Modeling Facet Nucleation and Growth of hut clusters on Ge/Si(001) John Venables1,3 Mike McKay2,4 and Jeff Drucker1,2 1) Physics, 2) Materials, Arizona State University, Tempe 3) LCN-UCL, London, 4) Lawrence Semiconductor, Tempe • • • • Ge/Si(001) hut clusters: Annealing in STM 2D Modeling of facet nucleation & growth Model details: Dm, i, DG(i), reconstruction Conclusion and References Explanation for NAN 546: April 09 This talk was given at two conferences in 2008-09: first at MRS Boston, 12/2/08 as a contributed 15 minute talk (13 slides), and then at the Surface Kinetics International (SKI) meeting at University of Utah, 3/20/09, as an invited 25 min talk (15 slides #1, 3-16 here). Slide #1 is an Agenda slide to guide one through the talk. The hyperlinks on slide #1 connect to Custom Shows, available under the Slide Show dropdown menu. The remaining slides #17-30 are in reserve for questions, and in practice were not used at the time. But they can be useful at a conference if one chooses to get into more detail. This material is copyright of the Authors. A summary paper is published in PRL 101 (2008) 216104 (M.R. McKay, J.A. Venables and J. Drucker) Reprints are available on request Ge/Si(001) STM Movies: watching paint dry at 450 OC gas-source MBE from Ge2H6 Ge = 5.0ML, 0.1 ML / min T = 450 °C, 26 min/frame 62 hrs total elapsed time first frame after 33min anneal Field of view 600nm x 600 nm Ge = 5.6ML, 0.2 ML / min T = 500 °C, 7 min/frame 14 hrs total elapsed time first frame after 160min anneal Field of view 400nm x 400 nm Mike McKay, John Venables and Jeff Drucker, 2007-08 7 4 30 nm 6 8 5 2 9 3 1 10 33 7 4 6 5 2 8 9 3 10 1,255 1 Ge/Si(001) hut clusters: Annealing at T = 450 oC 9 500 450 400 350 300 250 200 150 100 30 Volume 25 Length 20 15 Width 0 1000 2000 3000 10 4000 Anneal Time (min) 4 6 5 2 7 8 9 3 1 10 2,503 7 4 6 5 8 9 2 3 1 3,751 10 8 500 450 400 350 300 250 200 150 100 30 Volume 25 Length 20 15 Width 0 1000 2000 3000 10 4000 Anneal Time (min) Most islands static, smallest island grows (8). Volume, Length data and model result Volume, Density data for all huts in video field of view at selected times Density is constant Model Length increase DL(t), Average for 34 huts from STM video Background for real Ge/Si(001) • Wetting layer ~ 3+ ML; supersaturation on and in the WL, source of very mobile ad-dimers (Ed2 ~ 1 eV) for rapid growth eventually of dislocated islands • Low dimer formation energy (Ef2 ~ 0.3 eV) gives large i, even though condensation is complete • Stress grows with island size, sx decreases • Interdiffusion, and diffusion away from high stress regions around islands, reduces stress at higher T and lower F (e.g. at 600, not 450 oC for F ~1-3 ML/min.) • Specifics of {105} hut clusters, reconstruction, etc Chaparro et al. JAP 2000, McKay, Shumway & Drucker, JAP 2006. Outline of Si/Ge(001) hut growth model (105) facet nucleation or dissolution, nucleation at the apex, surface vacancy nucleation at the base Finite but low edge (or ridge) energy on the facet favors 2D facet nucleation case 1); Ad-dimers move everywhere, but perimeter barrier impedes replenishment on the hut Barrier height increases with hut size but almost saturates Mike McKay, John Venables and Jeff Drucker, 2008 2D Facet nucleation with perimeter barrier Elastic energy of Ge adatom on huts on s. c. model Si(001) substrate width 2r, length 2(s+r), 0 < s < (40-r) . Relative occupation of boundary sites at T= 450oC (723 K) J. Drucker and J. Shumway 2D Critical nuclei: i and DG(i) i = (X/(2Dm')2 and DG(i) = X2/(4Dm') How long does it take to nucleate a facet? Variables: Eb on the facet; Internal bias Vi; Barrier height Ep Values: 0.2-0.3 eV zero to -0.1 eV 0.3 - 0.7? eV very sensitive to supersaturation Dm/kT (10-30 meV at T = 673K) Additional elements in the model • Strain dependent adsorption energy at facet apex E = ahr2/2 + bhr3cos(g)/3s, g = 11.3o with a = 0.7 eV/nm3 and b = 0.81 eV/nm3 from Finite element elastic calculations • Extra "un-reconstruction" energy of {105} faces may increase DG(i): DFT calculations, ~ 0.5 eV for i > 3.5 dimers Cereda & Montalenti (2007) • Couple growth of facets to reduced dimer density n1 via Dm = kTln(n1/n1e) and nucleation rate Un to get dn1/dt = C.Un with known material constant C. McKay, Venables & Drucker, PRL 101, 216104 (2008). Effects of {105} reconstruction MD simulations with Tersoff potential + DFT (VASP): 3 Dimers + 2 Vacancies/2*2.5 unit cell; 1.23 dimers/nm2 Reconstruction P. Ratieri et al. PRL 88 (2002) 256103 Un-reconstruct S. Cereda & F Montalenti PRB 75 (2007) 195321 Coupling of 2D nucleation and annealing Supersaturation S = (n1/n1e); Dm = kT log (S) Nucleation rate/facet Un = AfZsiDn1ni, (1) with ni = n1exp(-DG(i)/kT), Af = area/facet Annealing reduces n1 and hence S, Dm via dn1/dt = -2NAfsdUn (2) with N huts/unit area, sd = facet dimers/area. Finding the critical nucleus size i and DG(i), the energy associated with Un is En = 2(Dm - L2) - [Ed + DG(i)] Length increase (t) data and model result Evolution of Dm, i and DG(i) with annealing time, t Model Length increase DL(t), Average for 34 huts from STM video Conclusions: Facet nucleation & hut growth • Growth rate limited by facet nucleation at hut apex • Energy gradient on the facet modifies 2D nucleation formulae, biases towards larger critical nucleus size. • Consumption of the adsorbed layer reduces the supersaturation Dm(t) and slows the growth rate: • Quantitative agreement with hut length L(t) annealing data for reasonable Dm and parameter values • Undoing the {105} reconstruction may account for a substantial part (~ 0.5 eV) of the critical nucleus energy • Ostwald ripening is suppressed (at 450 oC) because dimer supersaturation stays positive during annealing References Ge/Si(001) hut growth AFM and STM experiments S.A. Chaparro et al: JAP 87, 2245-2254 (2000) M.R. McKay, J. Shumway & J. Drucker: JAP 99, 094305 (2006) F. Montalenti et al: PRL 93, 216102 (2004) STM movies online at http://physics.asu.edu/jsdruck/stmanneal.htm Previous Ge/Si(001) {105} hut growth and step energy models D.E. Jesson et al.: PRL 80, 5156-5169 (1998) M. Kästner & B. Voightländer: PRL 82, 2745-2748 (1999) S. Cereda, F Montalenti & L. Miglio Surf. Sci. 591, 23-31 (2005) Details of {105} reconstruction and un-reconstructions Reconstruction P. Ratieri et al. PRL 88, 256103 (2002) Un-reconstruct S. Cereda & F. Montalenti PRB 75, 195321 (2007) Anisotropic elastic energy calculations fig 5a Approximately linear energy density on facet fig 5b Alternative approaches to modeling 1) Rate and rate-diffusion equations 2) Kinetic Monte Carlo simulations 3) Level-set and related methods plus 4) Correlation with ab-initio calculations Issues: Length and time scales, multi-scale; Parameter sets, lumped parameters; Ratsch and Venables, JVST A S96-109 (2003) Potentials due to strain, e Ovesson, D constant In general, D not constant, a2 depends on direction Demonstrate with 1D model & Lennard-Jones potential DFT calculations for Si, Ge/Si(001), and Si/Ge(001) D.J. Shu, X.G. Gong, L. Huang, F. Liu (2001) JCP 114, 10922; PRB 64, 245410; (2004) PRB 70, 155320 Transition rates in a 2D potential field - b Ei j 0 i j Wi j W e i, j on lattice s saddle point if if then but unfortunately this isn't true in general.... S. Ovesson PRL 88, 116102 (2002) Mean-field equations from microscopic dynamics Strain dependent Diffusion D and Drift velocity V as deduced by Grima, DeGraffenreid, Venables 2007 From Shu, Liu, Gong et al: For Ge/Si(001): a1 = -1.75 eV; at lattice sites a2 -a1 = 0.75 eV fast diffusion direction Ge/Si(001) concentration profiles a2= a1= -1.75 eV a2- a1= 0.75 eV a2- a1= 1.50 eV R. Grima, J. DeGraffenreid and J.A. Venables, 2007, PRB Visualization: Discussion points • 1D & 2D Graphics and Movies are excellent complement to RateDiffusion Equations; ideal for projects/talks, not so easy for papers Annealing, deposition, direct impingement, individual surface/edge processes. Nanowire systems using Ge/Si(001) model parameters. • Approximate solutions that concentrate on "Events" are great for understanding, and answer questions like: "What happens when and where?" So, how far do we want to go in the "realism" direction? • Hybrid FFT uses constant D in k-space + difference terms in real space and is more stable, but perhaps less accurate. Comparison of MED, hybrid FFT and multigrid methods for speed/accuracy done, But what are the general computational lessons to be drawn? • Tests on strongly non-linear problems (e.g. high-i* nucleation + growth) and "real systems", e.g. Ge/Si(001), work in progress. Need to include reconstructions, fluctuations, local environment, long timescales, etc: very complicated! But should we expect otherwise? Nanotechnology, modeling & education Interest in crystal growth, atomistic models and experiments in collaboration Interest in graduate education: web-based, web-enhanced courses, book See http://venables.asu.edu/ for details New Professional Science Masters (PSM) in Nanoscience degree program at ASU at http://physics.asu.edu/graduate/psm/overview • • • • height = 5 time = 90 Dt = 0.1 64*64 grid • (5*11) island • grows to • (19*33) • Dx = 5 • Dy = 10 MatLab Movie as *.avi (Quick time) Sizes and shapes in Ge/Si(001) TEM, AFM: Chaparro, Zhang, Drucker, Smith J. Appl. Phys (2000) Size distributions and alloying T = 600 °C Number of islands / cm2 / 2.5 nm bin 1.5 x109 Strain relief via 1) interdiffusion 2) change of shape (b ) X2 1 x109 5 x108 Hut-dome transitions reversible via alloying at high T > 500 oC 0 T = 450 °C (d) 4.8 x109 5 ML 6.5 ML 8.0 ML 9.5 ML 11.0 ML 12.5 ML 3.2 x109 1.6 x109 S. Chaparro, Jeff Drucker et al. PRL 1999, JAP 2000 0 0 40 80 120 Diameter (nm) 16 0 Nucleation of new facets - hut growth controlled by nucleation of new {105} planes on small facets - nucleation rate is how fast critical nuclei become supercritical nj = number density of nascent facets comprised of j dimers n i n1e-DG(i) kT number density of critical nuclei nucleation rate (number of new stable clusters per small {105} facet per second) facet area U n AZs i Dn1n i AZs i Dn12e-DG(i) kT Zeldovich factor (typically 0.1-0.5) dimer diffusion coeff. on {105} capture number (~# perimeter sites) D 4N o e Ed kT dimer sublimation energy from step edges (~0.3eV) Dm kTln n1 n1e & n1e Noe-L2 Un 1 [ AZ s N e 4 i o 2Dm-L 2 -E d DG(i) kT kT Why do smaller islands grow? • facet nucleation and growth depletes ad-dimer concentration on island • ‘refilling’ rate controlled elastic potential barrier at hut perimeter cf = dimer concentration on hut after facet nucleation event cn = dimer concentration required to nucleate stable facet co = dimer concentration outside of island V (r ) elastic potential energy at position r Vp = potential at hut perimeter. Vi = Vo. A new facet forms at t=0, depleting the dimer concentration on the hut surface to cf. How long is required for the island to refill to cn so that another stable facet can form? dc Island concentration, c, obeys A dt c o - c- c o - c . Solution for c is c co c f - co e- A t Use barrier form for boundary capture number, sB: time for hut to ‘refill’ to cn is tr p -V p -Vo kT a 2 -E d s B D e e a 4 4 A E d V p -Vo kT c f - c o e ln pa c n - c o so, large huts grow ~20 times slower than small huts kT