to the electron beam, similar to rods 1 and 2. The open hole edges were at different heights, as confirmed by adjusting the microscope focus.

The TEM image of rod 4 in Figure 4c showed spiral bend contour patterns similar to those found in rods 1 and 2. The dark contrast area observed near the center axis was hexagonal in shape but not free of material, as shown in Figure 4d .

The TEM image of nanowire 5 shown in Figure 4e did not exhibit spiral bend contour patterns. The light contrast area observed near the center axis is shown in the HRTEM image of Figure 4f . It appears to be two hexagons side by side in the same orientation. However, periodic variations in nanopipe direction could create a “double” appearance from a single screw dislocation line, as discussed below. The hexagonal feature was not free of material, but its lighter contrast suggested that it was thinner than the surrounding nanowire.

Nanowire 6 was similar in diameter to nanowire 5. Highresolution imaging could not be obtained for the nanowire 6 cross section, indicating that the sample was too thick at this location, and it is not discussed further.

Spiral bend contour patterns and hexagonal features near the rod/nanowire center axes have been observed in the GaN rods and nanowires in this study. The spiral contrast is interpreted as resulting from the presence of a screw dislocation with a giant Burgers vector. The presence of the spiral contrast pattern in a TEM cross-section image results from the relaxation that occurs around the screw dislocation when the thin TEM sample is produced. The elastic twisting of a thin disk containing a screw dislocation has been described by Eshelby and Stroh. 23 The contrast feature that results has also been described by Tunstall et al. 24 and Hirsch et al. 25 for a single dislocation in a thin TEM foil. Real radial displacements increase with decreasing distance from the dislocation line and are proportional to the Burgers vector. This results in diffracting columns being tilted close to the dislocation (in opposite directions on opposite sides of the dislocation), thereby creating diffraction contrast bend contours. The tilting of the columns decreases as the distance from the dislocation increases, resulting in the weaker bend contour contrast observed radially outward. It should also be noted that in the present case the imaging was carried out using a more coherent field emission gun, and subsequently the bend contours are more apparent than might be expected with the thermionic gun used by Tunstall et al.

Bend contour pairs were observed for rod 2, single bend contours were observed for rods 1 and 4, and bend contours were not observed for rod 3 and nanowire 5. Bend contour contrast varies significantly with thickness of the TEM sample for two reasons: changes with elastic relaxation with foil thickness and changes in TEM contrast effects with foil thickness. Where the cross sections are thinner, as indicated by well-resolved HRTEM lattice fringes, there is more relaxation and the spiraling bend contours associated with the twisting become evident. In very thin areas, the bend contours appear as pairs, due to kinematical electron diffraction imaging conditions. The pairs form at positive and negative Bragg conditions ((g) with zero deviation (s ) 0) from the Bragg condition, and the bright region between the bend contour pairs results from negative deviations from these Bragg conditions (s < 0). 24 With somewhat thicker foils, dynamical electron diffraction conditions apply, and diffracting conditions of s < 0 result in strong anomalous absorption, resulting in single bend contours appearing over conditions of +g to -g. 24 With thick samples, bend contours are difficult to observe due to less elastic relaxation in the cross section and lack of contrast. The thickness of the cross section at individual rod and nanowire locations could be inferred from HRTEM images. HRTEM images with well-resolved lattice fringes (not shown) for rods 1 and 2 indicated a thin cross section. HRTEM images with lattice fringes (not shown) were obtained with difficulty for rod 4 and nanowire 5 indicating thicker cross sections. HRTEM lattice fringes were not resolved for rod 3 indicating that the cross section was even thicker at the rod 3 location. The thickness observations correlate with appearance of the bend contours, except for nanowire 5. The smaller nanowire radius could decrease the elastic relaxation resulting in a lack of apparent bend contours. Furthermore, the bend contours were observed to move during tilting of the sample (i.e., Figure 4 , panels c and d), consistent with the well-known contrast behavior of bend contours. If the observed spiral patterns were caused by dislocations or other crystal features, one would expect very limited movement of the contrast from tilting.

Hexagonal features near the rod/nanowire center axes were observed for rods 1-4 and nanowire 5. Energy dispersive X-ray spectroscopy (EDS) carried out on the cross sections indicated no measurable differences in the chemistry of the hexagonal central regions and the outer regions of the cross sections, with only Ga and N peaks being observed. The HRTEM imaging was carried out under multibeam phase contrast conditions, but some diffraction contrast effects were also apparent. The contrast of the hexagonal features might therefore occur from phase contrast effects associated with differences in thickness and defocus and diffraction contrast from variations in crystal orientation. Atomic mass differences were ruled out by the EDS results. The HRTEM, SAED, and FFT images indicate that the central hexagonal features have very similar, if not identical, orientations, as the outer regions, which ruled out diffraction contrast. The observed contrast differences between the central hexagonal features and the outer regions are consistent with contrast variations associated with the central regions having different effective thickness than the outer regions. In the cross sections from rod 1 (Figure 3e ), rod 2 ( Figure 3d ), and rod 3 (Figure 4b ), the lightest contrast areas were completely transparent to the electron beam, indicative of holes.

In order for a nanopipe to form in GaN, the Burgers vector must exceed a critical dimension that is based on the shear modulus, µ (116 GPa for GaN 26 ) and surface energy, γ (1.9 J m -2 for the GaN {101 j 0} surface 27 ). This critical Burgers vector is given by the energy minimization approach described by Frank 17 and discussed for GaN by Pirouz. 28 The Burgers vector B is given by

B > 40πγ µ

where 40πγ/µ equals 2.06 nm for GaN; thus the Burgers vector magnitude must be greater than 2.06 nm in GaN for a hollow core to form. The unit cell length for wurtzite GaN is 0.52 nm along the c-direction, so 2.06 nm corresponds to about four unit cells. It should be noted that this minimum value is an estimate since the surface energy of a nanopipe surface is probably greater than that of a flat surface. 27, 28 Simple estimates of the Burgers vectors were obtained from Frank’s relation between B and hollow core radius r 0 17 with r 0 estimated from the sizes of the hexagonal features. The Burgers vectors for rods 1-4 and nanowire 5 were g5.1 nm. This simple estimate assumes isotropic elasticity and does not take into account the anisotropy of the GaN wurzite single crystal structure. All of the nanopipes investigated in this study had features that indicated a complex tube structure, rather than an idealized nanopipe with straight inner walls. In the cross sections from rod 4 and nanowire 5, significant contrast and no open holes were observed within the central hexagonal features. These observations are similar to the complex cores of nanopipes reported by Cherns and Hawkridge, with staggered pyramidal voids (ref 29 Figure 1b) or irregular nanopipes (ref 29 Figure 1c ). Staggered voids may also explain the “double” hexagonal feature in nanowire 5. The slight darker/lighter contrast variation between the hexagonal features suggests that their thicknesses are different. That they are side by side suggests a staggered structure consistent with an irregular nanopipe.

The observations in this study show evidence of hollow core screw dislocation formation consistent with recent observations by Bierman et al. 9 that identified an axial screw dislocation mechanism in PbS nanowires. In contrast to the Bierman study, our observations were carried out in cross section with the electron beam parallel to the Burgers vector. Experimental evidence in two different nanowire systems obtained using complementary parallel and perpendicular TEM investigations strongly suggests that axial screw dislocation mechanisms may be an important aspect of nanowire formation and should be investigated for many nanowire material systems. If confirmed as a general trend, axial screw dislocation mechanisms, with the possiblity of hollow core formation, will signficantly impact nanowire device design.

Nano Lett., Vol. 8, No. 12, 2008


  • Jacobs, B. W. Radiation Resilience in Reduced Dimensionality Systems. Ph.D. Thesis in Electrical & Computer Engineering, Michigan State University, 2007.
  • Hirsch, P. B.; Howie, A.; Nicholson, R. B.; Pashley, D. W.; Whelan, M. J. In Electron Microscopy of Thin Crystals; Butterworths:London, 1965.
  • Williams, D. B.; Carter, C. B. in Transmission Electron Microscopy: A Textbook for Materials Science; Spinger: New York, 1996. NL802198H

Leave a Reply