Atomic-Scale Observation of Vapor–Solid Nanowire Growth via

Dec 9, 2015 - (34) Based on this method and by using WinXMorph crystal morphology builder,(35) the preferred shape of W18O49 is predicted below (left ...
0 downloads 8 Views 5MB Size
Atomic-Scale Observation of Vapor−Solid Nanowire Growth via Oscillatory Mass Transport Zhengfei Zhang,† Yong Wang,*,† Hengbo Li,† Wentao Yuan,† Xiaofeng Zhang,‡ Chenghua Sun,*,§ and Ze Zhang*,† †

Center of Electron Microscopy and State Key Laboratory of Silicon Materials, School of Materials Science and Engineering, Zhejiang University, Hangzhou 310027, China ‡ Hitachi High Technologies America, Pleasanton, California 94588, United States § ARC Centre for Electromaterials Science, School of Chemistry, Monash University, Clayton, Victoria 3800, Australia S Supporting Information *

ABSTRACT: In situ atomic-scale transmission electron microscopy (TEM) can provide critical information regarding growth dynamics and kinetics of nanowires. A catalyst-aided nanowire growth mechanism has been welldemonstrated by this method. By contrast, the growth mechanism of nanowires without catalyst remains elusive because of a lack of crucial information on related growth dynamics at the atomic level. Herein, we present a real-time atomic-scale observation of the growth of tungsten oxide nanowires through an environmental TEM. Our results unambiguously demonstrate that the vapor−solid mechanism dominates the nanowire growth, and the oscillatory mass transport on the nanowire tip maintains the noncatalytic growth. Autocorrelation analysis indicates that adjacent nucleation events in the nanowire growth are independent of each other. These findings may improve the understanding of the vapor−solid growth mechanism of nanowires. KEYWORDS: in situ TEM, metal oxide, nanowire growth, vapor−solid, oscillatory mass transport

N

catalyst. In two similar experiments of tungsten oxide nanowires grown on tungsten filaments via thermal oxidation, Chen et al.17 ascribed the growth to the VS mechanism, although the tip growth of the nanowire was not identified, and Tokunaga et al.,18 however, claimed a solid diffusion mechanism via a detailed in situ TEM investigation. Similar contradictory mechanisms were also reported in the catalystfree growth of many other metal oxide nanowires on corresponding metal substrates such as CuO7,16,19 and ZnO.20−22 The key issue is that the oxide nanowires grow directly on the metal substrates, which makes it challenging to decently distinguish the VS mechanism and solid diffusion. On the other hand, mass transport plays an important role in the nanowire growth, and oscillatory mass transport associated with 2D island nucleation at the trijunction of vapor−liquid−solid has been widely observed in VLS-dominated nanowire growth.14,23,24 However, the growth dynamics and kinetics of

anowires may rely on a liquid or solid catalyst particle for nucleation and growth via supersaturation− precipitation; this process is called the vapor− liquid−solid (VLS)1,2 or vapor−solid−solid (VSS) mechanism.3 In the absence of catalyst, nanowires nucleate and grow on the substrate through a vapor−solid (VS) mechanism4,5 or solid diffusion.6,7 The spontaneous catalyst-free nanowire growth can be driven by directed screw dislocation,8 planar defects (twin boundary and stacking fault),9,10 and anisotropic surface energy of various facets11 or assisted by oxides.12 Probing into the exact processes of nanowire formation, including mass transport and growth dynamics, is of great scientific significance. In situ transmission electron microscopy (TEM) has been widely employed to investigate nanowire growth dynamics and kinetics. Valuable information at the atomic scale has been achieved for the growth through the VLS mechanism,2,3,13−15 which leads to an unprecedented understanding of the catalyst-aided nanowire growth. By contrast, few in situ works on nanowire growth without a catalyst have been reported, and critical information regarding the growth dynamics at the atomic level is limited in this case,16 leading to controversial results regarding growth mechanism success, in particular, in the growth of metal oxide nanowires without a © 2015 American Chemical Society

Received: September 17, 2015 Accepted: December 8, 2015 Published: December 9, 2015 763

DOI: 10.1021/acsnano.5b05851 ACS Nano 2016, 10, 763−769

Article

www.acsnano.org

Article

ACS Nano

Figure 1. W18O49 nanowires grown via thermal oxidation in ETEM. (a) SEM and (b,c) TEM images of nanowires grown on the tungsten filament by 2 h thermal oxidation under 0.095 Pa O2 at ∼700 °C obtained through ETEM. (d) Top-view SEM image of the tungsten lamella affixed on the Mo grid after being thermally oxidized for 3 h at ∼650 °C and 0.095 Pa O2 under TEM. (e) Growth kinetics curve of a nanowire grown on the tungsten filament based on in situ TEM observation (consecutive electron beam irradiation). The diameter of the nanowire increases from ∼10.8 nm (t = 0 s) to ∼12 nm (t = 2600 s). Conditions include a growth temperature of ∼700 °C, oxygen pressure of ∼0.095 Pa, and electron beam density of ∼5 × 104 Am−2. The scale bar in the inset of (e) measures 100 nm.

the nanowire via VS remains elusive due to the lack of direct observation at the atomic scale. Herein, we report the in situ high-resolution environmental TEM (ETEM) observation of the monomolecular layer (ML) by ML growth of tungsten oxide nanowires by the thermal oxidation of the tungsten filament. Our results show that the catalyst-free growth of tungsten oxide nanowires is based on the VS mechanism. The oscillatory mass transport on the tip governed by surface free energy plays a vital role in maintaining the layer-by-layer 1D growth. Furthermore, the adjacent monomolecular layer growth events are found to be independent of each other according to autocorrelation analysis.

truncated, and almost all nanowires grown in our experiments share this common feature. No catalysts are found on the tip of the nanowires; hence, the VLS and the self-catalytic growth mechanisms are excluded. We then explored the possibilities of the VS and the solid diffusion mechanisms, which are previously proposed for catalyst-free nanowire growth through a thermal oxidation method. The kind of metallic source is the key differentiating factor between the VS mechanism and solid diffusion mechanism in the case of nanowire growth via thermal oxidation of a metal substrate. This metallic source is provided either by vapor deposition or by a solid metal substrate obtained through the surface diffusion or twin boundary diffusion of metal ions.7 Although tip growth can be confirmed (see Movie 1), tungsten oxide nanowires exactly grow on the tungsten filament. Hence, excluding the effect of the diffusion of the metal ions from the metal substrate to nanowire tip along the nanowire sidewall is difficult for determining the growth mechanism. Indeed, controversial growth mechanisms were reported according to similar in situ TEM growth of tungsten oxide nanowires.16,17 To clarify the growth mechanism of W18O49 nanowires, we set up a dedicated system by affixing a tungsten (W) lamella on a molybdenum (Mo) grid, which was then loaded into the ETEM chamber by a heating TEM holder for nanowire growth via thermal oxidation. The experimental conditions were similar to that in the tungsten filament case (Figure S1a). Notably, tungsten oxide nanowire growth occurs on both the tungsten lamella and the Mo grid (Figure 1d and Figure S1). The nanowires mostly grow on the top portion of the Mo grid (downstream of the oxygen flow). This behavior demonstrates that the growth of the tungsten oxide nanowire is similar to that of a chemical vapor deposition (CVD) process in

RESULTS AND DISCUSSION Figure 1a shows a typical scanning electron microscopy (SEM) image of nanowires grown on a tungsten filament after a 2 h thermal oxidation at 700 °C in TEM. Figure 1b shows a lowresolution TEM image of the grown nanowires. The nanowires exhibit diameters of 10−50 nm and lengths of 100−1000 nm. The high-resolution TEM image of a typical nanowire displayed in the inset of Figure 1b indicates that the nanowire is composed of W18O49 with a monoclinic crystal structure (JCPDS# 712450). The growth direction of the nanowires is noted to follow the [010] direction, which corresponds to the close-packed plane (010) of the monoclinic W18O49.25 Figure 1c shows a high-resolution TEM image captured from a typical nanowire grown on a tungsten filament during thermal oxidation; no nanoparticle (or catalyst) can be seen on the tip of the nanowire (also see Figure 1b and Movie 1 in the Supporting Information). Interestingly, the nanowire tip rim is 764

DOI: 10.1021/acsnano.5b05851 ACS Nano 2016, 10, 763−769

Article

ACS Nano

Figure 2. Oscillatory mass transport during nanowire tip growth. (a−e) Series of magnified TEM snapshots captured during the period of one ML growth on the nanowire tip. (f) Theoretical analysis of the favored morphology. Left: Equilibrium shape derived from BFDH construction. Right: Structure of (010) surface terminated with oxygen. (g,h) Plots of the truncated area marked by dashed triangles in (a−e) and added MLs as a function of the growth time, respectively.

which has a much larger area than the sidewall and tip of the nanowire. Hence, the first manner (1) may dominate the growth. Fast growth (0.2−2 nms−1) can be expected at the beginning, as shown in Figure 1e. At nanowire length (L) greater than the diffusion length (LW) of WO3 molecules on the nanowire sidewall, nanowire growth slows down when the exchange of substrate and nanowire molecules ceases to contribute to the tip growth.30 In this regard, the nanowire continues to grow at a very low but nearly constant rate (0.01− 0.1 nms−1) for a long time under stable oxygen pressure (i.e., 0.095 Pa). It should be pointed out that the increase in nanowire diameter caused by the flow of the sidewall steps from the bottom to the tip is small (lateral growth rate ∼4.6 × 10−4 nms−1) and will not alter the characteristics of the growth kinetics. The tip growth dynamics of the nanowires grown on the tungsten filament is revealed by the in situ high-resolution TEM images. An oscillatory tip growth, as a key feature, is identified. Figure 2a−e is a series of TEM snapshots obtained from a video (refer to Movie 2 in Supporting Information) that show

which chemical vapors from the tungsten lamella deposit on the Mo substrate following the oxygen flow. Therefore, the VS mechanism should be responsible for the W18O49 nanowire growth. The vapor source for W18O49 nanowire growth is presumed to be gaseous WO3 (mainly W3O9 and W4O12),26 which originates from the sublimation of WO3 caused by the reaction of tungsten and oxygen at 700 °C (Figure S3).The low-pressure condition in the TEM column may promote the sublimation of WO3, although WO3 begins to sublime at 750 °C.27 Then, WO 3 vapor condenses on substrates to nucleate WO2.72(W18O49) by a reduction reaction as follows: WO3(g) → WO2.72(s) + 0.14O2(g).28,29 Increasing the temperature or oxygen pressure leads to higher nanowire growth rates (Figures S4 and S5). This trend also consolidates the VS mechanism. In VS growth, the vapor source is supplied for nanowire tip growth via three ways: (1) diffusion of those absorbed on the substrate to the growth tip, (2) diffusion of those absorbed on the nanowire sidewall to the tip, and (3) direct deposition on the tip. Most gaseous WO3 will deposit on the substrate surface, 765

DOI: 10.1021/acsnano.5b05851 ACS Nano 2016, 10, 763−769

Article

ACS Nano

Figure 3. Three typical step-flows (a−c) on the tip terrace of the same nanowire during layer-by-layer growth. The nucleation positions of new MLs are indicated by red arrows, and the directions of step-flow are indicated by white arrows.

3c. All of these three step-flows share approximately the same flow rate (500 nm2 s−1 or 20 nm s−1) and finish in about 1.42 s (presumably due to the time required for adatoms to detach and move from the corners to the nuclei). Similar shape oscillation growth has been widely observed in VLS,14,23,24 in which the triple-line region at the catalyst− solid−gas interface plays a key role. Wen et al.23 clarified that the balance of facet energies and droplet surface tension is responsible for the trijunction morphology, whether truncated or not in VLS. Furthermore, the time-varying catalyst chemical potential results in the periodically changing morphology of the growth interface. In VLS, the droplet plays the role of the providing source. In the catalyst-free VS growth, the role of droplet is replaced by the vapor source with a steady pressure. In this case, the time-varying catalyst chemical potential is substituted by the oscillatory truncation magnitude of the tip rim corner, which reflects the chemical potential of the crystallite.32 As stated above, most nanowires in our experiments share the common feature of a truncated tip rim with a very slow growth rate (0.01−0.05 nm s−1; slow rate region in Figure 1e). This finding indicates that the nanowire is close to its equilibrium shape.32,33 The initial truncated tip rim consists of small facets (Figure 2a), which subsequently grows into a flat facet (Figure 2b). We confirmed the presence of a (012) facet at the tip rim corner during the growth of another nanowire (Figure S7). In Tersoff’s model,32 the corner of the Ag crystal with a cubic structure is presumably atomically rough, which is favorable for vapor deposition during growth. Even so, the situation in our case is consistent with that in Tersoff’s model. During crystal growth, gaseous molecules first fill the truncated corner, whether rough or faceted, as this occurrence is energetically more favorable than forming a 2D island on the main facets, such as on the tip (010) surface. The driving force for the growth of tip rim corner is the difference in chemical potential between gaseous WO3 and crystal W18O49. However, adding materials to the corner will increase the total surface free energy of the system (assuming that the temperature and pressure of the atmosphere are constant during the growth).32,33 Afterward, 2D island nucleation takes place on the main facets when the resulting free energy rise is less than that of the atoms adding to the tip rim. The 2D island nucleation may favor its occurrence on the (010) facet, which is the close-packed plane. Once the nucleus forms on the (010) tip terrace, mass transport from the corners to the tip terrace will be triggered for the requirement of decreasing surface free

the morphological changes of the right tip rim of a nanowire during one ML growth. Initially (Figure 2a), the nanowire tip rim, which is made up of small facets, is truncated. During the incubation time (Figure 2a−c), vapor molecules fill the truncated tip rim, which results in the tip rim growth to a flat facet (Figure 2b) and a nearly orthogonal corner (Figure 2c). No step is found on the tip terrace during this incubation time (Movies 2 and 3), which indicates that the mechanism driven by screw dislocation,8 as reported previously, does not apply in this case. The nucleation and mass transport begin on the tip terrace of the nanowire at 66.87 s (Figure 2d; more information can be found in Movie 2 and Figure 3). Lattice atoms evidently diffuse from the corner to the tip terrace, again leaving the corner truncated (Figure 2e). Meanwhile, a new (010) ML is formed. The thickness of the newly formed layer is ∼0.378 nm (marked in Figure 2e), which is exactly equal to the thickness of one ML along the [010] of W18O49. The mass transport is finished in less than 1.5 s (Figure 3), whereas the growth time of the truncated tip rim corner is more than 29 s. This difference indicates that the rate-limiting step in one period of ML growth is the growth of a truncated tip rim corner rather than the formation of a new ML on the tip terrace. The left tip rim shows a similar shape oscillation synchronized with the right side, but the oscillation magnitude may be roughly different (see Movie 2 in Supporting Information). With such repeating shape oscillation, the nanowire continues to grow layer-by-layer (Figure 2g,h). The rate-limiting stage during nanowire growth is the growth of the truncated rim corner, which facilitates the observation of the nucleation and step-flow on the tip terrace. Figure 3 shows three typical step-flows on the tip of the same nanowire, indicating the stochastic nucleation on the topmost terrace. In Figure 3a, a step nucleates near the left side of the tip rim and flows toward the right side. However, the step-flow in Figure 3b shows a contrasting case. Both step-flows finish with mass transport from the tip rim to the step edge. In Figure 3c, a nucleus (∼2 nm in size) forms in the middle of the nanowire tip terrace and then propagates across the surrounding terrace. The nanowire in a TEM image is a two-dimensional projection along an electron beam direction. Hence, identifying exactly whether step nucleation occurs on the tip rim or on the center of the tip terrace is challenging. Nevertheless, nucleation on the center of the topmost terrace is more difficult than on the tip rim of the nanowire because of the Berg effect.31 Therefore, we believe that the step also nucleates at the tip rim edge in Figure 766

DOI: 10.1021/acsnano.5b05851 ACS Nano 2016, 10, 763−769

Article

ACS Nano

Figure 4. Autocorrelation analysis of the adjacent ML nucleation events. (a) Plot of the accumulated number of MLs vs time and (b) plot of the growth time of each layer vs the accumulated number of (010) layers based on 75 ML growth at 1000 °C. No obvious diameter (35 nm) change of the nanowire can be seen during the growth. (c) Sample autocorrelation with 95% confidence intervals for the 75 successive ML layer growths. The red dotted lines indicate the confidence intervals computed from the large lag standard error at a significance level of 95%. CI represents the confidence interval.

direction. Overall, it is clear that the growth of W18O49 is anisotropic, as confirmed by both experimental observations and theoretical calculations, which is essentially determined by the difference of relative stabilities between low-indexed surfaces. We note that such oscillatory mass transport was also observed in the Si and Ge nanowires grown by the VLS method and in cubic Ag crystallite growth by the VS method. The oscillatory mass transport might also maintain the catalyst-free growth of the nanowires with high crystal symmetry, such as cubic,36 tetragonal,37 and hexagonal,38 once the nanowire growth is induced by the minimization of the surface energy or the effect of the substrate. We then explore the correlation between adjacent ML nucleation, which is an important kinetics characteristic for the direction of controlling nanowire growth in VLS.39 In the growth of group III−V nanowires through the VLS mechanism, the rapid depletion of group V atoms during each ML growth may result in an anticorrelated relationship between adjacent nucleation events according to a sub-Poisson distribution of the nucleation statistics.39,40 In the VS growth case, an approximate linear relationship exists between the accumulated MLs and the duration at different temperatures (Figure 4a and Figure S5), but the time intervals between the nucleation events of adjacent MLs are found to be vibrational (Figure 4b). The distribution of the time interval appears to slightly deviate from a Poisson distribution normalized to the same total number (75 layers) of ML growth (Figure S9), which seemingly implies a slightly correlated relationship between adjacent 2D nucleation

energy, resulting in the growth of a new (010) ML layer on the tip and again leaving rough facets for the rim. Accordingly, the nanowire grows along the [010] direction layer by layer. To understand why the W18O49 grows into a nanowire, theoretical analysis and calculation are introduced. First, morphological prediction is performed according to the Bravais−Friedel− Donnay−Harker (BFDH) principle, which is a simple method to simulate a crystal shape based on the crystal lattice geometry.34 Based on this method and by using WinXMorph crystal morphology builder,35 the preferred shape of W18O49 is predicted below (left in Figure 2f), which shows rod-like morphology with [010] as the dominated growth direction and three facets including (100), (001), and (10−1) as the sidewall. The predicted shape agrees with the observed W18O 49 nanowires (Figure 1), confirming that the growth is anisotropic with [010] as the favored growth direction. To achieve better understanding of the nanowire growth, it is informative to compare the relative stabilities of these low-indexed surfaces because high-energy facets diminish rapidly during crystal growth and thus dominate the growth direction. Therefore, theoretical calculations based on density functional theory (DFT) have been performed, using fully relaxed and optimized slab models for these surfaces (more details can be found in the Supporting Information). Using the bulk W18O49 as a reference, the energy per WO2.72 unit is obtained as −33.65, −33.44, −33.25, and −33.03 eV for (−101), (100), (001), and (010), respectively, according to which the (010) covered partially by oxygen is the minority surface, being in line with the prediction by BFDH and confirming that [010] is the favored growth 767

DOI: 10.1021/acsnano.5b05851 ACS Nano 2016, 10, 763−769

Article

ACS Nano events.39,40 However, considering the fact that the deviation from Poisson statistics is hard to quantify and measure in terms of its statistical significance, we then do autocorrelation analysis, which can give a precise value on the correlation and would also give a confidence interval, to assess whether the random ML nucleation events are correlated with each other. The autocorrelogram of the time intervals between adjacent ML nucleation events displayed in Figure 4c indicates that the ML growth sequence is independent. The autocorrelation analysis results show that the first-order autocorrelation coefficient (0.1685) is within the confidence interval (−0.2398 to 0.2128) for a significance level of 95%. These results indicate the independence of adjacent stochastic ML nucleation events. The other calculation details can be found in the Supporting Information. In VLS, due to the low solubility of group V atoms in the catalyst droplet, group V atoms are depleted rapidly during each nucleation. In VS growth, since no catalyst droplet is present, the growth is determined directly by the external source pressure, which is almost steady. Therefore, 2D island nucleation modulation by an external source during growth is weak, and truly stochastic ML nucleation is expected.

Detailed information about the nanowire growth on tungsten lamella, the effect of the temperature, oxygen pressure and electron beam in the experiments, HRTEM of faceted truncated tip rim corner, DFT calculation concerning the stability of different facets and the effect of oxygen termination on the stability of W18O49 (010) facet, autocorrelation analysis of the successive ML nucleation events (PDF) Movie 1 (AVI) Movie 2 (AVI) Movie 3 (AVI)

AUTHOR INFORMATION Corresponding Authors

*E-mail: [email protected]. *E-mail: [email protected]. *E-mail: [email protected]. Author Contributions

Y.W. conceived the research and designed the experiments. C.S. was responsible for the theoretical calculations. Z.F.Z conducted the in situ TEM experiments with assistance from W.Y., H.L., and X.Z. All the authors participated in the analysis and discussion. Y.W. and Z.Z. supervised the overall research.

CONCLUSIONS In summary, we demonstrated the VS growth of tungsten oxide nanowires through thermal oxidation under ETEM. The atomic-scale in situ ETEM observations reveal that the oscillatory mass transport in the nanowire tip, which is governed by the minimization of surface free energy, maintains the layer-by-layer growth. Furthermore, autocorrelation analysis of the layer-by-layer growth kinetics statistics shows that stochastic ML nucleation events are independent from each other. The dynamic information obtained from the in situ ETEM observations is expected to broaden our current knowledge regarding the catalyst-free growth mechanism of nanowires. The findings may also be generalized to explain other VS-grown nanowires via the CVD and physical vapor deposition methods.

Notes

The authors declare no competing financial interest.

ACKNOWLEDGMENTS We acknowledge the support of National Natural Science Foundation of China (51390474, 11234011, 11327901) and the Ministry of Education (IRT13037). The authors are greatly indebted to Frances M. Ross and Jerry Tersoff at IBM, and Shengbai Zhang at RPI for their intensive discussion and valuable suggestion. C.S. acknowledges the financial support from ARC Discover Project (DP130100268) and Future Fellowship (FT130100076). C.S. also appreciates the generous grants of CPU time from Australian National Computational Infrastructure.

EXPERIMENTAL SECTION

REFERENCES

We used ETEM (Hitachi H9500, Japan), which was operated at 300 kV. The microscope was equipped with a gas-heating holder and a differential pumping system, which allows the increase of the pressure of the sample chamber to ∼0.1 Pa, with a spatial resolution of 0.2 nm. Digital video sequences can be recorded within a 6 frame/s time resolution. The temperatures given refer to the values of electric current that flows through the objective table. The following two in situ experiments were conducted in the ETEM: (1) A pure tungsten filament in the holder was first heated to ∼700 °C within the TEM column with a base pressure of ∼0.0003 Pa, followed by introducing pure oxygen to the filament area through a flow meter (monitors the oxygen pressure) and a gas inlet to keep the pressure of the filament area at ∼0.095 Pa. (2) A tungsten lamella was extracted from a fresh tungsten filament and affixed on a molybdenum grid in advance by an FEI-focused ion beam. The Mo grid was mounted to a double-tilt heating TEM holder, which was then loaded into the ETEM chamber. Thermal oxidation of tungsten lamella was carried out by heating the Mo grid with tungsten lamella to ∼650 °C under TEM. Pure oxygen was flowed into the TEM column to maintain a pressure of ∼0.095 Pa in the sample area.

(1) Wagner, R. S.; Ellis, W. C. Vapor-Liquid-Solid Mechanism of Single Crystal Growth. Appl. Phys. Lett. 1964, 4, 89−90. (2) Wu, Y. Y.; Yang, P. D. Direct Observation of Vapor-Liquid-Solid Nanowire Growth. J. Am. Chem. Soc. 2001, 123, 3165−3166. (3) Persson, A. I.; Larsson, M. W.; Stenstrom, S.; Ohlsson, B. J.; Samuelson, L.; Wallenberg, L. R. Solid-Phase Diffusion Mechanism for GaAs Nanowire Growth. Nat. Mater. 2004, 3, 677−681. (4) Pan, Z. W.; Dai, Z. R.; Wang, Z. L. Nanobelts of Semiconducting Oxides. Science 2001, 291, 1947−1949. (5) Sears, G. W. A Growth Mechanism for Mercury Whiskers. Acta Metall. 1955, 3, 361−366. (6) Barsoum, M. W.; Hoffman, E. N.; Doherty, R. D.; Gupta, S.; Zavaliangos, A. Driving Force and Mechanism for Spontaneous Metal Whisker Formation. Phys. Rev. Lett. 2004, 93, 206104. (7) Yuan, L.; Wang, Y.; Mema, R.; Zhou, G. Driving Force and Growth Mechanism for Spontaneous Oxide Nanowire Formation during the Thermal Oxidation of Metals. Acta Mater. 2011, 59, 2491− 2500. (8) Bierman, M. J.; Lau, Y. K. A.; Kvit, A. V.; Schmitt, A. L.; Jin, S. Dislocation-Driven Nanowire Growth and Eshelby Twist. Science 2008, 320, 1060−1063. (9) Meng, F.; Estruga, M.; Forticaux, A.; Morin, S. A.; Wu, Q.; Hu, Z.; Jin, S. Formation of Stacking Faults and the Screw DislocationDriven Growth: A Case Study of Aluminum Nitride Nanowires. ACS Nano 2013, 7, 11369−11378.

ASSOCIATED CONTENT S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acsnano.5b05851. 768

DOI: 10.1021/acsnano.5b05851 ACS Nano 2016, 10, 763−769

Article

ACS Nano

(33) Meakin, P. The Growth of Rough Surfaces and Interfaces. Phys. Rep. 1993, 235, 189−289. (34) Docherty, R.; Clydesdale, G.; Roberts, K. J.; Bennema, P. Application of Bravais-Friedel-Donnay-Harker, Attachment Energy and Ising Models to Predicting and Understanding the Morphology of Molecular Crystals. J. Phys. D: Appl. Phys. 1991, 24, 89−99. (35) Kaminsky, W. From CIF to Virtual Morphology Using the WinXMorph Program. J. Appl. Crystallogr. 2007, 40, 382−385. (36) Hsu, Y. J.; Lu, S. Y. Vapor-Solid Growth of Sn Nanowires: Growth Mechanism and Superconductivity. J. Phys. Chem. B 2005, 109, 4398−4403. (37) Park, H. D.; Prokes, S. M. Study of the Initial Nucleation and Growth of Catalyst-Free InAs and Ge Nanowires. Appl. Phys. Lett. 2007, 90, 203104. (38) Ye, C. H.; Meng, G. W.; Wang, Y. H.; Jiang, Z.; Zhang, L. D. On the Growth of CdS Nanowires by the Evaporation of CdS Nanopowders. J. Phys. Chem. B 2002, 106, 10338−10341. (39) Glas, F.; Harmand, J.-C.; Patriarche, G. Nucleation Antibunching in Catalyst-Assisted Nanowire Growth. Phys. Rev. Lett. 2010, 104, 135501. (40) Glas, F. Statistics of Sub-Poissonian Nucleation in a Nanophase. Phys. Rev. B: Condens. Matter Mater. Phys. 2014, 90, 125406.

(10) Smith, A. M.; Kast, M. G.; Nail, B. A.; Aloni, S.; Boettcher, S. W. Planar-Defect-Driven Growth Mechanism of Oxygen Deficient Tungsten Oxide Nanowires. J. Mater. Chem. A 2014, 2, 6121−6129. (11) Law, M.; Goldberger, J.; Yang, P. D. Semiconductor Nanowires and Nanotubes. Annu. Rev. Mater. Res. 2004, 34, 83−122. (12) Zhang, R. Q.; Lifshitz, Y.; Lee, S. T. Oxide-Assisted Growth of Semiconducting Nanowires. Adv. Mater. 2003, 15, 635−640. (13) Hofmann, S.; Sharma, R.; Wirth, C. T.; Cervantes-Sodi, F.; Ducati, C.; Kasama, T.; Dunin-Borkowski, R. E.; Drucker, J.; Bennett, P.; Robertson, J. Ledge-Flow-Controlled Catalyst Interface Dynamics during Si Nanowire Growth. Nat. Mater. 2008, 7, 372−375. (14) Oh, S. H.; Chisholm, M. F.; Kauffmann, Y.; Kaplan, W. D.; Luo, W.; Ruehle, M.; Scheu, C. Oscillatory Mass Transport in VaporLiquid-Solid Growth of Sapphire Nanowires. Science 2010, 330, 489− 493. (15) Ross, F. M. Controlling Nanowire Structures through Real Time Growth Studies. Rep. Prog. Phys. 2010, 73, 114501. (16) Rackauskas, S.; Jiang, H.; Wagner, J. B.; Shandakov, S. D.; Hansen, T. W.; Kauppinen, E. I.; Nasibulin, A. G. In situ Study of Noncatalytic Metal Oxide Nanowire Growth. Nano Lett. 2014, 14, 5810−5813. (17) Chen, C. L.; Mori, H. In situ TEM Observation of the Growth and Decomposition of Monoclinic W18O49 Nanowires. Nanotechnology 2009, 20, 285604. (18) Tokunaga, T.; Kawamoto, T.; Tanaka, K.; Nakamura, N.; Hayashi, Y.; Sasaki, K.; Kuroda, K.; Yamamoto, T. Growth and Structure Analysis of Tungsten Oxide Nanorods Using Environmental TEM. Nanoscale Res. Lett. 2012, 7, 85−85. (19) Jiang, X. C.; Herricks, T.; Xia, Y. N. CuO Nanowires Can Be Synthesized by Heating Copper Substrates in Air. Nano Lett. 2002, 2, 1333−1338. (20) Dang, H. Y.; Wang, J.; Fan, S. S. The Synthesis of Metal Oxide Nanowires by Directly Heating Metal Samples in Appropriate Oxygen Atmospheres. Nanotechnology 2003, 14, 738−741. (21) Sun, Y.; et al. In situ Observation of ZnO Nanowire Growth on Zinc Film in Environmental Scanning Electron Microscope. J. Chem. Phys. 2010, 132, 124705. (22) Zhao, C. X.; Li, Y. F.; Zhou, J.; Li, L. Y.; Deng, S. Z.; Xu, N. S.; Chen, J. Large-Scale Synthesis of Bicrystalline ZnO Nanowire Arrays by Thermal Oxidation of Zinc Film: Growth Mechanism and HighPerformance Field Emission. Cryst. Growth Des. 2013, 13, 2897−2905. (23) Wen, C. Y.; Tersoff, J.; Hillerich, K.; Reuter, M. C.; Park, J. H.; Kodambaka, S.; Stach, E. A.; Ross, F. M. Periodically Changing Morphology of the Growth Interface in Si, Ge, and GaP Nanowires. Phys. Rev. Lett. 2011, 107, 025503. (24) Gamalski, A. D.; Ducati, C.; Hofmann, S. Cyclic Supersaturation and Triple Phase Boundary Dynamics in Germanium Nanowire Growth. J. Phys. Chem. C 2011, 115, 4413−4417. (25) Gu, G.; Zheng, B.; Han, W. Q.; Roth, S.; Liu, J. Tungsten Oxide Nanowires on Tungsten Substrates. Nano Lett. 2002, 2, 849−851. (26) Blackburn, P. E.; Hoch, M.; Johnston, H. L. The Vaporization of Molybdenum and Tungsten Oxides. J. Phys. Chem. 1958, 62, 769−773. (27) Lassner, E.; Schubert, W.-D. Tungsten; Plenum Publishers: New York, 1999; pp 85−86. (28) Hashimoto, H.; Tanaka, K.; Yoda, E. Growth and Evaporation of Tungsten Oxide Crystals. J. Phys. Soc. Jpn. 1960, 15, 1006−1014. (29) Jeon, S.; Yong, K. Synthesis and Characterization of Tungsten Oxide Nanorods from Chemical Vapor Deposition-Grown Tungsten Film by Low-Temperature Thermal Annealing. J. Mater. Res. 2008, 23, 1320−1326. (30) Dayeh, S. A.; Yu, E. T.; Wang, D. Surface Diffusion and Substrate-Nanowire Adatom Exchange in InAs Nanowire Growth. Nano Lett. 2009, 9, 1967−1972. (31) Berg, W. F. Crystal Growth from Solutions. Proc. R. Soc. London, Ser. A 1938, 164, 79−95. (32) Tersoff, J.; Denier van der Gon, A. W.; Tromp, R. M. Shape Oscillations in Growth of Small Crystals. Phys. Rev. Lett. 1993, 70, 1143−1146. 769

DOI: 10.1021/acsnano.5b05851 ACS Nano 2016, 10, 763−769