1: \documentclass[twocolumn,prl,aps,showpacs,superscriptaddress]{revtex4}
2: \usepackage{epsfig}
3: \usepackage{fancyhdr}
4: \pagestyle{fancy}
5: \fancyhf{}
6: \lhead{Gai {\it et al.}, Physical Review Letters}
7: \rhead{\thepage}
8: %
9: \def\ang{\stackrel{\circ}{\rm A}}
10: %
11: \begin{document}
12: \title{Self-assembly of Nanometer-scale Magnetic Dots with Narrow Size
13: Distributions on an Insulating Substrate}
14: \author{Zheng Gai}
15: \affiliation{Solid State Division, Oak Ridge National Laboratory,
16: Oak Ridge, TN 37831}
17: \affiliation{Department of
18: Physics, Peking University, Beijing, 100871, P.R. China}
19: %
20: \author{Biao Wu}
21: \affiliation{Department of Physics, The University of Texas, Austin,
22: Texas 78712}
23: \affiliation{Solid State Division, Oak Ridge National Laboratory,
24: Oak Ridge, TN 37831}
25: %
26: \author{J.P. Pierce}
27: \affiliation{Department of Physics and Astronomy, University of Tennessee,
28: Knoxville,Tennessee 37996}
29: \affiliation{Solid State Division, Oak Ridge National Laboratory,
30: Oak Ridge, TN 37831}
31: %
32: \author{G.A. Farnan}
33: \affiliation{Solid State Division, Oak Ridge National Laboratory,
34: Oak Ridge, TN 37831}
35: %
36: \author{Dajun Shu}
37: \affiliation{International Center for Quantum Structures, Chinese Academy
38: of Sciences, Beijing 100080, P.R. China}
39: \affiliation{Department of Physics, Nanjing University, Nanjing, 210093, P.R.
40: China}
41: %
42: \author{Mu Wang}
43: \affiliation{Department of Physics, Nanjing University, Nanjing, 210093,
44: P.R. China}
45: \affiliation{International Center for Quantum Structures, Chinese Academy
46: of Sciences, Beijing 100080, P.R. China}
47: %
48: \author{Zhenyu Zhang}
49: \affiliation{Solid State Division, Oak Ridge National Laboratory,
50: Oak Ridge, TN 37831}
51: \affiliation{International Center for Quantum Structures, Chinese Academy
52: of Sciences, Beijing 100080, P.R. China}
53: %
54: \author{Jian Shen}
55: \affiliation{Solid State Division, Oak Ridge National Laboratory,
56: Oak Ridge, TN 37831}
57:
58: \date{\today}
59: \begin{abstract}
60: The self-assembly of iron dots on the insulating surface of NaCl(001) is investigated
61: experimentally and theoretically. Under proper growth conditions,
62: nanometer-scale magnetic iron dots with remarkably narrow size distributions can
63: be achieved in the absence of a wetting layer. Furthermore, both the vertical and
64: lateral sizes of the dots can be tuned with the iron dosage without introducing
65: apparent size broadening, even though the clustering is clearly in the strong
66: coarsening regime. These observations are interpreted using a phenomenological
67: mean-field theory, in which a coverage-dependent optimal dot size is selected by
68: strain-mediated dot-dot interactions.
69: \end{abstract}
70:
71: \pacs{81.07.Ta,61.46.+w,68.37.Ps,68.35.Md}
72: %81.15.Hi Molecular, atomic, ion, and chemical beam epitaxy
73: %81.15.Kk Vapor phase epitaxy; growth from vapor phase
74: %81.07.Ta Quantum dots
75: %68.55.Jk Structure and morphology; thickness; crystalline orientation and
76: %texture
77: %68.35.Md Surface thermodynamics, surface energies
78: \maketitle
79: Clustering on surfaces by nucleation and growth during atom deposition has been
80: an important subject in basic and applied science for decades
81: \cite{Venables,Zinke}. Recent efforts
82: have been focused on searching for methods to obtain nanometer-scale clusters with
83: narrow size distributions. Such clusters or quantum dots are potentially valuable
84: for optical, electronic, and magnetic device applications, but mass production of
85: such structures by lithography or etching-based fabrication has proved to be
86: exceptionally challenging \cite{Wolf,review}.
87: Alternatively, it has been realized that the strain
88: energy associated with the lattice mismatch between the dot and the substrate materials
89: can be exploited to induce self-assembled formation of quantum dots with narrow size
90: distributions. This has generated much excitement, particularly in the area of
91: semiconductor quantum dots \cite{Wolf,review,Mo,Jesson,Xie,Xin}.
92: In such cases, the growth of the dots often proceeds
93: in the Stranski-Krastanow (SK) mode, which is characterized by the presence of a wetting
94: layer prior to three-dimensional (3D) clustering. To date, the precise mechanism for size
95: selection in semiconductor quantum dot systems remains a subject of active debate
96: \cite{review,Mo,Jesson,Daruka}:
97: some attribute them to strain-induced thermodynamic equilibrium states, while others
98: associate them with metastable configurations due to kinetic limitations.
99:
100: Although improved size uniformity can be achieved in quantum dot formation via
101: the SK growth mode, the presence of a wetting layer is often undesirable, particularly
102: for electronic and magnetic device applications of metallic/magnetic quantum dots. For
103: this reason, it is preferred to fabricate quantum dots in the Volmer-Weber (VW) growth
104: mode, which is characterized by immediate 3D clustering on the substrate surface. Indeed,
105: considerable recent efforts have been devoted to metallic/magnetic quantum dot formation
106: on various substrates \cite{Fruchart,huang,Baumer}, but no significant size uniformity
107: has been achieved in such studies.
108:
109: In this Letter we investigate the self-assembly of iron dots on NaCl(001),
110: an insulating substrate, by thermal deposition and variable-temperature atomic
111: force microscopy in ultrahigh vacuum. We show that, by properly choosing
112: the growth conditions, nanometer-scale magnetic iron dots with remarkably narrow size
113: distributions can be achieved in the absence of a wetting layer (VW growth).
114: Moreover, by changing the dosage of iron, we can tune both the vertical and lateral
115: sizes of the dots without introducing apparent size broadening, even though the clustering
116: is already in the strong coarsening regime, signified by the decrease in dot density
117: as a function of the iron dosage. The preserved narrowness in the island size
118: distributions is in clear contradiction with the expectations of existing understanding
119: of clustering on surfaces in the coarsening regime \cite{Zinke}.
120: We interpret these observations
121: within a phenomenological mean-field theory, in which a coverage-dependent optimal dot
122: size is selected by the competition between the self-energy of a dot, i.e. the total
123: energy of an isolated Fe dot on NaCl(100), and the energy of the strain-induced dipolar
124: interaction between the dots.
125:
126: The experiments were performed in an ultrahigh vacuum (UHV) system with base pressure
127: of 1$\times$10$^{10}$ torr. The system is equipped with electron beam sources, laser
128: molecular beam epitaxy, and an \textit{in-situ} Omicron variable-temperature UHV beam
129: deflection atomic force microscope (AFM)/ scanning tunneling microscope (STM) with
130: cooling and heating facilities covering a temperature range of (13-1500) K. The
131: noncontact mode AFM was used to study the surface morphology in this work. The NaCl
132: single crystal substrates were cleaved in air, then were immediately loaded into the
133: UHV chamber and were annealed to 530K for one hour to remove surface contamination
134: prior to the experiments. AFM images of the cleaved NaCl(001) surface show monatomic
135: height steps and large terraces with no detectable adsorbates. The iron was evaporated
136: from an Fe wire (5N purity) heated by electron beam bombardment at a rate of
137: 0.04 ML/min (1ML is equivalent to the nominal surface atomic density of bcc Fe(110),
138: 1.7$\times$10$^{15}$ atoms/cm$^2$). The iron dots can be formed only within a
139: finite temperature window, since it is known that the morphology of the substrate
140: changes dramatically if the temperature is above 720K \cite{Szymonski},
141: whereas at low temperatures, only random clusters or percolated iron films form \cite{Gai}.
142: The dots discussed in this
143: Letter are grown at a substrate temperature of 530K.
144:
145: \begin{figure}[!htb]
146: \begin{center}
147: \resizebox *{7.0cm}{7.0cm}{\includegraphics*{fig1.eps}}
148: \end{center}
149: \caption{Nanometer-scale iron dots grown on NaCl(001) with a nominal Fe dose of 1.7 ML.
150: (a) A typical NC-AFM image. Scan area 500 nm$\times$500 nm. (b) Close up and line profile
151: (150 nm long) of the dots. Scan area 200 nm$\times$200 nm. The unit in the line
152: profile is nm. (c) and (d) Diameter and height distributions of the dots shown in (a)
153: and the corresponding Gaussian fits. The center
154: and dispersion for the diameter distribution
155: are 13.5 nm and 1.1 nm, respectively; and for the height distribution are 3.76 nm
156: and 0.36 nm, respectively.}
157: \label{fig1}
158: \end{figure}
159: Nanometer-scale dots can be formed directly on the NaCl(001) surface without a
160: wetting layer, indicating that the growth proceeds in the Volmer-Weber mode. As
161: shown in the AFM images in Figs. 1 and 2, deposited Fe atoms tend to nucleate and
162: form dots even at submonolayer coverage. The preferred VW growth mode is not too
163: surprising, given the large surface free energy difference (2.48 J/m$^{2}$ for
164: Fe versus 0.18 J/m$^{2}$for NaCl)\cite{Boer} and the large lattice mismatch between Fe
165: and NaCl\cite{ltt}. The most eye-catching characteristic of the dots is their size
166: uniformity. Fig. 1 (a) shows a representative AFM image of the dots with a nominal
167: Fe thickness of 1.7ML. The very uniform dots are randomly distributed on the terraces.
168: Fig. 1 (b) is a close up together with a line profile of the dots. Estimated from the
169: line scan, the dots are 3.5 to 4 nm in height, and around 14 nm in diameter \cite{ldia}.
170: Fig. 1 (c) and (d) are the lateral size and height distributions of the dots shown
171: in Fig.1 (a) along with the corresponding Gaussian fits. Consistent with the line
172: profile, the centers of the Gaussian fits of the height and diameter distributions of
173: the dots appear at 3.76 nm and 13.5 nm, respectively. Each distribution has a very narrow
174: width. The dispersions, $\Delta h=(\langle h^2\rangle-\langle h\rangle^2)^{1/2}$,
175: of the height and diameter distributions are only 0.36 nm and
176: 1.1 nm, respectively, which are less than 10\% of the average height and diameter values.
177: Such narrow dispersions are comparable with the narrowest dispersions achieved in
178: semiconductor quantum dot growth via the SK mode \cite{Xin}, and are particularly
179: striking because clusters grown via the VW mode normally have much broader distributions
180: \cite{Wolf,Fruchart,huang,Baumer}.
181:
182:
183: \begin{figure}[!htb]
184: \begin{center}
185: \resizebox *{7.0cm}{11.0cm}{\includegraphics*{fig2.eps}}
186: \end{center}
187: \caption{Typical NC-AFM images and height and diameter distributions for dots grown
188: on NaCl(001) at different Fe coverages $\theta$. The image area for each
189: coverage is 150 nm$\times$150 nm. The dispersions for the height distributions for the four
190: coverages are all 0.36nm; and the dispersions for the diameter distributions are
191: 0.9nm, 1.1nm, 1.1nm, and 1.2nm, for the four coverages, respectively. }
192: \label{fig2}
193: \end{figure}
194: Another striking observation lies in the coverage dependence of the dots shown in
195: Fig. 2. As the Fe coverage increases, the dots grow larger and taller, but an optimal
196: size of the dots is always selected while the narrowness of their lateral and vertical
197: size distributions is preserved. Our quantitative analysis of the dispersions does not
198: yield any significant additional (coverage-dependent) broadening of the dot size
199: distributions, consistent with the impression drawn by naked eye from the AFM images
200: shown in Fig. 2. In Fig. 3, the optimal height and lateral size of the dots are plotted
201: as functions of the iron coverage, together with the density of the dots. The decrease
202: of the dot density with increasing coverage indicates that the growth of the dots is
203: clearly in the coarsening regime. This observation on the
204: variation of the island density makes the persistence of the narrow size distributions
205: even more dramatic, because in the coarsening regime, the \textit{absolute} size
206: distribution of the islands is expected to broaden with increasing coverage \cite{Zinke}.
207:
208: Preliminary surface magneto-optical Kerr effect (SMOKE) measurements show that these
209: iron dots are superparamagnetic in nature. The details of the magnetic investigation,
210: however, are beyond the scope of the present paper, which focuses on the formation of
211: the highly uniform dots. In order to understand the underlying formation mechanism of
212: these nanoclusters, it is highly desirable to know both their detailed atomic structures
213: and their stability. Unfortunately, the iron dots formed on NaCl cannot be imaged by STM
214: with higher (atomic) resolution. On the other hand, \textit{in-situ} time (from immediately
215: after formation up to 10 days) and temperature (from room temperature up to 550 K)
216: dependent AFM studies reveal little change in the morphology of the Fe dots, indicating
217: that the dot arrays on the surface are in an energetically favorable configuration that
218: is at least metastable.
219: \begin{figure}[!htb]
220: \begin{center}
221: \resizebox *{5.0cm}{6.0cm}{\includegraphics*{fig3.eps}}
222: \end{center}
223: \caption{The density (a), the peak height (b), and the peak diameter (c) of
224: the iron dots as functions of the Fe coverage obtained from the Gaussian fits,
225: in comparison with the mean field theory predictions (straight lines). Only a single
226: parameter, $\alpha=0.46$, is used to fit all the slopes of the three independent plots.}
227: \label{fig3}
228: \end{figure}
229:
230: \begin{figure}[!htb]
231: \begin{center}
232: \resizebox *{4.0cm}{3.0cm}{\includegraphics*{fig4.eps}}
233: \end{center}
234: \caption{
235: Schematic drawing of the energy per atom for the Fe dot assembly as a function of
236: the volume $V$ at a given coverage, showing the existence of local energy minimum.
237: }
238: \label{fig4}
239: \end{figure}
240: At the present, the precise underlying mechanism for the formation of these narrow-sized
241: iron dots is not completely clear, but the following phenomenological mean-field theory,
242: stressing on strain-mediated dot-dot interactions, is able to reproduce most of the salient
243: features observed in the experiment. First, the fact that the preferred size of the Fe
244: dots is hardly changed upon annealing clearly indicates that the energy per atom
245: $E$, equivalently, the chemical potential, of the Fe dot assembly must have a
246: local minimum at the observed optimal size for a given coverage. This suggests that the
247: qualitative form of the energy $E$ follows the generic behavior shown in Fig. 4
248: for a given coverage: When the volume approaches zero, $E$ is close to the
249: energy of one adatom; when the dot grows very large, $E$ is about the energy
250: per atom in a bulk Fe crystal, which is much smaller than the energy of one adatom.
251: Between the two limits there exists a local minimum, mainly induced by the elastic
252: relaxation caused by the discontinuity of the intrinsic surface stress tensor at the
253: dot edges \cite{review}. Secondly, the observation of different optimal sizes at different
254: coverages shown in Fig. 2 suggests that the local energy minimum is a function of
255: the iron coverage, implying that dot-dot interactions should contribute substantially
256: to the energy $E$. Based on these considerations, we can express $E$ as
257: a function of both the iron coverage $\theta$ and the dot size $V$
258: within the mean field approximation as
259: \begin{equation}
260: E(\theta,V)=E_0(V)+P(\theta,V)\,,
261: \end{equation}
262: where $E_0$ is the self-energy of an isolated dot and $P$
263: is the dot-dot interaction energy.
264:
265:
266: The interaction energy per atom,$P(\theta,V)$, is originated from misfit-induced
267: strain within
268: the system. By adopting the standard dipole-dipole interaction between two dots as
269: $P(V_1,V_2)\propto V_1V_2/r^3$\cite{review}, we have
270: \begin{equation}
271: \label{eq:p}
272: P(\theta,V)=g{\theta^{3/2}\over V^{1/2}}\,,
273: \end{equation}
274: where $g$ is a constant related to the interaction strength and the effective number of
275: the nearest neighbors of a given dot. We note that, on one hand, the long range dot-dot
276: interaction is important enough in influencing the energy per atom of the dot assembly,
277: thereby the existence of an optimal dot size; on the other hand, it is not strong enough
278: to cause the center of mass of a given dot to change, thereby the lacking of the spatial
279: order among the dots\cite{Liu}.
280:
281: The self-energy $E_0(V)$ of an isolated dot of size $V$ includes all possible
282: contributions to
283: the dot, such as the strain energy, interface energy, step energy, and kink energy,
284: but it does not depend on the total coverage \textit{$\theta$}. In principle,
285: an explicit, but rather complex, expression can be obtained from existing studies
286: of strain-mediated dot formation in heteroepitaxy, such as Eq. (4.15) of
287: Ref.\cite{review}. But surprisingly, our detailed analysis based on Eq. (4.15)\cite{review}
288: shows that no qualitatively correct fitting could be obtained for the range of iron
289: coverages explored in the present experiment. In the following, we choose the simple
290: expression
291: \begin{equation}
292: E_0(V)=bV^{\alpha}\,.
293: \end{equation}
294: as an alternative phenomenological approach, and show that the corresponding energy
295: per atom is capable of explaining many of the salient features observed in the experiment.
296:
297: First, by setting the first derivative of the energy per atom $E(\theta,V)$ with respect
298: to the
299: volume $V$ equal to zero, we obtain a coverage-dependent local energy minimum,
300: at the volume $b\alpha V^{\alpha+1/2}={g\over 2}\theta^{3/2}$.
301: This, in turn, yields the relation between the density of the dots $\theta/V$
302: and the coverage $\theta$
303: \begin{equation}
304: b\alpha\Big({\theta\over V}\Big)^{-(\alpha+1/2)}={g\over 2}\theta^{1-\alpha}\,.
305: \end{equation}
306: With $\alpha$ = 0.46, our theory gives an excellent fit to the experimental data
307: as shown in Fig. 3(a). Next, assuming that the equilibrium shape of the dots does not
308: change with the volume \cite{Duport}, we can also obtain the dot height (\textit{h})
309: and the diameter (\textit{D}) as functions of the coverage, given by
310: $h\propto \theta^{0.52}$ and $D\propto \theta^{0.52}$,
311: respectively. These two relationships also find very good fits to the experimental
312: data as shown in Figs. 3(b) and 3(c). The fact that the exponent $\alpha$ as a single
313: fitting parameter yields very good fittings to the slopes of three independent sets
314: of experimental observations hints on the physical validity of the choice made in Eq. (3).
315: Whereas the precise underlying physics for the success of Eq. (3) remains to be
316: explored, we suspect that it could be related to the unusually large lattice mismatch
317: in the present system \cite{ltt}.
318: Finally, we note that the dispersion of the distribution
319: of the island heights (or the diameters) shows little change with the coverage, whereas
320: our mean field theory predicts that the dispersion should increase with the coverage,
321: $\Delta h\propto \theta^{0.4}$. This discrepancy is not too surprising, because
322: within the mean-field theory,
323: the resulting optimal size of the islands can be reliably obtained, but the exponent of
324: the scaling function around the optimal value is typically overestimated, a trend well
325: recognized in critical phenomena \cite{Wilson}.
326:
327: In summary, we have shown that, under proper growth conditions, the self-assembly of
328: iron dots on the insulating surface of NaCl(001) leads to the formation of nanometer-scale
329: magnetic iron dots with narrow size distributions in the absence of a wetting layer.
330: We have also demonstrated that the vertical and lateral sizes of the dots can both
331: be changed by the iron coverage while the narrow size distributions are preserved,
332: even though the clustering is clearly in the strong coarsening regime.
333: These striking observations have been interpreted
334: successfully with a phenomenological mean-field theory.
335:
336: We are grateful to E.W. Plummer for invaluable discussions.
337: This work was supported in part by the LDRD of ORNL,
338: managed by UT-Battelle, LLC for the USDOE (
339: DE-AC05-00OR22725), and by the NSF (DMR-0071893 (BW)
340: and DMR-0105232 (JP)).
341:
342:
343:
344: \begin{thebibliography}{99}
345: \bibitem{Venables}J.A. Venables, Phil. Mag. \textbf{27}, 697 (1973).
346:
347: \bibitem{Zinke} For an excellent review of earlier works, see: M. Zinke-Allmang,
348: L.C. Feldman, and M.H. Grabow, Surf. Sci. Rep. \textbf{16}, 377 (1992).
349:
350: \bibitem{Wolf}P. Moriarty, Rep. Prog. Phys. \textbf{64}, 297 (2001);
351: S. A. Wolf \textit{et al.}, Science, \textbf{294}, 1488 (2001).
352:
353: \bibitem{review}V.A. Shchukin and D. Bimberg, Rev. Mod. Phys.
354: {\bf 71}, 1125 (1999) and references therein.
355:
356: \bibitem{Mo}Y.W. Mo, D.E. Savage, B.S. Swartzentruber, and M.G. Lagally,
357: Phys. Rev. Lett. \textbf{65}, 1020 (1990); J. Tersoff, C. Teichert, and
358: M.G. Lagally, Phys. Rev. Lett. \textbf{76}, 1675 (1996).
359:
360: \bibitem{Jesson}D.E. Jesson, M. K\"{a}stner, and B. Voigtl\"{a}nder,
361: Phys. Rev. Lett. \textbf{84}, 330 (2000); F.M. Ross, R.M. Tromp, and
362: M.C. Reuter, Science,\textbf{ 286}, 1931 (1999).
363:
364: \bibitem{Xie}Q. Xie, A. Madhukar, P. Chen, and N.P. Kobayashi,
365: Phys. Rev. Lett. \textbf{75}, 2542 (1995).
366:
367: \bibitem{Xin} D. Leonard \textit{et al.}, Appl. Phys. Lett. \textbf{63}, 3203 (1993);
368: V.M. Ustinov \textit{et al.}, Appl. Phys. Lett. 72, 362 (1998); S.H. Xin
369: \textit{et al.}, Appl. Phys. Lett. 69, 3884 (1996).
370:
371: \bibitem{Daruka}I. Daruka and A.L. Barabasi, Phys. Rev. Lett. \textbf{79}, 3708 (1997).
372:
373: \bibitem{Fruchart}
374: O. Fruchart, M. Klaua, J. Barthel, and J. Kirschner, Phys. Rev. Lett.
375: \textbf{83}, 2769 (1999); Y. Park, S. Adenwalla, G.P. Felcher,
376: S.D. Bader, Phys. Rev. B \textbf{52}, 12779 (1995); C.T. Yu, D.Q. Li,
377: J. Pearson, S.D. Bader, Appl. Phys. Lett. \textbf{79}, 3848 (2001).
378:
379: \bibitem{huang}L. Huang, S.J. Chey, and J.H. Weaver.
380: Phys. Rev. Lett. {\bf 80}, 4095 (1998).
381:
382: \bibitem{Baumer}M. B\"{a}umer, and H.J. Freund, Prog. Surf. Sci., \textbf{61}, 127 (1999);
383: A. Sugawara, and M.R. Scheinfein, Phys. Rev. B \textbf{56}, R8499 (1997).
384:
385: \bibitem{Szymonski}
386: M. Szymonski \textit{et al.}, Prog. Surf. Sci. \textbf{67}, 123 (2001).
387:
388: \bibitem{Gai}Z. Gai, G.A. Farnan, J. Pierce, and J. Shen, to be published.
389:
390: \bibitem{Boer}
391: F.R. de Boer \textit{et al.}, \textit{Cohesion in Metals}
392: (North Holland, Amsterdam, 1988);
393: P.A. Mulheran, Modelling Simul. Mater. Sci. Eng. 2, 1123 (1994).
394:
395: \bibitem{ltt}
396: The lattice constants of bcc Fe and NaCl are 2.87 $\mathring{\rm A}$
397: and 5.46 $\mathring{\rm A}$,
398: the nearest neighbors of bcc Fe(110) and NaCl(100) are separated by 2.49
399: $\mathring{\rm A}$ and 3.99 $\mathring{\rm A}$,
400: respectively. The Fe dots are likely in bcc phase because its fcc phase exists only in
401: bulk at temperatures above 1186 K, or in Fe films grown on fcc substrates
402: (e.g. Cu(100)) whose lattice constants are very close to that of Fe.
403:
404: \bibitem{ldia}
405: The lateral diameters of the dots are overestimated because of the
406: tip effect. However, the lateral size can be estimated with an assumed
407: shape of the dots in addition to the known information of height, density, and total
408: dosages. For the 1.7 ML Fe dot, assuming a truncated
409: pyramid shape, the average lateral size of the dots is around 7 to 8 nm
410: instead of 14 nm.
411:
412: \bibitem{Liu}F. Liu, A.H. Li, and M.G. Lagally, Phys. Rev. Lett.
413: \textbf{87}, 126103 (2001).
414:
415: \bibitem{Duport}C. Duport, C. Priester, and J. Villain, in
416: \textit{Morphological Organization During growth and Removal,} edited by
417: Z.Y. Zhang and M.G. Lagally (World Scientific, Singapore 1998).
418:
419: \bibitem{Wilson}
420: K.G. Wilson, Rev. Mod. Phys. \textbf{55}, 583 (1983).
421: \end{thebibliography}
422: \end{document}
423: