1: \documentclass[twocolumn,aps,prb]{revtex4}
2: \usepackage{graphicx}
3:
4: \begin{document}
5: \title{X-ray diffraction peak profiles from threading dislocations in
6: GaN epitaxial films}
7: \author{V. M. Kaganer}
8: \author{O. Brandt}
9: \author{A. Trampert}
10: \author{K. H. Ploog}
11: \affiliation{Paul-Drude-Institut f\"{u}r Festk\"{o}rperelektronik,
12: Hausvogteiplatz 5--7, D-10117 Berlin, Germany}
13:
14: \begin{abstract}
15: We analyze the lineshape of x-ray diffraction profiles of GaN epitaxial
16: layers with large densities of randomly distributed threading
17: dislocations. The peaks are Gaussian only in the central, most intense
18: part of the peak, while the tails obey a power law. The $q^{-3}$ decay
19: typical for random dislocations is observed in double-crystal rocking
20: curves. The entire profile is well fitted by a restricted random
21: dislocation distribution. The densities of both edge and screw threading
22: dislocations and the ranges of dislocation correlations are obtained.
23: \end{abstract}
24:
25: \date{\today}
26:
27: \pacs{61.10.Nz, 61.12.Bt, 61.72.Ff, 68.55.-a}
28:
29: \maketitle
30:
31: \section{Introduction}
32:
33: GaN epitaxial layers grown on different substrates (e.g.,
34: Al$_{2}$O$_{3}$, SiC, or Si) possess very large densities of threading
35: dislocations which cross the layer along its normal, from the
36: layer-substrate interface to the surface.\cite{gilbook} The threading
37: dislocation density depends only marginally on the substrate material
38: (and hence on the misfit between the substrate and the layer) but rather
39: on the growth technique and conditions. For (0001) oriented layers of
40: wurtzite GaN, the overwhelming majority of dislocations are of edge type
41: with Burgers vectors $\mathbf{b}=\frac{1}{3} \left\langle
42: 11\overline{2}0\right\rangle $. Their
43: density\cite
44: {srikant4286jap,metzger1013pma,heinke2145apl,sun02,chierchia03} is
45: typically $10^{8}-10^{10}$~cm$^{-2}$. The density of screw dislocations
46: with Burgers vector $\mathbf{b}=[0001]$ is one to two orders of magnitude
47: lower than the density of edge dislocations.
48:
49: \begin{figure}[tbp]
50: \noindent
51: \includegraphics[width=\columnwidth]{skew.eps} \caption{Sketch of skew
52: geometry x-ray diffraction. The lattice planes of the actual reflection
53: are depicted in the left bottom part of the figure. The incident wave
54: $\mathbf{K}^{\rm in}$ and the diffracted wave $\mathbf{K}^{\rm out}$ make
55: the same angle $\Phi$ to the sample surface. The scattering vector
56: $\mathbf{Q}$ makes an angle $\Psi$ to the sample surface.}
57: \label{skew}
58: \end{figure}
59:
60: The dislocation density can be measured directly by plan-view
61: transmission electron microscopy (TEM). The actual virtue of TEM is not
62: the accurate determination of the dislocation density, but rather the
63: possibility to determine the type of the dislocation distribution (e.~g.,
64: random vs.\ granular/columnar structure). It is of limited statistical
65: significance considering the small area covered by TEM micrographs.
66: Alternatively, x-ray diffraction can be used to detect the lattice
67: distortions due to the presence of dislocations. In principle, both the
68: dislocation distribution and the actual dislocation density may be
69: obtained from x-ray diffraction profiles.
70:
71: The impact of screw and edge threading dislocations on the width of the
72: x-ray reflections in the limiting cases of lattice planes parallel and
73: perpendicular to the surface is commonly referred to as tilt and twist,
74: respectively.\cite
75: {srikant4286jap,metzger1013pma,heinke2145apl,sun02,chierchia03}
76: This designation stems from the model of misoriented
77: blocks\cite{chierchia03} which is \emph{not} appropriate for strains
78: caused by randomly distributed dislocations. In this case, the
79: description in terms of mean-squared distortions\cite{ratnikov00} is
80: actually more adequate. Moreover, all previous experimental studies
81: only use the full width at half-maximum (FWHM) of different
82: reflections as a measure of the dislocation density.
83:
84: The symmetric Bragg reflections from GaN layers are comparatively narrow,
85: since they are not influenced by the presence of edge dislocations. Edge
86: dislocations produce distortions within lattice planes parallel to the
87: surface but do not disturb positions of these planes along the layer
88: normal. The highest sensitivity to edge dislocations is obtained by
89: diffraction from lattice planes perpendicular to the surface. This
90: diffraction geometry requires grazing incidence illumination and is thus
91: commonly performed at a synchrotron.\cite{munkholm2972apl,yang4240jap} An
92: alternative geometry that can easily be realized in the laboratory is the
93: skew geometry,\cite{srikant4286jap,sun02} which is a quasi-symmetric (the
94: incident and the diffracted waves have the same angles to the surface)
95: non-coplanar (the surface normal does not lie in the plane defined by the
96: incident and the diffracted waves) geometry, as shown in Fig.\
97: \ref{skew}. By measuring different reflections with increasing lattice
98: plane inclination, one can extrapolate to lattice planes perpendicular to
99: the surface.\cite{srikant4286jap,sun02} A single reflection with a large
100: inclination can be regarded as a figure of merit.\cite{heinke2145apl} A
101: four-circle diffractometer\cite{busing457ac} is required for skew
102: geometry measurements, since the sample has to be tilted with respect to
103: its surface normal. Coplanar asymmetric reflections on a three-circle
104: diffractometer are much less sensitive to edge dislocations since they
105: only partially touch the lattice distortions parallel to the surface
106: plane.\cite{sun02}
107:
108: The FWHM of the diffraction peak depends not only on the dislocation
109: density, but also on the correlations between dislocations. Furthermore,
110: it depends on the mutual orientations of the scattering vector,
111: dislocation line direction, and the Burgers vector direction, which have
112: not been taken into consideration in previous diffraction studies of GaN
113: layers. In the present paper, we analyze the entire lineshape of the
114: diffraction profiles, and particularly their tails. These tails are due
115: to scattering in the close vicinity of the dislocation lines and are not
116: influenced by the correlations between dislocations. They follow
117: universal power laws and can be used to determine the dislocation
118: density. We fit the entire diffraction peak profile to the numerical
119: Fourier transform of the pair correlation function and simultaneously
120: obtain dislocation densities and the range of dislocation correlations.
121:
122: \section{Background}
123:
124: X-ray diffraction is a well-established technique to analyze crystal
125: lattice imperfections.\cite{warren:book69,krivoglaz:book} The
126: conventional and widely used methods (also in GaN
127: studies\cite{chierchia03}) are based on a comparison of the diffraction
128: peak widths of different reflections.\cite
129: {WilliamsonHall53,hordon61,ayers94} They aim to separate two
130: contributions to the peak broadening, the finite size of the crystalline
131: domains (grains) in the sample and the non-uniform strain within each
132: domain owing to lattice defects. The strain broadening of a diffraction
133: peak is proportional to the length of the scattering vector, while the
134: size effect does not depend on it. Thus, comparing the peak widths of
135: reflections of successive orders in the so-called Williamson-Hall plot,
136: one can separate both contributions. The separation method assumes that
137: all peaks are Gaussian,\cite{hordon61,ayers94} or all
138: Lorentzian.\cite{WilliamsonHall53} This is certainly not true, and the
139: methods based on the peak width give only rough, albeit instructive,
140: estimates of the crystal perfection. A recent development
141: \cite{ungar96,ungar99,ungar01} includes corrections for given
142: orientations of the dislocation lines and Burgers vectors with respect to
143: the scattering vector.
144:
145: The next milestone was the Fourier analysis of the diffraction peak
146: profile proposed by Warren and
147: Averbach.\cite{WarrenAverbach50,WarrenAverbach52} The starting point of
148: their analysis is the average over grain orientations in powder
149: diffraction. The corresponding integration of the scattered intensity in
150: reciprocal space is equivalent to a one-dimensional cut of the
151: correlation function in real space. We note that the powder diffraction
152: literature does not use the term correlation function but refers to the
153: Fourier coefficients of the intensity. We use the solid-state physics
154: terminology where this quantity is commonly called the pair correlation
155: function. In crystallography, the same function is referred to as
156: Patterson function.
157:
158: The Ansatz of the Warren-Averbach analysis is the assumption that the
159: correlation function is a product of two independent terms describing
160: the size and the strain effects, respectively. Furthermore, it is assumed
161: that the finite sizes of the grains give rise to a Lorentzian peak
162: (exponential correlation function) while non-uniform strain is described
163: by a Gaussian function. With these assumptions, the size and strain
164: effects are separated by differentiation of the Fourier-transformed peak
165: profile. In many cases, the latter step is not sufficiently accurate,
166: since it has to rely on a few Fourier coefficients obtained from noisy
167: experimental data. Balzar\cite{balzar92,balzar93} suggested to avoid this
168: difficulty by directly fitting the peak profile to a convolution of a
169: Lorentzian and a Gaussian, which is the Voigt function. The experimental
170: peak profiles are not always described by the Voigt function and various
171: other analytical functions were suggested on a purely phenomenological
172: basis.\cite {young82,keijser83}
173:
174: The assumption that non-uniform strain gives rise to a Gaussian
175: correlation function looks plausible, taking into account the stochastic
176: nature of this strain originating from randomly distributed lattice
177: defects. However, Krivoglaz and Ryboshapka\cite
178: {krivoglaz63,krivoglaz:book} have shown that this is not true for random
179: dislocations, which are the most common and most important source of
180: strain. Rather, they found that the slow ($\propto r^{-1}$) decay of the
181: strain with the distance $r$ from the dislocation line gives rise to a
182: correlation function
183: \begin{equation}
184: G(x)=\exp (-C\rho x^{2}\ln \frac{L}{\xi x}). \label{eq1}
185: \end{equation}
186: Here $C\sim 1$ is a dimensionless factor depending on the orientations of
187: the dislocation line and Burgers vector with respect to the $x$-direction
188: (an arbitrary direction along which the correlations are measured). Its
189: dependence on the scattering vector $\mathbf{Q}$ and the Burgers vector
190: $\mathbf{b}$ is given by $C\propto (\mathbf{Q}\cdot \mathbf{b})^{2}$. The
191: dislocation density $\rho $ is defined as a total length of the
192: dislocation lines per unit volume. For straight dislocations, $\rho $ is
193: equal to the number of dislocations crossing the unit area of a plane
194: perpendicular to the dislocation lines. $\xi \sim 1$ is another
195: dimensionless factor, for uncorrelated dislocations $\xi =\left|
196: \mathbf{Q}\cdot \mathbf{b} \right| /2\pi $.
197:
198: For completely random and uncorrelated dislocations, $L$ is the crystal
199: size, so that the diffraction peak width of an infinite crystal with
200: random uncorrelated dislocations tends to infinity. Wilkens\cite
201: {wilkens69,wilkens70nbs,wilkens70pss} pointed out that this divergence
202: has the same origin as the divergence of elastic energy of a crystal with
203: dislocations: the elastic energy is proportional to $\mu b^{2}\ln L/a$,
204: where $\mu $ is the shear modulus, $b$ is the length of the Burgers
205: vector, and $a$ is the lattice spacing. He suggested that the system can
206: drastically reduce elastic energy by a very minor rearrangement in the
207: dislocation ensemble: the positions of the dislocations remain random but
208: their Burgers vectors are correlated, so that the total Burgers vector in
209: a region exceeding some characteristic scale is zero. Then, the
210: dislocations screen each other and the elastic energy, as well as the
211: diffraction peak width, remain finite. Wilkens introduced a ``restricted
212: random distribution'' of dislocations by subdividing the crystal into
213: cells, such that the total Burgers vector in each cell is zero. He
214: found\cite {wilkens69,wilkens70nbs,wilkens70pss} that the functional form
215: of Eq.\ (\ref {eq1}) does not change but $L$ should be understood as a
216: finite size of the cells. Krivoglaz \emph{et
217: al.}\cite{krivoglaz83,krivoglaz:book} showed that the same result is
218: valid for a broad class of correlated dislocation distributions with
219: screening.
220:
221: We note also the two-dimensional crystal as a limiting case of
222: dislocation screening. Here the elastic energy of dislocations is to be
223: compared with the entropic term $-TS$, where $T$ is temperature and the
224: entropy $S=\ln (L/a)^{2}$ is the number of lattice sites where the
225: dislocation can be placed. Both elastic energy and entropy terms are
226: proportional to $\ln L$. As a result, above some temperature $T_{m}$,
227: dislocations are generated by unbinding of thermally excited dislocation
228: pairs, giving rise to the dislocation mediated
229: melting.\cite{KosterlitzThouless72,KosterlitzThouless73} The calculation
230: of the correlation function in this highly correlated dislocation
231: system\cite{kaganer94prl} shows that the logarithmic term in Eq.\
232: (\ref{eq1}) vanishes.
233:
234: Fourier transformation of the correlation function (\ref{eq1}) yields a
235: Gaussian shape only in the central part of the peak. The range of the
236: Gaussian peak shape is given by a dimensionless factor
237: \begin{equation}
238: M=L\sqrt{\rho }
239: \label{eq1a}
240: \end{equation}
241: and increases when $M$ is increased. The intensity distribution notably
242: deviates from the Gaussian shape at the tails of the diffraction peak.
243:
244: The tails of the diffraction peak due to dislocations follow a universal
245: law $I(q)\propto q^{-3}$, which can be obtained from Fourier
246: transformation of Eq.\ (\ref{eq1}).\cite{groma98} The $q^{-3}$ law is due
247: to the fact that at large $q$, the scattering takes place in the strained
248: regions close to dislocation lines, where the lattice is so strongly
249: misoriented that the Bragg law is locally satisfied for the wave vector
250: $q$. Calculation of the volume of these regions give the $q^{-3}$
251: dependence.\cite {wilkens63,krivoglaz:book,groma98} Groma and
252: co-workers\cite {groma98,groma00,borbely01} developed methods for the
253: peak profile analysis based on a calculation of the restricted moments of
254: $I(q)$. In particular, the second-order restricted moment $v(q)=
255: \int_{-q}^{q}q^{2}I(q)dq $ is proportional to $\ln q$, which they used to
256: determine the mean dislocation density. Higher-order moments describe
257: fluctuations of the dislocation density.
258:
259: GaN epitaxial film comprise a well-defined system where threading
260: dislocations are aligned perpendicular to the surface plane. The film is
261: oriented, contrary to a powder. However, the x-ray diffraction
262: measurements performed with an open detector give rise to an average very
263: similar to the powder average. The intensity is integrated over
264: directions of the outgoing beam, instead of the integration over
265: directions of the diffraction vector. The integrations in reciprocal
266: space gives rise to cuts in real space, which are different for the two
267: cases under consideration. The coordinate $x$ in Eq.\ (\ref{eq1}) runs
268: along the direction of the outgoing wave in case of the oriented sample
269: with open detector and in the direction of the diffraction vector for the
270: case of powder diffraction. This fact introduces a geometrical correction
271: in the orientational factor $C$. The skew diffraction geometry used for
272: the measurements gives rise to further corrections, which are calculated
273: below.
274:
275: Our approach consists of a direct fit of the measured intensities by the
276: numerical Fourier transformation of the correlation function (\ref{eq1}),
277: thus avoiding any transformation of the experimental data. We expect that
278: such a calculation is less influenced by the noise in the experimental
279: data and is more reliable. As shown below, we find good agreement between
280: measured and calculated peak profiles. From the fits, we reliably obtain
281: both the dislocation densities and the correlation range in the
282: restricted random dislocation distribution.
283:
284: \section{Theory}
285: \label{sec:theory}
286:
287: The intensity of x-ray scattering from a crystal disturbed by strain
288: fields of lattice defects can be represented as a Fourier transform
289: \begin{equation}
290: I(\mathbf{q})=\int G(\mathbf{r})\exp (i\mathbf{q}\cdot \mathbf{r})
291: d\mathbf{r}
292: \label{eq2}
293: \end{equation}
294: of the pair correlation function
295: \begin{equation}
296: G(\mathbf{r})=\left\langle \exp \{i\mathbf{Q}\cdot [\mathbf{U}
297: (\mathbf{r})- \mathbf{U}(0)]\}\right\rangle .
298: \label{eq3}
299: \end{equation}
300: Here $\mathbf{Q}=\mathbf{K}^{\mathrm{out}}-\mathbf{K}^{\mathrm{in}}$ is
301: the scattering vector ($\mathbf{K}^{\mathrm{in}}$ and
302: $\mathbf{K}^{\mathrm{out}}$ are the wave vectors of the incident and
303: scattered waves, respectively) and $\mathbf{q}=\mathbf{Q}-\mathbf{Q}_0$
304: is a small deviation of $\mathbf{Q}$ from the nearest reciprocal lattice
305: vector $\mathbf{Q}_0$, so that $q\ll Q$. $\mathbf{U}(\mathbf{r})$ is the
306: sum of displacements produced by all defects of the crystal in a given
307: point $\mathbf{r}$ and $\left\langle \ldots \right\rangle $ denotes the
308: average over the statistics of the defect distribution. Equation
309: (\ref{eq3}) implies an infinite and statistically uniform sample, so that
310: the choice of origin is arbitrary. When finite size effects are
311: essential, as for example for misfit dislocations in epitaxial
312: layers,\cite{kaganer97} the correlation function $G(\mathbf{r}_{1},
313: \mathbf{r}_{2})$ depends on the difference of displacements $\mathbf{U}
314: (\mathbf{r}_{1}) - \mathbf{U}(\mathbf{r}_{2})$ and Eq.\ (\ref{eq2})
315: contains the exponent $\exp [i\mathbf{q}\cdot
316: (\mathbf{r}_{1}-\mathbf{r}_{2})]$.
317:
318: We restrict ourselves to parallel straight dislocations and take
319: the direction of the dislocation lines as $z$ axis. Then,
320: $\mathbf{r}=(x,y)$ is a two-dimensional vector in the plane perpendicular
321: to the dislocation lines and Eq.\ (\ref{eq2}) can be written as
322: \begin{equation}
323: I(\mathbf{q})=\delta (q_{z})\int G(x,y)\exp (iq_{x}x+iq_{y}y)dxdy,
324: \label{eq2a}
325: \end{equation}
326: where the delta-function $\delta (q_{z})$ is due to the translational
327: invariance in the direction of the dislocation lines.
328:
329: In the experiments described in the subsequent sections, the x-ray
330: scattering measurements from oriented samples are performed with a wide
331: open detector. The intensity (\ref{eq2}) is then to be integrated over
332: all directions of the scattered wave $\mathbf{K}^{\mathrm{out}}$. The
333: result of this integration is very similar to the powder average. The
334: actual part of the sphere $\left| \mathbf{K}^{\mathrm{out}}\right| =k$
335: (where $k$ is the wave vector) can be replaced by the plane perpendicular
336: to the direction of $\mathbf{K}^{\mathrm{out}}$. Integration of the
337: intensity (\ref{eq2a}) over this plane gives rise to a one-dimensional
338: integral
339: \begin{equation}
340: \mathcal{I}(q)=\int G(x)\exp (iqx/\cos \Phi )dx, \label{eq7}
341: \end{equation}
342: where $\Phi $ is the angle between the $(x,y)$ plane and $\mathbf{K}^
343: {\mathrm{out}}$ (see Fig.\ \ref{skew}). It is given by $\sin \Phi =\sin
344: \Psi \sin \theta _{\mathrm{B}}$, where $\Psi $ is the angle between the
345: $(x,y)$ plane and the scattering vector $\mathbf{Q}$ and $\theta
346: _{\mathrm{B}}$ is the Bragg angle. The $x$ axis is chosen along the
347: projection of $\mathbf{K}^ {\mathrm{out}}$ on the plane perpendicular to
348: the dislocation lines. The corresponding expression for the case of
349: powder diffraction differs only by the direction of $x$: it runs along
350: $\mathbf{K}^{\mathrm{out}}$ for oriented films and along $\mathbf{Q}$ for
351: powder diffraction. The wave vector $q$ in Eq.\ (\ref{eq7}) is the
352: projection of $\mathbf{q}$ on the direction of
353: $\mathbf{K}^{\mathrm{out}}$, so that $q=Q \omega \cos \theta _{B}$, where
354: $\omega $ is the angular deviation from the peak center.
355:
356: Krivoglaz\cite{krivoglaz:book,krivoglaz61,krivoglaz63} performed the
357: average over an uncorrelated defect distribution and showed that the
358: correlation function can be represented as
359: \begin{equation}
360: G(\mathbf{r})=\exp \left\{ -\rho \int \left[ 1-e^{i\mathbf{Q}\cdot
361: [\mathbf{u} (\mathbf{R+r})-\mathbf{u}(\mathbf{R})]}\right]
362: d\mathbf{R}\right\} .
363: \label{eq4}
364: \end{equation}
365: Here $\mathbf{u}(\mathbf{r})$ is the displacement field produced at a
366: point $\mathbf{r}$ by a single defect located at the origin and $\rho $ is
367: the defect density. For straight dislocations, the integration in Eq.\
368: (\ref{eq4}) is performed over the $\mathbf{R}=(X,Y)$ plane perpendicular
369: to the dislocation lines and $\rho $ is the dislocation density per unit
370: area. Thus, according to Eq.\ (\ref{eq4}), in order to obtain positional
371: correlations between two points separated by a distance $\mathbf{r}$, one
372: has to place a dislocation in an arbitrary position $\mathbf{R}$ and
373: perform the integration over all $\mathbf{R}$.
374:
375: As a result of the slow decay of the dislocation strain, the main
376: contribution to the integral is due to remote dislocations, $R\gg r$, so
377: that the difference of displacements can be expanded as Taylor series,
378: $\mathbf{Q}\cdot [\mathbf{u}(\mathbf{R+r})-\mathbf{u}(\mathbf{R})]\approx
379: (\mathbf{r}\cdot \nabla )[\mathbf{Q}\cdot \mathbf{u}(\mathbf{R})]
380: =r_{i}Q_{j}\partial u_{j}/\partial X_{i}$. The distortion field of a
381: dislocation has a universal $r$-dependence, $\partial u_{j}/\partial
382: X_{i}=b\psi _{ij}/2\pi R$, where $b$ is the length of the Burgers vector
383: and $\psi _{ij}$ is a dimensionless factor of the order of unity which
384: depends only on the azimuth $\phi $. If the densities of dislocations
385: with opposite signs of the Burgers vectors are equal (there is no plastic
386: bend), the imaginary part of the correlation function (\ref{eq4}) is zero
387: and its real part, after expanding the exponent in curly brackets,
388: becomes\cite{krivoglaz63}
389: \begin{equation}
390: G(\mathbf{r})=\exp \left\{ -C\rho r^{2}\int \frac{dR}{R}\right\} ,
391: \label{eq5}
392: \end{equation}
393: where
394: \begin{equation}
395: C=\gamma (Qb)^{2}/4\pi ,\qquad \gamma =\frac{1}{2\pi }\int_{0}^{2\pi }
396: \left(\hat{r}_{i}\psi _{ij}\hat{Q}_{j}\right) ^{2}d\phi .
397: \label{eq6}
398: \end{equation}
399: $C$ and $\gamma $ are dimensionless factors of the order of unity. Here
400: $\mathbf{\hat{r}}$ and $\mathbf{\hat{Q}}$ are unit vectors in the
401: directions of $\mathbf{r}$ and $\mathbf{Q}$, respectively. The integral
402: in Eq.\ (\ref{eq5}) is taken from $\xi r$ (where $\xi \sim 1$ is a
403: dimensionless factor) to a limiting size $L$, which, for completely
404: uncorrelated dislocations, is equal to the crystal size. Then, the
405: integral is equal to $\ln (L/\xi r)$, and we arrive at Eq.\ (\ref{eq1}).
406: When the dislocations are correlated, so that the total Burgers vector
407: averaged over a certain characteristic scale is zero, the functional form
408: of the correlation function does not change but $L$ has the meaning of
409: that scale.\cite
410: {wilkens69,wilkens70nbs,wilkens70pss,krivoglaz83,krivoglaz:book}
411:
412: The integration range in Eq.\ (\ref{eq7}) is limited by distances smaller
413: than $L$. A finite integration limit introduced in Eq.\ (\ref{eq7})
414: leads, in a numerical evaluation of the integral, to unphysical
415: oscillations in $\mathcal{I}(q)$ commonly appearing in Fourier integrals
416: taken over a rigidly limited interval. We found that an appropriate
417: smoothing is obtained by substituting $\ln (L/\xi x)$ in Eq.\ (\ref{eq1})
418: with $\ln [(L+\xi x)/\xi x]$ and extending the integration range to
419: infinity. The expressions for $\xi =\xi (\mathbf{Q})$ are bulky and
420: depend on the type of correlations in dislocation positions.\cite
421: {wilkens70nbs,wilkens70pss,krivoglaz:book,krivoglaz83} We restrict
422: ourselves to the first approximation that does not depend on the type of
423: correlations, $\xi =\left| \mathbf{Q}\cdot \mathbf{b}\right| /2\pi $.
424:
425: Finally, combining the equations above, the diffracted intensity can be
426: represented as
427: \begin{eqnarray}
428: \mathcal{I}(\omega ) &=&I_{0}\int_{0}^{\infty }\exp (-Ax^{2}
429: \ln \frac{B+x}{x}) \cos (\omega x)dx \nonumber \\
430: &+&I_{\mathrm{backgr}}.
431: \label{eq9}
432: \end{eqnarray}
433: We proceed here to the angular deviation from the peak maximum $\omega $
434: and introduce the peak height $I_{0}$ and the background intensity
435: $I_{\mathrm{backgr}}$ to provide the exact formula that is used for the
436: fits of the experimental peaks presented below. Parameters $A$
437: (describing the dislocation density) and $B$\ (describing the
438: dislocation correlation range) are given by
439: \begin{equation}
440: A=f\rho b^{2},\qquad B=gL/b. \label{eq10}
441: \end{equation}
442: Here $f$ and $g$ are dimensionless quantities given by the diffraction
443: geometry,
444: \begin{equation}
445: f=\frac{\gamma }{4\pi }\frac{\cos ^{2}\Phi }{\cos ^{2}\theta _{B}},
446: \qquad g= \frac{2\pi \cos \theta _{\mathrm{B}}}{\cos \Phi \cos \Psi }.
447: \label{eq11}
448: \end{equation}
449: In Eq.~(\ref{eq11}), the expression for $g$ is written for edge
450: dislocations, taking into account the approximation $\xi =\left|
451: \mathbf{Q} \cdot \mathbf{b}\right| /2\pi $. For screw dislocations, $\cos
452: \Psi $ in this expression should be replaced by $\sin \Psi $. The length
453: of the Burgers vector $b$ in Eq.~(\ref{eq10}) is that of the relevant
454: Burgers vector for either edge or screw dislocations. The dimensionless
455: product $\rho b^{2}\ll 1$ is the mean number of dislocations crossing
456: each $b\times b$ cell in the plane perpendicular to the dislocation
457: lines. Equation (\ref{eq9}) with four parameters,
458: $A,B,I_{0},I_{\mathrm{backgr}}$ is used in Sec.\ \ref {sec:results}
459: below to fit the peak profiles and obtain the dislocation density $\rho $
460: and the length $L$.
461:
462: \begin{figure}[tbp]
463: \noindent \includegraphics[width=\columnwidth]{asymptotics.eps}
464: \caption{Behavior of the integral (\ref{eq12}) for different values of
465: the parameter $R$. All curves merge at a common $q^{-3}$ asymptotic. A
466: Gaussian peak profile is shown by the thin line for comparison.}
467: \label{asymptotics}
468: \end{figure}
469:
470: The behavior of the integral (\ref{eq9}) is illustrated in Fig.\ \ref
471: {asymptotics} where the function
472: \begin{equation}
473: I(q)=\int_{0}^{\infty }\exp \{-x^{2}\ln [(R+x)/x]\}\cos (qx)dx
474: \label{eq12}
475: \end{equation}
476: is numerically calculated for different values of the parameter $R$. The
477: curves merge at a common $q^{-3}$ asymptotic that does not depend on
478: $R$.\cite{krivoglaz:book,wilkens70pss,groma98,groma00} Then, Eq.\
479: (\ref{eq9}) has an asympotic behavior (for $\omega $ large in comparison
480: with the peak width)
481: \begin{equation}
482: \mathcal{I}(\omega )=A\frac{\pi I_{0}}{\omega ^{3}}+I_{\mathrm{backgr}}.
483: \label{eq13}
484: \end{equation}
485: We note that $\pi I_{0}$ is the integrated intensity of the peak.
486:
487: Figure \ref{asymptotics} also shows that the asymptotics (\ref{eq13}) is
488: reached quite close to the peak center for $R\sim 1$, while for $R\gg 1$
489: the central part of the peak is Gaussian and the angular range where the
490: Gaussian approximation is valid increases with increasing $R$. The
491: FWHM of the calculated peaks increases with
492: increasing $R$ and can be approximated by
493: \begin{equation}
494: \Delta q\approx 2.4+\ln R.
495: \label{eq13a}
496: \end{equation}
497:
498: Groma\cite{groma98} suggested to use the second restricted moment of the
499: intensity distribution
500: \begin{equation}
501: v_{2}(\omega )=\int_{-\omega }^{\omega }\varpi ^{2}[\mathcal{I}(\varpi )-
502: I_{\mathrm{backgr}}]d\varpi \label{eq14}
503: \end{equation}
504: to obtain the dislocation density from the asymptotic behavior
505: (\ref{eq13}). Note that the integral (\ref{eq14}) diverges, when taken in
506: infinite limits. One finds, by substituting (\ref{eq13}) into
507: (\ref{eq14}), that
508: \begin{equation}
509: v_{2}(\omega )=2\pi I_{0}A\ln \omega +\mathrm{const.}
510: \label{eq15}
511: \end{equation}
512:
513: It remains to calculate the orientational factor $C$ [Eq.\ (\ref{eq6})].
514: In the case of powder
515: diffraction,\cite{wilkens69,wilkens70nbs,wilkens70pss} the vector
516: $\mathbf{r}$ is directed along the projection $\mathbf{Q}_{\perp }$ of
517: the scattering vector $\mathbf{Q}$ on the $(x,y)$ plane. In our case, the
518: vector $\mathbf{r}$ is directed along the projection of
519: $\mathbf{K}^{\mathrm{out}}$ on the $(x,y)$ plane and makes an angle
520: $\alpha $ with the vector $\mathbf{Q}_{\perp }$, see Fig.\ \ref{skew}.
521: This angle is given by $\cos \alpha =\sin \theta _{\mathrm{B}}\cos \Psi
522: /\cos \Phi $. When evaluating the integral (\ref{eq6}) for edge
523: dislocations, we average $\gamma $ over possible orientations of the
524: Burgers vector in a hexagonal lattice (the dislocation lines are taken
525: along the sixfold axis) and obtain
526: \begin{equation}
527: \gamma _{\mathrm{e}}=\frac{9-8\nu +8\nu ^{2}-2(3-4\nu )\cos ^{2}\alpha }
528: {16(1-\nu )^{2}}\cos ^{2}\Psi , \label{eq8}
529: \end{equation}
530: where $\nu $ is the Poisson ratio. $\gamma _{\mathrm{e}}$ depends only
531: weakly on $\alpha $. Taking $\nu =1/3$, one can approximate $\gamma
532: _{\mathrm{e}}\approx \cos ^{2}\Psi $ in the whole range of reasonable
533: angles $\alpha $. The calculation for screw dislocations gives
534: \begin{equation}
535: \gamma _{\mathrm{s}}=\frac{1}{2}\sin ^{2}\Psi . \label{eq16}
536: \end{equation}
537:
538: Two limiting cases are of interest: a symmetric Bragg reflection to study
539: screw dislocations and a grazing incidence/grazing exit reflection as an
540: extreme case of a highly asymmetric skew geometry. For a symmetric Bragg
541: reflection, $\Psi =\pi /2$ and $\Phi =\theta _{B}$. Then, we obtain
542: $f=1/8\pi $ and, for screw dislocations, $g=2\pi $. The grazing
543: incidence/grazing exit geometry is the limit $\Phi =\Psi =0$, so that
544: $f=\gamma /(4\pi \cos ^{2}\theta _{B})$ and, for edge dislocations,
545: $g=2\pi \cos \theta _{\mathrm{B}}$.
546:
547: The finite thickness $T$ of the epitaxial layer (with threading
548: dislocations perpendicular to the layer plane)\ can be taken into account
549: by making the Warren-Averbach ansatz\cite{WarrenAverbach50}
550: \begin{equation}
551: G(x)=G_{\mathrm{d}}(x)G_{\mathrm{s}}(x), \label{eq17}
552: \end{equation}
553: where $G_{\mathrm{d}}(x)$ is the correlation function considered above.
554: The correlation function describing finite size effects can be written as
555: $G_{\mathrm{s}}(x)=\exp (-x/T)$. Its Fourier transformation is a
556: Lorentzian that is expected from the finite slit function $[\sin
557: (qT/2)/(qT/2)]^{2}$ after averaging over thickness variations. We note
558: that, in asymmetric reflections, Eq.\ (\ref{eq7}) gives rise to the
559: effective thickness $T/\cos \Phi $ along the direction of the diffracted
560: wave. The finite size correction (\ref{eq17}) does not complicate the
561: calculation of the intensity by numerical integration of Eq.\
562: (\ref{eq7}). If the resolution of the experiment cannot be neglected in
563: comparison with the peak width, the correlation function (\ref{eq17}) is
564: to be multiplied with the real-space resolution function
565: $\mathcal{R}(x)$, which also does not lead to additional complications of
566: the numerical integration.
567:
568: \section{Experiment}
569:
570: The GaN layers investigated here were grown on 6H-SiC(0001) by
571: plasma-assisted molecular beam epitaxy (PAMBE). The two PAMBE systems
572: employed are equipped with a solid-source effusion cell for Ga and a
573: radio-frequency nitrogen plasma source for producing active N. Both
574: systems have a base pressure of 5$\times $10$^{-11}$~Torr. We use 6N
575: N$_{2}$ gas as a precursor which is further purified to 5~ppb by a getter
576: filter. H$_2$-etched 6H-SiC(0001) wafers produced by Cree\texttrademark\
577: were used as substrates.\cite{brandt4019apl} An \emph{in situ} Ga
578: flash-off procedure was performed in order to remove residual suboxides
579: from the SiC substrate surface prior to growth.\cite{brandt4019apl} The
580: temperatures were calibrated by visual observation of the melting point
581: of Al (660$^{\circ }$C) attached to the substrate.
582:
583: \begin{figure}[tbp]
584: \noindent \includegraphics[width=\columnwidth]{omegascan.eps}
585: \caption{Double-crystal rocking curve with open detector (black line) and
586: triple-crystal $\omega$ scan with analyzer (gray line) recorded in skew
587: geometry across the 10.5 reflection of sample 1. The symmetric 004
588: reflection from a perfect Si crystal (broken line) is shown as a measure
589: of the resolution. The intensities from the GaN layer are presented in
590: counts per second, the Si(004) signal is scaled appropriately. The right
591: panel shows the same peaks in log-log scale.}
592: \label{omegascan}
593: \end{figure}
594:
595: Sample 1 was grown under Ga-stable conditions with a substrate
596: temperature of 740$^{\circ}$C. The growth rate employed was 140~nm/h, and
597: the film was grown to a final thickness of 340~nm. Under these growth
598: conditions, we observed an entirely streaky, (1$\times$1) RHEED pattern
599: throughout growth except for the initial nucleation stage. The surface
600: morphology of the film as observed by atomic force microscopy (AFM)
601: exhibits clearly resolved monatomic steps. The root-mean-square roughness
602: amounts to 0.3~nm over an area of $1 \times 1$~$\mu$m$^2$. Samples grown
603: under these conditions typically exhibit a narrow symmetric reflection,
604: suggesting a low density of screw dislocations. In contrast, sample 2 was
605: grown under near-stoichiometric conditions with a substrate temperature
606: of 780$^{\circ}$C. The growth rate employed was 445~nm/h and the total
607: thickness of the film amounts to 1660~nm. After the initial 100~nm of GaN
608: growth, a 10~nm thick AlN film was deposited prior to overgrowth by GaN.
609: The RHEED pattern exhibited a superposition of streaks and $\vee$-shaped
610: chevrons indicative of facetting. Indeed, the AFM micrographs of this
611: sample showed the characteristic plateau-valley morphology of GaN films
612: grown with insufficient Ga flux. Compared to films grown under the same
613: conditions as sample 1, films grown under these conditions exhibit
614: narrower asymmetric reflections, indicating a reduction of the edge
615: dislocation density.
616:
617: X-ray measurements were performed with a Philips X'Pert
618: PRO\texttrademark\ four-circle triple-axis diffractometer equipped with a
619: Cu$K_{\alpha 1}$ source in the focus of a multilayer x-ray mirror and a
620: Ge(022) hybrid monochromator. The detector was kept wide open and at
621: fixed position $2\theta_B$, leading to an angular acceptance of $1^{\circ
622: }$. All asymmetric rocking curves were recorded in skew
623: geometry.\cite{sun02}
624:
625: We denote the GaN reflections in the form \textit{hk.l} which is
626: equivalent to the four-index notation \textit{hkil} for hexagonal
627: crystals with $h+k+i=0 $.
628:
629: \section{Results}
630: \label{sec:results}
631:
632: \subsection{X-ray diffraction}
633:
634: \begin{figure}[tbp]
635: \noindent \includegraphics[width=\columnwidth]{asymm_peaks.eps}
636: \caption{Double-crystal rocking curves from sample 1 obtained in skew
637: geometry for various different reflections as indicated in the figure.
638: The profiles are shown in logarithmic (a) and log-log (b) scales. The
639: experimental data are shown by gray lines. The full black lines are fits
640: of the intensity by Eq.\ (\ref{eq9}). The dotted lines for the 10.1
641: profile show a Gaussian profile.}
642: \label{asymm_peaks}
643: \end{figure}
644:
645: Figure \ref{omegascan} compares a double-crystal rocking curve measured
646: with open detector (acceptance 1$^{\circ }$) and a triple-crystal
647: $\omega$ scan measured with a three-bounce Ge(022) analyzer across the
648: 10.5 reflection for the same GaN film (sample 1). The comparison with the
649: 004 rocking curve from a perfect Si crystal shows that the GaN peak
650: broadening due to instrumental resolution can be neglected. The log-log
651: plots (right column) show that the tails of the intensity distributions
652: follow the asymptotic laws expected for scattering from dislocations,
653: $I\propto q^{-3}$ for measurements with the open detector and $I\propto
654: q^{-4}$ in the case of a collimated diffracted beam. Thus, the comparison
655: of these two profiles shows that they contain the same information about
656: lattice distortions in the film. The rocking curve measurements with open
657: detector have, however, both experimental and theoretical advantages. The
658: experimental advantage is a two orders of magnitude higher intensity (the
659: intensities are plotted in Fig.\ \ref{omegascan} in counts per second).
660: The analysis of the rocking curve is also more simple since the intensity
661: distribution is described by the one-dimensional integral (\ref{eq7}),
662: while the analysis of the scans with the analyzer requires the
663: two-dimensional integration (\ref {eq2a}) of the correlation function.
664: Therefore, we restrict ourselves to the analysis of double-crystal
665: rocking curves.
666:
667: \begin{figure}[tbp]
668: \noindent \includegraphics[width=\columnwidth]{parametersAB.eps}
669: \caption{Parameters $A$ and $B$ obtained from the fits of the
670: experimental profiles by Eq.\ (\ref{eq9}). Samples 1 and 2 are denoted by
671: full and open symbols, respectively.}
672: \label{parametersAB}
673: \end{figure}
674:
675: Figure \ref{asymm_peaks} presents skew-geometry rocking curves for
676: various reflections from sample 1. Several conclusions can be drawn just
677: from a direct inspection of the peak profiles. First, the lineshapes are
678: far from being Gaussian at the tails of the intensity distribution: see
679: the comparison of the 10.1 profile with a Gaussian fit [dotted lines in
680: Figs.\ \ref {asymm_peaks} (a,b)]. When the dynamic range is larger than
681: two orders of magnitude, it is evident that the tails of the asymmetric
682: profiles rather follow the $q^{-3}$ law. For relatively weak reflections
683: (e.~g., 10.4 or 20.4), the $-3$ exponent is not reached. Thus, the
684: profiles obey the behavior typical for a random dislocation distribution.
685: In the following, we analyze them quantitatively to obtain the
686: characteristics of the dislocation ensemble.
687:
688: \begin{figure}[tbp]
689: \noindent \includegraphics[width=\columnwidth]{symm_peaks.eps}
690: \caption{Symmetric 00.2 and 00.4 x-ray diffraction profiles from samples
691: 1 and 2. The profiles are shown in logarithmic (a) and log-log (b) scales.}
692: \label{symm_peaks}
693: \end{figure}
694:
695: The solid lines in Figs.\ \ref{asymm_peaks} (a) and (b) are the fits of
696: the measured profiles by Eq.\ (\ref{eq9}). One can see that the peak
697: profiles are adequately described. In Fig.\ \ref{parametersAB}, we plot
698: the fit parameters $A$ and $B$ as functions of $f$ and $g$, respectively,
699: since according to Eq.\ (\ref{eq10}) both dependencies are expected to be
700: linear. Figure \ref{parametersAB} (a) can be considered as a refined
701: version of the Williamson-Hall plot. The linear fit of the data in Fig.\
702: \ref{parametersAB} for sample 1 crosses the axis of the ordinates at a
703: small but non-zero value of $A$, which indicates that, in addition to
704: threading dislocations, there is an additional source of peak broadening,
705: namely, size broadening. This effect is much smaller for sample 2. From
706: the slopes of the straight lines in Fig.\ \ref{parametersAB} (a), we
707: obtain $\rho _{\mathrm{e}}b_{\mathrm{e}}^{2}=A/f=4.7\times 10^{-5}$ and
708: $3.6\times 10^{-5} $ for samples 1 and 2, respectively, where
709: $b_{\mathrm{e}}=0.32$~nm is the length of the Burgers vector of edge
710: dislocations. This result yields a density of edge threading dislocations
711: $\rho _{\mathrm{e}}=4.6\times 10^{10}$ cm$^{-2}$ for sample 1 and
712: $3.5\times 10^{10}$ cm$^{-2}$ for sample 2. The mean distances between
713: edge dislocations are $r_{d}=1/\sqrt{\rho _{\mathrm{e}}}=47$~nm (sample
714: 1) and 53~nm (sample 2). From the slopes of the lines in Fig.\
715: \ref{parametersAB} (b), we obtain $L/b_{\mathrm{e}}=B/g=230$ and $330$,
716: which give the characteristic lengths of the dislocation correlations
717: $L=74$~nm and $106$~nm, respectively. The dimensionless parameter
718: characterizing the dislocation correlations\cite{wilkens70pss} possesses
719: the values $M=L/r_{d}=1.6$ for sample 1 and 2.0 for sample 2.
720:
721: We also fitted the profiles with Eqs.~(\ref{eq13}) and (\ref{eq15}), which
722: require less computational efforts. These fits give slightly larger
723: values for parameter $A$ (dislocation density) and are not able to give
724: parameter $B$ for dislocation correlations.
725:
726: \begin{figure}[tbp]
727: \noindent \includegraphics[width=\columnwidth]{twist.eps}
728: \caption{FWHM of the rocking curves as a function of inclination angle
729: $\Psi$ for the two GaN films. The symbols are experimental data and the
730: lines are fits to them.}
731: \label{twist}
732: \end{figure}
733:
734: The symmetric x-ray diffraction profiles shown in Fig.\ \ref{symm_peaks}
735: are narrow compared to the asymmetric reflections. Furthermore, the
736: profiles of samples 1 and 2 are qualitatively different. In sample 1, the
737: intensity distributions obey a $q^{-2}$ law in the intermediate range of
738: angular deviations (and intensities) that is followed by an even lower
739: exponent for large deviations (and very low intensities). In sample 2,
740: the intensity distributions are close to the $q^{-3}$ law.
741:
742: Edge threading dislocations with dislocation lines normal to the surface
743: and Burgers vectors in the surface plane, which are the main source of
744: the peak broadening in asymmetric reflections, do not distort the planes
745: parallel to the surface. They do thus not contribute to diffraction in
746: the symmetric reflections. The $q^{-2}$ law for the 00.2 and 00.4 peaks
747: of sample 1 points to the finite thickness of the epitaxial layer as the
748: main source of the broadening. The $q^{-1}$ law at the far tails of the
749: peaks are due to another source, possibly thermal diffuse scattering. We
750: did not investigate this part of the symmetric reflections since we
751: suppose that it is not related to threading dislocations which are the
752: topic of the present study. The profiles of sample 2, which is three
753: times thicker than sample 1, obey the $q^{-3}$ law indicating that the
754: broadening is primarily by dislocations.
755:
756: The solid lines in Fig.\ \ref{symm_peaks} are the fits of the
757: experimental curves to Eq.\ (\ref{eq17}), where the different thicknesses
758: of the layers are explicitly taken into account. We indeed obtain the
759: $q^{-2}$ intensity decay on the tails of the peak for sample 1 and the
760: $q^{-3}$ decay for sample 2. From the fit parameters, we obtain $\rho
761: _{\mathrm{s}}b_ {\mathrm{s}}^{2}=A/f=2.4\times 10^{-6}$ and $4.5\times
762: 10^{-6}$ for samples 1 and 2, respectively, where
763: $b_{\mathrm{s}}=0.52~$nm is the length of the Burgers vector of screw
764: dislocations. This results yields the screw threading dislocations
765: densities of $\rho _{\mathrm{s}}=9\times 10^{8}$ cm$^{-2}$ for sample 1
766: and $1.7\times 10^{9}$ cm$^{-2}$ for sample 2. From the values of the
767: parameter $B$ we obtain $L/b_{\mathrm{s}}=B/g=650$ and $440$, which
768: result in characteristic lengths of the dislocation correlations
769: $L=340$~nm and $230$~nm for sample 1 and 2, respectively. The parameter
770: $M$ is close to 1 for both samples. Note, however, that screw
771: dislocations are not the only source of the peak broadening in symmetric
772: reflections (see discussion below in Sec. \ref{sec:discuss}).
773:
774: \subsection{TEM}
775:
776: TEM is the method of choice to directly determine the character of
777: dislocations and their distribution in thin films. The $\mathbf{g \cdot
778: b}$ criterion in TEM is generally applied for the two-beam condition to
779: evaluate the alignment of the strain field of the dislocation with
780: Burgers vector $\mathbf{b}$ with respect to the diffraction vector
781: $\mathbf{g}$ producing the image contrast. This criterion is strictly
782: correct for screw dislocations and for edge dislocations only if their
783: line direction $\mathbf{l}$ and Burgers vector are in the imaging plane,
784: i.~e., if $\mathbf{g \cdot (b \times l)}$ is considered. In the present
785: case of GaN(0001) films it is well established that three types of
786: threading dislocations exist, having Burgers vectors $\frac{1}{3}
787: \left\langle 11\overline{2}0\right\rangle$, $\left\langle 0 0 0
788: 1\right\rangle$ and $\frac{1}{3} \left\langle
789: 11\overline{2}3\right\rangle$ representing edge, screw and mixed
790: dislocations, respectively, under the assumption that the dislocations
791: lines lie parallel to the $c$-axis. Mixed dislocations are not observed in
792: our samples.
793:
794: \begin{figure}[t!]
795: \noindent \includegraphics[width=\columnwidth]{tem.eps}
796: \caption{Cross-sectional (a) and plan-view (b) TEM images of sample 2.
797: The outcrops of screw dislocations are marked in (b) by arrows.}
798: \label{TEM}
799: \end{figure}
800:
801: In order to identify the Burgers vector, two images have to be recorded
802: with $\mathbf{g}$ parallel and perpendicular to the $c$-axis. The screw
803: dislocations are thus imaged if $\mathbf{g}$=[0002], an example of which
804: is shown in Fig.\ \ref{TEM}(a) for sample 2. From this image, we can
805: directly measure the dislocation density if we know the TEM specimen
806: thickness that is determined by tilting the interface from the end-on to
807: a well-defined inclined position. Edge dislocations appear in the
808: cross-sectional images if $\mathbf{g}$ is perpendicular to the $c$-axis,
809: e.~g., if $\mathbf{g}=[11\overline{2}0]$, and we are then able to measure
810: their density in the same way. Furthermore, plan-view TEM imaging is
811: applied to complement the measurement of the dislocation density. Figure
812: \ref{TEM}(b) shows a plan-view TEM image of sample 2, where the specimen
813: is tilted a few degrees off the [0001] zone axis to obtain two-beam
814: conditions with $\mathbf{g}=[11\overline{2}0]$ in order to bring the edge
815: dislocations in contrast. The edge threading dislocations are extended
816: along the film normal and homogeneously distributed. It is remarkable
817: that screw dislocations can be identified as well (marked by arrows),
818: although the contrast should vanish because of $\mathbf{g \cdot b}=0$.
819: However, this specific contrast is produced due to strain relaxation of
820: the screw dislocation at the free surfaces varying locally the lattice
821: plane distortions created by the strain field. Table \ref{table} compares
822: the densities of edge and screw dislocations obtained from the x-ray
823: diffraction and TEM measurements, revealing that the density of edge
824: dislocations is in fact slightly higher in sample 1 compared to sample 2.
825: The screw dislocation density in sample 1 is too low for reliable TEM
826: determination. Some of the screw dislocations are seen as hexagonal pits
827: in AFM micrographs. A lower limit of 5$\times$10$^7$~cm$^{-2}$ is thus
828: obtained for the screw dislocation density in sample 1.
829:
830: \begin{table}[tb]
831: \caption{Edge and screw dislocation densities determined by x-ray
832: diffraction and TEM.}
833: \label{table}
834: \begin{ruledtabular}
835: \begin{tabular}{lcllll}
836: &thickness& \multicolumn{2}{c}{$\rho _{\mathrm{e}}$ ($10^{10}$ cm$^{-2}$)}
837: &\multicolumn{2}{c}{$\rho _{\mathrm{s}}$ ($10^{ 9}$ cm$^{-2}$)} \\
838: &(nm)& X-ray & TEM & X-ray & TEM \\
839: \hline
840: sample 1 & 340 & 4.6 & 3.0$\pm$0.5 & 0.9 & --- \\
841: sample 2 & 1660 & 3.5 & 2.0$\pm$0.5 & 1.7 & 1.2 \\
842: \end{tabular}
843: \end{ruledtabular}
844: \end{table}
845:
846: \section{Discussion}
847:
848: \label{sec:discuss}
849:
850: The x-ray diffraction studies of GaN have used, up to now, only the
851: widths of diffraction peaks. Two quantities can be obtained from a series
852: of the reflections.\cite{sun02} One is the FWHM of the symmetric
853: reflection $\Delta \omega _{\mathrm{s}}$ which is influenced by screw
854: dislocations but insensitive to edge dislocations. The other quantity is
855: obtained by extrapolation of the FWHM of skew reflections to the limiting
856: case of grazing incidence/grazing exit diffraction, with both incident
857: and diffracted beams lying in the surface plane. This quantity ($\Delta
858: \omega _{\mathrm{e}}$) is sensitive to edge but insensitive to screw
859: dislocations. Figure \ref{twist} presents the widths of the diffraction
860: profiles of samples 1 and 2 as a function of the inclination angle
861: $\Psi$. The lines are fits by the model described in Ref.\
862: \onlinecite{sun02} to obtain $\Delta \omega _{\mathrm{s}}$ and $\Delta
863: \omega _{\mathrm{e}}$. While these quantities are determined
864: reliably and accurately, the question arises how they are related to the
865: actual dislocation densities. Most
866: commonly,\cite{metzger1013pma,chierchia03} the formulas initially
867: proposed in Refs.\ \onlinecite{gay53} and \onlinecite{dunn57} for a mosaic
868: crystal are used for this purpose:
869: \begin{equation}
870: \rho _{\mathrm{e}}=\frac{\Delta \omega _{\mathrm{e}}^{2}}{4.35b_
871: {\mathrm{e}}^{2}},\qquad \rho _{\mathrm{s}}=\frac{\Delta
872: \omega _{\mathrm{s}}^{2}} {4.35b_{\mathrm{s}}^{2}},
873: \label{eq18}
874: \end{equation}
875: with the coefficient $2\pi \ln 2\approx 4.35.$ In this context, the
876: quantities $\Delta \omega _{\mathrm{s}}$ and $\Delta \omega
877: _{\mathrm{e}}$ are often referred to as tilt and twist, respectively.
878:
879: Equations (\ref{eq18}) are valid for mosaic crystals, where the terms
880: tilt and twist describe different modes of relative block misorientation.
881: These terms loose their meaning when applied to a layer with randomly
882: distributed dislocations, since the broadening is then determined by a
883: complicated combination of misorientation of the lattice planes and
884: change of interplanar distances (strain). Hence, equations (\ref{eq18})
885: use an inappropriate approach and are not strictly valid in the case of a
886: random dislocation distribution. It remains correct that the dislocation
887: density is proportional to the square of the peak width and inversely
888: proportional to the square of the relevant Burgers vector, but the
889: coefficients need to be reconsidered.
890:
891: The relation between the dislocation density and the FWHM $\Delta \omega $
892: of the intensity distribution (\ref{eq9}) can be written, by making use of
893: Eqs.\ (\ref{eq10}) and (\ref{eq13a}), in a form similar to Eq. (\ref{eq18}):
894: \begin{equation}
895: \rho \approx \frac{\Delta \omega ^{2}}{[2.4+\ln (g\sqrt{f}M)]^{2}fb^{2}}.
896: \label{eq19}
897: \end{equation}
898: The dimensionless parameter $M=L\sqrt{\rho }$ that characterizes
899: dislocation correlations is slightly larger than 1 for the GaN films
900: studied here, which is an indication of a strong screening of the
901: long-range strain fields of dislocations by the neighboring dislocations.
902: One has $M\gg 1$ for uncorrelated dislocations. In this latter case, the
903: logarithmic term in Eq.\ (\ref{eq19}) is the one obtained by
904: Krivoglaz.\cite {krivoglaz63,krivoglaz:book}
905:
906: We can now simplify Eq. (\ref{eq19}) for the limiting cases of the
907: large-inclination skew diffraction and for symmetric Bragg diffraction by
908: using the expressions for the parameters $f$ and $g$ obtained in Sec.\
909: \ref {sec:theory} for these cases:
910: \begin{equation}
911: \rho _{\mathrm{e}}\approx \frac{4 \pi \Delta \omega _{\mathrm{e}}^{2}\cos
912: ^{2}\theta _{B}}{(3.0+\ln M)^2 b_{\mathrm{e}}^{2}},
913: \qquad \rho _{\mathrm{s}}\approx
914: \frac{8 \pi \Delta \omega _{\mathrm{s}}^{2}}{(2.6+\ln M)^2
915: b_{\mathrm{s}}^{2}}.
916: \label{eq20}
917: \end{equation}
918: The term $\ln M$ describing the range of dislocation correlations cannot
919: be neglected even when $M$ varies between 1 and 2, as it happens to be
920: the case for the samples studied in the present work. Compared to Eqs.\
921: (\ref{eq18}), Eqs.\ (\ref{eq20}) result in a four times higher edge
922: dislocation density and an order of magnitude higher screw dislocation
923: density and give dislocation densities that are in agreement with the TEM
924: results, see Table \ref{table}.
925:
926: There are at least three types of dislocations that may contribute to the
927: broadening of symmetric reflections. First, screw threading dislocations
928: distort the lattice planes parallel to the surface. Secondly, edge
929: threading dislocations can contribute to these reflections if the
930: dislocation lines deviate from the layer normal. Even if such deviations
931: are small, the effect cannot be neglected since the density of edge
932: threading dislocations is much larger than the density of screw threading
933: dislocations. Third, misfit dislocations at the layer-substrate interface
934: contribute to broadening of the symmetric reflections. Although misfit
935: dislocation lines are far from the surface, their non-uniform distortions
936: penetrate through the whole epitaxial layer.\cite {kaganer97} Finally,
937: the stress relaxation at the free surface gives rise to additional
938: strains around the outcrops of the edge threading dislocations,
939: contributing to an additional broadening to the symmetric reflections.
940: Thus, Eq.~(\ref{eq9}) and, particularly, Eqs.~(\ref{eq20}) provide an
941: upper estimate of the screw threading dislocation density since they are
942: assumed to be the only source of broadening for symmetric reflections.
943:
944: \section{Conclusions}
945:
946: The width of either symmetric or asymmetric reflection can be used as a
947: figure-of-merit for the dislocation density only if the dislocation
948: distribution is the same in all the samples to be compared. Even for
949: films having a spatially random distribution of dislocations, the width
950: of a given reflection depends not only on the dislocation density, but
951: also on the range of correlations in the restricted random dislocation
952: distribution. The lineshape analysis of the diffraction profile as
953: presented in this work returns the width as well as the correlation
954: range, and is thus a far more reliable approach for estimating the
955: dislocation density than a simple consideration of the width alone.
956:
957: The lineshape analysis has shown that the x-ray diffraction profiles of
958: the GaN films under investigation are Gaussian only in the central part
959: of the peak. The tails of the peak follow the power laws characteristic
960: to x-ray diffraction of crystals with randomly distributed dislocations.
961: The double-crystal rocking curves measured with a wide open detector
962: follow a $q^{-3}$ behavior, while the triple-crystal rocking curves with
963: an analyzer crystal obey a $q^{-4}$ behavior. The study of the
964: double-crystal rocking curves is more simple both experimentally and
965: theoretically, since the diffracted intensity is larger and the peak
966: profile is described by a one-dimensional Fourier transform of the pair
967: correlation function (\ref{eq9}).
968:
969: The $q^{-3}$ tails of the diffraction profiles are insensitive to
970: correlations between dislocations and allow a more reliable determination
971: of the dislocation densities. The entire diffraction profiles are
972: adequately fitted by Eq.\ (\ref{eq9}). The fits provide two parameters
973: characterizing the dislocation ensemble, the mean dislocation density
974: $\rho $ and the screening range $L$. The latter quantity corresponds to a
975: mean size of the cells with the total Burgers vector equal to zero. We
976: find that, for edge threading dislocations in GaN layers, $L$ is only
977: slightly larger than the mean distance between dislocations $\rho
978: ^{-1/2}$.
979:
980: %\bibliographystyle{apsrev}
981: %\bibliography{surface,twist}
982:
983: \begin{thebibliography}{41}
984: \expandafter\ifx\csname natexlab\endcsname\relax\def\natexlab#1{#1}\fi
985: \expandafter\ifx\csname bibnamefont\endcsname\relax
986: \def\bibnamefont#1{#1}\fi
987: \expandafter\ifx\csname bibfnamefont\endcsname\relax
988: \def\bibfnamefont#1{#1}\fi
989: \expandafter\ifx\csname citenamefont\endcsname\relax
990: \def\citenamefont#1{#1}\fi
991: \expandafter\ifx\csname url\endcsname\relax
992: \def\url#1{\texttt{#1}}\fi
993: \expandafter\ifx\csname urlprefix\endcsname\relax\def\urlprefix{URL }\fi
994: \providecommand{\bibinfo}[2]{#2}
995: \providecommand{\eprint}[2][]{\url{#2}}
996:
997: \bibitem[{\citenamefont{Gil}(1998)}]{gilbook}
998: \bibinfo{editor}{\bibfnamefont{B.}~\bibnamefont{Gil}}, ed.,
999: \emph{\bibinfo{title}{Group III Nitride Semiconductor Compounds: Physical and
1000: Applications}} (\bibinfo{publisher}{Oxford Univ. Press},
1001: \bibinfo{address}{New York}, \bibinfo{year}{1998}).
1002:
1003: \bibitem[{\citenamefont{Srikant et~al.}(1997)\citenamefont{Srikant, Speck, and
1004: Clarke}}]{srikant4286jap}
1005: \bibinfo{author}{\bibfnamefont{V.}~\bibnamefont{Srikant}},
1006: \bibinfo{author}{\bibfnamefont{J.~S.} \bibnamefont{Speck}}, \bibnamefont{and}
1007: \bibinfo{author}{\bibfnamefont{D.~R.} \bibnamefont{Clarke}},
1008: \bibinfo{journal}{J. Appl. Phys.} \textbf{\bibinfo{volume}{82}},
1009: \bibinfo{pages}{4286} (\bibinfo{year}{1997}).
1010:
1011: \bibitem[{\citenamefont{Metzger et~al.}(1998)\citenamefont{Metzger, H\"opler,
1012: Born, Ambacher, Stutzmann, St\"ommer, Schuster, G\"obel, Christiansen,
1013: Albrecht et~al.}}]{metzger1013pma}
1014: \bibinfo{author}{\bibfnamefont{T.}~\bibnamefont{Metzger}},
1015: \bibinfo{author}{\bibfnamefont{R.}~\bibnamefont{H\"opler}},
1016: \bibinfo{author}{\bibfnamefont{E.}~\bibnamefont{Born}},
1017: \bibinfo{author}{\bibfnamefont{O.}~\bibnamefont{Ambacher}},
1018: \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{Stutzmann}},
1019: \bibinfo{author}{\bibfnamefont{R.}~\bibnamefont{St\"ommer}},
1020: \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{Schuster}},
1021: \bibinfo{author}{\bibfnamefont{H.}~\bibnamefont{G\"obel}},
1022: \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{Christiansen}},
1023: \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{Albrecht}},
1024: \bibnamefont{et~al.}, \bibinfo{journal}{Philos. Mag. A}
1025: \textbf{\bibinfo{volume}{77}}, \bibinfo{pages}{1013} (\bibinfo{year}{1998}).
1026:
1027: \bibitem[{\citenamefont{Heinke et~al.}(2000)\citenamefont{Heinke, Kirchner,
1028: Einfeldt, and Hommel}}]{heinke2145apl}
1029: \bibinfo{author}{\bibfnamefont{H.}~\bibnamefont{Heinke}},
1030: \bibinfo{author}{\bibfnamefont{V.}~\bibnamefont{Kirchner}},
1031: \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{Einfeldt}}, \bibnamefont{and}
1032: \bibinfo{author}{\bibfnamefont{D.}~\bibnamefont{Hommel}},
1033: \bibinfo{journal}{Appl. Phys. Lett.} \textbf{\bibinfo{volume}{77}},
1034: \bibinfo{pages}{2145} (\bibinfo{year}{2000}).
1035:
1036: \bibitem[{\citenamefont{Sun et~al.}(2002)\citenamefont{Sun, Brandt, Liu,
1037: Trampert, and Ploog}}]{sun02}
1038: \bibinfo{author}{\bibfnamefont{Y.~J.} \bibnamefont{Sun}},
1039: \bibinfo{author}{\bibfnamefont{O.}~\bibnamefont{Brandt}},
1040: \bibinfo{author}{\bibfnamefont{T.~Y.} \bibnamefont{Liu}},
1041: \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{Trampert}}, \bibnamefont{and}
1042: \bibinfo{author}{\bibfnamefont{K.~H.} \bibnamefont{Ploog}},
1043: \bibinfo{journal}{Appl. Phys. Lett.} \textbf{\bibinfo{volume}{81}},
1044: \bibinfo{pages}{4928} (\bibinfo{year}{2002}).
1045:
1046: \bibitem[{\citenamefont{Chierchia et~al.}(2003)\citenamefont{Chierchia,
1047: B\"ottcher, Heinke, Einfeldt, Figge, and Hommel}}]{chierchia03}
1048: \bibinfo{author}{\bibfnamefont{R.}~\bibnamefont{Chierchia}},
1049: \bibinfo{author}{\bibfnamefont{T.}~\bibnamefont{B\"ottcher}},
1050: \bibinfo{author}{\bibfnamefont{H.}~\bibnamefont{Heinke}},
1051: \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{Einfeldt}},
1052: \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{Figge}}, \bibnamefont{and}
1053: \bibinfo{author}{\bibfnamefont{D.}~\bibnamefont{Hommel}},
1054: \bibinfo{journal}{J. Appl. Phys.} \textbf{\bibinfo{volume}{93}},
1055: \bibinfo{pages}{8918} (\bibinfo{year}{2003}).
1056:
1057: \bibitem[{\citenamefont{Ratnikov et~al.}(2000)\citenamefont{Ratnikov, Kyutt,
1058: Shubina, Paskova, Valcheva, and Monemar}}]{ratnikov00}
1059: \bibinfo{author}{\bibfnamefont{V.}~\bibnamefont{Ratnikov}},
1060: \bibinfo{author}{\bibfnamefont{R.}~\bibnamefont{Kyutt}},
1061: \bibinfo{author}{\bibfnamefont{T.}~\bibnamefont{Shubina}},
1062: \bibinfo{author}{\bibfnamefont{T.}~\bibnamefont{Paskova}},
1063: \bibinfo{author}{\bibfnamefont{E.}~\bibnamefont{Valcheva}}, \bibnamefont{and}
1064: \bibinfo{author}{\bibfnamefont{B.}~\bibnamefont{Monemar}},
1065: \bibinfo{journal}{J. Appl. Phys.} \textbf{\bibinfo{volume}{88}},
1066: \bibinfo{pages}{6252} (\bibinfo{year}{2000}).
1067:
1068: \bibitem[{\citenamefont{Munkholm et~al.}(1998)\citenamefont{Munkholm, Thompson,
1069: Foster, Eastman, Auciello, Fini, DenBaars, and Speck}}]{munkholm2972apl}
1070: \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{Munkholm}},
1071: \bibinfo{author}{\bibfnamefont{C.}~\bibnamefont{Thompson}},
1072: \bibinfo{author}{\bibfnamefont{C.~M.} \bibnamefont{Foster}},
1073: \bibinfo{author}{\bibfnamefont{J.~A.} \bibnamefont{Eastman}},
1074: \bibinfo{author}{\bibfnamefont{O.}~\bibnamefont{Auciello}},
1075: \bibinfo{author}{\bibfnamefont{P.}~\bibnamefont{Fini}},
1076: \bibinfo{author}{\bibfnamefont{S.~P.} \bibnamefont{DenBaars}},
1077: \bibnamefont{and} \bibinfo{author}{\bibfnamefont{J.~S.} \bibnamefont{Speck}},
1078: \bibinfo{journal}{Appl. Phys. Lett.} \textbf{\bibinfo{volume}{72}},
1079: \bibinfo{pages}{2972} (\bibinfo{year}{1998}).
1080:
1081: \bibitem[{\citenamefont{Yang et~al.}(2000)\citenamefont{Yang, Wu, Lee, and
1082: Chi}}]{yang4240jap}
1083: \bibinfo{author}{\bibfnamefont{C.~C.} \bibnamefont{Yang}},
1084: \bibinfo{author}{\bibfnamefont{M.~C.} \bibnamefont{Wu}},
1085: \bibinfo{author}{\bibfnamefont{C.~H.} \bibnamefont{Lee}}, \bibnamefont{and}
1086: \bibinfo{author}{\bibfnamefont{G.~C.} \bibnamefont{Chi}},
1087: \bibinfo{journal}{J. Appl. Phys.} \textbf{\bibinfo{volume}{87}},
1088: \bibinfo{pages}{2000} (\bibinfo{year}{2000}).
1089:
1090: \bibitem[{\citenamefont{Busing and Levy}(1967)}]{busing457ac}
1091: \bibinfo{author}{\bibfnamefont{W.~R.} \bibnamefont{Busing}} \bibnamefont{and}
1092: \bibinfo{author}{\bibfnamefont{H.~A.} \bibnamefont{Levy}},
1093: \bibinfo{journal}{Acta Cryst.} \textbf{\bibinfo{volume}{22}},
1094: \bibinfo{pages}{457} (\bibinfo{year}{1967}).
1095:
1096: \bibitem[{\citenamefont{Warren}(1969)}]{warren:book69}
1097: \bibinfo{author}{\bibfnamefont{B.~E.} \bibnamefont{Warren}},
1098: \emph{\bibinfo{title}{X-Ray Diffraction}}
1099: (\bibinfo{publisher}{Addison-Wesley}, \bibinfo{address}{Reading, Mass.},
1100: \bibinfo{year}{1969}).
1101:
1102: \bibitem[{\citenamefont{Krivoglaz}(1996)}]{krivoglaz:book}
1103: \bibinfo{author}{\bibfnamefont{M.~A.} \bibnamefont{Krivoglaz}},
1104: \emph{\bibinfo{title}{X-Ray and Neutron Diffraction in Nonideal Crystals}}
1105: (\bibinfo{publisher}{Springer}, \bibinfo{address}{Berlin},
1106: \bibinfo{year}{1996}).
1107:
1108: \bibitem[{\citenamefont{Williamson and Hall}(1953)}]{WilliamsonHall53}
1109: \bibinfo{author}{\bibfnamefont{G.~K.} \bibnamefont{Williamson}}
1110: \bibnamefont{and} \bibinfo{author}{\bibfnamefont{W.~H.} \bibnamefont{Hall}},
1111: \bibinfo{journal}{Acta Metall.} \textbf{\bibinfo{volume}{1}},
1112: \bibinfo{pages}{22} (\bibinfo{year}{1953}).
1113:
1114: \bibitem[{\citenamefont{Hordon and Averbach}(1961)}]{hordon61}
1115: \bibinfo{author}{\bibfnamefont{M.~J.} \bibnamefont{Hordon}} \bibnamefont{and}
1116: \bibinfo{author}{\bibfnamefont{B.~L.} \bibnamefont{Averbach}},
1117: \bibinfo{journal}{Acta Metall.} \textbf{\bibinfo{volume}{9}},
1118: \bibinfo{pages}{237} (\bibinfo{year}{1961}).
1119:
1120: \bibitem[{\citenamefont{Ayers}(1994)}]{ayers94}
1121: \bibinfo{author}{\bibfnamefont{J.~E.} \bibnamefont{Ayers}},
1122: \bibinfo{journal}{J. Cryst. Growth} \textbf{\bibinfo{volume}{135}},
1123: \bibinfo{pages}{71} (\bibinfo{year}{1994}).
1124:
1125: \bibitem[{\citenamefont{Ung\'ar and Borb\'ely}(1996)}]{ungar96}
1126: \bibinfo{author}{\bibfnamefont{T.}~\bibnamefont{Ung\'ar}} \bibnamefont{and}
1127: \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{Borb\'ely}},
1128: \bibinfo{journal}{Appl. Phys. Lett.} \textbf{\bibinfo{volume}{69}},
1129: \bibinfo{pages}{3173} (\bibinfo{year}{1996}).
1130:
1131: \bibitem[{\citenamefont{Ung\'ar and Tichy}(1999)}]{ungar99}
1132: \bibinfo{author}{\bibfnamefont{T.}~\bibnamefont{Ung\'ar}} \bibnamefont{and}
1133: \bibinfo{author}{\bibfnamefont{G.}~\bibnamefont{Tichy}},
1134: \bibinfo{journal}{Phys. Stat. Sol. (a)} \textbf{\bibinfo{volume}{171}},
1135: \bibinfo{pages}{425} (\bibinfo{year}{1999}).
1136:
1137: \bibitem[{\citenamefont{Ung\'ar et~al.}(2001)\citenamefont{Ung\'ar, Gubicza,
1138: Rib\'arik, and Borb\'ely}}]{ungar01}
1139: \bibinfo{author}{\bibfnamefont{T.}~\bibnamefont{Ung\'ar}},
1140: \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{Gubicza}},
1141: \bibinfo{author}{\bibfnamefont{G.}~\bibnamefont{Rib\'arik}},
1142: \bibnamefont{and}
1143: \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{Borb\'ely}},
1144: \bibinfo{journal}{J. Appl. Cryst.} \textbf{\bibinfo{volume}{34}},
1145: \bibinfo{pages}{298} (\bibinfo{year}{2001}).
1146:
1147: \bibitem[{\citenamefont{Warren and Averbach}(1950)}]{WarrenAverbach50}
1148: \bibinfo{author}{\bibfnamefont{B.~E.} \bibnamefont{Warren}} \bibnamefont{and}
1149: \bibinfo{author}{\bibfnamefont{B.~L.} \bibnamefont{Averbach}},
1150: \bibinfo{journal}{J. Appl. Phys.} \textbf{\bibinfo{volume}{21}},
1151: \bibinfo{pages}{595} (\bibinfo{year}{1950}).
1152:
1153: \bibitem[{\citenamefont{Warren and Averbach}(1952)}]{WarrenAverbach52}
1154: \bibinfo{author}{\bibfnamefont{B.~E.} \bibnamefont{Warren}} \bibnamefont{and}
1155: \bibinfo{author}{\bibfnamefont{B.~L.} \bibnamefont{Averbach}},
1156: \bibinfo{journal}{J. Appl. Phys.} \textbf{\bibinfo{volume}{23}},
1157: \bibinfo{pages}{497} (\bibinfo{year}{1952}).
1158:
1159: \bibitem[{\citenamefont{Balzar}(1992)}]{balzar92}
1160: \bibinfo{author}{\bibfnamefont{D.}~\bibnamefont{Balzar}}, \bibinfo{journal}{J.
1161: Appl. Cryst.} \textbf{\bibinfo{volume}{25}}, \bibinfo{pages}{559}
1162: (\bibinfo{year}{1992}).
1163:
1164: \bibitem[{\citenamefont{Balzar and Ledbetter}(1993)}]{balzar93}
1165: \bibinfo{author}{\bibfnamefont{D.}~\bibnamefont{Balzar}} \bibnamefont{and}
1166: \bibinfo{author}{\bibfnamefont{H.}~\bibnamefont{Ledbetter}},
1167: \bibinfo{journal}{J. App. Cryst.} \textbf{\bibinfo{volume}{26}},
1168: \bibinfo{pages}{97} (\bibinfo{year}{1993}).
1169:
1170: \bibitem[{\citenamefont{Young and Wiles}(1982)}]{young82}
1171: \bibinfo{author}{\bibfnamefont{R.~A.} \bibnamefont{Young}} \bibnamefont{and}
1172: \bibinfo{author}{\bibfnamefont{D.~B.} \bibnamefont{Wiles}},
1173: \bibinfo{journal}{J. Appl. Cryst.} \textbf{\bibinfo{volume}{15}},
1174: \bibinfo{pages}{430} (\bibinfo{year}{1982}).
1175:
1176: \bibitem[{\citenamefont{{Th. H. de Keijser} et~al.}(1983)\citenamefont{{Th. H.
1177: de Keijser}, Mittemeijer, and Rozendaal}}]{keijser83}
1178: \bibinfo{author}{\bibnamefont{{Th. H. de Keijser}}},
1179: \bibinfo{author}{\bibfnamefont{E.~J.} \bibnamefont{Mittemeijer}},
1180: \bibnamefont{and} \bibinfo{author}{\bibfnamefont{H.~C.~F.}
1181: \bibnamefont{Rozendaal}}, \bibinfo{journal}{J. Appl. Cryst.}
1182: \textbf{\bibinfo{volume}{16}}, \bibinfo{pages}{309} (\bibinfo{year}{1983}).
1183:
1184: \bibitem[{\citenamefont{Krivoglaz and Ryboshapka}(1963)}]{krivoglaz63}
1185: \bibinfo{author}{\bibfnamefont{M.~A.} \bibnamefont{Krivoglaz}}
1186: \bibnamefont{and} \bibinfo{author}{\bibfnamefont{K.~P.}
1187: \bibnamefont{Ryboshapka}}, \bibinfo{journal}{Fiz. Met. Metalloved.}
1188: \textbf{\bibinfo{volume}{15}}, \bibinfo{pages}{18} (\bibinfo{year}{1963}),
1189: \bibinfo{note}{[Phys. Met. Metallogr. {\bf 15}, 14 (1963)]}.
1190:
1191: \bibitem[{\citenamefont{Wilkens}(1969)}]{wilkens69}
1192: \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{Wilkens}},
1193: \bibinfo{journal}{Acta Metall.} \textbf{\bibinfo{volume}{17}},
1194: \bibinfo{pages}{1155} (\bibinfo{year}{1969}).
1195:
1196: \bibitem[{\citenamefont{Wilkens}(1970{\natexlab{a}})}]{wilkens70nbs}
1197: \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{Wilkens}},
1198: \emph{\bibinfo{title}{Fundamental Aspects of Dislocation Theory}}
1199: (\bibinfo{publisher}{Nat. Bur. Stand. (U.S.) Spec. Publ.},
1200: \bibinfo{address}{Washington, D.C.}, \bibinfo{year}{1970}{\natexlab{a}}), p.
1201: \bibinfo{pages}{1195}.
1202:
1203: \bibitem[{\citenamefont{Wilkens}(1970{\natexlab{b}})}]{wilkens70pss}
1204: \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{Wilkens}},
1205: \bibinfo{journal}{Phys. Stat. Sol. (a)} \textbf{\bibinfo{volume}{2}},
1206: \bibinfo{pages}{359} (\bibinfo{year}{1970}{\natexlab{b}}).
1207:
1208: \bibitem[{\citenamefont{Krivoglaz et~al.}(1983)\citenamefont{Krivoglaz,
1209: Martynenko, and Ryboshapka}}]{krivoglaz83}
1210: \bibinfo{author}{\bibfnamefont{M.~A.} \bibnamefont{Krivoglaz}},
1211: \bibinfo{author}{\bibfnamefont{O.~V.} \bibnamefont{Martynenko}},
1212: \bibnamefont{and} \bibinfo{author}{\bibfnamefont{K.~P.}
1213: \bibnamefont{Ryboshapka}}, \bibinfo{journal}{Fiz. Met. Metalloved.}
1214: \textbf{\bibinfo{volume}{55}}, \bibinfo{pages}{5} (\bibinfo{year}{1983}).
1215:
1216: \bibitem[{\citenamefont{Kosterlitz and Thouless}(1972)}]{KosterlitzThouless72}
1217: \bibinfo{author}{\bibfnamefont{J.~M.} \bibnamefont{Kosterlitz}}
1218: \bibnamefont{and} \bibinfo{author}{\bibfnamefont{D.~J.}
1219: \bibnamefont{Thouless}}, \bibinfo{journal}{J. Phys. C: Solid State Phys.}
1220: \textbf{\bibinfo{volume}{5}}, \bibinfo{pages}{L124} (\bibinfo{year}{1972}).
1221:
1222: \bibitem[{\citenamefont{Kosterlitz and Thouless}(1973)}]{KosterlitzThouless73}
1223: \bibinfo{author}{\bibfnamefont{J.~M.} \bibnamefont{Kosterlitz}}
1224: \bibnamefont{and} \bibinfo{author}{\bibfnamefont{D.~J.}
1225: \bibnamefont{Thouless}}, \bibinfo{journal}{J. Phys. C: Solid State Phys.}
1226: \textbf{\bibinfo{volume}{6}}, \bibinfo{pages}{1181} (\bibinfo{year}{1973}).
1227:
1228: \bibitem[{\citenamefont{Peterson and Kaganer}(1994)}]{kaganer94prl}
1229: \bibinfo{author}{\bibfnamefont{I.~R.} \bibnamefont{Peterson}} \bibnamefont{and}
1230: \bibinfo{author}{\bibfnamefont{V.~M.} \bibnamefont{Kaganer}},
1231: \bibinfo{journal}{Phys. Rev. Lett.} \textbf{\bibinfo{volume}{73}},
1232: \bibinfo{pages}{102} (\bibinfo{year}{1994}).
1233:
1234: \bibitem[{\citenamefont{Groma}(1998)}]{groma98}
1235: \bibinfo{author}{\bibfnamefont{I.}~\bibnamefont{Groma}},
1236: \bibinfo{journal}{Phys. Rev. B} \textbf{\bibinfo{volume}{57}},
1237: \bibinfo{pages}{7535} (\bibinfo{year}{1998}).
1238:
1239: \bibitem[{\citenamefont{Wilkens}(1963)}]{wilkens63}
1240: \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{Wilkens}},
1241: \bibinfo{journal}{Phys. Stat. Sol.} \textbf{\bibinfo{volume}{3}},
1242: \bibinfo{pages}{1718} (\bibinfo{year}{1963}).
1243:
1244: \bibitem[{\citenamefont{Groma}(2000)}]{groma00}
1245: \bibinfo{author}{\bibfnamefont{I.}~\bibnamefont{Groma}}, \bibinfo{journal}{J.
1246: Appl. Cryst.} \textbf{\bibinfo{volume}{33}}, \bibinfo{pages}{1329}
1247: (\bibinfo{year}{2000}).
1248:
1249: \bibitem[{\citenamefont{Borb\'ely and Groma}(2001)}]{borbely01}
1250: \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{Borb\'ely}} \bibnamefont{and}
1251: \bibinfo{author}{\bibfnamefont{I.}~\bibnamefont{Groma}},
1252: \bibinfo{journal}{Appl. Phys. Lett.} \textbf{\bibinfo{volume}{79}},
1253: \bibinfo{pages}{1772} (\bibinfo{year}{2001}).
1254:
1255: \bibitem[{\citenamefont{Kaganer et~al.}(1997)\citenamefont{Kaganer, K\"ohler,
1256: Schmidbauer, Opitz, and Jenichen}}]{kaganer97}
1257: \bibinfo{author}{\bibfnamefont{V.~M.} \bibnamefont{Kaganer}},
1258: \bibinfo{author}{\bibfnamefont{R.}~\bibnamefont{K\"ohler}},
1259: \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{Schmidbauer}},
1260: \bibinfo{author}{\bibfnamefont{R.}~\bibnamefont{Opitz}}, \bibnamefont{and}
1261: \bibinfo{author}{\bibfnamefont{B.}~\bibnamefont{Jenichen}},
1262: \bibinfo{journal}{Phys. Rev. B} \textbf{\bibinfo{volume}{55}},
1263: \bibinfo{pages}{1793} (\bibinfo{year}{1997}).
1264:
1265: \bibitem[{\citenamefont{Krivoglaz}(1961)}]{krivoglaz61}
1266: \bibinfo{author}{\bibfnamefont{M.~A.} \bibnamefont{Krivoglaz}},
1267: \bibinfo{journal}{Fiz. Met. Metalloved.} \textbf{\bibinfo{volume}{12}},
1268: \bibinfo{pages}{465} (\bibinfo{year}{1961}).
1269:
1270: \bibitem[{\citenamefont{Brandt et~al.}(1999)\citenamefont{Brandt, Muralidharan,
1271: Waltereit, Thamm, Trampert, von Kiedrowski, and Ploog}}]{brandt4019apl}
1272: \bibinfo{author}{\bibfnamefont{O.}~\bibnamefont{Brandt}},
1273: \bibinfo{author}{\bibfnamefont{R.}~\bibnamefont{Muralidharan}},
1274: \bibinfo{author}{\bibfnamefont{P.}~\bibnamefont{Waltereit}},
1275: \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{Thamm}},
1276: \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{Trampert}},
1277: \bibinfo{author}{\bibfnamefont{H.}~\bibnamefont{von Kiedrowski}},
1278: \bibnamefont{and} \bibinfo{author}{\bibfnamefont{K.~H.} \bibnamefont{Ploog}},
1279: \bibinfo{journal}{Appl. Phys. Lett.} \textbf{\bibinfo{volume}{75}},
1280: \bibinfo{pages}{4019} (\bibinfo{year}{1999}).
1281:
1282: \bibitem[{\citenamefont{Gay et~al.}(1953)\citenamefont{Gay, Hirsch, and
1283: Kelly}}]{gay53}
1284: \bibinfo{author}{\bibfnamefont{P.}~\bibnamefont{Gay}},
1285: \bibinfo{author}{\bibfnamefont{P.~B.} \bibnamefont{Hirsch}},
1286: \bibnamefont{and} \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{Kelly}},
1287: \bibinfo{journal}{Acta Metall.} \textbf{\bibinfo{volume}{1}},
1288: \bibinfo{pages}{315} (\bibinfo{year}{1953}).
1289:
1290: \bibitem[{\citenamefont{Dunn and Koch}(1957)}]{dunn57}
1291: \bibinfo{author}{\bibfnamefont{C.~G.} \bibnamefont{Dunn}} \bibnamefont{and}
1292: \bibinfo{author}{\bibfnamefont{E.~F.} \bibnamefont{Koch}},
1293: \bibinfo{journal}{Acta Metall.} \textbf{\bibinfo{volume}{5}},
1294: \bibinfo{pages}{548} (\bibinfo{year}{1957}).
1295:
1296: \end{thebibliography}
1297:
1298:
1299: \end{document}