1: %\documentclass[aps,preprint,superscriptaddress,]{revtex4}
2: \documentclass[aps,prl,twocolumn]{revtex4}
3: \usepackage{graphics}
4: \usepackage{graphicx}
5: \begin{document}
6: \title{Can vacancies lubricate dislocation motion in aluminum?}
7: \author{Gang Lu and Efthimios Kaxiras}
8: \affiliation
9: {Department of Physics and Division of Engineering and Applied Science,\\
10: Harvard University, Cambridge, MA 02138}
11: \begin{abstract}
12: The interaction of vacancy with dislocations in Al
13: is studied using the Semidiscrete Variational Peierls-Nabarro
14: model with {\it ab initio}
15: determined $\gamma$-surface. For the first
16: time, we confirm theoretically the so-called vacancy
17: lubrication effect on dislocation motion in Al, a discovery that
18: can settle a long-standing controversy in dislocation theory for
19: fcc metals. We provide insights on the lubrication effect
20: by exploring the connection between dislocation mobility and its
21: core width. We predict an increased dislocation splitting in
22: the presence of vacancy. We find that on average there is a
23: weak repulsion between vacancies and dislocations which
24: is independent of dislocation character.
25: \end{abstract}
26: \maketitle
27: Defects and their mutual interactions dominate the properties of
28: materials that host them. Vacancies as point defects,
29: have long been known to strongly interact with dislocations (line defects),
30: and the study of
31: their interactions represents one of most challenging problems
32: in material science and engineering
33: \cite{balluffi}. More than a decade ago, Benoit {\it et al.}
34: discovered an interesting phenomenon in ultra-high-purity
35: aluminum deformed at low temperature of 4.2 K.
36: They observed a marked decrease of elastic modulus in cold-worked Al,
37: which they attributed to vacancy
38: enhanced dislocation mobility in Al, a novel phenomenon
39: they termed dislocation lubrication effect \cite{lauzier,benoit}.
40: Corroborative experimental evidence led
41: these authors to conclude that vacancies,
42: generated from cold-work or irradiation, are solely responsible for
43: the enhanced dislocation mobility. This indeed is a quite intriguing
44: result because traditionally vacancies are thought to lock dislocation
45: motion by forming atmospheres around the dislocation \cite{nabarro}.
46: Furthermore the lubrication effect may hold the key to resolve
47: the long-standing controversy for Peierls
48: stress ($\sigma_p$) estimated from plastic deformation and from internal
49: friction measurements. It is generally believed that kink pair formation
50: (KPF) of dislocations is responsible for the Bordoni peaks observed in
51: internal friction measurements of fcc metals \cite{seeger,fantozzi}.
52: However $\sigma_p$ derived from KPF mechanism
53: is in the order of 10$^{-3} \mu$ ($\mu$ is the shear modulus)
54: far greater than the critical resolved shear stress (CRSS) estimated
55: from plastic deformation
56: experiments, which is around 10$^{-5}$ to 10$^{-4}\mu$ \cite{seeger2}.
57: This has been regarded as a serious problem because it casts doubt
58: on the well accepted KPF theory for the Bordoni peak in fcc metals.
59: The controversy is particularly troublesome in light of the good
60: agreement found in bcc and ionic crystals regarding $\sigma_p$ measured
61: from internal friction and low temperature CRSS
62: experiments \cite{fantozzi}. Therefore the vacancy lubrication
63: effect may settle the controversy by proposing that vacancies strongly
64: interact with dislocations and as a consequence lower their $\sigma_p$
65: to the level
66: that is consistent with the low temperature CRSS experiments and
67: thus bridge the gap in $\sigma_p$
68: \cite{lauzier,benoit}. Finally if the lubrication mechanism
69: turns out to be general, it may lead to an innovation in molding
70: technology for materials with high $\sigma_p$ by introducing
71: vacancies to the materials \cite{kosugi}. However interesting the
72: lubrication effect may seem to be, it has not been widely accepted,
73: which in our opinion, is due to poor
74: understanding of the phenomenon. In fact, there is no
75: complete theoretical
76: work ever published to address this problem, to the best of our
77: knowledge. Therefore it is the purpose of
78: this paper to present the first {\it ab initio} study of
79: the problem. As we will show in the following, our calculations not only
80: provide theoretical support for the lubrication effect,
81: they also reveal other important difference in
82: dislocation properties that are associated with the presence of
83: vacancies.
84:
85: In this paper, we employed the recently developed Semidiscrete Variational
86: Peierls-Nabarro (SVPN) model \cite{bulatov,lu1} in conjunction with
87: {\it ab initio} determined $\gamma$-surfaces \cite{gamma}. The SVPN model
88: provides an ideal framework for multiscale simulations of dislocation
89: properties, and it combines an atomistic ({\it ab initio}) treatment of
90: the interactions
91: across the slip plane and an elastic treatment of the continua on either
92: side of the slip plane.
93: The model has been shown to be quite successful in predicting
94: dislocation properties by comparing its predictions against the direct atomistic
95: simulations results \cite{bulatov,lu1}. For example, by using the $\gamma$-surface
96: calculated from Embedded Atom Method (EAM), we obtained $\sigma_p$ for dislocations
97: in Al that are in excellent agreement with that from
98: direct atomistic simulations employing the same EAM potential \cite{lu1}.
99: More remarkably, a good agreement is also achieved for dislocations
100: in Si \cite{bulatov} where the classic Peierls-Nabarro model fails
101: owing to the its insufficiency to deal with narrow dislocations,
102: such as dislocations in Si. Since one has no {\it a priori} knowledge
103: regarding the size of dislocations in the presence of vacancies, the SVPN
104: model seems to be particularly useful to explore the interaction of
105: vacancy with dislocations. Besides
106: $\sigma_p$, the model can also provide reliable results for other dislocation
107: properties, such as partial
108: separation distance and core width \cite{lu1}, as they are compared to direct
109: atomistic simulations \cite{bulatov2,fang}. Thus the strength
110: of this approach, when combined with {\it ab initio} calculations for
111: $\gamma$-surface
112: is that it produces essentially an atomistic simulation for dislocation
113: properties without suffering from uncertainties associated with empirical
114: potentials.
115:
116: In the SVPN approach, the equilibrium structure of a dislocation
117: is obtained by minimizing the dislocation energy functional \cite{bulatov,lu1}
118: \begin{equation}
119: U_{disl} = U_{elastic} + U_{misfit} + U_{stress} + Kb^2{\rm ln}L,
120: \end{equation}
121: where
122: \begin{equation}
123: U_{elastic} = \sum_{i,j}\frac{1}{2}\chi_{ij}[K_e(\rho_i^{(1)}\rho_j^{(1)} +
124: \rho_i^{(2)}\rho_j^{(2)}) + K_s\rho_i^{(3)}\rho_j^{(3)}],
125: \end{equation}
126: \begin{equation}
127: U_{misfit} = \sum_i\Delta x \gamma(\vec{f_i}),
128: \end{equation}
129: \begin{equation}
130: U_{stress} = - \sum_{i,l}\frac{x_i^2-x_{i-1}^2}{2}\rho_i^{(l)}\tau^{(l)},
131: \end{equation}
132: with respect to the dislocation Burgers vector density $\rho_i$.
133: Here, $\rho_i^{(1)}$, $\rho_i^{(2)}$ and $\rho_i^{(3)}$
134: are the edge, vertical and
135: screw components of the general Burgers vector density
136: defined at the $i$th nodal point as
137: $\rho_i = (f_i - f_{i-1})/(x_i - x_{i-1})$,
138: where $f_i$ and $x_i$ are the displacement vector
139: and the coordinate of the $i$th nodal point (atomic row).
140: $\gamma(\vec{f}_i)$ is the
141: $\gamma$-surface that is determined from {\it ab initio}
142: calculations.
143: $\tau^{(l)}$ is the external stress component interacting
144: with the corresponding Burgers vector density $\rho^{(l)}$.
145: $\chi_{ij}$ is the discretized elastic energy kernel,
146: and $K$, $K_e$ and $K_s$ are the
147: pre-logarithmic elastic energy factors \cite{bulatov,lu1}.
148: The quantity $L$ entering the last term
149: is the outer cutoff radius for the configuration-independent
150: part of the elastic energy \cite{hirth}.
151: We identify the dislocation {\it configuration-dependent}
152: part of the elastic energy and
153: the misfit energy as core energy, i.e., $U_{core} = U_{elastic} + U_{misfit}$.
154: The response of a dislocation to an external stress is achieved by minimization
155: of the energy functional at
156: the given value of the applied stress.
157: An instability is reached when an optimal solution for the Burgers vector density
158: distribution no longer exists, which
159: is manifested numerically by the failure of the minimization procedure to convergence.
160: $\sigma_p$ is then identified as the critical value of the applied stress
161: giving rise to this instability.
162:
163: In order to examine how vacancies change dislocation core structure by
164: modifying atomic bonding across the
165: slip plane, we carry out {\it ab inito}
166: calculations for the $\gamma$-surface of Al with vacancies at the slip plane.
167: Specifically, we select a supercell containing six Al layers in [111]
168: direction with four atoms per layer, and remove one Al atom from the
169: top layer (right below the designated slip plane) of the supercell to simulate a
170: vacancy concentration at 4 at.\%. We should emphasize that 4 at.\%
171: represents the vacancy concentration at the dislocation
172: core region that we are interested, therefore it is much greater than the average
173: vacancy concentration of the bulk material.
174: The {\it ab initio} calculations are based on the pseudopotential
175: plane-wave method with local density approximation \cite{kohn} to
176: the exchange-correlation functional
177: \cite{perdew}. A kinetic energy cutoff of 12 Ry for the plane-wave basis is used
178: and a $k$-point mesh consisting of (8,8,4) divisions along the reciprocal lattice
179: vectors is sampled for the Brillouin zone integration. Atomic relaxation is
180: performed before we initiate the sliding. During
181: the sliding process, atoms are allowed to move only along [111] direction
182: while the atoms at the innermost two layers are held fixed. Volume relaxation is
183: also performed for each sliding distance to minimize the tensile stress on
184: the supercell.
185:
186: The {\it ab initio} determined $\gamma$-surface for Al with and without
187: vacancies is presented in Fig. 1.
188: In order to highlight
189: the vacancy effect on $\gamma$-surface, we also summarize in
190: Table I some important stacking fault energies for both Al and Al+V
191: (Al with vacancies) systems. These special stacking faults correspond to the
192: various extremes along [12$\bar{1}$] and [101] directions of the
193: $\gamma$-surface.
194: It is clear that the vacancy lowers the intrinsic and unstable
195: stacking fault energy along [12$\bar{1}$] direction
196: while increases the run-on stacking fault energy and unstable stacking
197: fault energy along [101] direction. Therefore it is not immediately
198: clear how dislocation core structure can be changed by vacancies, and
199: a detailed analysis based on SVPN model is needed.
200: Since the experiments \cite{lauzier} have concluded that the
201: change in elastic constants due to vacancies is not responsible for the observed
202: lubrication effect, we will simply use the experimental elastic constants
203: of pure Al in our calculations for Al+V system \cite{lu1}.
204:
205: Having determined all the necessary parameters entering the model, we can
206: study the interaction of vacancy and dislocations using
207: the SVPN model. We select four, namely, screw (0$^\circ$),
208: 30$^\circ$, 60$^\circ$ and edge (90$^\circ$) dislocations,
209: all with the same Burgers vector, but different
210: orientations, in our calculations. $\sigma_p$ calculated
211: from the SVPN model for the dislocations with and without
212: vacancies is listed in Table II.
213: $\sigma_p$ represents the intrinsic mobility of a straight dislocation,
214: and it relates to the kink pair formation energy, 2$W_k$, by
215: \begin{equation}
216: 2W_k = \frac{(16\mu b^3 a^3 \sigma_p)^{1/2}}{\pi},
217: \end{equation}
218: where $b$ is the Burgers vector and $a$ the distance between
219: neighboring Peierls valleys \cite{seeger,fantozzi,benoit2}.
220: The corresponding values for Al are $\mu$ = 28.8 GPa,
221: $b = 2.85$ \AA~ and $a = 2.47 $\AA.
222: By measuring $2W_k$ (the activation energy for the Bordoni peak)
223: in internal friction experiments, one can derive $\sigma_p$
224: according to Eq. (5) for the relevant dislocation.
225: For example, using the experimentally measured value of
226: $2W_k$ (0.21 eV) for a screw dislocation, one obtains
227: $\sigma_p$ =
228: 8$\times 10^{-3}\mu$ (224 MPa) for the screw dislocation
229: in pure Al \cite{fantozzi,benoit2}. This value of $\sigma_p$ is in
230: excellent agreement with our model result, 8.82$\times 10^{-3}\mu$
231: for the same dislocation
232: (Table II). Furthermore, the less definite measurement for the
233: subsidiary peak (B1 peak) yields an activation energy ranging from
234: 0.12 to 0.16 eV, which corresponds to $\sigma_p$ in the
235: range of 2.8 to 4.6$\times 10^{-3}\mu$ (80 to 130 MPa) for the
236: 60$^\circ$ dislocation. This value
237: also agrees well with our result for the same
238: dislocation (3.40$\times 10^{-3}\mu$). The overall consistency between the
239: theoretical and experimental values for $\sigma_p$ indicates
240: the reliability of our model and establishes
241: the basis for further study of vacancy effect.
242:
243: When vacancies are introduced at the slip plane but are not adsorbed
244: by a dislocation line, we find that $\sigma_p$ for various
245: dislocation is lowered by more than one order of magnitude (except
246: for the edge dislocation), as shown in Table II. Therefore we
247: have confirmed the vacancy lubrication effect theoretically
248: for the first time since its proposal.
249: The fact that this lubrication effect is observed for
250: various dislocations suggests a generic nature of the
251: underlying mechanism. In order to shed light on this general
252: mechanism,
253: we have calculated dislocation core width which is defined
254: as the atomic spacing over which the relative displacement of the
255: dislocation changes from 1/4$b$ to 3/4$b$ \cite{lu1}.
256: It is generally believed that $\sigma_p$ is exponentially
257: lowered with the increase of dislocation half-width according
258: to the Peierls-Nabarro model \cite{hirth,lu1}.
259: The calculated dislocation half core width ($\zeta$) is
260: presented in Table II. It is found that in the
261: presence of vacancy, dislocation becomes 60\% to 90\% wider,
262: which we believe is due to the reduced slope of the $\gamma$-surface
263: along [12$\bar{1}$] direction as vacancies are
264: introduced. Since lattice restoring
265: force, represented by the slope of $\gamma$-surface, is
266: weakened by the vacancies, the repulsive elastic force
267: resulting from the continuous distribution of infinitesimal
268: dislocations dominates, leading to a wider dislocation core
269: and therefore enhanced dislocation mobility.
270: Although the vacancy lubrication effect may be qualitatively
271: understood from above argument by a careful inspection of
272: the $\gamma$-surface, one has to resort to
273: SVPN model to obtain reliable values of
274: $\sigma_p$ in order to make a quantitative comparison.
275: As shown in Table II, vacancies can bring $\sigma_p$
276: down to the values derived from the plastic deformation
277: experiments (10$^{-5}$ to 10$^{-4} \mu$), therefore bridge
278: the gap for $\sigma_p$ between the internal friction
279: measurement and the measurement of CRSS
280: at low temperature.
281:
282: In order to gain more insights into the interaction of vacancy
283: and dislocations, we have calculated dislocation Burgers vector
284: density for pure Al and Al+V, shown in Fig. 2. It is found
285: that dislocations tend to dissociate more into partials in
286: the presence of vacancy. This behavior is obviously associated
287: with the fact that the intrinsic stacking fault energy is reduced
288: by vacancy. The result cautions us to be more careful in
289: interpreting TEM data for partial separation distances because
290: accidentally introduced vacancies could change the result
291: significantly. We have also calculated
292: binding energies of vacancies to dislocation cores,
293: summarized in Table II. The binding energy is defined as difference between
294: dislocation core energy with and without the presence of vacancy.
295: Overall we find that the binding energies
296: are not sensitive to dislocation character,
297: and more importantly they are all marginally positive.
298: The positive binding energy indicates that dislocation is
299: energetically less stable with vacancies nearby.
300: Our result for positive value of binding energy qualitatively
301: agrees with the findings from atomistic simulations carried out
302: for Cu \cite{balluffi}. The small value of binding energy
303: we found in Al may relate to the fact that the dislocations are
304: dissociated into partials in Al. It is believed that the vacancy
305: binding energy to dissociated
306: dislocations is considerably lower than that to
307: the same dislocations in the non-dissociated condition
308: \cite{balluffi}.
309: Of course one has to bear in mind that the binding energy
310: we obtained represents an average value over all possible sites
311: for a vacancy at the core of the dislocation.
312:
313: To conclude, we have studied the interaction of vacancy with
314: dislocations in Al using the SVPN model with {\it ab initio}
315: determined $\gamma$-surface. We confirm the experimental
316: finding of vacancy lubrication effect in Al.
317: We propose that vacancies can weaken the
318: lattice restoring force across the slip
319: plane, which leads to a wider dislocation spreading,
320: and thus higher dislocation mobility. We find that $\sigma_p$
321: of the dislocations is lowered by more than one order of
322: magnitude in the presence of vacancy, which bridges the
323: gap between $\sigma_p$ observed from different experiments,
324: resolving one of the long-standing problems in dislocation
325: theory.
326: This work represents the first theoretical effort to
327: challenge the traditional point of view that regards vacancy as
328: a locking agent for dislocation motion.
329: We predict that vacancies can increase partial separation distance
330: in Al, and finally we find there exists a weak repulsion between
331: dislocations and vacancy, which is independent of
332: the dislocation character.
333:
334: \begin{acknowledgments}
335: We acknowledge the support from Grant No. F49620-99-1-0272
336: through the U.S. Air Force Office for
337: Scientific Research.
338: \end{acknowledgments}
339: \begin{thebibliography}{99}
340: \bibitem{balluffi}
341: R.W. Balluffi and A.V. Granato, in {\it Dislocations in Solids},
342: edited by F.R.N. Nabarro (North-Holland, Amsterdam, 1979), Vol. 4, p.1.
343: \bibitem{lauzier}
344: J. Lauzier, J. Hillairet, A. Vieux-Champagne, and W. Benoit, J. Phys.
345: Condens. Matter, {\bf 1}, 9273 (1989); J. Lauzier, J. Hillairet,
346: G. Gremaud and W. Benoit, {\it ibid}, {\bf 2}, 9247 (1990).
347: \bibitem{benoit}
348: W. Benoit, G. Gremaud and B. Quenet, Mater. Sci. Eng. A {\bf 164}, 42 (1993).
349: \bibitem{nabarro}
350: F.R.N. Nabarro, {\it Theory of Crystal Dislocations}, (Dover, New York, 1987).
351: \bibitem{seeger}
352: A. Seeger, Philos. Mag. {\bf 1}, 651 (1951).
353: \bibitem{fantozzi}
354: G. Fantozzi, C. Esnouf, W. Benoit and G. Ritchie, Prog. Mater. Sci, {\bf 27},
355: 311 (1982).
356: \bibitem{seeger2}
357: A. Seeger and P. Schiller, in {\it Physical Acoustics}, edited by
358: W. Mason (Academic Press, New York, 1966), Vol. 3, p.366.
359: \bibitem{kosugi}
360: T. Kosugi and T. Kino, Mater. Sci. Eng. A {\bf 164}, 368 (1993).
361: \bibitem{bulatov}
362: V. V. Bulatov and E. Kaxiras, Phys. Rev. Lett. {\bf78}, 4221 (1997).
363: \bibitem{lu1}
364: G. Lu, N. Kioussis, V. V. Bulatov, and E. Kaxiras, Phys. Rev. B {\bf62}, 3099 (2000),
365: Philos. Mag. Lett. {\bf80}, 675 (2000).
366: \bibitem{gamma}
367: When a crystal is cut along its slip plane and the upper half is displaced
368: relative to the lower by a vector $\vec{f}$, the energy increase per unit area
369: is defined as $\gamma$ and the energy surface
370: $\gamma(\vec{f})$, with $\vec{f}$ spanning the entire slip plane, is
371: called the $\gamma$-surface.
372: \bibitem{hirth}
373: J.P. Hirth and J. Lothe, {\it Theory of Dislocations}, 2nd ed. (Wiley, New York, 1992).
374: \bibitem{bulatov2}
375: V.V. Bulatov, O. Richmond, and M.V. Glazov, Acta Mater. {\bf 47}, 3507 (1999).
376: \bibitem{fang}
377: Q.F. Fang and R. Wang, Phys. Rev. B {\bf 62}, 9317 (2000).
378: \bibitem{kohn}
379: W. Kohn and L. Sham, Phys. Rev. {\bf 140}, A1133 (1965).
380: \bibitem{perdew}
381: J. Perdew and A. Zunger, Phys. Rev. B {\bf 23}, 5048 (1984).
382: \bibitem{benoit2}
383: W. Benoit, M. Bujard and G. Gremaud, Phys. Stat. Solidi. {\bf A104}, 427 (1987).
384: \end{thebibliography}
385:
386:
387: \begin{figure}
388: \includegraphics[width=300pt]{fig1a.eps}
389: \includegraphics[width=300pt]{fig1b.eps}
390: \caption{The $\gamma$-surface (J/m$^2$) for displacements along a (111) plane
391: for (a) pure Al and (b) Al+V systems.
392: The corners of the plane and its center correspond to identical equilibrium
393: configuration, i.e., the ideal lattice. The two surfaces are displayed in exactly
394: the same perspective and on the same energy scale to facilitate comparison.
395: The $\gamma$-surface of Al+V is truncated
396: to emphasize the more interesting region.}
397: \end{figure}
398:
399: \begin{figure}
400: \includegraphics[width=300pt]{fig2.eps}
401: \caption{Dislocation Burgers vector density for four dislocations (clockwise)
402: : screw (0$^\circ$), 30$^\circ$, 60$^\circ$ and edge (90$^\circ$) for the
403: pure Al (solid lines) and the Al+V (dashed lines) systems.
404: The peaks in the density plot represent partial dislocations.}
405: \end{figure}
406:
407: \begin{table}[p]
408: \caption{Fault vectors and energies (J/m$^2$) for some important stacking faults
409: of the pure Al and the Al+V systems.}
410: \begin{ruledtabular}
411: \begin{tabular*}{\columnwidth}{@{\extracolsep{\fill}}cccc}
412: & Vector & Al & Al+H \\ \hline
413: Intrinsic stacking & 1/6[12$\bar{1}$] & 0.164 & 0.105 \\
414: Unstable stacking & 1/10[12$\bar{1}$] & 0.224 & 0.143 \\
415: Unstable stacking & 1/4[101] & 0.250 & 0.427 \\
416: Run-on stacking & 1/3[12$\bar{1}$] & 0.400 & 0.831\\
417: \end{tabular*}
418: \end{ruledtabular}
419: \end{table}
420:
421: \begin{table}[p]
422: \caption{Peierls stress ($\sigma_p$, 10$^{-3}\mu$), half core width ($\zeta$,
423: \AA),
424: core energies ($U_{core}$, eV/\AA)
425: for the four dislocations
426: in the pure Al and the Al+V systems and
427: binding energy ($U_{b}$, eV/vacancy) for the four dislocations.}
428: \begin{ruledtabular}
429: \begin{tabular*}{\columnwidth}{@{\extracolsep{\fill}}cccccc}
430:
431: & & screw & 30$^\circ$ & 60$^\circ$ & edge \\ \hline
432: $\sigma_p$ & Al & 8.82 & 1.77 & 3.40 & 0.10 \\
433: & Al+V & 0.69 & 0.10 & 0.26 & 0.05 \\ \hline
434: $\zeta$ & Al & 2.1 &2.5 & 3.0 & 3.5 \\
435: & Al+V& 3.2 &4.0 & 5.6 & 6.3 \\ \hline
436: $U_{core}$ & Al& -0.084 & -0.110 & -0.168 & -0.198 \\
437: & Al+V&-0.046 & -0.073 & -0.133 & -0.164 \\ \hline
438: $U_{b}$ & &0.038 & 0.037 & 0.035 & 0.034 \\
439: \end{tabular*}
440: \end{ruledtabular}
441: \end{table}
442:
443: \end{document}
444:
445:
446:
447: