Morphology of InN films
A microscale SEM image of an InN film deposited on a Si substrate with 1.2 × 10
4 laser pulses in nitrogen (10
5 Pa) is shown in Figure . The InN composites aggregate into dendrite nanostructures (Figure ). Besides the fractal structure, some spherical droplets of InN or indium are scattered on the substrate (Figure ). The indium droplets are ejected from the indium target because of its relatively low melting point. The efficiency of droplet's formation depends on the experimental conditions. For example, the formation of InN droplets by metalorganic vapor phase epitaxy or metalorganic chemical vapor deposition was due to indium surface segregation when the ammonia concentration was insufficient [
26,
27]. Once a droplet attains its critical size, it becomes thermodynamically stable and continues to grow [
18]. At the same time, efficient nitridation of droplets is taking place either on the target's surface or in the plume following laser ablation at 157 nm [
17-
19].
The InN structure was first identified by EDXS, (Figure ), recorded 2 days after the growth of film. The spectrum indicates the presence of either the amorphous or crystalline InN [
28]. Oxygen was non-detectable, confirming the low oxidization rate at 2 days after the growth of film.
The growth of InN crystals by 157-nm PLD sidesteps some constraints imposed by the low reactivity of nitrogen. The InN crystalline phase is formed either through the ionic interaction of the molecular species on the target, or in the plume near the target's surface. Indeed, two 157-nm photons are required to ionize the N2 molecule from its ground electronic state. This condition is satisfied directly on the target because of the high-intensity-focused laser beam.
At longer film deposition times (2 h, approx. 105 laser pulses, 20 mJ per pulse), the morphology of the InN film appears to be denser and thicker, with two distinct morphological features (Figure ). The first feature type represents irregular dendrites (Figure ), while the second feature represents nanorod-like structures (Figure ), with mean width and height being 40 and 50 nm, respectively, Figure . The nanorods are growing on the top of 100-nm long (Z-direction) three-dimensional dendrite islands. The surface roughness distribution (Z-direction) (Figure ) is asymmetric for films grown for longer period of time. The surface roughness distribution has a major peak at 44 nm and a secondary one at 85 nm.
Irregular-shaped grains observed previously [
7,
29], were associated with nucleated InN hexagonal crystal islands (grown along the
Z-direction) that eventually were enlarged laterally along the InN
plane. Moreover, in previous studies, it was confirmed that the surface morphology of InN samples was influenced by the lattice mismatch (InN: hexagonal,
a = 0.354 nm,
c = 0.570 nm. Si: diamond cubic
a = 0.543 nm), between the samples and the Si (111) substrate leading to the growth of three-dimensional islands with size and density depending on the substrate temperature [
29].
The lattice mismatch is the source of shear stresses at the interface that tailors the morphology of the film [
30], e.g., dendrite structure in our case. The InN crystal has a lattice mismatch with most of monocrystalline substrates, and therefore different morphologies are expected to grow on different substrates [
31].
High-resolution images of InN nanostructures
The capturing of HRTEM images of InN crystals is not a trivial issue because InN is unstable under electron irradiation [
32], due to its low dissociation energy (0.073 eV) [
33]. The InN crystals first are transferred to a lacey-carbon-coated Ni grid using a scalpel knife, and then they are transferred to the carbon-coated TEM grid. The samples are examined under mild electron beam conditions; current density of the order of 10
-5 nA/nm
2, 50-μm condenser aperture, and spot size 3. With this method, although the information of relative orientation between the InN and Si substrate is lost, there are no artifacts such as structure alteration related to ion-milling. In addition, ion-milling is not recommended because InN is thermally dissociated. The electron beam destroys the InN structure by changing the morphology of the dendrite structures, and new aggregations of 20-40 nm nanospheres distributed selectively around the periphery of the grid cells are formed (Figure ). The change of the film morphology is accompanied by structural changes and the EDXS of the areas exposed to e-beam and a deficiency in nitrogen (Figure ). This is additionally confirmed by SAED, and the image of one indium nanosphere (Figure ) reveals the tetragonal crystal structure of pure indium (Figure ). In addition, no traces of oxygen are identified.
Using mild electron beam conditions two days after the PLD experiment, hexagonal hcp-InN crystal (Figure , circle A) and amorphous phases (Figure , circle B) approximately 10 nm wide are identified.
The long-term phase transformation from the hcp-InN phase to the metastable fcc-InN structure reported previously [
8] is undetectable because of the growth of the film on a different substrate where the lattice mismatch provided the required strain for phase transformation.
Structure and morphology of long-term-oxidized InN nanostructures
Three different methods are used for specifying the stoichiometric composition of the crystal nanodomains: EDXS, EELS, and SAED because results among these three methods may vary. This inconsistency is not only due to the presence of stray X-ray radiation at EDXS, especially at rough films with broad size distribution of nanodomains but also due to the difference between the cross sections of the electron and the X-ray beams (5 and 450 nm, respectively).
After storing the films for 6 months at ambient conditions, the co-existence of crystal In2O3 and amorphous In-Ox-Ny, together with the InN crystal phase is identified. The HRTEM images indicate the superposition of crystal areas of different contrast and parallel Laue levels, shown within the periphery of the circles C1, C2, and A in Figure and and in Figure , respectively. The image of Figure is recorded immediately after the growth of the film, while that of Figure after 6 months. It is likely that these images correspond to the growth of 2D crystal structures on the top of each other forming thus 3D structures. The picture resembles the collection of randomly overlaid coins with aligned heads.
This point is further strengthened by previous investigations on the size of monocrystals. It was found that they exhibit a sharp Gaussian distribution at the nanoscale (approx. 10 nm) [
34]. It is unlikely therefore that long 3D crystal structures are grown either in the plume, or directly on the Si substrate with PLD. The relatively small size of the crystal nanostructures is credited to the short range of ionic forces that are responsible for the short spatial correlation between the atoms of the crystal structure. Three different nanostructured groups with average size of 5 (A), 25 (B), and 40 nm (C) are identified, Figure , although larger crystal domains are grown through the aggregation of small nanocrystal domains (<5 nm). Furthermore, crystal structures with their Laue levels slightly rotated with respect to each other (Figure ) should correspond to a superposition of uncorrelated crystal nanostructures [
34] (separated by a relatively large distance in comparison to the correlation length). The HRTEM images indicate the superposition of crystal areas of different contrast and with parallel Laue levels, indicated within the periphery of the circles C1 and C2 (Figure ) and within the areas E and F (Figure ). Furthermore, various amorphous nanodomains are formed, such as the ones within the periphery of the triangle and the circle B (Figure ), and the areas within the periphery of the circles E and F (Figure ).
The nanoclusters tend to nucleate primarily on the substrate, but the size distribution of nanodomains is an indication that crystallization in the plume is possible too.
The concentration of oxygen in the film, measured with EDXS, is constantly increasing during the storage period and is uniform in areas of the same film morphology while the presence of oxygen is noticeable in the entire film area indicating the post-ablation oxidization process. A typical EDXS recorded 3 and 6 months after the film growth is shown in Figure .
The average experimental concentrations of nitrogen and oxygen are affected negligibly as observed from the X-ray absorbance, and the molar ratio of InN/In2O3 is estimated as 0.3/0.7. Indeed, owing to the MIV and MV absorption edges of In, the mass absorption coefficients (MACs) for OKα and NKα are 1.9 × 104 and 3.7 × 103 cm2/g, respectively. In addition, owing to the K absorption edge of nitrogen, the MACs of OKα and NKα lines in nitrogen are 1.7 × 104 and 1.5 × 103 cm2/g, respectively. The calculated attenuation of OKα line in a 10-nm-thick crystal (0.5InN/0.5In2O3) is approx. 10%, and the attenuation of the NKα line is approx. 2%.
In addition, the elemental analysis with EDXS depends on the film morphology, and therefore EELS elemental analysis is used on small (<5 nm), and larger crystal and amorphous domains. The average concentration of elements (mol%) of the small-sized domains with EELS is 40% In, 40% O, and 20% N, respectively (Figure ). No traces of nitrogen are detected on large-sized domains as can be seen in Figure . These results imply that oxidization starts by forming a small nucleus of the oxidized areas in the InN nanodomains with dimensions less than 5 nm. Taking into consideration that the surface of InN films is highly porous, the atmospheric oxygen diffuses first into the InN grid and then slowly oxidizes InN to an amorphous In-Ox-Ny phase (Figure ). The amorphous phase eventually undergoes a second-order phase transition to a crystal line In2O3 phase, when the size of the oxidized areas exceeds the critical value of approx. 5 nm. The results of EELS and HRTEM suggest that the chemical reaction is accompanied by agglomeration of small-sized oxidized domains, forming larger crystal domains of indium oxide only.
Besides EDXS and EELS, the levels of oxidization of InN films are measured by SAED for different areas of the samples (Figure ). The diffraction spots are fitted to the hcp-InN and bcc-In2O3 phases (the patterns are taken with 450-nm diameter of the electron beam). Taking into consideration the rotational average of the experimental SAED patterns, the intensity distribution of the diffracted electrons is deconvoluted using Diff Tools scripts in Gatan's Digital Micrograph software. Next, SAED intensity distributions for pure hcp-InN and bcc-In2O3 are calculated using the EMS program code. The best fit between the experimental diffraction patterns and the calculated intensity distributions is obtained by the linear combination approach (sum of weighted parts of both the simulated distributions for InN and In2O3). After an examination of the samples stored for 6 months in ambient conditions, we found approximately equal concentrations of hcp-InN and bcc-In2O3 phases (50% InN(hcp)-50% In2O3(bcc)). The best fit to the experimental patterns is achieved for this concentration, which is in agreement with the EDXS results, where the concentration of oxygen is constantly increasing. When the same sample is stored in isopropanol for the same duration, the best fit is for 80%hcp-InN/20%bcc-In2O3, confirming once more that oxidization takes place after the growth of film from the atmospheric oxygen.
The SAED images were obtained with fast Fourier transmission (FFT) on HRTEM images. This technique eliminates part of the information from the initial image, while preserving the periodical information, such as the lattice plane fringes. In the experimental SAED pattern, six clear rings made of discrete spots can be fitted (Figure ).
The experimental SAED patterns are compared to the simulated patterns of the intensity distribution of the diffracted beam (yellow inlets, Figure ) for three different phases (a, bcc-In2O3; b, hcp-InN; and c, bcc-In). The experimental SAED patterns are usually slightly distorted because of non-optimized alignment of the microscope and the nonlinearity of the CCD camera.
First, the d spacing values are calculated for cubic bcc In2O3 nanostructures. The poor matching of peak positions between the calculated and the experimental profiles of Figure (bcc-In2O3) indicates that additional crystalline phases besides bcc-In2O3 and hcp-InN are present, forming an alloy texture of hcp-InN/bcc-In2O3, and that a certain amount of oxygen is present in the form of amorphous In2O3 or In-Ox-Ny non-detectable with SAED. This result is in agreement with the direct identification of the amorphous phase from the HRTEM images (Figure ).
The same procedure is applied for pure hexagonal hcp-InN and tetragonal indium. In all the cases, the agreement with the experimental SAED patterns is lacking.
The sum of simulated patterns for 50% hcp-InN and 50% bcc-In2O3 is in agreement with the experimental ones, indicating the coexistence of cubic bcc-In2O3 and hexagonal hcp-InN phases in the films (Figure ).
Amorphous In-Ox-Ny and In2O3 nanostructures
It is generally accepted that InN is transformed first to the amorphous In-O
x-N
y/In
2O
3 phase and then subsequently to the crystalline In
2O
3 phase either by aging or annealing. It has been reported that after the initial growth of InN and subsequent formation of oxynitride, a mixed crystalline phase of In
2O
3 and InN crystals was grown after a short annealing time (10 min) at 550°C [
35]. The cubic In
2O
3 phase was formed layer-by-layer, after subsequent oxidization of the hexagonal InN layers [
36].
Substitution of nitrogen by oxygen is possible in principle, since both atoms have similar atomic radii [
37]. At room temperature and up to 800 K, the heat of formation of crystalline In
2O
3 (221 kcal/mol) is higher than the heat of formation of crystalline InN (34 kcal/mol) [
37], and therefore, even in the presence of oxygen, InN is preferably formed.
In the present study, InN is growing in pure nitrogen, and thus, the formation of In2O3 during the ablative phase is negligible. Furthermore, the oxidization of the post-ablated pure indium nanodomains from the atmospheric oxygen or the direct transformation of the crystal phase of InN to In2O3 are improbable to occur because oxidization via those pathways is energetically unfavorable. Following the film's exposure to atmospheric oxygen, the oxidization of InN is taking place slowly. The initially physisorbed oxygen eventually becomes chemisorbed inside the InN lattice. Therefore, the oxidization of the InN crystal should take place through the formation of an intermediate amorphous phase.
The single crystal layers are clearly identified, and the inset of FFT of real images and the experimental results indicate that the crystal In2O3, InN, and the amorphous In-Ox-Ny oxynitride phases coexist in the form of neighboring distinct nanodomains after the long-term exposure of InN to atmospheric conditions (Figure ).
1 Zoe Kollia,1 Goran Dražic,2 Spomenka Kobe,2 and Nicolaos Spyropoulos Antonakakis1
