Catalytic Effects of Cr on Nitridation of Silicon and Formation of One-dimensional Silicon Nitride Nanostructure

The catalytic effects of chromium (Cr) on the direct nitridation of silicon (Si) and morphology of nitridation product were investigated. Cr dramatically improved the conversation of Si to silicon nitride (Si3N4). The complete conversion was achieved at 1350 °C upon addition of 1.25 wt% Cr. This temperature was much lower than that required in the case without using a catalyst. Meanwhile, Cr played an important role in the in-situ growth of one-dimensional (1-D) α-Si3N4 nanostructures. α-Si3N4 nanowires and nanobelts became the primary product phases when 5 wt% Cr was used as the catalyst. The growth processes of the 1-D α-Si3N4 nanostructures were governed by the vapor-solid mechanism. First-principle calculations suggest that electrons can be transferred from Cr atoms to N atoms, facilitating the Si nitridation.

Scientific RepoRts | 6:31559 | DOI: 10.1038/srep31559 In the present work, the effects of Cr with various addition levels (up to 10%) on the direct nitridation of Si were investigated, and morphologies of Si 3 N 4 products examined in detail. To assist understanding the role played by Cr in the nitridation process, first-principle calculations were also carried out. The experimental and calculated results were discussed, based on which, the relevant catalytic reaction mechanisms proposed. Figure 1 illustrates the effects of Cr content on the overall conversion (OC) of Si and α -phase contents in samples resultant from 3 h nitridation at different temperatures. At 1200 °C and 1250 °C, the OC of Si to Si 3 N 4 in the reference samples without Cr was low (Fig. 1a). However, it increased evidently with increasing the Cr content. For example, at 1250 °C, the OC was only ~21% in the case of no Cr addition, but increased significantly to ~66% in the sample containing 10 wt% Cr. Upon further increasing the temperature to 1350 °C, the OC in the reference sample increased to 91% whereas nearly all of the Si had been nitrided in the sample containing 1.25 wt% Cr. Figure 1b shows α -phase contents in the samples containing different amounts of Cr after 3 h nitridation at different temperatures. With increasing the nitridation temperature, the α -Si 3 N 4 content decreased significantly, which was in the agreement with that reported previously 15 . The effect of Cr addition on the α -phase content appeared to be rather complicated. Overall, at low temperatures, the α -phase content decreased with the Cr addition, whereas at high temperatures, it initially decreased and then increased with the Cr addition. We think that the complicate effect of Cr on the α -phase content might be explained by the following two reasons: 1) one hand, considering the low thermal conductivity of Cr (93.7 W/mK) comparing with that of Si 3 N 4 (~450 W/mK) and Si (150 W/mK), the excess reaction heat arisen from the nitridation of Si might not be released immediately, resulting in local overheating and α → β phase transition of Si 3 N 4 ; and 2) on the other hand, the high amount of added Cr might act as diluents to absorb certain reaction heat generated from the nitridation process, then reduce the local overheating and decrease the phase transition of Si 3 N 4 . Figure 2 shows XRD patterns of samples containing 0-10% Cr after 3 h nitridation at 1250 °C and 1350 °C respectively. At 1250 °C, unreacted Si peaks remained as the main phase in the reference sample. With increasing the Cr addition, the α /β -Si 3 N 4 peaks increased evidently whereas Si peaks decreased, indicating the great accelerating effect of Cr on the Si nitridation. Minor Cr 2 N was detected in the sample containing 10 wt% Cr. At 1350 °C,  α -and β -Si 3 N 4 were identified in the reference sample, along with some unreacted Si. On the other hand, in the sample containing 1.25 wt% Cr, Si disappeared and only α -and β -Si 3 N 4 phases were present, indicating the complete conversion from Si to Si 3 N 4 . On increasing Cr to ≥ 5 wt%, α -and β -Si 3 N 4 remained as the primary phases, however, minor Cr 2 N and two other impurity phases (appeared to be Cr 3 Si and Cr 5 Si 3 ) were detected. Shown in Fig. 3 are SEM images of samples after 3 h nitridation at 1350 °C. Some unreacted Si was identified in the reference sample without Cr by EDS (the inset 1 in Fig. 3a), which was consistent with the XRD results shown in Fig. 2b, and a few one-dimensional (1-D) nanostructural phases were occasionally seen in the sample. EDS (the inset 2 in Fig. 3a), along with the XRD results ( Fig. 2b), confirmed that the crystalline phases surrounding the unreacted Si were Si 3 N 4 , suggesting that the nitridation of Si particles proceeded from the surface towards the center. Compared to the reference sample, much less unreacted Si phases but much more Si 3 N 4 phases were identified by EDS (the insets in Fig. 3b-d) in the samples containing Cr. Furthermore, with increasing the Cr content from 1.25 to 5 wt%, the quantity of 1-D nanophases also increased evidently.

Effects of Cr additions on Si nitridation.
High-magnification SEM images (Fig. 4) further reveal that most of the 1-D nanophases actually possessed wire-, belt-and branched belt-like morphologies. The nanowires were 50-200 nm in diameter and about 50 μ m in length, and the nanobelts were 300-1000 nm in width and about 10 μ m in length. However, upon addition of > 5% Cr, the quantity of 1-D nanowires/nanobelts appeared to be decreased and more Si 3 N 4 particles coexisted with them. Figure 5 further presents TEM, HRTEM, EDS and SAED of a representative nanowire formed in the sample containing 5 wt% Cr, after 3 h nitridation at 1350 °C, showing smooth surface and uniform diameter of the nanowire (Fig. 5a). EDS (Fig. 5b) further revealed that the nanowire contained Si and N in the atomic ratio of 0.751, almost the same as the stoichiometric ratio (0.750) of Si 3 N 4 (Cu peaks were from the copper grid sample holder), verifying that it was Si 3 N 4 . Also SAED (Fig. 5c) confirmed that is was single-crystal α -Si 3 N 4 . In addition, two-dimensional lattice fringes with d-spacing values of 0.56 nm and 0.67 (Fig. 5d) matched with the (001) and (100) planes, respectively, suggesting that the α -Si 3 N 4 nanowire grew along the [001] direction. Apart from Si 3 N 4 nanowires, as mentioned above, many nanobelts were also formed in the Cr added samples. As revealed by TEM (Fig. 6a), they had different widths but their widths were uniform along the entire length. EDS (Fig. 6b), SAED (Fig. 6c) and HRTEM (Fig. 6d) identified that they were also single crystal α -Si 3 N 4 but were grown along the [101] direction. It should be pointed out that no particles were observed and no Cr was detected by EDS at tips of the 1-D Si 3 N 4 nanowires and nanobelts (Figs 5b and 6b), suggesting that their growth processes should not have been dominated by the well-established vapor-liquid-solid (VLS) tip-growth mechanism 28 . The detailed growth mechanism regarding this will be discussed in more detail in Section 3.3 below.
In order to illustrate the role of Cr in the formation of these nanostructure materials, TEM images of the fired sample containing 10 wt% Cr were also taken. As shown in Fig. 7a, two types of α -Si 3 N 4 nanostructures, i.e. nanowire with 100 nm in diameter and nanobelt with 400 nm in width, simultaneously grew from a particle. EDS analysis (insets in Fig. 7b,c) confirmed that the main parts and tips of the 1-D nanostructures contained only Si and  N, but no Cr. Interestingly, Cr element was only detected at the root of the 1-D nanostructures (by EDS, Fig. 7d), which is believed to have acted as a "catalyst center" for the nucleation of the 1-D α -Si 3 N 4 nanostructures.

Mechanism of Cr catalyzed nitridation.
The results presented and discussed above (Figs 1~7) suggested that Cr had played significant roles in accelerating the Si nitridation. To assist clarifying these roles, DFT calculations at the GGA-PBE/USP level of theory were carried out to simulate the adsorption behavior of a N 2 molecule onto the Cr (001) surface and study the catalytic mechanism of Cr catalyst for nitridation reaction. The adsorption energy (E ad ) of N 2 on the para-position (90.24 kJ/mol) of the Cr (001) surface is higher than that on the ortho-position (66.24 kJ/mol) ( Table 1), suggesting that adsorption of N 2 onto the former is more favorable than the latter. Moreover, the bond lengths in a N 2 molecule adsorbed on the para-and ortho-positions of Cr (001) surface are respectively 1.176 and 1.195 Å, which are longer than the original bond length (1.158 Å) in a free-standing N 2 molecule (Table 1). Such increase in the bond length of N 2 , is believed to have assisted the dissociation of N 2 , as discussed previously 21 .
The Mulliken atomic charge distributions of N and Cr atoms ( Fig. 8 and Table 1) reveal that the N atoms are indeed negatively charged, whereas the Cr atoms are positively charged, providing further evidences on the electronic charge transfer from the latter to the former. In the case of para-position absorption, the two N atoms in a N 2 molecule possess identical electronic charge. Moreover, the two N atoms absorbed on the ortho-position of Cr (001) surface have negligibly different negative-charges, due to the asymmetrical surrounding of the two N atoms, as also reported in our previous paper 21 . Since the Π 2py * molecular orbital is an anti-bonding level, the bonding strength in a N 2 molecule will be weaken if the adsorbed N 2 molecule receives an electron from a Cr atom in the p-state. Changes in both bond length and Mulliken charge suggest that the N≡ N bond can be weakened and the relatively stable N 2 molecule activated when it is absorbed onto the surface of Cr.

Discussion
As well documented in the literature 6,14,19 , the catalytic growth of 1-D nanostructure is generally controlled by the well-established VLS mechanism. The presence of a catalyst particle at the tip of a nanowire/nanorod/nanotube is often regarded as one of the main supportive evidences for this mechanism 29 . In the present work, Cr was found at the roots of the 1-D α -Si 3 N 4 nanostructures, suggesting that Cr had played dominant roles in the nucleation of α -Si 3 N 4 and the subsequent growth of the 1-D nanowires and nanobelts (Fig. 4). The detailed mechanisms can be schematically illustrated in Fig. 9. For the case of 1-D α -Si 3 N 4 nanostructures, in the initial stage, N 2 molecules diffused onto the surfaces of Cr particles (Fig. 9a). As predicted by the first-principle calculations (Section 3.2), the bond length in the N 2 molecules would be increased and the bond strength decreased after their adsorption onto the Cr particles (see Fig. 8), resulting in activated N 2 molecules. Subsequently, such N 2 molecules would react with Si vapor generated from Reaction (1) (Fig. 9b-1), forming Si 3 N 4 which would nucleate forming "crystal seeds" (Reaction (2)) on the surface of Cr (Fig. 9c-1). Owing to the hexagonal structure of  30 . In this case, crystal surfaces with lower energies tend to serve as the enclosure surfaces, so the incoming Si and N preferred to diffuse to and deposit on the high energy surfaces (001) and (101) in the length directions [001] and [101], respectively, resulting in simultaneous formation of nanowires and nanobelts (Fig. 9d-1,e-1, Figs 5 and 6). Considering that the residual Cr catalyst was only detected in the roots of the nanostructures (Fig. 7) rather than at their tips, the VS mechanism should have dominated the growth process of the 1-D nanostructures.
Nevertheless, upon addition of > 5% Cr, lots of Si 3 N 4 particles rather than Si 3 N 4 nanowires/nanobelts were formed in the samples (Figs 2 and 3), which can be explained by the following two reasons: 1) the formation of lots of Cr x Si y alloy in the sample (Fig. 2) delayed the growth of 1-D α -Si 3 N 4 (Fig. 3d) as a result of the low activation energy of nitrogen diffusion through the alloy 31 , and 2) when the N 2 molecules diffused onto the Cr-Si interface (Fig. 9b-2), they would be activated by Cr forming active nitrogen species, thus accelerating the nitrdation rate of Si. With increasing the Cr addition, more activated N 2 molecules would be generated at the interface ( Fig. 9c-2). Consequently, the nitridation of Si would be promoted via a gas-solid reaction process (Reaction (3)), and more Si 3 N 4 particles generated via isotropic formation from Si surface to the center (Fig. 9d-2).
In summary, Cr exhibited a strong accelerating effect on the conversation of Si to Si 3 N 4 . At 1350 °C, the complete conversion of Si to Si 3 N 4 was achieved in the samples containing 1.25 wt% Cr. When the Cr addition was 5 wt%, the catalyst efficiently promoted the formation of 1-D α -Si 3 N 4 nanostructures. Si 3 N 4 nanowires about 50 μ m long and 50-200 nm in diameter, and nanobelts about 10 μ m long and 300-1000 nm in width, were simultaneously obtained. The 1-D α -Si 3 N 4 nanostructures grew from their Cr-containing roots via a VS mechanism. The accelerating effect of Cr on the nitridation of Si powder can be ascribed to the electron transfer from Cr to N, increasing the bond length and weakening bond strength in N 2 molecules, as predicted by the first-principle calculations.

Methods
Raw materials and sample preparation. Si (99.9% pure, ≤ 44 μ m, Naiou Nano Technology Co., Ltd., Shanghai, China) and Cr powders (≤ 6 μ m, 99.9% pure, Naiou Nano Technology Co., Ltd., Shanghai, China) were used as the main starting materials, and high purity N 2 (purity > 99.999%) was used as the nitrogen source. Si powders were mixed with various amounts of Cr (0-10 wt%) for 30 min in a ball mill at 300 rpm. The mixed batch was pressed under 30 MPa forming cylindrical samples with 20 mm in height and 20 mm in diameter. The samples were placed in an alumina-tube furnace and fired at 1200-1350 °C for 3 h in flowing N 2 . As the nitridation reaction of Si is strongly exothermic, to avoid overheating induced Si melting and its negative effect in nitridation, the furnace was initially heated to 1150 °C and then 1280 °C and held at each of these temperatures for 1 h, before being further heated to the final target nitridation temperature.
Sample characterization. Phase compositions in reacted samples were determined by X-ray diffraction (XRD, X'Pert Pro, Philips, Netherland). Spectra were recorded at 40 kV and 40 mA using Cu-Kα radiation, with a scanning rate of 2° (2θ)/min and a step size of 0.02° (2θ). ICDD cards used to identify Si, α -Si 3 N 4 and β -Si 3 N 4 are No. 01-089-5012, 01-073-1210 and 01-082-0697, respectively. The Rietveld refinement method was used to calculate crystalline phase contents in the fired samples. The overall conversion (OC) from Si to Si 3 N 4 was determined using the quantitative analysis values of the Si, α -and β -Si 3 N 4 , and α -phase contents in final products were calculated based on the quantitative analysis values of the α -and β -Si 3 N 4 .
Microstructures and morphologies of the final products were observed by using a field emission scanning electron microscope (FESEM, Nova 400 Nano FESEM, FEI Co., USA, 15 kV) and a high-resolution transmission electron microscope (HRTEM, 2000 F, Jeol Ltd., Japan, 200 kV). The samples for SEM were coated with gold, and those for TEM were prepared by ultra-sonic dispersion of the sample powders in EtOH, followed by dropping and drying the suspension onto a copper grid, respectively. Selected area electron diffraction (SAED), along with an energy dispersive X-ray spectroscopy (EDS, Noran 623 M-3SUT, Thermo Electron Corporation, Japan) was used for assisting phase identifications.
First-principle calculation. To assist understanding the catalytic nitridation mechanism in the case of using Cr as a catalyst, first-principle calculations based on the Cr (001) slab model were carried out using the CASTEP program based on the plane-wave pseudopotential (PW-PP) approach 32 A vacuum space of 10 Å was introduced to prevent interactions between slabs. The outmost six layers and absorbed N 2 molecules were fully relaxed.
The electron-ion interactions were represented by ultrasoft pseudopotentials (USP) 33 , and the electron-electron interactions were calculated by the generalized gradient approximation (GGA) with the Perdew-Burke-Ernzerhof (PBE) exchange correlation functional 34 . The equilibrium geometries were obtained by performing geometry optimization using the Broyden-Fletcher-Goldfarb-Shannon (BFGS) minimization method 35 , the energy cutoff for the plane wave basis set was set at 450 eV, and the Brillouin zone was sampled at 7 × 7 × 1 Monkhorst-Pack k-points. A Fermi smearing of 0.1 eV was utilized to speed up SCF convergence. The applied convergence criteria for geometry optimization were respectively 2.0× 10 −5 eV/atom, 0.05 eV/Å and 0.002 Å for energy, force and displacement.