Dislocation patterns in graphite

Dislocation patterns in graphite

JOURNAL A OF NUCLEAR det,ailed patterns analysis observed is single crystal dislocations are ribbons Shockloy fault. given by means in ...

9MB Sizes 3 Downloads 68 Views

JOURNAL

A

OF NUCLEAR

det,ailed

patterns

analysis

observed

is

single crystal

dislocations

are ribbons

Shockloy

fault.

given

by means

in graphite the

type

of

the

1 (1962) 17-66, NORTH-HOLLAND

consisting

of two

by

a strip

can be interpreted

de fines

microscopy

flakes. It is found

lamelles

presentBe;

that the

partials

PUBLISHIXG

of

I1 est montre

que les dislocations

de rubans limit&

Tous de

modele

et en supposant

faults

consisting

stacked

lamella

of

i.e.

rhombohedrally

fautes

stacking

width

a

occur,

material.

The

fault

energy

of the dislocation

of extended vector. the

of

faults

nodes

ribbons

angle

between

with

this can be used to determine

of the ribbon

and

l’aide

de la forma

that

partielles

The

possible

pared

from

with

vectors

observed

were

1200” C annealing

found

to

3000” C and annealed

contain

vacancy

energy,

It

Burgers

vect)or

that

associated

and

with

the

of

the

theoretical

point

as to the nature

dislocations of view.

It

is discussed is shown

loops tend to form preferentially

that

formation

nature

mechanism

networks

sre

of the twin bounda-ies.

and the formation

end

discussed

the

mesures

Dislocation

of constrictions

&

et 8,

composes

de

connus. La depenque forme

son vecteur

du coefficient

de Burgers

de Poisson effe&f.

entre

rubans

sont

et comparees

Les vecteurs

disavec

de Burgers

$, I’aide

Zt 3000” C et reeuit

de boucles

h. l’aide

d’effets

de mesures

emmagasinee,

minstion

de leur

entre

boucles

les

of

qu’il

est possible

vecteur et

movement

La

are

electriques

&

de tirer

mecanisme

une eon-

de la d&er-

de Burgers.

L’interaction “glissile”

11 est montrb

de formation

est

que les boucles

B se former

de faute

de pre-

d’empilement,.

de reseaux

aussi bien que leur structure,

m6me que la nature de joints 17

et de

associe

dislocations

ont tendance

dans les rubans

cations,

discussed.

les

theoriquement.

dues 8. des lacunes f&ence

dues 8, C, trouve

est done

elusion sur la nature des boucles & partir

fault

ib 1200” C

de dislocations

de boucles.

Xl est montre a

as the

in the ribbons

avec

trempe

d’energie

la formation

vacancy

structure

as well

Btendus

de Burgers

possibles

la presence

discutee

dislocation

des

mesuree

des lacunes. Le pie de recuit B 1200°-1300’

vacancy

in the stacking

est

de dislocations

experimentalement

Du graphite

of the from

de nceuds

observees.

determines

rev&

ribbons. The

interactions

precedemment & determination

a conclusion

glissile

by

loops. from

0.it.d.

en empile-

de contraste.

and meas~~rements

loops can be reached. The int,era.ction between loops

sont

The

C, found previously

is therefore

of vacancy

is shown

loops.

Bnergie,

cutbes d’un point de vue theorique

at

ce

I1 est montrti que celle-ci peut etre utilisee

les configurations from

peak at 1200-1300’

formation

Les

making

de

seules des

de couches

d’empilement

pour la d~te~in&tion

Burgers

by

faible

de rubans

du ruban

est verifiee.

are

and com-

The

experimentally

means of eleet~rical measL~remonts stored

ribbons

of view

base

dance de la largeur d’un ruban et de l’engle

Poisson’s

effects.

quenched

is

between point

configurations.

determined

use of contrast Graphite

interactions a theoretical

de

aux vecteurs

la direction

discussed

etre inter-

sur la

que se presentent

faute

and it is shown

ratio.

of

de

de la largeur

an effective

peuvent

rhomboedrique.

L’energie l’aide

of the ribbon

observ&

de groupes

on

direction

is checked,

oomposees

ment

Burgers

une faute d’empile-

satisfa~~te

d’empilement

fautes

t,he

the shape

known

of the width

the

vector

from

and from

of partials

The dependence

the Burgers

is obtained

les r&eaux maniere

du micros-

par deux partielles

et contenant

ment.

stacking

est

cope Blectronique.

pret&

energy

dans

graphite

& l’aide

way on the basis of this model and on the assumption small

de

a 6th effect&e

du type de Shockley

in a consistent

monocristaux

celle-ci

sont con&it&es

of stacking

de

that

on&

CO., AMSTERDAM

Une anslyse detaillt3e de reseaux de dislocations

dislocation

of electron

separated

All patterns

5, No.

MATERIALS

de m&es.

de dislo-

est discute,

de

Res mouve-

fXe

Energie

t3reite

der

der

lijrtgwbr Xnaten Burgers

Voktor.

Bantlbreite richtung &I

und

ergibt md

DW dem

gezeigt, werden

dass kann,

sich

der

an Versetzungsteile~ Winkel dieaar um die

as16 der

Form

ver-

mit bekanntem

Zu~arnrnen~~~

zwischen

zwischen

und dem BU~~FS Vektor

wird

wend&

Stapelfehler

V~rs~t~~~n~sb~nder

wurde

tier

dor

Band-

unt~rs~leht*

Zusammenhang Paisson’sche

Zahl

VWXIL

b&immen.

Methods have been developed in recent years Co observe dislocations directly in the electron microscope lT2J). A number of papers, all discussing results obtained by me&n@ of electron microscopy and diffraction have been devoted to this subject recently 4*5$6).It is the purpose of this paper to give a more detailed account of our obse~~tions. Preliminary results have been published elsewhere 7,s). .Apart from their i~~~rt~a~~e for the mecfianical behaviour of graphite, the dislocations in this material present aspects of a more fundamental interest. In a sense graphite is a choice material to be used as a model for a crystal with ELsmall stacking fault energy and with a singte glide plane. It was therefore thought wo~hwhile to study in detail the geometrical features of such a model. A number of contrast effects, typicsl of parti& were noticed, but they will not be considered here. unless required for the .inb~ pretation.

Specimens were obtained by cleavage of natural single erSystals (originating from di~erent sources: Ceylon, Ticonderoga Limestone (N.Y.) and ~adag~,~~~r. ~ynt~~eti~ graphite prepared in the Philips Laboratories hy the graphiti~~tion of silicon carbide single crystals was also examined. For the qutinching experiments use was made of the Ticonderoga graphite which is e~e~pt~~~nal~~pure_ Graphite is t-l,choice materid for tr~~~sn~i~s~on electron microscopy hecause it transmits electrons very readily and because of the ease of preparation of thin foils. Its great ther~~~al stability and its reasonable heat, and electrical conducti~~it~ along the foil plane make it qnit*e stable under electron irradiation. The observations were made with the standard Philips microscope operated aL LOO ItV. In order to allow ~o~ltrolled tilting of the specimen, a de&e was eo~~str~~~~~~ which made it possible to turn very slowly the spelecimen holder. Since rotation about one axis only is possible in this way, i6 was necessary in a few cases to semount the specimen in another orientation with respect

DISLOCATION

PATTERNS

IN

19

GRAPHITE

Some graphites have however the rhombohedral

to the axis. Graphite flakes have a tendency to fall off the grids ; specimens were therefore

stacking

always

section through this structure is given in fig. 2b.

mounted

between

two

grids,

one

of

symbolized

by a b c a b c . . . a cross

which had a larger mesh size.

Most graphites

3.

as can be deduced from the occurrence of streaks in the X-ray diffraction patterns. This is

The Structure of Graphite The

consists

most

common

structure

graphite

of sheets of linked carbon atoms. A schematic

of a hexagonal

hexagonally

of

view of the structure

stacking

is presented

in fig. 1 in

especially

contain

many

stacking

true for ground and hence deformed

crystals. The amount of rhombohedrally material

faults

is known

It will be evident

to increase

stacked

with grinding.

from our observations

why

this is so. A consideration of the structure suggests immediately that the glide plane should be the c-plane since the binding between successive sheets is weak. The binding within the sheets

Fig. 1.

Schematic

view of the structure of graphite in space.

space, whilst fig. 2a and 2c give two projections of interest. It is clear that the stacking sequence can be symbolized by a b a b . . . . The two positions a and b are indicated in fig. 2c, it is further evident that a third position c is possible. The sequences a c a c . . . and b c b c . . . represent also hexagonal graphite structures.

Rhombohedral

Hexagonal.

cb 1

(a)

,t-‘-9

d’ b‘j

‘Ld

(c) Pig. 2.

Structure

of graphite.

graphite;

on the other hand is strong and breaking of C-C bonds on glide is highly improbable. Models for dislocations which do not involve breaking of C - C bonds will be presented below. Since the c-plane is also a cleavage plane, the foil plane will be parallel to the glide plane and we therefore expect dislocation arrangements in the plane of observation. With respect to glide such specimens should behave more like bulk material than foils, where the glide plane is steeply inclined with respect to the surface.

a...

.

.‘c’;?

..0 .....’ (d

1

a) Cross section of hexagonal graphite; b) Cross section of rhombohedral c) Projection on the c-plane; d) Direction of the Burgers vector,

20

P.

4.

DELAVIGNETTE

AND

types of stacking

AMELINCKX

i.e. one Iayer is in a wrong position.

Stacking Faults in Graphite Three different

S,

faults are

can only be eliminated

This fault by two glide motions,

a priori possible in graphite, if one excludes the

or by the glide of a dipole,

“a over a” stacking.

partial dislocations

The first type of stacking fault would involve only o?ze infraction

against the stacking

it can be represented

rule ;

L

The relative

(1)

a b a c b c b c b ... L-.-l

(2)

a b c a c a c ... I

In an .., a b a b . . . structure these two stacking faults result when either an a or a b layer is shifted into c position. This fault will be shown to occur between the partials of the dissociated dislocations responsible for basal slip ; it can therefore be generated by slip only. The third type of stacking fault finally is represent,ed by t,he sequence a b a e b a 1) a lt . . . or

YII!!+ a b a b c a I) a t) . , .

-II

lattice

planes

of the energies of

stacking

fault energy for faults of the second

type is very small ; we can therefore that all types have a small energy.

We will see later that this type of stacking fault occurs in vacancy loops. The second type contains two infractions against the hexagonal stacking rule, it can be symholized as :

Or

magnitudes

sign and same

these stacking faults are roughly as the numbers of infractions 1: 2: 3. We will show that the

ababaca,cac u

a b

Burgers vector in the adjacent indicated by arrows.

by the sequences

ahababcbcbc OS

i.e. two coupled

of opposite

-

Both faults involve three infractions against the stacking rule. They can be described as the result of the insertion of a c-layer either between “a and b” or “between b and a”. We will see that this type of stacking fault occurs in prismatic loops due to in~rstitials. In a few cases we will meet stacking faults of a fourth type: L+ ... a b a b c b a b (4) IL.-J

5. 5.1.

Different Kinds of Dislocations BASAL

infere

in Graphite

I?ISLOCAT1OSS

The dislocations are invariably parallel to the foil plane and they consist of ribbons of two (or three) partials. Wo will now see how this can be understood. The only observed glide plane being the c-plane, glissile ~sloeations have Burgers vectors in this plane. The shortest among the vectors connecting one atom to the next crystallographically equivalent one are AB, AC and AD, as well as their negatives (fig. 2d). The vectors can be decomposed into two partial vectors, according to the reaction AB + Acr+oB, as demonstrated in fig. 2c, d. In the language of dislocations this means that the dislocations can split into two partials of of the Shockley type. The stacking fault resulting from the displacement over a partial vector like Ao, Ba or Co consists of a lamella in rhombohedral stacking, with two violations of t,he hexagonal stacking rule, as visualized in fig. 3a. Since a rhombohedral variety of graphite exists this stacking fault has a low energy and an observable dissociation into partials actually takes place. as will be demonstrated by our observations. A second type of dissociat#ion is geometrically possible. AC + A@‘+- /E or act + Ao as shown also in fig. 2d. A partial dislocation now brings one layer on top of the second one; we will call this the “a over a” or “b over b” position. It is to be expected that the energy associated

DISLOCATION

PATTERNS

21

IN GRAPHITE

tb) Model of extended 60” basal dislocations in graphite. No carbon-carbon bonds are broken. a) The Fig. 3. dislocation has a Burgers vector AC. Notice the lamella of rhombohedrally stacked material; b) Projection on the c-plane;

the Burgers vectors of the partials are bl and bz.

with this stacking fault is considerably higher than in the first case. We will demonstrate that all observed patterns can be explained without the necessity of assuming the occurrence of this type of stacking. On the contrary in order to explain certain patterns we will have to infer the fact that an “a over a” stacking is avoided. This may be regarded as an indirect proof. We conclude that the second type of dissociation does not take place. A model of an extended dislocation having the required Burgers vectors is represented in fig. 3a, b; it is clear that no C-C bonds are broken. Only deformation of hexagones is required to take up the strain. The presence of two kinds of c-planes i.e. a and b planes within the unit cell gives rise to two classes of partials. For those dislocations located between “a and b” (a underneath, b on top) the dissociation reactions are of the type AB -+ Ao+ uB, i.e. 0 inside; for those located between “b and a” the possible dissociation reactions are of the type AB + (TB+ Aa i.e. u outside. The stacking fault associated with this first dissociation is described by the

sequence ababacbcb... and for the second

type of dissociation

by:

ababcaca... One of these stacking sequences is represented in fig. 3a. The movement of a partial Aa, Bo or Co between

“a and b”

(a underneath,

b on top)

changes the stacking according to the prescription a + b + c + a, whilst the partials aA, oB and aC have the opposite effect, i.e. they change the stacking according to the scheme a + c --f b --f a. On the other hand the movement of the partials like aA, aB and aC between “b and a” (b underneath, a on top) changes the stacking according to a + b + c + a ; the partials Ao, Bo, Ca do the reverse. 5.2.

NON

BASAL

DISLOCATIONS

One could also consider perfect dislocations with a Burgers vector [oooc]. In the edge orientation they are equivalent to the insertion of two supplementary c-planes. Glide on other

P.

DELAVICNETTE

AXD

planes than the c-plane would however involve

S.

AXELINCKS

for such glide. It is therefore doubtful that such

graphy of grown crystal faces 9) and from the study of cleavage faces of graphite 10). These dislocations may play a role in the growth of

dislocations will occur as glide dislocations,

polytypes

the breaking of C - C bonds ; there is no evidence

may

however

form

sessile

(fig. 4a). A dissociation

prismatic

they loops

into two partials with

Partial having

of graphite. dislocations

a component

a Burgers

vector

in the c-direction

with

result

vector c/2 would reduce the strain energy. This

from the insertion

would

c-plane. Consider first the insertion of a c-plane

imply

growth

of one of the vacancy

layers in fig. 4a. Such a process would however give rise to a stacking “a over a” or “b over b” depending on which layer is assumed to grow. There will therefore be a tendency for the two layers to grow simultaneously. Dislocations with Burgers vectors having a component in the c-direction and emerging in the c-face would give rise to spiral growth on this face. There is evidence for the occurrence of such dislocations from studies of the topo-

a-

a

-a

(al

or the removal

as would result on the precipitation of int,erst,itials. The layers will be formed in c-position between an a and a b layer (fig. 681)). A prismatic dislocation loop cont,aining a stacking fault of type three and having a Burgers vector (WO&) will result. The stacking fault of the t)ype threth could be converted into one of type one by the nucleation of a partial, but since t’he stacking fault energy is small the corresponding stress y/b is too small and this will not happen; t*he loop is stable. The removal of a part of the c-plane, as would occur in the precipitation of vacancies would give rise to a prismatic dislocation loop containing an “a over a” stacking. Since the energy of such a stacking fault is too high it will be removed by the nucleation of a glissile partial at the periphery of the loop. After the partial has swept the loop the Burgers vector will be inclined with respect to the c-plane. A cross section through such a loop is shown in fig. 68a. The stacking

--1-a -b-a-at-L-

(b)

Fig. 4.

NOR basal dislocations in graphite. a) Perfect (000~) dislocation; b) dissociation of the dislocation on the c-plane

loops;

of the dissociated

fault within the loop

is now of type one. It, is clear that both interstitial and vacancy loop are sessile II). We can now discuss the possibility of dis-

b-----b-b-____b-a-a--___a-

in (a) into two partial prismatic

of a single

(e projection loop.

sociation for the double loop of fig. 4c. If glissile partials are nucleated and accompany the edges of the growing layers, t,wo concentric loops, separated by an anular stacking fault region. could be formed, as shown in cross section in fig. 4b. The system of concentric loops has a smaller strain energy than the double loop. Concentric loops have been observed (see fig. 5 in B) but it was not possible to decide unambiguously whether they were of the type described here, or the superposition of two independent loops, which may lead to the same contrast effects.

DISLOCATION

PATTERNS

IN

23

GRAPHITE

perfect dislocations

in the face centered lattice.

In order to be stable, a network

should not

contain adjacent nodes of the same type (K or P), no adjacent nodes are further allowed to be both extended

or both contracted.

In addition

aII parallel segments must repel each other and they will usually have the same Burgers vector. In discussing the stability of nets we will also have to stacking

take into faults of

account the presence of different specific surface

energies. Faulted areas tend to shrink in surface as much as allowed by the increase in line energy. 7.

Determination

of Burgers Vectors of Basal

Dislocations

Fig. 5. Loops due to quenched in vacancies after annealing. The insets show some hexagonal loops.

6.

Nodes between Partial Dislocations Basal Plane. Stability of Nets

in the

When using the notation introduced in fig. 2d two kinds of threefold nodes among partial dislocations can be distinguished. Their lettering scheme is given in fig. 6. The corresponding node conditions are oA + (TBf OC= 0 and Aa + Ba + Ca = 0. A distinction is clearly only possible if one adopts a convention concerning the sense in which the symbols are read. We accept the same rule as formulated earlier by Frank 12) for the face centered cubic lattice. When looking out from the node point we read from left to right of the line. The two kinds of nodes given here correspond in fact to the K and P nodes introduced by Frank for

(a) Fig.

6.

The two possible types

(b)

of threefold

nodes.

a) Corresponding to the K node introduced by Frank; b) Corresponding to the P node introduced by Frank.

It has been shown by Hirsch et al. 1) and Whelan 13) that the intensity of the diffraction contrast at dislocations depends on the value of n = b .g, where g is the diffraction vector. For n=O the dislocation does not show any contrast. It is clear that this effect, can be used to determine the direction of the Burgers vectors. A knowledge of the structure then allows one to make a plausible guess as to the magnitude. In the particular case of graphite one knows that glissile dislocations have a

Burgers vector in the c-plane. It is therefore sufficient to determine one diffraction vector for which contrast disappears at a given dislocation. This can be done either in light or dark field. In the first method one tries to find an inclination

of the specimen

for which contrast

disappears at the dislocation of interest, the others being visible. Without changing the position of the specimen, a selected area diffraction pattern is made of the region of interest. Care should be taken to select a region at some distance from the extinction contour whose proximity is responsible for the contrast. This precaution is necessary in order to obtain meaningful results because the kinematical theory, on which the effect is based, may become inapplicable near the extinction contour. In the diffraction pattern the spot which causes the contrast will be the most intense one, and can therefore generally be recognized.

24

P.

Pig.

7.

Successive

families

pattc !rns for each case.

DELAVIQNETTE

of partials

AND

in a network

a) S,tacking fault contrast;

go

S.

out

AMELINCKX

of contrast,.

The

b), c) and d) One set’ of partial

On the ot’her hand it is possible to make a dark field image using a predetermined reflection. Those dislocations for which b is in the plane corresponding to the selected spot will then disappear. In graphite, partials disappear for reflections of the type (1120) showing that the Burgers vectors are Aa, Bo and Ca as predicted by the consideration of $4.1. The set of photographs of fig. 7 shows the disappearance of contrast successively at each family of partials in a network. The insets give in each case the diffraction pattern in the correct orientation with respect to the image. From these observations it can be concluded

insets

show

dislocations

the

cliEraot,ic)n

is ollt of cnontrast,.

that the partials have the screw orientation at the extended nodes, and, further, that the ribbons as a whole have screw character, since the nodes are symmetrical. This distribution of Burgers vectors is represented in fig. 7. In the majority of cases isolated nodes have this orientation. In one case an extended node of edge ribbons was identified; it is shown in fig. 8. Stacking fault contrast occurs when n is fractional, the most intense contrast of this kind is found for (lOi0) reflections i.e. if n= Q or 3. This is demonstrated in fig. $1. In this case two partials can go out of contrast simultaneously in certain circumstances, ap-

DISLOCATION

parently violating of contrast.

PATTXRNS

the g ’ L -I 0 criterion for lack

These contrast

effects,

which

are

IN

25

GRAPHITE

specific for partial dislocations,

will be discussed

in a separate paper. 8.

Width of the Ribbons According

dislocation

to

Read 14) the

width

d of

a

ribbon depends on its character i.e.

on the angle q~between its total Burgers vector and the direction

of the line. The theoretical

dependence, as given by theory is represented by:

isotropic

elasticity

where Y is Poisson’s ratio. This formula has yet been verified experimentally because of di~~ulty of measuring d. F~lrthermore variation of the width as a function of 9

Fig, 8. Extended node of edge ribbons. a) The t,hree sets of partial djsloe~tions are visible; b), c) and d) One set of partials is out of contrast. The insets show the diffracting plane for each case.

not the the can

yield a value for Y. Such measurements are however only possible in those crystal where the width of the ribbon is sufficiently large to allow a direct measurement; graphite is such a crystal. One should however take care to avoid a number of sources of errors. Because of spherical aberration, the electron microscope only gives an undeformed image, and hence a reliable width, in the center of the field. The segments of ribbons to be measured were therefore always photographed in the center of the screen. In view of the one-sided nature of dislocation contrast, and because of the displacement of the dark-line with respect to the actual position of the dislocations,

care must be taken when

estimating the widths of ribbons when different contrast conditions are used. In our measurements these difficulties were eliminated by using one long dislocation line which was slightly curved so that a wide range of orientations was available. The contrast was continuously adjusted, by tilting the specimen, in such a way that the contrast conditions were as constant as possible, at the position of the segment to be measured. Although the ~splacement of the dark line was taken into account, a constant displacement would not cause a serious error Fig. 9. Network showing stacking fault cont,rast,. since it would only produce a parallel displaceThe inset is the corresponding diffraction pattern. ment of the straight line in fig. 10. The method

P.

DELAVICNETTE

ANT)

S.

AMELINCKS

.

Pig.

10.

the total

Plot

of t,he width

Burgers do in eq.

vector

d of a dislocation

and the direction

(5) whilst

ribbon

as a function

of the ribbon.

the slope of t,he straight

used also avoids the eventual lack of reproducibility of the microscope magnification. The Burgers vectors of the ribbons were determined using the procedure outlined above. The diffraction pattern fixes at the same time the orientation of the foil and hence rp. Fig. I1 reproduces at small magnification one of the ribbons used. whilst the insets illustrate the

The

R,bbon

I

of COB 2 q where

intercept

with

line is 2v,/(2---I*),where

p is t,he angle between

the d axis gives v is Poisson’s

the quantity

v&o.

with specific surface energy I’, one obtains (14)

different segments used for the measurements. The insets are all at the same magnification. The error introduced by the curvature of the ribbon is negligable provided this is small enough with respect, to the accuracy of the measurements. The data points for this ribbon are given as black dots in fig. 10 the open dots referring to a second ribbon. A few isolated measure~~ents are also included in the plot. It is clear that the functional dependence is reasonably well obeyed. The value of v deduced from the slope of the st,raight line is 0.24 5 0.04 which is significantly smaller than +. The value for do is do N 0.1 x 10-4 cm it is used to determine the specific stacking fault energy. 9.

The Stacking

Fault

Energy

Expressing the equilibrium separation the two partials, bordering a stacking

do of fault

One of the ribbons used for the measurement Fig. 11. of the width. The insets show some of the segment’s used. Noticft

the visible change orientation.

in width

with

DISLOCATION

PATTERNS

27

IN GRAPHITE

Fig. 12. Contrast at triple ribbon: the three ribbons go out of contrast simultaneously. respectively a T-node and a Y-node are visible. a) Line contrast; b) Stacking fault contrast;

In T and Y c) One set of

part,&1 dislocat,ions is out of cont,rast,.

where do has been defined in 5 6 ; p is the shear modulus and b is the Burgers vector of the partial dislocation. The value of d 0, which is the width of a 45” ribbon, was found by taking the intercept with the d axis in fig. 10; do 1: 10-5 cm. With ,u= 2.3 x 1010 dynes/cm2 15) and b= 1.42 x lo-scmthisleads toy = 3.5 x lo-2erg/cm2. alternatively lowing

one can make use of the fol-

approximate

procedure to obtain stacking fault energies from extended nodes. The interaction between partials is taken into account to a certain extend and the total energy is ~nimized by means of a variational procedure. The specific stacking fault energy is then given by Y=

formula 16)

involving the radius of curvature R of an extended node such as fig. 12 and shown schematically in fig. 13. This relation expresses the equilibrium curvature of a dislocation line under an applied force y per unit length. This relation however does not take into account the interaction between the partials forming the node and it is therefore only a rough approximation. Since R= 1.2 x 10-4 cm we find y = 2 x 10-z erg/cmz. _ _ ^. Siems et al. 17) have indicated a better

KYdYd3

(8)

where K=

pb2 2-i-v --

8n

Fig. 13.

Schematic

1-Y

view of extended node in order

t,o illustrate the notations

used.

28

P.

for edge ribbons

DELAVIGNETTE

AND

and K=

for screw ribbons;

“fusion” and “intersection”. By the first term we mean that the two ribbons

/lb2 2-3v -8n 1-Y

yo is a dimensionless

meet, number

of screw

and we find ZJO=0.3 x 10d4 cm;

this

leads to y = 3.5 x 10-s erg/cm2 in good agreement with the value deduced from the width of the ribbon. Using anisotropic theory this value would presumably increase. + The small value of y explains why rhombohedral graphite is sometimes found, especially after deformation. Heavy shear along the c-plane results in large networks of the type shown in fig. 14, of which practically half of t~he surface area is rhombohedral. 10.

without

reaction

The meaning of ~0 is clear from

iig. 13. The node of fig. 12 consists ribbons

AMELXNCKX

tinguish between

which is y0==4.55 for an edge and 5.95 for a screw ribbon.

S.

Interaction between Ribbons: Fusion Reactions between Simple Ribbons

We will now discuss in detail how ribbons can interact with one another. We will dis-

only

crossing takes

over

place

one

another;

between

the

a last

partial of the first and the first partial of the second ribbon. We will distinguish cases whereby the ribbons are between the same or between adjacent

lattice planes. In the first case they

contain the same type of stacking fault, whereas in the second case they contain different stacking faults. 10.1.

TWO

RIBBONS

STACKING LATTICE 10. I. I.

WITH

FAULT

THE

SAME

AND

IN

TYPE

THE

OF

SAME

PLANE

The. two complete Burgers vectors fu~t~~ an obtuse angle of 120”

10.1.1.1.

The last partial of the first ribbon is the same as the first of the second, e.g. Ao -t-oB and Bo + aC such ribbons always attract

On meeting, the two partials crB + Bo annihilate each other mutually and a single ribbon Acri crc! results. If the two ribbons are however pinned somewhere, or attached to the rest of

(b)

Large network consisting in small stacking

Fig. 14.

fault triangles. In Tw, one twin produces the inversion of the stacking

fault t,riangles.

t According to calculations by Siems, Delavignette and Amelinekx 17) and by Spence 25) the value for y becomes

0,7 erg/cmz.

(Cl

(df

Big. 15. Fusion of two ribbons, containing the same staaking fault giving rise to an isolated extended node. a) and b) the node contains a stacking fauIt of type I; c) and d) the node cont,ains a stacking fault of type II.

DISLOCATION

the network

an isolated

extended

PATTERNS

comes the repulsion due to the central partials.

node forms

As a result

as shown in fig. 15a, b ; fig. 12 is an observed example. This process is not responsible for the formation of the regular networks since it can only

produce

isolated

extended

nodes.

29

IN GRAPHITE

The

narrower

the ribbons

as they approach

become

gradually

one another

and a

complete dislocations CB is formed by the reaction AB + CA --f CB this dislocation immedi-

interior of the curved triangle contains a stacking

ately separates again into two partials and an

fault of the same type as the ribbons giving rise to it. Fig. 15b represents the lettering

isolated

pattern

for two ribbons both of which contain a stacking

if the stacking

fault is of type I and

The but the e.g.

node is formed

as shown

fault of the second type, for example

fig. 15d if it is of type II. 10.1.1.2.

contracted

in fig. 16. The same type of reaction can occur

two ribbons as a whole attract the last partial of the first repels first partial of the second ribbon Aa+ aB and Ca+ aA.

The question arises first whether for parallel ribbons an equilibrium configuration is possible, whereby the four partials remain separated. For example the attraction of the outer partials might compensate for the repulsion of the inner partials. A detailed consideration of the forces acting on the different partials shows however that as the ribbons approach one another their width also diminishes in such a way that the attraction of the outer partials always over-

and UC+ Bo.

The complete Burgers acute angle of 60”

10.1.2.

aB + Ao

The result is given in fig. 16d. vectors form an

The dislocations now repel as a whole and no fusion reaction is possible; however, we will see further that they can intersect under the influence

of a shear stress. The complete Burgers vectors are at an angle of 180”

10.1.3.

The dislocations attract since they are of opposite sign; they will now annihilate each other mutually. 10.2.

THETWORIBBONS STACKING ADJACENT

CONTAINADIFFERENT

FAULT, THAT

IS THEY

ARE

IN

PLANES

10.2.1. The two ribbons as a whole repel

Two different 10.2.1.1.

situations

Formation

may occur.

of threefold

ribbons

In the first situation, the last partial of the first ribbon attracts the first partial of the (a)

(Cl

Fig. 16.

Cd)

Fusion of two ribbons containing the same

stacking fault producing an isolated contracted node the

second ribbon

(b)

two ribbons contain a type I stacking fault. a) to c) different stages; d) stable node.

;

as for example

in Au+ aB and

OC + Aa. Although the ribbons repel as a whole the inner partials of the quartet aB and UC attract. A detailed consideration of the equilibrium condition reveals that a metastable equilibrium position is possible without recombination of the inner partials (see appendix I). This can be understood intuitively as follows. The outer partials are kept at a distance by the repulsive forces between them ; the inner partials attract? but they are bound to the outer partials by the stacking faults and they are therefore kept separated. If the stacking fault energy is small the equilibrium is very unstable and re-

P.

DELAVIQNETTE

AND

S.

AMELINC’KX

a-

-a -b

b-

--P

a-

-a-------a

a-a-

I

l -b~ccb~b-

b-----ba------a

-b

-

b-----ba-a-

-b-b

--a-a al8

(b)

t.31

(C)

Fig.

l’i.

Nodels

of t)hreefold

asymmetrical

ribbonr.

ribbon;

a) cross section

c) Jn projection

of the symmetrical

combination will take place under a small shear stress: a threefold ribbon will therefore result. This ribbon is now stable and will only dissociate again under a strong shear stress in a suitable direction.

ribbon;

showing

on the c-plane,

b) cr’oss srlc:t,ion or tlift

the notation

nsc~l.

the phase shift at the fault plane is 1/3z whilst. it is 213~ for the other half of the ribbon. As will be discussed in a separate paper Is)% this

The geometrical structure of the threefold ribbon is represented schematically in fig. 1Ta. The three dislocations have the same resultant Burgers vector Aa. The central dislocation is in fact the close superposition of two dislocations with vectors aB and OC in adjacent lattice planes. The resultant

dislocation

behaves

like

one with vector Ao. This is shown by the fact that contrast is found to disappea,r at the three dislocations simultaneously as illustrated by fig. 12~. The width of a threefold ribbon is about five times that of a single ribbon. In appendix I the equilibrium separation in a threefold ribbon is discussed. If the two reacting ribbons are part of a network, nodes of the shape shown in fig. 18, may result. We will call them Y-nodes. Observed examples can be found on several of the micrographs e.g. in fig. 12 and 19 and fig. 76a. The different nature of the stacking faults in the two ribbons is strikingly illustrated by fig. 19b. where the two halfs of the t’riple ribbon exhibit a very different contrast. For one ribbon

(I)

a-L-b-+--+--a b!

b-b I-_a-a---a

c-.+--a-+--b

-b-b

b-b--b-b ~-a-a--1

Cdl

(0

Frlsion of two ribbolls containing Fig. IX. st,acking fault, resulting in the pro(lrl’:tion fold ribbon fold b)

ribbon Two

(Y-node). have

the

a tliffererrt of a tlrrvv-

The tlrrce par%ials in t,hcxthrclc:-

same Burgers \-vc+or. a) ant1 c) stable c:onfiguration: different stages; (11 cross section throueh the ribboll.

DISLOCATION

Fig. 19.

Threefold ribbons.

PATTERNS

may lead to significant differences in diffracted intensity. The interacting dislocations need not necessarily be in adjacent planes for t,he formation of the threefold ribbons. If only a few unit cells separate the glide planes, the interaction is probably strong enough to ensure the stability of the threefold ribbons, The level difference between t!he two halves could also give rise to a small shade difference in the stacking fault contrast. This is however not the main origin as mentioned before. The equilibrium distance of the outer partials decreases slightly as the distance between glide planes increases. The meeting

partials repel

If the ribbons repel as a whole and the meeting partials also repel as e.g. in Aa+ oB and oB -t-Crr, no fusion reaction is possible. The ribbons can however cross over and form a network by intersection as will be shown later. 10.2.2.

The two ribbons as a whale attract

The first possibility

31

GRAPHITE

a) line contrast; b) stacking fault cont,rast. Notice the difference in shade between

the two parts of the ribbon. In A is a symmetrical

10.2.1.2.

IN

is now that the meeting

ribbon; in R an asymmetrical

one.

partials attract as e.g. in Atr + aB and GC + Ba. The two inner partials recombine according to oB +Co + Ao; as shown in fig. 20b. This is however only an intermediate stage because Ao can further react with Ba to form a partial oC. Since AGF and aC now repel the final situation of fig. 20d is stable. We will call these T-nodes, because in a more symmetrical form they are T-shaped. A number of examples can be found; one is shown in fig. 12. In all other eases the ribbons do not react by fusion, but they will react by “cross over” or by intersection as will be discussed in the next chapter. The fusion reactions are “unrepeatable”. By this we mean that the reaction can only take place once, because it results in a transformation of one of the ribbons into another one which does not react in the same way as the first, with a set of similar dislocations. Let us take the specific example of fig. 15b. Once the extended node formed, a ribbou ACJ+ aC has now to react with the next ribbon Aa+ aB or Ba+ UC (fig. 15a). However no fusion reaction is possible and so the process

I’.

DELAVIGNETTE

AND

S.

AMELI1JCXX

fdt

A,U

6, B

-a--&b-.--&-~-b-cJb-4&b--&---+”

-b----t_c+b-V a-*1;

b&b-

-a-a.&&.--+ekalc

Fig.

20. Fusion

stacking

of the formation

of an extended

node

takes

place only once. The same applies to the other reactions so far discussed. It is therefore clear that regular networks will not be formed by these processes. We will show that they form by an intersection reaction or by “cross over”. In summary we can conclude that fusion reactions can give rise to four types of isolated nodes. If the two reacting ribbons are in the same lattice plane, and hence contain the same stacking fault isolated extended or contracted nodes can form ; if the two ribbons are in adjacent lattice planes. and hence contain different stacking fault, T and Y nodes can be formed. 11,

Intersection of Two Simple Ribbons

If a fusion

reaction

is possible

it will take

place

of two ribbons

fault. resulting

preferentially

containing

in the production

rather

than

a different of a T-node.

intersection.

It is thus clear that we will only have to consider the cases where no fusion reaction is possible. il.

1.

THE

TWO

BURGERS

11.1.1.

DISLOCATIONS

HAVE

THE

SAME

VECTOR

The two ribbons contain the.sam,e stacking fault

The two dislocations will annihilate as shown in fig. 2 1. The process is a little more complicated than in the case of complete dislocations. On intersecting the region of overlap is stacked “a over a” and will therefore immediately shrink in area. In practice it will not even form at, all. The region of “a over a” stacked material will finally be eliminated completely by the

DISLOCATION

recombination

PATTERNS

of the four partials OC and Ccr,

two independent

ribbons

Aa + oB result.

IN

33

GRAPHITE

annihilate completely

since they are in adjacent

planes. Instead of annihilation

“dipoles”

form.

One of the dipoles consists of Aa and aA; the The

11.1.2.

two

ribbons

contain

a

dif7erent

to a row of interst,itials, the other to a row of

~~a~k~ng Gaels The situation

is represented

ribbons with an extended formation

in fig. 22. Two

jog are formed.

The

process is in fact rather complicated

if analysed locations

in detail.

On intersecting

the dis-

with the same Burgers vector

cannot

Intersection

-a

within

the two dipoles

contains the stacking a b a b c b a b . . . i.e. two III stacking faults (fig. 22~); it will therefore tend to shrink, especially since there is no long range interactions

between the dipoles. In doing this

(C)

(b)

8,a

C/U

sic

U;A b

b-

.--+----a-

---.-a

c-

-?---&--bb-

-b-b x-a

The region

of two ribbons of t,he same Burgers vector and same stacking fault,

(Cl)

--b

vacancies.

(b)

(a)

Fig. 21.

other of aB and Bo. One of them is equivalent

i c

a-r b-

-1) U-C ----b

t 4

a-v b-

-b

b-

-----a

a-

IVJ

Fig. 22.

Interse&ion

of two ribbons of the same t*otal Burgers vector but containing & different st,acking fault.

34

P.

the layer c is transformed normal stacking.

UELAVIGNETTE

ALUD

int,o a, restoring the

AMELINCKX

the ribbons.

The line

energy

of the dipoles,

the two dipoles

En. increases of course with the distance between

combine into a single dipole and rest,ore perfect This can

the tZwo components. In the partio~lIar case shown here the dist,ance is not known ; i41~ is

most easily be seen by assuming that Ba and Aa

found to be about, 1.5-1.7 times the line energy

combine

of a ribbon.

material

into

between into

On meeting

S.

the two

OC whilst

dislocations

of

ribbons.

oB and OA combine

opposite

sign

Co.

liiach

ribbon contains some type of extended jog. It is assumed that in fig. 23 the angular dislocations

resulted from interseot8ions of this

type ; with

every

sharp

bent) in one

of

the

ribbons corresponds a bent, of opposite sense in another ribbon, st,rongly suggesting that “something” has to connect the corner p0int.s. It, is suggested that the ~onne~t,in~ entities are the dipoles of fig. 22. Such dipoles should give a weak or no contrast since the stress fields cancel approximately. From the angle enclosed by the ribbons and the dipoles t,he line energy of the latter can be estimated as a function of the line energy of

11.2.

THE

TWO

AND

ALSO

RIBBONS THE

TWO

REPEL

AS

MEETING

A

WHOLE,

PARTIALS

REPEL

11.2.1.

They contuin the same stacking fad mad they are in the Same latrine $ane for ~xa~~~l~ Ba--toA and (‘n-i OA

New segments will form as indicated by the dotted lines in fig. 24a. The two segments aA will further annihilate each other and give rise to an extended node. The central region contains a stacking “a on a” as will be seen by referring to fig. 24d. In fig. 240 is represented the stacking within the ribbon l&i-. aA along the line UV. The passage of a dislocation Ca on the a-plane causes a change in stacking according to the scheme a -+ 1) -+ o -+ a. The stacking along the cut XP becomes therefore as represented in fig. 24d i.e. the central area is stacked “a on a”. This region will shrink since ‘(a on a” is a high energy stacking fault, and the result is therefore a contracted node as shown in fig. 24b. The reasoning followed above is in fact somewhat idealized in a way. The real sequence of events is more like the following.

On meeting

the two inner partials repel. They cannot cross one over the other since this would create “a on a” stacked material. Since the two ribbons as a whole also repel a shear stress is required to make them react. If the shear stress is such that it causes the ribbons to approach, it may either cause the ribbon BA or the ribbon CA

Fig.

23.

L-shaped

divloeation

bents are presumably

ribbons.

The

connected.

sharp

to recombine. Suppose that the direction of the shear stress is such that BA recombines, and CA does not. The perfect dislocation BA can now cross the partial Ca without di~oulty, since the Rurgers vectors are perpendicular, there is no large elastic interaction and the stacking is not

I)ISLOCATION

PATTERNS

IN

(a)

U

816 -iI_= -b

(b)

CIA

b-b ,,&_-,~:~“,~~

-b_=

v

x

-C---+.--4i--J---C-

c-j-.---c,st_b--c

-a -b-b-b-

-b-b-b-

--B-------a--a-

--a--a-a-

-b-------b-b-

-b-b-b(Cl

Fig.

24.

Y

61A

616 -b--+&-+--b_b--&..-A&~b-

--a,b+---b=c+---a,b CT

35

GRAPHITE

Intersection

Cd)

of two ribbons giving rise to one contracted

affected by the perfect dislocation. Once BA has crossed over CG it can however dissociate into GA -/- Bcr, since this restores perfect material between the two p&Gals. The partials will now be re~stributed into ribbons in a different way than at the start. The result is the formation of an extended and 8 contracted node. The important feature

and one extended

node.

In this case the process is slightly different; it is now CA that crosses over Bo and then dissociates again into Co + GA. The final result is however 11.2.2.

identical.

The two rabble wnta~n a d~~~ren~ stacking fault, e.g. Ao + at2 and oB-+ Aa

The two meeting partials now always attract,

of this process, which is represented in fig. 25 is however that the sam? ribbons are again available for new interactions the process is therefore repeatable. An observed example of nodes probably formed by this process is visible in fig. 26. If the direction of the shear stress is ~fferent, the ribbon CA may be forced to recombine.

(a)

Fig.

(b)

25.

Real sequence of events for the interaction

of fzig. 24. b) BA

(Cl

a) the ribbon

Bo + crA is compressed;

crosses over C; c) BA

dissociates

again.

Fig. 26.

Probable

example

process described in 3 11.2.1.

of node formed

by the

a.nd pictured in fig. 25.

36

P.

At?

DELAVIG

ETTE

AND

S.

AMELINCKX

PIG

-b---+b+-r-

-b+b-b-1-1-a61A (Cl

Fig. 27.

Intersection

of two ribbons containing

but since they are not in the same lattice plane the possibility for crossing over exists and will be easier the farther the glide planes are spaced. After crossing the new segments indicated by dotted lines in fig. 27a will develop. The segments Aa will not annihilate each other directly since they are in different planes; instead, a dipole Aaf- aA is formed as shown in fig. 27b. The area where the ribbons overlap contains a double stacking fault, i.e. two layers in cubic stacking fig. 270. It will therefore be eliminated. This process proceeds by glide of the dipole towards the rest of the dislocations, with the formation of constrictions in the two ribbons as

a different hacking

fault.

illustrated in fig. 27d. The two constrictions are stabilized by the other dislocations present. There is no net repulsion between the dipole and the isolated partial aC, the stacking fault will therefore be eliminated completely. The final situation is described by the lettering pattern in fig. 27f; an isolated example is visible in fig. 76a. The main feature of this interaction is the formation of a threefold ribbon, the three partials having the same Burgers vectors. pu’umerous examples of such threefold ribbons can be found in the photographs, e.g. in fig. 12, 28. The real sequence of events is again

DISLOCATION

PATTERNS

37

IN GRAPHITE

stress, the two outer partials attract. The result will be that the complete

dislocation

AB and

AC, which repel will cross, and after crossing they will dissociate

into Ao + aC and crB + Aa.

We have now reached the situation which was used as a starting point of our previous reasoning, the constrictions

are however

already present

and all that is needed is a slight rearrangement to superpose 12.

12.1.

the dislocations.

Reactions between Threefold Single Ribbons FORMATION IWSION

OF MULTIPLE

Ribbons and

RIBBONS;

REACTION

Quartet ribbons can be formed by fusion of a triple ribbon and a single ribbon for example

Fig. 28. Network formed by the intersection of two sets of dislocations in successive planes,

somewhat different if one considers the actual process of meeting. Suppose that the left ribbon is aB + Ao and the right ribbon Aa+ SC. On approaching, under the influence of a shear

(AU+ Aat Au) and (Ba+ oC) gives rise to a ribbon (Aa + Au + aC -t- oC). In appendix II the equilibrium configuration of such a, ribbon is calculated (fig. 29a). The process can go on inde~nitely and more complicated ribbons can be formed, a possible fivefold ribbon can be formed by combining the fourfold ribbon Ao + Aa f UC+ Oc with OB + Au giving rise to Au + Ao + aC + Au + Ao (fig. 29b). If the reacting ribbons are part of a network,

Fig. 29. Fusion reaction of threefold ribbon with single ribbon of different Burgers vector. a) Formation of a fourfold ribbon; b) Formation of a fivefold ribbon; o) Possible node involving a fourfold ribbon.

P.

DELAVIGNETTE

AND

S.

AMELINCKX

nodes like the one shown formed.

in fig. 2%

can be

We will now discuss a few more typical nodes that may result from such processess.

Let the

threefold ribbon have Burgers vectors Aa t- Ao -t Ao and the single ribbon On meeting (a)

(fig.

(bl

(Cl

30b),

Bo+ GA, (fig. 3Oa).

Aa reacts with Ba and forms which

in turn reacts

with

OC

OA to

lb)

td)

of threefold ribbon with Fig. 30. Fusion reaction simple ribbon with one common Burgers vector (the

of threefold ribbon with Fig. 31. Fusion reaction simple ribbon with one common Burgers vector (the

adjacent

adjacent

stacking

fault ribbons

are of different

type).

stacking

fault ribbons

are of the same type).





% -a-a-

a-,-

_,-_,---.-a-

--b-b-

b--b-

-b-b-b-b-

(I!)

IO)

Fig.

32.

Interaction

of threefold

ribbon

with

simple

ribbon

without

common

Burgers

vector.

DISLOCATION

reform

Ba (fig. 30~). Finally

PATTERNS

Ba reacts

IN

GRAPHITE

with

Aa and forms again UC. The resulting configuration and the lettering pattern are shown in fig. 30d. The same threefold ribbon can react with the single ribbon Ao

and

aA

oA+Ba

(fig. 31a). On meeting

annihilate

one

another

whereas

Acr and Bo react to form UC. The resulting node is presented in fig. 31b. It is in fact equivalent to the configuration

shown in fig. 18 and which

resulted from the fusion of two single ribbons. 12.Z

INTERSECTION AND

SINGLE

0F

THREEFOLD

RIBBONS

RIBBONS

In view of the frequency of such interactions it seemed worthwhile to investigate their geometry. Different relative positions of threefold ribbons and single ribbons have to be considered. 12.2.1.

No common

Burgers vector

We will first analyse in detail the case in which the single ribbon is in the same lattice

Fig.

33.

type

Observed

described

examples

in fig.

32.

of intersection In

A

of the

superposition

of

ribbons.

I’

r’

/

’ /& A

/

Dl’

//

A/

Qf

/

‘I

/

‘_______-_Jf______J

(b)

A_-..-_

/ (d)

(Cl

Fig.

34.

Intersection

of

threefold

ribbon

with

simple

ribbon

having

one

Burgers

vector

in common.

40

.I?. DELAVIGNETTE

plane

as one

threefold

of the ribbons

that

build

ANI)

the

ribbon.

They

have

no

common

Burgers

vector e.g. Ao -!-AC -;- Aa and sB + Co. The interaction is shown schematically

From

contain

through

fig. 32a along XY

this drawing

(fig. 32~) it is

&CO layers

in cubic

stacking.

On intersecting, segments will develop as shown by dotted lines in fig. 32a. The situation is shown in space in fig. 32d a cross section along X’T’ is further represented in fig. 32e. From this it, can be concluded t*hat sweeping of the area of overlap by the dipole Bo + aB will eliminat’e the wrong stacking and restore perfect material. The other region containing the stacking “1) over b” shrinkage it into a the other

leads to more

of the processes

complicated

n~hich \vill now be described

patterns,

described some of

and anslysetl.

in fig. 32a, whilst,

evident that one of the overlapping regions contains a stacking “b on b” and the other regions

Combined patterns

of layers is given in fig. 32b. c:

as a cross section and UV.

13.

AMELIP;CKX

The combination

however

the arrangement

Y.

‘l’he most striking one is perhaps what could he called the “st’ar pat,tern”.

Two examples of

it are visible in fig. 36. The let,tering pattern is reproduced in fig. 37. The correctness of the l&tering pattern is proved by referring to the extinct~ion pat,terns of fig. 36d and e. In fig. S6d dislocations lett,eretl Ao show all lack of contrast,.

can only be eliminated by complete of the hexagonal mesh, t,ransforming eont.racted node. The perfect area on hand will extend as much as allowed

by the increase in length of the dislocations bordering it,. The final situation is represent,ed in fig. 32g and the lettering pattern in fig. 3211. Xxamples of observat’ion are visible in fig. 33.

12.2.2.

C’omrnon Burgers vector

If the single ribbon contains a dislocation which has the same Burgers vector as the dislocations

in the threefold

ribbon,

say Xo,

the result is quite different. If the simple ribbon approaches the threefold ribbon in such a way that the meeting partials repel (fig. 34), no fusion reaction takes place between the Aa dislocat,ions which are in the same plane ; t8hose which are in adjacent planes form dipoles (IX) Aa + aA. In one of the overlapping regions, I. a multiple stacking fault is formed; in II a stacking “a over a” would result. Both st,acking faults will be eliminated. The first by sweeping the dipole Ao-t_aA over the faulted area, the second by shrinking to zero surface. The final result is as shown in fig. 34d. An example of an observation is given in fig. 35.

Fig. 35. ~bservecl exa,mple of intersect,ion described in fig. 34, a) Line corltrast ; b) Partial in tJha threefold ribbons arc out of contrast.

DISLOCATION

PATTERNS

IN

41

GRAPHITE

a) Line contrast of star pattern B; b) Line contrast of star pattern A; Fig. 36. Examples of star patt,erns. c) Stacking fault contrast of the t’wo stars A and B; d) One set of partial dislocations shows lack of contrast in star A; central

e) Another

hexagons

set of partial

in t,he two

dislocations

is out of contrast

st,ars are of opposit,e sign, giving stacking

fault

within

in star A. Notice

an indicat)ion

that the curvature

as to the presence

of the

or not of a

the lrexagon.

-b-b’_b._

(Cl

Fig.

37.

Analysis

b) Cross section

(d)

of the star pattern.

along

XY,

showing

the partial

a) Two interacting extended nodes giving rise to the “star” pattern; the presence of one layer in wrong position; c) Arrangement in space of dislocations;

d) Lettering

scheme

of the star.

42

P.

DELAVIGNETTE

whereas in fig. 36e the dislocations of contrast.

exhibit

S.

AMELINCKX

CO are out

of the stacking fault can nevertheless be deduced

Pig. 36a and b show the full line

from the inward curvature of sides of the central

contrast and fig. 36~ the stacking fault contrast. It’ is clear

AND

that

stacking

the central fault

of the configuration

region

contrast.

does not

The stability

hexagon;

this proves moreover that the specific

stacking fault energy is somewhat larger for the region inside the hexagon

than outside.

is assured by the repulsion

If the glide planes of the two extended nodes

between all the parallel segments of the central

are spaced by more than one unit cell a stacking

hexagon

fault

on the one hand,

and the stacking

faults on the other hand. Alternating the central hexagon

which

contains

~~~

infractions

against

nodes of

the stacking rule is formed. The centra,l region

are of the K and P type

will be smaller in this case, and bhe curvature

and the regions of stacking faults are alternatively of type I and type II, as indicated by different cross-hatching in fig. 3’id. The simplest way in which this pattern could be generated is by the superposition of two extended nodes of a different type, and hence in successive lattice planes, as demonstrated by fig. 3’ia. The central hexagon then contains a stacking fault with two violations of the stacking rule, i.e. the sequence is a b a b a o a b a b. This stacking fault cannot easily be Ginated except by complete disappearance of the central hexagon. The stacking fault energy is however only of the same order of Inagnitude as y; however parallel segments of the central hexagon repel more strongly than the partials in a ribbon and its cross section will therefore be larger than the ribbon width. The stacking fault will not be visible since it only consists of one layer which is in a wrong position; there is therefore no phase difference between waves diffracted above and below the fault plane. The presence

of the sides of the hexagon more pronounced, but again no cont’rast will be observable. Star patterns, without a stacking fauIt in the central hole, were however also observed. The absence of a stacking fault can be deduced from the lack of contrast and from t)he outward curvature of the central hole, which gives it a rounded shape. An example is presented in fig. 36 in B. Although the lettering pattern remains identical, the generation mechanism is different. A configuration of ribbons giving rise to the observed pattern is shown in fig. 38. The reaction involves two fusion reactions of pairs of ribbons in the same plane as described in par. 10.1.1.1. and two intersection reactions of the kind described in $ 11.2.2. The central region is now free of st,acking faults. The intersection of two families of ribbons, for example (Aa + crB) and (oil -t Ccr), containing a different stacking fault, also gives rise to star patterns of this type. The slight variations in shape are due to the varying distance between the two planes. If these are separated by more than c/2 (the separation distance is *z+ +, where n is an integer), the interaction becomes less pronounced and “stars” with short or vanishingly small radial segments result, like the one shown in fig. 36 in B. We can prove this statement by referring to this photograph. prom the pronouuted inward curvature of the central hole in A we conclude that it must contain a fault with more than two violations of the stacking rule; the only alternative is four. The stacking i + sequence is then a II a b c a c b a b. It is easy

Pig. 38. Configuration of ribbons giving rise to a star pattern without a dacking fault in the center.

uu L--II

DISLOCATION

PATTERNS

IN

43

GRAPHITE

to see that this implies that the two planes in which the two parts of the pattern are situated are separated by at least SC (the location of these planes is indicated by arrows). Prom the inter~onne~tioI1 of the two patterns through

contracted

A and B,

nodes C, which are known

to be planar, can be deduced that in the star B too, the two levels are separated by the same distance, i.e. by more than c/2, proving our statement. A somewhat related pattern is shown in fig. 39 and the corresponding lettering scheme is given in fig. 40. The simplest way by which this configuration may result is by a fusion reaction between two contracted nodes of a different type as demonstrated by fig. 40a. The central “hole” is now free of a stacking fault, as can be deduced from the “outward” curvature and its extension at the expense of the stacking faults is therefore only limited by the increase in line tension. The actual configuration of ribbons giving rise to this pattern, is drawn in fig, 41. It is clear that four different kinds of ribbons are involved in the “reaction”, two pairs react by “fusion” giving rise to the threefold ribbons; the other reactions take place by intersection, Left and right of the central pattern the nodes are of different kind as a consequence of the presence left and right of different ribbons, in the vertical set.

Fig.

39.

Pattern

in threefold

ribbon.

As a final example we will discuss the remark-

sidered as resulting from the interaction between

(bl

resulting from the fusion of two contracted nodes in adjacent hole is free of stacking

“hole”

able pattern of fig. 42 of which fig. 43 represents the lettering scheme. The whole can be con-

(a)

Fig. 40.

Sgmmet,rical

faults.

An examplo

lattice planes,

was presented in fig. 39.

The central

dislocations,

with a different

Burgers

vector.

If the two sets are in the same lattice plane, the result> is a netgvork of extended tracted nodes as shown somewhat fig. 44. An observed

example

and con-

idealized in

is represented

in

fig. 45. As the mesh size decreases the curvature of

t,he sides

of

the

staeking

fault

t,riangles

becomes smaller and smaller and finally a net work consisting of triangular meshes is formed.

I R

fC>

three nodes of one type and one larger node of the second Qpe and hence in a different plane. The areas of overlap indicated in fig. 42 by A contain faults of larger specific energy than that, of the type t,wo stacking faults. This can be deduced from the curvature of the lines towards the area of overlap. Such a stacking fault is a natural consequence of t’he proposed model. If the two interacting configurations are in adjacent planes the stacking fault is of the type . . . abab+c’bab uu

. ..

and it contains two violations against the ab sequence, i.e. the same number as the other fault regions, it is however of a different nature and the energy may therefore be slightly larger, However separated

it is probable

that they are in planes

by gc, the fault is then of the type + . ..ababcacbab l__-l L.-..Jli

j.

.._

l..i

and it contains four violat,ions of the s&eking rule. The energy will now be s~~bstantially larger. In neither ease will it exhibit fault contrast, in accord with the observations. 14.

Regular

networks

The networks of ~slocatio~~s are in fact twist boundaries with a. c-rotation axis. They result from the i~terse~tion of two families of basal

st,ur paWorn. The ~&tern results Fig. 42. Nultiple from t’he intoract,ion of one extended nude of a given kind and t.tlree nodes of’ the other f‘auit

contrast;

e) Partiesis

b) Partiats

kind.

a) Sacking

Bo show lack of contrast;

A(r are otrt of contrast fig. 43).

(see notation

in

DISLOCATION

PATTERNS

IN

45

GRAPHITE

(11

%ig. 43.

Analysis in terms of Burgers vectors of the pattern in fig. 42. star;

b) Configuration

(b)



Fig. 44.

a) Lettering pattern of the multiple

of nodes giving rise to the star.

a) Curved extended and contracted nodes; b) Degenerated Idealized regular networks. if either the &aeking fault energy or the mesh size becomes small.

form of (a)

Alternative triangle are faulted. This is shown schematically in fig. 44b and an observed example is visible in fig. 14. If the two intersecting sets of dislocations are in adjacent lattice planes a regular array of nodes of the type shown in fig. 27 results. The corresponding network is shown with its lettering pattern in fig. 46. An observed example is presented in fig. 28. The same network also results from the intersection of a family of threefold ribbons, say with Burgers vector aA + OA + aA, and a set of single ribbons which have a Burgers vector in common with tShe threefold ribbon, say oB+Aa. A related type of regular network consist,s of an array of “stars”, it is drawn in an idealized fashion in fig. 46, a whilst fig. 47 gives an observed example. Its formation necessitates

Fig. 45.

Regular

network

contracted

of curved extended nodes.

and

46

P.

DELAVIGNETTE

AND

S.

AMELINCKX

(b)

Fig.

46.

Idealized

regular

also the intersection

n&works.

a) Network

of stars;

of at least two families of

ribbons containing different stacking faults. The perfectly regular networks are of course the exception; ingeneral “stranger”or “singular” dislocations meander through the nets and cause deviations from the regularity, we call We will present only one them “singularities”. network containing singularities, together with t,he analysis in terms of Burgers vect’ors, as

47.

Observed

example

of a network

of stars.

of

segments

of threcfoltf

ribbons.

summarized in the lettering pattern. In most cases the lettering patterns have been proved by making use of the contrast effects discussed in 97. The network is shown in fig. 48, the analysis is given in fig. 49. The most striking feature is perhaps the presence of a line of discontinuit,y XY across the field of view. Along this line the orientation of the extended nodes changes by

Fig.

Fig.

b) Network

48.

Dislocation

network

as seen

in

stacking

fault, contrast,; it contains a number of singularities which are analysed in t,erms of Hurgers x,ectors ill fig. 49.

DISLOCATION

Fig.

180”

(or 60’).

49.

Lettering

This suggests

pattern

PATTERNS

corresponding

that this line is

either an isolated partial in the plane of the net, or a twin boundary. The perfect straightness points to the second interpretation. It is clear that the partials

of the net interact

strongly

with this line, but nevertheless cross it. This observation is in agreement with the model of a twin boundary suggested by Kennedy 19) and which consists of a symmetrical wall of partials. A number of threefold ribbons is present in the pattern; the difference between the two stacking faults in these ribbons is evident from their contrast. The holes in the threefold ribbons (in A) can be shown to be caused by the intersection with a simple ribbon, which contains no partial with the same Burgers vector as the partials in the ribbon. A star pattern is visible in B.

IN

GRAPHITE

to the observed pattern

15.

shown in fig. 48.

The Formation Mechanism for the Hexagonal Networks

A mechanism for the formation of hexagonal networks containing extended and contracted nodes has been proposed by Whelan et al. z”) and applied to stainless steel. The mechanism involves cross slip and it is therefore a priori inapplicable to graphite, where there is only one prominent glide plane. Moreover there is direct evidence that networks in graphite form by glide on the basal plane only. In fig. 50 for example it is clearly visible that a few nodes have been formed in between the points A and B in the interval between the two exposures. We propose here a mechanism for the formation of such nodes based on glide along the c-plane only. Consider the situation pictured in fig. 51 whereby a ribbon BaA and a ribbon CoA meet

P.

Fig. 50.

Observed

DELAVIGNETTE

AND

stages in the process of node fornlation.

(b) two more nodes have been formed;

c) One partial

c*clntrast, although

S.

AMELINCKX

a) and b) Two succossivo stages;

is ollt, of contrast;

all partials

(b)

between

d) The entire, dislocation

(a) and

1’ is otlt, of

are in contrast,.

,c

)

a) The ljartials CIA and CO combine to mechanism of extended and conkacted nodes. Fig. 51. Formation form a perfect dislocation CA; b) Dislocation CA crosses over the partial Hn; r) ‘l’hc perfcot, dislocation CA dissociates again according to the scheme CA --f Co -+ nA.

DISLOCATION

PATTERNS

49

IN GRAPHITE

as would be the case when a first set of ribbons

have a direction in sector II the process will be

with the same Burgers vector is intersected

slightly Now

by

different. the partials

aA

are forced

which first meet have vectors uA and Co; they repel. But, since there is a driving force for

together to form a perfect dislocation

RA, which

movement

again is possible

a ribbon

of %Ldifferent

kind. The two partials

of the ribbon

CcrA, we can assume

that GA and Co will combine

to form a perfect

is now pushed

Bo

through

Ca. Because of the repulsion

CA as shown in fig. 51a. This whole

oA, BA will now move

~slocation

CA is repelled by the right partial

instead

CoA, there is however

only

a small interaction with the left partial Bo, since vector Bo is perpendicular to vector CA. The partial Ba is further bound to the partial aA by the stacking fault that separated them. It is therefore reasonable to accept that the complete dislocation CA will be pushed “over” or “through” the partial Bo without difficulty, creating the situation shown in fig. 51b. The situation is now such that the ribbon BaA has effectively been reformed at the other side of CA. The perfect dislocation CA may now eventually dissociate again giving rise to the situation of fig. 510 and the whole process can start over again. The process described here requires that the applied shear stress, all dislocations in the correct sense. This restricts the of favourable orientations for the shear to sector I of fig. 52. Should the shear

under move range stress stress

uA

same situation

further

eventually

to

between BA and over oA, but

dissociate

again,

Co-t aA, into oA+ Bo. The

as before

has now effectively

If y is relatively large with respect to the applied shear stress, sector II will extend somewhat at the expense of sector I. Successive stages of the process pictured in fig. 51 can be observed in the left bottom corner of fig. 50. Between the two exposures two more nodes have been formed as can be judged from the configuration of dust partials. The dislocation segment, marked P, and which according to process I should be perfect does in fact exhibit contrast effects which differ from those observed at partials. In particular in fig. 50 (a and d) the segment P is out of contrast, whereas all three sets of partials are in contrast.

Crossing of Dislocations

A

e

u

/I f-.

52. Illustrating the relative directions of partial

Burgers vectors and applied shear stress. The process represented in fig. 51 takes place if the applied shear stress has a direction in the range I. In the range different process takes place,

DISTANT CROSSINGS

We will speak about crossing when the interacting dislocations are situated in the plane

ci

e

//

c

a slightly

Co. This

been reached. It is clear that the kind of process that will occur depends on the orientation of the shear stress and may be slightly on the value of y.

16.1.

Fig.

will

within the ribbon

16.

A

A

e

it

-.1.;~ e*

the partial

since BA is perpendicular

dislocation

aA of the ribbon

and

II

separated by at least c. On crossing in not too distant planes the dislocations excert locally strong forces one on the other. Usually there is a, torque tending to twist them into the anti parallel orientation, where they attract. Many examples of the typical configurations that result from such interactions were observed for ribbons. The behaviour depends of course on the Burgers vector of the dislocations. The vertical set of ribbons of fig. 53 contains ribbons with two different Burgers vectors ; this is

50

Fig.

Y.

Crossing

53.

of two sets of dislocations;

t)he twist

c

DELAVIGNETTE

at the crossing

1,

/

1)

,



S.

AAWELTSCKX

notice

points.

*I(T T T

6’

AND

Fig.

66.

threefold more

I

deformed

crossing

/

Crossing ribbon.

of

one

Sotice

single

that) the

t,han t)hrx t,luxtifold

of single ribbons

thv largest

ribboll

with

single ribboll.

one

ribbon

is

For

t)lrtx

i trrcl~~t’is on 1~11~

/

Fig. Fig.

54.

Lettering

scheme of the crossing in fig.

illustrated

56.

a) Line

Crossing contrast

is ollt

53.

reflected in the different behaviour at the crossing points with the same dislocations. Fig. 54 represents the lettering pattern for this situation. If a threefold ribbon is crossed by a single ribbon the largest deformation is on the single ribbon as in fig. 55.

of ribbons

16.2.

c:LUSE

showing

; b) 0 no set of partial

constrictioll<. tlislorn!iorls

of contt’ast.

CROSSIiYGS

There is in fact) a continuous change in pattern from distant crossing to crossing in adjacent planes. For close crossing the configurations resemble those characteristic of intersection. In fig. 56 e.g. constrictions are formed at every crossing point. The reason for the formation of

DISLOCATION

PATTERNS

51

IN GRAPHITE

-a

b=a-

-b ---a -b ---+-&----b-b’-b X--a -b --a -b b

-b

(bl

(a)

Fig. 57.

LeMering scheme for the patterns of fig. 56.

the constrictions

is to avoid the formation

of

stacking faults containing four violations of the moreover stacking rule ; elastic interaction stabilizes the constrictions. This can be shown by referring to fig. 57 which is a cross section along XY. Suppose that Ba would leave the combinatio~l of three parGals Aa, aA and Ba. The region sweeped by Bo would contain a stacking fault with four violations. Since there is no longer a long range repulsion between BG and GA, because of the presence of Ao, Ba

a) configuration

in space;

b) cross section.

will remain together with the dipole Aa+ aA which is in stable equilibrium when the dislocations are one above the other (at 45”). As a further example we discuss fig. 58 (node A) which resembles a contracted and extended node pair. In reality however a segment bisects the extended node, this segment produces only a very weak contrast because it consists in fact of a “dipole”. The presence of the segment can however be inferred from the “arrow point” shape of the apparently extended node. The analysis is pictured in fig. 59. The line energy of the dipole can be deduced from the angles at the sharp bent; it is found to be roughly 16.3.

equal to the energy THREEFOLD

RIBBONS

of a partial. RESULTING

FROM:

CROSSING

An interesting

pattern

involving

crossing

is

shown in fig. 19. The lettering pattern is shown in fig. 60. The most important feature of this pattern and which requires an explanation, is

Fig. 58.

iVet,work showing arrow point nodes due to the presence of dipoles in A.

the inequality in spacing for one of the threefold ribbons. This can best be judged from the stacking fault contrast (fig. 19b) which gives the true width. From the lettering patterns it is evident that the ribbon Itr consists of the overlapping of two stacking faults of the same kind and hence separated by a distance c (or nc -n: integer). The layer sequence in such i + a stacking is a b a b c a b c b c where the planes I_ III

P.

DELAVIGNETTE

AND

S.

AMELINCKX

(a)

Fig.

Fig.

60.

Lett~ering pattern

59.

(b)

Wxhanism

of formation

illustx-ating the formation

of the ribbons have been indicated by arrows. Such a stacking fault contains four violations of the ab sequence, against two for the usual stacking fault and it has therefore a higher energy. The structure of this threefold ribbon is shown in fig. t7b. The line AG which consists of the superposition of crC and aB is now one of the outer partiafs (fig. l?‘b), instead of the central one. The two stacking fault energies yl and yz are moreover different. From the asymmetry the relative values of the stacking

of “arrow

point”

of the asymmdrical

nodes.

threefdd

ribbon

shown in fig. 1.9.

fault energies can be deduced using formula (I, 10); one finds ~~~~~~=~~5 (see appendix I). The presence of the high energy stacking fault in RI is also the reason why its surface is reduced to a minimum by a displacement of the segment S (see fig. 60). The formation of t*his type of ribbon is possible because crB on arriving at
DISLOCATION

PATTERNS

It is clear that the stacking fault in ribbon Ri gives

contrast

under

the same circumstances

as the fault within the ribbon

Rz, since they

both result from the same net displacement vector R =Aa. The lamella ca does not show up of course.

The equality

of contrast

in Rr

and Ra is evident from fig. 19b. The conclusion is that the same pattern shows the occurrence

of a second

type

IN

53

GRAPHITE

taining two stacking faults of different energy. This is clearly the reason why the two halves of the ribbons A, have a different width as opposed to the threefold ribbons in B of the same figure ; which are of the type that contains two stacking

faults of the same kind.

This pattern shows another feature of interest. In the points marked D (fig. 62) some protruding

of threefold

ribbons, which is however less common, since the combination can only take place under a

A16 --c-1-c--

relatively large shear stress. It is further noteworthy that the energy of the two superposed faults is smaller than the two times the energy of a single fault, proving that some interaction exists between faults in neighbouring planes. 16.4.

CROSSING

OF TWO

RIBBONS

OF THE

6jB

-b&f+b-b-C-b-

b-bc

-l-c-c---

g

-.-,-a--

Y

-b-b-b--1-,-,-

-b-b-bCb,

SAME

TYPE

Under a shear stress two ribbons with the same stacking fault, for example (Ac++B) and (Co -I-oB), in glide planes separated by a distance c (or a multiple of c) may erosa as shown in fig. 6la. Segments will tend to develop as indicated by dotted lines. The region of overlap now contains a fault of the type ababacbacac... III IL--J (fig. 61b), i.e. containing fonr violations against the rule. No mutual annihilation of the segment oB and Bo in the extended part of the node is possible; instead a dipole (Ba+aB) is formed. Movement of the dipole does not change the area of the stacking fault; moreover there is only a very weak interaction with the other dislocations. In an isolated node as in fig. 61~ there is no preference as to which way the dipole will go. In a network it will take such a shape as to minimize the energy, this is roughly the same as to reduoe its length as muoh as possible. The arrangement shown in fig. 6ld is probably the one that is observed in fig. 62 in A. The three dislocations Ao+Ca -+(TB and two other OB now form a short segment of threefold ribbon of the kind discussed in 8 16.3 i.e. con-

Fig. 61.

Crossing of two ribbons of the same type

but in different planes (distant

of nc).

54

P.

DELAVJGNETTE

AND

S.

AMELIIiCKX

this net and acquired

a bent shape.

On the

one hand it generated the segments of threefold ribbon, as discussed above. where

intersecting

(Co-i-(TB),

it created

On the other hand

dislocations the dipoles

of

t,he

marked

set, D.

Also in this case it is possible to estimate t’he line energy of the dipoles. It now t’urns out to be very

small since the angle of the cusp is

hardly visible;

a rough estimate leads to G_ :!

of tjhe energy of a partial.

The much smaller

value found here as compared to $ I 1. I._). is presumably a consequence of the much smaller distance between the component’s of the dipole. 16.5.

Fig.

62.

Network

showing

different

singularities

analyzed in 3 16.4 and in fig. 61. In A asymmetrical and in H symmetrical

threefold ribbons; in 1) dipoles.

parts have been developed at both sides of a mesh, which contains no stacking faults. It is clear that these protrusions should not be stable unless they are connected by some line that stabilizes them. This line cannot be a partial dislocation since all partials are in contrast; from the geometry follows that it cannot be a perfect dislocation either. It is suggested that this invisible line is due to a dipole. From its occurrence across a perfectly stacked mesh, can

,?UPERPOSITION

014’ HIBBOh-S

Some superpositions of ribbons give rise to features which are worthwhile discussing. An example of interest is for instance visible in fig. 33 in D and represented schematically in fig. 6Ya. Ribbon

has a Burgers vector Aa- (TB and ribbon 2 a vector Ca-I-GA. The partial with vector Ca of ribbon 2 is attracted towards the partial Ao of ribbon 1 and is repelled by the other partials. Combination Aa + Co + aB will therefore take place. The partial Co of ribbon 2 is equally attracted by both partials aB, it is 1

be concluded that it has to be a dipole of perfect dislocations. Such dipoles do not change the stacking on passage, as do dipoles of partials. They have therefore no strong tendency to go one way or the other; this explains why it could remain across a mesh. Such dipole results when two ribbons of the same total Burgers vector cross in planes at a distance c ; a drawing similar to the one presented in fig. 59 would show this. The whole pattern of fig. 62 can now be understood. The planar network in the right top corner results from the intersection of ribbons (Ao+ oC) and (Co+ oB) in the same plane. One singular dislocation with vector (Bo+ UC) roughly parallel to (Co + oB) but lying in a plane differing by C in level, crossed

(‘onfiguration of partial dislocations resulting Fig. 63. from ihr superposition of ribbons leading to thr pattern observed in fig. 33 in A. a) Configuration partials to be compared se&on

with fig. 33 in A;

of

b) Cross

of the inkial situat,ion; c) C*o and Au combine

and form aB ; d) flnal situation

: cross section along XY.

DISLOCATION

however

connected

to the left partial

PATTERNS

aB by

IN

55

GRAPHITE

It is clear that the symmetrical

process might

the stacking fault and it will therefore combine

also happen. The first step would then consist

with it; the resulting cross section is shown in

in the reaction

fig. 63d.

Ca+Aa + aB. Fig. 64 shows a somewhat the

pattern

aA + aB + Ca; the second step

is represented

related feature; schematically

in

fig. 65a. It is suggested that in this case we have a superposition

of a ribbon aA + Ca and Aa + aB.

The two partials aA and Aa form a dipole. One of the partials either aB or Co can now glide towards this dipole and eliminate the stacking fault. Say that aB moved towards the dipole, then Ca will be repelled and we will have the situation pictured in fig. 65b. The ribbon will have approximately the normal width. The partial Co may just as well be attracted towards the dipole and then the results would be as shown in fig. 65~. The presence of one ribbon stabilizes in fact the recombination of the other. In fig. 64 there is now apparently a change over from configuration 65b to configuration 65~ as represented in fig. 65a. Superposition may also lead to a narrow threefold ribbon. Suppose that a ribbon Aa + aB

Superposition

Fig. 64.

of ribbons giving rise to cross

over of part~ialn.

UIA AIU

UiB

and a ribbon Ba+ aC combine by fusion into a threefold ribbon Aa+ (Ba+ oB) +aC. The central line is now a dipole and hence does not repel very strongly the two outer partials. These two partials repel however and a ribbon having the width of a single ribbon

will result.

The

central line however should now give very little contrast. The narrow triple ribbon of fig. 19 may be of this nature.

C(U

lil ’ I

Even fourfold ribbons may result from superA model of a fourfold ribbon of this

position.

UIAAla clu UIB (iL)

Fig.

65.

Con6guration

superposition. in A;

b) Cross section

situation;

of

partials

resulting

from

a) Pattern to be compared with fig. 64 of one possible

equilibrium

c) Cross section of alternative

equilibrium

situations.

(8,

Fig.

66.

(1:)

Fourfold position

ribbon resulting from the superof two twofold ribbons.

P. DELAVICNETTE

56

Fig.

67. Different contrasts of the same fourfold ribbon R. The two outerrpartials have the same l3urgers vector.

type is shown in fig. 66. The strong repulsion between the two inner partials aB makes this central ribbon to be larger, than the outer ribbon. The contrast effects observed in fig. 67 suggest that this may be a ribbon of this type. 17. 17.1.

AND S. AMELINCKX

Quenched-in

prismatic loops

QUENCHING PROCEDURE

For these experiments single crystals of pure natural graphite originating from the Ticonderoga Limestone formation (N.Y.) were used. According to Hennig, who kindly supplied the material, these crystals have been heated at elevated temperature in chlorine gas in order to purify them by volatizing the chlorides of eventual impurities. The crystal flakes were mounted in a slit at the tip of an outgassed and purified thin graphite rod and heated in vacuum by means of electron bombardment. The peak temperature was about 3000” C as measured with an optical pyrometer. A considerable amount of sublimation took place, indicating that at the temperature reached, defect formation should be appreciable. The current was then switched off, and the crystal was allowed to cool under vacuum. It was estimated by pyrometry that during the first seconds the cooling rate was about 1000” C/see; which should be sufficient to trap a number of vacancies. When examined directly after the quench no

loops

were found. The crystals were then annealed at 1200” C. For this operation they were enclosed in a pure graphite holder and sealed off under vacuum in a quartz capsule. There was no contact between the graphite crystals and the quartz. After this heat treatment the crystals were re-examined and now circular, slightly hexagonal features situated in the c-plane, as shown in fig. 5, were found. 17.2.

THE OBSERVATIONOF LOOPS. BURGERS VECTOR DETERIWINATION

We will now demonstrate contrast

effects

b

are

in

b

that the observed

agreement

with

the

b

(4

a

(b) Fig. 68. Schematic view of prismatic loops. a) Loops due to the precipitation of vacancies; the Burgers vector is inclined with respect to the c-plane; b) Loops due to the precipitation of interstitials. The Burgers vector is perpendicular to the c-plane.

DISLOCATION

assumption

that

condensation dislocations exhibit

the

are

loops

of vacancies.

PATTERNS

due to the

That the loops are

is deduced from the fact that they

diffraction

contrast.

The best contrast

is obtained when the interior of the loops exhibits stacking fault contrast, fig. 5 is taken in these circumstances.

Some loops are lighter,

others are darker than the environment. other inclinations

For

of the specimens, darker loops

may become lighter and vice versa. On overlap two dark discs may produce a lighter sector, as e.g. in fig. 5 loops A. This behaviour is typical for the diffraction contrast due to a stacking fault parallel to the c-plane. As shown in fig. 68 vacancy loops are characterized by Burgers vectors inclined with respect to the c-plane, whereas interstitial loops should have a perpendicular Burgers vector. The Burgers vector of the loops was determined by using ( 1120) reflection to make dark field images as demonstrated in fig. 69. Those loops which have a Burgers vector not lying in the reflecting (1120) plane, will show up; those which have their Burgers vector in that plane will not show up; but they will for another (1130) reflection. It is clear from fig. 69 that most of the loops show up in a dark field line contrast, proving that they have an inclined Burgers vector and

57

IN GRAPHITE

17.3.

DISCUSSION

OF RESULTS

of the loops

Most

although

are single and circular,

some of them are slightly hexagonal.

In fig. 5 loop B two concentric are seen, the central contrast

as the environment.

coincidence,

the same

This may be a

and the feature would then simply

be the superposition also possible

circular loops

part exhibits

of two loops. It is however

that the central region is perfect

and that we have a loop of the kind discussed in fig. 4 and which dissociated into two partial prismatic loops. It is difficult to distinguish between both possibilities. From the fact that an inclined Burgers vector is observed it can be concluded that the “a over a” stacking has a large energy. If this were not the case the Burgers vector would be perpendicular both for interstitial and for vacancy loops. An implication of these observations is clearly that at about 1200” C vacancies seem to become mobile: the annealing stage observed at about 1200-1300” C by means of electrical measurements 22) and as a release of stored energy 23) is therefore 18.

probably

due to this process.

Interaction between Glissile Dislocations and Sessile Loops due to Point Defects

hence are vacancy loops. Similar observations confirming our point of view have been published

The interaction presents two aspects: on the one hand glissile dislocations are pinned by

recently

prismatic

by

Williamson

and

Baker 21). The

authors have more over been able to identify interstitial loops in irradiated material.

loops

in their glide

plane ; on the

other hand vacancy loops tend to form preferentially in the stacking fault ribbons of extended dislocations. We will discuss both points. 18.1.

PREFERENTIAL

LOOP

NUCLEATION

IN

RIBBONS

Fig.

69.

image

Quenched-in

but

inverted

dislocation

contrast;

loops.

b) Dark

a) normal field image

using the (1120) reflection. The dislocation lines show inverted

contrast except for one loop, which has its

Burgers vector in the (1120)

plane used.)

A cross section through a stacking fault ribbon is shown in fig. 70. As opposed to what happens in the perfect crystal, precipitation of vacancies in a plane marked by a rectangle in fig. 70b, within the ribbon, does not give rise to a stacking “a over a” but instead produces immediately the low energy stacking fault of type one without requiring the nucleation of a partial. One can therefore conclude that

P.

DELAVIGNETTE

AND

9.

AXELINCKX

--a-a-a-

W

(a)

I

i

-b-a+b_a+c

-.+-aa’bII r--___71 I ,L__;_r;IR

-b

s-a -b

-b

a

-i)

c-ab-

-b-a -c-v “-l-/b

c-

-

-ii

a

a-

-b

b

b-

-a-* -b

70.

a) Precipitation

of vacancies

of the high energy stacking in the plane

surrounded t,hrough

4-

inside a ribbon

by a rectangle

without XY;

diffuse

the

produces

producing

point

defects

towards

the

dislocations through long range elastic interaction. We will further present direct evidence that preferential nucleation in extended dislocations does happen. We will now show that such loops effectively pin the dislocations. Fig. 70 represents the geometry of a loop formed within a ribbon; on expanding the edges of the loop will soon meet the partials of the ribbon and form segments like AR and CD which have an inclined Burgers vector AT, BT or CT. The other segments of the loop AC and BD, have a perpendicular Burgers vector (OOOq’2). Cross sections through the loop are shown in fig. 70~ and d.

a vacancy

a stacking

d) Cross section

vacancies will find it energetically more favourable to precipitate inside the ribbon, rather than in perfect crystal, and hence there is an interaction energy, which is to be added to the Cottrell type interaction. The latter tends to make

b-

td)

fault,; b) Cross section RS through

the loop along

b-

b

(b) Fig.

b-

a high

loop within

fault ribbon,

energy

for removal

oan precipitate

fault;

c) Cross scct.ion

stmking

t~hrough the loop

necessity

vacancies

along

Uf’.

The partials are pinned to the loops not only because of elastic interaction but also if they should det.ach themselves from the loop by crossing it, they would have to transform the stacking within the loop into “a over a”. The latter process makes it difficult to cross the loop. 18.2.

PIXNIXG

OF GLISSILE

DISLOCATIOM

BY

LOOPS A partial, approaching a vacanoy loop from outside on a plane marked by a rectangle in fig. 70b, would also have to transform the stacking within the loop into “a over a” and will therefore be hampered in its movement by the loops. Recombination of the two partials may take place before cutting through the loop; this would avoid creating a bad stacking. Elastic interaction between glissile partials and prismatic loops may lead to further pinning. Such interactions have been described in detail for

DISLOCATION

PATTERNS

perfect dislocations

in zinc II), with some modi-

fication

be applied

they

can

to partial

59

IN GRAPHITE

dis-

locations in graphite. We will discuss now the different possibilities which are shown schematically

in fig. 71 and 72.

In noting the Burgers vectors

of the vacancy

loops,

the

we will only

consider

component

parallel to the c-plane. The glissile dislocations are supposed to be in one of the two planes that emerge in the loop. If they are situated in more will corredistant planes the interactions spondingly be weaker.

-

The vacancy loop is in the same plane

1X.2.1.

(fs inside) In

fig.

(b)

‘ila, the ribbon

as a whole is repelled

r

by the loop, whereas in fig. 710, the first arriving A

B+AU

0

-

UA

0

-B+u

c

t-

(Cl

Fig.

72.

Interaction

with the loop

c) Reactions

-

68

A

with

the

first

between

vacancy

loops

a) The ribbon as a whole is repelled; b) The horizontal component of the Burgers vectors of the loop is annihilated by the first arriving c) The fist second

arriving partial is repelled by the partial

narrowing

of the

and

dislocations.

the

component

Burgers vector of the loop is annihilated by the first arriving partial, transforming the loop into one with a vertical Burgers vector. The second arriving partial interacts weakly with the loop since the Burgers vectors are perpendicular.

(C)

loop,

but

strong shear stress, the

In fig. 71b, the horizontal

artial;

partial,

ribbon can pass the loop, the second partial will be held back, leading to a local widening of the ribbon. -

glissile

arriving

of the second one.

If, under a sufficiently

Interaction

and

partial is repelled, but the second is attracted, resulting in a local narrowing of the ribbon.

(b)

71.

loops

B

0

Fig.

vacancy

; b) Repulsion of the ribbon as a whole ;

repulsion

al B+Ba

between

glissile dislocations. a) Both partials react successively

(a.1

A6

-

a

is attracted

of the ribbon.

resulting

in

18.2.2.

The vacancy loop is in an adjacent plane (0 outside)

The situation represented in fig. 72b, leads to repulsion of the whole ribbon. In fig. 72a, on the contrary, both partials react successively

60

P. DELAVIGNETTE

with the loop, pinned.

and the ribbon

In fig. 72c, only

reacts with the loop, repelled.

will be firmly

one of the partials

the other partial

being

AND

S. AMELINCKX

18.3. OBSERVATIONS

Direct evidence ential nucleation

was obtained

for the prefer-

of loops in the stacking fault

In the latter case, a reversal

of the

ribbon of extended

shear stress would result in movement

of the

$ 18.1. Fig. 73 shows a linear arrangement of loops. Although the dislocation ribbon itself is

partial Aa, the partial aB being held back. From the foregoing discussions it will be clear that vacancy the movement

loops are effective of dislocation

in hampering

not

visible

the

dislocations

linear

as discussed in

arrangement

strongly

suggests its presence.

ribbons either by

repulsion, by reaction, or by holding them back after passage. In this discussion we neglected the additional pinning caused in certain cases by the necessity of creating a high energy stacking fault. as pointed out in a previous paragraph. Since the interstitial loops have a Burgers vector which is perpendicular to the Burgers vectors of the glissile ribbon, the elastic interact’ion will be smaller. Furthermore on arriving at the interstitial loop a partial can always choose a plane, either above or below the loop so as to avoid the formation of a high energy with stacking fault. The elastic interaction

19.

Twinning

in Graphite

It is well known that graphite forms mechanical twins very easily, the composition plane being {liol). Whereas the layer sequence in one crystal is a b a b . . ., it is a c a c . . . in the twin crystal. A model for the twin boundary has been proposed by Kennedy 19). It is made up of a pure symmetrical tilt boundary consisting of partial dislocations one every two layers. As shown by Kennedy this leads to the correct

ribbons in more distant planes is comparable to that caused by vacancy loops. Direct reactions. leading to effective pinning, is expected to be less pronounced for interstitial loops, since the Burgers vector of loops and partials are mutually perpendicular. The tendency for interstitials to precipitate in dislocation ribbons is weaker than that for vacancies because of the small specific stacking fault energy

in graphite.

With

an interstitial

loop formed in a position, between the planes c and b within the ribbon of fig. 70 would be + associated a stacking fault b a b a c a c . . . i.e. containing one violation of the stacking rule. On the other hand when formed in perfect material the interstitial loop would contain a stacking fault . . . a b ac b a b with three infractions against the stacking rule. On this basis a weaker interaction is to be expected than for vacancies.

Fig.

73.

presumably

Linear

arrangement

nucleated

of

preferentially

prismatic

loops

along a ribbon.

DISLOCATION

orientation

difference

between

PATTERNS

the two

com-

ponents of the twin. No direct experimental evidence for this model was given however. The direction of the twin boundary required

is also as

by the model, i.e. perpendicular

Burgers vector

to a more

direct evidence that the model is indeed correct. shows a network

to fig. 14, which

of extended

and contracted

nodes lying on both sides, of a strip of twin TW and also within the strip. It is therefore reasonable to assume that the whole network resulted from the interaction between the same two families of dislocations. It is clear that the stacking fault triangles within the extended nodes differ by 60” (or 180”) in orientations in both regions. Since the two twin crystals differ in orientation one might object that this is due to a trivial contrast effect, i.e. that one region exhibits inverted contrast which causes inversion of the triangle. This is however not the case ; a careful inspection of the triangles reveals a slight curvature

61

GRAPHITE

indicating unambiguously

which side the stack-

ing fault is on. We will now show that this behaviour

is just

what

of the

one would

twin boundary that

of a partial dislocation.

We will present here some additional, We refer for this purpose

IN

the

model

dislocations boundary, locations

expect

is exact.

if the model

First of all it is clear

allows

through interaction

the

the

propagation

boundary.

with the boundary

is of course to be expected,

does not prevent

At

of the dis-

but this

penetration.

The same ribbon which in crystal II (fig. 74) is for instance between the planes a and b will be between planes a and c in the second crystal. The consequence of this is that dissociation into partials will be different in both crystals if in both cases a low energy stacking fault is to be formed. A dislocation ribbon, say with Burgers vector AB will dissociate in crystal I according to the scheme Ao+ oB whereas in crystal II the same dislocation will dissociate according to the scheme Ao’ + o’B or referred to the same reference triangle aB + Ao. The notation will therefore have to change when crossing the twin line. \c

A/

C

6

I

B

\\

Fig.

74.

Analysis

orientation

of the influence of a twin on the presence of a triangular network as seen in fig. 14. The of the nodes formed

from the same ribbon in crystal I and II differs by 180”.

62

P.

Fig.

shows

74

the intersection

which leads to the formation contracted lettering

DELAVIGNETTE

of ribbons,

of extended

nodes in the two crystals. pattern

AND

and

In the

we have taken into account

the changes in notation

at t,he twin line. Using

the results of 5 9.2.1 these intersections

will give

rise to the nodes also shown in fig. 14. It is now

clear

orientation

that

these

exhibit

the

required

difference.

The model as presented

S. AMELINCKX

the loop is evident from the successive stages shown. Small particles apparently lying in the surface locally

of

the

foil,

dislocation

are sufficient

movement

to

inhibit

as demonstrated

by fig. 76. Depending on the direction of the shear stress the moving

ribbons

This is clearly

are widened

or narrowed.

due to the following

effect.

If

the shear stress tends to move the two partials here can in fact be

in opposite

site sense, away from each other.

considered as a kink band; the two twin boundaries always occur in pairs of opposite sign. It was further found that the angular

the partial on which the resolved shear stress is largest will impose the sense of motion of the ribbon, the other partial being pulled in the

differences are not constant’. The observation can also be considered as evidence that the high energy stacking fault i.e. “a over a” will not form. If this were the

same sense because of the stacking fault,. ‘This is a case where widening occurs. The direction of the shear stress may also be such as to push the partials together. The sense of movement of the ribbon will again be dictated by t’he partial which has the largest resolved shear stress. The ribbon will now be narrower than normal. In fig. 77 the ribbons are clearly

case the dissociation scheme would be the same on both sides of the twin and the orientation difference would not be observed. The change in dissociation scheme at the twin line can also be deduced from the singular node observed at this line. Evidence was found that the same dislocation line can propagate through a twin boundary in accord with the model. 20.

Movement Sources

of

The dislocations

Dislocations are extremely

-

Dislocation

mobile along

the c-plane, proving in the most direct way that this is the glide plane. After a few minutes of irradiation with the electron beam the dislocations sumably

start moving almost inevitably, preas pointed out by Hirsch 24), as a of the stresses caused by the consequence deposition of a carbon film. The edges of a crystal flake or cleavage steps are the usual nucleation sites for dislocations. No interval sources of the Frank-Read type were ever noticed, although the circumstances are optimum as far as we restrict ourselves to thin foils. 3Iovement of ribbons is relatively smooth; at the stress level caused by the electron irradiation the movement is slow enough to be followed visually. Fig. 75 represents a typical example of “edge sources”; the expansion of

widened. This figure is remarkable for another reason. In the points marked C the ribbons are locally much narrower. They have apparently constrictions. Such constrictions can e.g. be caused by jogs resulting from the intersection of t,he ribbons with a dislocation having a Burgers vector with a component in t,he c-direction. These constrictions apparently move along the ribbon as can be judged from their displacement between the two successive expos~lrc~.

Appendix

I

EQUILIBRI~,~ THREEFOLD

SEPARATIOS

OF

PAHTIALS

IN

.i

RIBBON

We will introduce

the notation

and

where p is the shear modulus, b the Burgers vector of the partials and v Poisson’s ratio. The

DISLOCATION

PATTERNS

1N

GRAPHITE

Example of an “edge source” a, b and c are photographs taken at intervals of time of approximately Fig. 75. 30 seconds. The movement is probably caused by the stress due to the carbon film which forms on t,he specimen as a consequence of the decomposit,ion

Fig.

76.

induced beam.

Example by

of the movement

changing

a) Medium

b) High intensity,

t’he intensity

int,ensity

of dislocations of the electron

of the electron

all the dislocations

the right side; c) Low intensity,

beam;

are moving to

all dislocations

are

moving to the left side; d) Again high intensity,

all

dislocations are moving again to the right side of the photograph.

of organic vapours.

Fig. 77. Example of an “edge source”, showing constrictions in C, probably caused by jogs. Notice that the constrictions

move

along the ribbons.

64

P.

DELAVICNETTE

AND

S. AMELINCKX

repel. A small decrease will cause the combination of the inner partials. (a)

After recombination threefold ribbons are formed (fig. 17), the three partials having the same Burgers vector. If yi =ya we have also x = y from symmetry considerations. The equilirium separation

(b)

Fig.

78.

threefold bination ribbon

Two

ribbons

ribbon, of

before

to illustrate

Burgers

vector

of edge character;

combination

notat’ions

a

used. a) Comto

a threefold

b) Combination

of Burgers

vectors leading to a threefold

leading

into

is then given by the equation

ribbon of screw character.

equilibrium condition can then be formulated by expressing that the forces on all dislocations are zero. We will first consider the combination

where A = a sin2 CJJ +p cosz q, CJY being the angle between the total Burgers vector and the direction of the ribbon. In the pure edge orientation v== 90” and x=i~~/y whilst in the pure screw orientation v=O” and x=3 ,/?/y. As compared to the corre-

of Burgers vectors shown in fig. 78a which would lead after combination of the inner partials to a threefold pure edge ribbon. The

sponding single ribbons the threefold ribbons are about five times larger. If the stacking fault energies yi and y2 are different in the two halves of the triple ribbons.

equilibrium

the equilibrium

Lx 1 -2v+w

conditions

-_;&p& I

1 a%+?

are

:y&+y=o

A [y-1+(~+y)-11-y2

(L2)

where v and w have the meaning illustrated in

is

v = 0.55 x 10-4 and w = 0.05 x lo-4

,U= 2.3 x 1010 dyne/cma,

of Burgers Another simple combination vectors is possible, which would lead to a pure screw threefold ribbon (fig. 78b). In this case the metastable equilibrium configuration is given by v=O.4xlO-4 cm, w=O.14~10-4 cm. These equilibrium configurations are metastable. A small increase in spacing will make them separate further since as a whole the ribbons

(1,5) (W)

x=3A / (2yl-yz+g)

(L7)

y=3A

(1.8)

/ (2yz-yl+g)

with (1,s)

g=Vy1”+y22-y1y2

assuming y = 3 x 10-Z erg/cma; b= 1.42 A and v=Q.

=o

(L4)

This set of three compatible equations with two unknowns can easily be solved:

fig. 78. This system of equations can easily be reduced to a system of two quadratic equations of which the only real solution

become

A [-x-l-(x+y)-l]+yl=O i A [x-1-y-11+y2-yl =o

(Ll)

,;;I-+;&w-Y=O

conditions

The ratio yl/ya can be expressed directly in terms of observed quantities x and y. From (1,4) and (1,5) this ratio follows directly when putting y/x = r Yl _- r(:!+r)

yz- 2r+

1

(1. IO)

Applying this to the ribbon of fig. 191) gives for r N 2; y1/y2 N X/5 which is about what one would expect.

DISLOCATION

PATTERNS

IN

Appendix II

Appendix III

The equ~brium separation of partials in a fourfold ribbon of the type shown in fig. 79 can be obtained by solving the set of equations

CURVATURE FAULTED

65

GRAPHITE

OF

A

PARTIAL

The radius of curvature of the partial is

pb2 %2

3011

1

-+T ( 2‘u

Vi-W p

p1 4 v+w

B1 4 2w

61 --=

4 21--w

()

TWO

REQIONS

&4=

3&l ___------4 ?J--w

SEPARATING

-

rd

(III, 1)

and hence if d
1 -l-+ u-w >

+;? (

---- l 2u

l +_L

V-EW

v-w >

A -y=()

I

El:”

which express that the net force on each dislocation vanishes. The solution of these equations using the same numerical values as in appendix I, is w=O.l7xlO-4 cm and v=O.53~10-4 cm.

8 : Fig. 80. Curvature of a partial separating two regions containing a different stacking fault,, ill~tr&ting the notations used.

Fig. 79.

Model of a fourfold ribbon, illustrating the notations used.

References l) P. B. Hirsch, R. W. Horne and M. 5. Whelan, Phil. Mag. 1 (1956) 677 P. B. Hirsch, J. Silcox, R. E. Smallmann and K. H. Westmacott, Phil. Msg. 3 (1958) 897 P. B. Hirsch, A. Howie and M. J. Whelan, Phil. Trans. Roy. Sot. London 252 (1960) 499 2) W. BoUrnarm, Phys. Rev. 103 (1956) 1588 3) A. Berghezan and A. Fourdeux, Compt. Rend. 248 (1969) 1333 A. Berghezan and A. Fourdeux, Vierter Int. Kongr. El. Miw., Berlin 1958 (Springer Verlag, 1960), p. 567 4) A. Grena.11,Nature 182 (1958) 448; J. Metals11 (1959) 60

V. A. Phillips and P. Cannon, G. E. Ree. Report nr 60-RL-( 1473M) (1960) 9 G. K. Williamson, Proc. Roy. Soo. A 257 (1960) 457 ?) P. Deitbvignette and S. Amelinekx, J. Appl. Phys. 31 (1960) 1691 9 S. Amelinckx and P. Delavignette, J. Appl. Phys. 31 (1960) 2126 9) F. Hubbard Horn, Nature 170 (1952) 581 lo) G. R. Hennig, J. de Chim. Phys. 58 (1961) 12 11) A. Berghezan, A. Fourdeux and S. Amelinokx, Acta Met. 9 (1961) 464 12 ) F. C. Frank, Rep. Conf. Defects in Crys~lI~e Solids (Phys. Sot., London, 1955) p. 159 13 ) M. J. Whelan, J. Inst. Metals 87 (1958-59) 392

7

66 14)

P.

\V. J. Read, Dislocations Book

15)

Cy.,

M. G. Bowman Chem.

16)

Inc.,

Solids

‘7)

R.

I*)

R.

(McGraw-Hill

1953),

Siems,

2’)

1’. 131

and J. A. Krllmhansl, Proc.

4ND

J. Phys.

22)

367 Roy.

Sots., London,

A 249 23)

114

Zeitschr. Phil.

in Crystals York

6 (1958)

M. ?J. Whelan, (1958)

New

DELAVIGNETTE

I’.

I>elavignette

fiir Phys.

(in the

Uevers,

S. Amelinckx

Mag.

(in the press)

I!‘)

A. *J. Kennedy,

2”)

M. J. Whelan, W. Bollmann,

and

Proc.

G. K.

Williamson

(1961)

313

G. H. Kinchin,

Second

Hennig

and

Conf.

1955,

W.

K.

Sot., R.

75 (1960)

607

TV. Horne

and

SOT. A 240 (1957)

524

25)

P/756

P. B. Hirsch, Wiley

G. B. Spcnce, State

P/751

Woods

Dislocations

I)elavignet,tr

and

G. R.

(John

Roy. Roy.

24)

P.

AMELINCKX

1955,

Amelinckx

lxess) and

P. B. Hirsch, Pmt.

S.

S.

Univ.

C. Baker, Geneva

Phil.

Mag.

Conf. P/442

J. E. HOVP, Second

6

(1956) Geneva

(1956)

et al., Second

Geneva

Conference

(1956) R. T%‘. Horne and Mechanical and Fifth 1961,

Sons,

and M. U. Whelan, Properties

New

York,

of Crystals 1957)

Bien. Conf. on Carbon, (to be published)

p. 92 Penn.