Abstract
Growth of metallic nanorods by physical vapor deposition is a common practice, and the origin of their dimensions is a characteristic length scale that depends on the threedimensional EhrlichSchwoebel (3D ES) barrier. For most metals, the 3D ES barrier is large so the characteristic length scale is on the order of 200 nm. Using density functional theorybased ab initio calculations, this paper reports that the 3D ES barrier of Al is small, making it infeasible to grow Al nanorods. By analyzing electron density distributions, this paper shows that the small barrier is the result of covalent bonding in Al. Beyond the infeasibility of growing Al nanorods by physical vapor deposition, the results of this paper suggest a new mechanism of controlling the 3D ES barrier and thereby nanorod growth. The modification of local degree of covalent bonding, for example, via the introduction of surfactants, can increase the 3D ES barrier and promote nanorod growth, or decrease the 3D ES barrier and promote thin film growth.
Keywords:
metal surface steps; adatom; diffusion; covalent bonding; simulationIntroduction
The growth of metallic nanorods by physical vapor deposition (PVD) is a common practice. Limited diffusion is a critical condition for nanorod growth [1], and geometrical shadowing in glancing angle deposition further promotes nanorod growth [2,3]. In addition to the diffusion of adatoms on flat surfaces, when diffusing over monolayer surface steps, adatoms experience a large energy barrierhe EhrlichSchwoebel (ES) barrierwhich affects the resulting surface [4,5]. Adatoms diffusing over multiplelayer steps experience even larger energy barriers [69]; this barrier is referred to as threedimensional (3D) ES barrier; for comparison, the conventional ES barrier is called the twodimensional (2D) ES barrier. Further, even small clusters experience 3D ES barriers [10], and variations of 3D ES barriers exist when steps intersect [11,12].
The 3D ES barriers for metals are much larger than their 2D counterparts, for example, 0.40 versus 0.16 eV for Cu [8]. These 3D ES barriers, when sufficiently large, stabilize multiplelayer surface steps, which in turn enable the effective operation of the 3D ES barriers [13]. The dynamic competition of multiplelayer and monolayer surface steps and consequently the competition of diffusion over 3D and 2D ES barriers give rise to the characteristic length scale of nanorod dimension [14]. When this length scale is too large, larger than 1 μm, the growth of nanorods becomes unfeasible.
It is worth noting that the characteristic length scale of nanorods becomes much larger than 1 μm, when the 3D ES barrier is reduced from 0.40 to 0.16 eV [14]. A smaller 3D ES barrier is possible when covalent bonding, even in metals such as Al [15], becomes important. For metallic surfaces, the large 3D ES barrier is the result of low coordination of a diffusing adatom at the saddle point. In comparison, covalent bonding can lower the energy state of atoms even with low coordination. As an example, a bulk silicon atom has only four nearest neighbors, in contrast to 12 nearest neighbors of facecentered cubic (FCC) metals. Even in metals, some degree of covalent bonding exists. This is the case for Al, which has s and p outer electrons. The covalent nature of Al bonding has been found to be responsible for its ultra large ideal shear strength [15]. Even though the shear modulus of Al is lower than that of Cu, the ideal shear strength of Al is higher than that of Cu. The covalent nature of Al bonding is also responsible for its anomalous outward relaxation of {111} and {100} surfaces [16]. In contrast, similar surfaces of most metals exhibit inward relaxation due to the missing coordination and electrons at surfaces [17,18].
The logic connection from covalent bonding to small 3D ES barrier and then to increased characteristic length scale of growing nanorods may apply to Al. In contrast to numerous reports of Cu nanorods using PVD [2,3,19], there is the lack of report on Al nanorods using the same method. In appearance, Cu and Al have comparable melting temperatures and sublimation energies. One might expect similar feasibility of Cu and Al nanorods using PVD.
Based on the established logic connection, we hypothesize that the PVD growth of Al nanorods is not feasible without substrate cooling due to the small 3D ES barrier of Al. To verify our hypothesis, we use ab initio calculations to show that the 3D ES barrier of Al is substantially smaller than the value of Cu and that the smaller barrier correlates with the covalent bonding of Al. Taking this result one step further, we suggest that one can perturb the local degree of covalent bonding by introducing an impurity atom. If such atom also tends to float on the surface, this use of impurity atoms represents a new mechanism of surfactant application, the new mechanism being the control of local covalent bonding.
Simulation method
We describe the research method of this studydensity functional theorybased ab initio calculations. Our calculations rely on the ABINIT package, which uses a planewave basis [20,21] within the local density approximation of the normconserving TroullierMartins pseudopotentials [22]. We choose the cutoff energy in planewave expansion as 300 eV, for convergence within 0.01 eV [23]. The valence electrons of Al in ABINIT are 3s^{2}3p^{1}. The kpoint meshing follows the MonkhorstPack scheme with 2 × 4 × 1 grids [24].
In setting up the supercell, we focus on the configuration of multiplelayer (fourlayer) steps, but we also examine configurations of steps with fewer layers in thickness. Figure 1a is the supercell for a Btype <110> step of four layers, and it contains 277 atoms. The adatom diffuses from the left {111} facet to the right {111} facet over a fourlayer step. One horizontal x direction is [100] and the other horizontal y direction along the step is [], making the vertical z direction to be []. There are four double layers along x (1.5932 nm), six double layers along y (1.6898 nm), and six double layers in the flat substrate along z (1.6898 nm); the vacuum region is equivalent to five double layers, above the adatom; it is 1.4082 nm. To mimic a bulk environment, atoms (blue in Figure 1) in the three single layers at the bottom are fixed to their bulk positions. When the vertical (z) dimension is increased by four single layers, the results converge within 0.01 eV; when the dimension of the x direction is increased by four single layers, the results converge within 0.02 eV; and when the y direction is changed by two single layers, the results do not change. To calculate the diffusion barriers, we use the nudged elastic band method with eight images as in our previous studies [8].
Figure 1. A simulation cell. It contains (a) a Btype step and (b) an Atype step, with vacuum region above the adatom.
Based on the convergence tests, we use comparable dimensions in all calculations of various steps. When the Btype step is three layers, the x, y, and z directions are [], [], and [], respectively, so periodic boundary conditions along the horizontal directions still apply, and the cell contains 277 atoms. When the Btype step is two layers, the x, y, and z directions are [], [], and [], respectively, and the cell contains 277 atoms. When the Btype step is a monolayer, the x, y, and z directions are [], [], and [], respectively, and the cell contains 277 atoms. Figure 1b is the supercell for an Atype <110> step of four layers with the x, y, and z directions being [332], [], and [], respectively, and it contains 253 atoms. When the Atype step is three layers, the x, y, and z directions are [111], [], and [], respectively, and the cell contains 289 atoms. When the Atype step is two layers, the x, y, and z directions are [556], [], and [], respectively, and the cell contains 253 atoms. When the Atype step is a monolayer, the x, y, and z directions are [558], [], and [], respectively, and the cell contains 289 atoms.
Results and discussions
In presenting the results, we first look at the overall energy barriers for the Al metal surface steps. Figure 2a shows the energy variation along the diffusion path, or diffusion coordinate, for Btype <110> steps. As discussed before, the exchange mechanism leads to a smaller diffusion barrier [25] and is the only diffusion mechanism investigated here. When the number of steps becomes two or three, the diffusion barrier converges to the 3D ES barrier. For Btype steps, the 3D ES barrier is 0.13 eV, and the 2D ES barrier (corresponding to a monolayer step) is 0.07 eV; for comparison, the diffusion barrier of an adatom on flat Al{111} is 0.05 eV, consistent with earlier reports [26]. It is interesting to note that this 3D ES barrier of 0.13 eV for Al is in sharp contrast to the 0.40 eV for Cu [8]; we have repeated the Cu calculations using ABINIT for consistency and obtained 0.41 eV, which is within the uncertainty of 0.01 eV. This result confirms our hypothesis that the 3D ES barrier of Al is small. Consequently, multiplelayer surface steps are unstable [13], and the corresponding characteristic length scale is much larger than 200 nm for the growth of Al nanorods [14]. For Atype <110> steps, Figure 2b shows that the 3D ES barrier is 0.18 eV and the 2D ES barrier is 0.14 eV. While the 3D ES barrier of Atype steps is larger than that of Btype steps, both are substantially smaller than the 0.40 eV of Cu Btype steps.
Figure 2. Energy as a function of diffusion coordinate. For (a) Btype steps and (b) Atype steps; the energy of the initial configuration of diffusion is taken to be zero.
The results in Figure 2 are for diffusion from FCC to FCC sites. For Btype steps, when the diffusing adatom starts from a hexagonalclosepacked (HCP) site and ends at an FCC site, the 3D ES barrier is 0.14 eV. When the diffusing adatom starts from an FCC site and ends at an HCP site, the 3D ES barrier is 0.15 eV. When the diffusing adatom starts from an HCP site and ends at an HCP site, the 3D ES barrier is 0.15 eV. For an Atype step, when the diffusing adatom starts from an HCP site, the 3D ES barrier is reduced to 0.16 eV.
To gain insight of why the 3D ES barrier of Al is so much smaller than that of Cu, we compare the density of electrons in real space for the Btype step of four layers. As Figure 3a shows, there is substantial overlap of electrons between neighboring Al atoms and between Al adatom and its neighbors. This is also true when the Al adatom is at the saddle point (Figure 3b). In contrast, the overlap of electrons is substantially smaller in Cu (Figure 3c, d).
Figure 3. Electron density distributions. At the initial stage (a) and at the saddle point (b) of diffusion for Al and at the initial stage (c) and at the saddle point (d) of diffusion for Cu. The cross sections show the directional distribution with the covalent bonding between atoms in Al and the spherical charge distribution about a centered ion in Cu.
Accompanying the electron density distribution in real space is its distribution in energy space. Figure 4 shows the change of total density of state (DOS) from the starting configuration to the saddlepoint configuration of the diffusion process over the same fourlayer Btype step. In comparison, fewer electrons move into higher energy states for Al than in Cu. Even when the DOS of Al is scaled by 11/3, this comparison is still valid as shown by the inset of Figure 4; Al has three valence electrons and Cu has 11. In covalent bonding, electrons move with bonds during diffusion and thereby can stay in low energy states. The covalent bonding of Al correlates with the smaller change of DOS.
Figure 4. Change of total DOS from initial configuration to saddlepoint configuration. It corresponds to steps in Figure 3; with Fermi level indicated by a dotted line.
Having linked a smaller 3D ES barrier of Al with intrinsic covalent bonding, we now discuss the impacts of this link. It is possible to extrinsically modify local degree of covalent bonding. Surfactants have been used to modify diffusion, as well as nucleation, on surfaces [27,28]. If a surfactant atom can modify the degree of local covalent bonding, it will change the 3D ES barrier. For example, the introduction of Si atoms to Cu steps may promote the local sp electronic bonding and thereby the local covalent bonding, but to qualify as a surfactant, the atom (such as Si) also has to float on surfaces. That is, in addition to effects of atomic size and bonding strength, the control of degree of covalent bonding derived from surfactants is a new effect. This new effect may lead to another mechanism of controlling surface morphologies during growth, say by physical vapor deposition. There is a large body of literature reporting Cu nanorods by PVD [2,3,19], but none on Al nanorods by PVD without substrate cooling (at least none that we are aware of), consistent with the difference of intrinsic 3D ES barriers of the two metals. If a surfactant extrinsically promotes local covalent bonding, growing Cu nanorods may turn into thin films. Similarly, if a surfactant extrinsically reduces the degree of local covalent bonding, growing Al thin films may turn into nanorods.
Conclusion
In conclusion, we hypothesize that the lack of Al nanorods by PVD is due to the small 3D ES barrier of Al. Using ab initio calculations, we have shown that the 3D ES barrier of Al is indeed much smaller than that of Cu (0.13 versus 0.40 eV for the smaller barrier in each metal), confirming the hypothesis. Further, we have shown that the smaller 3D ES barrier of Al is the result of its covalent bonding. Finally, based on these two results, we suggest a new mechanism of surfactant applicationthat is, the change of degree of local covalent bonding through the introduction of surfactant atoms.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
SL carried out the calculations and analyses, and HH designed the project and participated in the analyses. All authors read and approved the final manuscript.
Acknowledgements
The authors acknowledge financial support from the Department of Energy Office of Basic Energy Sciences (DEFG0209ER46562). HH also acknowledges financial support of the National Science Foundation (CMMI0856426) in bridging the growth of thin films with that of nanorods.
References

Mullins WW: Theory of thermal grooving.
J Appl Phys 1957, 28:333. Publisher Full Text

Wang J, Huang HC, Kesapragada SV, Gall D: Growth of Yshaped nanorods through physical vapor deposition.
Nano Lett 2005, 5:2505. PubMed Abstract  Publisher Full Text

Kesapragada SV, Gall D: Anisotropic broadening of Cu nanorods during glancing angle deposition.
Appl Phys Lett 2006, 89:203121. Publisher Full Text

Ehrlich G, Hudda FG: Atomic view of surface selfdiffusion: tungsten on tungsten.
J Chem Phys 1966, 44:1039. Publisher Full Text

Schwoebel RL, Shipsey EJ: Step motion on crystal surfaces.
J Appl Phys 1966, 37:3682. Publisher Full Text

Liu SJ, Huang HC, Woo CH: SchwoebelEhrlich barrier: from two to three dimensions.
Appl Phys Lett 2002, 80:3295. Publisher Full Text

Lagally MG, Zhang ZY: Thinfilm cliffhanger.
Nature 2002, 417:907. PubMed Abstract  Publisher Full Text

Xiang SK, Huang HC: Ab initio determination of EhrlichSchwoebel barriers on Cu{111}.
Appl Phys Lett 2008, 92:101923. Publisher Full Text

Zhu WG, Mongeot FB, Valbusa U, Wang EG, Zhang Z: Adatom ascending at step edges and faceting on fcc metal (110) surfaces.
Phys Rev Lett 2004, 92:106102. PubMed Abstract  Publisher Full Text

Anguilar BH, Flores JC, Coronado AM, Huang HC: Atom diffusion of small Cu clusters across facetfacet barriers over Cu{111} surfaces.
Modell Simul Mater Sci Eng 2007, 15:419. Publisher Full Text

Han Y, Lu GH, Lee BJ, Liu F: Flatsurface, stepedge, facetfacet, and facetstep diffusion barriers in growth of a Pb mesa.
Surf Sci 2008, 602:2284. Publisher Full Text

Huang HC, Wang J: Surface kinetics: stepfacet barriers.
Appl Phys Lett 2003, 83:4752. Publisher Full Text

Zhang RX, Huang HC: Another kinetic mechanism of stabilizing multiplelayer surface steps.
Appl Phys Lett 2011, 98:221903. Publisher Full Text

Zhou LG, Huang HC: Characteristic length scale of nanorod diameter during growth.
Phys Rev Lett 2008, 101:266102. PubMed Abstract  Publisher Full Text

Ogata S, Li J, Yip S: Ideal pure shear strength of aluminum and copper.
Science 2002, 298:807. PubMed Abstract  Publisher Full Text

Finnis MW, Heine V: Theory of lattice contraction at aluminum surfaces.
J Phys F Met Phys 1974, 4:L37. Publisher Full Text

Noonan JR, Davis HL: Confirmation of an exception to the "general rule" of surface relaxations.
J Vac Sci Technol A 1990, 8:2671. Publisher Full Text

Davis HL, Hannon JB, Ray KB, Plummer EW: Anomalous interplanar expansion at the (0001) surface of Re.
Phys Rev Lett 1992, 68:2632. PubMed Abstract  Publisher Full Text

Wang PI, Parker TC, Karabacak T, Wang GC, Lu TM: Size control of Cu nanorods through oxygenmediated growth and low temperature sintering.
Nanotechnology 2009, 20:085605. PubMed Abstract  Publisher Full Text

Gonze X, Beuken JM, Caracas R, Detraux F, Fuchs M, Rignaness GM, Sindic L, Verstraete M, Zerah G, Jollet F, Torrent M, Mikami M, Ghosez Ph, Raty JY, Allen DC: Firstprinciples computation of material properties: the ABINIT software project.
Comput Mater Sci 2002, 25:478. Publisher Full Text

Gonze X, Beuken JM, Caracas R, Detraux F, Fuchs M, Rignaness GM, Sindic L, Verstraete M, Zerah G, Jollet F, Torrent M, Mikami M, Ghosez Ph, Raty JY, Allen DC: A brief introduction to the ABINIT software package.
Z. Kristallogr 2005, 220:558. Publisher Full Text

Troullier N, Martins JL: Efficient pseudopotentials for planewave calculations.
Phys Rev B 1991, 43:1993. Publisher Full Text

Mantina M, Wang Y, Arroyave R, Chen LQ, Liu ZK: Firstprinciples calculation of selfdiffusion coefficients.
Phys Rev Lett 2008, 100:215901. PubMed Abstract  Publisher Full Text

Monkhorst HJ, Pack JD: Special points for Brillouinzone integrations.
Phys Rev B 1976, 13:5188. Publisher Full Text

Wang J, Huang HC, Cale TS: Diffusion barriers on Cu surfaces and near steps.
Modell Simul Mater Sci Eng 2004, 12:1209. Publisher Full Text

Tiwary Y, Fichthorn KA: Mechanisms of atomic diffusion on the flat, stepped, and faceted surfaces on Al(110).

Xiang SK, Huang HC: Binding of In and Pb surfactants on Cu{111} surfaces.
Surf Sci 2010, 604:868. Publisher Full Text

Huang HC, Wei HL, Woo CH, Zhang XX: Engineering kinetic barriers in copper metallization.
Appl Phys Lett 2002, 81:4359. Publisher Full Text