Vous êtes sur la page 1sur 15

Pergamon PI1 S1359-6454(96)00132-2

Acta mater. Vol. 44, No. 12, pp. 47494163, 1996 Coovrirht RZ 1996 Acta Metallurzica Inc. Published by Elsevier S&ace Ltd
.I_ 1

Printed in Great Britain. All rights reserved 1359-6454196 $15.00 + 0.00

STEPS, DISLOCATIONS AND DISCONNECTIONS INTERFACE DEFECTS RELATING TO STRUCTURE PHASE TRANSFORMATIONS
J. P. HIRTH and R. C. POND*

AS AND

Mechanical and Materials Engineering Department, Washington State University, Pullman, WA 99164-2920, U.S.A. and 2Materials Science and Engineering Department, University of Liverpool, Liverpool L69 3BX, England (Received 21 March 1996) defects may exhibit both dislocation and step-like character. The Burgers vector of such a defect is an invariant topological quantity which can be determined mathematically or graphically by circuit mapping. Step character cannot be determined directly by such methods, but a new definition of scalar step heights is presented here in terms of crystallographic parameters. This partition of a disconnection into dislocation and step portions facilitates the analysis of processes such as phase Abstract-Interfacial transformations; the dislocation portion is associated with the long-range strain field and shape change, and the step portion with divergences in diffusional fluxes associated with growth. Examples are given in the h.c.p./f.c.c. case for various types of defect. Copyright 0 1996 Acta Metallurgica Inc.

1. INTRODUCTION

For many phase transformations, the mechanism of growth involves the motion of interfacial defects with dislocation and/or step character. Examples in diffusional transformations include precipitation from solid solution, discontinuous transformation, bainitic transformation and scaling reactions such as oxidation. Defects with step/dislocation character can also be important for diffusionless twinning and martensitic transformations. Different names have been associated with such defects, depending on whether one focuses on the step character (e.g. structural ledges, growth ledges) or the dislocation character (transformation dislocations). This nomenclature is enumerated in several reviews [l-4]. One of us [5] has ventured to define the generic defect with both dislocation and step character as a disconnection, consistent with a general discussion of symmetry by Kriiner [6]. In the present work, disconnections are re-examined on the basis of a recent generalized symmetry theory for the topological description of interface defects developed by one of the authors [3] and a new definition of step character. We also address the issue of possible redundancies in the definition of the interface plane for certain high symmetry cases.
2. BASIC SYMMETRY CONSIDERATIONS

In a recent paper [7] the topological properties of interfacial defects were set out in an integrated and comprehensive treatment based on the fundamental consequences of symmetry breaking in crystalline materials. Topological properties are intrinsic geo4749

metrical characteristics of linear and extended defects which are independent of defect orientation: an example is the Burgers vector of a dislocation. The treatment referred to above is comprehensive in the sense that all the symmetry of the adjacent crystals is taken into account, which allows surface, bulk and interface defects to be integrated into a single formal framework. For interface line-defects, the theory shows that admissible defects (i.e. those which separate crystallographically equivalent or identical interface structures) are characterized by operations 2$ = (QI/, q,,) that are given by the product of two symmetry operations, one from each of the two crystals. The range of distinct defects that can arise in a given interface depends on the extent of symmetry breaking when the bicrystal is formed. In the case of interface dislocations, which is of particular interest here, 9,, = (I, by) where I is the identity matrix and b,, is the Burgers vector of the defect. Such interface dislocations can arise as a result of breaking either translation or point symmetry operations and several distinct defects have been classified [3, 71; the reader is referred to the original work for details. Pure interface steps occur in the special case where the adjacent crystals have some symmetry in common, and hence no dislocation character arises, ~ = (I, 0). 2~ On an external (or free) surface of a single crystal, denoted by its outward pointing unit normal n, steps and demisteps can arise (ignoring other features that are caused by surface reconstruction), as indicated schematically in Fig. 1. In the case of a step [Fig. I(a)], the adjacent regions are related by a translation vector of the crystal (I, t) which is not parallel to the

4750

HIRTH and POND: STEPS, DISLOCATIONS AND DISCONNECTIONS

surface. The height of the step is h = n-t, and its motion would cause crystal growth or evaporation involving a flux of material. The step can be imagined to have been formed by the emergence of a dislocation with Burgers vector b onto the surface as indicated. The Burgers vector b and the step vector t are defined in the RH/FS sense [8, 91 relative to the sense vector 5, a unit vector parallel to the appropriate defect. According to this convention, t is the translation which relates the surface on the left-hand side of the defect (looking along 5 and onto the surface) to that on the right; thus cxb and gxt point to any extra material associated with the defect. Demisteps [Fig. l(b)] can only arise on certain surfaces of non-symmorphic crystals; the adjacent surfaces are now interrelated by a screw rotation or mirror-glide symmetry operation % (W, w); the = orientation of n is left invariant by this operation, i.e. Wn = n, but the surface is shifted by a vector w, analogous to t, and hence the demistep height is n-w. A step and a demistep are illustrated schematically in Fig. 2 for the case of an h.c.p. metal crystal; a step, S, is seen edge-on on a (liO0) surface, and a demistep, DD on the (0001) surface is viewed from above. , While t is a topological parameter, the height of surface features is not such a property; the height of a step characterized by (I, t) clearly depends on the orientation n of the surface on which the step exists; similarly for the case of demisteps. Nevertheless, the notion of a step or demistep is quite clear in the case of free surfaces. Moreover, the interrelation of bulk line defects and surface steps is also clear. When a dislocation characterized by (I, t), with b = t, emerges from the bulk onto a surface n, it creates a step of height n-t as indicated in Fig. l(a). Thus, the step and dislocation are both characterized by the same operation, (I, t), despite the dramatic change of atomic structure associated with the transition. In other words, the topological properties of a line

Fig. 2. Schematic illustration of a step, S, and demistep, DD on a (1100) and (0001) surface, respectively, , of an h.c.p. crystal.

~ f ,,,,,,,, // ,,,,
al

/Jh=J.l

b)

Fig. 1. Schematic illustration of surface features: (a) step, (b) demistep. The sign convention adopted is consistent with the RH/FS scheme for dislocations.

defect are conserved on emerging from the bulk onto a free surface. (Similar arguments apply to the emergence of dispirations onto certain surfaces leading to demisteps.) We now turn to the case of interface dislocations. These may be pure dislocations or may be the dislocation component of a disconnection that also has step character. The definition of a step in this instance is not as straightforward as for a step in a free surface. In previous work [3,7] it has been shown that sii can be developed from a Volterra-like operation applied to a bicrystal. A bicrystal is imagined to be created by joining a white (1) and black (p) crystal with prepared surfaces, as indicated in Fig. 3(a). The sign convention described previously for free surface discontinuities must now be reviewed since the white and black surface normals would point in opposite directions and we now need to choose a unique interfacial unit normal; here we adopt the convention used in previous work [3,7] where this normal, II, is taken to point into the white crystal (an alternative description which retains the free surface sign convention is outlined in Appendix A). Surface discontinuities like those on the white and black crystal surfaces in Fig. 3(b) and (c) are now characterized by operations that relate the crystal surfaces on the left to those on the right of each crystal when looking along 6 and into the white crystal. In the following analysis the characterization of interfacial defects is consistent with the RH/FS convention, and hence interactions between interface and crystal defects can be treated in the usual way.

HIRTH and POND: STEPS, DISLOCATIONS AND DISCONNECTIONS Interface dislocations arise when the black and white crystal surfaces to be joined exhibit surface features that are not complementary [Fig. 3(b) and (c)] or when the relative position of the crystals breaks otherwise coincident point symmetry elements. In Fig. 3(b), steps are present on the two crystal surfaces, and joining the crystals so that the interfacial structures are identical on either side of these perturbations introduces a dislocation characterized by so = (I, b,) with b, = t(l), - E%(p),

4751

(1)

where S = (P, p) expresses the relationship between the co-ordinate frames of the two crystals, p being the relative shift of their origins. The magnitude of the normal component of b,, designated b,, is equal to the difference between the two surface step heights expressed in the (1) frame, h(l), = n-t(A), and h(p), = n*Pt(p)i, i.e. b, is given by the scalar equation b. = h(l), - h(p),

(2)

In the case of the two misfitting demisteps in Fig. 3(c), the resulting dislocation, which can only arise on certain interfaces, is characterized by Z?,,= (I, b,) with

a)

b, = w(A), - Pw@), + (W - I)p

(3)

b)

where W is the mirror or rotation operation relating the surface structures on either side of the free surface demisteps, and w(n), and w(p)! are the associated, but unequal, shifts. As before, the magnitude of the normal component b. is given by expression (2) with h(1), = n*w(l), and h(l), = n*Pw(p), Other possibilities arise in special cases, such as steps on the white crystal surface juxtaposed with demisteps on the black, but we do not illustrate them here since the main principles are illustrated in Fig. 3. It is apparent that the interface location changes from one side of the defects in Fig. 3 to the other; i.e. there is an interfacial step. However, the height of such steps is not a topological property and requires further consideration. Pond [3] has argued that the interfacial step is given approximately by the average of the heights of the two free surface step heights taking their signs into account. This height, designated h,, is equal to h, = fn*[t(l), + Pt(p),]. (4)

cl

Fig. 3. (a) Schematic illustration of the formation of a bicrystal by joining crystals with prepared surfaces. (b) Formation of a disconnection by joining crystal surfaces with incompatible steps. (c) Incompatible demisteps.

The actual change of interface location would depend on the elastic parameters of the two crystals, and hence can only be given approximately by a geometrical expression such as equation (4). King and Smith [lo] have given another prescription for the case of defects in periodic grain boundaries. In the present work we wish to avoid uncertainties associated with elastic effects, and define a step height that is related to the transfer of material from one crystal to the other associated with motion of the defect along the interface, i.e. the differential diffusive flux between the crystals during phase transformation by defect motion. This step height, designated h, is simply the extent of material overlap normal to the interface when the adjacent crystals are placed together to make contact without elastic deformation to bond the interfaces, as depicted schematically in Fig. 4. For such an overlap to exist, h(i), and h(p), must have the same sign; the magnitude and sign of h is then simply that of the smaller of the two imaginary surface steps. Figure 4(a) corresponds to Fig. 3(b), where h(1), < h(p), and

4152 (a)

HIRTH

and POND:

STEPS, DISLOCATIONS

AND DISCONNECTIONS

fhrh(k)

Or

(b)

(cl or,

(d)

Fig. 4. Schematic illustrations to define overlap step height, (b) h = h(l) at a disconnection formed by incompatible demisteps as in Fig. 3(c); (c) h = 0 at a pure dislocation; and (d) h = h(l) = h(p) at a pure step.

h: (a),

both are positive. can also write

Consistently

with equation

(2) we

h = h(n), = n-t(A), = h(p), + b,

(5)

as is indicated schematically in the figure. Similarly, Fig. 4(b) corresponds to Fig. 3(c) where h(l), < h(p), and both are positive, so that h =
h(l)j = n*W(l)j = h(p), + bn. (6)

In cases where the surface steps have opposite signs, such as in Fig. 4(c), h = 0 because there is no material overlap in the sense defined above. The interfacial defect arising in this case would be a pure interface dislocation with b, given by expression (l), and this would include a substantial component normal to the interface, and hence its motion would require a flux of material. However, this flux would bring about the simultaneous growth (dissolution) of both crystals. On the other hand, for the defects in Fig. 4(a) and (b) the step height is associated with a differential flux required to transform one crystal into the other. A pure step is illustrated in Fig. 4(d), i.e. t(l), = Pt(ph and hence b, = 0. In this case the height of the step according to the two definitions above would be the same, i.e. h = h(l), = h(,u), = h,,. The crystallographic analysis described above is explicitly for the case of defects separating energeti-

cally degenerate regions of interface, but can be readily extended to include non-degenerate instances. Step heights h(i) and h(p) can still be identified in association with incompatible surface features in such cases, and hence so also can material overlap step heights, h. (We note that, unlike the degenerate case, extra material associated with non-degenerate defects may not comprise integral multiples of chemical formula units.) Burgers vectors can be defined by a-posteriori circuit mapping as illustrated later. In non-degenerate cases, it is possible that the excess volume of the interfacial structures on either side of a defect may not be identical [ 11, 121. Excess volume can be represented formally as a component of p that is antiparallel to n; thus a difference in the excess volumes on either side of a defect leads to a normal component of b that is perpendicular to the interface. We designate the magnitude of this component p., and hence write the total normal component of b as b, + fin. In the present context it is important to recognize that Pn is zero for the degenerate case, but may not be zero for the non-degenerate case. However, in the latter case, no material flux is associated with the climb of interfacial defects on account of the component Pn, and hence it need not be included in the modelling of such fluxes presented later. In summary, we suggest that the definition of step height in terms of material overlap, h, is useful in the context of phase transformations because its definition is independent of elastic distortion, and its magnitude is related to differential fluxes necessary to transform one phase into the other. Motion of a defect along an interface will require material fluxes related through the crystal densities to h(l), h(p) and h; the former two are associated with the growth/dissolution of the individual white and black crystals, and h is related to the extent of transformation from white to black or vice versa. A wide range of possibilities exists regarding the Burgers vectors and step heights for interfacial disconnections, and hence for the function of these defects during phase transformations. For example, consider first situations where h = 0; this case arises either as illustrated in Fig. 4(c) or, for example, when crystal dislocations enter the interface to relieve misfit stresses. The latter defects are often called misfit dislocations [13], and have b, = t(l), (or P*t(p),). Misfit is most efficiently accommodated when b, is parallel to the interface, and hence h(l) [or h(p)] = 0; no misfit would be accommodated if b, were perpendicular to the interface, in which case b, = nb, and b, = h(l) [or h(p)]. No differential flux is associated with the motion of disconnections if h = 0, but a flux is connected with the climb motion of defects if b, # 0. In the case of a pure step, i.e. b, = 0, h # 0, a differential flux will arise unless the densities and compositions of the (1) and (p) phases are identical, a topic we consider in detail in a later section.

HIRTH

and POND:

STEPS, DISLOCATIONS

AND DISCONNECTIONS

4753

3. DEFECT CHARACTERIZATION BY CIRCUIT

MAPPING The theory outlined in the previous section enables the Burgers vectors, b,, and step heights, h, for all admissible defects that separate crystallographically equivalent regions of an interface to be predicted when the symmetries of the adjacent crystals and the relationship between their co-ordinate frames are known. In the present work we wish to extend the characterization of defects to include those separating regions of distinct interfacial structures. To do this we use a method introduced by Pond and Hirth [7], which is outlined in this section. The method is based on circuit mapping, but incorporates the possibility of including any type of symmetry operation into a circuit, as opposed to the pioneering treatment [9] which considered only translations. First we describe the notation used for symmetry operations, and then briefly review defect characterization by circuit mapping. Symmetry operations are represented mathematically in this work in the notation of the International Tables for Crystallography [14]. Operations are represented in matrix form w = (W, w) where W is the orthogonal part (rotation, mirror, inversion or identity) and w is a displacement arising either due to translation intrinsically associated with the operation or due to the location of the symmetry element with respect to the chosen origin, or both. In the case of pure translation operations, ?V has the form (I, t) where I represents the identity and t the translation. Symmetry operations are defined with respect to some chosen origin in a crystal; in the present work we choose atomic sites for this purpose. For symmorphic crystals like f.c.c., which exhibit the FmJm, the value of w can be taken as spacegroup zero for all symmetry operations. On the other hand, for h.c.p. crystals, which are non-symmorphic and exhibit the spacegroup PB,/mmc, w = 0 for half of the operations and l/6 [2023] for the other half. For example, the (iOl0) mirror-glide operation relates the central a atom in Fig. 5(a) to the b atom above it in the diagram. If the a atom is taken as being at the origin, i.e. located at [0, 0, 0, 01, then the b atom is related to it by (Mc~olo,,l/6 [?023], i.e. the b atom has mirror related surroundings, as indicated by the operation M<~olo, and is located at l/6 [?023]. A convenient example is the formation of circuits in perfect crystals. We imagine the crystal now to be fixed, and an observer is transported on an excursion through the crystal by a sequence of operations enacted on him alone. If these operations are restricted to symmetry operations, each one takes the observer from one point to another crystallographitally equivalent one, and from the observer point of s view, the crystal appears unchanged at each stage. The starting point of the excursion, S, is located by the vector s with respect to the chosen origin. Initially, the observer is taken to be in his starting orientation (say facing along aI, with - a2 front right,

-a3 front left and co up in an h.c.p. crystal). At the end of the sequence of operations, the observer reaches the point F, located by the vector f, and may also have been rotated and/or inverted. Imagine that the observer has reached some intermediate point in his excursion; the next operation in the sequence, say the ith, has to operate on the observer at this point. If the operation is a translation, we simply have YV, = (I, t). However, in more general cases, the operation may involve rotation or reflection, for example, and furthermore this may act through some point r, other than the observer s location. Thus, in the general case we must write the ith operation as WY, where w? = (I, r)7Vn(I, r>- . (7)

(Such a redefinition does not modify the component, W,, but may change w,, although not in the case of translations.) A complete excursion can therefore be expressed mathematically as the following sequence of n operations:

We refer to % the circuit operation, and this relates as the observer final status to his starting orientation s

HCP

aA
0

b.

??

(a)

*
[oil 01

w
Fig. 5. Schematic projection of sites in an h.c.p. crystal
projected along (a) [OOOl] and (b) [2iiO]. Triangular and circular symbols represent a and b type atoms; large and small symbols represent the level of sites projected along

[2iTo].

4754

HIRTH and POND: STEPS, DISLOCATIONS AND DISCONNECTIONS

and location, specified by (I, s). Thus %(I, s) = (C, cs + c) (9)

where f = Cs + c is the final location of the observer and C is the resultant rotation and/or inversion of the observer frame. When %?= (I, 0) the circuit is closed s in both the translation and orientational senses. Illustrations of such circuits both in perfect crystals and to characterize crystal defects have been published elsewhere [ 151. The method described above for characterizing crystal defects can be extended readily to the case of interfacial defects. However, one must consider the following additional features: (i) the existence of two crystals which may have different symmetries and orientations, (ii) the selection of a bicrystalline reference space and (iii) the requirement that closed circuits around interfacial line-defects cross the interface at two places. These issues are discussed below. We distinguish the two crystals by designating one black (p) and the other white (2); co-ordinate frames and origins are selected for each of the crystals, and these are interrelated by B = (P, p) where P is the matrix which transforms white vectors into corresponding black ones (expressed in the white frame), and p is the rigid-body shift (expressed in the white frame) of the black origin away from the white origin. The characterization of interfacial defects ultimately needs to be expressed in the co-ordinate frame of one of the crystals, and we choose the white frame, arbitrarily in this capacity. A black symmetry operation, %Q),, re-expressed in the white frame, is given by ~+?# (~)@ -l. Defects are defined with respect to some reference space; in the case of single crystal defects, the crystal spacegroup is the most convenient reference. In the case of interfacial defects the reference space is the dichromatic complex [ 161 which can be visualized as the interpenetrating black and white crystals relatively oriented in some chosen way. (Reference bicrystal structures can be created from this reference space by choosing the location of the geometrical plane separating the two crystals, and the relative position of the crystals, as is discussed in the next section.) To characterize an interfacial defect, a line direction 5 is first assigned. A right-handed circuit is then constructed which comprises a white segment and a black segment, and these are linked by two displacements, white to black and black to white, across the interface as depicted schematically in Fig. 6. A starting point, S, is chosen in the white crystal, say close to the interface, and the observer is transported to another white point near the interface, X, at the end of the white segment. The sequence of operations in the white segment is represented by the white circuit operator, %?(A). Next, the observer experiences a displacement (I, p) across

P Fig. 6. Schematic illustration of a closed circuit constructed around an interfacial defect. The line direction of the defect is outwards from the page.

the interface from the white point X to the black point Y. The observer excursion then proceeds by means s of black symmetry operations until he reaches point Z; this black segment is represented by %?(p) which becomes p v(p)B- when re-expressed in the white frame. Finally, the circuit is closed by a displacement across the interface from Z to S. If the interfacial structures in the two regions are distinct, then, in general, so also will be the corresponding displacements p, and pz. However, if the interfacial structures are equivalent, there are two possibilities. The simplest situation is when the two interfacial structures are identical (except for their location); the displacement from Z to S is then simply (I, p)- The second possibility is that the two . structures are related by a symmetry operation, w; the second displacement then has the form (I, Wp)- The total circuit is now mapped into . the dichromatic complex (bicrystalline reference space), and for the simplest case is given by W@, /J) = (I, P)- ~v(~)~- (I, P)%(i) (10)

and the defect in question is characterized by % p)- . The circuit operator defined by equation (9) can be regarded as the irreducible expression of closure failure of the mapped circuit. In other words, it is the unique and most concise description of the closure failure after elimination of all the cancelling components in the initial circuit. Moreover, for defects separating degenerate regions of interface, this formulation can he shown to be identical to the characterization of interfacial defects obtained using the Volterra procedure, and hence %(A, p)- is equivalent to 9,. Illustrations of such interfacial defects characterized using the circuit operator formulation have been published elsewhere [3, 7, 151.

HIRTH 4. APPLICATION

and POND:

STEPS, DISLOCATIONS INTERFACES

AND DISCONNECTIONS

4755

TO F.C.C./H.C.P.

The object of this section is to illustrate characterization of interfacial defects separating degenerate and distinct regions of interfaces using the circuit operator method. As an introduction, the relatively simple case of coherent interfaces between f.c.c. (1) and h.c.p. (,u) crystals has been chosen. The discussion is set out in three parts: in the first, we describe our notation for the diagrammatic representation of f.c.c. and h.c.p. crystals, and then consider the number of distinct reference bicrystal structures; these reference structures are bicrystals with planar interfaces parallel to (11 l)/(OOOl). In the second part, we characterize defects for which h = 0, 1 and 2 [in spacing of the units of d(,,,), the interplanar (111)/(0002) atomic planes]. The third section is a discussion of the decomposition of a pure step in this interface (i.e. b, = 0) with h = 6dtlll) into various sequences of defects with h = 24,,). An f.c.c. crystal is depicted in Fig. 7(a) projected along [l 111.Square, circular and triangular symbols are used to represent the sequence of close-packed planes; following Thompson [17], we designate these a, b and c, respectively. In perfect crystals all atomic sites have identical surroundings, and are hence interrelated by translation operations (I, t). Figure 7(b) is a view of an f.c.c. crystal projected along [ilO]. We have retained the symbol types, but have introduced large and small symbols to represent the

FCC

an
bo
cc1

0
0

n
Z

AAAA
0

I
(b)

? ? ~17~0~0

Fig. 7. Schematic projection of sites projected along (a) [ll l] and (b) [ilO]. and square symbols represent a, b and and small symbols represent the level

in an f.c.c. crystal Triangular, circular c type atoms; large of sites along [ilO].

two different heights of atoms in the a, b, and c layers when observed from this direction. Figure 5(a) and (b) illustrates the corresponding projections of an h.c.p. crystal viewed along [OOOl] and [2iiO]. The plane sequence is . abab along [OOOl] in this case, and hence only circular and triangular symbols appear in the figure. In h.c.p. structures, a and b type atoms are not identical and are therefore not interrelated by translations. However, their surroundings are crystallography equivalent and related, for example, by c-mirror-glide operations. Atomic positions displaced by the passage of a dislocation are designated by adding a superscript to the principal symbol; thus ac, for example, indicates an atom in the a level along [11 11, but displaced by a vector of the type l/6(21 1) within the (111) plane so as to occupy the lateral position of a c type site. Adding a subscript, such as c,, denotes that a plane with c type lateral positioning in the appropriate crystal has changed to the level of a type planes in perfect material. Bicrystals are imagined to be created by placing the f.c.c. (1) crystal above the h.c.p. (p) and separated by an interface parallel to (11 l)/(OOOl) with the directions [ilO] and [2iiO] parallel. The f.c.c. crystal can be terminated by an a, b or c plane, although these terminations are all crystallographically identical. On the other hand, the h.c.p. crystal may be terminated by a or b planes, and these are crystallographically equivalent but not identical. It follows that six candidate bicrystals can be constructed, as illustrated in Fig. 8. Configurations (i) and (iv) can be eliminated directly since nearest-neighbour relationships are violated (although extrinsically faulted interfaces can be formed from them). Thus, if the f.c.c. and h.c.p. crystals [and the (111) and (0002) interplanar spacings] are chemically different, there are four distinct reference structures. If the two crystals are not chemically distinct, the notion of a geometrical plane separating one crystal from the other no longer holds because atoms can occupy sites belonging to both crystals, i.e. coincident sites. Either two layers, a and b, can be coincident, viz. structures (ii), (iii) and (v), or no layers are coincident, viz. structure (vi); it is not possible for one layer only to be coincident. Thus, if the atomic species in the two crystals [and the (111) and (0002) interplanar spacings] are the same, only two reference structures arise since (ii), (iii) and (v) are indistinguishable. To begin the description of the use of the circuit operator we consider the simple case of a circuit constructed in a reference structure. Reference structure (vi) is shown in Fig. 9(a), and a circuit SX = V(1) and YZ = W(p) with p, and p2 also depicted. If 5 is taken to point out of the page, then a right-handed circuit has the sense indicated in the diagram. This circuit has to be mapped into the reference space and, of course, maps onto itself since this reference structure has been obtained from the reference space. The elementary operations compris-

4156
b

HIRTH

and POND:

STEPS, DISLOCATIONS
b c

AND DISCONNECTIONS
000000~0Q0~0 ? ? ? ?Uoumuooou~
AAA~AAAAAAAAA 00000Q0000~0

~000~000~000
? ? outluoumu~u~
0 0 0 0 ~000~000

~AAAAAAAPAPAPA b c

a
b c -A A A A A A a

? ? ~U~UQO~U~U~ -

? ? U~unu~U~u~ AaAaAaAnAaAbA.

~AAAAAAAAAPA~A aA . A .

. A

b a b B b a

?? .

?? .
r

?? . ArAr

??

?? . A

?? -AA-

*..a

.o*o*m*o rArArArA

ArArA
0 ArA e .

aArArA

?? ?? eoeoooe
rArArArArA

?? 00eoeeeoe*e b ~AAAAAAAAAAAAA

*0*0~000oee

?? eoe*eoe*ooe

ArArArArArArA (

(U c

ii)

~u~uuo~u~uo

~A~AAAAAAAAAAA b c 000000000000 0 ~0~0~0~0~0~ -

c a
b c B

??

0 ?? 0 0

~00000~

AAAAAAAAAAAAA 000000000000 0

? ?0~00~0~0~

aAAAAAAAAA&AAA b 0 0 0 0 0 0 0 0 0 0 0 o--

AoAAAAAAAAAAA 000000000000 0

b -

aArArArArA.Ar~ b .o.o.o.o.o.*

??

??

.e.eoe*i-

AAAAAAAAAAAAA
?? e0eoo*e-e.o

aArArArArArArA b

?? o.o.o.e.o.o ?? eo*oeeeoeoe
(

AAAAAAAAAAAAA
?? eeeoeoeoe-

~AAAAAAAAAAAAA b

ArArArAAA.AAA

iii )
00~0~0~0~0~

( iv)
0

c
b c a b c

??

??

??

~0~0~0~

aAAAAAAAAAAAAA 000000000000 0 ~0~0~0~0~0~ AAAAAAAAAAAA 00~00000~000

APAAAAAAAAAAA 000000000000 0 0

??

??

0000~0~

APAAAAAAAAAAA 000000000000

? ? nOeOoOoO~Oll

b a b a

? ?OaOeO~Oo

T.A~A*A~AAAAA~T
b

~~~*~0~~~*-~-

?? e0e.e0e.e.esArArArArAri?? eoeoeoeoeoe rA.rA.rArA

AA
??

ArArArArArA
?? ?? e0eoeoe*e

BAA
b

A~AAAAAAAAAAA
?? oeeeeoeoo*e
ArArArArArArA

aA.ArA
b ?? e*e*eoe*e*e

b a

(VI

( vi 1

Fig. 8. Schematic projection along [T10]/[2iiO] of f.c.c. (I)/h.c.p. (p) bicrystals. The atom layer types, a, b or c are indicated on the left. The horizontal line on the left indicates the interfacial (dividing plane) level if the atomic species in (A) and (p) are distinct, otherwise structures (i), (iv) and (vi) exhibit a unique interfacial location, but in structures (ii), (iii) and (v) two coincident layers of atoms arise; the interfacial level is then chosen, by convention, to fall between the coincident layers as indicated by the solid and dashed lines, respectively, on the right-hand side. Structures (i) and (iv) exhibit a-a and b-b interfacial

structures in which nearest-neighbour

atomic separations are violated.

ing the white segment are translations, and those in the black segment are c-mirror-glide operations and translations. These operations are summarized in

Table

Bl

of Appendix is present

B. It can

be

seen

that

PI = - ~2, and hence %(a, p) = (I, 0), i.e. no defect


character and b, = 0 and h = 0.

HIRTH

and POND:

STEPS, DISLOCATIONS

AND DISCONNECTIONS

4157

Fig. 9. Schematic illustration of a closed circuit constructed in a reference bicrystal [structure (vi) in Fig. 81. When mapped into the reference space this circuit remains closed, i.e. V = (I, 0) (see Table BI in Appendix B).

Fig. 11. Schematic illustration of a closed circuit constructed around a disconnection with b, = l/6 [ii21 and h = 2dtl11,. The atom type symbols on the left and right of the diagram indicate that the interfacial structures on either side of the defect are identical, viz. c: b and ac: b representing reference structure (vi).

Defects separating degenerate regions of interface are depicted in Figs 10 and 11. The first is a pure step (b, = 0, h = 6dCIllJ, and the second is a transformation dislocation or disconnection (b,, = l/6 [ii2], h = 2d,,,,,). The character of these features can be t(l), = [222] and obtained using S?,,, with t(l), = [0003] for the step and t(k), = l/2 [112] and t(p); = [OOOl] for the dislocation, or by constructing circuits as indicated in the figures and mapping these into the reference space. Reference structure (vi) has been chosen arbitrarily in Figs 10 and 11, but the other reference structures would yield equivalent results. The circuit depicted in Fig. 11 is summarized after mapping into the reference space, in Table B2 of Appendix B. The structure of the step surface in Fig. 10 is of interest. Reading the f.c.c./h.c.p. layers from left to right, one sees that two planes, a/a and b/b are continuous close-packed planes, except for elastic distortions, across the step interface. The interface structures for the other planes are equivalent to so-called jog lines [17, 181 normal to the plane of view. These are equivalent to rows of partial vacancies (a/b, b/a, c/a) or partial interstitials (c/b), as is evident in Fig. 10. The jog lines can also be envisioned as partial dislocation dipoles. In the decomposition of a pure step into smaller steps,

discussed later, the formation of partial dislocations can be imagined as a glide dissociation of the partial dislocation dipole or jog line. Next we consider pure steps separating distinct interfacial structures. As described above a pure interfacial step with h = 6dClll) separates degenerate structures, and hence it is only necessary to consider steps with h < 6&,,). We show below that a step with h = 2dCllI) separates the two favourable reference structures (vi) and (ii) illustrated in Fig. 12. Moreover, because of the coincidence of two atomic layers in structure (ii) and our convention that we locate the geometric dividing plane between these layers for this structure, it is not necessary to consider steps with h = dClll)or 3dClll). These latter steps exhibit the same physical structure as the step with h = 2dClll), but where the geometric plane is relocated down and up one (111) atom plane layer, respectively. In addition, it is readily shown that a step with h = 4dull) leads to the b-b structure at the interface, and that with h = 5dflll) to the a-a structure. Because of the

L-.-c+-*-*++*--+ .A.AAA$AAA A .AAAAALAAAAA (

( vi 1

ii )

Fig. 12. Schematic illustration of a closed circuit constructed around a pure step with h = 2dol,) separating reference structures (vi) and (ii) (see Table B3 in Appendix B).

4758

HIRTH and POND: STEPS, DISLOCATIONS AND DISCONNECTIONS

aAa 00 0. A A f

P so0 Z.O

AA

rlA

AA

a b
b a b .+.

**+*a .A 0
AA

a****.
A

a**..
.A.A.A...A\.A

??

rlA
.;.. r:A

a
b

A
.

r/A

&-*-*-*~--.-*-~*-*t~-*-~
AAAAArAAA*A

aArArAr
b =

a b

my.0 AA A,A A A

Fig. 14. Schematic illustration of a pure dislocation with l/6 [ii21 and h = 0 separating reference structures (vi) and (iv). The asterisk locates the centre of the defect s displacement field.

rArA

AAAAAAAA~

( vi)

) ( iv

Fig. 13. Schematic illustrations of a pure step with h = 4&l,) separating reference structures (vi) and (iv); the circuit indicated remains closed when mapped into the reference space.

periodicity in the [lll]/[OOOl] direction, a step with h = + nd,,,,, separates the same interfacial structures as one with h = (+n f 6q)dclllj, where n and q are

integers. For the step with h = 2AClllJ, depicted schematically in Fig. 12, reference structure (vi) is on the left and structure (ii) is to the right of the step. The circuit indicated confirms that the discontinuity has no dislocation character, b = 0; the elements of the circuit are given in Table B3 of Appendix B, where it is shown that %?(a,p)- = (I, 0). However, it is also clear that the two interfacial structures are distinct, since pl # -pz.
The step with h = 4dclllj is illustrated in Fig. 13, and is seen to separate reference structure (vi) from the b-b interface structure. Again, the circuit indicated confirms that b = 0 since %(A, p)- = (I, 0). More-

the interfacial structures are now degenerate because p, = - pi modulo l/2 [Oli]. The shear introduced displaces the f.c.c. planes from a, b, c positions to ac, b, cb positions as shown in Fig. 11. The sequence of black principal letters and white superscripts to the right of the figure is now identical with the sequence of black and white principal letters to the left of the figure. Figure 14 shows a defect for which h = 0 separating the reference structures (vi) and (iv). The circuit indicated shows that b = l/6 [ii21see Table B4, Appendix B. Note that p, # p2, indicating that the structures on either side of the defect are not degenerate. We now consider the decomposition of a pure step with h = 6dClllj into smaller steps and dislocations. The initial step is shown in Fig. 10, and separates degenerate regions of reference structure (vi). Following decomposition into three steps with h = 2dClllj,as shown in Fig. 15, two regions of distinct interface appear, namely the reference structure (ii) corresponding to the step with h = 2dClllj,and the bb interface structure corresponding to the step with

over, the two interfacial structures are clearly distinct because p, # - pz; in addition pz = l/3 [ll l] is seen to be characteristic of a b-b (or a-a) interface structure. Defects separating distinct structures, as described above, may not be stable, and relaxation may arise. When such relaxation involves shear, this can be readily incorporated into the circuit operator formulation. As an example, consider the relaxation of the h = 2d~,,,, step illustrated in Fig. 12 to become the transformation dislocation shown in Fig. 11. If the f.c.c. material above the interface and to the right of the step is sheared by l/6 [l 121the configuration of Fig. 11 is obtained. Moreover, the circuit operator is now modified so that the new value for the displacement from black sites to white across the interface is pi = pz + l/6 [ll??], and hence %(A, p)- becomes (I, l/6 [ii2]). In addition, it is also clear that

(vi)

__

( ii

_ _ (iv)

_ _ (vi)

Fig. 15. Schematic illustration of a pure step with h = 6&l,, decomposing into two terraces of structure (ii) and (iv). Relaxation can transform these into regions degenerate with (vi), as explained in the text.

HIRTH and POND: STEPS, DISLOCATIONS AND DISCONNECTIONS 5. DIFFUSIONAL GROWTH RATES

4759

In terms of the preceding analysis, we can readily transcribe rate equations for diffusional growth at a disconnection. For convenience, the analysis can be separated into that for the step portion and that for the dislocation portion, as illustrated in Fig. 17. According to the crystallographic analysis, h is positive in this case. For an advance of the defect by 6x, the volume change for a length of defect, L, is 6V = hL 6x.
__~-~*- +--)~~~IC~~*. L(P) a b T . . ,A.A.. ..e*.= * *.rAA ..e** . , :

(11)

?? ??

?? b

Fig. 16. Schematic illustration of a disconnection with b,, = l/3 [ill] and h = 2&,1, separating degenerate structures. In this case h(i), = 3&11, and h(p), = 2dcoo02,.

Suppose that both phases are binary with components A and B with XA and Xe atoms of A and B per unit volume, respectively. The region transformed is originally p and finally 1, so the change in the number of A atoms in the volume swept is

/I = 4dclllj. Next we consider possible relaxation of this sequence of structures. The 2dclllJ step may relax to form structure (vi) by expanding a loop of dislocation with b = l/6 [ii21 over the terrace, thereby transforming the first step into a dislocation, as in Fig. 11. Similarly, by expanding the loop with b = l/6 [21 I] over the second terrace, this too is transformed to the reference structure (vi). When the superimposition of these two dislocation loops is considered at the step separating the two transformed terraces, the resulting Burgers vector, b,, is found to be l/6 [lzl]. Thus the total Burgers vector of the three dislocations forming the decomposed step is zero, consistent with the starting configuration, which was a pure step. As a final example, we consider an alternative relaxation of the middle two terraces, shown in Fig. 15, one having the reference structure (ii) and the other (iv), i.e. the b-b structure. It can be imagined that the white b layer could undergo an extrinsiclike shearing (which can be envisioned as occurring by motion of a dipole pair of glide partial dislocations) by l/6 [ll?] to form a b layer, thereby satisfying nearest-neighbour requirements. Such an extrinsic shear would not introduce dislocation character to the step. However, another possibility involving the removal of the white a plane from structure (ii) is illustrated in Fig. 16. Now it can be seen that the two interfacial structures are degenerate, both having the reference structure (vi). Removal of material in this way introduces dislocation character, and the circuit shown demonstrates that the Burgers vector is l/3 [ill]. Since the two structures are degenerate, this can be predicted using the Z$ formulation; substitution of t(l), = [ 11 l] and t(p)t = [OOOl]into expression (1) yields b, = l/3 [ 11 I]. We note that the free surface step heights in the white and black crystals are 3d~,,,) and 2d~000zj, respectively. In this case, therefore, h is equal to the smaller of these, i.e. 2dc0,x,z,, distinct from the value of h,, given as by expression (4) equal to l/2 (3dcIl11+ 2dc00aZj).

(Xi - X:)hL

6x = AXAhL 6x.

(12)

For a diffusion current ZA in atoms/s into the step, with the step velocity v = dx/dt, we have I.. = AXAhLv. Similarly, the countercurrent of B is (14) (13)

Is = AXshLv. For the dislocation normal component

part, the flux is produced by the b,, which is negative for the

Fig. 17. Schematic illustration of step and dislocation portions of a disconnection. (a) The juxtaposition of the imaginary free surface steps in the formation of the defect; note that h(%), and h(p), are both positive with h = h(l),. (b) Motion of the step part by a distance 6x, and (c) motion of the dislocation part by 6x.

4760
Table h + + _ _

HIRTH and POND: STEPS, DISLOCATIONS AND DISCONNECTIONS


1. Material fluxes disconnection b, _ + + and transformation associated motion parallel to nxc portion Step portion p-l p-1 1-P 1-p with

Dislocation p 1 p i

removal addition addition removal

example of Fig. 17(c). (Any component of b normal to the interface arising due to a difference in the excess volume of non-degenerate interfacial structures on either side of a defect, /3., does not lead to a diffusive flux.) Hence the volume of p removed by dislocation motion is
6V= b,L&x. (15)

In this case no i phase is created, so ANA = X;b,L 6x


(16)

and similarly for ANe. Proceeding as for the step case, we find that the current into the dislocation is IA = Xib,Lv and the total current into the disconnection IA = Lv[h AXA + b,XfJ and Ze = Lv[h AXr, + b,Xtf].
(18b) (17)

is
(184

Equations (18a) and (18b) describe the diffusive fluxes for a disconnection moving to the right (i.e. the direction nxg) in interfaces with the (a) crystal above the (p), as in Fig. 17, and where h and b, have opposite signs. When h and b, have the same sign (+ or -), the equations become IA = Lu[h AXA + b.r;] and ZB= Lu[h AXg + b&l.
(19b) (194

diffusional flux at the disconnection is evidently simple. The above analysis can be extended by inspection to multicomponent systems. We emphasize that the results in equations (18) and (19) follow from a mass balance which has physical meaning only for the combined step and dislocation fluxes. The individual fluxes are associated with changes in partial molar volume between the (p) and (A) planes and add to give the proper quantity. This point is illustrated in Appendix C for a diffusionless transformation, for which equations (18) and (19) are nonetheless applicable. To illustrate our analysis we consider the currents associated with defects related to the ones depicted in Figs 10, 11, 14 and 16. Let the 1 phase be pure A, and the p phase be the binary compound AB. In the p phase we take the a-atoms to be species A and the b-type to be B. (The space group of the p phase is now P6;n). For simplicity we take the lattice parameters to match on the (lll)/(OOOl) interface and doll) = d,Oooz, in Fig. 8; it follows that the total as number of atoms per unit volume, N, is the same in the two crystals. Let the atomic fractions of A and B be designated fA and fs. Hence Xi = j:N Xi=fgN and Xt;=fjlN with with fi=l; Ax,=N(f;-ft;)=N/2; ft;=fk= l/2; and Axe = N( f i -f t;) = - N/2. The currents IA and Z,, associated with the movement of the four types of defect are listed in Table 2. A simple hypothetical example of a diffusionless transformation defect is analogous to the disconnection shown in Fig. 11; let both phases be pure A, but the density of p be twice that of 1 because dtlll) = 2d~000z,. The component of b, parallel to the interface, l/6 [ii21 remains the same, but a normal component with b, = dcllu now arises. We have h(l) = 2dclll) and h(p) = 2d~0002,, hence h = d(,,,). and Moreover, AXA = NA- Nfl = -N since 2N. = NJ . Thus, the current ZA [equation (19)] is equal to Lu~,~,~,(--W + N = 0 as required. ) 6. DISCUSSION In this paper we have discussed the topological features of defects in interfaces. Symmetry theory has been used to characterize admissible defects, i.e. those that separate energetically degenerate (crystallographically equivalent) regions of interface. In addition, we have shown how this approach can be extended to include defects that separate distinct regions of interface; this involves taking into account the coordinates of the basis atoms in the adjacent
with illustrated 6. k&Ill) 0 0 0 +1 3 1 0 2 defects IS (LVN&I,) -3 -1 0 -1

The senses of the fluxes embodied in equations (18) and (19) in terms of 1/p material removal and transformation are summarized in Table 1. When the two steps have opposite signs, no overlap exists (i.e. h = 0), and the flux of A is given by IA = Lv[h(l)X~ - h(p)XfJ
(20)

and similarly for Is. These currents can be used as boundary conditions for the diffusion problem of the growth of a train of disconnections [18-201. Once h and b, have been determined, the phenomenological description of the
Table 2. Atomic Defect Step Disconnection Dislocation Disconnection currents

associated

Fig. no. 10 11 14 16

h (&llJ f6 f2 0 f2

HIRTH

and POND:

STEPS, DISLOCATIONS

AND DISCONNECTIONS

4761

crystals. Circuit mapping has been used as an a-posteriori method for characterizing both types of defect. A new definition of the step character of a defect has been introduced in terms of the extent of overlap when surface features on the two crystals are juxtaposed without atomic interactions arising across the interface. The height of such an overlap in the direction normal to the interface has been designated h, the step height. Dislocation character is quantified by the Burgers vector b. Using this framework, defects can be steps, b = 0 and h # 0, dislocations, b # 0, h = 0 or disconnections b # 0, and h # 0, and examples of all these types have been illustrated for both the degenerate and non-degenerate cases. The Burgers vector of an interfacial defect is invariant with the line direction of a defect and the interface orientation, signified by n. On the other hand, the scalar quantity h is not invariant with n, although it can be expressed as a function of vectors t(l) and t(p) that are so invariant. Nevertheless, the present formulation for h leads to the establishment of a simple expression for the material fluxes associated with defect movement. Equations (18) and (19) encapsulate both the flux due to dislocation climb, which is proportional to the normal component of the Burgers vector b, = n-b, and the differential flux associated with step movement, which is proportional to h. Hence, the present treatment explicitly quantifies diffusional fluxes in terms of the topological parameters of interfacial defects. Moreover, we have confirmed that the expression is valid even for the case of diffusionless transformations, i.e. where the overall flux is zero. The separation of the disconnection into dislocation and step portions has several advantages. First, the long-range strain field is associated solely with the dislocation portion. Second, the step and dislocation separation leads to an explicit expression for the divergence of the diffusion flux at disconnections moving during diffusional phase changes, i.e. the diffusional currents of each component at the disconnection. These diffusional currents provide the limiting conditions for the diffusional boundary value problem describing the phase transformation [19-211. Structurally, the separate characterization of the step and dislocation portions facilitates the description of processes that form disconnections or steps from dislocations, or vice versa. A simple example is shown in Fig. 1. Another example of such processes is the emission of a crystal dislocation from an interface disconnection [22]. A third arises in scaling reactions such as oxidation where the interface reactions lead to the possibility of interface dislocations combining to form disconnections that can act as source/sinks for ionic defects and metal vacancies [23]. For the examples selected for the h.c.p./f.c.c. case, the analysis shows that there is a rich variety of interfacial defects that can be classified by means of

crystallographic theory and circuit mapping. The defects fall into two categories: those that show degenerate interfacial structures on either side of the defect and those that do not. In the h.c.p./f.c.c. case, we suppose that one reference interface structure is favoured, analogous to extended dislocations with intrinsic vs extrinsic stacking faults in the parent structures. Since experimentally measured equilibrium fault energies do not differ much in these structures, the favouring of one structure is presumed to occur because of kinetic constraints associated with the more complex partial dislocation arrays associated with structures other than the favoured one. For interfaces between lower symmetry structures or with lower symmetry interfacial habit planes, one would expect larger differences in interfacial free energy to also favour one reference structure. However, the possibility exists of distinct interface structures with similar energies in which case both could be present. The analysis of specific defects for the f.c.c./h.c.p. case leads to pure steps that are six layers high, pure dislocations, disconnections that have steps two layers high and a dislocation portion that is either shear transformational in character or with Burgers vectors normal to the interface, and versions of the latter two-plane high defect in sequences that produce net shear and those that do not. From a symmetry viewpoint it is interesting that a sequence of two-high pure steps (Fig. 15) produces a non-degenerate situation with three different interfacial structures alternating in sequence and that a sequence of l/6 (112) pure misfit partial dislocations (Fig. 14) also produces an analogous non-degenerate sequence. Superposed as disconnections, the two arrays combine to give the degenerate arrays of Fig. 11. The h.c.p./f.c.c. case also provides insight for the role of defects in diffusional phase transformations. Crystal growth can occur in principle by the motion of the six-high step (or multiples thereof) without shear or shape change. Misfit could be accommodated in such a case by the presence of superposed misfit dislocations which would have to climb up the risers of the growth steps to maintain constant structure [24, 251. Figure 15 shows that such six-high steps can dissociate into a sequence of two-high disconnections while maintaining degenerate terrace structures. The individual disconnections have transformation dislocation character but have Burgers vectors that sum to zero in sets of three. Hence the motion of these growth defects also produces no microscopic shear or shape change. Disconnections with such alternate Burgers vectors have been observed by Howe et al. [26]. Crystal growth can also occur by the motion of a sequence of two-high disconnections (or multiples thereof) with transformational dislocation character [ 1, 3, IO]. Such defects would produce shear and shape change. Also, if properly spaced, these defects would correspond to

4762

HIRTH and POND: STEPS, DISLOCATIONS AND DISCONNECTIONS


11. J. M. Howe and D. A. Smith, Acta metall. mater. 40, 2343 (1992). 12. J. P. Hirth, Metall. Mater. Trans. A25, 1885 (1994). 13. F. C. Frank and J. H. van der Merwe, Proc. R. Sot. (Lond.) A198, 216 (1949). 14. T. Hahn (editor), International Tables for Crystallography. Reidel, Dordrecht (1983). 15. R. C. Pond, Interface Sci. 2, 299 (1995). 16. R. C. Pond and D. S. Vlachavas, Proc. R. Sot. (Lond.) A386, 95 (1983). 17. N. Thompson, Proc. Phys. Sot. (Lond.) 66B, 481

structural ledges [26, 271,accommodating misfit and establishing a particular habit plane [28,24]. Evidently, the habit plane could be changed if part of the misfit were accommodated by structural ledges and part by misfit dislocations [12]. Finally, we note that the present considerations also apply to martensitic phase transformations and the related deformation twinning process. As indicated in Fig. 11, the relevant transformation or twinning dislocations have disconnection character. Hence, allowable defects that maintain degenerate interfacial structures can also be determined. Also, by analogy with the alternate shape-change and non-shape-change defects for the diffusional growth case, such transformations in principle can occur without shape change or with varying shape change. In the particular case of twinning, this variation is exemplified by the observations of twinning by pure shear, by pure shuffle, or by combinations thereof [29, 301. 7. SUMMARY Interfacial disconnections with both step and dislocation character are analysed by crystallographic theory. The splitting into step and dislocation components yields a simple description of interface flux for diffusional growth. The symmetry theory is used to classify structures. In some cases there are degenerate structures, and a scheme to define a reference structure is proposed. The analysis shows that a limited number of step/dislocation combinations are possible for the example of an h.c.p./f.c.c. interface. The predicted structures are consistent with experimental observation.
Acknowledgements-The authors are grateful to the EPSRC for RCP financial support and to the National Science s Foundation for the support of JPH under Grant 942 0511. They also thank Mr F. Sarrazit for his assistance with the drawings. REFERENCES

(1953). 18. P. B. Hirsch, Phil. Mag. 7, 67 (1962). 19. G. J. Jones and R. K. Trivedi, J. Cryst. Growth 29, 155 (1975). 20. C. Atkinson, Proc. R. Sot. (Lond.) A384, 107 (1982). 21. M. Enomoto, Acta metall. 35, 947 (1987). 22. H. Gleiter, E. Hornbogen and G. Biro, Acta metall. 16, 1053 (1968). 23. J. P. Hirth, B. Pieraggi and R. A. Rapp, Acta metall. mater. 43, 1065 (1995). 24. H. I. Aaronson and W. T. Reynolds Jr, Scripta metall. 22, 567 & 575 (1988). 25. J. W. Christian and D. V. Edmonds, Scripta metall. 22,
573 & 577 (1988).

26. J. M. Howe, U. Dahmen and R. Gronsky, Phil. Mag.


56, 31 (1987).

27. M. G. Hall, H. I. Aaronson and K. R. Kinsman, Surf. Sci. 31, 257 (1972). 28. T. Furuhara, J. M. Howe and H. I. Aaronson, Acta metall. mater. 39, 2873 (1991). 29. R. J. Wasilewski, Metall. Trans. 1, 1617 (1970). 30. J. W. Christian and S. Mahajan, Prog. Mater. Sci. 39, 1 (1995). 31. J. P. Hirth, J. Phys. Chem. Solids 55, 985 (1994).

APPENDIX

As with the free surface step, one can relate vector character to an interface step [31], although the procedure is not direct. Consider two separated free surface steps as in Fig. Al. The surface steps are assigned translational vectors t, as already described, but now for each crystal relative to the same 5, a procedure differing from that in the main text. We emphasize that this procedure leads to a sign change for t(p). In general a circuit for either crystal could encompass demisteps and include symmetry operations other than translation: in such a case, only the translational component

1. R. W. Balluffi and G. B. Olson, Met. Trans. A16, 529 (1985). 2. J. P. Hirth, Met. Trans. A22, 1331 (1991). 3. R. C. Pond, in Dislocations in Solids (edited by F. R. N. Nabarro), Vol. 8, p. 1. North-Holland, Amsterdam (1989). 4. Proc. Pacific Rim Conference on the Roles of Shear and
D.@usion in the Foundation of Plate-Shaped Transformation Products. Metall. Trans. A25, 1787 & 2555 (1994). 5. J. P. Hirth, J. Phys. Chem. Solids 55, 985 (1994). 6. E. Kr(iner, in Dislocations and Properties of Real Crvstals (edited bv B. Evre). D. 67. Oxford Universitv

7&_

,,,,,,,,, /,

Press, Oxford, U.K. (1985): _ 7. R. C. Pond and J. P. Hirth, Solid State Phys. 47, 287 (1994). 8. J. P. Hirth and J. Lothe, Theory of Dislocations, 2nd edition. D. 22. Krieaer, Melbourne. FL f1992). 9. F. C. Frank, Phil.Mag. 42, 809 (1951): 10. A. H. King and D. A. Smith, Phil. Mag. A42, 495 (1980); A. H. King, Acta metall. 30, 419 (1982).

Fig. Al. Definition of surface discontinuities using free surface sign convention; g is taken to point outwards from the page.

HIRTH and POND: STEPS, DISLOCATIONS AND DISCONNECTIONS w is related to the vector character of the step and can be inserted for t where pertinent. We now imagine that the crystals are relatively translated in the y direction, but not x, so that an interface is created. The two steps become pure interface dislocations with no step character, i.e. b(p) = t(p) and b(l) = t(l) with h = 0 and relative to the same 4 as in Fig. Al. These vectors could be determined by Burgers circuits about the individual dislocations. These circuits would detect any added interface faults associated with the defects as well. Alternatively, the surfaces could be relatively translated in both the x and y directions to form an interface with a single defect, a disconnection. A circuit about the defect gives the Burgers vector b of the disconnection together with another possible defect character such as interface fault structure or disclination content. As for equation (1) but with the sign change for t(p), the vectors are interrelated by
b = t(l) + Pt(p). (Al)

4763 = (I, l/6 [411]) = (I, l/6 1iT21)

(1,P2) u(n, p)-


Table B3. For Fig. 12
W(2) = (I, SX) -

= (I, u6 m

= (I, l/6 (I>PI) U(p) = (I, YZ) = (I, l/3 [l, T?, 13, 9) = (I, l/6 = (I, l/6 (I, P2) = (I, oj Table B4. For Fig. 14 %(/I) = (I, SX) = (I, = (I, (I>PI) U(p) = (I, YZ) = I, l/3 [I, 20, 19,0]) = (I, = (I, (I, P2) = (I, UG., p)-

w 241) [JTT]) [19, 16,271) 13301)

l/6 l/6 l/6 l/6 i/6

[T8, IX, 361 [MT]) [18, 21,791) [222]) [ii21)

However, there is now also a step with height h = h(p). Since previously h was zero, this operation clearly shows that h is not a conserved quantity. One could instead create the same disconnection by allowing the two interface dislocations, created in the first process discussed above, to climb together to form a disconnection. The Burgers vector is conserved according to
b = b(l) + Pb(p) (AZ)

NB: p, = l/6 [zii] module l/2 [liO], i.e. it is characteristic for reference structure (vi).

APPENDIX

consistent with equation (Al). Although h is not conserved, one can define a vector [31] for interface steps that does satisfy equation (Al). For the disconnection, one can envision it to have been formed by merging the surface steps of Fig. Al. A step vector G is set equal to the shorter of the translational vectors, t(p) in the case illustrated. Then for the disconnection
b = t(l) + Pt(p) = t(A) + Pt. (Al)

If one now assigns a common normal vector n to the crystals as indicated in Fig. Al, then the quantities b. and h discussed in the main text are given by n-b and h = n*P1, respectively. In other words, the present procedure leads to the same sign for h and b as that in the main text. Obviously, for the combined defect in Fig. Al, the circuit does not detect the step vector, since a portion of the t or w vectors has cancelled. Hence, the scalar treatment is presented in the main text. However, referring the step height h to the vector e is useful in indicating that the final step may be inclined to the interface normal and thus have components both normal and parallel to the interface. Also, when the parallel component varies as a step is translated along the interface, it relates to the presence of jog lines, as discussed in the main text. APPENDIX Elements B after

Consider a diffusionless transformation occurring by movement of a disconnection as illustrated in Fig. Cl; the normal component of the Burgers vector, b., is negative and the step, h, is positive. According to equation (18) an atomic flux is to be expected for each chemical component of the system. However, in the case of diffusionless transformations we show that the fluxes associated with the dislocation and step portions are equal and opposite, so that the overall flux for each species is zero. We take the interfacial area per atom of species A to be go for both the 1 and n phases, i.e. consistent with the presence of an invariant plane-strain. Conservation of matter requires that

(Cl)
[We note that the volumes gAh(l) and gAh(p) correspond to the perfect 1, and p phases because any excess volume in the interfacial region is represented phenomenologically by p. as explained earlier, and this component does not lead to a material flux.] Recalling that h = h(L) and h(p) = h - b, for the case illustrated, and using expression (Cl) we have b. = - h AXA/X~. By substituting this value into equation (18a) we obtain IA = 0, and similarly for the other species present

of Circuits Depicted in the Figures Mapping into the Reference Space

Table Bl. For Fig. 9 u(n) = (I, SX) (I, PI) q(n) = (I, YZ) = (I, 5 10iiol) (I, P2) V(1, p)- Table B2. For Fig. 11 U(l) = (I, SX) (1, PI) U(n) = (I, YZ) = (I, [0 6611)

= (I, l/6 [f% n5, 301) = (I, l/6 [$ii]) = (I, l/6 [15, 15, 301) = (I, l/6 [411]) = (I, 0) 301) = (I, l/6 PIT]) = (I, l/6 122322,=I)
= (I, l/6 FL n,

Fig. Cl. Schematic illustration of disconnection motion in a diffusionless transformation. The step, h, is positive and b. is negative.

Vous aimerez peut-être aussi